Abstract

In a prospective cohort study of Finnish public sector employees, the authors examined the association between workplace social capital and depression. Data were obtained from 33,577 employees, who had no recent history of antidepressant treatment and who reported no history of physician-diagnosed depression at baseline in 2000–2002. Their risk of depression was measured with two indicators: recorded purchases of antidepressants until December 31, 2005, and self-reports of new-onset depression diagnosed by a physician in the follow-up survey in 2004–2005. Multilevel logistic regression analysis was used to explore whether self-reported and aggregate-level workplace social capital predicted indicators of depression at follow-up. The odds for antidepressant treatment and physician-diagnosed depression were 20–50% higher for employees with low self-reported social capital than for those reporting high social capital. These associations were not accounted for by sex, age, marital status, socioeconomic position, place of work, smoking, alcohol use, physical activity, and body mass index. The association between social capital and self-reported depression attenuated but remained significant after further adjustment for baseline psychological distress (a proxy for undiagnosed mental health problems). Aggregate-level social capital was not associated with subsequent depression.

“Social capital” can be defined as the norms of reciprocity and trust, formal and informal associations, and civic participation that facilitate collective action for mutual benefit (1). Such social capital has been posited as a protective factor for mental health (2, 3). In the United Kingdom, the Department of Health has cited developing social capital as an important feature of mental health promotion (4).

Social capital is mostly regarded as a property of the relations between individuals and groups of people (5–7). In empirical studies, the concept has been used at both individual and ecologic levels. A systematic review of 21 investigations (7) found that the 14 studies measuring social capital at the individual level supported an inverse relation between social capital and common mental disorders. There was no clear pattern to the results of the seven studies investigating the association between aggregate-level social capital and common mental disorders. Of the 21 studies included in this review, 18 were cross-sectional and thus did not provide information about the temporal order.

A further limitation in earlier studies is that the assessment covers social capital only in residential neighborhoods or communities but not in workplaces. Differences between people who are self-employed and employed as opposed to those who are not in the labor market may be an important source of differences in social capital. However, within working populations, sources of variation in social capital are likely to be found in settings where these people spend most of their time, that is, in workplaces (8, 9). The association of workplace social capital with self-rated health has only recently been demonstrated (8), but its association with mental health is unknown. Numerous studies have shown, however, that other psychosocial work factors, such as social support (10, 11) and organizational justice (12, 13), may contribute to mental health.

We examined low workplace social capital as a risk factor for new-onset depression using large prospective data of public sector employees initially free from depression.

MATERIALS AND METHODS

Study participants

Data were drawn from the ongoing Finnish Public Sector Study of the entire personnel of 10 towns and 21 hospitals in the areas where the towns are located. The purpose of this study is to explore the relation of behavioral and psychosocial factors to health (14). Similar methods of data collection were used in both subsamples (town and hospital). A baseline survey addressing workplace social capital, physician-diagnosed depression, health-risk behaviors, psychological distress, and other factors was conducted in 2000–2002; 32,293 town and 16,299 hospital employees responded (response rates: 67 percent and 69 percent, respectively; total response: 68 percent). Using personal identification numbers (unique number containing birth date and code for sex assigned to all citizens in Finland), we linked the survey responses to records from the National Prescription Register, kept by the Social Insurance Institute of Finland. Data on the purchase of antidepressants before and after the baseline survey were linked to all respondents.

In order to study the onset of new depression among healthy employees, we excluded participants who had a current or preexisting physician-diagnosed depression or a recent history of antidepressant treatment at baseline (n = 11,049) and those with missing data for any study variables (n = 3,043). We further excluded those who worked in work units with less than three employees (n = 334), because the distinction between individual- and aggregate-level social capital is meaningful only in groups. Thus, the final cohort included 33,577 employees. They did not differ substantially from all the respondents at baseline in terms of mean social capital (3.62 in the sample vs. 3.58 in the eligible population), mean age (44.8 years vs. 44.6 years), the proportion of women (80 percent vs. 81 percent), and occupational position (17 percent manual vs. 18 percent manual).

In 2004–2005, a follow-up questionnaire requesting physician-diagnosed depression was sent to all identifiable respondents of the baseline survey who were alive. Of the 35,914 respondents (response: 77 percent), 25,928 belonging to our cohort were free from depression at baseline. This subcohort was used in analyses of new-onset, physician-diagnosed depression.

The Ethics Committee of the Finnish Institute of Occupational Health approved the study.

Measures

Workplace social capital.

Workplace social capital was assessed with a validated and psychometrically tested self-assessment scale, which has been designed to measure social capital in the workplace. The scale includes the following eight items related to cognitive and structural components (7) of social capital (Cronbach's α = 0.87):

  1. “We have a ‘we are together’ attitude.”

  2. “People feel understood and accepted by each other.”

  3. “We can trust our supervisor.”

  4. “People in the work unit cooperate in order to help develop and apply new ideas.”

  5. “Do members of the work unit build on each other's ideas in order to achieve the best possible outcome?”

  6. “Our supervisor treats us with kindness and consideration.”

  7. “Our supervisor shows concern for our rights as an employee.”

  8. “People keep each other informed about work-related issues in the work unit.”

The responses were given on a 5-point rating scale (1 = “strongly disagree,” …, 5 = “strongly agree” in items 1–6 and 8; 1 = “to a very little extent,” …, 5 = “to a very great extent” in item 7). Confirmatory factor analysis showed that the fit of social capital as a single factor to the data was significantly better than that of social capital fitted as a two-dimensional construct (Δχ2 = 1,094.35, df = 1; p < 0.0001 from LISREL statistical software (Scientific Software International, Inc., Lincolnwood, Illinois)) (data not shown). For this reason, we used the measure as a single-factor construct, as in previous studies (8).

