Original Research

JAN

ORIGINAL RESEARCH

Burnout and physical and mental health among Swedish healthcare workers ˚ sberg & A ˚ ke Nygren Ulla Peterson, Evangelia Demerouti, Gunnar Bergstro¨m, Mats Samuelsson, Marie A Accepted for publication 6 December 2007

Correspondence to U. Peterson: e-mail: [email protected] Ulla Peterson BSc RN PhD Student Department of Clinical Neuroscience, Section for Personal Injury Prevention, Karolinska Institute, Stockholm, Sweden Evangelia Demerouti PhD Associate Professor Department of Social & Organizational Psychology and Research Institute Psychology & Health, Utrecht University, Utrecht, The Netherlands Gunnar Bergstro¨m PhD Associate Professor Department of Clinical Neuroscience, Section for Personal Injury Prevention, Karolinska Institute, Stockholm, Sweden Mats Samuelsson PhD RN Senior Lecturer, Director of Education The Swedish Red Cross University College of Nursing; and Department of Clinical Neuroscience, Department of Psychiatry Karolinska Institute, Stockholm, Sweden Marie A˚sberg MD PhD Professor Emeritus Department of Clinical Sciences, Danderyd Hospital, Stockholm, Sweden A˚ke Nygren MD PhD Professor Emeritus Department of Clinical Sciences, Danderyd Hospital, Stockholm, Sweden

˚ SBERG M. ¨ M G., SAMUELSSON M., A PETERSON U., DEMEROUTI E., BERGSTRO ˚ . (2008) & NYGREN A

Burnout and physical and mental health among Swedish healthcare workers. Journal of Advanced Nursing 62(1), 84–95 doi: 10.1111/j.1365-2648.2007.04580.x

Abstract Title. Burnout and physical and mental health among Swedish healthcare workers. Aim. This paper is a report of a study to investigate how burnout relates to self-reported physical and mental health, sleep disturbance, memory and lifestyle factors. Background. Previous research on the possible relationship between lifestyle factors and burnout has yielded somewhat inconsistent results. Most of the previous research on possible health implications of burnout has focused on its negative impact on mental health. Exhaustion appears to be the most obvious manifestation of burnout, which also correlates positively with workload and with other stressrelated outcomes. Method. A cross-sectional study was conducted, using questionnaires sent to all employees in a Swedish County Council (N = 6118) in 2002. The overall response rate was 65% (n = 3719). A linear discriminant analysis was used to look for different patterns of health indicators and lifestyle factors in four burnout groups (non-burnout, disengaged, exhausted and burnout). Results. Self-reported depression, anxiety, sleep disturbance, memory impairment and neck- and back pain most clearly discriminated burnout and exhausted groups from disengaged and non-burnout groups. Self-reported physical exercise and alcohol consumption played a minor role in discriminating between burnout and non-burnout groups, while physical exercise discriminated the exhausted from the disengaged group. Conclusion. Employees with burnout had most symptoms, compared with those who experienced only exhaustion, disengagement from work or no burnout, and the result underlines the importance of actions taken to prevent and combat burnout. Keywords: burnout, discriminant analysis, disengagement, exhaustion, healthcare workers, Job Demands–Resources model, questionnaire survey

Introduction Previous research on the possible relationship between burnout and sickness absence is rather limited (Schaufeli & 84

Enzmann 1998), and the impression gained from recent overviews of the burnout literature is that our knowledge is more extensive regarding predictors (work characteristics) of burnout than possible consequences for health. Maslach

 2008 The Authors. Journal compilation  2008 Blackwell Publishing Ltd

JAN: ORIGINAL RESEARCH

(2001) even raised the question of whether burnout actually has anything to do with health, and suggested that ‘burnout may be more significant in mediating job outcomes, such as behaviours that affect the quality of work’ (p. 611). From a recent review of the evidence of health consequences of burnout, it was concluded that ‘yet the accumulating evidence suggests that chronic burnout may harm physical health through different pathways’ (Shirom et al. 2005, p. 32). An association between burnout and increased risk of future sickness absence because of psychiatric disorders and diseases of the musculoskeletal system has also been found by Toppinen-Tanner et al. (2005). Psychiatric disorders, particularly depression, anxiety and stress-related conditions, are currently (2006) the predominant reason for long-term sick leave (>12 months) among women in Sweden (33Æ8%), and the second most common reason for men (25Æ5%) (National Social Insurance Board, 2006). Possible reasons for the increased rate of sick leave due to psychiatric disorders have been debated, and recent changes in the psychosocial work environment are one of the many factors that have been discussed (National Social Insurance Board, 2003; Theorell 2006). This suggests that burnout, defined as a work-related psychological stress reaction with the main components being exhaustion and disengagement (cynicism), might be an intervening link between an adverse psychosocial work environment and work stress-related psychiatric disorder (Tennant 2001, Paterniti et al. 2002, Borritz et al. 2005, Wang 2005, Ylipaavalniemi et al. 2005).

Background Most research on the possible health implications of burnout has focused on its negative impact on mental health (Melamed et al. 2006). An association between burnout and depression (Ahola et al. 2005, Toppinen-Tanner et al. 2005, Wang 2005), as well as between burnout and anxiety (Bargellini et al. 2000), has previously been described. Several authors have also concluded that depression and burnout are two distinct and separable constructs (Glass & McKnight 1996, Melamed et al. 2006, Ahola & Hakanen 2007), and that the exhaustion component of the burnout syndrome is more strongly related to depression (Glass & McKnight 1996, Schaufeli & Enzmann1998). Cognitive symptoms have also been associated with burnout; for example, Sandstro¨m et al. (2005) showed that patients with chronic burnout had specific cognitive impairments in non-verbal memory and attention. Similarly, Rydmark et al. (2006) found impaired working memory in women on long-term sick leave with stress-induced depression.

Burnout and physical and mental health

Previous research on the possible relationship between lifestyle and burnout has yielded somewhat inconsistent results. For example, Gorter et al. (2000) found that a highrisk group for burnout had an unhealthier lifestyle regarding sporting/physical exercise and increased alcohol consumption, while Burke (1994) and Shanafelt et al. (2002) found no associations between at-risk alcohol use and burnout. In a review by Shirom et al. (2005), it was concluded that current evidence did not support viewing health behaviours as either moderating or mediating the relationships between burnout and health. In the study reported in this paper, we investigated whether burnout and non-burnout respondents relate in different ways to health and lifestyle factors [focusing on exercise, alcohol consumption, anxiety and depression and on some physical conditions thought to be influenced by stress, namely neck and back pain (Bongers et al. 2006), tinnitus (Holgers 2003) and oral problems (Genco et al. 1998, Johannsen et al. 2006)].

