The Indicator Guide Health Profiles 2009
This document was uploaded by user and they confirmed that they have the permission to share it. If you are author or own the copyright of this book, please report to us by using this DMCA report form. Report DMCA
Overview
Download & View The Indicator Guide Health Profiles 2009 as PDF for free.
More details
- Words: 104,662
- Pages: 297
The Indicator Guide Health Profiles 2009
Health Profiles 2009 The Indicator Guide
2
Contents Section 1: Introduction Section 2: Our communities 1. Deprivation indicator 2. Children in poverty indicator 3. Statutory homelessness indicator 4. GCSE achieved (5A*–C inc. Eng and Maths) indicator 5. Violent Crime indicator 6. Carbon emissions indicator
6 12 18 24 31 38
Section 3: Children’s and young people’s health 7. Smoking in pregnancy indicator 8. Breast feeding initiation indicator 9. Physically active children indicator 10. Obese children indicator 11. Children’s tooth decay (at age 5) indicator 12. Teenage pregnancy (under 18) indicator
44 52 61 68 77 86
Section 4: Adults’ health and lifestyle 13. Adults who smoke indicator 14. Binge drinking adults indicator 15. Health eating adults indicator 16. Physically active adults indicator 17. Obese adults indicator
94 108 124 141 148
Section 5: Disease and poor health 18. Over 65s ‘not in good health’ indicator 19. Incapacity benefits for mental illness indicator 20. Hospital stays for alcohol related harm indicator 21. Drug misuse indicator 22. People diagnosed with diabetes indicator 23. New cases of tuberculosis indicator 24. Hip fracture in over-65s indicator
165 172 179 185 190 198 203
Section 6: Life expectancy and causes of death 25. Excess winter deaths indicator 26. Life expectancy – male indicator 27. Life expectancy – female indicator 28. Infant deaths indicator 29. Deaths from smoking indicator 30. Early deaths: heart disease and stroke indicator 31. Early deaths: cancer indicator 32. Road injuries and deaths indicator
212 221 227 233 238 248 253 258
Section 7: Charts and trend graphs 33. Deprivation chart 34. Life expectancy by deprivation quintile chart 35. Trend 1: all age, all cause mortality 36. Trend 2: early deaths from heart disease and stroke 37. Trend 3: early deaths from cancer 38. Health inequalities: ethnicity chart
266 271 276 281 286 291
Health Profiles 2009 The Indicator Guide
3
Section 1: Introduction
Health Profiles 2009 The Indicator Guide
4
Introduction The Health Profiles are created using national indicators which are selected to give a snapshot of the health in local authority areas. The metadata for the 32 indicators that have been selected for Health Profiles 2009 are presented within five domains:
1. 2. 3. 4. 5.
Our communities Children’s and young people’s health Adults’ health and lifestyle Disease and poor health Life expectancy and causes of death
The metadata for the charts and trend graphs within the central pages of the Heath Profiles 2009 are also presented. The Indicator Guide provides detailed information about each of these indicators. This includes a summary table with 10 key pieces of basic information about the indicator such as: • What is being measured? • Why is it being measured? • How is the indicator defined? and a further three sections with more detailed information: • The indicator description • The indicator specification • The indicator technical methods
Health Profiles 2009 The Indicator Guide
5
Section 2: Our communities
Health Profiles 2009 The Indicator Guide
6 Section 2: Our communities
1. DEPRIVATION INDICATOR Basic Information 1. What is being measured?
Level of deprivation of a population in an area, as measured by the percentage of people in that area living in the most deprived fifth of areas in England.
2. Why is it being measured?
The differences in deprivation between areas are a major determinant of health inequality in the United Kingdom.
3. How is this indicator actually defined?
IMD 2007 is a model of measuring deprivation in an area. It is underpinned by separate dimensions of deprivation; these dimensions are weighted and an overall deprivation score is given.
4. Who does it measure?
All persons, all ages in the relevant population (otherwise known as the ‘at risk’ population). (This is the population estimate used in the construction of the IMD 2007 – see Table 2)
5. When does it measure it?
Based on various indicators, mostly using 2005 data.
6. Will It measure absolute numbers or proportions?
Proportion based on a composite indicator.
7. Where does the data actually come from?
DCLG website: http://www.communities.gov.uk/communities/ neighbourhoodrenewal/deprivation/deprivation07/
8. How accurate and complete will the data be?
All indicators included in the IMD 2007 are considered to be accurate and complete.
9. Are there any caveats/ warnings/problems?
The indicators are based on mainly 2005. It is based on an average score of an area and it can’t be assumed to represent all individuals in that area. Although very comprehensive, some aspects of deprivation are not to be included in the indices, due to data being incomplete or not available. This may have a larger effect in some areas than others.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
Pg 4 Health Summary – Indicator No. 1
Subject category/ domain(s)
Our Communities
Indicator name (* Indicator title in health profile)
Deprivation
PHO with lead responsibility
Yorkshire and Humber
Health Profiles 2009 The Indicator Guide
7 Section 2: Our communities
Date of PHO dataset creation
Feb 2009
Indicator definition
% of the relevant population in this area living in the 20% most deprived areas in England. (The relevant population is the population estimate used in the construction of the IMD 2007)
Geography
Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs
Timeliness
Indicator is not regularly updated. Published in December 2007 as an update for IMD 2004, the definition is not completely consistent due to changes in some of the indicators used.
Rationale: What this indicator purports to measure
Level of deprivation of a population in an area, by measuring the percentage of the population living in the most deprived quintile of neighbourhoods in England.
Rationale: Public Health Importance
The difference in deprivation between areas is a major determinant of health inequality in the United Kingdom. Many studies and analyses have demonstrated the association of increasingly poor health with increasing deprivation. For instance, all cause mortality, smoking prevalence, self-reported long standing illness are all correlated with deprivation. If deprivation inequalities decrease, health inequalities are likely to decrease also.
Rationale: Purpose behind the inclusion of the indicator
To monitor and help reduce health inequalities.
Rationale: Policy relevance
•A cheson D. Report of the Independent Inquiry into Inequalities in Health. London: TSO, 1998. •D epartment of Health. The NHS Plan. London: TSO, 2000. www.dh.gov.uk/assetRoot/04/05/57/83/04055783.pdf •D epartment of Health. Tackling Health Inequalities: A Programme for Action. Department of Health, 2003. www.dh.gov.uk/assetRoot/04/01/93/62/04019362.pdf •H M Treasury 2007 PBR CSR: Public service agreements: http://www.hmtreasury.gov.uk/ Department for Communities and Local Government The New Performance Framework for Local Authorities & Local Authority Partnerships: Single Set of National Indicators Department for Communities and Local Government, 2007 http://www.communities.gov.uk/corporate/
Health Profiles 2009 The Indicator Guide
8 Section 2: Our communities
Interpretation: What a high/low level of indicator value means
An indicator value worse than average (red circle in health summary chart) represents a statistically significant worse level of deprivation for that local authority when compared to the national value. An indicator value better than average (green circle in health summary chart) represents a statistically significant better level of deprivation for that local authority when compared to the national value. Note: more than a quarter of local authorities have no areas that fall within the most deprived 20% of areas in England and therefore have a value of 0% for this indicator. This means that there is no best quartile range shown on the spine chart for this indicator as all local authorities in the best quartile are at 0%.
Interpretation: Potential for error due to type of measurement method
The indicators are based on mainly 2005 data and this is therefore at least 4 years out of date.
Interpretation: Potential for error due to bias and confounding
It is based on an average score of an area and it can’t be assumed to represent all individuals in that area.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate.
Although very comprehensive some aspects of deprivation will not be included in the indices, for example if the data is incomplete or not collected. This may have a larger effect in some areas than others.
This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Health Profiles 2009 The Indicator Guide
9 Section 2: Our communities
Table 2 – Indicator Specification Indicator definition: Variable
Deprivation
Indicator definition: Statistic
Percentage of the relevant population in this area living in the 20% most deprived areas in England. (The relevant population is the population estimate used in the construction of the IMD 2007)
Indicator definition: Gender
Persons
Indicator definition: age group
All ages
Indicator definition: period
2005
Indicator definition: scale
Percentage of population
Geography: geographies available for this indicator from other providers
Lower Super Output Area http://www.communities.gov.uk/communities/ neighbourhoodrenewal/deprivation/deprivation07/
Dimensions of inequality: subgroup analyses of this dataset available from other providers
IMD 2007 individual domains are available at Lower Super Output Area level and this measure could be derived for them. http://www.communities. gov.uk/communities/neighbourhoodrenewal/deprivation/deprivation07/
Data extraction: Source
Department of Communities and Local Government website
Data extraction: source URL
http://www.communities.gov.uk/communities/neighbourhoodrenewal/ deprivation/deprivation07/
Data extraction: date
Data extracted from source as at: February 2009
Numerator: definition
Number of the relevant population living in the most deprived quintile in England based on the IMD 2007 score. (The relevant population is the population estimate used in the construction of the IMD 2007).
Numerator: source
Department of Communities and Local Government
Denominator: definition
Denominator data – ‘at risk’ mid-2005 population estimates (elsewhere referred to as the ‘relevant’ population). The figures have been adjusted from the ONS mid-year estimate to exclude the prison population in order to fit the definition of ‘at risk’. The figures have been subject to disclosure control.
Denominator: source
Office for National Statistics (ONS) and Department of Communities and Local Government (DCLG).
Data quality: Accuracy and completeness
Criteria for inclusion of indicators to IMD 2004 and IMD 2007 included: • Up-to-date • Statistically robust • Available for the whole of England at a small level in a consistent form
Health Profiles 2009 The Indicator Guide
10 Section 2: Our communities
Table 3 – Indicator Technical Methods Numerator: extraction
The score was derived by YHPHO from the Indices of Deprivation 2007
Numerator: aggregation/ allocation
The number of relevant population in the worst quintile of deprivation was aggregated from Lower Super Output Area to Local Authority level. (The relevant population is the population estimate used in the construction of the IMD 2007).
Numerator data caveats
Deprivation level is based on the average for the area, not all individuals in the area will be deprived.
Denominator data caveats
Census data and mid-year estimates are known to be deficient in their estimates of: • Non-white populations • Full-time students • Men aged 20–39 • People living in nursing homes etc • Rough sleepers • Inner-city populations • Households of multiple occupation • Migrants
Methods used to calculate indicator value
The relevant population in the worst quintile of deprivation was aggregated from Lower Super Output Area to Local Authority level, and divided by the total for the Local Authority. (The relevant population is the population estimate used in the construction of the IMD 2007)
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Excluded at district level but included at regional and national level.
Disclosure Control
Every effort has been made by the DCLG and ONS to ensure that data does not allow the disclosure of confidential information.
Health Profiles 2009 The Indicator Guide
11 Section 2: Our communities
Confidence Intervals calculation method
The 95% confidence intervals are calculated with the method described by Wilson and by Newcombe which is a good approximation of the exact method. First calculate the estimated proportions of subjects with (p) and without (q) some feature of interest from a sample of size n. proportion with feature of interest = p = r/n proportion without feature of interest = q = 1 - p where r is the observed number of subjects with the feature of interest. Second, calculate the three quantities A = 2r + z2;
B = z z 2 + 4rq ;
and
C=2(n+z2),
where z is the appropriate value, z1-α/2, from the standard Normal distribution. Then the confidence interval for the population proportion is given by (A-B)/C
to
(A+B)/C
This method has the considerable advantage that it can be used for any data. When there are no observed events, r and hence p are both zero, and the recommended confidence interval simplifies to 0 to z2/(n+z2). When r = n so that p = 1, the interval becomes n/(n+z2) to 1. Wilson EB. J Am Stat Assoc 1927, 22, 209-212 Newcombe, RG. Two-sided confidence intervals for the single proportion: comparison of seven methods. Stat Med 1998;17:857-72.
Health Profiles 2009 The Indicator Guide
12 Section 2: Our communities
2. CHILDREN IN POVERTY INDICATOR Basic Information 1. What is being measured?
Children in Poverty
2. Why is it being measured?
Growing up in poverty damages children’s health and well-being, adversely affecting their future health and life chances as adults.
3. How is this indicator actually defined?
Income Deprivation Affecting Children Index: part of Indices of Deprivation 2007 – Income deprivation domain. Measuring the proportion of children under 16 years living in families receiving means-tested benefits.
4. Who does it measure?
Children under 16 years in the relevant population (otherwise known as the ‘at risk’ population). (This is the population estimate used in the construction of the IMD 2007 – see Table 2).
5. When does it measure it?
Based on 2005 data.
6. Will It measure absolute numbers or proportions?
Proportions
7. Where does the data actually come from?
Department of Communities and Local Government and ONS http:// www.communities.gov.uk/communities/neighbourhoodrenewal/ deprivation/deprivation07/
8. How accurate and complete will the data be?
The data are of very high quality as they are drawn from a 100% scan of administrative records and as a result are not subject to any sampling error. A small number of claimants whose details are held clerically are excluded. Comprehensive validation checks are undertaken.
9. Are there any caveats/ warnings/problems?
Rounding error is liable to occur in the process of aggregating from LSOAs to Local Authorities. Local authority denominator derived by aggregating LSOA-level relevant child population estimates, these are based on ONS experimental statistics.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
Pg 4 Health Summary – Indicator No. 2
Subject category/ domain(s)
Our Communities
Health Profiles 2009 The Indicator Guide
13 Section 2: Our communities
Indicator name (*Indicator title in health profile)
Income Deprivation Affecting Children Index (*Children in Poverty)
PHO with lead responsibility
Yorkshire and Humber
Date of PHO dataset creation
Feb 2009
Indicator definition
Prevalence of children living in families receiving means-tested benefits, under-16 years (2005). (part of Indices of Deprivation 2007 – Income deprivation domain)
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs
Timeliness
Indicator is not regularly updated. Published in December 2007 as an update for ID 2004, the definition is not completely consistent due to changes in some of the indicators used.
Rationale: What this indicator purports to measure
Child poverty operationalised as children living in families reliant on meanstested benefits and those in receipt of Working Tax Credit and/or Child Tax Credit with an equivalised income below 60 percent of the national median before housing costs. This is used as a proxy for the widely-used HBAI (Households Below Average Income) indicator, which is not available at local authority level.
Rationale: Public Health Importance
Growing up in poverty damages children’s health and well-being, adversely affecting their future health and life chances as adults. Ensuring a good environment in childhood, especially early childhood, is important. A considerable body of evidence links adverse childhood circumstances to poor child health outcomes and future adult ill health. Adverse outcomes include higher rates of: fatal accidents, poor dental health, child mortality, low educational attainment, low birth weight, childhood obesity, school exclusions, infant mortality, teenage pregnancy some infections, substance misuse, mental ill health. By international standards the comparative picture of child poverty in the UK has been poor. International variation in child poverty levels shows that child poverty is not inevitable. In other countries experiencing similar demographic changes and economic pressures to the UK, children have been protected from escalating child poverty by social policy favouring progressive taxation and higher spending on social protection for children. Eradicating child poverty is now a national policy target.
Rationale: Purpose behind the inclusion of the indicator
To monitor and help reduce health inequalities.
Rationale: Policy relevance
Opportunity for All Every Child Matters Children’s National Service Framework
Health Profiles 2009 The Indicator Guide
14 Section 2: Our communities
Interpretation: What a high/low level of indicator value means
An indicator value worse than average (red circle in health summary chart) represents a statistically significant worse rate of child poverty for that local authority when compared to the national value.
Interpretation: Potential for error due to type of measurement method
Benefits claims are an imperfect measure of income deprivation because some eligible families do not claim their entitlement. Others may be living in income deprivation but may not be entitled to claim.
Interpretation: Potential for error due to bias and confounding
Some groups are known to have a low propensity to claim the benefits to which they are entitled, and may be over-represented in certain areas.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself.
An indicator value better than average (green circle in health summary chart) represents a statistically significant better rate of child poverty for that local authority when compared to the national value.
The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Health Profiles 2009 The Indicator Guide
15 Section 2: Our communities
Table 2 – Indicator Specification Indicator definition: Variable
Children living in families receiving means-tested benefits
Indicator definition: Statistic
Percentage of relevant resident population. (This is the population estimate used in the construction of the IMD 2007)
Indicator definition: Gender
Persons
Indicator definition: age group
Under 16 years
Indicator definition: period
2005
Indicator definition: scale
Percentage of child population
Geography: geographies available for this indicator from other providers
Lower Super Output Area http://www.communities.gov.uk/communities/ neighbourhoodrenewal/ deprivation/deprivation07/
Dimensions of inequality: subgroup analyses of this dataset available from other providers
None.
Data extraction: Source
Department of Communities and Local Government and ONS
Data extraction: source URL
http://www.communities.gov.uk/communities/neighbourhoodrenewal/ deprivation/deprivation07/
Data extraction: date
Data extracted from source as at: February 2009.
Numerator: definition
The published data consists of the percentage of relevant children in the LSOA who are living in poverty. The number of such children was calculated by multiplying this proportion by the total number of children in the LSOA. This count was then aggregated from LSOA to LA level. Those counted as living in poverty are relevant children under 16 living in families receiving Income Support or Jobseekers Allowance (Income Based), or in families receiving Working Families Tax Credit/Child Tax Credit whose equivalised income is below 60% of median before housing costs. (The relevant population is the population estimate used in the construction if the IMD 2007)
Numerator: source
The proportion of relevant children living in poverty by LSOA can be found at http://www.communities.gov.uk/communities/neighbourhoodrenewal/ deprivation/deprivation07/ (The relevant population is the population estimate used in the construction if the IMD 2007)
Health Profiles 2009 The Indicator Guide
16 Section 2: Our communities
Denominator: definition
Denominator data – ‘at risk’ mid-2005 population estimates (age under 16) (elsewhere referred to as the ‘relevant’ population). The figures have been adjusted from the ONS mid-year estimate to exclude the prison population in order to fit the definition of ‘at risk’. The figures have been subject to disclosure control.
Denominator: source
http://www.communities.gov.uk/communities/neighbourhoodrenewal/ deprivation/deprivation07/
Data quality: Accuracy and completeness
The data are of very high quality as they are drawn from a 100% scan of administrative records and as a result are not subject to any sampling error. A small number of claimants whose details are held clerically are excluded. Comprehensive validation checks are undertaken by the DWP Information Centre to assess the accuracy, reliability, consistency and completeness of the data. Additional checks were undertaken by the University of Oxford to verify the quality of the data.
Table 3 – Indicator Technical Methods Numerator: extraction
Download from DCLG website and aggregation from LSOA to LA level.
Numerator: aggregation / allocation
The set of benefits is such that only one can be claimed at a time.
Numerator data caveats
Rounding error is liable to occur in the process of aggregating from LSOAs to Local Authorities.
Denominator data caveats
Local authority denominator derived by aggregating LSOA-level child population estimates; these are based on ONS experimental statistics
Methods used to calculate indicator value
The numerator and denominator were derived as explained above and a population weighted average of LSOAs was calculated for each LA. The quotient was expressed as a percentage.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Excluded at district level but included at regional and national level.
Disclosure Control
Every effort has been made by the DWP to ensure that data do not allow the disclosure of confidential information.
Health Profiles 2009 The Indicator Guide
17 Section 2: Our communities
Confidence Intervals calculation method
The 95% confidence intervals are calculated with the method described by Wilson and by Newcombe which is a good approximation of the exact method. First calculate the estimated proportions of subjects with (p) and without (q) some feature of interest from a sample of size n. proportion with feature of interest = p = r/n proportion without feature of interest = q = 1 - p where r is the observed number of subjects with the feature of interest. Second, calculate the three quantities A = 2r + z2;
B = z z 2 + 4rq ;
and
C=2(n+z2),
where z is the appropriate value, z1-α/2, from the standard Normal distribution. Then the confidence interval for the population proportion is given by (A-B)/C
to
(A+B)/C
This method has the considerable advantage that it can be used for any data. When there are no observed events, r and hence p are both zero, and the recommended confidence interval simplifies to 0 to z2/(n+z2). When r = n so that p = 1, the interval becomes n/(n+z2) to 1. Wilson EB. J Am Stat Assoc 1927, 22, 209-212 Newcombe, RG. Two-sided confidence intervals for the single proportion: comparison of seven methods. Stat Med 1998;17:857-72.
Health Profiles 2009 The Indicator Guide
18 Section 2: Our communities
3.1 HOMELESSNESS INDICATOR Basic Information 1. What is being measured?
Estimates of homelessness amongst the most needy and vulnerable groups.
2. Why is it being measured?
Homelessness is associated with severe poverty and is a social determinant of health. Statutorily homeless households contain some of the most vulnerable members of society. Tackling homelessness requires joint working across health and social care services.
3. How is this indicator actually defined?
Statutory homeless households, crude rate per 1000 estimated households, all ages, all persons, 2007 to 2008.
4. Who does it measure?
All persons, all ages.
5. When does it measure it?
Reported quarterly and updated every year.
6. Will It measure absolute numbers or proportions?
Proportions: number of statutorily homeless households per thousand estimated total households.
7. Where does the data actually come from?
Collection and collation from the Housing Strategy Statistical Appendix via the Department for Communities and Local Government.
8. How accurate and complete will the data be?
A built-in-validation system allows each LA to check data accuracy. Data is also validated manually by the DCLG. HSSA and P1E guidance notes provide universal definitions. Missing data is considered to be zeros.
9. Are there any caveats/warnings/ problems?
Data only count ‘statutory homeless’ so does not include the intentionally homeless or those who are not in priority need categories. The total number of households is an estimated figure.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
Pg 4 Health Summary – Indicator 3
Subject category/ domain(s)
Our Communities
Indicator name (*Indicator title in health profile)
Homelessness (*Statutory homelessness)
Health Profiles 2009 The Indicator Guide
19 Section 2: Our communities
PHO with lead responsibility
NEPHO
Date of PHO dataset creation
February 2009
Indicator definition
Statutory homeless households, crude rate per 1,000 estimated total households, all ages, 2007 to 2008, persons
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs
Timeliness
The Housing Strategy Statistical Appendix (HSSA) is updated annually, however it should be noted that quarterly statistics are published on the Department for Communities and Local Government website: www.communities.gov.uk
Rationale: What this indicator purports to measure
Estimates of homelessness amongst the most needy and vulnerable groups in society
Rationale: Public Health Importance
Homelessness is associated with severe poverty and is a social determinant of health. Homelessness is associated with adverse health, education and social outcomes, particularly for children. To be deemed statutorily homeless a household must have become unintentionally homeless and must be considered to be in priority need. As such, statutorily homeless households contain some of the most vulnerable and needy members of our communities. The statutory homeless statistics suggest that 62% of officially accepted homeless households include dependent children or an expectant mother. Preventing and tackling homelessness requires sustained and joined-up interventions by central and local government, health and social care and the voluntary sector.
Rationale: Purpose behind the inclusion of the indicator
To reduce the level of homelessness, particularly amongst the most vulnerable and needy groups in society.
Rationale: Policy relevance
The Department for Communities and Local Government (DCLG) has published a strategy document, ‘Sustainable Communities: Settled Homes: Changing Lives’ which sets out the Government’s plans on reducing homelessness with the aim of halving the number of homeless households in temporary accommodation by 2010.
Interpretation: What a high/low level of indicator value means
An indicator value worse than average (red circle in health summary chart) represents a statistically significant worse level of statutory homelessness for that local authority when compared to the national value. An indicator value better than average (green circle in health summary chart) represents a statistically significant better level of statutory homelessness for that local authority when compared to the national value. However no amount of homelessness is acceptable, and therefore a low indicator value should not mean that public health action is not needed.
Health Profiles 2009 The Indicator Guide
20 Section 2: Our communities
Interpretation: Potential for error due to type of measurement method
The statistic necessarily only measures the incidence of official homelessness. The number of households who are homeless but do not apply to the local authority and are therefore not considered under Housing Act legislation is not known. Reasons may include a lack of knowledge of the legislation, a correct or misplaced belief that they will not qualify for assistance, and/or a desire not to rely on state support. This statistic does not include households that have become unintentionally homeless but are not considered to be in priority need or households that have become intentionally homeless. Rough sleepers are also not included. Therefore, the measure is an underestimate of the extent of homelessness, both of those populations who would qualify for assistance and for the larger number of people who fall outside of the legislation. See: Poverty: the outcomes for children. 2001. ESRC. Edited by: Jonathan Bradshaw.
Interpretation: Potential for error due to bias and confounding
Potential confounding factors associated with the homelessness statistic include: housing affordability, housing capacity, variation in local authority methods of collection and collation of housing and homelessness statistics, local variation in demand for housing.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with a white symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or amber symbol depending on whether it is worse or better than the national value respectively.
Health Profiles 2009 The Indicator Guide
21 Section 2: Our communities
Table 2 – Indicator Specification Indicator definition: Variable
Statutory homeless households
Indicator definition: Statistic
Crude rate per 1,000 total estimated households
Indicator definition: Gender
Persons
Indicator definition: age group
All ages
Indicator definition: period
1 April 2007 to 31 March 2008
Indicator definition: scale
Per 1,000 total estimated households
Geography: geographies available for this indicator from other providers
No other geographies available from other providers.
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Data relating to the ethnicity of households is collected as part of the HSSA.
Data extraction: Source
Department for Communities and Local Government
Data extraction: source URL
http://www.communities.gov.uk/documents/housing/xls/hssa08sectione.xls http://www.communities.gov.uk/documents/housing/xls/hssa08sectioneh.xls
Data extraction: date
Data extracted from source as at: 3 February 2009
Numerator: definition
Count of households (2007/2008) who are eligible, unintentionally homeless and in priority need, for which the local authority accepts responsibility for securing accommodation under part VII of the Housing Act 1996 or part III of the Housing Act 1985.
Numerator: source
Department for Community and Local Government.
Denominator: definition
Total estimated households at 30th June 2008.
Denominator: source
2007/2008 Housing Strategy Statistical Appendix (HSSA) - Section F (Household Numbers) DCLG.
Health Profiles 2009 The Indicator Guide
22 Section 2: Our communities
Data quality: Accuracy and completeness
Each LA checks the data prior to sending it to the Housing Statistics Department of the DCLG. A built-in-validation system allows each LA to check the accuracy of the data (see detailed on the ‘Validation’ section). Once the data has been submitted to the DCLG, validation checks are carried out manually. The HSSA and P1E guidance notes help to provide LAs with universal definitions. Again, for financial management purposes, each LA has to maintain accurate and up-to-date information on homelessness. At LA level, 6% of total entries are missing. This is due to incomplete returns and methods used to safeguard the confidentiality of the data (see ‘Disclosure Control’ Section). There has been no imputation at LA, Regional or National level. Figures registered as missing values at LA level have not been estimated and have therefore been considered to be zeros when aggregating to a higher level.
Table 3 – Indicator Technical Methods Numerator: extraction
Simple download.
Numerator: aggregation / allocation
Counts had already been allocated to local authorities.
Numerator data caveats
To be classified as statutorily homeless, the following must be satisfied: • They are homeless, defined as those without any right to access secure accommodation for that night i.e. they are not legal tenants of any property, nor own any property. Or they can also be classed as ‘potentially homeless’ if they are about to lose their dwelling, be evicted, within 28 days. • They must have a local connection (lived or worked in the area, family in the area, have a care responsibility or need care from relatives in the area). • They are in priority need i.e. had dependent children in them (aged under 16 years) or are an older person household, or vulnerable. • The homeless household must not be intentionally homeless i.e. losing their previous accommodation through their own action such as not paying rent or a mortgage. By contrast the ‘non-statutory’ homeless are, those to whom no duty is owed either because they are deemed intentionally homeless, or are not in a priority need categories. These include the ‘single homeless’, many of whom are now young people of both sexes and, in the larger cities, of different ethnic groups, as well as some older white men.
Denominator data caveats
This is an estimated total number of households at the 30th June 2008.
Health Profiles 2009 The Indicator Guide
23 Section 2: Our communities
Methods used to calculate indicator value
Number of households deemed to be statutorily homeless during the period April 2007 to March 2008, divided by the estimated total number of households as at 30th June 2008, multiplied by 1000.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Isles of Scilly and City of London have been included in regional and England numerators and denominators. Isles of Scilly have been included in the numerator and denominator for the County of Cornwall.
Disclosure Control
Data have been suppressed in this dataset to protect both the confidentiality of individual information and the potential statistical instability caused by low counts. As a consequence, sums of counts may not equal related totals.
Confidence Intervals calculation method
The 95% confidence intervals are calculated with the method described by Wilson and by Newcombe which is a good approximation of the exact method. First calculate the estimated proportions of subjects with (p) and without (q) some feature of interest from a sample of size n. proportion with feature of interest = p = r/n proportion without feature of interest = q = 1 - p where r is the observed number of subjects with the feature of interest. Second, calculate the three quantities A = 2r + z2;
B = z z 2 + 4rq ;
and
C=2(n+z2),
where z is the appropriate value, z1-α/2, from the standard Normal distribution. Then the confidence interval for the population proportion is given by (A-B)/C
to
(A+B)/C
This method has the considerable advantage that it can be used for any data. When there are no observed events, r and hence p are both zero, and the recommended confidence interval simplifies to 0 to z2/(n+z2). When r = n so that p = 1, the interval becomes n/(n+z2) to 1. Wilson EB. J Am Stat Assoc 1927, 22, 209-212 Newcombe, RG. Two-sided confidence intervals for the single proportion: comparison of seven methods. Stat Med 1998;17:857-72.
Health Profiles 2009 The Indicator Guide
24 Section 2: Our communities
4. GCSE ACHIEVEMENT INDICATOR Basic Information 1. What is being measured?
GCSE achievement
2. Why is it being measured?
Educational attainment is influenced by both the quality of education children receive and their family’s socio-economic circumstances. Educational qualifications are a determinant of an individual’s labour market position, which in turn influences income, housing and other material resources. These are related to health and health inequalities.
3. How is this indicator actually defined?
Pupils achieving 5 or more GCSEs at grades A*-C (including English and Maths) or equivalent, percentage of pupils at end of Key Stage 4 in schools maintained by the Local Education Authority, at the end of the academic year 2007-08, persons.
4. Who does it measure?
Pupils at the end of Key Stage 4 in maintained schools.
5. When does it measure it?
This indicator is published annually.
6. Will it measure absolute numbers or proportions?
Proportions, percentage of pupils.
7. Where does the data actually come from?
Collection and collation by the Department for Children, Schools and Families (DCSF).
8. How accurate and complete will the data be?
This indicator only contains data for LEA maintained schools, therefore it excludes pupils educated in private schools.
9. Are there any caveats/warnings/ problems?
Data for Regions and Counties are aggregated on the basis of the schools’ administrating Local Education Authorities. Data for County Districts, Metropolitan County Districts (MCD), Unitary Authorities (UA) and London Boroughs (LB) are aggregated on the basis of Neighbourhood Renewal areas based on the geographic location of the school and not the location of the pupil’s residence.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Health Profiles 2009 The Indicator Guide
25 Section 2: Our communities
TABLE 1 – INDICATOR DESCRIPTION Information component
Pg 4 Health summary – Indicator No 4
Subject category/ domain(s)
Our Communities
Indicator name (*Indicator title in health profile)
GCSE and equivalent results, including maths and English, of pupils at the end of Key Stage 4 (*GCSE achieved 5A*–C inc. Eng & Maths).
PHO with lead responsibility
LHO
Date of PHO dataset creation
Feb 2009
Indicator definition
Pupils achieving 5 or more GCSEs, including maths and English, at grades A*–C or equivalent, percentage of pupils at end of Key Stage 4 in schools maintained by the Local Education Authority, at the end of academic year 2007–2008, persons.
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs, and Strategic Health Authorities for the South East.
Timeliness
Annual. Next data is expected at the end of January 2010. The data is suitable for time trend analysis.
Rationale: What this indicator purports to measure
This indicator measures the level of GCSE achievement in the area.
Rationale: Public Health Importance
Educational attainment is influenced by both the quality of education children receive and their family socio-economic circumstances. Educational qualifications are a determinant of an individual’s labour market position, which in turn influences income, housing and other material resources. These are related to health and health inequalities.
Rationale: Purpose behind the inclusion of the indicator
This indicator relates to the Department for Children, Schools and Families (DCSF) Public Service Agreement (PSA) target 10 which is to raise standards in schools and colleges so that: By 2008, 60% of those aged 16 achieve the equivalent of 5 GCSEs at grades A* to C; and in all schools at least 20% of pupils to achieve this standard by 2004, rising to 25% by 2006 and 30% by 2008. This indicator now uses the same definition as No 75 in the National Indicator Set.
Rationale: Policy relevance
This indicator is the DCSF PSA 10 target. Others include: Tackling health inequalities: A programme for action – Education. Local basket of inequalities indicators – Indicator 3.4. Best Value Performance Indicator 38 Opportunity for all – Children and young people – Indicator 6
Health Profiles 2009 The Indicator Guide
26 Section 2: Our communities
Interpretation: What a high/low level of indicator value means
A high indicator value (green circle in health summary chart) represents a statistically significant higher level of educational attainment for that local authority when compared to the England national value. An increase in this indicator indicates that education achievement has increased relative to the number of pupils (i.e. more children achieving at least 5 GCSEs or equivalent not same children achieving more GCSEs). A low indicator value (red circle in health summary chart) represents a statistically significant lower level of educational attainment for that local authority when compared to the national value. Where educational attainment is not satisfactory, this should prompt investigation as to the causes which may be related to quality of educational services; the socio-economic circumstances of families in the area or other factors known to influence achievement.
Interpretation: Potential for error due to type of measurement method
There are issues of equivalence of GCSE and other qualifications such as NVQs. For example a pupil with English, Mathematics, Double Science and German with top grades, currently counts equally with one who has a bare pass in an Intermediate GNVQ course in IT (being equivalent to 4 passes) and a GCSE grade C in any other subject. Some schools may show improvement in educational attainment through changing the type of qualifications being taken. See: Nuffield Review of 14-19 Education and Training Working Paper 4 (based on Discussion Paper given at Working Day I, 17 Dec 2003) CONTINUITY AND DISCONTINUITY IN THE ‘14-19 CURRICULUM’ Jeremy Higham, Post-14 Research Group, School of Education, University of Leeds. Available from: http://www.nuffield14-19review.org.uk/files/documents12-1.pdf
Interpretation: Potential for error due to bias and confounding
The statistic measures GCSE attainment in LEA maintained schools and therefore does not include attainment for children who are educated in private schools, approximately 7% of children in England. The statistic also does not take into account where pupils live. For some inner-London boroughs, this can mean up to 40% of resident pupils being educated privately or travelling to state schools in other boroughs. Estimating the likely impact of private education and pupils travelling to schools outside their resident local authorities is problematic. The annual schools’ census collects information on pupils educated in independent schools, regardless of where the pupils live. However, information is not available on the area of residence of pupils in independent schools and which local education authority would be responsible for them. Areas have different densities of independent schools. It should also be noted that the figure relates to the location of the school within a local authority area rather than the home residence of the pupils who attend the school.
Health Profiles 2009 The Indicator Guide
27 Section 2: Our communities
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider the confidence interval, the greater the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with a white symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or amber symbol depending on whether it is worse or better than the national value respectively.
TABLE 2 – INDICATOR SPECIFICATION Indicator definition: Variable
Pupils achieving 5 or more GCSEs, including maths and English, at grades A*–C or equivalent
Indicator definition: Statistic
Percentage of pupils at end of Key Stage 4 in schools maintained by the Local Education Authority,
Indicator definition: Gender
Persons
Indicator definition: age group
End of key stage 4 (The term is defined in the Education Act 2002 as “the period beginning at the same time as the school year in which the majority of pupils in his class attain the age of fifteen and ending at the same time as the school year in which the majority of pupils in his class cease to be of compulsory school age”)
Health Profiles 2009 The Indicator Guide
28 Section 2: Our communities
Indicator definition: period
At the end of the academic year 2007–2008
Indicator definition: scale Geography: geographies available for this indicator from other providers
Neighbourhood renewal area; Local Education Authority (LEA) http://www.dcsf.gov.uk/performancetables/
Dimensions of inequality: subgroup analyses of this dataset available from other providers
GCSE achievement: Available by school type and admission’s basis: http://www.dcsf.gov.uk/rsgateway/DB/SFR/s000826/SFR02_2009_SFRTables.xls Also available by sex (boys and girls) and for areas classified by level of deprivation, and for urban/rural areas: http://www.dcsf.gov.uk/rsgateway/DB/SFR/s000826/SFR02_2009_AdditionalTables_ Amended030309-2.xls Also available by free school meal status: http://www.dcsf.gov.uk/rsgateway/DB/SFR/s000826/GCSE_performance_by_FSM_ band.xls Also available by ethnic group: http://www.dcsf.gov.uk/rsgateway/DB/SFR/s000822/SFR322008-KS4FinalNITables_ Amended300109.xls
Data extraction: Source
Department for Children, Schools and Families (DCSF) tables 18 and 20 for 2007/08
Data extraction: source URL
http://www.dcsf.gov.uk/rsgateway/DB/SFR/s000826/SFR02_2009_AdditionalTables_ Amended160109.xls
Data extraction: date
Data extracted from source as at Feb 2009
Numerator: definition
Number of pupils at the end of key stage 4 at the end of the academic year 2007/08 achieving 5 or more GCSEs, including maths and English, at grades A*–C or equivalent in schools maintained by the Local Education Authority. The numerator will include achievements by these pupils in previous academic years. Numerators are not published by DCSF and therefore this number is estimated based on published percentage achievement and the total number pupils at the end of Key Stage 4.
Health Profiles 2009 The Indicator Guide
29 Section 2: Our communities
Numerator: source
DCSF tables 18 and 20 for 2006/07 Performance data at:
Denominator: definition
Number of pupils at the end of Key Stage 4 at the end of the academic year 2007/08 in schools maintained by the Local Education Authority.
Denominator: source
DCSF tables 18 and 20 for 2007/08 Performance data at: http://www.dcsf.gov.uk/rsgateway/DB/SFR/s000826/SFR02_2009_AdditionalTables_ Amended160109.xls
http://www.dcsf.gov.uk/rsgateway/DB/SFR/s000826/SFR02_2009_AdditionalTables_ Amended160109.xls Data for England, Regions and Counties are based on Table 18. Data for County Districts, Metropolitan County Districts, Unitary Authorities and London Boroughs are based on Table 20, with the exception of new UAs formed in April 2009 (County Durham, Northumberland, Shropshire, Wiltshire and Cornwall) which are taken from Table 18.
Data for England, Regions and Counties are based on Table 18. Data for County Districts, Metropolitan County Districts, Unitary Authorities and London Boroughs are based on Table 20. Data quality: Accuracy and completeness
County, Government Office Region and England Total (Maintained sector) figures are adjusted for pupils recently arrived from overseas. This indicator only contains data for LEA maintained schools. Data for Regions and Counties are aggregated on the basis of the schools’ administrating Local Education Authorities. Data for County Districts, Metropolitan County Districts (MCD), Unitary Authorities (UA) and London Boroughs (LB) are aggregated on the basis of Neighbourhood Renewal areas (geographic location of the school).
Table 3 – Indicator Technical Methods Numerator: extraction
Simple download from DCSF website
Numerator: aggregation/ allocation
Data for Regions and Counties are aggregated on the basis of the schools’ administrating Local Education Authorities. Data for County Districts, Metropolitan County Districts (MCD), Unitary Authorities (UA) and London Boroughs (LB) are aggregated on the basis of Neighbourhood Renewal areas (geographic location of the school). Data for some new Unitary Authorities, introduced in April 2009, did not correspond with existing geographies (Cheshire East, Cheshire West and Cheshire, Central Bedfordshire, and Cornwall). Published results are also not available for the two Strategic Health Authorities in the South East: South Central and South East Coast. For these areas data were derived from Table 20, where denominator data could be aggregated to the new geographies. Numerator data can be approximated from published percentages for local authorities and be aggregated to the new geographies to allow the calculation of percentages. Results for the other new UAs were taken from published data for local authorities (Bedford) or counties (County Durham, Northumberland, Shropshire and Wiltshire). None of the data are based on the postcode of the pupils.
Health Profiles 2009 The Indicator Guide
30 Section 2: Our communities
Numerator data caveats
The numerator values for local authorities and Regions will not add up to the total for England as the numerator is estimated. None of the data is based on the post code of the pupils. Excludes pupils educated in private schools
Denominator data caveats
None of the data is based on the post code of the pupils. Excludes pupils educated in private schools
Methods used to calculate indicator value
The number of pupils achieving 5 GCSEs, including maths and English, grade A*–C, divided by the number of pupils at the end of Key Stage 4, multiplied by 100.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Data for Cornwall excludes the Isles of Scilly. Therefore pupil numbers for County Districts will not add up to the data for Counties as a whole, and pupil numbers for MCD, UA, LB and Counties combined do not add up to the total for England. The Isles of Scilly is included in the England and South West Region totals.
Disclosure Control
Usage and dissemination of data is subject to Crown copyright. Data are suppressed if there are fewer than 3 pupils in that particular authority; however, there are no numbers less than 3 in this dataset.
Confidence Intervals calculation method
Confidence intervals have been calculated using the ‘recommended’ formula for a confidence interval of a proportion as described by Newcombe RG and Altman DG. Proportions and their differences. In Altman, DG et al (eds). Statistics with Confidence. 2nd ed. BMJ Books, 2000.
Health Profiles 2009 The Indicator Guide
31 Section 2: Our communities
5. VIOLENT CRIME INDICATOR Basic Information 1. What is being measured?
Recorded crimes of violence against the person.
2. Why is it being measured?
To help target policing and crime prevention resources and to reduce the incidence of violent crime.
3. How is this indicator actually defined?
Recorded violence against the person offences, crude rate per 1,000 population, all ages, 2007/08, persons.
4. Who does it measure?
All persons, all ages.
5. When does it measure it?
Continually reported and data is published annually.
6. Will It measure absolute numbers or proportions?
Proportion: Crude rate per 1,000 population.
7. Where does the data actually come from?
Home Office.
8. How accurate and complete will the data be?
Coverage is complete.
9. Are there any caveats/warnings/ problems?
This indicator only includes violent offences which are reported to the police. It is susceptible to changes in police crime reporting procedures.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
Pg 4 Health Summary – Indicator No 5
Subject category/ domain(s)
Our communities
Indicator name (*Indicator title in health profile)
Violent crime
PHO with lead responsibility
South West Public Health Observatory
Date of PHO dataset creation
November 2008 (revised February 2009 for Local Authorities as at April 2009)
Indicator definition
Recorded violence against the person offences, crude rate per 1,000 population, all ages, 2007/08, persons
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs
Health Profiles 2009 The Indicator Guide
32 Section 2: Our communities
Timeliness
The indicator presented in Health Profiles is routinely updated annually.
Rationale: What this indicator purports to measure
Level of reported violence against the person offences in an area
Rationale: Public Health Importance
The links between crime and health are complex (as outlined in the London Health Commission report http://www.londonshealth.gov.uk/pdf/lhs/crime.pdf and in the NACRO report “Better Health, Lower Crime” http://www.nacro. org.uk/data/resources/nacro-2004120264.pdf) but it is likely that crime is a determinant and a consequence of health. Research undertaken by the Home Office and a number of other organisations suggests that there is a relationship between violent crime and alcohol http:// www.fphm.org.uk/resources/AtoZ/bs_alcohol_violence.pdf Violent crime may result in temporary or permanent disability and in some cases death. Some victims of crime may suffer psychological distress and subsequent mental health problems. Crime and fear of crime can also alter people’s lifestyles and impact on their physical and psychological health Collectively, these consequences represent a burden to the healthcare services. This indicator specifically measures recorded ‘violence against the person’, the largest component of total ‘violent crime’ (which also includes robbery and sexual offences).
Rationale: Purpose behind the inclusion of the indicator
To help target policing and crime prevention resources and to reduce the incidence of violent crime.
Rationale: Policy relevance
Targets included in the Treasury’s Spending Review provide the basis for policy priorities. In 2002 the spending review included targets to reduce crime and the fear of crime and to reduce the gap between those Crime and Disorder Reduction Partnerships experiencing the highest crime and other areas.
Interpretation: What a high/low level of indicator value means
A high indicator value (red circle in health summary chart) represents a statistically significant higher rate of reported violent crime when compared to the England average value. A low indicator value (green circle in health summary chart) represents a statistically significant lower rate of reported violent crime when compared to the England value. However, it should be noted that high values reflect higher numbers of crimes recorded by the police. This may be a result of higher underlying incidence of violent offences, a greater proportion of incidents being reported to the police, or as a result of policing practice. The converse is true for low values of the indicator.
Interpretation: Potential for error due to type of measurement method
This indicator omits violent offences which are not reported to the police. It is susceptible to changes in police crime reporting procedures.
Health Profiles 2009 The Indicator Guide
33 Section 2: Our communities
Interpretation: Potential for error due to bias and confounding
The level and intensity of police service provision will affect rates of recorded violent crime offences, i.e. the higher the level of policing, the more likely it is that recorded crime figures will be elevated. Caution needs to be taken when considering crime rates in areas with low resident populations in which violent crimes are carried out by non-residents.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Table 2 – Indicator Specification Indicator definition: Variable
Recorded violence against the person offences
Indicator definition: Statistic
Crude rate
Indicator definition: Gender
Persons
Indicator definition: age group
All ages
Indicator definition: period
Numerator financial years 2007/08. Denominator mid-2006
Health Profiles 2009 The Indicator Guide
34 Section 2: Our communities
Indicator definition: scale
Per 1,000 population
Geography: geographies available for this indicator from other providers
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs Annual rates and numerators available for these geographies from www.homeoffice.gov.uk
Dimensions of inequality: subgroup analyses of this dataset available from other providers
None available.
Data extraction: Source
Home Office
Data extraction: source URL
http://www.homeoffice.gov.uk/rds/crimeew0708.html
Data extraction: date
Data extracted from source as at: 04/11/2008.
Numerator: definition
Annual count of recorded ‘violence against the person’ offences in the respective financial years. Violence against the person comprises the following offences: Murder, attempted murder, threat or conspiracy to murder, manslaughter, infanticide, child destruction, causing death by dangerous or careless driving, causing or allowing death of a child or vulnerable person, causing death by driving (unlicensed, disqualified or uninsured), wounding or other acts endangering life, endangering a railway passenger, endangering life at sea, other wounding, possession of weapons, harassment, racially or religiously aggravated other wounding, racially or religiously aggravated harassment, cruelty to and neglect of children, abandoning child under two years, child abduction, procuring illegal abortion, causing death by aggravated vehicle taking, assault on a constable, common assault, racially or religiously aggravated common assault. Please see the following for more details: http://www.homeoffice.gov.uk/rds/pdfs07/countviolence07.pdf Please note that for the 2007/08 data, the concealment of birth offence is no longer included in the Violence Against The Person group, as it has been moved to Miscellaneous Other Offences. This is a change for the 2007/08 data from the previous year. Numerator data have been used exactly as published in the Home Office supplied data. Where new Unitary Authorities that have been created as part of the April 2009 boundary changes are exactly co-terminous with pre-existing counties (Cornwall, Durham, Northumberland, Shropshire and Wiltshire) , the numerator data have been drawn from the relevant county figures. Where the new Unitary Authorities represent only a part of the existing counties, numerator data have been aggregated from the relevant published figures for the constituent districts. Please note that a similar approach has been applied to the denominator data (see Denominator: Definition for more information).
Numerator: source
Home Office
Health Profiles 2009 The Indicator Guide
35 Section 2: Our communities
Denominator: definition
2001 Census based mid-year population estimate for the mid-point year (2006). Data are based on the latest revisions of ONS mid-year population estimates for the respective year, supplied as part of the Home Office’s data. – Please note that the Home Office published statistics use mid-2006 population estimates as their denominator, rounded to the nearest 1,000. The data that have been entered into the Health Profiles tool, and the calculated confidence intervals, use the Home Office supplied data as published, except for newly created post-April 2009 Unitary Authorities. For new Unitary Authorities which are exactly coterminous with pre-existing county boundaries (Cornwall, Durham, Northumberland, Shropshire and Wiltshire), the relevant county population has been used in these calculations. For new Unitary Authorities which comprise only a part of the pre-existing county boundaries (Cheshire East, Cheshire West & Chester, Bedford and Central Bedfordshire), an aggregate of the populations for the constituent districts has been used for calculation purposes. Please note that a similar approach has been applied to the numerator data (see Numerator: Definition for more information).
Denominator: source
Home Office data.
Data quality: Accuracy and completeness
Coverage is complete. Historically there have been differences between police forces in procedures for recording, collecting and collating offence data. In April 2002, the National Crime Recording Standard (NCRS) was introduced to ensure greater consistency between forces in recording crime. As a result of the introduction of the NCRS, the recorded number of violence against the person offences increased by 23% in 2002/03. (http://www.crimestatistics.org.uk/output/Page107.asp). Home Office research (http://www.homeoffice.gov.uk/rds/pdfs06/ hosb1206.pdf) suggests that about 45% of violent crimes are reported to the police, with 68% of these being officially recorded as a crime. The Home Office provides specific counting rules regarding the counting and classification of violence against the person offences recorded by Police Forces in England and Wales (http://www. homeoffice.gov.uk/rds/countrules.html). Please also see the Numerator Data Caveats section for related information.
Health Profiles 2009 The Indicator Guide
36 Section 2: Our communities
Table 3 – Indicator Technical Methods Numerator: extraction
Downloaded from www.homeoffice.gov.uk
Numerator: aggregation/ allocation
Area where offence took place allocated by police using records with attached postcodes.
Numerator data caveats
These data exclude figures from British Transport Police and crimes committed at airports within the jurisdictions of Greater Manchester, Metropolitan and Kent police force areas. See Numerator Definition section for change in counting rules. The very nature of crime statistics make analysing the data effectively particularly challenging. Recorded crime, as presented here, is susceptible to fluctuations brought about by changes in recording, counting or even policing methods. In their annual reports entitled “Crime in England and Wales”, the Home Office balance these recorded crime figures with the results of the British Crime Survey (BCS). The BCS is an annual survey which asks respondents about their experience of crime during the previous twelve months, irrespective of whether it was reported to the police. This enables BCS data to be more resilient to changes in recording or policing. Because of survey size limitations, however, it is not possible to use the BCS data at the geographical level required by the Health Profiles. The latest annual report, “Crime in England and Wales 2007/08”, presents a thorough picture of the challenges around crime statistics, in particular how the recorded crime and BCS data may diverge and why, and should be read alongside this metadata document for a full understanding of the issues. The document is at http://www.homeoffice.gov.uk/rds/crimeew0708.html.
Denominator data caveats
Ideally, the denominator should reflect all people at risk of violent crime in the area, including visiting and or migratory persons. This data is not available and resident data has been used as a proxy. The use of resident population as a denominator is a proxy measure for population exposure and is consistent with how this indicator is presented elsewhere.
Methods used to calculate indicator value
Calculation of the numerator: Calculated as an annual of the number of the offences reported in 2007/08. Denominator count: Mid-2006 population estimates (all ages). The numerator was then divided by the denominator; the resulting value was then multiplied by 1000 to give a crude rate per 1000 population.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Data for the Isles of Scilly and City of London have not been included individually. Isles of Scilly is included in the totals for Cornwall, South West GOR and England, and City of London figures have been incorporated into totals for London GOR and England.
Health Profiles 2009 The Indicator Guide
37 Section 2: Our communities
Disclosure Control
Not applicable as no counts less than 5.
Confidence Intervals calculation method
The 95% confidence intervals for crude rate of violence against the person offences were constructed by assuming a Poisson distribution for the number of offences occurring in a specific period of time. The lower and upper limits for this confidence interval were then obtained using the following formulae relating the chi-square and Poisson distributions:
χ α , 2d χ 2α , 2d 2
LL = LL =
2
22
2
χ 1- α , 2(d+1) χ 1- α2 , 2(d+1) 2 2
UL =
22
UL = where LL and UL are the lower and upper2100*(1-a) per cent confidence limits and d denotes the number of observed events (e.g, serious injuries, 100* (1-α) violent offences, deaths) per unit of time exposed. χ² a/2,2d is the (100*a/2)th percentage point for a chi-squared distribution with 2d degrees of freedom 100* (1-α) th 2 and χ²(1-α/2),2(d+1) is the (100*(1-a a/2)) percentage point for a chi-squared χ α/2,2d distribution on 2(d+1) degrees of freedom. χ2 α/2,2d th The confidence limits for the rates were then obtained by dividing the upper (100*α/2) and lower limits for the counts by the person time exposed. (100*α/2)th χ2 (1-α/2),2(d+1)
Reference: χ2 (1-α/2),2(d+1) Dobson AJ, Kuulasmaa K, Eberle E, Scherer J. Confidence intervals for (100*(1-α/2))th weighted sums of Poisson parameters. Statistics in Medicine 1991;10:457-462.
(100*(1-α/2))th (100*(1-α/2))th (100*(1-α/2))th
Health Profiles 2009 The Indicator Guide
38 Section 2: Our communities
6. CARBON EMISSIONS INDICATOR
Basic Information 1. What is being measured?
Total end user CO2 emissions per capita (tonnes of CO2 per resident)
2. Why is it being measured?
Carbon dioxide emission is one of the main contributors to greenhouse gases which cause global warming. This indicator allows local authorities to monitor the effectiveness of their efforts to reduce carbon dioxide emissions.
3. How is this indicator actually defined?
Indicator 186: Per capita CO2 emissions in the local authority area.
4. Who does it measure?
All Persons, all ages.
5. When does it measure it?
Annually, 2006.
6. Will It measure absolute numbers or proportions?
Proportions: CO2 emissions per capita.
7. Where does the data actually come from?
AEA Energy & Environment for the Department for Environment Food and Rural Affairs (Defra)
8. How accurate and complete will the data be?
The indicator relies on centrally produced statistics to measure end user CO2 emissions in the Local Area. Some sectors of CO2 emissions have been excluded from the estimates as they could not be easily aggregated to local or regional levels.
9. Are there any caveats/warnings/ problems?
Analysis carried out by AEA Energy and Environment has confirmed that the data available for the construction of this local area Climate Change Indicator are sufficiently robust with relatively low levels of uncertainty. The data are based upon LA CO2 estimates produced by AEA technology on behalf of Defra.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
Confidence intervals have not been constructed for this indicator.
Table 1 – Indicator Description Information component
Page 4 Health Summary Indicator No. 6
Subject category/ domain(s)
Our Communities
Indicator name (* Indicator title in health profile)
Carbon emissions
Health Profiles 2009 The Indicator Guide
39 Section 2: Our communities
PHO with lead responsibility
WMPHO
Date of PHO dataset creation
23/02/2009
Indicator definition
Local and Regional CO2 Emissions Estimates for 2006 for the UK per capita
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs
Timeliness
Annual. The figures for 2007 will be available in Autumn 2009. It is expected that further improvements to the underlying data and methodology will be made in this dataset which will improve the level of accuracy in certain sectors.
Rationale: What this indicator purports to measure
This indicator purports to measure the contribution of each locality to Carbon dioxide emissions. It is based on estimates of activities resulting in carbon dioxide production (chiefly energy use). The carbon dioxide resulting from energy production is allocated not to the place where the energy is produced but to the place where the energy is used (eg carbon dioxide produced by a power station is allocated to the places using the electricity produced not to the place where the power station is located).
Rationale: Public Health Importance
Carbon dioxide is a major contributor to green house gases which cause global warming. Global warming is of the greatest public health importance and according to the worst predictions could result in millions of deaths across the world as a result of many changes including reduced capacity to produce food, flooding by sea of low lying lands and increased frequency of extreme weather events.
Rationale: Purpose behind the inclusion of the indicator
While the contribution of each single local authority area to carbon dioxide is a miniscule fraction of the global total, carbon dioxide emissions cannot be reduced unless each area accepts its responsibility to reduce emissions and does not simply assume that their contribution does not matter because someone else will make a reduction.
Rationale: Policy relevance
As explained in the previous section reducing carbon dioxide emissions require everyone to make an effort thinking globally but acting locally. This indicator allows each local authority and others to assess how successful their efforts to reduce carbon dioxide emissions have been. The indicator is one of those included in the National Indicator Set that can be used to monitor Local Area Agreements. The data are based upon LA CO2 estimates produced by AEA technology on behalf of Defra. These National Statistics estimate all emissions in an area and have been modified slightly for this indicator to exclude certain emissions of the following components: • Motorways • EU Emissions Trading Scheme sites (except energy suppliers, e.g. power stations in the scheme, whose emissions are indirectly included via the end user estimates, e.g. for electricity use) • Diesel railways • Land Use, Land Use Change, and Forestry In effect, these are emissions that LAs can be least expected to be responsible for. Arguments can be made for other exclusions, but a line has to be drawn somewhere that results in an indicator that is fair in terms of authority actions actually effecting change in the indicator, and where real change at the local level will be captured. Defra consulted with LAs during development of the indicator for LAA purposes.
Health Profiles 2009 The Indicator Guide
40 Section 2: Our communities
Interpretation: What a high/low level of indicator value means
A high indicator value represents a higher level of CO2 emissions per capita for that local authority. A low indicator value represents a lower level of CO2 emissions per capita for that local authority.
Interpretation: Potential for error due to type of measurement method
The indicator relies on accurately attributing emissions from energy use and emissions related to energy supply to the local level. This involves an iterative approach to estimate emissions to the end user, as energy producers will use energy from other producers and therefore also be consumers of energy themselves. Some sectors of CO2 emissions have been excluded from the estimates as they could not be easily aggregated to local or regional levels.
Interpretation: Potential for error due to bias and confounding
Unlike the 2005 Local Authority emissions estimates published last year, this dataset has now been classified as full National Statistics, and is no longer considered “experimental”. This gives users greater confidence in the estimates. In order to obtain this classification, a range of quality criteria set out by the UK Statistics Authority had to be met. To achieve this, there have been in particular, key improvements to the accuracy and comparability of the data. These include: • Re-classification of BERR Local Authority energy statistics as National Statistics; • Implementation of improved formal quality assurance procedures; • Methodological improvements; and • Reduced uncertainty in the accuracy of some of the data inputs. In terms of comparability, a consistent time series has now been produced by re-calculating the 2005 estimates to reflect the methodological changes used in calculating the 2006 estimates.
Confidence Intervals: Definition and purpose
Confidence intervals have not been constructed for this indicator.
Table 2 – Indicator Specification Indicator definition: Variable
CO2 emissions per capita
Indicator definition: Statistic
Tonnes of CO2 emissions per resident.
Indicator definition: Gender
Persons
Indicator definition: age group
All ages
Indicator definition: period
2006 calendar year
Indicator definition: scale
Tonnes per resident
Health Profiles 2009 The Indicator Guide
41 Section 2: Our communities
Geography: geographies available for this indicator from other providers
The original indicator is for the UK, so also includes Wales, Scotland and Northern Ireland. Complete dataset available from http://www.defra.gov.uk/ environment/statistics/globatmos/download/regionalrpt/local-regionalco2ni186indicator.xls
Dimensions of inequality: subgroup analyses of this dataset available from other providers
None available
Data extraction: Source
Local government performance framework NI 186 – Per capita CO2 emissions in the LA area Environment Statistics Service, Department for Environment, Food and Rural Affairs, Area 5F Ergon House, 17 Smith Square, London SW1P 3JR, 08459 33 55 77
Data extraction: source URL
http://www.defra.gov.uk/environment/localgovindicators/ni186.htm
Data extraction: date
5th February 2009
Numerator: definition
Total ktonnes of CO2 emissions, built up of Industry and Commercial, Domestic, Road Transport (excluding Motorways ,EU Emissions Trading Scheme sites (except energy suppliers, e.g. power stations in the scheme, whose emissions are indirectly included via the end user estimates, e.g. for electricity use) Diesel railways, Land Use, Land Use Change, and Forestry)
Numerator: source
Local gas, electricity and road transport fuel consumption estimates are published by BERR for 2005 (see http://www.berr.gov.uk/).
Denominator: definition
Mid Year 2006 population estimates, National Statistics
Denominator: source
Included in NI186 dataset though they have used National Statistics 2006 mid year population estimates.
Data quality: Accuracy and completeness
Unlike the 2005 Local Authority emissions estimates published last year, this dataset has now been classified as full National Statistics, and is no longer considered “experimental”. This gives users greater confidence in the estimates. In order to obtain this classification, a range of quality criteria set out by the UK Statistics Authority had to be met. To achieve this, there have been in particular, key improvements to the accuracy and comparability of the data. These include: • Re-classification of BERR Local Authority energy statistics as National Statistics; • Implementation of improved formal quality assurance procedures; • Methodological improvements; and • Reduced uncertainty in the accuracy of some of the data inputs. In terms of comparability, a consistent time series has now been produced by re-calculating the 2005 estimates to reflect the methodological changes used in calculating the 2006 estimates.
Health Profiles 2009 The Indicator Guide
42 Section 2: Our communities
Table 3 – Indicator Technical Methods Numerator: extraction
Download from Department for Environment Food and Rural Affairs (Defra) website
Numerator: aggregation /allocation
AEA Energy & Environment produce the data on behalf of the Department for Environment Food and Rural Affairs (Defra). They give the following description of the process that they use: “This dataset provides a spatial disaggregation of the national CO2 inventory on an End User basis in which emissions from the production and processing of fuels (including electricity) are reallocated to users of these fuels to reflect the total emissions relating to that fuel use. This is in contrast to ‘at source’ emissions in which all emissions are attributed to the sector that emits them directly. The End User basis for reporting emissions has been chosen for this dataset because it fully accounts for the emissions from energy use at the local level and does not penalise local areas for emissions from the production of energy which is then ‘exported’ to other areas. The method used follows as closely as possible that used for the End User emissions calculated as part of the NAEI and reported by Defra at the national level.”
Numerator data caveats
Data from some areas of CO2 emissions are excluded from the calculations due to the fact that the figures are not available disaggregated to required levels.
Denominator data caveats
Population figures are rounded to the nearest thousand and taken from the 2006 mid year population estimates
Methods used to calculate indicator value
The total of CO2 emissions in ktonnes per area, divided by the population rounded to the nearest thousand.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Scilly Isles and City of London has been included at LA level and included at County, Region and National levels.
Disclosure Control
Copyright of data and/or information presented or attached in this document may not reside solely with this Department. Please contact us or see guidance on copyright at: http://www.defra.gov.uk/environment/ statistics/help.htm
Confidence Intervals calculation method
Not applicable.
Health Profiles 2009 The Indicator Guide
43
Section 3: Children’s and young people’s health
Health Profiles 2009 The Indicator Guide
44 Section 3: Children’s and young people’s health
7. SMOKING IN PREGNANCY INDICATOR Basic Information 1. What is being measured?
Smoking in pregnancy
2. Why is it being measured?
Smoking in pregnancy has well known detrimental effects for the growth and development of the baby. Encouraging pregnant women to stop smoking during pregnancy may also help them kick the habit for good, and thus provide health benefits for the mother.
3. How is this indicator actually defined?
The percentage of women giving birth in 2007/08 who are current smokers at the time of delivery out of all maternities where smoking in pregnancy status is recorded.
4. Who does it measure?
Women giving birth in 2007/08 whose smoking in pregnancy status is recorded.
5. When does it measure it?
Financial year 2007/08.
6. Will It measure absolute numbers or proportions?
Proportions: Number of women who smoke in pregnancy per 100 maternities where smoking status is recorded.
7. Where does the data actually come from?
Care Quality Commission.
8. How accurate and complete will the data be?
93% of LAs have less than or equal to 5% mothers with unknown smoking status and were therefore deemed as having valid prevalence estimates to be included in the Health Profiles 2009.
9. Are there any caveats/warnings/ problems?
The indicator is based on the mother’s response and is therefore susceptible to responder bias. The data is originally recorded at PCT level which has been converted to LA level using birth weighting. If there are several LAs within one PCT they will all have the same prevalence, thereby masking any variation in prevalence which may exist within that PCT. The county, GOR, SHA and England prevalences were calculated from the LAs with valid data, therefore if these LAs are not truly representative of the areas they are being aggregated to, it would result in a biased estimate for aggregated areas.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Health Profiles 2009 The Indicator Guide
45 Section 3: Children’s and young people’s health
Table 1 – Indicator Description Information component
Page 4 Spine Chart – Indicator 7
Subject category/ domain(s)
Children’s and young people’s health.
Indicator name (* Indicator title in health profile)
Smoking in pregnancy
PHO with lead responsibility
ERPHO
Date of PHO dataset creation
11.01.2009
Indicator definition
The percentage of women giving birth in 2007/08 who are current smokers at the time of delivery out of all maternities where smoking in pregnancy status is recorded.
Geography
The data is available at the PCT level this has been converted into the appropriate areas for Health Profiles
Timeliness
The data is released on a quarterly basis.
Rationale: What this indicator purports to measure
Women were asked at the time of delivery whether they currently smoke. Due to the negative connotations surrounding smoking during pregnancy, this may be more susceptible to responder bias.
Rationale: Public Health Importance
Smoking in pregnancy has well-known detrimental effects for the growth and development of the baby. Encouraging pregnant women to stop smoking during pregnancy may also help them kick the habit for good, and thus provide health benefits for the mother.
Rationale: Purpose behind the inclusion of the indicator
To encourage women to quit smoking when pregnant and hopefully as a consequence of this to quit smoking permanently. To highlight LAs with high smoking in pregnancy prevalence to encourage intervention.
Rationale: Policy relevance
This indicator was judged to be a valid and an important measure of public health and was therefore included in the ‘children’s and young people’s health’ domain of the profiles. the NHS Priorities and Planning Framework 2003–06 target to deliver a one percentage point reduction per year in the proportion of women continuing to smoke throughout pregnancy, focussing especially on smokers from disadvantaged groups, is intended to contribute to the national target to reduce the gap in mortality between “routine and manual” groups and the population as a whole by 2010. It is also included in PCT Local Delivery Plans (LDP).
Interpretation: What a high/low level of indicator value means
An indicator value better than average (green circle in health summary chart) represents statistically significantly fewer women smoking during pregnancy for that local authority when compared to the national value. An indicator value worse than average (red circle in health summary chart) represents statistically significantly more women smoking during pregnancy for that local authority when compared to the national value.
Interpretation: Potential for error due to type of measurement method
It is based on the mother’s response and is therefore susceptible to responder bias, particularly as there is a stigma attached to smoking during pregnancy.
Health Profiles 2009 The Indicator Guide
46 Section 3: Children’s and young people’s health
Interpretation: Potential for error due to bias and confounding
As mentioned above, it is likely to be susceptible to responder bias if some mothers are reluctant to admit to smoking in pregnancy, which would make the estimates of smoking in pregnancy more likely to be an underestimate of the actual figure. Achieving good coverage minimises opportunity for non response bias. All PCTs that returned data had at least 95% coverage which was the cut off for inclusion. The calculation of county, region, SHA and England prevalence is based on only those LAs which returned data. They may therefore not be truly representative of the whole area if those LAs with invalid data have different prevalences compared with those LAs with valid data. When viewing amalgamated areas it should be taken into consideration that they may be subject to non response bias, and the overall regional and England prevalences are only estimates.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider the confidence interval the greater the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Health Profiles 2009 The Indicator Guide
47 Section 3: Children’s and young people’s health
Table 2 – Indicator Specification Indicator definition: Variable
The percentage of women giving birth in 2007/08 who are current smokers at the time of delivery out of all maternities where smoking in pregnancy status is recorded.
Indicator definition: Statistic
Percentage
Indicator definition: Gender
Female
Indicator definition: age group
n/a
Indicator definition: period
Financial year 2007/08
Indicator definition: scale
Percentage
Geography: geographies available for this indicator from other providers
PCT (commissioning PCT)
Dimensions of inequality: subgroup analyses of this dataset available from other providers
n/a
Data extraction: Source
Care Quality Commission.
Data extraction: source URL
Most of the data is available on the CQC website:
http://www.cqc.org.uk/publications.cfm?fde_id=1255 The denominator is different from that used by the CQC. In the health profiles only maternities with smoking in pregnancy status recorded is used as the denominator, where as all maternities are used as the denominator for the prevalence of smoking in pregnancy at PCT level calculated by the CQC. This means where you have LAs that are identical to PCTs, the two percentages will not be identical unless there is 100% data coverage (see supporting indicator). Where there is not 100% data coverage the health profiles percentage will be higher than the CQC percentage due to the smaller denominator in the former. The number of births by LA-PCT was extracted from the ONS 2007 birth file, which is sent to PHOs.
Data extraction: date
Received the CQC (formerly Health Care Commission) spreadsheet on 16/10/2008
Numerator: definition
Number of women known to smoke in pregnancy
Numerator: source
Care Quality Commission
Health Profiles 2009 The Indicator Guide
48 Section 3: Children’s and young people’s health
Denominator: definition
Number of maternities where smoking in pregnancy status is recorded
Denominator: source
Care Quality Commission
Data quality: Accuracy and completeness
The CQC data is based on commissioning PCT for the financial year 2007/08 and not PCT of residence, and therefore the number of maternities is slightly different from the ONS number of births in 2007 based on residence. There are estimated prevalence data for 93% of LAs. As the county, GOR, SHA and England prevalences are calculated from LAs which returned data, if these LAs are not truly representative of the areas they are being aggregated to, it would result in a biased estimate for aggregated areas. The data is originally recorded at PCT level which has been converted to LA level using birth weighting. If there are several LAs within one PCT they will all have the same prevalence, thereby masking any variation in prevalence which may exist within that PCT. The regional and SHA values for North West, Yorkshire and Humber, East Midlands, South East (South East Coast SHA and South Central SHA) and South West differ slightly from those produced from summing the CQC PCT data, due to how PCTs that straddle SHAs were dealt with. The regional and SHA data in the health profiles have been calculated by summing the LAs.
Health Profiles 2009 The Indicator Guide
49 Section 3: Children’s and young people’s health
Table 3 – Indicator Technical Methods Numerator and denominator: extraction
The numerator The numerator is converted from the PCT level numerator: “Actual number of women known to smoke in pregnancy (LDPR)”. There were no PCTs with >5% unknown smoking in pregnancy status to exclude but 7 PCTs did not return data (DNRs) and were therefore excluded from the analysis in Excel. The file was then loaded into Access to calculate the LA numerator. Weights based on births were used in order to convert the PCT data into LA data. Births by resident LA and PCT were extracted from the ONS 2007 birth file, to give the number of births in every LA-PCT overlapping block in 2007 by resident LA. This was carried out in SQL using the syntax: SELECT OSCTY, OSLAUA, PCT, COUNT(*) FROM D_births_96_07 LEFT JOIN HES.dbo.A_NationalPostcodes_Feb08 ON D_births_96_07.[Mat Pcode] = HES.dbo.A_NationalPostcodes_feb08.[PCD2] WHERE [SB Ind] <> ‘1’ and year(D_Births_96_07.DOBchild) = 2007 and (GOR = ‘A’ or GOR = ‘B’or GOR = ‘D’or GOR = ‘E’or GOR = ‘F’or GOR = ‘G’or GOR = ‘H’or GOR = ‘J’or GOR = ‘K’) GROUP BY OSCTY, OSLAUA, PCT ORDER BY OSCTY, OSLAUA, PCT To calculate the numerator we need to know the proportion of the LA-PCT overlap of each PCT and then multiply this proportion by the “Actual number of women known to smoke in pregnancy” in the PCT. An LA may overlap several PCTs so this has to be summed at the end. Expressed as an equation the numerator is calculated as follows: Smoking in pregnancy MumsLA = ∑ n/N * Smoking in pregnancy MumsPCT Smoking in pregnancy MumsLA = Estimated number of mothers known to smoke in pregnancy in the LA n= number of births in the LA-PCT overlapping block N = Number of births in the PCT
Health Profiles 2009 The Indicator Guide
50 Section 3: Children’s and young people’s health
Smoking in pregnancy MumsPCT = Number of mothers known to smoke in pregnancy in the PCT Before the data were entered into Access those PCTs who did not return data were excluded, and therefore those LAs entirely made up of PCTs who did not return data will not have an estimate. However if they are made up of several PCTs and only one of which has valid data they will have a numerator and denominator based on this one PCT. We do not want to include an estimate for this LA, so we need to select this out. To find out where this is the case, we worked out the LA-PCT birth overlap in each LA then summed this for the LA we had a prevalence estimate for. This is expressed in the following formula: ∑LA-PCT number of births/LA total number of births. There were no LAs partially made up with PCTs who did not return data so no LAs were excluded at this stage. The Denominator The denominator is calculated in the same way as the numerator. At PCT level the “number of maternities with smoking in pregnancy status known” is calculated in Excel (“actual number of maternities” – “number of maternities with status unknown”). The valid data is then loaded into Access and the denominator is calculated in the same way as the numerator using the same method of weighting and the formula: MaternitiesLA = ∑ n/N * MaternitiesPCT MaternitiesLA = Estimated number of maternities of known status for smoking in pregnancy in the LA n= number of births in the LA-PCT overlapping block N = number of births in the PCT MaternitiesPCT = number of maternities of known status for smoking in pregnancy in the PCT The denominator along with the numerator is then exported to Excel to complete the analysis. Numerator: aggregation/ allocation
Data was aggregated up to LA level using the method described above, in Access. Data was aggregated to county, region, SHA and England level using a look-up table and pivot table in Excel.
Numerator data caveats
We do not have a complete set of data due to 7 DNR PCTs, and not all PCTs had 100% coverage although all that returned data had at least 95% coverage.
Health Profiles 2009 The Indicator Guide
51 Section 3: Children’s and young people’s health
Denominator data caveats
The denominator and numerator are based on commissioning PCT whereas the weights are based on LA/PCT of residence. The denominator is different from that used by the CQC. In the health profiles only maternities with smoking in pregnancy status recorded is used as the denominator, where as all maternities are used as the denominator for the prevalence of smoking in pregnancy at PCT level calculated by the CQC. This means where you have LAs that are identical to PCTs, the two percentages will not be identical unless there is 100% data coverage (see supporting indicator). Where there is not 100% data coverage the health profiles percentage will be higher than the CQC percentage due to the smaller denominator in the former. For example, Barking and Dagenham PCT have a smoking in pregnancy prevalence of 11.3% using the CQC definition, and 11.4% according to the health profile definition.
Methods used to calculate indicator value
The indicator value is calculated as follows in Excel: Percentage of mothers who smoke in pregnancy: numerator/denominator *100
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Not used
Disclosure Control
None
Confidence Intervals calculation method
The 95% and 99.8% confidence intervals are calculated using Julian Flowers’ (erpho) confidence interval tool: http://www.erpho.org.uk/Download/Public/15374/1/Confidence_Intervals_ Wilson.xls This calculates confidence intervals using the following method for a confidence interval of a proportion as described by R.G. Newcombe. If r is the observed number of subjects with some feature in a sample of size n then the estimated proportion who have the feature is p = r/n. The proportion who do not have the feature is q = 1-p. First, calculate the three quantities A = 2r + z2;
B = z z 2 + 4rq ;
and
C=2(n+z2),
where z is z1-α/2, from the standard Normal distribution. Then the confidence interval for the population proportion is given by (A-B)/C
to
(A+B)/C
This method has the considerable advantage that it can be used for any data. When there are no observed events, r and hence p are both zero, and the recommended confidence interval simplifies to 0 to z2/(n+z2). When r = n so that p = 1, the interval becomes n/(n+z2) to 1. Reference: Newcombe, RG. Two-sided confidence intervals for the single proportion: comparison of seven methods. Stat Med 1998;17:857-72.
Health Profiles 2009 The Indicator Guide
52 Section 3: Children’s and young people’s health
8. BREAST FEEDING INITIATION INDICATOR Basic Information 1. What is being measured?
Breast Feeding Initiation
2. Why is it being measured?
Breast feeding has well known health benefits for the child and for the mother in later life. It costs nothing to implement and should be amenable to change through public health intervention.
3. How is this indicator actually defined?
The percentage of women giving birth in 2007/08 who put their baby to the breast in the first 48 hours after delivery, out of all maternities where breast feeding initiation status is recorded.
4. Who does it measure?
Women giving birth in 2007/08 with breast feeding initiation status recorded.
5. When does it measure it?
Financial year 2007/08
6. Will It measure absolute numbers or proportions?
Proportions: Number of women who initiate breast feeding per 100 maternities where breast feeding initiation status is recorded.
7. Where does the data actually come from?
Care Quality Commission
8. How accurate and complete will the data be?
92% of LAs have less than or equal to 5% mothers with unknown breast feeding initiation status and were therefore deemed as having valid prevalence estimates to be included in the Health Profiles 2009.
9. Are there any caveats/warnings/ problems?
The indicator is based on observation and is therefore susceptible to measurement bias. The data is originally recorded at PCT level which has been converted to LA level using birth weighting. If there are several LAs within one PCT they will all have the same prevalence, thereby masking any variation in prevalence which may exist within that PCT. A LA will not have an estimate, if any of its component PCTs have not returned data or have invalid data. The county, GOR, SHA and England prevalences were calculated from the LAs with data returned, therefore if these LAs are not truly representative of the areas they are being aggregated to, it would result in a biased estimate for aggregated areas.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Health Profiles 2009 The Indicator Guide
53 Section 3: Children’s and young people’s health
Table 1 – Indicator Description Information component
Page 4 Spine Chart – Indicator 8
Subject category/ domain(s)
Children’s and young people’s health.
Indicator name (* Indicator title in health profile)
Breast Feeding Initiation
PHO with lead responsibility
ERPHO
Date of PHO dataset creation
11.01.2009
Indicator definition
Measures the percentage of mothers who put their baby to the breast in the first 48 hours after delivery. The numerator is the number of women initiating breast feeding and the denominator the number of maternities with breast feeding initiation status recorded. This for the financial year 2007/08.
Geography
The data is available at the PCT level this has been converted into the appropriate areas for Health Profiles 2009.
Timeliness
The data is released on a quarterly basis.
Rationale: What this indicator purports to measure
It measures the percentage of babies put to the breast in the first 48 hours after delivery, i.e. initiation of breast feeding, directly. However, it does not necessarily reflect the level of sustained breast feeding. The indicator measures only whether baby received breast milk at least once within the first 48 hours rather than sustained breast feeding likely to deliver benefits. The argument is made that mothers who haven’t initiated breast feeding within the first 48 hours rarely do so later, but it is unknown what proportion persist. From a health inequalities perspective it is reasonable to focus attention on the early period to encourage disadvantaged groups to initiate breast feeding. Data is now being collected at 6–8 weeks (NI 53). With increasing quality in the future, this data can be included in future health profiles.
Rationale: Public Health Importance
Breast feeding has well known health benefits for the child and for the mother in later life (Breastfeeding – NHS ). It costs nothing to implement and should be amenable to change through public health intervention.
Rationale: Purpose behind the inclusion of the indicator
To encourage uptake of breast feeding.
Rationale: Policy relevance
This indicator was judged to be a valid and an important measure of public health and was therefore included in the ‘children’s and young people’s health’ domain of the profiles. The NHS Priorities and Planning Framework 2003–06 has a target to deliver a two percentage point increase in breast feeding initiation rates each year, with a particular focus on women from disadvantaged groups. A target for breast feeding initiation was also included in PCT Local Delivery Plans (LDPs).
To highlight LAs with low breast feeding prevalence to encourage intervention.
Health Profiles 2009 The Indicator Guide
54 Section 3: Children’s and young people’s health
Interpretation: What a high/low level of indicator value means
An indicator value better than average (green circle in health summary chart) represents statistically significantly more women initiating breast feeding for that local authority when compared to the national value. An indicator value worse than average (red circle in health summary chart) represents statistically significantly fewer women initiating breast feeding for that local authority when compared to the national value. High percentage of breast feeding initiation is good as it could be considered as a proxy measure for sustained breast feeding.
Interpretation: Potential for error due to type of measurement method
It is based on observation and therefore it prone to measurement bias.
Interpretation: Potential for error due to bias and confounding
It is likely to be subject to measurement bias by the midwives/nurses who record the data and their interpretation of whether breast feeding has been initiated. We have excluded those LAs with component PCTs with >5% missing data to try to avoid non response bias within the LA. The calculation of county, region, SHA and England prevalence is based on only those LAs which returned data. They may therefore not be truly representative of the whole area if those LAs which returned data have different prevalences compared with those LAs which did not. Where a LA consists of a PCT with complete data and a PCT with invalid data, no value was published for the LA (this was the case for Braintree and Runnymede LAs), although the complete PCT data was used for the calculation of the SHA,regional and England values. When viewing amalgamated areas it should be taken into consideration that they may be subject to non response bias, and the overall regional and England prevalences are only estimates.
Health Profiles 2009 The Indicator Guide
55 Section 3: Children’s and young people’s health
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider the confidence interval the greater the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Table 2 – Indicator Specification Indicator definition: Variable
The percentage of women giving birth in 2007/08 who put their baby to the breast in the first 48 hours after delivery, out of all maternities where breast feeding initiation status is recorded.
Indicator definition: Statistic
Percentage
Indicator definition: Gender
Female
Indicator definition: age group
n/a
Indicator definition: period
Financial year 2007–08
Health Profiles 2009 The Indicator Guide
56 Section 3: Children’s and young people’s health
Indicator definition: scale
Percentage
Geography: geographies available for this indicator from other providers
PCT (commissioning PCT)
Dimensions of inequality: subgroup analyses of this dataset available from other providers
n/a
Data extraction: Source
Care Quality Commission
Data extraction: source URL
The data are available on the CQC website: http://www.cqc.org.uk/publications.cfm?fde_id=1255 The denominator in the Health Profiles is different from that used by the CQC. In the health profiles only maternities with breast feeding initiation status recorded is used as the denominator, where as all maternities are used as the denominator for the prevalence of breast feeding initiation at PCT level calculated by the CQC. This means where you have LAs that are identical to PCTs, the two percentages will not be identical unless there is 100% data coverage (see supporting indicator). Where there is not 100% data coverage the health profiles percentage will be higher than the CQC percentage due to the smaller denominator in the former. The number of births by LA-PCT was extracted from the ONS 2007 birth file, which is sent to PHOs.
Data extraction: date
Received the CQC spreadsheet on 16/10/2008
Numerator: definition
Number of women giving birth in 2007/08 who initiate breast feeding in the first 48 hours after delivery.
Numerator: source
Care Quality Commission
Denominator: definition
Number of maternities in 2007/08 where breast feeding initiation status is recorded.
Denominator: source
Care Quality Commission
Health Profiles 2009 The Indicator Guide
57 Section 3: Children’s and young people’s health
Data quality: Accuracy and completeness
The CQC data is based on commissioning PCT for the financial year 2007/08 and not PCT of residence, and therefore the number of maternities is slightly different from the ONS number of births in 2007 based on residence and is for a slightly different time period. The ONS birth file is based on number of deliveries, where as the CQC maternity data is based on mothers and does not take into account multiple births. This factor should not affect prevalence estimates as multiple births are generally a random event and mothers with multiple births are likely to breast feed both babies or neither one. There are estimated prevalence data for 92% LAs. As the county, GOR, SHA and England prevalences are calculated from LAs with returned data, if these LAs are not truly representative of the areas they are being aggregated to, it would result in a biased estimate for aggregated areas. The data is originally recorded at PCT level which has been converted to LA level using birth weighting. If there are several LAs within one PCT they will all have the same prevalence, thereby masking any variation in prevalence which may exist within that PCT. The regional and SHA values for North West, Yorkshire and Humber, East Midlands, South East (South East Coast SHA and South Central SHA) and South West differ slightly from those produced from summing the CQC PCT data, due to how PCTs that straddle SHAs were dealt with. The regional and SHA data in the health profiles have been calculated by summing the LAs.
Health Profiles 2009 The Indicator Guide
58 Section 3: Children’s and young people’s health
Table 3 – Indicator Technical Methods Numerator and Denominator: extraction
The Numerator The numerator is converted from the PCT level numerator: “Actual number of women known to initiate breastfeeding (LDPR)”. There were no PCTs with more than 5% unknowns to exclude and so therefore only those PCTs who DNRs (did not return) were excluded in Excel which was 7 PCTs in total. The file was then loaded into Access to calculate the LA numerator. Weights based on births were used in order to convert the PCT data into LA data. Births by resident LA and PCT were extracted from the ONS 2007 birth file, to give the number of births in every LA-PCT overlapping block in 2006 by resident LA. This was carried out in SQL using the syntax: SELECT OSCTY, OSLAUA, PCT, COUNT(*) FROM D_births_96_07 LEFT JOIN HES.dbo.A_NationalPostcodes_Feb08 ON D_births_96_07.[Mat Pcode] = HES.dbo.A_NationalPostcodes_feb08.[PCD2] WHERE [SB Ind] <> ‘1’ and year(D_Births_96_07.DOBchild) = 2007 and (GOR = ‘A’ or GOR = ‘B’or GOR = ‘D’or GOR = ‘E’or GOR = ‘F’or GOR = ‘G’or GOR = ‘H’or GOR = ‘J’or GOR = ‘K’) GROUP BY OSCTY, OSLAUA, PCT ORDER BY OSCTY, OSLAUA, PCT To calculate the numerator we need to know the proportion of the LA-PCT overlap of each PCT and then multiply this proportion by the “Actual number of women known to initiate breast feeding” in the PCT. A LA may overlap several PCTs so this has to be summed at the end. Expressed as an equation the numerator is calculated as follows: Breast feeding MumsLA = ∑ n/N * Breast feeding MumsPCT Breast feeding MumsLA = Estimated number of mothers known to initiate breast feeding in the LA n= number of births in the LAD-PCT overlapping block N = Number of births in the PCT Breast feeding MumsPCT = number of mothers known to initiate breast feeding in the PCT. Before the data were entered into Access those PCTs which did not return data were excluded, therefore LAs entirely made up of PCTs with not data will not have
Health Profiles 2009 The Indicator Guide
59 Section 3: Children’s and young people’s health
an estimate. However if they are made up of several PCTs and only one of which has valid data they will have a numerator and denominator based on this one PCT. We do not want to include an estimate for this LA, so we need to select this out. To find out where this is the case, we worked out the LA-PCT birth overlap in each LA then summed this for the LA we had a prevalence estimate for. If it did not add up to 1 a prevalence estimate was not published (this was the case for Braintree and Runnymede LAs). This is expressed in the following formula: ∑LA-PCT number of births/LA total number of births. The denominator The denominator is calculated in the same way as the numerator. At PCT level the “number of maternities with breast feeding status known” is calculated in Excel (“actual number of maternities” – “number of maternities with status unknown”). The valid data is then loaded into Access and the denominator is calculated in the same way as the numerator using the same method of weighting and the formula: MaternitiesLA = ∑ n/N * MaternitiesPCT MaternitiesLA= Estimated number of maternities of known status for breast feeding in the LA n= number of births in the LAD-PCT overlapping block N = number of births in the PCT MaternitiesPCT = number of maternities of known status for breast feeding in the PCT The denominator along with the numerator is then exported to Excel to complete the analysis. Numerator: aggregation/ allocation
Data was aggregated up to LA level using the method described above in Access. Data was aggregated to county, region, SHA and England level using a look-up table and pivot table in Excel.
Numerator data caveats
We do not have a complete set of data as there was not 100% coverage from PCTs that returned data (although in all these cases coverage was over 95%) and more importantly 7 PCTs did not return data.
Denominator data caveats
The denominator and numerator are based on commissioning PCT whereas the weights are based on LA/PCT of residence. The denominator is different from that used by the CQC. In the health profiles only maternities with breast feeding initiation status recorded is used as the denominator, where as all maternities are used as the denominator for the prevalence of breast feeding initiation at PCT level calculated by the CQC. This means where you have LAs that are identical to PCTs, the two percentages will not be identical unless there is 100% data coverage (see supporting indicator). Where there is not 100% data coverage, the health profiles percentage will be higher than the CQC percentage due to the smaller denominator in the former. For example, Barking and Dagenham PCT have a breast feeding initiation prevalence of 72.0% according the CQC definition, and a slightly higher prevalence of 72.2% according to the health profile definition.
Methods used to calculate indicator value
The indicator value is calculated as follows in Excel: Percentage of women giving birth in 2007/08 initiating breast feeding: numerator/ denominator *100
Health Profiles 2009 The Indicator Guide
60 Section 3: Children’s and young people’s health
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Not used
Disclosure Control
none
Confidence Intervals calculation method
The 95% and 99.8% confidence intervals are calculated using Julian Flowers’ (erpho) confidence interval tool: http://www.erpho.org.uk/Download/
Public/15374/1/Confidence_Intervals_Wilson.xls This calculates confidence intervals using the following method for a confidence interval of a proportion as described by R.G. Newcombe. If r is the observed number of subjects with some feature in a sample of size n then the estimated proportion who have the feature is p = r/n. The proportion who do not have the feature is q = 1-p. First, calculate the three quantities A = 2r + z2;
B = z z 2 + 4rq ;
and
C=2(n+z2),
where z is z1-α/2, from the standard Normal distribution. Then the confidence interval for the population proportion is given by (A-B)/C
to
(A+B)/C
This method has the considerable advantage that it can be used for any data. When there are no observed events, r and hence p are both zero, and the recommended confidence interval simplifies to 0 to z2/(n+z2). When r = n so that p = 1, the interval becomes n/(n+z2) to 1. Reference Newcombe, RG. Two-sided confidence intervals for the single proportion: comparison of seven methods. Stat Med 1998;17:857-72.
Health Profiles 2009 The Indicator Guide
61 Section 3: Children’s and young people’s health
9. PHYSICALLY ACTIVE CHILDREN INDICATOR Basic Information 1. What is being measured?
Physical Activity provision for school children
2. Why is it being measured?
To help increase childhood participation in physical activity by highlighting areas with low participation rates in order to assess need and enable targeted intervention.
3. How is this indicator actually defined?
The percentage of children attending state schools belonging to a School Sport Partnership who participate in at least 2 hours of high quality PE and school sport within and beyond the curriculum in a typical week of the academic year.
4. Who does it measure?
The total number of school children in state schools who responded to the 2007/08 TNS School Sport Survey who participate in at least 2 hours of high quality PE and out of hours school sport in a typical week.
5. When does it measure it?
2007/08 academic year
6. Will It measure absolute numbers or proportions?
Percentage of all responses to the 2007/8 TNS School Sport Survey.
7. Where does the data actually come from?
TNS Social Research: Annual Survey of School Sport Partnerships on behalf of the Department for Children, Schools and Families.
8. How accurate and complete will the data be?
All partnership schools in the maintained sector in England were included in this survey, with response from over 99% of all such schools. Whilst some private schools have joined a School Sport Partnership and completed the TNS survey, their data is excluded from the DCSF national results dataset. Responses to the TNS School Sport Survey were selfreported by schools given the potential for positive response bias.
Health Profiles 2009 The Indicator Guide
62 Section 3: Children’s and young people’s health
9. Are there any caveats/warnings/ problems?
The indicator is a direct measure of service provision within state schools. It is important to acknowledge that this indicator does not take into account physical activity provision within private schools or physical activity undertaken by children outside of school, and so cannot be used as a total measure of physical activity for children. Although efforts have been made to clearly define ‘high quality PE’, the term is still open to individual interpretation and there is some potential for positive response bias as schools are self-reporting. It is important to make the distinction between physical activity and structured PE/sport. While PE/ sport may be physical activity, physical activity is not necessarily PE/sport; and this data should only be considered as a part of this wider issue within an area.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
Page 4: Health summary – Indicator No. 9
Subject category/ domain(s)
Children’s and Young People’s Health
Indicator name (*Indicator title in health profile)
Physically active children
PHO with lead responsibility
EMPHO
Date of PHO dataset creation
February 2009
Indicator definition
The percentage of children attending state schools belonging to a School Sport Partnership who participate in at least 2 hours of high quality PE and school sport within and beyond the curriculum in a typical week of the academic year.
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs
Timeliness
The survey was carried out between May and July 2008. Results were published in October 2008. The survey is carried out annually with the results from the 2009 survey due out in October 2009.
Health Profiles 2009 The Indicator Guide
63 Section 3: Children’s and young people’s health
Rationale: What this indicator purports to measure
The percentage of children in state maintained schools who participate in at least 2 hours of high quality PE and school sport per week
Rationale: Public Health Importance
All children, whatever their circumstance, should be able to participate in and enjoy PE and sport at school. Physical activity during childhood has a range of benefits including healthy growth and development, maintenance of energy balance, psychological well-being and social interaction. Through improved concentration and self-esteem, it can also improve school attendance, behaviour and attainment. The benefits continue well into adulthood by reducing, early in life, some of the key risk factors for diseases such as coronary heart disease, diabetes and osteoporosis. Some evidence also suggests that participation in physical activity during childhood can help to establish a physically active lifestyle in later life. Physical inactivity in childhood is a modifiable lifestyle risk factor.
Rationale: Purpose behind the inclusion of the indicator
To help increase childhood participation in physical activity by highlighting areas with low participation rates in order to assess need and enable targeted intervention.
Rationale: Policy relevance
This indicator relates directly to Indicator 5 of PSA Target 22: “Increase the percentage of 5 to 16 year olds participating in at least two hours per week of high-quality PE and sport at school” to 85% by 2008, and to at least 75% in each School Sport Partnership by 2008”. The long-term ambition is to offer all children at least 4 hours of sport every week by 2010. See http://www.teachernet.gov.uk/teachingandlearning/subjects/pe/ or http:// www.hm-treasury.gov.uk/d/pbr_csr07_psa22.pdf for further information. This indicator relates indirectly to indicator NI 57 in the National Indicator Set for Local Authorities and Local Authority Partnerships – “Children and young people’s participation in high-quality PE and sport”, which will measure both in school provision and community (out-of-school hours) provision. The in-school aspect of this indicator will be recorded through the School Sport Survey. See http://www.communities.gov.uk/documents/localgovernment/pdf/735125.pdf for further information. In 2004, the CMO Report “At least 5 a week: Evidence on the impact of physical activity and its relationship to health” recommended that children and young people need at least 60 minutes of moderate intensity physical activity each day. See “Chapter 4: Health benefits of physical activity in childhood and adolescence” at http://www.dh.gov.uk/en/Publicationsandstatistics/ Publications/PublicationsPolicyAndGuidance/DH_4080994 Increasing childhood participation in physical activity is also a key message in “Choosing Activity: A Physical Activity Action Plan”, which summarises how the government will deliver the commitments on physical activity that are presented in the public health white paper “Choosing Health” and “Tackling Health Inequalities: 2007 Status report on the Programme for Action”. See http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4105354 and http://www.dh.gov.uk/ en/Publicationsandstatistics/Publications/PublicationsPolicyAndGuidance/ DH_083471.
Health Profiles 2009 The Indicator Guide
64 Section 3: Children’s and young people’s health
Interpretation: What a high/low level of indicator value means
Indicator values should be considered in conjunction with the national target for 85% of school children to be participating in 2 or more hours of physical activity per week, and not just compared to the England average value when assessing the need for public health intervention.
Interpretation: Potential for error due to type of measurement method
The indicator is a direct measure of service provision within state schools. It is important to acknowledge that this indicator does not take into account physical activity provision within private schools or physical activity undertaken by children outside of school, and so cannot be used as a total measure of physical activity for children. Although efforts have been made to clearly define ‘high quality PE’, the term is still open to individual interpretation and there is some potential for positive response bias as schools are self-reporting.
Interpretation: Potential for error due to bias and confounding
Whilst private schools are able to join a School Sport Partnership and complete the TNS survey, their data is excluded from the DCSF national results dataset adding possible bias to the results. It is important to make the distinction between physical activity and structured PE/sport. While PE/sport may be physical activity, physical activity is not necessarily PE/sport; and this data should only be considered as a part of this wider issue within an area.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider the confidence interval the greater the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Health Profiles 2009 The Indicator Guide
65 Section 3: Children’s and young people’s health
TABLE 2 – INDICATOR SPECIFICATION Indicator definition: Variable
High quality PE and out-of-school hours sport
Indicator definition: Statistic
Percentage
Indicator definition: Gender
Persons
Indicator definition: age group
5–16 years
Indicator definition: period
2007/08 academic year
Indicator definition: scale
Percentage of school children in Year 1 – Year 11 attending state schools belonging to a School Sport Partnership
Geography: geographies available for this indicator from other providers
Results summarised by School, School Sport Partnership, Local Education Authority, England are available from the National PE, School Sport and Club Links website at https://dservuk.tns-global.com/schoolsports2007/default.aspx
Dimensions of inequality: subgroup analyses of this dataset available from other providers
The TNS 2008 School Sport Survey National Results report provides 2007/08 data broken down by school type (primary, secondary, special), school year (Year 1 – Year 11), and School Sport Partnership phase (ie when the school joined the School Sport Partnership programme). https://dservuk.tns-global. com/schoolsports2007/default.aspx
Data extraction: Source
TNS Social Research: Annual Survey of School Sport Partnerships on behalf of the Department for Children, Schools and Families.
Data extraction: source URL
Received directly
Data extraction: date
Received directly October 2008
Numerator: definition
The total number of school children in state schools who responded to the 2007/08 TNS School Sport Survey who participate in at least 2 hours of high quality PE and out of hours school sport in a typical week. (Includes compulsory PE curriculum time, optional PE curriculum time (e.g. for GCSE PE students), time spent participating in out of hours school led or school supervised sporting activities, including those taking place during break times within the school day. Excludes travelling time. As of the end of 2007 all state maintained schools in England are in School Sport Partnerships, arranged into 450 Partnerships
Numerator: source
TNS 2007/08 School Sport Survey Further information is available at https://dservuk.tns-global.com/ SchoolSports2008/Default.aspx Data is not available to download but was received directly from TNS.
Health Profiles 2009 The Indicator Guide
66 Section 3: Children’s and young people’s health
Denominator: definition
The total number of school children in state schools who responded to the 2007/08 TNS School Sport Survey. As of the end of 2007 all state maintained schools in England are in School Sport Partnerships, arranged into 450 Partnerships
Denominator: source
TNS 2007/08 School Sport Survey Further information is available at https://dservuk.tns-global.com/ SchoolSports2008/Default.aspx Data is not available to download but was received directly from TNS.
Data quality: Accuracy and completeness
Coverage: All partnership schools in the maintained sector in England were included in this survey, with response from over 99% of all such schools. Whilst some private schools have joined a School Sport Partnership and completed the TNS survey, their data is excluded from the DCSF national results dataset. Data Quality: Responses to the TNS School Sport Survey were self-reported by schools giving the potential for positive response bias. The questionnaire had a 99% response rate giving a relatively complete dataset. Data Reliability: Validation of the 2007/08 TNS School Sport Survey was carried out in 10% of School Sport Partnerships selected at random. The validation concluded that the majority of schools kept auditable records of the information they submitted in their survey responses, and that where auditable records were not available, reasonably robust results should have been produced using the verbally reported approach adopted by schools. For further information see the 2007/08 TNS School Sport Survey Results report, section 2.4 at http://www.dcsf.gov.uk/research/data/uploadfiles/DCSF-RW063. pdf The same questionnaire is used each year giving comparable results.
Table 3 – Indicator Technical Methods Numerator: extraction
Data received directly from TNS.
Numerator: aggregation/ allocation
Performed by TNS.
Numerator data caveats Denominator data caveats Methods used to calculate indicator value
The number of children in state maintained schools who participated in at least two hours of high quality PE/school sport per week for each local authority was divided by the total number of children within each surveyed school with valid responses to questions on physical activity in the TNS School Sport Survey and multiplied by 100. This generated the percentage achieving the recommended levels for school sport participation.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Data for the Isles of Scilly and City of London have not been presented. The Isles of Scilly are included in the Cornwall UA and South West regional and national totals and the City of London figures have been included in the regional London and national totals.
Health Profiles 2009 The Indicator Guide
67 Section 3: Children’s and young people’s health
Disclosure Control
None applied
Confidence Intervals calculation method
Confidence intervals have been calculated using the following method for a confidence interval of a proportion as described by RG Newcombe. If r is the observed number of subjects with some feature in a sample of size n then the estimated proportion who have the feature is p = r/n. The proportion who do not have the feature is q = 1-p. First, calculate the three quantities
A = 2r + z2; B = z z 2 + 4rq ; and C=2(n+z2), where z is z1-α/2, from the standard Normal distribution. Then the confidence interval for the population proportion is given by (A-B)/C
to
(A+B)/C
This method has the considerable advantage that it can be used for any data. When there are no observed events, r and hence p are both zero, and the recommended confidence interval simplifies to 0 to z2/(n+z2). When r = n so that p = 1, the interval becomes n/(n+z2) to 1. Reference Newcombe, RG. Two-sided confidence intervals for the single proportion: comparison of seven methods. Stat Med 1998;17:857-72.
Health Profiles 2009 The Indicator Guide
68 Section 3: Children’s and young people’s health
10. OBESE CHILDREN INDICATOR BASIC INFORMATION1. What is being
Prevalence of obesity in Reception year pupils
measured? 2. Why is it being measured?
To estimate and monitor prevalence of obesity in children. To help reduce the prevalence of childhood obesity; inform planning and delivery of services for children; ensure the proper targeting of resources to tackle obesity.
3. How is this indicator actually defined?
Prevalence of childhood obesity, percentage of school children in Reception year, ages 4–5, 2007/08, persons.
4. Who does it measure?
Children in Reception year (Year R, ages 4–5)
5. When does it measure it?
The National Child Measurement Programme (NCMP) takes place every school year. The data are usually made available around February following publication of the Information Centre National Report.
6. Will It measure absolute numbers or proportions?
Proportion: Percentage of school children in Reception year who are obese.
7. Where does the data actually come from?
The Information Centre for Health and Social Care (IC).
8. How accurate and complete will the data be?
All 152 PCTs provided data and the participation rate across all PCTs was 89% for Reception year (477,652 children measured). 125 PCTs exceeded 85% coverage for Reception year. The NCMP does not include children in the Independent sector, therefore, coverage of school children aged 4–5 is not complete.
9. Are there any caveats/warnings/ problems?
There is the potential for error in the collection, collation and interpretation of the data (bias may be introduced due to poor response rates and selective opt out of larger children which it is not possible to control for).
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Health Profiles 2009 The Indicator Guide
69 Section 3: Children’s and young people’s health
Table 1 – Indicator Description Information component
Pg 4 Health Summary – Indicator No. 10
Subject category/ domain(s)
Children’s and young people’s health
Indicator name (* Indicator title in health profile)
Prevalence of obesity in Reception year pupils (*Obese children)
PHO with lead responsibility
SEPHO
Date of PHO dataset creation
14/01/2009
Indicator definition
Prevalence of childhood obesity, percentage of school children in Reception year, ages 4–5, 2007–08, persons
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs.
Timeliness
The National Child Measurement Programme (NCMP) takes place every year. The 2007/08 data were made available in December 2008.
Rationale: What this indicator purports to measure
Estimate of prevalence of obesity in children in Reception year (Year R, ages 4–5).
Rationale: Public Health Importance
The UK is experiencing an epidemic of obesity affecting both adults and children. The Health Survey for England (HSE) found that among boys and girls aged 2 to 15, the proportion who were classified as obese increased from 10.9 per cent in 1995 to 18.0 per cent in 2005 among boys, and from 12.0 per cent to 18.1 per cent among girls. For those aged 2 to 10, the increase over the same period was from 9.6 per cent to 16.6 per cent for boys and 10.3 per cent to 16.7 per cent for girls. (see http://www.ic.nhs.uk/statistics-and-datacollections/health-and-lifestyles/obesity/statistics-on-obesity-physical-activityand-diet-england-2006)
Note: the definition for childhood obesity used for population monitoring is different to that used in a clinical setting, and produces higher prevalence figures. The prevalence figures here do not equate to the proportion of the child population that would be clinically classified as obese, but more to the proportion of children ‘classified as obese’ according to these population monitoring criteria.
There is concern about the rise of childhood obesity and the implications of such obesity persisting into adulthood. The health consequences of childhood obesity include: increased blood lipids, glucose intolerance, type 2 diabetes, hypertension, increases in liver enzymes associated with fatty liver, psychological problems – social isolation, low self-esteem, teasing and bullying, exacerbation of conditions such as asthma. The National Institute of Health and Clinical Excellence has produced guidelines to tackle obesity in adults in children – Obesity: the prevention, identification, assessment and management of overweight and obesity in adults and children. Available at http://guidance.nice.org.uk/CG43/guidance
Health Profiles 2009 The Indicator Guide
70 Section 3: Children’s and young people’s health
Rationale: Purpose behind the inclusion of the indicator
To estimate and monitor prevalence of obesity in children.
Rationale: Policy relevance
The National Child Measurement Programme (NCMP) was established in 2005 and is one element of the Government’s work programme on childhood obesity. It is operated by the Department of Health and the Department for Children, Schools and Families (DCSF) and central collection and analysis of the NCMP data is coordinated by the Information Centre for health and social care (The IC) which published the final report on its website along with a web tool to view the results by area: http://www.ic.nhs.uk/our-services/improvingpatient-care/national-child-measurement-programme.
To help reduce the prevalence of childhood obesity; inform planning and delivery of services for children; ensure the proper targeting of resources to tackle obesity.
Every year, as part of the NCMP, children in Reception (typically aged 4–5 years) and Year 6 (aged 10–11 years) are weighed and measured during the school year. The findings are used to inform local planning and delivery of services for children and gather population-level surveillance data to allow analysis of trends in excess weight. The programme also seeks to raise awareness of the importance of healthy weight in children. For further details on the NCMP see: http://www.dh.gov.uk/en/Publichealth/ Healthimprovement/Healthyliving/DH_073787 Childhood obesity was the subject of a Public Service Agreement (PSA) target set in July 2004 which aims to halt the year-on-year rise in obesity among children under 11 by 2010 in England. In September 2007, the government announced a new ambition: to reverse the rising tide of obesity and overweight in the population by ensuring that all individuals are able to maintain a healthy weight. The government’s initial focus is on children, and by 2020 they aim to have reduced the proportion of overweight and obese children to 2000 levels. The government strategy on excess weight is set out in “Healthy Weight, Healthy Lives: A Cross-Government Strategy for England” 2008 see: http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_082378 Performance Assessment Framework indicators will be abolished from April 2008 when a new performance framework for local authorities and local authority partnerships comes into place. This framework is supported by a national indicator set. Obesity among primary school age children in Reception Year is indicator NI 56 in the new national indicator set. See: http://www. communities.gov.uk/publications/localgovernment/nationalindicator Interpretation: What a high/low level of indicator value means
A high indicator value (red circle in health summary chart) represents a statistically significant higher level of estimated child obesity prevalence when compared to the national value. A low indicator value (yellow circle in health summary chart) represents a statistically significant lower level of estimated child obesity prevalence when compared to the national value. However obesity in children at any prevalence level greater than 0 is undesirable, and therefore a low indicator value should not mean that PH action is not needed.
Health Profiles 2009 The Indicator Guide
71 Section 3: Children’s and young people’s health
Interpretation: Potential for error due to type of measurement method
Measurement of children’s heights and weights, without shoes and coats and in normal, light, indoor clothing was overseen by healthcare professionals and undertaken in schools by trained staff. However, measurement bias may be introduced via the weighing and measuring and recording process, for example where an area did not use the recommended scales or ensure these were calibrated. There is also evidence that a systematic ‘rounding down’ of measures has occurred in certain areas, and that this has the effect of reducing the reported prevalence of obesity in those areas.
Interpretation: Potential for error due to bias and confounding
Bias may be introduced due to poor response rates and selective opt out of larger children which it is not possible to control for. For the 2007/08 NCMP, child measurements could be taken at any time during the 2007/08 academic year. Consequently, some children were almost two years older than others in the same school year at the point of measurement. As obesity prevalence is known to increase with age, there is potential for this to bias the results Ethnicity has been shown to be linked closely to reported prevalence of child obesity. This might be classed as a confounding factor, as it is not known how suitable the British 1990 Growth Reference is for ethnic populations. The dataset this reference is based on consists of only White British individuals. Further analysis is currently being undertaken to help better understand the links between child obesity and ethnicity. The reported prevalence of obesity is also closely related to the socio-economic characteristics of the children measured. The NCMP does not include pupils at independent schools and, as these pupils may be from a more affluent background, there is a possibility that this might bias prevalence figures using the NCMP.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider the confidence interval is the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health.
Health Profiles 2009 The Indicator Guide
72 Section 3: Children’s and young people’s health
The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with a white symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or amber symbol depending on whether it is worse or better than the national value respectively.
Table 2 – Indicator Specification Indicator definition: Variable
Prevalence of childhood obesity. Obesity in children is commonly defined for epidemiological purposes as having a Body Mass Index (BMI) greater than the 95th percentile (using the British 1990 growth reference). This is in line with the Health Survey for England and other population studies within the UK. Note: As prevalence rates have been calculated using the 95th percentile (see Methods used to calculate indicator value). This does not equate to the proportion of children clinically defined as obese, as the 98th percentile is used in clinical settings.
Indicator definition: Statistic
Percentage of school children in Reception year
Indicator definition: Gender
Persons
Indicator definition: age group
School children in Reception year (Year R), ages 4–5
Indicator definition: period
School year 2007–08 (September 2007 to September 2008)
Indicator definition: scale Geography: geographies available for this indicator from other providers
PCT, new SHA Available from http://www.ncmp.ic.nhs.uk/
Dimensions of inequality: subgroup analyses of this dataset available from other providers
The Information Centre’s summary report (see: http://www.ic.nhs.uk/) shows strong links between socio-economic status, ethnicity, urban/rural environment and prevalence. There are also some significant differences in prevalence by gender.
Data extraction: Source
The Information Centre for Health and Social Care (IC).
Data extraction: source URL
www.ic.nhs.uk/
Health Profiles 2009 The Indicator Guide
73 Section 3: Children’s and young people’s health
Data extraction: date
December 2008
Numerator: definition
The number of primary school age children in Year R (Reception year, ages 4–5) with valid height and weight recorded in the school year 2007–08 who are classified as obese.
Numerator: source
www.ic.nhs.uk/
Denominator: definition
The total number of primary school age children in Year R (Reception year, ages 4–5) with valid height and weight recorded in the school year 2007–08.
Denominator: source
www.ic.nhs.uk/
Data quality: Accuracy and completeness
Pupils eligible for inclusion in the NCMP were all children in Reception and Year 6 attending non-specialist maintained state schools in England. The NCMP does not include children in the Independent sector, therefore, coverage of school children aged 4–5 is not complete. Primary care trusts (PCTs) coordinate the data collection exercise with the support and cooperation of schools. The numbers of pupils eligible for inclusion at each school were provided by the DCSF but PCTs could edit these figures if necessary to ensure pupil denominators accurately reflected the number of children attending the school on the day of measurement. PCTs could also add or remove schools from their geograpically assigned list if, despite being within their PCT boundary, another PCT had undertaken measurement in that school. The IC conducted checks when all data was submitted to ensure schools had not been removed by one PCT but not added on by another. Primary Care Trust (PCT) staff used specially designed Excel spreadsheets (the NCMP Data-Capture tool) to enter these data and upload them to a central database at the Information Centre for Health and Social Care. All 152 PCTs provided data and the participation rate across all PCTs was 89% for Reception year (477,652 children measured). 125 PCTs exceeded 85% coverage for Reception year. PCTs were set a participation rate target for NCMP 2007/08 of 85% (for Reception and Year 6). Participation rates were assessed by comparing numbers measured with total numbers on school lists and are published in the IC National report (see http://www.ncmp.ic.nhs.uk/). Participation rates should be treated with caution because of difficulties in calculation of eligible pupil numbers particularly in Reception when pupils may join the school throughout the year. Low participation rates may bias prevalence figures and analysis has shown that PCTs with lower participation rates tend to have lower levels of prevalence than those with high participation rates. This suggests that there might be higher levels of opting out among children with higher BMIs. However, initial analysis by the IC shows participation rate to have little of no effect on prevalence for children in Reception.
Health Profiles 2009 The Indicator Guide
74 Section 3: Children’s and young people’s health
Table 3 – Indicator Technical Methods Numerator: extraction
Numerator extracted from an Access database provided by IC which included all data uploaded to the central NCMP database as at December 2008. The numerator was based on a pre-calculated field included within the database which categorised each child as “healthy weight”, “overweight” or “obese”.
Numerator: aggregation/ allocation
The figures have been aggregated to local authorities from school counts on the basis of postcode of school. Regional figures have been calculated by combining the Local Authority figures. SHA figures (provided for the South East Region only) have been calculated by assigning schools to SHAs according to the PCT who measured those pupils.
Numerator data caveats
The total number of children measured was the total number of records uploaded to the NCMP database after the following records had been removed: • Blank school Unique Reference Number, height, weight, sex or age. • Children outside the age range (age in months between 48 and 83 inclusive) • Independent/Private/Special Educational Needs (SEN) pupils • Records from Academies, Community Special, Foundation Special, Independent School Approved for SEN pupils, Non-maintained Special, Other Independent, Other Independent Special School, Pupil Referral Unit, where these schools had been measured and entered by PCTs • Records with extreme heights, weights and BMIs where value was further than 7 standard deviations from the mean After the above data cleaning, approximately 477,600 year R records remained. Further information on data quality and the validation process is included in the 2007/08 report from the Information Centre: http://www.ic.nhs.uk/ statistics-and-data-collections/supporting-information/health-and-lifestyles/ obesity/the-national-child-measurement-programme/national-childmeasurement-programme-2007/08/ncmp-2007-08-report
Denominator data caveats
As above.
Health Profiles 2009 The Indicator Guide
75 Section 3: Children’s and young people’s health
Methods used to calculate indicator value
The IC undertook the following steps to calculate the number of obese children: Every child’s Body Mass Index (BMI) was calculated as follows: BMI = 10,000 x weight (kg) height2 (cm) BMI was then referenced against the British 1990 Growth Reference (1) (2). This uses UK growth data from pre-1990 based on a large representative sample of 37,700 children constructed by combining data from 17 separate surveys. These data were then used to model the BMI distribution using Cole’s LMS method (3) for ages 0–20 for both males and females. This dataset can be used to express an individual child’s BMI as a percentile on the 1990 distribution. Each child’s BMI was referenced to the age (in decimal months) and sexspecific BMI growth curve in order to retrieve corresponding L, M and S values. These were used to calculate a z-score for each child using the following formula:
y
M
Where y is the BMI score. The z-score was converted to a p-score using the standardised normal distribution and children with a BMI p-score >=0.95 were flagged as obese. Using the data supplied by IC, prevalence rates were calculated by dividing the number of primary school age children in Year R (Reception year, ages 4–5) with valid height and weight measurements recorded as obese by the total number of primary school age children in Year R with valid height and weight measurements. This figure was multiplied by 100 to give percentage prevalence. (1) Cole, T.J., Freeman, J.V. and Preece, M. A. ‘British 1990 growth reference centiles for weight, height, body mass index and head circumference fitted by maximum penalized likelihood’, Statistics in Medicine, 17, 407-429 (1998). (2) Cole, T. J., Freeman, J. V. and Preece, M. A. ‘Body Mass Index reference curves for the UK, 1990’, Archives of Disease in Childhood, 73, 17-24 (1995). (3) Cole, T. J. and Green, P. J. ‘Smoothing reference centile curves: The LMS method and penalized likelihood’, Statistics in Medicine, 11, 1305-1319 (1992). Note: 40 records from Reception year had not been assigned to a local authority. On closer examination of school location, these were reassigned to Telford local authority.
Health Profiles 2009 The Indicator Guide
76 Section 3: Children’s and young people’s health
Small Populations: How Isles of Scilly and City of London populations have been dealt with
City of London has been included in the Hackney, Regional and National totals. Isles of Scilly have been included in the new Cornwall Unitary Authority.
Disclosure Control
None applied.
Confidence Intervals calculation method
The 95% confidence intervals are calculated with the method described by Wilson1 and Newcombe2 which is a good approximation of the exact method. The estimated proportions of subjects with and without the feature of interest were calculated: observed number of obese children in each area =r sample size = n proportion with feature of interest = p = r/n proportion without feature of interest = q (1 – p) Three values (A,B and C) were then calculated as follows: A = 2r + z2;
B = z z 2 + 4rq ;
and
C=2(n+z2)
where z is the appropriate value, z1-α/2, from the standard Normal distribution. Then the confidence interval for the population proportion is given by (A-B)/C
to
(A+B)/C
This method has the considerable advantage that it can be used for any data. When there are no observed events, r and hence p are both zero, and the recommended confidence interval simplifies to 0 to z2/(n+z2). When r = n so that p = 1, the interval becomes n/(n+z2) to 1. 1
Wilson EB. J Am Stat Assoc 1927, 22, 209-212
Newcombe, RG. Two-sided confidence intervals for the single proportion: comparison of seven methods. Stat Med 1998;17:857-72.
2
Health Profiles 2009 The Indicator Guide
77 Section 3: Children’s and young people’s health
11. CHILDREN’S TOOTH DECAY INDICATOR Basic Information 1. What is being measured?
Children’s tooth decay (at age 5)
2. Why is it being measured?
To draw attention to areas of high tooth decay. To improve oral health in children by reducing the prevalence of dental decay.
3. How is this indicator actually defined?
Mean number of teeth per child sampled which were either actively decayed or had been filled or extracted.
4. Who does it measure?
Sampled number of children at 5 years old with decayed/missing/filled teeth (dmft).
5. When does it measure it?
2005/06
6. Will It measure absolute numbers or proportions?
Numbers (the mean of)
7. Where does the data actually come from?
BASCD (British Association for the Study of Community Dentistry) www.bascd.org
8. How accurate and complete will the data be?
Data was only available at PCT level, and had to be apportioned to Local Authorities. This process can only be approximate where PCTs are not completely contained within Local Authorities. Data was missing for 22 PCTs and consequently it was not possible to present mean dmft for 26 local authorities and one county. Data for 12 PCTs was only available in combination with one or more other PCTs. Health Profiles 2009 are based on the new unitary authority boundaries and it has not been possible to present mean dmft for 3 of these new authorities.
9. Are there any caveats/warnings/ problems?
Positive/negative consent issue may cause bias.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
Pg 4 Health Summary – Indicator No. 11
Subject category/ domain(s)
Health and ill-health in our community
Indicator name (*Indicator title in health profile)
Mean number of decayed/missing/filled teeth in five-year-olds (*Children’s Tooth Decay)
Health Profiles 2009 The Indicator Guide
78 Section 3: Children’s and young people’s health
PHO with lead responsibility
Yorkshire and Humber
Date of PHO dataset creation
March 2008
Indicator definition
Mean number of teeth per child sampled which were either actively decayed or had been filled or extracted.
