Psychosocial Occupational Exposures and Mental Illness

Abstract

Background

The purpose of this study is to investigate the exposure-risk relationship between psychosocial occupational stress and mental illness.

Methods

We conducted a systematic review with meta-analyses as an update of a systematic review published in 2014. The study protocol was registered in PROSPERO (CRD42020170032). Literature searches were carried out in the MEDLINE, PsycINFO, and Embase databases. All procedural steps were performed independently by two reviewers; discordances were solved by consensus. All of the included full texts were subject to a methodological appraisal. Certainty of evidence was determined with the GRADE procedure.

Results

The pooled risk of depression was found to be approximately doubled in workers exposed to high job strain, which is defined as high work demands combined with low job control (effect estimate [EE] = 1.99, 95% CI [1.68; 2.35], heterogeneity [I2] = 24.7%, n = 8). In particular, high work demands are associated with incident depression (ES = 13.8 [1.19; 1.61], I2 = 69.0%, n = 9) and with incident anxiety disorder (ES = 1.79 [1.44; 2.23], I2 = 48.1%, n = 5). There were only a small number of methodologically adequate studies available on burnout, somatoform disorders, suicidal ideation, and suicide. Thus, no pooled risk estimates were calculated, although some individual studies showed a considerably increased risk.

Conclusion

Psychosocial occupational stress is clearly associated with depression and anxiety disorders.

Unfavorable psychosocial working conditions may pose a risk for the occurrence of mental illness (1– 6). Insight into the exposure-risk relationship is crucial for developing targeted preventive measures. The so-called doubling dose is important for the recognition of a new occupational disease: if a specific exposure is associated with a doubling of the risk of disease, it can be assumed—as a rule of thumb—that an occupational causation of this disease is more likely than unlikely. Our systematic review with meta-analyses investigates the association between psychosocial working conditions and an increased risk of mental distress or illness and analyzes the corresponding exposure-risk relationship. In addition, we examined which job groups or work-related activities are associated with a particularly high risk.

Methods

We conducted a systematic review, building on the methodology used in a systematic review published in 2014 which was carried out on behalf of the German Federal Institute for Occupational Safety and Health (BAuA, Bundesanstalt für Arbeitsschutz und Arbeitsmedizin) (7, 8).The PRISMA statement was followed for quality assurance and reporting (9). The study protocol was registered in PROSPERO (CRD42020170032).

Inclusion and exclusion criteria

The PECOS criteria were used to specify the selection criteria (10):

• Population: the working population

• Exposure: job strain (defined as high work demands combined with low job control [or decision latitude]), gratification crisis, effort-reward imbalance (defined as high efforts spent and low rewards received), lack of social support, mobbing, negative work climate, workplace conflict, job insecurity, organizational injustice, low discretionary power, long hours, shift and night work, experience of violence, sexual harassment.

• Comparator: workers not subjected to the respective exposure

• Outcome: depression, anxiety disorder, burnout, somatoform disorder, suicidal ideation, suicide

• Study designs: cohort studies, case-cohort studies, case-control studies, and mortality studies with follow-up duration longer than one year and reporting of effect estimates of the relative risks of disease; exclusion of cross-sectional studies, case series and case studies.

Search strategy

We conducted a literature search in the MEDLINE, PsycINFO and Embase electronic databases for the period 1 January 2014 to 26 June 2020 and combined this literature search with the literature search carried out in 2014. Latin-script publications in scientific journals with German or English abstracts available online were included. The search strings and further methodological information can be found in the eMethods section.

Study selection

Title/abstract screening and full-text screening of articles published after 2013 were each carried out by two persons independently (SD, AF, FH, MS). The studies included in the systematic review published in 2014 were also reviewed by two researchers independently to determine whether they met the more stringent inclusion criteria of the current review (follow-up period of at least one year, reporting of effect estimates of the relative risk of disease). Discordances were solved by consensus.

Data extraction

Data extraction was carried out by two researchers independently. Study information on reference, methodology, population, exposure, outcome, and results were documented in standardized data extraction tables.

Rating of methodological study quality

The methodological quality of the included studies was assessed by two researchers independently (SD, AF, FH, MS; solving of discordances: AS) using a tool that was developed for the 2014 systematic review based on the SIGN (Scottish Intercollegiate Guideline Network) and CASP (Critical Appraisal Skills Program) tools (8, eMethods).