The individual-level workplace social capital was the mean of response scores (theoretical range: 1–5) calculated for those participants who had responded to at least four items. A higher score indicates higher social capital. A range of psychometric methods has been used to evaluate the reliability and validity of this measure (8). Supporting convergent validity, the scale was associated with, but not redundant to, conceptually close constructs, such as procedural justice, job control, and effort-reward imbalance. Its associations with conceptually more distant concepts were weaker (divergence validity). In multilevel logistic regression models, social capital was significantly associated with self-rated health (criterion-related validity). The rwg index, which measures the extent to which raters assign the same ratings to a single target, was 0.88, which indicates a significant within-unit agreement.

In addition to individual-level social capital scores, aggregate-level social capital scores were calculated according to the work units. From employers' registers, we obtained information about the administrative units used, for example, to allocate organizational resources and to pay salaries. On the basis of this information, we determined 3,236 functional work units that were each typically at a single location (e.g., a school or a hospital ward). From the organizational hierarchies with multiple levels, we selected work units at the lowest organizational level but included only greater than two-person units. On this basis, the median unit size was 19 employees (interquartile range: 12–34; total range: 3–430). The response rates varied between 10 percent and 100 percent, but for only 8 percent of the units was it less than 50 percent. An aggregated social capital of the work unit (second level) was calculated as the mean of individual (first level) coworkers' responses from the same unit (self-estimation excluded), and then these mean scores were linked to each member of the unit.

Both individual- and aggregate-level social capital scores were divided into quartiles for the analysis, the highest quartile indicating the highest level of social capital.

Depression.

Depression at baseline and at follow-up was assessed from register data and from survey responses. We used prescription data from the National Prescription Register between January 1, 1994, and December 31, 2005, to identify antidepressant treatment. This register comprises outpatient prescription data classified according to the World Health Organization's Anatomical Therapeutic Chemical (ATC) classification code (15). The data consist of the date of the purchase of antidepressants (ATC code N06A) and the corresponding defined daily doses (the assumed average maintenance dose per day for a drug used on its main indication in adults).

In the surveys, respondents were asked to indicate preexisting or current diseases using a self administered checklist of 17 common chronic diseases (16). Physician-diagnosed depression was identified if the respondent reported that a physician had confirmed “depression.”

Cases of baseline depression were identified as respondents who made one or more purchases of antidepressants in the year of or within a 4-year period prior to baseline or who reported a history of physician-diagnosed depression at baseline. New-onset depression after baseline was assessed with annual purchases of antidepressants (>30 defined daily doses) in any subsequent year by the end of 2005 or with an affirmative response to the question of physician-diagnosed depression in the follow-up survey.

Covariates.

The demographic baseline characteristics obtained from the employers' registers included sex, age, place of work (town/hospital), and socioeconomic position based on the occupational-title classification of Statistics Finland, that is, upper-grade nonmanual workers (e.g., physicians, teachers), lower-grade nonmanual workers (e.g., technicians, registered nurses), and manual workers (e.g., cleaners, maintenance workers). In addition, marital status was measured.

Health behaviors have been found to be associated with depression (17–22) and, therefore, may potentially confound or mediate the relation between social capital and depression. We measured smoking (never smoker, former smoker, current smoker); consumption of alcohol (in grams of absolute alcohol per week) (23); physical activity measured as metabolic equivalent task hours (24); and body mass index (25) calculated from self-reported height and weight.

Psychological distress, representing an indicator of potential undiagnosed depression, was measured by a 12-item version of the General Health Questionnaire (26). Individuals with a score of 4 or higher were estimated to have psychological distress.

Statistical analysis

As our data are clustered on the workplace level, the data set was analyzed taking into account this multilevel structure. Adjusted odds ratios and their 95 percent confidence intervals for depression were obtained from multilevel logistic regression models for both the individual-level and aggregate-level scores of social capital at work. In addition to the main effects, the cross-product term of sex and social capital was entered in the models.

In model 1, we estimated the association between social capital and depression outcomes by adjusting for sociodemographic factors. In model 2, we additionally adjusted for health behaviors. Although participants who reported preexisting or current physician-diagnosed depression at baseline were excluded at baseline, many people with depression may not get diagnosed (12). In order to determine the temporal order between social capital and depression outcomes and to take into account undiagnosed mental disorders at baseline, we additionally adjusted for baseline psychological distress in model 3.

We used intraclass correlation (ICC) to study the resemblances of individual responses within work units (27). Technically, the multilevel ICC is a variation partition coefficient that indicates the proportion of the total variance of social capital that occurs at the work unit level (28).

All statistical analyses were conducted with SAS, version 9.1.3, statistical software (SAS Institute, Inc., Cary, North Carolina). The multilevel analyses were performed by using the GLIMMIX procedure.

RESULTS

Of all the 33,577 nondepressed respondents at baseline, 1,608 (5 percent) started a new antidepressant treatment after baseline (table 1). The treated were more often women, the nonmarried, smokers, consumers of more alcohol, those who exercised less, and persons who had psychological distress. Data on self-reports of physician-diagnosed depression were available for the subpopulation of 25,928 individuals who responded to the baseline and follow-up surveys. A total of 1,238 (5 percent) reported physician-diagnosed depression at follow-up. The onset of new, self-reported, physician-diagnosed depression was more likely in women, people living without a partner, among employees with adverse health behaviors, and those in the service of towns.

TABLE 1.