Job Demands–Resources model The theoretical basis for the study was the Job Demands– Resources model (JD-R; Demerouti et al. 2001a, Bakker & Demerouti 2007). One proposition in the model is that work characteristics may evoke two different processes. First, high job demands (e.g. work overload and emotional demands) may lead to depletion of energy, and might therefore lead to exhaustion and to health problems (e.g. Demerouti et al. 2000, 2001b, Lee & Ashforth 1996). Second, it is assumed that access to job resources (e.g. positive social climate at work, fair and empowering leadership) has a motivational potential and may lead to engagement at work, while lack of adequate job resources can preclude goal accomplishment, which might cause a sense of failure and frustration. Disengagement from work can then be seen as an important self-protection mechanism that may prevent the future frustration of not obtaining work-related goals (cf. Antonovsky 1987). This suggests that health problems might be differently related to the two main components – exhaustion and disengagement (cynicism) – of the burnout syndrome.

Measuring burnout Originally, professional burnout was observed and described among those who worked in the human services and health care. According to Maslach et al. (1996), burnout consists of three dimensions: emotional exhaustion, depersonalization (cynicism) and reduced personal accomplishment. However, research in the past decade has shown that the two core

 2008 The Authors. Journal compilation  2008 Blackwell Publishing Ltd

85

U. Peterson et al.

components of burnout – exhaustion, and cynicism or disengagement from work – can be observed in virtually any occupational group (Bakker et al. 2002, Demerouti et al. 2000, 2001a, 2001b; Leiter & Schaufeli 1996). The Maslach Burnout Inventory (MBI) is the most frequently used instrument to assess burnout (Schaufeli & Enzmann 1998). In the present study, we used an alternative measure of burnout, the Oldenburg Burnout Inventory (OLBI; Demerouti et al. 2001a, 2003), which conceives of burnout as a syndrome of work-related negative experiences, including feelings of exhaustion and disengagement from work. This instrument aims to overcome a psychometric shortcoming of the MBI, namely the one-sided wording of the items (Demerouti et al. 2001a). In the MBI scales, all exhaustion and cynicism items are phrased negatively, and all the professional efficacy items are phrased positively. From a psychometric point of view, such one-sided scales are inferior to scales that include both positively and negatively worded items (cf. Price & Mueller 1986), and might lead to artificial factor solutions (Lee & Ashforth 1990). The third burnout component, reduced personal accomplishment (Maslach et al. 1996), is excluded from this definition of burnout because it is not thought to constitute a core dimension of the condition (Shirom 2002).

The study Aim The aim of the study was to investigate how burnout relates to self-reported physical and mental health, sleep disturbance, memory and lifestyle factors. Two hypotheses were investigated: Hypothesis 1. As predicted by the JD-R model, impaired health will be more associated with the exhaustion component of burnout, than to the disengagement component. Hypothesis 2. Health impairment, impaired memory and sleep disturbance will be higher in the burnout and exhausted groups than in the non-burnout group.

Design The study was cross-sectional and based on questionnaires sent to all employees in a County Council area in Sweden during 2002.

Participants The study population included physicians, Registered Nurses, nursing assistants, social workers, occupational therapists, 86

physiotherapists, psychologists, dental nurses and hygienists, dentists, administrators, teachers and technicians. Registered Nurses, including midwives and biomedical technicians, constituted 33Æ7% (N = 1252) of the respondents. Questionnaires were sent to all of the employees (N = 6118) and resulted in an overall response rate of 65% (N = 3976) and a total of 3719 (61%) who had completed the OLBI. A total of 82% were women. Ages ranged from 22 to 66 years and the mean age for the sample was 47Æ0 years (SD 9Æ0), for women 46Æ9 years (SD 9Æ1), and for men 47Æ5 years (SD 8Æ9). The demographics of non-respondents were similar to those of the participants.

Measures The OLBI (Demerouti et al. 2001a, 2003) was used to assess burnout. The OLBI was translated into Swedish by two physicians at the Karolinska Institute, Stockholm. Translations from both the English and German versions of the OLBI were made in parallel, and the two versions were then compared to reach consensus on the Swedish translation. The Swedish text was then back-translated into German by a bilingual native German speaker, and the two German texts were compared by another German speaker experienced with the scale, and the scale’s original creator, showing acceptable correspondence between the two versions. The OLBI has two dimensions: exhaustion and disengagement. Each subscale consists of eight items, and four are positively worded and four negatively. Each item has four response alternatives, ranging from 1 (totally disagree) to 4 (totally agree). The positive and negative exhaustion and disengagement items were presented in mixed order. A study among 232 employees from different occupational groups confirmed the factorial and convergent validity of the OLBI and the MBI-GS (Demerouti et al. 2003). In the present study, Cronbach’s alpha for both exhaustion and disengagement was 0Æ83. The Hospital Anxiety and Depression Scale (HAD) is a self-assessment scale, developed for assessing change in non-psychiatric patients’ emotional state, as well as for assessing significant degrees of anxiety (HAD-A) and depression (HAD-D) (Zigmond & Snaith 1983). The HAD consists of 14 items, seven for depression and seven for anxiety. Each item has four response alternatives, ranging from 0 to 3. Cut-off scores for both anxiety and depression are, as suggested by Zigmond and Snaith (1983), 8–10 for doubtful cases and 11 or more for definite cases. A Swedish version of the HAD has shown acceptable validity and internal consistency (Lisspers et al. 1997). A review by Bjelland et al. (2002) showed that the

