Work-unit measures of organisational justice and risk of depression—a 2-year cohort study
- Matias Brødsgaard Grynderup1,
- Ole Mors2,
- Åse Marie Hansen3,4,
- Johan Hviid Andersen5,
- Jens Peter Bonde6,
- Anette Kærgaard5,
- Linda Kærlev7,
- Sigurd Mikkelsen6,
- Reiner Rugulies4,
- Jane Frølund Thomsen6,
- Henrik Albert Kolstad1
- 1Department of Occupational Medicine, Danish Ramazzini Centre, Aarhus University Hospital, Aarhus, Denmark
- 2Centre for Psychiatric Research, Aarhus University Hospital, Risskov, Denmark
- 3Department of Public Health, University of Copenhagen, Copenhagen, Denmark
- 4National Research Centre for the Working Environment, Copenhagen, Denmark
- 5Department of Occupational Medicine, Danish Ramazzini Centre, Regional Hospital Herning, Herning, Denmark
- 6Department of Occupational and Environmental Medicine, Bispebjerg University Hospital, Copenhagen, Denmark
- 7Center for Clinical Epidemiology, Odense University Hospital, Odense, Denmark
- Correspondence to Matias Brødsgaard Grynderup, Department of Occupational Medicine, Aarhus University Hospital, Nørrebrogade 44 byg. 2C, Aarhus 8000, Denmark;
- Received 20 June 2012
- Revised 7 February 2013
- Accepted 11 February 2013
- Published Online First 8 March 2013
Objectives The aim of this study is to analyse if low justice at work, analysed as aggregated workplace means, increases the risk of depression.
Methods A total of 4237 non-depressed Danish public employees within 378 different work units were enrolled in 2007. Mean levels of procedural and relational justice were computed for each work unit to obtain exposure measures that were robust to reporting bias related to depression. Two years later in 2009, 3047 (72%) participated at follow-up. Those reporting high levels of depressive, burn-out or stress symptoms were assigned to a psychiatric diagnostic interview. In the interview 58 cases of new onset depression were identified. Depression ORs by work unit level of procedural and relational justice were estimated by multivariable logistic regression accounting for established risk factors for depression.
Results Working in a work unit with low procedural justice (adjusted ORs of 2.50, 95% CI 1.06 to 5.88) and low relational justice (3.14, 95% CI 1.37 to 7.19) predicted onset of depression.
Conclusions Our results indicate that a work environment characterised by low levels of justice is a risk factor for depression.
Depression is the leading burden of disease assessed by disability-adjusted life years in middle-income and high-income countries.1 Strong evidence links bereavement and other emotionally painful life events causally with depression.2 Less distressing but long-lasting strenuous psychosocial working conditions may also be a risk factor for depression.3 ,4 However, most studies of this relation may have been hampered by biased self reports of working conditions related to individual characteristics such as personality traits or subclinical depressive symptoms.5 ,6 Measures of psychosocial working conditions that are obtained independently of the depressed participants may be the only option to circumvent the serious problem of reporting bias.3 ,6 This is relevant even in follow-up studies, because depression often has a long insidious preclinical stage.3 ,5 ,6 Averaging across work units6 ,7 or workplaces,8 assessment by experts9 or employers,10 or information on hospital overcrowding, reorganisation, work load or job titles11 are different approaches to obtain exposure information less affected by reporting bias.
Organisational justice at work is a relatively novel approach to an understanding of how psychological working conditions may affect health, wellbeing and productivity.12 Organisational justice is composed of two separate elements. Procedural justice describes the consistency of the decision-making procedures in a workplace, the accuracy of information collected to make decisions and the degree in which all involved are allowed to voice their concern and challenge any decisions. Relational justice describes the degree to which supervisors consider employees’ viewpoints, suppresses personal bias and treats the employees with kindness, consideration and truthfulness.12 Low levels of justice at work may increase the risk of depression13–15 and impact on other aspects of workers’ health, such as self-rated health,12 ,16 sickness absence,12 ,16 psychological distress,17 coronary heart disease18 and cardiovascular death.19 Prolonged stress has been suggested as a causal link between organisational justice and health problems.20 A work environment characterised by organisational justice may help employees cope with uncertainty and mistrust. Justice may also affect employees’ behaviours, feelings, beliefs, self-esteem and social identity.21 Increased inflammation, cardiac dysregulation, poor sleep quality and impaired cognitive function have also been suggested but are still hypothetical.20
In the present study we used mean scores of self-reported justice obtained in small work units with homogeneous working conditions. Participants in the work unit who were diagnosed with depression at baseline were excluded from the calculation of the mean scores because their depression could have influenced their assessment of the psychosocial work environment. We also excluded participants who were diagnosed with depression at follow-up from the calculation of the mean scores because they could have had preclinical depressive symptoms which could have influenced their assessment of their working conditions. By including only participants who were non-depressed throughout the study we should have avoided any reporting bias caused by depression.
