Purpose Work-related psychosocial factors may precipitate the onset of depression. In occupational mental health research, there are three widely used theoretical models, namely, job demand and control (JD-C), effort-reward imbalance (ERI) and work-family conflicts (WFC). However, the interaction between these models and their combined effect on the risk of major depression in the workplace is largely unknown. The aim of this study is to examine the longitudinal combined effects of JD-C, ERI and WFC on the risk of major depression in the working population.
Methods Longitudinal data (2008-2013) were collected on randomly selected participants (n=4200) from the working population of the province of Alberta, Canada, at baseline and 1-, 2-, 3- and 4-year follow-up. Data about JD-C, ERI, WFC and major depression were collected by trained interviewers using a computer-assisted telephone interviewing method. Generalised estimating equations for longitudinal modelling were used.
Results There was an independent association between high ERI and high WFC at tx and major depression at tx+1 (OR 1.56, 95% CI 1.25 to 1.96; OR 1.33, 95% CI 1.16 to 1.52), respectively. The combined effects of JD-C and ERI, ERI and WFC, and WFC and JD-C on the risk of major depression were as follows: OR 1.71, 95% CI 1.22 to 2.42, OR 2.47, 95% CI 1.99 to 3.49 and OR 2.21, 95% CI 1.48 to 3.30, respectively. The relative excess risks attributable to the interactions were statistically non-significant.
Conclusions Work-related psychosocial factors are associated with increased risk of major depression over time, but their combined effect is not synergistic. The effects of the factors depicted in the three occupational health models on the risk of major depression appear to be additive.
- mental health
- occupational health practice
- longitudinal studies
Statistics from Altmetric.com
What this paper adds
Little evidence exists whether three widely used occupational health models, that is, job demand and control (JD-C), effort–reward imbalance (ERI) and work–family conflicts (WFC), interact, thereby elevating or reducing the risk of major depression in the workplaces.
Our findings suggest that the three prominent models had independent cumulative effect on the increased risk of major depression over 4 years. The associations with the risk of major depression varied by models such that ERI had a stronger association with major depression compared to those of JD-C and WFC.
The joint exposure of workplace psychosocial factors elevates the risk of major depression, but the effect was additive.
Major depression is the leading cause of disability. Globally, an estimated 350 million people of all ages suffer from major depression, which may become a serious health condition.1 2 Major depression often negatively impacts occupational functioning.3 4 Work-related psychosocial factors likely precipitate the onset of mental disorders and aggravate symptoms as well as the course of mood disorders among workers.5 6 Occupational stress and mental health problems are highly correlated. For example, work stress scores above the 75th percentile is associated with increased risk for major depression.7–9 High job demand (excessive workload, extreme time pressures) is associated with the onset of major depression in workers with no previous diagnosis of major depression,10–14 and work stress increases the likelihood of the onset of pre-existing condition, for example, chronic pain. Among vulnerable individuals, work stress may play an additive role in the manifestation of such conditions.
In occupational mental health research, there are three widely used theoretical models, namely, the job demand and control (JD-C) model by Karasek et al 15 the effort–reward imbalance (ERI) model by Siegrist16 17 and work–family conflict (WFC) model.18–20 The JD-C model characterises the job according to the combination of two dimensions: job demand and decision latitude, which composed of skill use and decision authority (job control).15 16 According to this model, high job strain (JS) or high JD-C is the result of high demand and low control at work, which can result in psychological stress and chronic illness, including depression. Lack of control, in particular, is a relatively consistent predictor of depression. Overall, the proportion of variance in depression explained by JS factors varies from 10% to 50%,21 which might be due to the difference in measures used and study settings.
The ERI model is based on the broader concept of the job, including work role, and reflects the individual worker’s emotional needs and characteristics.16 This model suggests that an imbalance between reward (income, esteem and occupational status control) and effort (demands and obligations) can lead to adverse health effects in the worker.16 Studies suggest that there is a positive association between high ERI and depression. An extensive review of the ERI model concludes that employees working in a high-effort, low-reward environment are more likely to experience psychosomatic symptoms, including depression.22 Furthermore, work demand may affect family responsibilities (work-to-family conflict), and family role may also interfere with one’s work (family-to-work conflict). The competing demands of work and family roles can cause stress and performance loss and decrease work satisfaction and organisational commitment, which is described in the WFC model. This contradictory need of an individual (ie, time and resources) makes participation in both roles more difficult.18 19 Such inter-role conflict increases the risk of depression and general well-being, as shown in cross-sectional studies.20 23 24
It is possible that JD-C, ERI and WFC share common constructs and pathways that interact with one another to increase or reduce the risk of developing major depression over time. For instance, exposure to both high WFC and high ERI could interact, thereby increasing the risk of major depression substantially. The interaction between JD-C, WFC and ERI and how this influences the risk of major depression are largely unknown. For this reason, the objective of this study was to examine the longitudinal relationship of JD-C, ERI, WFC and their combinations on the risk of major depression in the working population.