 2008 The Authors. Journal compilation  2008 Blackwell Publishing Ltd

JAN: ORIGINAL RESEARCH

HAD performed well in screening for separate dimensions of anxiety and depression in both somatic, psychiatric and primary care patients and in the general population. In the present study, Cronbach’s alpha was 0Æ85 for HAD-D and 0Æ86 for HAD-A. Self-rated health (SRH) was measured with one single item from the Short Form Health Survey (SF-36; Sullivan et al. 1994, 1995): ‘In general, how would you describe your health?’ The item has a standardized response format with five alternatives, ranging from 1 to 5, where 1 = excellent, 2 = very good, 3 = good, 4 = fair and 5 = poor. In this analysis, the response alternatives ‘fair’ and ‘poor’ were considered as low self-rated health. Sleep disturbance was measured with three items (presented in Table 2), and with five response alternatives ˚ kerstedt ranging from ‘Never’ to ‘Always/Every day’ (A et al. 2002). Memory was assessed with 10 items (Nilsson et al. 1997), for example ‘I have lost things because I did not remember to take them with me when leaving’, ‘How do you find your memory today, compared to five years ago?’ Each item had five response alternatives ranging from ‘never’ to ‘usually’ or ‘much better’ to ‘much worse’. Reliability analysis and validation of this scale are not yet available. Cronbach’s alpha in the present study was 0Æ82. Tinnitus and oral problems were assessed with one question each, and the response alternatives were Never/ Sometimes/Often/Always or Yes/No. Questions about neck and back pain concern problems several times during the last year, or if neck and back pain is present right now and if the respondent several times has been or is right now on the sick list due to these problems. The Alcohol Use Disorders Identification Test (AUDIT) (Saunders et al. 1993) was used to measure problems with alcohol consumption. AUDIT includes three domains: alcohol consumption, drinking behaviour and alcohol-related problems. The scale consists of 10 items with five response alternatives ranging from ‘Never’ and ‘Less than monthly’ to ‘Daily’, giving a maximum possible score of 40. In the present study, the total score for the three domains was used in the discriminant analysis. A study among 997 persons randomly selected from the general Swedish population showed that the internal and test–retest reliability were satisfactory (Bergman & Ka¨llme´n 2002). Cronbach’s alpha in the present study was 0Æ75. Frequency of physical exercise was assessed on a five-point scale scoring from ‘never’ to ‘three times per week or more’. Exercising two times per week or more was classified as high, and ‘never’ or ‘irregularly’ was categorized as low physical activity (Engstro¨m et al. 1993).

Burnout and physical and mental health

Data collection Participants were informed about the purpose of the study via information meetings at the workplace and via the County Council’s intranet. The questionnaires were sent to home addresses and were accompanied by a letter explaining the purpose of the study and a prestamped envelope. The voluntary nature of participation was emphasized, and respondents were guaranteed confidentiality. Four weeks later, questionnaires and a follow-up letter were mailed to non-respondents.

Ethical considerations The study was approved by the Research Ethics Committee of Linko¨ping University, Sweden.

Data analysis To identify the four burnout groups, scores ‡2Æ25 on exhaustion were considered as having high exhaustion, while scores ‡2Æ1 on disengagement were considered as high. These scores on the OLBI corresponded to the mean scores on the MBI of a group of burned-out employees as diagnosed by a physician (Schaufeli et al. 2001). This yielded the following target groups for the discriminant analysis: • low exhaustion and low disengagement (non-burnout group) n = 1302; • low exhaustion and high disengagement (disengaged group) n = 469; • high exhaustion and low disengagement (exhausted group) n = 697; • high exhaustion and high disengagement (burnout group) n = 1251. A linear discriminant analysis was used to look for different generalized patterns of health impairment to indicate the differences between the four burnout groups. This form of statistical treatment of the problem offers the advantage of making use of the common variance of the individual aspects or items of health impairment. Pearson’s correlation coefficients and collinearity statistics were computed to evaluate potential association between the variables used (Table 1). To elucidate the response patterns among the four groups, some of the scales and items were analysed separately. Differences in self-rated health and sleep disturbance were estimated by comparing proportions, and were tested for statistical significance of the differences on the basis of 95% confidence intervals (Table 2). The non-burnout group was

 2008 The Authors. Journal compilation  2008 Blackwell Publishing Ltd

87

U. Peterson et al. Table 1 Pearson’s correlation coefficients, and tolerance and variance inflation factor (VIF), for the variables used in the discriminant analysis Correlations

1

1. Oral problems 2. Tinnitus 3. AUDIT 4. HAD-D 5. HAD-A 6. Neck-back pain 7. Exercise 8. Sleep disturbance 9. Memory 10. OLBI-Ex 11. OLBI-Dis

0Æ07 0Æ05 0Æ15 0Æ14 0Æ09 0Æ07 0Æ11 0Æ14 0Æ14 0Æ06

2

3

0Æ05 0Æ15 0Æ16 0Æ13 0Æ04 0Æ14 0Æ13 0Æ15 0Æ09

4

0Æ06 0Æ07 0Æ02 0Æ01 0Æ07 0Æ02 0Æ04 0Æ08

5

0Æ70 0Æ19 0Æ09 0Æ44 0Æ39 0Æ62 0Æ44

0Æ22 0Æ05 0Æ45 0Æ39 0Æ58 0Æ36

6

7

0Æ07 0Æ24 0Æ16 0Æ26 0Æ15

0Æ11 0Æ02 0Æ12 0Æ04

8

0Æ33 0Æ49 0Æ27

9

0Æ38 0Æ25

10

Tolerance

VIF

0Æ55

0Æ96 0Æ95 0Æ99 0Æ41 0Æ44 0Æ90 0Æ97 0Æ69 0Æ78 0Æ44 0Æ68

1Æ04 1Æ05 1Æ01 2Æ44 2Æ28 1Æ11 1Æ03 1Æ45 1Æ29 2Æ26 1Æ47

AUDIT, Alcohol Use Disorders Identification Test; HAD-A and HAD-D, Hospital Anxiety and Depression Scale; OLBI-Ex, Oldenburg Burnout Inventory-Exhaustion; OLBI-Dis, Oldenburg Burnout Inventory-Disengagement.

No burnout Disengaged Health 5Æ7 In general, how would you describe your health? Fair and poor (%) D(95% CI) Sleep disturbance I have difficulty falling asleep* (%) D(95% CI) I have difficulty waking up* (%) D(95% CI) I wake up a few times every night, and sometimes, I have difficulty falling asleep again* (%) D(95% CI)

6Æ1

Exhausted

Burnout

19Æ6

27Æ1

Table 2 Differences (D) in proportions, on the basis of 95% CIs, of self-reported health and of self-reported sleep disturbance

0Æ4 ( 2Æ0; 3Æ4) 13Æ9 (10Æ6; 17Æ3) 21Æ4 (18Æ5; 24Æ3)

5Æ7

5Æ9

16Æ2

21Æ7

6Æ5

0Æ2 ( 2Æ2; 3Æ2) 10Æ5 (7Æ4; 13Æ8) 9Æ1 18Æ3

16Æ0 (13Æ1; 18Æ7) 23Æ7

9Æ2

2Æ6 ( 0Æ3; 6Æ1) 11Æ8 (8Æ5; 15Æ3) 11Æ0 29Æ2

17Æ2 (14Æ2; 20Æ0) 31Æ9

1Æ8 ( 1Æ5; 5Æ5) 20Æ0 (16Æ0; 23Æ9) 22Æ7 (19Æ4; 25Æ8)

Reference group is the non-burnout group. *Response alternatives presented = several times/week and always.

used as the reference group. Differences between the four groups in age and HAD scores were tested by an analysis of variance (ANOVA ), and post hoc comparisons were made according to Dunnet’s t-test (Table 3).

