American Journal of Epidemiology Copyright 2006 by the Johns Hopkins Bloomberg School of Public Health All rights reserved;

; printed in U.S.A.

Vol. 163, No. 10 DOI: 10.1093/aje/kwj119 Advance Access publication March 29, 2006

Original Contribution Psychosocial Work Environment and Incidence of Severe Depressive Symptoms: Prospective Findings from a 5-Year Follow-up of the Danish Work Environment Cohort Study

ltmann, Birgit Aust, and Hermann Burr Reiner Rugulies, Ute Bu

From the National Institute of Occupational Health, Copenhagen, Denmark. Received for publication June 22, 2005; accepted for publication October 10, 2005.

The authors analyzed the impact of psychosocial work characteristics on the incidence of severe depressive symptoms among 4,133 (49% women) employees from a representative sample of the Danish workforce between 1995 and 2000. Psychosocial work characteristics at baseline included quantitative demands, inuence at work, possibilities for development, social support from supervisors and coworkers, and job insecurity. Severe depressive symptoms were measured with the ve-item Mental Health Inventory of the 36-item Short-Form Health Survey, with a cutoff point of 52. Women with low inuence at work (relative risk (RR) 2.17, 95% condence interval (CI): 1.23, 3.82) and low supervisor support (RR 2.03, 95% CI: 1.20, 3.43) were at increased risk for severe depressive symptoms after exclusion of cases at baseline and adjustment for sociodemographic factors, baseline depression score, and health behaviors. Further adjustments for socioeconomic position did not change the result substantially. Additional analyses showed that a one-standard deviation increase on the inuence scale resulted in a 27% decreased risk of severe depressive symptoms. Among men, job insecurity predicted severe depressive symptoms (RR 2.04, 95% CI: 1.02, 4.07). The ndings indicate that the work environment inuences the risk of developing severe depressive symptoms and that different factors play a role for men and women. anxiety; depression; longitudinal studies; mental health; social support; stress, psychological; workplace

Abbreviations: CI, condence interval; DWECS, Danish Work Environment Cohort Study; MHI-5, ve-item Mental Health Inventory of the 36-item Short-Form Health Survey; RR, relative risk.

Editors note: An invited commentary on this article appears on page 888, and the authors response appears on page 891.

Depression is a major burden of disease and includes such diagnoses as unipolar (major) depression, dysthymia, or depressive episodes, according to the fourth edition of the Diagnostic and Statistical Manual of Mental Health Disorders (1) and the International Classication of Diseases, Tenth Revision (2). The World Health Organization reports

that the point prevalence for unipolar depression is 3.2 percent for women and 1.9 percent for men, with a 12-month prevalence of 9.5 percent for women and 5.8 percent for men (3). Unipolar depression is the fourth leading cause of disability-adjusted life years (3, 4) and the leading cause for years of life lived with disability in men and women in the world (3). Depression is not only adverse in itself but also triggers severe negative consequences. On the societal level, depression contributes to long-term sickness absence and early retirement (5, 6). On the individual level, depression might lead to life-threatening diseases. Two recent meta-analyses

105, DK-2100 Copenhagen, Denmark Correspondence to Dr. Reiner Rugulies, National Institute of Occupational Health, Lers Parkalle (e-mail: rer@ami.dk).

877

Am J Epidemiol 2006;163:877887

878 Rugulies et al.

found that depression was an independent risk factor for myocardial infarction and death due to coronary heart disease in initially healthy people (7, 8). Numerous studies have investigated whether psychosocial work characteristics are associated with a higher probability of major depression and depressive symptoms; for an overview, refer to the article by Tennant entitled, Workrelated Stress and Depressive Disorders (9). However, most of these studies are cross-sectional in design and therefore do not allow causal inference. Only a few studies have utilized prospective data so far. Using components of the demand-control-support model, that is, high psychological demands, low decision authority, and low social support at work (10, 11), in a French cohort with 11,552 employees of an electricity and gas company, Niedhammer et al. (12) found that these components predicted the onset of depression over a 12-month follow-up. Stansfeld et al. (13) found, in a 5-year follow-up with 10,308 British civil servants, that demands, control, and support, as well as a high effortreward imbalance (14), predicted the incidence of minor psychiatric disorders (which were identied as mainly depressive symptoms) in men and women. Although these and other studies with selected occupational groups provide insight into the association between the psychosocial work environment and depression, to our knowledge, no prospective study has yet been conducted that investigated this association in the general working population. This paper aims to ll this void by analyzing the impact of psychosocial work characteristics on the incidence of severe depressive symptoms in a 5-year prospective study with a representative sample of the Danish workforce.
MATERIALS AND METHODS Study sample

the variables of interest and 104 participants who had shown severe depressive symptoms in 1995, resulting in a nal study sample of 4,133 participants (gure 1).
Measurement of severe depressive symptoms

Severe depressive symptoms at baseline and at follow-up were assessed with the Danish version of the ve-item Mental Health Inventory (MHI-5) of the 36-item Short-Form Health Survey (1619). Although the scale is meant to measure mental health in general, a review of the literature showed that the instrument seems to be most appropriate for measuring severe depressive symptoms (16, 2022). The MHI-5 consists of ve items on the frequency of depressive symptoms in the past 4 weeks. A complete list of the items is provided in the Appendix. Scores were summed up and standardized, yielding a scale with scores ranging from 0 to 100, with higher scores indicating fewer depressive symptoms. In accordance with studies from others (20, 23), we classied participants scoring 52 points or less as cases of severe depressive symptoms.
Measurement of psychosocial work characteristics