Baseline characteristics and the proportion of antidepressant treatment after baseline and new-onset, self-reported, physician-diagnosed depression at follow-up among participants free from depression at baseline, the Finnish Public Sector Study, 2000–2005

 All participants Respondents to baseline and follow-up surveys 
 Total cohort (n = 33,577) Cases with antidepressant treatment after baseline (n = 1,608) Total cohort (n = 25,928) Cases with self-reported depression at follow-up (n = 1,238) 
 No. Mean (SD*) No. Mean (SD) p value† No. Mean (SD) No. Mean (SD) p value† 
Sex       <0.001       0.012 
    Women 26,954 80  1,383 86   21,259 82  1,048 85   
    Men 6,623 20  225 14   4,669 18  190 15   
Age (years) 33,577  43.8 (9.5) 1,608  43.9 (9.1) 0.696 25,928  44.4 (9.2) 1,238  44.6 (8.8) 0.123 
Socioeconomic position       0.139       0.272 
    Upper nonmanual 10,064 30  453 28   7,667 30  377 30   
    Lower-level nonmanual 17,725 53  887 55   14,022 54  645 52   
    Manual 5,788 17  268 17   4,239 16  216 17   
Place of work       0.356       <0.001 
    Town 21,971 65  1,035 64   16,711 64  965 78   
    Hospital 11,606 35  573 36   9,217 36  273 32   
Married or cohabiting       <0.001        
    Yes 26,131 78  1,193 74   20,293 78  903 73  <0.001 
    No 7,446 22  415 26   5,635 22  335 27   
Current smoking       <0.001       <0.001 
    Never smoker 22,040 66  963 60   17,252 67  728 59   
    Former smoker 5,595 17  294 18   4,514 17  237 19   
    Current smoker 5,742 17  351 22   4,162 16  273 22   
Alcohol use (g/week) 33,577  63.4 (96.0) 1,608  69.1 (113.6) 0.014 25,928  61.5 (92.4) 1,238  70.3 (107.2) 0.001 
Physical activity (MET*-hours/week) 33,577  4.8 (4.3) 1,608  4.3 (3.8) <0.001 25,928  4.8 (4.2) 1,238  4.2 (4.1) <0.001 
Body mass index 33,577  24.9 (4.0) 1,608  25.0 (4.0) 0.116 25,928  24.9 (4.0) 1,238  25.2 (4.1) <0.001 
Psychological distress       <0.001       <0.001 
    No 26,270 78  1,003 62   20,350 78  715 58   
    Yes 7,307 22  605 38   5,578 22  523 42   
 All participants Respondents to baseline and follow-up surveys 
 Total cohort (n = 33,577) Cases with antidepressant treatment after baseline (n = 1,608) Total cohort (n = 25,928) Cases with self-reported depression at follow-up (n = 1,238) 
 No. Mean (SD*) No. Mean (SD) p value† No. Mean (SD) No. Mean (SD) p value† 
Sex       <0.001       0.012 
    Women 26,954 80  1,383 86   21,259 82  1,048 85   
    Men 6,623 20  225 14   4,669 18  190 15   
Age (years) 33,577  43.8 (9.5) 1,608  43.9 (9.1) 0.696 25,928  44.4 (9.2) 1,238  44.6 (8.8) 0.123 
Socioeconomic position       0.139       0.272 
    Upper nonmanual 10,064 30  453 28   7,667 30  377 30   
    Lower-level nonmanual 17,725 53  887 55   14,022 54  645 52   
    Manual 5,788 17  268 17   4,239 16  216 17   
Place of work       0.356       <0.001 
    Town 21,971 65  1,035 64   16,711 64  965 78   
    Hospital 11,606 35  573 36   9,217 36  273 32   
Married or cohabiting       <0.001        
    Yes 26,131 78  1,193 74   20,293 78  903 73  <0.001 
    No 7,446 22  415 26   5,635 22  335 27   
Current smoking       <0.001       <0.001 
    Never smoker 22,040 66  963 60   17,252 67  728 59   
    Former smoker 5,595 17  294 18   4,514 17  237 19   
    Current smoker 5,742 17  351 22   4,162 16  273 22   
Alcohol use (g/week) 33,577  63.4 (96.0) 1,608  69.1 (113.6) 0.014 25,928  61.5 (92.4) 1,238  70.3 (107.2) 0.001 
Physical activity (MET*-hours/week) 33,577  4.8 (4.3) 1,608  4.3 (3.8) <0.001 25,928  4.8 (4.2) 1,238  4.2 (4.1) <0.001 
Body mass index 33,577  24.9 (4.0) 1,608  25.0 (4.0) 0.116 25,928  24.9 (4.0) 1,238  25.2 (4.1) <0.001 
Psychological distress       <0.001       <0.001 
    No 26,270 78  1,003 62   20,350 78  715 58   
    Yes 7,307 22  605 38   5,578 22  523 42   
*

SD, standard deviation; MET, metabolic equivalent.

Difference between the cases and noncases.

No interactions between sex and social capital were found for antidepressant treatment (p ≥ 0.810) or self-reported depression (p ≥ 0.310). Table 2 shows multilevel logistic regression models for the associations between social capital and antidepressant treatment. After adjustment for sociodemographic characteristics and health behaviors, low individual social capital was associated with a 34 percent higher odds of antidepressant treatment after baseline, but the relation attenuated after further adjustment for psychological distress. Tests for linear trends supported these results (table 2).

TABLE 2.