Results Collinearity statistics and Pearson’s correlation coefficients for the study variables are displayed in Table 1. The highest correlation between the variables was found between HAD-D and HAD-A (r = 0Æ70), and the correlations between the 88

others were considerably lower. The collinearity statistics revealed that the variance inflation factors (VIF) ranged from 1Æ01 to 2Æ44, and thus multicollinearity might not be a serious threat in this study (O’Brien 2007). Moreover, an analysis of variance revealed no difference in age (F = 2Æ13, d.f. = 3, P = 0Æ09) between the four burnout groups.

Mean score analyses Overall, there was an increase in severity over the four groups in SRH, anxiety and depression, in the following order (from

 2008 The Authors. Journal compilation  2008 Blackwell Publishing Ltd

JAN: ORIGINAL RESEARCH Table 3 Mean, standard deviation of the Hospital Anxiety (Anx) and Depression (Dep) Scale and the proportions (%) scoring higher than 8 and 11 among the four groups

Burnout and physical and mental health

HAD HAD HAD HAD HAD HAD

Anx mean (SD ) Anx ‡8 and £10 (%)* Anx ‡11 (%)* Dep mean (SD ) Dep ‡8 and £10 (%)* Dep ‡11 (%)*

No burnout

Disengaged

Exhausted

Burnout

P-value

3Æ0 (2Æ5) 3Æ9 1Æ0 1Æ9 (2Æ0) 1Æ6 0Æ2

3Æ6 (2Æ8) 5Æ5 2Æ6 2Æ8 (2Æ3) 2Æ7 0Æ7

5Æ7 (3Æ3) 19Æ2 7Æ5 4Æ3 (3Æ1) 11Æ1 4Æ5

6Æ8 (3Æ8) 23Æ4 16Æ2 5Æ8 (3Æ6) 18Æ1 10Æ4

<0Æ001

<0Æ001

HAD-A and HAD-D, Hospital Anxiety and Depression Scale. *According to the cut-off scores suggested by Zigmond & Snaith (1983).

Discriminant analysis The discriminant analysis yielded a statistically significant separation of the four groups, Wilk’s k = 0Æ67, v2 (d.f. = 27) = 1241Æ1, P < 0Æ001. More specifically, two discriminant functions were significant for an optimal discrimination between the groups. The eigenvalue of the first discriminant function was 0Æ47 and the canonical correlation was 0Æ57. For the second discriminant function, these values were 0Æ01 and 0Æ11 respectively. As can be seen from the plot of the group centroids (Figure 1), the first discriminant function clearly discriminates the burnout and exhausted groups from the

1·00

Discriminant function 2 life-style

the lowest to the highest): non-burnout, disengaged, exhausted and burnout. The proportion of respondents rating their health as poor or fair was statistically significantly higher in the burnout (27Æ1%) and exhausted (19Æ6%) groups, compared with the non-burnout group (5Æ7%), while there was no significant difference between the disengaged (6Æ1%), and non-burnout groups (Table 2). An overview of self-reported sleep disturbance for the four groups (Table 2) shows a similar pattern in all three items for the burnout and exhausted groups, which reported statistically significantly more sleep disturbance, compared with the non-burnout group. Means, standard deviations of anxiety and depression and proportions of respondents scoring above the cut-off scores suggested by Zigmond and Snaith (1983) are presented in Table 3. This shows that the exhausted and burnout groups have the highest proportions of respondents scoring equal to or above 11 (definite cases), in both depression and anxiety. There was a statistically significant difference between the four groups (HAD-A, F = 424Æ26, P < 0Æ001; HAD-D, F = 348Æ89, P < 0Æ001). Dunnet’s post hoc test revealed a statistically significant difference in both depression and anxiety between the burnout, exhausted and disengaged groups, compared with the non-burnout group.

0·50 Disengaged Burn-out 0·00 Non Burn-out Exhausted –0·50

–1·00 –1·00

–0·50

0·00

0·50

1·00

Discriminant function 1 health Figure 1 Group centroids for the four burnout groups in a twodimensional space.

non-burnout and disengaged groups, while the second discriminant function separates the disengaged from the exhausted group. Overall, 47Æ5% of the total sample could be correctly classified, indicating that the classification by these discriminant functions is obviously superior to a random assignment (which would be only 25% of the sample). The plot of the group centroids shows that the health variables could more accurately discriminate between the burnout and non-burnout groups, as the distance between these groups is the largest. Discrimination between the disengaged and nonburnout groups, as well as between the exhausted and burnout groups, was less accurate as the distances between these two pairs of groups are rather small. The standardized canonical coefficients in the discriminant functions are displayed in Table 4. The loadings are considered substantial when they are ‡|0Æ30|. The first canonical variable – discriminating for the presence or absence of burnout – represents the scales measuring depression,

 2008 The Authors. Journal compilation  2008 Blackwell Publishing Ltd

89

U. Peterson et al. Table 4 Standardized canonical coefficients for scales in the discriminant functions based on the health indicators (N = 3090)

Depression Anxiety Sleep disturbances Memory Neck and back pain Oral problems or problems with teeth Physical activity (exercise) Tinnitus Alcohol consumption

0Æ84* 0Æ77* 0Æ64* 0Æ49* 0Æ33* 0Æ28* 0Æ12 0Æ22* 0Æ05

Function 2 0Æ33 0Æ17 0Æ21 0Æ06 0Æ16 0Æ11 0Æ40* 0Æ03 0Æ48*

Burnout Actual group membership

Function 1

Disengaged

Exhausted

Non-burnout 0%

*Largest absolute correlations between each variable and any discriminant function.