Statistical analysis

Only studies with adequate methodological quality were included in the synthesis of evidence. A random-effects model was used for the meta-analytical calculation of pooled effect estimates with 95% confidence intervals if at least three individual risk estimates were available for comparable exposure-outcome combinations. Heterogeneity was evaluated by I². The Stata metan package was used for the analyses. Presence of publication bias was examined using funnel plots (for at least 10 individual risk estimates). Confidence of evidence was determined with the GRADE procedure (11).

Results

Results of the literature search

The results of the literature search are shown in the PRISMA flow chart (efigure 1). The literature search for studies published after 2013 yielded 17 713 hits. Title/abstract screening yielded 464 publications for full-text screening. The inclusion criteria were met by 41 publications which were included in the systematic review (eTabelle 1). An additional 18 publications were included from the 2017 review (7, 8) (eTabelle 1). The excluded publications with references and reasons for exclusion are listed in eTable 2.

eFIGURE 1.

PRISMA flowchart (according to Moher et al. [9]) on search and inclusion of studies

Study characteristics

A total of 59 publications from 32 studies were available for evidence synthesis. With the exception of two studies on suicide risks, only cohort studies were included. Most of the included studies were from European countries, especially from the Scandinavian region, and usually included men and women with no limitation to a specific profession. eTabelle 1 provides an overview of the included studies, specifying the respective included population, exposure and outcome. Detailed extraction tables for the review update are presented in eTabelle 3 and for the 2014 review in the German Federal Institute for Occupational Safety and Health (BAuA) report (8) where the included confounders are listed, too.

Results of the rating of methodological study quality

No study was found to have a very low risk of bias. Study-specific information can be found in the eMethods section.

Results for depression

Jobs characterized by high work demands and low job control (high strain)

A total of ten studies investigated the association between job strain and depression (eTable 1). A more detailed definition of job strain can be found in the eMethods section. Eight studies were included in the meta-analysis (table 1). Four studies used the traditional classification into low-strain jobs, passive jobs, active jobs, and high-strain jobs (12– 15). The results of these four studies show that working in high strain jobs is associated with a 73% increase in risk of depression compared to working in low-strain jobs (characterized by high job control and low work demands) (pooled effect estimate [EE] for risk of disease: 1.73; 95% confidence interval: [1.32; 2.27]; I²= 15.8%; n = 4, Table 2, eFigures 2a– c). A dichotomous analysis of job strain (based on the ratio of job control to work demands) found a doubling of risk of depression with high job strain (EE: 1.99 [1.68; 2.35]; I²= 24.7%; n = 8, Figure 1). The risk estimates for men and women were about comparable (eFigures 3a, b). When different courses of job strain over time were examined, continuously high job strain was associated with the highest risk (EE: 2.73 [1.96; 3.80]; I2 = 11.8%; n = 3) (figure 2). The second highest risk was found for job strain levels that increase over time (efigure 4a), followed by job strain levels that decrease over time (efigure 4b).

Table 1. Results of the meta-analyses—Research question 1.

Exposure	Studies	Odds ratio [95% CI]	Heterogeneity (I2) Classification
Question: Are certain psychosocial working conditions associated with an increased risk of developing mental distress or illness (especially burnout or depression)?
Depression, studies included in meta-analysis
High work demands	14, e1–e8	1.38 [1.19; 1.61]	69.0%, considerable
Only women	e2, e4, e6	1.31 [1.17; 1.47]	0%, no heterogeneity
Only men	e2, e4, e6	0.93 [0.48; 1.79]	87.4%, considerable– substantial
Low job control	14, e1–e12	1.14 [1.04; 1.24]	54.8%, moderate–considerable
Only women	e2, e4, e6, e9	1.18 [0.90; 1.56]	56.9%, moderate–considerable
Only men	e2, e4, e6, e9	1.15 [1.15; 1.27]	0%, no heterogeneity
High job strain	17, 28, 30, e2, e13–e16	1.99 [1.68; 2.35]	24.7%, might not be important
Only women	e2, e14, e15	1.68 [1.21; 2.32]	0%, no heterogeneity
Only men	e2, e14, e15	1.88 [1.00; 3.57]	51.9%, moderate–considerable
Low social support	15, e1, e2, e17	1.39 [1.13; 1.72]	28.9%, might not be important
Low social support from superiors	e10, e16, e18	1.15 [0.96; 1.38]	0%, no heterogeneity
Low social support from coworkers	e7, e10, e15, e18	1.27 [0.99; 1.64]	48.1%, moderate–considerable
Effort-reward imbalance	30, e11, e19, e20	1.72 [1.50; 1.97]	23.9%, might not be important
High emotional demands	e1, e7, e21	1.42 [1.09; 1.84]	0%, no heterogeneity
Job insecurity	e1, e6–e8, e22	1.51 [1.19; 1.91]	84.2%, considerable–substantial
Anxiety disorders
High work demands	e1, e7, e8, e23, e24	1.79 [1.44; 2.23]	48.1%, moderate–considerable
Low job control	e1, e7–e9, e23, e24	1.17 [1.00; 1.36]	38.0%, moderate
Low social support	e1, e7, e23	1.34 [0.81; 2.22]	88.7%, considerable– substantial
Job insecurity	e1, e7, e8, e23	1.55 [1.33; 1.81]	16.6%, might not be important