Associations between individual- and aggregate-level social capital at baseline and antidepressant treatment at follow-up in 33,577 participants initially free from depression, the Finnish Public Sector Study, 2000–2005†

Social capital at baseline Model 1‡ Model 2§ Model 3¶ 
Odds ratio 95% confidence interval Odds ratio 95% confidence interval Odds ratio 95% confidence interval 
Individual level       
    1 (low) 1.36** 1.17, 1.57 1.34** 1.16, 1.55 1.13 0.97, 1.31 
    2 1.17* 1.00, 1.34 1.16 1.00, 1.35 1.07 0.92, 1.25 
    3 1.07 0.92, 1.24 1.06 0.92, 1.23 1.02 0.88, 1.19 
    4 (high) 1.00 Referent 1.00 Referent 1.00 Referent 
        Test for linear trend# p < 0.0001 p < 0.0001 p = 0.08 
Aggregate level       
    1 (low) 1.00 0.86, 1.17 1.00 0.85, 1.16 0.95 0.81, 1.12 
    2 0.98 0.84, 1.15 0.98 0.84, 1.14 0.95 0.81, 1.10 
    3 0.93 0.80, 1.09 0.93 0.80, 1.09 0.91 0.78, 1.06 
    4 (high) 1.00 Referent 1.00 Referent 1.00 Referent 
        Test for linear trend# p = 0.34 p = 0.37 p = 0.75 
Social capital at baseline Model 1‡ Model 2§ Model 3¶ 
Odds ratio 95% confidence interval Odds ratio 95% confidence interval Odds ratio 95% confidence interval 
Individual level       
    1 (low) 1.36** 1.17, 1.57 1.34** 1.16, 1.55 1.13 0.97, 1.31 
    2 1.17* 1.00, 1.34 1.16 1.00, 1.35 1.07 0.92, 1.25 
    3 1.07 0.92, 1.24 1.06 0.92, 1.23 1.02 0.88, 1.19 
    4 (high) 1.00 Referent 1.00 Referent 1.00 Referent 
        Test for linear trend# p < 0.0001 p < 0.0001 p = 0.08 
Aggregate level       
    1 (low) 1.00 0.86, 1.17 1.00 0.85, 1.16 0.95 0.81, 1.12 
    2 0.98 0.84, 1.15 0.98 0.84, 1.14 0.95 0.81, 1.10 
    3 0.93 0.80, 1.09 0.93 0.80, 1.09 0.91 0.78, 1.06 
    4 (high) 1.00 Referent 1.00 Referent 1.00 Referent 
        Test for linear trend# p = 0.34 p = 0.37 p = 0.75 
*

p < 0.05; **p < 0.001.

Of the 33,577 participants, 1,608 were prescribed antidepressant treatments during follow-up.

Adjusted for sociodemographics (sex, age, marital status, socioeconomic position, and place of work (town/hospital)).

§

Additionally adjusted for health behaviors (smoking, alcohol use, physical activity, and body mass index).

Additionally adjusted for psychological distress.

#

Linear trend tested with the continuous social capital variable.

In the subpopulation of respondents to baseline and follow-up surveys (table 3), as in all participants, the association between low individual-level social capital and antidepressant treatment attenuated after the adjustment for psychological distress. Before adjustment for psychological distress, the odds of self-reported, physician-diagnosed depression were approximately 50 percent higher for employees with low social capital compared with those with high social capital. When psychological distress was added into the model, the association between low social capital and self-reported, physician-diagnosed depression was reduced (odds ratio = 1.20, 95 percent confidence interval: 1.01, 1.42). Again, tests for linear trends produced corresponding results (table 3).

TABLE 3.

Associations of individual- and aggregate-level social capital at baseline with self-reported, physician-diagnosed depression and antidepressant treatment at follow-up in 25,928 respondents to the baseline and follow-up surveys who were initially free from depression, the Finnish Public Sector Study, 2000–2005†