20%

Non-burnout

anxiety, sleep disturbance, self-reported memory problems and neck and back pain. The second canonical variable – separating exhaustion from disengagement – represents the scales measuring self-reported physical exercise and alcohol consumption. As positive values in a discriminant function mean predominance of the respective canonical (or discriminant) variables and negative values an under-representation of these variables, Figure 1 and Table 4 can be interpreted as follows: positive values on the first discriminant function delineate an over-representation of impaired mental health, sleep disturbances, experience of impaired memory and neck and back pain. Negative values on this discriminant function indicate an under-representation or absence of symptoms. Similarly, positive values on the second discriminant function stand for an over-representation of physical exercise and of hazardous alcohol consumption, while negative values indicate under-representation of physical exercise and hazardous alcohol consumption. Specifically, the burnout group is characterized by the presence of impaired mental health (located on the positive part of the first discriminant function), while the non-burnout group is characterized by the absence of impaired mental health (located on the negative part of the first discriminant function). These groups have almost zero values on the second discriminant function, indicating that physical exercise and reported alcohol consumption played a minor role in discriminating between them. Moreover, the distinctive characteristic of the exhausted group is the experience of impaired mental health (positive values on the first discriminant function), low physical exercise and less self-reported harmful alcohol consumption (negative values on the second discriminant function), whereas the disengaged group is characterized by the opposite combination: low impairment of mental health (negative values on the first discriminant function), high 90

40%

60%

80%

100%

Predicted group membership (based on health) Exhausted

Disengaged

Burnout

Figure 2 Results of classification procedure.

physical exercise and reports of more hazardous alcohol consumption (positive values on the second discriminant function). Classifications Figure 2 shows the proportion of participants classified correctly and falsely. The best prediction can be made for the burnout and non-burnout groups (correct classifications are above 50%). The worst prediction is made for the exhausted participants, where 29% were correctly classified and 33% were classified as being burnt out. Thirty-seven per cent of the disengaged participants were correctly classified, while 35% were classified as being non-burnout. Figure 2 shows that the proportion of extremely wrong classifications was relatively low (predicted non-burnout when participants were actually burned out).

Discussion The data revealed that self-reported depression, anxiety, sleep disturbance, impaired memory and neck and back pain were the health indicators that most clearly discriminated the burnout and exhausted groups from the disengaged and nonburnout groups. The results also indicated that self-reported physical exercise and alcohol consumption played a minor role in discriminating between the burnout and non-burnout groups, while physical exercise and alcohol consumption seem to discriminate the exhausted from the disengaged group.

Lifestyle factors Alcohol consumption discriminated the disengaged group from the exhausted group, indicating that the former group

 2008 The Authors. Journal compilation  2008 Blackwell Publishing Ltd

JAN: ORIGINAL RESEARCH

had more hazardous or harmful alcohol use. As a gender difference in AUDIT scores has been found previously (Bergman & Ka¨llme´n 2000), a post hoc linear regression analysis was performed. When gender was taken into account, there was no difference in self-reported alcohol consumption between the exhausted and disengaged groups. A higher frequency of exercise was reported from the disengaged group, compared with the exhausted. As a gender difference was previously found by Fusilier and Manning (2005), a post hoc linear regression analysis for gender was performed, revealing that this did not affect our result. One possible interpretation might be that the exhausted group put too much effort into work, and do not have enough energy left over for exercise in leisure time, and/or that they experience exercise as a further increase of the burden. Indeed, the highest proportion of overtime was found in the exhausted group. Another possible interpretation could be that exercise might serve as a protective factor against developing burnout for the disengaged group (cf. Fusilier & Manning 2005). However, previous research into lifestyle factors and burnout has shown somewhat inconsistent results (e.g. Gorter et al. 2000, Shanafelt et al. 2002).

Health indicators and sleep disturbance Consistent with previous research (Melamed et al. 1999, Grossi et al. 2003, Ekstedt et al. 2006), the burnout group reported more sleep disturbance compared with the disengaged and non-burnout groups. Interestingly, a higher frequency of respondents in the exhausted group also reported sleep disturbance. This indicates that sleep disturbance might be more associated with the exhaustion component in burnout, which also seems reasonable. The result revealed a statistically significantly higher frequency of respondents with poor self-rated health in the exhausted and burnout groups, compared with the nonburnout group. SRH has previously been found to be a powerful predictor of morbidity (Kaplan et al. 1996) and mortality (Idler & Benyamini 1997), and the result therefore highlights the importance of interventions aimed at preventing exhaustion and burnout. Although tinnitus and oral problems did not discriminate between the four burnout groups, there was an increase in occurrence over the four groups, in the following order (from the lowest to the highest): non-burnout, disengaged, exhausted and burn out. There was no statistically significant difference between the exhausted and burnout groups, but both these groups reported statistically significantly higher frequency of tinnitus and oral problems, compared with the non-burnout group (data not presented). Stress has previously

Burnout and physical and mental health

been found to be an indicator of risk of periodontal disease in adults (Genco et al. 1998), and Johannsen et al. (2006) found that the amount of plaque and gingival inflammation was higher among women with burnout. In our total sample (N = 3719), 4Æ5% were defined as definite cases of depression (HAD-D score ‡11, as suggested by Zigmond & Snaith 1983). Definite cases of depression occurred twice as often in the burnout group (10Æ4%), and definite cases of anxiety were even more common (16Æ2%). The increase in severity of depression over the four groups supports the findings of Glass and McKnight (1996) that depression might be more related to the exhaustion component of burnout. A relationship between exhaustion and job demands has previously been found (Demerouti et al. 2001a), as well as an association between high job demands and an increased risk of psychiatric disorder (Stansfeld et al. 1997).

Relationship between burnout patterns and health The present study revealed similarities in the pattern of answers to questions about health, sleep disturbance and memory between, on the one hand, the exhausted and burnout groups, and, on the other, between the disengaged and non-burnout groups. The two former groups reported poorer health, more sleep disturbance and impaired memory, and interestingly 33% in the exhausted group were actually classified as being burnt out in the discriminant analysis. Our results therefore indicate that depression, anxiety, neck and back pain, poor SRH, sleep disturbance and perceived memory impairment might be more associated with the exhaustion component of burnout. Exhaustion has previously been found to correlate positively with other stress-related outcomes (Maslach 2003), but several authors have also suggested that stress and burnout are two different constructs (Pines & Keinan 2005, Smith et al. 2006). However, it has been emphasized that focusing exclusively on the exhaustion dimension might lead to loss of appreciation of other aspects of the burnout syndrome (Maslach et al. 2001). Burnout is thought to be a work-related stress syndrome (Maslach et al. 1996), and a relationship between exhaustion and job demands, as well as between disengagement and job resources, has been found previously (Demerouti et al. 2001a). Peterson et al. (2008) indicated that it was access to/lack of relevant resources at work (a fair and empowering leadership, a positive social climate at the workplace, control of decision and support from superiors) that were crucial for whether a respondent was classified as burnt out or not. That study also revealed that the exhausted group experienced high job demands and also reported good access to job

 2008 The Authors. Journal compilation  2008 Blackwell Publishing Ltd

91

U. Peterson et al.

• Previous research on the possible relationship between lifestyle factors and burnout has yielded somewhat inconsistent results. • Exhaustion appears to be the most obvious manifestation of burnout, which also correlates positively with workload and other stress-related outcomes.

necessarily indicate that the respondent is a case of ‘clinical burnout’, this is a large and important target group for interventions aimed at preventing stress and burnout. A previous study (Peterson et al. 2008), revealed that, e.g. a fair and empowering leadership, a positive social climate at the workplace, control of decision and support from superiors, as well as reasonable work load, was important in avoiding burnout.