The aim of this article is to investigate if low levels of justice at work, aggregated at the work unit level, increase the risk of depression in a prospective cohort study of Danish employees.
Materials and methods
We measured relational and procedural justice in 2007 and analysed if lower levels predicted new-onset depression present at follow-up in 2009. Cases of depression were identified in 2007 and 2009 by a two-step procedure: First, we identified participants reporting mental symptoms (symptoms of depression, stress or burn-out) in a questionnaire. Second, these participants were invited to take part in a standardised psychiatric interview to clinically diagnose cases with depression.
In 2007, the Danish PRISME (Psychological risk factors in the work environment and biological mechanism for the development of stress, burnout and depression) cohort of 10 036 public employees from 502 work units in Aarhus, Denmark, was recruited for the baseline study, and 4489 employees (44.7%) from 474 work units participated by filling in a postal questionnaire concerning working conditions and health. Participants with depression at baseline (as defined below) were excluded (n=100). We also excluded five participants from five work units for which we could not identify the work unit leader and participants from work units with less than three responders who were non-depressed at baseline and follow-up (147 workers from 90 work units) to avoid unstable work unit measures of exposure. A total of 4237 participants from 378 work units were eligible for follow-up. In 2009, all participants from 2007 were approached again, and 3047 (72%) participated, comprising our final study population. Further details of the study have been reported in more detail elsewhere.6 ,22
Measures of psychosocial working conditions
Procedural and relational justice were measured with a Danish version of the organisational justice questionnaire originally developed by Moorman23 and modified by Kivimäki et al.16 The questionnaire contained four items about procedural justice and four items about relational justice (figure 1). All items were rated on a 5-point scale from 1 (‘strongly disagree’) to 5 (‘strongly agree’). Mean values of all items on both justice scales were calculated for each work unit after exclusion of participants with depression at baseline or follow-up. The mean values of each work unit were assigned to all employees working in the particular work unit.
Measures of mental symptoms
We assessed depressive symptoms by the Common Mental Disorder Questionnaire subscale for depression (six items),24 stress by the Perceived Stress Scale (four items),25 and burn-out by the Copenhagen Burn-Out Inventory (six items).26 All questions concerned the last 4 weeks, and responses were given on 5-point scales.
At baseline in 2007, participants were selected for the psychiatric interview if a) their point score was 3 or higher on three or more of the six depressive symptoms items, b) the mean score was 2.5 or more on the perceived stress scale or c) the mean score was 4 or more on the Copenhagen Burn-Out Inventory. The selection criteria for depressive symptoms were chosen to obtain optimal validity.24 We expected that participants with depression would also have high perceived stress and burn-out levels and therefore included participants based on these mental symptom scales.
At follow-up in 2009, we redefined the selection criteria for the psychiatric interviews based on tabulation of the frequency of a depression diagnosis by different cut-off levels of depressive, stress and burn-out scores in the 2007 data. We did this to identify the largest number of depression cases with the lowest number of interviews. We selected participants with high scores on at least two of the three mental symptom scales (depressive scores of 3 or higher on two or more of the six questions, average stress and burn-out scores of 2.5 or higher).
Diagnosis of depression
Diagnoses of depression were obtained by the Schedules for Clinical Assessment in Neuropsychiatry interview (V.2.1 part I, sections 6, 7, 8 and 10)27 according to the International Classification of Disease, 10th revision, Diagnostic Criteria for Research (ICD-10-DCR). All questions referred to the previous 3 months. The interviews were conducted by 10 students of medicine or psychology who had been trained at a 1 week course by a WHO certified trainer (OM). Inter-rater reliability on item level was satisfactory (κ=0.71).
Cases of depression
In 2007, a total of 100 participants were diagnosed with depression and excluded from the study. The ICD-10-DCR diagnostic criteria for a mild, moderate and severe depressive episode were fulfilled for 40, 43 and 17 participants, respectively. In 2009, a total of 58 of 3047 participants were diagnosed with a new onset of depression. The ICD-10-DCR diagnostic criteria for a mild, moderate and severe depressive episode were fulfilled for 15, 32 and 11 participants, respectively.
ORs of depression were calculated by logistic regression analyses with robust clusters based on the work unit of the participants.28 As the data were cluster-sampled the analyses must account for this. Since the main focus of the analyses were not to provide an apportionment of the variance into between and within clusters, but to report risk estimates on a population level, we used robust variance estimation. Analyses were performed with continuous-scale exposure information (linear, quadratic and cubic transformations) and tertile categorisation. Associations were further explored with restricted cubic spline regression analysis (four knots on percentiles 5, 35, 65 and 95). Linearity of the relation between exposure variables and depression was tested with likelihood-ratio testing.