Study design and population
A longitudinal study was conducted among randomly selected participants from the working population in the province of Alberta, Canada, between 2008 and 2013. At baseline, 4300 eligible participants, aged 25–65 years who were residing or working in Alberta at the time of the interview, were included. The participants were recruited using random digit dialling. At the recruitment stage, 9776 potentially eligible individuals were contacted by telephone. These individuals were read a script that indicated that they were considered eligible if they were between the ages of 25 and 64 years, were currently working and residing in Alberta, would agree to be contacted for two annual follow-up interviews (through 2011) and would provide primary and secondary contact information. At the end of recruitment, 4302 eligible participants completed the baseline interviews. Among the 4302 participants, there were three duplications and 60 people who did not provide their names. These participants were excluded from the cohort, which resulted in 4239 baseline participants in the cohort. Out of these, 39 were also excluded for incomplete data on the variables of interest at baseline. For this study, a total of 4200 eligible participants were included at baseline, with the overall response rate of 71% at 12-month follow-up. These participants were re-interviewed at 2-, 3- and 4-year follow-up. Data were collected at baseline and follow-ups using a computer-assisted telephone interview. Detailed information about recruitment and follow-ups can be found in previous publications.12 13 The study was approved by the Conjoint Health Research Ethics Board of the University of Calgary.
We used the WHO’s Composite International Diagnostic Interview (CIDI) to assess major depression. The CIDI-Auto V.2.1 is a computerised program developed and supported by the WHO. After the completion of the interview, the scoring algorithm program in the CIDI-Auto V.2.1 automatically generated diagnoses based on criteria of the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition.25 The lifetime version of the CIDI-Auto V.2.1 was used at baseline to assess lifetime prevalence of major depression. During the follow-up, the 12-month version of the CIDI-Auto V.2.1 was used to determine whether major depression had occurred in the past 12 months.
Work-related psychosocial factors
Job demand and control
JD-C was assessed using the full version of the Job Content Questionnaire (JCQ) developed by Karasek et al.15 This 25-item scale evaluated JS in five dimensions of work stress: decision latitude (three items), skill discretion (six items), psychological demand (five items), job security (three items) and social support from supervisors and coworkers (eight items). Respondents’ answers to the questions on this 12-item instrument were based on a 5-point scale, from 0=strongly agree to 4=strongly disagree. A higher score indicates a higher level of JS. Reversed coding for some items was used. The dimensional scores were calculated using the formula provided in the JCQ manual.12 Higher scores on the JCQ scales indicate more job stress or strain. We used the scores of psychological demand, skill discretion and decision authority to compute a JS ratio (JSR) using the equation JSR=psychological demand/((skill discretion+decision authority)/2). As reflected by the formula, demand exceeds control when JSR scores increase. As the scores were skewed, we defined high JD-C as having a JSR score more than 75th percentile during analyses.
ERI was assessed by the ERI questionnaire.16 Determination of ERI focused on two elements: perceived effort (five items) and rewards (11 items). According to Siegrist et al the imbalance between effort and reward was determined by a ratio according to the formula: e/(r×c) where ‘e’ is the sum score of the effort scale, ‘r’ is the sum score of the reward scale and ‘c’ defines a correction factor (0.454545) for different number of items in the denominator and numerator.16 A higher ERI ratio indicated a higher level of imbalance. To compare across the scales and consistent with previous research,11 14 we defined high ERI as having a ratio score more than 75th percentile.
WFC were assessed using an eight-item work-to-family conflict scale.13 The work-to-family conflict questionnaire included separate scales for work-to-family conflict and family-to-work conflict, each of which contained four questions. Each question assessed potential conflict and had three possible answers: not at all, to some extent and a great deal. Each subscale had a possible summary score that ranged from 4 to 12. The Cronbach’s α values of the two subscales were 0.56 and 0.73, respectively. The work-to family conflict scores were highly skewed. To make meaningful comparisons, we defined high WFC as having scores more than 75th percentile.
Other variables, including age (continuous variable), sex, marital status (married/common-law partnership, single/never married or separated/divorced/widowed), educational level (less than high school, high school and college, or university), personal annual income (<$C30 000, $C30 000–$C59 999, $C60 000–$C79 999 or $C80 000), race/citizenship, job type (full time, part time and other), job position, major depression in the past year and average weekly number of working hours,6 9 13 23 26 were assessed at baseline.
