Impact Of Change Forecaster Documentation
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 Impact Of Change Forecaster Documentation as PDF for free.
More details
- Words: 1,709
- Pages: 17
Impact of Change® Forecaster Product Documentation
SG-2 Forecasting Philosophy 1. Comprehensive (forecasts should incorporate a broad range of signficant drivers of utilization; e.g. a forecast that only focuses on one driver cannot reliably forecast total utilization) 2. Timely (forecasts should incorporate the latest information available about emerging technologies, population trends, etc. 3. Locally specific (forecasts should reflect local conditions as the practice of health care is highly regional) 4. Precise (Precision in the input data and empirical timings and impact factors is critical)
The Impact of Change™
About the Impact of Change® Forecaster The Impact of Change® Forecaster quantifies the impact of changes in the business and technology of health care on utilization of health care services. In addition to population changes, the IoC forecasts utilization based on technology, economic and sociocultural drivers as well.
Technology
Inpatient Discharges
Impact of Change® Database
Inpatient Days
Economy
Socio-cultural
2002 - 2010
Outpatient Services
Specific Examples of Factors
Technology Minimally Invasive Surgery Imaging Immune Modulation Targeted Drug Therapies Genetic Engineering Implantables Biologicals Medical Informatics
Economy
Socio-Cultural
Insurance Coverage Unemployment Total Employment Consumer Confidence Gross Domestic Product Cost Inflation
Obesity Smoking Physical Activity Lifestyle
The Impact of Change™
Data Sources Demographics Utilization Inpatient Outpatient
Mathematical Approach The most powerful approach to modelling changes over time, given initial conditions (e.g. initial population, use-rates/volumes), is to use differential equations. The time variable, however, is typically discrete (e.g. one-year intervals) so it is extremely common on economic and social sciences modelling to use what are termed difference equations.1 This is the basic approach used with the IoC. Essentially, one takes the initial conditions, along with factors (developing over time) that affect the initial conditions and generates a sequential evolution over time of utilization. Some details on how these drivers are constructed is described in the pages following.
1
Goldberg, S. "Introduction to Difference Equations," Dover Press, 1986. The Impact of Change™
Technology Impact
SG-2 Technical Brief
Assumptions 1. Most models of technology adoption involve a "logistic" or S-shaped diffusion & adoption curve. 2. "National" adoption curves are the "sum" of many individual regional/institutional curves and are hence "spread" out. By definition, then, the local adoption metrics will be different than the national one.
Approach: 1. Timing: Assign the following variables for each "technology item" a. ts = Start Year b. ti = Inflection Year c. td = Decline Year d. r = rate of growth e. Item scale factor 2. Mapping: Map all technologies to utilization variables 3. Impact Factors: Assign an impact factor (IF) for each mapping (most are zero). 4. Calculation parameters: Most importantly the probability scale factor.
Timing:
The Impact of Change™
The equation:
Sgn( ( tb + t ) − ts ) + 1 − aPs (( Sgn( Sgn(ts − tb ))+ 1)(ts − tb ) ) T (t ) = • e 2
(
2
) (
1 • − r ( ( tb + t ) − ti ) 2 1+ e
)
[ Sgn( td ) • Sgn( Sgn( ( tb + t ) − td ) + 1) ] • − r ( t2b − th ) •e
The curve: Timing Progression Chart 1.2
Fractional Impact
1.
.8
.6
.4
.2
. 2003
2004
2005
2006
2007
2008
2009
2010
2011
Year
The Impact of Change™
The curve (another example): Timing Progression Chart .7
.6
Fractional Impact
.5
.4
.3
.2
.1
. 2001
2002
2003
2004
2005
2006
2007
2008
2009
Year
The Impact of Change™
Impact Factors:
" Based on titrating to additional 800 procedures in 2002 (est.) and approximately 4000 on "waiting-list" with given national timings 0.06 IF was determined to be optimal"
The result: Yearly Discharge Growth Chart 45,000 40,000
Cumulative Percent Change
35,000 30,000 25,000 20,000 15,000 10,000 5,000 0 2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
Year
The Impact of Change™
Economic Effect
SG-2 Technical Brief
Assumptions 3. Studies show a 25% reduction in utilization (inpatient) with an increase in unemployment. Effect is indirect via loss of employment determined health insurance benefits. 4. The effect is probably "across-the-board" – e.g. even non-discretionary utilization is affected. The only exception is probably acute, trauma-related care. 5. By corollary, does utilization increase by 25% if unemployment declines? Probably by not as much.