What this paper adds

Study limitations

Burnout is associated with poorer self-rated health, more sleep disturbance and impaired memory. • Self-reported physical exercise and alcohol consumption play a minor role in discriminating between the burnout and non-burnout groups. • Health impairment and sleep disturbance may be more strongly related to the exhaustion component of the burnout syndrome. • Burnout might be an intervening link between an adverse psychosocial work environment and the medical condition ‘exhaustion disorder’.

The use of self-report data might be a limitation due to the risk of common method variance (Podsakoff et al. 2003, Spector 2006). The study was cross-sectional, and consequently the results allow more than one interpretation of the direction of causality. Despite that, they do show the respondents’ pattern of answers, and the four groups actually related in different ways to questions about health and lifestyle factors. The HAD-D and HAD-A were correlated (r = 0Æ70), and high correlations between these two subscales were also found in the review by Bjelland et al. (2002). According to Licht (1995, p. 45), ‘most investigators would probably agree that correlations of r > 0Æ80 between predictors should be considered very problematic’ in terms of multicollinearity. As the correlation did not exceed 0Æ70 (the other correlations were considerably lower), and both conditions were of interest among the four groups, we estimated that they had sufficient unique variance and decided to include both subscales in the analysis. Although a response rate of 65% may appear lower than desirable, it compares well with several recent studies in burnout and nursing (e.g. 48Æ8% in Kilfedder et al. 2001 and 59% in Demerouti et al. 2000). Only public service employees were involved, but as blue and white collar workers were included, we believe that the results will nonetheless be fairly generalizable. Another limitation is that 82% of the respondents were women. This is not unexpected, as women are generally over-represented in healthcare professions such as those included in our study. Most of the instruments used are validated, but the instrument used to assess memory has not been formally tested and this result should therefore be interpreted with caution.

What is already known about this topic

•

resources, while the burnout group reported high job demands but poor access to job resources. This raises the question of whether burnout initially is more associated with access to/lack of relevant job resources. In that case, and if burnout is considered as a gradually developing condition in response to chronic stress at work (cf. Maslach & Goldberg 1998), health impairment is probably not the initial problem. Instead, experience of poor access to relevant job resources might cause disengagement, failure and frustration. Disengagement might also lead to withdrawal from work and to reduced motivation or commitment (Demerouti et al. 2001a, Bakker et al. 2003), and might in turn affect the quality of work. However, it seems reasonable that long-term experience of lack of resources might gradually result in exhaustion and subsequent health impairment. Taken together, the evidence suggests that burnout may be an intervening link between an adverse psychosocial work environment and the medical condition ‘Exhaustion disorder’, for which tentative diagnostic criteria were recently formulated in Sweden (National Board of Health and Welfare, 2003). This issue needs to be further examined using a longitudinal approach. The highest frequency of respondents in the present study were Registered Nurses, including midwives and biomedical technicians, and 32% of them were actually located in the burnout group, while 39Æ4% were considered as non-burnout. Although an OLBI score above the cut-off limits does not 92

Conclusion Using a large sample of healthcare workers, it was possible to predict membership in a specific burnout category on the basis of health indicators. The result convincingly showed that burnout was associated with poorer self-rated health,

 2008 The Authors. Journal compilation  2008 Blackwell Publishing Ltd

JAN: ORIGINAL RESEARCH

more depression and anxiety, more sleep disturbance and perceived impaired memory. Therefore, our results underline the importance of finding adequate strategies to prevent and combat stress and burnout. This would not only prevent from much unnecessary suffering for the individual employee, but would also benefit the organization as a whole.

Author contributions UP and ED were responsible for the study conception and design. UP performed the data collection. UP and ED performed the data analysis. UP, ED, MA˚ and A˚N were responsible for drafting the manuscript. ED, GB, MS, MA˚ ˚ N made critical revisions to the paper for important and A intellectual content. ED and GB provided statistical expertise. MA˚ and A˚N obtained funding. ED, GB, MS, MA˚ and A˚N supervised the study.

References Ahola K. & Hakanen J. (2007) Job strain, burnout, and depressive symptoms: a prospective study among dentists. Journal of Affective Disorders 104, 103–110. doi: 10.1016/j.jad.2007.03.004. Ahola K., Honkonen T., Isometsa¨ E., Kalimo R., Nykyri E., Aromaa A. & Lo¨nnqvist J. (2005) The relationship between jobrelated burnout and depressive disorders – results from the Finnish Health 2000 Study. Journal of Affective Disorders 88, 55–62. A˚kerstedt T., Knutsson A., Westerholm P., Theorell T., Alfredsson L. & Kecklund G. (2002). To predict sleep disturbances, risk to fall asleep and mental tiredness. Three studies. (In Swedish), In Psychosocial Load and Risk Factors for Cardiovascular Diseases. Work and Health 2002:7. (Westerholm P., ed.), Arbetslivsinstitutet, Stockholm, pp. 51–53. Antonovsky A. (1987). Unraveling the Mystery of Health. How People Manage Stress and Stay Well. Jossey-Bass, San Fransisco, CA. Bakker A.B. & Demerouti E. (2007) The Job Demands–Resources model: state of the art. Journal of Managerial Psychology 22, 309– 328. Bakker A.B., Demerouti E. & Schaufeli W. (2002) Validation of the Maslach Burnout Inventory – general survey: an Internet study. Anxiety, Stress and Coping: An International Journal 15, 245–260. Bakker A.B., Demerouti E., de Boer E. & Schaufeli W.B. (2003) Job demands and job resources as predictors of absence duration and frequency. Journal of Vocational Behavior 62, 341–356. Bargellini A., Barbieri A., Rovesti S., Vivoli R., Roncaglia R. & Borella P. (2000) Relation between immune variables and burnout in a sample of physicians. Occupational Environment Medicine 27, 453–457. Bergman H. & Ka¨llme´n H. (2000) Alcohol use among Swedes as assessed by the Alcohol Use Disorders Identification Test (AUDIT): a psychometric study (In Swedish). La¨kartidningen 97, 2078– 2084.