Social capital at baseline Self-reported, physician-diagnosed depression Antidepressant treatment 
Model 1‡ Model 2§ Model 3¶ Model 1‡ Model 2§ Model 3¶ 
Odds ratio 95% confidence interval Odds ratio 95% confidence interval Odds ratio 95% confidence interval Odds ratio 95% confidence interval Odds ratio 95% confidence interval Odds ratio 95% confidence interval 
Individual level             
    1 (low) 1.53*** 1.30, 1.81 1.51*** 1.27, 1.78 1.20* 1.01, 1.42 1.34** 1.12, 1.59 1.32** 1.11, 1.57 1.09 0.91, 1.31 
    2 1.16 0.97, 1.38 1.15 0.96, 1.37 1.04 0.87, 1.24 1.23* 1.03, 1.47 1.21* 1.02, 1.45 1.12 0.93, 1.34 
    3 1.10 0.92, 1.30 1.09 0.92, 1.30 1.03 0.87, 1.23 1.12 0.94, 1.33 1.12 0.94, 1.33 1.07 0.89, 1.27 
    4 (high) 1.00 Referent 1.00 Referent 1.00 Referent 1.00 Referent 1.00 Referent 1.00 Referent 
        Test for linear trend# p < 0.0001 p < 0.0001 p = 0.007 p < 0.0001 p < 0.001 p = 0.24 
Aggregate level             
    1 (low) 1.02 0.86, 1.22 1.01 0.84, 1.20 0.95 0.79, 1.14 0.98 0.82, 1.18 0.97 0.81, 1.17 0.93 0.78, 1.12 
    2 0.98 0.83, 1.17 0.98 0.82, 1.16 0.94 0.79, 1.12 0.98 0.82, 1.17 0.98 0.82, 1.17 0.95 0.79, 1.13 
    3 0.98 0.82, 1.16 0.97 0.82, 1.16 0.95 0.79, 1.13 0.92 0.77, 1.10 0.92 0.77, 1.10 0.89 0.75, 1.07 
    4 (high) 1.00 Referent 1.00 Referent 1.00 Referent 1.00 Referent 1.00 Referent 1.00 Referent 
        Test for linear trend# p = 0.73 p = 0.84 p = 0.64 p = 0.39 p = 0.43 p = 0.78 
Social capital at baseline Self-reported, physician-diagnosed depression Antidepressant treatment 
Model 1‡ Model 2§ Model 3¶ Model 1‡ Model 2§ Model 3¶ 
Odds ratio 95% confidence interval Odds ratio 95% confidence interval Odds ratio 95% confidence interval Odds ratio 95% confidence interval Odds ratio 95% confidence interval Odds ratio 95% confidence interval 
Individual level             
    1 (low) 1.53*** 1.30, 1.81 1.51*** 1.27, 1.78 1.20* 1.01, 1.42 1.34** 1.12, 1.59 1.32** 1.11, 1.57 1.09 0.91, 1.31 
    2 1.16 0.97, 1.38 1.15 0.96, 1.37 1.04 0.87, 1.24 1.23* 1.03, 1.47 1.21* 1.02, 1.45 1.12 0.93, 1.34 
    3 1.10 0.92, 1.30 1.09 0.92, 1.30 1.03 0.87, 1.23 1.12 0.94, 1.33 1.12 0.94, 1.33 1.07 0.89, 1.27 
    4 (high) 1.00 Referent 1.00 Referent 1.00 Referent 1.00 Referent 1.00 Referent 1.00 Referent 
        Test for linear trend# p < 0.0001 p < 0.0001 p = 0.007 p < 0.0001 p < 0.001 p = 0.24 
Aggregate level             
    1 (low) 1.02 0.86, 1.22 1.01 0.84, 1.20 0.95 0.79, 1.14 0.98 0.82, 1.18 0.97 0.81, 1.17 0.93 0.78, 1.12 
    2 0.98 0.83, 1.17 0.98 0.82, 1.16 0.94 0.79, 1.12 0.98 0.82, 1.17 0.98 0.82, 1.17 0.95 0.79, 1.13 
    3 0.98 0.82, 1.16 0.97 0.82, 1.16 0.95 0.79, 1.13 0.92 0.77, 1.10 0.92 0.77, 1.10 0.89 0.75, 1.07 
    4 (high) 1.00 Referent 1.00 Referent 1.00 Referent 1.00 Referent 1.00 Referent 1.00 Referent 
        Test for linear trend# p = 0.73 p = 0.84 p = 0.64 p = 0.39 p = 0.43 p = 0.78 
*

p < 0.05; **p < 0.01; ***p < 0.001.

Of the 25,928 respondents, 1,238 reported physician-diagnosed depression at follow-up, and 1,153 were prescribed antidepressant medication during follow-up.

Adjusted for sociodemographics (sex, age, marital status, socioeconomic position, and place of work (town/hospital)).

§

Additionally adjusted for health behaviors (smoking, alcohol drinking, physical activity, and body mass index).

Additionally adjusted for psychological distress.

#

Linear trend tested with the continuous social capital variable.

We repeated the analysis with two dichotomous outcomes combining the indicators of new-onset depression: 1) self-reported, physician-diagnosed depression or antidepressant treatment at follow-up (n = 1,375) versus neither of them (an indicator identifying those depressive employees who did not self-report depression but were on antidepressant treatment, as well as those who reported depression but were not treated by antidepressant medication); and 2) self-reported, physician-diagnosed depression and antidepressant treatment at follow-up (n = 508) versus others (an indicator tapping individuals with depression treated by antidepressant medication). The results were in the same direction as when the two depression indicators were examined separately (table 4).

TABLE 4.

Associations of individual- and aggregate-level social capital at baseline with self-reported, physician-diagnosed depression and/or antidepressant treatment at follow-up in 25,928 respondents to baseline and follow-up surveys who were initially free from depression, the Finnish Public Sector Study, 2000–2005†

Social capital at baseline Self-reported, physician-diagnosed depression or antidepressant treatment Self-reported, physician-diagnosed depression and antidepressant treatment‡ 
Model 1§ Model 2¶ Model 1§ Model 2¶ 
Odds ratio 95% confidence interval Odds ratio 95% confidence interval Odds ratio 95% confidence interval Odds ratio 95% confidence interval 
Individual level         
    1 (low) 1.51** 1.31, 1.73 1.48** 1.29, 1.70 1.29 1.00, 1.67 1.27 0.98, 1.64 
    2 1.20* 1.04, 1.39 1.19* 1.03, 1.38 1.19 0.91, 1.55 1.17 0.90, 1.53 
    3 1.09 0.95, 1.26 1.09 0.95, 1.26 1.16 0.90, 1.50 1.16 0.90, 1.50 
    4 (high) 1.00 Referent 1.00 Referent 1.00 Referent 1.00 Referent 
Aggregate level         
    1 (low) 1.03 0.90, 1.19 1.02 0.89, 1.18 0.92 0.71, 1.21 0.91 0.70, 1.19 
    2 1.00 0.87, 1.15 1.00 0.87, 1.15 0.93 0.72, 1.20 0.92 0.70, 1.19 
    3 1.00 0.87, 1.15 1.00 0.87, 1.15 0.79 0.61, 1.04 0.79 0.61, 1.04 
    4 (high) 1.00 Referent 1.00 Referent 1.00 Referent 1.00 Referent 
Social capital at baseline Self-reported, physician-diagnosed depression or antidepressant treatment Self-reported, physician-diagnosed depression and antidepressant treatment‡ 
Model 1§ Model 2¶ Model 1§ Model 2¶ 
Odds ratio 95% confidence interval Odds ratio 95% confidence interval Odds ratio 95% confidence interval Odds ratio 95% confidence interval 
Individual level         
    1 (low) 1.51** 1.31, 1.73 1.48** 1.29, 1.70 1.29 1.00, 1.67 1.27 0.98, 1.64 
    2 1.20* 1.04, 1.39 1.19* 1.03, 1.38 1.19 0.91, 1.55 1.17 0.90, 1.53 
    3 1.09 0.95, 1.26 1.09 0.95, 1.26 1.16 0.90, 1.50 1.16 0.90, 1.50 
    4 (high) 1.00 Referent 1.00 Referent 1.00 Referent 1.00 Referent 
Aggregate level         
    1 (low) 1.03 0.90, 1.19 1.02 0.89, 1.18 0.92 0.71, 1.21 0.91 0.70, 1.19 
    2 1.00 0.87, 1.15 1.00 0.87, 1.15 0.93 0.72, 1.20 0.92 0.70, 1.19 
    3 1.00 0.87, 1.15 1.00 0.87, 1.15 0.79 0.61, 1.04 0.79 0.61, 1.04 
    4 (high) 1.00 Referent 1.00 Referent 1.00 Referent 1.00 Referent 
*