Approach: 5. Get change in unemployment forecasts from Bureau of Labor Statistics and SG-2 analysis. Need. This is δ. Increasing unemployment is a positive number; while decreasing unemployment is a negative number. 6. The "utilization depressor" is µ. Usually set to -0.25 7. The "negative bias factor" is κ. Default is 25. Using this in an exponential accentuates the change on the positive (or increasing unemployment) side.
∆ = δ • µ • e (κ • δ ) Current Defaults:
The Impact of Change™
δ: Our current national set looks as follows:
Eastern Mass, SSM/Cent. Missouri and ANOVA were all run with these national numbers. Unemployment Curve Chart
"Change in unemployment" Impact
.02
.015
.01
.005
.
-.005
-.01 2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
Year
The Impact of Change™
∆: Applying the above equation (with defaults for κ and µ) to the national δ's, we get the following results:
∆
δ
Year 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
No kappa .0042 .0167 -.002 -.004 -.002 0 0 -.004 -.001 -.001
(with Kappa
-0.00105 -0.00417 0.0005 0.001 0.0005 0 0 0.001 0.00025 0.00025
-0.00117 -0.00634 0.000476 0.000905 0.000476 0 0 0.000905 0.000244 0.000244
"Delta" Chart ∆ 0.02
Pure
0.015
Kappa No Kappa 0.01
0.005
0
-0.005
-0.01 2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
Year
These ∆ 's are the "changes in rate" that will be applied to the utilization volumes (non-Medicare component – see below).
The Impact of Change™
Implementation: These "rates" are then integrated and applied to the non-Medicare component (for DRGs) or total (for OPCs) utilization volumes. Currently the Medicare fractions are "hard-coded" (non time & space dependent) into the core DRG table. Example of DRG Medicare fractions: DRG
DRG Description
Medicare Fraction
1.00 Craniotomy >17 X Trauma
0.3429158
1.10 Craniotomy >17 X Trauma
0.3429158
2.00 Craniotomy For Trauma >17
0.5328467
2.10 Craniotomy For Trauma >17
0.5328467
3.00 Craniotomy Age 0-17
0
3.10 Craniotomy Age 0-17
0
4.00 Spinal Proc
0.1917476
4.10 Spinal Proc
0.1917476
The rates can be applied to: 1. initial utilization volumes or 2. "nonlinearly" multiplied against population-driven utilization.
Treatment of Medicare Fractions There are two sources for medicare fraction: a. "Hard-coded"in the DRG tables directly or b. derived directly from the age-group identifier of the relevant dataset. E.g. If the age group is greater than 65 then the medicare fraction is 1; other wise it is zero. If the age grouping spans 65 then the medicare fraction is some number between 0 and 1 – appropriately weighted at the time that the age group designation is defined. The default is to use option (2); if option (2) does not yield a medicare fraction (e.g. for undefined or global age group designations) then option (1) will be applied. Option (2) also allows for the medicare fraction to change over time as the proportion of the population over 65 likewise changes.
The Impact of Change™
Results: National data; linear run. Using κ, m defaults Economic Population
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
157,348
157,170
156,203
156,244
156,383
156,460
156,462
156,462
156,599
156,640
156,679
157,348
161,549
166,054
170,891
176,088
181,679
187,701
194,194
201,204
208,782
216,987
-0.4% 37.9%
Population / Economic "Wedge" NHDS National 2000 - 2010 SG-2 Forecast Economic Population
250 200 150 100 50
0 2000 2001 2002 2003 2004 2005 2006 2007
2008
2009
2010
Economic Effect Detail 157,600 157,400 157,200 157,000 156,800 156,600 156,400 156,200 156,000 155,800 155,600 2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
The Impact of Change™
Consumer Effect
SG-2 Technical Brief
Assumptions 1. The consumer effect is closely related to the economic effect. 2. Applies to discretionary DRGs/OPCs only 3. Grows over the decade by an SG-2 determined curve
Approach: 1. Convolute D's from economic analysis with the SG-2 Consumerism curve. 2. These are the new rates 3. Apply only to discretionary DRGs (non-Medicare fraction) or discretionary OPCs (total fraction)
Consumerism Curve Chart 12.