The analyses are based on data from the Danish Work Environment Cohort Study (DWECS). In 1990, a random sample of 9,653 Danish residents was drawn from the Central Population Register of Denmark and interviewed by phone on various aspects of work and health, including the application of standardized scales on psychosocial work characteristics. In 1995 and 2000, the same people were contacted again, and new participants additionally supplemented the cohort. A more detailed description of DWECS has been published elsewhere (15). Depressive symptoms were measured in 1995 and 2000 but not in the 1990 survey. Therefore, analyses in this paper are based on baseline variables in 1995 and the incidence of severe depressive symptoms in 2000. In 1995, 10,702 Danish residents were approached to participate in DWECS. Of these, 8,583 participated in the survey (response rate: 80 percent). Among the respondents, 5,423 were employed at the time of the survey and therefore eligible for the study. Of these, 11 emigrated or died during the follow-up period, and 942 did not respond to the follow-up questionnaire, yielding a follow-up sample of 4,470 (response rate: 83 percent; estimated combined response rate: 0.80 3 0.83 66 percent). For the purpose of the analyses in this paper, we further excluded 233 participants with missing values on

We measured six psychosocial work characteristics: quantitative demands, inuence at work, possibilities for development, social support from supervisors, social support from coworkers, and job insecurity. All six factors are well recognized as important psychosocial exposures at work and have been used in several other studies (9, 12, 13, 2428). However, at the time of the baseline assessment, no reliable Danish instrument measuring these factors existed, and therefore the items and the scales for this study were constructed by a research group at the National Institute of Occupational Health in Denmark (15). The items and response categories of the psychosocial work characteristics are provided in the Appendix. Inuence at work and possibilities for development were measured with scales consisting of three items each. The scales were built by summing up the numerical values of the response options, with higher values indicating higher inuence and higher possibilities for development, respectively. Respondents scoring below the median on the two scales were categorized as exposed to low inuence and low possibilities for development, respectively. Quantitative demands, social support from supervisors, and social support from coworkers were measured with single items. We asked participants if the amount of their work is so extensive that they do not have time to think and talk about anything else than work. Participants who answered that this is true for at least half of the working hours were categorized as having high quantitative demands. Regarding social support, participants were asked if they receive support and encouragement from their supervisors and from their coworkers. Respondents who answered usually not, never, or I do not have a supervisor/I do not have coworkers were categorized as exposed to low supervisor support and low coworker support, respectively. Job insecurity was measured with four dichotomous (yes/ no) items asking the participants if they were worried about
Am J Epidemiol 2006;163:877887

Psychosocial Work Environment and Depressive Symptoms

879

any of the following situations: 1) becoming unemployed; 2) being transferred to another job against their will; 3) becoming redundant as a result of new technology; and 4) having difculty in nding another job if they became unemployed. Respondents who answered yes to at least one of the four items were categorized as exposed to job insecurity.
Measurement of covariates

We recorded the covariates age, gender, cohabitation, number of children living at home, highest level of school education, change in employment status, health behaviors, and socioeconomic position. We included these variables, because other studies have found them to be related with psychosocial work characteristics, mental health, or both (2935). We combined cohabitation and children living at home into the variable family status, with the following categories: 1) cohabiting with child; 2) cohabiting without child; 3) single with child; and 4) single without child. The highest level of school education was categorized in accordance with the Danish educational system, yielding three groups: 1) left school without completing the 10-year examination; 2) completed the 10-year examination; and 3) earned advanced education. The variable change in employment status indicated if a respondent had experienced any of the following circumstances: 1) had been at the same workplace between baseline and the follow-up survey; 2) had changed workplaces during this period; or 3) was no longer employed at follow-up. Health behaviors included smoking, alcohol consumption, and leisure-time physical activity. Whereas smoking was assessed in the 1995 baseline survey, alcohol consumption and leisure-time physical activity had to be derived from the 2000 follow-up survey. Smoking was dichotomized as current smokers and nonsmokers. Alcohol consumption was categorized by three groups: 1) no consumption of alcohol, 2) moderate consumption, and 3) heavy consumption, with heavy consumption dened as drinking more than two and three units per day, respectively, for women and men. A unit was dened as one small bottle of beer (0.33 liter), one glass of wine, or one shot of liquor. With four response categories corresponding to sedentary, light, moderate, and strenuous physical activity, leisure-time physical activity was assessed with the question, When you should describe your leisure-time physical activity in the last year, including commuting to or from work, to what group do you belong? Socioeconomic position was dened by employment grade, job title, and education, yielding ve categories: I (executives and/or academics); II (middle managers and/ or having more than 34 years of further education); III (other white collar workers); IV (skilled blue collar workers); and V (semi- or unskilled blue collar workers).
Statistical analysis

Station, Texas). Internal consistency of the depression scale and the work environment scales was estimated with Cronbachs alpha. Correlations between depression scores at baseline and follow-up were calculated with Pearsons correlation coefcients. Using the six categorical psychosocial workplace characteristics as predictors and the incidence of severe depressive symptoms as the outcome, we calculated gender-stratied relative risks and 95 percent condence intervals with multivariate logistic regression models. In addition, we used the two scales on inuence at work and possibilities for development as continuous variables to assess the relative risk for severe depressive symptoms for each one-standard deviation change of the scales. Covariates were included in three different models. Model 1 was adjusted for age, family status, school education, change in employment status, and depression score at baseline. Model 2 was additionally adjusted for health behaviors, that is, smoking, alcohol consumption, and leisure-time physical activity. Model 3 was further adjusted for socioeconomic position. In an additional analysis, we repeated model 2 for a study sample of 4,237 participants that also included the 104 persons with severe depressive symptoms at baseline.