Burnout and physical and mental health Bergman H. & Ka¨llme´n H. (2002) Alcohol use among Swedes and a psychometric evaluation of the Alcohol Use Disorders Identification Test. Alcohol & Alcoholism 37(1), 245–251. Bjelland I., Dahl A.A., Tangen Haug T. & Neckelmann D. (2002) The validity of the Hospital Anxiety and Depression Scale. An updated literature review. Journal of Psychosomatic Research 52, 69–77. Bongers P.M., Ijmker S., van den Heuvel S. & Blatter B.M. (2006) Epidemiology of work related neck and upper limb problems: psychosocial and personal risk factors (Part I) and effective interventions from a bio behavioural perspective (Part II). Journal of Occupational Rehabilitation 16, 279–302. Borritz M., Bu¨ltmann U., Rugulies R., Christensen K.B., Villadsen E. & Kristensen T.S. (2005) Psychosocial work characteristics as predictors for burnout: findings from 3-year follow up of the PUMA study. Journal of Occupational and Environmental Medicine 47(10), 1015–1025. Burke R.J. (1994) Stressful events, work–family conflict, coping, psychological burnout, and well-being among police officers. Psychological Reports 75, 787–800. Demerouti E., Bakker A.B., Nachreiner F. & Schaufeli W.B. (2000) A model of burnout and life satisfaction amongst nurses. Journal of Advanced Nursing 32(2), 454–464. Demerouti E., Bakker A.B., Nachreiner F. & Schaufeli W.B. (2001a) The Job Demands–Resources model of burnout. Journal of Applied Psychology 86(3), 499–512. Demerouti E., Bakker A.B., deJonge J., Janssen P.P.M. & Schaufeli W.B. (2001b) Burnout and engagement at work as a function of demands and control. Scandinavian Journal of Work, Environment & Health 27(4), 279–286. Demerouti E., Bakker A.B., Vardakou I. & Kantas A. (2003) The convergent validity of two burnout instruments. A multitrait– multimethod analysis. European Journal of Psychological Assessment 19(1), 12–23. Ekstedt M., So¨derstro¨m M., A˚kerstedt T., Nilsson J., Søndergaard H.-P. & Perski A. (2006) Disturbed sleep and fatigue in occupational burnout. Scandinavian Journal of Work Environment & Health 32(2), 121–131. Engstro¨m L.-M., Ekblom B., von Koch M. & Seger J. (1993). Lifestyle – Achievement Health Liv 90 (In Swedish) (No. 1). Folksam. Ho¨gskolan fo¨r La¨rarutbildning. Idrottsho¨gskolan. Karolinska Institutet. Korpen. Riksidrottsfo¨rbundet, Stockholm. Fusilier M. & Manning M.R. (2005) Psychosocial predictors of health status revisited. Journal of Behavioral Medicine 28(4), 347–358. Genco R.J., Ho A.W., Kopman J., Grossi S.G., Dunford R.G. & Tedesco L. (1998) Models to evaluate the role of stress in periodontal disease. Annals of Periodontology 3(1), 288– 302. Glass D.C. & McKnight J.D. (1996) Perceived control, depressive symptomatology, and professional burnout: a review of the evidence. Psychology and Health 11, 23–48. Gorter R.C., Eijkman M.A.J. & Hoogstraten J. (2000) Burnout and health among Dutch dentists. European Journal of Oral Sciences 108, 261–267. Grossi G., Perski A., Evenga˚rd B., Blomkvist V. & Orth-Gome´r K. (2003) Physiological correlates of burnout among women. Journal of Psychosomatic Research 55, 309–316.

 2008 The Authors. Journal compilation  2008 Blackwell Publishing Ltd

93

U. Peterson et al. Holgers K.M. (2003) Tinnitus treatment is guided by etiology. Noise, stress or anxiety/depression plausible causes. (In Swedish). La¨kartidningen 100(46), 3744–3749. Idler E.L. & Benyamini Y. (1997) Self-rated health and mortality: a review of twenty-seven community studies. Journal of Health and Social Behavior 38, 21–37. Johannsen A., Rylander G., So¨der B. & A˚sberg M. (2006) Dental plaque, gingival inflammation, and elevated levels of interleukin-6 and cortisol in gingival crevicular fluid from women with stressrelated depression and exhaustion. Journal of Periodontol 77, 1403–1409. Kaplan G.A., Goldberg D.E., Everson S.A., Cohen R.D., Salonen R., Tuomilehto J. & Salonen J. (1996) Perceived health status and morbidity and mortality: evidence from the Kuopio ischaemic heart disease risk factor study. International Journal of Epidemiology 25(2), 259–265. Kilfedder C.J., Power K.G. & Wells T.J. (2001) Burnout in psychiatric nursing. Journal of Advanced Nursing 34(3), 383– 396. Lee R.T. & Ashforth B.E. (1990) On the meaning of Maslach’s three dimensions of burnout. Journal of Applied Psychology 75(6), 743– 747. Lee R.T. & Ashforth B.E. (1996) A meta-analytic examination of the correlates of the three dimensions of job burnout. Journal of Applied Psychology 81, 123–133. Leiter M.P. & Schaufeli W.B. (1996) Consistency of the burnout construct across occupations. Anxiety, Stress and Coping 9, 229– 243. Licht M. (1995) Multiple regression and correlation. In Reading and Understanding Multivariate Statistics (Grimm L. & Yarnold P., eds), American Psychological Association, Washington, DC, pp. 19–64. Lisspers J., Nygren A˚. & So¨derman E. (1997) Hospital Anxiety and Depression Scale (HAD): some psychometric data for a Swedish sample. Acta Psychiatrica Scandinavia 96(4), 281–286. Maslach C. (2001) What have we learned about burnout and health? Psychology and Health 16, 607–611. Maslach C. (2003) Job burnout: new directions in research and intervention. Current Directions in Psychological Science 12(5), 189–192. Maslach C. & Goldberg J. (1998) Prevention of burnout: new perspectives. Applied & Preventive Psychology 7, 63–74. Maslach C., Jackson S.E. & Leiter M.P. (1996) Maslach Burnout Inventory Manual, 3rd edn. Consulting Psychologists Press, Palo Alto, CA. Maslach C., Schaufeli W.B. & Leiter M.P. (2001) Job burnout. Annual Reviews Psychology 52, 397–422. Melamed S., Ugarten U., Shirom A., Kahana L., Lerman Y. & Froom P. (1999) Chronic burnout, somatic arousal and elevated salivary cortisol levels. Journal of Psychosomatic Research 46(6), 591–598. Melamed S., Shirom A., Toker S., Berliner S. & Shapira I. (2006) Burnout and risk of cardiovascular disease: evidence, possible causal paths, and promising research directions. Psychological Bulletin 132(3), 327–353. National Board of Health and Welfare (2003). Exhaustion disorder. [In Swedish. Utmattningssyndrom – Stressrelaterad psykisk oha¨lsa] [Website: http://www.socialstyrelsen.se].