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs. Children are allocated to these geographies according to the location of their school.
Timeliness
BASCD (British Association for the Study of Community Dentistry) conduct a survey of five-year-olds every two years. The 2007/08 survey results are due for publication later in 2009.
Rationale: What this indicator purports to measure
The mean number of decayed, missing or filled teeth per child.
Rationale: Public Health Importance
Dental caries (tooth decay) and periodontal (gum) disease are the most common dental pathologies in the UK. Tooth decay has become less common over the past two decades, but is still a significant health and social problem. It results in destruction of the crowns of teeth and frequently leads to pain and infection. Dental disease is more common in deprived, compared with affluent, communities. The indicator is a good direct measure of dental health and an indirect, proxy measure of child health and diet.
Rationale: Purpose behind the inclusion of the indicator
To draw attention to areas of high tooth decay. To improve oral health in children by reducing the prevalence of dental decay.
Rationale: Policy relevance
Children’s National Service Framework
Interpretation: What a high/low level of indicator value means
An indicator value worse than average (red circle in health summary chart) represents a statistically significant worse rate of children’s tooth decay for that local authority when compared to the national value.
This indicator supports Choosing Health and LAAs.
An indicator value better than average (green circle in health summary chart) represents a statistically significant better rate of children’s tooth decay for that local authority when compared to the national value. From this statistic alone, it cannot be deduced whether the problem is evenly distributed or confined to a small pocket of children.
Health Profiles 2009 The Indicator Guide
79 Section 3: Children’s and young people’s health
Interpretation: Potential for error due to type of measurement method
Data was only available at PCT level, and had to be apportioned to Local Authorities. This process can only be approximate where PCTs are not completely contained within Local Authorities. Data was missing for 22 PCTs listed below and consequently it was not possible to present mean dmft for 26 local authorities and one county. Data for 12 PCTs was only available in combination with one or more other PCTs, these are also listed below. 45 PCTs where positive consent operated for all or part of the sampling period are listed below. Data missing for these PCTs PCT 5A7 5A8 5AK 5CE 5CV 5CW 5FN 5FQ 5FR 5FT 5FV 5FY 5GF 5J8 5J9 5JP 5KA 5KC 5KD 5KE 5KV TAK
PCT name Bromley PCT Greenwich PCT Southend on Sea PCT Bournemouth Teaching PCT South Hams and West Devon PCT Torbay Care Trust South and East Dorset PCT North Devon PCT Exeter PCT East Devon PCT Mid Devon PCT Teignbridge PCT Huntingdonshire PCT Durham Dales PCT Darlington PCT Castle Point and Rochford PCT Derwentside PCT Durham and Chester-le-Street PCT Easington PCT Sedgefield PCT Poole PCT Bexley Care Trust
Data available for these PCTs in combination only South East Hertfordshire PCT (5GJ), Royston & Buntingford and Bishops Stortford PCT (5GK) Ashford PCT (5LL), Canterbury and Coastal PCT (5LM), East Kent Coastal PCT (5LN), Shepway PCT (5LP) Watford and Three Rivers PCT (5GV), Dacorum PCT (5GW) Hertsmere PCT (5CP), St Albans and Harpenden PCT (5GX) Welwyn and Hatfield (5GG) North Hertfordshire and Stevenage PCT (5GH)
Health Profiles 2009 The Indicator Guide
80 Section 3: Children’s and young people’s health
Positive/Mixed consent in operation during sampling period PCT PCT name 5DF North Hampshire 5DK Newbury 5E1 North Tees 5G6 Blackwater Valley & Hart PCT 5HF Wyre PCT 5KJ Part of Craven Harrogate & Rural (Airedale) 5KL Sunderland 5KM Middlesborough 5L5 Guildford & Waverley PCT 5L7 Surrey Heath and Woking Area PCT 5AC Daventry & South Northants 5AW Airedale 5CC Blackburn with Darwen PCT 5CF Bradford City 5CG South,West Bradford 5CH North Bradford 5CK Doncaster Central 5CX Trafford South PCT 5EE North Sheffield 5EG North Eastern Derbyshire 5EN Sheffield West 5EP Sheffield South West 5EQ South East Sheffield 5F5 Salford PCT 5F6 Trafford North PCT 5G7 Hyndburn & Ribble Valley PCT 5HA Central Liverpool PCT 5HE Fylde PCT 5HG Ashton, Leigh & Wigan PCT 5HH Leeds West 5HJ Leeds North East 5HK East Leeds 5HL South Leeds 5HM Leeds North West 5HN High Peak and Dales 5HQ Bolton PCT 5HX Ealing 5J6 Calderdale 5J7 North Kirklees 5LJ Huddersfield Central 5LK South Huddersfield 5LV Northamptonshire Heartlands 5LW Northampton 5M5 South Sefton PCT 5M7 Sutton and Merton
Form of Consent Positive Positive Positive Positive Positive Positive Positive Positive Positive Positive Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed
The data source is a series of nationally co-ordinated dental epidemiological surveys commissioned by individual PCTs to standardised national protocols and diagnostic standards and involving the dental examination of children in the specified age-group, in state schools. The data source is part of a cycle of nationally co-ordinated dental epidemiological surveys as outlined in Health Service Guidelines (93)25. Surveys are conducted every second year for 5 year olds and every fourth year for 12 years olds and 14 year olds.
Health Profiles 2009 The Indicator Guide
81 Section 3: Children’s and young people’s health
The data relate to children attending state schools in an area. It cannot be assumed that all children necessarily live in the same area. National minimum standards are set for the random sampling of children to obtain a sample representative of the age-group in the area. Many Health Authorities commission larger samples in order to obtain data on intra-district variations in dental caries for local planning purposes. Data are collected and analysed locally. Summary data items are reported nationally to the British Association for the Study of Community Dentistry, which produces national tables through the Dental Health Services Research Unit at the University of Dundee. These data are published in the journal of the British Association of the Study of Community Dentistry, Community Dental Health. Interpretation: Potential for error due to bias and confounding
For the first time in the history of the BASCD survey of the dental caries experience of 5-year-old children in England and Wales, some parents were required to give positive consent for their children’s teeth to be inspected. In previous surveys all children in state schools were, by default, eligible for inclusion unless their parents submitted a form stating that they did not wish for their children’s teeth to be inspected. This has led to serious concerns that the results of the 2005/06 survey may be biased and not comparable to earlier surveys. Bias may result for several reasons. Firstly, there may be variation between schools generally in how pro-active they are at encouraging parents to return consent forms. Secondly, parents living in more deprived areas (where mean dmft is higher) may be less likely to return consent forms, than those living in more affluent areas (where mean dmft is lower). The picture is further confused by the fact that positive consent was not universally introduced. In 10 PCTs all schools operated positive consent throughout the sampling period, in 35 PCTs some of the schools operated positive consent for some of the sampling period, and in the remaining PCTs no schools operated positive consent. It will be difficult to disentangle the impact of the introduction of positive consent from genuine reductions in the prevalence of dental caries. Anecdotal evidence so far suggests that positive consent had a large effect in some areas.
Health Profiles 2009 The Indicator Guide
82 Section 3: Children’s and young people’s health
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Table 2 – Indicator Specification Indicator definition: Variable
Decayed, filled or missing teeth.
Indicator definition: Statistic
Mean number per child
Indicator definition: Gender
Persons
Indicator definition: age group
Five-year-olds
Indicator definition: period
Sampled during winter months of 2005–06.
Indicator definition: scale
Per child.
Geography: geographies available for this indicator from other providers
PCT and SHA. Available from www.bascd.org
Health Profiles 2009 The Indicator Guide
83 Section 3: Children’s and young people’s health
Dimensions of inequality: subgroup analyses of this dataset available from other providers
None.
Data extraction: Source
2006/06 BASCD survey of the dental caries experience of 5-year-old children in England and Wales.
Data extraction: source URL
Supplied privately by BASCD.
Data extraction: date
March 2007
Numerator: definition
Number of decayed/missing/filled teeth in the survey sample of children in the respective academic year. This was not available at local authority level, therefore the calculation did not proceed by dividing this number by a local authority population.
Numerator: source
Dental epidemiological survey programme undertaken by Health Authorities and co-ordinated nationally for the UK Health Departments by the British Association for the Study of Community Dentistry (BASCD).
Denominator: definition
Number of children in the survey sample. This was not available at local authority level, therefore the calculation did not proceed by dividing the number of decayed, filled or missing teeth by this number.
Denominator: source
Provided by BASCD.
Data quality: Accuracy and completeness
See: Interpretation: Potential for error sections and Methods used in Table 3
Table 3 – Indicator Technical Methods Numerator: extraction
Provided by BASCD.
Numerator: aggregation/ allocation
See Methods used below
Numerator data caveats
The data for 5-year-olds relate to deciduous (milk) teeth.
Denominator data caveats
See: Interpretation: potential for error sections and Methods used below
Health Profiles 2009 The Indicator Guide
84 Section 3: Children’s and young people’s health
Methods used to calculate indicator value
dmft for 2005/06 was calculated at PCT level. A PCT may contain or overlap several LADs. The following steps were taken to estimate dmft for Local Authorities (LADs): • ‘Building blocks’ were created based on intersections between PCTs and LADs. • There is a strong relationship between deprivation and dmft. Rather than base the estimated dmft just on the ‘parent’ PCT which may include areas with different levels of deprivation to the building block belonging to the LAD, the calculation was adjusted for deprivation. • This adjustment was made by firstly applying the regression equation between all PCTs dmft and IMD score, to give an expected dmft for the PCT. Secondly, a dmft estimate was arrived at for each building block which reflected the fact that the more deprived ones would carry more than their ‘fair share’ of their parent PCT’s burden of dental decay. The building block estimate is calculated as follows:
[PCT dmft estimate from survey ]∗ [building block expected dmft ] PCT expected dmft • The calculation needed to reflect the different populations of five-year-olds by area, however this was not available for the building blocks and a proxy weight was derived from the 2005 Royal Mail Postcode Address File (PAF) which is considered a good estimate of population distribution • The final LAD estimate was obtained by multiplying each building block’s dmft estimate by the number of residences in the building block, adding up all the building blocks in the LAD, and dividing the total number of residencies in the LAD.
y LAD = ∑
dmft i Ni ∗ ∗ yi dmft PCT N LAD
Where:
y
• LAD is the estimated mean dmft in the LAD, y is the mean dmft found in the ith PCT (supplied by BASCD) • i • dmfti is the expected dmft in the ith ‘building block’ (LAD by PCT intersection) • dmftPCT is the expected dmft in the ‘parent’ PCT • Ni is number of residences in the ith ‘building block’ (LAD by PCT intersection) • NLAD is the total number of residences in the LAD This was the method used for LADs and Counties. A much simpler method was used for GORs and England because they are made up of whole PCTs. Estimated mean dmft in GORs and England was a weighted average of the mean dmft in the PCTs geographically contained within the GOR or England, using the total five-year-old population (supplied by (BASCD) in the PCT as weights.
y GOR = ∑
Ni ∗ yi N GOR
y GOR is the estimated mean dmft in the GOR
yi
is the mean dmft found in the ith PCT (supplied by BASCD)
Ni is the total number of 5-yr-old children in the ith PCT (supplied by BASCD) NGOR is the total number of 5-yr-old children in the GOR (supplied by BASCD)
Health Profiles 2009 The Indicator Guide
85 Section 3: Children’s and young people’s health
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Excluded.
Disclosure Control
Not applicable
Confidence Intervals calculation method
The 95% confidence intervals were calculated based on the following method: Standard error of mean dmft
SE LAD =
2
dmft i Ni σ i2 ∗ ∗ ∑ dmft N LAD ni PCT
Where the sum is taken over all building blocks in the LAD, and… • SELAD is the estimated standard error of mean dmft in the LAD 2 • σ i is the dmft variance in the ith PCT (supplied by BASCD) • ni is the number of 5-yr-old children sampled (supplied by BASCD) • dmfti is the expected dmft in the ith ‘building block’ (LAD by PCT intersection) • dmftPCT is the expected dmft in the ‘parent’ PCT • Ni is number of residences in the ith ‘building block’ (LAD by PCT intersection) • NLAD is the total number of residences in the LAD The confidence interval for the estimate is: 0.95 ± 1.96* SELAD
Health Profiles 2009 The Indicator Guide
86 Section 3: Children’s and young people’s health
12. TEENAGE PREGNANCY (UNDER 18) INDICATOR Basic Information 1. What is being measured?
Teenage conceptions
2. Why is it being measured?
Teenage parents are prone to poor antenatal health, their babies are more likely to have lower birth weight and are more likely to die in infancy. Teenage mothers are less likely to finish their education, less likely to find a good job, and more likely to end up as single parents or bringing up their children in poverty. Children born to teenage mothers run a much greater risk of poor health and have a much higher chance of becoming teenage mothers themselves.
3. How is this indicator actually defined?
Under-18 conception rate per 1,000 females aged 15–17, 2005–2007 (provisional).
4. Who does it measure?
Females aged under 18 who conceive
5. When does it measure it?
2005–2007
6. Will It measure absolute numbers or proportions?
Measures the rate per 1,000 females aged 15–17
7. Where does the data actually come from?
Teenage Pregnancy Unit & ONS.
8. How accurate and complete will the data be?
Data relating to legal abortions and births is collated through mandatory reporting processes and is of sound data quality.
9. Are there any caveats/warnings/ problems?
Miscarriages and illegal abortions are not included in the conception rates, resulting in rates that may be an under estimation.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
Pg 4 Health Summary – Indicator No 12
Subject category/ domain(s)
Children’s and young people’s health
Indicator name (*Indicator title in health profile)
Teenage conceptions (“Teenage pregnancy [under 18]”)
PHO with lead responsibility
EMPHO
Health Profiles 2009 The Indicator Guide
87 Section 3: Children’s and young people’s health
Date of PHO dataset creation
February 2009
Indicator definition
Under-18 conception rate per 1,000 females aged 15–17 (crude rate) 2005–2007 (provisional).
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs
Timeliness
Data published 14 months after period end, updated annually. (Raw data is available up to 11 months after the event)
Rationale: What this indicator purports to measure
This indicator measures the level of teenage conceptions in the area.
Rationale: Public Health Importance
Teenage pregnancy is a significant public health issue in England. Teenage parents are prone to poor antenatal health, lower birth weight babies and higher infant mortality rates. Teenage mothers are less likely to finish their education, less likely to find a good job, and more likely to end up as single parents or bringing up their children in poverty. Children born to teenage mothers run a much greater risk of poor health and have a much higher chance of becoming teenage mothers themselves. However, it is worth remembering that many young people are successful in adapting to the role of parenthood and have happy, healthy children. For further information see “Teenage Pregnancy: Accelerating the Strategy to 2010” (http://www.everychildmatters.gov.uk/_files/94C1FA 2E9D4C9717E5D0AF1413A329A4.pdf) and Health Statistics Quarterly No. 33 (page 34) (http://www.statistics.gov.uk/downloads/theme_health/ hsq33web.pdf)
Rationale: Purpose behind the inclusion of the indicator
The purpose of including this indicator is to highlight local authorities with high teenage pregnancy rates in order to assess need and enable targeted intervention. In order to assist local areas in reducing under-18 conception rates the following toolkit was produced by DfES. It provides local areas with detailed guidance on the key ingredients that need to be in place locally to reduce under-18 conception rates. www.everychildmatters.gov.uk/resources-and-practice/IG00198/
Rationale: Policy relevance
Teenage conceptions are an important public health target. There is a national PSA target (PSA 14) to reduce teenage conception rates by 50% by 2010 (1998 baseline). See http://www.hm-treasury.gov.uk/d/ pbr_csr07_psa14.pdf for more information. There is an equivalent indicator in the National Indicator Set for Local Authorities and Local Authority Partnerships (NI 112). See http://www. communities.gov.uk/documents/localgovernment/pdf/735125.pdf for more information. This indicator supports Choosing Health and Programme for Action.
Health Profiles 2009 The Indicator Guide
88 Section 3: Children’s and young people’s health
Interpretation: What a high/low level of indicator value means
A high indicator value (red circle) represents a statistically significant higher rate of teenage conceptions for that local authority when compared to the national average. A low indicator value (green circle) represents a statistically significant lower rate of teenage conceptions for that local authority when compared to the national average. Rates that are lower than the England average may still represent a large number of teenage pregnancies and therefore a low indicator value should not be interpreted as meaning that public health action is not needed.
Interpretation: Potential for error due to type of measurement method
Miscarriages and illegal abortions are not included in the conception rates, resulting in rates that may be an under estimation. Data relating to legal abortions and births is collated through mandatory reporting processes and is of sound data quality. Whilst it is acknowledged that the non-inclusion of miscarriages reflects conceptions that are not included in these figures, comparable data in this area is not available. Conception cannot be inferred from the prescription of emergency contraception, and early loss of pregnancy may not be recognised or require medical attention. The omission of this data retains the quality of the indicator rather than limiting it.
Interpretation: Potential for error due to bias and confounding
Teenage pregnancy is often a cause and a consequence of social exclusion. The risk of teenage parenthood is greatest for young people who have grown up in poverty and disadvantage or those with poor educational attainment.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Health Profiles 2009 The Indicator Guide
89 Section 3: Children’s and young people’s health
Table 2 – Indicator Specification Indicator definition: Variable
Estimates of conceptions (excluding pregnancies leading to spontaneous abortion before 24 weeks gestation), based on pregnancies which lead to a delivery at which one or more live or still births occurs and is registered in England and Wales, or termination of pregnancy by abortion under the 1967 Act in England and Wales.
Indicator definition: Statistic
Crude rate
Indicator definition: Gender
Females
Indicator definition: age group
Under 18
Indicator definition: period
2005–2007 (provisional) pooled data
Indicator definition: scale
Per 1,000 female population aged 15–17
Geography: geographies available for this indicator from other providers
ONS publish figures for England and Wales, England, GOR and LA. Available to download from http://www.everychildmatters.gov.uk/resources/IG00200/
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Data is available broken down into under 16 years and under 18 years. Indicator cannot be broken down by ethnicity or socio-economic group.
Data extraction: Source
Produced by the Office for National Statistics and disseminated via the Teenage Pregnancy Unit in the Department for Children, Schools and Families
Data extraction: source URL
Data downloaded from http://www.everychildmatters.gov.uk/resources/ IG00200/ and http://www.statistics.gov.uk/StatBase/Product.asp?vlnk=15055& Pos=1&ColRank=2&Rank=576)
Data extraction: date
Data downloaded on 26th February 2009
Health Profiles 2009 The Indicator Guide
90 Section 3: Children’s and young people’s health
Numerator: definition
The number of conceptions estimated to have occurred to females aged under-18 during 2005–2007. Actual dates of conception are not directly available but have been estimated from a dataset of birth registrations and legal terminations of pregnancy recorded in the respective calendar years plus the following year (i.e. 2005– 2008) as follows: a) For maternities resulting in one or more live births, the date of conception is assumed to be 38 weeks prior to the date of birth (no gestation is recorded at live birth registration); b) For maternities resulting in stillbirths, the date of conception is assumed to be the recorded gestation minus 2 weeks prior to the date of the stillbirth (recorded gestation is time since last menstrual period; conception is assumed to occur 2 weeks after); c) For abortions under the 1967 Act, the date of conception is assumed to be the recorded gestation minus 2 weeks prior to the date of the abortion (recorded gestation is time since last menstrual period; conception is assumed to occur 2 weeks after). The mother’s age at conception is then derived from the mother’s date of birth and the estimated date of conception.
Numerator: source
Produced by the Office for National Statistics and disseminated via the Teenage Pregnancy Unit in the Department for Children, Schools and Families. Data for South Central SHA and South East Coast SHA were subsequently calculated from Local Authority data. Data for the Local Authorities with new (2009) boundaries were also calculated from previous (2008) Local Authority data.
Denominator: definition
Number of females aged 15–17.
Denominator: source
The resident population figures used are ONS revised (that take into account improved estimates of international migration) mid year estimates 2005–2007, as derived from the 2001 census with allowance for subsequent births deaths, migration and ageing of the population. The population numbers were not explicitly presented in the Teenage Pregnancy Unit data, but were back calculated for all areas by using the indicator value and the numerator value. This was with the exception of the South Central SHA and the South Coast SHA, whose populations were subsequently calculated from their constituent local authority populations. Again, Local Authorities with new boundaries had their denominators calculated from pre-2009 Local Authority data. Due to the rounding of the published crude rates, the back-calculated denominators for a geographic area may not equal the sum of the denominators of its constituent areas
Data quality: Accuracy and completeness
Data relating to legal abortions and births is collated through mandatory reporting processes and is of sound data quality.
Health Profiles 2009 The Indicator Guide
91 Section 3: Children’s and young people’s health
TABLE 3 – INDICATOR TECHNICAL METHODS Numerator: extraction
Received directly from the Teenage Pregnancy Unit
Numerator: aggregation/ allocation
No aggregation was required, with the exception of South Central SHA and South East Coast SHA, which were calculated by aggregating from their constituent Local Authorities. Data for the Local Authorities with new (2009) boundaries were also calculated from previous (2008) Local Authority data.
Numerator data caveats
The conception figures exclude any pregnancies leading to spontaneous abortion before 24 weeks gestation and illegal abortions.
Denominator data caveats
In under-18s, a three year age group only (15–17) is used as the denominator, because the vast majority of conceptions to under-18 year olds occur in this age group (95%). To include younger populations would produce misleading results. The 15–17 group is effectively treated as the “population at risk”. As the age at which conception can take place varies from child to child it is impossible to correctly define the population at risk (i.e. data identifying the female population aged 0–17 who are fertile is not available).
Methods used to calculate indicator value
The teenage conceptions indicator is presented as a crude rate. A crude rate is defined as the number of observed events divided by the population-years at risk. For presentation purposes rates are usually multiplied by a scaling factor, in this case 1,000. A rate, r, expressed per 1,000 population is given by:
r=
O × 1,000 n
where: O is the number of observed events (i.e. estimated number of under-18 conceptions, 2005-2007); n is the population-years at risk (i.e. the sum of the mid-year female 15-17 years population estimate for each of the years in the period 2005-2007. As ONS published only the estimated number of under-18 conceptions, and the crude rate per 1,000 population to 1 decimal place, the denominator, the population-years at risk was estimated by back calculating using the above formula.) Small Populations: How Isles of Scilly and City of London populations have been dealt with
Conception data supplied by the ONS does not include data for City of London and Isles of Scilly Local Authorities owing to the small number of events in these small populations. Instead conception figures are combined with those of Hackney LB and Cornwall UA respectively.
Disclosure Control
Teenage conceptions data is subject to disclosure control and any LA with a conception count of fewer than five individuals needs to be suppressed. Because of the size of LA populations however it is only the City of London and Isles of Scilly Local Authorities due to their small population, which are affected. Their conception figures are combined with those of Hackney LB and Cornwall UA respectively.
Health Profiles 2009 The Indicator Guide
92 Section 3: Children’s and young people’s health
Confidence Intervals calculation method
Confidence intervals have been calculated using the Wilson ‘Score’ method for a confidence interval of a proportion as described by Newcombe RG If r is the observed number of subjects with some feature in a sample of size n then the estimated proportion who have the feature is p = r/n. The proportion who do not have the feature is q = 1-p. First, calculate the three quantities 2
A = 2r + z2; B = z z + 4rq ; and C=2(n+z2), where z is z1-α/2, from the standard Normal distribution. Then the confidence interval for the population proportion is given by (A-B)/C to (A+B)/C This method has the considerable advantage that it can be used for any data. When there are no observed events, r and hence p are both zero, and the recommended confidence interval simplifies to 0 to z2/(n+z2). When r = n so that p = 1, the interval becomes n/(n+z2) to 1. Reference Wilson, EB. Probable inference, the law of succession, and statistical inference. Journal of American Statistical Association 1927;22:209-212 quoted in Newcombe, RG. Two-sided confidence intervals for the single proportion: comparison of seven methods. Stat Med 1998;17:857-72.
Health Profiles 2009 The Indicator Guide
93
Section 4: Adult’s Health and Lifestyle
Health Profiles 2009 The Indicator Guide
94 Section 4: adult’s health and lifestyle
13a ADULTS WHO SMOKE UPPER TIER INDICATOR Basic Information 1. What is being measured?
Prevalence of adult smoking.
2. Why is it being measured?
To estimate the prevalence of adult smokers. Smoking prevalence is a direct measure of health care need i.e. the ability to benefit from tobacco control interventions, including smoking cessation services.
3. How is this indicator actually defined?
Prevalence of smoking, percentage of resident population, adults, 2003–2005, persons.
4. Who does it measure?
Adults (aged 16 and over).
5. When does it measure it?
The Health Survey for England (HSE) is carried out annually.
6. Will It measure absolute numbers or proportions?
Percentage of resident adult population aged 16 and over.
7. Where does the data actually come from?
Health Surveys for England, National Centre for Social Research (NatCen). Published by The Information Centre for Health and Social Care (IC), 2007.
8. How accurate and complete will the data be?
The HSE was designed to be representative of the general, non-institutional population living in England. The current “full” sample size of the HSE comprises about 16,000 adults aged 16 and over.
9. Are there any caveats/warnings/ problems?
HSE numerator data are broadly based on observed self-reported current smoking status and as such are subject to responder bias. The Health Survey for England under-samples younger people, people in employment, ethnic minorities, women, those who are healthier but exhibit less healthy behaviour.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
County Health Profiles: Pg 4 Health Summary – Indicator 13
Subject category/ domain(s)
Adults health and lifestyle
Indicator name (*Indicator title in health profile)
Prevalence of adult smoking (*Adults who smoke)
Health Profiles 2009 The Indicator Guide
95 Section 4: adult’s health and lifestyle
PHO with lead responsibility
SEPHO
Date of PHO dataset creation
21/01/2009
Indicator definition
Prevalence of smoking, percentage of resident population, adults, 2003–2005, persons
Geography
England, GOR, County, South East SHAs, new 2009 UAs (Bedford UA, Central Bedfordshire UA, Cheshire East UA, Cheshire West and Chester UA, County Durham, Northumberland UA, Shropshire UA, Wiltshire UA, Cornwall UA).
Timeliness
Updated annually.
Rationale: What this indicator purports to measure
Estimate of smoking prevalence in adults.
Rationale: Public Health Importance
Smoking is the most important cause of preventable ill health and premature mortality in the UK. It is linked to respiratory illness, cancer and coronary heart disease. Smoking not only affects the smoker; over 17,000 children under the age of five are admitted to hospital every year with illnesses resulting from passive smoking. A list of disease specific conditions attributable to smoking is published in The Smoking Epidemic in England, HDA, 2004 http://www.nice.org.uk/page. aspx?o=502811 Smoking is a modifiable lifestyle risk factor; effective tobacco control measures can reduce the prevalence of smoking in the population.
Rationale: Purpose behind the inclusion of the indicator
To help reduce the prevalence of smoking.
Rationale: Policy relevance
Choosing Health: Making healthy choices easier. http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4094550
Smoking prevalence is a direct measure of health care need i.e. the ability to benefit from tobacco control interventions, including smoking cessation services.
Smoking Kills. A White Paper on Tobacco http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4006684 Tackling Health Inequalities: A Programme for Action http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4008268 Interpretation: What a high/low level of indicator value means
A high indicator value (red circle in health summary chart) represents a statistically significant higher level (worse) of estimated adult smoking prevalence when compared to the national value. A low indicator value (green circle in health summary chart) represents a statistically significant lower (better) level of estimated adult smoking prevalence when compared to the national value. However smoking at any prevalence level greater than 0 is undesirable, and therefore a low indicator value should not mean that PH action is not needed.
Health Profiles 2009 The Indicator Guide
96 Section 4: adult’s health and lifestyle
Interpretation: Potential for error due to type of measurement method
Each participant in the Health Survey for England was asked if they currently smoked cigarettes.
Interpretation: Potential for error due to bias and confounding
The Health Survey for England under-samples younger people, people in employment, ethnic minorities, women, those who are healthier but exhibit less healthy behaviour.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate.
Self-reported smoking status may be prone to respondent bias. In order to ensure agreement between these estimates at GOR and SHA level, the lower tier synthetic estimates (districts), which are based on modelled data, have been calibrated. However, the synthetic estimates have not been calibrated to ensure agreement at County level. As a result, there may be inconsistencies between lower tier and county estimates for some areas as the datasets are derived using different methods.
These data have not been age-standardised and, therefore, variation between area values may be a result of differences in population structure.
This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Health Profiles 2009 The Indicator Guide
97 Section 4: adult’s health and lifestyle
Table 2 – Indicator Specification Indicator definition: Variable
Prevalence of smoking. Smoking is defined as self-reported current cigarette smoking.
Indicator definition: Statistic
Percentage of resident adult population aged 16 and over
Indicator definition: Gender
Persons
Indicator definition: age group
Adults (aged 16 and over)
Indicator definition: period
2003–2005
Indicator definition: scale Geography: geographies available for this indicator from other providers
Strategic Health Authority. http://www.ic.nhs.uk/webfiles/Popgeog/Direct%20Estimates%20%20of%20 Obesity%20(adults)%202003-2005.pdf
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Age, gender, ethnicity, social class http://www.dh.gov.uk/en/Publicationsandstatistics/PublishedSurvey/ HealthSurveyForEngland/index.htm
Data extraction: Source
Health Surveys for England, National Centre for Social Research (NatCen). Published by The Information Centre for Health and Social Care (IC), 2007
Data extraction: source URL
Data received directly from NatCen.
Data extraction: date
16 January 2008.
Numerator: definition
The number of persons aged 16+ who are self-reported smokers in a sample survey of the health of the population of England 2003–05.
Numerator: source
Health Survey for England (HSE), commissioned by the Department of Health and carried out by the Joint Health Survey Unit of Social and Community Planning Research and of the Department of Epidemiology and Public Health at University College, London.
Denominator: definition
Total number of respondents (with valid recorded smoking status) aged 16+ in the Health Survey for England 2003–2005.
Denominator: source
Health Survey for England (HSE), commissioned by the Department of Health and carried out by the Joint Health Survey Unit of Social and Community Planning Research and of the Department of Epidemiology and Public Health at University College, London.
Data for 4 new 2009 UAs (Bedford UA, Central Bedfordshire UA, Cheshire East UA and Cheshire West UA) were calculated by NatCen as a separate exercise on 21/01/2009 and received on the same date.
Health Profiles 2009 The Indicator Guide
98 Section 4: adult’s health and lifestyle
Data quality: Accuracy and completeness
The Health Survey for England is designed to provide data at both national and regional level about the population living in private households in England. It uses a clustered, stratified multi-stage sample design. In the 2005 HSE, for example, a random sample of over 7,200 addresses (around 16,000 people) were selected from the Postcode Address File (PAF) to ensure households were sampled proportionately across the nine Government Office Regions in England. 720 postcode sectors were selected and 26 addresses within each sector. Each individual within a selected household was eligible for inclusion. One of the effects of using a complex sample design is that standard errors for survey estimates are generally higher than would be derived from a simple random sample of the same size. There was a full adult sample of around 16,000 in the 2003 HSE. However, the 2004 and 2005 Health Surveys had only 8,000 adults in the normal ‘general population’ sample as these two surveys included boost samples. The 2004 HSE included a boost sample to increase the number of participants from minority ethnic groups and a special Chinese boost sample. The 2005 HSE included a boost sample for older people living in private households and for five months of the year a boost of children aged 2–15 was included. To ensure that each year’s sample was given an approximately equal weight in the calculation of the 3-year estimates (2003–2005) respondents in the boost sample years were weighted up by two. The numerator and denominator counts used to estimate prevalence are based on a sample of the population in each area and, as such, are not true counts. For this reason the numerator and denominator data are not shown in the data sheet.
Table 3 – Indicator Technical Methods Numerator: extraction
Not Applicable
Numerator: aggregation/ allocation
Residency by local authority of each respondent is allocated by postcode of residency.
Numerator data caveats
Questions about current cigarette smoking were asked by the interviewer. For those aged 16 and 17, the questions were asked through a self-completion questionnaire to allow for greater privacy. These data have not been age-standardised and, therefore, variation between area values may be a result of differences in population structure.
Denominator data caveats
The HSE is a series of annual surveys that began in 1991 with the aim of monitoring the health of the population. It was designed to be representative of the general, non-institutional population living in England. The current “full” sample size of the HSE comprises about 16,000 adults aged 16 and over. For each participant, the survey included an interview and a physical examination by a nurse, at which various physical measurements, tests, and samples of blood and saliva were collected. These measurements provided biomedical information about known risk factors associated with disease and objective validation for self-reported health behaviour.
Health Profiles 2009 The Indicator Guide
99 Section 4: adult’s health and lifestyle
Methods used to calculate indicator value
Estimates are based on pooling together three consecutive years of Health Survey for England data (2003-2005). The general population sample size in 2004 and 2005 was about half the sample size in 2001 owing to the sampling of specific population groups – namely, ethnic minority populations (2004) and older people living in private households (2005). To ensure that each year’s sample was given an approximately equal weight in the calculation of the 2003-2005 estimates, respondents in 2004 and 2005 were weighted up by two. Data for 4 new 2009 UAs (Bedford UA, Central Bedfordshire UA, Cheshire East UA and Cheshire West UA) were calculated by NatCen as a separate exercise on 21/01/2009 using the same method as described above. As geographical coverage for the remaining 6 new 2009 UAs (County Durham UA, Northumberland UA, Shropshire UA, Wiltshire UA, Cornwall UA) remains the same as their old County equivalents data for these areas has not been recalculated as part of this exercise.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
The Health Survey for England sample does not cover the Isles of Scilly and no HSE respondents over 2003–2005 were located in the City of London.
Disclosure Control
Not applicable.
Confidence Intervals calculation method
The standard errors, and 95% confidence intervals, have been calculated using STATA’s survey module (the svy:mean commands), further details can be obtained from Shaun Scholes at NatCen ([email protected]). One of the effects of using a complex design is that standard errors for survey estimates are generally higher than the standard errors that would be derived from a simple random sample of the same size.
Health Profiles 2009 The Indicator Guide
100 Section 4: adult’s health and lifestyle
13b ADULTS WHO SMOKE LOWER TIER INDICATOR Basic Information 1. What is being measured?
Estimated prevalence of adult smoking.
2. Why is it being measured?
To estimate the expected proportion of adult smokers in local authorities given the characteristics of local authority populations. Smoking prevalence is a direct measure of health care need i.e. the ability to benefit from tobacco control interventions.
3. How is this indicator actually defined?
Prevalence of smoking, percentage of resident population, adults, 2003–2005, persons.
4. Who does it measure?
Adults (aged 16 and over).
5. When does it measure it?
Updated as ad-hoc.
6. Will It measure absolute numbers or proportions?
Percentage of resident adult population aged 16 and over.
7. Where does the data actually come from?
Modelled by the National Centre for Social Research (NatCen). Published by The Information Centre for Health and Social Care (IC), 2007.
8. How accurate and complete will the data be?
These are modelled estimates based on national survey data. The model is non-aetiological (not based on known casual factors). The estimates do not take into account additional local factors that may impact on the true smoking prevalence rate in an area and may not match with local lifestyle survey results or modelled estimates which use known risk factors such as socio-economic status, age, gender and ethnicity.
9. Are there any caveats/warnings/ problems?
As these estimates are modelled they should be used and interpreted with caution (see above).
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
TABLE 1 – INDICATOR DESCRIPTION Information component
Pg 4 Health Summary – Indicator 13
Subject category/ domain(s)
Adults health and lifestyle
Indicator name (* Indicator title in health profile)
Estimated prevalence of adult smoking (*Adults who smoke)
Health Profiles 2009 The Indicator Guide
101 Section 4: adult’s health and lifestyle
PHO with lead responsibility
SEPHO
Date of PHO dataset creation
21/01/2009
Indicator definition
Prevalence of smoking, percentage of resident population, adults, 2003-2005, persons.
Geography
Local Authority: County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs. Exceptions: Data for the 9 new 2009 UAs (Bedford UA, Central Bedfordshire UA, Cheshire East UA, Cheshire West and Chester UA, County Durham, Northumberland UA, Shropshire UA, Wiltshire UA, Cornwall UA) are not based on modelled estimates but are calculated directly using results from the Health Surveys for England 2003–05.
Timeliness
Updated as ad-hoc.
Rationale: What this indicator purports to measure
Expected prevalence of adult smoking.
Rationale: Public Health Importance
Smoking is the most important cause of preventable ill health and premature mortality in the UK. It is linked to respiratory illness, cancer and coronary heart disease. Smoking not only affects the smoker; over 17,000 children under the age of five are admitted to hospital every year with illnesses resulting from passive smoking. A list of disease specific conditions attributable to smoking is published in The Smoking Epidemic in England, HDA, 2004 http://www.nice.org.uk/page. aspx?o=502811 Smoking is a modifiable lifestyle risk factor; effective tobacco control measures can reduce the prevalence of smoking in the population.
Rationale: Purpose behind the inclusion of the indicator
To estimate the expected proportion of adult smokers in local authorities given the characteristics of local authority populations.
Rationale: Policy relevance
Choosing Health: Making healthy choices easier. http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4094550
Smoking prevalence is a direct measure of health care need i.e. the ability to benefit from tobacco control interventions, including smoking cessation services.
Smoking Kills. A White Paper on Tobacco http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4006684 Tackling Health Inequalities: A Programme for Action http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4008268
Health Profiles 2009 The Indicator Guide
102 Section 4: adult’s health and lifestyle
Interpretation: What a high/low level of indicator value means
Given the characteristics of the local population, a high indicator value (red circle in health summary chart) represents a statistically significant higher (worse) level of expected estimated adult smoking prevalence when compared to the national value. A low indicator value (green circle in health summary chart) represents a statistically significant lower level (better) of expected estimated adult smoking prevalence when compared to the national value. However smoking at any prevalence level greater than 0 is undesirable, and therefore a low indicator value should not mean that PH action is not needed.
Interpretation: Potential for error due to type of measurement method
It is important that users note that these model-based estimates do not take account of any additional local factors that may impact on the true smoking prevalence rate in an area (e.g. local initiatives designed to reduce smoking). They will almost certainly not mirror precisely any available measures from local studies or surveys (although research by NatCen and others have shown that they tend to be related). The figures, therefore, cannot be used to monitor performance or change over time. The model used is a non-aetiological model i.e. is not based on known aetiological risk factors. This may lead to estimated smoking levels which are at odds with, for example, local lifestyle survey results or modelled estimates which use known co-variates (e.g. socio-economic status, age, gender and ethnicity) such as the smoking prevalence estimates modelled in the Health Poverty Index available at www.hpi.org.uk (see variables used in generation of model in calculation of indicator section below). There may also be a discrepancy between the modelled lower tier estimates (districts) and upper tier estimates (County geographies and above, plus new 2009 UA areas) which are based on actual Health Survey for England data. This may lead to inconsistencies between lower tier and county estimates for some areas as the datasets are derived using different methods.
Interpretation: Potential for error due to bias and confounding
These model based healthy lifestyle indicators are derived using the Health Survey for England data and are subject to both sampling and non-sampling error. Sampling errors arise solely as a result of drawing a sample rather than conducting a full survey of the population. Generally, the smaller the sample size the larger the variability in the estimates that one would expect to obtain from all the possible samples. Non-sampling errors arise during the course of the survey activities and there is no simple direct way of estimating the size of these errors. Non-sampling error may include e.g. respondent bias, interview bias and refusal to participate. The use of statistical models for prediction involves making assumptions about relationships in the data. The suitability of the chosen models for the given data and the validity of the model in describing real world dynamics have a bearing on the nature and magnitude of the errors introduced. A key source of modelling error arises from omitting variables that would otherwise help improve the model predictions either by error or because there is no available or reliable data source for them.
Health Profiles 2009 The Indicator Guide
103 Section 4: adult’s health and lifestyle
The model-based estimate generated for a particular area is the expected measure for that area based on its population characteristics – and not an estimate of the actual prevalence. In statistical terms, the model-based estimate is actually a biased estimate of the true value for the area and, as such, should be treated with caution. As mentioned above, the model-based estimates are unable to take account of any additional local factors that may impact on the true prevalence rate. To interpret the estimates, NatCen recommend that users adopt statements such as “given the characteristics of the local population we would expect approximately x% of adults within LA Y to be smokers”. Validation exercises were used to check the appropriateness of the chosen models. Confidence intervals have been calculated for the model-based estimates to capture both sampling and modelling error. The confidence intervals provide a range within which we can be fairly sure the ‘true’ value for that area lies. It is recommended that users look at the confidence interval for the estimates, not just the estimate. Estimates for two areas can only be described as significantly different if the confidence intervals for the estimates do not overlap. Users should also note that the potential sources of bias and error also apply to any ranking or banding of the small-area estimates. NatCen do not encourage any ranking of small area estimates within larger areas such as Local Authorities, Primary Care Organisations and Strategic Health Authorities. The 2003–2005 model based estimates are not comparable with the preceding estimates for 2000–2002 owing to differences in geography and modelling methodology: The 2000–2002 LA estimates were calculated by aggregating the model-based estimates for the component wards. The 2003–05 LA estimates have been calculated by modelling directly at the LA level. The choice of co-variate data was different as both the Index of Multiple deprivation 2004 and ONS area classifications were excluded in the 2003–05 estimates owing to their statistical relationship with other census-based covariates. The 2003–05 based estimates were adjusted to be consistent with the direct survey estimates at GOR/SHA level. The model-based estimates have been produced solely for LAs and cannot be translated onto any other geographical boundary system (i.e. aggregated or averaged over any other spatial unit). As a result of limitations in the model based estimates they are published as “experimental statistics”. This term is applied to any set of ONS statistics that do not meet the rigorous quality standards of National Statistics and/or may be subject to change due to methodological development.
Health Profiles 2009 The Indicator Guide
104 Section 4: adult’s health and lifestyle
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
TABLE 2 – INDICATOR SPECIFICATION Indicator definition: Variable
Prevalence of smoking.
Indicator definition: Statistic
Percentage
Indicator definition: Gender
Persons
Indicator definition: age group
Adults (aged 16 and over)
Indicator definition: period
2003–2005
Indicator definition: scale
Per resident adult population aged 16 and over
Smoking is defined as self-reported current cigarette smoking.
Health Profiles 2009 The Indicator Guide
105 Section 4: adult’s health and lifestyle
Geography: geographies available for this indicator from other providers
MSOA available from http://www.ic.nhs.uk/pubs/healthylifestyles05
Dimensions of inequality: subgroup analyses of this dataset available from other providers
None.
Data extraction: Source
National Centre for Social Research (NatCen).
Data extraction: source URL
Data received directly from NatCen. Also published by IC, December 2007 (see link above).
Data extraction: date
16/01/2008
Numerator: definition
Model estimates by NatCen using data from a number of sources including Health Survey for England 2003–2005, Census 2001.
Numerator: source
NatCen.
Denominator: definition
Not applicable.
Denominator: source
Not applicable.
Data quality: Accuracy and completeness
The model-based approach generates estimates that are of a different nature from standard survey estimates because they are dependent upon how well the relationship between healthy lifestyle behaviours for individuals and the Census/administrative information about the area in which they live are specified in the model. The accuracy and completeness of the information will be subject to the same constraints surrounding the Health Survey for England and Census data sets on which they are based (see Interpretation and potential sources of error section).
Table 3 – Indicator Technical Methods Numerator: extraction
Not applicable
Numerator: aggregation/ allocation
Not applicable
Numerator data caveats
See Interpretation: potential sources of error section.
Denominator data caveats
Not applicable.
Health Profiles 2009 The Indicator Guide
106 Section 4: adult’s health and lifestyle
Methods used to calculate indicator value
The process of creating the model-based estimates of healthy lifestyle behaviours involved three main stages: • A statistical model was used to represent the relationships between current smoking and area-level characteristics in the small areas covered by the HSfE. The 2001 Census provided the main source for demographic and social covariate data. Other routine sources of data providing area-level characteristics for LAs included all age-all cause mortality, diversity index, life expectancy, emergency hospital admissions, hospital admissions attributable to alcohol, job seekers allowance claimant counts and educational attainment. • The model outputs were applied in conjunction with covariate data (available for all LAs) to estimate the ‘expected’ prevalence given the characteristics of the area. • An adjustment factor was applied to ensure that the model-based estimates for each LA corresponded with the 2003–2005 direct estimates at GOR/SHA level taken from the Health Survey for England data. The area level characteristics associated with increased propensity for an adult to be a current smoker were: a higher proportion of residents of White ethnic origin and a higher proportion of residents aged 16–74 whose highest qualification was an NVQ Level 1 (or with no qualifications). The area-level characteristics associated with decreased propensity for an adult to be a current smoker were: a relatively higher proportion of Income Support claimants who were classified as “carers and others”; a higher proportion of residents who were unpaid carers; a higher proportion of household residents aged 16 or over who were living as a couple and a higher proportion of residents aged 70–74. The model-based estimates were constrained to the direct GOR/SHA estimates taken from the Health Survey for England data. This was done by aggregating the model-based estimates to GOR/SHA and comparing to the direct estimates. The relevant ratios of the HSfE direct estimates to the aggregated model-based estimates at GOR/SHA level were then used to scale the model based LA level estimates. However, the model-based estimates have not been calibrated to ensure agreement at County level. As a result, there may be inconsistencies between lower tier and county estimates for some areas as the datasets are derived using different methods. For a fuller technical description of the methodology see the Model-Based Estimates User Guide and other reports available on the Information Centre website: www.ic.nhs.uk/webfiles/Popgeog/Healthy%20Lifestyle%20Behaviours-%20 Model%20Based%20Estimates%20for%20Middle%20Layer%20 Super%20Output%20Areas%20and%20Local%20Authorities%20in%20England_2003-2005__%20User%20Guide.pdf For methods used to calculate data for new 2009 UAs please see metadata for upper tier geographies.
Health Profiles 2009 The Indicator Guide
107 Section 4: adult’s health and lifestyle
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Model based estimates were not produced for Isles of Scilly or City of London.
Disclosure Control
Not applicable.
Confidence Intervals calculation method
The model-based estimate generated for a particular LA is the expected measure for that LA based on its characteristics as measured by the covariates in the model. In statistical terms, the model-based estimate is actually a biased estimate of the true value for an area and, as such, should be treated with caution. By placing confidence intervals around a model-based estimate, however, we can generate a range within which we can be fairly sure the ‘true’ value for that area lies. In order to generate the confidence interval, an estimate of the variance of the difference between the model-based estimate and the true LA measure is required. The estimate of the variance has two components which correspond to the variance that is not explained by the model (and hence is not predicted by the model-based estimate) and the uncertainty of the model-based estimate itself. Obtaining the confidence interval for the LA estimates required computing two area-level variance terms: one for the Primary Sampling Units and one for the LA. The first term was required to allow for the clustering in the sample and the second term to estimate the residual variance at the LA level that was not explained by the model. The confidence interval for the model-based LA estimates was estimated to be: βˆ T X i T e −1 log it αˆ + βˆ X i ± 1.96 T βˆ X i 1+ e
2
M i N ij 2 2 T 2 ˆ ∑ ˆ ˆ σ σ X Var ( β ) X + + i i v N i u j =1
1
where: Χi is the vector of covariates values for LA i,
Xi
βˆ
βˆ is the vector of parameter estimates for the LA-level covariates,
σˆ 2 is the estimate of the between PSU-level variance,
σˆ u2u
σˆ 2 is the estimate of the between LA-level variance.
σˆ v2v
N is the population estimate of the number of adults in LA i,
Ni i
Nijij is the population estimate of the number of adults in postcode sector N M i and
Mi is the number of postcode sectors in LA i.
ij,
2
Health Profiles 2009 The Indicator Guide
108 Section 4: adult’s health and lifestyle
14a BINGE DRINKING ADULTS UPPER TIER INDICATOR Basic Information 1. What is being measured?
Prevalence of adult binge drinking.
2. Why is it being measured?
To estimate the proportion of binge drinking adults in local authorities. To help reduce the prevalence of excessive alcohol consumption and the health risks associated with single episodes of intoxication.
3. How is this indicator actually defined?
Prevalence of binge drinking, percentage of resident population, adults, 2003–2005, persons.
4. Who does it measure?
Adults (aged 16 and over).
5. When does it measure it?
The Health Survey for England (HSE) is carried out annually.
6. Will It measure absolute numbers or proportions?
Percentage of resident adult population aged 16 and over.
7. Where does the data actually come from?
Health Surveys for England, National Centre for Social Research (NatCen). Published by The Information Centre for Health and Social Care (IC), 2007.
8. How accurate and complete will the data be?
The HSE was designed to be representative of the general, non-institutional population living in England. The current “full” sample size of the HSE comprises about 16,000 adults aged 16 and over.
9. Are there any caveats/warnings/ problems?
HSE numerator data are based on observed self-reported drinking behaviour. Self-reported consumption may be prone to respondent bias. The Health Survey for England under-samples younger people, people in employment, ethnic minorities, women, those who are healthier but exhibit less healthy behaviour.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
County Health Profiles: Pg 4 Health Summary – Indicator 14
Subject category/ domain(s)
Adults health and lifestyle
Indicator name (*Indicator title in health profile)
Prevalence of adults who binge drink (*Binge drinking adults).
Health Profiles 2009 The Indicator Guide
109 Section 4: adult’s health and lifestyle
PHO with lead responsibility
SEPHO
Date of PHO dataset creation
21/01/2009
Indicator definition
Prevalence of binge drinking, percentage of resident population, adults, 20032005, persons.
Geography
England, GOR, County, South East SHAs, new 2009 UAs (Bedford UA, Central Bedfordshire UA, Cheshire East UA, Cheshire West and Chester UA, County Durham UA, Northumberland UA, Shropshire UA, Wiltshire UA, Cornwall UA).
Timeliness
Updated annually.
Rationale: What this indicator purports to measure
Prevalence of adult binge drinking.
Rationale: Public Health Importance
Harmful drinking is a significant public health problem in the UK and is associated with a wide range of health problems, including brain damage, alcohol poisoning, chronic liver disease, breast cancer, skeletal muscle damage, mental ill-health and social problems. Alcohol plays a role in many accidents, acts of violence and other instances of criminal behaviour. Nationally between 780,000 and 1.3 million children are affected by their parents’ alcohol misuse. Such children are four times more likely to suffer from a psychiatric disorder by the age of 15 than the national average and are at increased risk of aggressive behaviour, delinquency, hyperactivity and other forms of conduct disorder. There are particular risks associated with drink-driving, alcohol consumption in the workplace or during the working day and drinking during pregnancy. Alcohol-related problems contribute to social and health inequalities, and reducing harmful drinking is one important element in the broad policy thrust to reduce health inequalities following the recommendations of the Acheson Report (1998). For some people, binge drinking is an occasional event. For others, it is part of a chronic drinking pattern Binge drinking and severe intoxication can cause muscular in-coordination, blurred vision, stupor, hypothermia, convulsions, depressed reflexes, respiratory depression, hypotension and coma. Death can occur from respiratory or circulatory failure or if binge drinkers inhale their own vomit. It is well known that binge drinkers are at increased risk of accidents and alcohol poisoning. A growing body of research suggests that binge drinkers also have a higher all-cause mortality rate than those who have the same average alcohol consumption but drink more frequently. Binge drinking is specifically related to accidents and violence, both of which impact on the health service. The Strategy Unit has estimated that, at peak times, up to 70 per cent of all admissions to accident and emergency units are related to alcohol consumption. The total cost of alcohol misuse to the health service is estimated to be in the region of £1.7 billion a year. Effective interventions to reduce alcohol consumption and alcohol related harm exist. The evidence suggests that multi-sectoral, multi-faceted interventions work best. These include: • population level measures such as restricting the availability and price of alcohol, drink driving legislation and taxation; • school based alcohol education programmes; • brief interventions in a variety of settings, such as primary health care, workbased training programmes etc.
Health Profiles 2009 The Indicator Guide
110 Section 4: adult’s health and lifestyle
For further information please see Choosing Health in the South East: Alcohol p35 for summary list of effective interventions. http://www.sepho.org.uk/ Download/Public/10571/1/sepho%20alcohol%20report%20Jan%2007.pdf Rationale: Purpose behind the inclusion of the indicator
To estimate the proportion of binge drinking adults in local authorities.
Rationale: Policy relevance
Choosing Health: Making healthy choices easier. http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4094550
To help reduce the prevalence of excessive alcohol consumption and the health risks associated with single episodes of intoxication.
Alcohol Harm Reduction Strategy for England (2004) http://www.cabinetoffice.gov.uk/upload/assets/www.cabinetoffice.gov.uk/ strategy/caboffce%20alcoholhar.pdf Interpretation: What a high/low level of indicator value means
A high indicator value (red circle in health summary chart) represents a statistically significant higher level (worse) of estimated adult binge drinking prevalence when compared to the national value. A low indicator value (green circle in health summary chart) represents a statistically significant lower (better) level of estimated adult binge drinking prevalence when compared to the national value. However binge drinking at any prevalence level greater than 0 is undesirable, and therefore a low indicator value should not mean that PH action is not needed.
Interpretation: Potential for error due to type of measurement method
HSE numerator data are broadly based on observed self-reported drinking behaviour. Men were defined as having indulged in binge drinking if they had consumed 8 or more units of alcohol on the heaviest drinking day in the previous seven days; for women the cut-off was 6 or more units of alcohol. Self-reported consumption may be prone to respondent bias. In order to ensure agreement between these estimates at GOR and SHA level, the lower tier synthetic estimates (districts), which are based on modelled data, have been calibrated. However, the synthetic estimates have not been calibrated to ensure agreement at County level. As a result, there may be inconsistencies between lower tier and county estimates for some areas as the datasets are derived using different methods.
Interpretation: Potential for error due to bias and confounding
The Health Survey for England under-samples younger people, people in employment, ethnic minorities, women, those who are healthier but exhibit less healthy behaviour. These data have not been age-standardised and, therefore, variation between area values may be a result of differences in population structure.
Health Profiles 2009 The Indicator Guide
111 Section 4: adult’s health and lifestyle
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Table 2 – Indicator Specification Indicator definition: Variable
Prevalence of binge drinking. Binge drinking in adults is defined separately for men and women. Men are defined as having indulged in binge drinking if they had consumed 8 or more units of alcohol on the heaviest drinking day in the previous seven days; for women the cut-off was 6 or more units of alcohol.
Indicator definition: Statistic
Percentage of resident adult population aged 16 and over
Indicator definition: Gender
Persons
Indicator definition: age group
Adults (aged 16 and over)
Indicator definition: period
2003–2005
Indicator definition: scale
Health Profiles 2009 The Indicator Guide
112 Section 4: adult’s health and lifestyle
Geography: geographies available for this indicator from other providers
Strategic Health Authority. http://www.ic.nhs.uk/webfiles/Popgeog/Direct%20Estimates%20%20of%20 Obesity%20(adults)%202003-2005.pdf
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Age, gender, ethnicity, social class http://www.dh.gov.uk/en/Publicationsandstatistics/PublishedSurvey/ HealthSurveyForEngland/index.htm
Data extraction: Source
Health Surveys for England, National Centre for Social Research (NatCen). Published by The Information Centre for Health and Social Care (IC), 2007
Data extraction: source URL
Data received directly from NatCen.
Data extraction: date
16 January 2008. Data for 4 new 2009 UAs (Bedford UA, Central Bedfordshire UA, Cheshire East UA and Cheshire West UA) were calculated by NatCen as a separate exercise on 21/01/2009 and received on the same date.
Numerator: definition
Proportion of adult men who drank 8 or more units of alcohol on the heaviest drinking day in the previous seven days at time of survey and adult women who drank 6 or more units of alcohol on the heaviest drinking day in the previous seven days at time of survey, 2003–2005.
Numerator: source
Health Survey for England (HSE), commissioned by the Department of Health and carried out by the Joint Health Survey Unit of Social and Community Planning Research and of the Department of Epidemiology and Public Health at University College, London.
Denominator: definition
Total number of respondents (with valid measurements on drinking habits in the last week) aged 16+ in the Health Survey for England 2003–2005.
Denominator: source
Health Survey for England (HSE), commissioned by the Department of Health/ IC and carried out by the Joint Health Surveys Unit of the National Centre for Social Research (NatCen) and the Department of Epidemiology and Public Health at the Royal Free and University College Medical School, London.
Health Profiles 2009 The Indicator Guide
113 Section 4: adult’s health and lifestyle
Data quality: Accuracy and completeness
The Health Survey for England is designed to provide data at both national and regional level about the population living in private households in England. It uses a clustered, stratified multi-stage sample design. In the 2005 HSE, for example, a random sample of over 7,200 addresses (around 16,000 people) were selected from the Postcode Address File (PAF) to ensure households were sampled proportionately across the nine Government Office Regions in England. 720 postcode sectors were selected and 26 addresses within each sector. Each individual within a selected household was eligible for inclusion. One of the effects of using a complex sample design is that standard errors for survey estimates are generally higher than would be derived from a simple random sample of the same size. There was a full adult sample of around 16,000 in the 2003 HSE. However, the 2004 and 2005 Health Surveys had only 8,000 adults in the normal ‘general population’ sample as these two surveys included boost samples. The 2004 HSE included a boost sample to increase the number of participants from minority ethnic groups and a special Chinese boost sample. The 2005 HSE included a boost sample for older people living in private households and for five months of the year a boost of children aged 2–15 was included. To ensure that each year’s sample was given an approximately equal weight in the calculation of the 3-year estimates (2003–2005) respondents in the boost sample years were weighted up by two. The numerator and denominator counts used to estimate prevalence are based on a sample of the population in each area and, as such, are not true counts. For this reason the numerator and denominator data are not shown in the data sheet.
Table 3 – Indicator Technical Methods Numerator: extraction
Not applicable.
Numerator: aggregation/ allocation
Residency by local authority of each respondent is allocated by postcode of residency.
Numerator data caveats
These data have not been age-standardised and, therefore, variation between area values may be a result of differences in population structure. The HSE question module concerning “mean weekly alcohol consumption” aims to classify respondents into broad consumption bands based on “usual” behaviour, rather than offer a precise estimate of actual weekly consumption. Adults were first asked how often they drank each of five types of alcoholic drinks (e.g. beer, spirits, wine) in the last 12 months, and how much of each type they had usually drank in one day. From these two sets of questions, an estimated weekly consumption expressed in terms of units of alcohol was derived. Over the years the list of drinks included in the survey has changed to reflect the emergence of new brands and types of drinks. Respondents for whom any information on drinking was not answered, refused or not known were excluded.
Health Profiles 2009 The Indicator Guide
114 Section 4: adult’s health and lifestyle
Denominator data caveats
The HSE is a series of annual surveys that began in 1991 with the aim of monitoring the health of the population. It was designed to be representative of the general, non-institutional population living in England. The current “full” sample size of the HSE comprises about 16,000 adults aged 16 and over. For each participant, the survey included an interview and a physical examination by a nurse, at which various physical measurements, tests, and samples of blood and saliva were collected. These measurements provided biomedical information about known risk factors associated with disease and objective validation for self-reported health behaviour.
Methods used to calculate indicator value
Estimates are based on pooling together three consecutive years of Health Survey for England data (2003–2005). The general population sample size in 2004 and 2005 was about half the sample size in 2001 owing to the sampling of specific population groups – namely, ethnic minority populations (2004) and older people living in private households (2005). To ensure that each year’s sample was given an approximately equal weight in the calculation of the 2003–2005 estimates, respondents in 2004 and 2005 were weighted up by two. Data for 4 new 2009 UAs (Bedford UA, Central Bedfordshire UA, Cheshire East UA and Cheshire West UA) were calculated by NatCen as a separate exercise on 21/01/2009 using the same method as described above. As geographical coverage for the remaining 6 new 2009 UAs (County Durham UA, Northumberland UA, Shropshire UA, Wiltshire UA, Cornwall UA) remains the same as their old County equivalents, data for these areas has not been recalculated as part of this exercise.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
The Health Survey for England sample does not cover the Isles of Scilly and no HSE respondents over 2003–2005 were located in the City of London.
Disclosure Control
Not applicable.
Confidence Intervals calculation method
The standard errors, and 95% confidence intervals, have been calculated using STATA’s survey module (the svy:mean commands), further details can be obtained from Shaun Scholes at NatCen ([email protected]). One of the effects of using a complex design is that standard errors for survey estimates are generally higher than the standard errors that would be derived from a simple random sample of the same size.
Health Profiles 2009 The Indicator Guide
115 Section 4: adult’s health and lifestyle
14b BINGE DRINKING ADULTS LOWER TIER INDICATOR Basic Information 1. What is being measured?
Prevalence of adult binge drinking.
2. Why is it being measured?
To estimate the proportion of binge drinking adults in local authorities. To help reduce the prevalence of excessive alcohol consumption and the health risks associated with single episodes of intoxication.
3. How is this indicator actually defined?
Prevalence of binge drinking, percentage of resident population, adults, 2003–2005, persons.
4. Who does it measure?
Adults (aged 16 and over).
5. When does it measure it?
Updated as ad-hoc.
6. Will It measure absolute numbers or proportions?
Percentage of resident adult population aged 16 and over.
7. Where does the data actually come from?
Modelled by the National Centre for Social Research (NatCen). Published by The Information Centre for Health and Social Care (IC), 2007.
8. How accurate and complete will the data be?
These are modelled estimates based on national survey data. The model is non-aetiological (not based on known casual factors). The estimates do not take into account additional local factors that may impact on the true prevalence of binge drinking in an area and may not match with local lifestyle survey results or modelled estimates which use known risk factors such as socio-economic status, age, gender and ethnicity.
9. Are there any caveats/warnings/ problems?
As these estimates are modelled they should be used and interpreted with caution (see above).
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
Pg 4 Health Summary – Indicator No 14
Subject category/ domain(s)
Adults health and lifestyle
Indicator name (*Indicator title in health profile)
Estimated prevalence of adults who binge drink (*Binge drinking adults).
PHO with lead responsibility
SEPHO
Health Profiles 2009 The Indicator Guide
116 Section 4: adult’s health and lifestyle
Date of PHO dataset creation
21/01/2009
Indicator definition
Prevalence of binge drinking, percentage of resident population, adults, 20032005, persons.
Geography
Local Authority: County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs. Exceptions: Data for the 9 new 2009 UAs (Bedford UA, Central Bedfordshire UA, Cheshire East UA, Cheshire West and Chester UA, County Durham, Northumberland UA, Shropshire UA, Wiltshire UA, Cornwall UA) are not based on modelled estimates but are calculated directly using results from the Health Surveys for England 2003–05.
Timeliness
Updated as ad-hoc.
Rationale: What this indicator purports to measure
Expected prevalence of adult binge drinking.
Health Profiles 2009 The Indicator Guide
117 Section 4: adult’s health and lifestyle
Rationale: Public Health Importance
Harmful drinking is a significant public health problem in the UK and is associated with a wide range of health problems, including brain damage, alcohol poisoning, chronic liver disease, breast cancer, skeletal muscle damage, mental ill-health and social problems. Alcohol plays a role in many accidents, acts of violence and other instances of criminal behaviour. Nationally between 780,000 and 1.3 million children are affected by their parents’ alcohol misuse. Such children are four times more likely to suffer from a psychiatric disorder by the age of 15 than the national average and are at increased risk of aggressive behaviour, delinquency, hyperactivity and other forms of conduct disorder. There are particular risks associated with drink-driving, alcohol consumption in the workplace or during the working day and drinking during pregnancy. Alcohol-related problems contribute to social and health inequalities, and reducing harmful drinking is one important element in the broad policy thrust to reduce health inequalities following the recommendations of the Acheson Report (1998). For some people, binge drinking is an occasional event. For others, it is part of a chronic drinking pattern Binge drinking and severe intoxication can cause muscular in-coordination, blurred vision, stupor, hypothermia, convulsions, depressed reflexes, respiratory depression, hypotension and coma. Death can occur from respiratory or circulatory failure or if binge drinkers inhale their own vomit. It is well known that binge drinkers are at increased risk of accidents and alcohol poisoning. A growing body of research suggests that binge drinkers also have a higher all-cause mortality rate than those who have the same average alcohol consumption but drink more frequently Binge drinking is specifically related to accidents and violence, both of which impact on the health service. The Strategy Unit has estimated that, at peak times, up to 70 per cent of all admissions to accident and emergency units are related to alcohol consumption. The total cost of alcohol misuse to the health service is estimated to be in the region of £1.7 billion a year. Effective interventions to reduce alcohol consumption and alcohol related harm exist. The evidence suggests that multi-sectoral, multi-faceted interventions work best. These include: • population level measures such as restricting the availability and price of alcohol, drink driving legislation and taxation; • school based alcohol education programmes; • brief interventions in a variety of settings, such as primary health care, workbased training programmes etc. For further information please see Choosing Health in the South East: Alcohol p35 for summary list of effective interventions. http://www.sepho.org.uk/ Download/Public/10571/1/sepho%20alcohol%20report%20Jan%2007.pdf
Rationale: Purpose behind the inclusion of the indicator
To estimate the expected proportion of binge drinking adults in local authorities given the characteristics of local authority populations.
Rationale: Policy relevance
Choosing Health: Making healthy choices easier. http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4094550
To help reduce the prevalence of excessive alcohol consumption and the health risks associated with single episodes of intoxication.
Alcohol Harm Reduction Strategy for England (2004) http://www.cabinetoffice.gov.uk/upload/assets/www.cabinetoffice.gov.uk/ strategy/caboffce%20alcoholhar.pdf
Health Profiles 2009 The Indicator Guide
118 Section 4: adult’s health and lifestyle
Interpretation: What a high/low level of indicator value means
Given the characteristics of the local population, a high indicator value (red circle in health summary chart) represents a statistically significant higher (worse) level of expected estimated binge drinking prevalence when compared to the national value. A low indicator value (green circle in health summary chart) represents a statistically significant lower level (better) of expected estimated adult binge drinking prevalence when compared to the national value. However binge drinking at any prevalence level greater than 0 is undesirable, and therefore a low indicator value should not mean that PH action is not needed.
Interpretation: Potential for error due to type of measurement method
It is important that users note that these model-based do not take account of any additional local factors that may impact on the true binge drinking prevalence rate in an area (e.g. local initiatives designed to reduce alcohol consumption). They will almost certainly not mirror precisely any available measures from local studies or surveys (although research by NatCen and others have shown that they tend to be related). The figures, therefore, cannot be used to monitor performance or change over time. The model used is a non-aetiological model i.e. is not based on risk factors such as physical activity levels and calorie intake. This may lead to estimated binge drinking levels which are at odds with other risk factors estimates such as healthy eating and physical activity; local lifestyle survey results and modelled estimates which use known co-variates (see variables used in generation of model in calculation of indicator section below). There may also be a discrepancy between the modelled lower tier estimates (districts) and upper tier estimates (County geographies and above, plus new 2009 UA areas) which are based on actual Health Survey for England data. This has led to inconsistencies between lower tier and county estimates for some areas as the datasets are derived using different methods. It should be noted that the North West Public Health Observatory (NWPHO) have recently generated LA estimates of hazardous and harmful drinking using Health Survey for England data. These estimates are not comparable with the NatCen modelled estimates owing to differences in time period, choice of alcohol indicator, modelling methodology and covariate data. See http://www. nwph.net/alcohol/lape/ for the NWPHO profiles and further information.
Interpretation: Potential for error due to bias and confounding
These model based healthy lifestyle indicators are derived using the Health Survey for England data and are subject to both sampling and non-sampling error. Sampling errors arise solely as a result of drawing a sample rather than conducting a full survey of the population. Generally, the smaller the sample size the larger the variability in the estimates that one would expect to obtain from all the possible samples.
Health Profiles 2009 The Indicator Guide
119 Section 4: adult’s health and lifestyle
Non-sampling errors arise during the course of the survey activities and there is no simple direct way of estimating the size of these errors. Non sampling error may include e.g. respondent bias, interview bias and refusal to participate. The use of statistical models for prediction involves making assumptions about relationships in the data. The suitability of the chosen models for the given data and the validity of the model in describing real world dynamics have a bearing on the nature and magnitude of the errors introduced. A key source of modelling error arises from omitting variables that would otherwise help improve the model predictions either by error or because there is no available or reliable data source for them. The model-based estimate generated for a particular area is the expected measure for that area based on its population characteristics - and not an estimate of the actual prevalence. In statistical terms, the model-based estimate is actually a biased estimate of the true value for the area and, as such, should be treated with caution. As mentioned above, the model-based estimates are unable to take account of any additional local factors that may impact on the true prevalence rate. To interpret the estimates, NatCen recommend that users adopt statements such as “given the characteristics of the local population we would expect approximately x% of adults within LA Y to indulge in binge drinking”. Validation exercises were used to check the appropriateness of the chosen models. Confidence intervals have been calculated for the model-based estimates to capture both sampling and modelling error. The confidence intervals provide a range within which we can be fairly sure the ‘true’ value for that area lies. It is recommended that users look at the confidence interval for the estimates, not just the estimate. Estimates for two areas can only be described as significantly different if the confidence intervals for the estimates do not overlap. Users should also note that the potential sources of bias and error also apply to any ranking or banding of the small-area estimates. NatCen do not encourage any ranking of small area estimates within larger areas such as Local Authorities, Primary Care Organisations and Strategic Health Authorities. • The 2003–2005 model based estimates are not comparable with the preceding estimates for 2000–2002 owing to differences in geography and modelling methodology: • The 2000–2002 LA estimates were calculated by aggregating the modelbased estimates for the component wards. The 2003–05 LA estimates have been calculated by modelling directly at the LA level. • The choice of co-variate data was different as both the Index of Multiple deprivation 2004 and ONS area classifications were excluded in the 2003– 05 estimates owing to their statistical relationship with other census-based covariates. • The 2003–05 based estimates were adjusted to be consistent with the direct survey estimates at GOR/SHA level. The model-based estimates have been produced solely for LAs and cannot be translated onto any other geographical boundary system (i.e. aggregated or averaged over any other spatial unit). As a result of limitations in the model based estimates they are published as “experimental statistics”. This term is applied to any set of ONS statistics that do not meet the rigorous quality standards of National Statistics and/or may be subject to change due to methodological development.
Health Profiles 2009 The Indicator Guide
120 Section 4: adult’s health and lifestyle
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
TABLE 2 – INDICATOR SPECIFICATION Indicator definition: Variable
Prevalence of binge drinking. Binge drinking in adults is defined separately for men and women. Men are defined as having indulged in binge drinking if they had consumed 8 or more units of alcohol on the heaviest drinking day in the previous seven days; for women the cut-off was 6 or more units of alcohol.
Indicator definition: Statistic
Percentage of resident adult population aged 16 and over
Indicator definition: Gender
Persons
Indicator definition: age group
Adults (aged 16 and over)
Indicator definition: period
2003–2005
Indicator definition: scale
Health Profiles 2009 The Indicator Guide
121 Section 4: adult’s health and lifestyle
Geography: geographies available for this indicator from other providers
MSOA available from http://www.ic.nhs.uk/pubs/healthylifestyles05
Dimensions of inequality: subgroup analyses of this dataset available from other providers
None.
Data extraction: Source
National Centre for Social Research (NatCen).
Data extraction: source URL
Data received directly from NatCen. Also published by IC, December 2007 (see link above).
Data extraction: date
16/01/2008
Numerator: definition
Model estimates by NatCen using data from a number of sources including Health Survey for England 2003–2005, Census 2001.
Numerator: source
NatCen.
Denominator: definition
Not applicable.
Denominator: source
Not applicable.
Data quality: Accuracy and completeness
The model-based approach generates estimates that are of a different nature from standard survey estimates because they are dependent upon how well the relationship between healthy lifestyle behaviours for individuals and the Census/administrative information about the area in which they live are specified in the model. The accuracy and completeness of the information will be subject to the same constraints surrounding the Health Survey for England and Census data sets on which they are based (see Interpretation: potential sources of error section).
TABLE 3 – INDICATOR TECHNICAL METHODS Numerator: extraction
Not applicable.
Numerator: aggregation/ allocation
Not applicable.
Numerator data caveats
See Interpretation: potential sources of error section.
Denominator data caveats
Not applicable.
Health Profiles 2009 The Indicator Guide
122 Section 4: adult’s health and lifestyle
Methods used to calculate indicator value
The process of creating the model-based estimates of healthy lifestyle behaviours involved three main stages: • A statistical model was used to represent the relationships between binge drinking and area-level characteristics in the small areas covered by the HSfE. The 2001 Census provided the main source for demographic and social covariate data. Other routine sources of data providing area-level characteristics for LAs included all age-all cause mortality, diversity index, life expectancy, emergency hospital admissions, hospital admissions attributable to alcohol, job seekers allowance claimant counts and educational attainment. • The model outputs were applied in conjunction with covariate data (available for all LAs) to estimate the ‘expected’ prevalence given the characteristics of the area. • An adjustment factor was applied to ensure that the model-based estimates for each LA corresponded with the 2003–2005 direct estimates at GOR/SHA level taken from the Health Survey for England data. The model contained an interaction between the percentage of residents aged 20–24 and the percentage of households headed by a single-parent. An interaction term in a statistical model arises were the effect of a covariate on a healthy lifestyle behaviour depends on the values of another covariate. The area level characteristics associated with increased propensity for an adult to indulge in binge drinking were: a higher proportion of residents aged 25– 29; a higher proportion of residents aged 20–24; a higher proportion of male residents aged 55–59 and a higher proportion of residents of White ethnic origin. The area-level characteristics associated with decreased propensity for an adult to indulge in binge drinking were: a higher proportion of female residents aged 55–59; and a higher proportion of residents residing in areas classified as 1) Town and Fringe or 2) Village, Hamlet or Isolated Dwelling (relative to living in urban areas containing more than 10,000 residents). The model-based estimates were constrained to the direct GOR/SHA estimates taken from the Health Survey for England data. This was done by aggregating the model-based estimates to GOR/SHA and comparing to the direct estimates. The relevant ratios of the HSfE direct estimates to the aggregated model-based estimates at GOR/SHA level were then used to scale the model based LA level estimates. However, the model-based estimates have not been calibrated to ensure agreement at County level. As a result, there may be inconsistencies between lower tier and county estimates for some areas as the datasets are derived using different methods. For a fuller technical description of the methodology see the Model-Based Estimates User Guide and other reports available on the Information Centre website: http://www.ic.nhs.uk/webfiles/Popgeog/Healthy%20Lifestyle%20 Behaviours-%20Model%20Based%20Estimates%20for%20Middle%20 Layer%20Super%20Output%20Areas%20and%20Local%20Authorities%20 in%20England_2003-2005__%20User%20Guide.pdf For methods used to calculate data for new 2009 UAs please see metadata for upper tier geographies.
Health Profiles 2009 The Indicator Guide
123 Section 4: adult’s health and lifestyle
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Model based estimates were not produced for Isles of Scilly or City of London.
Disclosure Control
Not applicable.
Confidence Intervals calculation method
The model-based estimate generated for a particular LA is the expected measure for that LA based on its characteristics as measured by the covariates in the model. In statistical terms, the model-based estimate is actually a biased estimate of the true value for an area and, as such, should be treated with caution. By placing confidence intervals around a model-based estimate, however, we can generate a range within which we can be fairly sure the ‘true’ value for that area lies. In order to generate the confidence interval, an estimate of the variance of the difference between the model-based estimate and the true LA measure is required. The estimate of the variance has two components which correspond to the variance that is not explained by the model (and hence is not predicted by the model-based estimate) and the uncertainty of the model-based estimate itself. Obtaining the confidence interval for the LA estimates required computing two area-level variance terms: one for the Primary Sampling Units and one for the LA. The first term was required to allow for the clustering in the sample and the second term to estimate the residual variance at the LA level that was not explained by the model. The confidence interval for the model-based LA estimates was estimated to be:
βˆ T X i T e −1 ˆ ˆ log it α + β X i ± 1.96 T βˆ X i 1+ e
2
N ij 2 T 2 ˆ ∑ σˆ + σˆ v + X i Var ( β ) X i N i u j =1 Mi
2
1
where: Χi is the vector of covariates values for LA i,
Xi
β
βˆ is the vector of parameter estimates for the LA-level covariates, ˆ
σˆ2u2 is the estimate of the between PSU-level variance,
σˆ u
σˆ 2 is the estimate of the between LA-level variance.
σˆ v2v
N is the population estimate of the number of adults in LA i,
Ni i
Nijij is the population estimate of the number of adults in postcode sector N M i and
Mi is the number of postcode sectors in LA i.
ij,
2
Health Profiles 2009 The Indicator Guide
124 Section 4: adult’s health and lifestyle
15a HEALTHY EATING ADULTS UPPER TIER INDICATOR Basic Information 1. What is being measured?
Prevalence of adult healthy eating, that is adults who consume 5 or more portions of fruit and vegetables per day.
2. Why is it being measured?
To estimate the proportion of healthy eating adults. There is evidence that eating at least 5 portions of a variety of fruit and vegetables a day could reduce the risk of deaths from chronic diseases such as heart disease, stroke, and cancer by up to 20%.
3. How is this indicator actually defined?
Prevalence of healthy eating, percentage of resident population, adults, 2003–2005, persons.
4. Who does it measure?
Adults (aged 16 and over).
5. When does it measure it?
The Health Survey for England (HSE) is carried out annually.
6. Will It measure absolute numbers or proportions?
Percentage of resident adult population aged 16 and over.
7. Where does the data actually come from?
Health Surveys for England, National Centre for Social Research (NatCen). Published by The Information Centre for Health and Social Care (IC), 2007.
8. How accurate and complete will the data be?
The HSE was designed to be representative of the general, non-institutional population living in England. The current “full” sample size of the HSE comprises about 16,000 adults aged 16 and over.
9. Are there any caveats/warnings/ problems?
HSE numerator data are broadly based on observed self-reported daily consumption of fruit and vegetables. Self-reported consumption may be prone to respondent bias. The Health Survey for England under-samples younger people, people in employment, ethnic minorities, women, those who are healthier but exhibit less healthy behaviour.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Health Profiles 2009 The Indicator Guide
125 Section 4: adult’s health and lifestyle
Table 1 – Indicator Description Information component
County Health Profiles: Pg 4 Health Summary – Indicator 15
Subject category/ domain(s)
Adults health and lifestyle
Indicator name (*Indicator title in health profile)
Prevalence of adults who eat healthily (*Healthy eating adults).
PHO with lead responsibility
SEPHO
Date of PHO dataset creation
21/01/2008
Indicator definition
Prevalence of healthy eating, percentage of resident population, adults, 20032005, persons.
Geography
England, GOR, County, South East SHAs, new 2009 UAs (Bedford UA, Central Bedfordshire UA, Cheshire East UA, Cheshire West and Chester UA, County Durham, Northumberland UA, Shropshire UA, Wiltshire UA, Cornwall UA).
Timeliness
Updated annually.
Rationale: What this indicator purports to measure
Prevalence of adult healthy eating, that is, adults who consume 5 or more portions of fruit and vegetables per day.
Rationale: Public Health Importance
The indicator is a measure of a protective lifestyle factor. A diet rich in fruit and vegetables confers protective effects against the development of heart disease and certain cancers. It has been estimated that eating at least 5 portions of a variety of fruit and vegetables a day could reduce the risk of deaths from chronic diseases such as heart disease, stroke, and cancer by up to 20%. It has been estimated that diet might contribute to the development of onethird of all cancers, and that increasing fruit and vegetable consumption is the second most important cancer prevention strategy, after reducing smoking. In 1998, the Department of Health’s Committee on Medical Aspects of Food Policy and Nutrition reviewed the evidence and concluded that higher vegetable consumption would reduce the risk of colorectal cancer and gastric cancer. There was also weakly consistent evidence that higher fruit and vegetable consumption would reduce the risk of breast cancer. These cancers combined represent about 18% of the cancer burden in men and about 30% in women. Higher consumption of fruit and vegetables also reduces the risk of coronary heart disease and stroke. A recent study found that each increase of 1 portion of fruit and vegetables a day lowered the risk of coronary heart disease by 4% and the risk of stroke by 6%. Evidence also suggests an increase in fruit and vegetable intake can help lower blood pressure. Research suggests that there are other health benefits, including delaying the development of cataracts, reducing the symptoms of asthma, improving bowel function, and helping to manage diabetes. As well as the direct health benefits, eating fruit and vegetables can help to achieve other dietary goals including increasing fibre intake, reducing fat intake, help maintain a healthy weight, and substituting for foods with added sugars (as frequent consumption of foods with added sugars can contribute to tooth decay).
Health Profiles 2009 The Indicator Guide
126 Section 4: adult’s health and lifestyle
Rationale: Purpose behind the inclusion of the indicator
To estimate the proportion of adults who consume 5 or more portions of fruit and vegetables per day in local authorities.
Rationale: Policy relevance
Choosing Health: Making healthy choices easier. http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4094550
To help increase the prevalence of healthy eating and the health benefits associated with eating a healthy diet.
Department of Health National 5 A Day programme http://www.dh.gov.uk/en/Policyandguidance/Healthandsocialcaretopics/ FiveADay/FiveADaygeneralinformation/index.htm The NHS Plan http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4002960 The NHS Cancer Plan http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4009609 National Service Framework for Coronary Heart Disease http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4094275 National Service Framework for Diabetes http://www.dh.gov.uk/en/Policyandguidance/Healthandsocialcaretopics/ Diabetes/DH_4015717 National Service Framework for Older People http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4003066 Interpretation: What a high/low level of indicator value means
A high indicator value (green circle in health summary chart) represents a statistically significant higher level (better) of adults who are estimated to consume 5 or more portions of fruit and vegetables per day when compared to the national value. A low indicator value (red circle in health summary chart) represents a statistically significant lower (worse) level of adults who are estimated to consume 5 or more portions of fruit and vegetables per day when compared to the national value.
Interpretation: Potential for error due to type of measurement method
HSE numerator data are broadly based on observed self-reported daily consumption of fruit and vegetables. Self-reported consumption may be prone to respondent bias. There may be variation in how informants defined and reported the amount of food consumed. Although everyday measures were used to help informants to define how much they had consumed, this task may have been difficult for certain food items, such as fruit in composite foods like apple pie. In order to ensure agreement between these estimates at GOR and SHA level, the lower tier synthetic estimates (districts), which are based on modelled data, have been calibrated. However, the synthetic estimates have not been calibrated to ensure agreement at County level. As a result, there may be inconsistencies between lower tier and county estimates for some areas as the datasets are derived using different methods.
Health Profiles 2009 The Indicator Guide
127 Section 4: adult’s health and lifestyle
Interpretation: Potential for error due to bias and confounding
The Health Survey for England under-samples younger people, people in employment, ethnic minorities, women, those who are healthier but exhibit less healthy behaviour.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate.
These data have not been age-standardised and, therefore, variation between area values may be a result of differences in population structure.
This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
TABLE 2 – INDICATOR SPECIFICATION Indicator definition: Variable
Prevalence of healthy eating.
Indicator definition: Statistic
Percentage of resident adult population aged 16 and over
Indicator definition: Gender
Persons
Indicator definition: age group
Adults (aged 16 and over)
Indicator definition: period
2003–2005
Healthy eating is defined as those who consume 5 or more portions of fruit and vegetables per day.
Health Profiles 2009 The Indicator Guide
128 Section 4: adult’s health and lifestyle
Indicator definition: scale Geography: geographies available for this indicator from other providers
Strategic Health Authority. http://www.ic.nhs.uk/webfiles/Popgeog/Direct%20Estimates%20%20of%20 Obesity%20(adults)%202003-2005.pdf
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Age, gender, ethnicity, social class http://www.dh.gov.uk/en/Publicationsandstatistics/PublishedSurvey/ HealthSurveyForEngland/index.htm
Data extraction: Source
Health Surveys for England, National Centre for Social Research (NatCen). Published by The Information Centre for Health and Social Care (IC), 2007
Data extraction: source URL
Data received directly from NatCen.
Data extraction: date
16 January 2008. Data for 4 new 2009 UAs (Bedford UA, Central Bedfordshire UA, Cheshire East UA and Cheshire West UA) were calculated by NatCen as a separate exercise on 21/01/2009 and received on the same date.
Numerator: definition
Proportion of adults who reported consumption of 5 or more portions of fruit and vegetables per day at the time of the survey, 2003–2005. A portion of fruit or vegetables was defined as an 80g serving.
Numerator: source
Health Survey for England (HSE), commissioned by the Department of Health/ IC and carried out by the Joint Health Surveys Unit of the National Centre for Social Research (NatCen) and the Department of Epidemiology and Public Health at the Royal Free and University College Medical School, London.
Denominator: definition
Total number of respondents (with valid record of consumption of fruit and vegetables on the day before the interview) aged 16+ in the Health Survey for England 2003–2005.
Denominator: source
Health Survey for England (HSE), commissioned by the Department of Health/ IC and carried out by the Joint Health Surveys Unit of the National Centre for Social Research (NatCen) and the Department of Epidemiology and Public Health at the Royal Free and University College Medical School, London.
Health Profiles 2009 The Indicator Guide
129 Section 4: adult’s health and lifestyle
Data quality: Accuracy and completeness
The Health Survey for England is designed to provide data at both national and regional level about the population living in private households in England. It uses a clustered, stratified multi-stage sample design. In the 2005 HSE, for example, a random sample of over 7,200 addresses (around 16,000 people) were selected from the Postcode Address File (PAF) to ensure households were sampled proportionately across the nine Government Office Regions in England. 720 postcode sectors were selected and 26 addresses within each sector. Each individual within a selected household was eligible for inclusion. One of the effects of using a complex sample design is that standard errors for survey estimates are generally higher than would be derived from a simple random sample of the same size. There was a full adult sample of around 16,000 in the 2003 HSE. However, the 2004 and 2005 Health Surveys had only 8,000 adults in the normal ‘general population’ sample as these two surveys included boost samples. The 2004 HSE included a boost sample to increase the number of participants from minority ethnic groups and a special Chinese boost sample. The 2005 HSE included a boost sample for older people living in private households and for five months of the year a boost of children aged 2–15 was included. To ensure that each year’s sample was given an approximately equal weight in the calculation of the 3-year estimates (2003–2005) respondents in the boost sample years were weighted up by two. The numerator and denominator counts used to estimate prevalence are based on a sample of the population in each area and, as such, are not true counts. For this reason the numerator and denominator data are not shown in the data sheet.
Table 3 – Indicator Technical Methods Numerator: extraction
Not Applicable
Numerator: aggregation/ allocation
Residency by local authority of each respondent is allocated by postcode of residency.
Numerator data caveats
These data have not been age-standardised and, therefore, variation between area values may be a result of differences in population structure. The definition of portion size for pulses and very small fruits was changed in 2002, as studies of foods in these categories indicated that an 80g portion is larger than that defined in HSE 2001. It is likely that HSE 2001 overestimated the consumption of pulses and very small fruits. From 2002 onwards, a portion of pulses was defined as 3 tablespoons (rather than 2) and a portion of very small fruit as 2 handfuls (rather than 1).
Denominator data caveats
The HSE is a series of annual surveys that began in 1991 with the aim of monitoring the health of the population. It was designed to be representative of the general, non-institutional population living in England. The current “full” sample size of the HSE comprises about 16,000 adults aged 16 and over. For each participant, the survey included an interview and a physical examination by a nurse, at which various physical measurements, tests, and samples of blood and saliva were collected. These measurements provided biomedical information about known risk factors associated with disease and objective validation for self-reported health behaviour.
Health Profiles 2009 The Indicator Guide
130 Section 4: adult’s health and lifestyle
Methods used to calculate indicator value
Questions about fruit and vegetable consumption have been included in the Health Survey for England since 2001. The questions are interviewer administered as part of the Computer Assisted Personal Interview (CAPI). Questions are designed to assess fruit and vegetable consumption and focus on consumption on the day before the interview, which was defined as the 24 hours from midnight to midnight. This time period was selected to ensure that variations in informant work patterns and times of meals did not affect the average measure of daily consumption. Fruit and vegetable consumption is measured in portions per day, where a portion is defined as an 80g serving. A range of foods, including fruit, vegetables, pulses, salads and fruit juice contribute to the total number of portions consumed. Portion size was translated into everyday measures to help informants to report how much they had consumed. For example, informants were asked how many tablespoons of vegetables, cereal bowls of salad, or pieces of medium sized fruit (such as apples) they had consumed in the previous 24 hours. These everyday measures were converted back to portions prior to analysis. The table below shows portion sizes for the different food items included in the questionnaire.
Estimates are based on pooling together three consecutive years of Health Survey for England data (2003–2005). The general population sample size in 2004 and 2005 was about half the sample size in 2003 owing to the sampling of specific population groups – namely, ethnic minority populations (2004) and older people living in private households (2005). To ensure that each year’s sample was given an approximately equal weight in the calculation of the 2003– 2005 estimates, respondents in 2004 and 2005 were weighted up by two. Data for 4 new 2009 UAs (Bedford UA, Central Bedfordshire UA, Cheshire East UA and Cheshire West UA) were calculated by NatCen as a separate exercise on 21/01/2009 using the same method as described above. As geographical coverage for the remaining 6 new 2009 UAs (County Durham UA, Northumberland UA, Shropshire UA, Wiltshire UA, Cornwall UA) remains the same as their old County equivalents, data for these areas has not been recalculated as part of this exercise.
Health Profiles 2009 The Indicator Guide
131 Section 4: adult’s health and lifestyle
Small Populations: How Isles of Scilly and City of London populations have been dealt with
The Health Survey for England sample does not cover the Isles of Scilly and no HSE respondents over 2003–2005 were located in the City of London.
Disclosure Control
Not applicable.
Confidence Intervals calculation method
The standard errors, and 95% confidence intervals, have been calculated using STATA’s survey module (the svy:mean commands), further details can be obtained from Shaun Scholes at NatCen ([email protected]). One of the effects of using a complex design is that standard errors for survey estimates are generally higher than the standard errors that would be derived from a simple random sample of the same size.
Health Profiles 2009 The Indicator Guide
132 Section 4: adult’s health and lifestyle
15b. HEALTHY EATING ADULTS LOWER TIER INDICATOR BASIC INFORMATION 1. What is being measured?
Estimated prevalence of adult healthy eating, that is adults who consume 5 or more portions of fruit and vegetables per day
2. Why is it being measured?
To estimate the proportion of healthy eating adults. There is evidence that eating at least 5 portions of a variety of fruit and vegetables a day could reduce the risk of deaths from chronic diseases such as heart disease, stroke, and cancer by up to 20%.
3. How is this indicator actually defined?
Prevalence of healthy eating, percentage of resident population, adults, 2003–2005, persons.
4. Who does it measure?
Adults (aged 16 and over).
5. When does it measure it?
Updated as ad-hoc.
6. Will It measure absolute numbers or proportions?
Percentage of resident adult population aged 16 and over
7. Where does the data actually come from?
Modelled by the National Centre for Social Research (NatCen). Published by The Information Centre for Health and Social Care (IC), 2007.
8. How accurate and complete will the data be?
These are modelled estimates based on national survey data. The model is non-aetiological (not based on known casual factors). The estimates do not take into account additional local factors that may impact on the true prevalence of healthy eating in an area and may not match with local lifestyle survey results or modelled estimates which use known risk factors such as socio-economic status, age, gender and ethnicity.
9. Are there any caveats/warnings/ problems?
As these estimates are modelled they should be used and interpreted with caution (see above).
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
Pg 4 Health Summary – Indicator No 15
Subject category/ domain(s)
Adults health and lifestyle
Indicator name (*Indicator title in health profile)
Estimated prevalence of adults who eat healthily (*Healthy eating adults)
Health Profiles 2009 The Indicator Guide
133 Section 4: adult’s health and lifestyle
PHO with lead responsibility
SEPHO
Date of PHO dataset creation
21/01/2009
Indicator definition
Prevalence of healthy eating, percentage of resident population, adults, 20032005, persons.
Geography
Local Authority: County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs. Exceptions: Data for the 9 new 2009 UAs (Bedford UA, Central Bedfordshire UA, Cheshire East UA, Cheshire West and Chester UA, County Durham, Northumberland UA, Shropshire UA, Wiltshire UA, Cornwall UA) are not based on modelled estimates but are calculated directly using results from the Health Surveys for England 2003–05.
Timeliness
Updated as ad-hoc.
Rationale: What this indicator purports to measure
Expected prevalence of adult healthy eating that is adults who consume 5 or more portions of fruit and vegetables per day.
Rationale: Public Health Importance
The indicator is a measure of a protective lifestyle factor. A diet rich in fruit and vegetables confers protective effects against the development of heart disease and certain cancers. It has been estimated that eating at least 5 portions of a variety of fruit and vegetables a day could reduce the risk of deaths from chronic diseases such as heart disease, stroke, and cancer by up to 20%. It has been estimated that diet might contribute to the development of onethird of all cancers, and that increasing fruit and vegetable consumption is the second most important cancer prevention strategy, after reducing smoking. In 1998, the Department of Health’s Committee on Medical Aspects of Food Policy and Nutrition reviewed the evidence and concluded that higher vegetable consumption would reduce the risk of colorectal cancer and gastric cancer. There was also weakly consistent evidence that higher fruit and vegetable consumption would reduce the risk of breast cancer. These cancers combined represent about 18% of the cancer burden in men and about 30% in women. Higher consumption of fruit and vegetables also reduces the risk of coronary heart disease and stroke. A recent study found that each increase of 1 portion of fruit and vegetables a day lowered the risk of coronary heart disease by 4% and the risk of stroke by 6%. Evidence also suggests an increase in fruit and vegetable intake can help lower blood pressure. Research suggests that there are other health benefits, including delaying the development of cataracts, reducing the symptoms of asthma, improving bowel function, and helping to manage diabetes. As well as the direct health benefits, eating fruit and vegetables can help to achieve other dietary goals including increasing fibre intake, reducing fat intake, help maintain a healthy weight, and substituting for foods with added sugars (as frequent consumption of foods with added sugars can contribute to tooth decay).
Health Profiles 2009 The Indicator Guide
134 Section 4: adult’s health and lifestyle
Rationale: Purpose behind the inclusion of the indicator
To estimate the expected proportion of adults who 5 or more portions of fruit and vegetables per day in local authorities given the characteristics of local authority populations.
Rationale: Policy relevance
Choosing Health: Making healthy choices easier. http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4094550
To help increase the prevalence of healthy eating and the health benefits associated with eating a healthy diet.
Department of Health National 5 A Day programme http://www.dh.gov.uk/en/Policyandguidance/Healthandsocialcaretopics/ FiveADay/FiveADaygeneralinformation/index.htm The NHS Plan http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4002960 The NHS Cancer Plan http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4009609 National Service Framework for Coronary Heart Disease http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4094275 National Service Framework for Diabetes http://www.dh.gov.uk/en/Policyandguidance/Healthandsocialcaretopics/ Diabetes/DH_4015717 National Service Framework for Older People http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4003066 Interpretation: What a high/low level of indicator value means
Given the characteristics of the local population, a high indicator value (green circle in health summary chart) represents a statistically significant higher (better) proportion of adults who are estimated to consume 5 or more portions of fruit and vegetables per day when compared to the national value. Given the characteristics of the local population, a low indicator value (red circle in health summary chart) represents a statistically significant lower (worse) proportion of adults who are estimated to consume 5 or more portions of fruit and vegetables per day when compared to the national value.
Health Profiles 2009 The Indicator Guide
135 Section 4: adult’s health and lifestyle
Interpretation: Potential for error due to type of measurement method
It is important that users note that these model-based do not take account of any additional local factors that may impact on the true prevalence rate of fruit and vegetable consumption in an area (e.g. local initiatives designed to increase fruit and vegetable consumption). They will almost certainly not mirror precisely any available measures from local studies or surveys (although research by NatCen and others have shown that they tend to be related). The figures, therefore, cannot be used to monitor performance or change over time. The model used is a non-aetiological model i.e. is not based on known aetiological risk factors. This may lead to estimated fruit and vegetable consumption levels which are at odds with, for example, local lifestyle survey results or modelled estimates which use known co-variates such as socioeconomic status, age, gender and ethnicity such as the fruit and vegetable consumption estimates modelled in the Health Poverty Index available at www. hpi.org.uk (see variables used in generation of model in calculation of indicator section below). There may also be a discrepancy between the modelled lower tier estimates (districts) and upper tier estimates (County geographies and above, plus new 2009 UA areas) which are based on actual Health Survey for England data. This has lead to inconsistencies between lower tier and county estimates for some areas as the datasets are derived using different methods.
Interpretation: Potential for error due to bias and confounding
These model based healthy lifestyle indicators are derived using the Health Survey for England data and are subject to both sampling and non-sampling error. Sampling errors arise solely as a result of drawing a sample rather than conducting a full survey of the population. Generally, the smaller the sample size the larger the variability in the estimates that one would expect to obtain from all the possible samples. Non-sampling errors arise during the course of the survey activities and there is no simple direct way of estimating the size of these errors. Non sampling error may include e.g. respondent bias, interview bias and refusal to participate. The use of statistical models for prediction involves making assumptions about relationships in the data. The suitability of the chosen models for the given data and the validity of the model in describing real world dynamics have a bearing on the nature and magnitude of the errors introduced. A key source of modelling error arises from omitting variables that would otherwise help improve the model predictions either by error or because there is no available or reliable data source for them. The model-based estimate generated for a particular area is the expected measure for that area based on its population characteristics - and not an estimate of the actual prevalence. In statistical terms, the model-based estimate is actually a biased estimate of the true value for the area and, as such, should be treated with caution. As mentioned above, the model-based estimates are unable to take account of any additional local factors that may impact on the true prevalence rate. To interpret the estimates, NatCen recommend that users adopt statements such as “given the characteristics of the local population we would expect approximately x% of adults within LA Y to consume 5 or more portions of fruit and vegetables per day ”.
Health Profiles 2009 The Indicator Guide
136 Section 4: adult’s health and lifestyle
Validation exercises were used to check the appropriateness of the chosen models. Confidence intervals have been calculated for the model-based estimates to capture both sampling and modelling error. The confidence intervals provide a range within which we can be fairly sure the ‘true’ value for that area lies. It is recommended that users look at the confidence interval for the estimates, not just the estimate. Estimates for two areas can only be described as significantly different if the confidence intervals for the estimates do not overlap. Users should also note that the potential sources of bias and error also apply to any ranking or banding of the small-area estimates. NatCen do not encourage any ranking of small area estimates within larger areas such as Local Authorities, Primary Care Organisations and Strategic Health Authorities. The 2003–2005 model based estimates are not comparable with the preceding estimates for 2000-2002 owing to differences in geography and modelling methodology: The 2000–2002 LA estimates were calculated by aggregating the model-based estimates for the component wards. The 2003–05 LA estimates have been calculated by modelling directly at the LA level. The choice of co-variate data was different as both the Index of Multiple deprivation 2004 and ONS area classifications were excluded in the 2003–05 estimates owing to their statistical relationship with other census-based covariates. The 2003–05 based estimates were adjusted to be consistent with the direct survey estimates at GOR/SHA level. The model-based estimates have been produced solely for LAs and cannot be translated onto any other geographical boundary system (i.e. aggregated or averaged over any other spatial unit). As a result of limitations in the model based estimates they are published as “experimental statistics”. This term is applied to any set of ONS statistics that do not meet the rigorous quality standards of National Statistics and/or may be subject to change due to methodological development. Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health.
Health Profiles 2009 The Indicator Guide
137 Section 4: adult’s health and lifestyle
The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Table 2 – Indicator Specification Indicator definition: Variable
Prevalence of healthy eating. Healthy eating is defined as those who consume 5 or more portions of fruit and vegetables per day.
Indicator definition: Statistic
Percentage of resident adult population aged 16 and over
Indicator definition: Gender
Persons
Indicator definition: age group
Adults (aged 16 and over)
Indicator definition: period
2003–2005
Indicator definition: scale Geography: geographies available for this indicator from other providers
MSOA available from http://www.ic.nhs.uk/pubs/healthylifestyles05
Dimensions of inequality: subgroup analyses of this dataset available from other providers
None.
Data extraction: Source
National Centre for Social Research (NatCen).
Data extraction: source URL
Data received directly from NatCen. Also published by IC, December 2007 (see link above).
Data extraction: date
16/01/2008
Numerator: definition
Model estimates by NatCen using data from a number of sources including Health Survey for England 2003–2005, Census 2001.
Numerator: source
NatCen.
Health Profiles 2009 The Indicator Guide
138 Section 4: adult’s health and lifestyle
Denominator: definition
Not applicable.
Denominator: source
Not applicable.
Data quality: Accuracy and completeness
The model-based approach generates estimates that are of a different nature from standard survey estimates because they are dependent upon how well the relationship between healthy lifestyle behaviours for individuals and the Census/administrative information about the area in which they live are specified in the model. The accuracy and completeness of the information will be subject to the same constraints surrounding the Health Survey for England and Census data sets on which they are based (see Interpretation: potential sources of error section).
TABLE 3 – INDICATOR TECHNICAL METHODS Numerator: extraction
Not applicable.
Numerator: aggregation / allocation
Not applicable.
Numerator data caveats
See Interpretation: potential sources of error section.
Denominator data caveats
Not applicable.
Methods used to calculate indicator value
The process of creating the model-based estimates of healthy lifestyle behaviours involved three main stages: • A statistical model was used to represent the relationships between healthy eating and area-level characteristics in the small areas covered by the HSfE. The 2001 Census provided the main source for demographic and social covariate data. Other routine sources of data providing area-level characteristics for LAs included all age-all cause mortality, diversity index, life expectancy, emergency hospital admissions, hospital admissions attributable to alcohol, job seekers allowance claimant counts and educational attainment. • The model outputs were applied in conjunction with covariate data (available for all LAs) to estimate the ‘expected’ prevalence given the characteristics of the area. • An adjustment factor was applied to ensure that the model-based estimates for each LA corresponded with the 2003–2005 direct estimates at GOR/SHA level taken from the Health Survey for England data. The model contained an interaction term between SHA and the percentage of residents claiming Job Seekers Allowance over 12 months. The model also contained an interaction term between the percentage of residents aged 16–74 whose highest qualification attained was an NVQ Level 1 (or with no qualifications) and living in areas characterised as containing Village, Hamlet and Isolated dwellings.
Health Profiles 2009 The Indicator Guide
139 Section 4: adult’s health and lifestyle
The area level characteristics associated with increased propensity for an adult to consume 5 or more portions of fruit and vegetables were: a higher proportion of residents of Pakistani ethnic origin; a higher hospital admission rate for acute conditions usually managed in primary care settings; and a higher proportion of residents with limiting long-term illness. The area-level characteristics associated with decreased propensity for an adult to consume 5 or more portions of fruit and vegetables were: residing in an area with a high standardised mortality ratio and areas where a relatively higher proportion of households have no regular access to a car. The model-based estimates were constrained to the direct GOR/SHA estimates taken from the Health Survey for England data. This was done by aggregating the model-based estimates to GOR/SHA and comparing to the direct estimates. The relevant ratios of the HSfE direct estimates to the aggregated model-based estimates at GOR/SHA level were then used to scale the model based LA level estimates. However, the model-based estimates have not been calibrated to ensure agreement at County level. As a result, there may be inconsistencies between lower tier and county estimates for some areas as the datasets are derived using different methods. For a fuller technical description of the methodology see the Model-Based Estimates User Guide and other reports available on the Information Centre website: http://www.ic.nhs.uk/webfiles/Popgeog/Healthy%20Lifestyle%20 Behaviours-%20Model%20Based%20Estimates%20for%20Middle%20 Layer%20Super%20Output%20Areas%20and%20Local%20Authorities%20 in%20England_2003-2005__%20User%20Guide.pdf For methods used to calculate data for new 2009 UAs please see metadata for upper tier geographies. Small Populations: How Isles of Scilly and City of London populations have been dealt with
Model based estimates were not produced for Isles of Scilly or City of London.
Disclosure Control
Not applicable.
Health Profiles 2009 The Indicator Guide
140 Section 4: adult’s health and lifestyle
Confidence Intervals calculation method
The model-based estimate generated for a particular LA is the expected measure for that LA based on its characteristics as measured by the covariates in the model. In statistical terms, the model-based estimate is actually a biased estimate of the true value for an area and, as such, should be treated with caution. By placing confidence intervals around a model-based estimate, however, we can generate a range within which we can be fairly sure the ‘true’ value for that area lies. In order to generate the confidence interval, an estimate of the variance of the difference between the model-based estimate and the true LA measure is required. The estimate of the variance has two components which correspond to the variance that is not explained by the model (and hence is not predicted by the model-based estimate) and the uncertainty of the model-based estimate itself. Obtaining the confidence interval for the LA estimates required computing two area-level variance terms: one for the Primary Sampling Units and one for the LA. The first term was required to allow for the clustering in the sample and the second term to estimate the residual variance at the LA level that was not explained by the model. The confidence interval for the model-based LA estimates was estimated to be:
βˆ T X i T e −1 log it αˆ + βˆ X i ± 1.96 T βˆ X i 1+ e
2
M i N ij 2 2 ∑ σˆ u + σˆ v2 + X TiVar ( βˆ ) X i Ni j =1
1
where: Χi is the vector of covariates values for LA i,
Xi
βˆ
βˆ is the vector of parameter estimates for the LA-level covariates,
σˆ 2 is the estimate of the between PSU-level variance,
σˆ u2u
σˆ 2 is the estimate of the between LA-level variance.
σˆ v2v
N is the population estimate of the number of adults in LA i,
Ni i
Nijij is the population estimate of the number of adults in postcode sector N M i and
Mi is the number of postcode sectors in LA i.
ij,
2
Health Profiles 2009 The Indicator Guide
141 Section 4: adult’s health and lifestyle
16. PHYSICALLY ACTIVE ADULTS INDICATOR Basic Information 1. What is being measured?
Adults participating in recommended levels of physical activity.
2. Why is it being measured?
To estimate prevalence of physical activity beneficial to health in the population. To monitor the effectiveness of programmes aiming to increase participation of adults in sport and active recreation.
3. How is this indicator actually defined?
Participation in moderate intensity sport and active recreation on 20 or more days in the previous 4 weeks, (averaging 5 or more times per week).
4. Who does it measure?
Adults (aged 16 and over).
5. When does it measure it?
The survey is to be repeated in 2008/9 and 2009/10.
6. Will It measure absolute numbers or proportions?
Proportions: persons, aged 16 and over, 2007/08, as percentage respondents of the Sport England Active People Survey 2.
7. Where does the data actually come from?
Data received directly from Sport England.
8. How accurate and complete will the data be?
The Active People Survey is a telephone survey conducted using CATI (Computer Aided Telephone Interviewing). At least 500 interviews are undertaken per local authority. The exceptions were Isles of Scilly, City of London (due to small populations) and Birmingham where 5,000 interviews were conducted and Liverpool where 2,500 interviews were conducted.
9. Are there any caveats/warnings/ problems?
This measure is a crude proportion and no agestandardisation has been applied to the results to adjust for differences in age structure between areas. It is likely that a greater proportion of younger people undertake levels of physical activity at the recommended levels than older people. Sport England numerator data are based on observed self-reported physical activity levels in the previous 4 weeks and self-reported physical activity levels may be prone to respondent bias. In addition, the indicator does not include active recreation such as housework, DIY, activity in ones job or active transport.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Health Profiles 2009 The Indicator Guide
142 Section 4: adult’s health and lifestyle
TABLE 1 – INDICATOR DESCRIPTION Information component
Pg 4 Health Summary – Indicator 16
Subject category/ domain(s)
Adults health and lifestyle
Indicator name (* Indicator title in health profile)
Adults participating in recommended levels of physical activity (*Physically active adults)
PHO with lead responsibility
SEPHO
Date of PHO dataset creation
04/03/2009
Indicator definition
Participation in *moderate intensity sport and active recreation on 20 or more days in the previous 4 weeks, (averaging 5 or more times per week), percentage, persons, aged 16 and over, 2007/08, as percentage respondents of the Sport England Active People Survey 2
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs, new 2009 Unitary Authorities.
Timeliness
The survey is to be repeated in 2008/9 (APS3) and 2009/10 (APS4)
Rationale: What this indicator purports to measure
This indicator estimates the proportion of adults participating in physical activity beneficial to health.
Rationale: Public Health Importance
People who have a physically active lifestyle are at approximately half the risk of developing coronary heart disease compared to those who have a sedentary lifestyle. Regular physical activity is also associated with a reduced risk of diabetes, obesity, osteoporosis and colon cancer and with improved mental health. In older adults physical activity is associated with increased functional capacities. In terms of mortality, morbidity and quality of life, the Chief Medical Officer has estimated the cost of inactivity in England to be £8.2 billion annually. Evidence for the effectiveness of interventions to increase the population levels of physical activity is summarised by Kahn E, Ramsey L, Brownson R, Heath G, Howze E, Powell K, et al. ‘The Effectiveness of Interventions to Increase Physical Activity: A Systematic Review’. Am J Prev Med 2002; 22 (4S)
Rationale: Purpose behind the inclusion of the indicator
To estimate prevalence of physical activity beneficial to health in the population To monitor the effectiveness of programmes aiming to increase participation of adults in sport and active recreation The indicator is a measure of health need i.e. the ability to benefit from public health interventions aiming to improve levels of physical activity beneficial to health in the adult population.
Health Profiles 2009 The Indicator Guide
143 Section 4: adult’s health and lifestyle
Rationale: Policy relevance
Choosing Health: Making Healthy Choices Easier, Department of Health, 2004 http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4094550 Choosing Health: a physical activity action plan, Department of Health, 2005 http://www.dh.gov.uk/prod_consum_dh/groups/dh_digitalassets/@dh/@en/ documents/digitalasset/dh_4105710.pdf At Least Five a Week: Evidence on the Impact of Physical Activity and its Relationship to Health, 2004 – report by the CMO http://www.dh.gov.uk/prod_consum_dh/groups/dh_digitalassets/@dh/@en/ documents/digitalasset/dh_4080981.pdf Game Plan: A Strategy for Delivering Government’s Sport and Physical Activity Objectives, 2002 http://www.sportdevelopment.org.uk/html/gameplan.html The Sport England Strategy 2008–2001 was published in June 2008 and commits Sport England to deliver on a series of demanding targets by 2012/13: • one million people doing more sport • a 25% reduction in the number of 16–18 year olds who drop out of five key sports • improved talent development systems in at least 25 sports • a measurable increase in people’s satisfaction with their experience of sport – the first time the organisation has set such a qualitative measure • a major contribution to the delivery of the five hour sports offer for children and young people. The strategy can be found at: http://www.sportengland.org/index/get_resources/resource_downloads/sport_ england_strategy.htm
Interpretation: What a high/low level of indicator value means
A high indicator value (green circle in health summary chart) represents a statistically significant higher (better) estimated percentage of adults participating in physical activity when compared to the national value. A low indicator value (red circle in health summary chart) represents a statistically significant lower (worse) estimated percentage of adults participating in physical activity when compared to the national value.
Interpretation: Potential for error due to type of measurement method
Sport England numerator data are based on observed self-reported physical activity levels in the previous 4 weeks through a series of questions which focus on walking, cycling and other types of sport and recreational physical activity. Self-reported physical activity levels may be prone to respondent bias. Respondents are required to remember how long each session of physical activity lasted and describe the intensity level of the activity they undertook. As a result, levels of physical activity are likely to be over-reported. The Active People Survey measures sport and active recreation but excludes other forms of physical activity such as housework, DIY, activity in ones job, active transport etc. and this may lead to under-estimation of levels.
Health Profiles 2009 The Indicator Guide
144 Section 4: adult’s health and lifestyle
Interpretation: Potential for error due to bias and confounding
This measure of adults participating in recommended levels of physical activity is a crude proportion and no age-standardisation has been applied to the results to adjust for differences in age structure between areas. It is likely that a greater proportion of younger people undertake levels of physical activity at the recommended levels than older people. Details of the sampling frame used and which groups may have been under sampled (e.g. ethnic minorities, those without telephone access etc) were not available at the time of creation of this indicator. Therefore, it is not known how representative of the general population the survey results are.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with a amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Health Profiles 2009 The Indicator Guide
145 Section 4: adult’s health and lifestyle
Table 2 – Indicator Specification Indicator definition: Variable
Participation in *moderate intensity sport and active recreation on 20 or more days (averaging 5 or more times per week) in the previous 4 weeks *Moderate intensity sport and active recreation is defined as a 30 minute session of activity such as cycling, walking, sport and recreational physical activity. For walking, moderate activity was defined as walking at “a fairly brisk pace” or “a fast pace”. For cycling and other sport and recreational activity, moderate activity was defined as where the effort required was enough to raise the individual’s breathing rate, make the respondent out of breath or sweat. For further details see http://www.sportengland.org/index/ get_resources/research/active_people.htm
Indicator definition: Statistic
Percentage of respondents of the Sport England Active People Survey 2 (APS2)
Indicator definition: Gender
Persons
Indicator definition: age group
16 and over
Indicator definition: period
October 2007–October 2008
Indicator definition: scale Geography: geographies available for this indicator from other providers
At time of creation of indicator: County Sport Partnerships available from Sport England www.sportengland.org
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Age, gender, ethnicity, education, housing tenure, disability, working status, household income, National Statistics Socio-economic Classification (NS-SEC) available from Sport England.
Data extraction: Source
Sport England. Ipsos MORI undertook the survey on behalf of Sport England.
Data extraction: source URL
Data received directly from Sport England.
Data extraction: date
21/02/2009 – Local authority district estimates 25/02/2009 – Regional/County data 4/3/2009 – Central Bedfordshire UA, Cheshire East, Cheshire West and Chester
Numerator: definition
Respondents aged 16 and over, with valid responses to questions on physical activity in the Active People Survey 2, who participated in moderate intensity sport or recreational activities for 30 minutes or more on 20 or more days in the previous four weeks.
Numerator: source
Sport England Active People Survey 2 www.sportengland.org/
Denominator: definition
Respondents aged 16 and over, with valid responses to questions on physical activity in the Active People Survey 2
Health Profiles 2009 The Indicator Guide
146 Section 4: adult’s health and lifestyle
Denominator: source
Sport England Active People Survey 2 www.sportengland.org/
Data quality: Accuracy and completeness
The Active People Survey was undertaken by Ipsos MORI and conducted across every local authority in England. The survey is conducted by telephone using Random Digit Dialling (RDD) to generate a sample of telephone numbers. The RDD sample is drawn by selecting numbers from a database comprising all exchange codes allocated for residential use in the UK. A representative sample is then drawn by randomising the last four digits of each number. The telephone survey is conducted using CATI (Computer Aided Telephone Interviewing) and in APS2 achieved 188,800 interviews overall with a minimum of 500 interviews per local authority. This number of interviews will be repeated in the 2008/9 (Active People Survey 3) and 2009/10 (Active People Survey 4) surveys giving a cumulative local authority sample of 1,000 (equivalent to the first Active People Survey) every two years. All local authorities were offered the opportunity to boost their APS2 from 500 to 1,000. This was taken up by 14 local authorities. Two LAs choose larger boosts: Liverpool boosted by 2,000 to provide a 2,500 overall sample size and Birmingham boosted by 3,800 to provide a 5,000 overall sample size. Further exceptions to the 500 sample target in APS2 were Isles of Scilly and City of London which have small populations and do not require such a large sample. Respondents were matched to local authority of residence using postcode of residents. Where postcode was not known (around 1 in 10), respondents were asked to state in which local authority they lived or dialling codes/exchange area telephone number were used to look this up. The Active People Survey data are weighted to be representative of the 16+ population of each reporting geography (e.g. local authority, Government Region etc). Data within each reporting geography are weighted by the following factors: age within gender, ethnicity (White/Non White), National Statistics socio-economic classification (NS SEC). To ensure summer and winter responses are given equal weight within the annual data, a final weight is applied to distribute half the sample size to the summer period and half the sample size to the winter period. Further information is available on request from Sport England.
Table 3 – Indicator Technical Methods Numerator: extraction
Extraction by Sport England.
Numerator: aggregation / allocation
Performed by Sport England.
Numerator data caveats
Numerator data are weighted totals and as such do not sum to overall totals.
Denominator data caveats
Denominator data are weighted totals and as such do not sum to overall totals.
Health Profiles 2009 The Indicator Guide
147 Section 4: adult’s health and lifestyle
Methods used to calculate indicator value
The number of respondents aged 16 and over, with valid responses to questions on physical activity in the Active People Survey 2006, who participated in moderate intensity sport or recreational activities for 30 minutes or more on 20 or more days in the previous four weeks for each local authority was divided by the number of respondents aged 16 and over, with valid responses to questions on physical activity in the Active People Survey, 2006 and multiplied by 100. This generated the percentage participating in recommended levels of physical activity.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Isles of Scilly and City of London are excluded from the lower tier datasets but included in England, Regional and County figures.
Disclosure Control
None applied.
Confidence Intervals calculation method
95% confidence intervals have been calculated using a formula based on a normal approximation provided by Sport England as follows: Upper and lower 95% confidence interval = 100*(1.96*SQRT((Share*(1Share))/Total)) where share is the question proportion and total is the sample size.
Health Profiles 2009 The Indicator Guide
148 Section 4: adult’s health and lifestyle
17a. OBESE ADULTS UPPER TIER INDICATOR Basic Information 1. What is being measured?
Prevalence of obese adults
2. Why is it being measured?
To estimates the proportion of adults who are classified as obese. Obesity has serious health consequences and is association with all cause mortality and decreased life expectancy.
3. How is this indicator actually defined?
Prevalence of obesity, percentage of resident population, adults, 2003–2005, persons.
4. Who does it measure?
Adults (aged 16 and over).
5. When does it measure it?
The Health Survey for England (HSE) is carried out annually.
6. Will It measure absolute numbers or proportions?
Percentage of resident adult population aged 16 and over.
7. Where does the data actually come from?
Health Surveys for England, National Centre for Social Research (NatCen). Published by The Information Centre for Health and Social Care (IC), 2007.