RESULTS

All analyses were conducted with the STATA, version 7.0, statistical program package (Stata Corporation, College
Am J Epidemiol 2006;163:877887

Figure 1 gives an illustration of the exclusion process and shows the percentage of participants with severe depressive symptoms in 1995 and 2000 at each step. The percentages of cases decreased throughout the exclusion process. Caseness of severe depressive symptoms in 2000 was found among 3.9 percent of the 8,583 responders to the 1995 baseline survey but only among 2.5 percent of the 4,133 persons in the nal study sample. This reduction was caused mainly by excluding persons who were not employees in 1995. For example, among the 3,160 nonemployees in 1995, caseness of severe depressive symptoms was 8.1 percent in 1995 and 5.7 percent in 2000. The highest prevalence rates with 10.4 percent in 1995 and 7.1 percent in 2000 were found among the 2,348 persons who were out of the labor market (not shown in the gure). The mean age of the sample was 39 years with a standard deviation of 11 years. Slightly more than one fourth had an advanced school education, and the majority worked as white collar workers (table 1). Depression scores in 1995 and in 2000 were positively correlated with a correlation coefcient of 0.31 (p < 0.001). The mean score on the depression scale at follow-up was 87.06 with a standard deviation of 12.37. The score was not normally distributed but heavily skewed to the right (i.e., in the direction of fewer depressive symptoms), with 84 percent of the participants scoring 80 points or more. There were 105 participants (68 women and 37 men) with severe depressive symptoms at follow-up, yielding an incidence rate of 2.5 (3.4 for women and 1.7 for men). Cronbachs alpha values for the inuence at work and the possibilities for development scales were 0.53 and 0.64, respectively. Cronbachs alpha for the depression scales was 0.78 in 1995 and 0.81 in 2000.

880 Rugulies et al.

FIGURE 1. Illustration of the exclusion process in the determination of the nal study sample and the percentage of participants with severe depressive symptoms in 1995 and 2000 at each step, Danish Work Environment Cohort Study (DWECS).

Table 2 shows the relative risks for severe depressive symptoms for psychosocial work characteristics among women. Low inuence at work (relative risk (RR) 2.17, 95 percent condence interval (CI): 1.23, 3.82) and low social support from supervisors (RR 2.03, 95 percent CI: 1.20, 3.43) were predictive for the onset of severe depressive symptoms, when adjusted for sociodemographic factors, health behaviors, and depression score at baseline (model 2). When socioeconomic position was added

(model 3), effect estimates decreased somewhat more but remained substantial (for low inuence: RR 1.96, 95 percent CI: 1.10, 3.47; for low supervisor support: RR 1.92, 95 percent CI: 1.13, 3.26). There were no effects for quantitative demands, possibilities for development, social support from coworkers, and job insecurity. When we analyzed inuence at work and possibilities for development as continuous variables (data not shown in table), we a found a dose-response relation between increasing
Am J Epidemiol 2006;163:877887

Psychosocial Work Environment and Depressive Symptoms

881

TABLE 1. Sociodemographic characteristics of the 4,133 Danish Work Environment Cohort Study participants in 1995 and the incidence rate of severe depressive symptoms in 2000
No. %

Gender Women Men Age (years) <35 3544 4554 55 Family status Cohabiting with child Cohabiting without child Single with child Single without child School education Left school without completing 10-year examination Completed 10-year examination Advanced education Socioeconomic position I. Executives/academics II. Middle managers/white collar workers with further education III. Other white collar workers IV. Skilled blue collar workers V. Semi- or unskilled blue collar workers Incidence of severe depressive symptoms at follow-up in 2000 Yes No 105 4,028 2.5 97.5 564 664 1,548 482 875 13.6 16.1 37.4 11.7 21.2 1,483 1,576 1,074 35.9 38.1 26.0 1,976 1,247 178 732 47.8 30.2 4.3 17.7 1,568 1,171 1,004 390 37.9 28.3 24.3 9.4 2,004 2,129 48.5 51.5

Repeating the analyses of model 2 for a study sample also including persons with severe depressive symptoms at baseline did not change the effect estimates substantially for most variables. Changes of relative risks of 10 percent or more were observed among women for inuence at work (from 2.17 to 2.41 when cases at baseline were included) and possibilities for development (from 1.11 to 1.30) and among men for supervisor support (from 1.17 to 1.46) and coworker support (from 1.29 to 1.45).