94

National Social Insurance Board (2003). Psychosocial Work Environment & Long Term Sick-Leave. (In Swedish). Retrieved from http://www.forsakringskassan.se on 15 May 2004. National Social Insurance Board (2006). Statistics. (In Swedish). Retrieved from http://www.forsakringskassan.se on 15 February 2007. Nilsson L.-G., Ba¨ckman L., Erngrund K., Nyberg L., Adolfsson R., Bucht G., Karlsson S., Widing M. & Winblad B. (1997) The Betula prospective cohort study: memory, health, and aging. Aging, Neuropsychology, and Cognition 4(1), 1–32. O’Brien R.M. (2007) A caution regarding rules of thumb for variance inflation factors. Quality & Quantity. Springer, doi: 10.1007/ s11135-006-9018-6. Paterniti S., Niedhammer I., Lang T. & Consoli S.M. (2002) Psychosocial factors at work, personality traits and depressive symptoms. Longitudinal results from the GAZEL Study. British Journal of Psychiatry 181, 111–117. ˚. Peterson U., Demerouti E., Bergstro¨m G., A˚sberg M. & Nygren A (2008) Work characteristics and sickness absence in burnout and non burnout groups: a study of Swedish health care workers. International Journal of Stress Management, accepted for publication. Pines A.M. & Keinan G. (2005) Stress and burnout: the significant difference. Personality and individual differences 39, 625–635. Podsakoff M., MacKenzie S., Lee J.Y. & Podsakoff N. (2003) Common method biases in behavioural research: a critical review of the literature and recommended remedies. Journal of Applied Psychology 88(5), 879–903. Price J.L. & Mueller C.W. (1986) Handbook of Organizational Measurement. Pittman, Marshfield, MA. Rydmark I., Wahlberg K., Ghatan P.H., Modell S., Nygren A˚., Ingvar M., A˚sberg M. & Heilig M. (2006) Neuroendocrine, cognitive and structural imaging characteristics of women on longterm sickleave with job stress-induced depression. Biological Psychiatry 60, 867– 873. Sandstro¨m A., Rhodin I.N., Lundberg M., Olsson T. & Nyberg L. (2005) Impaired cognitive performance in patients with chronic burnout syndrome. Biological Psychology 69(3), 271–279. Saunders J.B., Aasland O.G., Babor T.F., de la Fuente J.R. & Grant M. (1993) Development of the Alcohol Use Disorders Identification Test (AUDIT): WHO collaborative project on early detection of persons with harmful alcohol consumption. Addiction 88, 791– 804. Schaufeli W. & Enzmann D. (1998) The Burnout Companion to Study and Practice: A Critical Analysis. Taylor & Francis Ltd, London. Schaufeli W.B., Bakker A.B., Hoogduin K., Schaap C. & Kladler A. (2001) On the clinical validity of the Maslach Burnout Inventory and the burnout measure. Psychology and Health 16, 565–582. Shanafelt T.D., Bradley K.A., Wipf J.E. & Back A.L. (2002) Burnout and self-reported patient care in an internal medicine residency program. Annals of Internal Medicine 136, 358–367. Shirom A. (2002) Job related burnout: a review. In Handbook of Occupational Health Psychology (Quick J.C. & Tetrick L.E., eds), American Psychological Association, Washington, DC, pp. 245– 264. Shirom A., Melamed S., Toker S., Berliner S. & Shapira I. (2005) Burnout, mental and physical health: a review of the evidence and

 2008 The Authors. Journal compilation  2008 Blackwell Publishing Ltd

JAN: ORIGINAL RESEARCH a proposed explanatory model. International review of Industrial and Organizational Psychology 20, 269–309. Smith K., Davy J. & Everly Jr G. (2006) An assessment of the construct distinctiveness of stress arousal and burnout. Psychological Reports 99, 396–406. Spector P. (2006) Method variance in organizational research Truth or urban legend? Organizational Research Methods 9(2), 221–232. Stansfeld S.A., Fuhrer R., Head J., Ferrie J. & Shipley M. (1997) Work and psychiatric disorder in the Whitehall II Study. Journal of Psychosomatic Research 43(1), 73–81. Sullivan M., Karlsson J. & Ware J.E. (1994) SF-36 Ha¨lsoenka¨t: Svensk Manual och Tolkningsguide (Swedish Manual and Interpretation Guide). Sahlgrenska University Hospital, Gothenburg. Sullivan M., Karlsson J. & Ware Jr J.E. (1995) The Swedish SF-36 Health Survey. Evaluation of data Quality, scaling assumptions, reliability and construct validity across general populations in Sweden. Social Science & Medicine 41(10), 1349–1358.

Burnout and physical and mental health Tennant C. (2001) Work-related stress and depressive disorders. Journal of Psychosomatic Research 51, 697–704. Theorell T. (2006). I spa˚ren av 90-talet. (Consequences of the 1990’ths in Sweden) (In Swedish). Karolinska Institutet University Press, Stockholm. Toppinen-Tanner S., Ojaja¨rvi A., Va¨a¨na¨nen A., Kalimo R. & Ja¨ppinen P. (2005) Burnout as a predictor of medically certified sick-leave absence and their diagnosed causes. Behavioral Medicine 31, 18–27. Wang J. (2005) Work stress as a risk factor for major depressive episode(s). Psychological Medicine 35(6), 865–871. Ylipaavalniemi J., Kivima¨ki M., Elovainio M., Virtanen M., Keltikangas-Ja¨rvinen L. & Vahtera J. (2005) Psychosocial work characteristics and incidence of newly diagnosed depression: a prospective cohort study of three different models. Social Science & Medicine 61, 111–122. Zigmond A.S. & Snaith R.P. (1983) The Hospital Anxiety and Depression Scale. Acta Psychiatrica Scandinavia 67, 361–370.

 2008 The Authors. Journal compilation  2008 Blackwell Publishing Ltd

95