8. How accurate and complete will the data be?
The HSE was designed to be representative of the general, non-institutional population living in England. The current “full” sample size of the HSE comprises about 16,000 adults aged 16 and over.
9. Are there any caveats/warnings/ problems?
BMI is calculated using measured height and weight which may be subject to measurement bias. In some sections of the population, applying BMI classification is not always straightforward e.g. when looking at the elderly or different ethnic groups. A definition based on waist-hip ratio is often considered a better measure of obesity. The Health Survey for England under-samples younger people, people in employment, ethnic minorities, women, those who are healthier but exhibit less healthy behaviour.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Health Profiles 2009 The Indicator Guide
149 Section 4: adult’s health and lifestyle
Table 1 – Indicator Description Information component
Pg 4 Health Summary – Indicator Number 17
Subject category/ domain(s)
Adults health and lifestyle
Indicator name (* Indicator title in health profile)
Prevalence of obese adults (*Obese adults)
PHO with lead responsibility
SEPHO
Date of PHO dataset creation
21/01/2009
Indicator definition
Prevalence of obesity, percentage of resident population, adults, 2003-2005, persons
Geography
England, GOR, County, South East SHAs, new 2009 UAs (Bedford UA, Central Bedfordshire UA, Cheshire East UA, Cheshire West and Chester UA, County Durham, Northumberland UA, Shropshire UA, Wiltshire UA, Cornwall UA).
Timeliness
Updated annually.
Rationale: What this indicator purports to measure
Prevalence of adult obesity
Rationale: Public Health Importance
Obesity in adults is defined for epidemiological purposes as body mass index (BMI) > 30 kg/m2. There is an association between all cause mortality and obesity. Obesity decreases life expectancy by up to nine years. Obesity causes insulin insensitivity, which is an important causal factor in diabetes, heart disease, hypertension and stroke. Obesity is associated with the development of hormone-sensitive cancers; the increased mechanical load increases liability to osteoarthritis and sleep apnoea. Obesity carries psychosocial penalties. Thus there are many routes by which obesity is a detriment to wellbeing. All these penalties as outlined in the table below (except the risk of gallstones and hip fracture) decrease with weight loss.
Health Profiles 2009 The Indicator Guide
150 Section 4: adult’s health and lifestyle
Proportion of various diseases attributable to obesity (BMI > 27 kg/m2) Disease Obesity Hypertension Myocardial infarcation Angina pectoris Stroke Venous thrombosis NIDDM Hyperlipidaemia Gout Osteoarthritis Gall-bladder disease Colorectal cancer Breast cancer Genitourinary cancer Hip fracture
Relative Risk 2.9 1.9 2.5 3.1 1.5 2.9 1.5 2.5 1.8 2.0 1.3 1.2 1.6 0.8
Attributable proportion (%) 100.0 24.1 13.9 20.5 25.8 7.7 24.1 7.7 20.0 11.8 14.3 4.7 3.2 9.1 –3.5
Source: http://hcna.radcliffe-oxford.com/obframe.html It is estimated that obesity costs the NHS over £1 billion per year and society as a whole up to £3.5 billion per year. Effective interventions exist to prevent and treat obesity. See NICE guidance – Obesity: the prevention, identification, assessment and management of overweight and obesity in adults and children. http://guidance.nice.org.uk/CG43 Rationale: Purpose behind the inclusion of the indicator
To estimate the proportion of obese adults in local authorities.
Rationale: Policy relevance
Choosing Health: Making healthy choices easier (Dept. Health, 2004). http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4094550
To help reduce the prevalence of obesity.
Obesity: Defusing the Health Time Bomb (from the Annual Report of the Chief Medical Officer, 2002). http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/AnnualReports/ Browsable/DH_4875027 Tackling Obesity: A Toolbox for Local Partnership Action (The National Heart Forum, 2007) http://www.fphm.org.uk/resources/AtoZ/toolkit_obesity/obesity_toolkit_intro. pdf Tackling Obesity in England, 2001 (National Audit Office, 2001) http://www.nao.org.uk/publications/nao_reports/00-01/0001220.pdf
Health Profiles 2009 The Indicator Guide
151 Section 4: adult’s health and lifestyle
Interpretation: What a high/low level of indicator value means
A high indicator value (red circle in health summary chart) represents a statistically significant higher level (worse) of estimated adult obesity prevalence when compared to the national value. A low indicator value (green circle in health summary chart) represents a statistically significant lower level (better) of estimated adult obesity prevalence when compared to the national value. However obesity at any prevalence level greater than 0 is undesirable, and therefore a low indicator value should not mean that PH action is not needed.
Interpretation: Potential for error due to type of measurement method
For each participant in the Health Survey for England, height and weight was recorded by a nurse. BMI was then calculated for all informants who had valid height and weight measurements, those considered to have unreliable measurements were excluded from the analysis (e.g. pregnant, chair-bound, unsteady or those who could not stand straight). Those who weighed more than 130kg were asked for an “estimated weight” because the scales were unreliable above this level. These were included in the analysis. In some sections of the population, applying the BMI classification described above is not always straightforward e.g. when looking at the elderly or different ethnic groups. For example, in certain Asian populations a given BMI equates to a higher percentage of body fat than the same BMI in a white European population. In some Black populations, however, the converse is true. A definition based on waist-hip ratio is often considered a better measure of obesity. However, BMI is most commonly used and easier to measure routinely. In order to ensure agreement between these estimates at GOR and SHA level, the lower tier synthetic estimates (districts), which are based on modelled data, have been calibrated. However, the synthetic estimates have not been calibrated to ensure agreement at County level. As a result, there may be inconsistencies between lower tier and county estimates for some areas as the datasets are derived using different methods.
Interpretation: Potential for error due to bias and confounding
The Health Survey for England under-samples younger people, people in employment, ethnic minorities, women, those who are healthier but exhibit less healthy behaviour. These data have not been age-standardised and, therefore, variation between area values may be a result of differences in population structure.
Health Profiles 2009 The Indicator Guide
152 Section 4: adult’s health and lifestyle
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Table 2 – Indicator Specification Indicator definition: Variable
Prevalence of obesity. Obesity in adults is defined for epidemiological purposes as body mass index (BMI) > 30 kg/m2.
Indicator definition: Statistic
Percentage of resident adult population aged 16 and over
Indicator definition: Gender
Person
Indicator definition: age group
Adults (aged 16 and over)
Indicator definition: period
2003–2005
Indicator definition: scale
Health Profiles 2009 The Indicator Guide
153 Section 4: adult’s health and lifestyle
Geography: geographies available for this indicator from other providers
Strategic Health Authority. http://www.ic.nhs.uk/webfiles/Popgeog/Direct%20Estimates%20%20of%20 Obesity%20(adults)%202003-2005.pdf
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Age, gender, ethnicity, social class http://www.dh.gov.uk/en/Publicationsandstatistics/PublishedSurvey/ HealthSurveyForEngland/index.htm
Data extraction: Source
Health Surveys for England, National Centre for Social Research (NatCen). Published by The Information Centre for Health and Social Care (IC), 2007
Data extraction: source URL
Data received directly from NatCen.
Data extraction: date
16 January 2008. Data for 4 new 2009 UAs (Bedford UA, Central Bedfordshire UA, Cheshire East UA and Cheshire West UA) were calculated by NatCen as a separate exercise on 21/01/2009 and received on the same date.
Numerator: definition
The number of persons aged 16+ who are obese in a sample survey of the health of the population of England.
Numerator: source
Health Survey for England (HSE), commissioned by the Department of Health/ IC and carried out by the Joint Health Surveys Unit of the National Centre for Social Research (NatCen) and the Department of Epidemiology and Public Health at the Royal Free and University College Medical School, London.
Denominator: definition
Total number of respondents (with valid measurements for height and weight) aged 16+ in the Health Survey for England 2003_2005.
Denominator: source
Health Survey for England (HSE), commissioned by the Department of Health/ IC and carried out by the Joint Health Surveys Unit of the National Centre for Social Research (NatCen) and the Department of Epidemiology and Public Health at the Royal Free and University College Medical School, London.
Data quality: Accuracy and completeness
The Health Survey for England is designed to provide data at both national and regional level about the population living in private households in England. It uses a clustered, stratified multi-stage sample design. In the 2005 HSE, for example, a random sample of over 7,200 addresses (around 16,000 people) were selected from the Postcode Address File (PAF) to ensure households were sampled proportionately across the nine Government Office Regions in England. 720 postcode sectors were selected and 26 addresses within each sector. Each individual within a selected household was eligible for inclusion. One of the effects of using a complex sample design is that standard errors for survey estimates are generally higher than would be derived from a simple random sample of the same size.
Health Profiles 2009 The Indicator Guide
154 Section 4: adult’s health and lifestyle
There was a full adult sample of around 16,000 in the 2003 HSE. However, the 2004 and 2005 Health Surveys had only 8,000 adults in the normal ‘general population’ sample as these two surveys included boost samples. The 2004 HSE included a boost sample to increase the number of participants from minority ethnic groups and a special Chinese boost sample. The 2005 HSE included a boost sample for older people living in private households and for five months of the year a boost of children aged 2–15 was included. To ensure that each year’s sample was given an approximately equal weight in the calculation of the 3-year estimates (2003–2005) respondents in the boost sample years were weighted up by two. The numerator and denominator counts used to estimate prevalence are based on a sample of the population in each area and, as such, are not true counts. For this reason the numerator and denominator data are not shown in the data sheet.
Table 3 – Indicator Technical Methods Numerator: extraction
Not applicable.
Numerator: aggregation/ allocation
Residency by local authority of each respondent is allocated by postcode of residency.
Numerator data caveats
Body mass index is defined as weight (kg) divided by height squared (m2). For adults, four groups are defined according to their BMI: • • • •
Underweight - under 18.5kg/m2; Desirable weight – 18.5 to 25kg/m2; Overweight - 25 to 30kg/m2; Obese - over 30kg/m2
BMI was calculated for all informants who had valid height and weight measurements. The height of informants who were chair-bound, unsteady, or could not stand straight was not measured. Data for those who were considered by the interviewer to have unreliable measurements (e.g. wearing a wig, turban) were excluded from the analysis. The weight of informants who were pregnant, chair bound, unsteady, or could not stand was not measured. Those who weighed more than 130 kg were asked for an “estimated weight” because the scales were unreliable above this level. These have been included in the analysis. These data have not been age-standardised and, therefore, variation between area values may be a result of differences in population structure. Denominator data caveats
The HSE is a series of annual surveys that began in 1991 with the aim of monitoring the health of the population. It was designed to be representative of the general, non-institutional population living in England. The current “full” sample size of the HSE comprises about 16,000 adults aged 16 and over. For each participant, the survey included an interview and a physical examination by a nurse, at which various physical measurements, tests, and samples of blood and saliva were collected. These measurements provided biomedical information about known risk factors associated with disease and objective validation for self-reported health behaviour.
Health Profiles 2009 The Indicator Guide
155 Section 4: adult’s health and lifestyle
Methods used to calculate indicator value
Estimates are based on pooling together three consecutive years of Health Survey for England data (2003–2005). The general population sample size in 2004 and 2005 was about half the sample size in 2003 owing to the sampling of specific population groups – namely, ethnic minority populations (2004) and older people living in private households (2005). To ensure that each year’s sample was given an approximately equal weight in the calculation of the 2003-2005 estimates, respondents in 2004 and 2005 were weighted up by two. Data for 4 new 2009 UAs (Bedford UA, Central Bedfordshire UA, Cheshire East UA and Cheshire West UA) were calculated by NatCen as a separate exercise on 21/01/2009 using the same method as described above. As geographical coverage for the remaining 6 new 2009 UAs (County Durham UA, Northumberland UA, Shropshire UA, Wiltshire UA, Cornwall UA) remains the same as their old County equivalents, data for these areas has not been recalculated as part of this exercise.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
The Health Survey for England sample does not cover the Isles of Scilly and no HSE respondents over 2003–2005 were located in the City of London.
Disclosure Control
Not applicable.
Confidence Intervals calculation method
The standard errors, and 95% confidence intervals, have been calculated using STATA’s survey module (the svy:mean commands), further details can be obtained from Shaun Scholes at NatCen ([email protected]). One of the effects of using a complex design is that standard errors for survey estimates are generally higher than the standard errors that would be derived from a simple random sample of the same size.
Health Profiles 2009 The Indicator Guide
156 Section 4: adult’s health and lifestyle
17b. OBESE ADULTS LOWER TIER INDICATOR Basic Information 1. What is being measured?
Estimated prevalence of obese adults.
2. Why is it being measured?
To estimate the expected proportion of obese adults in local authorities given the characteristics of local authority populations. To help reduce the prevalence of obesity.
3. How is this indicator actually defined?
Prevalence of obesity, percentage of resident population, adults, 2003–2005, persons.
4. Who does it measure?
Adults (aged 16 and over).
5. When does it measure it?
Updated as ad-hoc.
6. Will It measure absolute numbers or proportions?
Percentage of resident adult population aged 16 and over.
7. Where does the data actually come from?
Modelled by the National Centre for Social Research (NatCen). Published by The Information Centre for Health and Social Care (IC), 2007.
8. How accurate and complete will the data be?
These are modelled estimates based on national survey data. The model is non-aetiological (not based on known casual factors). The estimates do not take into account additional local factors that may impact on the true prevalence of obesity in an area and may not match with local lifestyle survey results or modelled estimates which use known risk factors.
9. Are there any caveats/warnings/ problems?
As these estimates are modelled they should be used and interpreted with caution (see above).
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
TABLE 1 – INDICATOR DESCRIPTION Information component
Pg 4 Health Summary – Indicator 17
Subject category/ domain(s)
Adults health and lifestyle
Indicator name (* Indicator title in health profile)
Estimated prevalence of obese adults (*Obese adults)
PHO with lead responsibility
SEPHO
Health Profiles 2009 The Indicator Guide
157 Section 4: adult’s health and lifestyle
Date of PHO dataset creation
21/01/2009
Indicator definition
Prevalence of obesity, percentage of resident population, adults, 2003–2005, persons.
Geography
Local Authority: County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs. Exceptions: Data for the 9 new 2009 UAs (Bedford UA, Central Bedfordshire UA, Cheshire East UA, Cheshire West and Chester UA, County Durham, Northumberland UA, Shropshire UA, Wiltshire UA, Cornwall UA) are not based on modelled estimates but are calculated directly using results from the Health Surveys for England 2003-05.
Timeliness
Updated as ad-hoc.
Rationale: What this indicator purports to measure
Expected prevalence of adult obesity.
Rationale: Public Health Importance
Obesity in adults is defined for epidemiological purposes as body mass index (BMI) > 30 kg/m2. There is an association between all cause mortality and obesity. Obesity decreases life expectancy by up to nine years. Obesity causes insulin insensitivity, which is an important causal factor in diabetes, heart disease, hypertension and stroke. Obesity is associated with the development of hormone-sensitive cancers; the increased mechanical load increases liability to osteoarthritis and sleep apnoea. Obesity carries psychosocial penalties. Thus there are many routes by which obesity is a detriment to wellbeing. All these penalties as outlined in the table below (except the risk of gallstones and hip fracture) decrease with weight loss. Proportion of various diseases attributable to obesity (BMI > 27 kg/m2) Disease Obesity Hypertension Myocardial infarcation Angina pectoris Stroke Venous thrombosis NIDDM Hyperlipidaemia Gout Osteoarthritis Gall-bladder disease Colorectal cancer Breast cancer Genitourinary cancer Hip fracture
Relative Risk 2.9 1.9 2.5 3.1 1.5 2.9 1.5 2.5 1.8 2.0 1.3 1.2 1.6 0.8
Attributable proportion (%) 100.0 24.1 13.9 20.5 25.8 7.7 24.1 7.7 20.0 11.8 14.3 4.7 3.2 9.1 -3.5
Source: http://hcna.radcliffe-oxford.com/obframe.html
Health Profiles 2009 The Indicator Guide
158 Section 4: adult’s health and lifestyle
It is estimated that obesity costs the NHS over £1 billion per year and society as a whole up to £3.5 billion per year. Effective interventions exist to prevent and treat obesity. See NICE guidance – Obesity: the prevention, identification, assessment and management of overweight and obesity in adults and children. http://guidance.nice.org.uk/CG43 Rationale: Purpose behind the inclusion of the indicator
To estimate the expected proportion of obese adults in local authorities given the characteristics of local authority populations.
Rationale: Policy relevance
Choosing Health: Making healthy choices easier (Dept. Health, 2004). http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4094550
To help reduce the prevalence of obesity.
Obesity: Defusing the Health Time Bomb (from the Annual Report of the Chief Medical Officer, 2002). http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/AnnualReports/ Browsable/DH_4875027 Tackling Obesity: A Toolbox for Local Partnership Action (The National Heart Forum, 2007) http://www.fphm.org.uk/resources/AtoZ/toolkit_obesity/obesity_toolkit_intro. pdf Tackling Obesity in England, 2001 (National Audit Office, 2001) http://www.nao.org.uk/publications/nao_reports/00-01/0001220.pdf Interpretation: What a high/low level of indicator value means
A high indicator value (red circle in health summary chart) represents a statistically significant higher level (worse) of expected estimated adult obesity prevalence given the characteristics of the population for that local authority when compared to the national value. A low indicator value (green circle in health summary chart) represents a statistically significant lower level (better) of expected estimated adult obesity prevalence given the characteristics of the population for that local authority when compared to the national value. However obesity at any prevalence level greater than 0 is undesirable, and therefore a low indicator value should not mean that PH action is not needed.
Interpretation: Potential for error due to type of measurement method
It is important that users note that these model based do not take account of any additional local factors that may impact on the true obesity prevalence rate in an area (e.g. local initiatives designed to reduce obesity). They will almost certainly not mirror precisely any available measures from local studies or surveys (although research by NatCen and others have shown that they tend to be related). The figures, therefore, cannot be used to monitor performance or change over time. The model used is a non-aetiological model i.e. is not based on risk factors such as physical activity levels and calorie intake. This may lead to estimated obesity levels which are at odds with other risk factors estimates such as healthy eating and physical activity; local lifestyle survey results and modelled estimates which use known co-variates (see variables used in generation of model in calculation of indicator section below). There may also be a discrepancy between the modelled lower tier estimates (districts) and upper tier estimates (County geographies and above, plus new 2009 UA areas) which are based on actual Health Survey for England data. This has led to inconsistencies between lower tier and county estimates for some areas as the datasets are derived using different methods.
Health Profiles 2009 The Indicator Guide
159 Section 4: adult’s health and lifestyle
Interpretation: Potential for error due to bias and confounding
These model based healthy lifestyle indicators are derived using the Health Survey for England data and are subject to both sampling and non-sampling error. Sampling errors arise solely as a result of drawing a sample rather than conducting a full survey of the population. Generally, the smaller the sample size the larger the variability in the estimates that one would expect to obtain from all the possible samples. Non-sampling errors arise during the course of the survey activities and there is no simple direct way of estimating the size of these errors. Non sampling error may include e.g. respondent bias, interview bias and refusal to participate. The use of statistical models for prediction involves making assumptions about relationships in the data. The suitability of the chosen models for the given data and the validity of the model in describing real world dynamics have a bearing on the nature and magnitude of the errors introduced. A key source of modelling error arises from omitting variables that would otherwise help improve the model predictions either by error or because there is no available or reliable data source for them. The model-based estimate generated for a particular area is the expected measure for that area based on its population characteristics - and not an estimate of the actual prevalence. In statistical terms, the model-based estimate is actually a biased estimate of the true value for the area and, as such, should be treated with caution. As mentioned above, the model-based estimates are unable to take account of any additional local factors that may impact on the true prevalence rate (e.g. local initiatives designed to reduce obesity). To interpret the estimates, NatCen recommend that users adopt statements such as “given the characteristics of the local population we would expect approximately x% of adults within LA Y to be obese”. Validation exercises were used to check the appropriateness of the chosen models. Confidence intervals have been calculated for the model-based estimates to capture both sampling and modelling error. The confidence intervals provide a range within which we can be fairly sure the ‘true’ value for that area lies. It is recommended that users look at the confidence interval for the estimates, not just the estimate. Estimates for two areas can only be described as significantly different if the confidence intervals for the estimates do not overlap. Users should also note that the potential sources of bias and error also apply to any ranking or banding of the small-area estimates. NatCen do not encourage any ranking of small area estimates within larger areas such as Local Authorities, Primary Care Organisations and Strategic Health Authorities. • The 2003–2005 model based estimates are not comparable with the preceding estimates for 2000–2002 owing to differences in geography and modelling methodology: • The 2000–2002 LA estimates were calculated by aggregating the modelbased estimates for the component wards. The 2003–05 LA estimates have been calculated by modelling directly at the LA level. • The choice of co-variate data was different as both the Index of Multiple deprivation 2004 and ONS area classifications were excluded in the 2003– 05 estimates owing to their statistical relationship with other census-based covariates. • The 2003–05 based estimates were adjusted to be consistent with the direct survey estimates at GOR/SHA level.
Health Profiles 2009 The Indicator Guide
160 Section 4: adult’s health and lifestyle
The model-based estimates have been produced solely for LAs and cannot be translated onto any other geographical boundary system (i.e. aggregated or averaged over any other spatial unit). As a result of limitations in the model based estimates they are published as “experimental statistics”. This term is applied to any set of ONS statistics that do not meet the rigorous quality standards of National Statistics and/or may be subject to change due to methodological development. Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
TABLE 2 – INDICATOR SPECIFICATION Indicator definition: Variable
Prevalence of obesity. Obesity in adults is defined for epidemiological purposes as body mass index (BMI) > 30 kg/m2.
Indicator definition: Statistic
Percentage of resident adult population aged 16 and over
Indicator definition: Gender
Persons
Indicator definition: age group
Adults (aged 16 and over)
Health Profiles 2009 The Indicator Guide
161 Section 4: adult’s health and lifestyle
Indicator definition: period
2003–2005
Indicator definition: scale Geography: geographies available for this indicator from other providers
MSOA available from http://www.ic.nhs.uk/pubs/healthylifestyles05
Dimensions of inequality: subgroup analyses of this dataset available from other providers
None.
Data extraction: Source
National Centre for Social Research (NatCen).
Data extraction: source URL
Data received directly from NatCen. Also published by IC, December 2007 (see link above).
Data extraction: date
16/01/2008
Numerator: definition
Model estimates by NatCen using data from a number of sources including Health Survey for England 2003–2005, Census 2001.
Numerator: source
NatCen.
Denominator: definition
Not applicable.
Denominator: source
Not applicable.
Data quality: Accuracy and completeness
The model-based approach generates estimates that are of a different nature from standard survey estimates because they are dependent upon how well the relationship between healthy lifestyle behaviours for individuals and the Census/administrative information about the area in which they live are specified in the model. The accuracy and completeness of the information will be subject to the same constraints surrounding the Health Survey for England and Census data sets on which they are based (see Interpretation: potential sources of error section).
Table 3 – Indicator Technical Methods Numerator: extraction
Not applicable.
Numerator: aggregation / allocation
Not applicable.
Numerator data caveats
See Interpretation: potential sources of error section.
Health Profiles 2009 The Indicator Guide
162 Section 4: adult’s health and lifestyle
Denominator data caveats
Not applicable.
Methods used to calculate indicator value
The process of creating the model-based estimates of healthy lifestyle behaviours involved three main stages: • A statistical model was used to represent the relationships between obesity and area-level characteristics in the small areas covered by the HSfE. The 2001 Census provided the main source for demographic and social covariate data. Other routine sources of data providing area-level characteristics for LAs included all age-all cause mortality, diversity index, life expectancy, emergency hospital admissions, hospital admissions attributable to alcohol, job seekers allowance claimant counts and educational attainment. • The model outputs were applied in conjunction with covariate data (available for all LAs) to estimate the ‘expected’ prevalence given the characteristics of the area. • An adjustment factor was applied to ensure that the model-based estimates for each LA corresponded with the 2003–2005 direct estimates at GOR/SHA level taken from the Health Survey for England data. The area level characteristics associated with increased propensity for a person to be classed as obese were: a higher proportion of residents of Black or ethnic origin; and a higher proportion of male residents. The area-level characteristics associated with decreased propensity for an adult to be classed as obese were: a higher proportion of residents aged 16–74 whose highest qualification attained was Level 2, 4 or 5; a relatively higher proportion of residents aged 20–24; a relatively higher proportion of Income Support claimants who were classifed as ‘carers and others’; a higher proportion of residents aged 16–74 who had never worked or were long-term unemployed; and a higher proportion of households without basic amenties (central heating and/or sole use of a bath). The model-based estimates were constrained to the direct GOR/SHA estimates taken from the Health Survey for England data. This was done by aggregating the model-based estimates to GOR/SHA and comparing to the direct estimates. The relevant ratios of the HSfE direct estimates to the aggregated model-based estimates at GOR/SHA level were then used to scale the model based LA level estimates. However, the model-based estimates have not been calibrated to ensure agreement at County level. As a result, there may be inconsistencies between lower tier and county estimates for some areas as the datasets are derived using different methods. For a fuller technical description of the methodology see the Model-Based Estimates User Guide and other reports available on the Information Centre website: http://www.ic.nhs.uk/webfiles/Popgeog/Healthy%20Lifestyle%20 Behaviours-%20Model%20Based%20Estimates%20for%20Middle%20 Layer%20Super%20Output%20Areas%20and%20Local%20Authorities%20 in%20England_2003-2005__%20User%20Guide.pdf For methods used to calculate data for new 2009 UAs please see metadata for upper tier geographies.
Health Profiles 2009 The Indicator Guide
163 Section 4: adult’s health and lifestyle
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Model based estimates were not produced for Isles of Scilly or City of London.
Disclosure Control
Not applicable.
Confidence Intervals calculation method
The model based estimate generated for a particular LA is the expected measure for that LA based on its characteristics as measured by the covariates in the model. In statistical terms, the model-based estimate is actually a biased estimate of the true value for an area and, as such, should be treated with caution. By placing confidence intervals around a model-based estimate, however, we can generate a range within which we can be fairly sure the ‘true’ value for that area lies. In order to generate the confidence interval, an estimate of the variance of the difference between the model-based estimate and the true LA measure is required. The estimate of the variance has two components which correspond to the variance that is not explained by the model (and hence is not predicted by the model-based estimate) and the uncertainty of the model-based estimate itself. Obtaining the confidence interval for the LA estimates required computing two area-level variance terms: one for the Primary Sampling Units and one for the LA. The first term was required to allow for the clustering in the sample and the second term to estimate the residual variance at the LA level that was not explained by the model. The confidence interval for the model-based LA estimates was estimated to be:
βˆ T X i T e −1 ˆ ˆ log it α + β X i ± 1.96 T βˆ X i 1+ e
2
N ij 2 T 2 ˆ ∑ σˆ + σˆ v + X i Var ( β ) X i N i u j =1 Mi
2
1
where: Χi is the vector of covariates values for LA i,
Xi
β
βˆ is the vector of parameter estimates for the LA-level covariates, ˆ
σˆ2u2 is the estimate of the between PSU-level variance,
σˆ u
σˆ 2 is the estimate of the between LA-level variance.
σˆ v2v
N is the population estimate of the number of adults in LA i,
Ni i
Nijij is the population estimate of the number of adults in postcode sector N M i and
Mi is the number of postcode sectors in LA i.
ij,
2
Health Profiles 2009 The Indicator Guide
164
Section 5: Disease and poor health
Health Profiles 2009 The Indicator Guide
165 Section 5: disease and poor health
18. OVER 65s ‘NOT IN GOOD HEALTH’ INDICATOR Basic Information 1. What is being measured?
Over 65s Self-Assessed General Health: ‘Not Good’.
2. Why is it being measured?
As a perception of general health. This indicator may help monitor likely health care burden.
3. How is this indicator actually defined?
Self assessed general health: ‘Not Good’, directly age and sex standardised percentage, over 65s, 2001, persons.
4. Who does it measure?
All persons, 65 years and over.
5. When does it measure it?
2001, next Census will be in 2011.
6. Will it measure absolute numbers or proportions?
Age and sex standardised percentage.
7. Where does the data actually come from?
2001 Census, ONS.
8. How accurate and complete will the data be?
This census data has been subject to edit and imputation procedures to correct for incorrect or missing data.
9. Are there any caveats/ warnings/ problems?
‘Not in good health’ is the self-assessed perception of health and therefore subject to differing thresholds applied by individuals.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Health Profiles 2009 The Indicator Guide
166 Section 5: disease and poor health
Table 1 – Indicator Description Information component
Pg 4 Health Summary – Indicator No 18
Subject category/ domain(s)
Disease and poor health.
Indicator name (* Indicator title in health profile)
Over-65s ‘not in good health’
PHO with lead responsibility
WMPHO
Date of PHO dataset creation
12/02/2009
Indicator definition
Self Assessed General Health: ‘Not Good’, directly age and sex standardised percentage, over 65s, 2001, persons
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs
Timeliness
Every 10 years, next update available in 2012. Time trend analysis is appropriate.
Rationale: What this indicator purports to measure
Perception of General Health.
Rationale: Public Health Importance
The indicator was chosen as the best available measure of self assessed population health. Self reported single item of health has a good correlation with mortality and health care utilisation. For further information see: Where Wealth means Health (www.nwpho.org. uk/inequalities)
Rationale: Purpose behind the inclusion of the indicator
To help monitor likely health care burden.
Rationale: Policy relevance
No direct policy driver.
Interpretation: What a high/low level of indicator value means
A high indicator value (red blob in spine chart) represents a statistically significant higher level of estimated self assessed “not good” health for that local authority when compared to the national value. A low indicator value (green blob) represents a statistically significant lower level of estimated self assessed “not good” health for that local authority when compared to the national value.
Health Profiles 2009 The Indicator Guide
167 Section 5: disease and poor health
Interpretation: Potential for error due to type of measurement method
Self reported health status can be subject to variation according to non causative effects (e.g. good weather).
Interpretation: Potential for error due to bias and confounding
The following groups may be under-sampled within the census:
Age standardisation does assume that minor differences in age structure within age bands are unimportant and in general this is true for younger age groups. However with older groups minor differences in average age within age band may become important, especially as all over 85’s are grouped together. The European Standard Population is younger than the UK population and therefore biases comparisons towards younger age groups. Although age standardisation will reduce the effect of age differences in the population it cannot be assumed to eliminate them.
• • • • • • • • •
Areas with high non-white population Full-time students aged 18–74 (out of term time residents) Prisoners Men aged 20–39 Residential homes, nursing homes, hospitals Rough sleepers Areas with high population density Areas with high numbers of multi-occupancy households Migrants: someone who spends 3 to 12 months in the country for certain purposes (excluding tourism), asylum seekers, migrant/seasonal workers,
This can result in an underestimate or overestimate of self assessed ill-health in some areas. Age standardisation does assume that minor differences in age structure within age bands are unimportant and in general this is true for younger age groups. However with older groups minor differences in average age within age band may become important, especially as all over 85s are grouped together. The European Standard Population is younger than the UK population and therefore biases comparisons towards younger age groups. Although age standardisation will reduce the effect of age differences in the population it cannot be assumed to eliminate them.
Health Profiles 2009 The Indicator Guide
168 Section 5: disease and poor health
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider the confidence interval, the greater the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Health Profiles 2009 The Indicator Guide
169 Section 5: disease and poor health
Table 2 – Indicator Specification Indicator definition: Variable
Self Assessed General Health: ‘Not Good’ (over 65s)
Indicator definition: Statistic
Directly age & sex standardised percentage
Indicator definition: Gender
Persons
Indicator definition: age group
Over 65s
Indicator definition: period
2001
Indicator definition: scale
Age & sex standardised percentage
Geography: geographies available for this indicator from other providers
Unique dataset not available elsewhere.
Dimensions of inequality: subgroup analyses of this dataset available from other providers
None.
Data extraction: Source
Nomis Official Labour Market Statistics (a service provided by the Office for National Statistics)
Data extraction: source URL
https://www.nomisweb.co.uk/Default.asp
Data extraction: date
Data extracted from source as at: 12/02/2009
Numerator: definition
All people aged over 65 usually resident in the area at the time of the 2001 Census, who described their general health in the 12 months before Census day as ‘Not good’.
Numerator: source
Office for National Statistics (ONS).
Denominator: definition
All people aged over 65 counted as usually resident in the area at the time of the 2001 Census by 5 year age band. A usual resident was generally defined as someone who spent most of their time at a specific address. It included: people who usually lived at that address but were temporarily away (on holiday, visiting friends or relatives, or temporarily in a hospital or similar establishment); people who worked away from home for part of the time; students, if it was their term-time address; a baby born before 30 April 2001 even if it was still in hospital; and people present on Census Day, even if temporarily, who had no other usual address. However, it did not include anyone present on Census Day who had another usual address or anyone who had been living or intended to live in a special establishment, such as a residential home, nursing home or hospital, for six months or more.
Health Profiles 2009 The Indicator Guide
170 Section 5: disease and poor health
Denominator: source
Office for National Statistics (ONS).
Data quality: Accuracy and completeness
The Census was followed by the Census Coverage Survey (CCS), an independent doorstep survey of a sample of a third of a million households, covering every local authority, which was used to adjust the Census counts for under-enumeration. Under-enumeration in the 2001 Census did not occur uniformly across all areas. Response rates were lowest for inner city areas where characteristics known to be related to non-response such as multi-occupancy and higher proportions of non-English speaking population, are most prevalent. This census data has been subject to edit and imputation procedures to correct for incorrect or missing data. There has been an extensive quality assurance process, including checks against administrative records and sources of information on particular groups such as students and the armed forces.
Table 3 – Indicator Technical Methods Numerator: extraction
Downloaded from Nomis website.
Numerator: aggregation/ allocation
Where available data published for higher geographies has been used and not aggregated from lower geographies. This census data has been subject to edit and imputation procedures to correct for incorrect or missing data as a result summation of areas to higher geographies may not result in the total stated in the published census data. However, since the 2001 census new Unitary Authorities have been created. The new Unitary Authorities of Northumberland, Shropshire, Wiltshire and Durham have the same boundaries as the counties of the same name so have been allocated these data. Data for five other new unitary Authorities have been calculated from the aggregation of previous Local Authorities. New UA name Bedford Central Bedfordshire Cheshire East
Cheshire West and Chester
Cornwall*
Old LA name Bedford Mid Bedfordshire South Bedfordshire Congleton Crewe and Nantwich Macclesfield Chester Ellesmere Port & Neston Vale Royal Caradon Carrick Kerrier North Cornwall Penwith Restormel
*=Cornwall Unitary Authority is the same as the previous Cornwall County however the data for this county in the data source was aggregated with Isles of Scilly and so has instead been aggregated from its local authorities.
Health Profiles 2009 The Indicator Guide
171 Section 5: disease and poor health
Numerator data caveats
See earlier comments on Data quality: Accuracy and completeness
Denominator data caveats
See earlier comments on Data quality: Accuracy and completeness
Methods used to calculate indicator value
The directly age & sex standardised rate is the rate of events that would occur in a population with a standard age and sex structure if that population were to experience the age & sex specific rates of the subject population. The standard population used is the European Standard Population for age standardisation. Sex standardisation has been used, based on a 1:1 ratio of males and females, because females have a longer life expectancy and therefore make up a larger proportion of responses in older age groups. This has been expressed as a rate per 100 i.e. a percentage. The methodology is explained in detail in the APHO Technical Briefing on Commonly Used Public Health Statistics and their Confidence Intervals http://www.apho.org.uk/resource/item.aspx?RID=48457 and the associated Excel tool http://www.apho.org.uk/resource/view.aspx?RID=48617. Age standardisation does assume that minor differences in age structure within age bands are unimportant and in general this is true for younger age groups. However with older groups minor differences in average age within age band may become important, especially as all over 85’s are grouped together. The European Standard Population is younger than the UK population and therefore biases comparisons towards younger age groups. Although age standardisation will reduce the effect of age differences in the population it cannot be assumed to eliminate them.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Isles of Scilly and City of London are excluded from local authority level but are included for higher level geographies.
Disclosure Control
Not applicable
Confidence Intervals calculation method
95% and 99.8% confidence intervals were calculated using a method described by Dobson1. The Byar’s1 method has been used when the numerator is greater than 388. This is less accurate for small numerators so for numerators less than 389 an exact method based on the Poisson distribution2 has been used. The methodology is explained in detail in the APHO Technical Briefing on Commonly Used Public Health Statistics and their Confidence Intervals http://www.apho.org.uk/resource/item.aspx?RID=48457 and the associated Excel tool http://www.apho.org.uk/resource/view.aspx?RID=48617. 1. Dobson A et al. Confidence intervals for weighted sums of Poisson parameters. Stat Med 1991;10:457-62. 2. Breslow NE, Day NE. Statistical methods in cancer research, volume II: The design and analysis of cohort studies. Lyon: International Agency for Research on Cancer, World Health Organisation; 1987. 3. Armitage P, Berry G. Statistical methods in medical research (3rd edn). Oxford: Blackwell; 1994.
Health Profiles 2009 The Indicator Guide
172 Section 5: disease and poor health
19. MENTAL HEALTH INDICATOR Basic Information 1. What is being measured?
Working age people who are in receipt of benefits for mental health conditions.
2. Why is it being measured?
People with long term psychiatric disabilities are less likely to be in employment than those with longterm physical disabilities, despite indications that most people with severe mental illness would like to work. To help improve the provision of services to help mentally ill people find work and reduce social exclusion.
3. How is this indicator actually defined?
Claimants/beneficiaries of incapacity benefit/severe disablement allowance, with mental or behavioural disorders. Crude rate, all persons of working age, per thousand working age population, 2007.
4. Who does it measure?
Working age adults: 16–64 males and 16–59 females.
5. When does it measure it?
Reported quarterly and updated every year.
6. Will It measure absolute numbers or proportions?
Proportions: number of working age claimants per thousand working age population.
7. Where does the data actually come from?
Collection and collation from incapacity benefits data via the Department for Work and Pensions.
8. How accurate and complete will the data be?
The data is based on all claims for benefit (100% sample) and double counts are removed.
9. Are there any caveats/warnings/ problems?
Benefit counts are rounded quarterly point estimates from the end of February, May, August and November 2008. Population data are mid-2007 and rounded to the nearest 100.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
Pg 4 Health Summary: Indicator No 19.
Subject category/ domain(s)
Disease and poor health.
Indicator name (*Indicator title in health profile)
Claimants/beneficiaries of incapacity benefit/severe disablement allowance with mental or behavioural disorders (*Incapacity benefits for mental illness).
Health Profiles 2009 The Indicator Guide
173 Section 5: disease and poor health
PHO with lead responsibility
NEPHO
Date of PHO dataset creation
February 2009
Indicator definition
Claimants/beneficiaries of incapacity benefit/severe disablement allowance with mental or behavioural disorders, crude rate, males and females, working age, 2007, Per 1000 working age population.
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs.
Timeliness
This indicator is potentially suitable for time-series analysis.
Rationale: What this indicator purports to measure
Prevalence of working age people with severe mental illness who are in receipt of benefits for mental health conditions.
Rationale: Public Health Importance
This is a proxy measure of levels of severe mental illness in the community, and a direct measure of socio-economic disadvantage in those ‘not in work’ because of mental illness. Severe mental illness severely restricts the capacity to fully participate in society and in particular the employment market. Unemployment rates are high amongst people with severe mental illness. In the UK unemployment rates of 60–100% have been reported. These high rates reflect the disability caused by severe mental illness, but they also reflect discrimination (unemployment rates are higher than in other disabled group) and the low priority given to employment by psychiatric services. People with long-term psychiatric disabilities are even less likely to be in employment than those with long-term physical disabilities. Despite high unemployment rates amongst the severely mentally ill, surveys have consistently shown that most want to work. These low rates of employment should be considered against the facts that at least 30-40% of people who are significantly disabled by enduring mental illness are capable of holding down a job. More than 900,000 adults in England claim sickness and disability benefits for mental health conditions. This group is now larger than the total number of unemployed people claiming Jobseeker’s Allowance in England. Vocational rehabilitation services can help mentally ill people find work. See: Crowther R, Marshall M, Bond G, Huxley P. Vocational rehabilitation for people with severe mental illness. Cochrane Database of Systematic Reviews 2001, Issue 2. Art. No.: CD003080. DOI: 10.1002/14651858.CD003080 See Royal College of Psychiatrists Memorandum to Select Committee on Work and Pensions available at: http://www.publications.parliament.uk/pa/cm200506/cmselect/ cmworpen/616/6022719.htm
Rationale: Purpose behind the inclusion of the indicator
To estimate the prevalence of those with severe mental illness who are not in work because of their mental ill health. To help improve the provision of services for helping mentally ill people find work
Health Profiles 2009 The Indicator Guide
174 Section 5: disease and poor health
Rationale: Policy relevance
In June 2004 the Social Exclusion Unit of the Office of the Deputy Prime Minister (ODPM) published the ‘Mental Health and Social Exclusion’ report. This report highlights the large number of adults in England claiming sickness and disability benefits for mental health conditions (approximately 40% of all claims) with the statistics showing that more adults now fall into this group than the total number of unemployed people claiming Jobseeker’s Allowance. The report draws attention to the fact that adults with longterm mental health problems are one of the most excluded groups in society facing numerous barriers that only serve to stop them from achieving their full potential as individuals and members of the community as a whole. Current government policy is focusing on finding ways of returning people to work via initiatives such as Pathways to Work, New Deal for Disabled People, and Disability Employment Brokers. Mental Health National Service Framework
Interpretation: What a high/low level of indicator value means
An indicator value worse than average (red circle in health summary chart) represents a statistically significant worse rate of benefit claimants for mental and behavioural disorders for that local authority when compared to the national value. An indicator value better than average (green circle in health summary chart) represents a statistically significant better rate of benefit claimants for mental and behavioural disorders for that local authority when compared to the national value.
Interpretation: Potential for error due to type of measurement method
The first 28 weeks of incapacity are assessed under the “own occupation test” which looks at a person’s ability to do their usual job and is based on medical certificates from a GP. After this time, the personal capability assessment (PCA) applies which involves completing an incapacity questionnaire (IB50) that assesses ability to do any work. Forms are difficult to understand and complete. There may be an “institutional bias” against people with mental health problems in the incapacity benefit questionnaire which does not establish information about fluctuating conditions. The DWP uses a system that allocates points to certain activities and tasks, with 10 points needed to determine a person’s eligibility on grounds of mental incapacity. Mental health descriptors are not itemised in the same detail as physical descriptors and a client is therefore less likely to answer in the way most helpful to a successful outcome in their case. Problems those with severe mental illness face with personal capability assessments conducted by Medical Services include: • doctors not listening to clients; • poor recording of clinical findings; • incorrect assumptions based on information from the client and from the medical examination; • effects of mental illness not appropriately taken into account by the scoring system employed; and • difficulties in arranging home visits for some clients. The assessment process may lead to underestimation of unemployment due to severe mental illness. See Royal College of Psychiatrists Memorandum to Select Committee on Work and Pensions available at: http://www.publications.parliament.uk/pa/cm200506/cmselect/ cmworpen/616/6022719.htm
Health Profiles 2009 The Indicator Guide
175 Section 5: disease and poor health
Interpretation: Potential for error due to bias and confounding
There may be disincentives to return to work: These include: • Medical review process: There are concerns that engagement in voluntary work, education and training – which can help get back to work - may trigger the medical review process. • Permitted work rules: These may deter people who will be at risk of losing incapacity benefit and associated benefits. • Transition to work may disrupt income: People returning to work will lose housing benefit. Should the return to work fail, they will be at risk of losing their home. • Fear of drop in income. • 52-week linking rule: The time period may not be long enough for people whose illness has a relapsing and remitting course. (For example, the average time to clinical relapse for people being treated for schizophrenia is approximately two years.) • Difficulties in getting financial support for practical needs at work. • Lack of expert, independent benefits advice on return to work. Therefore severe mental illness as a cause of inability to participate in the labour market may be overestimated due to these disincentives to return to work. See Royal College of Psychiatrists Memorandum to Select Committee on Work and Pensions available at: http://www.publications.parliament.uk/pa/cm200506/cmselect/ cmworpen/616/6022719.htm There are groups of people who may be less likely to engage with, or be able to benefit from, the benefit system eg certain ethnic minorities, married women, those with functional illiteracy, rough sleepers. This may underestimate the problem.
Health Profiles 2009 The Indicator Guide
176 Section 5: disease and poor health
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Table 2 – Indicator Specification Indicator definition: Variable
Claimants/beneficiaries of incapacity benefit/severe disablement allowance with mental or behavioural disorders.
Indicator definition: Statistic
Crude rate.
Indicator definition: Gender
Males and Females.
Indicator definition: age group
16–64 yrs (males working age) 16– 59 yrs (females working age)
Indicator definition: period
2007
Indicator definition: scale
Per 1000 working age population.
Health Profiles 2009 The Indicator Guide
177 Section 5: disease and poor health
Geography: geographies available for this indicator from other providers
Super Output Area, Wards, England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs.
Dimensions of inequality: subgroup analyses of this dataset available from other providers
None.
Data extraction: Source
Numerator: NOMIS Denominator: Office for National Statistics (ONS)
Data extraction: source URL
Numerator: http://www.nomisweb.co.uk Denominator: http://www.statistics.gov.uk/statbase/Product.asp?vlnk=15106
Data extraction: date
4 February 2009.
Numerator: definition
Count of working age claimants/beneficiaries of incapacity benefit/severe disablement allowance with mental or behavioural problems, 2007.
Numerator: source
NOMIS/Department for Work and Pensions.
Denominator: definition
Estimated count of working age people (rounded to nearest 100). For the purposes of calculating this indicator working age has been defined as 16 to 64 years for males and 16 to 59 years for females, 2007.
Denominator: source
Office for National Statistics (ONS).
Data quality: Accuracy and completeness
This is a 100% data source with double-counts removed.
Available from Department of Work & Pensions.
Data relating to the age and gender of benefit claimants is collected by the DWP and may be available upon request.
Table 3 – Indicator Technical Methods Numerator: extraction
Quarterly benefit claimant statistics are published routinely on the NOMIS website (http://www.nomisweb.co.uk). The advanced query tool provided on this website was used to extract relevant data at local authority district level.
Numerator: aggregation/ allocation
Average annual benefit count figures at local authority district level were derived from quarterly benefit count figures taken at the end of February, May, August and November 2007.
Numerator data caveats
The numerator counts are derived from quarterly figures recorded as point estimates taken at the end of February, May, August and November 2007. Aggregated numerator data is based upon rounded data. This can result in a margin of error.
Denominator data caveats
Local authority level working-age mid-2007 population estimates. Data are rounded to nearest 100 to comply with ONS publication policy.
Health Profiles 2009 The Indicator Guide
178 Section 5: disease and poor health
Methods used to calculate indicator value
Mental Health Indicator: Calculation of the numerator: average annual count of claimants/beneficiaries of incapacity benefit/severe disablement allowance February to November 2007 was calculated by adding four quarterly counts together and dividing the result by four. This value was then rounded to the nearest value of ten. Totals for counties and regions were calculated by aggregating up the local authority district level data to the appropriate geography. Calculation of the denominator: Persons level working age population data was downloaded from the ONS website. The local authority values were extracted from this table and these values were then multiplied by 1000 to give unit level figures rounded to the nearest number of 100. For new the unitary authorities where totals of old local authority district level populations did not total up to ONS county council population, the ONS total for the counties was used. The numerator was then divided by the denominator; the resulting value was then multiplied by 1000 to give a crude rate per 1,000 working age population.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Isles of Scilly and City of London have been included in regional and England numerators and denominators. Isles of Scilly has been included in the numerator and denominator for the County of Cornwall.
Disclosure Control
To preserve claimant confidentiality this data is rounded by the DWP prior to publication. Numerator values have been rounded to the nearest value of ten as a means of disclosure control. Denominator values are rounded to the nearest value of 100 as a means of disclosure control.
Confidence Intervals calculation method
The 95% confidence intervals are calculated with the method described by Wilson and by Newcombe which is a good approximation of the exact method. First calculate the estimated proportions of subjects with (p) and without (q) some feature of interest from a sample of size n. proportion with feature of interest = p = r/n proportion without feature of interest = q = 1 - p where r is the observed number of subjects with the feature of interest. Second, calculate the three quantities A = 2r + z2;
B = z z 2 + 4rq ; and C=2(n+z2),
where z is the appropriate value, z1-α/2, from the standard Normal distribution. Then the confidence interval for the population proportion is given by (A-B)/C to (A+B)/C This method has the considerable advantage that it can be used for any data. When there are no observed events, r and hence p are both zero, and the recommended confidence interval simplifies to 0 to z2/(n+z2). When r = n so that p = 1, the interval becomes n/(n+z2) to 1. Wilson EB. J Am Stat Assoc 1927, 22, 209-212 Newcombe, RG. Two-sided confidence intervals for the single proportion: comparison of seven methods. Stat Med 1998;17:857-72.
Health Profiles 2009 The Indicator Guide
179 Section 5: disease and poor health
20. HOSPITAL STAYS FOR ALCOHOL RELATED HARM INDICATOR Basic Information 1. What is being measured?
Hospital stays for Alcohol Related Harm: NI39.
2. Why is it being measured?
The acute or long term effects of excessive alcohol consumption are a major cause of avoidable hospital admissions. This indicator may help to monitor likely health care burden.
3. How is this indicator actually defined?
Hospital Admissions for Alcohol Related Harm (2007/08), directly age and sex standardised rate, all ages, admissions per 100,000 European Standard population.
4. Who does it measure?
All admissions, all ages.
5. When does it measure it?
Continually reported and updated every year.
6. Will it measure absolute numbers or proportions?
Proportions: numbers of admissions per hundred thousand European standard population.
7. Where does the data actually come from?
Collection and collation from Hospital Episode Statistics via the DH.
8. How accurate and complete will the data be?
HES Data and ONS population statistics are considered to be complete and robust.
9. Are there any caveats/ warnings/ problems?
Hospital admission data can be coded differently in different parts of the country.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
Pg 4 Health Summary – Indicator No. 20.
Subject category/ domain(s)
Disease and poor health.
Indicator name (*Indicator title in health profile)
Hospital stays for alcohol-related harm.
Health Profiles 2009 The Indicator Guide
180 Section 5: disease and poor health
PHO with lead responsibility
NWPHO.
Date of PHO dataset creation
March 2009.
Indicator definition
Hospital admissions for alcohol-related harm 2007/08, directly age and sex standardised hospital admissions for all ages, per 100,000 European Standard population, based on residence in the area. Full indicator definition can be found at: http://www.nwph.net/alcohol/lape/NI39Technical_Dec2008.pdf
Geography
The following geographies: England, GOR, Local Authority: County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs, for this and other alcohol related indicators are available from the NWPHO web site. www.nwph.net/alcohol/lape
Timeliness
Updated every year, next update available in 2010. Time trend analysis is appropriate from data available from the NWPHO. Time trend analysis should not be done between Health Profiles 2006, Health Profiles 2007 and Health Profiles 2008 because the indicator calculation changed from multiple years to a single year.
Rationale: What this indicator purports to measure
This indicator measures the number of admissions to hospital for alcohol-related harm. A comprehensive list of ICD10 codes, age and sex specific attributable fractions can be found at: http://www.nwph.net/alcohol/lape/NI39Technical_ Dec2008.pdf
Rationale: Public Health Importance
The acute or long term effects of excessive alcohol consumption are a major cause of avoidable hospital admissions.
Rationale: Purpose behind the inclusion of the indicator
To help monitor likely health care burden.
Rationale: Policy relevance
Alcohol Harm Reduction Strategy for England (2004): http://www.cabinetoffice. gov.uk/media/cabinetoffice/strategy/assets/caboffce%20alcoholhar.pdf. Reducing Alcohol Harm (2008) http://www.dh.gov.uk/en/Publichealth/Healthimprovement/Alcoholmisuse/ DH_085387 This indicator is included in three key indicator sets and performance management frameworks, these are: Vital Signs Indicator VSC26, National Indicator NI39 and Public Service Agreement Indicator 25.2.
Interpretation: What a high/low level of indicator value means
A high indicator value (red circle in health summary chart) represents a statistically significant higher level of hospital admissions for alcohol-related harm for that local authority when compared to the national value. A low indicator value (green circle in health summary chart) represents a statistically significant lower level of hospital admissions for alcohol-related harm for that local authority when compared to the national value.
Interpretation: Potential for error due to type of measurement method
Hospital admission data can be coded differently in different parts of the country.
Health Profiles 2009 The Indicator Guide
181 Section 5: disease and poor health
Interpretation: Potential for error due to bias and confounding Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated significance level. In Health Profiles this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with a white symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or amber symbol depending on whether it is worse or better than the national value respectively.
Table 2 – Indicator Specification Indicator definition: Variable
Hospital Admissions for Alcohol-Related Harm.
Indicator definition: Statistic
Directly age and sex standardised rate.
Indicator definition: Gender
Persons.
Indicator definition: age group
All Ages.
Indicator definition: period
2007/08.
Indicator definition: scale
Per 100,000 European Standard population.
Health Profiles 2009 The Indicator Guide
182 Section 5: disease and poor health
Geography: geographies available for this indicator from other providers
The following geographies: England, GOR, Local Authority: County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs, for this and other alcohol related indicators are available from the NWPHO web site. www.nwph.net/alcohol/lape
Dimensions of inequality: subgroup analyses of this dataset available from other providers
HES data is available by LSOA, therefore analysis by deprivation and geodemographics is possible.
Data extraction: Source
The Department of Health (from HES).
Data extraction: source URL
Hospital Episodes Statistics (HES). Record level access is restricted to authorised users, for example PHOs. Further information regarding HES can be found at: www.hesonline.org.uk
Data extraction: date
February 2009
Numerator: definition
A detailed definition of the numerator data used for this indicator can be found at: http://www.nwph.net/alcohol/lape/NI39Technical_Dec2008.pdf
Numerator: source
Produced by the Department of Health.
Denominator: definition
Mid-year population estimates (2007) by 5-year age band.
Denominator: source
Office for National Statistics (ONS).
Data quality: Accuracy and completeness
HES Data and ONS population statistics are considered to be complete and robust.
TABLE 3 – INDICATOR TECHNICAL METHODS Numerator: extraction
A detailed definition of the extraction method and application of attributable fractions to produce the numerator data used for this indicator can be found at: http://www.nwph.net/alcohol/lape/NI39Technical_Dec2008.pdf
Numerator: aggregation / allocation
Data extracted using area of residence (GOR, County and LA) of patient.
Numerator data caveats
Hospital admission data can be coded differently in different parts of the country. Identification of persons in any one year can also affect the analysis if people are prone to moving within a local authority on a regular basis In some cases the PCT of residence is recorded in HES but not the LA. To ensure that the figures for coterminous PCTs and LAs were the same, details of LA were added where this information could be ascertained from PCT of residency.
Denominator data caveats
Health Profiles 2009 The Indicator Guide
183 Section 5: disease and poor health
Methods used to calculate indicator value
The directly age-standardised rate is the rate of events that would occur in a standard population if that population were to experience the age/sex-specific rates of the subject population. Explicitly:
∑w r DSR = ∑w
i i
i
i
× 100,000
i
(expressed per 100,000 population) where: wi is the number, or proportion, of individuals in the standard population in age/sex group i. ri is the crude age/sex-specific rate in the subject population in age/sex group i, given by:
ri =
Oi ni
where: Oi is the observed number of events in the subject population in age/sex group i. ni is the number of individuals in the subject population in age/sex group i. The standard population generally used for the direct method is the European Standard Population. Small Populations: How Isles of Scilly and City of London populations have been dealt with
Isles of Scilly and City of London were excluded. Though these are available from www.nwph.net/alcohol/lape .
Disclosure Control
Not applicable
Health Profiles 2009 The Indicator Guide
184 Section 5: disease and poor health
Confidence Intervals calculation method
Confidence intervals for the rates were calculated using the method described in the NCHOD Compendium for directly standardised rates. www.nchod.nhs.uk 95% confidence intervals for the age-standardised rates were calculated using a normal approximation. Standard errors are obtained using the method described by Breslow and Day but modified to use the binomial variance for a proportion to estimate the variances of the crude age/sex-specific rates. This method is likely to be unreliable when there are fewer than 50 cases in an area, hence confidence intervals for rates based on less than 50 cases should be viewed with caution. The lower and upper limits for the rates are denoted by DSRLL and DSRUL respectively.
DSRLL / UL = DSR ± 1.96 × 100,000 ×
1 ∑ wi ij
2
×∑ ij
wi2 ⋅ rij (1 − rij ) nij
(expressed per 100,000 population)
where: wi is the number, or proportion, of individuals in the standard population in age/sex group i. rij is the crude age/sex-specific rate in the subject population in age/sex group i, in year j. nij is the number of individuals in the subject population in age/sex group i, in year j. Ref: Breslow NE and Day NE. Statistical Methods in Cancer Research, Volume II: The Design and Analysis of Cohort Studies. Lyon: International Agency for Research on Cancer, World Health Organization, 1987: 59 Keyfitz N. Sampling variance of age-standardised mortality rates. Human Biology. 1966; 38: 309-317.
Health Profiles 2009 The Indicator Guide
185 Section 5: disease and poor health
21. DRUG MISUSE Basic Information 1. What is being measured?
Estimated Problem Drug Users (Crack &/or Opiates)
2. Why is it being measured?
To help monitor likely health care burden from drug misuse.
3. How is this indicator actually defined?
Estimated problem drug users (Crack &/or Opiates), crude rate, 15–64 Ages, 2006/07, persons.
4. Who does it measure?
All persons, 15–64 Ages.
5. When does it measure it?
Continually reported and updated every year
6. Will it measure absolute numbers or proportions?
Proportions: number of cases Per 1,000 residents aged 15–64 years
7. Where does the data actually come from?
Collection and collation from The National Treatment Agency.
8. How accurate and complete will the data be?
County district estimates are based on a model applied to top tier local authorities. This model relied on a probabilistic allocation of cases to Local Authorities based on postal sector of residence. A regression model uses number of users in treatment (2006/07) to disaggregate the County level estimates to County Districts.
9. Are there any caveats/ warnings/ problems?
The number of users in treatment will have impacted on the prevalence estimates published by the Home Office and the regression model which may introduce bias to the estimates.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
Pg 4 Health Summary – Indicator No. 21
Subject category/ domain(s)
Disease and poor health
Indicator name (*Indicator title in health profile)
Drug misuse
PHO with lead responsibility
NWPHO
Date of PHO dataset creation
Revised January 09
Health Profiles 2009 The Indicator Guide
186 Section 5: disease and poor health
Indicator definition
Estimated Problem Drug Users (Crack and/or Opiates), Crude Rate, 15–64 Ages, 2006/07, persons
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs
Timeliness
Every year. Time trend analysis is not appropriate.
Rationale: What this indicator purports to measure
This indicator estimates the number of problem drug users (Crack and/or Opiates) in an area.
Rationale: Public Health Importance
The indicator was chosen as the best available estimate of drug use prevalence in an area.
Rationale: Purpose behind the inclusion of the indicator
To help monitor likely health care burden from drug misuse.
Rationale: Policy relevance
The Department of Health has introduced a ten year drug strategy (20082018) which is aimed at restricting the supply of illegal drugs and reducing the demand for them. http://www.dh.gov.uk/en/Publichealth/Healthimprovement/ Drugmisuse/DH_085886
Interpretation: What a high/low level of indicator value means
A high indicator value (red circle in health summary chart) represents a statistically significant higher estimate of problem (crack and/or opiates) drug users for that local authority when compared to the national value. A low indicator value (amber circle in health summary chart) represents a statistically significant lower estimate of problem (crack and/or opiates) drug users for that local authority when compared to the national value. Confidence Intervals for Top Tier Local authorities (Counties, MCDs, UAs, LBs) were taken from the HO published prevalence data but cannot currently be calculated for County Districts. Therefore interpretation of significance for County Districts cannot be made.
Interpretation: Potential for error due to type of measurement method
The base estimates of the number of problem drug users were published by the Home Office (2004/05) and issues with the methods are outlined in the report (www.homeoffice.gov.uk/rds/pdfs06/rdsolr1606.pdf). A regression model uses number of users in treatment (2006/07) to disaggregate the County level estimates to County Districts. Treatment data is only available with postcode sector of residence which is not available for a large proportion of cases reported by DAATs, consequently the allocation of cases to LAs is likely to be less accurate for some areas than for others.
Interpretation: Potential for error due to bias and confounding
The number of users in treatment will have impacted on the prevalence estimates published by the Home Office and the regression model which may introduce bias to the estimates.
Health Profiles 2009 The Indicator Guide
187 Section 5: disease and poor health
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with a white symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or amber symbol depending on whether it is worse or better than the national value respectively.
Table 2 – Indicator Specification Indicator definition: Variable
Problem Drug Users (Crack and/or Opiates)
Indicator definition: Statistic
Crude Rate
Indicator definition: Gender
Persons
Indicator definition: age group
15–64
Indicator definition: period
2006/07
Indicator definition: scale
Per 1,000 residents aged 15–64 years
Geography: geographies available for this indicator from other providers
The data is available by DAAT from the Home Office and NTA.
Health Profiles 2009 The Indicator Guide
188 Section 5: disease and poor health
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Analysis of the 2006/07 PDU estimates by Index of Multiple Deprivation (IMD, 2007) is available in the APHO publication ‘Indications of Public Health in English Regions 10: Drug Use which will be published in April 2009.
Data extraction: Source
Source of data: The National Treatment Agency
Data extraction: source URL
The data extraction source URL is not available to the public domain
Data extraction: date
Data extracted from source July. 2007
Numerator: definition
Estimate of resident persons aged 15–64 believed to be problem drug users in 2006/07.
Numerator: source
The National Treatment Agency
Denominator: definition
2006 Mid-year population estimates (persons aged 15–64) by Local Authority District, County, Region and England.
Denominator: source
Office for National Statistics (ONS).
Data quality: Accuracy and completeness
This is a modelled estimate of the number of problem drug users in a local authority district. The number of users in treatment is based on probabilistic allocation of cases to Local Authorities from postal sector of residence. The National Treatment Agency provided data on individuals in contact with structured treatment with their DAT of residence and PCT of residence. Each case was allocated into LA and GOR of residence using a schema provided by NWPHO. There were 451 cases in which there were no DAT and PCT of residence, these were not used in the regression model. Coverage includes all DAAT areas.
Table 3 – Indicator Technical Methods Numerator: extraction
Downloaded from NTA website.
Numerator: aggregation/ allocation
See Methods used to calculate indicator value
Numerator data caveats
The number of users in treatment will have impacted on the prevalence estimates published by the Home Office and the regression model which may introduce bias to the numerator (also see above).
Health Profiles 2009 The Indicator Guide
189 Section 5: disease and poor health
Denominator data caveats
The Mid 2006 Resident Population Estimates for All persons15 to 64 year age group has been used. Census data and mid-year estimates are known to be deficient in their estimates of: • Non-white populations • Full-time students • Men aged 20–39 • People living in nursing homes etc • Rough sleepers • Inner-city populations • Households of multiple occupation • Migrants
Methods used to calculate indicator value
The analysis uses the problem drug user prevalence data (2006/07) provided by the HO & NTA for Top Tier Local Authorities (Counties, MCDs, UAs, LBs) and the number of opiate and/or crack users in treatment data by Local Authority District provided by the NTA (2006/07). To allow for disaggregating of the County level prevalence data to County Districts, a model was created based on the number of opiate and/or crack users in treatment and the prevalence for Unitary Authorities and Metropolitan County Districts. The results of this model were then used to estimate the likely prevalence of problem drug users for the given number of opiate and crack users in treatment for County Districts. These totals for County Districts were then reconciled to ensure consistency with the County Prevalence Total. In future the Home Office will calculate direct Multiple Indicator Methods (MIM) and Capture Recapture (CR) estimates of drug users for all LADs. These current estimates (for HP3) are an interim measure until the new HO data are available.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Isles of Scilly and City of London were excluded.
Disclosure Control
Not applicable
Confidence Intervals calculation method
The Confidence Intervals were calculated for the estimated prevalence of drug users (opiate and/or crack users) with the Exact method. This method takes no account of uncertainty in the model used to estimate numbers of drug users at the County Districts.
Health Profiles 2009 The Indicator Guide
190 Section 5: disease and poor health
22. PEOPLE DIAGNOSED WITH DIABETES INDICATOR Basic Information 1. What is being measured?
Prevalence of recorded diabetes
2. Why is it being measured?
Diabetes is a common disease with serious consequences. It is the 5th leading cause of death globally and accounts for about 10% of NHS costs.
3. How is this indicator actually defined?
The prevalence of QOF-recorded diabetes (in adults aged 17+) in the population.
4. Who does it measure?
All persons, all ages
5. When does it measure it?
Patients registered with GP practices, aged 17 and over at midnight on the 31st March 2008, with a coded diagnosis of diabetes on the 1st April 2007. (QOF DM1).
6. Will It measure absolute numbers or proportions?
Proportion (displayed as a percentage): Number of recorded cases of diabetes in adults aged 17+ per 100 resident population.
7. Where does the data actually come from?
The published QOF information is derived from the Quality Management Analysis System (QMAS), a national system developed by NHS Connecting for Health, and is published online by the Information Centre.
8. How accurate and complete will the data be?
The data covers more than 99% of GP-registered patients in England, although not everyone is registered with a GP (especially some groups with particular needs).
9. Are there any caveats/warnings/ problems?
Potential errors in collection, collation and interpretation. It is a measure of recorded diabetes prevalence and not actual prevalence and therefore under-reports groups who are less likely to be registered with a GP, such as ethnic populations, young people, homeless people, migrants and travellers. Record level data is not available therefore we have to apportion practices to LAs based on a look up file created from NSTS data. The NSTS practice file does not completely match the QOF practices and therefore the apportioning may not mirror exactly where the patients come from, but it is the best approximation with the data provided. The diabetes prevalence will not match up with the QOF published prevalence where the areas are identical due to a different denominator being used (QOF uses the practice list size whereas the Health Profiles use the resident population).
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Health Profiles 2009 The Indicator Guide
191 Section 5: disease and poor health
TABLE 1 – INDICATOR DESCRIPTION Information component
P4 Health Summary – Indicator No 22
Subject category/ domain(s)
Disease and poor health
Indicator name (*Indicator title in health profile)
Prevalence of recorded diabetes (“People diagnosed with diabetes”)
PHO with lead responsibility
ERPHO
Date of PHO dataset creation
18th February 2009
Indicator definition
The prevalence of QOF-recorded diabetes (in adults aged 17+) in the population.
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs
Timeliness
Data is extracted from the QMAS system annually in June and published in QPID (quality and prevalence indicators database) in September-October each year.
Rationale: What this indicator purports to measure
Prevalence of recorded diabetes.
Rationale: Public Health Importance
Diabetes is a common disease with serious consequences. It is the 5th leading cause of death globally and accounts for about 10% of NHS costs. The burden falls disproportionately on elderly and ethnic populations. We use the indicator in this context as a proxy for healthcare need and demand (a high prevalence of diabetes can indicate a less healthy population with higher service utilisation). The sequelae of diabetes include blindness, amputation, neuropathy, renal disease, heart disease and other complications. It is treatable and in most cases preventable. Important modifiable risk factors are obesity, diet and lack of physical activity.
Rationale: Purpose behind the inclusion of the indicator
To encourage better collection of the primary care data to give more accurate estimates of disease prevalence. To monitor diabetes prevalence To emphasise the burden of disease To encourage preventative action
Rationale: Policy relevance
The twelve standards of the Diabetes National Service Framework cover all aspects of diabetes care and prevention, and together with the Delivery Strategy, set out a ten-year programme of change and improvement which will raise the quality of services and reduce unacceptable variations. http://www.dh.gov.uk/en/Healthcare/NationalServiceFrameworks/Diabetes/ index.htm
Health Profiles 2009 The Indicator Guide
192 Section 5: disease and poor health
Interpretation: What a high/low level of indicator value means
An indicator value better than average (green circle in health summary chart) represents a statistically significant lower number of people diagnosed with diabetes compared with the national average. An indicator value worse than average (red circle in health summary chart) represents a statistically significant higher number of people diagnosed with diabetes compared with the national average. A high value can indicate a genuinely high prevalence and/or better detection and recording. Conversely a low value may indicate genuinely low prevalence and/or poor detection and recording.
Interpretation: Potential for error due to type of measurement method
See above.
Interpretation: Potential for error due to bias and confounding
There may be under-representation of young people, ethnic populations and other vulnerable groups e.g. homeless, travellers in the numerator as they are less likely to be registered with the GP.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate.
Due to the fact that recording is rewarded through QOF points there may be potential for “gaming”. There are a large number of codes used to record diabetes on GP systems which may lead to counting errors depending on how the data is extracted (see the QOF definitions for the codes used). There may also be potential biases in the attribution of practice populations to local authority areas but these are probably small.
This uncertainty arises because factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider the confidence interval the greater the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the confidence interval can be used to test whether the value is statistically significantly different to the national value. If the confidence interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the confidence interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Health Profiles 2009 The Indicator Guide
193 Section 5: disease and poor health
Table 2 – Indicator Specification Indicator definition: Variable
The prevalence of QOF-recorded diabetes (in adults aged 17+) in the population.
Indicator definition: Statistic
Proportion (displayed as a percentage): Number of recorded cases of diabetes in adults aged 17+ per 100 resident population
Indicator definition: Gender
Persons
Indicator definition: age group
Numerator 17+ Denominator all ages
Indicator definition: period
Financial year 2007/08
Indicator definition: scale
Per 100 resident population
Geography: geographies available for this indicator from other providers
National, SHA, PCT, general practice.
Dimensions of inequality: subgroup analyses of this dataset available from other providers
No (but could be grouped by deprivation score at practice level using derived IMD scores)
Data extraction: Source
Numerator: The Information Centre for health and social care. Denominator: ONS 2007 population estimates Weight: A special extract (mapping general practice populations to local authorities for February 2007) of the NHS Strategic Tracing Service (NSTS) database commissioned from ATOS Origin (available on request). (The NSTS is part of NHS Connecting for Health and is an amalgamation of the Exeter System and the NHS central register.)
Data extraction: source URL
http://www.ic.nhs.uk/statistics-and-data-collections/supporting-information/ audits-and-performance/the-quality-and-outcomes-framework/qof-2007/08/datatables http://www.statistics.gov.uk/statbase/Product.asp?vlnk=15106 http://www.connectingforhealth.nhs.uk/nsts
Data extraction: date
Numerator: QOF data downloaded in October 2008 Denominator: August 2008 Attribution table supplied: February 2009
Health Profiles 2009 The Indicator Guide
194 Section 5: disease and poor health
Numerator: definition
Patients registered with GP practices, aged 17 and over at midnight on the 31st March 2008, with a coded diagnosis of diabetes on the 1st April 2007. (QOF DM1). The case definition of diabetes (with Read codes) can be found at: http://www. primarycarecontracting.nhs.uk/uploads/QOF/qof_bus_rules_v9/diabetes_ruleset_ r4_v9.0.pdf
Numerator: source
Quality and Outcomes Framework (QOF). http://www.ic.nhs.uk/webfiles/QOF/2007-08/Data%20tables/Practice/ Domain%20level%20Spreadsheets/LR/QOF0708_Pracs_Prevalence.xls
Denominator: definition
Mid-year 2007 LA estimates (latest year available). We used a resident population as a denominator rather than an apportioned registered population to avoid the list inflation in registered populations. Denominator data has been used exactly as published in the Home Office supplied data. Where new Unitary Authorities that have been created as part of the April 2009 boundary changes are exactly co-terminous with pre-existing counties (Durham, Northumberland, Shropshire and Wiltshire), the denominator data has been drawn from the relevant county figures. Where the new Unitary Authorities represent only a part of the existing counties (Cornwall, Cheshire East, Cheshire West & Chester, Bedford and Central Bedfordshire), denominator data has been aggregated from the relevant published figures for the constituent districts. We assumed the problems of list counting and inflation did not apply to the numerator (it is very unlikely that diabetics will be counted twice and there are QA checks in QOF to avoid this).
Denominator: source
Office for National Statistics (ONS). http://www.statistics.gov.uk/downloads/ theme_population/Mid_2007_UK_England_&_Wales_Scotland_and_Northern_ Ireland%20_21_08_08.zip
Data quality: Accuracy and completeness
QOF represents 99.8 per cent of registered patients in England (based on registration data from the ePACT system of the Prescription Pricing Division of the NHS Business Services Authority, January to March 2008). There were 8294 practices in the QOFR 200708 dataset, and 99.9% of these could be matched to a LA(s) using the NSTS attribution lookup. Users of data derived from QMAS should recognise that QMAS was established as a mechanism to support the calculation of practice QOF payments and not as a person-based epidemiological tool. It is not a comprehensive source of data on quality of care in general practice, but it is potentially a rich and valuable source of such information, providing that the limitations of the data are acknowledged. See also QOF assessor validation reports available at: http://www. connectingforhealth.nhs.uk/delivery/programmes/qof/
Health Profiles 2009 The Indicator Guide
195 Section 5: disease and poor health
Table 3 – Indicator Technical Methods Numerator: extraction
Numerators extracted from downloaded QOF prevalence by general practice spreadsheets available on the IC website.
Numerator: aggregation/ allocation
We assigned counts of patients with recorded diabetes to give an estimated resident number of diabetics as follows: The weights We obtained a cross boundary flow table from NSTS (February 2007) enabling us to assign GP-registered populations to LA of residence, then for each practice we calculated the number of persons resident in relevant LAs using the following SQL: We discounted any null returns or practices that didn’t match on to QOF, and excluded populations where the number of patients in a practice from a particular ≤50. This LA was made the weighting simpler. In addition, as the practice population is always greater than the actual population of the country, this exclusion brings the total practice population closer to the ONS estimate. We calculated the proportion of each GP-registered population in each LA. We used these proportions to distribute the counts of diabetic patients in each LA, assuming a uniform spatial distribution of diabetes patients practice to each within each practice. We summed the LA estimated counts to whole LAs to give an overall figure for the LA.
Ʃ
This can be expressed in a formula as:
Ʃ
KEY P-LApop: population of a practice from a particular LA in the cases where this
population is >50 The look up table is derived from a Practice–LA look-up table provided by the NSTS Ppop: Total practice population excluding P-LA populations which are ≤50 Pnum: Practice number of patients registered with disease (QOF)
Health Profiles 2009 The Indicator Guide
196 Section 5: disease and poor health
Where the PCT has an identical boundary to the LA, the values we generated for the LA were slightly different from those published for the PCT by QOF because the practice–PCT look-up used by QOF is different from the one we derived from NSTS, therefore to keep our data consistent with the published data, we replaced our calculated figures with the published ones for these 91 PCTs. Due to the fact that a few practices were not included in the apportioning, and the differing look-ups used by the QOF team compared to our generated look-up, the overall figure for registered diabetes patients derived from calculations was less than the overall QOF published figure, and so the remaining LAs that were not identical to PCTs were scaled up slightly using a weight calculated by:
This scaling weight was 1.003, which meant that the non-PCT-identical LA figures were being scaled up by 0.3%. The LA level data was then aggregated to higher levels using look-ups. Numerator data caveats
The allocation method may have incorrectly apportioned patients to LAs particularly for practices straddling LA boundaries. We have not found a satisfactory way of assessing this, as there is no definitive list of practices in England, and the NSTS file does not quite include all practices in QOF and vice versa. There is no age adjustment so variations in prevalence need to be interpreted in the light of variations in age structures. (It may for example, be helpful to correlate the crude prevalence presented here with the proportion of the population over 65)
Denominator data caveats
Subject to limitations of ONS population estimation methods. The denominator does not quite correspond with ageband of the numerator. To keep it consistent with previous Health Profiles all ages were used as the denominator but diabetes was only recorded in patients aged 17+.
Methods used to calculate indicator value
The indicator value is presented as a percentage although strictly is a ratio. There is no age adjustment. The calculation was performed as LA count of diabetes/total LA resident population
Small Populations: How Isles of Scilly and City of London populations have been dealt with
City of London and the Isles of Scilly data were included in the calculation of the regional and national prevalences but they do not contribute to the LA prevalences.
Disclosure Control
Not relevant
Health Profiles 2009 The Indicator Guide
197 Section 5: disease and poor health
Confidence Intervals calculation method
The 95% and 99.8% confidence intervals are calculated using Julian Flowers’ (erpho) confidence interval tool: http://www.erpho.org.uk/Download/
Public/15374/1/Confidence_Intervals_Wilson.xls This calculates confidence intervals using the following method for a confidence interval of a proportion as described by R.G. Newcombe. If r is the observed number of subjects with some feature in a sample of size n then the estimated proportion who have the feature is p = r/n. The proportion who do not have the feature is q = 1-p. First, calculate the three quantities A = 2r + z2;
B = z z 2 + 4rq ; and C=2(n+z2),
where z is z1-α/2, from the standard Normal distribution. Then the confidence interval for the population proportion is given by (A-B)/C to (A+B)/C This method has the considerable advantage that it can be used for any data. When there are no observed events, r and hence p are both zero, and the recommended confidence interval simplifies to 0 to z2/(n+z2). When r = n so that p = 1, the interval becomes n/(n+z2) to 1. Reference Newcombe, R.G. Two-sided confidence intervals for the single proportion: comparison of seven methods. Stat Med 1998;17:857-72.
Health Profiles 2009 The Indicator Guide
198 Section 5: disease and poor health
23. NEW CASES OF TUBERCULOSIS INDICATOR Basic Information 1. What is being measured?
New cases of tuberculosis (TB)
2. Why is it being measured?
To reduce the spread of TB by identifying areas where rates of TB are high.
3. How is this indicator actually defined?
3-year average of TB incidence per 100,000 population, all ages, 2004–06, persons.
4. Who does it measure?
All persons, all ages.
5. When does it measure it?
Continuous monitoring locally, validated periodically at regional level. The most recent year for which data are available is 2006.
6. Will It measure absolute numbers or proportions?
Proportion: Crude rate per 100,000 population.
7. Where does the data actually come from?
Health Protection Agency.
8. How accurate and complete will the data be?
The national surveillance system has a high level of completeness and accuracy.
9. Are there any caveats/warnings/ problems?
A small number of cases may be assigned to the wrong local authority due to missing/incorrect postcodes. TB incidence is highest in inner-cities and concentrated in certain population groups such as the homeless, prisoners, substance mis-users and communities linked to high-TB incidence countries.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
Pg 4 Health Summary – Indicator No 24
Subject category/ domain(s)
Disease and poor health
Indicator name (*Indicator title in health profile)
New cases of tuberculosis
PHO with lead responsibility
South West Public Health Observatory
Date of PHO dataset creation
February 2008, recalculated for new April 2009 boundaries by the Health Protection Agency and SWPHO in February 2009.
Indicator definition
3-year average of TB incidence per 100,000 population, 2004–06.
Health Profiles 2009 The Indicator Guide
199 Section 5: disease and poor health
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs.
Timeliness
Continuous monitoring locally, validated periodically at regional level. Most recent complete data available is 2006.
Rationale: What this indicator purports to measure
The incidence (number of new cases per 100,000 population) of TB notified during the year.
Rationale: Public Health Importance
TB has re-emerged as a major public health problem and is the leading cause of death worldwide among curable infectious diseases. In England cases fell progressively until 1987 but started to rise again in the late 1980s. Over 8,000 new cases are now being reported each year in the UK. TB remains a disease associated with socio-economic deprivation and largely affects migrants from high incidence countries and deprived sub-groups of the population.
Rationale: Purpose behind the inclusion of the indicator
To reduce the spread of TB by identifying areas where rates of TB are high. Incidence is high amongst [first and subsequent generation] migrant populations so this indicator could identify where resources should be targeted to support TB control including the prompt identification of cases and measures to ensure treatment completion. Other high risk groups include the homeless, problem drug users and prisoners. It is very important that there are Health Protection indicators incorporated in to the wider health profiles and although many Health Protection data are not yet available at LA, work is being undertaken to ensure that this happens. TB therefore offers a good opportunity to provide information on a re-emerging PH issue and support the Chief Medical Officer’s (CMO) Action Plan.
Rationale: Policy relevance
The CMO has identified TB as a re-emerging threat which needs concerted action to address. In response to this, the National Institute for Health and Clinical Excellence (NICE) published guidelines which describe recommended measures for the control of TB. The Department of Health recently published “Tuberculosis prevention and treatment: a toolkit for planning, commissioning and delivering high-quality services in England”, outlining the standards for the delivery of tuberculosis services, in June 2007. This indicator, therefore, provides a marker to monitor progress towards the achievement of the objectives of the CMO’s Action Plan and supports the implementation of the TB toolkit. In addition, the European Centre for Disease Control has also published an action plan which outlines the disease prevention and control measures expected of European Union member states. This indicator will help in supporting the implementation of the plan.
Interpretation: What a high/low level of indicator value means
A high indicator value (red circle in health summary chart) represents a statistically significant worse rate of TB incidence when compared to the England average value. A low indicator value (green circle in health summary chart) represents a statistically significant better rate of TB incidence when compared to the England value. However a low indicator value should not mean that public health action is not needed as TB is preventable and any cases of tuberculosis need to be treated. The prompt diagnosis and treatment of infectious cases is key to halting the health burden which TB causes in the UK.
Health Profiles 2009 The Indicator Guide
200 Section 5: disease and poor health
Interpretation: Potential for error due to type of measurement method
New cases of TB are reported through a voluntary enhanced surveillance system to collect detailed data on each case. It provides the most accurate measurement of TB incidence. The quality of national TB surveillance improved significantly since the introduction of this system in 1999. A national study, undertaken in 2003, on the completeness of national surveillance suggests that there may be undernotification of as much as 15% of cases.
Interpretation: Potential for error due to bias and confounding
The rates are not age-standardised or adjusted to take into account variation between areas in risk factors associated with TB. In the UK, TB is highest in inner cities and concentrated in certain population groups such as the homeless, prisoners, substance mis-users and communities linked to high-TB incidence countries
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Table 2 – Indicator Specification Indicator definition: Variable
New cases of tuberculosis
Indicator definition: Statistic
3-year average (mean) of TB incidence per 100,000 population
Indicator definition: Gender
All persons
Health Profiles 2009 The Indicator Guide
201 Section 5: disease and poor health
Indicator definition: age group
All ages
Indicator definition: period
Numerator 2004–2006, denominator mid-year population estimates
Indicator definition: scale
Per 100,000 resident population
Geography: geographies available for this indicator from other providers
Additional data, mostly at national level, available on Health Protection Agency website www.hpa.org.uk
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Not available
Data extraction: Source
Health Protection Agency
Data extraction: source URL
Data at Local Authority level not published elsewhere.
Data extraction: date
Data extracted from source (Enhanced Tuberculosis Surveillance system ) as at: 27/11/2007
Numerator: definition
Average number of TB cases reported between 2004–2006 in accordance with the agreed case definition
Numerator: source
Health Protection Agency
Denominator: definition
ONS mid-year population estimates
Denominator: source
Office for National Statistics (ONS).
Data quality: Accuracy and completeness
Data are provided through the Enhanced Tuberculosis Surveillance system, which commenced in January 1999. It is therefore robust, standardised and comprehensive throughout England, Wales and Northern Ireland.
TABLE 3 – INDICATOR TECHNICAL METHODS Numerator: extraction
Health Protection Agency
Numerator: aggregation/ allocation
Cases were allocated to Local Authority based on their postcode of residence and summarised as a 3-year (mean) average 2004–06.
Numerator data caveats
The TB incidence was obtained by deriving the local authority of residence for each TB case. Incorrectly entered postcode of residence or local authority may have resulted in assigning some cases to the incorrect local authority or county. A small number of cases for whom the local authority was not known were included only in the England total.
Health Profiles 2009 The Indicator Guide
202 Section 5: disease and poor health
Denominator data caveats
As the data were for a period of three years (2004-2006), the mid-year population estimate for the middle year (2005) was used. This middle year may not always be representative of the three year period.
Methods used to calculate indicator value
Calculation of the numerator: Calculated a 3-year average (mean) for 2004– 2006 by summing the individual years and dividing by 3. Denominator count: Mid-year 2005 population estimates (all ages). The numerator was then divided by the denominator; the resulting value was then multiplied by 100,000 to give a crude rate per 100,000 population. The resulting rate was rounded down to an integer value.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
These are included in the totals for their regions (respectively, South West and London).
Disclosure Control
Indicator values for areas where there were <5 cases over the 3-year period (giving an average of 1 or 2 per year) are suppressed and these areas are given a value and rate of zero. The rates for all areas were rounded down to an integer value.
Confidence Intervals calculation method
The confidence intervals for these crude rates were constructed using the following formula that relates the chi-square and Poisson distributions:
χ α ,2d 2
LL =
2
2
χ 1- α ,2(d+1) 2
2
UL =
2
where LL and UL are the lower and upper 100*(1-a) per cent confidence limits and d denotes the number of observed events (e.g, serious injuries, (100*α)offences, violent deaths) per unit of time exposed. χ² a/2,2d is the (100*a/2)th percentage point for a chi-squared distribution with 2d degrees of freedom and χ²(1-α/2),2(d+1) is the (100*(1-a/2))th percentage point for a chi-squared distribution on 2(d+1) degrees of freedom.
The confidence limits for the rates were then obtained by dividing the upper and lower limits for the counts by the person time exposed. Reference: Dobson AJ, Kuulasmaa K, Eberle E, Scherer J. Confidence intervals for weighted sums of Poisson parameters. Statistics in Medicine 1991;10:457-462.
Health Profiles 2009 The Indicator Guide
203 Section 5: disease and poor health
24. HIP FRACUTRE IN OVER-65s INDICATOR Basic Information 1. What is being measured?
Hip fractures in over 65 year olds.
2. Why is it being measured?
Hip fracture is the most common injury related to falls in older people. More than 95% of hip fractures in adults ages 65 and older are caused by a fall. Hip fractures in the elderly and frail can lead to loss of mobility and loss of independence. For many older people it is the event that forces them to leave their homes and move into residential care. Mortality after hip fracture is high: around 30% for one year. To stimulate discussion and encourage local investigation, and to lead to improvement in data quality and quality of care.
3. How is this indicator actually defined?
Emergency Hospital Admission for fractured neck of femur, directly age-standardised rate, 65 year and over, 2006–07, persons.
4. Who does it measure?
All persons, 65 years and over.
5. When does it measure it?
Created specifically for HP3 2008. Regularly updated. HES data for 2007/2008 was not available for inclusion in HP4 2009. However, HP3 2008 data have been updated to accommodate the newly created Unitary Authorities in England.
6. Will It measure absolute numbers or proportions?
Proportions: numbers of cases per hundred thousand European standard population.
7. Where does the data actually come from?
Collection and collation from Hospital Episodes Statistics (HES) via the NHS Information Centre.
8. How accurate and complete will the data be?
HES data and ONS population statistics are considered to be complete and robust.
9. Are there any caveats/warnings/ problems?
There may be variation between Trusts in the way hospital admissions are coded. Routine data do not allow for all of these aspects to be identified and removed from the indicator, however, this may be done through local audit. Population estimates used are Persons rounded and therefore may produce slight differences in rate calculation from those done locally. Population estimates of higher geographies such as County, SHA, GOR and England are aggregated of the LA populations.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Health Profiles 2009 The Indicator Guide
204 Section 5: disease and poor health
Table 1 – Indicator Description Information component
Pg 4 Health Summary Indicator No 25
Subject category/ domain(s)
Diseases and poor health
Indicator name (*Indicator title in health profile)
Hip fracture in over-65s
PHO with lead responsibility
WMPHO
Date of PHO dataset creation
11/02/08
Indicator definition
Emergency Hospital Admission for fracture neck of femur, directly agestandardised rate, 65 year and over, 2006-07, Persons
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs
Timeliness
Created specifically for HP3 2008. Not regularly updated.
Rationale: What this indicator purports to measure
It is an estimate of serious falls in older people.
Rationale: Public Health Importance
Hip fracture is a major cause of disability and the leading cause of mortality due to injury in older people aged over 75. Hospital admission for fractured neck of femur is a good proxy measure of the incidence of hip fracture in older people. Falls prevention programmes aim to reduce the incidence of fractured neck of femur in the community. Hip fracture is the most common injury related to falls in older people. More than 95% of hip fractures in adults ages 65 and older are caused by a fall. Hip fractures in the elderly and frail can lead to loss of mobility and loss of independence. For many older people it is the event that forces them to leave their homes and move into residential care. Mortality after hip fracture is high: around 30% for one year.
Rationale: Purpose behind the inclusion of the indicator
To monitor the incidence of fracture neck of femur. To monitor public health programmes aiming to reduce the risk of older people falling. To stimulate discussion and encourage local investigation, and to lead to improvement in data quality and quality of care.
Health Profiles 2009 The Indicator Guide
205 Section 5: disease and poor health
Rationale: Policy relevance
Standard 6 of the National Service Framework for Older People aims to “reduce the number of falls which result in serious injury and ensure effective treatment and rehabilitation for those who have fallen”. A New ambition for old age (DH, 2006), which outlines the next steps in implementing the NSF, lists ‘falls and bone health’ as one of its 10 programmes, and outlines the components of integrated falls services. There is NICE Guidance on the assessment and prevention of falls in older people (NICE, 2004). Studies have indicated that falls prevention services can reduce falls (Interventions to prevent falls in elderly people can be effective. Cochrane Collaboration. 2003. http://www.cochrane.org/reviews/en/ab000340.html)
Interpretation: What a high/low level of indicator value means
A high indicator value (red blob in spine chart) represents a statistically significant higher level of estimated incidence of fractured neck of femur for that local authority when compared to the national value.
Interpretation: Potential for error due to type of measurement method
There may be variation between Trusts in the way hospital admissions are coded. Routine data do not allow for all of these aspects to be identified and removed from the indicator, however, this may be done through local audit.
Interpretation: Potential for error due to bias and confounding
In order to allow comparison of groups with different age structures it is common to present “age standardised” rates. These are calculated by summing the product of age specific rates for each age band in the group by the number in that age band in the standard population. The sum is then divided by the total number in all age bands in the standard population to obtain the age standardised rate. Any difference between groups in age standardised rate is then not due to difference in age structure since the same standard population was used to calculate all age standardised rates. The method does however assume that minor differences in age structure within age bands are unimportant and in general this is true for younger age groups. For the age standardised rates given in this report five year age bands have generally been used. However with older groups minor differences in average within age band may become important and some authorities argue that for older age bands one year groups should be used.
A low indicator value (green blob) represents a statistically significant lower level of estimated incidence of fractured neck of femur for that local authority when compared to the national value.
The problem is further compounded by the use of the European Standardised Population which has a top age band of 85 years and over. This means that an age band of 85 years and over also has to be used for age specific rates and this is clearly unsatisfactory. Further the European Standard population is much younger than the UK population and therefore biases comparisons towards younger age groups. The rank order of age standardised rates may be appreciably changed by use of different standard populations. Therefore for groups above 65 and particularly for groups including those over age of 85 the use of age standardised rates will reduce the effect of age differences in the populations but cannot be assumed to have eliminated it.
Health Profiles 2009 The Indicator Guide
206 Section 5: disease and poor health
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider the confidence interval, the greater the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Table 2 – Indicator Specification Indicator definition: Variable
Emergency Hospital Admission for fractured neck of femur
Indicator definition: Statistic
Directly age-standardised rate
Indicator definition: Gender
Persons
Indicator definition: age group
65 years and over
Indicator definition: period
2006–07 Financial year
Indicator definition: scale
Per 100,000 European Standard population
Geography: geographies available for this indicator from other providers
None. Dataset unique to WMPHO.
Health Profiles 2009 The Indicator Guide
207 Section 5: disease and poor health
Dimensions of inequality: subgroup analyses of this dataset available from other providers
None available
Data extraction: Source
Hospital Episode Statistics (HES) for the respective financial year, England, NHS Health and Social Care Information Centre. HES Universe 2006/2007 (universe ID HIP0606) This material is Crown Copyright but may be reproduced without formal permission or charge for personal or in-house use. © Crown Copyright 2004
Data extraction: source URL
Extracted through direct link from HES, NHS Health and Social Care Information Centre.
Data extraction: date
Data extracted from source as at: 28th January 2008
Numerator: definition
Emergency Hospital Admissions for primary diagnosis of fractured neck of femur in 65+ age group. Diagnosis of fractured neck of femur classified by primary diagnosis (ICD10 S72.0, S72.1, S72.2), admitted in the respective financial year. ICD10 codes for fractured proximal femur refer to the following diagnoses: • S72.0 Fracture of neck of femur; • S72.1 Pertrochanteric fracture; • S72.2 Subtrochanteric fracture
Numerator: source
Hospital Episode Statistics (HES) for the respective financial year, England, NHS Health and Social Care Information Centre. HES Universe 2006/2007 (universe ID HIP0606) This material is Crown Copyright but may be reproduced without formal permission or charge for personal or in-house use. © Crown Copyright 2004
Denominator: definition
Denominator data – Mid year population estimates 2006 for persons aged 65+ rounded. Data are based on the latest revisions of ONS mid-year population estimates for the respective years, last updated 21/08/07.
Denominator: source
Office for National Statistics (ONS).
Data quality: Accuracy and completeness
Hospital Episode Statistics (HES) are compiled from data sent by over 300 NHS Trusts and Primary Care Trusts (PCTs) in England. The Health and Social Care Information Centre liaises closely with these organisations to encourage submission of complete and valid data and seeks to minimise inaccuracies and the effect of missing and invalid data via HES processes. Whilst this brings about improvement over time, some shortcomings remain. HP2007 data for this indicator was extracted for all admissions. There was considerable variation in patterns of admission for fractured neck of femur across the country with the numbers of transfer admissions between hospitals varying from 2.6% in North East Region to 21.6% in South West Region. Specific problems with over counting due to incorrect coding of internal transfers giving new admissions were identified in Ashfield and Nottingham City. These were examined locally and admission to patient ratios calculated. WMPHO then investigated using emergency admissions only and produced almost identical results in these two LAs. The decision to extract emergency admissions only should reduce erroneous geographical variations.
Health Profiles 2009 The Indicator Guide
208 Section 5: disease and poor health
Table 3 – Indicator Technical Methods Numerator: extraction
HES extraction The number of finished admission episodes including transfers, for 65+ ages with any emergency method of admission and with any of the following primary diagnoses (DIAG_01, ICD 10 codes) in the respective financial year:Fractured proximal femur • S72.0 Fracture of neck of femur; • S72.1 Pertrochanteric fracture; • S72.2 Subtrochanteric fracture Data were extracted as Finished Admission Consultant Episodes The data was ungrossed Counts are by: age/sex/organisation of residence in Finished Admission Episode (values for England are aggregates of these) where: age bands 65–69, 70–74, 75–79, 80-84, 85+; sex is 1, 2 (male and female) Method of admission = Emergency.
Numerator: aggregation /allocation
Residency (by Local Authority) of each Finished Admission Episodes is allocated by HES. Since the 2001 census new Unitary Authorities have been created. The new Unitary Authorities of Northumberland, Shropshire, Wiltshire and Durham have the same boundaries as the counties of the same name so have been allocated these data. Data for five other new unitary Authorities have been calculated from the aggregation of previous Local Authorities. New UA name Bedford Central Bedfordshire Cheshire East
Cheshire West and Chester
Cornwall*
Old LA name Bedford Mid Bedfordshire South Bedfordshire Congleton Crewe and Nantwich Macclesfield Chester Ellesmere Port & Neston Vale Royal Caradon Carrick Kerrier North Cornwall Penwith Restormel
Health Profiles 2009 The Indicator Guide
209 Section 5: disease and poor health
Numerator data caveats
The value shown in the health profiles is for emergency admissions with a primary diagnosis of fractured neck of femur as defined by ICD10 codes S72.0 to S72.2. Fractured neck of femur is sometimes defined as S72.0 only, or as the whole S72.0-S72.9 grouping, or as various other combinations within the S72.0–S72.9 range. The value shown in the health profiles is for primary diagnosis of fractured neck of femur only. If local values have used all diagnosis of fractured neck of femur they will have higher values. HP2007 data for this indicator was extracted for all admissions. There was considerable variation in patterns of admission for fractured neck of femur across the country with the numbers of transfer admissions between hospitals varying from 2.6% in North East Region to 21.6% in South West Region. Specific problems with over counting due to incorrect coding of internal transfers giving new admissions were identified in Ashfield and Nottingham City. These were examined locally and admission to patient ratios calculated. WMPHO then investigated using emergency admissions only and produced almost identical results in these two LAs. The decision to extract emergency admissions only should reduce erroneous geographical variations.
Denominator Population estimates used are Persons rounded and therefore may produce slight data caveats differences in rate calculation from those done locally. Population estimates of higher geographies such as County, SHA, GOR and England are aggregates of the LA populations. Methods used to calculate indicator value
The directly age-standardised rate is the rate of events that would occur in a standard population if that population were to experience the age-specific rates of the subject population. Explicitly:
∑w r DSR = ∑w
i i
i
i
× 100,000
i
(expressed per 100,000 population) where: wi is the number, or proportion, of individuals in the standard population in age group i. ri is the crude age-specific rate in the subject population in age group i, given by:
ri =
Oi ni
where: Oi is the observed number of events in the subject population in age group i. ni is the number of individuals in the subject population in age group i. The standard population generally used for the direct method is the European Standard Population. The age groups used are: 65–69, 70–74, 75–79, 80–84, 85+.
Health Profiles 2009 The Indicator Guide
210 Section 5: disease and poor health
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Scilly Isles and City of London has been excluded at LA level and included at Region and National levels.
Disclosure Control
Not applicable
Confidence Intervals calculation method
Confidence intervals for the rates were calculated using the method described in the NCHOD Compendium for directly standardised rates. www.nchod.nhs.uk 95% confidence intervals for the age-standardised rates were calculated using a normal approximation. Standard errors are obtained using the method described by Breslow and Day but modified to use the binomial variance for a proportion to estimate the variances of the crude age-specific rates. This method is likely to be unreliable when there are fewer than 50 cases in an area, hence confidence intervals for rates based on less than 50 cases should be viewed with caution. The lower and upper limits for the rates are denoted by DSRLL and DSRUL respectively.
DSRLL / UL = DSR ± 1.96 × 100,000 ×
1 ∑ wi ij
2
×∑ ij
wi2 ⋅ rij (1 − rij ) nij
(expressed per 100,000 population)
where: wi is the number, or proportion, of individuals in the standard population in age group i. rij is the crude age-specific rate in the subject population in age group i, in year j. nij is the number of individuals in the subject population in age group i, in year j. Ref: Breslow NE and Day NE. Statistical Methods in Cancer Research, Volume II: The Design and Analysis of Cohort Studies. Lyon: International Agency for Research on Cancer, World Health Organization, 1987: 59 Keyfitz N. Sampling variance of age-standardised mortality rates. Human Biology. 1966; 38: 309-317.
Health Profiles 2009 The Indicator Guide
211
Section 6: Life expectancy and cause of death
Health Profiles 2009 The Indicator Guide
212 Section 6: life expectancy and cause of death
25. EXCESS WINTER DEATHS INDICATOR Basic Information 1. What is being measured?
Excess winter deaths
2. Why is it being measured?
England, in common with other European countries, experiences higher levels of mortality in the winter than in the summer. There is some evidence to suggest that excess winter mortality (EWM) is preventable. Mortality in winter increases more in England compared to other European countries with colder climates, suggesting that it is more than just lower temperatures responsible for the excess mortality in winter1,2.
3. How is this indicator actually defined?
The Excess Winter Mortality Index (EWM Index) is the excess winter deaths expressed as a ratio of the expected deaths based on the non-winter death rate for the period 1.8.04 to 31.7.07.
4. Who does it measure?
The ratio of extra deaths from all causes that occur in the winter months compared to the average of the number of non-winter deaths of the same period.
5. When does it measure it?
All those that died between 1.8.04 to 31.7.07
6. Will It measure absolute numbers or proportions?
Ratio
7. Where does the data actually come from?
The Annual Mortality File provided by ONS for PHOs.
8. How accurate and complete will the data be?
It is the most reliable mortality data source available, and apart from the small number of deaths registered with a long delay due to the coroner’s investigation, it is supposed to be complete.
9. Are there any caveats/warnings/ problems?
EWM Index was calculated based on the “ONS Method” which defines the winter period as December to March, and the non-winter period as August to November of that same year and April to July of the following year. This winter period was selected as they are the months which over the last 50 years have displayed above average monthly mortality. However, if mortality starts to increase prior to this, for example in November, the number of deaths in the non-winter period will increase, which in turn will decrease the estimate of excess winter mortality. The EWM Index will be partly dependent on the proportion of older people in the population as most excess winter deaths effect older people (there is no standardisation in this calculation by age or any other factor).
Health Profiles 2009 The Indicator Guide
213 Section 6: life expectancy and cause of death
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
25
Subject category/ domain(s)
Life expectancy and causes of death
Indicator name (* Indicator title in health profile)
Excess winter deaths
PHO with lead responsibility
ERPHO
Date of PHO dataset creation
19.2.08
Indicator definition
Excess Winter Mortality Index (EWM Index) is the excess winter deaths as a ratio of the expected deaths based on the non-winter death rate for the period 1.8.04 to 31.7.07.
Geography
LA/UA
Timeliness
The most recent year of data available is 2007
Rationale: What this indicator purports to measure
This indicator measures excess winter deaths expressed as the EWM Index, in order that comparisons can be made easily between different geographies. It indicates whether there are higher than expected deaths in the winter compared to the rest of the year.
Rationale: Public Health Importance
The number of excess winter deaths depends on the temperature and the level of disease in the population as well as other factors, such as how well equipped people are to cope with the drop in temperature. Most excess winter deaths are due to circulatory and respiratory diseases, and the majority occur amongst the elderly population. Research carried out by the Eurowinter Group1 and Curwen2 found that mortality during winter increases more in England and Wales compared to other European countries with colder climates, suggesting that many more deaths could be preventable in England and Wales.
Rationale: Purpose behind the inclusion of the indicator
To highlight areas with higher than expected levels of excess winter deaths, to give an idea where interventions need to be improved or instigated to cope with the change of seasonal temperature.
Rationale: Policy relevance
The Department of Health’s annual Keep Warm Keep Well campaign is aimed at financially disadvantaged older or disabled people and their carers, and families with young children on low incomes. It gives information on the health benefits of keeping warm in winter, providing advice on healthy eating and exercise, home heating and energy efficiency, and details of the grants and benefits available.
Health Profiles 2009 The Indicator Guide
214 Section 6: life expectancy and cause of death
Interpretation: What a high/low level of indicator value means
An indicator value better than average (green circle in health summary chart) indicates that excess winter deaths are statistically significantly lower for that local authority compared to the national average.
Interpretation: Potential for error due to type of measurement method
EWM Index was calculated based on the “ONS Method” which defines the winter period as December to March, and the non-winter period as August to November of that same year and April to July of the following year. This winter period was selected as they are the months which over the last 50 years have displayed above average monthly mortality. However, if mortality starts to increase prior to this, for example in November, the number of deaths in the non-winter period will increase, which in turn will decrease the estimate of excess winter mortality.
An indicator value worse than average (red circle in health summary chart) indicates that excess winter deaths are statistically significantly higher for this area compared to the national average.
Interpretation: Potential for error due to bias and confounding Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider the confidence interval the greater the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Health Profiles 2009 The Indicator Guide
215 Section 6: life expectancy and cause of death
Table 2 – Indicator Specification Indicator definition: Variable
The Excess Winter Mortality Index (EWM Index) is the excess winter deaths expressed as a ratio of the expected deaths based on the non-winter deaths for the period 1.8.04 to 31.7.07.
Indicator definition: Statistic
Ratio expressed as a percentage
Indicator definition: Gender
Person
Indicator definition: age group
All ages
Indicator definition: period
Deaths occurring in the period: 1.8.04 to 31.7.07
Indicator definition: scale
In per cent
Geography: geographies available for this indicator from other providers Dimensions of inequality: subgroup analyses of this dataset available from other providers Data extraction: Source
Annual Mortality File provided by ONS for PHOs (2001-2007)
Data extraction: source URL
This is not publically available
Data extraction: date Numerator: definition
The numerator, excess winter deaths, calculated as: Total number of winter deaths i.e. deaths occurring in the months December to March for the period 1.08.04 to 31.07.07 minus half number of deaths in the non-winter months (August to November, April to July).
Numerator: source
Annual Mortality File (ONS)
Denominator: definition
The average number of deaths occurring in the equivalent non-winter period i.e. the deaths occurring in August to November, April to July divided by 2)
Denominator: source
Annual Mortality File (ONS)
Data quality: Accuracy and completeness
It is the most reliable mortality data source available, and apart from the small number of deaths registered with a long delay due to the coroner’s investigation, it is supposed to be complete.
Health Profiles 2009 The Indicator Guide
216 Section 6: life expectancy and cause of death
Table 3 – Indicator Technical Methods Numerator: extraction
The number of winter deaths and non-winter deaths was extracted from the Annual Mortality File using the following SQL: CREATE TABLE P25xx_Winterdeaths_LA (Y_MOD varchar(10), LACode varchar(4), LAName varchar(50), Deaths int) INSERT INTO P25xx_Winterdeaths_LA (Y_MOD, LACode, LAName, Deaths) SELECT Left([DOD],6), [LA Code now], [LA name], COUNT(*) FROM M_DEATHS2001to2007 LEFT JOIN HES.dbo.A_LACounty ON M_DEATHS2001to2007.ResCty + M_ DEATHS2001to2007.ResLAUA = HES.dbo.A_LACounty.[LA Code] LEFT JOIN HES.dbo.A_NationalPostcodes_Feb08 ON M_DEATHS2001to2007. Pcode_8dig = HES.dbo.A_NationalPostcodes_Feb08.[PCD2] where [DOD]>=20040801 and [DOD]<=20070731 Group by [LA Code now], [LA name], Left([DOD],6) ----------------------------------------------------------------------UPDATE P25xx_Winterdeaths_LA SET Y_MOD = ‘S’ WHERE Y_MOD = ‘200408’ or Y_MOD = ‘200409’ or Y_MOD = ‘200410’ or Y_MOD = ‘200411’ or Y_MOD = ‘200504’ or Y_MOD = ‘200505’ or Y_MOD = ‘200506’ or Y_MOD = ‘200507’ or Y_MOD = ‘200508’ or Y_MOD = ‘200509’ or Y_MOD = ‘200510’ or Y_MOD = ‘200511’ or Y_MOD = ‘200604’ or Y_MOD = ‘200605’ or Y_MOD = ‘200606’ or Y_MOD = ‘200607’ or Y_MOD = ‘200608’ or Y_MOD = ‘200609’
Health Profiles 2009 The Indicator Guide
217 Section 6: life expectancy and cause of death
or Y_MOD = ‘200610’ or Y_MOD = ‘200611’ or Y_MOD = ‘200704’ or Y_MOD = ‘200705’ or Y_MOD = ‘200706’ or Y_MOD = ‘200707’ UPDATE P25xx_Winterdeaths_LA SET Y_MOD = ‘W’ WHERE Y_MOD = ‘200412’ or Y_MOD = ‘200501’ or Y_MOD = ‘200502’ or Y_MOD = ‘200503’ or Y_MOD = ‘200512’ or Y_MOD = ‘200601’ or Y_MOD = ‘200602’ or Y_MOD = ‘200603’ or Y_MOD = ‘200612’ or Y_MOD = ‘200701’ or Y_MOD = ‘200702’ or Y_MOD = ‘200703’ -------------------------------------------------------Select LAcode, Y_MOD, sum(deaths) as [Sum] into P25xx_Winterdeaths_EWDI from P25xx_Winterdeaths_LA group by LAcode, Y_MOD order by LAcode, Y_MOD ---------------------------------------------------CREATE TABLE P25xx_Winterdeaths_LA_W (Y_MOD varchar(10), LACode varchar(4), LAName varchar(50), W_Deaths int) Insert into P25xx_Winterdeaths_LA_W (Y_MOD, LACode, LAName, W_Deaths) select Y_MOD, LACode, LAName, Sum(2*Deaths) from P25xx_Winterdeaths_LA where Y_MOD = ‘W’ group by LACode, LAName
Health Profiles 2009 The Indicator Guide
218 Section 6: life expectancy and cause of death
CREATE TABLE P25xx_Winterdeaths_LA_S (Y_MOD varchar(10), LACode varchar(4), LAName varchar(50), S_Deaths int) Insert into P25xx_Winterdeaths_LA_S (Y_MOD, LACode, LAName, S_Deaths) select Y_MOD, LACode, LAName, sum(Deaths) from P25xx_Winterdeaths_LA where Y_MOD = ‘S’ group by LACode, LAName ----------------------------------------------CREATE TABLE P25xx_Winterdeaths_EWDI (LACode varchar(4), LAName varchar(50), [W_deaths_times_2] int, S_deaths int, EWDI decimal(8,4)) Insert into P25xx_Winterdeaths_EWDI (P25xx_Winterdeaths_LA_W.LACode, P25xx_Winterdeaths_LA_W.LAName, [W_deaths_times_2], S_deaths, EWDI) select P25xx_Winterdeaths_LA_W.LACode, P25xx_Winterdeaths_LA_W.LAName, W_deaths, S_deaths, W_deaths/S_deaths from P25xx_Winterdeaths_LA_W join P25xx_Winterdeaths_LA_S on P25xx_Winterdeaths_LA_W.LACode = P25xx_ Winterdeaths_LA_S.LAcode The numerator, excess winter deaths, was then calculated in excel as part of the EWM Index calculation (see below). The numerator published in the template is the average (yearly) excess winter deaths for the three year period, which is calculated: Winter deaths (Dec-Mar)- 0.5 summer deaths (Aug-Nov,Apr-Jul) for the period 1.08.04-31.07.07
3
Health Profiles 2009 The Indicator Guide
219 Section 6: life expectancy and cause of death
Numerator: aggregation/ allocation
The LA/UA data was aggregated up to new UA, county, region, SHA and national level. In April 2009 new Unitary Authorities have been created. The new Unitary Authorities of Northumberland, Shropshire, Wiltshire, and Durham have the same boundaries as the counties of the same name so have been allocated these data. Data for five other new unitary Authorities have been calculated from the aggregation of previous Local Authorities (see table below) New UA name
Old LA name
Bedford
Bedford
Central Bedfordshire
Mid Bedfordshire South Bedfordshire
Cheshire East
Congleton Crewe and Nantwich Macclesfield
Cheshire West and Chester
Chester Ellesmere Port & Neston Vale Royal
Cornwall
Caradon Carrick Kerrier North Cornwall Penwith Restormel
Numerator data caveats Denominator data caveats Methods used to calculate indicator value
Method to calculate the EWM Index was as defined by ONS: excess winter deaths expected number of deaths based on the non-winter death rates = winter deaths (Dec-Mar)- 0.5 non-winter deaths (Aug–Nov,Apr–Jul) 0.5 non-winter deaths (Aug–Nov,Apr–Jul) = 2 * winter deaths (Dec-Mar) non-winter deaths (Aug-Nov, Apr-Jul)
-1
The last arrangement of the formula was used in this calculation. Small Populations: How Isles of Scilly and City of London populations have been dealt with
Isles of Scilly and the City of London were excluded from local authority calculations. The Isles of Scilly and the City of London were included in the higher geographies.
Disclosure Control
Not relevant
Health Profiles 2009 The Indicator Guide
220 Section 6: life expectancy and cause of death
Confidence Intervals calculation method
The EWM index is treated as an odds (a/b) i.e. 2 x a/b -1 Where: a = number of winter deaths, b = number of summer deaths, 2 is a constant and therefore does not need to be included in the confidence interval calculation as a constant has no variance, and -1 is also a constant and can be subtracted at the end. The formula for a calculation of 95% confidence intervals was taken from Kirkwood and Sterne3. 95% CI = odds/EF to odds x EF Where EF = exp[1.96 x s.e.(log odds)] And s.e.(log odds) = √[1/a+1/b] For 99.8% confidence intervals the Standard Error was changed to 3.091
1 The Eurowinter group (1997) Cold exposure and winter mortality from ischaemic heart disease, cerebrovascular disease, respiratory disease, and all causes in warm and cold regions in Europe. The Lancet 349, 1341-1346 2 Curwen M (1990/91) Excess winter mortality: a British phenomenon? Health Trends 4, 169-75 3 Kirkwood B.R. and Sterne J.A.C (2003) Essential Medical Statistics, 2nd Edition, p.164
Health Profiles 2009 The Indicator Guide
221 Section 6: life expectancy and cause of death
26. LIFE EXPECTANCY – MALE INDICATOR Basic Information 1. What is being measured?
Life Expectancy
2. Why is it being measured?
All cause mortality is a fundamental and probably the oldest measure of the health status of a population. Differences in levels of all-cause mortality reflect health inequalities between different population groups, e.g. between genders, social classes and ethnic groups. Life expectancy at birth is chosen as the preferred summary measure of all cause mortality as it quantifies the differences between areas in units (years of life) that are more readily understood and meaningful to the audience than those of other measures.
3. How is this indicator actually defined?
Life expectancy at birth, years, all ages, 2005–07, males.
4. Who does it measure?
Males, all ages.
5. When does it measure it?
This indicator is updated annually.
6. Will it measure absolute numbers or proportions?
Measures life expectancy in years.
7. Where does the data actually come from?
Office for National Statistics (ONS).
8. How accurate and complete will the data be?
Data on deaths are considered to be complete and robust.
9. Are there any caveats/warnings/ problems?
Death records without a valid area code are excluded, but the number of such records is negligible. The populations used for the calculation of the figures for this indicator are based on the 2001 Census, and are estimates.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
Pg 4 Health Summary – Indicator No. 26
Subject category/ domain(s)
Life expectancy and causes of death
Health Profiles 2009 The Indicator Guide
222 Section 6: life expectancy and cause of death
Indicator name (*Indicator title in health profile)
Life expectancy – male
PHO with lead responsibility
LHO
Date of PHO dataset creation
February 2009
Indicator definition
Life expectancy at birth, years, all ages, 2005–07, males
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs.
Timeliness
ONS produced data are updated annually in the Autumn of the following year.
Rationale: What this indicator purports to measure
Life expectancy at birth is a summary measure of the all cause mortality rates in an area in a given period. It is the average number of years a new-born baby would survive, were he or she to experience the particular area’s age-specific mortality rates for that time period throughout his or her life.
Rationale: Public Health Importance
All cause mortality is a fundamental and probably the oldest measure of the health status of a population. It represents the cumulative effect of the prevalence of risk factors, prevalence and severity of disease, and the effectiveness of interventions and treatment. Differences in levels of all-cause mortality reflect health inequalities between different population groups, e.g. between genders, social classes and ethnic groups. Life expectancy at birth is chosen as the preferred summary measure of all cause mortality as it quantifies the differences between areas in units (years of life) that are more readily understood and meaningful to the audience than those of other measures.
Rationale: Purpose behind the inclusion of the indicator
To help reduce premature mortality and facilitate planning of health services at local level.
Rationale: Policy relevance
There is a national health inequalities target for life expectancy which aims to increase average life expectancy at birth in England to 78.6 years for men and to 82.5 years for women, and to reduce health inequalities by 10% by 2010 as measured by life expectancy at birth (Department of Health PSA priority 1). Also life expectancy is an indicator in the following: • Local basket of inequalities indicators - Indicator 13.12. • Opportunity for all - Communities - Indicator 39. • Quality of life indicators - Health and social well-being - Indicator 33
Interpretation: What a high/low level of indicator value means
The higher the life expectancy, the longer the estimated expectation of life for males living in that area at that time.
Health Profiles 2009 The Indicator Guide
223 Section 6: life expectancy and cause of death
Interpretation: Potential for error due to type of measurement method
The figures reflect the contemporary mortality among those living in the area in each time period. They are not the number of years a baby born in the area in each time period could actually expect to live, both because the death rates of the area are likely to change in the future and because many of those born in the area will live elsewhere for at least some part of their lives. Life expectancy at birth is also not a guide to the remaining expectancy of life at any other given age. For example, if female life expectancy at birth was 80 years for a particular area, the life expectancy of women aged 75 in that area would exceed 5 years. This reflects the fact that survival from a particular age depends only on the mortality rates beyond that age, whereas survival from birth is based on mortality rates for all ages.
Interpretation: Potential for error due to bias and confounding
Older people living in nursing homes tend to be in poorer health than those not living in nursing homes. As these homes are unevenly distributed across the country, a higher death rate – consequently lower life expectancy level – in one area could simply be the result of migration of frail older people moving into nursing homes in that area.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider the confidence interval, the greater the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with a white symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or amber symbol depending on whether it is worse or better than the national value respectively.
Health Profiles 2009 The Indicator Guide
224 Section 6: life expectancy and cause of death
TABLE 2 – INDICATOR SPECIFICATION Indicator definition: Variable
Life expectancy at birth
Indicator definition: Statistic
Number of years
Indicator definition: Gender
Males
Indicator definition: age group
All ages
Indicator definition: period
2005–2007
Indicator definition: scale
Years life expectancy
Geography: geographies available for this indicator from other providers
England & Wales, UK and constituent countries, GORs, counties and local authorities Available from the Office for National Statistics (ONS)
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Social Class. Available from the Office for National Statistics (ONS)
Data extraction: Source
Office for National Statistics (ONS)
Data extraction: source URL
http://www.statistics.gov.uk/statbase/Product.asp?vlnk=8841 (See “Results for England and Wales”)
Data extraction: date
January 2009
Numerator: definition
Deaths from all causes, all ages, males
Numerator: source
Office for National Statistics (ONS)
Denominator: definition
Estimated resident population estimates, mid-year, all ages, males
Denominator: source
Office for National Statistics (ONS), available from http://www.statistics.gov.uk/statbase/Product.asp?vlnk=15106 (See mid-year estimates for 2005 to 2007) The population estimates used for the calculation of figures for this indicator are based on the 2001 Census. The population estimates used to calculate the LE figures were published on the following dates: • Mid-year population estimates for 2005 – 22 August 2007; • Mid-year population estimates for 2006 – 25 October 2007; • Mid-year population estimates for 2007 – 21 August 2008.