DISCUSSION

inuence at work and decreasing risk of developing severe depressive symptoms. After adjustment for the covariates from model 2, a one-standard deviation increase on the inuence at work scale resulted in a 27 percent decreased risk of severe depressive symptoms (RR 0.73, 95 percent CI: 0.59, 0.91; p 0.006). Additional adjustment for socioeconomic position resulted in a relative risk of 0.80 (95 percent CI: 0.63, 1.01; p 0.06). For possibilities for development, no statistically signicant effect was found. Table 3 shows the relative risks of severe depressive symptoms for psychosocial work characteristics among men. Job insecurity was predictive in all three models (model 2: RR 2.04, 95 percent CI: 1.02, 4.07; model 3: RR 2.09, 95 percent CI: 1.04, 4.20). No statistically signicant associations were found for quantitative demands, inuence at work, possibilities for development, and social support from supervisors and from coworkers. Inuence at work and possibilities for development as continuous variables were also unrelated to the risk of severe depressive symptoms (data not shown in table).
Am J Epidemiol 2006;163:877887

Psychosocial work characteristics were prospectively associated with the development of severe depressive symptoms in this representative sample of the Danish workforce. Among women, low inuence at work and low social support from supervisors resulted in a twofold increased risk. Moreover, low inuence at work showed a dose-response relation with risk of severe depressive symptoms. These results are in line with other prospective studies that found low control at work and low social support predictive for depression, psychiatric disorders, and psychological distress (12, 13, 28). Among men, the association between psychosocial work characteristics and severe depressive symptoms was remarkably different. Neither inuence at work nor social support from supervisors was associated with severe depressive symptoms. Job insecurity, on the other hand, was predictive, which is in line with ndings from Ferrie et al. (36) that job insecurity is a risk factor for minor psychiatric disorders and depression. However, Ferrie et al. found job insecurity to be predictive in both men and women, whereas in our study the effect of job insecurity was restricted to men. The different patterns of association between psychosocial work characteristics and severe depressive symptoms for men and women in our study underline the importance of conducting gender-stratied analyses when studying depression. Whether these gender differences may be partly explained by differences in the structure of the occupations and thus different work characteristics for men and women (37) or by differences in the perception of or vulnerability to work conditions (38) or whether they emerge because of a different time course of psychosocial work characteristics in the onset of depressive symptoms cannot yet be determined. Certainly, these gender differences merit further exploration.
Strengths of the study

To our knowledge, this is the rst time that prospective analyses on the impact of psychosocial work characteristics on severe depressive symptoms were conducted in a representative sample of the general working population. The prospective nature of the study allows causal inference, and the use of the representative sample allows generalizing the ndings for the Danish workforce. In the analyses, we excluded persons with severe depressive symptoms at baseline and also adjusted for baseline

882 Rugulies et al.

TABLE 2. Impact of psychosocial work characteristics on the incidence of severe depressive symptoms in 2,004 employed women, Danish National Work Environment Cohort Study, 19952000
Incident severe depressive symptoms No. % Relative risk Model 1* 95% condence interval Relative risk Model 2y 95% condence interval Relative risk Model 3z 95% condence interval

Psychosocial work characteristics

Exposed (no.)

Quantitative demands Low High Inuence at work High Low Possibilities for development High Low Social support from supervisor High Low Social support from coworkers High Low Job insecurity No Yes 1,129 875 33 35 2.92 4.00 1.00 1.21 0.73, 1.99 1.00 1.21 0.73, 1.99 1.00 1.04 0.62, 1.74 1,803 201 59 9 3.27 4.48 1.00 1.07 0.51, 2.25 1.00 1.05 0.50, 2.23 1.00 0.98 0.46, 2.11 1,611 393 43 25 2.67 6.36 1.00 2.05 1.22, 3.46 1.00 2.03 1.20, 3.43 1.00 1.92 1.13, 3.26 961 1,043 28 40 2.91 3.84 1.00 1.14 0.68, 1.91 1.00 1.11 0.66, 1.87 1.00 0.86 0.49, 1.50 947 1,057 18 50 1.90 4.73 1.00 2.23 1.27, 3.92 1.00 2.17 1.23, 3.82 1.00 1.96 1.10, 3.47 1,402 602 48 20 3.42 3.32 1.00 0.80 0.46, 1.39 1.00 0.83 0.48, 1.44 1.00 0.97 0.55, 1.70

* Model 1 was adjusted for age (continuous), family status (cohabiting and having children), school education, change in employment status during follow-up, and depression score at baseline (continuous). y Model 2 was adjusted for all the variables in model 1 plus smoking (at baseline), alcohol consumption (at follow-up), and leisure-time physical activity (at follow-up). z Model 3 was adjusted for all the variables in model 2 plus socioeconomic position. Referent.

depression score to reduce the possibility that depressive symptoms affected the perception of psychosocial work characteristics. We further adjusted for other potential confounders, including family status, health behaviors (model 2), and socioeconomic position (model 3). A priori, we viewed model 2 as the most appropriate model. As shown in Materials and Methods, socioeconomic position was based mainly on occupational status. It has been argued that low occupational status increases the likelihood of exposure to adverse psychosocial work characteristics, which then increases the risk of disease (3942). Consequently, researchers in the eld of work and health have pointed out that controlling for occupational status when analyzing associations between psychosocial work characteristics and health outcomes could cause overadjustment (43, 44). We nd this argument convincing and therefore believe that the analysis in model 2 is the most appropriate. However, we also showed the ndings for model 3 in order to address recent criticism that associations between psychosocial work characteristics and health outcomes are confounded by more favorable psychosocial exposure patterns among people of higher socioeconomic position (4547).