We included the following potential confounders as measured at the individual level at baseline in all models: gender, age (continuous), previous episodes of depression (yes, no), family history of depression (yes, no), income (continuous), years of education beyond primary or high school (<3, 3–4, >4), alcohol consumption (continuous), living alone (yes, no), neuroticism (continuous with quadratic term; from the neuroticism scale of the Eysenck Personality Questionnaire Revised-Abbreviated version29), depressive symptoms (continuous with quadratic term; from the Common Mental Disorder Questionnaire), body mass index (continuous), years of smoking (continuous). Traumatic life events defined as serious illness or injury, assault, death of a relative or friend, marital problems, or serious illness or assault of a close relative30 during the last 6 months were also included and measured at follow-up. The selection of these potential confounders was based upon a review of the literature.31–35 We examined all continuous covariates for linearity by likelihood-ratio testing. Linearity was not accepted for neuroticism and baseline depressive symptoms, so these potential confounders were included as linear and quadratic terms. We tested for interaction between gender and procedural and relational justice, respectively, and performed subanalyses for female participants only.
We used likelihood-ratio testing to identify the strongest potential confounders of new-onset depression, and performed similar analyses on a model only including these variables. The homogeneity of self-reported procedural and relational justice within the work units were assessed by intraclass correlation and within-group inter-rater agreement indices.36 Since our screening criteria for being invited to the psychiatric interviews changed slightly from baseline to follow-up we checked if exclusion of cases of depression that would not have been identified by the baseline criteria changed our findings. All analyses were conducted with the STATA 11 statistical software (StataCorp LP, College Station, Texas, USA).
Nurses (30%), social workers (18%), teachers (11%), managers (7%) and medical doctors (6%) were the most prevalent professions among the participants. Previous depression, a family history of depression, traumatic life events, neuroticism and subclinical depressive symptoms at baseline predicted depression at follow-up (table 1). There were only small differences between responders and non-responders at follow-up. Responders had a mean level of procedural justice of 2.82 and a mean level of relational justice of 2.20 compared with non-responders with 2.82 and 2.23. Responders had a mean age of 43 years, 80% were women and 83% had 3 or more years of education. Non-responders had a mean age of 45 years, 79% were women, and 74% had 3 or more years of education. The 100 depressed participants who were excluded at baseline had a mean level of procedural justice of 2.88, a mean level of relational justice of 2.30, a mean age of 44.5 years, 83% were women and 78% had 3 or more years of education.
The risk of depression increased monotonously by lower levels of procedural and relational justice. The adjusted ORs for a 1-point decrease on the 5-point justice scales were 2.96 (1.19 to 7.34) and 4.84 (2.15 to 10.90) for procedural and relational justice, respectively (table 2). Neither quadratic, nor cubic, nor spline models fitted the data significantly better than the linear models of exposure. The adjusted ORs for the lowest tertile compared with the highest tertile were 2.50 (1.06 to 5.88) for procedural justice and 3.14 (1.37 to 7.19) for relational justice (table 2).
In a model only including the strongest potential confounders (gender, previous depression, traumatic life events, living alone, depressive symptoms at baseline and neuroticism) we observed similar results as those obtained by the fully adjusted model (data not shown). We found a medium to large intraclass correlation of 0.16 and 0.15 for procedural and relational justice, respectively. We found an average inter-rater agreement of 0.75 for procedural justice and 0.77 for relational justice, indicating a strong homogeneity within work units.
We found no interaction between gender and procedural justice (p=0.84) and gender and relational justice (p=0.85). We found very similar results when only examining female participants (data not shown). One depressed participant would not have been included among the cases if we had applied the same screening criteria for being invited to the psychiatric interviews at baseline as at follow-up. Excluding this single participant did not change the results (data not shown).
Members of work units with low levels of procedural or relational justice had a substantially increased risk of developing depression over a 2-year period. The results showed an exposure-response relationship.
The baseline participation rate was low (45%), which could have biased results, if participation was associated with level of justice as well as depression. We investigated this by extrapolating the work unit justice estimates to the non-responding members of the work units and by accessing registry information on redeemed antidepressant medication for the entire source population that has been published elsewhere.37 We found relative ORs of antidepressant use of 1.01 (0.75 to 1.37) for low procedural justice and 1.01 (0.74 to 1.38) for low relational justice when comparing responders with the entire source population. This indicates that the low baseline participation did not distort the estimates of the associations between justice and depression, since the relation between justice and antidepressant use are almost identical for participants and non-participants.
Based on previously reported prevalence and recurrence rates of depression we had expected twice the number of cases.33 ,38 Our lower than expected number of cases may in part be due to a healthy worker effect. It may also in part be due to a lower participation rate of depressive employees as non-participants at baseline were more often prescribed antidepressant medication,37 and in part due to the low baseline participation rate. Additionally, some participants with depression may have not been identified by our screening procedure for the psychiatric interview. However, even if we had indeed missed participants with depression, this can hardly explain the strong associations between organisational justice and depression that we observed.