Health Profiles 2009 The Indicator Guide
225 Section 6: life expectancy and cause of death
Data quality: Accuracy and completeness
Mortality data quality and coverage is extremely high. The figures are threeyear averages so as to provide large enough numbers to ensure that the presented figures are sufficiently robust. Two authorities, City of London and Isles of Scilly, are excluded from the results because of small numbers of deaths and populations in these areas.
Table 3 – Indicator Technical Methods Numerator: extraction
Extraction of the completed indicator from ONS http://www.statistics.gov.uk/statbase/Product.asp?vlnk=8841
Numerator: aggregation / allocation
Deaths were assigned to local authority boundaries by ONS using the National Statistics Postcode Directory.
Numerator data caveats
Source material used to calculate LE: Area of residence is allocated by ONS using the postcode and the National Statistics Postcode Directory – records without a valid area code are excluded but the number of such records is negligible.
Denominator data caveats
Denominators are estimates of mid-year populations, based on results of the 2001 Census.
Methods used to calculate indicator value
The figures are rolling three-year averages produced by aggregating deaths and population estimates for 2005–2007. Abridged life tables were constructed using standard methods. Separate tables were constructed for males and females. The tables were created using annual mid-year population estimates and deaths registered in each year. All figures presented are for life expectancy at birth. A life table template which illustrates the method used to calculate the life expectancy results is available on the ONS website: http://www.statistics.gov.uk/statbase/ssdataset.asp?vlnk=6949 Data for some new Unitary Authorities, introduced in April 2009, did not correspond with existing geographies (Cheshire East, Cheshire West and Cheshire, Central Bedfordshire, and Cornwall). Results for these areas were calculated using the South East Public Health Observatory life expectancy calculator: http://www.sepho.org.uk/viewResource.aspx?id=8943 This was also used to calculate figures for the two Strategic Health Authorities in the South East: South Central and South East Coast. Results for the other new UAs were taken from published data for local authorities (Bedford) or counties (County Durham, Northumberland, Shropshire, and Wiltshire).
Small Populations: How Isles of Scilly and City of London populations have been dealt with
City of London and Isles of Scilly are excluded from the results because of small numbers of deaths and populations in these areas. Isles of Scilly has not been included in the new UA for Cornwall and City of London has not been included with Hackney. However, Isles of Scilly and City of London data are all included in the regional figures for the South West and London, and England figures.
Disclosure Control
Not applicable
Health Profiles 2009 The Indicator Guide
226 Section 6: life expectancy and cause of death
Confidence Intervals calculation method
The calculation of the confidence intervals was made using the method developed by Chiang. A report which details research undertaken by the Office for National Statistics on comparing methodologies to enable the calculation of confidence intervals for life expectancy at birth has now been published as No 33 in the National Statistics Methodological Series. This report, “Life expectancy at birth: methodological options for small populations”, also presents research carried out to establish if there is a minimum population size below which the calculation of life expectancy may not be considered feasible. It concludes with a summary of methodological conclusions and considers how these could be applied to the calculation of life expectancy at birth for wards in England and Wales. A copy of the report can be found on the NS website at: http://www.statistics.gov.uk/methods_quality/publications.asp The SEPHO calculator however also includes an adjustment to include a term for the variance associated with the final age interval as developed by Silcocks. Chiang CL. The Life Table and its Construction. In: Introduction to Stochastic Processes in Biostatistics. New York, John Wiley & Sons, 1968: 189-214. Silcocks PBS, Jenner DA, Reza R. Life expectancy as a summary of mortality in a population: statistical considerations and suitability for use by health authorities. J Epidemiol Community Health 2001; 55: 38-4
Health Profiles 2009 The Indicator Guide
227 Section 6: life expectancy and cause of death
27. LIFE EXPECTANCY – FEMALE INDICATOR Basic Information 1. What is being measured?
Life Expectancy
2. Why is it being measured?
All cause mortality is a fundamental and probably the oldest measure of the health status of a population. Differences in levels of all-cause mortality reflect health inequalities between different population groups, e.g. between genders, social classes and ethnic groups. Life expectancy at birth is chosen as the preferred summary measure of all cause mortality as it quantifies the differences between areas in units (years of life) that are more readily understood and meaningful to the audience than those of other measures.
3. How is this indicator actually defined?
Life expectancy at birth, years, all ages, 2005–07, females.
4. Who does it measure?
Females, all ages.
5. When does it measure it?
This indicator is updated annually.
6. Will it measure absolute numbers or proportions?
Measures life expectancy in years.
7. Where does the data actually come from?
Office for National Statistics (ONS)
8. How accurate and complete will the data be?
Data on deaths are considered to be complete and robust.
9. Are there any caveats/warnings/ problems?
Death records without a valid area code are excluded, but the number of such records is negligible. The populations used for the calculation of the figures for this indicator are based on the 2001 Census, and are estimates.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
Pg 4 Health Summary – Indicator No. 27
Subject category/ domain(s)
Life expectancy and causes of death
Health Profiles 2009 The Indicator Guide
228 Section 6: life expectancy and cause of death
Indicator name (* Indicator title in health profile)
Life expectancy – female
PHO with lead responsibility
LHO
Date of PHO dataset creation
February 2009
Indicator definition
Life expectancy at birth, years, all ages, 2005–07, females
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs.
Timeliness
ONS produced data are updated annually in the Autumn of the following year.
Rationale: What this indicator purports to measure
Life expectancy at birth is a summary measure of the all cause mortality rates in an area in a given period. It is the average number of years a new-born baby would survive, were he or she to experience the particular area’s age-specific mortality rates for that time period throughout his or her life.
Rationale: Public Health Importance
All cause mortality is a fundamental and probably the oldest measure of the health status of a population. It represents the cumulative effect of the prevalence of risk factors, prevalence and severity of disease, and the effectiveness of interventions and treatment. Differences in levels of all-cause mortality reflect health inequalities between different population groups, e.g. between genders, social classes and ethnic groups. Life expectancy at birth is chosen as the preferred summary measure of all cause mortality as it quantifies the differences between areas in units (years of life) that are more readily understood and meaningful to the audience than those of other measures.
Rationale: Purpose behind the inclusion of the indicator
To help reduce premature mortality and facilitate planning of health services at local level.
Rationale: Policy relevance
There is a national health inequalities target for life expectancy which aims to increase average life expectancy at birth in England to 78.6 years for men and to 82.5 years for women, and to reduce health inequalities by 10% by 2010 as measured by life expectancy at birth (Department of Health PSA priority 1). Also life expectancy is an indicator in the following: • Local basket of inequalities indicators - Indicator 13.12. • Opportunity for all - Communities - Indicator 39. • Quality of life indicators - Health and social well-being - Indicator 33
Interpretation: What a high/low level of indicator value means
The higher the life expectancy, the longer the estimated expectation of life for females living in that area at that time.
Health Profiles 2009 The Indicator Guide
229 Section 6: life expectancy and cause of death
Interpretation: Potential for error due to type of measurement method
The figures reflect the contemporary mortality among those living in the area in each time period. They are not the number of years a baby born in the area in each time period could actually expect to live, both because the death rates of the area are likely to change in the future and because many of those born in the area will live elsewhere for at least some part of their lives. Life expectancy at birth is also not a guide to the remaining expectancy of life at any other given age. For example, if female life expectancy at birth was 80 years for a particular area, the life expectancy of women aged 75 in that area would exceed 5 years. This reflects the fact that survival from a particular age depends only on the mortality rates beyond that age, whereas survival from birth is based on mortality rates for all ages.
Interpretation: Potential for error due to bias and confounding
Older people living in nursing homes tend to be in poorer health than those not living in nursing homes. As these homes are unevenly distributed across the country, a higher death rate – consequently lower life expectancy level – in one area could simply be the result of migration of frail older people moving into nursing homes in that area.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider the confidence interval, the greater the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with a white symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or amber symbol depending on whether it is worse or better than the national value respectively.
Health Profiles 2009 The Indicator Guide
230 Section 6: life expectancy and cause of death
Table 2 – Indicator Specification Indicator definition: Variable
Life expectancy at birth
Indicator definition: Statistic
Number of years
Indicator definition: Gender
Females
Indicator definition: age group
All ages
Indicator definition: period
2005–2007
Indicator definition: scale
Years life expectancy
Geography: geographies available for this indicator from other providers
England & Wales, UK and constituent countries, GORs, counties and local authorities
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Social Class.
Data extraction: Source
Office for National Statistics (ONS)
Data extraction: source URL
http://www.statistics.gov.uk/statbase/Product.asp?vlnk=8841 (See “Results for England and Wales”)
Data extraction: date
January 2009
Numerator: definition
Deaths from all causes, all ages, females
Numerator: source
Office for National Statistics (ONS)
Denominator: definition
Estimated resident population estimates, mid-year, all ages, females
Denominator: source
Office for National Statistics (ONS), available from http://www.statistics.gov.uk/statbase/Product.asp?vlnk=15106 (See mid-year estimates for 2005 to 2007)
Available from the Office for National Statistics (ONS)
Available from the Office for National Statistics (ONS)
The population estimates used for the calculation of figures for this indicator are based on the 2001 Census. The population estimates used to calculate the LE figures were published on the following dates: • Mid-year population estimates for 2005 - 22 August 2007; • Mid-year population estimates for 2006 – 25 October 2007; • Mid-year population estimates for 2007 – 21 August 2008. Data quality: Accuracy and completeness
Mortality data quality and coverage is extremely high. The figures are three-year averages so as to provide large enough numbers to ensure that the presented figures are sufficiently robust. Two authorities, City of London and Isles of Scilly, are excluded from the results because of small numbers of deaths and populations in these areas.
Health Profiles 2009 The Indicator Guide
231 Section 6: life expectancy and cause of death
Table 3 – Indicator Technical Methods Numerator: extraction
Extraction of the completed indicator from ONS http://www.statistics.gov.uk/statbase/Product.asp?vlnk=8841
Numerator: aggregation / allocation
Deaths were assigned to local authority boundaries by ONS using the National Statistics Postcode Directory.
Numerator data caveats
Source material used to calculate LE: Area of residence is allocated by ONS using the postcode and the National Statistics Postcode Directory – records without a valid area code are excluded but the number of such records is negligible.
Denominator data caveats
Denominators are estimates of mid-year populations, based on results of the 2001 Census.
Methods used to calculate indicator value
The figures are rolling three-year averages produced by aggregating deaths and population estimates for 2005–2007. Abridged life tables were constructed using standard methods. Separate tables were constructed for males and females. The tables were created using annual mid-year population estimates and deaths registered in each year. All figures presented are for life expectancy at birth. A life table template which illustrates the method used to calculate the life expectancy results is available on the ONS website: http://www.statistics.gov.uk/statbase/ssdataset.asp?vlnk=6949 Data for some new Unitary Authorities, introduced in April 2009, did not correspond with existing geographies (Cheshire East, Cheshire West and Cheshire, Central Bedfordshire, and Cornwall). Results for these areas were calculated using the South East Public Health Observatory life expectancy calculator: http://www.sepho.org.uk/viewResource.aspx?id=8943 This was also used to calculate figures for the two Strategic Health Authorities in the South East: South Central and South East Coast. Results for the other new UAs were taken from published data for local authorities (Bedford) or counties (County Durham, Northumberland, Shropshire, and Wiltshire).
Small Populations: How Isles of Scilly and City of London populations have been dealt with
City of London and Isles of Scilly are excluded from the results because of small numbers of deaths and populations in these areas. Isles of Scilly has not been included in the new UA for Cornwall and City of London has not been included with Hackney. However, Isles of Scilly and City of London data are all included in the regional figures for the South West and London, and England figures.
Disclosure Control
Not applicable
Health Profiles 2009 The Indicator Guide
232 Section 6: life expectancy and cause of death
Confidence Intervals calculation method
The calculation of the confidence intervals was made using the method developed by Chiang. A report which details research undertaken by the Office for National Statistics on comparing methodologies to enable the calculation of confidence intervals for life expectancy at birth has now been published as No 33 in the National Statistics Methodological Series. This report, “Life expectancy at birth: methodological options for small populations”, also presents research carried out to establish if there is a minimum population size below which the calculation of life expectancy may not be considered feasible. It concludes with a summary of methodological conclusions and considers how these could be applied to the calculation of life expectancy at birth for wards in England and Wales. A copy of the report can be found on the NS website at: http://www.statistics.gov.uk/methods_quality/publications.asp The SEPHO calculator however also includes an adjustment to include a term for the variance associated with the final age interval as developed by Silcocks. Chiang CL. The Life Table and its Construction. In: Introduction to Stochastic Processes in Biostatistics. New York, John Wiley & Sons, 1968: 189-214. Silcocks PBS, Jenner DA, Reza R. Life expectancy as a summary of mortality in a population: statistical considerations and suitability for use by health authorities. J Epidemiol Community Health 2001; 55: 38-4
Health Profiles 2009 The Indicator Guide
233 Section 6: life expectancy and cause of death
28. INFANT DEATHS INDICATOR Basic Information 1. What is being measured?
Infant Deaths
2. Why is it being measured?
Infant mortality is an indicator of the general health of an entire population. It reflects the relationship between causes of infant mortality and upstream determinants of population health such as economic, social and environmental conditions. Deaths occurring during the first 28 days of life (the neonatal period) in particular, are considered to reflect the health and care of both mother and newborn.
3. How is this indicator actually defined?
Infant deaths, crude rate, persons, aged less than 1 year, 2005–07, per 1,000 live births.
4. Who does it measure?
Infants aged less than 1 year.
5. When does it measure it?
This indicator is updated annually.
6. Will it measure absolute numbers or proportions?
Proportions: numbers per 1,000 live births.
7. Where does the data actually come from?
Office for National Statistics (ONS)
8. How accurate and complete will the data be?
Data on births and deaths are considered to be complete and robust.
9. Are there any caveats/warnings/ problems?
Records without a valid area code are excluded but the number of such records is negligible.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
Page 4 Health summary – Indicator no. 28
Subject category/ domain(s)
Life expectancy and causes of death
Indicator name (*Indicator title in health profile)
Infant deaths
PHO with lead responsibility
LHO
Date of PHO dataset creation
February 2009
Health Profiles 2009 The Indicator Guide
234 Section 6: life expectancy and cause of death
Indicator definition
Infant deaths, crude rate, persons, aged less than 1 year, 2005–07, per 1,000 live births
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs, and Strategic Health Authorities for the South East.
Timeliness
The Compendium infant mortality indicator is updated annually, usually around September of the following year.
Rationale: What this indicator purports to measure
This indicator measures the level of infant deaths in the area.
Rationale: Public Health Importance
Infant mortality is an indicator of the general health of an entire population. It reflects the relationship between causes of infant mortality and upstream determinants of population health such as economic, social and environmental conditions. Deaths occurring during the first 28 days of life (the neonatal period) in particular, are considered to reflect the health and care of both mother and newborn.
Rationale: Purpose behind the inclusion of the indicator
There is a national health inequalities target for infant mortality which aims to reduce the gap between the infant mortality rate in the ‘routine and manual classes’ and the population as a whole. However, this target is difficult to monitor at local level as the number of infant deaths in any given local authority or primary care trust (PCT) among a particular social class group is very small and subject to random fluctuations from year to year. Therefore we have chosen to include overall infant mortality as an indicator. There are wide inequalities in infant mortality rates by local authority in England and monitoring these inequalities is essential to understanding trends in inequalities in infant mortality.
Rationale: Policy relevance
There is a national health inequalities target for infant mortality which aims to reduce the gap between the infant mortality rate in the ‘routine and manual classes’ and the population as a whole.
Interpretation: What a high/low level of indicator value means
A high indicator value (red circle in health summary chart) represents a statistically significant higher level of infant deaths for that local authority when compared to the national value. A low indicator value (green circle in health summary chart) represents a statistically significant lower level of infant deaths for that local authority when compared to the national value. A reduction in the infant death rate indicates a reduction in the number of deaths, relative to the number of live births. However, as the number of infant deaths in any given area is small, fluctuations from year to year are possible and may not indicate a statistically significant trend.
Interpretation: Potential for error due to type of measurement method
Coverage can be considered to be complete as the registration of deaths is a legal requirement. Data quality for the relevant fields (age and area of residence) is extremely high. The small number of infant deaths at a local authority level means that pooling of three years data is required. Even with pooled rates, however, numbers may still be small and large random fluctuations possible.
Health Profiles 2009 The Indicator Guide
235 Section 6: life expectancy and cause of death
Interpretation: Potential for error due to bias and confounding
The rates are not standardised or adjusted to take into account any potential confounding variables such as the age or ethnicity of the mother. Whether or not such variables need to be considered depends on the purpose to which the indicator is being put.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider the confidence interval, the greater the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with a white symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or amber symbol depending on whether it is worse or better than the national value respectively.
TABLE 2 – INDICATOR SPECIFICATION Indicator definition: Variable
Infant deaths
Indicator definition: Statistic
Crude rate
Indicator definition: Gender
Persons
Indicator definition: age group
Less than 1 year
Indicator definition: period
2005–2007
Indicator definition: scale
Per 1,000 live births
Health Profiles 2009 The Indicator Guide
236 Section 6: life expectancy and cause of death
Geography: geographies available for this indicator from other providers
England & Wales, ONS area classification, Strategic Health Authority, Primary Care Organisation
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Infant death rates by the National Statistics Socio-Economic Classification (based on father’s occupation at death registration) are available at a national level from the Office for National Statistics. As only a 10 per cent sample of live births are coded for father’s occupation, local infant death rates by NS-SEC are not available.
Data extraction: Source
Compendium of Clinical and Health Indicators/Clinical and Health Outcomes Knowledge Base (www.nchod.nhs.uk or nww.nchod.nhs.uk) Source of data: National Statistics
Data extraction: source URL
http://www.nchod.nhs.uk/
Data extraction: date
Data extracted from source as at Jan 2009
Numerator: definition
The number of infant deaths aged under 1 year registered in 2005–07
Numerator: source
Office for National Statistics (ONS).
Denominator: definition
The number of live births registered in 2005-07
Denominator: source
Office for National Statistics (ONS).
Data quality: Accuracy and completeness
Statistics on births and deaths are derived from the registration of births and deaths. The Office for National Statistics complete a variety of quality checks on the data before making them available for analysis. Data on births and deaths are considered to be largely complete.
Available from: www.nchod.nhs.uk
Infant death rates are available at a national level by mother’s country of birth from ONS.
Table 3 – Indicator Technical Methods Numerator: extraction
Infant mortality counts by area were extracted by ONS and supplied to NCHOD.
Numerator: aggregation / allocation
Deaths were assigned to geographical areas by ONS using the postcode of residence and the National Statistics Postcode Directory (NSPD).
Numerator data caveats
Records without a valid area code are excluded but the number of such records is negligible.
Denominator data caveats
Live births were assigned to geographical areas by ONS using the postcode of residence and the National Statistics Postcode Directory (NSPD). Records without a valid area code are excluded but the number of such records is negligible.
Health Profiles 2009 The Indicator Guide
237 Section 6: life expectancy and cause of death
Methods used to calculate indicator value
Crude rate per 1,000 live births: The number of infant deaths is divided by the number of live births in the same area and multiplied by 1,000. Data were sourced from NCHOD except for the new Unitary Authorities, introduced in April 2009, which did not correspond with existing geographies (Cheshire East, Cheshire West and Cheshire, Central Bedfordshire, and Cornwall). Results for these areas were calculated by the LHO. Results for the other new UAs were taken from published data for local authorities (Bedford) or counties (County Durham, Northumberland, Shropshire, and Wiltshire).
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Data for Isles of Scilly and City of London have not been included due to small populations; however, data for regions (South West and London) and for England do include these data. Data for the Isles of Scilly have not been included in the new Cornwall unitary authority.
Disclosure Control
None applied.
Confidence Intervals calculation method
Within the Compendium the 95% confidence intervals for crude rates and percentages are calculated using the likelihood-based method described by Aitken et al, which is a good approximation of the exact method. For example lower and upper limits for a rate expressed per 1,000 are given by:
rLL
rUL
r 1.96 exp ln − 1 − r nr (1 − r ) = × 1,000 r 1.96 1 + exp ln − 1 − r nr (1 − r ) r 1.96 exp ln + 1 − r nr (1 − r ) = × 1,000 1 + exp ln r + 1.96 1 − r nr (1 − r )
where: r is the crude rate; n is the population-years at risk. Aitken M et al. Statistical Modelling in GLIM. Oxford: Oxford University Press, 1990.
Health Profiles 2009 The Indicator Guide
238 Section 6: life expectancy and cause of death
29. DEATHS FROM SMOKING INDICATOR Basic Information 1. What is being measured?
Rates of deaths, attributable to smoking.
2. Why is it being measured?
Because smoking is the single biggest cause of death and there are big inequalities between and within communities in smoking and in death rates due to smoking.
3. How is this indicator actually defined?
Deaths attributable to smoking per 100,000 population.
4. Who does it measure?
Input data are annual death files provided by ONS, rates are calculated using (ex-) smoking rates and estimates for smoking attributable fractions of various causes of deaths.
5. When does it measure it?
1/1/2005–31/12/2007
6. Will It measure absolute numbers or proportions?
Proportions: number of cases per 100,000 population.
7. Where does the data actually come from?
Office of National Statistics.
8. How accurate and complete will the data be?
Although the input data is very reliable, the results also depend on estimated (ex-) smoking rates and the estimated proportion of causes of death which can be attributed to smoking.
9. Are there any caveats/warnings/ problems?
No
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
Page 4 Spine Chart – Indicator 29
Subject category/ domain(s)
Life expectancy and causes of death
Indicator name (* Indicator title in health profile)
Smoking attributable mortality (“Deaths from smoking”)
PHO with lead responsibility
ERPHO
Date of PHO dataset creation
5 Feb 2009
Indicator definition
Deaths attributable to smoking, directly age standardised rate, 35 years +, 2005–07, persons.
Health Profiles 2009 The Indicator Guide
239 Section 6: life expectancy and cause of death
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs
Timeliness
Annual updates available from erpho
Rationale: What this indicator purports to measure
Rate of deaths that can be attributed to smoking in persons 35 and over
Rationale: Public Health Importance
Smoking still accounts for between 1 in 6 and 1 in 10 of all deaths in England, and accounts for about half of the inequality in death rates between spearhead and non-spearhead areas. It remains the biggest single cause of preventable mortality and morbidity in the world.
Rationale: Purpose behind the inclusion of the indicator
To encourage smoking prevention. To focus action on tackling smoking related disease and to help prioritise actions to tackle health inequalities. To promote better measurement of smoking prevalence. Smoking related deaths is a powerful proxy measure of overall health and predictor of health care demand.
Rationale: Policy relevance
Smoking Kills Choosing Health Smokefree England
Interpretation: What a high/low level of indicator value means
An indicator value better than average (green circle in health summary chart) represents a statistically significant better level of smoking attributable mortality for that local authority when compared to the national value. An indicator value worse than average (red circle in health summary chart) represents a statistically significant worse level of smoking attributable mortality for that local authority when compared to the national value. High smoking attributable death rates are indicative of poor population health and high smoking rates. A zero rate is unachievable but lower rates than the best in England are seen in California and Scandinavian countries.
Interpretation: Potential for error due to type of measurement method
The method relies on the use of estimates of the contribution of smoking to a range of causes of death derived from the American Cancer Prevention Society II study. It presumes a degree of generalisability of these estimates. Assuming that the contribution of smoking to deaths in Americans in the 1980’s is the same as its contribution to an English population in 2005-07. These estimates do not take into account any degree of uncertainty meaning the estimates will be over precise.
Interpretation: Potential for error due to bias and confounding
Smoking prevalence rates for England were used because the model required estimates of ex-smoking and non-smoking as well as current smoking rates. These figures are not available at local authority level. The method will tend to overestimate smoking related deaths in low prevalence areas and underestimate in high prevalence areas.
Health Profiles 2009 The Indicator Guide
240 Section 6: life expectancy and cause of death
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider the confidence interval is the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so it is said that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Table 2 – Indicator Specification Indicator definition: Variable
Deaths attributable to smoking
Indicator definition: Statistic
Directly age-standardised rate
Indicator definition: Gender
Persons
Indicator definition: age group
35 and over
Indicator definition: period
2005–07
Health Profiles 2009 The Indicator Guide
241 Section 6: life expectancy and cause of death
Indicator definition: scale
Per 100,000 European Standard population
Geography: geographies available for this indicator from other providers
SHAs Available from erpho on request
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Data can be calculated at MSOA level, and aggregated into deprivation quintiles to calculate within area inequalities. See for example http://www.erpho.org.uk/ viewResource.aspx?id=18243
Data extraction: Source
ONS for mortality data and population estimates. Prevalence of smoking, ex-smoking and non-smoking derived from Health Survey for England 2003–05 SAMMEC website (for relative risks from the American Cancer Prevention Society II Study 1982-1988)
Data extraction: source URL
http://apps.nccd.cdc.gov/sammec/ ONS mortality extracts are supplied to PHOs under a non-disclosure agreement and the provisions of the Health Act 2000. Health Survey for England http://www.ic.nhs.uk/statistics-and-data-collections/healthand-lifestyles-related-surveys/health-survey-for-england/health-survey-for-england2005-latest-trends
Data extraction: date
December 2008 (HSE, Sammec) December 2008 (mortality data)
Health Profiles 2009 The Indicator Guide
242 Section 6: life expectancy and cause of death
Numerator: definition
Disease Category
ICD10
Malignant Neoplasms Lip, Oral Cavity, Pharynx
C00–C14
Oesophagus
C15
Stomach
C16
Pancreas
C25
Larynx
C32
Trachea, Lung, Bronchus
C33–C34
Cervix Uteri
C53
Kidney and Renal Pelvis
C64–C65
Urinary Bladder
C67
Acute Myeloid Leukemia
C92.0
Cardiovascular Diseases Ischemic Heart Disease
I20–I25
Other Heart Disease
I00–I09, I26–I51
Cerebrovascular Disease
I60–I69
Atherosclerosis
I70
Aortic Aneurysm
I71
Other Arterial Disease
I72–I78
Respiratory Diseases Pneumonia, Influenza
J10–J18
Bronchitis, Emphysema
J40–J42, J43
Chronic Airway Obstruction
J44
Numerator: source
Death extracts from ONS: ONS PHO death extract 2005 ONS PHO death extract 2006 ONS PHO death extract 2007
Denominator: definition
Mid-year local authority population estimates
Denominator: source
ONS Mid-year population estimates 2005, ONS Mid year population estimates 2006, ONS Mid-year population estimates 2007, Estimates are as published for Persons rounded. The 3 years data is pooled.
Data quality: Accuracy and completeness
The accuracy of these estimates is contingent on the underlying accuracy of the three components: 1. Mortality data – because this is cause specific there may be coding variation between places although for the key contributors e.g. cancer deaths, this is less likely. 2. The relative risks are unpublished and only made available through the SAMMEC website 3. Smoking prevalence is estimated using national survey data. Future updates will prove how accurate these assumptions are.
Health Profiles 2009 The Indicator Guide
243 Section 6: life expectancy and cause of death
Table 3 – Indicator Technical Methods Numerator: extraction
CREATE TABLE P2564_HPSAMMEC_LA_ICD10 (RegPeriod varchar(8), LACode varchar(4), LAName varchar(30), Sex varchar(1), AgeGroup varchar(5), SAFAgeGrp varchar(5), ICDCode varchar(4), Deaths decimal(8,3)) INSERT INTO P2564_HPSAMMEC_LA_ICD10 (RegPeriod, LACode, LAName, Sex, AgeGroup, ICDCode, Deaths) SELECT ‘2005-07’, [LA Code now], [LA name], Sex, [FiveYearTo85], SUBSTRING([Und COD (non-neo)],1,3), COUNT(*) FROM M_Deaths2001to2007 LEFT JOIN A_DeathAgeBands ON M_Deaths2001to2007.Age = A_ DeathAgeBands.AgeInUnits AND M_Deaths2001to2007.[Age Unit] = A_ DeathAgeBands.Units LEFT JOIN HES.dbo.A_LACounty ON M_Deaths2001to2007.ResCty + M_ Deaths2001to2007.ResLAUA = HES.dbo.A_LACounty.[LA Code] LEFT JOIN HES.dbo.A_NationalPostcodes_Feb08 ON M_Deaths2001to2007. Pcode_8dig = HES.dbo.A_NationalPostcodes_Feb08.[pcd2] WHERE YEAR([Reg Date]) > 2004 AND YEAR([Reg Date]) < 2008 AND [LA Code now] <> ‘NULL’ AND [LA Code now] <> ‘15UH’ AND ([FiveYearTo85] = ‘35-39’ OR [FiveYearTo85] = ‘40-44’ OR [FiveYearTo85] = ‘45-49’ OR [FiveYearTo85] = ‘50-54’ OR [FiveYearTo85] = ‘55-59’ OR [FiveYearTo85] = ‘60-64’ OR [FiveYearTo85] = ‘65-69’ OR [FiveYearTo85] = ‘70-74’ OR [FiveYearTo85] = ‘75-79’ OR [FiveYearTo85] = ‘80-84’ OR [FiveYearTo85] = ‘85+’) AND (SUBSTRING([Und COD (non-neo)],1,3) = ‘C00’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C01’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C02’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C03’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C04’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C05’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C06’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C07’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C08’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C09’
Health Profiles 2009 The Indicator Guide
244 Section 6: life expectancy and cause of death
OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C10’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C11’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C12’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C13’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C14’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C15’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C16’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C25’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C32’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C33’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C34’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C53’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C64’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C65’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C67’ OR [Und COD (non-neo)] = ‘C920’ OR SUBSTRING([Und COD (nonneo)],1,2) = ‘I0’ OR SUBSTRING([Und COD (non-neo)],1,2) = ‘I2’ OR SUBSTRING([Und COD (non-neo)],1,2) = ‘I3’ OR SUBSTRING([Und COD (non-neo)],1,2) = ‘I4’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘I50’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘I51’ OR SUBSTRING([Und COD (non-neo)],1,2) = ‘I6’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘I70’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘I71’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘I72’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘I73’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘I74’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘I75’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘I76’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘I77’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘I78’ OR SUBSTRING([Und COD (non-neo)],1,2) = ‘J1’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘J40’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘J41’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘J42’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘J43’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘J44’) GROUP BY [LA Code now], [LA Name], Sex, [FiveYearTo85] , SUBSTRING([Und COD (non-neo)],1,3) ORDER BY [LA Code now], Sex, [FiveYearTo85], SUBSTRING([Und COD (non-neo)],1,3)
Health Profiles 2009 The Indicator Guide
245 Section 6: life expectancy and cause of death
Numerator: aggregation/ allocation
Deaths in the smoking related ICD10 categories were multiplied by the Smoking Attributable Fraction (SAF), according to the age of the deceased (see below). Since the 2001 census new Unitary Authorities have been created. The new Unitary Authorities of Northumberland, Shropshire, Wiltshire, and Durham have the same boundaries as the counties of the same name so have been allocated these data. Data for five other new unitary Authorities have been calculated from the aggregation of previous Local Authorities. New UA name Bedford Central Bedfordshire Cheshire East
Cheshire West and Chester
Cornwall
Old LA name Bedford Mid Bedfordshire South Bedfordshire Congleton Crewe and Nantwich Macclesfield Chester Ellesmere Port & Neston Vale Royal Caradon Carrick Kerrier North Cornwall Penwith Restormel
Numerator data caveats Denominator data caveats
Population estimates used are Persons rounded and pooled across the 3 years. This may produce slight differences in rate calculations from those done locally. Population estimates of higher geographies are aggregated from LA populations.
Health Profiles 2009 The Indicator Guide
246 Section 6: life expectancy and cause of death
Methods used to calculate indicator value
Age-specific attributable deaths were calculated as follows: The SAF multiplier applied to deaths in each smoking related ICD10 category were determined using the formula: SAF = [p1(RR1 - 1) + p2(RR2 - 1)]/[1 + p1(RR1 - 1) + p2(RR2 - 1)] Where p1 = prevalence of current smokers; p2= ex-smoker prevalence; RR1 = relative risk in current smokers and RR2 = relative risk in former smokers. The following relative risks for each disease-age-sex group were used. RR Male Disease Category
Current Smoker
RR Female
Former Smoker
Current Smoker
Former Smoker
Malignant Neoplasms Lip, Oral Cavity, Pharynx
10.89
3.40
5.08
2.29
Esophagus
6.76
4.46
7.75
2.79
Stomach
1.96
1.47
1.36
1.32
Pancreas
2.31
1.15
2.25
1.55
Larynx
14.6
6.34
13.02
5.16
23.26
8.70
12.69
4.53
Cervix Uteri
1.00
1.00
1.59
1.14
Kidney and Renal Pelvis
2.72
1.73
1.29
1.05
Urinary Bladder
3.27
2.09
2.22
1.89
Acute Myeloid Leukemia
1.86
1.33
1.13
1.38
Persons Aged 35–64
2.80
1.64
3.08
1.32
Persons Aged 65+
1.51
1.21
1.60
1.20
Other Heart Disease
1.78
1.22
1.49
1.14
Persons Aged 35–64
3.27
1.04
4.00
1.30
Persons Aged 65+
1.63
1.04
1.49
1.03
Atherosclerosis
2.44
1.33
1.83
1.00
Aortic Aneurysm
6.21
3.07
7.07
2.07
Other Arterial Disease
2.07
1.01
2.17
1.12
1.75
1.36
2.17
1.10
Bronchitis, Emphysema
17.10
15.64
12.04
11.77
Chronic Airway Obstruction
10.58
6.80
13.08
6.78
Trachea, Lung, Bronchus
Cardiovascular Diseases Ischemic Heart Disease
Cerebrovascular Disease
Respiratory Diseases Pneumonia, Influenza
Health Profiles 2009 The Indicator Guide
247 Section 6: life expectancy and cause of death
Together with the following England smoking prevalence estimates Smoking Prevalence Sex
Age
Smoker
Former smoker
M
35-44
29.7%
19.6%
M
45-54
25.2%
31.4%
M
55-64
20.8%
44.5%
M
65-74
12.7%
54.7%
M
75+
8.2%
60.4%
F
35-44
28.9%
16.2%
F
45-54
26.9%
21.8%
F
55-64
22.0%
28.8%
F
65-74
14.8%
30.1%
F
75+
9.4%
36.0%
The counts of deaths were aggregated by relevant ICD codes, age and sex groups Relative risks were assumed to be constant within broader age bands Death counts in 5-year age bands were multiplied by the appropriate SAF to give smoking attributable deaths in each age-sex band. Directly age standardised rates were calculated using the batch calculator available from the erpho website http://www.erpho.org.uk/topics/tools/rates. aspx#12474 Small Populations: How Isles of Scilly and City of London populations have been dealt with
Isles of Scilly and the City of London were excluded from local authority calculations. City of London was included in the higher geographies (the Isles of Scilly were not included in higher geographies).
Disclosure Control
Not relevant
Confidence Intervals calculation method
Confidence intervals have been calculated using Byars method. See http://www.yhpho.org.uk/standardisedratecalculater.aspx
Health Profiles 2009 The Indicator Guide
248 Section 6: life expectancy and cause of death
30. EARLY DEATHS: HEART DISEASE AND STROKE INDICATOR Basic Information 1. What is being measured?
Mortality rate from all circulatory diseases (including heart disease and stroke).
2. Why is it being measured?
Circulatory disease accounts for 40% of all deaths (30% under 75). Mortality is a direct measure of health care need and indicates the overall circulatory disease burden on the population, reflecting both the incidence of disease and the ability to treat it.
3. How is this indicator actually defined?
Mortality from all circulatory diseases, directly age-standardised rate, persons, under 75, 200507 (pooled), per 100,000 European Standard population.
4. Who does it measure?
People aged under 75.
5. When does it measure it?
Updated annually.
6. Will it measure absolute numbers or proportions?
Directly age-standardised rate.
7. Where does the data actually come from?
Office for National Statistics (ONS).
8. How accurate and complete will the data be?
Mortality counts are derived from an annual mortality extract supplied by ONS and are based on the original underlying cause of death for which there is nearly 100% coverage on the mortality register.
9. Are there any caveats/warnings/ problems?
Area of residence is allocated by ONS using the postcode and the National Statistics Postcode Directory – records without a valid area code are excluded but the number of such records is negligible. There is the potential for the underlying cause of death to be incorrectly attributed on the death certificate and, therefore, the cause of death misclassified.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
Pg 4 Health Summary – Indicator No 30
Subject category/ domain(s)
Life expectancy and causes of death
Health Profiles 2009 The Indicator Guide
249 Section 6: life expectancy and cause of death
Indicator name (*Indicator title in health profile)
Mortality rate from all circulatory diseases (*Early deaths: heart disease & stroke)
PHO with lead responsibility
SEPHO
Date of PHO dataset creation
05 February 2009
Indicator definition
Mortality from all circulatory diseases, directly age-standardised rate, persons, under 75, 2005–07 (pooled), per 100,000 European Standard population
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs (boundaries as at April 2008, new 2009 Unitary Authorities as at April 2009).
Timeliness
The Compendium mortality from all circulatory diseases indicator is updated annually, usually around November/December following the publication by ONS of the new year’s mortality extract (usually in May) and mid-year population estimates (usually August-September).
Rationale: What this indicator purports to measure
Early mortality from all circulatory diseases.
Rationale: Public Health Importance
Circulatory disease accounts for 40% of all deaths (30% under 75). Mortality is a direct measure of health care need reflecting the overall circulatory disease burden on the population, both the incidence of disease and the ability to treat it. The mortality rate may be improved by reducing the population’s risk (e.g. encouraging healthier lifestyles and reducing exposure to smoking), by earlier detection of disease and by more effective treatment.
Rationale: Purpose behind the inclusion of the indicator
To estimate premature mortality due to circulatory diseases.
Rationale: Policy relevance
The under 75 circulatory disease mortality rate is a key target indicator in the 1999 Public Health White Paper ‘Saving Lives: Our Healthier Nation’. The target is to reduce the number of deaths from circulatory disease in people aged under 75 years by at least two-fifths by 2010. The baseline for monitoring this target was the three year period 1995–97.
To reduce premature deaths from circulatory diseases.
This measure supports delivery of the Department of Health PSA targets and LDP and is relevant to Choosing Health, Coronary Heart Disease NSF and Programme for Action. Interpretation: What a high/low level of indicator value means
A high indicator value (red circle in health summary chart) represents a statistically significant higher (worse) rate of early deaths from circulatory disease when compared to the national value. A low indicator value (green circle in health summary chart) represents a statistically significant lower (better) rate of early deaths from circulatory disease when compared to the national value.
Health Profiles 2009 The Indicator Guide
250 Section 6: life expectancy and cause of death
Interpretation: Potential for error due to type of measurement method
Coverage can be considered to be complete as the registration of deaths is a legal requirement. Data quality for the relevant fields (age, sex, underlying cause of death, area of residence) is extremely high.
Interpretation: Potential for error due to bias and confounding
The rates are age-standardised. This improves the comparability of rates for different areas, or between different time periods, by taking into account differences in the age structures of the populations being compared.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate.
There is the potential for the underlying cause of death to be incorrectly attributed on the death certificate and, therefore, the cause of death misclassified.
This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
TABLE 2 – INDICATOR SPECIFICATION Indicator definition: Variable
Mortality from all circulatory diseases (ICD10 I00 –I99)
Indicator definition: Statistic
Directly age-standardised rate
Indicator definition: Gender
Persons
Health Profiles 2009 The Indicator Guide
251 Section 6: life expectancy and cause of death
Indicator definition: age group
Under 75
Indicator definition: period
2005–07 (pooled**) ** average of annual rates
Indicator definition: scale
Per 100,000 European Standard population
Geography: geographies available for this indicator from other providers
England & Wales, ONS area, Primary Care Organisation, Strategic Health Authority.
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Age, gender available from NCHOD.
Data extraction: Source
NCHOD.
Data extraction: source URL
Data received directly from NCHOD.
Data extraction: date
14 January 2009
Numerator: definition
Deaths from all circulatory disease, classified by underlying cause of death (ICD10 I00 – I99), registered in the respective calendar years 2005-07, in people aged under 75.
Numerator: source
Office for National Statistics (ONS)
Denominator: definition
2001 census based mid-year population estimates for respective calendar years 2005 to 2007, people aged under 75 (current as at 29 September, 2008).
Denominator: source
ONS
Data quality: Accuracy and completeness
Coverage can be considered to be complete as the registration of deaths is a legal requirement. Data quality for the relevant fields (age, sex, underlying cause of death, area of residence) is extremely high. Area of residence is allocated by ONS using the postcode and the National Statistics Postcode Directory – records without a valid area code are excluded but the number of such records is negligible.
Available from National Centre for Health Outcomes Development (NCHOD) website www.nchod.nhs.uk Data can also be found at Neighbourhood Renewal Unit Public Service Agreement Floor Targets (http://www.fti.communities.gov.uk/fti/).
Table 3 – Indicator Technical Methods Numerator: extraction
Extraction by NCHOD.
Numerator: aggregation / allocation
Deaths were assigned to geographical areas using the area code supplied in the mortality extract. This is derived from postcode of residence by the ONS using the National Statistics Postcode Directory (NSPD).
Health Profiles 2009 The Indicator Guide
252 Section 6: life expectancy and cause of death
Numerator data caveats
Area of residence is allocated by ONS using the postcode and the National Statistics Postcode Directory – records without a valid area code are excluded but the number of such records is negligible. Mortality counts are derived from the annual DH mortality extract supplied by ONS and are based on the original underlying cause of death for which there is nearly 100% coverage on the mortality register.
Denominator data caveats
Data are based on the latest revisions of ONS mid-year population estimates for the respective years, current as at 29 September 2008.
Methods used to calculate indicator value
The directly age-standardised rate (DSR) is the rate of events that would occur in a population with a standard age structure if that population were to experience the age-specific rates of the subject population. The standard population used is the European Standard Population. The age groups used are: Under 1, 1–4, 5-9,…, 80–84, 85+. The rate for 2005–07 has been calculated as the simple average of the individual annual rates. The rate is expressed per 100,000 population.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Isles of Scilly and City of London are excluded from the lower tier datasets but included in England and Regional figures.
Disclosure Control
None applied.
Confidence Intervals calculation method
Confidence intervals for the age-standardised rates were calculated using a normal approximation, i.e. +/- 1.96 standard errors. The standard errors are obtained using the method described by Breslow and Day, but modified to use the binomial variance for a proportion to estimate the variances of the crude age-specific rates. This method is likely to be unreliable when there are fewer than 50 cases in an area, hence confidence intervals for rates based on less than 50 cases should be viewed with caution. The lower and upper limits for the rates are denoted by DSRLL and DSRUL respectively. The lower and upper limits for the rates are denoted by DSRLL and DSRUL respectively.
DSR LL / UL = DSR ± 1.96 × 100,000 ×
1 ∑ wi ij
2
×∑ ij
wi2 ⋅ rij (1 − rij ) nij
(expressed per 100,000 population) where: wi is the number, or proportion, of individuals in the standard population in age group i. rij is the crude age-specific rate in the subject population in age group i, in year j. nij is the number of individuals in the subject population in age group i, in year j.
Health Profiles 2009 The Indicator Guide
253 Section 6: life expectancy and cause of death
31. EARLY DEATHS: CANCER INDICATOR Basic Information 1. What is being measured?
Mortality rate from all cancers.
2. Why is it being measured?
Early mortality from cancer is a direct measure of health care need as public health interventions for prevention, early diagnosis, and effective treatment can all reduce the burden of cancer morbidity and mortality.
3. How is this indicator actually defined?
Mortality from all cancers, directly age-standardised rate, persons, under 75, 2005–07 (pooled), per 100,000 European Standard population.
4. Who does it measure?
People aged under 75.
5. When does it measure it?
Updated annually.
6. Will It measure absolute numbers or proportions?
Directly age-standardised rate.
7. Where does the data actually come from?
Office for National Statistics (ONS).
8. How accurate and complete will the data be?
Mortality counts are derived from an annual mortality extract supplied by ONS and are based on the original underlying cause of death for which there is nearly 100% coverage on the mortality register.
9. Are there any caveats/warnings/ problems?
Area of residence is allocated by ONS using the postcode and the National Statistics Postcode Directory – records without a valid area code are excluded but the number of such records is negligible. There is the potential for the underlying cause of death to be incorrectly attributed on the death certificate and, therefore, the cause of death misclassified.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
TABLE 1 – INDICATOR DESCRIPTION Information component
Pg 4 Health Summary – Indicator No 31
Subject category/ domain(s)
Life expectancy and causes of death
Health Profiles 2009 The Indicator Guide
254 Section 6: life expectancy and cause of death
Indicator name (*Indicator title in health profile)
Mortality rate from all cancers (*Early deaths: cancer)
PHO with lead responsibility
SEPHO
Date of PHO dataset creation
05 February 2009
Indicator definition
Mortality from all cancers, directly age-standardised rate, persons, under 75, 2005–07 (pooled), per 100,000 European Standard population
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs (boundaries as at April 2008, new 2009 Unitary Authorities as at April 2009).
Timeliness
The Compendium mortality from all cancers indicator is updated annually, usually around November/December following the publication by ONS of the new year’s mortality extract (usually in May) and mid-year population estimates (usually August–September).
Rationale: What this indicator purports to measure
Early mortality from all cancers.
Rationale: Public Health Importance
Cancer is amongst the three leading causes of death at all ages except for pre-school age children in the UK. It accounts for 26% all deaths. If current incidence rates remain the same, by 2025 there will be an additional 100,000 cases of cancer diagnosed each year as a result of the ageing population. Inequalities exist in cancer rates between the most deprived areas and the most affluent. Early mortality from cancer is a direct measure of health care need as public health interventions for prevention, early diagnosis, effective treatment can all reduce the burden of cancer morbidity and mortality.
Rationale: Purpose behind the inclusion of the indicator
To estimate premature mortality due to cancer.
Rationale: Policy relevance
The directly age-standardised mortality rate from all cancers for persons aged under 75 is a target indicator in the Saving Lives: Our Healthier Nation strategy. The target is a 20% reduction by the year 2010 from the baseline rate in 1995–97.
To reduce premature deaths from cancer.
This measure supports delivery of the Department of Health PSA targets and LDP and is relevant to Choosing Health, Cancer NSF and Programme for Action. Interpretation: What a high/low level of indicator value means
A high indicator value (red circle in health summary chart) represents a statistically significant higher (worse) rate of early deaths from cancer for that local authority when compared to the national value. A low indicator value (green circle in health summary chart) represents a statistically significant lower (better) rate of early deaths from cancer when compared to the national value.
Health Profiles 2009 The Indicator Guide
255 Section 6: life expectancy and cause of death
Interpretation: Potential for error due to type of measurement method
Coverage can be considered to be complete as the registration of deaths is a legal requirement. Data quality for the relevant fields (age, sex, underlying cause of death, area of residence) is extremely high.
Interpretation: Potential for error due to bias and confounding
The rates are age-standardised. This improves the comparability of rates for different areas, or between different time periods, by taking into account differences in the age structures of the populations being compared.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate.
There is the potential for the underlying cause of death to be incorrectly attributed on the death certificate and, therefore, the cause of death misclassified.
This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Table 2 – Indicator Specification Indicator definition: Variable
Mortality from all cancers (ICD10 C00-C97)
Indicator definition: Statistic
Directly age-standardised rate
Indicator definition: Gender
Persons
Health Profiles 2009 The Indicator Guide
256 Section 6: life expectancy and cause of death
Indicator definition: age group
Under 75
Indicator definition: period
2005–07 (pooled**) ** average of annual rates
Indicator definition: scale
Per 100,000 European Standard population
Geography: geographies available for this indicator from other providers
ONS area, Primary Care Organisation, Strategic Health Authority.
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Age, gender available from NCHOD.
Data extraction: Source
NCHOD.
Data extraction: source URL
Data received directly from NCHOD.
Data extraction: date
14 January 2009
Numerator: definition
Deaths from all malignant neoplasms, classified by underlying cause of death (ICD10 C00-C97), registered in the respective calendar years 2005–07, in people aged under 75.
Numerator: source
Office for National Statistics (ONS)
Denominator: definition
2001 census based mid-year population estimates for respective calendar years 2005 to 2007, people aged under 75 (current as at 29 September, 2008).
Denominator: source
ONS
Data quality: Accuracy and completeness
Coverage can be considered to be complete as the registration of deaths is a legal requirement. Data quality for the relevant fields (age, sex, underlying cause of death, area of residence) is extremely high. Area of residence is allocated by ONS using the postcode and the National Statistics Postcode Directory – records without a valid area code are excluded but the number of such records is negligible.
Available from National Centre for Health Outcomes Development (NCHOD) website www.nchod.nhs.uk Data can also be found at Neighbourhood Renewal Unit Public Service Agreement Floor Targets (http://www.fti.communities.gov.uk/fti/).
Table 3 – Indicator Technical Methods Numerator: extraction
Extraction by NCHOD.
Numerator: aggregation/ allocation
Deaths were assigned to geographical areas using the area code supplied in the mortality extract. This is derived from postcode of residence by the ONS using the National Statistics Postcode Directory (NSPD).
Health Profiles 2009 The Indicator Guide
257 Section 6: life expectancy and cause of death
Numerator data caveats
Area of residence is allocated by ONS using the postcode and the National Statistics Postcode Directory - records without a valid area code are excluded but the number of such records is negligible. Mortality counts are derived from the annual DH mortality extract supplied by ONS and are based on the original underlying cause of death for which there is nearly 100% coverage on the mortality register.
Denominator data caveats
Data are based on the latest revisions of ONS mid-year population estimates for the respective years, current as at 29 September, 2008.
Methods used to calculate indicator value
The directly age-standardised rate (DSR) is the rate of events that would occur in a population with a standard age structure if that population were to experience the age-specific rates of the subject population. The standard population used is the European Standard Population. The age groups used are: Under 1, 1–4, 5-9,…, 80–84, 85+. The rate for 2005–07 has been calculated as the simple average of the individual annual rates. The rate is expressed per 100,000 population.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Isles of Scilly and City of London are excluded from the lower tier datasets but included in England and Regional figures.
Disclosure Control
None applied.
Confidence Intervals calculation method
Confidence intervals for the age-standardised rates were calculated using a normal approximation, i.e. +/- 1.96 standard errors. The standard errors are obtained using the method described by Breslow and Day, but modified to use the binomial variance for a proportion to estimate the variances of the crude age-specific rates. This method is likely to be unreliable when there are fewer than 50 cases in an area, hence confidence intervals for rates based on less than 50 cases should be viewed with caution. The lower and upper limits for the rates are denoted by DSRLL and DSRUL respectively.
DSR LL / UL = DSR ± 1.96 × 100,000 ×
1 ∑ wi ij
2
×∑ ij
wi2 ⋅ rij (1 − rij ) nij
(expressed per 100,000 population) where: wi is the number, or proportion, of individuals in the standard population in age group i. rij is the crude age-specific rate in the subject population in age group i, in year j. nij is the number of individuals in the subject population in age group i, in year j.
Health Profiles 2009 The Indicator Guide
258 Section 6: life expectancy and cause of death
32. ROAD INJURIES AND DEATHS INDICATOR Basic Information 1. What is being measured?
People killed or seriously injured on the road.
2. Why is it being measured?
To help reduce road traffic collisions which are a major cause of preventable death and serious injury.
3. How is this indicator actually defined?
People killed or seriously injured on the roads of the area, crude rate per 100,000 resident population, all ages, 2005–2007.
4. Who does it measure?
All persons, all ages
5. When does it measure it?
Continually reported and data is published annually
6. Will It measure absolute numbers or proportions?
Proportion: Crude rate of people killed or seriously injured per 100,000 resident population
7. Where does the data actually come from?
The data is collected by the Police and published by Department for Transport
8. How accurate and complete will the data be?
Coverage is complete.
9. Are there any caveats/warnings/ problems?
Not all road casualties are reported to police.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Data quality varies as there are differences between police forces in procedures for recording, collecting and collating.
The figures are for road casualties that occurred within a local authority area irrespective of the home residence of the person involved in the accident. Areas with high inflows of people or traffic may have artificially high rates because the at-risk resident population is not an accurate measure of exposure to transport. This is likely to affect the results for employment centres and sparsely populated rural areas which have high numbers of visitors or through traffic.
Table 1 – Indicator Description Information component
Pg 4 Health Summary – Indicator No 32
Subject category/ domain(s)
Life expectancy and causes of death
Indicator name (*Indicator title in health profile)
Road injuries and deaths
Health Profiles 2009 The Indicator Guide
259 Section 6: life expectancy and cause of death
PHO with lead responsibility
South West Public Health Observatory
Date of PHO dataset creation
January 2009 (revised February 2009 for new April 2009 area boundaries)
Indicator definition
People killed or seriously injured on the roads of the area, crude rate per 100,000 resident population, all ages, 2005–2007.
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs.
Timeliness
The indicator presented in Health Profiles is not routinely updated but may become available in future Health Profiles. The rates and numerators published by DfT are available annually (usually October-December).
Rationale: What this indicator purports to measure
Unintentional deaths and serious injuries on public roads caused by road traffic collisions.
Rationale: Public Health Importance
Motor vehicle traffic accidents are a major cause of preventable deaths and morbidity, particularly in younger age groups. For children and for men aged 20–64 years, mortality rates for motor vehicle traffic accidents are higher in lower socioeconomic groups. For instance, there would be 600 fewer deaths in men aged 20–64 years from motor vehicle traffic accidents each year if all men had the same death rates as those in social classes I and II combined (Acheson, D. Report of the Independent Inquiry into Inequalities in Health. London: TSO, 1998. Department of Health. The NHS Plan. London:TSO). The vast majority of road traffic collisions are preventable and can be avoided through improved education, awareness, road infrastructure and vehicle safety (Peden M, Scurfield R, Sleet D, Mohan D, Hyder AA, Jarawan E et al. 2004, World report on road traffic injury prevention. Geneva: World Health Organisation).
Rationale: Purpose behind the inclusion of the indicator
To help reduce road injury and death.
Rationale: Policy relevance
One of the Department of Transport’s PSA targets is to reduce the number of people killed or seriously injured by 40%, and the number of children killed or seriously injured by 50% by 2010 compared with the baseline of 1994–8. In Saving Lives: Our Healthier Nation, ‘accidents’, including road traffic collisions, were identified as one of the four main priority areas. Tackling Health Inequalities: A Programme For Action included a headline target to reduce road casualty numbers in disadvantaged areas.
Interpretation: What a high/low level of indicator value means
A high indicator value (red circle in health summary chart) represents a statistically significant higher rate of road injury and death when compared to the national value. A low indicator value (amber circle in health summary chart) represents a statistically significant rate of road injury and death when compared to the national value. However, as the vast majority of road traffic collisions are preventable, any number of deaths and injuries greater than zero is undesirable and therefore a low indicator value should not mean that PH action is not needed. Factors that should be considered when interpreting this indicator include the nature of the road network, volumes of traffic and number of pedestrians.
Health Profiles 2009 The Indicator Guide
260 Section 6: life expectancy and cause of death
Interpretation: Potential for error due to type of measurement method
Research has shown that not all road casualties are reported to police. Therefore, this indicator may be an under-estimate of the true level of serious injury.
Interpretation: Potential for error due to bias and confounding
The completeness of reporting road injury to the police has been shown to vary by factors such as age, injury severity and type of road user. For example, research suggests that cyclists are less likely to report injury to the police. There are also differences by age group, with higher reporting rates amongst younger and older people. Age also impacts on injury severity, with older people more likely to be seriously injured or killed when involved in a road collision. Therefore, population structure and mix of traffic in areas may contribute to variation in Local Authority values.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate.
Districts with low resident populations but which have high inflows of people or traffic may have artefactually high rates because the at-risk resident population is not an accurate measure of exposure to transport. This is likely to affect employment centres and sparsely populated rural areas which have high numbers of visitors or through traffic.
This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Health Profiles 2009 The Indicator Guide
261 Section 6: life expectancy and cause of death
TABLE 2 – INDICATOR SPECIFICATION Indicator definition: Variable
People killed or seriously injured on the roads of the area.
Indicator definition: Statistic
Crude rate.
Indicator definition: Gender
Persons.
Indicator definition: age group
All ages.
Indicator definition: period
Numerator 2005–2007, denominator mid-2006.
Indicator definition: scale
Per 100,000 resident population.
Geography: geographies available for this indicator from other providers
England, GOR, Local Authority: Counties, Unitary Authorities, London Boroughs.
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Numerator data is available from www.dft.gov.uk for Counties, Unitary Authorities, London Boroughs, broken down by age group and road user type.
Data extraction: Source
Department for Transport.
Data extraction: source URL
http://www.dft.gov.uk/pgr/statistics/datatablespublications/accidents/ casualtieslatables/
Data extraction: date
Data extracted from source as at: 29/12/2008
Annual rates and numerators are available for these geographies from www.dft.gov. uk.
Health Profiles 2009 The Indicator Guide
262 Section 6: life expectancy and cause of death
Numerator: definition
3-year average of the number of people (all ages) killed or seriously injured on the roads of a LA in the period 2005–2007. The indicator is based on casualties who incur injury on the public highway (including footways) in which at least one road vehicle or a vehicle in collision with a pedestrian is involved and which becomes known to the police within 30 days of its occurrence. The vehicle need not be moving and accidents involving stationary vehicles and pedestrians or other road users are included. One accident may give rise to more than one casualty. This indicator includes only on casualties who are fatally or seriously injured and these categories are defined as follows. Fatal casualties are those who sustained injuries which caused death less than 30 days after the accident; confirmed suicides are excluded. Seriously injured casualties are those who sustained an injury for which they are detained in hospital as an in-patient, or any of the following injuries, whether or not they are admitted to hospital: fractures, concussion, internal injuries, crushings, burns (excluding friction burns), severe cuts and lacerations, severe general shock requiring medical treatment and injuries causing death 30 or more days after the accident. A casualty is recorded as seriously or slightly injured by the police on the basis of information available within a short time of the accident. This generally will not reflect the results of a medical examination, but may be influenced according to whether the casualty is hospitalised or not. Hospitalisation procedures will vary regionally.
Numerator: source
Department for Transport.
Denominator: definition
2001 Census based mid-year population estimate for the year 2006.
Denominator: source
Office for National Statistics (ONS).
Data quality: Accuracy and completeness
Coverage is complete. Data quality varies as there are differences between police forces in procedures for recording, collecting and collating. There is no information available about the extent of this variation. Serious injury is not recorded on the basis of a clinical diagnosis. This may cause problems in classification of serious versus slight injury. Other sources of data which measure road injury and death are death registrations data (which matches well with police reports of road deaths) and hospital episode statistics. The match between hospital admissions and police reports of serious injury is discussed comprehensively in recent national reports: http://www.dft.gov.uk/pgr/roadsafety/research/rsrr/theme5/trendsfatalcar76.pdf http://eprints.ucl.ac.uk/archive/00003373/01/3373.pdf There are differences between police data and hospital records of road casualties due to differences in both definitions and reporting.
Table 3 – Indicator Technical Methods Numerator: extraction
Downloaded from www.dft.gov.uk
Numerator: aggregation /allocation
Calculated a 3-year average for 2005–2007 by summing the individual years and dividing by 3.
Health Profiles 2009 The Indicator Guide
263 Section 6: life expectancy and cause of death
Numerator data caveats
Research has shown that not all road casualties are reported to police. DfT published two papers on the level of under-reporting on 23 June 2006, which can be found at the addresses below: http://www.dft.gov.uk/pgr/roadsafety/research/rsrr/theme5/trendsfatalcar76.pdf http://eprints.ucl.ac.uk/archive/00003373/01/3373.pdf These reports suggest that the serious group of casualties could be up to twice as large as indicated by the ‘serious’ category in police records. Not all of this shortfall in the serious group of casualties is due to under-reporting because in the slight category are casualties which should be in the serious category and have been misclassified or misrecorded. These could add up to another 25% to the serious category. The police definition of ‘serious’ injury has remained consistent and therefore this data can be used to look at time trends. However, it should be noted that recording practices may have changed over time. An alternative to using casualties and deaths occurring on the roads in the area would have been to use a measure of resident casualties and deaths, regardless of where they occurred, using hospital admissions data. This would be a good measure of the health effects upon the population which lives in the district but would be less useful for targeting of road safety interventions. For this reason, occurring casualties and deaths were used in the Health Profiles.
Denominator data caveats
Ideally, the denominator should include all people travelling on public roads in the area (in vehicles and as pedestrians) and take account of the distances travelled. This measure is not available. The use of resident population as a denominator is a proxy measure for population exposure and is consistent with how this indicator is presented elsewhere.
Methods used to calculate indicator value
Calculation of the numerator: Calculated a 3-year average for 2005–2007 by summing the individual years and dividing by 3. Denominator count: Mid-2006 population estimates (all ages). The numerator was then divided by the denominator; the resulting value was then multiplied by 100,000 to give a crude rate per 100,000 population.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Any casualties recorded on the Isles of Scilly are included in the Cornwall numerator, consistent with how this data is published by DfT. Therefore the Isles of Scilly population is included in the Cornwall denominator.
Disclosure Control
Not applicable as no counts less than 5.
The City of London and Heathrow counts (numerator and denominator) are included in the London regional total and the England total.
Health Profiles 2009 The Indicator Guide
264 Section 6: life expectancy and cause of death
Confidence Intervals calculation method
The confidence intervals for these crude rates were constructed using the three year total in the following formula that relates the chi-square and Poisson distributions:
χ α ,2d 2
LL =
2
2
χ 1- α ,2(d+1) 2
2
UL =
2
where LL and UL are the lower and upper 100*(1-a) per cent confidence limits and d denotes the number of observed events (e.g, serious injuries, violent offences, (100*α) deaths) per unit of time exposed. χ² a/2,2d is the (100*a/2)th percentage point for a chisquared distribution with 2d degrees of freedom and χ²(1-α/2),2(d+1) is the (100*(1-a/2))th percentage point for a chi-squared distribution on 2(d+1) degrees of freedom.
The confidence limits for the rates were then obtained by dividing the upper and lower limits for the counts by the person time exposed. Reference: Dobson AJ, Kuulasmaa K, Eberle E, Scherer J. Confidence intervals for weighted sums of Poisson parameters. Statistics in Medicine 1991;10:457-462.
Health Profiles 2009 The Indicator Guide
265
Section 7: Charts and trend graphs
Health Profiles 2009 The Indicator Guide
266 Section 7: charts and trend graphs
33. DEPRIVATION CHART Basic Information 1. What is being measured?
Proportion of the population of an area living in each of the five national deprivation quintiles, based on the Indices of Deprivation 2007.
2. Why is it being measured?
The differences in deprivation between areas are a major determinant of health inequality in the UK.
3. How is this indicator actually defined?
These charts are based on the Index of Multiple Deprivation (IMD) 2007, which is a model measuring deprivation at the neighbourhood level. It is underpinned by separate dimensions of deprivation, these dimensions are weighted and an overall deprivation score is given. Each neighbourhood, or Lower Layer Super Output Area (LSOA), in England is allocated an IMD score. These are ranked for all 32482 LSOAs in England and divided into five equal groups or ‘national quintiles’. The charts are shaded according to the proportion of the relevant population in an area living in neighbourhoods belonging to each national deprivation quintile (where 1 is the most deprived quintile and 5 is the least deprived). (The relevant population is the population estimate used in the construction of the IMD 2007)
4. Who does it measure?
All persons, all ages in the relevant population (otherwise known as the ‘at risk’ population). (This is the population estimate used in the construction of the IMD 2007 – see Table 2)
5. When does it measure it?
2007 Based on various indicators from 2005 and 2001
6. Will It measure absolute numbers or proportions?
Proportion based on a composite indicator.
7. Where does the data actually come from?
DCLG
8. How accurate and complete will the data be?
All indicators included in the IMD 2007 were chosen using criteria that included accuracy and completion.
9. Are there any caveats/warnings/ problems?
The indicators are based on mainly 2005 data and this is therefore at least 4 years out of date. It is based on an average score of an area and it can’t be assumed to represent all individuals in that area. Although very comprehensive, some aspects of deprivation will not be included in the indices, for example if the data is incomplete or not collected. This may have a larger effect in some areas than others.
Health Profiles 2009 The Indicator Guide
267 Section 7: charts and trend graphs
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
n/a
Table 1 – Indicator Description Information component
Chart on p2 of District and County profiles.
Subject category/ domain(s)
Deprivation
Indicator name (*Indicator title in health profile)
Deprivation: a national perspective
PHO with lead responsibility
Yorkshire & Humber
Date of PHO dataset creation
February 2009
Indicator definition
This is a measure of the IMD 2007 which is a model measuring deprivation in an area. It is underpinned by separate dimensions of deprivation, these dimensions are weighted and an overall deprivation score is given. The charts are shaded according to the proportion of the relevant population in an area living in neighbourhoods belonging to each national deprivation quintile (where 1 is the most deprived quintile and 5 is the least deprived). (The relevant population is the population estimate used in the construction of the IMD 2007)
Geography
National, Regional, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs
Timeliness
Indicator is not regularly updated. Published 2004 and 2007, the definition is not completely consistent due to changes in some of the indicators used.
Rationale: What this indicator purports to measure
Level of deprivation of the relevant population of an area using a composite of domains which are made up of a range of indicators. LSOA areas are divided into quintiles where 1 is the most deprived quintile and 5 is the least.
Rationale: Public Health Importance
The differences in deprivation between areas are a major determinant of health inequality in the United Kingdom.
(The relevant population is the population estimate used in the construction of the IMD 2007)
Many studies and analyses have demonstrated the association of increasingly poor health with increasing deprivation. For instance, all cause mortality, smoking prevalence, self-reported long standing illness and a diet rich in fruit and vegetables are all positively or negatively correlated with deprivation. If deprivation inequality does not decrease, Acheson takes the view that health inequalities are also unlikely to decrease.
Health Profiles 2009 The Indicator Guide
268 Section 7: charts and trend graphs
Rationale: Purpose behind the inclusion of the indicator
To monitor and help reduce health inequalities.
Rationale: Policy relevance
• Acheson D. Report of the Independent Inquiry into Inequalities in Health. London: TSO, 1998. • Department of Health. The NHS Plan. London: TSO, 2000. • www.dh.gov.uk/assetRoot/04/05/57/83/04055783.pdf • Department of Health. Tackling Health Inequalities: A Programme for Action. Department of Health, 2003. • www.dh.gov.uk/assetRoot/04/01/93/62/04019362.pdf • HM Treasury 2007 PBR CSR: Public service agreements: • http://www.hm-treasury.gov.uk/ • Department for Communities and Local Government The New Performance Framework for Local Authorities & Local Authority Partnerships: Single Set of National Indicators Department for Communities and Local Government, 2007 http://www.communities.gov.uk/localgovernment/
Interpretation: What a high/low level of indicator value means
A high indicator value (dark shading) represents a high level of deprivation by national standards.
Interpretation: Potential for error due to type of measurement method
The indicators are based on 2005 and 2001 data and this is therefore at least 4 years out of date.
Interpretation: Potential for error due to bias and confounding
It is based on an average score of an area and can’t be assumed to represent all individuals in that area. Although very comprehensive, some aspects of deprivation will not be included in the indices, for example if the data is incomplete or not collected. This may have a larger effect in some area than others.
Confidence Intervals: Definition and purpose
N/A
Table 2 – Indicator Specification Indicator definition: Variable
Indices of Deprivation 2007
Indicator definition: Statistic
National quintile: All LSOAs in England are ranked according to deprivation score and split into five equal groups
Indicator definition: Gender
Persons
Indicator definition: age group
All ages
Indicator definition: period
2001 and 2005
Health Profiles 2009 The Indicator Guide
269 Section 7: charts and trend graphs
Indicator definition: scale
Proportion of relevant population in each national quintile.
Geography: geographies available for this indicator from other providers
IMD 2007 overall score and individual domains are available at Lower Layer Super Output Area (LSOA) level. Also, some summary scores are available at LA and County level from http://www.communities.gov.uk/communities/ neighbourhoodrenewal/deprivation/deprivation07/
Dimensions of inequality: subgroup analyses of this dataset available from other providers
IMD 2007 individual domains are available at Lower Layer Super Output Area (LSOA) level from http://www.communities.gov.uk/communities/ neighbourhoodrenewal/deprivation/deprivation07/
Data extraction: source
Department of Communities and Local Government
Data extraction: source URL
http://www.communities.gov.uk/communities/neighbourhoodrenewal/ deprivation/deprivation07/
Data extraction: date
February 2009
Numerator: definition
Number of the relevant population living in each quintile in England based on the IMD 2007 score.
(The relevant population is the population estimate used in the construction of the IMD 2007)
(The relevant population is the population estimate used in the construction of the IMD 2007). Numerator: source
Department of Communities and Local Government
Denominator: definition
Denominator data – ‘at risk’ mid-2005 population estimates (elsewhere referred to as the ‘relevant’ population). The figures have been adjusted from the ONS mid-year estimate to exclude the prison population in order to fit the definition of ‘at risk’. The figures have been subject to disclosure control. They are aggregated from LSOA level to the LA, county, region or national level.
Denominator: source
Office for National Statistics (ONS) and Department of Communities and Local Government (DCLG).
Data quality: accuracy and completeness
Criteria for inclusion of indicators to ID 2004 and ID 2007 included: • Up-to-date • Statistically robust • Available for the whole of England at a small level in a consistent form
TABLE 3 – INDICATOR TECHNICAL METHODS Numerator: extraction
Department of Communities and Local Government website http://www.communities.gov.uk/communities/neighbourhoodrenewal/deprivation/ deprivation07/
Numerator: aggregation/ allocation
The relevant population by national quintile of deprivation was aggregated from Lower Super Output Area to Local Authority, county, regional and national level. (The relevant population is the population estimate used in the construction of the IMD 2007).
Health Profiles 2009 The Indicator Guide
270 Section 7: charts and trend graphs
Numerator data caveats
Deprivation level is based on the average for the area; not all individuals in the area will be deprived.
Denominator data caveats
Census data and mid-year estimates are known to be deficient in their estimates of: • Non-white populations • Full-time students • Men aged 20-39 • People living in nursing homes etc • Rough sleepers • Inner-city populations • Households of multiple occupation • Migrants
Methods used to calculate indicator value
From the downloaded file with LSOA data. All LSOAs in England are ranked according to deprivation score and split into five equal quintiles. The relevant population from each LSOA is then aggregated to the area being measured and a proportion calculated for the population within that area in each national quintile.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Only included at regional and national level
Disclosure Control
Every effort has been made by the DCLG to ensure that data do not allow the disclosure of confidential information.
Confidence Intervals calculation method
N/A
(The relevant population is the population estimate used in the construction of the IMD 2007)
Health Profiles 2009 The Indicator Guide
271 Section 7: charts and trend graphs
34. LIFE EXPECTANCY BY DEPRIVATION QUINTILE CHART Basic Information 1. What is being measured?
Life expectancy by deprivation quintile.
2. Why is it being measured?
All cause mortality is a fundamental and probably the oldest measure of the health status of a population. Differences in levels of all-cause mortality reflect health inequalities between different population groups, e.g. between genders, social classes and ethnic groups. Life expectancy at birth is chosen as the preferred summary measure of all cause mortality as it quantifies the differences between areas in units (years of life) that are more readily understood and meaningful to the audience than those of other measures. Within every local authority there are different levels of deprivation. This indicator therefore looks at the within area inequalities in life expectancy.
3. How is this indicator actually defined?
Life expectancy at birth, years, all ages, 2003–2007, by national deprivation quintile.
4. Who does it measure?
Males, females, all ages.
5. When does it measure it?
Data for this indicator is updated annually.
6. Will it measure absolute numbers or proportions?
Measures life expectancy at birth in years.
7. Where does the data actually come from?
Office for National Statistics (ONS).
8. How accurate and complete will the data be?
Data on deaths are considered to be complete and robust.
9. Are there any caveats/warnings/ problems?
Death records without a valid area code are excluded but the number of such records is negligible. The populations used for the calculation of figures for this indicator are based on the 2001 Census, and are experimental estimates. Deprivation quintiles were assigned to Lower Layer Super Output Areas (LSOAs) at a national level. Results will be missing for areas where a local authority had none of its population living in a particular quintile. Results are also not presented for areas where the population at risk for the years 2003–07 was less than 5,000 i.e. where the total aggregated population for these years fell below this threshold. These areas are displayed as “n/a” in the data table.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
95% confidence intervals should be considered when making comparisons between areas or genders.
Health Profiles 2009 The Indicator Guide
272 Section 7: charts and trend graphs
Table 1 – Indicator Description Information component
Pg 2 Health Inequalities: life expectancy
Subject category/ domain(s)
Health Inequalities
Indicator name (* Indicator title in health profile)
Life expectancy
PHO with lead responsibility
LHO
Date of PHO dataset creation
March 2009
Indicator definition
Life expectancy at birth, years, all ages, 2003–07, males and females, by national deprivation quintiles (IMD 2007).
Geography
England, GOR, SHAs in the South East, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs.
Timeliness
ONS produced data are updated annually in the autumn of the following year.
Rationale: What this indicator purports to measure
Life expectancy at birth is a summary measure of the all cause mortality rates in an area in a given period. It is the average number of years a new-born baby would survive, were he or she to experience the particular area’s age-specific mortality rates for that time period throughout his or her life.
Rationale: Public Health Importance
All cause mortality is a fundamental and probably the oldest measure of the health status of a population. It represents the cumulative effect of the prevalence of risk factors, prevalence and severity of disease, and the effectiveness of interventions and treatment. Differences in levels of all-cause mortality reflect health inequalities between different population groups, e.g. between genders, social classes and ethnic groups. Life expectancy at birth is chosen as the preferred summary measure of all cause mortality as it quantifies the differences between areas in units (years of life) that are more readily understood and meaningful to the audience than those of other measures.
Rationale: Purpose behind the inclusion of the indicator
To help reduce premature mortality and facilitate planning of health services at local level.
Rationale: Policy relevance
There is a national health inequalities target for life expectancy which aims to increase average life expectancy at birth in England to 78.6 years for men and to 82.5 years for women, and to reduce health inequalities by 10% by 2010 as measured by life expectancy at birth (Department of Health PSA priority 1). The focus of the life expectancy target is on the quintile with the worst indicators of mortality and highest levels of deprivation, known as the Spearhead group. However, within every local authority there are different levels of deprivation. This indicator therefore looks at the within area inequalities in life expectancy by national deprivation quintile. Life expectancy is also an indicator in the following: Local basket of inequalities indicators - Indicator 13.12. Opportunity for all - Communities - Indicator 39. Quality of life indicators - Health and social well-being - Indicator 33
Health Profiles 2009 The Indicator Guide
273 Section 7: charts and trend graphs
Interpretation: What a high/low level of indicator value means
The higher the life expectancy, the longer the estimated expectation of life for males and females living in that area and deprivation quintile at that time.
Interpretation: Potential for error due to type of measurement method
The figures reflect the contemporary mortality among those living in the area in each time period. They are not the number of years a baby born in the area in each time period could actually expect to live, both because the death rates of the area are likely to change in the future and because many of those born in the area will live elsewhere for at least some part of their lives. Life expectancy at birth is also not a guide to the remaining expectancy of life at any other given age. For example, if female life expectancy at birth was 80 years for a particular area, life expectancy of women aged exactly 75 years in that area would exceed 5 years. This reflects the fact that survival from a particular age depends only on the mortality rates beyond that age, whereas survival from birth is based on mortality rates for all ages.
Interpretation: Potential for error due to bias and confounding
Older people living in nursing homes tend to be in poorer health than those not living in nursing homes. As these homes are unevenly distributed across the country, a higher death rate – consequently lower life expectancy level – in one area could simply be the result of migration of frail older people moving into nursing homes in that area.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider the confidence interval is, the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals can also be used to make comparisons between deprivation quintiles within each local authority. If the confidence intervals don’t overlap, this indicates that the difference in life expectancy is statistically significant.
Health Profiles 2009 The Indicator Guide
274 Section 7: charts and trend graphs
Table 2 – Indicator Specification Indicator definition: Variable
Life expectancy by national deprivation quintile (IMD 2007)
Indicator definition: Statistic
Life expectancy by national deprivation quintile (IMD 2007)
Indicator definition: Gender
Males and Females
Indicator definition: age group
All ages
Indicator definition: period
2003 to 2007
Indicator definition: scale Geography: geographies available for this indicator from other providers
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs.
Dimensions of inequality: subgroup analyses of this dataset available from other providers
None
Data extraction: Source
Calculated by LHO
Data extraction: source URL
N/A
Data extraction: date
N/A
Numerator: definition
Number of deaths by age, sex (2003-2007) and deprivation quintile (IMD 2007). Deprivation quintiles were assigned to LSOAs at a national level.
Numerator: source
Mortality data from Office for National Statistics (ONS), analysed by LHO
Denominator: definition
Lower Layer Super Output Area (LSOA) population estimates by age, sex (2003–2007) and deprivation quintile (IMD 2007). Deprivation quintiles were assigned to LSOAs at a national level. 2006 populations were revised estimates provided by ONS in February 2009. Population estimates for LSOAs were not available by single year of age from ONS for 2003. To provide estimates for the age groups under 1 and 1–4 (needed for the life table calculation), the age group 0–4 was divided into 5. This was based on an analysis of estimates from 2004–07 which had shown that populations were evenly distributed, at national level, across each year of age between 0–4. Population estimates for LSOAs are not published by single year of age by ONS, but are made available to PHOs for analysis purposes.
Health Profiles 2009 The Indicator Guide
275 Section 7: charts and trend graphs
Denominator: source
Small Area Population Estimates Unit, Office for National Statistics (ONS) Centre for Demography.
Data quality: Accuracy and completeness
Mortality data quality and coverage is extremely high. The figures are five-year averages so as to provide large enough numbers to ensure that the presented figures are sufficiently robust. However, areas with a population below 5,000 person years at risk across the period 2003–07, have not been included in line with ONS recommendations and are displayed as “n/a” in the data table. Where a local authority had none of its population living in a particular quintile no bar will be displayed in the chart for that quintile and no value presented in the data table. Two authorities, City of London and Isles of Scilly, are excluded from the results because of small numbers of deaths and populations in these areas.
Table 3 – Indicator Technical Methods Numerator: extraction
Data from ONS, extraction by LHO.
Numerator: aggregation/ allocation
Deaths were assigned to Lower Layer Super Output Areas (LSOA) by the LHO using the November 2008 version of the National Statistics Postcode Directory. LSOAs were each assigned to a national deprivation quintile and these were then aggregated within each geographical area.
Numerator data caveats Denominator data caveats
The populations used to calculate life expectancy are experimental estimates.
Methods used to calculate indicator value
Life expectancy was calculated using the South East Public Health Observatory life expectancy calculator: http://www.sepho.org.uk/viewResource. aspx?id=8943
Small Populations: How Isles of Scilly and City of London populations have been dealt with
City of London and Isles of Scilly are excluded from the results because of small numbers of deaths and populations in these areas.
Disclosure Control
Not applicable
Confidence Intervals calculation method
The calculation of the confidence intervals was made using the method developed by Chiang. The SEPHO calculator however also includes an adjustment to include a term for the variance associated with the final age interval as developed by Silcocks. Chiang CL. The Life Table and its Construction. In: Introduction to Stochastic Processes in Biostatistics. New York, John Wiley & Sons, 1968: 189-214.Silcocks PBS, Jenner DA, Reza R. Life expectancy as a summary of mortality in a population: statistical considerations and suitability for use by health authorities. J Epidemiol Community Health 2001; 55: 38-4
Health Profiles 2009 The Indicator Guide
276 Section 7: charts and trend graphs
35. TREND 1: ALL AGE, ALL CAUSE MORTALITY Basic Information 1. What is being measured?
Trend in death rates from all causes
2. Why is it being measured?
All cause mortality is a fundamental and probably the oldest measure of the health status of a population. It represents the cumulative effect of the prevalence of risk factors, prevalence and severity of disease, and the effectiveness of interventions and treatment. Differences in levels of all-cause mortality reflect health inequalities between different population groups, e.g. between genders, social classes and ethnic groups.
3. How is this indicator actually defined?
Mortality from all causes, directly age-standardised rate, males and females, all ages, 1996–98 to 2005– 07 (average of annual rates) per 100,000 European Standard Population.
4. Who does it measure?
All persons, all ages
5. When does it measure it?
This indicator is updated annually
6. Will it measure absolute numbers or proportions?
Proportions: numbers of deaths per hundred thousand European Standard Population
7. Where does the data actually come from?
Office for National Statistics (ONS)
8. How accurate and complete will the data be?
Data on deaths are considered to be complete and robust
9. Are there any caveats/warnings/ problems?
Area of residence is allocated by ONS using the postcode and the National Statistics Postcode Directory - records without a valid area code are excluded but the number of such records is negligible. The populations used for the calculation of the figures for this indicator are based on the 2001 Census, and are estimates.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
Data are age-standardised to allow comparisons between areas. 95% confidence intervals should be considered when making comparisons between areas or over time.
Health Profiles 2009 The Indicator Guide
277 Section 7: charts and trend graphs
TABLE 1 – INDICATOR DESCRIPTION Information component
Pg 3 Health Inequalities: changes over time
Subject category/ domain(s)
Health inequalities: changes over time
Indicator name (*Indicator title in health profile)
Trend in death rates from all causes.
PHO with lead responsibility
LHO
Date of PHO dataset creation
March 2009
Indicator definition
Mortality from all causes, directly age-standardised rate, males and females, all ages, 1996-98 to 2005-07 (average of annual rates), per 100,000 European Standard Population
Geography
England, GOR, Strategic Health Authorities for the South East, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs (boundaries as at April 2009).
Timeliness
The Compendium mortality from all causes indicator is updated annually, usually around November following the publication by ONS of the new year’s mortality extract (usually in May) and mid-year population estimates (usually August-September).
Rationale: What this indicator purports to measure
Trend in mortality from all causes.
Rationale: Public Health Importance
All cause mortality is a fundamental and probably the oldest measure of the health status of a population. It represents the cumulative effect of the prevalence of risk factors, prevalence and severity of disease, and the effectiveness of interventions and treatment. Differences in levels of all-cause mortality reflect health inequalities between different population groups, e.g. between genders, social classes and ethnic groups.
Rationale: Purpose behind the inclusion of the indicator
To monitor mortality rates due to all causes over time.
Rationale: Policy relevance
There is a national health inequalities target for life expectancy which aims to increase average life expectancy at birth in England to 78.6 years for men and to 82.5 years for women, and to reduce health inequalities by 10% by 2010 as measured by life expectancy at birth (Department of Health PSA priority 1). All age all cause mortality is a good proxy for life expectancy and is being used to measure progress towards meeting the life expectancy target. The baseline for monitoring this target is the three year period 1995–97. The ten year trend charts presented in the health profiles show the three year moving averages of annual mortality rates from all causes from the 1996–98 to the most recently available years of data (2005–07 average) compared against the national trend. Earlier trend years are available from previous Health Profiles
Health Profiles 2009 The Indicator Guide
278 Section 7: charts and trend graphs
Interpretation: What a high/low level of indicator value means
An upward sloping trend line indicates that the mortality rate from all causes is worsening. A downward sloping trend line indicates that the mortality rate from all causes is improving. If the trend line for the local authority is consistently above the trend line for England then mortality rates from all causes have been consistently higher than those for England for the stated period. If the trend line for the local authority is consistently below the trend line for England then mortality rates from all causes have been consistently lower than those for England for the stated period. If the local authority line is above that of England and gap between the local authority trend line and the England trend line is widening then inequalities in mortality rates from all causes are worsening. If the authority line is above that of England and the gap between the local authority trend line and the England trend line is narrowing then inequalities in mortality rates from all causes are improving.
Interpretation: Potential for error due to type of measurement method
Coverage can be considered to be complete as the registration of deaths is a legal requirement. Data quality for the relevant fields (age, sex, area of residence) is extremely high.
Interpretation: Potential for error due to bias and confounding
The rates are age-standardised. This improves the comparability of rates for different areas, or between different time periods, by taking into account differences in the age structures of the populations being compared.
TABLE 2 – INDICATOR SPECIFICATION Indicator definition: Variable
Mortality from all causes
Indicator definition: Statistic
Directly age-standardised rate
Indicator definition: Gender
Male and Female
Indicator definition: age group
All ages
Indicator definition: period
1996-98 to 2005-07 -average of annual rates, with the exception of new 2009 Unitary Authorities: Cheshire East, Cheshire West and Chester, Central Bedfordshire, and Cornwall. Rates for these areas were based on data pooled for each three-year period.
Indicator definition: scale
Per 100,000 European Standard population
Geography: geographies available for this indicator from other providers
England & Wales, ONS area classification, Primary Care Organisation, Strategic Health Authority. Available from National Centre for Health Outcomes Development (NCHOD) website www.nchod.nhs.uk
Health Profiles 2009 The Indicator Guide
279 Section 7: charts and trend graphs
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Age, gender available from NCHOD.
Data extraction: Source
NCHOD.
Data extraction: source URL
Data received directly from NCHOD with the exception of new 2009 Unitary Authorities: Cheshire East, Cheshire West and Chester, Central Bedfordshire, and Cornwall. Rates for these areas were calculated by the LHO using annual mortality files, supplied to PHOs by ONS, and published ONS mid-year population estimates.
Data extraction: date
February 2009
Numerator: definition
Deaths from all causes, registered in the respective calendar years 1996–98 to 2005–07.
Numerator: source
Office for National Statistics (ONS)
Denominator: definition
2001 Census based mid-year population estimates for respective calendar years 1996 to 2007, people of all ages, current as at 21 August 2008.
Denominator: source
ONS
Data quality: Accuracy and completeness
Coverage can be considered to be complete as the registration of deaths is a legal requirement. Data quality for the relevant fields (age, sex, area of residence) is extremely high. Area of residence is allocated by ONS using the postcode and the National Statistics Postcode Directory - records without a valid area code are excluded but the number of such records is negligible.
Health Profiles 2009 The Indicator Guide
280 Section 7: charts and trend graphs
Table 3 – Indicator Technical Methods Numerator: extraction
Extraction by NCHOD with the exception of new 2009 Unitary Authorities: Cheshire East, Cheshire West and Chester, Central Bedfordshire, and Cornwall. Deaths for these areas were extracted by the LHO from annual mortality files, supplied to PHOs by ONS
Numerator: aggregation/ allocation
Deaths were assigned to geographical areas using the area code supplied in the mortality extract. This is derived from postcode of residence by the ONS using the National Statistics Postcode Directory (NSPD).
Numerator data caveats
Area of residence is allocated by ONS using the postcode and the National Statistics Postcode Directory - records without a valid area code are excluded but the number of such records is negligible.
Denominator data caveats
Data are based on the latest revisions of ONS mid-year population estimates for the respective years, current as at 21 August 2008.
Methods used to calculate indicator value
The directly age-standardised rate (DSR) is the rate of events that would occur in a population with a standard age structure if that population were to experience the age-specific rates of the subject population. The standard population used is the European Standard Population. The rate is expressed per 100,000 population. Data for some new Unitary Authorities, introduced in April 2009, did not correspond with existing geographies (Cheshire East, Cheshire West and Chester, Central Bedfordshire, and Cornwall). Results for these areas were calculated by the LHO. Deaths from 1995 to 2007 were assigned to current boundaries using the November 2008 release of the National Statistics Postcode Directory. Rates were based on pooled deaths and populations for each three year period from 1995–97 to 2005–07. Results for the other new UAs were taken from published data from NCHOD for local authorities (Bedford) or counties (County Durham, Northumberland, Shropshire, and Wiltshire).
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Isles of Scilly and City of London are excluded from the lower tier datasets but included in England, Regional and County figures.
Disclosure Control
None applied.
Confidence Intervals calculation method
Not applicable – Trend chart presented without Confidence Intervals
Health Profiles 2009 The Indicator Guide
281 Section 7: charts and trend graphs
36. T REND 2: EARLY DEATHS FROM HEART DISEASE AND STROKE Basic Information 1. What is being measured?
Trend in early death rates from all circulatory diseases (including heart disease and stroke)
2. Why is it being measured?
Circulatory disease accounts for 40% of all deaths (30% under 75). Mortality is a direct measure of health care need indicating the overall circulatory disease burden on the population, reflecting both the incidence of disease and the ability to treat it.
3. How is this indicator actually defined?
Mortality from all circulatory diseases, directly age-standardised rate, persons, under 75, 1996–98 to 2005–07 (average of annual rates), per 100,000 European Standard population
4. Who does it measure?
People aged under 75
5. When does it measure it?
Updated annually
6. Will It measure absolute numbers or proportions?
Directly age-standardised rate
7. Where does the data actually come from?
Office for National Statistics (ONS)
8. How accurate and complete will the data be?
Mortality counts are derived from an annual mortality extract supplied by ONS and are based on the original underlying cause of death for which there is nearly 100% coverage on the mortality register.
9. Are there any caveats/ warnings/problems?
Area of residence is allocated by ONS using the postcode and the National Statistics Postcode Directory – records without a valid area code are excluded but the number of such records is negligible. There is the potential for the underlying cause of death to be incorrectly attributed on the death certificate and, therefore, the cause of death misclassified.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
Data are age-standardised to allow comparisons between areas. 95% confidence intervals should be considered when making comparisons between areas or over time.
TABLE 1 – INDICATOR DESCRIPTION Information component
Pg 3 Health Inequalities: changes over time
Subject category/ domain(s)
Health inequalities: changes over time
Indicator name (*Indicator title in health profile)
Trend in early mortality rates from all circulatory diseases
Health Profiles 2009 The Indicator Guide
282 Section 7: charts and trend graphs
PHO with lead responsibility
SEPHO
Date of PHO dataset creation
05 February 2009
Indicator definition
Mortality from all circulatory diseases, directly age-standardised rate, persons, under 75, 1996–98 to 2005–07 (average of annual rates), per 100,000 European Standard population
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs (boundaries as at April 2008, new 2009 Unitary Authorities as at April 2009).
Timeliness
The Compendium mortality from all circulatory diseases indicator is updated annually, usually around November/December following the publication by ONS of the new year’s mortality extract (usually in May) and mid-year population estimates (usually August–September).
Rationale: What this indicator purports to measure
Trend in early mortality from all circulatory diseases.
Rationale: Public Health Importance
Circulatory disease accounts for 40% of all deaths (30% under 75). Mortality is a direct measure of health care need reflecting the overall circulatory disease burden on the population, both the incidence of disease and the ability to treat it. The mortality rate may be improved by reducing the population’s risk (e.g. encouraging healthier lifestyles and reducing exposure to smoking), by earlier detection of disease and by more effective treatment. The Our Healthier Nation target is to reduce the number of deaths from circulatory disease in people aged under 75 years by at least 40% by 2010. The baseline for monitoring this target is the three year period 1995–97. The ten year trend charts presented in the health profiles show three year moving averages of annual mortality rates from all circulatory disease from 1996–8 to the most recently available years of data (2005–07 average) compared against the national trend. Earlier trend years are available from previous Health Profiles and on the Health Profiles website.
Rationale: Purpose behind the inclusion of the indicator
To monitor premature mortality due to circulatory diseases over time. To reduce premature deaths from circulatory diseases.
Rationale: Policy relevance
The under 75 circulatory disease mortality rate is a key target indicator in the 1999 Public Health White Paper ‘Saving Lives: Our Healthier Nation’. The target is to reduce the number of deaths from circulatory disease in people aged under 75 years by at least two-fifths by 2010. The baseline for monitoring this target is the three year period 1995–97. This measure supports delivery of the Department of Health PSA targets and LDP and is relevant to Choosing Health, Coronary Heart Disease NSF and Programme for Action.
Health Profiles 2009 The Indicator Guide
283 Section 7: charts and trend graphs
Interpretation: What a high/low level of indicator value means
An upward sloping trend line indicates that the early mortality rate from all circulatory diseases is worsening. A downward sloping trend line indicates that the early mortality rate from all circulatory diseases is improving. If the trend line for the local authority is consistently above the trend line for England then early death rates from circulatory diseases have been consistently higher than those for England for the stated period. If the trend line for the local authority is consistently below the trend line for England then early death rates from circulatory diseases have been consistently lower than those for England for the stated period. If the gap between the local authority trend line and the England trend line is widening then inequalities in early mortality rates from all circulatory diseases are worsening. If the gap between the local authority trend line and the England trend line is narrowing then inequalities in early mortality rates from all circulatory diseases are improving.
Interpretation: Potential for error due to type of measurement method
Coverage can be considered to be complete as the registration of deaths is a legal requirement. Data quality for the relevant fields (age, sex, underlying cause of death, area of residence) is extremely high.
Interpretation: Potential for error due to bias and confounding
The rates are age-standardised. This improves the comparability of rates for different areas, or between different time periods, by taking into account differences in the age structures of the populations being compared.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate.
There is the potential for the underlying cause of death to be incorrectly attributed on the death certificate and, therefore, the cause of death misclassified.
This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health.
Health Profiles 2009 The Indicator Guide
284 Section 7: charts and trend graphs
Table 2 – Indicator Specification Indicator definition: Variable
Mortality from all circulatory diseases (ICD10 I00 –I99, ICD9 390-459 adjusted)
Indicator definition: Statistic
Directly age-standardised rate
Indicator definition: Gender
Persons
Indicator definition: age group
Under 75
Indicator definition: period
1996–98 to 2005–07 (average of annual rates)
Indicator definition: scale
Per 100,000 European Standard population
Geography: geographies available for this indicator from other providers
England & Wales, ONS area, Primary Care Organisation, Strategic Health Authority.
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Age, gender available from NCHOD.
Data extraction: Source
NCHOD.
Data extraction: source URL
Data received directly from NCHOD.
Data extraction: date
14 January 2009
Numerator: definition
Deaths from all circulatory disease, classified by underlying cause of death (ICD10 I00 – I99, ICD9 390-459 adjusted), registered in the respective calendar years 1996–98 to 2005–07, in people aged under 75.
Numerator: source
Office for National Statistics (ONS)
Denominator: definition
2001 Census based mid-year population estimates for respective calendar years 1996 to 2007, people aged under 75, current as at 29 September 2008.
Denominator: source
ONS
Data quality: Accuracy and completeness
Coverage can be considered to be complete as the registration of deaths is a legal requirement. Data quality for the relevant fields (age, sex, underlying cause of death, area of residence) is extremely high. Area of residence is allocated by ONS using the postcode and the National Statistics Postcode Directory – records without a valid area code are excluded but the number of such records is negligible.
Available from National Centre for Health Outcomes Development (NCHOD) website www.nchod.nhs.uk Data can also be found at Neighbourhood Renewal Unit Public Service Agreement Floor Targets (http://www.fti.communities.gov.uk/fti/).
Health Profiles 2009 The Indicator Guide
285 Section 7: charts and trend graphs
Table 3 – Indicator Technical Methods Numerator: extraction
Extraction by NCHOD.
Numerator: aggregation / allocation
Deaths were assigned to geographical areas using the area code supplied in the mortality extract. This is derived from postcode of residence by the ONS using the National Statistics Postcode Directory (NSPD).
Numerator data caveats
Area of residence is allocated by ONS using the postcode and the National Statistics Postcode Directory – records without a valid area code are excluded but the number of such records is negligible. Mortality counts are derived from the annual DH mortality extract supplied by ONS and are based on the original underlying cause of death for which there is nearly 100% coverage on the mortality register. In January 2001, the ONS implemented a change from ICD-9 to ICD-10 for coding causes of death in England & Wales. As part of an exercise to investigate the effects of this change, the ONS also re-coded all deaths registered in 1999. Deaths for years prior to 1999 and for year 2000 have not been re-coded. The numbers of deaths observed in the years 1996-98 and 2000 have, therefore, been adjusted to give “expected” numbers of deaths which would have been coded to this cause in ICD-10. This was done by multiplying the ICD-9 based death counts by the appropriate ICD-10/9 comparability ratio published by the ONS. For this indicator the following ICD-10/9 comparability ratios were used: Males aged 0-74: 1.012; Females aged 0–74: 1.015. Adjusted person counts are the sum of the adjusted male and female counts
Denominator data caveats
Data are based on the latest revisions of ONS mid-year population estimates for the respective years, current as at 29 September 2008.
Methods used to calculate indicator value
The directly age-standardised rate (DSR) is the rate of events that would occur in a population with a standard age structure if that population were to experience the age-specific rates of the subject population. The standard population used is the European Standard Population. The age groups used are: Under 1, 1–4, 5–9,…, 80–84, 85+. The rates for each time period e.g. 2005-07 have been calculated as the simple average of the individual annual rates. The rate is expressed per 100,000 population.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Isles of Scilly and City of London are excluded from the lower tier datasets but included in England and Regional figures.
Disclosure Control
None applied.
Confidence Intervals calculation method
Not applicable – Trend chart presented without Confidence Intervals
Health Profiles 2009 The Indicator Guide
286 Section 7: charts and trend graphs
37. TREND 3: EARLY DEATH RATES FROM CANCER Basic Information 1. What is being measured?
Trend in early death rates from cancer
2. Why is it being measured?
Early mortality from cancer is a direct measure of health care need as public health interventions for prevention, early diagnosis, and effective treatment can all reduce the burden of cancer morbidity and mortality.
3. How is this indicator actually defined?
Mortality from all cancers, directly age-standardised rate, persons, under 75, 1996–98 to 2005–07 (average of annual rates), per 100,000 European Standard population
4. Who does it measure?
People aged under 75
5. When does it measure it?
Updated annually
6. Will It measure absolute numbers or proportions?
Directly age-standardised rate
7. Where does the data actually come from?
Office for National Statistics (ONS)
8. How accurate and complete will the data be?
Mortality counts are derived from an annual mortality extract supplied by ONS and are based on the original underlying cause of death for which there is nearly 100% coverage on the mortality register.
9. Are there any caveats/warnings/ problems?
Area of residence is allocated by ONS using the postcode and the National Statistics Postcode Directory – records without a valid area code are excluded but the number of such records is negligible. There is the potential for the underlying cause of death to be incorrectly attributed on the death certificate and, therefore, the cause of death misclassified.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
Data are age-standardised to allow comparisons between areas. 95% confidence intervals should be considered when making comparisons between areas or over time.
Health Profiles 2009 The Indicator Guide
287 Section 7: charts and trend graphs
Table 1 – Indicator Description Information component
County Health Profiles: Pg 3 Health Summary – Indicator 34
Subject category/ domain(s)
Health inequalities: changes over time
Indicator name (*Indicator title in health profile)
Trend in early death rates from cancer
PHO with lead responsibility
SEPHO
Date of PHO dataset creation
05 February 2009
Indicator definition
Mortality from all cancers, directly age-standardised rate, persons, under 75, 1996-98 to 2005-07 (average of annual rates), per 100,000 European Standard population
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs (boundaries as at April 2008, new 2009 Unitary Authorities as at April 2009).
Timeliness
Updated annually.
Rationale: What this indicator purports to measure
Trend in early mortality from all cancers.
Rationale: Public Health Importance
Cancer is amongst the three leading causes of death at all ages except for pre-school age children in the UK. It accounts for 26% all deaths. If current incidence rates remain the same, by 2025 there will be an additional 100,000 cases of cancer diagnosed each year as a result of the ageing population. Inequalities exist in cancer rates between the most deprived areas and the most affluent. Early mortality from cancer is a direct measure of health care need as public health interventions for prevention, early diagnosis, effective treatment can all reduce the burden of cancer morbidity and mortality. The Our Healthier Nation target is to reduce the number of deaths from cancer in people aged under 75 years by at least a fifth by 2010. The baseline for monitoring this target is the three year period 1995–97. The ten year trend charts presented in the health profiles show the three year moving averages of annual mortality rates from all cancer from 1996–8 to the most recently available years of data (2005–07 average) compared against the national trend. Earlier trend years are available from previous Health Profiles and on the Health Profiles website.
Rationale: Purpose behind the inclusion of the indicator
To monitor premature mortality due to cancer over time. To reduce premature deaths from cancer.
Health Profiles 2009 The Indicator Guide
288 Section 7: charts and trend graphs
Rationale: Policy relevance
The directly age-standardised mortality rate from all cancers for persons aged under 75 is a target indicator in the Saving Lives: Our Healthier Nation strategy. The target is a 20% reduction by the year 2010 from the baseline rate in 1995–97. This measure supports delivery of the Department of Health PSA targets and LDP and is relevant to Choosing Health, Cancer NSF and Programme for Action.
Interpretation: What a high/low level of indicator value means
An upward sloping trend line indicates that the early mortality rate from all cancers is worsening. A downward sloping trend line indicates that the early mortality rate from all cancers is improving. If the trend line for the local authority is consistently above the trend line for England then early death rates from cancer have been consistently higher than those for England for the stated period. If the trend line for the local authority is consistently below the trend line for England then early death rates from cancer have been consistently lower than those for England for the stated period. If the gap between the local authority trend line and the England trend line is widening then inequalities in early mortality rates from all cancers are worsening. If the gap between the local authority trend line and the England trend line is narrowing then inequalities in early mortality rates from all cancers are improving.
Interpretation: Potential for error due to type of measurement method
Coverage can be considered to be complete as the registration of deaths is a legal requirement. Data quality for the relevant fields (age, sex, underlying cause of death, area of residence) is extremely high.
Interpretation: Potential for error due to bias and confounding
The rates are age-standardised. This improves the comparability of rates for different areas, or between different time periods, by taking into account differences in the age structures of the populations being compared.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate.
There is the potential for the underlying cause of death to be incorrectly attributed on the death certificate and, therefore, the cause of death misclassified.
This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health.
Health Profiles 2009 The Indicator Guide
289 Section 7: charts and trend graphs
Table 2 – Indicator Specification Indicator definition: Variable
Mortality from all cancers (ICD10 C00-C97, ICD9 140-208 adjusted)
Indicator definition: Statistic
Directly age-standardised rate
Indicator definition: Gender
Persons
Indicator definition: age group
Under 75
Indicator definition: period
1996-98 to 2005-07 (average of annual rates)
Indicator definition: scale
Per 100,000 European Standard population
Geography: geographies available for this indicator from other providers
England & Wales, ONS area, Primary Care Organisation, Strategic Health Authority.
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Age, gender available from NCHOD.
Data extraction: Source
NCHOD.
Data extraction: source URL
Data received directly from NCHOD.
Data extraction: date
14 January 2009
Numerator: definition
Deaths from all malignant neoplasms, classified by underlying cause of death (ICD10 C00-C97, ICD9 140-208 adjusted), registered in the respective calendar years 1996–98 to 2005–07, in people aged under 75.
Numerator: source
Office for National Statistics (ONS)
Denominator: definition
2001 Census based mid-year population estimates for respective calendar years 1996 to 2007, people aged under 75, current as at 29 September, 2008.
Denominator: source
ONS
Data quality: Accuracy and completeness
Coverage can be considered to be complete as the registration of deaths is a legal requirement. Data quality for the relevant fields (age, sex, underlying cause of death, area of residence) is extremely high. Area of residence is allocated by ONS using the postcode and the National Statistics Postcode Directory – records without a valid area code are excluded but the number of such records is negligible.
Available from National Centre for Health Outcomes Development (NCHOD) website www.nchod.nhs.uk Data can also be found at Neighbourhood Renewal Unit Public Service Agreement Floor Targets (http://www.fti.communities.gov.uk/fti/).
Figures for various NRF and NDC and Spearhead Groups are available via the Neighbourhood Renewal Unit http://www.neighbourhood.gov.uk/
Health Profiles 2009 The Indicator Guide
290 Section 7: charts and trend graphs
Table 3 – Indicator Technical Methods Numerator: extraction
Extraction by NCHOD.
Numerator: aggregation / allocation
Deaths were assigned to geographical areas using the area code supplied in the mortality extract. This is derived from postcode of residence by the ONS using the National Statistics Postcode Directory (NSPD).
Numerator data caveats
Area of residence is allocated by ONS using the postcode and the National Statistics Postcode Directory -– records without a valid area code are excluded but the number of such records is negligible. Mortality counts are derived from the annual DH mortality extract supplied by ONS and are based on the original underlying cause of death for which there is nearly 100% coverage on the mortality register. In January 2001, the ONS implemented a change from ICD-9 to ICD-10 for coding causes of death in England & Wales. As part of an exercise to investigate the effects of this change, the ONS also re-coded all deaths registered in 1999. Deaths for years prior to 1999 and for year 2000 have not been re-coded. The numbers of deaths observed in the years 1996–98 and 2000 have, therefore, been adjusted to give “expected” numbers of deaths which would have been coded to this cause in ICD-10. This was done by multiplying the ICD-9 based death counts by the appropriate ICD-10/9 comparability ratio published by the ONS. For this indicator the following ICD-10/9 comparability ratios were used: Males aged 0–74: 1.013; Females aged 0–74: 1.009. Adjusted person counts are the sum of the adjusted male and female counts.
Denominator data caveats
Data are based on the latest revisions of ONS mid-year population estimates for the respective years, current as at 29 September, 2008.
Methods used to calculate indicator value
The directly age-standardised rate (DSR) is the rate of events that would occur in a population with a standard age structure if that population were to experience the age-specific rates of the subject population. The standard population used is the European Standard Population. The age groups used are: Under 1, 1–4, 5–9,…, 80–84, 85+. The rates for each time period e.g. 2005-2007 have been calculated as the simple average of the individual annual rates. The rate is expressed per 100,000 population.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Isles of Scilly and City of London are excluded from the lower tier datasets but included in England and Regional figures.
Disclosure Control
None applied.
Confidence Intervals calculation method
Not applicable – Trend chart presented without Confidence Intervals
Health Profiles 2009 The Indicator Guide
291 Section 7: charts and trend graphs
38. HEALTH INEQUALITIES: ETHNICITY CHART Basic Information 1. What is being measured?
The percentage of children in each ethnic group who are eligible for free school meals
2. Why is it being measured?
Eligibility for free school meals is an indicator of deprivation and people who suffer more deprivation tend to have poorer health.
3. How is this indicator actually defined?
The percentage of children by each ethnic group who are eligible for free school meals.
4. Who does it measure?
School age children aged 5–15
5. When does it measure it?
January 2008
6. Will It measure absolute numbers or proportions?
Both proportions and absolute numbers are presented.
7. Where does the data actually come from?
Department for Children, Schools and Families
8. How accurate and complete will the data be?
This indicator only contains data for LEA maintained Primary, Secondary and Special schools.
9. Are there any caveats/warnings/ problems?
Excludes pupils educated in private schools. Numbers presented are rounded to the nearest 10.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
95% confidence intervals should be considered when making comparisons between areas or ethnic groups.
Table 1 – Indicator Description Information component
Page 3 - Health Inequalities: ethnicity
Subject category/ domain(s)
Health inequalities
Indicator name (* Indicator title in health profile)
Ethnicity
PHO with lead responsibility
EMPHO
Date of PHO dataset creation
March 2009
Indicator definition
Percentage of children in maintained schools eligible for free school meals by ethnic group.
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs
Health Profiles 2009 The Indicator Guide
292 Section 7: charts and trend graphs
Timeliness
Every school is required to supply information to the DCSF each January through the Pupil Level Annual Schools Census (PLASC).
Rationale: What this indicator purports to measure
Percentage of children in low socio-economic groups as measured by those eligible for free school meals.
Rationale: Public Health Importance
Ethnic inequalities in health are well known1, but are difficult to monitor and address due to limited data2. Eligibility for free school meals (FSM) has frequently been used as a proxy measure of poverty, disadvantage, and social exclusion as information on adult’s income is not collected. Eligibility for Free School Meals is strongly associated with low educational achievement. 3 Aspinall P, Jacobson B. Ethnic disparities in health and health care: A focused review of the evidence and selected examples of good practice. London: LHO, 2004. Aspinall P, Jacobson B, Polato GM. Missing record: The case for recording ethnicity at birth and death registration. London: LHO, 2003. Tackling low educational achievement, Robert Cassen and Geeta Kingdon, 2007 www.jrf.org.uk/bookshop/eBooks/2063-education-schools-achievement. pdf
Rationale: Purpose behind the inclusion of the indicator
This indicator can help to indicate within a local area which ethnic groups are likely to experience poorer health than others and could be used to predict health care need and target resources.
Rationale: Policy relevance
There are four national indicators which relate to the educational achievement of children eligible for free school meals. National indicator no.81 ‘Inequality gap in the achievement of a Level 3 qualification by the age of 19’. This indicator reports the gap in attainment of level 3 at age 19 in each Local Authority between those young people who were in receipt of free school meals at academic age 15 and those who were not. At the moment significantly fewer young people in receipt of free school meals at the age of 15 achieve L3 qualifications by the age of 19 than their peers who were not in receipt of free school meals at age 15. National indicator 82 ‘Inequality gap in the achievement of a Level 2 qualification by the age of 19’. This indicator reports the percentages of young people who were in receipt of free school meals at academic age 15 who attain level 2 qualifications by the age of 19. At the moment significantly fewer young people in receipt of free school meals at the academic age of 15 achieve L2 qualifications by the age of 19 than their peers who were not in receipt of free school meals at academic age 15. National indicator no. 102 ‘Achievement gap between pupils eligible for free school meals and their peers achieving the expected level at Key Stages 2 and 4’. This indicator seeks to narrow the gap in achievement between children from disadvantaged backgrounds and their peers. National Indicator 106: Young people from low income backgrounds progressing to higher education. The indicator takes the form of the gap between the proportions of 15 year olds eligible for free school meals (FSM) and those not eligible for FSM progressing to higher education at the age of 18 or 19.
Health Profiles 2009 The Indicator Guide
293 Section 7: charts and trend graphs
Interpretation: What a high/low level of indicator value means
A high percentage indicates an ethnic group with a high risk of poor educational achievement. The numbers provided allow local authorities to estimate the burden of this risk in their populations. Ethnic groups with a higher percentage than the England average, which is shown on the chart for each ethnic group, are at increased risk of poor educational achievement when compared to the England average. Ethnic groups with a lower percentage than the England average, which is shown on the chart for each ethnic group, are at decreased risk of poor educational achievement when compared to the England average.
Interpretation: Potential for error due to type of measurement method
The subjective, multi-faceted and changing nature of ethnic identification makes it a particularly difficult piece of information to collect. There is no consensus on what constitutes an ‘ethnic group’. Membership of any ethnic group is something that is subjectively meaningful to the person concerned and the terminology used to describe ethnic group has changed markedly over time. As a result, ethnic group, however defined or measured, will tend to change over time depending on social and political attitudes or developments. Basing ethnic identification upon an objective and rigid classification of ethnic groups is not therefore achievable in practice. In particular research has shown that people with mixed backgrounds may vary their answer to questions about ethnicity to suit their perception of the form they are completing.
Interpretation: Potential for error due to bias and confounding
The statistic measures eligibility for free school meals attainment in LEA maintained schools only.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant. If the interval does not include the national value, the difference is statistically significant.
Health Profiles 2009 The Indicator Guide
294 Section 7: charts and trend graphs
Table 2 – Indicator Specification Indicator definition: Variable
For each ethnic group the indicator is defined as the number of pupils who are eligible for free school meals (numerator), divided by all pupils (denominator), expressed as a percentage.
Indicator definition: Statistic
Percentage by ethnic group
Indicator definition: Gender
Persons
Indicator definition: age group
Aged 5–15
Indicator definition: period
January 2008
Indicator definition: scale
Percentage and number
Geography: geographies available for this indicator from other providers
Number and percent of pupils eligible for free school meals is available for England, Regional and LEA levels on the DFES website. http://www.dcsf.gov. uk/rsgateway/DB/SFR/s000786/index.shtml
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Number and percent of pupils eligible for free school meals is available for by primary, secondary and special school for England, Regional and LEA levels on the DFES website. http://www.dcsf.gov.uk/rsgateway/DB/SFR/s000786/index. shtml
Data extraction: Source
Pupil level annual schools census (PLASC).
Data extraction: source URL
Received directly from Department for Children, Schools and Families
Data extraction: date
Final dataset received March 2009.
Numerator: definition
Eligibility to receive free school meals is dependant on a child’s parent(s) or carer(s) receiving: Income Support (IS); Income Based Jobseekers Allowance (IBJSA); Support under part VI of the Immigration and Asylum Act 1999; Child Tax Credit, provided they are not entitled to working Tax Credit and have an annual income, as assessed by the Inland Revenue, that (for 2007/2008) does not exceed £14,495 and £15,575 (for 2008/2009); the guaranteed element of State Pension Credit. Children who receive IS or IBJSA in their own right are also entitled to free school meals.
Numerator: source
Pupil level annual schools census (PLASC).
Number and percent of pupils by ethnic group is available for England, Regional and LEA levels on the DFES website. http://www.dcsf.gov.uk/ rsgateway/DB/SFR/s000786/index.shtml
Number and percent of pupils by ethnic group by primary and secondary schools is available for England, Regional and LEA levels on the DFES website. http://www.dcsf.gov.uk/rsgateway/DB/SFR/s000786/index.shtml
Health Profiles 2009 The Indicator Guide
295 Section 7: charts and trend graphs
Denominator: definition
Data for five ethnic groups have been presented. Data is collected for 18 groups and have been combined as follows. • White Ethnic group includes those classified as: White British, Irish, Traveller of Irish Heritage, Gypsy/Roma, Any other White background. • Mixed Ethnic group includes those classified as: White and Black Caribbean, White and Black African, White and Asian, Any other Mixed background. • Asian Ethnic group includes those classified as: Indian, Pakistani, Bangladeshi, Any other Asian background. • Black Ethnic group includes those classified as: Black Caribbean, Black African, Any other Black background. • Chinese and any other ethnic group includes: Chinese and any other ethnic group.
Denominator: source
Pupil level annual schools census (PLASC).
Data quality: Accuracy and completeness
Every school is required to supply information to the DCSF each January through the Pupil Level Annual Schools Census (PLASC). The Census collects school level data about classes being taught and staff in schools and also pupil level information from Maintained Primary, Middle, Secondary and Specials schools. The data items collected in PLASC include a unique pupil number, pupil name, gender, date of birth, home postcode, ethnic group, first language, free school meal eligibility and special educational needs information. Approximately 1.5% of pupils were unclassified by ethnic group that is the information was either refused on not obtained.
TABLE 3 – INDICATOR TECHNICAL METHODS Numerator: extraction
Received directly from Department for Children, Schools and Families
Health Profiles 2009 The Indicator Guide
296 Section 7: charts and trend graphs
Numerator: aggregation / allocation
County, Regional and England numerators were aggregated from their constituent Districts, UAs, MCDs and London Boroughs. Since the 2008 PLASC, there have been new Unitary Authorities created. The new Unitary Authorities of Northumberland, Shropshire, Wiltshire and Durham have the same boundaries as the former counties of the same name so have been allocated these data. Data for five other new unitary Authorities have been calculated from the aggregation of previous Local Authorities. New UA name
Old LA name
Bedford
Bedford
Central Bedfordshire
Mid Bedfordshire South Bedfordshire
Cheshire East
Congleton Crewe and Nantwich Macclesfield
Cheshire West and Chester
Chester Ellesmere Port & Neston Vale Royal
Cornwall
Caradon Carrick Kerrier North Cornwall Penwith Restormel
Numerator data caveats
Where the numerator is less than 5 no percentage is shown. (Data supplied by DCSF was rounded to the nearest 10, so this applies to values received that were rounded to 0.)
Denominator data caveats
Where the ethnic population is less than 30 the data has not been presented. (Data supplied by DCSF was rounded to the nearest 10, so this applies to values received that were rounded to 0, 10 or 20.)
Methods used to calculate indicator value
The indicator is expressed as the percentage of children (aged 5–15) eligible for free school meals by ethnic groups. A percentage is defined as the number of observed events divided by the total population of interest multiplied by 100. A percentage, p, is given by:
p=
O × 100 n
where: O is the number of observed events (i.e. number of children eligible for free school meals for each ethnic group); n is the size of the population of interest (i.e. the number of children in ethnic group aged 5-15)
Health Profiles 2009 The Indicator Guide
297 Section 7: charts and trend graphs
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Data for the Isles of Scilly and City of London have not been presented. The Isles of Scilly are included in the South West regional and national totals and the City of London figures have been included in the regional London and national totals.
Disclosure Control
Data was supplied by the department for Children, Schools and Families for each of the total groups (White, Mixed, Asian, Black and Chinese and any other ethnic group) rounded to the nearest ten. Thus, for all areas, the numbers presented are rounded to the nearest ten.
Confidence Intervals calculation method
Confidence intervals have been calculated using the following method for a confidence interval of a proportion as described by Newcombe RG If r is the observed number of subjects with some feature in a sample of size n then the estimated proportion who have the feature is p = r/n. The proportion who do not have the feature is q = 1-p. First, calculate the three quantities A = 2r + z2;
B = z z 2 + 4rq ;
and
C=2(n+z2),
where z is z1-α/2, from the standard Normal distribution. Then the confidence interval for the population proportion is given by (A-B)/C
to
(A+B)/C
This method has the considerable advantage that it can be used for any data. When there are no observed events, r and hence p are both zero, and the recommended confidence interval simplifies to 0 to z2/(n+z2). When r = n so that p = 1, the interval becomes n/(n+z2) to 1. Reference Newcombe, RG. Two-sided confidence intervals for the single proprotion: comparison of seven methods. Stat Med 1998;17:857-72.