Limitations of the study

A limitation of this study is that severe depressive symptoms were not assessed with the gold standard, a diagnostic interview, but with a questionnaire. This questionnaire, the MHI-5, was developed to measure mental health and has been used as a measurement of both mental health in general and depression. After a review of the literature, we concluded that the MHI-5 is an acceptable instrument to measure severe depressive symptoms. Rumpf et al. (21) validated the MHI-5 against clinical diagnoses with a diagnostic interview and found that the sensitivity and specicity of the MHI-5 were most satisfactory for mood disorders (which includes major depression and dysthymia), with an area under the receiver operating characteristics curve of 0.88. Anxiety disorders, somatoform disorders, and substance use disorders had considerably lower receiver operating characteristics curves of 0.71, 0.65, and 0.65, respectively. Strand et al. (20) and Berwick et al. (16) compared the MHI-5 with several other diagnostic instruments, and both concluded that the MHI-5 is a valid instrument to measure depressive disorders.
Am J Epidemiol 2006;163:877887

Psychosocial Work Environment and Depressive Symptoms

883

TABLE 3. Impact of psychosocial work characteristics on the incidence of severe depressive symptoms in 2,129 employed men, Danish National Work Environment Cohort Study, 19952000
Incident severe depressive symptoms No. % Relative risk Model 1* 95% condence interval Relative risk Model 2y 95% condence interval Relative risk Model 3z 95% condence interval

Psychosocial work characteristics

Exposed (no.)

Quantitative demands Low High Inuence at work High Low Possibilities for development High Low Social support from supervisor High Low Social support from coworkers High Low Job insecurity No Yes 1,331 798 15 22 1.13 2.76 1.00 2.04 1.02, 4.06 1.00 2.04 1.02, 4.07 1.00 2.09 1.04, 4.20 1,803 326 27 10 1.50 3.07 1.00 1.33 0.61, 2.92 1.00 1.29 0.59, 2.84 1.00 1.26 0.57, 2.82 1,519 610 22 15 1.45 2.46 1.00 1.20 0.60, 2.40 1.00 1.17 0.58, 2.35 1.00 1.15 0.57, 2.32 1,136 993 15 22 1.32 2.22 1.00 1.18 0.58, 2.39 1.00 1.16 0.57, 2.37 1.00 1.26 0.59, 2.67 1,074 1,055 19 18 1.77 1.71 1.00 0.61 0.30, 1.23 1.00 0.60 0.30, 1.22 1.00 0.60 0.29, 1.24 1,628 501 31 6 1.90 1.20 1.00 0.47 0.18, 1.19 1.00 0.46 0.18, 1.19 1.00 0.48 0.19, 1.25

* Model 1 was adjusted for age (continuous), family status (cohabiting and having children), school education, change in employment status during follow-up, and depression score at baseline (continuous). y Model 2 was adjusted for all the variables in model 1 plus smoking (at baseline), alcohol consumption (at follow-up), and leisure-time physical activity (at follow-up). z Model 3 was adjusted for all the variables in model 2 plus socioeconomic position. Referent.

On the basis of these ndings, we are condent that the MHI-5 is an acceptable instrument to identify persons with severe depressive symptoms. However, the literature on the MHI-5 does not provide cutoff points for identifying mild and moderate depressive symptoms, and we therefore did not conduct analyses of these less severe, albeit important, forms of depressive disorders. Furthermore, it has to be noted that the MHI-5 scale covers exclusively affective symptoms and mood but not disturbance in daily function, which is an important criterion in assessing depression in psychiatric interviews. Consequently, we did not use the terms clinical depression, unipolar depression, or major depression for caseness in our study but used the more general term severe depressive symptoms. We examined the relation between psychosocial work characteristics and the onset of severe depressive symptoms by use of a 5-year follow-up period. It should be taken into account that, within the follow-up period, changes may have occurred in both exposure and outcome variables. With regard to the exposure measures, this would mean that, for some participants, psychosocial work characteristics measured at baseline would differ from the exposure during some
Am J Epidemiol 2006;163:877887

phases of the follow-up period. This would be a nondifferential misclassication in the measurement of the exposure variable, which would result in a potential underestimation of the effect sizes. With regard to the outcome variable, it is possible that some participants developed severe depressive symptoms after the 1995 baseline survey but were no longer cases when we conducted the follow-up survey in 2000. It has further to be noted that we had a relatively low number of incident cases. Consequently, some of the condence intervals were wide, especially among men, increasing the probability of a type II error (i.e., we might have failed to detect an effect for some exposure variables because of lack of statistical power). For example, high quantitative demands among men showed a relative risk of 0.48, indicating an unexpected protective effect, which might have been statistically signicant if the numbers of cases would have been higher. Similar to most research on working conditions and health, our study is prone to selection bias because of the healthy worker effect (48). As shown in gure 1, people out of the labor market were at higher risk for severe depressive symptoms. Although it is reasonable to assume that a certain

884 Rugulies et al.

proportion of these cases were not participating in the workforce because of depression, another proportion of these cases might have developed depression and left the workforce because of previous exposure to adverse psychosocial work characteristics. Two covariates, alcohol consumption and leisure-time physical activity, were not measured at baseline but at follow-up only. By including these variables in the analyses, we might have biased the results toward the null hypothesis, because alcohol consumption and leisure-time physical activity measured in 2000 might have been inuenced by the development of severe depressive symptoms during followup. We were aware of this possible bias; however, we believed it would be more important to address the potential confounding effects of these two important health-related behaviors and therefore included them in the analyses. Studies analyzing the associations between self-reported exposure to psychosocial work characteristics and selfreported measures of health are generally prone to what has been called bias due to common method or triviality trap (24, 49, 50). This means that statistical associations between exposure and outcome might be the result of poor healths causing a more negative view of the work environment or of a general negative response style that includes the self-assessment of both work environment and health. In this study, however, we controlled for this bias, by adjusting for baseline depression scores in the analyses. Finally, our study is limited with regard to the inclusiveness of the constructs we used to measure the psychosocial work environment. Although we used well-established constructs, recent research indicates that a more comprehensive assessment of the psychosocial work environment is needed. There is accumulating evidence that the imbalance between high efforts and low reward at work increases the risk of ill health, including mental health disorders (14, 51, 52). It has further been argued that concepts such as meaning of work and predictability of work can contribute to a better understanding of the psychosocial work environment (53, 54). There is also an ongoing debate on the need for an expansion of the psychological demands concept, including a differentiation in quantitative, cognitive, sensorial, and emotional demands (55, 56) and a distinction between intensive (high work pace) and extensive (long working hours) quantitative demands (57). Most of these new developments and discussions emerged after the 1995 survey of DWECS. Fortunately, we were able to address several of these issues in the 2000 survey by including substantially expanded and revised psychosocial work environment measurements (15). When new follow-up data on depressive symptoms are available from the DWECS 2005 survey, we will be able to conduct analyses with a more comprehensive assessment of the psychosocial work environment.