Fractional Impact
10.
8.
6.
4.
2.
. 2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
Year
DRG
DRG Description
Discretionary DRG?
1.00 Craniotomy >17 X Trauma
N
1.10 Craniotomy >17 X Trauma
N
2.00 Craniotomy For Trauma >17 N 2.10 Craniotomy For Trauma >17 N 3.00 Craniotomy Age 0-17
N
3.10 Craniotomy Age 0-17
N
4.00 Spinal Proc
N
4.10 Spinal Proc
N
5.00 Extracranial Vascular Procs N 5.10 Extracranial Vascular Procs N 6.00 Carpal Tunnel Release
Y The Impact of Change™
DRG
DRG Description
6.10 Carpal Tunnel Release
Discretionary DRG? Y
The Impact of Change™
Results:
Population / Consumerism "Wedge" NHDS National 2000 - 2010 SG-2 Forecast
250 Consumerism Population
200
150
100
50
0 2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
Consumerism Effect Detail 157,500 157,400 157,300 157,200 157,100 157,000 156,900 2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
The Impact of Change™
Sociocultural Impacts
SG-2 Technical Brief
Assumptions 1. Many effects act roughly progressively over the forecast period (e.g. without a logistic curve, etc.). 2. These "meta-trends" often involve socio-cultural shifts within a population that are similar to demographic changes but involve factors beyond simply population change. Examples, include "obesity", increasing reliance of ERs for primary care, increasing acceptability of cosmetic surgery, etc. 3.
Approach: 1. Identify the sociocultural factor (potentially locally specific) 2. Identify which utilization parameters (DRG, OPC, etc.) are affected 3. Determine from the literature and/or "micro-model" an estimation of th percent yearly change in utilization attributable to that sociocultural factor. This number is termed the impact factor (IF).
Example sociocultural factors:
The Impact of Change™
Example Result: Effect of obesity on DRG 127 Heart Failure & Shock Yearly Discharge Growth Chart 1,050,000 1,040,000
Cumulative Percent Change
1,030,000 1,020,000 1,010,000 1,000,000 990,000 980,000 970,000 960,000 2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
Year
The Impact of Change™
SG-2 Forecasting Philosophy 1. Comprehensive (forecasts should incorporate a broad range of signficant drivers of utilization; e.g. a forecast that only focuses on one driver cannot reliably forecast total utilization) 2. Timely (forecasts should incorporate the latest information available about emerging technologies, population trends, etc. 3. Locally specific (forecasts should reflect local conditions as the practice of health care is highly regional) 4. Precise (Precision in the input data and empirical timings and impact factors is critical)
The Impact of Change™
About the Impact of Change® Forecaster The Impact of Change® Forecaster quantifies the impact of changes in the business and technology of health care on utilization of health care services. In addition to population changes, the IoC forecasts utilization based on technology, economic and sociocultural drivers as well.
Technology
Inpatient Discharges
Impact of Change® Database
Inpatient Days
Economy
Socio-cultural
2002 - 2010
Outpatient Services
Specific Examples of Factors
Technology Minimally Invasive Surgery Imaging Immune Modulation Targeted Drug Therapies Genetic Engineering Implantables Biologicals Medical Informatics
Economy
Socio-Cultural
Insurance Coverage Unemployment Total Employment Consumer Confidence Gross Domestic Product Cost Inflation
Obesity Smoking Physical Activity Lifestyle
The Impact of Change™
Data Sources Demographics Utilization Inpatient Outpatient
Mathematical Approach The most powerful approach to modelling changes over time, given initial conditions (e.g. initial population, use-rates/volumes), is to use differential equations. The time variable, however, is typically discrete (e.g. one-year intervals) so it is extremely common on economic and social sciences modelling to use what are termed difference equations.1 This is the basic approach used with the IoC. Essentially, one takes the initial conditions, along with factors (developing over time) that affect the initial conditions and generates a sequential evolution over time of utilization. Some details on how these drivers are constructed is described in the pages following.
1
Goldberg, S. "Introduction to Difference Equations," Dover Press, 1986. The Impact of Change™
Technology Impact
SG-2 Technical Brief
Assumptions 1. Most models of technology adoption involve a "logistic" or S-shaped diffusion & adoption curve. 2. "National" adoption curves are the "sum" of many individual regional/institutional curves and are hence "spread" out. By definition, then, the local adoption metrics will be different than the national one.