Health Profiles 2009 The Indicator Guide
2
Contents Section 1: Introduction Section 2: Our communities 1. Deprivation indicator 2. Children in poverty indicator 3. Statutory homelessness indicator 4. GCSE achieved (5A*–C inc. Eng and Maths) indicator 5. Violent Crime indicator 6. Carbon emissions indicator
6 12 18 24 31 38
Section 3: Children’s and young people’s health 7. Smoking in pregnancy indicator 8. Breast feeding initiation indicator 9. Physically active children indicator 10. Obese children indicator 11. Children’s tooth decay (at age 5) indicator 12. Teenage pregnancy (under 18) indicator
44 52 61 68 77 86
Section 4: Adults’ health and lifestyle 13. Adults who smoke indicator 14. Binge drinking adults indicator 15. Health eating adults indicator 16. Physically active adults indicator 17. Obese adults indicator
94 108 124 141 148
Section 5: Disease and poor health 18. Over 65s ‘not in good health’ indicator 19. Incapacity benefits for mental illness indicator 20. Hospital stays for alcohol related harm indicator 21. Drug misuse indicator 22. People diagnosed with diabetes indicator 23. New cases of tuberculosis indicator 24. Hip fracture in over-65s indicator
165 172 179 185 190 198 203
Section 6: Life expectancy and causes of death 25. Excess winter deaths indicator 26. Life expectancy – male indicator 27. Life expectancy – female indicator 28. Infant deaths indicator 29. Deaths from smoking indicator 30. Early deaths: heart disease and stroke indicator 31. Early deaths: cancer indicator 32. Road injuries and deaths indicator
212 221 227 233 238 248 253 258
Section 7: Charts and trend graphs 33. Deprivation chart 34. Life expectancy by deprivation quintile chart 35. Trend 1: all age, all cause mortality 36. Trend 2: early deaths from heart disease and stroke 37. Trend 3: early deaths from cancer 38. Health inequalities: ethnicity chart
266 271 276 281 286 291
Health Profiles 2009 The Indicator Guide
3
Section 1: Introduction
Health Profiles 2009 The Indicator Guide
4
Introduction The Health Profiles are created using national indicators which are selected to give a snapshot of the health in local authority areas. The metadata for the 32 indicators that have been selected for Health Profiles 2009 are presented within five domains:
1. 2. 3. 4. 5.
Our communities Children’s and young people’s health Adults’ health and lifestyle Disease and poor health Life expectancy and causes of death
The metadata for the charts and trend graphs within the central pages of the Heath Profiles 2009 are also presented. The Indicator Guide provides detailed information about each of these indicators. This includes a summary table with 10 key pieces of basic information about the indicator such as: • What is being measured? • Why is it being measured? • How is the indicator defined? and a further three sections with more detailed information: • The indicator description • The indicator specification • The indicator technical methods
Health Profiles 2009 The Indicator Guide
5
Section 2: Our communities
Health Profiles 2009 The Indicator Guide
6 Section 2: Our communities
1. DEPRIVATION INDICATOR Basic Information 1. What is being measured?
Level of deprivation of a population in an area, as measured by the percentage of people in that area living in the most deprived fifth of areas in England.
2. Why is it being measured?
The differences in deprivation between areas are a major determinant of health inequality in the United Kingdom.
3. How is this indicator actually defined?
IMD 2007 is a model of measuring deprivation in an area. It is underpinned by separate dimensions of deprivation; these dimensions are weighted and an overall deprivation score is given.
4. Who does it measure?
All persons, all ages in the relevant population (otherwise known as the ‘at risk’ population). (This is the population estimate used in the construction of the IMD 2007 – see Table 2)
5. When does it measure it?
Based on various indicators, mostly using 2005 data.
6. Will It measure absolute numbers or proportions?
Proportion based on a composite indicator.
7. Where does the data actually come from?
DCLG website: http://www.communities.gov.uk/communities/ neighbourhoodrenewal/deprivation/deprivation07/
8. How accurate and complete will the data be?
All indicators included in the IMD 2007 are considered to be accurate and complete.
9. Are there any caveats/ warnings/problems?
The indicators are based on mainly 2005. It is based on an average score of an area and it can’t be assumed to represent all individuals in that area. Although very comprehensive, some aspects of deprivation are not to be included in the indices, due to data being incomplete or not available. This may have a larger effect in some areas than others.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
Pg 4 Health Summary – Indicator No. 1
Subject category/ domain(s)
Our Communities
Indicator name (* Indicator title in health profile)
Deprivation
PHO with lead responsibility
Yorkshire and Humber
Health Profiles 2009 The Indicator Guide
7 Section 2: Our communities
Date of PHO dataset creation
Feb 2009
Indicator definition
% of the relevant population in this area living in the 20% most deprived areas in England. (The relevant population is the population estimate used in the construction of the IMD 2007)
Geography
Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs
Timeliness
Indicator is not regularly updated. Published in December 2007 as an update for IMD 2004, the definition is not completely consistent due to changes in some of the indicators used.
Rationale: What this indicator purports to measure
Level of deprivation of a population in an area, by measuring the percentage of the population living in the most deprived quintile of neighbourhoods in England.
Rationale: Public Health Importance
The difference in deprivation between areas is a major determinant of health inequality in the United Kingdom. Many studies and analyses have demonstrated the association of increasingly poor health with increasing deprivation. For instance, all cause mortality, smoking prevalence, self-reported long standing illness are all correlated with deprivation. If deprivation inequalities decrease, health inequalities are likely to decrease also.
Rationale: Purpose behind the inclusion of the indicator
To monitor and help reduce health inequalities.
Rationale: Policy relevance
•A cheson D. Report of the Independent Inquiry into Inequalities in Health. London: TSO, 1998. •D epartment of Health. The NHS Plan. London: TSO, 2000. www.dh.gov.uk/assetRoot/04/05/57/83/04055783.pdf •D epartment of Health. Tackling Health Inequalities: A Programme for Action. Department of Health, 2003. www.dh.gov.uk/assetRoot/04/01/93/62/04019362.pdf •H M Treasury 2007 PBR CSR: Public service agreements: http://www.hmtreasury.gov.uk/ Department for Communities and Local Government The New Performance Framework for Local Authorities & Local Authority Partnerships: Single Set of National Indicators Department for Communities and Local Government, 2007 http://www.communities.gov.uk/corporate/
Health Profiles 2009 The Indicator Guide
8 Section 2: Our communities
Interpretation: What a high/low level of indicator value means
An indicator value worse than average (red circle in health summary chart) represents a statistically significant worse level of deprivation for that local authority when compared to the national value. An indicator value better than average (green circle in health summary chart) represents a statistically significant better level of deprivation for that local authority when compared to the national value. Note: more than a quarter of local authorities have no areas that fall within the most deprived 20% of areas in England and therefore have a value of 0% for this indicator. This means that there is no best quartile range shown on the spine chart for this indicator as all local authorities in the best quartile are at 0%.
Interpretation: Potential for error due to type of measurement method
The indicators are based on mainly 2005 data and this is therefore at least 4 years out of date.
Interpretation: Potential for error due to bias and confounding
It is based on an average score of an area and it can’t be assumed to represent all individuals in that area.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate.
Although very comprehensive some aspects of deprivation will not be included in the indices, for example if the data is incomplete or not collected. This may have a larger effect in some areas than others.
This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Health Profiles 2009 The Indicator Guide
9 Section 2: Our communities
Table 2 – Indicator Specification Indicator definition: Variable
Deprivation
Indicator definition: Statistic
Percentage of the relevant population in this area living in the 20% most deprived areas in England. (The relevant population is the population estimate used in the construction of the IMD 2007)
Indicator definition: Gender
Persons
Indicator definition: age group
All ages
Indicator definition: period
2005
Indicator definition: scale
Percentage of population
Geography: geographies available for this indicator from other providers
Lower Super Output Area http://www.communities.gov.uk/communities/ neighbourhoodrenewal/deprivation/deprivation07/
Dimensions of inequality: subgroup analyses of this dataset available from other providers
IMD 2007 individual domains are available at Lower Super Output Area level and this measure could be derived for them. http://www.communities. gov.uk/communities/neighbourhoodrenewal/deprivation/deprivation07/
Data extraction: Source
Department of Communities and Local Government website
Data extraction: source URL
http://www.communities.gov.uk/communities/neighbourhoodrenewal/ deprivation/deprivation07/
Data extraction: date
Data extracted from source as at: February 2009
Numerator: definition
Number of the relevant population living in the most deprived quintile in England based on the IMD 2007 score. (The relevant population is the population estimate used in the construction of the IMD 2007).
Numerator: source
Department of Communities and Local Government
Denominator: definition
Denominator data – ‘at risk’ mid-2005 population estimates (elsewhere referred to as the ‘relevant’ population). The figures have been adjusted from the ONS mid-year estimate to exclude the prison population in order to fit the definition of ‘at risk’. The figures have been subject to disclosure control.
Denominator: source
Office for National Statistics (ONS) and Department of Communities and Local Government (DCLG).
Data quality: Accuracy and completeness
Criteria for inclusion of indicators to IMD 2004 and IMD 2007 included: • Up-to-date • Statistically robust • Available for the whole of England at a small level in a consistent form
Health Profiles 2009 The Indicator Guide
10 Section 2: Our communities
Table 3 – Indicator Technical Methods Numerator: extraction
The score was derived by YHPHO from the Indices of Deprivation 2007
Numerator: aggregation/ allocation
The number of relevant population in the worst quintile of deprivation was aggregated from Lower Super Output Area to Local Authority level. (The relevant population is the population estimate used in the construction of the IMD 2007).
Numerator data caveats
Deprivation level is based on the average for the area, not all individuals in the area will be deprived.
Denominator data caveats
Census data and mid-year estimates are known to be deficient in their estimates of: • Non-white populations • Full-time students • Men aged 20–39 • People living in nursing homes etc • Rough sleepers • Inner-city populations • Households of multiple occupation • Migrants
Methods used to calculate indicator value
The relevant population in the worst quintile of deprivation was aggregated from Lower Super Output Area to Local Authority level, and divided by the total for the Local Authority. (The relevant population is the population estimate used in the construction of the IMD 2007)
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Excluded at district level but included at regional and national level.
Disclosure Control
Every effort has been made by the DCLG and ONS to ensure that data does not allow the disclosure of confidential information.
Health Profiles 2009 The Indicator Guide
11 Section 2: Our communities
Confidence Intervals calculation method
The 95% confidence intervals are calculated with the method described by Wilson and by Newcombe which is a good approximation of the exact method. First calculate the estimated proportions of subjects with (p) and without (q) some feature of interest from a sample of size n. proportion with feature of interest = p = r/n proportion without feature of interest = q = 1 - p where r is the observed number of subjects with the feature of interest. Second, calculate the three quantities A = 2r + z2;
B = z z 2 + 4rq ;
and
C=2(n+z2),
where z is the appropriate value, z1-α/2, from the standard Normal distribution. Then the confidence interval for the population proportion is given by (A-B)/C
to
(A+B)/C
This method has the considerable advantage that it can be used for any data. When there are no observed events, r and hence p are both zero, and the recommended confidence interval simplifies to 0 to z2/(n+z2). When r = n so that p = 1, the interval becomes n/(n+z2) to 1. Wilson EB. J Am Stat Assoc 1927, 22, 209-212 Newcombe, RG. Two-sided confidence intervals for the single proportion: comparison of seven methods. Stat Med 1998;17:857-72.
Health Profiles 2009 The Indicator Guide
12 Section 2: Our communities
2. CHILDREN IN POVERTY INDICATOR Basic Information 1. What is being measured?
Children in Poverty
2. Why is it being measured?
Growing up in poverty damages children’s health and well-being, adversely affecting their future health and life chances as adults.
3. How is this indicator actually defined?
Income Deprivation Affecting Children Index: part of Indices of Deprivation 2007 – Income deprivation domain. Measuring the proportion of children under 16 years living in families receiving means-tested benefits.
4. Who does it measure?
Children under 16 years in the relevant population (otherwise known as the ‘at risk’ population). (This is the population estimate used in the construction of the IMD 2007 – see Table 2).
5. When does it measure it?
Based on 2005 data.
6. Will It measure absolute numbers or proportions?
Proportions
7. Where does the data actually come from?
Department of Communities and Local Government and ONS http:// www.communities.gov.uk/communities/neighbourhoodrenewal/ deprivation/deprivation07/
8. How accurate and complete will the data be?
The data are of very high quality as they are drawn from a 100% scan of administrative records and as a result are not subject to any sampling error. A small number of claimants whose details are held clerically are excluded. Comprehensive validation checks are undertaken.
9. Are there any caveats/ warnings/problems?
Rounding error is liable to occur in the process of aggregating from LSOAs to Local Authorities. Local authority denominator derived by aggregating LSOA-level relevant child population estimates, these are based on ONS experimental statistics.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
Pg 4 Health Summary – Indicator No. 2
Subject category/ domain(s)
Our Communities
Health Profiles 2009 The Indicator Guide
13 Section 2: Our communities
Indicator name (*Indicator title in health profile)
Income Deprivation Affecting Children Index (*Children in Poverty)
PHO with lead responsibility
Yorkshire and Humber
Date of PHO dataset creation
Feb 2009
Indicator definition
Prevalence of children living in families receiving means-tested benefits, under-16 years (2005). (part of Indices of Deprivation 2007 – Income deprivation domain)
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs
Timeliness
Indicator is not regularly updated. Published in December 2007 as an update for ID 2004, the definition is not completely consistent due to changes in some of the indicators used.
Rationale: What this indicator purports to measure
Child poverty operationalised as children living in families reliant on meanstested benefits and those in receipt of Working Tax Credit and/or Child Tax Credit with an equivalised income below 60 percent of the national median before housing costs. This is used as a proxy for the widely-used HBAI (Households Below Average Income) indicator, which is not available at local authority level.
Rationale: Public Health Importance
Growing up in poverty damages children’s health and well-being, adversely affecting their future health and life chances as adults. Ensuring a good environment in childhood, especially early childhood, is important. A considerable body of evidence links adverse childhood circumstances to poor child health outcomes and future adult ill health. Adverse outcomes include higher rates of: fatal accidents, poor dental health, child mortality, low educational attainment, low birth weight, childhood obesity, school exclusions, infant mortality, teenage pregnancy some infections, substance misuse, mental ill health. By international standards the comparative picture of child poverty in the UK has been poor. International variation in child poverty levels shows that child poverty is not inevitable. In other countries experiencing similar demographic changes and economic pressures to the UK, children have been protected from escalating child poverty by social policy favouring progressive taxation and higher spending on social protection for children. Eradicating child poverty is now a national policy target.
Rationale: Purpose behind the inclusion of the indicator
To monitor and help reduce health inequalities.
Rationale: Policy relevance
Opportunity for All Every Child Matters Children’s National Service Framework
Health Profiles 2009 The Indicator Guide
14 Section 2: Our communities
Interpretation: What a high/low level of indicator value means
An indicator value worse than average (red circle in health summary chart) represents a statistically significant worse rate of child poverty for that local authority when compared to the national value.
Interpretation: Potential for error due to type of measurement method
Benefits claims are an imperfect measure of income deprivation because some eligible families do not claim their entitlement. Others may be living in income deprivation but may not be entitled to claim.
Interpretation: Potential for error due to bias and confounding
Some groups are known to have a low propensity to claim the benefits to which they are entitled, and may be over-represented in certain areas.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself.
An indicator value better than average (green circle in health summary chart) represents a statistically significant better rate of child poverty for that local authority when compared to the national value.
The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Health Profiles 2009 The Indicator Guide
15 Section 2: Our communities
Table 2 – Indicator Specification Indicator definition: Variable
Children living in families receiving means-tested benefits
Indicator definition: Statistic
Percentage of relevant resident population. (This is the population estimate used in the construction of the IMD 2007)
Indicator definition: Gender
Persons
Indicator definition: age group
Under 16 years
Indicator definition: period
2005
Indicator definition: scale
Percentage of child population
Geography: geographies available for this indicator from other providers
Lower Super Output Area http://www.communities.gov.uk/communities/ neighbourhoodrenewal/ deprivation/deprivation07/
Dimensions of inequality: subgroup analyses of this dataset available from other providers
None.
Data extraction: Source
Department of Communities and Local Government and ONS
Data extraction: source URL
http://www.communities.gov.uk/communities/neighbourhoodrenewal/ deprivation/deprivation07/
Data extraction: date
Data extracted from source as at: February 2009.
Numerator: definition
The published data consists of the percentage of relevant children in the LSOA who are living in poverty. The number of such children was calculated by multiplying this proportion by the total number of children in the LSOA. This count was then aggregated from LSOA to LA level. Those counted as living in poverty are relevant children under 16 living in families receiving Income Support or Jobseekers Allowance (Income Based), or in families receiving Working Families Tax Credit/Child Tax Credit whose equivalised income is below 60% of median before housing costs. (The relevant population is the population estimate used in the construction if the IMD 2007)
Numerator: source
The proportion of relevant children living in poverty by LSOA can be found at http://www.communities.gov.uk/communities/neighbourhoodrenewal/ deprivation/deprivation07/ (The relevant population is the population estimate used in the construction if the IMD 2007)
Health Profiles 2009 The Indicator Guide
16 Section 2: Our communities
Denominator: definition
Denominator data – ‘at risk’ mid-2005 population estimates (age under 16) (elsewhere referred to as the ‘relevant’ population). The figures have been adjusted from the ONS mid-year estimate to exclude the prison population in order to fit the definition of ‘at risk’. The figures have been subject to disclosure control.
Denominator: source
http://www.communities.gov.uk/communities/neighbourhoodrenewal/ deprivation/deprivation07/
Data quality: Accuracy and completeness
The data are of very high quality as they are drawn from a 100% scan of administrative records and as a result are not subject to any sampling error. A small number of claimants whose details are held clerically are excluded. Comprehensive validation checks are undertaken by the DWP Information Centre to assess the accuracy, reliability, consistency and completeness of the data. Additional checks were undertaken by the University of Oxford to verify the quality of the data.
Table 3 – Indicator Technical Methods Numerator: extraction
Download from DCLG website and aggregation from LSOA to LA level.
Numerator: aggregation / allocation
The set of benefits is such that only one can be claimed at a time.
Numerator data caveats
Rounding error is liable to occur in the process of aggregating from LSOAs to Local Authorities.
Denominator data caveats
Local authority denominator derived by aggregating LSOA-level child population estimates; these are based on ONS experimental statistics
Methods used to calculate indicator value
The numerator and denominator were derived as explained above and a population weighted average of LSOAs was calculated for each LA. The quotient was expressed as a percentage.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Excluded at district level but included at regional and national level.
Disclosure Control
Every effort has been made by the DWP to ensure that data do not allow the disclosure of confidential information.
Health Profiles 2009 The Indicator Guide
17 Section 2: Our communities
Confidence Intervals calculation method
The 95% confidence intervals are calculated with the method described by Wilson and by Newcombe which is a good approximation of the exact method. First calculate the estimated proportions of subjects with (p) and without (q) some feature of interest from a sample of size n. proportion with feature of interest = p = r/n proportion without feature of interest = q = 1 - p where r is the observed number of subjects with the feature of interest. Second, calculate the three quantities A = 2r + z2;
B = z z 2 + 4rq ;
and
C=2(n+z2),
where z is the appropriate value, z1-α/2, from the standard Normal distribution. Then the confidence interval for the population proportion is given by (A-B)/C
to
(A+B)/C
This method has the considerable advantage that it can be used for any data. When there are no observed events, r and hence p are both zero, and the recommended confidence interval simplifies to 0 to z2/(n+z2). When r = n so that p = 1, the interval becomes n/(n+z2) to 1. Wilson EB. J Am Stat Assoc 1927, 22, 209-212 Newcombe, RG. Two-sided confidence intervals for the single proportion: comparison of seven methods. Stat Med 1998;17:857-72.
Health Profiles 2009 The Indicator Guide
18 Section 2: Our communities
3.1 HOMELESSNESS INDICATOR Basic Information 1. What is being measured?
Estimates of homelessness amongst the most needy and vulnerable groups.
2. Why is it being measured?
Homelessness is associated with severe poverty and is a social determinant of health. Statutorily homeless households contain some of the most vulnerable members of society. Tackling homelessness requires joint working across health and social care services.
3. How is this indicator actually defined?
Statutory homeless households, crude rate per 1000 estimated households, all ages, all persons, 2007 to 2008.
4. Who does it measure?
All persons, all ages.
5. When does it measure it?
Reported quarterly and updated every year.
6. Will It measure absolute numbers or proportions?
Proportions: number of statutorily homeless households per thousand estimated total households.
7. Where does the data actually come from?
Collection and collation from the Housing Strategy Statistical Appendix via the Department for Communities and Local Government.
8. How accurate and complete will the data be?
A built-in-validation system allows each LA to check data accuracy. Data is also validated manually by the DCLG. HSSA and P1E guidance notes provide universal definitions. Missing data is considered to be zeros.
9. Are there any caveats/warnings/ problems?
Data only count ‘statutory homeless’ so does not include the intentionally homeless or those who are not in priority need categories. The total number of households is an estimated figure.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
Pg 4 Health Summary – Indicator 3
Subject category/ domain(s)
Our Communities
Indicator name (*Indicator title in health profile)
Homelessness (*Statutory homelessness)
Health Profiles 2009 The Indicator Guide
19 Section 2: Our communities
PHO with lead responsibility
NEPHO
Date of PHO dataset creation
February 2009
Indicator definition
Statutory homeless households, crude rate per 1,000 estimated total households, all ages, 2007 to 2008, persons
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs
Timeliness
The Housing Strategy Statistical Appendix (HSSA) is updated annually, however it should be noted that quarterly statistics are published on the Department for Communities and Local Government website: www.communities.gov.uk
Rationale: What this indicator purports to measure
Estimates of homelessness amongst the most needy and vulnerable groups in society
Rationale: Public Health Importance
Homelessness is associated with severe poverty and is a social determinant of health. Homelessness is associated with adverse health, education and social outcomes, particularly for children. To be deemed statutorily homeless a household must have become unintentionally homeless and must be considered to be in priority need. As such, statutorily homeless households contain some of the most vulnerable and needy members of our communities. The statutory homeless statistics suggest that 62% of officially accepted homeless households include dependent children or an expectant mother. Preventing and tackling homelessness requires sustained and joined-up interventions by central and local government, health and social care and the voluntary sector.
Rationale: Purpose behind the inclusion of the indicator
To reduce the level of homelessness, particularly amongst the most vulnerable and needy groups in society.
Rationale: Policy relevance
The Department for Communities and Local Government (DCLG) has published a strategy document, ‘Sustainable Communities: Settled Homes: Changing Lives’ which sets out the Government’s plans on reducing homelessness with the aim of halving the number of homeless households in temporary accommodation by 2010.
Interpretation: What a high/low level of indicator value means
An indicator value worse than average (red circle in health summary chart) represents a statistically significant worse level of statutory homelessness for that local authority when compared to the national value. An indicator value better than average (green circle in health summary chart) represents a statistically significant better level of statutory homelessness for that local authority when compared to the national value. However no amount of homelessness is acceptable, and therefore a low indicator value should not mean that public health action is not needed.
Health Profiles 2009 The Indicator Guide
20 Section 2: Our communities
Interpretation: Potential for error due to type of measurement method
The statistic necessarily only measures the incidence of official homelessness. The number of households who are homeless but do not apply to the local authority and are therefore not considered under Housing Act legislation is not known. Reasons may include a lack of knowledge of the legislation, a correct or misplaced belief that they will not qualify for assistance, and/or a desire not to rely on state support. This statistic does not include households that have become unintentionally homeless but are not considered to be in priority need or households that have become intentionally homeless. Rough sleepers are also not included. Therefore, the measure is an underestimate of the extent of homelessness, both of those populations who would qualify for assistance and for the larger number of people who fall outside of the legislation. See: Poverty: the outcomes for children. 2001. ESRC. Edited by: Jonathan Bradshaw.
Interpretation: Potential for error due to bias and confounding
Potential confounding factors associated with the homelessness statistic include: housing affordability, housing capacity, variation in local authority methods of collection and collation of housing and homelessness statistics, local variation in demand for housing.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with a white symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or amber symbol depending on whether it is worse or better than the national value respectively.
Health Profiles 2009 The Indicator Guide
21 Section 2: Our communities
Table 2 – Indicator Specification Indicator definition: Variable
Statutory homeless households
Indicator definition: Statistic
Crude rate per 1,000 total estimated households
Indicator definition: Gender
Persons
Indicator definition: age group
All ages
Indicator definition: period
1 April 2007 to 31 March 2008
Indicator definition: scale
Per 1,000 total estimated households
Geography: geographies available for this indicator from other providers
No other geographies available from other providers.
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Data relating to the ethnicity of households is collected as part of the HSSA.
Data extraction: Source
Department for Communities and Local Government
Data extraction: source URL
http://www.communities.gov.uk/documents/housing/xls/hssa08sectione.xls http://www.communities.gov.uk/documents/housing/xls/hssa08sectioneh.xls
Data extraction: date
Data extracted from source as at: 3 February 2009
Numerator: definition
Count of households (2007/2008) who are eligible, unintentionally homeless and in priority need, for which the local authority accepts responsibility for securing accommodation under part VII of the Housing Act 1996 or part III of the Housing Act 1985.
Numerator: source
Department for Community and Local Government.
Denominator: definition
Total estimated households at 30th June 2008.
Denominator: source
2007/2008 Housing Strategy Statistical Appendix (HSSA) - Section F (Household Numbers) DCLG.
Health Profiles 2009 The Indicator Guide
22 Section 2: Our communities
Data quality: Accuracy and completeness
Each LA checks the data prior to sending it to the Housing Statistics Department of the DCLG. A built-in-validation system allows each LA to check the accuracy of the data (see detailed on the ‘Validation’ section). Once the data has been submitted to the DCLG, validation checks are carried out manually. The HSSA and P1E guidance notes help to provide LAs with universal definitions. Again, for financial management purposes, each LA has to maintain accurate and up-to-date information on homelessness. At LA level, 6% of total entries are missing. This is due to incomplete returns and methods used to safeguard the confidentiality of the data (see ‘Disclosure Control’ Section). There has been no imputation at LA, Regional or National level. Figures registered as missing values at LA level have not been estimated and have therefore been considered to be zeros when aggregating to a higher level.
Table 3 – Indicator Technical Methods Numerator: extraction
Simple download.
Numerator: aggregation / allocation
Counts had already been allocated to local authorities.
Numerator data caveats
To be classified as statutorily homeless, the following must be satisfied: • They are homeless, defined as those without any right to access secure accommodation for that night i.e. they are not legal tenants of any property, nor own any property. Or they can also be classed as ‘potentially homeless’ if they are about to lose their dwelling, be evicted, within 28 days. • They must have a local connection (lived or worked in the area, family in the area, have a care responsibility or need care from relatives in the area). • They are in priority need i.e. had dependent children in them (aged under 16 years) or are an older person household, or vulnerable. • The homeless household must not be intentionally homeless i.e. losing their previous accommodation through their own action such as not paying rent or a mortgage. By contrast the ‘non-statutory’ homeless are, those to whom no duty is owed either because they are deemed intentionally homeless, or are not in a priority need categories. These include the ‘single homeless’, many of whom are now young people of both sexes and, in the larger cities, of different ethnic groups, as well as some older white men.
Denominator data caveats
This is an estimated total number of households at the 30th June 2008.
Health Profiles 2009 The Indicator Guide
23 Section 2: Our communities
Methods used to calculate indicator value
Number of households deemed to be statutorily homeless during the period April 2007 to March 2008, divided by the estimated total number of households as at 30th June 2008, multiplied by 1000.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Isles of Scilly and City of London have been included in regional and England numerators and denominators. Isles of Scilly have been included in the numerator and denominator for the County of Cornwall.
Disclosure Control
Data have been suppressed in this dataset to protect both the confidentiality of individual information and the potential statistical instability caused by low counts. As a consequence, sums of counts may not equal related totals.
Confidence Intervals calculation method
The 95% confidence intervals are calculated with the method described by Wilson and by Newcombe which is a good approximation of the exact method. First calculate the estimated proportions of subjects with (p) and without (q) some feature of interest from a sample of size n. proportion with feature of interest = p = r/n proportion without feature of interest = q = 1 - p where r is the observed number of subjects with the feature of interest. Second, calculate the three quantities A = 2r + z2;
B = z z 2 + 4rq ;
and
C=2(n+z2),
where z is the appropriate value, z1-α/2, from the standard Normal distribution. Then the confidence interval for the population proportion is given by (A-B)/C
to
(A+B)/C
This method has the considerable advantage that it can be used for any data. When there are no observed events, r and hence p are both zero, and the recommended confidence interval simplifies to 0 to z2/(n+z2). When r = n so that p = 1, the interval becomes n/(n+z2) to 1. Wilson EB. J Am Stat Assoc 1927, 22, 209-212 Newcombe, RG. Two-sided confidence intervals for the single proportion: comparison of seven methods. Stat Med 1998;17:857-72.
Health Profiles 2009 The Indicator Guide
24 Section 2: Our communities
4. GCSE ACHIEVEMENT INDICATOR Basic Information 1. What is being measured?
GCSE achievement
2. Why is it being measured?
Educational attainment is influenced by both the quality of education children receive and their family’s socio-economic circumstances. Educational qualifications are a determinant of an individual’s labour market position, which in turn influences income, housing and other material resources. These are related to health and health inequalities.
3. How is this indicator actually defined?
Pupils achieving 5 or more GCSEs at grades A*-C (including English and Maths) or equivalent, percentage of pupils at end of Key Stage 4 in schools maintained by the Local Education Authority, at the end of the academic year 2007-08, persons.
4. Who does it measure?
Pupils at the end of Key Stage 4 in maintained schools.
5. When does it measure it?
This indicator is published annually.
6. Will it measure absolute numbers or proportions?
Proportions, percentage of pupils.
7. Where does the data actually come from?
Collection and collation by the Department for Children, Schools and Families (DCSF).
8. How accurate and complete will the data be?
This indicator only contains data for LEA maintained schools, therefore it excludes pupils educated in private schools.
9. Are there any caveats/warnings/ problems?
Data for Regions and Counties are aggregated on the basis of the schools’ administrating Local Education Authorities. Data for County Districts, Metropolitan County Districts (MCD), Unitary Authorities (UA) and London Boroughs (LB) are aggregated on the basis of Neighbourhood Renewal areas based on the geographic location of the school and not the location of the pupil’s residence.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Health Profiles 2009 The Indicator Guide
25 Section 2: Our communities
TABLE 1 – INDICATOR DESCRIPTION Information component
Pg 4 Health summary – Indicator No 4
Subject category/ domain(s)
Our Communities
Indicator name (*Indicator title in health profile)
GCSE and equivalent results, including maths and English, of pupils at the end of Key Stage 4 (*GCSE achieved 5A*–C inc. Eng & Maths).
PHO with lead responsibility
LHO
Date of PHO dataset creation
Feb 2009
Indicator definition
Pupils achieving 5 or more GCSEs, including maths and English, at grades A*–C or equivalent, percentage of pupils at end of Key Stage 4 in schools maintained by the Local Education Authority, at the end of academic year 2007–2008, persons.
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs, and Strategic Health Authorities for the South East.
Timeliness
Annual. Next data is expected at the end of January 2010. The data is suitable for time trend analysis.
Rationale: What this indicator purports to measure
This indicator measures the level of GCSE achievement in the area.
Rationale: Public Health Importance
Educational attainment is influenced by both the quality of education children receive and their family socio-economic circumstances. Educational qualifications are a determinant of an individual’s labour market position, which in turn influences income, housing and other material resources. These are related to health and health inequalities.
Rationale: Purpose behind the inclusion of the indicator
This indicator relates to the Department for Children, Schools and Families (DCSF) Public Service Agreement (PSA) target 10 which is to raise standards in schools and colleges so that: By 2008, 60% of those aged 16 achieve the equivalent of 5 GCSEs at grades A* to C; and in all schools at least 20% of pupils to achieve this standard by 2004, rising to 25% by 2006 and 30% by 2008. This indicator now uses the same definition as No 75 in the National Indicator Set.
Rationale: Policy relevance
This indicator is the DCSF PSA 10 target. Others include: Tackling health inequalities: A programme for action – Education. Local basket of inequalities indicators – Indicator 3.4. Best Value Performance Indicator 38 Opportunity for all – Children and young people – Indicator 6
Health Profiles 2009 The Indicator Guide
26 Section 2: Our communities
Interpretation: What a high/low level of indicator value means
A high indicator value (green circle in health summary chart) represents a statistically significant higher level of educational attainment for that local authority when compared to the England national value. An increase in this indicator indicates that education achievement has increased relative to the number of pupils (i.e. more children achieving at least 5 GCSEs or equivalent not same children achieving more GCSEs). A low indicator value (red circle in health summary chart) represents a statistically significant lower level of educational attainment for that local authority when compared to the national value. Where educational attainment is not satisfactory, this should prompt investigation as to the causes which may be related to quality of educational services; the socio-economic circumstances of families in the area or other factors known to influence achievement.
Interpretation: Potential for error due to type of measurement method
There are issues of equivalence of GCSE and other qualifications such as NVQs. For example a pupil with English, Mathematics, Double Science and German with top grades, currently counts equally with one who has a bare pass in an Intermediate GNVQ course in IT (being equivalent to 4 passes) and a GCSE grade C in any other subject. Some schools may show improvement in educational attainment through changing the type of qualifications being taken. See: Nuffield Review of 14-19 Education and Training Working Paper 4 (based on Discussion Paper given at Working Day I, 17 Dec 2003) CONTINUITY AND DISCONTINUITY IN THE ‘14-19 CURRICULUM’ Jeremy Higham, Post-14 Research Group, School of Education, University of Leeds. Available from: http://www.nuffield14-19review.org.uk/files/documents12-1.pdf
Interpretation: Potential for error due to bias and confounding
The statistic measures GCSE attainment in LEA maintained schools and therefore does not include attainment for children who are educated in private schools, approximately 7% of children in England. The statistic also does not take into account where pupils live. For some inner-London boroughs, this can mean up to 40% of resident pupils being educated privately or travelling to state schools in other boroughs. Estimating the likely impact of private education and pupils travelling to schools outside their resident local authorities is problematic. The annual schools’ census collects information on pupils educated in independent schools, regardless of where the pupils live. However, information is not available on the area of residence of pupils in independent schools and which local education authority would be responsible for them. Areas have different densities of independent schools. It should also be noted that the figure relates to the location of the school within a local authority area rather than the home residence of the pupils who attend the school.
Health Profiles 2009 The Indicator Guide
27 Section 2: Our communities
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider the confidence interval, the greater the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with a white symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or amber symbol depending on whether it is worse or better than the national value respectively.
TABLE 2 – INDICATOR SPECIFICATION Indicator definition: Variable
Pupils achieving 5 or more GCSEs, including maths and English, at grades A*–C or equivalent
Indicator definition: Statistic
Percentage of pupils at end of Key Stage 4 in schools maintained by the Local Education Authority,
Indicator definition: Gender
Persons
Indicator definition: age group
End of key stage 4 (The term is defined in the Education Act 2002 as “the period beginning at the same time as the school year in which the majority of pupils in his class attain the age of fifteen and ending at the same time as the school year in which the majority of pupils in his class cease to be of compulsory school age”)
Health Profiles 2009 The Indicator Guide
28 Section 2: Our communities
Indicator definition: period
At the end of the academic year 2007–2008
Indicator definition: scale Geography: geographies available for this indicator from other providers
Neighbourhood renewal area; Local Education Authority (LEA) http://www.dcsf.gov.uk/performancetables/
Dimensions of inequality: subgroup analyses of this dataset available from other providers
GCSE achievement: Available by school type and admission’s basis: http://www.dcsf.gov.uk/rsgateway/DB/SFR/s000826/SFR02_2009_SFRTables.xls Also available by sex (boys and girls) and for areas classified by level of deprivation, and for urban/rural areas: http://www.dcsf.gov.uk/rsgateway/DB/SFR/s000826/SFR02_2009_AdditionalTables_ Amended030309-2.xls Also available by free school meal status: http://www.dcsf.gov.uk/rsgateway/DB/SFR/s000826/GCSE_performance_by_FSM_ band.xls Also available by ethnic group: http://www.dcsf.gov.uk/rsgateway/DB/SFR/s000822/SFR322008-KS4FinalNITables_ Amended300109.xls
Data extraction: Source
Department for Children, Schools and Families (DCSF) tables 18 and 20 for 2007/08
Data extraction: source URL
http://www.dcsf.gov.uk/rsgateway/DB/SFR/s000826/SFR02_2009_AdditionalTables_ Amended160109.xls
Data extraction: date
Data extracted from source as at Feb 2009
Numerator: definition
Number of pupils at the end of key stage 4 at the end of the academic year 2007/08 achieving 5 or more GCSEs, including maths and English, at grades A*–C or equivalent in schools maintained by the Local Education Authority. The numerator will include achievements by these pupils in previous academic years. Numerators are not published by DCSF and therefore this number is estimated based on published percentage achievement and the total number pupils at the end of Key Stage 4.
Health Profiles 2009 The Indicator Guide
29 Section 2: Our communities
Numerator: source
DCSF tables 18 and 20 for 2006/07 Performance data at:
Denominator: definition
Number of pupils at the end of Key Stage 4 at the end of the academic year 2007/08 in schools maintained by the Local Education Authority.
Denominator: source
DCSF tables 18 and 20 for 2007/08 Performance data at: http://www.dcsf.gov.uk/rsgateway/DB/SFR/s000826/SFR02_2009_AdditionalTables_ Amended160109.xls
http://www.dcsf.gov.uk/rsgateway/DB/SFR/s000826/SFR02_2009_AdditionalTables_ Amended160109.xls Data for England, Regions and Counties are based on Table 18. Data for County Districts, Metropolitan County Districts, Unitary Authorities and London Boroughs are based on Table 20, with the exception of new UAs formed in April 2009 (County Durham, Northumberland, Shropshire, Wiltshire and Cornwall) which are taken from Table 18.
Data for England, Regions and Counties are based on Table 18. Data for County Districts, Metropolitan County Districts, Unitary Authorities and London Boroughs are based on Table 20. Data quality: Accuracy and completeness
County, Government Office Region and England Total (Maintained sector) figures are adjusted for pupils recently arrived from overseas. This indicator only contains data for LEA maintained schools. Data for Regions and Counties are aggregated on the basis of the schools’ administrating Local Education Authorities. Data for County Districts, Metropolitan County Districts (MCD), Unitary Authorities (UA) and London Boroughs (LB) are aggregated on the basis of Neighbourhood Renewal areas (geographic location of the school).
Table 3 – Indicator Technical Methods Numerator: extraction
Simple download from DCSF website
Numerator: aggregation/ allocation
Data for Regions and Counties are aggregated on the basis of the schools’ administrating Local Education Authorities. Data for County Districts, Metropolitan County Districts (MCD), Unitary Authorities (UA) and London Boroughs (LB) are aggregated on the basis of Neighbourhood Renewal areas (geographic location of the school). Data for some new Unitary Authorities, introduced in April 2009, did not correspond with existing geographies (Cheshire East, Cheshire West and Cheshire, Central Bedfordshire, and Cornwall). Published results are also not available for the two Strategic Health Authorities in the South East: South Central and South East Coast. For these areas data were derived from Table 20, where denominator data could be aggregated to the new geographies. Numerator data can be approximated from published percentages for local authorities and be aggregated to the new geographies to allow the calculation of percentages. Results for the other new UAs were taken from published data for local authorities (Bedford) or counties (County Durham, Northumberland, Shropshire and Wiltshire). None of the data are based on the postcode of the pupils.
Health Profiles 2009 The Indicator Guide
30 Section 2: Our communities
Numerator data caveats
The numerator values for local authorities and Regions will not add up to the total for England as the numerator is estimated. None of the data is based on the post code of the pupils. Excludes pupils educated in private schools
Denominator data caveats
None of the data is based on the post code of the pupils. Excludes pupils educated in private schools
Methods used to calculate indicator value
The number of pupils achieving 5 GCSEs, including maths and English, grade A*–C, divided by the number of pupils at the end of Key Stage 4, multiplied by 100.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Data for Cornwall excludes the Isles of Scilly. Therefore pupil numbers for County Districts will not add up to the data for Counties as a whole, and pupil numbers for MCD, UA, LB and Counties combined do not add up to the total for England. The Isles of Scilly is included in the England and South West Region totals.
Disclosure Control
Usage and dissemination of data is subject to Crown copyright. Data are suppressed if there are fewer than 3 pupils in that particular authority; however, there are no numbers less than 3 in this dataset.
Confidence Intervals calculation method
Confidence intervals have been calculated using the ‘recommended’ formula for a confidence interval of a proportion as described by Newcombe RG and Altman DG. Proportions and their differences. In Altman, DG et al (eds). Statistics with Confidence. 2nd ed. BMJ Books, 2000.
Health Profiles 2009 The Indicator Guide
31 Section 2: Our communities
5. VIOLENT CRIME INDICATOR Basic Information 1. What is being measured?
Recorded crimes of violence against the person.
2. Why is it being measured?
To help target policing and crime prevention resources and to reduce the incidence of violent crime.
3. How is this indicator actually defined?
Recorded violence against the person offences, crude rate per 1,000 population, all ages, 2007/08, persons.
4. Who does it measure?
All persons, all ages.
5. When does it measure it?
Continually reported and data is published annually.
6. Will It measure absolute numbers or proportions?
Proportion: Crude rate per 1,000 population.
7. Where does the data actually come from?
Home Office.
8. How accurate and complete will the data be?
Coverage is complete.
9. Are there any caveats/warnings/ problems?
This indicator only includes violent offences which are reported to the police. It is susceptible to changes in police crime reporting procedures.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
Pg 4 Health Summary – Indicator No 5
Subject category/ domain(s)
Our communities
Indicator name (*Indicator title in health profile)
Violent crime
PHO with lead responsibility
South West Public Health Observatory
Date of PHO dataset creation
November 2008 (revised February 2009 for Local Authorities as at April 2009)
Indicator definition
Recorded violence against the person offences, crude rate per 1,000 population, all ages, 2007/08, persons
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs
Health Profiles 2009 The Indicator Guide
32 Section 2: Our communities
Timeliness
The indicator presented in Health Profiles is routinely updated annually.
Rationale: What this indicator purports to measure
Level of reported violence against the person offences in an area
Rationale: Public Health Importance
The links between crime and health are complex (as outlined in the London Health Commission report http://www.londonshealth.gov.uk/pdf/lhs/crime.pdf and in the NACRO report “Better Health, Lower Crime” http://www.nacro. org.uk/data/resources/nacro-2004120264.pdf) but it is likely that crime is a determinant and a consequence of health. Research undertaken by the Home Office and a number of other organisations suggests that there is a relationship between violent crime and alcohol http:// www.fphm.org.uk/resources/AtoZ/bs_alcohol_violence.pdf Violent crime may result in temporary or permanent disability and in some cases death. Some victims of crime may suffer psychological distress and subsequent mental health problems. Crime and fear of crime can also alter people’s lifestyles and impact on their physical and psychological health Collectively, these consequences represent a burden to the healthcare services. This indicator specifically measures recorded ‘violence against the person’, the largest component of total ‘violent crime’ (which also includes robbery and sexual offences).
Rationale: Purpose behind the inclusion of the indicator
To help target policing and crime prevention resources and to reduce the incidence of violent crime.
Rationale: Policy relevance
Targets included in the Treasury’s Spending Review provide the basis for policy priorities. In 2002 the spending review included targets to reduce crime and the fear of crime and to reduce the gap between those Crime and Disorder Reduction Partnerships experiencing the highest crime and other areas.
Interpretation: What a high/low level of indicator value means
A high indicator value (red circle in health summary chart) represents a statistically significant higher rate of reported violent crime when compared to the England average value. A low indicator value (green circle in health summary chart) represents a statistically significant lower rate of reported violent crime when compared to the England value. However, it should be noted that high values reflect higher numbers of crimes recorded by the police. This may be a result of higher underlying incidence of violent offences, a greater proportion of incidents being reported to the police, or as a result of policing practice. The converse is true for low values of the indicator.
Interpretation: Potential for error due to type of measurement method
This indicator omits violent offences which are not reported to the police. It is susceptible to changes in police crime reporting procedures.
Health Profiles 2009 The Indicator Guide
33 Section 2: Our communities
Interpretation: Potential for error due to bias and confounding
The level and intensity of police service provision will affect rates of recorded violent crime offences, i.e. the higher the level of policing, the more likely it is that recorded crime figures will be elevated. Caution needs to be taken when considering crime rates in areas with low resident populations in which violent crimes are carried out by non-residents.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Table 2 – Indicator Specification Indicator definition: Variable
Recorded violence against the person offences
Indicator definition: Statistic
Crude rate
Indicator definition: Gender
Persons
Indicator definition: age group
All ages
Indicator definition: period
Numerator financial years 2007/08. Denominator mid-2006
Health Profiles 2009 The Indicator Guide
34 Section 2: Our communities
Indicator definition: scale
Per 1,000 population
Geography: geographies available for this indicator from other providers
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs Annual rates and numerators available for these geographies from www.homeoffice.gov.uk
Dimensions of inequality: subgroup analyses of this dataset available from other providers
None available.
Data extraction: Source
Home Office
Data extraction: source URL
http://www.homeoffice.gov.uk/rds/crimeew0708.html
Data extraction: date
Data extracted from source as at: 04/11/2008.
Numerator: definition
Annual count of recorded ‘violence against the person’ offences in the respective financial years. Violence against the person comprises the following offences: Murder, attempted murder, threat or conspiracy to murder, manslaughter, infanticide, child destruction, causing death by dangerous or careless driving, causing or allowing death of a child or vulnerable person, causing death by driving (unlicensed, disqualified or uninsured), wounding or other acts endangering life, endangering a railway passenger, endangering life at sea, other wounding, possession of weapons, harassment, racially or religiously aggravated other wounding, racially or religiously aggravated harassment, cruelty to and neglect of children, abandoning child under two years, child abduction, procuring illegal abortion, causing death by aggravated vehicle taking, assault on a constable, common assault, racially or religiously aggravated common assault. Please see the following for more details: http://www.homeoffice.gov.uk/rds/pdfs07/countviolence07.pdf Please note that for the 2007/08 data, the concealment of birth offence is no longer included in the Violence Against The Person group, as it has been moved to Miscellaneous Other Offences. This is a change for the 2007/08 data from the previous year. Numerator data have been used exactly as published in the Home Office supplied data. Where new Unitary Authorities that have been created as part of the April 2009 boundary changes are exactly co-terminous with pre-existing counties (Cornwall, Durham, Northumberland, Shropshire and Wiltshire) , the numerator data have been drawn from the relevant county figures. Where the new Unitary Authorities represent only a part of the existing counties, numerator data have been aggregated from the relevant published figures for the constituent districts. Please note that a similar approach has been applied to the denominator data (see Denominator: Definition for more information).
Numerator: source
Home Office
Health Profiles 2009 The Indicator Guide
35 Section 2: Our communities
Denominator: definition
2001 Census based mid-year population estimate for the mid-point year (2006). Data are based on the latest revisions of ONS mid-year population estimates for the respective year, supplied as part of the Home Office’s data. – Please note that the Home Office published statistics use mid-2006 population estimates as their denominator, rounded to the nearest 1,000. The data that have been entered into the Health Profiles tool, and the calculated confidence intervals, use the Home Office supplied data as published, except for newly created post-April 2009 Unitary Authorities. For new Unitary Authorities which are exactly coterminous with pre-existing county boundaries (Cornwall, Durham, Northumberland, Shropshire and Wiltshire), the relevant county population has been used in these calculations. For new Unitary Authorities which comprise only a part of the pre-existing county boundaries (Cheshire East, Cheshire West & Chester, Bedford and Central Bedfordshire), an aggregate of the populations for the constituent districts has been used for calculation purposes. Please note that a similar approach has been applied to the numerator data (see Numerator: Definition for more information).
Denominator: source
Home Office data.
Data quality: Accuracy and completeness
Coverage is complete. Historically there have been differences between police forces in procedures for recording, collecting and collating offence data. In April 2002, the National Crime Recording Standard (NCRS) was introduced to ensure greater consistency between forces in recording crime. As a result of the introduction of the NCRS, the recorded number of violence against the person offences increased by 23% in 2002/03. (http://www.crimestatistics.org.uk/output/Page107.asp). Home Office research (http://www.homeoffice.gov.uk/rds/pdfs06/ hosb1206.pdf) suggests that about 45% of violent crimes are reported to the police, with 68% of these being officially recorded as a crime. The Home Office provides specific counting rules regarding the counting and classification of violence against the person offences recorded by Police Forces in England and Wales (http://www. homeoffice.gov.uk/rds/countrules.html). Please also see the Numerator Data Caveats section for related information.
Health Profiles 2009 The Indicator Guide
36 Section 2: Our communities
Table 3 – Indicator Technical Methods Numerator: extraction
Downloaded from www.homeoffice.gov.uk
Numerator: aggregation/ allocation
Area where offence took place allocated by police using records with attached postcodes.
Numerator data caveats
These data exclude figures from British Transport Police and crimes committed at airports within the jurisdictions of Greater Manchester, Metropolitan and Kent police force areas. See Numerator Definition section for change in counting rules. The very nature of crime statistics make analysing the data effectively particularly challenging. Recorded crime, as presented here, is susceptible to fluctuations brought about by changes in recording, counting or even policing methods. In their annual reports entitled “Crime in England and Wales”, the Home Office balance these recorded crime figures with the results of the British Crime Survey (BCS). The BCS is an annual survey which asks respondents about their experience of crime during the previous twelve months, irrespective of whether it was reported to the police. This enables BCS data to be more resilient to changes in recording or policing. Because of survey size limitations, however, it is not possible to use the BCS data at the geographical level required by the Health Profiles. The latest annual report, “Crime in England and Wales 2007/08”, presents a thorough picture of the challenges around crime statistics, in particular how the recorded crime and BCS data may diverge and why, and should be read alongside this metadata document for a full understanding of the issues. The document is at http://www.homeoffice.gov.uk/rds/crimeew0708.html.
Denominator data caveats
Ideally, the denominator should reflect all people at risk of violent crime in the area, including visiting and or migratory persons. This data is not available and resident data has been used as a proxy. The use of resident population as a denominator is a proxy measure for population exposure and is consistent with how this indicator is presented elsewhere.
Methods used to calculate indicator value
Calculation of the numerator: Calculated as an annual of the number of the offences reported in 2007/08. Denominator count: Mid-2006 population estimates (all ages). The numerator was then divided by the denominator; the resulting value was then multiplied by 1000 to give a crude rate per 1000 population.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Data for the Isles of Scilly and City of London have not been included individually. Isles of Scilly is included in the totals for Cornwall, South West GOR and England, and City of London figures have been incorporated into totals for London GOR and England.
Health Profiles 2009 The Indicator Guide
37 Section 2: Our communities
Disclosure Control
Not applicable as no counts less than 5.
Confidence Intervals calculation method
The 95% confidence intervals for crude rate of violence against the person offences were constructed by assuming a Poisson distribution for the number of offences occurring in a specific period of time. The lower and upper limits for this confidence interval were then obtained using the following formulae relating the chi-square and Poisson distributions:
χ α , 2d χ 2α , 2d 2
LL = LL =
2
22
2
χ 1- α , 2(d+1) χ 1- α2 , 2(d+1) 2 2
UL =
22
UL = where LL and UL are the lower and upper2100*(1-a) per cent confidence limits and d denotes the number of observed events (e.g, serious injuries, 100* (1-α) violent offences, deaths) per unit of time exposed. χ² a/2,2d is the (100*a/2)th percentage point for a chi-squared distribution with 2d degrees of freedom 100* (1-α) th 2 and χ²(1-α/2),2(d+1) is the (100*(1-a a/2)) percentage point for a chi-squared χ α/2,2d distribution on 2(d+1) degrees of freedom. χ2 α/2,2d th The confidence limits for the rates were then obtained by dividing the upper (100*α/2) and lower limits for the counts by the person time exposed. (100*α/2)th χ2 (1-α/2),2(d+1)
Reference: χ2 (1-α/2),2(d+1) Dobson AJ, Kuulasmaa K, Eberle E, Scherer J. Confidence intervals for (100*(1-α/2))th weighted sums of Poisson parameters. Statistics in Medicine 1991;10:457-462.
(100*(1-α/2))th (100*(1-α/2))th (100*(1-α/2))th
Health Profiles 2009 The Indicator Guide
38 Section 2: Our communities
6. CARBON EMISSIONS INDICATOR
Basic Information 1. What is being measured?
Total end user CO2 emissions per capita (tonnes of CO2 per resident)
2. Why is it being measured?
Carbon dioxide emission is one of the main contributors to greenhouse gases which cause global warming. This indicator allows local authorities to monitor the effectiveness of their efforts to reduce carbon dioxide emissions.
3. How is this indicator actually defined?
Indicator 186: Per capita CO2 emissions in the local authority area.
4. Who does it measure?
All Persons, all ages.
5. When does it measure it?
Annually, 2006.
6. Will It measure absolute numbers or proportions?
Proportions: CO2 emissions per capita.
7. Where does the data actually come from?
AEA Energy & Environment for the Department for Environment Food and Rural Affairs (Defra)
8. How accurate and complete will the data be?
The indicator relies on centrally produced statistics to measure end user CO2 emissions in the Local Area. Some sectors of CO2 emissions have been excluded from the estimates as they could not be easily aggregated to local or regional levels.
9. Are there any caveats/warnings/ problems?
Analysis carried out by AEA Energy and Environment has confirmed that the data available for the construction of this local area Climate Change Indicator are sufficiently robust with relatively low levels of uncertainty. The data are based upon LA CO2 estimates produced by AEA technology on behalf of Defra.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
Confidence intervals have not been constructed for this indicator.
Table 1 – Indicator Description Information component
Page 4 Health Summary Indicator No. 6
Subject category/ domain(s)
Our Communities
Indicator name (* Indicator title in health profile)
Carbon emissions
Health Profiles 2009 The Indicator Guide
39 Section 2: Our communities
PHO with lead responsibility
WMPHO
Date of PHO dataset creation
23/02/2009
Indicator definition
Local and Regional CO2 Emissions Estimates for 2006 for the UK per capita
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs
Timeliness
Annual. The figures for 2007 will be available in Autumn 2009. It is expected that further improvements to the underlying data and methodology will be made in this dataset which will improve the level of accuracy in certain sectors.
Rationale: What this indicator purports to measure
This indicator purports to measure the contribution of each locality to Carbon dioxide emissions. It is based on estimates of activities resulting in carbon dioxide production (chiefly energy use). The carbon dioxide resulting from energy production is allocated not to the place where the energy is produced but to the place where the energy is used (eg carbon dioxide produced by a power station is allocated to the places using the electricity produced not to the place where the power station is located).
Rationale: Public Health Importance
Carbon dioxide is a major contributor to green house gases which cause global warming. Global warming is of the greatest public health importance and according to the worst predictions could result in millions of deaths across the world as a result of many changes including reduced capacity to produce food, flooding by sea of low lying lands and increased frequency of extreme weather events.
Rationale: Purpose behind the inclusion of the indicator
While the contribution of each single local authority area to carbon dioxide is a miniscule fraction of the global total, carbon dioxide emissions cannot be reduced unless each area accepts its responsibility to reduce emissions and does not simply assume that their contribution does not matter because someone else will make a reduction.
Rationale: Policy relevance
As explained in the previous section reducing carbon dioxide emissions require everyone to make an effort thinking globally but acting locally. This indicator allows each local authority and others to assess how successful their efforts to reduce carbon dioxide emissions have been. The indicator is one of those included in the National Indicator Set that can be used to monitor Local Area Agreements. The data are based upon LA CO2 estimates produced by AEA technology on behalf of Defra. These National Statistics estimate all emissions in an area and have been modified slightly for this indicator to exclude certain emissions of the following components: • Motorways • EU Emissions Trading Scheme sites (except energy suppliers, e.g. power stations in the scheme, whose emissions are indirectly included via the end user estimates, e.g. for electricity use) • Diesel railways • Land Use, Land Use Change, and Forestry In effect, these are emissions that LAs can be least expected to be responsible for. Arguments can be made for other exclusions, but a line has to be drawn somewhere that results in an indicator that is fair in terms of authority actions actually effecting change in the indicator, and where real change at the local level will be captured. Defra consulted with LAs during development of the indicator for LAA purposes.
Health Profiles 2009 The Indicator Guide
40 Section 2: Our communities
Interpretation: What a high/low level of indicator value means
A high indicator value represents a higher level of CO2 emissions per capita for that local authority. A low indicator value represents a lower level of CO2 emissions per capita for that local authority.
Interpretation: Potential for error due to type of measurement method
The indicator relies on accurately attributing emissions from energy use and emissions related to energy supply to the local level. This involves an iterative approach to estimate emissions to the end user, as energy producers will use energy from other producers and therefore also be consumers of energy themselves. Some sectors of CO2 emissions have been excluded from the estimates as they could not be easily aggregated to local or regional levels.
Interpretation: Potential for error due to bias and confounding
Unlike the 2005 Local Authority emissions estimates published last year, this dataset has now been classified as full National Statistics, and is no longer considered “experimental”. This gives users greater confidence in the estimates. In order to obtain this classification, a range of quality criteria set out by the UK Statistics Authority had to be met. To achieve this, there have been in particular, key improvements to the accuracy and comparability of the data. These include: • Re-classification of BERR Local Authority energy statistics as National Statistics; • Implementation of improved formal quality assurance procedures; • Methodological improvements; and • Reduced uncertainty in the accuracy of some of the data inputs. In terms of comparability, a consistent time series has now been produced by re-calculating the 2005 estimates to reflect the methodological changes used in calculating the 2006 estimates.
Confidence Intervals: Definition and purpose
Confidence intervals have not been constructed for this indicator.
Table 2 – Indicator Specification Indicator definition: Variable
CO2 emissions per capita
Indicator definition: Statistic
Tonnes of CO2 emissions per resident.
Indicator definition: Gender
Persons
Indicator definition: age group
All ages
Indicator definition: period
2006 calendar year
Indicator definition: scale
Tonnes per resident
Health Profiles 2009 The Indicator Guide
41 Section 2: Our communities
Geography: geographies available for this indicator from other providers
The original indicator is for the UK, so also includes Wales, Scotland and Northern Ireland. Complete dataset available from http://www.defra.gov.uk/ environment/statistics/globatmos/download/regionalrpt/local-regionalco2ni186indicator.xls
Dimensions of inequality: subgroup analyses of this dataset available from other providers
None available
Data extraction: Source
Local government performance framework NI 186 – Per capita CO2 emissions in the LA area Environment Statistics Service, Department for Environment, Food and Rural Affairs, Area 5F Ergon House, 17 Smith Square, London SW1P 3JR, 08459 33 55 77
Data extraction: source URL
http://www.defra.gov.uk/environment/localgovindicators/ni186.htm
Data extraction: date
5th February 2009
Numerator: definition
Total ktonnes of CO2 emissions, built up of Industry and Commercial, Domestic, Road Transport (excluding Motorways ,EU Emissions Trading Scheme sites (except energy suppliers, e.g. power stations in the scheme, whose emissions are indirectly included via the end user estimates, e.g. for electricity use) Diesel railways, Land Use, Land Use Change, and Forestry)
Numerator: source
Local gas, electricity and road transport fuel consumption estimates are published by BERR for 2005 (see http://www.berr.gov.uk/).
Denominator: definition
Mid Year 2006 population estimates, National Statistics
Denominator: source
Included in NI186 dataset though they have used National Statistics 2006 mid year population estimates.
Data quality: Accuracy and completeness
Unlike the 2005 Local Authority emissions estimates published last year, this dataset has now been classified as full National Statistics, and is no longer considered “experimental”. This gives users greater confidence in the estimates. In order to obtain this classification, a range of quality criteria set out by the UK Statistics Authority had to be met. To achieve this, there have been in particular, key improvements to the accuracy and comparability of the data. These include: • Re-classification of BERR Local Authority energy statistics as National Statistics; • Implementation of improved formal quality assurance procedures; • Methodological improvements; and • Reduced uncertainty in the accuracy of some of the data inputs. In terms of comparability, a consistent time series has now been produced by re-calculating the 2005 estimates to reflect the methodological changes used in calculating the 2006 estimates.
Health Profiles 2009 The Indicator Guide
42 Section 2: Our communities
Table 3 – Indicator Technical Methods Numerator: extraction
Download from Department for Environment Food and Rural Affairs (Defra) website
Numerator: aggregation /allocation
AEA Energy & Environment produce the data on behalf of the Department for Environment Food and Rural Affairs (Defra). They give the following description of the process that they use: “This dataset provides a spatial disaggregation of the national CO2 inventory on an End User basis in which emissions from the production and processing of fuels (including electricity) are reallocated to users of these fuels to reflect the total emissions relating to that fuel use. This is in contrast to ‘at source’ emissions in which all emissions are attributed to the sector that emits them directly. The End User basis for reporting emissions has been chosen for this dataset because it fully accounts for the emissions from energy use at the local level and does not penalise local areas for emissions from the production of energy which is then ‘exported’ to other areas. The method used follows as closely as possible that used for the End User emissions calculated as part of the NAEI and reported by Defra at the national level.”
Numerator data caveats
Data from some areas of CO2 emissions are excluded from the calculations due to the fact that the figures are not available disaggregated to required levels.
Denominator data caveats
Population figures are rounded to the nearest thousand and taken from the 2006 mid year population estimates
Methods used to calculate indicator value
The total of CO2 emissions in ktonnes per area, divided by the population rounded to the nearest thousand.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Scilly Isles and City of London has been included at LA level and included at County, Region and National levels.
Disclosure Control
Copyright of data and/or information presented or attached in this document may not reside solely with this Department. Please contact us or see guidance on copyright at: http://www.defra.gov.uk/environment/ statistics/help.htm
Confidence Intervals calculation method
Not applicable.
Health Profiles 2009 The Indicator Guide
43
Section 3: Children’s and young people’s health
Health Profiles 2009 The Indicator Guide
44 Section 3: Children’s and young people’s health
7. SMOKING IN PREGNANCY INDICATOR Basic Information 1. What is being measured?
Smoking in pregnancy
2. Why is it being measured?
Smoking in pregnancy has well known detrimental effects for the growth and development of the baby. Encouraging pregnant women to stop smoking during pregnancy may also help them kick the habit for good, and thus provide health benefits for the mother.
3. How is this indicator actually defined?
The percentage of women giving birth in 2007/08 who are current smokers at the time of delivery out of all maternities where smoking in pregnancy status is recorded.
4. Who does it measure?
Women giving birth in 2007/08 whose smoking in pregnancy status is recorded.
5. When does it measure it?
Financial year 2007/08.
6. Will It measure absolute numbers or proportions?
Proportions: Number of women who smoke in pregnancy per 100 maternities where smoking status is recorded.
7. Where does the data actually come from?
Care Quality Commission.
8. How accurate and complete will the data be?
93% of LAs have less than or equal to 5% mothers with unknown smoking status and were therefore deemed as having valid prevalence estimates to be included in the Health Profiles 2009.
9. Are there any caveats/warnings/ problems?
The indicator is based on the mother’s response and is therefore susceptible to responder bias. The data is originally recorded at PCT level which has been converted to LA level using birth weighting. If there are several LAs within one PCT they will all have the same prevalence, thereby masking any variation in prevalence which may exist within that PCT. The county, GOR, SHA and England prevalences were calculated from the LAs with valid data, therefore if these LAs are not truly representative of the areas they are being aggregated to, it would result in a biased estimate for aggregated areas.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Health Profiles 2009 The Indicator Guide
45 Section 3: Children’s and young people’s health
Table 1 – Indicator Description Information component
Page 4 Spine Chart – Indicator 7
Subject category/ domain(s)
Children’s and young people’s health.
Indicator name (* Indicator title in health profile)
Smoking in pregnancy
PHO with lead responsibility
ERPHO
Date of PHO dataset creation
11.01.2009
Indicator definition
The percentage of women giving birth in 2007/08 who are current smokers at the time of delivery out of all maternities where smoking in pregnancy status is recorded.
Geography
The data is available at the PCT level this has been converted into the appropriate areas for Health Profiles
Timeliness
The data is released on a quarterly basis.
Rationale: What this indicator purports to measure
Women were asked at the time of delivery whether they currently smoke. Due to the negative connotations surrounding smoking during pregnancy, this may be more susceptible to responder bias.
Rationale: Public Health Importance
Smoking in pregnancy has well-known detrimental effects for the growth and development of the baby. Encouraging pregnant women to stop smoking during pregnancy may also help them kick the habit for good, and thus provide health benefits for the mother.
Rationale: Purpose behind the inclusion of the indicator
To encourage women to quit smoking when pregnant and hopefully as a consequence of this to quit smoking permanently. To highlight LAs with high smoking in pregnancy prevalence to encourage intervention.
Rationale: Policy relevance
This indicator was judged to be a valid and an important measure of public health and was therefore included in the ‘children’s and young people’s health’ domain of the profiles. the NHS Priorities and Planning Framework 2003–06 target to deliver a one percentage point reduction per year in the proportion of women continuing to smoke throughout pregnancy, focussing especially on smokers from disadvantaged groups, is intended to contribute to the national target to reduce the gap in mortality between “routine and manual” groups and the population as a whole by 2010. It is also included in PCT Local Delivery Plans (LDP).
Interpretation: What a high/low level of indicator value means
An indicator value better than average (green circle in health summary chart) represents statistically significantly fewer women smoking during pregnancy for that local authority when compared to the national value. An indicator value worse than average (red circle in health summary chart) represents statistically significantly more women smoking during pregnancy for that local authority when compared to the national value.
Interpretation: Potential for error due to type of measurement method
It is based on the mother’s response and is therefore susceptible to responder bias, particularly as there is a stigma attached to smoking during pregnancy.
Health Profiles 2009 The Indicator Guide
46 Section 3: Children’s and young people’s health
Interpretation: Potential for error due to bias and confounding
As mentioned above, it is likely to be susceptible to responder bias if some mothers are reluctant to admit to smoking in pregnancy, which would make the estimates of smoking in pregnancy more likely to be an underestimate of the actual figure. Achieving good coverage minimises opportunity for non response bias. All PCTs that returned data had at least 95% coverage which was the cut off for inclusion. The calculation of county, region, SHA and England prevalence is based on only those LAs which returned data. They may therefore not be truly representative of the whole area if those LAs with invalid data have different prevalences compared with those LAs with valid data. When viewing amalgamated areas it should be taken into consideration that they may be subject to non response bias, and the overall regional and England prevalences are only estimates.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider the confidence interval the greater the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Health Profiles 2009 The Indicator Guide
47 Section 3: Children’s and young people’s health
Table 2 – Indicator Specification Indicator definition: Variable
The percentage of women giving birth in 2007/08 who are current smokers at the time of delivery out of all maternities where smoking in pregnancy status is recorded.
Indicator definition: Statistic
Percentage
Indicator definition: Gender
Female
Indicator definition: age group
n/a
Indicator definition: period
Financial year 2007/08
Indicator definition: scale
Percentage
Geography: geographies available for this indicator from other providers
PCT (commissioning PCT)
Dimensions of inequality: subgroup analyses of this dataset available from other providers
n/a
Data extraction: Source
Care Quality Commission.
Data extraction: source URL
Most of the data is available on the CQC website:
http://www.cqc.org.uk/publications.cfm?fde_id=1255 The denominator is different from that used by the CQC. In the health profiles only maternities with smoking in pregnancy status recorded is used as the denominator, where as all maternities are used as the denominator for the prevalence of smoking in pregnancy at PCT level calculated by the CQC. This means where you have LAs that are identical to PCTs, the two percentages will not be identical unless there is 100% data coverage (see supporting indicator). Where there is not 100% data coverage the health profiles percentage will be higher than the CQC percentage due to the smaller denominator in the former. The number of births by LA-PCT was extracted from the ONS 2007 birth file, which is sent to PHOs.
Data extraction: date
Received the CQC (formerly Health Care Commission) spreadsheet on 16/10/2008
Numerator: definition
Number of women known to smoke in pregnancy
Numerator: source
Care Quality Commission
Health Profiles 2009 The Indicator Guide
48 Section 3: Children’s and young people’s health
Denominator: definition
Number of maternities where smoking in pregnancy status is recorded
Denominator: source
Care Quality Commission
Data quality: Accuracy and completeness
The CQC data is based on commissioning PCT for the financial year 2007/08 and not PCT of residence, and therefore the number of maternities is slightly different from the ONS number of births in 2007 based on residence. There are estimated prevalence data for 93% of LAs. As the county, GOR, SHA and England prevalences are calculated from LAs which returned data, if these LAs are not truly representative of the areas they are being aggregated to, it would result in a biased estimate for aggregated areas. The data is originally recorded at PCT level which has been converted to LA level using birth weighting. If there are several LAs within one PCT they will all have the same prevalence, thereby masking any variation in prevalence which may exist within that PCT. The regional and SHA values for North West, Yorkshire and Humber, East Midlands, South East (South East Coast SHA and South Central SHA) and South West differ slightly from those produced from summing the CQC PCT data, due to how PCTs that straddle SHAs were dealt with. The regional and SHA data in the health profiles have been calculated by summing the LAs.
Health Profiles 2009 The Indicator Guide
49 Section 3: Children’s and young people’s health
Table 3 – Indicator Technical Methods Numerator and denominator: extraction
The numerator The numerator is converted from the PCT level numerator: “Actual number of women known to smoke in pregnancy (LDPR)”. There were no PCTs with >5% unknown smoking in pregnancy status to exclude but 7 PCTs did not return data (DNRs) and were therefore excluded from the analysis in Excel. The file was then loaded into Access to calculate the LA numerator. Weights based on births were used in order to convert the PCT data into LA data. Births by resident LA and PCT were extracted from the ONS 2007 birth file, to give the number of births in every LA-PCT overlapping block in 2007 by resident LA. This was carried out in SQL using the syntax: SELECT OSCTY, OSLAUA, PCT, COUNT(*) FROM D_births_96_07 LEFT JOIN HES.dbo.A_NationalPostcodes_Feb08 ON D_births_96_07.[Mat Pcode] = HES.dbo.A_NationalPostcodes_feb08.[PCD2] WHERE [SB Ind] <> ‘1’ and year(D_Births_96_07.DOBchild) = 2007 and (GOR = ‘A’ or GOR = ‘B’or GOR = ‘D’or GOR = ‘E’or GOR = ‘F’or GOR = ‘G’or GOR = ‘H’or GOR = ‘J’or GOR = ‘K’) GROUP BY OSCTY, OSLAUA, PCT ORDER BY OSCTY, OSLAUA, PCT To calculate the numerator we need to know the proportion of the LA-PCT overlap of each PCT and then multiply this proportion by the “Actual number of women known to smoke in pregnancy” in the PCT. An LA may overlap several PCTs so this has to be summed at the end. Expressed as an equation the numerator is calculated as follows: Smoking in pregnancy MumsLA = ∑ n/N * Smoking in pregnancy MumsPCT Smoking in pregnancy MumsLA = Estimated number of mothers known to smoke in pregnancy in the LA n= number of births in the LA-PCT overlapping block N = Number of births in the PCT
Health Profiles 2009 The Indicator Guide
50 Section 3: Children’s and young people’s health
Smoking in pregnancy MumsPCT = Number of mothers known to smoke in pregnancy in the PCT Before the data were entered into Access those PCTs who did not return data were excluded, and therefore those LAs entirely made up of PCTs who did not return data will not have an estimate. However if they are made up of several PCTs and only one of which has valid data they will have a numerator and denominator based on this one PCT. We do not want to include an estimate for this LA, so we need to select this out. To find out where this is the case, we worked out the LA-PCT birth overlap in each LA then summed this for the LA we had a prevalence estimate for. This is expressed in the following formula: ∑LA-PCT number of births/LA total number of births. There were no LAs partially made up with PCTs who did not return data so no LAs were excluded at this stage. The Denominator The denominator is calculated in the same way as the numerator. At PCT level the “number of maternities with smoking in pregnancy status known” is calculated in Excel (“actual number of maternities” – “number of maternities with status unknown”). The valid data is then loaded into Access and the denominator is calculated in the same way as the numerator using the same method of weighting and the formula: MaternitiesLA = ∑ n/N * MaternitiesPCT MaternitiesLA = Estimated number of maternities of known status for smoking in pregnancy in the LA n= number of births in the LA-PCT overlapping block N = number of births in the PCT MaternitiesPCT = number of maternities of known status for smoking in pregnancy in the PCT The denominator along with the numerator is then exported to Excel to complete the analysis. Numerator: aggregation/ allocation
Data was aggregated up to LA level using the method described above, in Access. Data was aggregated to county, region, SHA and England level using a look-up table and pivot table in Excel.
Numerator data caveats
We do not have a complete set of data due to 7 DNR PCTs, and not all PCTs had 100% coverage although all that returned data had at least 95% coverage.
Health Profiles 2009 The Indicator Guide
51 Section 3: Children’s and young people’s health
Denominator data caveats
The denominator and numerator are based on commissioning PCT whereas the weights are based on LA/PCT of residence. The denominator is different from that used by the CQC. In the health profiles only maternities with smoking in pregnancy status recorded is used as the denominator, where as all maternities are used as the denominator for the prevalence of smoking in pregnancy at PCT level calculated by the CQC. This means where you have LAs that are identical to PCTs, the two percentages will not be identical unless there is 100% data coverage (see supporting indicator). Where there is not 100% data coverage the health profiles percentage will be higher than the CQC percentage due to the smaller denominator in the former. For example, Barking and Dagenham PCT have a smoking in pregnancy prevalence of 11.3% using the CQC definition, and 11.4% according to the health profile definition.
Methods used to calculate indicator value
The indicator value is calculated as follows in Excel: Percentage of mothers who smoke in pregnancy: numerator/denominator *100
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Not used
Disclosure Control
None
Confidence Intervals calculation method
The 95% and 99.8% confidence intervals are calculated using Julian Flowers’ (erpho) confidence interval tool: http://www.erpho.org.uk/Download/Public/15374/1/Confidence_Intervals_ Wilson.xls This calculates confidence intervals using the following method for a confidence interval of a proportion as described by R.G. Newcombe. If r is the observed number of subjects with some feature in a sample of size n then the estimated proportion who have the feature is p = r/n. The proportion who do not have the feature is q = 1-p. First, calculate the three quantities A = 2r + z2;
B = z z 2 + 4rq ;
and
C=2(n+z2),
where z is z1-α/2, from the standard Normal distribution. Then the confidence interval for the population proportion is given by (A-B)/C
to
(A+B)/C
This method has the considerable advantage that it can be used for any data. When there are no observed events, r and hence p are both zero, and the recommended confidence interval simplifies to 0 to z2/(n+z2). When r = n so that p = 1, the interval becomes n/(n+z2) to 1. Reference: Newcombe, RG. Two-sided confidence intervals for the single proportion: comparison of seven methods. Stat Med 1998;17:857-72.
Health Profiles 2009 The Indicator Guide
52 Section 3: Children’s and young people’s health
8. BREAST FEEDING INITIATION INDICATOR Basic Information 1. What is being measured?
Breast Feeding Initiation
2. Why is it being measured?
Breast feeding has well known health benefits for the child and for the mother in later life. It costs nothing to implement and should be amenable to change through public health intervention.
3. How is this indicator actually defined?
The percentage of women giving birth in 2007/08 who put their baby to the breast in the first 48 hours after delivery, out of all maternities where breast feeding initiation status is recorded.
4. Who does it measure?
Women giving birth in 2007/08 with breast feeding initiation status recorded.
5. When does it measure it?
Financial year 2007/08
6. Will It measure absolute numbers or proportions?
Proportions: Number of women who initiate breast feeding per 100 maternities where breast feeding initiation status is recorded.
7. Where does the data actually come from?
Care Quality Commission
8. How accurate and complete will the data be?
92% of LAs have less than or equal to 5% mothers with unknown breast feeding initiation status and were therefore deemed as having valid prevalence estimates to be included in the Health Profiles 2009.
9. Are there any caveats/warnings/ problems?
The indicator is based on observation and is therefore susceptible to measurement bias. The data is originally recorded at PCT level which has been converted to LA level using birth weighting. If there are several LAs within one PCT they will all have the same prevalence, thereby masking any variation in prevalence which may exist within that PCT. A LA will not have an estimate, if any of its component PCTs have not returned data or have invalid data. The county, GOR, SHA and England prevalences were calculated from the LAs with data returned, therefore if these LAs are not truly representative of the areas they are being aggregated to, it would result in a biased estimate for aggregated areas.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Health Profiles 2009 The Indicator Guide
53 Section 3: Children’s and young people’s health
Table 1 – Indicator Description Information component
Page 4 Spine Chart – Indicator 8
Subject category/ domain(s)
Children’s and young people’s health.
Indicator name (* Indicator title in health profile)
Breast Feeding Initiation
PHO with lead responsibility
ERPHO
Date of PHO dataset creation
11.01.2009
Indicator definition
Measures the percentage of mothers who put their baby to the breast in the first 48 hours after delivery. The numerator is the number of women initiating breast feeding and the denominator the number of maternities with breast feeding initiation status recorded. This for the financial year 2007/08.
Geography
The data is available at the PCT level this has been converted into the appropriate areas for Health Profiles 2009.
Timeliness
The data is released on a quarterly basis.
Rationale: What this indicator purports to measure
It measures the percentage of babies put to the breast in the first 48 hours after delivery, i.e. initiation of breast feeding, directly. However, it does not necessarily reflect the level of sustained breast feeding. The indicator measures only whether baby received breast milk at least once within the first 48 hours rather than sustained breast feeding likely to deliver benefits. The argument is made that mothers who haven’t initiated breast feeding within the first 48 hours rarely do so later, but it is unknown what proportion persist. From a health inequalities perspective it is reasonable to focus attention on the early period to encourage disadvantaged groups to initiate breast feeding. Data is now being collected at 6–8 weeks (NI 53). With increasing quality in the future, this data can be included in future health profiles.
Rationale: Public Health Importance
Breast feeding has well known health benefits for the child and for the mother in later life (Breastfeeding – NHS ). It costs nothing to implement and should be amenable to change through public health intervention.
Rationale: Purpose behind the inclusion of the indicator
To encourage uptake of breast feeding.
Rationale: Policy relevance
This indicator was judged to be a valid and an important measure of public health and was therefore included in the ‘children’s and young people’s health’ domain of the profiles. The NHS Priorities and Planning Framework 2003–06 has a target to deliver a two percentage point increase in breast feeding initiation rates each year, with a particular focus on women from disadvantaged groups. A target for breast feeding initiation was also included in PCT Local Delivery Plans (LDPs).
To highlight LAs with low breast feeding prevalence to encourage intervention.
Health Profiles 2009 The Indicator Guide
54 Section 3: Children’s and young people’s health
Interpretation: What a high/low level of indicator value means
An indicator value better than average (green circle in health summary chart) represents statistically significantly more women initiating breast feeding for that local authority when compared to the national value. An indicator value worse than average (red circle in health summary chart) represents statistically significantly fewer women initiating breast feeding for that local authority when compared to the national value. High percentage of breast feeding initiation is good as it could be considered as a proxy measure for sustained breast feeding.
Interpretation: Potential for error due to type of measurement method
It is based on observation and therefore it prone to measurement bias.
Interpretation: Potential for error due to bias and confounding
It is likely to be subject to measurement bias by the midwives/nurses who record the data and their interpretation of whether breast feeding has been initiated. We have excluded those LAs with component PCTs with >5% missing data to try to avoid non response bias within the LA. The calculation of county, region, SHA and England prevalence is based on only those LAs which returned data. They may therefore not be truly representative of the whole area if those LAs which returned data have different prevalences compared with those LAs which did not. Where a LA consists of a PCT with complete data and a PCT with invalid data, no value was published for the LA (this was the case for Braintree and Runnymede LAs), although the complete PCT data was used for the calculation of the SHA,regional and England values. When viewing amalgamated areas it should be taken into consideration that they may be subject to non response bias, and the overall regional and England prevalences are only estimates.
Health Profiles 2009 The Indicator Guide
55 Section 3: Children’s and young people’s health
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider the confidence interval the greater the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Table 2 – Indicator Specification Indicator definition: Variable
The percentage of women giving birth in 2007/08 who put their baby to the breast in the first 48 hours after delivery, out of all maternities where breast feeding initiation status is recorded.
Indicator definition: Statistic
Percentage
Indicator definition: Gender
Female
Indicator definition: age group
n/a
Indicator definition: period
Financial year 2007–08
Health Profiles 2009 The Indicator Guide
56 Section 3: Children’s and young people’s health
Indicator definition: scale
Percentage
Geography: geographies available for this indicator from other providers
PCT (commissioning PCT)
Dimensions of inequality: subgroup analyses of this dataset available from other providers
n/a
Data extraction: Source
Care Quality Commission
Data extraction: source URL
The data are available on the CQC website: http://www.cqc.org.uk/publications.cfm?fde_id=1255 The denominator in the Health Profiles is different from that used by the CQC. In the health profiles only maternities with breast feeding initiation status recorded is used as the denominator, where as all maternities are used as the denominator for the prevalence of breast feeding initiation at PCT level calculated by the CQC. This means where you have LAs that are identical to PCTs, the two percentages will not be identical unless there is 100% data coverage (see supporting indicator). Where there is not 100% data coverage the health profiles percentage will be higher than the CQC percentage due to the smaller denominator in the former. The number of births by LA-PCT was extracted from the ONS 2007 birth file, which is sent to PHOs.
Data extraction: date
Received the CQC spreadsheet on 16/10/2008
Numerator: definition
Number of women giving birth in 2007/08 who initiate breast feeding in the first 48 hours after delivery.
Numerator: source
Care Quality Commission
Denominator: definition
Number of maternities in 2007/08 where breast feeding initiation status is recorded.
Denominator: source
Care Quality Commission
Health Profiles 2009 The Indicator Guide
57 Section 3: Children’s and young people’s health
Data quality: Accuracy and completeness
The CQC data is based on commissioning PCT for the financial year 2007/08 and not PCT of residence, and therefore the number of maternities is slightly different from the ONS number of births in 2007 based on residence and is for a slightly different time period. The ONS birth file is based on number of deliveries, where as the CQC maternity data is based on mothers and does not take into account multiple births. This factor should not affect prevalence estimates as multiple births are generally a random event and mothers with multiple births are likely to breast feed both babies or neither one. There are estimated prevalence data for 92% LAs. As the county, GOR, SHA and England prevalences are calculated from LAs with returned data, if these LAs are not truly representative of the areas they are being aggregated to, it would result in a biased estimate for aggregated areas. The data is originally recorded at PCT level which has been converted to LA level using birth weighting. If there are several LAs within one PCT they will all have the same prevalence, thereby masking any variation in prevalence which may exist within that PCT. The regional and SHA values for North West, Yorkshire and Humber, East Midlands, South East (South East Coast SHA and South Central SHA) and South West differ slightly from those produced from summing the CQC PCT data, due to how PCTs that straddle SHAs were dealt with. The regional and SHA data in the health profiles have been calculated by summing the LAs.
Health Profiles 2009 The Indicator Guide
58 Section 3: Children’s and young people’s health
Table 3 – Indicator Technical Methods Numerator and Denominator: extraction
The Numerator The numerator is converted from the PCT level numerator: “Actual number of women known to initiate breastfeeding (LDPR)”. There were no PCTs with more than 5% unknowns to exclude and so therefore only those PCTs who DNRs (did not return) were excluded in Excel which was 7 PCTs in total. The file was then loaded into Access to calculate the LA numerator. Weights based on births were used in order to convert the PCT data into LA data. Births by resident LA and PCT were extracted from the ONS 2007 birth file, to give the number of births in every LA-PCT overlapping block in 2006 by resident LA. This was carried out in SQL using the syntax: SELECT OSCTY, OSLAUA, PCT, COUNT(*) FROM D_births_96_07 LEFT JOIN HES.dbo.A_NationalPostcodes_Feb08 ON D_births_96_07.[Mat Pcode] = HES.dbo.A_NationalPostcodes_feb08.[PCD2] WHERE [SB Ind] <> ‘1’ and year(D_Births_96_07.DOBchild) = 2007 and (GOR = ‘A’ or GOR = ‘B’or GOR = ‘D’or GOR = ‘E’or GOR = ‘F’or GOR = ‘G’or GOR = ‘H’or GOR = ‘J’or GOR = ‘K’) GROUP BY OSCTY, OSLAUA, PCT ORDER BY OSCTY, OSLAUA, PCT To calculate the numerator we need to know the proportion of the LA-PCT overlap of each PCT and then multiply this proportion by the “Actual number of women known to initiate breast feeding” in the PCT. A LA may overlap several PCTs so this has to be summed at the end. Expressed as an equation the numerator is calculated as follows: Breast feeding MumsLA = ∑ n/N * Breast feeding MumsPCT Breast feeding MumsLA = Estimated number of mothers known to initiate breast feeding in the LA n= number of births in the LAD-PCT overlapping block N = Number of births in the PCT Breast feeding MumsPCT = number of mothers known to initiate breast feeding in the PCT. Before the data were entered into Access those PCTs which did not return data were excluded, therefore LAs entirely made up of PCTs with not data will not have
Health Profiles 2009 The Indicator Guide
59 Section 3: Children’s and young people’s health
an estimate. However if they are made up of several PCTs and only one of which has valid data they will have a numerator and denominator based on this one PCT. We do not want to include an estimate for this LA, so we need to select this out. To find out where this is the case, we worked out the LA-PCT birth overlap in each LA then summed this for the LA we had a prevalence estimate for. If it did not add up to 1 a prevalence estimate was not published (this was the case for Braintree and Runnymede LAs). This is expressed in the following formula: ∑LA-PCT number of births/LA total number of births. The denominator The denominator is calculated in the same way as the numerator. At PCT level the “number of maternities with breast feeding status known” is calculated in Excel (“actual number of maternities” – “number of maternities with status unknown”). The valid data is then loaded into Access and the denominator is calculated in the same way as the numerator using the same method of weighting and the formula: MaternitiesLA = ∑ n/N * MaternitiesPCT MaternitiesLA= Estimated number of maternities of known status for breast feeding in the LA n= number of births in the LAD-PCT overlapping block N = number of births in the PCT MaternitiesPCT = number of maternities of known status for breast feeding in the PCT The denominator along with the numerator is then exported to Excel to complete the analysis. Numerator: aggregation/ allocation
Data was aggregated up to LA level using the method described above in Access. Data was aggregated to county, region, SHA and England level using a look-up table and pivot table in Excel.
Numerator data caveats
We do not have a complete set of data as there was not 100% coverage from PCTs that returned data (although in all these cases coverage was over 95%) and more importantly 7 PCTs did not return data.
Denominator data caveats
The denominator and numerator are based on commissioning PCT whereas the weights are based on LA/PCT of residence. The denominator is different from that used by the CQC. In the health profiles only maternities with breast feeding initiation status recorded is used as the denominator, where as all maternities are used as the denominator for the prevalence of breast feeding initiation at PCT level calculated by the CQC. This means where you have LAs that are identical to PCTs, the two percentages will not be identical unless there is 100% data coverage (see supporting indicator). Where there is not 100% data coverage, the health profiles percentage will be higher than the CQC percentage due to the smaller denominator in the former. For example, Barking and Dagenham PCT have a breast feeding initiation prevalence of 72.0% according the CQC definition, and a slightly higher prevalence of 72.2% according to the health profile definition.
Methods used to calculate indicator value
The indicator value is calculated as follows in Excel: Percentage of women giving birth in 2007/08 initiating breast feeding: numerator/ denominator *100
Health Profiles 2009 The Indicator Guide
60 Section 3: Children’s and young people’s health
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Not used
Disclosure Control
none
Confidence Intervals calculation method
The 95% and 99.8% confidence intervals are calculated using Julian Flowers’ (erpho) confidence interval tool: http://www.erpho.org.uk/Download/
Public/15374/1/Confidence_Intervals_Wilson.xls This calculates confidence intervals using the following method for a confidence interval of a proportion as described by R.G. Newcombe. If r is the observed number of subjects with some feature in a sample of size n then the estimated proportion who have the feature is p = r/n. The proportion who do not have the feature is q = 1-p. First, calculate the three quantities A = 2r + z2;
B = z z 2 + 4rq ;
and
C=2(n+z2),
where z is z1-α/2, from the standard Normal distribution. Then the confidence interval for the population proportion is given by (A-B)/C
to
(A+B)/C
This method has the considerable advantage that it can be used for any data. When there are no observed events, r and hence p are both zero, and the recommended confidence interval simplifies to 0 to z2/(n+z2). When r = n so that p = 1, the interval becomes n/(n+z2) to 1. Reference Newcombe, RG. Two-sided confidence intervals for the single proportion: comparison of seven methods. Stat Med 1998;17:857-72.
Health Profiles 2009 The Indicator Guide
61 Section 3: Children’s and young people’s health
9. PHYSICALLY ACTIVE CHILDREN INDICATOR Basic Information 1. What is being measured?
Physical Activity provision for school children
2. Why is it being measured?
To help increase childhood participation in physical activity by highlighting areas with low participation rates in order to assess need and enable targeted intervention.
3. How is this indicator actually defined?
The percentage of children attending state schools belonging to a School Sport Partnership who participate in at least 2 hours of high quality PE and school sport within and beyond the curriculum in a typical week of the academic year.
4. Who does it measure?
The total number of school children in state schools who responded to the 2007/08 TNS School Sport Survey who participate in at least 2 hours of high quality PE and out of hours school sport in a typical week.
5. When does it measure it?
2007/08 academic year
6. Will It measure absolute numbers or proportions?
Percentage of all responses to the 2007/8 TNS School Sport Survey.
7. Where does the data actually come from?
TNS Social Research: Annual Survey of School Sport Partnerships on behalf of the Department for Children, Schools and Families.
8. How accurate and complete will the data be?
All partnership schools in the maintained sector in England were included in this survey, with response from over 99% of all such schools. Whilst some private schools have joined a School Sport Partnership and completed the TNS survey, their data is excluded from the DCSF national results dataset. Responses to the TNS School Sport Survey were selfreported by schools given the potential for positive response bias.
Health Profiles 2009 The Indicator Guide
62 Section 3: Children’s and young people’s health
9. Are there any caveats/warnings/ problems?
The indicator is a direct measure of service provision within state schools. It is important to acknowledge that this indicator does not take into account physical activity provision within private schools or physical activity undertaken by children outside of school, and so cannot be used as a total measure of physical activity for children. Although efforts have been made to clearly define ‘high quality PE’, the term is still open to individual interpretation and there is some potential for positive response bias as schools are self-reporting. It is important to make the distinction between physical activity and structured PE/sport. While PE/ sport may be physical activity, physical activity is not necessarily PE/sport; and this data should only be considered as a part of this wider issue within an area.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
Page 4: Health summary – Indicator No. 9
Subject category/ domain(s)
Children’s and Young People’s Health
Indicator name (*Indicator title in health profile)
Physically active children
PHO with lead responsibility
EMPHO
Date of PHO dataset creation
February 2009
Indicator definition
The percentage of children attending state schools belonging to a School Sport Partnership who participate in at least 2 hours of high quality PE and school sport within and beyond the curriculum in a typical week of the academic year.
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs
Timeliness
The survey was carried out between May and July 2008. Results were published in October 2008. The survey is carried out annually with the results from the 2009 survey due out in October 2009.
Health Profiles 2009 The Indicator Guide
63 Section 3: Children’s and young people’s health
Rationale: What this indicator purports to measure
The percentage of children in state maintained schools who participate in at least 2 hours of high quality PE and school sport per week
Rationale: Public Health Importance
All children, whatever their circumstance, should be able to participate in and enjoy PE and sport at school. Physical activity during childhood has a range of benefits including healthy growth and development, maintenance of energy balance, psychological well-being and social interaction. Through improved concentration and self-esteem, it can also improve school attendance, behaviour and attainment. The benefits continue well into adulthood by reducing, early in life, some of the key risk factors for diseases such as coronary heart disease, diabetes and osteoporosis. Some evidence also suggests that participation in physical activity during childhood can help to establish a physically active lifestyle in later life. Physical inactivity in childhood is a modifiable lifestyle risk factor.
Rationale: Purpose behind the inclusion of the indicator
To help increase childhood participation in physical activity by highlighting areas with low participation rates in order to assess need and enable targeted intervention.
Rationale: Policy relevance
This indicator relates directly to Indicator 5 of PSA Target 22: “Increase the percentage of 5 to 16 year olds participating in at least two hours per week of high-quality PE and sport at school” to 85% by 2008, and to at least 75% in each School Sport Partnership by 2008”. The long-term ambition is to offer all children at least 4 hours of sport every week by 2010. See http://www.teachernet.gov.uk/teachingandlearning/subjects/pe/ or http:// www.hm-treasury.gov.uk/d/pbr_csr07_psa22.pdf for further information. This indicator relates indirectly to indicator NI 57 in the National Indicator Set for Local Authorities and Local Authority Partnerships – “Children and young people’s participation in high-quality PE and sport”, which will measure both in school provision and community (out-of-school hours) provision. The in-school aspect of this indicator will be recorded through the School Sport Survey. See http://www.communities.gov.uk/documents/localgovernment/pdf/735125.pdf for further information. In 2004, the CMO Report “At least 5 a week: Evidence on the impact of physical activity and its relationship to health” recommended that children and young people need at least 60 minutes of moderate intensity physical activity each day. See “Chapter 4: Health benefits of physical activity in childhood and adolescence” at http://www.dh.gov.uk/en/Publicationsandstatistics/ Publications/PublicationsPolicyAndGuidance/DH_4080994 Increasing childhood participation in physical activity is also a key message in “Choosing Activity: A Physical Activity Action Plan”, which summarises how the government will deliver the commitments on physical activity that are presented in the public health white paper “Choosing Health” and “Tackling Health Inequalities: 2007 Status report on the Programme for Action”. See http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4105354 and http://www.dh.gov.uk/ en/Publicationsandstatistics/Publications/PublicationsPolicyAndGuidance/ DH_083471.
Health Profiles 2009 The Indicator Guide
64 Section 3: Children’s and young people’s health
Interpretation: What a high/low level of indicator value means
Indicator values should be considered in conjunction with the national target for 85% of school children to be participating in 2 or more hours of physical activity per week, and not just compared to the England average value when assessing the need for public health intervention.
Interpretation: Potential for error due to type of measurement method
The indicator is a direct measure of service provision within state schools. It is important to acknowledge that this indicator does not take into account physical activity provision within private schools or physical activity undertaken by children outside of school, and so cannot be used as a total measure of physical activity for children. Although efforts have been made to clearly define ‘high quality PE’, the term is still open to individual interpretation and there is some potential for positive response bias as schools are self-reporting.
Interpretation: Potential for error due to bias and confounding
Whilst private schools are able to join a School Sport Partnership and complete the TNS survey, their data is excluded from the DCSF national results dataset adding possible bias to the results. It is important to make the distinction between physical activity and structured PE/sport. While PE/sport may be physical activity, physical activity is not necessarily PE/sport; and this data should only be considered as a part of this wider issue within an area.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider the confidence interval the greater the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Health Profiles 2009 The Indicator Guide
65 Section 3: Children’s and young people’s health
TABLE 2 – INDICATOR SPECIFICATION Indicator definition: Variable
High quality PE and out-of-school hours sport
Indicator definition: Statistic
Percentage
Indicator definition: Gender
Persons
Indicator definition: age group
5–16 years
Indicator definition: period
2007/08 academic year
Indicator definition: scale
Percentage of school children in Year 1 – Year 11 attending state schools belonging to a School Sport Partnership
Geography: geographies available for this indicator from other providers
Results summarised by School, School Sport Partnership, Local Education Authority, England are available from the National PE, School Sport and Club Links website at https://dservuk.tns-global.com/schoolsports2007/default.aspx
Dimensions of inequality: subgroup analyses of this dataset available from other providers
The TNS 2008 School Sport Survey National Results report provides 2007/08 data broken down by school type (primary, secondary, special), school year (Year 1 – Year 11), and School Sport Partnership phase (ie when the school joined the School Sport Partnership programme). https://dservuk.tns-global. com/schoolsports2007/default.aspx
Data extraction: Source
TNS Social Research: Annual Survey of School Sport Partnerships on behalf of the Department for Children, Schools and Families.
Data extraction: source URL
Received directly
Data extraction: date
Received directly October 2008
Numerator: definition
The total number of school children in state schools who responded to the 2007/08 TNS School Sport Survey who participate in at least 2 hours of high quality PE and out of hours school sport in a typical week. (Includes compulsory PE curriculum time, optional PE curriculum time (e.g. for GCSE PE students), time spent participating in out of hours school led or school supervised sporting activities, including those taking place during break times within the school day. Excludes travelling time. As of the end of 2007 all state maintained schools in England are in School Sport Partnerships, arranged into 450 Partnerships
Numerator: source
TNS 2007/08 School Sport Survey Further information is available at https://dservuk.tns-global.com/ SchoolSports2008/Default.aspx Data is not available to download but was received directly from TNS.
Health Profiles 2009 The Indicator Guide
66 Section 3: Children’s and young people’s health
Denominator: definition
The total number of school children in state schools who responded to the 2007/08 TNS School Sport Survey. As of the end of 2007 all state maintained schools in England are in School Sport Partnerships, arranged into 450 Partnerships
Denominator: source
TNS 2007/08 School Sport Survey Further information is available at https://dservuk.tns-global.com/ SchoolSports2008/Default.aspx Data is not available to download but was received directly from TNS.
Data quality: Accuracy and completeness
Coverage: All partnership schools in the maintained sector in England were included in this survey, with response from over 99% of all such schools. Whilst some private schools have joined a School Sport Partnership and completed the TNS survey, their data is excluded from the DCSF national results dataset. Data Quality: Responses to the TNS School Sport Survey were self-reported by schools giving the potential for positive response bias. The questionnaire had a 99% response rate giving a relatively complete dataset. Data Reliability: Validation of the 2007/08 TNS School Sport Survey was carried out in 10% of School Sport Partnerships selected at random. The validation concluded that the majority of schools kept auditable records of the information they submitted in their survey responses, and that where auditable records were not available, reasonably robust results should have been produced using the verbally reported approach adopted by schools. For further information see the 2007/08 TNS School Sport Survey Results report, section 2.4 at http://www.dcsf.gov.uk/research/data/uploadfiles/DCSF-RW063. pdf The same questionnaire is used each year giving comparable results.
Table 3 – Indicator Technical Methods Numerator: extraction
Data received directly from TNS.
Numerator: aggregation/ allocation
Performed by TNS.
Numerator data caveats Denominator data caveats Methods used to calculate indicator value
The number of children in state maintained schools who participated in at least two hours of high quality PE/school sport per week for each local authority was divided by the total number of children within each surveyed school with valid responses to questions on physical activity in the TNS School Sport Survey and multiplied by 100. This generated the percentage achieving the recommended levels for school sport participation.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Data for the Isles of Scilly and City of London have not been presented. The Isles of Scilly are included in the Cornwall UA and South West regional and national totals and the City of London figures have been included in the regional London and national totals.
Health Profiles 2009 The Indicator Guide
67 Section 3: Children’s and young people’s health
Disclosure Control
None applied
Confidence Intervals calculation method
Confidence intervals have been calculated using the following method for a confidence interval of a proportion as described by RG Newcombe. If r is the observed number of subjects with some feature in a sample of size n then the estimated proportion who have the feature is p = r/n. The proportion who do not have the feature is q = 1-p. First, calculate the three quantities
A = 2r + z2; B = z z 2 + 4rq ; and C=2(n+z2), where z is z1-α/2, from the standard Normal distribution. Then the confidence interval for the population proportion is given by (A-B)/C
to
(A+B)/C
This method has the considerable advantage that it can be used for any data. When there are no observed events, r and hence p are both zero, and the recommended confidence interval simplifies to 0 to z2/(n+z2). When r = n so that p = 1, the interval becomes n/(n+z2) to 1. Reference Newcombe, RG. Two-sided confidence intervals for the single proportion: comparison of seven methods. Stat Med 1998;17:857-72.
Health Profiles 2009 The Indicator Guide
68 Section 3: Children’s and young people’s health
10. OBESE CHILDREN INDICATOR BASIC INFORMATION1. What is being
Prevalence of obesity in Reception year pupils
measured? 2. Why is it being measured?
To estimate and monitor prevalence of obesity in children. To help reduce the prevalence of childhood obesity; inform planning and delivery of services for children; ensure the proper targeting of resources to tackle obesity.
3. How is this indicator actually defined?
Prevalence of childhood obesity, percentage of school children in Reception year, ages 4–5, 2007/08, persons.
4. Who does it measure?
Children in Reception year (Year R, ages 4–5)
5. When does it measure it?
The National Child Measurement Programme (NCMP) takes place every school year. The data are usually made available around February following publication of the Information Centre National Report.
6. Will It measure absolute numbers or proportions?
Proportion: Percentage of school children in Reception year who are obese.
7. Where does the data actually come from?
The Information Centre for Health and Social Care (IC).
8. How accurate and complete will the data be?
All 152 PCTs provided data and the participation rate across all PCTs was 89% for Reception year (477,652 children measured). 125 PCTs exceeded 85% coverage for Reception year. The NCMP does not include children in the Independent sector, therefore, coverage of school children aged 4–5 is not complete.
9. Are there any caveats/warnings/ problems?
There is the potential for error in the collection, collation and interpretation of the data (bias may be introduced due to poor response rates and selective opt out of larger children which it is not possible to control for).
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Health Profiles 2009 The Indicator Guide
69 Section 3: Children’s and young people’s health
Table 1 – Indicator Description Information component
Pg 4 Health Summary – Indicator No. 10
Subject category/ domain(s)
Children’s and young people’s health
Indicator name (* Indicator title in health profile)
Prevalence of obesity in Reception year pupils (*Obese children)
PHO with lead responsibility
SEPHO
Date of PHO dataset creation
14/01/2009
Indicator definition
Prevalence of childhood obesity, percentage of school children in Reception year, ages 4–5, 2007–08, persons
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs.
Timeliness
The National Child Measurement Programme (NCMP) takes place every year. The 2007/08 data were made available in December 2008.
Rationale: What this indicator purports to measure
Estimate of prevalence of obesity in children in Reception year (Year R, ages 4–5).
Rationale: Public Health Importance
The UK is experiencing an epidemic of obesity affecting both adults and children. The Health Survey for England (HSE) found that among boys and girls aged 2 to 15, the proportion who were classified as obese increased from 10.9 per cent in 1995 to 18.0 per cent in 2005 among boys, and from 12.0 per cent to 18.1 per cent among girls. For those aged 2 to 10, the increase over the same period was from 9.6 per cent to 16.6 per cent for boys and 10.3 per cent to 16.7 per cent for girls. (see http://www.ic.nhs.uk/statistics-and-datacollections/health-and-lifestyles/obesity/statistics-on-obesity-physical-activityand-diet-england-2006)
Note: the definition for childhood obesity used for population monitoring is different to that used in a clinical setting, and produces higher prevalence figures. The prevalence figures here do not equate to the proportion of the child population that would be clinically classified as obese, but more to the proportion of children ‘classified as obese’ according to these population monitoring criteria.
There is concern about the rise of childhood obesity and the implications of such obesity persisting into adulthood. The health consequences of childhood obesity include: increased blood lipids, glucose intolerance, type 2 diabetes, hypertension, increases in liver enzymes associated with fatty liver, psychological problems – social isolation, low self-esteem, teasing and bullying, exacerbation of conditions such as asthma. The National Institute of Health and Clinical Excellence has produced guidelines to tackle obesity in adults in children – Obesity: the prevention, identification, assessment and management of overweight and obesity in adults and children. Available at http://guidance.nice.org.uk/CG43/guidance
Health Profiles 2009 The Indicator Guide
70 Section 3: Children’s and young people’s health
Rationale: Purpose behind the inclusion of the indicator
To estimate and monitor prevalence of obesity in children.
Rationale: Policy relevance
The National Child Measurement Programme (NCMP) was established in 2005 and is one element of the Government’s work programme on childhood obesity. It is operated by the Department of Health and the Department for Children, Schools and Families (DCSF) and central collection and analysis of the NCMP data is coordinated by the Information Centre for health and social care (The IC) which published the final report on its website along with a web tool to view the results by area: http://www.ic.nhs.uk/our-services/improvingpatient-care/national-child-measurement-programme.
To help reduce the prevalence of childhood obesity; inform planning and delivery of services for children; ensure the proper targeting of resources to tackle obesity.
Every year, as part of the NCMP, children in Reception (typically aged 4–5 years) and Year 6 (aged 10–11 years) are weighed and measured during the school year. The findings are used to inform local planning and delivery of services for children and gather population-level surveillance data to allow analysis of trends in excess weight. The programme also seeks to raise awareness of the importance of healthy weight in children. For further details on the NCMP see: http://www.dh.gov.uk/en/Publichealth/ Healthimprovement/Healthyliving/DH_073787 Childhood obesity was the subject of a Public Service Agreement (PSA) target set in July 2004 which aims to halt the year-on-year rise in obesity among children under 11 by 2010 in England. In September 2007, the government announced a new ambition: to reverse the rising tide of obesity and overweight in the population by ensuring that all individuals are able to maintain a healthy weight. The government’s initial focus is on children, and by 2020 they aim to have reduced the proportion of overweight and obese children to 2000 levels. The government strategy on excess weight is set out in “Healthy Weight, Healthy Lives: A Cross-Government Strategy for England” 2008 see: http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_082378 Performance Assessment Framework indicators will be abolished from April 2008 when a new performance framework for local authorities and local authority partnerships comes into place. This framework is supported by a national indicator set. Obesity among primary school age children in Reception Year is indicator NI 56 in the new national indicator set. See: http://www. communities.gov.uk/publications/localgovernment/nationalindicator Interpretation: What a high/low level of indicator value means
A high indicator value (red circle in health summary chart) represents a statistically significant higher level of estimated child obesity prevalence when compared to the national value. A low indicator value (yellow circle in health summary chart) represents a statistically significant lower level of estimated child obesity prevalence when compared to the national value. However obesity in children at any prevalence level greater than 0 is undesirable, and therefore a low indicator value should not mean that PH action is not needed.
Health Profiles 2009 The Indicator Guide
71 Section 3: Children’s and young people’s health
Interpretation: Potential for error due to type of measurement method
Measurement of children’s heights and weights, without shoes and coats and in normal, light, indoor clothing was overseen by healthcare professionals and undertaken in schools by trained staff. However, measurement bias may be introduced via the weighing and measuring and recording process, for example where an area did not use the recommended scales or ensure these were calibrated. There is also evidence that a systematic ‘rounding down’ of measures has occurred in certain areas, and that this has the effect of reducing the reported prevalence of obesity in those areas.
Interpretation: Potential for error due to bias and confounding
Bias may be introduced due to poor response rates and selective opt out of larger children which it is not possible to control for. For the 2007/08 NCMP, child measurements could be taken at any time during the 2007/08 academic year. Consequently, some children were almost two years older than others in the same school year at the point of measurement. As obesity prevalence is known to increase with age, there is potential for this to bias the results Ethnicity has been shown to be linked closely to reported prevalence of child obesity. This might be classed as a confounding factor, as it is not known how suitable the British 1990 Growth Reference is for ethnic populations. The dataset this reference is based on consists of only White British individuals. Further analysis is currently being undertaken to help better understand the links between child obesity and ethnicity. The reported prevalence of obesity is also closely related to the socio-economic characteristics of the children measured. The NCMP does not include pupils at independent schools and, as these pupils may be from a more affluent background, there is a possibility that this might bias prevalence figures using the NCMP.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider the confidence interval is the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health.
Health Profiles 2009 The Indicator Guide
72 Section 3: Children’s and young people’s health
The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with a white symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or amber symbol depending on whether it is worse or better than the national value respectively.
Table 2 – Indicator Specification Indicator definition: Variable
Prevalence of childhood obesity. Obesity in children is commonly defined for epidemiological purposes as having a Body Mass Index (BMI) greater than the 95th percentile (using the British 1990 growth reference). This is in line with the Health Survey for England and other population studies within the UK. Note: As prevalence rates have been calculated using the 95th percentile (see Methods used to calculate indicator value). This does not equate to the proportion of children clinically defined as obese, as the 98th percentile is used in clinical settings.
Indicator definition: Statistic
Percentage of school children in Reception year
Indicator definition: Gender
Persons
Indicator definition: age group
School children in Reception year (Year R), ages 4–5
Indicator definition: period
School year 2007–08 (September 2007 to September 2008)
Indicator definition: scale Geography: geographies available for this indicator from other providers
PCT, new SHA Available from http://www.ncmp.ic.nhs.uk/
Dimensions of inequality: subgroup analyses of this dataset available from other providers
The Information Centre’s summary report (see: http://www.ic.nhs.uk/) shows strong links between socio-economic status, ethnicity, urban/rural environment and prevalence. There are also some significant differences in prevalence by gender.
Data extraction: Source
The Information Centre for Health and Social Care (IC).
Data extraction: source URL
www.ic.nhs.uk/
Health Profiles 2009 The Indicator Guide
73 Section 3: Children’s and young people’s health
Data extraction: date
December 2008
Numerator: definition
The number of primary school age children in Year R (Reception year, ages 4–5) with valid height and weight recorded in the school year 2007–08 who are classified as obese.
Numerator: source
www.ic.nhs.uk/
Denominator: definition
The total number of primary school age children in Year R (Reception year, ages 4–5) with valid height and weight recorded in the school year 2007–08.
Denominator: source
www.ic.nhs.uk/
Data quality: Accuracy and completeness
Pupils eligible for inclusion in the NCMP were all children in Reception and Year 6 attending non-specialist maintained state schools in England. The NCMP does not include children in the Independent sector, therefore, coverage of school children aged 4–5 is not complete. Primary care trusts (PCTs) coordinate the data collection exercise with the support and cooperation of schools. The numbers of pupils eligible for inclusion at each school were provided by the DCSF but PCTs could edit these figures if necessary to ensure pupil denominators accurately reflected the number of children attending the school on the day of measurement. PCTs could also add or remove schools from their geograpically assigned list if, despite being within their PCT boundary, another PCT had undertaken measurement in that school. The IC conducted checks when all data was submitted to ensure schools had not been removed by one PCT but not added on by another. Primary Care Trust (PCT) staff used specially designed Excel spreadsheets (the NCMP Data-Capture tool) to enter these data and upload them to a central database at the Information Centre for Health and Social Care. All 152 PCTs provided data and the participation rate across all PCTs was 89% for Reception year (477,652 children measured). 125 PCTs exceeded 85% coverage for Reception year. PCTs were set a participation rate target for NCMP 2007/08 of 85% (for Reception and Year 6). Participation rates were assessed by comparing numbers measured with total numbers on school lists and are published in the IC National report (see http://www.ncmp.ic.nhs.uk/). Participation rates should be treated with caution because of difficulties in calculation of eligible pupil numbers particularly in Reception when pupils may join the school throughout the year. Low participation rates may bias prevalence figures and analysis has shown that PCTs with lower participation rates tend to have lower levels of prevalence than those with high participation rates. This suggests that there might be higher levels of opting out among children with higher BMIs. However, initial analysis by the IC shows participation rate to have little of no effect on prevalence for children in Reception.
Health Profiles 2009 The Indicator Guide
74 Section 3: Children’s and young people’s health
Table 3 – Indicator Technical Methods Numerator: extraction
Numerator extracted from an Access database provided by IC which included all data uploaded to the central NCMP database as at December 2008. The numerator was based on a pre-calculated field included within the database which categorised each child as “healthy weight”, “overweight” or “obese”.
Numerator: aggregation/ allocation
The figures have been aggregated to local authorities from school counts on the basis of postcode of school. Regional figures have been calculated by combining the Local Authority figures. SHA figures (provided for the South East Region only) have been calculated by assigning schools to SHAs according to the PCT who measured those pupils.
Numerator data caveats
The total number of children measured was the total number of records uploaded to the NCMP database after the following records had been removed: • Blank school Unique Reference Number, height, weight, sex or age. • Children outside the age range (age in months between 48 and 83 inclusive) • Independent/Private/Special Educational Needs (SEN) pupils • Records from Academies, Community Special, Foundation Special, Independent School Approved for SEN pupils, Non-maintained Special, Other Independent, Other Independent Special School, Pupil Referral Unit, where these schools had been measured and entered by PCTs • Records with extreme heights, weights and BMIs where value was further than 7 standard deviations from the mean After the above data cleaning, approximately 477,600 year R records remained. Further information on data quality and the validation process is included in the 2007/08 report from the Information Centre: http://www.ic.nhs.uk/ statistics-and-data-collections/supporting-information/health-and-lifestyles/ obesity/the-national-child-measurement-programme/national-childmeasurement-programme-2007/08/ncmp-2007-08-report
Denominator data caveats
As above.
Health Profiles 2009 The Indicator Guide
75 Section 3: Children’s and young people’s health
Methods used to calculate indicator value
The IC undertook the following steps to calculate the number of obese children: Every child’s Body Mass Index (BMI) was calculated as follows: BMI = 10,000 x weight (kg) height2 (cm) BMI was then referenced against the British 1990 Growth Reference (1) (2). This uses UK growth data from pre-1990 based on a large representative sample of 37,700 children constructed by combining data from 17 separate surveys. These data were then used to model the BMI distribution using Cole’s LMS method (3) for ages 0–20 for both males and females. This dataset can be used to express an individual child’s BMI as a percentile on the 1990 distribution. Each child’s BMI was referenced to the age (in decimal months) and sexspecific BMI growth curve in order to retrieve corresponding L, M and S values. These were used to calculate a z-score for each child using the following formula:
y
M
Where y is the BMI score. The z-score was converted to a p-score using the standardised normal distribution and children with a BMI p-score >=0.95 were flagged as obese. Using the data supplied by IC, prevalence rates were calculated by dividing the number of primary school age children in Year R (Reception year, ages 4–5) with valid height and weight measurements recorded as obese by the total number of primary school age children in Year R with valid height and weight measurements. This figure was multiplied by 100 to give percentage prevalence. (1) Cole, T.J., Freeman, J.V. and Preece, M. A. ‘British 1990 growth reference centiles for weight, height, body mass index and head circumference fitted by maximum penalized likelihood’, Statistics in Medicine, 17, 407-429 (1998). (2) Cole, T. J., Freeman, J. V. and Preece, M. A. ‘Body Mass Index reference curves for the UK, 1990’, Archives of Disease in Childhood, 73, 17-24 (1995). (3) Cole, T. J. and Green, P. J. ‘Smoothing reference centile curves: The LMS method and penalized likelihood’, Statistics in Medicine, 11, 1305-1319 (1992). Note: 40 records from Reception year had not been assigned to a local authority. On closer examination of school location, these were reassigned to Telford local authority.
Health Profiles 2009 The Indicator Guide
76 Section 3: Children’s and young people’s health
Small Populations: How Isles of Scilly and City of London populations have been dealt with
City of London has been included in the Hackney, Regional and National totals. Isles of Scilly have been included in the new Cornwall Unitary Authority.
Disclosure Control
None applied.
Confidence Intervals calculation method
The 95% confidence intervals are calculated with the method described by Wilson1 and Newcombe2 which is a good approximation of the exact method. The estimated proportions of subjects with and without the feature of interest were calculated: observed number of obese children in each area =r sample size = n proportion with feature of interest = p = r/n proportion without feature of interest = q (1 – p) Three values (A,B and C) were then calculated as follows: A = 2r + z2;
B = z z 2 + 4rq ;
and
C=2(n+z2)
where z is the appropriate value, z1-α/2, from the standard Normal distribution. Then the confidence interval for the population proportion is given by (A-B)/C
to
(A+B)/C
This method has the considerable advantage that it can be used for any data. When there are no observed events, r and hence p are both zero, and the recommended confidence interval simplifies to 0 to z2/(n+z2). When r = n so that p = 1, the interval becomes n/(n+z2) to 1. 1
Wilson EB. J Am Stat Assoc 1927, 22, 209-212
Newcombe, RG. Two-sided confidence intervals for the single proportion: comparison of seven methods. Stat Med 1998;17:857-72.
2
Health Profiles 2009 The Indicator Guide
77 Section 3: Children’s and young people’s health
11. CHILDREN’S TOOTH DECAY INDICATOR Basic Information 1. What is being measured?
Children’s tooth decay (at age 5)
2. Why is it being measured?
To draw attention to areas of high tooth decay. To improve oral health in children by reducing the prevalence of dental decay.
3. How is this indicator actually defined?
Mean number of teeth per child sampled which were either actively decayed or had been filled or extracted.
4. Who does it measure?
Sampled number of children at 5 years old with decayed/missing/filled teeth (dmft).
5. When does it measure it?
2005/06
6. Will It measure absolute numbers or proportions?
Numbers (the mean of)
7. Where does the data actually come from?
BASCD (British Association for the Study of Community Dentistry) www.bascd.org
8. How accurate and complete will the data be?
Data was only available at PCT level, and had to be apportioned to Local Authorities. This process can only be approximate where PCTs are not completely contained within Local Authorities. Data was missing for 22 PCTs and consequently it was not possible to present mean dmft for 26 local authorities and one county. Data for 12 PCTs was only available in combination with one or more other PCTs. Health Profiles 2009 are based on the new unitary authority boundaries and it has not been possible to present mean dmft for 3 of these new authorities.
9. Are there any caveats/warnings/ problems?
Positive/negative consent issue may cause bias.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
Pg 4 Health Summary – Indicator No. 11
Subject category/ domain(s)
Health and ill-health in our community
Indicator name (*Indicator title in health profile)
Mean number of decayed/missing/filled teeth in five-year-olds (*Children’s Tooth Decay)
Health Profiles 2009 The Indicator Guide
78 Section 3: Children’s and young people’s health
PHO with lead responsibility
Yorkshire and Humber
Date of PHO dataset creation
March 2008
Indicator definition
Mean number of teeth per child sampled which were either actively decayed or had been filled or extracted.
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs. Children are allocated to these geographies according to the location of their school.
Timeliness
BASCD (British Association for the Study of Community Dentistry) conduct a survey of five-year-olds every two years. The 2007/08 survey results are due for publication later in 2009.
Rationale: What this indicator purports to measure
The mean number of decayed, missing or filled teeth per child.
Rationale: Public Health Importance
Dental caries (tooth decay) and periodontal (gum) disease are the most common dental pathologies in the UK. Tooth decay has become less common over the past two decades, but is still a significant health and social problem. It results in destruction of the crowns of teeth and frequently leads to pain and infection. Dental disease is more common in deprived, compared with affluent, communities. The indicator is a good direct measure of dental health and an indirect, proxy measure of child health and diet.
Rationale: Purpose behind the inclusion of the indicator
To draw attention to areas of high tooth decay. To improve oral health in children by reducing the prevalence of dental decay.
Rationale: Policy relevance
Children’s National Service Framework
Interpretation: What a high/low level of indicator value means
An indicator value worse than average (red circle in health summary chart) represents a statistically significant worse rate of children’s tooth decay for that local authority when compared to the national value.