p < 0.05; **p < 0.001.

Of the 25,928 respondents, 1,883 reported physician-diagnosed depression at follow-up or were prescribed antidepressant medication; 508 reported physician-diagnosed depression at follow-up and were prescribed antidepressant medication.

After exclusion of those who either had self-reported, physician-diagnosed depression or were on antidepressant treatment (n = 1,375).

§

Adjusted for sociodemographics (sex, age, marital status, socioeconomic position, and place of work (town/hospital)).

Additionally adjusted for health behaviors (smoking, alcohol drinking, physical activity, and body mass index).

The ICC for social capital was 22 percent in the crude model. This percentage indicates that a substantial proportion of the variance of individual social capital was between work units.

As displayed in tables 2–4, there was no association between aggregate-level social capital and depression outcomes in the total cohort or among follow-up respondents.

DISCUSSION

Data from a large cohort of public sector employees initially free from depression suggest that low individual-level social capital at work is associated with the onset of depression. The odds for new antidepressant medication and/or reported physician-diagnosed depression were approximately 20–50 percent higher for employees with low self-reported social capital than for those reporting high social capital. This association was not accounted for by sex, age, marital status, socioeconomic position, place of work, smoking, alcohol use, physical activity, or body mass index. The association between social capital and self-reported, physician-diagnosed depression was also independent of baseline psychological distress, although the association became weaker after this adjustment. Our results confirm earlier findings in community contexts that individual perceptions of social capital may play a role in shaping people's mental health (2, 7, 29).

The mechanisms underlying the association between workplace social capital and depression may be largely similar to those in the neighborhood context. First, although low social capital can be distinguished from the concept of social support (29–31), there may be a relation between the two, with lower workplace social capital decreasing the likelihood of accessing various forms of support (3). Low social capital could also reflect poorer access to local services and amenities (30, 32, 33), and it could be an obstacle for an effective dissemination of mental health information and knowledge at the workplace (34). Second, a low level of integration within a social network may produce negative psychological states, which decrease motivation for self-care (3), and it could increase vulnerability to the adverse health effects of chronic stress (35, 36). Third, communities with low levels of social capital have been suggested to be less effective at exercising social control over health-risk behaviors (34, 37); members of these communities may not obtain normative guidance about healthy behaviors, which can in turn affect mental health (3). However, this explanation is not highly plausible for the present findings, given that the association between workplace social capital and depression was little affected by adjustment for health behaviors.

We found no association between aggregate-level social capital and depression, suggesting that the processes determining the causes and consequences of social capital are different at various levels of aggregation (31). It is possible that workplace social capital increases the risk of depression through influences on the individual's own perception of social capital. The lack of observed association between aggregate-level social capital and depression may also relate to exposure misclassification or measurement imprecision. The aggregate-level social capital score was constructed on the basis of administrative records. Although these records provide information about formal functional work units, informal work groups might provide a more accurate proxy for aggregate-level social capital in some cases. Finally, we cannot exclude the possibility that there is no meaningful effect of workplace social capital on depression. As the associations between individual-level social capital and depression were reduced upon adjustment for psychological distress at baseline, the associations could be attributable to reverse causation, with undiagnosed depression influencing self-reports of social capital. Even with a longitudinal design it is difficult to distinguish between lack of social capital as an antecedent of or as a concomitant cause or consequence of mental health problems (3).

Strengths and limitations

As far as we are aware, this is the first study focusing on the longitudinal association between workplace social capital and mental health. The respondents represented the target population well in terms of sex, age, and socioeconomic position. Although the sample of public sector employees was not truly representative of the general working population in Finland, it represents a heterogeneous group of public sector workers in both manual and nonmanual occupations. We used multilevel modeling to take into account the hierarchical data, and information on antidepressant treatment was obtained from a national register. The prescriptions in our study were based on a physician's examination and covered virtually all purchases of prescribed antidepressants for the cohort.

At least five limitations of this study are noteworthy. First, our baseline response rate was satisfactory but not high. Differences in age, sex, occupational position, and social capital at baseline between included and excluded cohort members were small and unlikely to introduce a major selection bias. Poorer psychological health could lead to social withdrawal and lower participation (7). However, as all participants included in the study were free from self-reported depression and had no recent history of antidepressant use, variation in depression at baseline is an unlikely source of major bias. The onset of major depression after baseline might have contributed to selective dropout from the follow-up survey, but it could not bias records of antidepressant prescriptions, as they were obtained for the entire baseline cohort. In those work units where the response rate was low, aggregate-level social capital may not have given reliable results. However, only 8 percent of the participants were in work units with less than a 50 percent response rate.