women. Reviews indicate that intervention strategies to improve the psychosocial work environment are available and effective (5860). Hence, improving the psychosocial work environment might be an important tool for reducing severe depressive symptoms.

ACKNOWLEDGMENTS

Conict of interest: none declared.

REFERENCES 1. American Psychiatric Association. Diagnostic and statistical manual of mental health disorders: DSM-IV-TR. 4th ed, text revision. Arlington, VA: American Psychiatric Association, 2000. (http://www.psych.org/research/dor/dsm/index.cfm). 2. World Health Organization. International classication of diseases. Tenth Revision, 2005 update. Geneva, Switzerland: World Health Organization, 2005. (http://www.who.int/ classications/icd/en/). 3. World Health Organization. The world health report 2001. Mental health: new understanding, new hope. Geneva, Switzerland: World Health Organization, 2001. 4. Murray CJL, Lopez AD, Mathers CD, et al. The global burden of disease 2000 project: aims, methods and data sources. Geneva, Switzerland: World Health Organization, 2002. (Paper 36). (http://www3.who.int/whosis/menu.cfm?pathwhosis, burden,burden_gbd2000&languageenglish). 5. Henderson M, Glozier N, Holland Elliott K. Long term sickness absence. BMJ 2005;330:8023. 6. Karpansalo M, Kauhanen J, Lakka TA, et al. Depression and early retirement: prospective population based study in middle aged men. J Epidemiol Community Health 2005;59: 704. 7. Rugulies R. Depression as a predictor for coronary heart disease. A review and meta-analysis. Am J Prev Med 2002;23: 5161. 8. Wulsin LR, Singal BM. Do depressive symptoms increase the risk for the onset of coronary disease? A systematic quantitative review. Psychosom Med 2003;65:20110. 9. Tennant C. Work-related stress and depressive disorders. J Psychosom Res 2001;51:697704. 10. Karasek R, Theorell T. Healthy work: stress, productivity, and the reconstruction of working life. New York, NY: Basic Books, 1990. 11. Theorell T, Karasek R. Current issues relating to psychological job strain and cardiovascular disease research. J Occup Health Psychol 1996;1:926. 12. Niedhammer I, Goldberg M, Leclerc A, et al. Psychosocial factors at work and subsequent depressive symptoms in the Gazel cohort. Scand J Work Environ Health 1998;24:197205. 13. Stansfeld SA, Fuhrer R, Shipley MJ, et al. Work characteristics predict psychiatric disorder: prospective results from the Whitehall II study. Occup Environ Med 1999;56:3027. 14. Siegrist J, Starke D, Chandola T, et al. The measurement of effort-reward imbalance at work: European comparisons. Soc Sci Med 2004;58:148399. 15. Burr H, Bjorner JB, Kristensen TS, et al. Trends in the Danish work environment in 19902000 and their associations with labor-force changes. Scand J Work Environ Health 2003;29: 2709. Am J Epidemiol 2006;163:877887

Conclusion

We conclude that psychosocial work characteristics contribute to the development of severe depressive symptoms in the Danish workforce. Different aspects of the psychosocial work environment seem to be important for men and