First, we described the baseline sample characteristics. Second, in order to examine the association between earlier work conditions and later major depressive episode, we used generalised estimating equations (GEE) time-lag models. Unlike standard models, such time-lag model provides the relationships between changes between subsequent time-points (‘within-subject’ relationships), which takes into account the temporal sequence of a possible cause and effect.27 GEE allows correlated observations over time and missing values at different measurement points. We used an exchangeable correlation structure to take within-subject dependencies into account. GEE (assuming missing completely at random) give unbiased effect estimates taking into account missing data.27 In this structure, we assumed the correlations between subsequent measurements to be the same, irrespective of the length of the time interval. Moreover, the dataset was reconstructed from broad structure into long format, since there is no statistical software capable of performing the analysis automatically. The output of the GEE analysis used to answer the question of whether there is a longitudinal relationship between outcome variable Y (ie, major depression) and the predictor variables X1 to X4, in which each patient ID had four observations (time-points). The response rate during follow-up was around 70%, and the total number of observations would be four times that of the remaining sample. In GEE time-lag models, the value of outcome at time-point Tj are predicted by predictor X at time-point Tj−1. Thus, we examined the longitudinal associations of JD-C at tx, ERI at tx and WFC at tx with major depression tx+1 using with tx being a given wave and tx+1 being the next wave.27 All analyses were adjusted for job sector, sex, age, education (low, medium, high) and job type (full time vs part time), and history of major depression. Age (continuous variable), sex (male/female), marital status (married/common-law partnership, single/never married or separated/divorced/widowed), educational level (less than high school, high school and college, or university), personal annual income (<$C30 000, $C30 000–$C59 999, $C60 000–$C79 999 or $C80 000), job type (full time, part time and other), history of depression and average weekly number of working hours.
Third, we examined whether the joint association of high JD-C tx and high ERI tx or high JD-C tx and high WFC tx or high ERI tx and high WFC tx with major depression tx+1 is stronger than the sum of separate associations of high JS and ERI with major depression. We created a four-category variable: low ERI and low JS, high JS, high ERI, and high ERI and high JS. If high JS was present, then i=1; otherwise, i=0. If high ERI was present, then j=1; otherwise, j=0. Then, ORij represented the OR in both high JS and high ERI category i, j. The three OR estimates ((ie, OR11, OR10, OR01) and OR00 (reference category)) were computed from GEE analyses. The relative excess risk due to interaction (RERI) was used to assess the additional risk due to joint exposures of work stress from JS, WFC and ERI compared with their separate risks for major depression. RERI was defined as RERI=OR11−OR10−OR01+1.28 29 CIs for the RERI were calculated using covariances of parameter estimates from GEE models. All analyses were performed using STATA V.14 and SPSS V.24 software.
Table 1 shows the baseline characteristics of the participants (n=4200) interviewed and included in the study at baseline (2008/2009). Of these, 55% were women, and majority (82%) of the participants was full-time workers. The average working hours of the participants at baseline was 41.3 hours per week with an SD of 12.4. The lifetime and 12-month prevalence rates of major depression at baseline were 23% and 7%, respectively.
The loss to follow-up regarding depression (availability of fewer than two assessments) was 22% and was associated with lower age, lower educational status and income and being non-Canadian, but not associated with gender, working hours and lifetime and 12-month depression diagnosis at baseline. Major depression at baseline was significantly correlated with major depression at 1-, 2-, 3- and 4-year follow-up (p<0.001), indicating stability of major depression over time.
The separate longitudinal effect of JD-C, ERI and WFC on the risk of depression
We found longitudinal associations between high ERI at tx and major depression at tx+1 (OR 1.56, 95% CI 1.25 to 1.96) after adjustment of age, gender, educational status, marital status, income and race/citizenship, working hours and job type, and history of depression (table 2). In addition, we found a significant association between high WFC at tx and major depression at tx+1 (OR 1.33, 95% CI 1.16 to 1.52). However, the longitudinal association between high JD-C and major depression was not statistically significant.
The combined effect of ERI, JS and WFC on major depression
The longitudinal combined effect of high ERI and high JS at tx and major depression at tx+1 compared with those reported low JS and low ERI was as follows: OR 2.18, 95% CI 1.60 to 2.96 (table 3). After adjustment of age, gender, educational status, marital status, income, working hours, job type, history of depression and race/citizenship, the association attenuated but remained statistically significant (OR 1.71, 95% CI 1.22 to 2.42). We also found a longitudinal combined effect of high ERI and high WFC with major depression (OR 2.47, 95% CI 1.99 to 3.49), and of high JS and high WFC (OR 2.21, 95% CI 1.48 to 3.30; table 3). However, the respective RERIs for these job stress models with respect to major depression were not statistically significant, indicating that there was no synergistic effect between these models on the risk of major depression.