Second, although we assessed social capital with a psychometrically validated measure, specifically designed to measure social capital in a work context, it was based on self-reports and therefore subject to response and recall bias. Further studies of workplace social capital could benefit from more objective measures, such as the number of times informally socialized with coworkers, in assessment of the structural elements of social capital.

Third, we had no information on the diagnosis for which antidepressants were prescribed, preventing us from excluding those prescriptions that were for indications other than mental disorders, such as chronic pain or sleeping problems. Further, we had no information on the prescribed dosage of the treatment, which could have acted as a proxy for diagnosis, since in general lower doses are used for the management of non-mental health conditions. We believe, however, that this is unlikely to be a major source of bias, as depressive and anxiety disorders are the main causes for recommended antidepressant use. In addition, we included only purchases lasting more than 1 month assessed by commonly used doses for treating depression. There was an indication of slightly stronger associations of low individual-level social capital with the outcome of self-reported, physician-diagnosed depression, as compared with the outcome of antidepressant treatment. Considering that antidepressants are not recommended for the initial treatment of mild depression (38), this raises the hypothesis that social capital may be more strongly related to mild than moderate or severe depression.

Fourth, although we performed multiple adjustments, it is still possible that some unmeasured factors are behind the observed associations between individual social capital and depression. For example, as no data on neighborhood social capital were available in this study, we cannot rule out the possibility that neighborhood contextual factors may have contributed to residual confounding of the observed associations.

Finally, all participants were from the public sector and had a full-time job, limiting the generalizability of our findings. People with severe or chronic depression may not be able to enter the labor market, and the differences between employed and nonemployed people may be even more important as a source of differences in social capital between different segments of the population than workplace. Further studies in the general population are needed to examine this issue.

Conclusions and practical implications

Psychiatric disorders are among the most common causes of disability retirement in workers (39). According to estimates of the World Health Organization, approximately 121 million people suffer from depression, and it will account for 15 percent of the disease burden throughout the world by 2020 (40). In addition to individual suffering, depression leads to substantial loss of productivity.

Building or sustaining healthy communities has been seen as an important weapon in a state's strategy to prevent mental ill health (4). Our findings showed that low individual social capital at work is associated with the onset of depression, but this association was not confirmed by the work-unit coworkers' assessments of social capital. Thus, this study failed to provide unambiguous support for interventions to increase social capital at work as a means of preventing depression among employees.

Abbreviations

    Abbreviations
  • ATC

    Anatomical Therapeutic Chemical

  • ICC

    intraclass correlation

Funding was from the Academy of Finland (projects 105195, 110451, 117604, 124322, and 124271); the Finnish Work Environment Fund (project 103432); and the participating towns and hospitals.

Conflict of interest: none declared.

References

1.
Putnam
R
Bowling alone: the collapse and revival of American community
2000
New York, NY
Simon Schuster
2.
McKenzie
K
Whitley
R
Weich
S
Social capital and mental health
Br J Psychiatry
 , 
2002
, vol. 
181
 (pg. 
280
-
3
)
3.
Kawachi
I
Berkman
L
Social ties and mental health
J Urban Health
 , 
2001
, vol. 
78
 (pg. 
458
-
67
)
4.
Department of Health
Making it happen: a guide to mental health promotion
2001
London, United Kingdom
The Stationery Office
5.
Almedom
AM
Social capital and mental health: an interdisciplinary review of primary evidence
Soc Sci Med
 , 
2005
, vol. 
61
 (pg. 
943
-
64
)
6.
Islam
MK
Merlo
J
Kawachi
I
, et al.  . 
Social capital and health: does egalitarianism matter? A literature review
Int J Equity Health
 , 
2006
, vol. 
5
 (pg. 
1
-
28
)
7.
De Silva
MJ
McKenzie
K
Harpham
T
, et al.  . 
Social capital and mental illness: a systematic review
J Epidemiol Community Health
 , 
2005
, vol. 
59
 (pg. 
619
-
27
)
8.
Kouvonen
A
Kivimäki
M
Vahtera
J
, et al.  . 
Psychometric evaluation of a short measure of social capital at work. (Electronic article)
BMC Public Health
 , 
2006
, vol. 
6
 pg. 
251
 