This indicator supports Choosing Health and LAAs.
An indicator value better than average (green circle in health summary chart) represents a statistically significant better rate of children’s tooth decay for that local authority when compared to the national value. From this statistic alone, it cannot be deduced whether the problem is evenly distributed or confined to a small pocket of children.
Health Profiles 2009 The Indicator Guide
79 Section 3: Children’s and young people’s health
Interpretation: Potential for error due to type of measurement method
Data was only available at PCT level, and had to be apportioned to Local Authorities. This process can only be approximate where PCTs are not completely contained within Local Authorities. Data was missing for 22 PCTs listed below and consequently it was not possible to present mean dmft for 26 local authorities and one county. Data for 12 PCTs was only available in combination with one or more other PCTs, these are also listed below. 45 PCTs where positive consent operated for all or part of the sampling period are listed below. Data missing for these PCTs PCT 5A7 5A8 5AK 5CE 5CV 5CW 5FN 5FQ 5FR 5FT 5FV 5FY 5GF 5J8 5J9 5JP 5KA 5KC 5KD 5KE 5KV TAK
PCT name Bromley PCT Greenwich PCT Southend on Sea PCT Bournemouth Teaching PCT South Hams and West Devon PCT Torbay Care Trust South and East Dorset PCT North Devon PCT Exeter PCT East Devon PCT Mid Devon PCT Teignbridge PCT Huntingdonshire PCT Durham Dales PCT Darlington PCT Castle Point and Rochford PCT Derwentside PCT Durham and Chester-le-Street PCT Easington PCT Sedgefield PCT Poole PCT Bexley Care Trust
Data available for these PCTs in combination only South East Hertfordshire PCT (5GJ), Royston & Buntingford and Bishops Stortford PCT (5GK) Ashford PCT (5LL), Canterbury and Coastal PCT (5LM), East Kent Coastal PCT (5LN), Shepway PCT (5LP) Watford and Three Rivers PCT (5GV), Dacorum PCT (5GW) Hertsmere PCT (5CP), St Albans and Harpenden PCT (5GX) Welwyn and Hatfield (5GG) North Hertfordshire and Stevenage PCT (5GH)
Health Profiles 2009 The Indicator Guide
80 Section 3: Children’s and young people’s health
Positive/Mixed consent in operation during sampling period PCT PCT name 5DF North Hampshire 5DK Newbury 5E1 North Tees 5G6 Blackwater Valley & Hart PCT 5HF Wyre PCT 5KJ Part of Craven Harrogate & Rural (Airedale) 5KL Sunderland 5KM Middlesborough 5L5 Guildford & Waverley PCT 5L7 Surrey Heath and Woking Area PCT 5AC Daventry & South Northants 5AW Airedale 5CC Blackburn with Darwen PCT 5CF Bradford City 5CG South,West Bradford 5CH North Bradford 5CK Doncaster Central 5CX Trafford South PCT 5EE North Sheffield 5EG North Eastern Derbyshire 5EN Sheffield West 5EP Sheffield South West 5EQ South East Sheffield 5F5 Salford PCT 5F6 Trafford North PCT 5G7 Hyndburn & Ribble Valley PCT 5HA Central Liverpool PCT 5HE Fylde PCT 5HG Ashton, Leigh & Wigan PCT 5HH Leeds West 5HJ Leeds North East 5HK East Leeds 5HL South Leeds 5HM Leeds North West 5HN High Peak and Dales 5HQ Bolton PCT 5HX Ealing 5J6 Calderdale 5J7 North Kirklees 5LJ Huddersfield Central 5LK South Huddersfield 5LV Northamptonshire Heartlands 5LW Northampton 5M5 South Sefton PCT 5M7 Sutton and Merton
Form of Consent Positive Positive Positive Positive Positive Positive Positive Positive Positive Positive Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed Mixed
The data source is a series of nationally co-ordinated dental epidemiological surveys commissioned by individual PCTs to standardised national protocols and diagnostic standards and involving the dental examination of children in the specified age-group, in state schools. The data source is part of a cycle of nationally co-ordinated dental epidemiological surveys as outlined in Health Service Guidelines (93)25. Surveys are conducted every second year for 5 year olds and every fourth year for 12 years olds and 14 year olds.
Health Profiles 2009 The Indicator Guide
81 Section 3: Children’s and young people’s health
The data relate to children attending state schools in an area. It cannot be assumed that all children necessarily live in the same area. National minimum standards are set for the random sampling of children to obtain a sample representative of the age-group in the area. Many Health Authorities commission larger samples in order to obtain data on intra-district variations in dental caries for local planning purposes. Data are collected and analysed locally. Summary data items are reported nationally to the British Association for the Study of Community Dentistry, which produces national tables through the Dental Health Services Research Unit at the University of Dundee. These data are published in the journal of the British Association of the Study of Community Dentistry, Community Dental Health. Interpretation: Potential for error due to bias and confounding
For the first time in the history of the BASCD survey of the dental caries experience of 5-year-old children in England and Wales, some parents were required to give positive consent for their children’s teeth to be inspected. In previous surveys all children in state schools were, by default, eligible for inclusion unless their parents submitted a form stating that they did not wish for their children’s teeth to be inspected. This has led to serious concerns that the results of the 2005/06 survey may be biased and not comparable to earlier surveys. Bias may result for several reasons. Firstly, there may be variation between schools generally in how pro-active they are at encouraging parents to return consent forms. Secondly, parents living in more deprived areas (where mean dmft is higher) may be less likely to return consent forms, than those living in more affluent areas (where mean dmft is lower). The picture is further confused by the fact that positive consent was not universally introduced. In 10 PCTs all schools operated positive consent throughout the sampling period, in 35 PCTs some of the schools operated positive consent for some of the sampling period, and in the remaining PCTs no schools operated positive consent. It will be difficult to disentangle the impact of the introduction of positive consent from genuine reductions in the prevalence of dental caries. Anecdotal evidence so far suggests that positive consent had a large effect in some areas.
Health Profiles 2009 The Indicator Guide
82 Section 3: Children’s and young people’s health
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Table 2 – Indicator Specification Indicator definition: Variable
Decayed, filled or missing teeth.
Indicator definition: Statistic
Mean number per child
Indicator definition: Gender
Persons
Indicator definition: age group
Five-year-olds
Indicator definition: period
Sampled during winter months of 2005–06.
Indicator definition: scale
Per child.
Geography: geographies available for this indicator from other providers
PCT and SHA. Available from www.bascd.org
Health Profiles 2009 The Indicator Guide
83 Section 3: Children’s and young people’s health
Dimensions of inequality: subgroup analyses of this dataset available from other providers
None.
Data extraction: Source
2006/06 BASCD survey of the dental caries experience of 5-year-old children in England and Wales.
Data extraction: source URL
Supplied privately by BASCD.
Data extraction: date
March 2007
Numerator: definition
Number of decayed/missing/filled teeth in the survey sample of children in the respective academic year. This was not available at local authority level, therefore the calculation did not proceed by dividing this number by a local authority population.
Numerator: source
Dental epidemiological survey programme undertaken by Health Authorities and co-ordinated nationally for the UK Health Departments by the British Association for the Study of Community Dentistry (BASCD).
Denominator: definition
Number of children in the survey sample. This was not available at local authority level, therefore the calculation did not proceed by dividing the number of decayed, filled or missing teeth by this number.
Denominator: source
Provided by BASCD.
Data quality: Accuracy and completeness
See: Interpretation: Potential for error sections and Methods used in Table 3
Table 3 – Indicator Technical Methods Numerator: extraction
Provided by BASCD.
Numerator: aggregation/ allocation
See Methods used below
Numerator data caveats
The data for 5-year-olds relate to deciduous (milk) teeth.
Denominator data caveats
See: Interpretation: potential for error sections and Methods used below
Health Profiles 2009 The Indicator Guide
84 Section 3: Children’s and young people’s health
Methods used to calculate indicator value
dmft for 2005/06 was calculated at PCT level. A PCT may contain or overlap several LADs. The following steps were taken to estimate dmft for Local Authorities (LADs): • ‘Building blocks’ were created based on intersections between PCTs and LADs. • There is a strong relationship between deprivation and dmft. Rather than base the estimated dmft just on the ‘parent’ PCT which may include areas with different levels of deprivation to the building block belonging to the LAD, the calculation was adjusted for deprivation. • This adjustment was made by firstly applying the regression equation between all PCTs dmft and IMD score, to give an expected dmft for the PCT. Secondly, a dmft estimate was arrived at for each building block which reflected the fact that the more deprived ones would carry more than their ‘fair share’ of their parent PCT’s burden of dental decay. The building block estimate is calculated as follows:
[PCT dmft estimate from survey ]∗ [building block expected dmft ] PCT expected dmft • The calculation needed to reflect the different populations of five-year-olds by area, however this was not available for the building blocks and a proxy weight was derived from the 2005 Royal Mail Postcode Address File (PAF) which is considered a good estimate of population distribution • The final LAD estimate was obtained by multiplying each building block’s dmft estimate by the number of residences in the building block, adding up all the building blocks in the LAD, and dividing the total number of residencies in the LAD.
y LAD = ∑
dmft i Ni ∗ ∗ yi dmft PCT N LAD
Where:
y
• LAD is the estimated mean dmft in the LAD, y is the mean dmft found in the ith PCT (supplied by BASCD) • i • dmfti is the expected dmft in the ith ‘building block’ (LAD by PCT intersection) • dmftPCT is the expected dmft in the ‘parent’ PCT • Ni is number of residences in the ith ‘building block’ (LAD by PCT intersection) • NLAD is the total number of residences in the LAD This was the method used for LADs and Counties. A much simpler method was used for GORs and England because they are made up of whole PCTs. Estimated mean dmft in GORs and England was a weighted average of the mean dmft in the PCTs geographically contained within the GOR or England, using the total five-year-old population (supplied by (BASCD) in the PCT as weights.
y GOR = ∑
Ni ∗ yi N GOR
y GOR is the estimated mean dmft in the GOR
yi
is the mean dmft found in the ith PCT (supplied by BASCD)
Ni is the total number of 5-yr-old children in the ith PCT (supplied by BASCD) NGOR is the total number of 5-yr-old children in the GOR (supplied by BASCD)
Health Profiles 2009 The Indicator Guide
85 Section 3: Children’s and young people’s health
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Excluded.
Disclosure Control
Not applicable
Confidence Intervals calculation method
The 95% confidence intervals were calculated based on the following method: Standard error of mean dmft
SE LAD =
2
dmft i Ni σ i2 ∗ ∗ ∑ dmft N LAD ni PCT
Where the sum is taken over all building blocks in the LAD, and… • SELAD is the estimated standard error of mean dmft in the LAD 2 • σ i is the dmft variance in the ith PCT (supplied by BASCD) • ni is the number of 5-yr-old children sampled (supplied by BASCD) • dmfti is the expected dmft in the ith ‘building block’ (LAD by PCT intersection) • dmftPCT is the expected dmft in the ‘parent’ PCT • Ni is number of residences in the ith ‘building block’ (LAD by PCT intersection) • NLAD is the total number of residences in the LAD The confidence interval for the estimate is: 0.95 ± 1.96* SELAD
Health Profiles 2009 The Indicator Guide
86 Section 3: Children’s and young people’s health
12. TEENAGE PREGNANCY (UNDER 18) INDICATOR Basic Information 1. What is being measured?
Teenage conceptions
2. Why is it being measured?
Teenage parents are prone to poor antenatal health, their babies are more likely to have lower birth weight and are more likely to die in infancy. Teenage mothers are less likely to finish their education, less likely to find a good job, and more likely to end up as single parents or bringing up their children in poverty. Children born to teenage mothers run a much greater risk of poor health and have a much higher chance of becoming teenage mothers themselves.
3. How is this indicator actually defined?
Under-18 conception rate per 1,000 females aged 15–17, 2005–2007 (provisional).
4. Who does it measure?
Females aged under 18 who conceive
5. When does it measure it?
2005–2007
6. Will It measure absolute numbers or proportions?
Measures the rate per 1,000 females aged 15–17
7. Where does the data actually come from?
Teenage Pregnancy Unit & ONS.
8. How accurate and complete will the data be?
Data relating to legal abortions and births is collated through mandatory reporting processes and is of sound data quality.
9. Are there any caveats/warnings/ problems?
Miscarriages and illegal abortions are not included in the conception rates, resulting in rates that may be an under estimation.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
Pg 4 Health Summary – Indicator No 12
Subject category/ domain(s)
Children’s and young people’s health
Indicator name (*Indicator title in health profile)
Teenage conceptions (“Teenage pregnancy [under 18]”)
PHO with lead responsibility
EMPHO
Health Profiles 2009 The Indicator Guide
87 Section 3: Children’s and young people’s health
Date of PHO dataset creation
February 2009
Indicator definition
Under-18 conception rate per 1,000 females aged 15–17 (crude rate) 2005–2007 (provisional).
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs
Timeliness
Data published 14 months after period end, updated annually. (Raw data is available up to 11 months after the event)
Rationale: What this indicator purports to measure
This indicator measures the level of teenage conceptions in the area.
Rationale: Public Health Importance
Teenage pregnancy is a significant public health issue in England. Teenage parents are prone to poor antenatal health, lower birth weight babies and higher infant mortality rates. Teenage mothers are less likely to finish their education, less likely to find a good job, and more likely to end up as single parents or bringing up their children in poverty. Children born to teenage mothers run a much greater risk of poor health and have a much higher chance of becoming teenage mothers themselves. However, it is worth remembering that many young people are successful in adapting to the role of parenthood and have happy, healthy children. For further information see “Teenage Pregnancy: Accelerating the Strategy to 2010” (http://www.everychildmatters.gov.uk/_files/94C1FA 2E9D4C9717E5D0AF1413A329A4.pdf) and Health Statistics Quarterly No. 33 (page 34) (http://www.statistics.gov.uk/downloads/theme_health/ hsq33web.pdf)
Rationale: Purpose behind the inclusion of the indicator
The purpose of including this indicator is to highlight local authorities with high teenage pregnancy rates in order to assess need and enable targeted intervention. In order to assist local areas in reducing under-18 conception rates the following toolkit was produced by DfES. It provides local areas with detailed guidance on the key ingredients that need to be in place locally to reduce under-18 conception rates. www.everychildmatters.gov.uk/resources-and-practice/IG00198/
Rationale: Policy relevance
Teenage conceptions are an important public health target. There is a national PSA target (PSA 14) to reduce teenage conception rates by 50% by 2010 (1998 baseline). See http://www.hm-treasury.gov.uk/d/ pbr_csr07_psa14.pdf for more information. There is an equivalent indicator in the National Indicator Set for Local Authorities and Local Authority Partnerships (NI 112). See http://www. communities.gov.uk/documents/localgovernment/pdf/735125.pdf for more information. This indicator supports Choosing Health and Programme for Action.
Health Profiles 2009 The Indicator Guide
88 Section 3: Children’s and young people’s health
Interpretation: What a high/low level of indicator value means
A high indicator value (red circle) represents a statistically significant higher rate of teenage conceptions for that local authority when compared to the national average. A low indicator value (green circle) represents a statistically significant lower rate of teenage conceptions for that local authority when compared to the national average. Rates that are lower than the England average may still represent a large number of teenage pregnancies and therefore a low indicator value should not be interpreted as meaning that public health action is not needed.
Interpretation: Potential for error due to type of measurement method
Miscarriages and illegal abortions are not included in the conception rates, resulting in rates that may be an under estimation. Data relating to legal abortions and births is collated through mandatory reporting processes and is of sound data quality. Whilst it is acknowledged that the non-inclusion of miscarriages reflects conceptions that are not included in these figures, comparable data in this area is not available. Conception cannot be inferred from the prescription of emergency contraception, and early loss of pregnancy may not be recognised or require medical attention. The omission of this data retains the quality of the indicator rather than limiting it.
Interpretation: Potential for error due to bias and confounding
Teenage pregnancy is often a cause and a consequence of social exclusion. The risk of teenage parenthood is greatest for young people who have grown up in poverty and disadvantage or those with poor educational attainment.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Health Profiles 2009 The Indicator Guide
89 Section 3: Children’s and young people’s health
Table 2 – Indicator Specification Indicator definition: Variable
Estimates of conceptions (excluding pregnancies leading to spontaneous abortion before 24 weeks gestation), based on pregnancies which lead to a delivery at which one or more live or still births occurs and is registered in England and Wales, or termination of pregnancy by abortion under the 1967 Act in England and Wales.
Indicator definition: Statistic
Crude rate
Indicator definition: Gender
Females
Indicator definition: age group
Under 18
Indicator definition: period
2005–2007 (provisional) pooled data
Indicator definition: scale
Per 1,000 female population aged 15–17
Geography: geographies available for this indicator from other providers
ONS publish figures for England and Wales, England, GOR and LA. Available to download from http://www.everychildmatters.gov.uk/resources/IG00200/
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Data is available broken down into under 16 years and under 18 years. Indicator cannot be broken down by ethnicity or socio-economic group.
Data extraction: Source
Produced by the Office for National Statistics and disseminated via the Teenage Pregnancy Unit in the Department for Children, Schools and Families
Data extraction: source URL
Data downloaded from http://www.everychildmatters.gov.uk/resources/ IG00200/ and http://www.statistics.gov.uk/StatBase/Product.asp?vlnk=15055& Pos=1&ColRank=2&Rank=576)
Data extraction: date
Data downloaded on 26th February 2009
Health Profiles 2009 The Indicator Guide
90 Section 3: Children’s and young people’s health
Numerator: definition
The number of conceptions estimated to have occurred to females aged under-18 during 2005–2007. Actual dates of conception are not directly available but have been estimated from a dataset of birth registrations and legal terminations of pregnancy recorded in the respective calendar years plus the following year (i.e. 2005– 2008) as follows: a) For maternities resulting in one or more live births, the date of conception is assumed to be 38 weeks prior to the date of birth (no gestation is recorded at live birth registration); b) For maternities resulting in stillbirths, the date of conception is assumed to be the recorded gestation minus 2 weeks prior to the date of the stillbirth (recorded gestation is time since last menstrual period; conception is assumed to occur 2 weeks after); c) For abortions under the 1967 Act, the date of conception is assumed to be the recorded gestation minus 2 weeks prior to the date of the abortion (recorded gestation is time since last menstrual period; conception is assumed to occur 2 weeks after). The mother’s age at conception is then derived from the mother’s date of birth and the estimated date of conception.
Numerator: source
Produced by the Office for National Statistics and disseminated via the Teenage Pregnancy Unit in the Department for Children, Schools and Families. Data for South Central SHA and South East Coast SHA were subsequently calculated from Local Authority data. Data for the Local Authorities with new (2009) boundaries were also calculated from previous (2008) Local Authority data.
Denominator: definition
Number of females aged 15–17.
Denominator: source
The resident population figures used are ONS revised (that take into account improved estimates of international migration) mid year estimates 2005–2007, as derived from the 2001 census with allowance for subsequent births deaths, migration and ageing of the population. The population numbers were not explicitly presented in the Teenage Pregnancy Unit data, but were back calculated for all areas by using the indicator value and the numerator value. This was with the exception of the South Central SHA and the South Coast SHA, whose populations were subsequently calculated from their constituent local authority populations. Again, Local Authorities with new boundaries had their denominators calculated from pre-2009 Local Authority data. Due to the rounding of the published crude rates, the back-calculated denominators for a geographic area may not equal the sum of the denominators of its constituent areas
Data quality: Accuracy and completeness
Data relating to legal abortions and births is collated through mandatory reporting processes and is of sound data quality.
Health Profiles 2009 The Indicator Guide
91 Section 3: Children’s and young people’s health
TABLE 3 – INDICATOR TECHNICAL METHODS Numerator: extraction
Received directly from the Teenage Pregnancy Unit
Numerator: aggregation/ allocation
No aggregation was required, with the exception of South Central SHA and South East Coast SHA, which were calculated by aggregating from their constituent Local Authorities. Data for the Local Authorities with new (2009) boundaries were also calculated from previous (2008) Local Authority data.
Numerator data caveats
The conception figures exclude any pregnancies leading to spontaneous abortion before 24 weeks gestation and illegal abortions.
Denominator data caveats
In under-18s, a three year age group only (15–17) is used as the denominator, because the vast majority of conceptions to under-18 year olds occur in this age group (95%). To include younger populations would produce misleading results. The 15–17 group is effectively treated as the “population at risk”. As the age at which conception can take place varies from child to child it is impossible to correctly define the population at risk (i.e. data identifying the female population aged 0–17 who are fertile is not available).
Methods used to calculate indicator value
The teenage conceptions indicator is presented as a crude rate. A crude rate is defined as the number of observed events divided by the population-years at risk. For presentation purposes rates are usually multiplied by a scaling factor, in this case 1,000. A rate, r, expressed per 1,000 population is given by:
r=
O × 1,000 n
where: O is the number of observed events (i.e. estimated number of under-18 conceptions, 2005-2007); n is the population-years at risk (i.e. the sum of the mid-year female 15-17 years population estimate for each of the years in the period 2005-2007. As ONS published only the estimated number of under-18 conceptions, and the crude rate per 1,000 population to 1 decimal place, the denominator, the population-years at risk was estimated by back calculating using the above formula.) Small Populations: How Isles of Scilly and City of London populations have been dealt with
Conception data supplied by the ONS does not include data for City of London and Isles of Scilly Local Authorities owing to the small number of events in these small populations. Instead conception figures are combined with those of Hackney LB and Cornwall UA respectively.
Disclosure Control
Teenage conceptions data is subject to disclosure control and any LA with a conception count of fewer than five individuals needs to be suppressed. Because of the size of LA populations however it is only the City of London and Isles of Scilly Local Authorities due to their small population, which are affected. Their conception figures are combined with those of Hackney LB and Cornwall UA respectively.
Health Profiles 2009 The Indicator Guide
92 Section 3: Children’s and young people’s health
Confidence Intervals calculation method
Confidence intervals have been calculated using the Wilson ‘Score’ method for a confidence interval of a proportion as described by Newcombe RG If r is the observed number of subjects with some feature in a sample of size n then the estimated proportion who have the feature is p = r/n. The proportion who do not have the feature is q = 1-p. First, calculate the three quantities 2
A = 2r + z2; B = z z + 4rq ; and C=2(n+z2), where z is z1-α/2, from the standard Normal distribution. Then the confidence interval for the population proportion is given by (A-B)/C to (A+B)/C This method has the considerable advantage that it can be used for any data. When there are no observed events, r and hence p are both zero, and the recommended confidence interval simplifies to 0 to z2/(n+z2). When r = n so that p = 1, the interval becomes n/(n+z2) to 1. Reference Wilson, EB. Probable inference, the law of succession, and statistical inference. Journal of American Statistical Association 1927;22:209-212 quoted in Newcombe, RG. Two-sided confidence intervals for the single proportion: comparison of seven methods. Stat Med 1998;17:857-72.
Health Profiles 2009 The Indicator Guide
93
Section 4: Adult’s Health and Lifestyle
Health Profiles 2009 The Indicator Guide
94 Section 4: adult’s health and lifestyle
13a ADULTS WHO SMOKE UPPER TIER INDICATOR Basic Information 1. What is being measured?
Prevalence of adult smoking.
2. Why is it being measured?
To estimate the prevalence of adult smokers. Smoking prevalence is a direct measure of health care need i.e. the ability to benefit from tobacco control interventions, including smoking cessation services.
3. How is this indicator actually defined?
Prevalence of smoking, percentage of resident population, adults, 2003–2005, persons.
4. Who does it measure?
Adults (aged 16 and over).
5. When does it measure it?
The Health Survey for England (HSE) is carried out annually.
6. Will It measure absolute numbers or proportions?
Percentage of resident adult population aged 16 and over.
7. Where does the data actually come from?
Health Surveys for England, National Centre for Social Research (NatCen). Published by The Information Centre for Health and Social Care (IC), 2007.
8. How accurate and complete will the data be?
The HSE was designed to be representative of the general, non-institutional population living in England. The current “full” sample size of the HSE comprises about 16,000 adults aged 16 and over.
9. Are there any caveats/warnings/ problems?
HSE numerator data are broadly based on observed self-reported current smoking status and as such are subject to responder bias. The Health Survey for England under-samples younger people, people in employment, ethnic minorities, women, those who are healthier but exhibit less healthy behaviour.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
County Health Profiles: Pg 4 Health Summary – Indicator 13
Subject category/ domain(s)
Adults health and lifestyle
Indicator name (*Indicator title in health profile)
Prevalence of adult smoking (*Adults who smoke)
Health Profiles 2009 The Indicator Guide
95 Section 4: adult’s health and lifestyle
PHO with lead responsibility
SEPHO
Date of PHO dataset creation
21/01/2009
Indicator definition
Prevalence of smoking, percentage of resident population, adults, 2003–2005, persons
Geography
England, GOR, County, South East SHAs, new 2009 UAs (Bedford UA, Central Bedfordshire UA, Cheshire East UA, Cheshire West and Chester UA, County Durham, Northumberland UA, Shropshire UA, Wiltshire UA, Cornwall UA).
Timeliness
Updated annually.
Rationale: What this indicator purports to measure
Estimate of smoking prevalence in adults.
Rationale: Public Health Importance
Smoking is the most important cause of preventable ill health and premature mortality in the UK. It is linked to respiratory illness, cancer and coronary heart disease. Smoking not only affects the smoker; over 17,000 children under the age of five are admitted to hospital every year with illnesses resulting from passive smoking. A list of disease specific conditions attributable to smoking is published in The Smoking Epidemic in England, HDA, 2004 http://www.nice.org.uk/page. aspx?o=502811 Smoking is a modifiable lifestyle risk factor; effective tobacco control measures can reduce the prevalence of smoking in the population.
Rationale: Purpose behind the inclusion of the indicator
To help reduce the prevalence of smoking.
Rationale: Policy relevance
Choosing Health: Making healthy choices easier. http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4094550
Smoking prevalence is a direct measure of health care need i.e. the ability to benefit from tobacco control interventions, including smoking cessation services.
Smoking Kills. A White Paper on Tobacco http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4006684 Tackling Health Inequalities: A Programme for Action http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4008268 Interpretation: What a high/low level of indicator value means
A high indicator value (red circle in health summary chart) represents a statistically significant higher level (worse) of estimated adult smoking prevalence when compared to the national value. A low indicator value (green circle in health summary chart) represents a statistically significant lower (better) level of estimated adult smoking prevalence when compared to the national value. However smoking at any prevalence level greater than 0 is undesirable, and therefore a low indicator value should not mean that PH action is not needed.
Health Profiles 2009 The Indicator Guide
96 Section 4: adult’s health and lifestyle
Interpretation: Potential for error due to type of measurement method
Each participant in the Health Survey for England was asked if they currently smoked cigarettes.
Interpretation: Potential for error due to bias and confounding
The Health Survey for England under-samples younger people, people in employment, ethnic minorities, women, those who are healthier but exhibit less healthy behaviour.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate.
Self-reported smoking status may be prone to respondent bias. In order to ensure agreement between these estimates at GOR and SHA level, the lower tier synthetic estimates (districts), which are based on modelled data, have been calibrated. However, the synthetic estimates have not been calibrated to ensure agreement at County level. As a result, there may be inconsistencies between lower tier and county estimates for some areas as the datasets are derived using different methods.
These data have not been age-standardised and, therefore, variation between area values may be a result of differences in population structure.
This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Health Profiles 2009 The Indicator Guide
97 Section 4: adult’s health and lifestyle
Table 2 – Indicator Specification Indicator definition: Variable
Prevalence of smoking. Smoking is defined as self-reported current cigarette smoking.
Indicator definition: Statistic
Percentage of resident adult population aged 16 and over
Indicator definition: Gender
Persons
Indicator definition: age group
Adults (aged 16 and over)
Indicator definition: period
2003–2005
Indicator definition: scale Geography: geographies available for this indicator from other providers
Strategic Health Authority. http://www.ic.nhs.uk/webfiles/Popgeog/Direct%20Estimates%20%20of%20 Obesity%20(adults)%202003-2005.pdf
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Age, gender, ethnicity, social class http://www.dh.gov.uk/en/Publicationsandstatistics/PublishedSurvey/ HealthSurveyForEngland/index.htm
Data extraction: Source
Health Surveys for England, National Centre for Social Research (NatCen). Published by The Information Centre for Health and Social Care (IC), 2007
Data extraction: source URL
Data received directly from NatCen.
Data extraction: date
16 January 2008.
Numerator: definition
The number of persons aged 16+ who are self-reported smokers in a sample survey of the health of the population of England 2003–05.
Numerator: source
Health Survey for England (HSE), commissioned by the Department of Health and carried out by the Joint Health Survey Unit of Social and Community Planning Research and of the Department of Epidemiology and Public Health at University College, London.
Denominator: definition
Total number of respondents (with valid recorded smoking status) aged 16+ in the Health Survey for England 2003–2005.
Denominator: source
Health Survey for England (HSE), commissioned by the Department of Health and carried out by the Joint Health Survey Unit of Social and Community Planning Research and of the Department of Epidemiology and Public Health at University College, London.
Data for 4 new 2009 UAs (Bedford UA, Central Bedfordshire UA, Cheshire East UA and Cheshire West UA) were calculated by NatCen as a separate exercise on 21/01/2009 and received on the same date.
Health Profiles 2009 The Indicator Guide
98 Section 4: adult’s health and lifestyle
Data quality: Accuracy and completeness
The Health Survey for England is designed to provide data at both national and regional level about the population living in private households in England. It uses a clustered, stratified multi-stage sample design. In the 2005 HSE, for example, a random sample of over 7,200 addresses (around 16,000 people) were selected from the Postcode Address File (PAF) to ensure households were sampled proportionately across the nine Government Office Regions in England. 720 postcode sectors were selected and 26 addresses within each sector. Each individual within a selected household was eligible for inclusion. One of the effects of using a complex sample design is that standard errors for survey estimates are generally higher than would be derived from a simple random sample of the same size. There was a full adult sample of around 16,000 in the 2003 HSE. However, the 2004 and 2005 Health Surveys had only 8,000 adults in the normal ‘general population’ sample as these two surveys included boost samples. The 2004 HSE included a boost sample to increase the number of participants from minority ethnic groups and a special Chinese boost sample. The 2005 HSE included a boost sample for older people living in private households and for five months of the year a boost of children aged 2–15 was included. To ensure that each year’s sample was given an approximately equal weight in the calculation of the 3-year estimates (2003–2005) respondents in the boost sample years were weighted up by two. The numerator and denominator counts used to estimate prevalence are based on a sample of the population in each area and, as such, are not true counts. For this reason the numerator and denominator data are not shown in the data sheet.
Table 3 – Indicator Technical Methods Numerator: extraction
Not Applicable
Numerator: aggregation/ allocation
Residency by local authority of each respondent is allocated by postcode of residency.
Numerator data caveats
Questions about current cigarette smoking were asked by the interviewer. For those aged 16 and 17, the questions were asked through a self-completion questionnaire to allow for greater privacy. These data have not been age-standardised and, therefore, variation between area values may be a result of differences in population structure.
Denominator data caveats
The HSE is a series of annual surveys that began in 1991 with the aim of monitoring the health of the population. It was designed to be representative of the general, non-institutional population living in England. The current “full” sample size of the HSE comprises about 16,000 adults aged 16 and over. For each participant, the survey included an interview and a physical examination by a nurse, at which various physical measurements, tests, and samples of blood and saliva were collected. These measurements provided biomedical information about known risk factors associated with disease and objective validation for self-reported health behaviour.
Health Profiles 2009 The Indicator Guide
99 Section 4: adult’s health and lifestyle
Methods used to calculate indicator value
Estimates are based on pooling together three consecutive years of Health Survey for England data (2003-2005). The general population sample size in 2004 and 2005 was about half the sample size in 2001 owing to the sampling of specific population groups – namely, ethnic minority populations (2004) and older people living in private households (2005). To ensure that each year’s sample was given an approximately equal weight in the calculation of the 2003-2005 estimates, respondents in 2004 and 2005 were weighted up by two. Data for 4 new 2009 UAs (Bedford UA, Central Bedfordshire UA, Cheshire East UA and Cheshire West UA) were calculated by NatCen as a separate exercise on 21/01/2009 using the same method as described above. As geographical coverage for the remaining 6 new 2009 UAs (County Durham UA, Northumberland UA, Shropshire UA, Wiltshire UA, Cornwall UA) remains the same as their old County equivalents data for these areas has not been recalculated as part of this exercise.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
The Health Survey for England sample does not cover the Isles of Scilly and no HSE respondents over 2003–2005 were located in the City of London.
Disclosure Control
Not applicable.
Confidence Intervals calculation method
The standard errors, and 95% confidence intervals, have been calculated using STATA’s survey module (the svy:mean commands), further details can be obtained from Shaun Scholes at NatCen ([email protected]). One of the effects of using a complex design is that standard errors for survey estimates are generally higher than the standard errors that would be derived from a simple random sample of the same size.
Health Profiles 2009 The Indicator Guide
100 Section 4: adult’s health and lifestyle
13b ADULTS WHO SMOKE LOWER TIER INDICATOR Basic Information 1. What is being measured?
Estimated prevalence of adult smoking.
2. Why is it being measured?
To estimate the expected proportion of adult smokers in local authorities given the characteristics of local authority populations. Smoking prevalence is a direct measure of health care need i.e. the ability to benefit from tobacco control interventions.
3. How is this indicator actually defined?
Prevalence of smoking, percentage of resident population, adults, 2003–2005, persons.
4. Who does it measure?
Adults (aged 16 and over).
5. When does it measure it?
Updated as ad-hoc.
6. Will It measure absolute numbers or proportions?
Percentage of resident adult population aged 16 and over.
7. Where does the data actually come from?
Modelled by the National Centre for Social Research (NatCen). Published by The Information Centre for Health and Social Care (IC), 2007.
8. How accurate and complete will the data be?
These are modelled estimates based on national survey data. The model is non-aetiological (not based on known casual factors). The estimates do not take into account additional local factors that may impact on the true smoking prevalence rate in an area and may not match with local lifestyle survey results or modelled estimates which use known risk factors such as socio-economic status, age, gender and ethnicity.
9. Are there any caveats/warnings/ problems?
As these estimates are modelled they should be used and interpreted with caution (see above).
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
TABLE 1 – INDICATOR DESCRIPTION Information component
Pg 4 Health Summary – Indicator 13
Subject category/ domain(s)
Adults health and lifestyle
Indicator name (* Indicator title in health profile)
Estimated prevalence of adult smoking (*Adults who smoke)
Health Profiles 2009 The Indicator Guide
101 Section 4: adult’s health and lifestyle
PHO with lead responsibility
SEPHO
Date of PHO dataset creation
21/01/2009
Indicator definition
Prevalence of smoking, percentage of resident population, adults, 2003-2005, persons.
Geography
Local Authority: County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs. Exceptions: Data for the 9 new 2009 UAs (Bedford UA, Central Bedfordshire UA, Cheshire East UA, Cheshire West and Chester UA, County Durham, Northumberland UA, Shropshire UA, Wiltshire UA, Cornwall UA) are not based on modelled estimates but are calculated directly using results from the Health Surveys for England 2003–05.
Timeliness
Updated as ad-hoc.
Rationale: What this indicator purports to measure
Expected prevalence of adult smoking.
Rationale: Public Health Importance
Smoking is the most important cause of preventable ill health and premature mortality in the UK. It is linked to respiratory illness, cancer and coronary heart disease. Smoking not only affects the smoker; over 17,000 children under the age of five are admitted to hospital every year with illnesses resulting from passive smoking. A list of disease specific conditions attributable to smoking is published in The Smoking Epidemic in England, HDA, 2004 http://www.nice.org.uk/page. aspx?o=502811 Smoking is a modifiable lifestyle risk factor; effective tobacco control measures can reduce the prevalence of smoking in the population.
Rationale: Purpose behind the inclusion of the indicator
To estimate the expected proportion of adult smokers in local authorities given the characteristics of local authority populations.
Rationale: Policy relevance
Choosing Health: Making healthy choices easier. http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4094550
Smoking prevalence is a direct measure of health care need i.e. the ability to benefit from tobacco control interventions, including smoking cessation services.
Smoking Kills. A White Paper on Tobacco http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4006684 Tackling Health Inequalities: A Programme for Action http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4008268
Health Profiles 2009 The Indicator Guide
102 Section 4: adult’s health and lifestyle
Interpretation: What a high/low level of indicator value means
Given the characteristics of the local population, a high indicator value (red circle in health summary chart) represents a statistically significant higher (worse) level of expected estimated adult smoking prevalence when compared to the national value. A low indicator value (green circle in health summary chart) represents a statistically significant lower level (better) of expected estimated adult smoking prevalence when compared to the national value. However smoking at any prevalence level greater than 0 is undesirable, and therefore a low indicator value should not mean that PH action is not needed.
Interpretation: Potential for error due to type of measurement method
It is important that users note that these model-based estimates do not take account of any additional local factors that may impact on the true smoking prevalence rate in an area (e.g. local initiatives designed to reduce smoking). They will almost certainly not mirror precisely any available measures from local studies or surveys (although research by NatCen and others have shown that they tend to be related). The figures, therefore, cannot be used to monitor performance or change over time. The model used is a non-aetiological model i.e. is not based on known aetiological risk factors. This may lead to estimated smoking levels which are at odds with, for example, local lifestyle survey results or modelled estimates which use known co-variates (e.g. socio-economic status, age, gender and ethnicity) such as the smoking prevalence estimates modelled in the Health Poverty Index available at www.hpi.org.uk (see variables used in generation of model in calculation of indicator section below). There may also be a discrepancy between the modelled lower tier estimates (districts) and upper tier estimates (County geographies and above, plus new 2009 UA areas) which are based on actual Health Survey for England data. This may lead to inconsistencies between lower tier and county estimates for some areas as the datasets are derived using different methods.
Interpretation: Potential for error due to bias and confounding
These model based healthy lifestyle indicators are derived using the Health Survey for England data and are subject to both sampling and non-sampling error. Sampling errors arise solely as a result of drawing a sample rather than conducting a full survey of the population. Generally, the smaller the sample size the larger the variability in the estimates that one would expect to obtain from all the possible samples. Non-sampling errors arise during the course of the survey activities and there is no simple direct way of estimating the size of these errors. Non-sampling error may include e.g. respondent bias, interview bias and refusal to participate. The use of statistical models for prediction involves making assumptions about relationships in the data. The suitability of the chosen models for the given data and the validity of the model in describing real world dynamics have a bearing on the nature and magnitude of the errors introduced. A key source of modelling error arises from omitting variables that would otherwise help improve the model predictions either by error or because there is no available or reliable data source for them.
Health Profiles 2009 The Indicator Guide
103 Section 4: adult’s health and lifestyle
The model-based estimate generated for a particular area is the expected measure for that area based on its population characteristics – and not an estimate of the actual prevalence. In statistical terms, the model-based estimate is actually a biased estimate of the true value for the area and, as such, should be treated with caution. As mentioned above, the model-based estimates are unable to take account of any additional local factors that may impact on the true prevalence rate. To interpret the estimates, NatCen recommend that users adopt statements such as “given the characteristics of the local population we would expect approximately x% of adults within LA Y to be smokers”. Validation exercises were used to check the appropriateness of the chosen models. Confidence intervals have been calculated for the model-based estimates to capture both sampling and modelling error. The confidence intervals provide a range within which we can be fairly sure the ‘true’ value for that area lies. It is recommended that users look at the confidence interval for the estimates, not just the estimate. Estimates for two areas can only be described as significantly different if the confidence intervals for the estimates do not overlap. Users should also note that the potential sources of bias and error also apply to any ranking or banding of the small-area estimates. NatCen do not encourage any ranking of small area estimates within larger areas such as Local Authorities, Primary Care Organisations and Strategic Health Authorities. The 2003–2005 model based estimates are not comparable with the preceding estimates for 2000–2002 owing to differences in geography and modelling methodology: The 2000–2002 LA estimates were calculated by aggregating the model-based estimates for the component wards. The 2003–05 LA estimates have been calculated by modelling directly at the LA level. The choice of co-variate data was different as both the Index of Multiple deprivation 2004 and ONS area classifications were excluded in the 2003–05 estimates owing to their statistical relationship with other census-based covariates. The 2003–05 based estimates were adjusted to be consistent with the direct survey estimates at GOR/SHA level. The model-based estimates have been produced solely for LAs and cannot be translated onto any other geographical boundary system (i.e. aggregated or averaged over any other spatial unit). As a result of limitations in the model based estimates they are published as “experimental statistics”. This term is applied to any set of ONS statistics that do not meet the rigorous quality standards of National Statistics and/or may be subject to change due to methodological development.
Health Profiles 2009 The Indicator Guide
104 Section 4: adult’s health and lifestyle
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
TABLE 2 – INDICATOR SPECIFICATION Indicator definition: Variable
Prevalence of smoking.
Indicator definition: Statistic
Percentage
Indicator definition: Gender
Persons
Indicator definition: age group
Adults (aged 16 and over)
Indicator definition: period
2003–2005
Indicator definition: scale
Per resident adult population aged 16 and over
Smoking is defined as self-reported current cigarette smoking.
Health Profiles 2009 The Indicator Guide
105 Section 4: adult’s health and lifestyle
Geography: geographies available for this indicator from other providers
MSOA available from http://www.ic.nhs.uk/pubs/healthylifestyles05
Dimensions of inequality: subgroup analyses of this dataset available from other providers
None.
Data extraction: Source
National Centre for Social Research (NatCen).
Data extraction: source URL
Data received directly from NatCen. Also published by IC, December 2007 (see link above).
Data extraction: date
16/01/2008
Numerator: definition
Model estimates by NatCen using data from a number of sources including Health Survey for England 2003–2005, Census 2001.
Numerator: source
NatCen.
Denominator: definition
Not applicable.
Denominator: source
Not applicable.
Data quality: Accuracy and completeness
The model-based approach generates estimates that are of a different nature from standard survey estimates because they are dependent upon how well the relationship between healthy lifestyle behaviours for individuals and the Census/administrative information about the area in which they live are specified in the model. The accuracy and completeness of the information will be subject to the same constraints surrounding the Health Survey for England and Census data sets on which they are based (see Interpretation and potential sources of error section).
Table 3 – Indicator Technical Methods Numerator: extraction
Not applicable
Numerator: aggregation/ allocation
Not applicable
Numerator data caveats
See Interpretation: potential sources of error section.
Denominator data caveats
Not applicable.
Health Profiles 2009 The Indicator Guide
106 Section 4: adult’s health and lifestyle
Methods used to calculate indicator value
The process of creating the model-based estimates of healthy lifestyle behaviours involved three main stages: • A statistical model was used to represent the relationships between current smoking and area-level characteristics in the small areas covered by the HSfE. The 2001 Census provided the main source for demographic and social covariate data. Other routine sources of data providing area-level characteristics for LAs included all age-all cause mortality, diversity index, life expectancy, emergency hospital admissions, hospital admissions attributable to alcohol, job seekers allowance claimant counts and educational attainment. • The model outputs were applied in conjunction with covariate data (available for all LAs) to estimate the ‘expected’ prevalence given the characteristics of the area. • An adjustment factor was applied to ensure that the model-based estimates for each LA corresponded with the 2003–2005 direct estimates at GOR/SHA level taken from the Health Survey for England data. The area level characteristics associated with increased propensity for an adult to be a current smoker were: a higher proportion of residents of White ethnic origin and a higher proportion of residents aged 16–74 whose highest qualification was an NVQ Level 1 (or with no qualifications). The area-level characteristics associated with decreased propensity for an adult to be a current smoker were: a relatively higher proportion of Income Support claimants who were classified as “carers and others”; a higher proportion of residents who were unpaid carers; a higher proportion of household residents aged 16 or over who were living as a couple and a higher proportion of residents aged 70–74. The model-based estimates were constrained to the direct GOR/SHA estimates taken from the Health Survey for England data. This was done by aggregating the model-based estimates to GOR/SHA and comparing to the direct estimates. The relevant ratios of the HSfE direct estimates to the aggregated model-based estimates at GOR/SHA level were then used to scale the model based LA level estimates. However, the model-based estimates have not been calibrated to ensure agreement at County level. As a result, there may be inconsistencies between lower tier and county estimates for some areas as the datasets are derived using different methods. For a fuller technical description of the methodology see the Model-Based Estimates User Guide and other reports available on the Information Centre website: www.ic.nhs.uk/webfiles/Popgeog/Healthy%20Lifestyle%20Behaviours-%20 Model%20Based%20Estimates%20for%20Middle%20Layer%20 Super%20Output%20Areas%20and%20Local%20Authorities%20in%20England_2003-2005__%20User%20Guide.pdf For methods used to calculate data for new 2009 UAs please see metadata for upper tier geographies.
Health Profiles 2009 The Indicator Guide
107 Section 4: adult’s health and lifestyle
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Model based estimates were not produced for Isles of Scilly or City of London.
Disclosure Control
Not applicable.
Confidence Intervals calculation method
The model-based estimate generated for a particular LA is the expected measure for that LA based on its characteristics as measured by the covariates in the model. In statistical terms, the model-based estimate is actually a biased estimate of the true value for an area and, as such, should be treated with caution. By placing confidence intervals around a model-based estimate, however, we can generate a range within which we can be fairly sure the ‘true’ value for that area lies. In order to generate the confidence interval, an estimate of the variance of the difference between the model-based estimate and the true LA measure is required. The estimate of the variance has two components which correspond to the variance that is not explained by the model (and hence is not predicted by the model-based estimate) and the uncertainty of the model-based estimate itself. Obtaining the confidence interval for the LA estimates required computing two area-level variance terms: one for the Primary Sampling Units and one for the LA. The first term was required to allow for the clustering in the sample and the second term to estimate the residual variance at the LA level that was not explained by the model. The confidence interval for the model-based LA estimates was estimated to be: βˆ T X i T e −1 log it αˆ + βˆ X i ± 1.96 T βˆ X i 1+ e
2
M i N ij 2 2 T 2 ˆ ∑ ˆ ˆ σ σ X Var ( β ) X + + i i v N i u j =1
1
where: Χi is the vector of covariates values for LA i,
Xi
βˆ
βˆ is the vector of parameter estimates for the LA-level covariates,
σˆ 2 is the estimate of the between PSU-level variance,
σˆ u2u
σˆ 2 is the estimate of the between LA-level variance.
σˆ v2v
N is the population estimate of the number of adults in LA i,
Ni i
Nijij is the population estimate of the number of adults in postcode sector N M i and
Mi is the number of postcode sectors in LA i.
ij,
2
Health Profiles 2009 The Indicator Guide
108 Section 4: adult’s health and lifestyle
14a BINGE DRINKING ADULTS UPPER TIER INDICATOR Basic Information 1. What is being measured?
Prevalence of adult binge drinking.
2. Why is it being measured?
To estimate the proportion of binge drinking adults in local authorities. To help reduce the prevalence of excessive alcohol consumption and the health risks associated with single episodes of intoxication.
3. How is this indicator actually defined?
Prevalence of binge drinking, percentage of resident population, adults, 2003–2005, persons.
4. Who does it measure?
Adults (aged 16 and over).
5. When does it measure it?
The Health Survey for England (HSE) is carried out annually.
6. Will It measure absolute numbers or proportions?
Percentage of resident adult population aged 16 and over.
7. Where does the data actually come from?
Health Surveys for England, National Centre for Social Research (NatCen). Published by The Information Centre for Health and Social Care (IC), 2007.
8. How accurate and complete will the data be?
The HSE was designed to be representative of the general, non-institutional population living in England. The current “full” sample size of the HSE comprises about 16,000 adults aged 16 and over.
9. Are there any caveats/warnings/ problems?
HSE numerator data are based on observed self-reported drinking behaviour. Self-reported consumption may be prone to respondent bias. The Health Survey for England under-samples younger people, people in employment, ethnic minorities, women, those who are healthier but exhibit less healthy behaviour.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
County Health Profiles: Pg 4 Health Summary – Indicator 14
Subject category/ domain(s)
Adults health and lifestyle
Indicator name (*Indicator title in health profile)
Prevalence of adults who binge drink (*Binge drinking adults).
Health Profiles 2009 The Indicator Guide
109 Section 4: adult’s health and lifestyle
PHO with lead responsibility
SEPHO
Date of PHO dataset creation
21/01/2009
Indicator definition
Prevalence of binge drinking, percentage of resident population, adults, 20032005, persons.
Geography
England, GOR, County, South East SHAs, new 2009 UAs (Bedford UA, Central Bedfordshire UA, Cheshire East UA, Cheshire West and Chester UA, County Durham UA, Northumberland UA, Shropshire UA, Wiltshire UA, Cornwall UA).
Timeliness
Updated annually.
Rationale: What this indicator purports to measure
Prevalence of adult binge drinking.
Rationale: Public Health Importance
Harmful drinking is a significant public health problem in the UK and is associated with a wide range of health problems, including brain damage, alcohol poisoning, chronic liver disease, breast cancer, skeletal muscle damage, mental ill-health and social problems. Alcohol plays a role in many accidents, acts of violence and other instances of criminal behaviour. Nationally between 780,000 and 1.3 million children are affected by their parents’ alcohol misuse. Such children are four times more likely to suffer from a psychiatric disorder by the age of 15 than the national average and are at increased risk of aggressive behaviour, delinquency, hyperactivity and other forms of conduct disorder. There are particular risks associated with drink-driving, alcohol consumption in the workplace or during the working day and drinking during pregnancy. Alcohol-related problems contribute to social and health inequalities, and reducing harmful drinking is one important element in the broad policy thrust to reduce health inequalities following the recommendations of the Acheson Report (1998). For some people, binge drinking is an occasional event. For others, it is part of a chronic drinking pattern Binge drinking and severe intoxication can cause muscular in-coordination, blurred vision, stupor, hypothermia, convulsions, depressed reflexes, respiratory depression, hypotension and coma. Death can occur from respiratory or circulatory failure or if binge drinkers inhale their own vomit. It is well known that binge drinkers are at increased risk of accidents and alcohol poisoning. A growing body of research suggests that binge drinkers also have a higher all-cause mortality rate than those who have the same average alcohol consumption but drink more frequently. Binge drinking is specifically related to accidents and violence, both of which impact on the health service. The Strategy Unit has estimated that, at peak times, up to 70 per cent of all admissions to accident and emergency units are related to alcohol consumption. The total cost of alcohol misuse to the health service is estimated to be in the region of £1.7 billion a year. Effective interventions to reduce alcohol consumption and alcohol related harm exist. The evidence suggests that multi-sectoral, multi-faceted interventions work best. These include: • population level measures such as restricting the availability and price of alcohol, drink driving legislation and taxation; • school based alcohol education programmes; • brief interventions in a variety of settings, such as primary health care, workbased training programmes etc.
Health Profiles 2009 The Indicator Guide
110 Section 4: adult’s health and lifestyle
For further information please see Choosing Health in the South East: Alcohol p35 for summary list of effective interventions. http://www.sepho.org.uk/ Download/Public/10571/1/sepho%20alcohol%20report%20Jan%2007.pdf Rationale: Purpose behind the inclusion of the indicator
To estimate the proportion of binge drinking adults in local authorities.
Rationale: Policy relevance
Choosing Health: Making healthy choices easier. http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4094550
To help reduce the prevalence of excessive alcohol consumption and the health risks associated with single episodes of intoxication.
Alcohol Harm Reduction Strategy for England (2004) http://www.cabinetoffice.gov.uk/upload/assets/www.cabinetoffice.gov.uk/ strategy/caboffce%20alcoholhar.pdf Interpretation: What a high/low level of indicator value means
A high indicator value (red circle in health summary chart) represents a statistically significant higher level (worse) of estimated adult binge drinking prevalence when compared to the national value. A low indicator value (green circle in health summary chart) represents a statistically significant lower (better) level of estimated adult binge drinking prevalence when compared to the national value. However binge drinking at any prevalence level greater than 0 is undesirable, and therefore a low indicator value should not mean that PH action is not needed.
Interpretation: Potential for error due to type of measurement method
HSE numerator data are broadly based on observed self-reported drinking behaviour. Men were defined as having indulged in binge drinking if they had consumed 8 or more units of alcohol on the heaviest drinking day in the previous seven days; for women the cut-off was 6 or more units of alcohol. Self-reported consumption may be prone to respondent bias. In order to ensure agreement between these estimates at GOR and SHA level, the lower tier synthetic estimates (districts), which are based on modelled data, have been calibrated. However, the synthetic estimates have not been calibrated to ensure agreement at County level. As a result, there may be inconsistencies between lower tier and county estimates for some areas as the datasets are derived using different methods.
Interpretation: Potential for error due to bias and confounding
The Health Survey for England under-samples younger people, people in employment, ethnic minorities, women, those who are healthier but exhibit less healthy behaviour. These data have not been age-standardised and, therefore, variation between area values may be a result of differences in population structure.
Health Profiles 2009 The Indicator Guide
111 Section 4: adult’s health and lifestyle
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Table 2 – Indicator Specification Indicator definition: Variable
Prevalence of binge drinking. Binge drinking in adults is defined separately for men and women. Men are defined as having indulged in binge drinking if they had consumed 8 or more units of alcohol on the heaviest drinking day in the previous seven days; for women the cut-off was 6 or more units of alcohol.
Indicator definition: Statistic
Percentage of resident adult population aged 16 and over
Indicator definition: Gender
Persons
Indicator definition: age group
Adults (aged 16 and over)
Indicator definition: period
2003–2005
Indicator definition: scale
Health Profiles 2009 The Indicator Guide
112 Section 4: adult’s health and lifestyle
Geography: geographies available for this indicator from other providers
Strategic Health Authority. http://www.ic.nhs.uk/webfiles/Popgeog/Direct%20Estimates%20%20of%20 Obesity%20(adults)%202003-2005.pdf
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Age, gender, ethnicity, social class http://www.dh.gov.uk/en/Publicationsandstatistics/PublishedSurvey/ HealthSurveyForEngland/index.htm
Data extraction: Source
Health Surveys for England, National Centre for Social Research (NatCen). Published by The Information Centre for Health and Social Care (IC), 2007
Data extraction: source URL
Data received directly from NatCen.
Data extraction: date
16 January 2008. Data for 4 new 2009 UAs (Bedford UA, Central Bedfordshire UA, Cheshire East UA and Cheshire West UA) were calculated by NatCen as a separate exercise on 21/01/2009 and received on the same date.
Numerator: definition
Proportion of adult men who drank 8 or more units of alcohol on the heaviest drinking day in the previous seven days at time of survey and adult women who drank 6 or more units of alcohol on the heaviest drinking day in the previous seven days at time of survey, 2003–2005.
Numerator: source
Health Survey for England (HSE), commissioned by the Department of Health and carried out by the Joint Health Survey Unit of Social and Community Planning Research and of the Department of Epidemiology and Public Health at University College, London.
Denominator: definition
Total number of respondents (with valid measurements on drinking habits in the last week) aged 16+ in the Health Survey for England 2003–2005.
Denominator: source
Health Survey for England (HSE), commissioned by the Department of Health/ IC and carried out by the Joint Health Surveys Unit of the National Centre for Social Research (NatCen) and the Department of Epidemiology and Public Health at the Royal Free and University College Medical School, London.
Health Profiles 2009 The Indicator Guide
113 Section 4: adult’s health and lifestyle
Data quality: Accuracy and completeness
The Health Survey for England is designed to provide data at both national and regional level about the population living in private households in England. It uses a clustered, stratified multi-stage sample design. In the 2005 HSE, for example, a random sample of over 7,200 addresses (around 16,000 people) were selected from the Postcode Address File (PAF) to ensure households were sampled proportionately across the nine Government Office Regions in England. 720 postcode sectors were selected and 26 addresses within each sector. Each individual within a selected household was eligible for inclusion. One of the effects of using a complex sample design is that standard errors for survey estimates are generally higher than would be derived from a simple random sample of the same size. There was a full adult sample of around 16,000 in the 2003 HSE. However, the 2004 and 2005 Health Surveys had only 8,000 adults in the normal ‘general population’ sample as these two surveys included boost samples. The 2004 HSE included a boost sample to increase the number of participants from minority ethnic groups and a special Chinese boost sample. The 2005 HSE included a boost sample for older people living in private households and for five months of the year a boost of children aged 2–15 was included. To ensure that each year’s sample was given an approximately equal weight in the calculation of the 3-year estimates (2003–2005) respondents in the boost sample years were weighted up by two. The numerator and denominator counts used to estimate prevalence are based on a sample of the population in each area and, as such, are not true counts. For this reason the numerator and denominator data are not shown in the data sheet.
Table 3 – Indicator Technical Methods Numerator: extraction
Not applicable.
Numerator: aggregation/ allocation
Residency by local authority of each respondent is allocated by postcode of residency.
Numerator data caveats
These data have not been age-standardised and, therefore, variation between area values may be a result of differences in population structure. The HSE question module concerning “mean weekly alcohol consumption” aims to classify respondents into broad consumption bands based on “usual” behaviour, rather than offer a precise estimate of actual weekly consumption. Adults were first asked how often they drank each of five types of alcoholic drinks (e.g. beer, spirits, wine) in the last 12 months, and how much of each type they had usually drank in one day. From these two sets of questions, an estimated weekly consumption expressed in terms of units of alcohol was derived. Over the years the list of drinks included in the survey has changed to reflect the emergence of new brands and types of drinks. Respondents for whom any information on drinking was not answered, refused or not known were excluded.
Health Profiles 2009 The Indicator Guide
114 Section 4: adult’s health and lifestyle
Denominator data caveats
The HSE is a series of annual surveys that began in 1991 with the aim of monitoring the health of the population. It was designed to be representative of the general, non-institutional population living in England. The current “full” sample size of the HSE comprises about 16,000 adults aged 16 and over. For each participant, the survey included an interview and a physical examination by a nurse, at which various physical measurements, tests, and samples of blood and saliva were collected. These measurements provided biomedical information about known risk factors associated with disease and objective validation for self-reported health behaviour.
Methods used to calculate indicator value
Estimates are based on pooling together three consecutive years of Health Survey for England data (2003–2005). The general population sample size in 2004 and 2005 was about half the sample size in 2001 owing to the sampling of specific population groups – namely, ethnic minority populations (2004) and older people living in private households (2005). To ensure that each year’s sample was given an approximately equal weight in the calculation of the 2003–2005 estimates, respondents in 2004 and 2005 were weighted up by two. Data for 4 new 2009 UAs (Bedford UA, Central Bedfordshire UA, Cheshire East UA and Cheshire West UA) were calculated by NatCen as a separate exercise on 21/01/2009 using the same method as described above. As geographical coverage for the remaining 6 new 2009 UAs (County Durham UA, Northumberland UA, Shropshire UA, Wiltshire UA, Cornwall UA) remains the same as their old County equivalents, data for these areas has not been recalculated as part of this exercise.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
The Health Survey for England sample does not cover the Isles of Scilly and no HSE respondents over 2003–2005 were located in the City of London.
Disclosure Control
Not applicable.
Confidence Intervals calculation method
The standard errors, and 95% confidence intervals, have been calculated using STATA’s survey module (the svy:mean commands), further details can be obtained from Shaun Scholes at NatCen ([email protected]). One of the effects of using a complex design is that standard errors for survey estimates are generally higher than the standard errors that would be derived from a simple random sample of the same size.
Health Profiles 2009 The Indicator Guide
115 Section 4: adult’s health and lifestyle
14b BINGE DRINKING ADULTS LOWER TIER INDICATOR Basic Information 1. What is being measured?
Prevalence of adult binge drinking.
2. Why is it being measured?
To estimate the proportion of binge drinking adults in local authorities. To help reduce the prevalence of excessive alcohol consumption and the health risks associated with single episodes of intoxication.
3. How is this indicator actually defined?
Prevalence of binge drinking, percentage of resident population, adults, 2003–2005, persons.
4. Who does it measure?
Adults (aged 16 and over).
5. When does it measure it?
Updated as ad-hoc.
6. Will It measure absolute numbers or proportions?
Percentage of resident adult population aged 16 and over.
7. Where does the data actually come from?
Modelled by the National Centre for Social Research (NatCen). Published by The Information Centre for Health and Social Care (IC), 2007.
8. How accurate and complete will the data be?
These are modelled estimates based on national survey data. The model is non-aetiological (not based on known casual factors). The estimates do not take into account additional local factors that may impact on the true prevalence of binge drinking in an area and may not match with local lifestyle survey results or modelled estimates which use known risk factors such as socio-economic status, age, gender and ethnicity.
9. Are there any caveats/warnings/ problems?
As these estimates are modelled they should be used and interpreted with caution (see above).
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
Pg 4 Health Summary – Indicator No 14
Subject category/ domain(s)
Adults health and lifestyle
Indicator name (*Indicator title in health profile)
Estimated prevalence of adults who binge drink (*Binge drinking adults).
PHO with lead responsibility
SEPHO
Health Profiles 2009 The Indicator Guide
116 Section 4: adult’s health and lifestyle
Date of PHO dataset creation
21/01/2009
Indicator definition
Prevalence of binge drinking, percentage of resident population, adults, 20032005, persons.
Geography
Local Authority: County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs. Exceptions: Data for the 9 new 2009 UAs (Bedford UA, Central Bedfordshire UA, Cheshire East UA, Cheshire West and Chester UA, County Durham, Northumberland UA, Shropshire UA, Wiltshire UA, Cornwall UA) are not based on modelled estimates but are calculated directly using results from the Health Surveys for England 2003–05.
Timeliness
Updated as ad-hoc.
Rationale: What this indicator purports to measure
Expected prevalence of adult binge drinking.
Health Profiles 2009 The Indicator Guide
117 Section 4: adult’s health and lifestyle
Rationale: Public Health Importance
Harmful drinking is a significant public health problem in the UK and is associated with a wide range of health problems, including brain damage, alcohol poisoning, chronic liver disease, breast cancer, skeletal muscle damage, mental ill-health and social problems. Alcohol plays a role in many accidents, acts of violence and other instances of criminal behaviour. Nationally between 780,000 and 1.3 million children are affected by their parents’ alcohol misuse. Such children are four times more likely to suffer from a psychiatric disorder by the age of 15 than the national average and are at increased risk of aggressive behaviour, delinquency, hyperactivity and other forms of conduct disorder. There are particular risks associated with drink-driving, alcohol consumption in the workplace or during the working day and drinking during pregnancy. Alcohol-related problems contribute to social and health inequalities, and reducing harmful drinking is one important element in the broad policy thrust to reduce health inequalities following the recommendations of the Acheson Report (1998). For some people, binge drinking is an occasional event. For others, it is part of a chronic drinking pattern Binge drinking and severe intoxication can cause muscular in-coordination, blurred vision, stupor, hypothermia, convulsions, depressed reflexes, respiratory depression, hypotension and coma. Death can occur from respiratory or circulatory failure or if binge drinkers inhale their own vomit. It is well known that binge drinkers are at increased risk of accidents and alcohol poisoning. A growing body of research suggests that binge drinkers also have a higher all-cause mortality rate than those who have the same average alcohol consumption but drink more frequently Binge drinking is specifically related to accidents and violence, both of which impact on the health service. The Strategy Unit has estimated that, at peak times, up to 70 per cent of all admissions to accident and emergency units are related to alcohol consumption. The total cost of alcohol misuse to the health service is estimated to be in the region of £1.7 billion a year. Effective interventions to reduce alcohol consumption and alcohol related harm exist. The evidence suggests that multi-sectoral, multi-faceted interventions work best. These include: • population level measures such as restricting the availability and price of alcohol, drink driving legislation and taxation; • school based alcohol education programmes; • brief interventions in a variety of settings, such as primary health care, workbased training programmes etc. For further information please see Choosing Health in the South East: Alcohol p35 for summary list of effective interventions. http://www.sepho.org.uk/ Download/Public/10571/1/sepho%20alcohol%20report%20Jan%2007.pdf
Rationale: Purpose behind the inclusion of the indicator
To estimate the expected proportion of binge drinking adults in local authorities given the characteristics of local authority populations.
Rationale: Policy relevance
Choosing Health: Making healthy choices easier. http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4094550
To help reduce the prevalence of excessive alcohol consumption and the health risks associated with single episodes of intoxication.
Alcohol Harm Reduction Strategy for England (2004) http://www.cabinetoffice.gov.uk/upload/assets/www.cabinetoffice.gov.uk/ strategy/caboffce%20alcoholhar.pdf
Health Profiles 2009 The Indicator Guide
118 Section 4: adult’s health and lifestyle
Interpretation: What a high/low level of indicator value means
Given the characteristics of the local population, a high indicator value (red circle in health summary chart) represents a statistically significant higher (worse) level of expected estimated binge drinking prevalence when compared to the national value. A low indicator value (green circle in health summary chart) represents a statistically significant lower level (better) of expected estimated adult binge drinking prevalence when compared to the national value. However binge drinking at any prevalence level greater than 0 is undesirable, and therefore a low indicator value should not mean that PH action is not needed.
Interpretation: Potential for error due to type of measurement method
It is important that users note that these model-based do not take account of any additional local factors that may impact on the true binge drinking prevalence rate in an area (e.g. local initiatives designed to reduce alcohol consumption). They will almost certainly not mirror precisely any available measures from local studies or surveys (although research by NatCen and others have shown that they tend to be related). The figures, therefore, cannot be used to monitor performance or change over time. The model used is a non-aetiological model i.e. is not based on risk factors such as physical activity levels and calorie intake. This may lead to estimated binge drinking levels which are at odds with other risk factors estimates such as healthy eating and physical activity; local lifestyle survey results and modelled estimates which use known co-variates (see variables used in generation of model in calculation of indicator section below). There may also be a discrepancy between the modelled lower tier estimates (districts) and upper tier estimates (County geographies and above, plus new 2009 UA areas) which are based on actual Health Survey for England data. This has led to inconsistencies between lower tier and county estimates for some areas as the datasets are derived using different methods. It should be noted that the North West Public Health Observatory (NWPHO) have recently generated LA estimates of hazardous and harmful drinking using Health Survey for England data. These estimates are not comparable with the NatCen modelled estimates owing to differences in time period, choice of alcohol indicator, modelling methodology and covariate data. See http://www. nwph.net/alcohol/lape/ for the NWPHO profiles and further information.
Interpretation: Potential for error due to bias and confounding
These model based healthy lifestyle indicators are derived using the Health Survey for England data and are subject to both sampling and non-sampling error. Sampling errors arise solely as a result of drawing a sample rather than conducting a full survey of the population. Generally, the smaller the sample size the larger the variability in the estimates that one would expect to obtain from all the possible samples.
Health Profiles 2009 The Indicator Guide
119 Section 4: adult’s health and lifestyle
Non-sampling errors arise during the course of the survey activities and there is no simple direct way of estimating the size of these errors. Non sampling error may include e.g. respondent bias, interview bias and refusal to participate. The use of statistical models for prediction involves making assumptions about relationships in the data. The suitability of the chosen models for the given data and the validity of the model in describing real world dynamics have a bearing on the nature and magnitude of the errors introduced. A key source of modelling error arises from omitting variables that would otherwise help improve the model predictions either by error or because there is no available or reliable data source for them. The model-based estimate generated for a particular area is the expected measure for that area based on its population characteristics - and not an estimate of the actual prevalence. In statistical terms, the model-based estimate is actually a biased estimate of the true value for the area and, as such, should be treated with caution. As mentioned above, the model-based estimates are unable to take account of any additional local factors that may impact on the true prevalence rate. To interpret the estimates, NatCen recommend that users adopt statements such as “given the characteristics of the local population we would expect approximately x% of adults within LA Y to indulge in binge drinking”. Validation exercises were used to check the appropriateness of the chosen models. Confidence intervals have been calculated for the model-based estimates to capture both sampling and modelling error. The confidence intervals provide a range within which we can be fairly sure the ‘true’ value for that area lies. It is recommended that users look at the confidence interval for the estimates, not just the estimate. Estimates for two areas can only be described as significantly different if the confidence intervals for the estimates do not overlap. Users should also note that the potential sources of bias and error also apply to any ranking or banding of the small-area estimates. NatCen do not encourage any ranking of small area estimates within larger areas such as Local Authorities, Primary Care Organisations and Strategic Health Authorities. • The 2003–2005 model based estimates are not comparable with the preceding estimates for 2000–2002 owing to differences in geography and modelling methodology: • The 2000–2002 LA estimates were calculated by aggregating the modelbased estimates for the component wards. The 2003–05 LA estimates have been calculated by modelling directly at the LA level. • The choice of co-variate data was different as both the Index of Multiple deprivation 2004 and ONS area classifications were excluded in the 2003– 05 estimates owing to their statistical relationship with other census-based covariates. • The 2003–05 based estimates were adjusted to be consistent with the direct survey estimates at GOR/SHA level. The model-based estimates have been produced solely for LAs and cannot be translated onto any other geographical boundary system (i.e. aggregated or averaged over any other spatial unit). As a result of limitations in the model based estimates they are published as “experimental statistics”. This term is applied to any set of ONS statistics that do not meet the rigorous quality standards of National Statistics and/or may be subject to change due to methodological development.
Health Profiles 2009 The Indicator Guide
120 Section 4: adult’s health and lifestyle
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
TABLE 2 – INDICATOR SPECIFICATION Indicator definition: Variable
Prevalence of binge drinking. Binge drinking in adults is defined separately for men and women. Men are defined as having indulged in binge drinking if they had consumed 8 or more units of alcohol on the heaviest drinking day in the previous seven days; for women the cut-off was 6 or more units of alcohol.
Indicator definition: Statistic
Percentage of resident adult population aged 16 and over
Indicator definition: Gender
Persons
Indicator definition: age group
Adults (aged 16 and over)
Indicator definition: period
2003–2005
Indicator definition: scale
Health Profiles 2009 The Indicator Guide
121 Section 4: adult’s health and lifestyle
Geography: geographies available for this indicator from other providers
MSOA available from http://www.ic.nhs.uk/pubs/healthylifestyles05
Dimensions of inequality: subgroup analyses of this dataset available from other providers
None.
Data extraction: Source
National Centre for Social Research (NatCen).
Data extraction: source URL
Data received directly from NatCen. Also published by IC, December 2007 (see link above).
Data extraction: date
16/01/2008
Numerator: definition
Model estimates by NatCen using data from a number of sources including Health Survey for England 2003–2005, Census 2001.
Numerator: source
NatCen.
Denominator: definition
Not applicable.
Denominator: source
Not applicable.
Data quality: Accuracy and completeness
The model-based approach generates estimates that are of a different nature from standard survey estimates because they are dependent upon how well the relationship between healthy lifestyle behaviours for individuals and the Census/administrative information about the area in which they live are specified in the model. The accuracy and completeness of the information will be subject to the same constraints surrounding the Health Survey for England and Census data sets on which they are based (see Interpretation: potential sources of error section).
TABLE 3 – INDICATOR TECHNICAL METHODS Numerator: extraction
Not applicable.
Numerator: aggregation/ allocation
Not applicable.
Numerator data caveats
See Interpretation: potential sources of error section.
Denominator data caveats
Not applicable.
Health Profiles 2009 The Indicator Guide
122 Section 4: adult’s health and lifestyle
Methods used to calculate indicator value
The process of creating the model-based estimates of healthy lifestyle behaviours involved three main stages: • A statistical model was used to represent the relationships between binge drinking and area-level characteristics in the small areas covered by the HSfE. The 2001 Census provided the main source for demographic and social covariate data. Other routine sources of data providing area-level characteristics for LAs included all age-all cause mortality, diversity index, life expectancy, emergency hospital admissions, hospital admissions attributable to alcohol, job seekers allowance claimant counts and educational attainment. • The model outputs were applied in conjunction with covariate data (available for all LAs) to estimate the ‘expected’ prevalence given the characteristics of the area. • An adjustment factor was applied to ensure that the model-based estimates for each LA corresponded with the 2003–2005 direct estimates at GOR/SHA level taken from the Health Survey for England data. The model contained an interaction between the percentage of residents aged 20–24 and the percentage of households headed by a single-parent. An interaction term in a statistical model arises were the effect of a covariate on a healthy lifestyle behaviour depends on the values of another covariate. The area level characteristics associated with increased propensity for an adult to indulge in binge drinking were: a higher proportion of residents aged 25– 29; a higher proportion of residents aged 20–24; a higher proportion of male residents aged 55–59 and a higher proportion of residents of White ethnic origin. The area-level characteristics associated with decreased propensity for an adult to indulge in binge drinking were: a higher proportion of female residents aged 55–59; and a higher proportion of residents residing in areas classified as 1) Town and Fringe or 2) Village, Hamlet or Isolated Dwelling (relative to living in urban areas containing more than 10,000 residents). The model-based estimates were constrained to the direct GOR/SHA estimates taken from the Health Survey for England data. This was done by aggregating the model-based estimates to GOR/SHA and comparing to the direct estimates. The relevant ratios of the HSfE direct estimates to the aggregated model-based estimates at GOR/SHA level were then used to scale the model based LA level estimates. However, the model-based estimates have not been calibrated to ensure agreement at County level. As a result, there may be inconsistencies between lower tier and county estimates for some areas as the datasets are derived using different methods. For a fuller technical description of the methodology see the Model-Based Estimates User Guide and other reports available on the Information Centre website: http://www.ic.nhs.uk/webfiles/Popgeog/Healthy%20Lifestyle%20 Behaviours-%20Model%20Based%20Estimates%20for%20Middle%20 Layer%20Super%20Output%20Areas%20and%20Local%20Authorities%20 in%20England_2003-2005__%20User%20Guide.pdf For methods used to calculate data for new 2009 UAs please see metadata for upper tier geographies.
Health Profiles 2009 The Indicator Guide
123 Section 4: adult’s health and lifestyle
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Model based estimates were not produced for Isles of Scilly or City of London.
Disclosure Control
Not applicable.
Confidence Intervals calculation method
The model-based estimate generated for a particular LA is the expected measure for that LA based on its characteristics as measured by the covariates in the model. In statistical terms, the model-based estimate is actually a biased estimate of the true value for an area and, as such, should be treated with caution. By placing confidence intervals around a model-based estimate, however, we can generate a range within which we can be fairly sure the ‘true’ value for that area lies. In order to generate the confidence interval, an estimate of the variance of the difference between the model-based estimate and the true LA measure is required. The estimate of the variance has two components which correspond to the variance that is not explained by the model (and hence is not predicted by the model-based estimate) and the uncertainty of the model-based estimate itself. Obtaining the confidence interval for the LA estimates required computing two area-level variance terms: one for the Primary Sampling Units and one for the LA. The first term was required to allow for the clustering in the sample and the second term to estimate the residual variance at the LA level that was not explained by the model. The confidence interval for the model-based LA estimates was estimated to be:
βˆ T X i T e −1 ˆ ˆ log it α + β X i ± 1.96 T βˆ X i 1+ e
2
N ij 2 T 2 ˆ ∑ σˆ + σˆ v + X i Var ( β ) X i N i u j =1 Mi
2
1
where: Χi is the vector of covariates values for LA i,
Xi
β
βˆ is the vector of parameter estimates for the LA-level covariates, ˆ
σˆ2u2 is the estimate of the between PSU-level variance,
σˆ u
σˆ 2 is the estimate of the between LA-level variance.
σˆ v2v
N is the population estimate of the number of adults in LA i,
Ni i
Nijij is the population estimate of the number of adults in postcode sector N M i and
Mi is the number of postcode sectors in LA i.
ij,
2
Health Profiles 2009 The Indicator Guide
124 Section 4: adult’s health and lifestyle
15a HEALTHY EATING ADULTS UPPER TIER INDICATOR Basic Information 1. What is being measured?
Prevalence of adult healthy eating, that is adults who consume 5 or more portions of fruit and vegetables per day.
2. Why is it being measured?
To estimate the proportion of healthy eating adults. There is evidence that eating at least 5 portions of a variety of fruit and vegetables a day could reduce the risk of deaths from chronic diseases such as heart disease, stroke, and cancer by up to 20%.
3. How is this indicator actually defined?
Prevalence of healthy eating, percentage of resident population, adults, 2003–2005, persons.
4. Who does it measure?
Adults (aged 16 and over).
5. When does it measure it?
The Health Survey for England (HSE) is carried out annually.
6. Will It measure absolute numbers or proportions?
Percentage of resident adult population aged 16 and over.
7. Where does the data actually come from?
Health Surveys for England, National Centre for Social Research (NatCen). Published by The Information Centre for Health and Social Care (IC), 2007.
8. How accurate and complete will the data be?
The HSE was designed to be representative of the general, non-institutional population living in England. The current “full” sample size of the HSE comprises about 16,000 adults aged 16 and over.
9. Are there any caveats/warnings/ problems?
HSE numerator data are broadly based on observed self-reported daily consumption of fruit and vegetables. Self-reported consumption may be prone to respondent bias. The Health Survey for England under-samples younger people, people in employment, ethnic minorities, women, those who are healthier but exhibit less healthy behaviour.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Health Profiles 2009 The Indicator Guide
125 Section 4: adult’s health and lifestyle
Table 1 – Indicator Description Information component
County Health Profiles: Pg 4 Health Summary – Indicator 15
Subject category/ domain(s)
Adults health and lifestyle
Indicator name (*Indicator title in health profile)
Prevalence of adults who eat healthily (*Healthy eating adults).
PHO with lead responsibility
SEPHO
Date of PHO dataset creation
21/01/2008
Indicator definition
Prevalence of healthy eating, percentage of resident population, adults, 20032005, persons.
Geography
England, GOR, County, South East SHAs, new 2009 UAs (Bedford UA, Central Bedfordshire UA, Cheshire East UA, Cheshire West and Chester UA, County Durham, Northumberland UA, Shropshire UA, Wiltshire UA, Cornwall UA).
Timeliness
Updated annually.
Rationale: What this indicator purports to measure
Prevalence of adult healthy eating, that is, adults who consume 5 or more portions of fruit and vegetables per day.
Rationale: Public Health Importance
The indicator is a measure of a protective lifestyle factor. A diet rich in fruit and vegetables confers protective effects against the development of heart disease and certain cancers. It has been estimated that eating at least 5 portions of a variety of fruit and vegetables a day could reduce the risk of deaths from chronic diseases such as heart disease, stroke, and cancer by up to 20%. It has been estimated that diet might contribute to the development of onethird of all cancers, and that increasing fruit and vegetable consumption is the second most important cancer prevention strategy, after reducing smoking. In 1998, the Department of Health’s Committee on Medical Aspects of Food Policy and Nutrition reviewed the evidence and concluded that higher vegetable consumption would reduce the risk of colorectal cancer and gastric cancer. There was also weakly consistent evidence that higher fruit and vegetable consumption would reduce the risk of breast cancer. These cancers combined represent about 18% of the cancer burden in men and about 30% in women. Higher consumption of fruit and vegetables also reduces the risk of coronary heart disease and stroke. A recent study found that each increase of 1 portion of fruit and vegetables a day lowered the risk of coronary heart disease by 4% and the risk of stroke by 6%. Evidence also suggests an increase in fruit and vegetable intake can help lower blood pressure. Research suggests that there are other health benefits, including delaying the development of cataracts, reducing the symptoms of asthma, improving bowel function, and helping to manage diabetes. As well as the direct health benefits, eating fruit and vegetables can help to achieve other dietary goals including increasing fibre intake, reducing fat intake, help maintain a healthy weight, and substituting for foods with added sugars (as frequent consumption of foods with added sugars can contribute to tooth decay).
Health Profiles 2009 The Indicator Guide
126 Section 4: adult’s health and lifestyle
Rationale: Purpose behind the inclusion of the indicator
To estimate the proportion of adults who consume 5 or more portions of fruit and vegetables per day in local authorities.
Rationale: Policy relevance
Choosing Health: Making healthy choices easier. http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4094550
To help increase the prevalence of healthy eating and the health benefits associated with eating a healthy diet.
Department of Health National 5 A Day programme http://www.dh.gov.uk/en/Policyandguidance/Healthandsocialcaretopics/ FiveADay/FiveADaygeneralinformation/index.htm The NHS Plan http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4002960 The NHS Cancer Plan http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4009609 National Service Framework for Coronary Heart Disease http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4094275 National Service Framework for Diabetes http://www.dh.gov.uk/en/Policyandguidance/Healthandsocialcaretopics/ Diabetes/DH_4015717 National Service Framework for Older People http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4003066 Interpretation: What a high/low level of indicator value means
A high indicator value (green circle in health summary chart) represents a statistically significant higher level (better) of adults who are estimated to consume 5 or more portions of fruit and vegetables per day when compared to the national value. A low indicator value (red circle in health summary chart) represents a statistically significant lower (worse) level of adults who are estimated to consume 5 or more portions of fruit and vegetables per day when compared to the national value.
Interpretation: Potential for error due to type of measurement method
HSE numerator data are broadly based on observed self-reported daily consumption of fruit and vegetables. Self-reported consumption may be prone to respondent bias. There may be variation in how informants defined and reported the amount of food consumed. Although everyday measures were used to help informants to define how much they had consumed, this task may have been difficult for certain food items, such as fruit in composite foods like apple pie. In order to ensure agreement between these estimates at GOR and SHA level, the lower tier synthetic estimates (districts), which are based on modelled data, have been calibrated. However, the synthetic estimates have not been calibrated to ensure agreement at County level. As a result, there may be inconsistencies between lower tier and county estimates for some areas as the datasets are derived using different methods.
Health Profiles 2009 The Indicator Guide
127 Section 4: adult’s health and lifestyle
Interpretation: Potential for error due to bias and confounding
The Health Survey for England under-samples younger people, people in employment, ethnic minorities, women, those who are healthier but exhibit less healthy behaviour.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate.
These data have not been age-standardised and, therefore, variation between area values may be a result of differences in population structure.
This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
TABLE 2 – INDICATOR SPECIFICATION Indicator definition: Variable
Prevalence of healthy eating.
Indicator definition: Statistic
Percentage of resident adult population aged 16 and over
Indicator definition: Gender
Persons
Indicator definition: age group
Adults (aged 16 and over)
Indicator definition: period
2003–2005
Healthy eating is defined as those who consume 5 or more portions of fruit and vegetables per day.
Health Profiles 2009 The Indicator Guide
128 Section 4: adult’s health and lifestyle
Indicator definition: scale Geography: geographies available for this indicator from other providers
Strategic Health Authority. http://www.ic.nhs.uk/webfiles/Popgeog/Direct%20Estimates%20%20of%20 Obesity%20(adults)%202003-2005.pdf
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Age, gender, ethnicity, social class http://www.dh.gov.uk/en/Publicationsandstatistics/PublishedSurvey/ HealthSurveyForEngland/index.htm
Data extraction: Source
Health Surveys for England, National Centre for Social Research (NatCen). Published by The Information Centre for Health and Social Care (IC), 2007
Data extraction: source URL
Data received directly from NatCen.
Data extraction: date
16 January 2008. Data for 4 new 2009 UAs (Bedford UA, Central Bedfordshire UA, Cheshire East UA and Cheshire West UA) were calculated by NatCen as a separate exercise on 21/01/2009 and received on the same date.
Numerator: definition
Proportion of adults who reported consumption of 5 or more portions of fruit and vegetables per day at the time of the survey, 2003–2005. A portion of fruit or vegetables was defined as an 80g serving.
Numerator: source
Health Survey for England (HSE), commissioned by the Department of Health/ IC and carried out by the Joint Health Surveys Unit of the National Centre for Social Research (NatCen) and the Department of Epidemiology and Public Health at the Royal Free and University College Medical School, London.
Denominator: definition
Total number of respondents (with valid record of consumption of fruit and vegetables on the day before the interview) aged 16+ in the Health Survey for England 2003–2005.
Denominator: source
Health Survey for England (HSE), commissioned by the Department of Health/ IC and carried out by the Joint Health Surveys Unit of the National Centre for Social Research (NatCen) and the Department of Epidemiology and Public Health at the Royal Free and University College Medical School, London.
Health Profiles 2009 The Indicator Guide
129 Section 4: adult’s health and lifestyle
Data quality: Accuracy and completeness
The Health Survey for England is designed to provide data at both national and regional level about the population living in private households in England. It uses a clustered, stratified multi-stage sample design. In the 2005 HSE, for example, a random sample of over 7,200 addresses (around 16,000 people) were selected from the Postcode Address File (PAF) to ensure households were sampled proportionately across the nine Government Office Regions in England. 720 postcode sectors were selected and 26 addresses within each sector. Each individual within a selected household was eligible for inclusion. One of the effects of using a complex sample design is that standard errors for survey estimates are generally higher than would be derived from a simple random sample of the same size. There was a full adult sample of around 16,000 in the 2003 HSE. However, the 2004 and 2005 Health Surveys had only 8,000 adults in the normal ‘general population’ sample as these two surveys included boost samples. The 2004 HSE included a boost sample to increase the number of participants from minority ethnic groups and a special Chinese boost sample. The 2005 HSE included a boost sample for older people living in private households and for five months of the year a boost of children aged 2–15 was included. To ensure that each year’s sample was given an approximately equal weight in the calculation of the 3-year estimates (2003–2005) respondents in the boost sample years were weighted up by two. The numerator and denominator counts used to estimate prevalence are based on a sample of the population in each area and, as such, are not true counts. For this reason the numerator and denominator data are not shown in the data sheet.
Table 3 – Indicator Technical Methods Numerator: extraction
Not Applicable
Numerator: aggregation/ allocation
Residency by local authority of each respondent is allocated by postcode of residency.
Numerator data caveats
These data have not been age-standardised and, therefore, variation between area values may be a result of differences in population structure. The definition of portion size for pulses and very small fruits was changed in 2002, as studies of foods in these categories indicated that an 80g portion is larger than that defined in HSE 2001. It is likely that HSE 2001 overestimated the consumption of pulses and very small fruits. From 2002 onwards, a portion of pulses was defined as 3 tablespoons (rather than 2) and a portion of very small fruit as 2 handfuls (rather than 1).
Denominator data caveats
The HSE is a series of annual surveys that began in 1991 with the aim of monitoring the health of the population. It was designed to be representative of the general, non-institutional population living in England. The current “full” sample size of the HSE comprises about 16,000 adults aged 16 and over. For each participant, the survey included an interview and a physical examination by a nurse, at which various physical measurements, tests, and samples of blood and saliva were collected. These measurements provided biomedical information about known risk factors associated with disease and objective validation for self-reported health behaviour.
Health Profiles 2009 The Indicator Guide
130 Section 4: adult’s health and lifestyle
Methods used to calculate indicator value
Questions about fruit and vegetable consumption have been included in the Health Survey for England since 2001. The questions are interviewer administered as part of the Computer Assisted Personal Interview (CAPI). Questions are designed to assess fruit and vegetable consumption and focus on consumption on the day before the interview, which was defined as the 24 hours from midnight to midnight. This time period was selected to ensure that variations in informant work patterns and times of meals did not affect the average measure of daily consumption. Fruit and vegetable consumption is measured in portions per day, where a portion is defined as an 80g serving. A range of foods, including fruit, vegetables, pulses, salads and fruit juice contribute to the total number of portions consumed. Portion size was translated into everyday measures to help informants to report how much they had consumed. For example, informants were asked how many tablespoons of vegetables, cereal bowls of salad, or pieces of medium sized fruit (such as apples) they had consumed in the previous 24 hours. These everyday measures were converted back to portions prior to analysis. The table below shows portion sizes for the different food items included in the questionnaire.
Estimates are based on pooling together three consecutive years of Health Survey for England data (2003–2005). The general population sample size in 2004 and 2005 was about half the sample size in 2003 owing to the sampling of specific population groups – namely, ethnic minority populations (2004) and older people living in private households (2005). To ensure that each year’s sample was given an approximately equal weight in the calculation of the 2003– 2005 estimates, respondents in 2004 and 2005 were weighted up by two. Data for 4 new 2009 UAs (Bedford UA, Central Bedfordshire UA, Cheshire East UA and Cheshire West UA) were calculated by NatCen as a separate exercise on 21/01/2009 using the same method as described above. As geographical coverage for the remaining 6 new 2009 UAs (County Durham UA, Northumberland UA, Shropshire UA, Wiltshire UA, Cornwall UA) remains the same as their old County equivalents, data for these areas has not been recalculated as part of this exercise.
Health Profiles 2009 The Indicator Guide
131 Section 4: adult’s health and lifestyle
Small Populations: How Isles of Scilly and City of London populations have been dealt with
The Health Survey for England sample does not cover the Isles of Scilly and no HSE respondents over 2003–2005 were located in the City of London.
Disclosure Control
Not applicable.
Confidence Intervals calculation method
The standard errors, and 95% confidence intervals, have been calculated using STATA’s survey module (the svy:mean commands), further details can be obtained from Shaun Scholes at NatCen ([email protected]). One of the effects of using a complex design is that standard errors for survey estimates are generally higher than the standard errors that would be derived from a simple random sample of the same size.
Health Profiles 2009 The Indicator Guide
132 Section 4: adult’s health and lifestyle
15b. HEALTHY EATING ADULTS LOWER TIER INDICATOR BASIC INFORMATION 1. What is being measured?
Estimated prevalence of adult healthy eating, that is adults who consume 5 or more portions of fruit and vegetables per day
2. Why is it being measured?
To estimate the proportion of healthy eating adults. There is evidence that eating at least 5 portions of a variety of fruit and vegetables a day could reduce the risk of deaths from chronic diseases such as heart disease, stroke, and cancer by up to 20%.
3. How is this indicator actually defined?
Prevalence of healthy eating, percentage of resident population, adults, 2003–2005, persons.
4. Who does it measure?
Adults (aged 16 and over).
5. When does it measure it?
Updated as ad-hoc.
6. Will It measure absolute numbers or proportions?
Percentage of resident adult population aged 16 and over
7. Where does the data actually come from?
Modelled by the National Centre for Social Research (NatCen). Published by The Information Centre for Health and Social Care (IC), 2007.
8. How accurate and complete will the data be?
These are modelled estimates based on national survey data. The model is non-aetiological (not based on known casual factors). The estimates do not take into account additional local factors that may impact on the true prevalence of healthy eating in an area and may not match with local lifestyle survey results or modelled estimates which use known risk factors such as socio-economic status, age, gender and ethnicity.
9. Are there any caveats/warnings/ problems?
As these estimates are modelled they should be used and interpreted with caution (see above).
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
Pg 4 Health Summary – Indicator No 15
Subject category/ domain(s)
Adults health and lifestyle
Indicator name (*Indicator title in health profile)
Estimated prevalence of adults who eat healthily (*Healthy eating adults)
Health Profiles 2009 The Indicator Guide
133 Section 4: adult’s health and lifestyle
PHO with lead responsibility
SEPHO
Date of PHO dataset creation
21/01/2009
Indicator definition
Prevalence of healthy eating, percentage of resident population, adults, 20032005, persons.
Geography
Local Authority: County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs. Exceptions: Data for the 9 new 2009 UAs (Bedford UA, Central Bedfordshire UA, Cheshire East UA, Cheshire West and Chester UA, County Durham, Northumberland UA, Shropshire UA, Wiltshire UA, Cornwall UA) are not based on modelled estimates but are calculated directly using results from the Health Surveys for England 2003–05.
Timeliness
Updated as ad-hoc.
Rationale: What this indicator purports to measure
Expected prevalence of adult healthy eating that is adults who consume 5 or more portions of fruit and vegetables per day.
Rationale: Public Health Importance
The indicator is a measure of a protective lifestyle factor. A diet rich in fruit and vegetables confers protective effects against the development of heart disease and certain cancers. It has been estimated that eating at least 5 portions of a variety of fruit and vegetables a day could reduce the risk of deaths from chronic diseases such as heart disease, stroke, and cancer by up to 20%. It has been estimated that diet might contribute to the development of onethird of all cancers, and that increasing fruit and vegetable consumption is the second most important cancer prevention strategy, after reducing smoking. In 1998, the Department of Health’s Committee on Medical Aspects of Food Policy and Nutrition reviewed the evidence and concluded that higher vegetable consumption would reduce the risk of colorectal cancer and gastric cancer. There was also weakly consistent evidence that higher fruit and vegetable consumption would reduce the risk of breast cancer. These cancers combined represent about 18% of the cancer burden in men and about 30% in women. Higher consumption of fruit and vegetables also reduces the risk of coronary heart disease and stroke. A recent study found that each increase of 1 portion of fruit and vegetables a day lowered the risk of coronary heart disease by 4% and the risk of stroke by 6%. Evidence also suggests an increase in fruit and vegetable intake can help lower blood pressure. Research suggests that there are other health benefits, including delaying the development of cataracts, reducing the symptoms of asthma, improving bowel function, and helping to manage diabetes. As well as the direct health benefits, eating fruit and vegetables can help to achieve other dietary goals including increasing fibre intake, reducing fat intake, help maintain a healthy weight, and substituting for foods with added sugars (as frequent consumption of foods with added sugars can contribute to tooth decay).
Health Profiles 2009 The Indicator Guide
134 Section 4: adult’s health and lifestyle
Rationale: Purpose behind the inclusion of the indicator
To estimate the expected proportion of adults who 5 or more portions of fruit and vegetables per day in local authorities given the characteristics of local authority populations.
Rationale: Policy relevance
Choosing Health: Making healthy choices easier. http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4094550
To help increase the prevalence of healthy eating and the health benefits associated with eating a healthy diet.
Department of Health National 5 A Day programme http://www.dh.gov.uk/en/Policyandguidance/Healthandsocialcaretopics/ FiveADay/FiveADaygeneralinformation/index.htm The NHS Plan http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4002960 The NHS Cancer Plan http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4009609 National Service Framework for Coronary Heart Disease http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4094275 National Service Framework for Diabetes http://www.dh.gov.uk/en/Policyandguidance/Healthandsocialcaretopics/ Diabetes/DH_4015717 National Service Framework for Older People http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4003066 Interpretation: What a high/low level of indicator value means
Given the characteristics of the local population, a high indicator value (green circle in health summary chart) represents a statistically significant higher (better) proportion of adults who are estimated to consume 5 or more portions of fruit and vegetables per day when compared to the national value. Given the characteristics of the local population, a low indicator value (red circle in health summary chart) represents a statistically significant lower (worse) proportion of adults who are estimated to consume 5 or more portions of fruit and vegetables per day when compared to the national value.
Health Profiles 2009 The Indicator Guide
135 Section 4: adult’s health and lifestyle
Interpretation: Potential for error due to type of measurement method
It is important that users note that these model-based do not take account of any additional local factors that may impact on the true prevalence rate of fruit and vegetable consumption in an area (e.g. local initiatives designed to increase fruit and vegetable consumption). They will almost certainly not mirror precisely any available measures from local studies or surveys (although research by NatCen and others have shown that they tend to be related). The figures, therefore, cannot be used to monitor performance or change over time. The model used is a non-aetiological model i.e. is not based on known aetiological risk factors. This may lead to estimated fruit and vegetable consumption levels which are at odds with, for example, local lifestyle survey results or modelled estimates which use known co-variates such as socioeconomic status, age, gender and ethnicity such as the fruit and vegetable consumption estimates modelled in the Health Poverty Index available at www. hpi.org.uk (see variables used in generation of model in calculation of indicator section below). There may also be a discrepancy between the modelled lower tier estimates (districts) and upper tier estimates (County geographies and above, plus new 2009 UA areas) which are based on actual Health Survey for England data. This has lead to inconsistencies between lower tier and county estimates for some areas as the datasets are derived using different methods.
Interpretation: Potential for error due to bias and confounding
These model based healthy lifestyle indicators are derived using the Health Survey for England data and are subject to both sampling and non-sampling error. Sampling errors arise solely as a result of drawing a sample rather than conducting a full survey of the population. Generally, the smaller the sample size the larger the variability in the estimates that one would expect to obtain from all the possible samples. Non-sampling errors arise during the course of the survey activities and there is no simple direct way of estimating the size of these errors. Non sampling error may include e.g. respondent bias, interview bias and refusal to participate. The use of statistical models for prediction involves making assumptions about relationships in the data. The suitability of the chosen models for the given data and the validity of the model in describing real world dynamics have a bearing on the nature and magnitude of the errors introduced. A key source of modelling error arises from omitting variables that would otherwise help improve the model predictions either by error or because there is no available or reliable data source for them. The model-based estimate generated for a particular area is the expected measure for that area based on its population characteristics - and not an estimate of the actual prevalence. In statistical terms, the model-based estimate is actually a biased estimate of the true value for the area and, as such, should be treated with caution. As mentioned above, the model-based estimates are unable to take account of any additional local factors that may impact on the true prevalence rate. To interpret the estimates, NatCen recommend that users adopt statements such as “given the characteristics of the local population we would expect approximately x% of adults within LA Y to consume 5 or more portions of fruit and vegetables per day ”.
Health Profiles 2009 The Indicator Guide
136 Section 4: adult’s health and lifestyle
Validation exercises were used to check the appropriateness of the chosen models. Confidence intervals have been calculated for the model-based estimates to capture both sampling and modelling error. The confidence intervals provide a range within which we can be fairly sure the ‘true’ value for that area lies. It is recommended that users look at the confidence interval for the estimates, not just the estimate. Estimates for two areas can only be described as significantly different if the confidence intervals for the estimates do not overlap. Users should also note that the potential sources of bias and error also apply to any ranking or banding of the small-area estimates. NatCen do not encourage any ranking of small area estimates within larger areas such as Local Authorities, Primary Care Organisations and Strategic Health Authorities. The 2003–2005 model based estimates are not comparable with the preceding estimates for 2000-2002 owing to differences in geography and modelling methodology: The 2000–2002 LA estimates were calculated by aggregating the model-based estimates for the component wards. The 2003–05 LA estimates have been calculated by modelling directly at the LA level. The choice of co-variate data was different as both the Index of Multiple deprivation 2004 and ONS area classifications were excluded in the 2003–05 estimates owing to their statistical relationship with other census-based covariates. The 2003–05 based estimates were adjusted to be consistent with the direct survey estimates at GOR/SHA level. The model-based estimates have been produced solely for LAs and cannot be translated onto any other geographical boundary system (i.e. aggregated or averaged over any other spatial unit). As a result of limitations in the model based estimates they are published as “experimental statistics”. This term is applied to any set of ONS statistics that do not meet the rigorous quality standards of National Statistics and/or may be subject to change due to methodological development. Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health.
Health Profiles 2009 The Indicator Guide
137 Section 4: adult’s health and lifestyle
The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Table 2 – Indicator Specification Indicator definition: Variable
Prevalence of healthy eating. Healthy eating is defined as those who consume 5 or more portions of fruit and vegetables per day.
Indicator definition: Statistic
Percentage of resident adult population aged 16 and over
Indicator definition: Gender
Persons
Indicator definition: age group
Adults (aged 16 and over)
Indicator definition: period
2003–2005
Indicator definition: scale Geography: geographies available for this indicator from other providers
MSOA available from http://www.ic.nhs.uk/pubs/healthylifestyles05
Dimensions of inequality: subgroup analyses of this dataset available from other providers
None.
Data extraction: Source
National Centre for Social Research (NatCen).
Data extraction: source URL
Data received directly from NatCen. Also published by IC, December 2007 (see link above).
Data extraction: date
16/01/2008
Numerator: definition
Model estimates by NatCen using data from a number of sources including Health Survey for England 2003–2005, Census 2001.
Numerator: source
NatCen.
Health Profiles 2009 The Indicator Guide
138 Section 4: adult’s health and lifestyle
Denominator: definition
Not applicable.
Denominator: source
Not applicable.
Data quality: Accuracy and completeness
The model-based approach generates estimates that are of a different nature from standard survey estimates because they are dependent upon how well the relationship between healthy lifestyle behaviours for individuals and the Census/administrative information about the area in which they live are specified in the model. The accuracy and completeness of the information will be subject to the same constraints surrounding the Health Survey for England and Census data sets on which they are based (see Interpretation: potential sources of error section).
TABLE 3 – INDICATOR TECHNICAL METHODS Numerator: extraction
Not applicable.
Numerator: aggregation / allocation
Not applicable.
Numerator data caveats
See Interpretation: potential sources of error section.
Denominator data caveats
Not applicable.
Methods used to calculate indicator value
The process of creating the model-based estimates of healthy lifestyle behaviours involved three main stages: • A statistical model was used to represent the relationships between healthy eating and area-level characteristics in the small areas covered by the HSfE. The 2001 Census provided the main source for demographic and social covariate data. Other routine sources of data providing area-level characteristics for LAs included all age-all cause mortality, diversity index, life expectancy, emergency hospital admissions, hospital admissions attributable to alcohol, job seekers allowance claimant counts and educational attainment. • The model outputs were applied in conjunction with covariate data (available for all LAs) to estimate the ‘expected’ prevalence given the characteristics of the area. • An adjustment factor was applied to ensure that the model-based estimates for each LA corresponded with the 2003–2005 direct estimates at GOR/SHA level taken from the Health Survey for England data. The model contained an interaction term between SHA and the percentage of residents claiming Job Seekers Allowance over 12 months. The model also contained an interaction term between the percentage of residents aged 16–74 whose highest qualification attained was an NVQ Level 1 (or with no qualifications) and living in areas characterised as containing Village, Hamlet and Isolated dwellings.
Health Profiles 2009 The Indicator Guide
139 Section 4: adult’s health and lifestyle
The area level characteristics associated with increased propensity for an adult to consume 5 or more portions of fruit and vegetables were: a higher proportion of residents of Pakistani ethnic origin; a higher hospital admission rate for acute conditions usually managed in primary care settings; and a higher proportion of residents with limiting long-term illness. The area-level characteristics associated with decreased propensity for an adult to consume 5 or more portions of fruit and vegetables were: residing in an area with a high standardised mortality ratio and areas where a relatively higher proportion of households have no regular access to a car. The model-based estimates were constrained to the direct GOR/SHA estimates taken from the Health Survey for England data. This was done by aggregating the model-based estimates to GOR/SHA and comparing to the direct estimates. The relevant ratios of the HSfE direct estimates to the aggregated model-based estimates at GOR/SHA level were then used to scale the model based LA level estimates. However, the model-based estimates have not been calibrated to ensure agreement at County level. As a result, there may be inconsistencies between lower tier and county estimates for some areas as the datasets are derived using different methods. For a fuller technical description of the methodology see the Model-Based Estimates User Guide and other reports available on the Information Centre website: http://www.ic.nhs.uk/webfiles/Popgeog/Healthy%20Lifestyle%20 Behaviours-%20Model%20Based%20Estimates%20for%20Middle%20 Layer%20Super%20Output%20Areas%20and%20Local%20Authorities%20 in%20England_2003-2005__%20User%20Guide.pdf For methods used to calculate data for new 2009 UAs please see metadata for upper tier geographies. Small Populations: How Isles of Scilly and City of London populations have been dealt with
Model based estimates were not produced for Isles of Scilly or City of London.
Disclosure Control
Not applicable.
Health Profiles 2009 The Indicator Guide
140 Section 4: adult’s health and lifestyle
Confidence Intervals calculation method
The model-based estimate generated for a particular LA is the expected measure for that LA based on its characteristics as measured by the covariates in the model. In statistical terms, the model-based estimate is actually a biased estimate of the true value for an area and, as such, should be treated with caution. By placing confidence intervals around a model-based estimate, however, we can generate a range within which we can be fairly sure the ‘true’ value for that area lies. In order to generate the confidence interval, an estimate of the variance of the difference between the model-based estimate and the true LA measure is required. The estimate of the variance has two components which correspond to the variance that is not explained by the model (and hence is not predicted by the model-based estimate) and the uncertainty of the model-based estimate itself. Obtaining the confidence interval for the LA estimates required computing two area-level variance terms: one for the Primary Sampling Units and one for the LA. The first term was required to allow for the clustering in the sample and the second term to estimate the residual variance at the LA level that was not explained by the model. The confidence interval for the model-based LA estimates was estimated to be:
βˆ T X i T e −1 log it αˆ + βˆ X i ± 1.96 T βˆ X i 1+ e
2
M i N ij 2 2 ∑ σˆ u + σˆ v2 + X TiVar ( βˆ ) X i Ni j =1
1
where: Χi is the vector of covariates values for LA i,
Xi
βˆ
βˆ is the vector of parameter estimates for the LA-level covariates,
σˆ 2 is the estimate of the between PSU-level variance,
σˆ u2u
σˆ 2 is the estimate of the between LA-level variance.
σˆ v2v
N is the population estimate of the number of adults in LA i,
Ni i
Nijij is the population estimate of the number of adults in postcode sector N M i and
Mi is the number of postcode sectors in LA i.
ij,
2
Health Profiles 2009 The Indicator Guide
141 Section 4: adult’s health and lifestyle
16. PHYSICALLY ACTIVE ADULTS INDICATOR Basic Information 1. What is being measured?
Adults participating in recommended levels of physical activity.
2. Why is it being measured?
To estimate prevalence of physical activity beneficial to health in the population. To monitor the effectiveness of programmes aiming to increase participation of adults in sport and active recreation.
3. How is this indicator actually defined?
Participation in moderate intensity sport and active recreation on 20 or more days in the previous 4 weeks, (averaging 5 or more times per week).
4. Who does it measure?
Adults (aged 16 and over).
5. When does it measure it?
The survey is to be repeated in 2008/9 and 2009/10.
6. Will It measure absolute numbers or proportions?
Proportions: persons, aged 16 and over, 2007/08, as percentage respondents of the Sport England Active People Survey 2.
7. Where does the data actually come from?
Data received directly from Sport England.
8. How accurate and complete will the data be?
The Active People Survey is a telephone survey conducted using CATI (Computer Aided Telephone Interviewing). At least 500 interviews are undertaken per local authority. The exceptions were Isles of Scilly, City of London (due to small populations) and Birmingham where 5,000 interviews were conducted and Liverpool where 2,500 interviews were conducted.
9. Are there any caveats/warnings/ problems?
This measure is a crude proportion and no agestandardisation has been applied to the results to adjust for differences in age structure between areas. It is likely that a greater proportion of younger people undertake levels of physical activity at the recommended levels than older people. Sport England numerator data are based on observed self-reported physical activity levels in the previous 4 weeks and self-reported physical activity levels may be prone to respondent bias. In addition, the indicator does not include active recreation such as housework, DIY, activity in ones job or active transport.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Health Profiles 2009 The Indicator Guide
142 Section 4: adult’s health and lifestyle
TABLE 1 – INDICATOR DESCRIPTION Information component
Pg 4 Health Summary – Indicator 16
Subject category/ domain(s)
Adults health and lifestyle
Indicator name (* Indicator title in health profile)
Adults participating in recommended levels of physical activity (*Physically active adults)
PHO with lead responsibility
SEPHO
Date of PHO dataset creation
04/03/2009
Indicator definition
Participation in *moderate intensity sport and active recreation on 20 or more days in the previous 4 weeks, (averaging 5 or more times per week), percentage, persons, aged 16 and over, 2007/08, as percentage respondents of the Sport England Active People Survey 2
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs, new 2009 Unitary Authorities.
Timeliness
The survey is to be repeated in 2008/9 (APS3) and 2009/10 (APS4)
Rationale: What this indicator purports to measure
This indicator estimates the proportion of adults participating in physical activity beneficial to health.
Rationale: Public Health Importance
People who have a physically active lifestyle are at approximately half the risk of developing coronary heart disease compared to those who have a sedentary lifestyle. Regular physical activity is also associated with a reduced risk of diabetes, obesity, osteoporosis and colon cancer and with improved mental health. In older adults physical activity is associated with increased functional capacities. In terms of mortality, morbidity and quality of life, the Chief Medical Officer has estimated the cost of inactivity in England to be £8.2 billion annually. Evidence for the effectiveness of interventions to increase the population levels of physical activity is summarised by Kahn E, Ramsey L, Brownson R, Heath G, Howze E, Powell K, et al. ‘The Effectiveness of Interventions to Increase Physical Activity: A Systematic Review’. Am J Prev Med 2002; 22 (4S)
Rationale: Purpose behind the inclusion of the indicator
To estimate prevalence of physical activity beneficial to health in the population To monitor the effectiveness of programmes aiming to increase participation of adults in sport and active recreation The indicator is a measure of health need i.e. the ability to benefit from public health interventions aiming to improve levels of physical activity beneficial to health in the adult population.
Health Profiles 2009 The Indicator Guide
143 Section 4: adult’s health and lifestyle
Rationale: Policy relevance
Choosing Health: Making Healthy Choices Easier, Department of Health, 2004 http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4094550 Choosing Health: a physical activity action plan, Department of Health, 2005 http://www.dh.gov.uk/prod_consum_dh/groups/dh_digitalassets/@dh/@en/ documents/digitalasset/dh_4105710.pdf At Least Five a Week: Evidence on the Impact of Physical Activity and its Relationship to Health, 2004 – report by the CMO http://www.dh.gov.uk/prod_consum_dh/groups/dh_digitalassets/@dh/@en/ documents/digitalasset/dh_4080981.pdf Game Plan: A Strategy for Delivering Government’s Sport and Physical Activity Objectives, 2002 http://www.sportdevelopment.org.uk/html/gameplan.html The Sport England Strategy 2008–2001 was published in June 2008 and commits Sport England to deliver on a series of demanding targets by 2012/13: • one million people doing more sport • a 25% reduction in the number of 16–18 year olds who drop out of five key sports • improved talent development systems in at least 25 sports • a measurable increase in people’s satisfaction with their experience of sport – the first time the organisation has set such a qualitative measure • a major contribution to the delivery of the five hour sports offer for children and young people. The strategy can be found at: http://www.sportengland.org/index/get_resources/resource_downloads/sport_ england_strategy.htm
Interpretation: What a high/low level of indicator value means
A high indicator value (green circle in health summary chart) represents a statistically significant higher (better) estimated percentage of adults participating in physical activity when compared to the national value. A low indicator value (red circle in health summary chart) represents a statistically significant lower (worse) estimated percentage of adults participating in physical activity when compared to the national value.
Interpretation: Potential for error due to type of measurement method
Sport England numerator data are based on observed self-reported physical activity levels in the previous 4 weeks through a series of questions which focus on walking, cycling and other types of sport and recreational physical activity. Self-reported physical activity levels may be prone to respondent bias. Respondents are required to remember how long each session of physical activity lasted and describe the intensity level of the activity they undertook. As a result, levels of physical activity are likely to be over-reported. The Active People Survey measures sport and active recreation but excludes other forms of physical activity such as housework, DIY, activity in ones job, active transport etc. and this may lead to under-estimation of levels.
Health Profiles 2009 The Indicator Guide
144 Section 4: adult’s health and lifestyle
Interpretation: Potential for error due to bias and confounding
This measure of adults participating in recommended levels of physical activity is a crude proportion and no age-standardisation has been applied to the results to adjust for differences in age structure between areas. It is likely that a greater proportion of younger people undertake levels of physical activity at the recommended levels than older people. Details of the sampling frame used and which groups may have been under sampled (e.g. ethnic minorities, those without telephone access etc) were not available at the time of creation of this indicator. Therefore, it is not known how representative of the general population the survey results are.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with a amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Health Profiles 2009 The Indicator Guide
145 Section 4: adult’s health and lifestyle
Table 2 – Indicator Specification Indicator definition: Variable
Participation in *moderate intensity sport and active recreation on 20 or more days (averaging 5 or more times per week) in the previous 4 weeks *Moderate intensity sport and active recreation is defined as a 30 minute session of activity such as cycling, walking, sport and recreational physical activity. For walking, moderate activity was defined as walking at “a fairly brisk pace” or “a fast pace”. For cycling and other sport and recreational activity, moderate activity was defined as where the effort required was enough to raise the individual’s breathing rate, make the respondent out of breath or sweat. For further details see http://www.sportengland.org/index/ get_resources/research/active_people.htm
Indicator definition: Statistic
Percentage of respondents of the Sport England Active People Survey 2 (APS2)
Indicator definition: Gender
Persons
Indicator definition: age group
16 and over
Indicator definition: period
October 2007–October 2008
Indicator definition: scale Geography: geographies available for this indicator from other providers
At time of creation of indicator: County Sport Partnerships available from Sport England www.sportengland.org
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Age, gender, ethnicity, education, housing tenure, disability, working status, household income, National Statistics Socio-economic Classification (NS-SEC) available from Sport England.
Data extraction: Source
Sport England. Ipsos MORI undertook the survey on behalf of Sport England.
Data extraction: source URL
Data received directly from Sport England.
Data extraction: date
21/02/2009 – Local authority district estimates 25/02/2009 – Regional/County data 4/3/2009 – Central Bedfordshire UA, Cheshire East, Cheshire West and Chester
Numerator: definition
Respondents aged 16 and over, with valid responses to questions on physical activity in the Active People Survey 2, who participated in moderate intensity sport or recreational activities for 30 minutes or more on 20 or more days in the previous four weeks.
Numerator: source
Sport England Active People Survey 2 www.sportengland.org/
Denominator: definition
Respondents aged 16 and over, with valid responses to questions on physical activity in the Active People Survey 2
Health Profiles 2009 The Indicator Guide
146 Section 4: adult’s health and lifestyle
Denominator: source
Sport England Active People Survey 2 www.sportengland.org/
Data quality: Accuracy and completeness
The Active People Survey was undertaken by Ipsos MORI and conducted across every local authority in England. The survey is conducted by telephone using Random Digit Dialling (RDD) to generate a sample of telephone numbers. The RDD sample is drawn by selecting numbers from a database comprising all exchange codes allocated for residential use in the UK. A representative sample is then drawn by randomising the last four digits of each number. The telephone survey is conducted using CATI (Computer Aided Telephone Interviewing) and in APS2 achieved 188,800 interviews overall with a minimum of 500 interviews per local authority. This number of interviews will be repeated in the 2008/9 (Active People Survey 3) and 2009/10 (Active People Survey 4) surveys giving a cumulative local authority sample of 1,000 (equivalent to the first Active People Survey) every two years. All local authorities were offered the opportunity to boost their APS2 from 500 to 1,000. This was taken up by 14 local authorities. Two LAs choose larger boosts: Liverpool boosted by 2,000 to provide a 2,500 overall sample size and Birmingham boosted by 3,800 to provide a 5,000 overall sample size. Further exceptions to the 500 sample target in APS2 were Isles of Scilly and City of London which have small populations and do not require such a large sample. Respondents were matched to local authority of residence using postcode of residents. Where postcode was not known (around 1 in 10), respondents were asked to state in which local authority they lived or dialling codes/exchange area telephone number were used to look this up. The Active People Survey data are weighted to be representative of the 16+ population of each reporting geography (e.g. local authority, Government Region etc). Data within each reporting geography are weighted by the following factors: age within gender, ethnicity (White/Non White), National Statistics socio-economic classification (NS SEC). To ensure summer and winter responses are given equal weight within the annual data, a final weight is applied to distribute half the sample size to the summer period and half the sample size to the winter period. Further information is available on request from Sport England.
Table 3 – Indicator Technical Methods Numerator: extraction
Extraction by Sport England.
Numerator: aggregation / allocation
Performed by Sport England.
Numerator data caveats
Numerator data are weighted totals and as such do not sum to overall totals.
Denominator data caveats
Denominator data are weighted totals and as such do not sum to overall totals.
Health Profiles 2009 The Indicator Guide
147 Section 4: adult’s health and lifestyle
Methods used to calculate indicator value
The number of respondents aged 16 and over, with valid responses to questions on physical activity in the Active People Survey 2006, who participated in moderate intensity sport or recreational activities for 30 minutes or more on 20 or more days in the previous four weeks for each local authority was divided by the number of respondents aged 16 and over, with valid responses to questions on physical activity in the Active People Survey, 2006 and multiplied by 100. This generated the percentage participating in recommended levels of physical activity.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Isles of Scilly and City of London are excluded from the lower tier datasets but included in England, Regional and County figures.
Disclosure Control
None applied.
Confidence Intervals calculation method
95% confidence intervals have been calculated using a formula based on a normal approximation provided by Sport England as follows: Upper and lower 95% confidence interval = 100*(1.96*SQRT((Share*(1Share))/Total)) where share is the question proportion and total is the sample size.
Health Profiles 2009 The Indicator Guide
148 Section 4: adult’s health and lifestyle
17a. OBESE ADULTS UPPER TIER INDICATOR Basic Information 1. What is being measured?
Prevalence of obese adults
2. Why is it being measured?
To estimates the proportion of adults who are classified as obese. Obesity has serious health consequences and is association with all cause mortality and decreased life expectancy.
3. How is this indicator actually defined?
Prevalence of obesity, percentage of resident population, adults, 2003–2005, persons.
4. Who does it measure?
Adults (aged 16 and over).
5. When does it measure it?
The Health Survey for England (HSE) is carried out annually.
6. Will It measure absolute numbers or proportions?
Percentage of resident adult population aged 16 and over.
7. Where does the data actually come from?
Health Surveys for England, National Centre for Social Research (NatCen). Published by The Information Centre for Health and Social Care (IC), 2007.
8. How accurate and complete will the data be?
The HSE was designed to be representative of the general, non-institutional population living in England. The current “full” sample size of the HSE comprises about 16,000 adults aged 16 and over.
9. Are there any caveats/warnings/ problems?
BMI is calculated using measured height and weight which may be subject to measurement bias. In some sections of the population, applying BMI classification is not always straightforward e.g. when looking at the elderly or different ethnic groups. A definition based on waist-hip ratio is often considered a better measure of obesity. The Health Survey for England under-samples younger people, people in employment, ethnic minorities, women, those who are healthier but exhibit less healthy behaviour.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Health Profiles 2009 The Indicator Guide
149 Section 4: adult’s health and lifestyle
Table 1 – Indicator Description Information component
Pg 4 Health Summary – Indicator Number 17
Subject category/ domain(s)
Adults health and lifestyle
Indicator name (* Indicator title in health profile)
Prevalence of obese adults (*Obese adults)
PHO with lead responsibility
SEPHO
Date of PHO dataset creation
21/01/2009
Indicator definition
Prevalence of obesity, percentage of resident population, adults, 2003-2005, persons
Geography
England, GOR, County, South East SHAs, new 2009 UAs (Bedford UA, Central Bedfordshire UA, Cheshire East UA, Cheshire West and Chester UA, County Durham, Northumberland UA, Shropshire UA, Wiltshire UA, Cornwall UA).