The purpose of this longitudinal study was to examine the combined effects of JD-C, WFC and ERI on major depression in a representative working population. Overall, the findings show that these occupational health models had independent cumulative effect on the risk of major depression over 4 years.
Previous studies show that each of these occupational health models are independently associated with major depression.7 11 14 26 The present longitudinal study confirmed previous cross-sectional findings regarding the independent effect of these models on the risk of major depression. Furthermore, the present study provides evidence for the combined effect of individual occupational health models on risk of major depression. We found that high ERI along with high JD-C was associated with an increased risk of major depression and was slightly larger than what would have been expected on the additive scale. For employees reporting both high ERI and high JD-C, the additional risk of major depression was 100%, while the risk attributable to high ERI and high JD-C alone was 92% (18% to JD-C and 74% to ERI), leading to a RERI of 0.06. Similarly, the additional risk of major depression due to joint exposures of high ERI and WFC was 190%, leading to a RERI of 1.03. However, these additional risks were statistically non-significant (ie, no synergistic effect). The non-significant RERI may be partly due to the construct overlap between these models. For instance, high efforts and low rewards are the strongest predictors of emotional exhaustion, psychosomatic health complaints (psychological demands) and physical health symptoms,30 31 which might indirectly explain the effect of JS on major depression, since JS is a ratio of psychological demands and decision control according to JD-C model.15 The non-significant RERI may also be due to a relatively small sample size for the subgroup analysis of high ERI, JD-C and WFC status categories with major depression.
One important finding was that risk of depression was more strongly associated with ERI compared with JD-C and WFC, suggesting that an imbalance between effort and reward puts individuals at a higher risk for developing major depression when compared with other factors. This is likely because effort is the strongest predictor of emotional exhaustion, psychosomatic health complaints (psychological demands) and physical health symptoms. Efforts such as psychological and physical demands at work and rewards such as money (ie, adequate salary), esteem (eg, respect and support) and security/career opportunities (eg, promotion prospects, job security and status consistency) provide a more accurate estimate of experienced stress than a restriction to one of these sources.16 17 32 In a high-effort, low-reward condition, individuals are at a higher risk for emotional distress, which often leads to somatic health problems and other strain reactions (ie, poor subjective health and sickness absence). For example, having a demanding but unstable job or achieving a high performance at work without receiving a promotion in response to this accomplishment can result in stress imbalance. Sustained stress imbalance elevates corticosteroids, which directly affects certain brain structures and induces a downregulation of glucocorticoid receptors, thereby impairing affect regulation.33 Interestingly, similar physiological responses have been shown in patients with job-stress-related depression. Moreover, when there is high job demand and a lack of decision latitude or sustained work and family role conflict, the individuals may have feelings of helplessness and hopelessness. If this persists for longer period of time, an individual can be conditioned to the phenomena called learnt helplessness, which is a well-known psychological cause and manifestation of major depression.
Strength and limitations
The major strength of our study is its longitudinal design. A longitudinal study design is effective in delineating the temporal relationship between job stress exposures and major depression. In this study, we were able to examine longitudinal associations between occupational exposures at one point in time with major depression 1 year later over a 4-year follow-up period by using time-lag models (ie, temporal relationship was maintained). These support assumptions on causality, even though not conclusive ones. Unlike previous studies, this longitudinal study simultaneously incorporated and compared the models of JD-C, ERI and WFC.
There were a few limitations in this study. First, the sample size was relatively small for subgroup analysis and computing interaction. This is likely a result of the strict inclusion criteria for the cohort during the recruitment stage. Second, a random digit dialling method was used to recruit participants, which results in selection bias. However, given the good response rate in the follow-up and the use of weights, it is unlikely that selection bias had a significant effect on the results. Third, the present study relied on self-report. Since the aetiology of major depression is complex and multifactorial, the observed risks could be related to other factors, such as chronic physical conditions, and stressful negative events. This may be an alternative explanation for the association between the work stress model and the risk of depression. Although the history of depression was controlled during analysis, some people with minor depression may have been included. The perception of the work environment could be affected by milder depressive symptoms, leading to confounding results. On the other hand, minor depression at baseline could be the result of a negative work environment, meaning that minor depression may become part of a causal chain linking work environmental factors and depression.
In conclusion, exposure to high JD-C, high ERI and high WFC may independently increase the risk of major depression. The relative excess risk of major depression due to the combined effects of these models was not larger than the sum of their separate effects (ie, no synergistic effect).
Contributors YTN and JW conceived and designed the study. YTN performed statistical analyses and drafted the manuscript. JW critically reviewed the article for important intellectual content. All authors have approved the final article and accept full responsibility for the design and the conduct of the study, and all approved the decision to publish.
Competing interests None declared.
Ethics approval The study was approved by the Conjoint Health Research Ethics Board of the University of Calgary.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.