Approach: 1. Timing: Assign the following variables for each "technology item" a. ts = Start Year b. ti = Inflection Year c. td = Decline Year d. r = rate of growth e. Item scale factor 2. Mapping: Map all technologies to utilization variables 3. Impact Factors: Assign an impact factor (IF) for each mapping (most are zero). 4. Calculation parameters: Most importantly the probability scale factor.
Timing:
The Impact of Change™
The equation:
Sgn( ( tb + t ) − ts ) + 1 − aPs (( Sgn( Sgn(ts − tb ))+ 1)(ts − tb ) ) T (t ) = • e 2
(
2
) (
1 • − r ( ( tb + t ) − ti ) 2 1+ e
)
[ Sgn( td ) • Sgn( Sgn( ( tb + t ) − td ) + 1) ] • − r ( t2b − th ) •e
The curve: Timing Progression Chart 1.2
Fractional Impact
1.
.8
.6
.4
.2
. 2003
2004
2005
2006
2007
2008
2009
2010
2011
Year
The Impact of Change™
The curve (another example): Timing Progression Chart .7
.6
Fractional Impact
.5
.4
.3
.2
.1
. 2001
2002
2003
2004
2005
2006
2007
2008
2009
Year
The Impact of Change™
Impact Factors:
" Based on titrating to additional 800 procedures in 2002 (est.) and approximately 4000 on "waiting-list" with given national timings 0.06 IF was determined to be optimal"
The result: Yearly Discharge Growth Chart 45,000 40,000
Cumulative Percent Change
35,000 30,000 25,000 20,000 15,000 10,000 5,000 0 2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
Year
The Impact of Change™
Economic Effect
SG-2 Technical Brief
Assumptions 3. Studies show a 25% reduction in utilization (inpatient) with an increase in unemployment. Effect is indirect via loss of employment determined health insurance benefits. 4. The effect is probably "across-the-board" – e.g. even non-discretionary utilization is affected. The only exception is probably acute, trauma-related care. 5. By corollary, does utilization increase by 25% if unemployment declines? Probably by not as much.
Approach: 5. Get change in unemployment forecasts from Bureau of Labor Statistics and SG-2 analysis. Need. This is δ. Increasing unemployment is a positive number; while decreasing unemployment is a negative number. 6. The "utilization depressor" is µ. Usually set to -0.25 7. The "negative bias factor" is κ. Default is 25. Using this in an exponential accentuates the change on the positive (or increasing unemployment) side.
∆ = δ • µ • e (κ • δ ) Current Defaults:
The Impact of Change™
δ: Our current national set looks as follows:
Eastern Mass, SSM/Cent. Missouri and ANOVA were all run with these national numbers. Unemployment Curve Chart
"Change in unemployment" Impact
.02
.015
.01
.005
.
-.005
-.01 2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
Year
The Impact of Change™
∆: Applying the above equation (with defaults for κ and µ) to the national δ's, we get the following results:
∆
δ
Year 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
No kappa .0042 .0167 -.002 -.004 -.002 0 0 -.004 -.001 -.001
(with Kappa
-0.00105 -0.00417 0.0005 0.001 0.0005 0 0 0.001 0.00025 0.00025
-0.00117 -0.00634 0.000476 0.000905 0.000476 0 0 0.000905 0.000244 0.000244
"Delta" Chart ∆ 0.02
Pure
0.015
Kappa No Kappa 0.01
0.005
0
-0.005
-0.01 2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
Year
These ∆ 's are the "changes in rate" that will be applied to the utilization volumes (non-Medicare component – see below).
The Impact of Change™
Implementation: These "rates" are then integrated and applied to the non-Medicare component (for DRGs) or total (for OPCs) utilization volumes. Currently the Medicare fractions are "hard-coded" (non time & space dependent) into the core DRG table. Example of DRG Medicare fractions: DRG
DRG Description
Medicare Fraction
1.00 Craniotomy >17 X Trauma
0.3429158
1.10 Craniotomy >17 X Trauma
0.3429158
2.00 Craniotomy For Trauma >17
0.5328467
2.10 Craniotomy For Trauma >17
0.5328467
3.00 Craniotomy Age 0-17
0
3.10 Craniotomy Age 0-17
0
4.00 Spinal Proc
0.1917476
4.10 Spinal Proc
0.1917476
The rates can be applied to: 1. initial utilization volumes or 2. "nonlinearly" multiplied against population-driven utilization.