At follow-up, the participation rate was higher (72%) than at baseline, but selection may still have biased our findings. However, we found only a small difference between responders’ and non-responders’ levels of justice and depressive symptoms at baseline. Thus, baseline justice and depressive symptoms did not predict participation at follow-up, indicating that bias due to selective loss to follow-up is unlikely.
Participants who were diagnosed with depression at baseline were excluded from the calculation of the mean scores since their depression could influence their assessment of the psychosocial work environment. We also excluded participants that were diagnosed with depression at follow-up because they could have preclinical depressive symptoms which could influence their assessment of their working conditions. By including only participants who are non-depressed throughout the study we circumvent the serious problem of biased self-reporting of working conditions which may have hampered most previous studies of psychosocial factors and the risk of depression.3 ,6
Justice at work is likely to be related to social class and thereby to lifestyle factors and the associations between justice and depression may thus be confounded. We therefore adjusted for income, educational level, alcohol consumption, body mass index and smoking, and any effects of confounding from non-controlled socioeconomic and lifestyle factors therefore seem small.
Likewise, personality factors may be related to perceived justice at work as well as to depression.33 ,39 We adjusted for neuroticism that is a risk factor of depression,33 but did not take other personality traits into account. However, hostility and trait anxiety, did not have strong confounding effects on the relation between perceived justice and depression in a recent study.39 This makes confounding due to these personality traits unlikely in our study.
We did not adjust for other psychosocial work factors, and it is possible that the association between justice and depression was, at least partly, mediated by other work factors, but we did not find any association between psychological demands, decision latitude and depression in a recent analysis of this population.22
The adjusted association between relational justice and depression was stronger than the crude association, though the difference was not statistically significant. This increase was primarily because women and participants with previous depression and low income were more prevalent in work units with higher levels of relational justice. These factors were all related to depression and adjusting for them increased the association between relational justice and depression. A similar pattern was also found for procedural justice, although to a smaller degree.
Working conditions may vary significantly between workers within a work unit, and one may argue that this variance was not captured by our work-unit average exposure measure. Therefore, we explicitly identified units of workers that shared leadership, colleagues and work content and thus were expected to experience similar levels of justice. We found a strong homogeneity within work units, which may justify aggregation in a multilevel analysis.36 Furthermore, risk estimates obtained from grouped exposures are not expected to be attenuated because grouping accounts for random misclassification and leads to predominance of Berkson-type error in exposure assessment.40
So far, only few studies linking organisational justice with depression have been published, but earlier results are in line with our findings.13–15 Low justice at work has also been related to minor psychiatric morbidity, doctor-diagnosed psychiatric disorder, coronary heart disease, cardiovascular death, sickness absence and other health effects.20 ,41
This study provides evidence that a work environment characterised by low procedural and relational justice is a risk factor for depression. This is an important finding that may open new possibilities for prevention of depression because unfair working conditions are amendable to change.
A management style characterised by a clearly articulated concern for being fair reinforced through use of accurate and transparent procedures has been suggested to increase justice at work.42 Further studies are needed for investigating the exact factors that contribute to an unjust workplace.
What this paper adds
It has been suggested that low levels of justice at work increase the risk of depression.
However, studies may have been hampered by biased self-reports of working conditions. Thus, measures of justice at work, analysed independently of the perceptions of the depressed participants, are needed.
We classified employees of 378 small work units with similar working conditions by the average levels of procedural and relational justice obtained among the non-depressed workers and examined the risk of depression 3 years later. This design should be robust to reporting bias related to depression status.
We observed that low levels of procedural and relational justice were associated with increased risk of depression. This is an important finding that may open new possibilities for prevention of depression because unfair working conditions are amendable to change.
Contributors MBG: performed the statistical analysis for the present paper, interpreted the analyses and drafted the manuscript. All coauthors have helped with interpretation of the analyses and with revising the manuscript critically. HAK, ÅMH, AK, JFT, LK, SM and MBG: coordinated the data collection, enrolment of participants, data acquisition and quality assurance. OM: designed and assured the quality of the psychiatric interviews. All authors participated in the design and coordination of the PRISME study, development of the questionnaire and have made substantial contributions to interpretation of data; and read and approved the final manuscript.
Funding This study was supported through grants from the Danish Work Environment Research Fund (5-2005-09), The Lundbeck foundation and H Lundbeck A/S. The funding bodies had no role in the design or conduct of the study; collection, management, analysis, or interpretation of the data; or preparation, review, or approval of the manuscript.
Competing interests None.
Ethics approval The Regional Scientific Ethical Committee for Southern Denmark.
Provenance and peer review Not commissioned; externally peer reviewed.