Psychosocial Work Environment and Depressive Symptoms

885

16. Berwick DM, Murphy JM, Goldman PA, et al. Performance of a ve-item mental health screening test. Med Care 1991;29: 16976. 17. Ware JE, Sherbourne CD. The MOS 36-Item Short-Form Health Survey (SF-36). I. Conceptual framework and item selection. Med Care 1992;30:47383. 18. Bjorner JB, Damsgaard MT, Watt T, et al. Tests of data quality, scaling assumptions, and reliability of the Danish SF-36. J Clin Epidemiol 1998;51:100111. 19. Bjorner JB, Thunedborg K, Kristensen TS, et al. The Danish SF-36 Health Survey: translation and preliminary validity studies. J Clin Epidemiol 1998;51:9919. 20. Strand BH, Dalgard OS, Tambs K, et al. Measuring the mental health status of the Norwegian population: a comparison of the instruments SCL-25, SCL-10, SCL-5, and MHI-5 (SF-36). Nord J Psychiatry 2003;57:11318. 21. Rumpf HJ, Meyer C, Hapke U, et al. Screening for mental health: validity of the MHI-5 using DSM-IV axis I psychiatric disorders as gold standard. Psychiatry Res 2001;105: 24353. 22. Grosch JW, Murphy LR. Occupational differences in depression and global health: results from a national sample of US workers. J Occup Environ Med 1998;40:15364. 23. Holmes WC. A short, psychiatric, case-nding measure for HIV seropositive outpatients: performance characteristics of the 5-item mental health subscale of the SF-20 in a male, seropositive sample. Med Care 1998;36:23743. 24. Kasl SV. Measuring job stressors and studying the health impact of the work environment: an epidemiologic commentary. J Occup Health Psychol 1998;3:390401. 25. Hurrell JJ Jr, Nelson DL, Simmons BL. Measuring job stressors and strains: where we have been, where we are, and where we need to go. J Occup Health Psychol 1998;3:36889. 26. Belkic KL, Landsbergis PA, Schnall PL, et al. Is job strain a major source of cardiovascular disease risk? Scand J Work Environ Health 2004;30:85128. 27. Steenland K, Fine L, Belkic K, et al. Research ndings linking workplace factors to CVD outcomes. Occup Med 2000; 15:768. 28. Bu ltmann U, Kant IJ, Van den Brandt PA, et al. Psychosocial work characteristics as risk factors for the onset of fatigue and psychological distress: prospective results from the Maastricht Cohort Study. Psychol Med 2002;32:33345. 29. Lorant V, Deliege D, Eaton W, et al. Socioeconomic inequalities in depression: a meta-analysis. Am J Epidemiol 2003; 157:98112. 30. Stansfeld SA, Head J, Marmot MG. Explaining social class differences in depression and well-being. Soc Psychiatry Psychiatr Epidemiol 1998;33:19. 31. Allgo wer A, Wardle J, Steptoe A. Depressive symptoms, social support, and personal health behaviors in young men and women. Health Psychol 2001;20:2237. 32. Albertsen K, Hannerz H, Borg V, et al. Work environment and smoking cessation over a ve-year period. Scand J Public Health 2004;32:16471. 33. Olsen LR, Mortensen EL, Bech P. Prevalence of major depression and stress indicators in the Danish general population. Acta Psychiatr Scand 2004;109:96103. 34. Backmand H, Kaprio J, Kujala U, et al. Inuence of physical activity on depression and anxiety of former elite athletes. Int J Sports Med 2003;24:60919. 35. Haarasilta LM, Marttunen MJ, Kaprio JA, et al. Correlates of depression in a representative nationwide sample of adolescents (1519 years) and young adults (2024 years). Eur J Public Health 2004;14:2805. Am J Epidemiol 2006;163:877887

36. Ferrie JE, Shipley MJ, Stansfeld SA, et al. Effects of chronic job insecurity and change in job security on self reported health, minor psychiatric morbidity, physiological measures, and health related behaviours in British civil servants: the Whitehall II study. J Epidemiol Community Health 2002;56: 4504. 37. Lennon MC. Work conditions as explanations for the relation between socioeconomic status, gender, and psychological disorders. Epidemiol Rev 1995;17:1207. 38. Roxburgh S. Gender differences in work and well-being: effects of exposure and vulnerability. J Health Soc Behav 1996;37:26577. 39. Marmot MG. The social pattern of health and illness. In: Blane D, Brunner E, Wilkenson R, eds. Health and social organization: towards a health policy in the twenty-rst century. London, United Kingdom: Routledge, 1996:4267. 40. Rugulies R, Aust B, Syme SL. The epidemiology of health and illness. A socio-psycho-physiological perspective. In: Sutton S, Baum A, Johnston M, eds. The Sage handbook of health psychology. London, United Kingdom: Sage Publishers, 2004: 2768. 41. Syme SL, Balfour JL. Social determinants of disease. In: Wallace RB, ed. Maxcy-Rosenau-Last public health & preventive medicine. Stamford, CT: Appelton & Lange, 1998: 795810. 42. Marmot M, Wilkinson RG. Psychosocial and material pathways in the relation between income and health: a response to Lynch et al. BMJ 2001;322:12336. 43. Stansfeld SA, Bosma H, Hemingway H, et al. Psychosocial work characteristics and social support as predictors of SF-36 health functioning: the Whitehall II study. Psychosom Med 1998;60:24755. 44. North FM, Syme SL, Feeney A, et al. Psychosocial work environment and sickness absence among British civil servants: the Whitehall II study. Am J Public Health 1996;86: 33240. 45. Macleod J, Davey Smith G. Psychosocial factors and public health: a suitable case for treatment? J Epidemiol Community Health 2003;57:56570. 46. Macleod J, Davey Smith G, Heslop P, et al. Are the effects of psychosocial exposures attributable to confounding? Evidence from a prospective observational study on psychological stress and mortality. J Epidemiol Community Health 2001;55:87884. 47. Macleod J, Davey Smith G, Heslop P, et al. Psychological stress and cardiovascular disease: empirical demonstration of bias in a prospective observational study of Scottish men. BMJ 2002;324:124751. 48. Checkoway H, Pearce N, Crawford-Brown DJ. Research methods in occupational epidemiology. New York, NY: Oxford University Press, 1989. 49. Kristensen TS. Job stress and cardiovascular disease: a theoretic critical review. J Occup Health Psychol 1996;1: 24660. 50. Kasl SV. Methodologies in stress and health: past difculties, present dilemmas, future directions. In: Kasl SV, Cooper CL, eds. Stress and health: issues in research methodology. New York, NY: Wiley, 1987:30718. 51. von dem Knesebeck O, Siegrist J. Reported nonreciprocity of social exchange and depressive symptoms. Extending the model of effort-reward imbalance beyond work. J Psychosom Res 2003;55:20914. 52. Larisch M, Joksimovic L, von dem Knesebeck O, et al. Beruiche Gratikationskrisen und depressive Symptome: Eine Querschnittsstudie bei Erwerbsta tigen im mittleren

886 Rugulies et al.

53.