Timeliness
Updated annually.
Rationale: What this indicator purports to measure
Prevalence of adult obesity
Rationale: Public Health Importance
Obesity in adults is defined for epidemiological purposes as body mass index (BMI) > 30 kg/m2. There is an association between all cause mortality and obesity. Obesity decreases life expectancy by up to nine years. Obesity causes insulin insensitivity, which is an important causal factor in diabetes, heart disease, hypertension and stroke. Obesity is associated with the development of hormone-sensitive cancers; the increased mechanical load increases liability to osteoarthritis and sleep apnoea. Obesity carries psychosocial penalties. Thus there are many routes by which obesity is a detriment to wellbeing. All these penalties as outlined in the table below (except the risk of gallstones and hip fracture) decrease with weight loss.
Health Profiles 2009 The Indicator Guide
150 Section 4: adult’s health and lifestyle
Proportion of various diseases attributable to obesity (BMI > 27 kg/m2) Disease Obesity Hypertension Myocardial infarcation Angina pectoris Stroke Venous thrombosis NIDDM Hyperlipidaemia Gout Osteoarthritis Gall-bladder disease Colorectal cancer Breast cancer Genitourinary cancer Hip fracture
Relative Risk 2.9 1.9 2.5 3.1 1.5 2.9 1.5 2.5 1.8 2.0 1.3 1.2 1.6 0.8
Attributable proportion (%) 100.0 24.1 13.9 20.5 25.8 7.7 24.1 7.7 20.0 11.8 14.3 4.7 3.2 9.1 –3.5
Source: http://hcna.radcliffe-oxford.com/obframe.html It is estimated that obesity costs the NHS over £1 billion per year and society as a whole up to £3.5 billion per year. Effective interventions exist to prevent and treat obesity. See NICE guidance – Obesity: the prevention, identification, assessment and management of overweight and obesity in adults and children. http://guidance.nice.org.uk/CG43 Rationale: Purpose behind the inclusion of the indicator
To estimate the proportion of obese adults in local authorities.
Rationale: Policy relevance
Choosing Health: Making healthy choices easier (Dept. Health, 2004). http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4094550
To help reduce the prevalence of obesity.
Obesity: Defusing the Health Time Bomb (from the Annual Report of the Chief Medical Officer, 2002). http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/AnnualReports/ Browsable/DH_4875027 Tackling Obesity: A Toolbox for Local Partnership Action (The National Heart Forum, 2007) http://www.fphm.org.uk/resources/AtoZ/toolkit_obesity/obesity_toolkit_intro. pdf Tackling Obesity in England, 2001 (National Audit Office, 2001) http://www.nao.org.uk/publications/nao_reports/00-01/0001220.pdf
Health Profiles 2009 The Indicator Guide
151 Section 4: adult’s health and lifestyle
Interpretation: What a high/low level of indicator value means
A high indicator value (red circle in health summary chart) represents a statistically significant higher level (worse) of estimated adult obesity prevalence when compared to the national value. A low indicator value (green circle in health summary chart) represents a statistically significant lower level (better) of estimated adult obesity prevalence when compared to the national value. However obesity at any prevalence level greater than 0 is undesirable, and therefore a low indicator value should not mean that PH action is not needed.
Interpretation: Potential for error due to type of measurement method
For each participant in the Health Survey for England, height and weight was recorded by a nurse. BMI was then calculated for all informants who had valid height and weight measurements, those considered to have unreliable measurements were excluded from the analysis (e.g. pregnant, chair-bound, unsteady or those who could not stand straight). Those who weighed more than 130kg were asked for an “estimated weight” because the scales were unreliable above this level. These were included in the analysis. In some sections of the population, applying the BMI classification described above is not always straightforward e.g. when looking at the elderly or different ethnic groups. For example, in certain Asian populations a given BMI equates to a higher percentage of body fat than the same BMI in a white European population. In some Black populations, however, the converse is true. A definition based on waist-hip ratio is often considered a better measure of obesity. However, BMI is most commonly used and easier to measure routinely. In order to ensure agreement between these estimates at GOR and SHA level, the lower tier synthetic estimates (districts), which are based on modelled data, have been calibrated. However, the synthetic estimates have not been calibrated to ensure agreement at County level. As a result, there may be inconsistencies between lower tier and county estimates for some areas as the datasets are derived using different methods.
Interpretation: Potential for error due to bias and confounding
The Health Survey for England under-samples younger people, people in employment, ethnic minorities, women, those who are healthier but exhibit less healthy behaviour. These data have not been age-standardised and, therefore, variation between area values may be a result of differences in population structure.
Health Profiles 2009 The Indicator Guide
152 Section 4: adult’s health and lifestyle
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Table 2 – Indicator Specification Indicator definition: Variable
Prevalence of obesity. Obesity in adults is defined for epidemiological purposes as body mass index (BMI) > 30 kg/m2.
Indicator definition: Statistic
Percentage of resident adult population aged 16 and over
Indicator definition: Gender
Person
Indicator definition: age group
Adults (aged 16 and over)
Indicator definition: period
2003–2005
Indicator definition: scale
Health Profiles 2009 The Indicator Guide
153 Section 4: adult’s health and lifestyle
Geography: geographies available for this indicator from other providers
Strategic Health Authority. http://www.ic.nhs.uk/webfiles/Popgeog/Direct%20Estimates%20%20of%20 Obesity%20(adults)%202003-2005.pdf
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Age, gender, ethnicity, social class http://www.dh.gov.uk/en/Publicationsandstatistics/PublishedSurvey/ HealthSurveyForEngland/index.htm
Data extraction: Source
Health Surveys for England, National Centre for Social Research (NatCen). Published by The Information Centre for Health and Social Care (IC), 2007
Data extraction: source URL
Data received directly from NatCen.
Data extraction: date
16 January 2008. Data for 4 new 2009 UAs (Bedford UA, Central Bedfordshire UA, Cheshire East UA and Cheshire West UA) were calculated by NatCen as a separate exercise on 21/01/2009 and received on the same date.
Numerator: definition
The number of persons aged 16+ who are obese in a sample survey of the health of the population of England.
Numerator: source
Health Survey for England (HSE), commissioned by the Department of Health/ IC and carried out by the Joint Health Surveys Unit of the National Centre for Social Research (NatCen) and the Department of Epidemiology and Public Health at the Royal Free and University College Medical School, London.
Denominator: definition
Total number of respondents (with valid measurements for height and weight) aged 16+ in the Health Survey for England 2003_2005.
Denominator: source
Health Survey for England (HSE), commissioned by the Department of Health/ IC and carried out by the Joint Health Surveys Unit of the National Centre for Social Research (NatCen) and the Department of Epidemiology and Public Health at the Royal Free and University College Medical School, London.
Data quality: Accuracy and completeness
The Health Survey for England is designed to provide data at both national and regional level about the population living in private households in England. It uses a clustered, stratified multi-stage sample design. In the 2005 HSE, for example, a random sample of over 7,200 addresses (around 16,000 people) were selected from the Postcode Address File (PAF) to ensure households were sampled proportionately across the nine Government Office Regions in England. 720 postcode sectors were selected and 26 addresses within each sector. Each individual within a selected household was eligible for inclusion. One of the effects of using a complex sample design is that standard errors for survey estimates are generally higher than would be derived from a simple random sample of the same size.
Health Profiles 2009 The Indicator Guide
154 Section 4: adult’s health and lifestyle
There was a full adult sample of around 16,000 in the 2003 HSE. However, the 2004 and 2005 Health Surveys had only 8,000 adults in the normal ‘general population’ sample as these two surveys included boost samples. The 2004 HSE included a boost sample to increase the number of participants from minority ethnic groups and a special Chinese boost sample. The 2005 HSE included a boost sample for older people living in private households and for five months of the year a boost of children aged 2–15 was included. To ensure that each year’s sample was given an approximately equal weight in the calculation of the 3-year estimates (2003–2005) respondents in the boost sample years were weighted up by two. The numerator and denominator counts used to estimate prevalence are based on a sample of the population in each area and, as such, are not true counts. For this reason the numerator and denominator data are not shown in the data sheet.
Table 3 – Indicator Technical Methods Numerator: extraction
Not applicable.
Numerator: aggregation/ allocation
Residency by local authority of each respondent is allocated by postcode of residency.
Numerator data caveats
Body mass index is defined as weight (kg) divided by height squared (m2). For adults, four groups are defined according to their BMI: • • • •
Underweight - under 18.5kg/m2; Desirable weight – 18.5 to 25kg/m2; Overweight - 25 to 30kg/m2; Obese - over 30kg/m2
BMI was calculated for all informants who had valid height and weight measurements. The height of informants who were chair-bound, unsteady, or could not stand straight was not measured. Data for those who were considered by the interviewer to have unreliable measurements (e.g. wearing a wig, turban) were excluded from the analysis. The weight of informants who were pregnant, chair bound, unsteady, or could not stand was not measured. Those who weighed more than 130 kg were asked for an “estimated weight” because the scales were unreliable above this level. These have been included in the analysis. These data have not been age-standardised and, therefore, variation between area values may be a result of differences in population structure. Denominator data caveats
The HSE is a series of annual surveys that began in 1991 with the aim of monitoring the health of the population. It was designed to be representative of the general, non-institutional population living in England. The current “full” sample size of the HSE comprises about 16,000 adults aged 16 and over. For each participant, the survey included an interview and a physical examination by a nurse, at which various physical measurements, tests, and samples of blood and saliva were collected. These measurements provided biomedical information about known risk factors associated with disease and objective validation for self-reported health behaviour.
Health Profiles 2009 The Indicator Guide
155 Section 4: adult’s health and lifestyle
Methods used to calculate indicator value
Estimates are based on pooling together three consecutive years of Health Survey for England data (2003–2005). The general population sample size in 2004 and 2005 was about half the sample size in 2003 owing to the sampling of specific population groups – namely, ethnic minority populations (2004) and older people living in private households (2005). To ensure that each year’s sample was given an approximately equal weight in the calculation of the 2003-2005 estimates, respondents in 2004 and 2005 were weighted up by two. Data for 4 new 2009 UAs (Bedford UA, Central Bedfordshire UA, Cheshire East UA and Cheshire West UA) were calculated by NatCen as a separate exercise on 21/01/2009 using the same method as described above. As geographical coverage for the remaining 6 new 2009 UAs (County Durham UA, Northumberland UA, Shropshire UA, Wiltshire UA, Cornwall UA) remains the same as their old County equivalents, data for these areas has not been recalculated as part of this exercise.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
The Health Survey for England sample does not cover the Isles of Scilly and no HSE respondents over 2003–2005 were located in the City of London.
Disclosure Control
Not applicable.
Confidence Intervals calculation method
The standard errors, and 95% confidence intervals, have been calculated using STATA’s survey module (the svy:mean commands), further details can be obtained from Shaun Scholes at NatCen ([email protected]). One of the effects of using a complex design is that standard errors for survey estimates are generally higher than the standard errors that would be derived from a simple random sample of the same size.
Health Profiles 2009 The Indicator Guide
156 Section 4: adult’s health and lifestyle
17b. OBESE ADULTS LOWER TIER INDICATOR Basic Information 1. What is being measured?
Estimated prevalence of obese adults.
2. Why is it being measured?
To estimate the expected proportion of obese adults in local authorities given the characteristics of local authority populations. To help reduce the prevalence of obesity.
3. How is this indicator actually defined?
Prevalence of obesity, percentage of resident population, adults, 2003–2005, persons.
4. Who does it measure?
Adults (aged 16 and over).
5. When does it measure it?
Updated as ad-hoc.
6. Will It measure absolute numbers or proportions?
Percentage of resident adult population aged 16 and over.
7. Where does the data actually come from?
Modelled by the National Centre for Social Research (NatCen). Published by The Information Centre for Health and Social Care (IC), 2007.
8. How accurate and complete will the data be?
These are modelled estimates based on national survey data. The model is non-aetiological (not based on known casual factors). The estimates do not take into account additional local factors that may impact on the true prevalence of obesity in an area and may not match with local lifestyle survey results or modelled estimates which use known risk factors.
9. Are there any caveats/warnings/ problems?
As these estimates are modelled they should be used and interpreted with caution (see above).
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
TABLE 1 – INDICATOR DESCRIPTION Information component
Pg 4 Health Summary – Indicator 17
Subject category/ domain(s)
Adults health and lifestyle
Indicator name (* Indicator title in health profile)
Estimated prevalence of obese adults (*Obese adults)
PHO with lead responsibility
SEPHO
Health Profiles 2009 The Indicator Guide
157 Section 4: adult’s health and lifestyle
Date of PHO dataset creation
21/01/2009
Indicator definition
Prevalence of obesity, percentage of resident population, adults, 2003–2005, persons.
Geography
Local Authority: County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs. Exceptions: Data for the 9 new 2009 UAs (Bedford UA, Central Bedfordshire UA, Cheshire East UA, Cheshire West and Chester UA, County Durham, Northumberland UA, Shropshire UA, Wiltshire UA, Cornwall UA) are not based on modelled estimates but are calculated directly using results from the Health Surveys for England 2003-05.
Timeliness
Updated as ad-hoc.
Rationale: What this indicator purports to measure
Expected prevalence of adult obesity.
Rationale: Public Health Importance
Obesity in adults is defined for epidemiological purposes as body mass index (BMI) > 30 kg/m2. There is an association between all cause mortality and obesity. Obesity decreases life expectancy by up to nine years. Obesity causes insulin insensitivity, which is an important causal factor in diabetes, heart disease, hypertension and stroke. Obesity is associated with the development of hormone-sensitive cancers; the increased mechanical load increases liability to osteoarthritis and sleep apnoea. Obesity carries psychosocial penalties. Thus there are many routes by which obesity is a detriment to wellbeing. All these penalties as outlined in the table below (except the risk of gallstones and hip fracture) decrease with weight loss. Proportion of various diseases attributable to obesity (BMI > 27 kg/m2) Disease Obesity Hypertension Myocardial infarcation Angina pectoris Stroke Venous thrombosis NIDDM Hyperlipidaemia Gout Osteoarthritis Gall-bladder disease Colorectal cancer Breast cancer Genitourinary cancer Hip fracture
Relative Risk 2.9 1.9 2.5 3.1 1.5 2.9 1.5 2.5 1.8 2.0 1.3 1.2 1.6 0.8
Attributable proportion (%) 100.0 24.1 13.9 20.5 25.8 7.7 24.1 7.7 20.0 11.8 14.3 4.7 3.2 9.1 -3.5
Source: http://hcna.radcliffe-oxford.com/obframe.html
Health Profiles 2009 The Indicator Guide
158 Section 4: adult’s health and lifestyle
It is estimated that obesity costs the NHS over £1 billion per year and society as a whole up to £3.5 billion per year. Effective interventions exist to prevent and treat obesity. See NICE guidance – Obesity: the prevention, identification, assessment and management of overweight and obesity in adults and children. http://guidance.nice.org.uk/CG43 Rationale: Purpose behind the inclusion of the indicator
To estimate the expected proportion of obese adults in local authorities given the characteristics of local authority populations.
Rationale: Policy relevance
Choosing Health: Making healthy choices easier (Dept. Health, 2004). http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4094550
To help reduce the prevalence of obesity.
Obesity: Defusing the Health Time Bomb (from the Annual Report of the Chief Medical Officer, 2002). http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/AnnualReports/ Browsable/DH_4875027 Tackling Obesity: A Toolbox for Local Partnership Action (The National Heart Forum, 2007) http://www.fphm.org.uk/resources/AtoZ/toolkit_obesity/obesity_toolkit_intro. pdf Tackling Obesity in England, 2001 (National Audit Office, 2001) http://www.nao.org.uk/publications/nao_reports/00-01/0001220.pdf Interpretation: What a high/low level of indicator value means
A high indicator value (red circle in health summary chart) represents a statistically significant higher level (worse) of expected estimated adult obesity prevalence given the characteristics of the population for that local authority when compared to the national value. A low indicator value (green circle in health summary chart) represents a statistically significant lower level (better) of expected estimated adult obesity prevalence given the characteristics of the population for that local authority when compared to the national value. However obesity at any prevalence level greater than 0 is undesirable, and therefore a low indicator value should not mean that PH action is not needed.
Interpretation: Potential for error due to type of measurement method
It is important that users note that these model based do not take account of any additional local factors that may impact on the true obesity prevalence rate in an area (e.g. local initiatives designed to reduce obesity). They will almost certainly not mirror precisely any available measures from local studies or surveys (although research by NatCen and others have shown that they tend to be related). The figures, therefore, cannot be used to monitor performance or change over time. The model used is a non-aetiological model i.e. is not based on risk factors such as physical activity levels and calorie intake. This may lead to estimated obesity levels which are at odds with other risk factors estimates such as healthy eating and physical activity; local lifestyle survey results and modelled estimates which use known co-variates (see variables used in generation of model in calculation of indicator section below). There may also be a discrepancy between the modelled lower tier estimates (districts) and upper tier estimates (County geographies and above, plus new 2009 UA areas) which are based on actual Health Survey for England data. This has led to inconsistencies between lower tier and county estimates for some areas as the datasets are derived using different methods.
Health Profiles 2009 The Indicator Guide
159 Section 4: adult’s health and lifestyle
Interpretation: Potential for error due to bias and confounding
These model based healthy lifestyle indicators are derived using the Health Survey for England data and are subject to both sampling and non-sampling error. Sampling errors arise solely as a result of drawing a sample rather than conducting a full survey of the population. Generally, the smaller the sample size the larger the variability in the estimates that one would expect to obtain from all the possible samples. Non-sampling errors arise during the course of the survey activities and there is no simple direct way of estimating the size of these errors. Non sampling error may include e.g. respondent bias, interview bias and refusal to participate. The use of statistical models for prediction involves making assumptions about relationships in the data. The suitability of the chosen models for the given data and the validity of the model in describing real world dynamics have a bearing on the nature and magnitude of the errors introduced. A key source of modelling error arises from omitting variables that would otherwise help improve the model predictions either by error or because there is no available or reliable data source for them. The model-based estimate generated for a particular area is the expected measure for that area based on its population characteristics - and not an estimate of the actual prevalence. In statistical terms, the model-based estimate is actually a biased estimate of the true value for the area and, as such, should be treated with caution. As mentioned above, the model-based estimates are unable to take account of any additional local factors that may impact on the true prevalence rate (e.g. local initiatives designed to reduce obesity). To interpret the estimates, NatCen recommend that users adopt statements such as “given the characteristics of the local population we would expect approximately x% of adults within LA Y to be obese”. Validation exercises were used to check the appropriateness of the chosen models. Confidence intervals have been calculated for the model-based estimates to capture both sampling and modelling error. The confidence intervals provide a range within which we can be fairly sure the ‘true’ value for that area lies. It is recommended that users look at the confidence interval for the estimates, not just the estimate. Estimates for two areas can only be described as significantly different if the confidence intervals for the estimates do not overlap. Users should also note that the potential sources of bias and error also apply to any ranking or banding of the small-area estimates. NatCen do not encourage any ranking of small area estimates within larger areas such as Local Authorities, Primary Care Organisations and Strategic Health Authorities. • The 2003–2005 model based estimates are not comparable with the preceding estimates for 2000–2002 owing to differences in geography and modelling methodology: • The 2000–2002 LA estimates were calculated by aggregating the modelbased estimates for the component wards. The 2003–05 LA estimates have been calculated by modelling directly at the LA level. • The choice of co-variate data was different as both the Index of Multiple deprivation 2004 and ONS area classifications were excluded in the 2003– 05 estimates owing to their statistical relationship with other census-based covariates. • The 2003–05 based estimates were adjusted to be consistent with the direct survey estimates at GOR/SHA level.
Health Profiles 2009 The Indicator Guide
160 Section 4: adult’s health and lifestyle
The model-based estimates have been produced solely for LAs and cannot be translated onto any other geographical boundary system (i.e. aggregated or averaged over any other spatial unit). As a result of limitations in the model based estimates they are published as “experimental statistics”. This term is applied to any set of ONS statistics that do not meet the rigorous quality standards of National Statistics and/or may be subject to change due to methodological development. Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
TABLE 2 – INDICATOR SPECIFICATION Indicator definition: Variable
Prevalence of obesity. Obesity in adults is defined for epidemiological purposes as body mass index (BMI) > 30 kg/m2.
Indicator definition: Statistic
Percentage of resident adult population aged 16 and over
Indicator definition: Gender
Persons
Indicator definition: age group
Adults (aged 16 and over)
Health Profiles 2009 The Indicator Guide
161 Section 4: adult’s health and lifestyle
Indicator definition: period
2003–2005
Indicator definition: scale Geography: geographies available for this indicator from other providers
MSOA available from http://www.ic.nhs.uk/pubs/healthylifestyles05
Dimensions of inequality: subgroup analyses of this dataset available from other providers
None.
Data extraction: Source
National Centre for Social Research (NatCen).
Data extraction: source URL
Data received directly from NatCen. Also published by IC, December 2007 (see link above).
Data extraction: date
16/01/2008
Numerator: definition
Model estimates by NatCen using data from a number of sources including Health Survey for England 2003–2005, Census 2001.
Numerator: source
NatCen.
Denominator: definition
Not applicable.
Denominator: source
Not applicable.
Data quality: Accuracy and completeness
The model-based approach generates estimates that are of a different nature from standard survey estimates because they are dependent upon how well the relationship between healthy lifestyle behaviours for individuals and the Census/administrative information about the area in which they live are specified in the model. The accuracy and completeness of the information will be subject to the same constraints surrounding the Health Survey for England and Census data sets on which they are based (see Interpretation: potential sources of error section).
Table 3 – Indicator Technical Methods Numerator: extraction
Not applicable.
Numerator: aggregation / allocation
Not applicable.
Numerator data caveats
See Interpretation: potential sources of error section.
Health Profiles 2009 The Indicator Guide
162 Section 4: adult’s health and lifestyle
Denominator data caveats
Not applicable.
Methods used to calculate indicator value
The process of creating the model-based estimates of healthy lifestyle behaviours involved three main stages: • A statistical model was used to represent the relationships between obesity and area-level characteristics in the small areas covered by the HSfE. The 2001 Census provided the main source for demographic and social covariate data. Other routine sources of data providing area-level characteristics for LAs included all age-all cause mortality, diversity index, life expectancy, emergency hospital admissions, hospital admissions attributable to alcohol, job seekers allowance claimant counts and educational attainment. • The model outputs were applied in conjunction with covariate data (available for all LAs) to estimate the ‘expected’ prevalence given the characteristics of the area. • An adjustment factor was applied to ensure that the model-based estimates for each LA corresponded with the 2003–2005 direct estimates at GOR/SHA level taken from the Health Survey for England data. The area level characteristics associated with increased propensity for a person to be classed as obese were: a higher proportion of residents of Black or ethnic origin; and a higher proportion of male residents. The area-level characteristics associated with decreased propensity for an adult to be classed as obese were: a higher proportion of residents aged 16–74 whose highest qualification attained was Level 2, 4 or 5; a relatively higher proportion of residents aged 20–24; a relatively higher proportion of Income Support claimants who were classifed as ‘carers and others’; a higher proportion of residents aged 16–74 who had never worked or were long-term unemployed; and a higher proportion of households without basic amenties (central heating and/or sole use of a bath). The model-based estimates were constrained to the direct GOR/SHA estimates taken from the Health Survey for England data. This was done by aggregating the model-based estimates to GOR/SHA and comparing to the direct estimates. The relevant ratios of the HSfE direct estimates to the aggregated model-based estimates at GOR/SHA level were then used to scale the model based LA level estimates. However, the model-based estimates have not been calibrated to ensure agreement at County level. As a result, there may be inconsistencies between lower tier and county estimates for some areas as the datasets are derived using different methods. For a fuller technical description of the methodology see the Model-Based Estimates User Guide and other reports available on the Information Centre website: http://www.ic.nhs.uk/webfiles/Popgeog/Healthy%20Lifestyle%20 Behaviours-%20Model%20Based%20Estimates%20for%20Middle%20 Layer%20Super%20Output%20Areas%20and%20Local%20Authorities%20 in%20England_2003-2005__%20User%20Guide.pdf For methods used to calculate data for new 2009 UAs please see metadata for upper tier geographies.
Health Profiles 2009 The Indicator Guide
163 Section 4: adult’s health and lifestyle
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Model based estimates were not produced for Isles of Scilly or City of London.
Disclosure Control
Not applicable.
Confidence Intervals calculation method
The model based estimate generated for a particular LA is the expected measure for that LA based on its characteristics as measured by the covariates in the model. In statistical terms, the model-based estimate is actually a biased estimate of the true value for an area and, as such, should be treated with caution. By placing confidence intervals around a model-based estimate, however, we can generate a range within which we can be fairly sure the ‘true’ value for that area lies. In order to generate the confidence interval, an estimate of the variance of the difference between the model-based estimate and the true LA measure is required. The estimate of the variance has two components which correspond to the variance that is not explained by the model (and hence is not predicted by the model-based estimate) and the uncertainty of the model-based estimate itself. Obtaining the confidence interval for the LA estimates required computing two area-level variance terms: one for the Primary Sampling Units and one for the LA. The first term was required to allow for the clustering in the sample and the second term to estimate the residual variance at the LA level that was not explained by the model. The confidence interval for the model-based LA estimates was estimated to be:
βˆ T X i T e −1 ˆ ˆ log it α + β X i ± 1.96 T βˆ X i 1+ e
2
N ij 2 T 2 ˆ ∑ σˆ + σˆ v + X i Var ( β ) X i N i u j =1 Mi
2
1
where: Χi is the vector of covariates values for LA i,
Xi
β
βˆ is the vector of parameter estimates for the LA-level covariates, ˆ
σˆ2u2 is the estimate of the between PSU-level variance,
σˆ u
σˆ 2 is the estimate of the between LA-level variance.
σˆ v2v
N is the population estimate of the number of adults in LA i,
Ni i
Nijij is the population estimate of the number of adults in postcode sector N M i and
Mi is the number of postcode sectors in LA i.
ij,
2
Health Profiles 2009 The Indicator Guide
164
Section 5: Disease and poor health
Health Profiles 2009 The Indicator Guide
165 Section 5: disease and poor health
18. OVER 65s ‘NOT IN GOOD HEALTH’ INDICATOR Basic Information 1. What is being measured?
Over 65s Self-Assessed General Health: ‘Not Good’.
2. Why is it being measured?
As a perception of general health. This indicator may help monitor likely health care burden.
3. How is this indicator actually defined?
Self assessed general health: ‘Not Good’, directly age and sex standardised percentage, over 65s, 2001, persons.
4. Who does it measure?
All persons, 65 years and over.
5. When does it measure it?
2001, next Census will be in 2011.
6. Will it measure absolute numbers or proportions?
Age and sex standardised percentage.
7. Where does the data actually come from?
2001 Census, ONS.
8. How accurate and complete will the data be?
This census data has been subject to edit and imputation procedures to correct for incorrect or missing data.
9. Are there any caveats/ warnings/ problems?
‘Not in good health’ is the self-assessed perception of health and therefore subject to differing thresholds applied by individuals.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Health Profiles 2009 The Indicator Guide
166 Section 5: disease and poor health
Table 1 – Indicator Description Information component
Pg 4 Health Summary – Indicator No 18
Subject category/ domain(s)
Disease and poor health.
Indicator name (* Indicator title in health profile)
Over-65s ‘not in good health’
PHO with lead responsibility
WMPHO
Date of PHO dataset creation
12/02/2009
Indicator definition
Self Assessed General Health: ‘Not Good’, directly age and sex standardised percentage, over 65s, 2001, persons
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs
Timeliness
Every 10 years, next update available in 2012. Time trend analysis is appropriate.
Rationale: What this indicator purports to measure
Perception of General Health.
Rationale: Public Health Importance
The indicator was chosen as the best available measure of self assessed population health. Self reported single item of health has a good correlation with mortality and health care utilisation. For further information see: Where Wealth means Health (www.nwpho.org. uk/inequalities)
Rationale: Purpose behind the inclusion of the indicator
To help monitor likely health care burden.
Rationale: Policy relevance
No direct policy driver.
Interpretation: What a high/low level of indicator value means
A high indicator value (red blob in spine chart) represents a statistically significant higher level of estimated self assessed “not good” health for that local authority when compared to the national value. A low indicator value (green blob) represents a statistically significant lower level of estimated self assessed “not good” health for that local authority when compared to the national value.
Health Profiles 2009 The Indicator Guide
167 Section 5: disease and poor health
Interpretation: Potential for error due to type of measurement method
Self reported health status can be subject to variation according to non causative effects (e.g. good weather).
Interpretation: Potential for error due to bias and confounding
The following groups may be under-sampled within the census:
Age standardisation does assume that minor differences in age structure within age bands are unimportant and in general this is true for younger age groups. However with older groups minor differences in average age within age band may become important, especially as all over 85’s are grouped together. The European Standard Population is younger than the UK population and therefore biases comparisons towards younger age groups. Although age standardisation will reduce the effect of age differences in the population it cannot be assumed to eliminate them.
• • • • • • • • •
Areas with high non-white population Full-time students aged 18–74 (out of term time residents) Prisoners Men aged 20–39 Residential homes, nursing homes, hospitals Rough sleepers Areas with high population density Areas with high numbers of multi-occupancy households Migrants: someone who spends 3 to 12 months in the country for certain purposes (excluding tourism), asylum seekers, migrant/seasonal workers,
This can result in an underestimate or overestimate of self assessed ill-health in some areas. Age standardisation does assume that minor differences in age structure within age bands are unimportant and in general this is true for younger age groups. However with older groups minor differences in average age within age band may become important, especially as all over 85s are grouped together. The European Standard Population is younger than the UK population and therefore biases comparisons towards younger age groups. Although age standardisation will reduce the effect of age differences in the population it cannot be assumed to eliminate them.
Health Profiles 2009 The Indicator Guide
168 Section 5: disease and poor health
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider the confidence interval, the greater the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Health Profiles 2009 The Indicator Guide
169 Section 5: disease and poor health
Table 2 – Indicator Specification Indicator definition: Variable
Self Assessed General Health: ‘Not Good’ (over 65s)
Indicator definition: Statistic
Directly age & sex standardised percentage
Indicator definition: Gender
Persons
Indicator definition: age group
Over 65s
Indicator definition: period
2001
Indicator definition: scale
Age & sex standardised percentage
Geography: geographies available for this indicator from other providers
Unique dataset not available elsewhere.
Dimensions of inequality: subgroup analyses of this dataset available from other providers
None.
Data extraction: Source
Nomis Official Labour Market Statistics (a service provided by the Office for National Statistics)
Data extraction: source URL
https://www.nomisweb.co.uk/Default.asp
Data extraction: date
Data extracted from source as at: 12/02/2009
Numerator: definition
All people aged over 65 usually resident in the area at the time of the 2001 Census, who described their general health in the 12 months before Census day as ‘Not good’.
Numerator: source
Office for National Statistics (ONS).
Denominator: definition
All people aged over 65 counted as usually resident in the area at the time of the 2001 Census by 5 year age band. A usual resident was generally defined as someone who spent most of their time at a specific address. It included: people who usually lived at that address but were temporarily away (on holiday, visiting friends or relatives, or temporarily in a hospital or similar establishment); people who worked away from home for part of the time; students, if it was their term-time address; a baby born before 30 April 2001 even if it was still in hospital; and people present on Census Day, even if temporarily, who had no other usual address. However, it did not include anyone present on Census Day who had another usual address or anyone who had been living or intended to live in a special establishment, such as a residential home, nursing home or hospital, for six months or more.
Health Profiles 2009 The Indicator Guide
170 Section 5: disease and poor health
Denominator: source
Office for National Statistics (ONS).
Data quality: Accuracy and completeness
The Census was followed by the Census Coverage Survey (CCS), an independent doorstep survey of a sample of a third of a million households, covering every local authority, which was used to adjust the Census counts for under-enumeration. Under-enumeration in the 2001 Census did not occur uniformly across all areas. Response rates were lowest for inner city areas where characteristics known to be related to non-response such as multi-occupancy and higher proportions of non-English speaking population, are most prevalent. This census data has been subject to edit and imputation procedures to correct for incorrect or missing data. There has been an extensive quality assurance process, including checks against administrative records and sources of information on particular groups such as students and the armed forces.
Table 3 – Indicator Technical Methods Numerator: extraction
Downloaded from Nomis website.
Numerator: aggregation/ allocation
Where available data published for higher geographies has been used and not aggregated from lower geographies. This census data has been subject to edit and imputation procedures to correct for incorrect or missing data as a result summation of areas to higher geographies may not result in the total stated in the published census data. However, since the 2001 census new Unitary Authorities have been created. The new Unitary Authorities of Northumberland, Shropshire, Wiltshire and Durham have the same boundaries as the counties of the same name so have been allocated these data. Data for five other new unitary Authorities have been calculated from the aggregation of previous Local Authorities. New UA name Bedford Central Bedfordshire Cheshire East
Cheshire West and Chester
Cornwall*
Old LA name Bedford Mid Bedfordshire South Bedfordshire Congleton Crewe and Nantwich Macclesfield Chester Ellesmere Port & Neston Vale Royal Caradon Carrick Kerrier North Cornwall Penwith Restormel
*=Cornwall Unitary Authority is the same as the previous Cornwall County however the data for this county in the data source was aggregated with Isles of Scilly and so has instead been aggregated from its local authorities.
Health Profiles 2009 The Indicator Guide
171 Section 5: disease and poor health
Numerator data caveats
See earlier comments on Data quality: Accuracy and completeness
Denominator data caveats
See earlier comments on Data quality: Accuracy and completeness
Methods used to calculate indicator value
The directly age & sex standardised rate is the rate of events that would occur in a population with a standard age and sex structure if that population were to experience the age & sex specific rates of the subject population. The standard population used is the European Standard Population for age standardisation. Sex standardisation has been used, based on a 1:1 ratio of males and females, because females have a longer life expectancy and therefore make up a larger proportion of responses in older age groups. This has been expressed as a rate per 100 i.e. a percentage. The methodology is explained in detail in the APHO Technical Briefing on Commonly Used Public Health Statistics and their Confidence Intervals http://www.apho.org.uk/resource/item.aspx?RID=48457 and the associated Excel tool http://www.apho.org.uk/resource/view.aspx?RID=48617. Age standardisation does assume that minor differences in age structure within age bands are unimportant and in general this is true for younger age groups. However with older groups minor differences in average age within age band may become important, especially as all over 85’s are grouped together. The European Standard Population is younger than the UK population and therefore biases comparisons towards younger age groups. Although age standardisation will reduce the effect of age differences in the population it cannot be assumed to eliminate them.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Isles of Scilly and City of London are excluded from local authority level but are included for higher level geographies.
Disclosure Control
Not applicable
Confidence Intervals calculation method
95% and 99.8% confidence intervals were calculated using a method described by Dobson1. The Byar’s1 method has been used when the numerator is greater than 388. This is less accurate for small numerators so for numerators less than 389 an exact method based on the Poisson distribution2 has been used. The methodology is explained in detail in the APHO Technical Briefing on Commonly Used Public Health Statistics and their Confidence Intervals http://www.apho.org.uk/resource/item.aspx?RID=48457 and the associated Excel tool http://www.apho.org.uk/resource/view.aspx?RID=48617. 1. Dobson A et al. Confidence intervals for weighted sums of Poisson parameters. Stat Med 1991;10:457-62. 2. Breslow NE, Day NE. Statistical methods in cancer research, volume II: The design and analysis of cohort studies. Lyon: International Agency for Research on Cancer, World Health Organisation; 1987. 3. Armitage P, Berry G. Statistical methods in medical research (3rd edn). Oxford: Blackwell; 1994.
Health Profiles 2009 The Indicator Guide
172 Section 5: disease and poor health
19. MENTAL HEALTH INDICATOR Basic Information 1. What is being measured?
Working age people who are in receipt of benefits for mental health conditions.
2. Why is it being measured?
People with long term psychiatric disabilities are less likely to be in employment than those with longterm physical disabilities, despite indications that most people with severe mental illness would like to work. To help improve the provision of services to help mentally ill people find work and reduce social exclusion.
3. How is this indicator actually defined?
Claimants/beneficiaries of incapacity benefit/severe disablement allowance, with mental or behavioural disorders. Crude rate, all persons of working age, per thousand working age population, 2007.
4. Who does it measure?
Working age adults: 16–64 males and 16–59 females.
5. When does it measure it?
Reported quarterly and updated every year.
6. Will It measure absolute numbers or proportions?
Proportions: number of working age claimants per thousand working age population.
7. Where does the data actually come from?
Collection and collation from incapacity benefits data via the Department for Work and Pensions.
8. How accurate and complete will the data be?
The data is based on all claims for benefit (100% sample) and double counts are removed.
9. Are there any caveats/warnings/ problems?
Benefit counts are rounded quarterly point estimates from the end of February, May, August and November 2008. Population data are mid-2007 and rounded to the nearest 100.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
Pg 4 Health Summary: Indicator No 19.
Subject category/ domain(s)
Disease and poor health.
Indicator name (*Indicator title in health profile)
Claimants/beneficiaries of incapacity benefit/severe disablement allowance with mental or behavioural disorders (*Incapacity benefits for mental illness).
Health Profiles 2009 The Indicator Guide
173 Section 5: disease and poor health
PHO with lead responsibility
NEPHO
Date of PHO dataset creation
February 2009
Indicator definition
Claimants/beneficiaries of incapacity benefit/severe disablement allowance with mental or behavioural disorders, crude rate, males and females, working age, 2007, Per 1000 working age population.
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs.
Timeliness
This indicator is potentially suitable for time-series analysis.
Rationale: What this indicator purports to measure
Prevalence of working age people with severe mental illness who are in receipt of benefits for mental health conditions.
Rationale: Public Health Importance
This is a proxy measure of levels of severe mental illness in the community, and a direct measure of socio-economic disadvantage in those ‘not in work’ because of mental illness. Severe mental illness severely restricts the capacity to fully participate in society and in particular the employment market. Unemployment rates are high amongst people with severe mental illness. In the UK unemployment rates of 60–100% have been reported. These high rates reflect the disability caused by severe mental illness, but they also reflect discrimination (unemployment rates are higher than in other disabled group) and the low priority given to employment by psychiatric services. People with long-term psychiatric disabilities are even less likely to be in employment than those with long-term physical disabilities. Despite high unemployment rates amongst the severely mentally ill, surveys have consistently shown that most want to work. These low rates of employment should be considered against the facts that at least 30-40% of people who are significantly disabled by enduring mental illness are capable of holding down a job. More than 900,000 adults in England claim sickness and disability benefits for mental health conditions. This group is now larger than the total number of unemployed people claiming Jobseeker’s Allowance in England. Vocational rehabilitation services can help mentally ill people find work. See: Crowther R, Marshall M, Bond G, Huxley P. Vocational rehabilitation for people with severe mental illness. Cochrane Database of Systematic Reviews 2001, Issue 2. Art. No.: CD003080. DOI: 10.1002/14651858.CD003080 See Royal College of Psychiatrists Memorandum to Select Committee on Work and Pensions available at: http://www.publications.parliament.uk/pa/cm200506/cmselect/ cmworpen/616/6022719.htm
Rationale: Purpose behind the inclusion of the indicator
To estimate the prevalence of those with severe mental illness who are not in work because of their mental ill health. To help improve the provision of services for helping mentally ill people find work
Health Profiles 2009 The Indicator Guide
174 Section 5: disease and poor health
Rationale: Policy relevance
In June 2004 the Social Exclusion Unit of the Office of the Deputy Prime Minister (ODPM) published the ‘Mental Health and Social Exclusion’ report. This report highlights the large number of adults in England claiming sickness and disability benefits for mental health conditions (approximately 40% of all claims) with the statistics showing that more adults now fall into this group than the total number of unemployed people claiming Jobseeker’s Allowance. The report draws attention to the fact that adults with longterm mental health problems are one of the most excluded groups in society facing numerous barriers that only serve to stop them from achieving their full potential as individuals and members of the community as a whole. Current government policy is focusing on finding ways of returning people to work via initiatives such as Pathways to Work, New Deal for Disabled People, and Disability Employment Brokers. Mental Health National Service Framework
Interpretation: What a high/low level of indicator value means
An indicator value worse than average (red circle in health summary chart) represents a statistically significant worse rate of benefit claimants for mental and behavioural disorders for that local authority when compared to the national value. An indicator value better than average (green circle in health summary chart) represents a statistically significant better rate of benefit claimants for mental and behavioural disorders for that local authority when compared to the national value.
Interpretation: Potential for error due to type of measurement method
The first 28 weeks of incapacity are assessed under the “own occupation test” which looks at a person’s ability to do their usual job and is based on medical certificates from a GP. After this time, the personal capability assessment (PCA) applies which involves completing an incapacity questionnaire (IB50) that assesses ability to do any work. Forms are difficult to understand and complete. There may be an “institutional bias” against people with mental health problems in the incapacity benefit questionnaire which does not establish information about fluctuating conditions. The DWP uses a system that allocates points to certain activities and tasks, with 10 points needed to determine a person’s eligibility on grounds of mental incapacity. Mental health descriptors are not itemised in the same detail as physical descriptors and a client is therefore less likely to answer in the way most helpful to a successful outcome in their case. Problems those with severe mental illness face with personal capability assessments conducted by Medical Services include: • doctors not listening to clients; • poor recording of clinical findings; • incorrect assumptions based on information from the client and from the medical examination; • effects of mental illness not appropriately taken into account by the scoring system employed; and • difficulties in arranging home visits for some clients. The assessment process may lead to underestimation of unemployment due to severe mental illness. See Royal College of Psychiatrists Memorandum to Select Committee on Work and Pensions available at: http://www.publications.parliament.uk/pa/cm200506/cmselect/ cmworpen/616/6022719.htm
Health Profiles 2009 The Indicator Guide
175 Section 5: disease and poor health
Interpretation: Potential for error due to bias and confounding
There may be disincentives to return to work: These include: • Medical review process: There are concerns that engagement in voluntary work, education and training – which can help get back to work - may trigger the medical review process. • Permitted work rules: These may deter people who will be at risk of losing incapacity benefit and associated benefits. • Transition to work may disrupt income: People returning to work will lose housing benefit. Should the return to work fail, they will be at risk of losing their home. • Fear of drop in income. • 52-week linking rule: The time period may not be long enough for people whose illness has a relapsing and remitting course. (For example, the average time to clinical relapse for people being treated for schizophrenia is approximately two years.) • Difficulties in getting financial support for practical needs at work. • Lack of expert, independent benefits advice on return to work. Therefore severe mental illness as a cause of inability to participate in the labour market may be overestimated due to these disincentives to return to work. See Royal College of Psychiatrists Memorandum to Select Committee on Work and Pensions available at: http://www.publications.parliament.uk/pa/cm200506/cmselect/ cmworpen/616/6022719.htm There are groups of people who may be less likely to engage with, or be able to benefit from, the benefit system eg certain ethnic minorities, married women, those with functional illiteracy, rough sleepers. This may underestimate the problem.
Health Profiles 2009 The Indicator Guide
176 Section 5: disease and poor health
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Table 2 – Indicator Specification Indicator definition: Variable
Claimants/beneficiaries of incapacity benefit/severe disablement allowance with mental or behavioural disorders.
Indicator definition: Statistic
Crude rate.
Indicator definition: Gender
Males and Females.
Indicator definition: age group
16–64 yrs (males working age) 16– 59 yrs (females working age)
Indicator definition: period
2007
Indicator definition: scale
Per 1000 working age population.
Health Profiles 2009 The Indicator Guide
177 Section 5: disease and poor health
Geography: geographies available for this indicator from other providers
Super Output Area, Wards, England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs.
Dimensions of inequality: subgroup analyses of this dataset available from other providers
None.
Data extraction: Source
Numerator: NOMIS Denominator: Office for National Statistics (ONS)
Data extraction: source URL
Numerator: http://www.nomisweb.co.uk Denominator: http://www.statistics.gov.uk/statbase/Product.asp?vlnk=15106
Data extraction: date
4 February 2009.
Numerator: definition
Count of working age claimants/beneficiaries of incapacity benefit/severe disablement allowance with mental or behavioural problems, 2007.
Numerator: source
NOMIS/Department for Work and Pensions.
Denominator: definition
Estimated count of working age people (rounded to nearest 100). For the purposes of calculating this indicator working age has been defined as 16 to 64 years for males and 16 to 59 years for females, 2007.
Denominator: source
Office for National Statistics (ONS).
Data quality: Accuracy and completeness
This is a 100% data source with double-counts removed.
Available from Department of Work & Pensions.
Data relating to the age and gender of benefit claimants is collected by the DWP and may be available upon request.
Table 3 – Indicator Technical Methods Numerator: extraction
Quarterly benefit claimant statistics are published routinely on the NOMIS website (http://www.nomisweb.co.uk). The advanced query tool provided on this website was used to extract relevant data at local authority district level.
Numerator: aggregation/ allocation
Average annual benefit count figures at local authority district level were derived from quarterly benefit count figures taken at the end of February, May, August and November 2007.
Numerator data caveats
The numerator counts are derived from quarterly figures recorded as point estimates taken at the end of February, May, August and November 2007. Aggregated numerator data is based upon rounded data. This can result in a margin of error.
Denominator data caveats
Local authority level working-age mid-2007 population estimates. Data are rounded to nearest 100 to comply with ONS publication policy.
Health Profiles 2009 The Indicator Guide
178 Section 5: disease and poor health
Methods used to calculate indicator value
Mental Health Indicator: Calculation of the numerator: average annual count of claimants/beneficiaries of incapacity benefit/severe disablement allowance February to November 2007 was calculated by adding four quarterly counts together and dividing the result by four. This value was then rounded to the nearest value of ten. Totals for counties and regions were calculated by aggregating up the local authority district level data to the appropriate geography. Calculation of the denominator: Persons level working age population data was downloaded from the ONS website. The local authority values were extracted from this table and these values were then multiplied by 1000 to give unit level figures rounded to the nearest number of 100. For new the unitary authorities where totals of old local authority district level populations did not total up to ONS county council population, the ONS total for the counties was used. The numerator was then divided by the denominator; the resulting value was then multiplied by 1000 to give a crude rate per 1,000 working age population.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Isles of Scilly and City of London have been included in regional and England numerators and denominators. Isles of Scilly has been included in the numerator and denominator for the County of Cornwall.
Disclosure Control
To preserve claimant confidentiality this data is rounded by the DWP prior to publication. Numerator values have been rounded to the nearest value of ten as a means of disclosure control. Denominator values are rounded to the nearest value of 100 as a means of disclosure control.
Confidence Intervals calculation method
The 95% confidence intervals are calculated with the method described by Wilson and by Newcombe which is a good approximation of the exact method. First calculate the estimated proportions of subjects with (p) and without (q) some feature of interest from a sample of size n. proportion with feature of interest = p = r/n proportion without feature of interest = q = 1 - p where r is the observed number of subjects with the feature of interest. Second, calculate the three quantities A = 2r + z2;
B = z z 2 + 4rq ; and C=2(n+z2),
where z is the appropriate value, z1-α/2, from the standard Normal distribution. Then the confidence interval for the population proportion is given by (A-B)/C to (A+B)/C This method has the considerable advantage that it can be used for any data. When there are no observed events, r and hence p are both zero, and the recommended confidence interval simplifies to 0 to z2/(n+z2). When r = n so that p = 1, the interval becomes n/(n+z2) to 1. Wilson EB. J Am Stat Assoc 1927, 22, 209-212 Newcombe, RG. Two-sided confidence intervals for the single proportion: comparison of seven methods. Stat Med 1998;17:857-72.
Health Profiles 2009 The Indicator Guide
179 Section 5: disease and poor health
20. HOSPITAL STAYS FOR ALCOHOL RELATED HARM INDICATOR Basic Information 1. What is being measured?
Hospital stays for Alcohol Related Harm: NI39.
2. Why is it being measured?
The acute or long term effects of excessive alcohol consumption are a major cause of avoidable hospital admissions. This indicator may help to monitor likely health care burden.
3. How is this indicator actually defined?
Hospital Admissions for Alcohol Related Harm (2007/08), directly age and sex standardised rate, all ages, admissions per 100,000 European Standard population.
4. Who does it measure?
All admissions, all ages.
5. When does it measure it?
Continually reported and updated every year.
6. Will it measure absolute numbers or proportions?
Proportions: numbers of admissions per hundred thousand European standard population.
7. Where does the data actually come from?
Collection and collation from Hospital Episode Statistics via the DH.
8. How accurate and complete will the data be?
HES Data and ONS population statistics are considered to be complete and robust.
9. Are there any caveats/ warnings/ problems?
Hospital admission data can be coded differently in different parts of the country.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
Pg 4 Health Summary – Indicator No. 20.
Subject category/ domain(s)
Disease and poor health.
Indicator name (*Indicator title in health profile)
Hospital stays for alcohol-related harm.
Health Profiles 2009 The Indicator Guide
180 Section 5: disease and poor health
PHO with lead responsibility
NWPHO.
Date of PHO dataset creation
March 2009.
Indicator definition
Hospital admissions for alcohol-related harm 2007/08, directly age and sex standardised hospital admissions for all ages, per 100,000 European Standard population, based on residence in the area. Full indicator definition can be found at: http://www.nwph.net/alcohol/lape/NI39Technical_Dec2008.pdf
Geography
The following geographies: England, GOR, Local Authority: County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs, for this and other alcohol related indicators are available from the NWPHO web site. www.nwph.net/alcohol/lape
Timeliness
Updated every year, next update available in 2010. Time trend analysis is appropriate from data available from the NWPHO. Time trend analysis should not be done between Health Profiles 2006, Health Profiles 2007 and Health Profiles 2008 because the indicator calculation changed from multiple years to a single year.
Rationale: What this indicator purports to measure
This indicator measures the number of admissions to hospital for alcohol-related harm. A comprehensive list of ICD10 codes, age and sex specific attributable fractions can be found at: http://www.nwph.net/alcohol/lape/NI39Technical_ Dec2008.pdf
Rationale: Public Health Importance
The acute or long term effects of excessive alcohol consumption are a major cause of avoidable hospital admissions.
Rationale: Purpose behind the inclusion of the indicator
To help monitor likely health care burden.
Rationale: Policy relevance
Alcohol Harm Reduction Strategy for England (2004): http://www.cabinetoffice. gov.uk/media/cabinetoffice/strategy/assets/caboffce%20alcoholhar.pdf. Reducing Alcohol Harm (2008) http://www.dh.gov.uk/en/Publichealth/Healthimprovement/Alcoholmisuse/ DH_085387 This indicator is included in three key indicator sets and performance management frameworks, these are: Vital Signs Indicator VSC26, National Indicator NI39 and Public Service Agreement Indicator 25.2.
Interpretation: What a high/low level of indicator value means
A high indicator value (red circle in health summary chart) represents a statistically significant higher level of hospital admissions for alcohol-related harm for that local authority when compared to the national value. A low indicator value (green circle in health summary chart) represents a statistically significant lower level of hospital admissions for alcohol-related harm for that local authority when compared to the national value.
Interpretation: Potential for error due to type of measurement method
Hospital admission data can be coded differently in different parts of the country.
Health Profiles 2009 The Indicator Guide
181 Section 5: disease and poor health
Interpretation: Potential for error due to bias and confounding Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated significance level. In Health Profiles this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with a white symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or amber symbol depending on whether it is worse or better than the national value respectively.
Table 2 – Indicator Specification Indicator definition: Variable
Hospital Admissions for Alcohol-Related Harm.
Indicator definition: Statistic
Directly age and sex standardised rate.
Indicator definition: Gender
Persons.
Indicator definition: age group
All Ages.
Indicator definition: period
2007/08.
Indicator definition: scale
Per 100,000 European Standard population.
Health Profiles 2009 The Indicator Guide
182 Section 5: disease and poor health
Geography: geographies available for this indicator from other providers
The following geographies: England, GOR, Local Authority: County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs, for this and other alcohol related indicators are available from the NWPHO web site. www.nwph.net/alcohol/lape
Dimensions of inequality: subgroup analyses of this dataset available from other providers
HES data is available by LSOA, therefore analysis by deprivation and geodemographics is possible.
Data extraction: Source
The Department of Health (from HES).
Data extraction: source URL
Hospital Episodes Statistics (HES). Record level access is restricted to authorised users, for example PHOs. Further information regarding HES can be found at: www.hesonline.org.uk
Data extraction: date
February 2009
Numerator: definition
A detailed definition of the numerator data used for this indicator can be found at: http://www.nwph.net/alcohol/lape/NI39Technical_Dec2008.pdf
Numerator: source
Produced by the Department of Health.
Denominator: definition
Mid-year population estimates (2007) by 5-year age band.
Denominator: source
Office for National Statistics (ONS).
Data quality: Accuracy and completeness
HES Data and ONS population statistics are considered to be complete and robust.
TABLE 3 – INDICATOR TECHNICAL METHODS Numerator: extraction
A detailed definition of the extraction method and application of attributable fractions to produce the numerator data used for this indicator can be found at: http://www.nwph.net/alcohol/lape/NI39Technical_Dec2008.pdf
Numerator: aggregation / allocation
Data extracted using area of residence (GOR, County and LA) of patient.
Numerator data caveats
Hospital admission data can be coded differently in different parts of the country. Identification of persons in any one year can also affect the analysis if people are prone to moving within a local authority on a regular basis In some cases the PCT of residence is recorded in HES but not the LA. To ensure that the figures for coterminous PCTs and LAs were the same, details of LA were added where this information could be ascertained from PCT of residency.
Denominator data caveats
Health Profiles 2009 The Indicator Guide
183 Section 5: disease and poor health
Methods used to calculate indicator value
The directly age-standardised rate is the rate of events that would occur in a standard population if that population were to experience the age/sex-specific rates of the subject population. Explicitly:
∑w r DSR = ∑w
i i
i
i
× 100,000
i
(expressed per 100,000 population) where: wi is the number, or proportion, of individuals in the standard population in age/sex group i. ri is the crude age/sex-specific rate in the subject population in age/sex group i, given by:
ri =
Oi ni
where: Oi is the observed number of events in the subject population in age/sex group i. ni is the number of individuals in the subject population in age/sex group i. The standard population generally used for the direct method is the European Standard Population. Small Populations: How Isles of Scilly and City of London populations have been dealt with
Isles of Scilly and City of London were excluded. Though these are available from www.nwph.net/alcohol/lape .
Disclosure Control
Not applicable
Health Profiles 2009 The Indicator Guide
184 Section 5: disease and poor health
Confidence Intervals calculation method
Confidence intervals for the rates were calculated using the method described in the NCHOD Compendium for directly standardised rates. www.nchod.nhs.uk 95% confidence intervals for the age-standardised rates were calculated using a normal approximation. Standard errors are obtained using the method described by Breslow and Day but modified to use the binomial variance for a proportion to estimate the variances of the crude age/sex-specific rates. This method is likely to be unreliable when there are fewer than 50 cases in an area, hence confidence intervals for rates based on less than 50 cases should be viewed with caution. The lower and upper limits for the rates are denoted by DSRLL and DSRUL respectively.
DSRLL / UL = DSR ± 1.96 × 100,000 ×
1 ∑ wi ij
2
×∑ ij
wi2 ⋅ rij (1 − rij ) nij
(expressed per 100,000 population)
where: wi is the number, or proportion, of individuals in the standard population in age/sex group i. rij is the crude age/sex-specific rate in the subject population in age/sex group i, in year j. nij is the number of individuals in the subject population in age/sex group i, in year j. Ref: Breslow NE and Day NE. Statistical Methods in Cancer Research, Volume II: The Design and Analysis of Cohort Studies. Lyon: International Agency for Research on Cancer, World Health Organization, 1987: 59 Keyfitz N. Sampling variance of age-standardised mortality rates. Human Biology. 1966; 38: 309-317.
Health Profiles 2009 The Indicator Guide
185 Section 5: disease and poor health
21. DRUG MISUSE Basic Information 1. What is being measured?
Estimated Problem Drug Users (Crack &/or Opiates)
2. Why is it being measured?
To help monitor likely health care burden from drug misuse.
3. How is this indicator actually defined?
Estimated problem drug users (Crack &/or Opiates), crude rate, 15–64 Ages, 2006/07, persons.
4. Who does it measure?
All persons, 15–64 Ages.
5. When does it measure it?
Continually reported and updated every year
6. Will it measure absolute numbers or proportions?
Proportions: number of cases Per 1,000 residents aged 15–64 years
7. Where does the data actually come from?
Collection and collation from The National Treatment Agency.
8. How accurate and complete will the data be?
County district estimates are based on a model applied to top tier local authorities. This model relied on a probabilistic allocation of cases to Local Authorities based on postal sector of residence. A regression model uses number of users in treatment (2006/07) to disaggregate the County level estimates to County Districts.
9. Are there any caveats/ warnings/ problems?
The number of users in treatment will have impacted on the prevalence estimates published by the Home Office and the regression model which may introduce bias to the estimates.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
Pg 4 Health Summary – Indicator No. 21
Subject category/ domain(s)
Disease and poor health
Indicator name (*Indicator title in health profile)
Drug misuse
PHO with lead responsibility
NWPHO
Date of PHO dataset creation
Revised January 09
Health Profiles 2009 The Indicator Guide
186 Section 5: disease and poor health
Indicator definition
Estimated Problem Drug Users (Crack and/or Opiates), Crude Rate, 15–64 Ages, 2006/07, persons
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs
Timeliness
Every year. Time trend analysis is not appropriate.
Rationale: What this indicator purports to measure
This indicator estimates the number of problem drug users (Crack and/or Opiates) in an area.
Rationale: Public Health Importance
The indicator was chosen as the best available estimate of drug use prevalence in an area.
Rationale: Purpose behind the inclusion of the indicator
To help monitor likely health care burden from drug misuse.
Rationale: Policy relevance
The Department of Health has introduced a ten year drug strategy (20082018) which is aimed at restricting the supply of illegal drugs and reducing the demand for them. http://www.dh.gov.uk/en/Publichealth/Healthimprovement/ Drugmisuse/DH_085886
Interpretation: What a high/low level of indicator value means
A high indicator value (red circle in health summary chart) represents a statistically significant higher estimate of problem (crack and/or opiates) drug users for that local authority when compared to the national value. A low indicator value (amber circle in health summary chart) represents a statistically significant lower estimate of problem (crack and/or opiates) drug users for that local authority when compared to the national value. Confidence Intervals for Top Tier Local authorities (Counties, MCDs, UAs, LBs) were taken from the HO published prevalence data but cannot currently be calculated for County Districts. Therefore interpretation of significance for County Districts cannot be made.
Interpretation: Potential for error due to type of measurement method
The base estimates of the number of problem drug users were published by the Home Office (2004/05) and issues with the methods are outlined in the report (www.homeoffice.gov.uk/rds/pdfs06/rdsolr1606.pdf). A regression model uses number of users in treatment (2006/07) to disaggregate the County level estimates to County Districts. Treatment data is only available with postcode sector of residence which is not available for a large proportion of cases reported by DAATs, consequently the allocation of cases to LAs is likely to be less accurate for some areas than for others.
Interpretation: Potential for error due to bias and confounding
The number of users in treatment will have impacted on the prevalence estimates published by the Home Office and the regression model which may introduce bias to the estimates.
Health Profiles 2009 The Indicator Guide
187 Section 5: disease and poor health
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with a white symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or amber symbol depending on whether it is worse or better than the national value respectively.
Table 2 – Indicator Specification Indicator definition: Variable
Problem Drug Users (Crack and/or Opiates)
Indicator definition: Statistic
Crude Rate
Indicator definition: Gender
Persons
Indicator definition: age group
15–64
Indicator definition: period
2006/07
Indicator definition: scale
Per 1,000 residents aged 15–64 years
Geography: geographies available for this indicator from other providers
The data is available by DAAT from the Home Office and NTA.
Health Profiles 2009 The Indicator Guide
188 Section 5: disease and poor health
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Analysis of the 2006/07 PDU estimates by Index of Multiple Deprivation (IMD, 2007) is available in the APHO publication ‘Indications of Public Health in English Regions 10: Drug Use which will be published in April 2009.
Data extraction: Source
Source of data: The National Treatment Agency
Data extraction: source URL
The data extraction source URL is not available to the public domain
Data extraction: date
Data extracted from source July. 2007
Numerator: definition
Estimate of resident persons aged 15–64 believed to be problem drug users in 2006/07.
Numerator: source
The National Treatment Agency
Denominator: definition
2006 Mid-year population estimates (persons aged 15–64) by Local Authority District, County, Region and England.
Denominator: source
Office for National Statistics (ONS).
Data quality: Accuracy and completeness
This is a modelled estimate of the number of problem drug users in a local authority district. The number of users in treatment is based on probabilistic allocation of cases to Local Authorities from postal sector of residence. The National Treatment Agency provided data on individuals in contact with structured treatment with their DAT of residence and PCT of residence. Each case was allocated into LA and GOR of residence using a schema provided by NWPHO. There were 451 cases in which there were no DAT and PCT of residence, these were not used in the regression model. Coverage includes all DAAT areas.
Table 3 – Indicator Technical Methods Numerator: extraction
Downloaded from NTA website.
Numerator: aggregation/ allocation
See Methods used to calculate indicator value
Numerator data caveats
The number of users in treatment will have impacted on the prevalence estimates published by the Home Office and the regression model which may introduce bias to the numerator (also see above).
Health Profiles 2009 The Indicator Guide
189 Section 5: disease and poor health
Denominator data caveats
The Mid 2006 Resident Population Estimates for All persons15 to 64 year age group has been used. Census data and mid-year estimates are known to be deficient in their estimates of: • Non-white populations • Full-time students • Men aged 20–39 • People living in nursing homes etc • Rough sleepers • Inner-city populations • Households of multiple occupation • Migrants
Methods used to calculate indicator value
The analysis uses the problem drug user prevalence data (2006/07) provided by the HO & NTA for Top Tier Local Authorities (Counties, MCDs, UAs, LBs) and the number of opiate and/or crack users in treatment data by Local Authority District provided by the NTA (2006/07). To allow for disaggregating of the County level prevalence data to County Districts, a model was created based on the number of opiate and/or crack users in treatment and the prevalence for Unitary Authorities and Metropolitan County Districts. The results of this model were then used to estimate the likely prevalence of problem drug users for the given number of opiate and crack users in treatment for County Districts. These totals for County Districts were then reconciled to ensure consistency with the County Prevalence Total. In future the Home Office will calculate direct Multiple Indicator Methods (MIM) and Capture Recapture (CR) estimates of drug users for all LADs. These current estimates (for HP3) are an interim measure until the new HO data are available.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Isles of Scilly and City of London were excluded.
Disclosure Control
Not applicable
Confidence Intervals calculation method
The Confidence Intervals were calculated for the estimated prevalence of drug users (opiate and/or crack users) with the Exact method. This method takes no account of uncertainty in the model used to estimate numbers of drug users at the County Districts.
Health Profiles 2009 The Indicator Guide
190 Section 5: disease and poor health
22. PEOPLE DIAGNOSED WITH DIABETES INDICATOR Basic Information 1. What is being measured?
Prevalence of recorded diabetes
2. Why is it being measured?
Diabetes is a common disease with serious consequences. It is the 5th leading cause of death globally and accounts for about 10% of NHS costs.
3. How is this indicator actually defined?
The prevalence of QOF-recorded diabetes (in adults aged 17+) in the population.
4. Who does it measure?
All persons, all ages
5. When does it measure it?
Patients registered with GP practices, aged 17 and over at midnight on the 31st March 2008, with a coded diagnosis of diabetes on the 1st April 2007. (QOF DM1).
6. Will It measure absolute numbers or proportions?
Proportion (displayed as a percentage): Number of recorded cases of diabetes in adults aged 17+ per 100 resident population.
7. Where does the data actually come from?
The published QOF information is derived from the Quality Management Analysis System (QMAS), a national system developed by NHS Connecting for Health, and is published online by the Information Centre.
8. How accurate and complete will the data be?
The data covers more than 99% of GP-registered patients in England, although not everyone is registered with a GP (especially some groups with particular needs).
9. Are there any caveats/warnings/ problems?
Potential errors in collection, collation and interpretation. It is a measure of recorded diabetes prevalence and not actual prevalence and therefore under-reports groups who are less likely to be registered with a GP, such as ethnic populations, young people, homeless people, migrants and travellers. Record level data is not available therefore we have to apportion practices to LAs based on a look up file created from NSTS data. The NSTS practice file does not completely match the QOF practices and therefore the apportioning may not mirror exactly where the patients come from, but it is the best approximation with the data provided. The diabetes prevalence will not match up with the QOF published prevalence where the areas are identical due to a different denominator being used (QOF uses the practice list size whereas the Health Profiles use the resident population).
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Health Profiles 2009 The Indicator Guide
191 Section 5: disease and poor health
TABLE 1 – INDICATOR DESCRIPTION Information component
P4 Health Summary – Indicator No 22
Subject category/ domain(s)
Disease and poor health
Indicator name (*Indicator title in health profile)
Prevalence of recorded diabetes (“People diagnosed with diabetes”)
PHO with lead responsibility
ERPHO
Date of PHO dataset creation
18th February 2009
Indicator definition
The prevalence of QOF-recorded diabetes (in adults aged 17+) in the population.
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs
Timeliness
Data is extracted from the QMAS system annually in June and published in QPID (quality and prevalence indicators database) in September-October each year.
Rationale: What this indicator purports to measure
Prevalence of recorded diabetes.
Rationale: Public Health Importance
Diabetes is a common disease with serious consequences. It is the 5th leading cause of death globally and accounts for about 10% of NHS costs. The burden falls disproportionately on elderly and ethnic populations. We use the indicator in this context as a proxy for healthcare need and demand (a high prevalence of diabetes can indicate a less healthy population with higher service utilisation). The sequelae of diabetes include blindness, amputation, neuropathy, renal disease, heart disease and other complications. It is treatable and in most cases preventable. Important modifiable risk factors are obesity, diet and lack of physical activity.
Rationale: Purpose behind the inclusion of the indicator
To encourage better collection of the primary care data to give more accurate estimates of disease prevalence. To monitor diabetes prevalence To emphasise the burden of disease To encourage preventative action
Rationale: Policy relevance
The twelve standards of the Diabetes National Service Framework cover all aspects of diabetes care and prevention, and together with the Delivery Strategy, set out a ten-year programme of change and improvement which will raise the quality of services and reduce unacceptable variations. http://www.dh.gov.uk/en/Healthcare/NationalServiceFrameworks/Diabetes/ index.htm
Health Profiles 2009 The Indicator Guide
192 Section 5: disease and poor health
Interpretation: What a high/low level of indicator value means
An indicator value better than average (green circle in health summary chart) represents a statistically significant lower number of people diagnosed with diabetes compared with the national average. An indicator value worse than average (red circle in health summary chart) represents a statistically significant higher number of people diagnosed with diabetes compared with the national average. A high value can indicate a genuinely high prevalence and/or better detection and recording. Conversely a low value may indicate genuinely low prevalence and/or poor detection and recording.
Interpretation: Potential for error due to type of measurement method
See above.
Interpretation: Potential for error due to bias and confounding
There may be under-representation of young people, ethnic populations and other vulnerable groups e.g. homeless, travellers in the numerator as they are less likely to be registered with the GP.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate.
Due to the fact that recording is rewarded through QOF points there may be potential for “gaming”. There are a large number of codes used to record diabetes on GP systems which may lead to counting errors depending on how the data is extracted (see the QOF definitions for the codes used). There may also be potential biases in the attribution of practice populations to local authority areas but these are probably small.
This uncertainty arises because factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider the confidence interval the greater the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the confidence interval can be used to test whether the value is statistically significantly different to the national value. If the confidence interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the confidence interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Health Profiles 2009 The Indicator Guide
193 Section 5: disease and poor health
Table 2 – Indicator Specification Indicator definition: Variable
The prevalence of QOF-recorded diabetes (in adults aged 17+) in the population.
Indicator definition: Statistic
Proportion (displayed as a percentage): Number of recorded cases of diabetes in adults aged 17+ per 100 resident population
Indicator definition: Gender
Persons
Indicator definition: age group
Numerator 17+ Denominator all ages
Indicator definition: period
Financial year 2007/08
Indicator definition: scale
Per 100 resident population
Geography: geographies available for this indicator from other providers
National, SHA, PCT, general practice.
Dimensions of inequality: subgroup analyses of this dataset available from other providers
No (but could be grouped by deprivation score at practice level using derived IMD scores)
Data extraction: Source
Numerator: The Information Centre for health and social care. Denominator: ONS 2007 population estimates Weight: A special extract (mapping general practice populations to local authorities for February 2007) of the NHS Strategic Tracing Service (NSTS) database commissioned from ATOS Origin (available on request). (The NSTS is part of NHS Connecting for Health and is an amalgamation of the Exeter System and the NHS central register.)
Data extraction: source URL
http://www.ic.nhs.uk/statistics-and-data-collections/supporting-information/ audits-and-performance/the-quality-and-outcomes-framework/qof-2007/08/datatables http://www.statistics.gov.uk/statbase/Product.asp?vlnk=15106 http://www.connectingforhealth.nhs.uk/nsts
Data extraction: date
Numerator: QOF data downloaded in October 2008 Denominator: August 2008 Attribution table supplied: February 2009
Health Profiles 2009 The Indicator Guide
194 Section 5: disease and poor health
Numerator: definition
Patients registered with GP practices, aged 17 and over at midnight on the 31st March 2008, with a coded diagnosis of diabetes on the 1st April 2007. (QOF DM1). The case definition of diabetes (with Read codes) can be found at: http://www. primarycarecontracting.nhs.uk/uploads/QOF/qof_bus_rules_v9/diabetes_ruleset_ r4_v9.0.pdf
Numerator: source
Quality and Outcomes Framework (QOF). http://www.ic.nhs.uk/webfiles/QOF/2007-08/Data%20tables/Practice/ Domain%20level%20Spreadsheets/LR/QOF0708_Pracs_Prevalence.xls
Denominator: definition
Mid-year 2007 LA estimates (latest year available). We used a resident population as a denominator rather than an apportioned registered population to avoid the list inflation in registered populations. Denominator data has been used exactly as published in the Home Office supplied data. Where new Unitary Authorities that have been created as part of the April 2009 boundary changes are exactly co-terminous with pre-existing counties (Durham, Northumberland, Shropshire and Wiltshire), the denominator data has been drawn from the relevant county figures. Where the new Unitary Authorities represent only a part of the existing counties (Cornwall, Cheshire East, Cheshire West & Chester, Bedford and Central Bedfordshire), denominator data has been aggregated from the relevant published figures for the constituent districts. We assumed the problems of list counting and inflation did not apply to the numerator (it is very unlikely that diabetics will be counted twice and there are QA checks in QOF to avoid this).
Denominator: source
Office for National Statistics (ONS). http://www.statistics.gov.uk/downloads/ theme_population/Mid_2007_UK_England_&_Wales_Scotland_and_Northern_ Ireland%20_21_08_08.zip
Data quality: Accuracy and completeness
QOF represents 99.8 per cent of registered patients in England (based on registration data from the ePACT system of the Prescription Pricing Division of the NHS Business Services Authority, January to March 2008). There were 8294 practices in the QOFR 200708 dataset, and 99.9% of these could be matched to a LA(s) using the NSTS attribution lookup. Users of data derived from QMAS should recognise that QMAS was established as a mechanism to support the calculation of practice QOF payments and not as a person-based epidemiological tool. It is not a comprehensive source of data on quality of care in general practice, but it is potentially a rich and valuable source of such information, providing that the limitations of the data are acknowledged. See also QOF assessor validation reports available at: http://www. connectingforhealth.nhs.uk/delivery/programmes/qof/
Health Profiles 2009 The Indicator Guide
195 Section 5: disease and poor health
Table 3 – Indicator Technical Methods Numerator: extraction
Numerators extracted from downloaded QOF prevalence by general practice spreadsheets available on the IC website.
Numerator: aggregation/ allocation
We assigned counts of patients with recorded diabetes to give an estimated resident number of diabetics as follows: The weights We obtained a cross boundary flow table from NSTS (February 2007) enabling us to assign GP-registered populations to LA of residence, then for each practice we calculated the number of persons resident in relevant LAs using the following SQL: We discounted any null returns or practices that didn’t match on to QOF, and excluded populations where the number of patients in a practice from a particular ≤50. This LA was made the weighting simpler. In addition, as the practice population is always greater than the actual population of the country, this exclusion brings the total practice population closer to the ONS estimate. We calculated the proportion of each GP-registered population in each LA. We used these proportions to distribute the counts of diabetic patients in each LA, assuming a uniform spatial distribution of diabetes patients practice to each within each practice. We summed the LA estimated counts to whole LAs to give an overall figure for the LA.
Ʃ
This can be expressed in a formula as:
Ʃ
KEY P-LApop: population of a practice from a particular LA in the cases where this
population is >50 The look up table is derived from a Practice–LA look-up table provided by the NSTS Ppop: Total practice population excluding P-LA populations which are ≤50 Pnum: Practice number of patients registered with disease (QOF)
Health Profiles 2009 The Indicator Guide
196 Section 5: disease and poor health
Where the PCT has an identical boundary to the LA, the values we generated for the LA were slightly different from those published for the PCT by QOF because the practice–PCT look-up used by QOF is different from the one we derived from NSTS, therefore to keep our data consistent with the published data, we replaced our calculated figures with the published ones for these 91 PCTs. Due to the fact that a few practices were not included in the apportioning, and the differing look-ups used by the QOF team compared to our generated look-up, the overall figure for registered diabetes patients derived from calculations was less than the overall QOF published figure, and so the remaining LAs that were not identical to PCTs were scaled up slightly using a weight calculated by:
This scaling weight was 1.003, which meant that the non-PCT-identical LA figures were being scaled up by 0.3%. The LA level data was then aggregated to higher levels using look-ups. Numerator data caveats
The allocation method may have incorrectly apportioned patients to LAs particularly for practices straddling LA boundaries. We have not found a satisfactory way of assessing this, as there is no definitive list of practices in England, and the NSTS file does not quite include all practices in QOF and vice versa. There is no age adjustment so variations in prevalence need to be interpreted in the light of variations in age structures. (It may for example, be helpful to correlate the crude prevalence presented here with the proportion of the population over 65)
Denominator data caveats
Subject to limitations of ONS population estimation methods. The denominator does not quite correspond with ageband of the numerator. To keep it consistent with previous Health Profiles all ages were used as the denominator but diabetes was only recorded in patients aged 17+.
Methods used to calculate indicator value
The indicator value is presented as a percentage although strictly is a ratio. There is no age adjustment. The calculation was performed as LA count of diabetes/total LA resident population
Small Populations: How Isles of Scilly and City of London populations have been dealt with
City of London and the Isles of Scilly data were included in the calculation of the regional and national prevalences but they do not contribute to the LA prevalences.
Disclosure Control
Not relevant
Health Profiles 2009 The Indicator Guide
197 Section 5: disease and poor health
Confidence Intervals calculation method
The 95% and 99.8% confidence intervals are calculated using Julian Flowers’ (erpho) confidence interval tool: http://www.erpho.org.uk/Download/
Public/15374/1/Confidence_Intervals_Wilson.xls This calculates confidence intervals using the following method for a confidence interval of a proportion as described by R.G. Newcombe. If r is the observed number of subjects with some feature in a sample of size n then the estimated proportion who have the feature is p = r/n. The proportion who do not have the feature is q = 1-p. First, calculate the three quantities A = 2r + z2;
B = z z 2 + 4rq ; and C=2(n+z2),
where z is z1-α/2, from the standard Normal distribution. Then the confidence interval for the population proportion is given by (A-B)/C to (A+B)/C This method has the considerable advantage that it can be used for any data. When there are no observed events, r and hence p are both zero, and the recommended confidence interval simplifies to 0 to z2/(n+z2). When r = n so that p = 1, the interval becomes n/(n+z2) to 1. Reference Newcombe, R.G. Two-sided confidence intervals for the single proportion: comparison of seven methods. Stat Med 1998;17:857-72.
Health Profiles 2009 The Indicator Guide
198 Section 5: disease and poor health
23. NEW CASES OF TUBERCULOSIS INDICATOR Basic Information 1. What is being measured?
New cases of tuberculosis (TB)
2. Why is it being measured?
To reduce the spread of TB by identifying areas where rates of TB are high.
3. How is this indicator actually defined?
3-year average of TB incidence per 100,000 population, all ages, 2004–06, persons.
4. Who does it measure?
All persons, all ages.
5. When does it measure it?
Continuous monitoring locally, validated periodically at regional level. The most recent year for which data are available is 2006.
6. Will It measure absolute numbers or proportions?
Proportion: Crude rate per 100,000 population.
7. Where does the data actually come from?
Health Protection Agency.
8. How accurate and complete will the data be?
The national surveillance system has a high level of completeness and accuracy.
9. Are there any caveats/warnings/ problems?
A small number of cases may be assigned to the wrong local authority due to missing/incorrect postcodes. TB incidence is highest in inner-cities and concentrated in certain population groups such as the homeless, prisoners, substance mis-users and communities linked to high-TB incidence countries.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
Pg 4 Health Summary – Indicator No 24
Subject category/ domain(s)
Disease and poor health
Indicator name (*Indicator title in health profile)
New cases of tuberculosis
PHO with lead responsibility
South West Public Health Observatory
Date of PHO dataset creation
February 2008, recalculated for new April 2009 boundaries by the Health Protection Agency and SWPHO in February 2009.
Indicator definition
3-year average of TB incidence per 100,000 population, 2004–06.
Health Profiles 2009 The Indicator Guide
199 Section 5: disease and poor health
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs.
Timeliness
Continuous monitoring locally, validated periodically at regional level. Most recent complete data available is 2006.
Rationale: What this indicator purports to measure
The incidence (number of new cases per 100,000 population) of TB notified during the year.
Rationale: Public Health Importance
TB has re-emerged as a major public health problem and is the leading cause of death worldwide among curable infectious diseases. In England cases fell progressively until 1987 but started to rise again in the late 1980s. Over 8,000 new cases are now being reported each year in the UK. TB remains a disease associated with socio-economic deprivation and largely affects migrants from high incidence countries and deprived sub-groups of the population.
Rationale: Purpose behind the inclusion of the indicator
To reduce the spread of TB by identifying areas where rates of TB are high. Incidence is high amongst [first and subsequent generation] migrant populations so this indicator could identify where resources should be targeted to support TB control including the prompt identification of cases and measures to ensure treatment completion. Other high risk groups include the homeless, problem drug users and prisoners. It is very important that there are Health Protection indicators incorporated in to the wider health profiles and although many Health Protection data are not yet available at LA, work is being undertaken to ensure that this happens. TB therefore offers a good opportunity to provide information on a re-emerging PH issue and support the Chief Medical Officer’s (CMO) Action Plan.
Rationale: Policy relevance
The CMO has identified TB as a re-emerging threat which needs concerted action to address. In response to this, the National Institute for Health and Clinical Excellence (NICE) published guidelines which describe recommended measures for the control of TB. The Department of Health recently published “Tuberculosis prevention and treatment: a toolkit for planning, commissioning and delivering high-quality services in England”, outlining the standards for the delivery of tuberculosis services, in June 2007. This indicator, therefore, provides a marker to monitor progress towards the achievement of the objectives of the CMO’s Action Plan and supports the implementation of the TB toolkit. In addition, the European Centre for Disease Control has also published an action plan which outlines the disease prevention and control measures expected of European Union member states. This indicator will help in supporting the implementation of the plan.
Interpretation: What a high/low level of indicator value means
A high indicator value (red circle in health summary chart) represents a statistically significant worse rate of TB incidence when compared to the England average value. A low indicator value (green circle in health summary chart) represents a statistically significant better rate of TB incidence when compared to the England value. However a low indicator value should not mean that public health action is not needed as TB is preventable and any cases of tuberculosis need to be treated. The prompt diagnosis and treatment of infectious cases is key to halting the health burden which TB causes in the UK.
Health Profiles 2009 The Indicator Guide
200 Section 5: disease and poor health
Interpretation: Potential for error due to type of measurement method
New cases of TB are reported through a voluntary enhanced surveillance system to collect detailed data on each case. It provides the most accurate measurement of TB incidence. The quality of national TB surveillance improved significantly since the introduction of this system in 1999. A national study, undertaken in 2003, on the completeness of national surveillance suggests that there may be undernotification of as much as 15% of cases.
Interpretation: Potential for error due to bias and confounding
The rates are not age-standardised or adjusted to take into account variation between areas in risk factors associated with TB. In the UK, TB is highest in inner cities and concentrated in certain population groups such as the homeless, prisoners, substance mis-users and communities linked to high-TB incidence countries
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Table 2 – Indicator Specification Indicator definition: Variable
New cases of tuberculosis
Indicator definition: Statistic
3-year average (mean) of TB incidence per 100,000 population
Indicator definition: Gender
All persons
Health Profiles 2009 The Indicator Guide
201 Section 5: disease and poor health
Indicator definition: age group
All ages
Indicator definition: period
Numerator 2004–2006, denominator mid-year population estimates
Indicator definition: scale
Per 100,000 resident population
Geography: geographies available for this indicator from other providers
Additional data, mostly at national level, available on Health Protection Agency website www.hpa.org.uk
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Not available
Data extraction: Source
Health Protection Agency
Data extraction: source URL
Data at Local Authority level not published elsewhere.
Data extraction: date
Data extracted from source (Enhanced Tuberculosis Surveillance system ) as at: 27/11/2007
Numerator: definition
Average number of TB cases reported between 2004–2006 in accordance with the agreed case definition
Numerator: source
Health Protection Agency
Denominator: definition
ONS mid-year population estimates
Denominator: source
Office for National Statistics (ONS).
Data quality: Accuracy and completeness
Data are provided through the Enhanced Tuberculosis Surveillance system, which commenced in January 1999. It is therefore robust, standardised and comprehensive throughout England, Wales and Northern Ireland.
TABLE 3 – INDICATOR TECHNICAL METHODS Numerator: extraction
Health Protection Agency
Numerator: aggregation/ allocation
Cases were allocated to Local Authority based on their postcode of residence and summarised as a 3-year (mean) average 2004–06.
Numerator data caveats
The TB incidence was obtained by deriving the local authority of residence for each TB case. Incorrectly entered postcode of residence or local authority may have resulted in assigning some cases to the incorrect local authority or county. A small number of cases for whom the local authority was not known were included only in the England total.
Health Profiles 2009 The Indicator Guide
202 Section 5: disease and poor health
Denominator data caveats
As the data were for a period of three years (2004-2006), the mid-year population estimate for the middle year (2005) was used. This middle year may not always be representative of the three year period.
Methods used to calculate indicator value
Calculation of the numerator: Calculated a 3-year average (mean) for 2004– 2006 by summing the individual years and dividing by 3. Denominator count: Mid-year 2005 population estimates (all ages). The numerator was then divided by the denominator; the resulting value was then multiplied by 100,000 to give a crude rate per 100,000 population. The resulting rate was rounded down to an integer value.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
These are included in the totals for their regions (respectively, South West and London).
Disclosure Control
Indicator values for areas where there were <5 cases over the 3-year period (giving an average of 1 or 2 per year) are suppressed and these areas are given a value and rate of zero. The rates for all areas were rounded down to an integer value.
Confidence Intervals calculation method
The confidence intervals for these crude rates were constructed using the following formula that relates the chi-square and Poisson distributions:
χ α ,2d 2
LL =
2
2
χ 1- α ,2(d+1) 2
2
UL =
2
where LL and UL are the lower and upper 100*(1-a) per cent confidence limits and d denotes the number of observed events (e.g, serious injuries, (100*α)offences, violent deaths) per unit of time exposed. χ² a/2,2d is the (100*a/2)th percentage point for a chi-squared distribution with 2d degrees of freedom and χ²(1-α/2),2(d+1) is the (100*(1-a/2))th percentage point for a chi-squared distribution on 2(d+1) degrees of freedom.
The confidence limits for the rates were then obtained by dividing the upper and lower limits for the counts by the person time exposed. Reference: Dobson AJ, Kuulasmaa K, Eberle E, Scherer J. Confidence intervals for weighted sums of Poisson parameters. Statistics in Medicine 1991;10:457-462.
Health Profiles 2009 The Indicator Guide
203 Section 5: disease and poor health
24. HIP FRACUTRE IN OVER-65s INDICATOR Basic Information 1. What is being measured?
Hip fractures in over 65 year olds.
2. Why is it being measured?
Hip fracture is the most common injury related to falls in older people. More than 95% of hip fractures in adults ages 65 and older are caused by a fall. Hip fractures in the elderly and frail can lead to loss of mobility and loss of independence. For many older people it is the event that forces them to leave their homes and move into residential care. Mortality after hip fracture is high: around 30% for one year. To stimulate discussion and encourage local investigation, and to lead to improvement in data quality and quality of care.
3. How is this indicator actually defined?
Emergency Hospital Admission for fractured neck of femur, directly age-standardised rate, 65 year and over, 2006–07, persons.
4. Who does it measure?
All persons, 65 years and over.
5. When does it measure it?
Created specifically for HP3 2008. Regularly updated. HES data for 2007/2008 was not available for inclusion in HP4 2009. However, HP3 2008 data have been updated to accommodate the newly created Unitary Authorities in England.
6. Will It measure absolute numbers or proportions?
Proportions: numbers of cases per hundred thousand European standard population.
7. Where does the data actually come from?
Collection and collation from Hospital Episodes Statistics (HES) via the NHS Information Centre.
8. How accurate and complete will the data be?
HES data and ONS population statistics are considered to be complete and robust.
9. Are there any caveats/warnings/ problems?
There may be variation between Trusts in the way hospital admissions are coded. Routine data do not allow for all of these aspects to be identified and removed from the indicator, however, this may be done through local audit. Population estimates used are Persons rounded and therefore may produce slight differences in rate calculation from those done locally. Population estimates of higher geographies such as County, SHA, GOR and England are aggregated of the LA populations.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Health Profiles 2009 The Indicator Guide
204 Section 5: disease and poor health
Table 1 – Indicator Description Information component
Pg 4 Health Summary Indicator No 25
Subject category/ domain(s)
Diseases and poor health
Indicator name (*Indicator title in health profile)
Hip fracture in over-65s
PHO with lead responsibility
WMPHO
Date of PHO dataset creation
11/02/08
Indicator definition
Emergency Hospital Admission for fracture neck of femur, directly agestandardised rate, 65 year and over, 2006-07, Persons
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs
Timeliness
Created specifically for HP3 2008. Not regularly updated.
Rationale: What this indicator purports to measure
It is an estimate of serious falls in older people.
Rationale: Public Health Importance
Hip fracture is a major cause of disability and the leading cause of mortality due to injury in older people aged over 75. Hospital admission for fractured neck of femur is a good proxy measure of the incidence of hip fracture in older people. Falls prevention programmes aim to reduce the incidence of fractured neck of femur in the community. Hip fracture is the most common injury related to falls in older people. More than 95% of hip fractures in adults ages 65 and older are caused by a fall. Hip fractures in the elderly and frail can lead to loss of mobility and loss of independence. For many older people it is the event that forces them to leave their homes and move into residential care. Mortality after hip fracture is high: around 30% for one year.
Rationale: Purpose behind the inclusion of the indicator
To monitor the incidence of fracture neck of femur. To monitor public health programmes aiming to reduce the risk of older people falling. To stimulate discussion and encourage local investigation, and to lead to improvement in data quality and quality of care.
Health Profiles 2009 The Indicator Guide
205 Section 5: disease and poor health
Rationale: Policy relevance
Standard 6 of the National Service Framework for Older People aims to “reduce the number of falls which result in serious injury and ensure effective treatment and rehabilitation for those who have fallen”. A New ambition for old age (DH, 2006), which outlines the next steps in implementing the NSF, lists ‘falls and bone health’ as one of its 10 programmes, and outlines the components of integrated falls services. There is NICE Guidance on the assessment and prevention of falls in older people (NICE, 2004). Studies have indicated that falls prevention services can reduce falls (Interventions to prevent falls in elderly people can be effective. Cochrane Collaboration. 2003. http://www.cochrane.org/reviews/en/ab000340.html)
Interpretation: What a high/low level of indicator value means
A high indicator value (red blob in spine chart) represents a statistically significant higher level of estimated incidence of fractured neck of femur for that local authority when compared to the national value.
Interpretation: Potential for error due to type of measurement method
There may be variation between Trusts in the way hospital admissions are coded. Routine data do not allow for all of these aspects to be identified and removed from the indicator, however, this may be done through local audit.
Interpretation: Potential for error due to bias and confounding
In order to allow comparison of groups with different age structures it is common to present “age standardised” rates. These are calculated by summing the product of age specific rates for each age band in the group by the number in that age band in the standard population. The sum is then divided by the total number in all age bands in the standard population to obtain the age standardised rate. Any difference between groups in age standardised rate is then not due to difference in age structure since the same standard population was used to calculate all age standardised rates. The method does however assume that minor differences in age structure within age bands are unimportant and in general this is true for younger age groups. For the age standardised rates given in this report five year age bands have generally been used. However with older groups minor differences in average within age band may become important and some authorities argue that for older age bands one year groups should be used.
A low indicator value (green blob) represents a statistically significant lower level of estimated incidence of fractured neck of femur for that local authority when compared to the national value.
The problem is further compounded by the use of the European Standardised Population which has a top age band of 85 years and over. This means that an age band of 85 years and over also has to be used for age specific rates and this is clearly unsatisfactory. Further the European Standard population is much younger than the UK population and therefore biases comparisons towards younger age groups. The rank order of age standardised rates may be appreciably changed by use of different standard populations. Therefore for groups above 65 and particularly for groups including those over age of 85 the use of age standardised rates will reduce the effect of age differences in the populations but cannot be assumed to have eliminated it.
Health Profiles 2009 The Indicator Guide
206 Section 5: disease and poor health
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider the confidence interval, the greater the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Table 2 – Indicator Specification Indicator definition: Variable
Emergency Hospital Admission for fractured neck of femur
Indicator definition: Statistic
Directly age-standardised rate
Indicator definition: Gender
Persons
Indicator definition: age group
65 years and over
Indicator definition: period
2006–07 Financial year
Indicator definition: scale
Per 100,000 European Standard population
Geography: geographies available for this indicator from other providers
None. Dataset unique to WMPHO.
Health Profiles 2009 The Indicator Guide
207 Section 5: disease and poor health
Dimensions of inequality: subgroup analyses of this dataset available from other providers
None available
Data extraction: Source
Hospital Episode Statistics (HES) for the respective financial year, England, NHS Health and Social Care Information Centre. HES Universe 2006/2007 (universe ID HIP0606) This material is Crown Copyright but may be reproduced without formal permission or charge for personal or in-house use. © Crown Copyright 2004
Data extraction: source URL
Extracted through direct link from HES, NHS Health and Social Care Information Centre.
Data extraction: date
Data extracted from source as at: 28th January 2008
Numerator: definition
Emergency Hospital Admissions for primary diagnosis of fractured neck of femur in 65+ age group. Diagnosis of fractured neck of femur classified by primary diagnosis (ICD10 S72.0, S72.1, S72.2), admitted in the respective financial year. ICD10 codes for fractured proximal femur refer to the following diagnoses: • S72.0 Fracture of neck of femur; • S72.1 Pertrochanteric fracture; • S72.2 Subtrochanteric fracture
Numerator: source
Hospital Episode Statistics (HES) for the respective financial year, England, NHS Health and Social Care Information Centre. HES Universe 2006/2007 (universe ID HIP0606) This material is Crown Copyright but may be reproduced without formal permission or charge for personal or in-house use. © Crown Copyright 2004
Denominator: definition
Denominator data – Mid year population estimates 2006 for persons aged 65+ rounded. Data are based on the latest revisions of ONS mid-year population estimates for the respective years, last updated 21/08/07.
Denominator: source
Office for National Statistics (ONS).
Data quality: Accuracy and completeness
Hospital Episode Statistics (HES) are compiled from data sent by over 300 NHS Trusts and Primary Care Trusts (PCTs) in England. The Health and Social Care Information Centre liaises closely with these organisations to encourage submission of complete and valid data and seeks to minimise inaccuracies and the effect of missing and invalid data via HES processes. Whilst this brings about improvement over time, some shortcomings remain. HP2007 data for this indicator was extracted for all admissions. There was considerable variation in patterns of admission for fractured neck of femur across the country with the numbers of transfer admissions between hospitals varying from 2.6% in North East Region to 21.6% in South West Region. Specific problems with over counting due to incorrect coding of internal transfers giving new admissions were identified in Ashfield and Nottingham City. These were examined locally and admission to patient ratios calculated. WMPHO then investigated using emergency admissions only and produced almost identical results in these two LAs. The decision to extract emergency admissions only should reduce erroneous geographical variations.
Health Profiles 2009 The Indicator Guide
208 Section 5: disease and poor health
Table 3 – Indicator Technical Methods Numerator: extraction
HES extraction The number of finished admission episodes including transfers, for 65+ ages with any emergency method of admission and with any of the following primary diagnoses (DIAG_01, ICD 10 codes) in the respective financial year:Fractured proximal femur • S72.0 Fracture of neck of femur; • S72.1 Pertrochanteric fracture; • S72.2 Subtrochanteric fracture Data were extracted as Finished Admission Consultant Episodes The data was ungrossed Counts are by: age/sex/organisation of residence in Finished Admission Episode (values for England are aggregates of these) where: age bands 65–69, 70–74, 75–79, 80-84, 85+; sex is 1, 2 (male and female) Method of admission = Emergency.
Numerator: aggregation /allocation
Residency (by Local Authority) of each Finished Admission Episodes is allocated by HES. Since the 2001 census new Unitary Authorities have been created. The new Unitary Authorities of Northumberland, Shropshire, Wiltshire and Durham have the same boundaries as the counties of the same name so have been allocated these data. Data for five other new unitary Authorities have been calculated from the aggregation of previous Local Authorities. New UA name Bedford Central Bedfordshire Cheshire East
Cheshire West and Chester
Cornwall*
Old LA name Bedford Mid Bedfordshire South Bedfordshire Congleton Crewe and Nantwich Macclesfield Chester Ellesmere Port & Neston Vale Royal Caradon Carrick Kerrier North Cornwall Penwith Restormel
Health Profiles 2009 The Indicator Guide
209 Section 5: disease and poor health
Numerator data caveats
The value shown in the health profiles is for emergency admissions with a primary diagnosis of fractured neck of femur as defined by ICD10 codes S72.0 to S72.2. Fractured neck of femur is sometimes defined as S72.0 only, or as the whole S72.0-S72.9 grouping, or as various other combinations within the S72.0–S72.9 range. The value shown in the health profiles is for primary diagnosis of fractured neck of femur only. If local values have used all diagnosis of fractured neck of femur they will have higher values. HP2007 data for this indicator was extracted for all admissions. There was considerable variation in patterns of admission for fractured neck of femur across the country with the numbers of transfer admissions between hospitals varying from 2.6% in North East Region to 21.6% in South West Region. Specific problems with over counting due to incorrect coding of internal transfers giving new admissions were identified in Ashfield and Nottingham City. These were examined locally and admission to patient ratios calculated. WMPHO then investigated using emergency admissions only and produced almost identical results in these two LAs. The decision to extract emergency admissions only should reduce erroneous geographical variations.
Denominator Population estimates used are Persons rounded and therefore may produce slight data caveats differences in rate calculation from those done locally. Population estimates of higher geographies such as County, SHA, GOR and England are aggregates of the LA populations. Methods used to calculate indicator value
The directly age-standardised rate is the rate of events that would occur in a standard population if that population were to experience the age-specific rates of the subject population. Explicitly:
∑w r DSR = ∑w
i i
i
i
× 100,000
i
(expressed per 100,000 population) where: wi is the number, or proportion, of individuals in the standard population in age group i. ri is the crude age-specific rate in the subject population in age group i, given by:
ri =
Oi ni
where: Oi is the observed number of events in the subject population in age group i. ni is the number of individuals in the subject population in age group i. The standard population generally used for the direct method is the European Standard Population. The age groups used are: 65–69, 70–74, 75–79, 80–84, 85+.
Health Profiles 2009 The Indicator Guide
210 Section 5: disease and poor health
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Scilly Isles and City of London has been excluded at LA level and included at Region and National levels.
Disclosure Control
Not applicable
Confidence Intervals calculation method
Confidence intervals for the rates were calculated using the method described in the NCHOD Compendium for directly standardised rates. www.nchod.nhs.uk 95% confidence intervals for the age-standardised rates were calculated using a normal approximation. Standard errors are obtained using the method described by Breslow and Day but modified to use the binomial variance for a proportion to estimate the variances of the crude age-specific rates. This method is likely to be unreliable when there are fewer than 50 cases in an area, hence confidence intervals for rates based on less than 50 cases should be viewed with caution. The lower and upper limits for the rates are denoted by DSRLL and DSRUL respectively.
DSRLL / UL = DSR ± 1.96 × 100,000 ×
1 ∑ wi ij
2
×∑ ij
wi2 ⋅ rij (1 − rij ) nij
(expressed per 100,000 population)
where: wi is the number, or proportion, of individuals in the standard population in age group i. rij is the crude age-specific rate in the subject population in age group i, in year j. nij is the number of individuals in the subject population in age group i, in year j. Ref: Breslow NE and Day NE. Statistical Methods in Cancer Research, Volume II: The Design and Analysis of Cohort Studies. Lyon: International Agency for Research on Cancer, World Health Organization, 1987: 59 Keyfitz N. Sampling variance of age-standardised mortality rates. Human Biology. 1966; 38: 309-317.
Health Profiles 2009 The Indicator Guide
211
Section 6: Life expectancy and cause of death
Health Profiles 2009 The Indicator Guide
212 Section 6: life expectancy and cause of death
25. EXCESS WINTER DEATHS INDICATOR Basic Information 1. What is being measured?
Excess winter deaths
2. Why is it being measured?
England, in common with other European countries, experiences higher levels of mortality in the winter than in the summer. There is some evidence to suggest that excess winter mortality (EWM) is preventable. Mortality in winter increases more in England compared to other European countries with colder climates, suggesting that it is more than just lower temperatures responsible for the excess mortality in winter1,2.
3. How is this indicator actually defined?
The Excess Winter Mortality Index (EWM Index) is the excess winter deaths expressed as a ratio of the expected deaths based on the non-winter death rate for the period 1.8.04 to 31.7.07.
4. Who does it measure?
The ratio of extra deaths from all causes that occur in the winter months compared to the average of the number of non-winter deaths of the same period.
5. When does it measure it?
All those that died between 1.8.04 to 31.7.07
6. Will It measure absolute numbers or proportions?
Ratio
7. Where does the data actually come from?
The Annual Mortality File provided by ONS for PHOs.
8. How accurate and complete will the data be?
It is the most reliable mortality data source available, and apart from the small number of deaths registered with a long delay due to the coroner’s investigation, it is supposed to be complete.
9. Are there any caveats/warnings/ problems?
EWM Index was calculated based on the “ONS Method” which defines the winter period as December to March, and the non-winter period as August to November of that same year and April to July of the following year. This winter period was selected as they are the months which over the last 50 years have displayed above average monthly mortality. However, if mortality starts to increase prior to this, for example in November, the number of deaths in the non-winter period will increase, which in turn will decrease the estimate of excess winter mortality. The EWM Index will be partly dependent on the proportion of older people in the population as most excess winter deaths effect older people (there is no standardisation in this calculation by age or any other factor).
Health Profiles 2009 The Indicator Guide
213 Section 6: life expectancy and cause of death
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
25
Subject category/ domain(s)
Life expectancy and causes of death
Indicator name (* Indicator title in health profile)
Excess winter deaths
PHO with lead responsibility
ERPHO
Date of PHO dataset creation
19.2.08
Indicator definition
Excess Winter Mortality Index (EWM Index) is the excess winter deaths as a ratio of the expected deaths based on the non-winter death rate for the period 1.8.04 to 31.7.07.
Geography
LA/UA
Timeliness
The most recent year of data available is 2007
Rationale: What this indicator purports to measure
This indicator measures excess winter deaths expressed as the EWM Index, in order that comparisons can be made easily between different geographies. It indicates whether there are higher than expected deaths in the winter compared to the rest of the year.
Rationale: Public Health Importance
The number of excess winter deaths depends on the temperature and the level of disease in the population as well as other factors, such as how well equipped people are to cope with the drop in temperature. Most excess winter deaths are due to circulatory and respiratory diseases, and the majority occur amongst the elderly population. Research carried out by the Eurowinter Group1 and Curwen2 found that mortality during winter increases more in England and Wales compared to other European countries with colder climates, suggesting that many more deaths could be preventable in England and Wales.
Rationale: Purpose behind the inclusion of the indicator
To highlight areas with higher than expected levels of excess winter deaths, to give an idea where interventions need to be improved or instigated to cope with the change of seasonal temperature.
Rationale: Policy relevance
The Department of Health’s annual Keep Warm Keep Well campaign is aimed at financially disadvantaged older or disabled people and their carers, and families with young children on low incomes. It gives information on the health benefits of keeping warm in winter, providing advice on healthy eating and exercise, home heating and energy efficiency, and details of the grants and benefits available.
Health Profiles 2009 The Indicator Guide
214 Section 6: life expectancy and cause of death
Interpretation: What a high/low level of indicator value means
An indicator value better than average (green circle in health summary chart) indicates that excess winter deaths are statistically significantly lower for that local authority compared to the national average.
Interpretation: Potential for error due to type of measurement method
EWM Index was calculated based on the “ONS Method” which defines the winter period as December to March, and the non-winter period as August to November of that same year and April to July of the following year. This winter period was selected as they are the months which over the last 50 years have displayed above average monthly mortality. However, if mortality starts to increase prior to this, for example in November, the number of deaths in the non-winter period will increase, which in turn will decrease the estimate of excess winter mortality.
An indicator value worse than average (red circle in health summary chart) indicates that excess winter deaths are statistically significantly higher for this area compared to the national average.
Interpretation: Potential for error due to bias and confounding Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider the confidence interval the greater the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Health Profiles 2009 The Indicator Guide
215 Section 6: life expectancy and cause of death
Table 2 – Indicator Specification Indicator definition: Variable
The Excess Winter Mortality Index (EWM Index) is the excess winter deaths expressed as a ratio of the expected deaths based on the non-winter deaths for the period 1.8.04 to 31.7.07.
Indicator definition: Statistic
Ratio expressed as a percentage
Indicator definition: Gender
Person
Indicator definition: age group
All ages
Indicator definition: period
Deaths occurring in the period: 1.8.04 to 31.7.07
Indicator definition: scale
In per cent
Geography: geographies available for this indicator from other providers Dimensions of inequality: subgroup analyses of this dataset available from other providers Data extraction: Source
Annual Mortality File provided by ONS for PHOs (2001-2007)
Data extraction: source URL
This is not publically available
Data extraction: date Numerator: definition
The numerator, excess winter deaths, calculated as: Total number of winter deaths i.e. deaths occurring in the months December to March for the period 1.08.04 to 31.07.07 minus half number of deaths in the non-winter months (August to November, April to July).
Numerator: source
Annual Mortality File (ONS)
Denominator: definition
The average number of deaths occurring in the equivalent non-winter period i.e. the deaths occurring in August to November, April to July divided by 2)
Denominator: source
Annual Mortality File (ONS)
Data quality: Accuracy and completeness
It is the most reliable mortality data source available, and apart from the small number of deaths registered with a long delay due to the coroner’s investigation, it is supposed to be complete.
Health Profiles 2009 The Indicator Guide
216 Section 6: life expectancy and cause of death
Table 3 – Indicator Technical Methods Numerator: extraction
The number of winter deaths and non-winter deaths was extracted from the Annual Mortality File using the following SQL: CREATE TABLE P25xx_Winterdeaths_LA (Y_MOD varchar(10), LACode varchar(4), LAName varchar(50), Deaths int) INSERT INTO P25xx_Winterdeaths_LA (Y_MOD, LACode, LAName, Deaths) SELECT Left([DOD],6), [LA Code now], [LA name], COUNT(*) FROM M_DEATHS2001to2007 LEFT JOIN HES.dbo.A_LACounty ON M_DEATHS2001to2007.ResCty + M_ DEATHS2001to2007.ResLAUA = HES.dbo.A_LACounty.[LA Code] LEFT JOIN HES.dbo.A_NationalPostcodes_Feb08 ON M_DEATHS2001to2007. Pcode_8dig = HES.dbo.A_NationalPostcodes_Feb08.[PCD2] where [DOD]>=20040801 and [DOD]<=20070731 Group by [LA Code now], [LA name], Left([DOD],6) ----------------------------------------------------------------------UPDATE P25xx_Winterdeaths_LA SET Y_MOD = ‘S’ WHERE Y_MOD = ‘200408’ or Y_MOD = ‘200409’ or Y_MOD = ‘200410’ or Y_MOD = ‘200411’ or Y_MOD = ‘200504’ or Y_MOD = ‘200505’ or Y_MOD = ‘200506’ or Y_MOD = ‘200507’ or Y_MOD = ‘200508’ or Y_MOD = ‘200509’ or Y_MOD = ‘200510’ or Y_MOD = ‘200511’ or Y_MOD = ‘200604’ or Y_MOD = ‘200605’ or Y_MOD = ‘200606’ or Y_MOD = ‘200607’ or Y_MOD = ‘200608’ or Y_MOD = ‘200609’
Health Profiles 2009 The Indicator Guide
217 Section 6: life expectancy and cause of death
or Y_MOD = ‘200610’ or Y_MOD = ‘200611’ or Y_MOD = ‘200704’ or Y_MOD = ‘200705’ or Y_MOD = ‘200706’ or Y_MOD = ‘200707’ UPDATE P25xx_Winterdeaths_LA SET Y_MOD = ‘W’ WHERE Y_MOD = ‘200412’ or Y_MOD = ‘200501’ or Y_MOD = ‘200502’ or Y_MOD = ‘200503’ or Y_MOD = ‘200512’ or Y_MOD = ‘200601’ or Y_MOD = ‘200602’ or Y_MOD = ‘200603’ or Y_MOD = ‘200612’ or Y_MOD = ‘200701’ or Y_MOD = ‘200702’ or Y_MOD = ‘200703’ -------------------------------------------------------Select LAcode, Y_MOD, sum(deaths) as [Sum] into P25xx_Winterdeaths_EWDI from P25xx_Winterdeaths_LA group by LAcode, Y_MOD order by LAcode, Y_MOD ---------------------------------------------------CREATE TABLE P25xx_Winterdeaths_LA_W (Y_MOD varchar(10), LACode varchar(4), LAName varchar(50), W_Deaths int) Insert into P25xx_Winterdeaths_LA_W (Y_MOD, LACode, LAName, W_Deaths) select Y_MOD, LACode, LAName, Sum(2*Deaths) from P25xx_Winterdeaths_LA where Y_MOD = ‘W’ group by LACode, LAName
Health Profiles 2009 The Indicator Guide
218 Section 6: life expectancy and cause of death
CREATE TABLE P25xx_Winterdeaths_LA_S (Y_MOD varchar(10), LACode varchar(4), LAName varchar(50), S_Deaths int) Insert into P25xx_Winterdeaths_LA_S (Y_MOD, LACode, LAName, S_Deaths) select Y_MOD, LACode, LAName, sum(Deaths) from P25xx_Winterdeaths_LA where Y_MOD = ‘S’ group by LACode, LAName ----------------------------------------------CREATE TABLE P25xx_Winterdeaths_EWDI (LACode varchar(4), LAName varchar(50), [W_deaths_times_2] int, S_deaths int, EWDI decimal(8,4)) Insert into P25xx_Winterdeaths_EWDI (P25xx_Winterdeaths_LA_W.LACode, P25xx_Winterdeaths_LA_W.LAName, [W_deaths_times_2], S_deaths, EWDI) select P25xx_Winterdeaths_LA_W.LACode, P25xx_Winterdeaths_LA_W.LAName, W_deaths, S_deaths, W_deaths/S_deaths from P25xx_Winterdeaths_LA_W join P25xx_Winterdeaths_LA_S on P25xx_Winterdeaths_LA_W.LACode = P25xx_ Winterdeaths_LA_S.LAcode The numerator, excess winter deaths, was then calculated in excel as part of the EWM Index calculation (see below). The numerator published in the template is the average (yearly) excess winter deaths for the three year period, which is calculated: Winter deaths (Dec-Mar)- 0.5 summer deaths (Aug-Nov,Apr-Jul) for the period 1.08.04-31.07.07
3
Health Profiles 2009 The Indicator Guide
219 Section 6: life expectancy and cause of death
Numerator: aggregation/ allocation
The LA/UA data was aggregated up to new UA, county, region, SHA and national level. In April 2009 new Unitary Authorities have been created. The new Unitary Authorities of Northumberland, Shropshire, Wiltshire, and Durham have the same boundaries as the counties of the same name so have been allocated these data. Data for five other new unitary Authorities have been calculated from the aggregation of previous Local Authorities (see table below) New UA name
Old LA name
Bedford
Bedford
Central Bedfordshire
Mid Bedfordshire South Bedfordshire
Cheshire East
Congleton Crewe and Nantwich Macclesfield
Cheshire West and Chester
Chester Ellesmere Port & Neston Vale Royal
Cornwall
Caradon Carrick Kerrier North Cornwall Penwith Restormel
Numerator data caveats Denominator data caveats Methods used to calculate indicator value
Method to calculate the EWM Index was as defined by ONS: excess winter deaths expected number of deaths based on the non-winter death rates = winter deaths (Dec-Mar)- 0.5 non-winter deaths (Aug–Nov,Apr–Jul) 0.5 non-winter deaths (Aug–Nov,Apr–Jul) = 2 * winter deaths (Dec-Mar) non-winter deaths (Aug-Nov, Apr-Jul)
-1
The last arrangement of the formula was used in this calculation. Small Populations: How Isles of Scilly and City of London populations have been dealt with
Isles of Scilly and the City of London were excluded from local authority calculations. The Isles of Scilly and the City of London were included in the higher geographies.
Disclosure Control
Not relevant
Health Profiles 2009 The Indicator Guide
220 Section 6: life expectancy and cause of death
Confidence Intervals calculation method
The EWM index is treated as an odds (a/b) i.e. 2 x a/b -1 Where: a = number of winter deaths, b = number of summer deaths, 2 is a constant and therefore does not need to be included in the confidence interval calculation as a constant has no variance, and -1 is also a constant and can be subtracted at the end. The formula for a calculation of 95% confidence intervals was taken from Kirkwood and Sterne3. 95% CI = odds/EF to odds x EF Where EF = exp[1.96 x s.e.(log odds)] And s.e.(log odds) = √[1/a+1/b] For 99.8% confidence intervals the Standard Error was changed to 3.091
1 The Eurowinter group (1997) Cold exposure and winter mortality from ischaemic heart disease, cerebrovascular disease, respiratory disease, and all causes in warm and cold regions in Europe. The Lancet 349, 1341-1346 2 Curwen M (1990/91) Excess winter mortality: a British phenomenon? Health Trends 4, 169-75 3 Kirkwood B.R. and Sterne J.A.C (2003) Essential Medical Statistics, 2nd Edition, p.164
Health Profiles 2009 The Indicator Guide
221 Section 6: life expectancy and cause of death
26. LIFE EXPECTANCY – MALE INDICATOR Basic Information 1. What is being measured?
Life Expectancy
2. Why is it being measured?
All cause mortality is a fundamental and probably the oldest measure of the health status of a population. Differences in levels of all-cause mortality reflect health inequalities between different population groups, e.g. between genders, social classes and ethnic groups. Life expectancy at birth is chosen as the preferred summary measure of all cause mortality as it quantifies the differences between areas in units (years of life) that are more readily understood and meaningful to the audience than those of other measures.
3. How is this indicator actually defined?
Life expectancy at birth, years, all ages, 2005–07, males.
4. Who does it measure?
Males, all ages.
5. When does it measure it?
This indicator is updated annually.
6. Will it measure absolute numbers or proportions?
Measures life expectancy in years.
7. Where does the data actually come from?
Office for National Statistics (ONS).
8. How accurate and complete will the data be?
Data on deaths are considered to be complete and robust.
9. Are there any caveats/warnings/ problems?
Death records without a valid area code are excluded, but the number of such records is negligible. The populations used for the calculation of the figures for this indicator are based on the 2001 Census, and are estimates.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
Pg 4 Health Summary – Indicator No. 26
Subject category/ domain(s)
Life expectancy and causes of death
Health Profiles 2009 The Indicator Guide
222 Section 6: life expectancy and cause of death
Indicator name (*Indicator title in health profile)
Life expectancy – male
PHO with lead responsibility
LHO
Date of PHO dataset creation
February 2009
Indicator definition
Life expectancy at birth, years, all ages, 2005–07, males
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs.
Timeliness
ONS produced data are updated annually in the Autumn of the following year.
Rationale: What this indicator purports to measure
Life expectancy at birth is a summary measure of the all cause mortality rates in an area in a given period. It is the average number of years a new-born baby would survive, were he or she to experience the particular area’s age-specific mortality rates for that time period throughout his or her life.
Rationale: Public Health Importance
All cause mortality is a fundamental and probably the oldest measure of the health status of a population. It represents the cumulative effect of the prevalence of risk factors, prevalence and severity of disease, and the effectiveness of interventions and treatment. Differences in levels of all-cause mortality reflect health inequalities between different population groups, e.g. between genders, social classes and ethnic groups. Life expectancy at birth is chosen as the preferred summary measure of all cause mortality as it quantifies the differences between areas in units (years of life) that are more readily understood and meaningful to the audience than those of other measures.
Rationale: Purpose behind the inclusion of the indicator
To help reduce premature mortality and facilitate planning of health services at local level.
Rationale: Policy relevance
There is a national health inequalities target for life expectancy which aims to increase average life expectancy at birth in England to 78.6 years for men and to 82.5 years for women, and to reduce health inequalities by 10% by 2010 as measured by life expectancy at birth (Department of Health PSA priority 1). Also life expectancy is an indicator in the following: • Local basket of inequalities indicators - Indicator 13.12. • Opportunity for all - Communities - Indicator 39. • Quality of life indicators - Health and social well-being - Indicator 33
Interpretation: What a high/low level of indicator value means
The higher the life expectancy, the longer the estimated expectation of life for males living in that area at that time.
Health Profiles 2009 The Indicator Guide
223 Section 6: life expectancy and cause of death
Interpretation: Potential for error due to type of measurement method
The figures reflect the contemporary mortality among those living in the area in each time period. They are not the number of years a baby born in the area in each time period could actually expect to live, both because the death rates of the area are likely to change in the future and because many of those born in the area will live elsewhere for at least some part of their lives. Life expectancy at birth is also not a guide to the remaining expectancy of life at any other given age. For example, if female life expectancy at birth was 80 years for a particular area, the life expectancy of women aged 75 in that area would exceed 5 years. This reflects the fact that survival from a particular age depends only on the mortality rates beyond that age, whereas survival from birth is based on mortality rates for all ages.
Interpretation: Potential for error due to bias and confounding
Older people living in nursing homes tend to be in poorer health than those not living in nursing homes. As these homes are unevenly distributed across the country, a higher death rate – consequently lower life expectancy level – in one area could simply be the result of migration of frail older people moving into nursing homes in that area.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider the confidence interval, the greater the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with a white symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or amber symbol depending on whether it is worse or better than the national value respectively.
Health Profiles 2009 The Indicator Guide
224 Section 6: life expectancy and cause of death
TABLE 2 – INDICATOR SPECIFICATION Indicator definition: Variable
Life expectancy at birth
Indicator definition: Statistic
Number of years
Indicator definition: Gender
Males
Indicator definition: age group
All ages
Indicator definition: period
2005–2007
Indicator definition: scale
Years life expectancy
Geography: geographies available for this indicator from other providers
England & Wales, UK and constituent countries, GORs, counties and local authorities Available from the Office for National Statistics (ONS)
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Social Class. Available from the Office for National Statistics (ONS)
Data extraction: Source
Office for National Statistics (ONS)
Data extraction: source URL
http://www.statistics.gov.uk/statbase/Product.asp?vlnk=8841 (See “Results for England and Wales”)
Data extraction: date
January 2009
Numerator: definition
Deaths from all causes, all ages, males
Numerator: source
Office for National Statistics (ONS)
Denominator: definition
Estimated resident population estimates, mid-year, all ages, males
Denominator: source
Office for National Statistics (ONS), available from http://www.statistics.gov.uk/statbase/Product.asp?vlnk=15106 (See mid-year estimates for 2005 to 2007) The population estimates used for the calculation of figures for this indicator are based on the 2001 Census. The population estimates used to calculate the LE figures were published on the following dates: • Mid-year population estimates for 2005 – 22 August 2007; • Mid-year population estimates for 2006 – 25 October 2007; • Mid-year population estimates for 2007 – 21 August 2008.
Health Profiles 2009 The Indicator Guide
225 Section 6: life expectancy and cause of death
Data quality: Accuracy and completeness
Mortality data quality and coverage is extremely high. The figures are threeyear averages so as to provide large enough numbers to ensure that the presented figures are sufficiently robust. Two authorities, City of London and Isles of Scilly, are excluded from the results because of small numbers of deaths and populations in these areas.
Table 3 – Indicator Technical Methods Numerator: extraction
Extraction of the completed indicator from ONS http://www.statistics.gov.uk/statbase/Product.asp?vlnk=8841
Numerator: aggregation / allocation
Deaths were assigned to local authority boundaries by ONS using the National Statistics Postcode Directory.
Numerator data caveats
Source material used to calculate LE: Area of residence is allocated by ONS using the postcode and the National Statistics Postcode Directory – records without a valid area code are excluded but the number of such records is negligible.
Denominator data caveats
Denominators are estimates of mid-year populations, based on results of the 2001 Census.