9.
Kawachi
I
Social capital and community effects on population and individual health
Ann N Y Acad Sci
 , 
1999
, vol. 
896
 (pg. 
120
-
30
)
10.
Stansfeld
SA
Fuhrer
R
Shipley
MJ
, et al.  . 
Work characteristics predict psychiatric disorder: prospective results from the Whitehall II study
Occup Environ Med
 , 
1999
, vol. 
56
 (pg. 
302
-
7
)
11.
Stansfeld
S
Candy
B
Psychosocial work environment and mental health—a meta-analytic review
Scand J Work Environ Health
 , 
2006
, vol. 
32
 (pg. 
443
-
62
)
12.
Kivimaki
M
Elovainio
M
Vahtera
J
, et al.  . 
Association between organizational inequity and incidence of psychiatric disorders in female employees
Psychol Med
 , 
2003
, vol. 
33
 (pg. 
319
-
26
)
13.
Ferrie
JE
Head
J
Shipley
MJ
, et al.  . 
Injustice at work and incidence of psychiatric morbidity: the Whitehall II study
Occup Environ Med
 , 
2006
, vol. 
63
 (pg. 
443
-
50
)
14.
Kivimäki
M
Lawlor
DA
Davey Smith
G
, et al.  . 
Socioeconomic position, co-occurrence of behavior-related risk factors, and coronary heart disease: the Finnish Public Sector Study
Am J Public Health
 , 
2007
, vol. 
97
 (pg. 
874
-
9
)
15.
World Health Organization
Guidelines for ATC classification and DDD assignment
2004
7th ed.
Oslo, Norway
WHO Collaborating Center for Drug Statistics Methodology
16.
Koskenvuo
M
Kaprio
J
Rose
RJ
, et al.  . 
Hostility as a risk factor for mortality and ischemic heart disease in men
Psychosom Med
 , 
1988
, vol. 
50
 (pg. 
330
-
40
)
17.
Strawbridge
WJ
Deleger
S
Roberts
RE
, et al.  . 
Physical activity reduces the risk of subsequent depression for older adults
Am J Epidemiol
 , 
2002
, vol. 
156
 (pg. 
328
-
34
)
18.
Paperwalla
KN
Levin
TT
Weiner
J
, et al.  . 
Smoking and depression
Med Clin North Am
 , 
2004
, vol. 
88
 (pg. 
1483
-
94
)
19.
Klungsoyr
O
Nygård
JF
Sørensen
T
, et al.  . 
Cigarette smoking and incidence of first depressive episode: an 11-year, population-based follow-up study
Am J Epidemiol
 , 
2006
, vol. 
163
 (pg. 
421
-
32
)
20.
Paschall
MJ
Freisthler
B
Lipton
RI
Moderate alcohol use and depression in young adults: findings from a national longitudinal study
Am J Public Health
 , 
2005
, vol. 
95
 (pg. 
453
-
7
)
21.
Dixon
JB
Dixon
ME
O'Brien
PE
Depression in association with severe obesity: changes with weight loss
Arch Intern Med
 , 
2003
, vol. 
163
 (pg. 
2058
-
65
)
22.
Onyike
CU
Crum
RM
Lee
HB
, et al.  . 
Is obesity associated with major depression? Results from the Third National Health and Nutrition Examination Survey
Am J Epidemiol
 , 
2003
, vol. 
158
 (pg. 
1139
-
47
)
23.
Kouvonen
A
Kivimäki
M
Cox
SJ
, et al.  . 
Job strain, effort-reward imbalance and heavy drinking: a study in 40,851 employees
J Occup Environ Med
 , 
2005
, vol. 
47
 (pg. 
503
-
13
)
24.
Kujala
UM
Kaprio
J
Koskenvuo
M
Modifiable risk factors as predictors of all-cause mortality: the roles of genetics and childhood environment
Am J Epidemiol
 , 
2002
, vol. 
156
 (pg. 
985
-
93
)
25.
Obesity: preventing and managing the global epidemic. Report of a WHO consultation
World Health Organ Tech Rep Ser
 , 
2000
, vol. 
894
 (pg. 
i
-
xii
1–253
26.
Goldberg
D
Williams
P
A user's guide to the General Health Questionnaire
1988
Berkshire, United Kingdom
NFER-Nelson Publishing Co
27.
Bryk
AS
Raudenbush
SW
Hierarchical linear models: applications and data analysis methods
1992
Newbury Park, CA
Sage Publishing
28.
Merlo
J
Chaix
B
Yang
M
, et al.  . 
A brief conceptual tutorial of multilevel analysis in social epidemiology: linking the statistical concept of clustering to the idea of contextual phenomenon
J Epidemiol Community Health
 , 
2005
, vol. 
59
 (pg. 
443
-
9
)
29.
Kawachi
I
Kim
D
Coutts
A
, et al.  . 
Commentary: reconciling the three accounts of social capital
Int J Epidemiol
 , 
2004
, vol. 
33
 (pg. 
682
-
90
)
30.
Harpham
T
Grant
E
Thomas
E
Measuring social capital within health surveys: key issues
Health Policy Plan
 , 
2002
, vol. 
17
 (pg. 
106
-
11
)
31.
Lochner
K
Kawachi
I
Kennedy
BP
Social capital. A guide to its measurement
Health Place
 , 
1999
, vol. 
5
 (pg. 
259
-
70
)
32.
Kawachi
I
Kennedy
BP
Glass
R
Social capital and self-rated health: a contextual analysis
Am J Public Health
 , 
1999
, vol. 
89
 (pg. 
1187
-
93
)
33.
Sampson
RJ
Raudenbush
SW
Earls
F
Neighborhoods and violent crime: a multilevel study of collective efficacy
Science
 , 
1997
, vol. 
277
 (pg. 
918
-
24
)
34.
Kawachi
I
Berkman
L
Berkman
L
Kawachi
I
Social cohesion, social capital, and health
Social epidemiology
 , 
2000
New York, NY
Oxford University Press
(pg. 
174
-
90
)
35.
Brunner
E
Marmot
M
Marmot
M
Wilkinson
RG
Social organisation, stress, and health
Social determinants of health
 , 
1999
Oxford, United Kingdom
Oxford University Press
(pg. 
17
-
43
)
36.
Wilkinson
R
Unhealthy societies: the afflictions of inequality
1996
London, United Kingdom
Routledge
37.
Subramanian
SV
Lochner
K
Kawachi
I
Neighbourhood differences in social capital: a compositional artifact or a contextual construct?
Health Place
 , 
2003
, vol. 
9
 (pg. 
33
-
44
)
38.
National Institute for Clinical Excellence
Depression: management of depression in primary and secondary care. Clinical guideline 23
2004
London, United Kingdom
National Institute for Clinical Excellence
39.
Pattani
S
Constantinovici
N
Williams
S
Who retires early from the NHS because of ill health and what does it cost? A national cross sectional study
BMJ
 , 
2001
, vol. 
322
 (pg. 
208
-
9
)
40.
Murray
CJ
Lopez
AD
Alternative projections of mortality and disability by cause 1990 –2020: Global Burden of Disease Study
Lancet
 , 
1997
, vol. 
349
 (pg. 
1498
-
504
)

Author notes

Editor's note: An invited commentary on this article appears on page1152.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/2.0/uk/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.