54.

55.

56. 57.

58. 59. 60.

Erwachsenenalter. (In German). Psychother Psychosom Med Psychol 2003;53:2238. Nielsen ML, Kristensen TS, Smith-Hansen L. The Intervention Project on Absence and Well-being (IPAW): design and results from the baseline of a 5-year study. Work Stress 2002; 16:191206. Nielsen ML, Rugulies R, Christensen KB, et al. Impact of the psychosocial work environment on registered absence from work: a two-year longitudinal study using the IPAW cohort. Work Stress 2004;18:32335. de Jonge J, Mulder MJ, Nijhuis FJ. The incorporation of different demand concepts in the job demand-control model: effects on health care professionals. Soc Sci Med 1999;48: 114960. Kristensen TS, Borg V, Hannerz H. Socioeconomic status and psychosocial work environment: results from a Danish national study. Scand J Public Health 2002;30:418. Kristensen TS, Bjorner JB, Christensen KB, et al. The distinction between work pace and working hours in the measurement of quantitative demands. Work Stress 2004;18: 30522. Aust B, Ducki A. Comprehensive health promotion interventions at the workplace: experiences with health circles in Germany. J Occup Health Psychol 2004;9:25870. Kompier M, Cooper C. Preventing stress, improving productivity. European case studies in the workplace. New York, NY: Routledge, 1999. Kompier MAJ, Aust B, van den Berg AM, et al. Stress prevention in bus drivers: evaluation of 13 natural experiments. J Occup Health Psychol 2000;5:1131.

Quantitative demands Item and response categories. Is your amount of work so extensive that you do not have time to think and talk about anything else than work? (1 almost all working hours; 2 3 4 of working hours; 3 12 of working hours; 4 14 of working hours; 5 seldom; 6 never). Scoring. Participants scoring 1, 2, or 3 were coded as having high quantitative demands. Inuence at work Items and response categories. In your work, is it possible for you to decide your work pace? (1 almost all working hours; 2 34 of working hours; 3 12 of working hours; 4 14 of working hours; 5 seldom; 6 never). Are you involved in planning your work? (1 always; 2 usually; 3 usually not; 4 never). Do you receive information on those decisions that affect your workplace? (1 always; 2 usually; 3 usually not; 4 never). Scoring. Response categories were recoded and summed up to build a scale on inuence at work. Scores on the scale were standardized from 0 to 100, with higher scores indicating more inuence at work. Participants scoring below the median on the scale were coded as having low inuence at work. Possibilities for development Items and response categories. Does your work require that you repeat the same work tasks many times per hour? (1 almost all working hours; 2 34 of working hours; 3 12 of working hours; 4 14 of working hours; 5 seldom; 6 never). Do you have the possibility to learn new things and to qualify yourself at work? (1 always; 2 usually; 3 usually not; 4 never). Is you work varied? (1 to a large extent; 2 to some extent; 3 only to a less extent; 4 no, or only to a small extent). Scoring. Response categories were recoded and summed up to build a scale on possibilities for development. Scores on the scale were standardized from 0 to 100, with higher scores indicating more possibilities for development. Participants scoring below the median on the scale were coded as having low possibilities for development. Social support from supervisors Item and response categories. Do you receive support and encouragement from your supervisor? (1 always; 2 usually; 3 usually not; 4 never; 5 do not have a supervisor). Scoring. Participants scoring 3, 4, or 5 were coded as having low supervisor support. Social support from coworkers Item and response categories. Do you receive support and encouragement from your coworkers? (1 always; 2 usually; 3 usually not; 4 never; 5 do not have coworkers). Am J Epidemiol 2006;163:877887

APPENDIX

This appendix describes how the endpoint variable (severe depressive symptoms) and the six predictor variables (quantitative demands, inuence at work, possibilities for development, social support from supervisors, social support from coworkers, job insecurity) were measured in the survey questionnaire and how scores were calculated.
Severe depressive symptoms Items and response categories. How much of the time during the past 4 weeks have you been a very nervous person? How much of the time during the past 4 weeks have you felt so down in the dumps that nothing could cheer you up? How much of the time during the past 4 weeks have you felt calm and peaceful? How much of the time during the past 4 weeks have you felt downhearted and blue? How much of the time during the past 4 weeks have you been a happy person? Response categories for all the preceding questions include the following: 1 all of the time; 2 most of the time; 3 a good bit of the time; 4 some of the time; 5 a little of the time; 6 none of the time. Scoring. Response categories were recoded and summed up to build a scale on depressive mood. Scores on the scale were standardized from 0 to 100, with higher scores indicating less depressive mood. Participants scoring 52 points or below on the scale were coded as having severe depressive symptoms.

Psychosocial Work Environment and Depressive Symptoms Scoring. Participants scoring 3, 4, or 5 were coded as having low coworker support. Job insecurity Items and response categories. Are you worried about any of the following outcomes: becoming unemployed;

887

being transferred to another job against your will; becoming redundant because of new technology; having difculty in securing another job if you became unemployed? (1 yes; 2 no). Scoring. Participants responding yes to at least one item were scored as having job insecurity.

Am J Epidemiol 2006;163:877887