Treatment of Medicare Fractions There are two sources for medicare fraction: a. "Hard-coded"in the DRG tables directly or b. derived directly from the age-group identifier of the relevant dataset. E.g. If the age group is greater than 65 then the medicare fraction is 1; other wise it is zero. If the age grouping spans 65 then the medicare fraction is some number between 0 and 1 – appropriately weighted at the time that the age group designation is defined. The default is to use option (2); if option (2) does not yield a medicare fraction (e.g. for undefined or global age group designations) then option (1) will be applied. Option (2) also allows for the medicare fraction to change over time as the proportion of the population over 65 likewise changes.
The Impact of Change™
Results: National data; linear run. Using κ, m defaults Economic Population
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
157,348
157,170
156,203
156,244
156,383
156,460
156,462
156,462
156,599
156,640
156,679
157,348
161,549
166,054
170,891
176,088
181,679
187,701
194,194
201,204
208,782
216,987
-0.4% 37.9%
Population / Economic "Wedge" NHDS National 2000 - 2010 SG-2 Forecast Economic Population
250 200 150 100 50
0 2000 2001 2002 2003 2004 2005 2006 2007
2008
2009
2010
Economic Effect Detail 157,600 157,400 157,200 157,000 156,800 156,600 156,400 156,200 156,000 155,800 155,600 2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
The Impact of Change™
Consumer Effect
SG-2 Technical Brief
Assumptions 1. The consumer effect is closely related to the economic effect. 2. Applies to discretionary DRGs/OPCs only 3. Grows over the decade by an SG-2 determined curve
Approach: 1. Convolute D's from economic analysis with the SG-2 Consumerism curve. 2. These are the new rates 3. Apply only to discretionary DRGs (non-Medicare fraction) or discretionary OPCs (total fraction)
Consumerism Curve Chart 12.
Fractional Impact
10.
8.
6.
4.
2.
. 2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
Year
DRG
DRG Description
Discretionary DRG?
1.00 Craniotomy >17 X Trauma
N
1.10 Craniotomy >17 X Trauma
N
2.00 Craniotomy For Trauma >17 N 2.10 Craniotomy For Trauma >17 N 3.00 Craniotomy Age 0-17
N
3.10 Craniotomy Age 0-17
N
4.00 Spinal Proc
N
4.10 Spinal Proc
N
5.00 Extracranial Vascular Procs N 5.10 Extracranial Vascular Procs N 6.00 Carpal Tunnel Release
Y The Impact of Change™
DRG
DRG Description
6.10 Carpal Tunnel Release
Discretionary DRG? Y
The Impact of Change™
Results:
Population / Consumerism "Wedge" NHDS National 2000 - 2010 SG-2 Forecast
250 Consumerism Population
200
150
100
50
0 2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
Consumerism Effect Detail 157,500 157,400 157,300 157,200 157,100 157,000 156,900 2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
The Impact of Change™
Sociocultural Impacts
SG-2 Technical Brief
Assumptions 1. Many effects act roughly progressively over the forecast period (e.g. without a logistic curve, etc.). 2. These "meta-trends" often involve socio-cultural shifts within a population that are similar to demographic changes but involve factors beyond simply population change. Examples, include "obesity", increasing reliance of ERs for primary care, increasing acceptability of cosmetic surgery, etc. 3.
Approach: 1. Identify the sociocultural factor (potentially locally specific) 2. Identify which utilization parameters (DRG, OPC, etc.) are affected 3. Determine from the literature and/or "micro-model" an estimation of th percent yearly change in utilization attributable to that sociocultural factor. This number is termed the impact factor (IF).
Example sociocultural factors:
The Impact of Change™
Example Result: Effect of obesity on DRG 127 Heart Failure & Shock Yearly Discharge Growth Chart 1,050,000 1,040,000
Cumulative Percent Change
1,030,000 1,020,000 1,010,000 1,000,000 990,000 980,000 970,000 960,000 2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
Year
The Impact of Change™
Related Documents
Impact Of Change Forecaster Documentation
November 2019 5
Impact Of Change Methodology
November 2019 9
Change Impact Assessment.doc
December 2019 7
The Impact Of Change Sell Sheet V2
November 2019 6
Impact Of Climate Change On Water Resources.pdf
August 2019 46