Methods used to calculate indicator value
The figures are rolling three-year averages produced by aggregating deaths and population estimates for 2005–2007. Abridged life tables were constructed using standard methods. Separate tables were constructed for males and females. The tables were created using annual mid-year population estimates and deaths registered in each year. All figures presented are for life expectancy at birth. A life table template which illustrates the method used to calculate the life expectancy results is available on the ONS website: http://www.statistics.gov.uk/statbase/ssdataset.asp?vlnk=6949 Data for some new Unitary Authorities, introduced in April 2009, did not correspond with existing geographies (Cheshire East, Cheshire West and Cheshire, Central Bedfordshire, and Cornwall). Results for these areas were calculated using the South East Public Health Observatory life expectancy calculator: http://www.sepho.org.uk/viewResource.aspx?id=8943 This was also used to calculate figures for the two Strategic Health Authorities in the South East: South Central and South East Coast. Results for the other new UAs were taken from published data for local authorities (Bedford) or counties (County Durham, Northumberland, Shropshire, and Wiltshire).
Small Populations: How Isles of Scilly and City of London populations have been dealt with
City of London and Isles of Scilly are excluded from the results because of small numbers of deaths and populations in these areas. Isles of Scilly has not been included in the new UA for Cornwall and City of London has not been included with Hackney. However, Isles of Scilly and City of London data are all included in the regional figures for the South West and London, and England figures.
Disclosure Control
Not applicable
Health Profiles 2009 The Indicator Guide
226 Section 6: life expectancy and cause of death
Confidence Intervals calculation method
The calculation of the confidence intervals was made using the method developed by Chiang. A report which details research undertaken by the Office for National Statistics on comparing methodologies to enable the calculation of confidence intervals for life expectancy at birth has now been published as No 33 in the National Statistics Methodological Series. This report, “Life expectancy at birth: methodological options for small populations”, also presents research carried out to establish if there is a minimum population size below which the calculation of life expectancy may not be considered feasible. It concludes with a summary of methodological conclusions and considers how these could be applied to the calculation of life expectancy at birth for wards in England and Wales. A copy of the report can be found on the NS website at: http://www.statistics.gov.uk/methods_quality/publications.asp The SEPHO calculator however also includes an adjustment to include a term for the variance associated with the final age interval as developed by Silcocks. Chiang CL. The Life Table and its Construction. In: Introduction to Stochastic Processes in Biostatistics. New York, John Wiley & Sons, 1968: 189-214. Silcocks PBS, Jenner DA, Reza R. Life expectancy as a summary of mortality in a population: statistical considerations and suitability for use by health authorities. J Epidemiol Community Health 2001; 55: 38-4
Health Profiles 2009 The Indicator Guide
227 Section 6: life expectancy and cause of death
27. LIFE EXPECTANCY – FEMALE INDICATOR Basic Information 1. What is being measured?
Life Expectancy
2. Why is it being measured?
All cause mortality is a fundamental and probably the oldest measure of the health status of a population. Differences in levels of all-cause mortality reflect health inequalities between different population groups, e.g. between genders, social classes and ethnic groups. Life expectancy at birth is chosen as the preferred summary measure of all cause mortality as it quantifies the differences between areas in units (years of life) that are more readily understood and meaningful to the audience than those of other measures.
3. How is this indicator actually defined?
Life expectancy at birth, years, all ages, 2005–07, females.
4. Who does it measure?
Females, all ages.
5. When does it measure it?
This indicator is updated annually.
6. Will it measure absolute numbers or proportions?
Measures life expectancy in years.
7. Where does the data actually come from?
Office for National Statistics (ONS)
8. How accurate and complete will the data be?
Data on deaths are considered to be complete and robust.
9. Are there any caveats/warnings/ problems?
Death records without a valid area code are excluded, but the number of such records is negligible. The populations used for the calculation of the figures for this indicator are based on the 2001 Census, and are estimates.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
Pg 4 Health Summary – Indicator No. 27
Subject category/ domain(s)
Life expectancy and causes of death
Health Profiles 2009 The Indicator Guide
228 Section 6: life expectancy and cause of death
Indicator name (* Indicator title in health profile)
Life expectancy – female
PHO with lead responsibility
LHO
Date of PHO dataset creation
February 2009
Indicator definition
Life expectancy at birth, years, all ages, 2005–07, females
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs.
Timeliness
ONS produced data are updated annually in the Autumn of the following year.
Rationale: What this indicator purports to measure
Life expectancy at birth is a summary measure of the all cause mortality rates in an area in a given period. It is the average number of years a new-born baby would survive, were he or she to experience the particular area’s age-specific mortality rates for that time period throughout his or her life.
Rationale: Public Health Importance
All cause mortality is a fundamental and probably the oldest measure of the health status of a population. It represents the cumulative effect of the prevalence of risk factors, prevalence and severity of disease, and the effectiveness of interventions and treatment. Differences in levels of all-cause mortality reflect health inequalities between different population groups, e.g. between genders, social classes and ethnic groups. Life expectancy at birth is chosen as the preferred summary measure of all cause mortality as it quantifies the differences between areas in units (years of life) that are more readily understood and meaningful to the audience than those of other measures.
Rationale: Purpose behind the inclusion of the indicator
To help reduce premature mortality and facilitate planning of health services at local level.
Rationale: Policy relevance
There is a national health inequalities target for life expectancy which aims to increase average life expectancy at birth in England to 78.6 years for men and to 82.5 years for women, and to reduce health inequalities by 10% by 2010 as measured by life expectancy at birth (Department of Health PSA priority 1). Also life expectancy is an indicator in the following: • Local basket of inequalities indicators - Indicator 13.12. • Opportunity for all - Communities - Indicator 39. • Quality of life indicators - Health and social well-being - Indicator 33
Interpretation: What a high/low level of indicator value means
The higher the life expectancy, the longer the estimated expectation of life for females living in that area at that time.
Health Profiles 2009 The Indicator Guide
229 Section 6: life expectancy and cause of death
Interpretation: Potential for error due to type of measurement method
The figures reflect the contemporary mortality among those living in the area in each time period. They are not the number of years a baby born in the area in each time period could actually expect to live, both because the death rates of the area are likely to change in the future and because many of those born in the area will live elsewhere for at least some part of their lives. Life expectancy at birth is also not a guide to the remaining expectancy of life at any other given age. For example, if female life expectancy at birth was 80 years for a particular area, the life expectancy of women aged 75 in that area would exceed 5 years. This reflects the fact that survival from a particular age depends only on the mortality rates beyond that age, whereas survival from birth is based on mortality rates for all ages.
Interpretation: Potential for error due to bias and confounding
Older people living in nursing homes tend to be in poorer health than those not living in nursing homes. As these homes are unevenly distributed across the country, a higher death rate – consequently lower life expectancy level – in one area could simply be the result of migration of frail older people moving into nursing homes in that area.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider the confidence interval, the greater the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with a white symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or amber symbol depending on whether it is worse or better than the national value respectively.
Health Profiles 2009 The Indicator Guide
230 Section 6: life expectancy and cause of death
Table 2 – Indicator Specification Indicator definition: Variable
Life expectancy at birth
Indicator definition: Statistic
Number of years
Indicator definition: Gender
Females
Indicator definition: age group
All ages
Indicator definition: period
2005–2007
Indicator definition: scale
Years life expectancy
Geography: geographies available for this indicator from other providers
England & Wales, UK and constituent countries, GORs, counties and local authorities
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Social Class.
Data extraction: Source
Office for National Statistics (ONS)
Data extraction: source URL
http://www.statistics.gov.uk/statbase/Product.asp?vlnk=8841 (See “Results for England and Wales”)
Data extraction: date
January 2009
Numerator: definition
Deaths from all causes, all ages, females
Numerator: source
Office for National Statistics (ONS)
Denominator: definition
Estimated resident population estimates, mid-year, all ages, females
Denominator: source
Office for National Statistics (ONS), available from http://www.statistics.gov.uk/statbase/Product.asp?vlnk=15106 (See mid-year estimates for 2005 to 2007)
Available from the Office for National Statistics (ONS)
Available from the Office for National Statistics (ONS)
The population estimates used for the calculation of figures for this indicator are based on the 2001 Census. The population estimates used to calculate the LE figures were published on the following dates: • Mid-year population estimates for 2005 - 22 August 2007; • Mid-year population estimates for 2006 – 25 October 2007; • Mid-year population estimates for 2007 – 21 August 2008. Data quality: Accuracy and completeness
Mortality data quality and coverage is extremely high. The figures are three-year averages so as to provide large enough numbers to ensure that the presented figures are sufficiently robust. Two authorities, City of London and Isles of Scilly, are excluded from the results because of small numbers of deaths and populations in these areas.
Health Profiles 2009 The Indicator Guide
231 Section 6: life expectancy and cause of death
Table 3 – Indicator Technical Methods Numerator: extraction
Extraction of the completed indicator from ONS http://www.statistics.gov.uk/statbase/Product.asp?vlnk=8841
Numerator: aggregation / allocation
Deaths were assigned to local authority boundaries by ONS using the National Statistics Postcode Directory.
Numerator data caveats
Source material used to calculate LE: Area of residence is allocated by ONS using the postcode and the National Statistics Postcode Directory – records without a valid area code are excluded but the number of such records is negligible.
Denominator data caveats
Denominators are estimates of mid-year populations, based on results of the 2001 Census.
Methods used to calculate indicator value
The figures are rolling three-year averages produced by aggregating deaths and population estimates for 2005–2007. Abridged life tables were constructed using standard methods. Separate tables were constructed for males and females. The tables were created using annual mid-year population estimates and deaths registered in each year. All figures presented are for life expectancy at birth. A life table template which illustrates the method used to calculate the life expectancy results is available on the ONS website: http://www.statistics.gov.uk/statbase/ssdataset.asp?vlnk=6949 Data for some new Unitary Authorities, introduced in April 2009, did not correspond with existing geographies (Cheshire East, Cheshire West and Cheshire, Central Bedfordshire, and Cornwall). Results for these areas were calculated using the South East Public Health Observatory life expectancy calculator: http://www.sepho.org.uk/viewResource.aspx?id=8943 This was also used to calculate figures for the two Strategic Health Authorities in the South East: South Central and South East Coast. Results for the other new UAs were taken from published data for local authorities (Bedford) or counties (County Durham, Northumberland, Shropshire, and Wiltshire).
Small Populations: How Isles of Scilly and City of London populations have been dealt with
City of London and Isles of Scilly are excluded from the results because of small numbers of deaths and populations in these areas. Isles of Scilly has not been included in the new UA for Cornwall and City of London has not been included with Hackney. However, Isles of Scilly and City of London data are all included in the regional figures for the South West and London, and England figures.
Disclosure Control
Not applicable
Health Profiles 2009 The Indicator Guide
232 Section 6: life expectancy and cause of death
Confidence Intervals calculation method
The calculation of the confidence intervals was made using the method developed by Chiang. A report which details research undertaken by the Office for National Statistics on comparing methodologies to enable the calculation of confidence intervals for life expectancy at birth has now been published as No 33 in the National Statistics Methodological Series. This report, “Life expectancy at birth: methodological options for small populations”, also presents research carried out to establish if there is a minimum population size below which the calculation of life expectancy may not be considered feasible. It concludes with a summary of methodological conclusions and considers how these could be applied to the calculation of life expectancy at birth for wards in England and Wales. A copy of the report can be found on the NS website at: http://www.statistics.gov.uk/methods_quality/publications.asp The SEPHO calculator however also includes an adjustment to include a term for the variance associated with the final age interval as developed by Silcocks. Chiang CL. The Life Table and its Construction. In: Introduction to Stochastic Processes in Biostatistics. New York, John Wiley & Sons, 1968: 189-214. Silcocks PBS, Jenner DA, Reza R. Life expectancy as a summary of mortality in a population: statistical considerations and suitability for use by health authorities. J Epidemiol Community Health 2001; 55: 38-4
Health Profiles 2009 The Indicator Guide
233 Section 6: life expectancy and cause of death
28. INFANT DEATHS INDICATOR Basic Information 1. What is being measured?
Infant Deaths
2. Why is it being measured?
Infant mortality is an indicator of the general health of an entire population. It reflects the relationship between causes of infant mortality and upstream determinants of population health such as economic, social and environmental conditions. Deaths occurring during the first 28 days of life (the neonatal period) in particular, are considered to reflect the health and care of both mother and newborn.
3. How is this indicator actually defined?
Infant deaths, crude rate, persons, aged less than 1 year, 2005–07, per 1,000 live births.
4. Who does it measure?
Infants aged less than 1 year.
5. When does it measure it?
This indicator is updated annually.
6. Will it measure absolute numbers or proportions?
Proportions: numbers per 1,000 live births.
7. Where does the data actually come from?
Office for National Statistics (ONS)
8. How accurate and complete will the data be?
Data on births and deaths are considered to be complete and robust.
9. Are there any caveats/warnings/ problems?
Records without a valid area code are excluded but the number of such records is negligible.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
Page 4 Health summary – Indicator no. 28
Subject category/ domain(s)
Life expectancy and causes of death
Indicator name (*Indicator title in health profile)
Infant deaths
PHO with lead responsibility
LHO
Date of PHO dataset creation
February 2009
Health Profiles 2009 The Indicator Guide
234 Section 6: life expectancy and cause of death
Indicator definition
Infant deaths, crude rate, persons, aged less than 1 year, 2005–07, per 1,000 live births
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs, and Strategic Health Authorities for the South East.
Timeliness
The Compendium infant mortality indicator is updated annually, usually around September of the following year.
Rationale: What this indicator purports to measure
This indicator measures the level of infant deaths in the area.
Rationale: Public Health Importance
Infant mortality is an indicator of the general health of an entire population. It reflects the relationship between causes of infant mortality and upstream determinants of population health such as economic, social and environmental conditions. Deaths occurring during the first 28 days of life (the neonatal period) in particular, are considered to reflect the health and care of both mother and newborn.
Rationale: Purpose behind the inclusion of the indicator
There is a national health inequalities target for infant mortality which aims to reduce the gap between the infant mortality rate in the ‘routine and manual classes’ and the population as a whole. However, this target is difficult to monitor at local level as the number of infant deaths in any given local authority or primary care trust (PCT) among a particular social class group is very small and subject to random fluctuations from year to year. Therefore we have chosen to include overall infant mortality as an indicator. There are wide inequalities in infant mortality rates by local authority in England and monitoring these inequalities is essential to understanding trends in inequalities in infant mortality.
Rationale: Policy relevance
There is a national health inequalities target for infant mortality which aims to reduce the gap between the infant mortality rate in the ‘routine and manual classes’ and the population as a whole.
Interpretation: What a high/low level of indicator value means
A high indicator value (red circle in health summary chart) represents a statistically significant higher level of infant deaths for that local authority when compared to the national value. A low indicator value (green circle in health summary chart) represents a statistically significant lower level of infant deaths for that local authority when compared to the national value. A reduction in the infant death rate indicates a reduction in the number of deaths, relative to the number of live births. However, as the number of infant deaths in any given area is small, fluctuations from year to year are possible and may not indicate a statistically significant trend.
Interpretation: Potential for error due to type of measurement method
Coverage can be considered to be complete as the registration of deaths is a legal requirement. Data quality for the relevant fields (age and area of residence) is extremely high. The small number of infant deaths at a local authority level means that pooling of three years data is required. Even with pooled rates, however, numbers may still be small and large random fluctuations possible.
Health Profiles 2009 The Indicator Guide
235 Section 6: life expectancy and cause of death
Interpretation: Potential for error due to bias and confounding
The rates are not standardised or adjusted to take into account any potential confounding variables such as the age or ethnicity of the mother. Whether or not such variables need to be considered depends on the purpose to which the indicator is being put.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider the confidence interval, the greater the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with a white symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or amber symbol depending on whether it is worse or better than the national value respectively.
TABLE 2 – INDICATOR SPECIFICATION Indicator definition: Variable
Infant deaths
Indicator definition: Statistic
Crude rate
Indicator definition: Gender
Persons
Indicator definition: age group
Less than 1 year
Indicator definition: period
2005–2007
Indicator definition: scale
Per 1,000 live births
Health Profiles 2009 The Indicator Guide
236 Section 6: life expectancy and cause of death
Geography: geographies available for this indicator from other providers
England & Wales, ONS area classification, Strategic Health Authority, Primary Care Organisation
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Infant death rates by the National Statistics Socio-Economic Classification (based on father’s occupation at death registration) are available at a national level from the Office for National Statistics. As only a 10 per cent sample of live births are coded for father’s occupation, local infant death rates by NS-SEC are not available.
Data extraction: Source
Compendium of Clinical and Health Indicators/Clinical and Health Outcomes Knowledge Base (www.nchod.nhs.uk or nww.nchod.nhs.uk) Source of data: National Statistics
Data extraction: source URL
http://www.nchod.nhs.uk/
Data extraction: date
Data extracted from source as at Jan 2009
Numerator: definition
The number of infant deaths aged under 1 year registered in 2005–07
Numerator: source
Office for National Statistics (ONS).
Denominator: definition
The number of live births registered in 2005-07
Denominator: source
Office for National Statistics (ONS).
Data quality: Accuracy and completeness
Statistics on births and deaths are derived from the registration of births and deaths. The Office for National Statistics complete a variety of quality checks on the data before making them available for analysis. Data on births and deaths are considered to be largely complete.
Available from: www.nchod.nhs.uk
Infant death rates are available at a national level by mother’s country of birth from ONS.
Table 3 – Indicator Technical Methods Numerator: extraction
Infant mortality counts by area were extracted by ONS and supplied to NCHOD.
Numerator: aggregation / allocation
Deaths were assigned to geographical areas by ONS using the postcode of residence and the National Statistics Postcode Directory (NSPD).
Numerator data caveats
Records without a valid area code are excluded but the number of such records is negligible.
Denominator data caveats
Live births were assigned to geographical areas by ONS using the postcode of residence and the National Statistics Postcode Directory (NSPD). Records without a valid area code are excluded but the number of such records is negligible.
Health Profiles 2009 The Indicator Guide
237 Section 6: life expectancy and cause of death
Methods used to calculate indicator value
Crude rate per 1,000 live births: The number of infant deaths is divided by the number of live births in the same area and multiplied by 1,000. Data were sourced from NCHOD except for the new Unitary Authorities, introduced in April 2009, which did not correspond with existing geographies (Cheshire East, Cheshire West and Cheshire, Central Bedfordshire, and Cornwall). Results for these areas were calculated by the LHO. Results for the other new UAs were taken from published data for local authorities (Bedford) or counties (County Durham, Northumberland, Shropshire, and Wiltshire).
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Data for Isles of Scilly and City of London have not been included due to small populations; however, data for regions (South West and London) and for England do include these data. Data for the Isles of Scilly have not been included in the new Cornwall unitary authority.
Disclosure Control
None applied.
Confidence Intervals calculation method
Within the Compendium the 95% confidence intervals for crude rates and percentages are calculated using the likelihood-based method described by Aitken et al, which is a good approximation of the exact method. For example lower and upper limits for a rate expressed per 1,000 are given by:
rLL
rUL
r 1.96 exp ln − 1 − r nr (1 − r ) = × 1,000 r 1.96 1 + exp ln − 1 − r nr (1 − r ) r 1.96 exp ln + 1 − r nr (1 − r ) = × 1,000 1 + exp ln r + 1.96 1 − r nr (1 − r )
where: r is the crude rate; n is the population-years at risk. Aitken M et al. Statistical Modelling in GLIM. Oxford: Oxford University Press, 1990.
Health Profiles 2009 The Indicator Guide
238 Section 6: life expectancy and cause of death
29. DEATHS FROM SMOKING INDICATOR Basic Information 1. What is being measured?
Rates of deaths, attributable to smoking.
2. Why is it being measured?
Because smoking is the single biggest cause of death and there are big inequalities between and within communities in smoking and in death rates due to smoking.
3. How is this indicator actually defined?
Deaths attributable to smoking per 100,000 population.
4. Who does it measure?
Input data are annual death files provided by ONS, rates are calculated using (ex-) smoking rates and estimates for smoking attributable fractions of various causes of deaths.
5. When does it measure it?
1/1/2005–31/12/2007
6. Will It measure absolute numbers or proportions?
Proportions: number of cases per 100,000 population.
7. Where does the data actually come from?
Office of National Statistics.
8. How accurate and complete will the data be?
Although the input data is very reliable, the results also depend on estimated (ex-) smoking rates and the estimated proportion of causes of death which can be attributed to smoking.
9. Are there any caveats/warnings/ problems?
No
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
Page 4 Spine Chart – Indicator 29
Subject category/ domain(s)
Life expectancy and causes of death
Indicator name (* Indicator title in health profile)
Smoking attributable mortality (“Deaths from smoking”)
PHO with lead responsibility
ERPHO
Date of PHO dataset creation
5 Feb 2009
Indicator definition
Deaths attributable to smoking, directly age standardised rate, 35 years +, 2005–07, persons.
Health Profiles 2009 The Indicator Guide
239 Section 6: life expectancy and cause of death
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs
Timeliness
Annual updates available from erpho
Rationale: What this indicator purports to measure
Rate of deaths that can be attributed to smoking in persons 35 and over
Rationale: Public Health Importance
Smoking still accounts for between 1 in 6 and 1 in 10 of all deaths in England, and accounts for about half of the inequality in death rates between spearhead and non-spearhead areas. It remains the biggest single cause of preventable mortality and morbidity in the world.
Rationale: Purpose behind the inclusion of the indicator
To encourage smoking prevention. To focus action on tackling smoking related disease and to help prioritise actions to tackle health inequalities. To promote better measurement of smoking prevalence. Smoking related deaths is a powerful proxy measure of overall health and predictor of health care demand.
Rationale: Policy relevance
Smoking Kills Choosing Health Smokefree England
Interpretation: What a high/low level of indicator value means
An indicator value better than average (green circle in health summary chart) represents a statistically significant better level of smoking attributable mortality for that local authority when compared to the national value. An indicator value worse than average (red circle in health summary chart) represents a statistically significant worse level of smoking attributable mortality for that local authority when compared to the national value. High smoking attributable death rates are indicative of poor population health and high smoking rates. A zero rate is unachievable but lower rates than the best in England are seen in California and Scandinavian countries.
Interpretation: Potential for error due to type of measurement method
The method relies on the use of estimates of the contribution of smoking to a range of causes of death derived from the American Cancer Prevention Society II study. It presumes a degree of generalisability of these estimates. Assuming that the contribution of smoking to deaths in Americans in the 1980’s is the same as its contribution to an English population in 2005-07. These estimates do not take into account any degree of uncertainty meaning the estimates will be over precise.
Interpretation: Potential for error due to bias and confounding
Smoking prevalence rates for England were used because the model required estimates of ex-smoking and non-smoking as well as current smoking rates. These figures are not available at local authority level. The method will tend to overestimate smoking related deaths in low prevalence areas and underestimate in high prevalence areas.
Health Profiles 2009 The Indicator Guide
240 Section 6: life expectancy and cause of death
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider the confidence interval is the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so it is said that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Table 2 – Indicator Specification Indicator definition: Variable
Deaths attributable to smoking
Indicator definition: Statistic
Directly age-standardised rate
Indicator definition: Gender
Persons
Indicator definition: age group
35 and over
Indicator definition: period
2005–07
Health Profiles 2009 The Indicator Guide
241 Section 6: life expectancy and cause of death
Indicator definition: scale
Per 100,000 European Standard population
Geography: geographies available for this indicator from other providers
SHAs Available from erpho on request
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Data can be calculated at MSOA level, and aggregated into deprivation quintiles to calculate within area inequalities. See for example http://www.erpho.org.uk/ viewResource.aspx?id=18243
Data extraction: Source
ONS for mortality data and population estimates. Prevalence of smoking, ex-smoking and non-smoking derived from Health Survey for England 2003–05 SAMMEC website (for relative risks from the American Cancer Prevention Society II Study 1982-1988)
Data extraction: source URL
http://apps.nccd.cdc.gov/sammec/ ONS mortality extracts are supplied to PHOs under a non-disclosure agreement and the provisions of the Health Act 2000. Health Survey for England http://www.ic.nhs.uk/statistics-and-data-collections/healthand-lifestyles-related-surveys/health-survey-for-england/health-survey-for-england2005-latest-trends
Data extraction: date
December 2008 (HSE, Sammec) December 2008 (mortality data)
Health Profiles 2009 The Indicator Guide
242 Section 6: life expectancy and cause of death
Numerator: definition
Disease Category
ICD10
Malignant Neoplasms Lip, Oral Cavity, Pharynx
C00–C14
Oesophagus
C15
Stomach
C16
Pancreas
C25
Larynx
C32
Trachea, Lung, Bronchus
C33–C34
Cervix Uteri
C53
Kidney and Renal Pelvis
C64–C65
Urinary Bladder
C67
Acute Myeloid Leukemia
C92.0
Cardiovascular Diseases Ischemic Heart Disease
I20–I25
Other Heart Disease
I00–I09, I26–I51
Cerebrovascular Disease
I60–I69
Atherosclerosis
I70
Aortic Aneurysm
I71
Other Arterial Disease
I72–I78
Respiratory Diseases Pneumonia, Influenza
J10–J18
Bronchitis, Emphysema
J40–J42, J43
Chronic Airway Obstruction
J44
Numerator: source
Death extracts from ONS: ONS PHO death extract 2005 ONS PHO death extract 2006 ONS PHO death extract 2007
Denominator: definition
Mid-year local authority population estimates
Denominator: source
ONS Mid-year population estimates 2005, ONS Mid year population estimates 2006, ONS Mid-year population estimates 2007, Estimates are as published for Persons rounded. The 3 years data is pooled.
Data quality: Accuracy and completeness
The accuracy of these estimates is contingent on the underlying accuracy of the three components: 1. Mortality data – because this is cause specific there may be coding variation between places although for the key contributors e.g. cancer deaths, this is less likely. 2. The relative risks are unpublished and only made available through the SAMMEC website 3. Smoking prevalence is estimated using national survey data. Future updates will prove how accurate these assumptions are.
Health Profiles 2009 The Indicator Guide
243 Section 6: life expectancy and cause of death
Table 3 – Indicator Technical Methods Numerator: extraction
CREATE TABLE P2564_HPSAMMEC_LA_ICD10 (RegPeriod varchar(8), LACode varchar(4), LAName varchar(30), Sex varchar(1), AgeGroup varchar(5), SAFAgeGrp varchar(5), ICDCode varchar(4), Deaths decimal(8,3)) INSERT INTO P2564_HPSAMMEC_LA_ICD10 (RegPeriod, LACode, LAName, Sex, AgeGroup, ICDCode, Deaths) SELECT ‘2005-07’, [LA Code now], [LA name], Sex, [FiveYearTo85], SUBSTRING([Und COD (non-neo)],1,3), COUNT(*) FROM M_Deaths2001to2007 LEFT JOIN A_DeathAgeBands ON M_Deaths2001to2007.Age = A_ DeathAgeBands.AgeInUnits AND M_Deaths2001to2007.[Age Unit] = A_ DeathAgeBands.Units LEFT JOIN HES.dbo.A_LACounty ON M_Deaths2001to2007.ResCty + M_ Deaths2001to2007.ResLAUA = HES.dbo.A_LACounty.[LA Code] LEFT JOIN HES.dbo.A_NationalPostcodes_Feb08 ON M_Deaths2001to2007. Pcode_8dig = HES.dbo.A_NationalPostcodes_Feb08.[pcd2] WHERE YEAR([Reg Date]) > 2004 AND YEAR([Reg Date]) < 2008 AND [LA Code now] <> ‘NULL’ AND [LA Code now] <> ‘15UH’ AND ([FiveYearTo85] = ‘35-39’ OR [FiveYearTo85] = ‘40-44’ OR [FiveYearTo85] = ‘45-49’ OR [FiveYearTo85] = ‘50-54’ OR [FiveYearTo85] = ‘55-59’ OR [FiveYearTo85] = ‘60-64’ OR [FiveYearTo85] = ‘65-69’ OR [FiveYearTo85] = ‘70-74’ OR [FiveYearTo85] = ‘75-79’ OR [FiveYearTo85] = ‘80-84’ OR [FiveYearTo85] = ‘85+’) AND (SUBSTRING([Und COD (non-neo)],1,3) = ‘C00’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C01’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C02’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C03’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C04’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C05’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C06’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C07’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C08’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C09’
Health Profiles 2009 The Indicator Guide
244 Section 6: life expectancy and cause of death
OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C10’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C11’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C12’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C13’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C14’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C15’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C16’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C25’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C32’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C33’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C34’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C53’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C64’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C65’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘C67’ OR [Und COD (non-neo)] = ‘C920’ OR SUBSTRING([Und COD (nonneo)],1,2) = ‘I0’ OR SUBSTRING([Und COD (non-neo)],1,2) = ‘I2’ OR SUBSTRING([Und COD (non-neo)],1,2) = ‘I3’ OR SUBSTRING([Und COD (non-neo)],1,2) = ‘I4’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘I50’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘I51’ OR SUBSTRING([Und COD (non-neo)],1,2) = ‘I6’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘I70’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘I71’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘I72’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘I73’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘I74’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘I75’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘I76’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘I77’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘I78’ OR SUBSTRING([Und COD (non-neo)],1,2) = ‘J1’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘J40’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘J41’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘J42’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘J43’ OR SUBSTRING([Und COD (non-neo)],1,3) = ‘J44’) GROUP BY [LA Code now], [LA Name], Sex, [FiveYearTo85] , SUBSTRING([Und COD (non-neo)],1,3) ORDER BY [LA Code now], Sex, [FiveYearTo85], SUBSTRING([Und COD (non-neo)],1,3)
Health Profiles 2009 The Indicator Guide
245 Section 6: life expectancy and cause of death
Numerator: aggregation/ allocation
Deaths in the smoking related ICD10 categories were multiplied by the Smoking Attributable Fraction (SAF), according to the age of the deceased (see below). Since the 2001 census new Unitary Authorities have been created. The new Unitary Authorities of Northumberland, Shropshire, Wiltshire, and Durham have the same boundaries as the counties of the same name so have been allocated these data. Data for five other new unitary Authorities have been calculated from the aggregation of previous Local Authorities. New UA name Bedford Central Bedfordshire Cheshire East
Cheshire West and Chester
Cornwall
Old LA name Bedford Mid Bedfordshire South Bedfordshire Congleton Crewe and Nantwich Macclesfield Chester Ellesmere Port & Neston Vale Royal Caradon Carrick Kerrier North Cornwall Penwith Restormel
Numerator data caveats Denominator data caveats
Population estimates used are Persons rounded and pooled across the 3 years. This may produce slight differences in rate calculations from those done locally. Population estimates of higher geographies are aggregated from LA populations.
Health Profiles 2009 The Indicator Guide
246 Section 6: life expectancy and cause of death
Methods used to calculate indicator value
Age-specific attributable deaths were calculated as follows: The SAF multiplier applied to deaths in each smoking related ICD10 category were determined using the formula: SAF = [p1(RR1 - 1) + p2(RR2 - 1)]/[1 + p1(RR1 - 1) + p2(RR2 - 1)] Where p1 = prevalence of current smokers; p2= ex-smoker prevalence; RR1 = relative risk in current smokers and RR2 = relative risk in former smokers. The following relative risks for each disease-age-sex group were used. RR Male Disease Category
Current Smoker
RR Female
Former Smoker
Current Smoker
Former Smoker
Malignant Neoplasms Lip, Oral Cavity, Pharynx
10.89
3.40
5.08
2.29
Esophagus
6.76
4.46
7.75
2.79
Stomach
1.96
1.47
1.36
1.32
Pancreas
2.31
1.15
2.25
1.55
Larynx
14.6
6.34
13.02
5.16
23.26
8.70
12.69
4.53
Cervix Uteri
1.00
1.00
1.59
1.14
Kidney and Renal Pelvis
2.72
1.73
1.29
1.05
Urinary Bladder
3.27
2.09
2.22
1.89
Acute Myeloid Leukemia
1.86
1.33
1.13
1.38
Persons Aged 35–64
2.80
1.64
3.08
1.32
Persons Aged 65+
1.51
1.21
1.60
1.20
Other Heart Disease
1.78
1.22
1.49
1.14
Persons Aged 35–64
3.27
1.04
4.00
1.30
Persons Aged 65+
1.63
1.04
1.49
1.03
Atherosclerosis
2.44
1.33
1.83
1.00
Aortic Aneurysm
6.21
3.07
7.07
2.07
Other Arterial Disease
2.07
1.01
2.17
1.12
1.75
1.36
2.17
1.10
Bronchitis, Emphysema
17.10
15.64
12.04
11.77
Chronic Airway Obstruction
10.58
6.80
13.08
6.78
Trachea, Lung, Bronchus
Cardiovascular Diseases Ischemic Heart Disease
Cerebrovascular Disease
Respiratory Diseases Pneumonia, Influenza
Health Profiles 2009 The Indicator Guide
247 Section 6: life expectancy and cause of death
Together with the following England smoking prevalence estimates Smoking Prevalence Sex
Age
Smoker
Former smoker
M
35-44
29.7%
19.6%
M
45-54
25.2%
31.4%
M
55-64
20.8%
44.5%
M
65-74
12.7%
54.7%
M
75+
8.2%
60.4%
F
35-44
28.9%
16.2%
F
45-54
26.9%
21.8%
F
55-64
22.0%
28.8%
F
65-74
14.8%
30.1%
F
75+
9.4%
36.0%
The counts of deaths were aggregated by relevant ICD codes, age and sex groups Relative risks were assumed to be constant within broader age bands Death counts in 5-year age bands were multiplied by the appropriate SAF to give smoking attributable deaths in each age-sex band. Directly age standardised rates were calculated using the batch calculator available from the erpho website http://www.erpho.org.uk/topics/tools/rates. aspx#12474 Small Populations: How Isles of Scilly and City of London populations have been dealt with
Isles of Scilly and the City of London were excluded from local authority calculations. City of London was included in the higher geographies (the Isles of Scilly were not included in higher geographies).
Disclosure Control
Not relevant
Confidence Intervals calculation method
Confidence intervals have been calculated using Byars method. See http://www.yhpho.org.uk/standardisedratecalculater.aspx
Health Profiles 2009 The Indicator Guide
248 Section 6: life expectancy and cause of death
30. EARLY DEATHS: HEART DISEASE AND STROKE INDICATOR Basic Information 1. What is being measured?
Mortality rate from all circulatory diseases (including heart disease and stroke).
2. Why is it being measured?
Circulatory disease accounts for 40% of all deaths (30% under 75). Mortality is a direct measure of health care need and indicates the overall circulatory disease burden on the population, reflecting both the incidence of disease and the ability to treat it.
3. How is this indicator actually defined?
Mortality from all circulatory diseases, directly age-standardised rate, persons, under 75, 200507 (pooled), per 100,000 European Standard population.
4. Who does it measure?
People aged under 75.
5. When does it measure it?
Updated annually.
6. Will it measure absolute numbers or proportions?
Directly age-standardised rate.
7. Where does the data actually come from?
Office for National Statistics (ONS).
8. How accurate and complete will the data be?
Mortality counts are derived from an annual mortality extract supplied by ONS and are based on the original underlying cause of death for which there is nearly 100% coverage on the mortality register.
9. Are there any caveats/warnings/ problems?
Area of residence is allocated by ONS using the postcode and the National Statistics Postcode Directory – records without a valid area code are excluded but the number of such records is negligible. There is the potential for the underlying cause of death to be incorrectly attributed on the death certificate and, therefore, the cause of death misclassified.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Table 1 – Indicator Description Information component
Pg 4 Health Summary – Indicator No 30
Subject category/ domain(s)
Life expectancy and causes of death
Health Profiles 2009 The Indicator Guide
249 Section 6: life expectancy and cause of death
Indicator name (*Indicator title in health profile)
Mortality rate from all circulatory diseases (*Early deaths: heart disease & stroke)
PHO with lead responsibility
SEPHO
Date of PHO dataset creation
05 February 2009
Indicator definition
Mortality from all circulatory diseases, directly age-standardised rate, persons, under 75, 2005–07 (pooled), per 100,000 European Standard population
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs (boundaries as at April 2008, new 2009 Unitary Authorities as at April 2009).
Timeliness
The Compendium mortality from all circulatory diseases indicator is updated annually, usually around November/December following the publication by ONS of the new year’s mortality extract (usually in May) and mid-year population estimates (usually August-September).
Rationale: What this indicator purports to measure
Early mortality from all circulatory diseases.
Rationale: Public Health Importance
Circulatory disease accounts for 40% of all deaths (30% under 75). Mortality is a direct measure of health care need reflecting the overall circulatory disease burden on the population, both the incidence of disease and the ability to treat it. The mortality rate may be improved by reducing the population’s risk (e.g. encouraging healthier lifestyles and reducing exposure to smoking), by earlier detection of disease and by more effective treatment.
Rationale: Purpose behind the inclusion of the indicator
To estimate premature mortality due to circulatory diseases.
Rationale: Policy relevance
The under 75 circulatory disease mortality rate is a key target indicator in the 1999 Public Health White Paper ‘Saving Lives: Our Healthier Nation’. The target is to reduce the number of deaths from circulatory disease in people aged under 75 years by at least two-fifths by 2010. The baseline for monitoring this target was the three year period 1995–97.
To reduce premature deaths from circulatory diseases.
This measure supports delivery of the Department of Health PSA targets and LDP and is relevant to Choosing Health, Coronary Heart Disease NSF and Programme for Action. Interpretation: What a high/low level of indicator value means
A high indicator value (red circle in health summary chart) represents a statistically significant higher (worse) rate of early deaths from circulatory disease when compared to the national value. A low indicator value (green circle in health summary chart) represents a statistically significant lower (better) rate of early deaths from circulatory disease when compared to the national value.
Health Profiles 2009 The Indicator Guide
250 Section 6: life expectancy and cause of death
Interpretation: Potential for error due to type of measurement method
Coverage can be considered to be complete as the registration of deaths is a legal requirement. Data quality for the relevant fields (age, sex, underlying cause of death, area of residence) is extremely high.
Interpretation: Potential for error due to bias and confounding
The rates are age-standardised. This improves the comparability of rates for different areas, or between different time periods, by taking into account differences in the age structures of the populations being compared.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate.
There is the potential for the underlying cause of death to be incorrectly attributed on the death certificate and, therefore, the cause of death misclassified.
This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
TABLE 2 – INDICATOR SPECIFICATION Indicator definition: Variable
Mortality from all circulatory diseases (ICD10 I00 –I99)
Indicator definition: Statistic
Directly age-standardised rate
Indicator definition: Gender
Persons
Health Profiles 2009 The Indicator Guide
251 Section 6: life expectancy and cause of death
Indicator definition: age group
Under 75
Indicator definition: period
2005–07 (pooled**) ** average of annual rates
Indicator definition: scale
Per 100,000 European Standard population
Geography: geographies available for this indicator from other providers
England & Wales, ONS area, Primary Care Organisation, Strategic Health Authority.
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Age, gender available from NCHOD.
Data extraction: Source
NCHOD.
Data extraction: source URL
Data received directly from NCHOD.
Data extraction: date
14 January 2009
Numerator: definition
Deaths from all circulatory disease, classified by underlying cause of death (ICD10 I00 – I99), registered in the respective calendar years 2005-07, in people aged under 75.
Numerator: source
Office for National Statistics (ONS)
Denominator: definition
2001 census based mid-year population estimates for respective calendar years 2005 to 2007, people aged under 75 (current as at 29 September, 2008).
Denominator: source
ONS
Data quality: Accuracy and completeness
Coverage can be considered to be complete as the registration of deaths is a legal requirement. Data quality for the relevant fields (age, sex, underlying cause of death, area of residence) is extremely high. Area of residence is allocated by ONS using the postcode and the National Statistics Postcode Directory – records without a valid area code are excluded but the number of such records is negligible.
Available from National Centre for Health Outcomes Development (NCHOD) website www.nchod.nhs.uk Data can also be found at Neighbourhood Renewal Unit Public Service Agreement Floor Targets (http://www.fti.communities.gov.uk/fti/).
Table 3 – Indicator Technical Methods Numerator: extraction
Extraction by NCHOD.
Numerator: aggregation / allocation
Deaths were assigned to geographical areas using the area code supplied in the mortality extract. This is derived from postcode of residence by the ONS using the National Statistics Postcode Directory (NSPD).
Health Profiles 2009 The Indicator Guide
252 Section 6: life expectancy and cause of death
Numerator data caveats
Area of residence is allocated by ONS using the postcode and the National Statistics Postcode Directory – records without a valid area code are excluded but the number of such records is negligible. Mortality counts are derived from the annual DH mortality extract supplied by ONS and are based on the original underlying cause of death for which there is nearly 100% coverage on the mortality register.
Denominator data caveats
Data are based on the latest revisions of ONS mid-year population estimates for the respective years, current as at 29 September 2008.
Methods used to calculate indicator value
The directly age-standardised rate (DSR) is the rate of events that would occur in a population with a standard age structure if that population were to experience the age-specific rates of the subject population. The standard population used is the European Standard Population. The age groups used are: Under 1, 1–4, 5-9,…, 80–84, 85+. The rate for 2005–07 has been calculated as the simple average of the individual annual rates. The rate is expressed per 100,000 population.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Isles of Scilly and City of London are excluded from the lower tier datasets but included in England and Regional figures.
Disclosure Control
None applied.
Confidence Intervals calculation method
Confidence intervals for the age-standardised rates were calculated using a normal approximation, i.e. +/- 1.96 standard errors. The standard errors are obtained using the method described by Breslow and Day, but modified to use the binomial variance for a proportion to estimate the variances of the crude age-specific rates. This method is likely to be unreliable when there are fewer than 50 cases in an area, hence confidence intervals for rates based on less than 50 cases should be viewed with caution. The lower and upper limits for the rates are denoted by DSRLL and DSRUL respectively. The lower and upper limits for the rates are denoted by DSRLL and DSRUL respectively.
DSR LL / UL = DSR ± 1.96 × 100,000 ×
1 ∑ wi ij
2
×∑ ij
wi2 ⋅ rij (1 − rij ) nij
(expressed per 100,000 population) where: wi is the number, or proportion, of individuals in the standard population in age group i. rij is the crude age-specific rate in the subject population in age group i, in year j. nij is the number of individuals in the subject population in age group i, in year j.
Health Profiles 2009 The Indicator Guide
253 Section 6: life expectancy and cause of death
31. EARLY DEATHS: CANCER INDICATOR Basic Information 1. What is being measured?
Mortality rate from all cancers.
2. Why is it being measured?
Early mortality from cancer is a direct measure of health care need as public health interventions for prevention, early diagnosis, and effective treatment can all reduce the burden of cancer morbidity and mortality.
3. How is this indicator actually defined?
Mortality from all cancers, directly age-standardised rate, persons, under 75, 2005–07 (pooled), per 100,000 European Standard population.
4. Who does it measure?
People aged under 75.
5. When does it measure it?
Updated annually.
6. Will It measure absolute numbers or proportions?
Directly age-standardised rate.
7. Where does the data actually come from?
Office for National Statistics (ONS).
8. How accurate and complete will the data be?
Mortality counts are derived from an annual mortality extract supplied by ONS and are based on the original underlying cause of death for which there is nearly 100% coverage on the mortality register.
9. Are there any caveats/warnings/ problems?
Area of residence is allocated by ONS using the postcode and the National Statistics Postcode Directory – records without a valid area code are excluded but the number of such records is negligible. There is the potential for the underlying cause of death to be incorrectly attributed on the death certificate and, therefore, the cause of death misclassified.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
TABLE 1 – INDICATOR DESCRIPTION Information component
Pg 4 Health Summary – Indicator No 31
Subject category/ domain(s)
Life expectancy and causes of death
Health Profiles 2009 The Indicator Guide
254 Section 6: life expectancy and cause of death
Indicator name (*Indicator title in health profile)
Mortality rate from all cancers (*Early deaths: cancer)
PHO with lead responsibility
SEPHO
Date of PHO dataset creation
05 February 2009
Indicator definition
Mortality from all cancers, directly age-standardised rate, persons, under 75, 2005–07 (pooled), per 100,000 European Standard population
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs (boundaries as at April 2008, new 2009 Unitary Authorities as at April 2009).
Timeliness
The Compendium mortality from all cancers indicator is updated annually, usually around November/December following the publication by ONS of the new year’s mortality extract (usually in May) and mid-year population estimates (usually August–September).
Rationale: What this indicator purports to measure
Early mortality from all cancers.
Rationale: Public Health Importance
Cancer is amongst the three leading causes of death at all ages except for pre-school age children in the UK. It accounts for 26% all deaths. If current incidence rates remain the same, by 2025 there will be an additional 100,000 cases of cancer diagnosed each year as a result of the ageing population. Inequalities exist in cancer rates between the most deprived areas and the most affluent. Early mortality from cancer is a direct measure of health care need as public health interventions for prevention, early diagnosis, effective treatment can all reduce the burden of cancer morbidity and mortality.
Rationale: Purpose behind the inclusion of the indicator
To estimate premature mortality due to cancer.
Rationale: Policy relevance
The directly age-standardised mortality rate from all cancers for persons aged under 75 is a target indicator in the Saving Lives: Our Healthier Nation strategy. The target is a 20% reduction by the year 2010 from the baseline rate in 1995–97.
To reduce premature deaths from cancer.
This measure supports delivery of the Department of Health PSA targets and LDP and is relevant to Choosing Health, Cancer NSF and Programme for Action. Interpretation: What a high/low level of indicator value means
A high indicator value (red circle in health summary chart) represents a statistically significant higher (worse) rate of early deaths from cancer for that local authority when compared to the national value. A low indicator value (green circle in health summary chart) represents a statistically significant lower (better) rate of early deaths from cancer when compared to the national value.
Health Profiles 2009 The Indicator Guide
255 Section 6: life expectancy and cause of death
Interpretation: Potential for error due to type of measurement method
Coverage can be considered to be complete as the registration of deaths is a legal requirement. Data quality for the relevant fields (age, sex, underlying cause of death, area of residence) is extremely high.
Interpretation: Potential for error due to bias and confounding
The rates are age-standardised. This improves the comparability of rates for different areas, or between different time periods, by taking into account differences in the age structures of the populations being compared.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate.
There is the potential for the underlying cause of death to be incorrectly attributed on the death certificate and, therefore, the cause of death misclassified.
This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Table 2 – Indicator Specification Indicator definition: Variable
Mortality from all cancers (ICD10 C00-C97)
Indicator definition: Statistic
Directly age-standardised rate
Indicator definition: Gender
Persons
Health Profiles 2009 The Indicator Guide
256 Section 6: life expectancy and cause of death
Indicator definition: age group
Under 75
Indicator definition: period
2005–07 (pooled**) ** average of annual rates
Indicator definition: scale
Per 100,000 European Standard population
Geography: geographies available for this indicator from other providers
ONS area, Primary Care Organisation, Strategic Health Authority.
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Age, gender available from NCHOD.
Data extraction: Source
NCHOD.
Data extraction: source URL
Data received directly from NCHOD.
Data extraction: date
14 January 2009
Numerator: definition
Deaths from all malignant neoplasms, classified by underlying cause of death (ICD10 C00-C97), registered in the respective calendar years 2005–07, in people aged under 75.
Numerator: source
Office for National Statistics (ONS)
Denominator: definition
2001 census based mid-year population estimates for respective calendar years 2005 to 2007, people aged under 75 (current as at 29 September, 2008).
Denominator: source
ONS
Data quality: Accuracy and completeness
Coverage can be considered to be complete as the registration of deaths is a legal requirement. Data quality for the relevant fields (age, sex, underlying cause of death, area of residence) is extremely high. Area of residence is allocated by ONS using the postcode and the National Statistics Postcode Directory – records without a valid area code are excluded but the number of such records is negligible.
Available from National Centre for Health Outcomes Development (NCHOD) website www.nchod.nhs.uk Data can also be found at Neighbourhood Renewal Unit Public Service Agreement Floor Targets (http://www.fti.communities.gov.uk/fti/).
Table 3 – Indicator Technical Methods Numerator: extraction
Extraction by NCHOD.
Numerator: aggregation/ allocation
Deaths were assigned to geographical areas using the area code supplied in the mortality extract. This is derived from postcode of residence by the ONS using the National Statistics Postcode Directory (NSPD).
Health Profiles 2009 The Indicator Guide
257 Section 6: life expectancy and cause of death
Numerator data caveats
Area of residence is allocated by ONS using the postcode and the National Statistics Postcode Directory - records without a valid area code are excluded but the number of such records is negligible. Mortality counts are derived from the annual DH mortality extract supplied by ONS and are based on the original underlying cause of death for which there is nearly 100% coverage on the mortality register.
Denominator data caveats
Data are based on the latest revisions of ONS mid-year population estimates for the respective years, current as at 29 September, 2008.
Methods used to calculate indicator value
The directly age-standardised rate (DSR) is the rate of events that would occur in a population with a standard age structure if that population were to experience the age-specific rates of the subject population. The standard population used is the European Standard Population. The age groups used are: Under 1, 1–4, 5-9,…, 80–84, 85+. The rate for 2005–07 has been calculated as the simple average of the individual annual rates. The rate is expressed per 100,000 population.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Isles of Scilly and City of London are excluded from the lower tier datasets but included in England and Regional figures.
Disclosure Control
None applied.
Confidence Intervals calculation method
Confidence intervals for the age-standardised rates were calculated using a normal approximation, i.e. +/- 1.96 standard errors. The standard errors are obtained using the method described by Breslow and Day, but modified to use the binomial variance for a proportion to estimate the variances of the crude age-specific rates. This method is likely to be unreliable when there are fewer than 50 cases in an area, hence confidence intervals for rates based on less than 50 cases should be viewed with caution. The lower and upper limits for the rates are denoted by DSRLL and DSRUL respectively.
DSR LL / UL = DSR ± 1.96 × 100,000 ×
1 ∑ wi ij
2
×∑ ij
wi2 ⋅ rij (1 − rij ) nij
(expressed per 100,000 population) where: wi is the number, or proportion, of individuals in the standard population in age group i. rij is the crude age-specific rate in the subject population in age group i, in year j. nij is the number of individuals in the subject population in age group i, in year j.
Health Profiles 2009 The Indicator Guide
258 Section 6: life expectancy and cause of death
32. ROAD INJURIES AND DEATHS INDICATOR Basic Information 1. What is being measured?
People killed or seriously injured on the road.
2. Why is it being measured?
To help reduce road traffic collisions which are a major cause of preventable death and serious injury.
3. How is this indicator actually defined?
People killed or seriously injured on the roads of the area, crude rate per 100,000 resident population, all ages, 2005–2007.
4. Who does it measure?
All persons, all ages
5. When does it measure it?
Continually reported and data is published annually
6. Will It measure absolute numbers or proportions?
Proportion: Crude rate of people killed or seriously injured per 100,000 resident population
7. Where does the data actually come from?
The data is collected by the Police and published by Department for Transport
8. How accurate and complete will the data be?
Coverage is complete.
9. Are there any caveats/warnings/ problems?
Not all road casualties are reported to police.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
The data point is green or red when the figure in a local authority is statistically significantly better or worse respectively than the England average, based on the 95% confidence intervals of the figure compared to the England value.
Data quality varies as there are differences between police forces in procedures for recording, collecting and collating.
The figures are for road casualties that occurred within a local authority area irrespective of the home residence of the person involved in the accident. Areas with high inflows of people or traffic may have artificially high rates because the at-risk resident population is not an accurate measure of exposure to transport. This is likely to affect the results for employment centres and sparsely populated rural areas which have high numbers of visitors or through traffic.
Table 1 – Indicator Description Information component
Pg 4 Health Summary – Indicator No 32
Subject category/ domain(s)
Life expectancy and causes of death
Indicator name (*Indicator title in health profile)
Road injuries and deaths
Health Profiles 2009 The Indicator Guide
259 Section 6: life expectancy and cause of death
PHO with lead responsibility
South West Public Health Observatory
Date of PHO dataset creation
January 2009 (revised February 2009 for new April 2009 area boundaries)
Indicator definition
People killed or seriously injured on the roads of the area, crude rate per 100,000 resident population, all ages, 2005–2007.
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs.
Timeliness
The indicator presented in Health Profiles is not routinely updated but may become available in future Health Profiles. The rates and numerators published by DfT are available annually (usually October-December).
Rationale: What this indicator purports to measure
Unintentional deaths and serious injuries on public roads caused by road traffic collisions.
Rationale: Public Health Importance
Motor vehicle traffic accidents are a major cause of preventable deaths and morbidity, particularly in younger age groups. For children and for men aged 20–64 years, mortality rates for motor vehicle traffic accidents are higher in lower socioeconomic groups. For instance, there would be 600 fewer deaths in men aged 20–64 years from motor vehicle traffic accidents each year if all men had the same death rates as those in social classes I and II combined (Acheson, D. Report of the Independent Inquiry into Inequalities in Health. London: TSO, 1998. Department of Health. The NHS Plan. London:TSO). The vast majority of road traffic collisions are preventable and can be avoided through improved education, awareness, road infrastructure and vehicle safety (Peden M, Scurfield R, Sleet D, Mohan D, Hyder AA, Jarawan E et al. 2004, World report on road traffic injury prevention. Geneva: World Health Organisation).
Rationale: Purpose behind the inclusion of the indicator
To help reduce road injury and death.
Rationale: Policy relevance
One of the Department of Transport’s PSA targets is to reduce the number of people killed or seriously injured by 40%, and the number of children killed or seriously injured by 50% by 2010 compared with the baseline of 1994–8. In Saving Lives: Our Healthier Nation, ‘accidents’, including road traffic collisions, were identified as one of the four main priority areas. Tackling Health Inequalities: A Programme For Action included a headline target to reduce road casualty numbers in disadvantaged areas.
Interpretation: What a high/low level of indicator value means
A high indicator value (red circle in health summary chart) represents a statistically significant higher rate of road injury and death when compared to the national value. A low indicator value (amber circle in health summary chart) represents a statistically significant rate of road injury and death when compared to the national value. However, as the vast majority of road traffic collisions are preventable, any number of deaths and injuries greater than zero is undesirable and therefore a low indicator value should not mean that PH action is not needed. Factors that should be considered when interpreting this indicator include the nature of the road network, volumes of traffic and number of pedestrians.
Health Profiles 2009 The Indicator Guide
260 Section 6: life expectancy and cause of death
Interpretation: Potential for error due to type of measurement method
Research has shown that not all road casualties are reported to police. Therefore, this indicator may be an under-estimate of the true level of serious injury.
Interpretation: Potential for error due to bias and confounding
The completeness of reporting road injury to the police has been shown to vary by factors such as age, injury severity and type of road user. For example, research suggests that cyclists are less likely to report injury to the police. There are also differences by age group, with higher reporting rates amongst younger and older people. Age also impacts on injury severity, with older people more likely to be seriously injured or killed when involved in a road collision. Therefore, population structure and mix of traffic in areas may contribute to variation in Local Authority values.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate.
Districts with low resident populations but which have high inflows of people or traffic may have artefactually high rates because the at-risk resident population is not an accurate measure of exposure to transport. This is likely to affect employment centres and sparsely populated rural areas which have high numbers of visitors or through traffic.
This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant and the value is shown on the health summary chart with an amber symbol. If the interval does not include the national value, the difference is statistically significant and the value is shown on the health summary chart with a red or green symbol depending on whether it is worse or better than the national value respectively.
Health Profiles 2009 The Indicator Guide
261 Section 6: life expectancy and cause of death
TABLE 2 – INDICATOR SPECIFICATION Indicator definition: Variable
People killed or seriously injured on the roads of the area.
Indicator definition: Statistic
Crude rate.
Indicator definition: Gender
Persons.
Indicator definition: age group
All ages.
Indicator definition: period
Numerator 2005–2007, denominator mid-2006.
Indicator definition: scale
Per 100,000 resident population.
Geography: geographies available for this indicator from other providers
England, GOR, Local Authority: Counties, Unitary Authorities, London Boroughs.
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Numerator data is available from www.dft.gov.uk for Counties, Unitary Authorities, London Boroughs, broken down by age group and road user type.
Data extraction: Source
Department for Transport.
Data extraction: source URL
http://www.dft.gov.uk/pgr/statistics/datatablespublications/accidents/ casualtieslatables/
Data extraction: date
Data extracted from source as at: 29/12/2008
Annual rates and numerators are available for these geographies from www.dft.gov. uk.
Health Profiles 2009 The Indicator Guide
262 Section 6: life expectancy and cause of death
Numerator: definition
3-year average of the number of people (all ages) killed or seriously injured on the roads of a LA in the period 2005–2007. The indicator is based on casualties who incur injury on the public highway (including footways) in which at least one road vehicle or a vehicle in collision with a pedestrian is involved and which becomes known to the police within 30 days of its occurrence. The vehicle need not be moving and accidents involving stationary vehicles and pedestrians or other road users are included. One accident may give rise to more than one casualty. This indicator includes only on casualties who are fatally or seriously injured and these categories are defined as follows. Fatal casualties are those who sustained injuries which caused death less than 30 days after the accident; confirmed suicides are excluded. Seriously injured casualties are those who sustained an injury for which they are detained in hospital as an in-patient, or any of the following injuries, whether or not they are admitted to hospital: fractures, concussion, internal injuries, crushings, burns (excluding friction burns), severe cuts and lacerations, severe general shock requiring medical treatment and injuries causing death 30 or more days after the accident. A casualty is recorded as seriously or slightly injured by the police on the basis of information available within a short time of the accident. This generally will not reflect the results of a medical examination, but may be influenced according to whether the casualty is hospitalised or not. Hospitalisation procedures will vary regionally.
Numerator: source
Department for Transport.
Denominator: definition
2001 Census based mid-year population estimate for the year 2006.
Denominator: source
Office for National Statistics (ONS).
Data quality: Accuracy and completeness
Coverage is complete. Data quality varies as there are differences between police forces in procedures for recording, collecting and collating. There is no information available about the extent of this variation. Serious injury is not recorded on the basis of a clinical diagnosis. This may cause problems in classification of serious versus slight injury. Other sources of data which measure road injury and death are death registrations data (which matches well with police reports of road deaths) and hospital episode statistics. The match between hospital admissions and police reports of serious injury is discussed comprehensively in recent national reports: http://www.dft.gov.uk/pgr/roadsafety/research/rsrr/theme5/trendsfatalcar76.pdf http://eprints.ucl.ac.uk/archive/00003373/01/3373.pdf There are differences between police data and hospital records of road casualties due to differences in both definitions and reporting.
Table 3 – Indicator Technical Methods Numerator: extraction
Downloaded from www.dft.gov.uk
Numerator: aggregation /allocation
Calculated a 3-year average for 2005–2007 by summing the individual years and dividing by 3.
Health Profiles 2009 The Indicator Guide
263 Section 6: life expectancy and cause of death
Numerator data caveats
Research has shown that not all road casualties are reported to police. DfT published two papers on the level of under-reporting on 23 June 2006, which can be found at the addresses below: http://www.dft.gov.uk/pgr/roadsafety/research/rsrr/theme5/trendsfatalcar76.pdf http://eprints.ucl.ac.uk/archive/00003373/01/3373.pdf These reports suggest that the serious group of casualties could be up to twice as large as indicated by the ‘serious’ category in police records. Not all of this shortfall in the serious group of casualties is due to under-reporting because in the slight category are casualties which should be in the serious category and have been misclassified or misrecorded. These could add up to another 25% to the serious category. The police definition of ‘serious’ injury has remained consistent and therefore this data can be used to look at time trends. However, it should be noted that recording practices may have changed over time. An alternative to using casualties and deaths occurring on the roads in the area would have been to use a measure of resident casualties and deaths, regardless of where they occurred, using hospital admissions data. This would be a good measure of the health effects upon the population which lives in the district but would be less useful for targeting of road safety interventions. For this reason, occurring casualties and deaths were used in the Health Profiles.
Denominator data caveats
Ideally, the denominator should include all people travelling on public roads in the area (in vehicles and as pedestrians) and take account of the distances travelled. This measure is not available. The use of resident population as a denominator is a proxy measure for population exposure and is consistent with how this indicator is presented elsewhere.
Methods used to calculate indicator value
Calculation of the numerator: Calculated a 3-year average for 2005–2007 by summing the individual years and dividing by 3. Denominator count: Mid-2006 population estimates (all ages). The numerator was then divided by the denominator; the resulting value was then multiplied by 100,000 to give a crude rate per 100,000 population.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Any casualties recorded on the Isles of Scilly are included in the Cornwall numerator, consistent with how this data is published by DfT. Therefore the Isles of Scilly population is included in the Cornwall denominator.
Disclosure Control
Not applicable as no counts less than 5.
The City of London and Heathrow counts (numerator and denominator) are included in the London regional total and the England total.
Health Profiles 2009 The Indicator Guide
264 Section 6: life expectancy and cause of death
Confidence Intervals calculation method
The confidence intervals for these crude rates were constructed using the three year total in the following formula that relates the chi-square and Poisson distributions:
χ α ,2d 2
LL =
2
2
χ 1- α ,2(d+1) 2
2
UL =
2
where LL and UL are the lower and upper 100*(1-a) per cent confidence limits and d denotes the number of observed events (e.g, serious injuries, violent offences, (100*α) deaths) per unit of time exposed. χ² a/2,2d is the (100*a/2)th percentage point for a chisquared distribution with 2d degrees of freedom and χ²(1-α/2),2(d+1) is the (100*(1-a/2))th percentage point for a chi-squared distribution on 2(d+1) degrees of freedom.
The confidence limits for the rates were then obtained by dividing the upper and lower limits for the counts by the person time exposed. Reference: Dobson AJ, Kuulasmaa K, Eberle E, Scherer J. Confidence intervals for weighted sums of Poisson parameters. Statistics in Medicine 1991;10:457-462.
Health Profiles 2009 The Indicator Guide
265
Section 7: Charts and trend graphs
Health Profiles 2009 The Indicator Guide
266 Section 7: charts and trend graphs
33. DEPRIVATION CHART Basic Information 1. What is being measured?
Proportion of the population of an area living in each of the five national deprivation quintiles, based on the Indices of Deprivation 2007.
2. Why is it being measured?
The differences in deprivation between areas are a major determinant of health inequality in the UK.
3. How is this indicator actually defined?
These charts are based on the Index of Multiple Deprivation (IMD) 2007, which is a model measuring deprivation at the neighbourhood level. It is underpinned by separate dimensions of deprivation, these dimensions are weighted and an overall deprivation score is given. Each neighbourhood, or Lower Layer Super Output Area (LSOA), in England is allocated an IMD score. These are ranked for all 32482 LSOAs in England and divided into five equal groups or ‘national quintiles’. The charts are shaded according to the proportion of the relevant population in an area living in neighbourhoods belonging to each national deprivation quintile (where 1 is the most deprived quintile and 5 is the least deprived). (The relevant population is the population estimate used in the construction of the IMD 2007)
4. Who does it measure?
All persons, all ages in the relevant population (otherwise known as the ‘at risk’ population). (This is the population estimate used in the construction of the IMD 2007 – see Table 2)
5. When does it measure it?
2007 Based on various indicators from 2005 and 2001
6. Will It measure absolute numbers or proportions?
Proportion based on a composite indicator.
7. Where does the data actually come from?
DCLG
8. How accurate and complete will the data be?
All indicators included in the IMD 2007 were chosen using criteria that included accuracy and completion.
9. Are there any caveats/warnings/ problems?
The indicators are based on mainly 2005 data and this is therefore at least 4 years out of date. It is based on an average score of an area and it can’t be assumed to represent all individuals in that area. Although very comprehensive, some aspects of deprivation will not be included in the indices, for example if the data is incomplete or not collected. This may have a larger effect in some areas than others.
Health Profiles 2009 The Indicator Guide
267 Section 7: charts and trend graphs
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
n/a
Table 1 – Indicator Description Information component
Chart on p2 of District and County profiles.
Subject category/ domain(s)
Deprivation
Indicator name (*Indicator title in health profile)
Deprivation: a national perspective
PHO with lead responsibility
Yorkshire & Humber
Date of PHO dataset creation
February 2009
Indicator definition
This is a measure of the IMD 2007 which is a model measuring deprivation in an area. It is underpinned by separate dimensions of deprivation, these dimensions are weighted and an overall deprivation score is given. The charts are shaded according to the proportion of the relevant population in an area living in neighbourhoods belonging to each national deprivation quintile (where 1 is the most deprived quintile and 5 is the least deprived). (The relevant population is the population estimate used in the construction of the IMD 2007)
Geography
National, Regional, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs
Timeliness
Indicator is not regularly updated. Published 2004 and 2007, the definition is not completely consistent due to changes in some of the indicators used.
Rationale: What this indicator purports to measure
Level of deprivation of the relevant population of an area using a composite of domains which are made up of a range of indicators. LSOA areas are divided into quintiles where 1 is the most deprived quintile and 5 is the least.
Rationale: Public Health Importance
The differences in deprivation between areas are a major determinant of health inequality in the United Kingdom.
(The relevant population is the population estimate used in the construction of the IMD 2007)
Many studies and analyses have demonstrated the association of increasingly poor health with increasing deprivation. For instance, all cause mortality, smoking prevalence, self-reported long standing illness and a diet rich in fruit and vegetables are all positively or negatively correlated with deprivation. If deprivation inequality does not decrease, Acheson takes the view that health inequalities are also unlikely to decrease.
Health Profiles 2009 The Indicator Guide
268 Section 7: charts and trend graphs
Rationale: Purpose behind the inclusion of the indicator
To monitor and help reduce health inequalities.
Rationale: Policy relevance
• Acheson D. Report of the Independent Inquiry into Inequalities in Health. London: TSO, 1998. • Department of Health. The NHS Plan. London: TSO, 2000. • www.dh.gov.uk/assetRoot/04/05/57/83/04055783.pdf • Department of Health. Tackling Health Inequalities: A Programme for Action. Department of Health, 2003. • www.dh.gov.uk/assetRoot/04/01/93/62/04019362.pdf • HM Treasury 2007 PBR CSR: Public service agreements: • http://www.hm-treasury.gov.uk/ • Department for Communities and Local Government The New Performance Framework for Local Authorities & Local Authority Partnerships: Single Set of National Indicators Department for Communities and Local Government, 2007 http://www.communities.gov.uk/localgovernment/
Interpretation: What a high/low level of indicator value means
A high indicator value (dark shading) represents a high level of deprivation by national standards.
Interpretation: Potential for error due to type of measurement method
The indicators are based on 2005 and 2001 data and this is therefore at least 4 years out of date.
Interpretation: Potential for error due to bias and confounding
It is based on an average score of an area and can’t be assumed to represent all individuals in that area. Although very comprehensive, some aspects of deprivation will not be included in the indices, for example if the data is incomplete or not collected. This may have a larger effect in some area than others.
Confidence Intervals: Definition and purpose
N/A
Table 2 – Indicator Specification Indicator definition: Variable
Indices of Deprivation 2007
Indicator definition: Statistic
National quintile: All LSOAs in England are ranked according to deprivation score and split into five equal groups
Indicator definition: Gender
Persons
Indicator definition: age group
All ages
Indicator definition: period
2001 and 2005
Health Profiles 2009 The Indicator Guide
269 Section 7: charts and trend graphs
Indicator definition: scale
Proportion of relevant population in each national quintile.
Geography: geographies available for this indicator from other providers
IMD 2007 overall score and individual domains are available at Lower Layer Super Output Area (LSOA) level. Also, some summary scores are available at LA and County level from http://www.communities.gov.uk/communities/ neighbourhoodrenewal/deprivation/deprivation07/
Dimensions of inequality: subgroup analyses of this dataset available from other providers
IMD 2007 individual domains are available at Lower Layer Super Output Area (LSOA) level from http://www.communities.gov.uk/communities/ neighbourhoodrenewal/deprivation/deprivation07/
Data extraction: source
Department of Communities and Local Government
Data extraction: source URL
http://www.communities.gov.uk/communities/neighbourhoodrenewal/ deprivation/deprivation07/
Data extraction: date
February 2009
Numerator: definition
Number of the relevant population living in each quintile in England based on the IMD 2007 score.
(The relevant population is the population estimate used in the construction of the IMD 2007)
(The relevant population is the population estimate used in the construction of the IMD 2007). Numerator: source
Department of Communities and Local Government
Denominator: definition
Denominator data – ‘at risk’ mid-2005 population estimates (elsewhere referred to as the ‘relevant’ population). The figures have been adjusted from the ONS mid-year estimate to exclude the prison population in order to fit the definition of ‘at risk’. The figures have been subject to disclosure control. They are aggregated from LSOA level to the LA, county, region or national level.
Denominator: source
Office for National Statistics (ONS) and Department of Communities and Local Government (DCLG).
Data quality: accuracy and completeness
Criteria for inclusion of indicators to ID 2004 and ID 2007 included: • Up-to-date • Statistically robust • Available for the whole of England at a small level in a consistent form
TABLE 3 – INDICATOR TECHNICAL METHODS Numerator: extraction
Department of Communities and Local Government website http://www.communities.gov.uk/communities/neighbourhoodrenewal/deprivation/ deprivation07/
Numerator: aggregation/ allocation
The relevant population by national quintile of deprivation was aggregated from Lower Super Output Area to Local Authority, county, regional and national level. (The relevant population is the population estimate used in the construction of the IMD 2007).
Health Profiles 2009 The Indicator Guide
270 Section 7: charts and trend graphs
Numerator data caveats
Deprivation level is based on the average for the area; not all individuals in the area will be deprived.
Denominator data caveats
Census data and mid-year estimates are known to be deficient in their estimates of: • Non-white populations • Full-time students • Men aged 20-39 • People living in nursing homes etc • Rough sleepers • Inner-city populations • Households of multiple occupation • Migrants
Methods used to calculate indicator value
From the downloaded file with LSOA data. All LSOAs in England are ranked according to deprivation score and split into five equal quintiles. The relevant population from each LSOA is then aggregated to the area being measured and a proportion calculated for the population within that area in each national quintile.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Only included at regional and national level
Disclosure Control
Every effort has been made by the DCLG to ensure that data do not allow the disclosure of confidential information.
Confidence Intervals calculation method
N/A
(The relevant population is the population estimate used in the construction of the IMD 2007)
Health Profiles 2009 The Indicator Guide
271 Section 7: charts and trend graphs
34. LIFE EXPECTANCY BY DEPRIVATION QUINTILE CHART Basic Information 1. What is being measured?
Life expectancy by deprivation quintile.
2. Why is it being measured?
All cause mortality is a fundamental and probably the oldest measure of the health status of a population. Differences in levels of all-cause mortality reflect health inequalities between different population groups, e.g. between genders, social classes and ethnic groups. Life expectancy at birth is chosen as the preferred summary measure of all cause mortality as it quantifies the differences between areas in units (years of life) that are more readily understood and meaningful to the audience than those of other measures. Within every local authority there are different levels of deprivation. This indicator therefore looks at the within area inequalities in life expectancy.
3. How is this indicator actually defined?
Life expectancy at birth, years, all ages, 2003–2007, by national deprivation quintile.
4. Who does it measure?
Males, females, all ages.
5. When does it measure it?
Data for this indicator is updated annually.
6. Will it measure absolute numbers or proportions?
Measures life expectancy at birth in years.
7. Where does the data actually come from?
Office for National Statistics (ONS).
8. How accurate and complete will the data be?
Data on deaths are considered to be complete and robust.
9. Are there any caveats/warnings/ problems?
Death records without a valid area code are excluded but the number of such records is negligible. The populations used for the calculation of figures for this indicator are based on the 2001 Census, and are experimental estimates. Deprivation quintiles were assigned to Lower Layer Super Output Areas (LSOAs) at a national level. Results will be missing for areas where a local authority had none of its population living in a particular quintile. Results are also not presented for areas where the population at risk for the years 2003–07 was less than 5,000 i.e. where the total aggregated population for these years fell below this threshold. These areas are displayed as “n/a” in the data table.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
95% confidence intervals should be considered when making comparisons between areas or genders.
Health Profiles 2009 The Indicator Guide
272 Section 7: charts and trend graphs
Table 1 – Indicator Description Information component
Pg 2 Health Inequalities: life expectancy
Subject category/ domain(s)
Health Inequalities
Indicator name (* Indicator title in health profile)
Life expectancy
PHO with lead responsibility
LHO
Date of PHO dataset creation
March 2009
Indicator definition
Life expectancy at birth, years, all ages, 2003–07, males and females, by national deprivation quintiles (IMD 2007).
Geography
England, GOR, SHAs in the South East, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs.
Timeliness
ONS produced data are updated annually in the autumn of the following year.
Rationale: What this indicator purports to measure
Life expectancy at birth is a summary measure of the all cause mortality rates in an area in a given period. It is the average number of years a new-born baby would survive, were he or she to experience the particular area’s age-specific mortality rates for that time period throughout his or her life.
Rationale: Public Health Importance
All cause mortality is a fundamental and probably the oldest measure of the health status of a population. It represents the cumulative effect of the prevalence of risk factors, prevalence and severity of disease, and the effectiveness of interventions and treatment. Differences in levels of all-cause mortality reflect health inequalities between different population groups, e.g. between genders, social classes and ethnic groups. Life expectancy at birth is chosen as the preferred summary measure of all cause mortality as it quantifies the differences between areas in units (years of life) that are more readily understood and meaningful to the audience than those of other measures.
Rationale: Purpose behind the inclusion of the indicator
To help reduce premature mortality and facilitate planning of health services at local level.
Rationale: Policy relevance
There is a national health inequalities target for life expectancy which aims to increase average life expectancy at birth in England to 78.6 years for men and to 82.5 years for women, and to reduce health inequalities by 10% by 2010 as measured by life expectancy at birth (Department of Health PSA priority 1). The focus of the life expectancy target is on the quintile with the worst indicators of mortality and highest levels of deprivation, known as the Spearhead group. However, within every local authority there are different levels of deprivation. This indicator therefore looks at the within area inequalities in life expectancy by national deprivation quintile. Life expectancy is also an indicator in the following: Local basket of inequalities indicators - Indicator 13.12. Opportunity for all - Communities - Indicator 39. Quality of life indicators - Health and social well-being - Indicator 33
Health Profiles 2009 The Indicator Guide
273 Section 7: charts and trend graphs
Interpretation: What a high/low level of indicator value means
The higher the life expectancy, the longer the estimated expectation of life for males and females living in that area and deprivation quintile at that time.
Interpretation: Potential for error due to type of measurement method
The figures reflect the contemporary mortality among those living in the area in each time period. They are not the number of years a baby born in the area in each time period could actually expect to live, both because the death rates of the area are likely to change in the future and because many of those born in the area will live elsewhere for at least some part of their lives. Life expectancy at birth is also not a guide to the remaining expectancy of life at any other given age. For example, if female life expectancy at birth was 80 years for a particular area, life expectancy of women aged exactly 75 years in that area would exceed 5 years. This reflects the fact that survival from a particular age depends only on the mortality rates beyond that age, whereas survival from birth is based on mortality rates for all ages.
Interpretation: Potential for error due to bias and confounding
Older people living in nursing homes tend to be in poorer health than those not living in nursing homes. As these homes are unevenly distributed across the country, a higher death rate – consequently lower life expectancy level – in one area could simply be the result of migration of frail older people moving into nursing homes in that area.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider the confidence interval is, the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals can also be used to make comparisons between deprivation quintiles within each local authority. If the confidence intervals don’t overlap, this indicates that the difference in life expectancy is statistically significant.
Health Profiles 2009 The Indicator Guide
274 Section 7: charts and trend graphs
Table 2 – Indicator Specification Indicator definition: Variable
Life expectancy by national deprivation quintile (IMD 2007)
Indicator definition: Statistic
Life expectancy by national deprivation quintile (IMD 2007)
Indicator definition: Gender
Males and Females
Indicator definition: age group
All ages
Indicator definition: period
2003 to 2007
Indicator definition: scale Geography: geographies available for this indicator from other providers
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs.
Dimensions of inequality: subgroup analyses of this dataset available from other providers
None
Data extraction: Source
Calculated by LHO
Data extraction: source URL
N/A
Data extraction: date
N/A
Numerator: definition
Number of deaths by age, sex (2003-2007) and deprivation quintile (IMD 2007). Deprivation quintiles were assigned to LSOAs at a national level.
Numerator: source
Mortality data from Office for National Statistics (ONS), analysed by LHO
Denominator: definition
Lower Layer Super Output Area (LSOA) population estimates by age, sex (2003–2007) and deprivation quintile (IMD 2007). Deprivation quintiles were assigned to LSOAs at a national level. 2006 populations were revised estimates provided by ONS in February 2009. Population estimates for LSOAs were not available by single year of age from ONS for 2003. To provide estimates for the age groups under 1 and 1–4 (needed for the life table calculation), the age group 0–4 was divided into 5. This was based on an analysis of estimates from 2004–07 which had shown that populations were evenly distributed, at national level, across each year of age between 0–4. Population estimates for LSOAs are not published by single year of age by ONS, but are made available to PHOs for analysis purposes.
Health Profiles 2009 The Indicator Guide
275 Section 7: charts and trend graphs
Denominator: source
Small Area Population Estimates Unit, Office for National Statistics (ONS) Centre for Demography.
Data quality: Accuracy and completeness
Mortality data quality and coverage is extremely high. The figures are five-year averages so as to provide large enough numbers to ensure that the presented figures are sufficiently robust. However, areas with a population below 5,000 person years at risk across the period 2003–07, have not been included in line with ONS recommendations and are displayed as “n/a” in the data table. Where a local authority had none of its population living in a particular quintile no bar will be displayed in the chart for that quintile and no value presented in the data table. Two authorities, City of London and Isles of Scilly, are excluded from the results because of small numbers of deaths and populations in these areas.
Table 3 – Indicator Technical Methods Numerator: extraction
Data from ONS, extraction by LHO.
Numerator: aggregation/ allocation
Deaths were assigned to Lower Layer Super Output Areas (LSOA) by the LHO using the November 2008 version of the National Statistics Postcode Directory. LSOAs were each assigned to a national deprivation quintile and these were then aggregated within each geographical area.
Numerator data caveats Denominator data caveats
The populations used to calculate life expectancy are experimental estimates.
Methods used to calculate indicator value
Life expectancy was calculated using the South East Public Health Observatory life expectancy calculator: http://www.sepho.org.uk/viewResource. aspx?id=8943
Small Populations: How Isles of Scilly and City of London populations have been dealt with
City of London and Isles of Scilly are excluded from the results because of small numbers of deaths and populations in these areas.
Disclosure Control
Not applicable
Confidence Intervals calculation method
The calculation of the confidence intervals was made using the method developed by Chiang. The SEPHO calculator however also includes an adjustment to include a term for the variance associated with the final age interval as developed by Silcocks. Chiang CL. The Life Table and its Construction. In: Introduction to Stochastic Processes in Biostatistics. New York, John Wiley & Sons, 1968: 189-214.Silcocks PBS, Jenner DA, Reza R. Life expectancy as a summary of mortality in a population: statistical considerations and suitability for use by health authorities. J Epidemiol Community Health 2001; 55: 38-4
Health Profiles 2009 The Indicator Guide
276 Section 7: charts and trend graphs
35. TREND 1: ALL AGE, ALL CAUSE MORTALITY Basic Information 1. What is being measured?
Trend in death rates from all causes
2. Why is it being measured?
All cause mortality is a fundamental and probably the oldest measure of the health status of a population. It represents the cumulative effect of the prevalence of risk factors, prevalence and severity of disease, and the effectiveness of interventions and treatment. Differences in levels of all-cause mortality reflect health inequalities between different population groups, e.g. between genders, social classes and ethnic groups.
3. How is this indicator actually defined?
Mortality from all causes, directly age-standardised rate, males and females, all ages, 1996–98 to 2005– 07 (average of annual rates) per 100,000 European Standard Population.
4. Who does it measure?
All persons, all ages
5. When does it measure it?
This indicator is updated annually
6. Will it measure absolute numbers or proportions?
Proportions: numbers of deaths per hundred thousand European Standard Population
7. Where does the data actually come from?
Office for National Statistics (ONS)
8. How accurate and complete will the data be?
Data on deaths are considered to be complete and robust
9. Are there any caveats/warnings/ problems?
Area of residence is allocated by ONS using the postcode and the National Statistics Postcode Directory - records without a valid area code are excluded but the number of such records is negligible. The populations used for the calculation of the figures for this indicator are based on the 2001 Census, and are estimates.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
Data are age-standardised to allow comparisons between areas. 95% confidence intervals should be considered when making comparisons between areas or over time.
Health Profiles 2009 The Indicator Guide
277 Section 7: charts and trend graphs
TABLE 1 – INDICATOR DESCRIPTION Information component
Pg 3 Health Inequalities: changes over time
Subject category/ domain(s)
Health inequalities: changes over time
Indicator name (*Indicator title in health profile)
Trend in death rates from all causes.
PHO with lead responsibility
LHO
Date of PHO dataset creation
March 2009
Indicator definition
Mortality from all causes, directly age-standardised rate, males and females, all ages, 1996-98 to 2005-07 (average of annual rates), per 100,000 European Standard Population
Geography
England, GOR, Strategic Health Authorities for the South East, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs (boundaries as at April 2009).
Timeliness
The Compendium mortality from all causes indicator is updated annually, usually around November following the publication by ONS of the new year’s mortality extract (usually in May) and mid-year population estimates (usually August-September).
Rationale: What this indicator purports to measure
Trend in mortality from all causes.
Rationale: Public Health Importance
All cause mortality is a fundamental and probably the oldest measure of the health status of a population. It represents the cumulative effect of the prevalence of risk factors, prevalence and severity of disease, and the effectiveness of interventions and treatment. Differences in levels of all-cause mortality reflect health inequalities between different population groups, e.g. between genders, social classes and ethnic groups.
Rationale: Purpose behind the inclusion of the indicator
To monitor mortality rates due to all causes over time.
Rationale: Policy relevance
There is a national health inequalities target for life expectancy which aims to increase average life expectancy at birth in England to 78.6 years for men and to 82.5 years for women, and to reduce health inequalities by 10% by 2010 as measured by life expectancy at birth (Department of Health PSA priority 1). All age all cause mortality is a good proxy for life expectancy and is being used to measure progress towards meeting the life expectancy target. The baseline for monitoring this target is the three year period 1995–97. The ten year trend charts presented in the health profiles show the three year moving averages of annual mortality rates from all causes from the 1996–98 to the most recently available years of data (2005–07 average) compared against the national trend. Earlier trend years are available from previous Health Profiles
Health Profiles 2009 The Indicator Guide
278 Section 7: charts and trend graphs
Interpretation: What a high/low level of indicator value means
An upward sloping trend line indicates that the mortality rate from all causes is worsening. A downward sloping trend line indicates that the mortality rate from all causes is improving. If the trend line for the local authority is consistently above the trend line for England then mortality rates from all causes have been consistently higher than those for England for the stated period. If the trend line for the local authority is consistently below the trend line for England then mortality rates from all causes have been consistently lower than those for England for the stated period. If the local authority line is above that of England and gap between the local authority trend line and the England trend line is widening then inequalities in mortality rates from all causes are worsening. If the authority line is above that of England and the gap between the local authority trend line and the England trend line is narrowing then inequalities in mortality rates from all causes are improving.
Interpretation: Potential for error due to type of measurement method
Coverage can be considered to be complete as the registration of deaths is a legal requirement. Data quality for the relevant fields (age, sex, area of residence) is extremely high.
Interpretation: Potential for error due to bias and confounding
The rates are age-standardised. This improves the comparability of rates for different areas, or between different time periods, by taking into account differences in the age structures of the populations being compared.
TABLE 2 – INDICATOR SPECIFICATION Indicator definition: Variable
Mortality from all causes
Indicator definition: Statistic
Directly age-standardised rate
Indicator definition: Gender
Male and Female
Indicator definition: age group
All ages
Indicator definition: period
1996-98 to 2005-07 -average of annual rates, with the exception of new 2009 Unitary Authorities: Cheshire East, Cheshire West and Chester, Central Bedfordshire, and Cornwall. Rates for these areas were based on data pooled for each three-year period.
Indicator definition: scale
Per 100,000 European Standard population
Geography: geographies available for this indicator from other providers
England & Wales, ONS area classification, Primary Care Organisation, Strategic Health Authority. Available from National Centre for Health Outcomes Development (NCHOD) website www.nchod.nhs.uk
Health Profiles 2009 The Indicator Guide
279 Section 7: charts and trend graphs
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Age, gender available from NCHOD.
Data extraction: Source
NCHOD.
Data extraction: source URL
Data received directly from NCHOD with the exception of new 2009 Unitary Authorities: Cheshire East, Cheshire West and Chester, Central Bedfordshire, and Cornwall. Rates for these areas were calculated by the LHO using annual mortality files, supplied to PHOs by ONS, and published ONS mid-year population estimates.
Data extraction: date
February 2009
Numerator: definition
Deaths from all causes, registered in the respective calendar years 1996–98 to 2005–07.
Numerator: source
Office for National Statistics (ONS)
Denominator: definition
2001 Census based mid-year population estimates for respective calendar years 1996 to 2007, people of all ages, current as at 21 August 2008.
Denominator: source
ONS
Data quality: Accuracy and completeness
Coverage can be considered to be complete as the registration of deaths is a legal requirement. Data quality for the relevant fields (age, sex, area of residence) is extremely high. Area of residence is allocated by ONS using the postcode and the National Statistics Postcode Directory - records without a valid area code are excluded but the number of such records is negligible.
Health Profiles 2009 The Indicator Guide
280 Section 7: charts and trend graphs
Table 3 – Indicator Technical Methods Numerator: extraction
Extraction by NCHOD with the exception of new 2009 Unitary Authorities: Cheshire East, Cheshire West and Chester, Central Bedfordshire, and Cornwall. Deaths for these areas were extracted by the LHO from annual mortality files, supplied to PHOs by ONS
Numerator: aggregation/ allocation
Deaths were assigned to geographical areas using the area code supplied in the mortality extract. This is derived from postcode of residence by the ONS using the National Statistics Postcode Directory (NSPD).
Numerator data caveats
Area of residence is allocated by ONS using the postcode and the National Statistics Postcode Directory - records without a valid area code are excluded but the number of such records is negligible.
Denominator data caveats
Data are based on the latest revisions of ONS mid-year population estimates for the respective years, current as at 21 August 2008.
Methods used to calculate indicator value
The directly age-standardised rate (DSR) is the rate of events that would occur in a population with a standard age structure if that population were to experience the age-specific rates of the subject population. The standard population used is the European Standard Population. The rate is expressed per 100,000 population. Data for some new Unitary Authorities, introduced in April 2009, did not correspond with existing geographies (Cheshire East, Cheshire West and Chester, Central Bedfordshire, and Cornwall). Results for these areas were calculated by the LHO. Deaths from 1995 to 2007 were assigned to current boundaries using the November 2008 release of the National Statistics Postcode Directory. Rates were based on pooled deaths and populations for each three year period from 1995–97 to 2005–07. Results for the other new UAs were taken from published data from NCHOD for local authorities (Bedford) or counties (County Durham, Northumberland, Shropshire, and Wiltshire).
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Isles of Scilly and City of London are excluded from the lower tier datasets but included in England, Regional and County figures.
Disclosure Control
None applied.
Confidence Intervals calculation method
Not applicable – Trend chart presented without Confidence Intervals
Health Profiles 2009 The Indicator Guide
281 Section 7: charts and trend graphs
36. T REND 2: EARLY DEATHS FROM HEART DISEASE AND STROKE Basic Information 1. What is being measured?
Trend in early death rates from all circulatory diseases (including heart disease and stroke)
2. Why is it being measured?
Circulatory disease accounts for 40% of all deaths (30% under 75). Mortality is a direct measure of health care need indicating the overall circulatory disease burden on the population, reflecting both the incidence of disease and the ability to treat it.
3. How is this indicator actually defined?
Mortality from all circulatory diseases, directly age-standardised rate, persons, under 75, 1996–98 to 2005–07 (average of annual rates), per 100,000 European Standard population
4. Who does it measure?
People aged under 75
5. When does it measure it?
Updated annually
6. Will It measure absolute numbers or proportions?
Directly age-standardised rate
7. Where does the data actually come from?
Office for National Statistics (ONS)
8. How accurate and complete will the data be?
Mortality counts are derived from an annual mortality extract supplied by ONS and are based on the original underlying cause of death for which there is nearly 100% coverage on the mortality register.
9. Are there any caveats/ warnings/problems?
Area of residence is allocated by ONS using the postcode and the National Statistics Postcode Directory – records without a valid area code are excluded but the number of such records is negligible. There is the potential for the underlying cause of death to be incorrectly attributed on the death certificate and, therefore, the cause of death misclassified.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
Data are age-standardised to allow comparisons between areas. 95% confidence intervals should be considered when making comparisons between areas or over time.
TABLE 1 – INDICATOR DESCRIPTION Information component
Pg 3 Health Inequalities: changes over time
Subject category/ domain(s)
Health inequalities: changes over time
Indicator name (*Indicator title in health profile)
Trend in early mortality rates from all circulatory diseases
Health Profiles 2009 The Indicator Guide
282 Section 7: charts and trend graphs
PHO with lead responsibility
SEPHO
Date of PHO dataset creation
05 February 2009
Indicator definition
Mortality from all circulatory diseases, directly age-standardised rate, persons, under 75, 1996–98 to 2005–07 (average of annual rates), per 100,000 European Standard population
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs (boundaries as at April 2008, new 2009 Unitary Authorities as at April 2009).
Timeliness
The Compendium mortality from all circulatory diseases indicator is updated annually, usually around November/December following the publication by ONS of the new year’s mortality extract (usually in May) and mid-year population estimates (usually August–September).
Rationale: What this indicator purports to measure
Trend in early mortality from all circulatory diseases.
Rationale: Public Health Importance
Circulatory disease accounts for 40% of all deaths (30% under 75). Mortality is a direct measure of health care need reflecting the overall circulatory disease burden on the population, both the incidence of disease and the ability to treat it. The mortality rate may be improved by reducing the population’s risk (e.g. encouraging healthier lifestyles and reducing exposure to smoking), by earlier detection of disease and by more effective treatment. The Our Healthier Nation target is to reduce the number of deaths from circulatory disease in people aged under 75 years by at least 40% by 2010. The baseline for monitoring this target is the three year period 1995–97. The ten year trend charts presented in the health profiles show three year moving averages of annual mortality rates from all circulatory disease from 1996–8 to the most recently available years of data (2005–07 average) compared against the national trend. Earlier trend years are available from previous Health Profiles and on the Health Profiles website.
Rationale: Purpose behind the inclusion of the indicator
To monitor premature mortality due to circulatory diseases over time. To reduce premature deaths from circulatory diseases.
Rationale: Policy relevance
The under 75 circulatory disease mortality rate is a key target indicator in the 1999 Public Health White Paper ‘Saving Lives: Our Healthier Nation’. The target is to reduce the number of deaths from circulatory disease in people aged under 75 years by at least two-fifths by 2010. The baseline for monitoring this target is the three year period 1995–97. This measure supports delivery of the Department of Health PSA targets and LDP and is relevant to Choosing Health, Coronary Heart Disease NSF and Programme for Action.
Health Profiles 2009 The Indicator Guide
283 Section 7: charts and trend graphs
Interpretation: What a high/low level of indicator value means
An upward sloping trend line indicates that the early mortality rate from all circulatory diseases is worsening. A downward sloping trend line indicates that the early mortality rate from all circulatory diseases is improving. If the trend line for the local authority is consistently above the trend line for England then early death rates from circulatory diseases have been consistently higher than those for England for the stated period. If the trend line for the local authority is consistently below the trend line for England then early death rates from circulatory diseases have been consistently lower than those for England for the stated period. If the gap between the local authority trend line and the England trend line is widening then inequalities in early mortality rates from all circulatory diseases are worsening. If the gap between the local authority trend line and the England trend line is narrowing then inequalities in early mortality rates from all circulatory diseases are improving.
Interpretation: Potential for error due to type of measurement method
Coverage can be considered to be complete as the registration of deaths is a legal requirement. Data quality for the relevant fields (age, sex, underlying cause of death, area of residence) is extremely high.
Interpretation: Potential for error due to bias and confounding
The rates are age-standardised. This improves the comparability of rates for different areas, or between different time periods, by taking into account differences in the age structures of the populations being compared.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate.
There is the potential for the underlying cause of death to be incorrectly attributed on the death certificate and, therefore, the cause of death misclassified.
This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health.
Health Profiles 2009 The Indicator Guide
284 Section 7: charts and trend graphs
Table 2 – Indicator Specification Indicator definition: Variable
Mortality from all circulatory diseases (ICD10 I00 –I99, ICD9 390-459 adjusted)
Indicator definition: Statistic
Directly age-standardised rate
Indicator definition: Gender
Persons
Indicator definition: age group
Under 75
Indicator definition: period
1996–98 to 2005–07 (average of annual rates)
Indicator definition: scale
Per 100,000 European Standard population
Geography: geographies available for this indicator from other providers
England & Wales, ONS area, Primary Care Organisation, Strategic Health Authority.
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Age, gender available from NCHOD.
Data extraction: Source
NCHOD.
Data extraction: source URL
Data received directly from NCHOD.
Data extraction: date
14 January 2009
Numerator: definition
Deaths from all circulatory disease, classified by underlying cause of death (ICD10 I00 – I99, ICD9 390-459 adjusted), registered in the respective calendar years 1996–98 to 2005–07, in people aged under 75.
Numerator: source
Office for National Statistics (ONS)
Denominator: definition
2001 Census based mid-year population estimates for respective calendar years 1996 to 2007, people aged under 75, current as at 29 September 2008.
Denominator: source
ONS
Data quality: Accuracy and completeness
Coverage can be considered to be complete as the registration of deaths is a legal requirement. Data quality for the relevant fields (age, sex, underlying cause of death, area of residence) is extremely high. Area of residence is allocated by ONS using the postcode and the National Statistics Postcode Directory – records without a valid area code are excluded but the number of such records is negligible.
Available from National Centre for Health Outcomes Development (NCHOD) website www.nchod.nhs.uk Data can also be found at Neighbourhood Renewal Unit Public Service Agreement Floor Targets (http://www.fti.communities.gov.uk/fti/).
Health Profiles 2009 The Indicator Guide
285 Section 7: charts and trend graphs
Table 3 – Indicator Technical Methods Numerator: extraction
Extraction by NCHOD.
Numerator: aggregation / allocation
Deaths were assigned to geographical areas using the area code supplied in the mortality extract. This is derived from postcode of residence by the ONS using the National Statistics Postcode Directory (NSPD).
Numerator data caveats
Area of residence is allocated by ONS using the postcode and the National Statistics Postcode Directory – records without a valid area code are excluded but the number of such records is negligible. Mortality counts are derived from the annual DH mortality extract supplied by ONS and are based on the original underlying cause of death for which there is nearly 100% coverage on the mortality register. In January 2001, the ONS implemented a change from ICD-9 to ICD-10 for coding causes of death in England & Wales. As part of an exercise to investigate the effects of this change, the ONS also re-coded all deaths registered in 1999. Deaths for years prior to 1999 and for year 2000 have not been re-coded. The numbers of deaths observed in the years 1996-98 and 2000 have, therefore, been adjusted to give “expected” numbers of deaths which would have been coded to this cause in ICD-10. This was done by multiplying the ICD-9 based death counts by the appropriate ICD-10/9 comparability ratio published by the ONS. For this indicator the following ICD-10/9 comparability ratios were used: Males aged 0-74: 1.012; Females aged 0–74: 1.015. Adjusted person counts are the sum of the adjusted male and female counts
Denominator data caveats
Data are based on the latest revisions of ONS mid-year population estimates for the respective years, current as at 29 September 2008.
Methods used to calculate indicator value
The directly age-standardised rate (DSR) is the rate of events that would occur in a population with a standard age structure if that population were to experience the age-specific rates of the subject population. The standard population used is the European Standard Population. The age groups used are: Under 1, 1–4, 5–9,…, 80–84, 85+. The rates for each time period e.g. 2005-07 have been calculated as the simple average of the individual annual rates. The rate is expressed per 100,000 population.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Isles of Scilly and City of London are excluded from the lower tier datasets but included in England and Regional figures.
Disclosure Control
None applied.
Confidence Intervals calculation method
Not applicable – Trend chart presented without Confidence Intervals
Health Profiles 2009 The Indicator Guide
286 Section 7: charts and trend graphs
37. TREND 3: EARLY DEATH RATES FROM CANCER Basic Information 1. What is being measured?
Trend in early death rates from cancer
2. Why is it being measured?
Early mortality from cancer is a direct measure of health care need as public health interventions for prevention, early diagnosis, and effective treatment can all reduce the burden of cancer morbidity and mortality.
3. How is this indicator actually defined?
Mortality from all cancers, directly age-standardised rate, persons, under 75, 1996–98 to 2005–07 (average of annual rates), per 100,000 European Standard population
4. Who does it measure?
People aged under 75
5. When does it measure it?
Updated annually
6. Will It measure absolute numbers or proportions?
Directly age-standardised rate
7. Where does the data actually come from?
Office for National Statistics (ONS)
8. How accurate and complete will the data be?
Mortality counts are derived from an annual mortality extract supplied by ONS and are based on the original underlying cause of death for which there is nearly 100% coverage on the mortality register.
9. Are there any caveats/warnings/ problems?
Area of residence is allocated by ONS using the postcode and the National Statistics Postcode Directory – records without a valid area code are excluded but the number of such records is negligible. There is the potential for the underlying cause of death to be incorrectly attributed on the death certificate and, therefore, the cause of death misclassified.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
Data are age-standardised to allow comparisons between areas. 95% confidence intervals should be considered when making comparisons between areas or over time.
Health Profiles 2009 The Indicator Guide
287 Section 7: charts and trend graphs
Table 1 – Indicator Description Information component
County Health Profiles: Pg 3 Health Summary – Indicator 34
Subject category/ domain(s)
Health inequalities: changes over time
Indicator name (*Indicator title in health profile)
Trend in early death rates from cancer
PHO with lead responsibility
SEPHO
Date of PHO dataset creation
05 February 2009
Indicator definition
Mortality from all cancers, directly age-standardised rate, persons, under 75, 1996-98 to 2005-07 (average of annual rates), per 100,000 European Standard population
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs (boundaries as at April 2008, new 2009 Unitary Authorities as at April 2009).
Timeliness
Updated annually.
Rationale: What this indicator purports to measure
Trend in early mortality from all cancers.
Rationale: Public Health Importance
Cancer is amongst the three leading causes of death at all ages except for pre-school age children in the UK. It accounts for 26% all deaths. If current incidence rates remain the same, by 2025 there will be an additional 100,000 cases of cancer diagnosed each year as a result of the ageing population. Inequalities exist in cancer rates between the most deprived areas and the most affluent. Early mortality from cancer is a direct measure of health care need as public health interventions for prevention, early diagnosis, effective treatment can all reduce the burden of cancer morbidity and mortality. The Our Healthier Nation target is to reduce the number of deaths from cancer in people aged under 75 years by at least a fifth by 2010. The baseline for monitoring this target is the three year period 1995–97. The ten year trend charts presented in the health profiles show the three year moving averages of annual mortality rates from all cancer from 1996–8 to the most recently available years of data (2005–07 average) compared against the national trend. Earlier trend years are available from previous Health Profiles and on the Health Profiles website.
Rationale: Purpose behind the inclusion of the indicator
To monitor premature mortality due to cancer over time. To reduce premature deaths from cancer.
Health Profiles 2009 The Indicator Guide
288 Section 7: charts and trend graphs
Rationale: Policy relevance
The directly age-standardised mortality rate from all cancers for persons aged under 75 is a target indicator in the Saving Lives: Our Healthier Nation strategy. The target is a 20% reduction by the year 2010 from the baseline rate in 1995–97. This measure supports delivery of the Department of Health PSA targets and LDP and is relevant to Choosing Health, Cancer NSF and Programme for Action.
Interpretation: What a high/low level of indicator value means
An upward sloping trend line indicates that the early mortality rate from all cancers is worsening. A downward sloping trend line indicates that the early mortality rate from all cancers is improving. If the trend line for the local authority is consistently above the trend line for England then early death rates from cancer have been consistently higher than those for England for the stated period. If the trend line for the local authority is consistently below the trend line for England then early death rates from cancer have been consistently lower than those for England for the stated period. If the gap between the local authority trend line and the England trend line is widening then inequalities in early mortality rates from all cancers are worsening. If the gap between the local authority trend line and the England trend line is narrowing then inequalities in early mortality rates from all cancers are improving.
Interpretation: Potential for error due to type of measurement method
Coverage can be considered to be complete as the registration of deaths is a legal requirement. Data quality for the relevant fields (age, sex, underlying cause of death, area of residence) is extremely high.
Interpretation: Potential for error due to bias and confounding
The rates are age-standardised. This improves the comparability of rates for different areas, or between different time periods, by taking into account differences in the age structures of the populations being compared.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate.
There is the potential for the underlying cause of death to be incorrectly attributed on the death certificate and, therefore, the cause of death misclassified.
This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health.
Health Profiles 2009 The Indicator Guide
289 Section 7: charts and trend graphs
Table 2 – Indicator Specification Indicator definition: Variable
Mortality from all cancers (ICD10 C00-C97, ICD9 140-208 adjusted)
Indicator definition: Statistic
Directly age-standardised rate
Indicator definition: Gender
Persons
Indicator definition: age group
Under 75
Indicator definition: period
1996-98 to 2005-07 (average of annual rates)
Indicator definition: scale
Per 100,000 European Standard population
Geography: geographies available for this indicator from other providers
England & Wales, ONS area, Primary Care Organisation, Strategic Health Authority.
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Age, gender available from NCHOD.
Data extraction: Source
NCHOD.
Data extraction: source URL
Data received directly from NCHOD.
Data extraction: date
14 January 2009
Numerator: definition
Deaths from all malignant neoplasms, classified by underlying cause of death (ICD10 C00-C97, ICD9 140-208 adjusted), registered in the respective calendar years 1996–98 to 2005–07, in people aged under 75.
Numerator: source
Office for National Statistics (ONS)
Denominator: definition
2001 Census based mid-year population estimates for respective calendar years 1996 to 2007, people aged under 75, current as at 29 September, 2008.
Denominator: source
ONS
Data quality: Accuracy and completeness
Coverage can be considered to be complete as the registration of deaths is a legal requirement. Data quality for the relevant fields (age, sex, underlying cause of death, area of residence) is extremely high. Area of residence is allocated by ONS using the postcode and the National Statistics Postcode Directory – records without a valid area code are excluded but the number of such records is negligible.
Available from National Centre for Health Outcomes Development (NCHOD) website www.nchod.nhs.uk Data can also be found at Neighbourhood Renewal Unit Public Service Agreement Floor Targets (http://www.fti.communities.gov.uk/fti/).
Figures for various NRF and NDC and Spearhead Groups are available via the Neighbourhood Renewal Unit http://www.neighbourhood.gov.uk/
Health Profiles 2009 The Indicator Guide
290 Section 7: charts and trend graphs
Table 3 – Indicator Technical Methods Numerator: extraction
Extraction by NCHOD.
Numerator: aggregation / allocation
Deaths were assigned to geographical areas using the area code supplied in the mortality extract. This is derived from postcode of residence by the ONS using the National Statistics Postcode Directory (NSPD).
Numerator data caveats
Area of residence is allocated by ONS using the postcode and the National Statistics Postcode Directory -– records without a valid area code are excluded but the number of such records is negligible. Mortality counts are derived from the annual DH mortality extract supplied by ONS and are based on the original underlying cause of death for which there is nearly 100% coverage on the mortality register. In January 2001, the ONS implemented a change from ICD-9 to ICD-10 for coding causes of death in England & Wales. As part of an exercise to investigate the effects of this change, the ONS also re-coded all deaths registered in 1999. Deaths for years prior to 1999 and for year 2000 have not been re-coded. The numbers of deaths observed in the years 1996–98 and 2000 have, therefore, been adjusted to give “expected” numbers of deaths which would have been coded to this cause in ICD-10. This was done by multiplying the ICD-9 based death counts by the appropriate ICD-10/9 comparability ratio published by the ONS. For this indicator the following ICD-10/9 comparability ratios were used: Males aged 0–74: 1.013; Females aged 0–74: 1.009. Adjusted person counts are the sum of the adjusted male and female counts.
Denominator data caveats
Data are based on the latest revisions of ONS mid-year population estimates for the respective years, current as at 29 September, 2008.
Methods used to calculate indicator value
The directly age-standardised rate (DSR) is the rate of events that would occur in a population with a standard age structure if that population were to experience the age-specific rates of the subject population. The standard population used is the European Standard Population. The age groups used are: Under 1, 1–4, 5–9,…, 80–84, 85+. The rates for each time period e.g. 2005-2007 have been calculated as the simple average of the individual annual rates. The rate is expressed per 100,000 population.
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Isles of Scilly and City of London are excluded from the lower tier datasets but included in England and Regional figures.
Disclosure Control
None applied.
Confidence Intervals calculation method
Not applicable – Trend chart presented without Confidence Intervals
Health Profiles 2009 The Indicator Guide
291 Section 7: charts and trend graphs
38. HEALTH INEQUALITIES: ETHNICITY CHART Basic Information 1. What is being measured?
The percentage of children in each ethnic group who are eligible for free school meals
2. Why is it being measured?
Eligibility for free school meals is an indicator of deprivation and people who suffer more deprivation tend to have poorer health.
3. How is this indicator actually defined?
The percentage of children by each ethnic group who are eligible for free school meals.
4. Who does it measure?
School age children aged 5–15
5. When does it measure it?
January 2008
6. Will It measure absolute numbers or proportions?
Both proportions and absolute numbers are presented.
7. Where does the data actually come from?
Department for Children, Schools and Families
8. How accurate and complete will the data be?
This indicator only contains data for LEA maintained Primary, Secondary and Special schools.
9. Are there any caveats/warnings/ problems?
Excludes pupils educated in private schools. Numbers presented are rounded to the nearest 10.
10. Are particular tests needed such as standardisation, significance tests, or statistical process control to test the meaning of the data and the variation they show?
95% confidence intervals should be considered when making comparisons between areas or ethnic groups.
Table 1 – Indicator Description Information component
Page 3 - Health Inequalities: ethnicity
Subject category/ domain(s)
Health inequalities
Indicator name (* Indicator title in health profile)
Ethnicity
PHO with lead responsibility
EMPHO
Date of PHO dataset creation
March 2009
Indicator definition
Percentage of children in maintained schools eligible for free school meals by ethnic group.
Geography
England, GOR, Local Authority: Counties, County Districts, Metropolitan County Districts, Unitary Authorities, London Boroughs
Health Profiles 2009 The Indicator Guide
292 Section 7: charts and trend graphs
Timeliness
Every school is required to supply information to the DCSF each January through the Pupil Level Annual Schools Census (PLASC).
Rationale: What this indicator purports to measure
Percentage of children in low socio-economic groups as measured by those eligible for free school meals.
Rationale: Public Health Importance
Ethnic inequalities in health are well known1, but are difficult to monitor and address due to limited data2. Eligibility for free school meals (FSM) has frequently been used as a proxy measure of poverty, disadvantage, and social exclusion as information on adult’s income is not collected. Eligibility for Free School Meals is strongly associated with low educational achievement. 3 Aspinall P, Jacobson B. Ethnic disparities in health and health care: A focused review of the evidence and selected examples of good practice. London: LHO, 2004. Aspinall P, Jacobson B, Polato GM. Missing record: The case for recording ethnicity at birth and death registration. London: LHO, 2003. Tackling low educational achievement, Robert Cassen and Geeta Kingdon, 2007 www.jrf.org.uk/bookshop/eBooks/2063-education-schools-achievement. pdf
Rationale: Purpose behind the inclusion of the indicator
This indicator can help to indicate within a local area which ethnic groups are likely to experience poorer health than others and could be used to predict health care need and target resources.
Rationale: Policy relevance
There are four national indicators which relate to the educational achievement of children eligible for free school meals. National indicator no.81 ‘Inequality gap in the achievement of a Level 3 qualification by the age of 19’. This indicator reports the gap in attainment of level 3 at age 19 in each Local Authority between those young people who were in receipt of free school meals at academic age 15 and those who were not. At the moment significantly fewer young people in receipt of free school meals at the age of 15 achieve L3 qualifications by the age of 19 than their peers who were not in receipt of free school meals at age 15. National indicator 82 ‘Inequality gap in the achievement of a Level 2 qualification by the age of 19’. This indicator reports the percentages of young people who were in receipt of free school meals at academic age 15 who attain level 2 qualifications by the age of 19. At the moment significantly fewer young people in receipt of free school meals at the academic age of 15 achieve L2 qualifications by the age of 19 than their peers who were not in receipt of free school meals at academic age 15. National indicator no. 102 ‘Achievement gap between pupils eligible for free school meals and their peers achieving the expected level at Key Stages 2 and 4’. This indicator seeks to narrow the gap in achievement between children from disadvantaged backgrounds and their peers. National Indicator 106: Young people from low income backgrounds progressing to higher education. The indicator takes the form of the gap between the proportions of 15 year olds eligible for free school meals (FSM) and those not eligible for FSM progressing to higher education at the age of 18 or 19.
Health Profiles 2009 The Indicator Guide
293 Section 7: charts and trend graphs
Interpretation: What a high/low level of indicator value means
A high percentage indicates an ethnic group with a high risk of poor educational achievement. The numbers provided allow local authorities to estimate the burden of this risk in their populations. Ethnic groups with a higher percentage than the England average, which is shown on the chart for each ethnic group, are at increased risk of poor educational achievement when compared to the England average. Ethnic groups with a lower percentage than the England average, which is shown on the chart for each ethnic group, are at decreased risk of poor educational achievement when compared to the England average.
Interpretation: Potential for error due to type of measurement method
The subjective, multi-faceted and changing nature of ethnic identification makes it a particularly difficult piece of information to collect. There is no consensus on what constitutes an ‘ethnic group’. Membership of any ethnic group is something that is subjectively meaningful to the person concerned and the terminology used to describe ethnic group has changed markedly over time. As a result, ethnic group, however defined or measured, will tend to change over time depending on social and political attitudes or developments. Basing ethnic identification upon an objective and rigid classification of ethnic groups is not therefore achievable in practice. In particular research has shown that people with mixed backgrounds may vary their answer to questions about ethnicity to suit their perception of the form they are completing.
Interpretation: Potential for error due to bias and confounding
The statistic measures eligibility for free school meals attainment in LEA maintained schools only.
Confidence Intervals: Definition and purpose
A confidence interval is a range of values that is normally used to describe the uncertainty around a point estimate of a quantity, for example, a mortality rate. This uncertainty arises as factors influencing the indicator are subject to chance occurrences that are inherent in the world around us. These occurrences result in random fluctuations in the indicator value between different areas and time periods. In the case of indicators based on a sample of the population, uncertainty also arises from random differences between the sample and the population itself. The stated value should therefore be considered as only an estimate of the true or ‘underlying’ value. Confidence intervals quantify the uncertainty in this estimate and, generally speaking, describe how much different the point estimate could have been if the underlying conditions stayed the same, but chance had led to a different set of data. The wider is the confidence interval the greater is the uncertainty in the estimate. Confidence intervals are given with a stated probability level. In Health Profiles 2009 this is 95%, and so we say that there is a 95% probability that the interval covers the true value. The use of 95% is arbitrary but is conventional practice in medicine and public health. The confidence intervals have also been used to make comparisons against the national value. For this purpose the national value has been treated as an exact reference value rather than as an estimate and, under these conditions, the interval can be used to test whether the value is statistically significantly different to the national. If the interval includes the national value, the difference is not statistically significant. If the interval does not include the national value, the difference is statistically significant.
Health Profiles 2009 The Indicator Guide
294 Section 7: charts and trend graphs
Table 2 – Indicator Specification Indicator definition: Variable
For each ethnic group the indicator is defined as the number of pupils who are eligible for free school meals (numerator), divided by all pupils (denominator), expressed as a percentage.
Indicator definition: Statistic
Percentage by ethnic group
Indicator definition: Gender
Persons
Indicator definition: age group
Aged 5–15
Indicator definition: period
January 2008
Indicator definition: scale
Percentage and number
Geography: geographies available for this indicator from other providers
Number and percent of pupils eligible for free school meals is available for England, Regional and LEA levels on the DFES website. http://www.dcsf.gov. uk/rsgateway/DB/SFR/s000786/index.shtml
Dimensions of inequality: subgroup analyses of this dataset available from other providers
Number and percent of pupils eligible for free school meals is available for by primary, secondary and special school for England, Regional and LEA levels on the DFES website. http://www.dcsf.gov.uk/rsgateway/DB/SFR/s000786/index. shtml
Data extraction: Source
Pupil level annual schools census (PLASC).
Data extraction: source URL
Received directly from Department for Children, Schools and Families
Data extraction: date
Final dataset received March 2009.
Numerator: definition
Eligibility to receive free school meals is dependant on a child’s parent(s) or carer(s) receiving: Income Support (IS); Income Based Jobseekers Allowance (IBJSA); Support under part VI of the Immigration and Asylum Act 1999; Child Tax Credit, provided they are not entitled to working Tax Credit and have an annual income, as assessed by the Inland Revenue, that (for 2007/2008) does not exceed £14,495 and £15,575 (for 2008/2009); the guaranteed element of State Pension Credit. Children who receive IS or IBJSA in their own right are also entitled to free school meals.
Numerator: source
Pupil level annual schools census (PLASC).
Number and percent of pupils by ethnic group is available for England, Regional and LEA levels on the DFES website. http://www.dcsf.gov.uk/ rsgateway/DB/SFR/s000786/index.shtml
Number and percent of pupils by ethnic group by primary and secondary schools is available for England, Regional and LEA levels on the DFES website. http://www.dcsf.gov.uk/rsgateway/DB/SFR/s000786/index.shtml
Health Profiles 2009 The Indicator Guide
295 Section 7: charts and trend graphs
Denominator: definition
Data for five ethnic groups have been presented. Data is collected for 18 groups and have been combined as follows. • White Ethnic group includes those classified as: White British, Irish, Traveller of Irish Heritage, Gypsy/Roma, Any other White background. • Mixed Ethnic group includes those classified as: White and Black Caribbean, White and Black African, White and Asian, Any other Mixed background. • Asian Ethnic group includes those classified as: Indian, Pakistani, Bangladeshi, Any other Asian background. • Black Ethnic group includes those classified as: Black Caribbean, Black African, Any other Black background. • Chinese and any other ethnic group includes: Chinese and any other ethnic group.
Denominator: source
Pupil level annual schools census (PLASC).
Data quality: Accuracy and completeness
Every school is required to supply information to the DCSF each January through the Pupil Level Annual Schools Census (PLASC). The Census collects school level data about classes being taught and staff in schools and also pupil level information from Maintained Primary, Middle, Secondary and Specials schools. The data items collected in PLASC include a unique pupil number, pupil name, gender, date of birth, home postcode, ethnic group, first language, free school meal eligibility and special educational needs information. Approximately 1.5% of pupils were unclassified by ethnic group that is the information was either refused on not obtained.
TABLE 3 – INDICATOR TECHNICAL METHODS Numerator: extraction
Received directly from Department for Children, Schools and Families
Health Profiles 2009 The Indicator Guide
296 Section 7: charts and trend graphs
Numerator: aggregation / allocation
County, Regional and England numerators were aggregated from their constituent Districts, UAs, MCDs and London Boroughs. Since the 2008 PLASC, there have been new Unitary Authorities created. The new Unitary Authorities of Northumberland, Shropshire, Wiltshire and Durham have the same boundaries as the former counties of the same name so have been allocated these data. Data for five other new unitary Authorities have been calculated from the aggregation of previous Local Authorities. New UA name
Old LA name
Bedford
Bedford
Central Bedfordshire
Mid Bedfordshire South Bedfordshire
Cheshire East
Congleton Crewe and Nantwich Macclesfield
Cheshire West and Chester
Chester Ellesmere Port & Neston Vale Royal
Cornwall
Caradon Carrick Kerrier North Cornwall Penwith Restormel
Numerator data caveats
Where the numerator is less than 5 no percentage is shown. (Data supplied by DCSF was rounded to the nearest 10, so this applies to values received that were rounded to 0.)
Denominator data caveats
Where the ethnic population is less than 30 the data has not been presented. (Data supplied by DCSF was rounded to the nearest 10, so this applies to values received that were rounded to 0, 10 or 20.)
Methods used to calculate indicator value
The indicator is expressed as the percentage of children (aged 5–15) eligible for free school meals by ethnic groups. A percentage is defined as the number of observed events divided by the total population of interest multiplied by 100. A percentage, p, is given by:
p=
O × 100 n
where: O is the number of observed events (i.e. number of children eligible for free school meals for each ethnic group); n is the size of the population of interest (i.e. the number of children in ethnic group aged 5-15)
Health Profiles 2009 The Indicator Guide
297 Section 7: charts and trend graphs
Small Populations: How Isles of Scilly and City of London populations have been dealt with
Data for the Isles of Scilly and City of London have not been presented. The Isles of Scilly are included in the South West regional and national totals and the City of London figures have been included in the regional London and national totals.
Disclosure Control
Data was supplied by the department for Children, Schools and Families for each of the total groups (White, Mixed, Asian, Black and Chinese and any other ethnic group) rounded to the nearest ten. Thus, for all areas, the numbers presented are rounded to the nearest ten.
Confidence Intervals calculation method
Confidence intervals have been calculated using the following method for a confidence interval of a proportion as described by Newcombe RG If r is the observed number of subjects with some feature in a sample of size n then the estimated proportion who have the feature is p = r/n. The proportion who do not have the feature is q = 1-p. First, calculate the three quantities A = 2r + z2;
B = z z 2 + 4rq ;
and
C=2(n+z2),
where z is z1-α/2, from the standard Normal distribution. Then the confidence interval for the population proportion is given by (A-B)/C
to
(A+B)/C
This method has the considerable advantage that it can be used for any data. When there are no observed events, r and hence p are both zero, and the recommended confidence interval simplifies to 0 to z2/(n+z2). When r = n so that p = 1, the interval becomes n/(n+z2) to 1. Reference Newcombe, RG. Two-sided confidence intervals for the single proprotion: comparison of seven methods. Stat Med 1998;17:857-72.
Related Documents
The Indicator Guide Health Profiles 2009
June 2020 1
1. Indicator Health Status.docx
June 2020 17
The Union's Health And Fitness Guide 2009
April 2020 0
Health Guide
August 2019 35
Health Beat User Guide 2009
December 2019 10