95% CI, 95% confidence interval

Table 2. Results of the meta-analyses—Research question 2.

Exposure	Studies	Odds ratio [95% CI]	Heterogeneity (I2), Classification
Question: How can the exposure-risk relationships between psychosocial working conditions and the occurrence of mental illness be described?
Depression
High work demands
Highest vs. lowest category	12, e1, e3, e4, e5, e8	1.53 [1.09; 2.14]	65.2%, considerable
Second highest vs. lowest category		1.27 [1.01; 1.60]	25.1%, might not be important
Low job control
Lowest vs. highest job control	e1, e3, e4, e5, e8, e10	1.05 [0.73; 1.52]	65.1%, considerable
Second lowest vs. highest job control		0.91 [0.87; 0.95]	0%, no heterogeneity
Job strain
High vs. low job strain	12–15	1.73 [1.32; 2.27]	15.8%, might not be important
Active job vs. low job strain		1.31 [0.95; 1.80]	34.6%, moderate
Passive job vs. low job strain		1.27 [0.97; 1.66]	9.1%, might not be important
Change of job strain
From high to low	e2, e15, e16	1.48 [1.11; 1.98]	0%, no heterogeneity
From low to high		2.15 [1.61; 2.88]	1.6%, might not be important
Constantly high		2.73 [1.96; 3.80]	11.8%, might not be important

95% CI, 95% confidence interval

eFigure 2a.

Risk of depression: high vs. low job strain

eFigure 2c.

Risk of depression: Passive job vs. low job strain

eFigure 3a.

Risk of depression and high job strain for women (dichotomous: high vs. low)

eFigure 3b.

Risk of depression and high job strain for men (dichotomous: high vs. low)

Figure 2.

Risk of depression and continuously high job strain.

Continuously high job strain is found to be associated with the highest risk of depression.

95% CI, 95% confidence interval; m, men; w, women

eFigure 4a.

Risk of depression and increasingly high job strain

eFigure 4b.

Risk of depression and decreasingly high job strain

Most studies included various occupations without specifying them more precisely. In the study of Wieclaw et al. (16), the occupational groups with the highest job strain are laborers, tradespeople, and cleaners among both men and women, as well as jobs in construction, customer service, and operating machinery.

Nine studies investigated high work demands as one dimension of job strain (eTable 1). The result of the meta-analysis shows that with dichotomous evaluation, high psychosocial work demands increase the risk of depression by 38% (Table 1, eFigure 5a) compared to low work demands. In the gender-separated analysis, an increased risk is found in women but not (with marked heterogeneity) in men (eFigures 5b, c). The comparison of the highest with the lowest category of psychosocial work demands (efigure 6a) finds an increase in risk of 53%, while the comparison of the second highest with the lowest category finds an increase in risk by 27% (efigure 6b).

eFigure 5a.

Risk of depression and high psychosocial work demands (dichotomous: low vs. high)

eFigure 5b.

Risk of depression and high psychosocial work demands for women (dichotomous: high vs. low)

eFigure 5c.

Risk of depression and high psychosocial work demands for men (dichotomous: high vs. low)

eFigure 6a.

Risk of depression and high psychosocial work demands (dichotomous: highest vs. lowest category)

eFigure 6b.

Risk of depression and high psychosocial work demands (dichotomous: second highest vs. lowest category

The dimension of job control is examined in 14 studies (eTable 1). The result of the meta-analysis shows that lower job control as a further dimension of job strain increases the risk of depression by 14% (efigure 7). All meta-analyses are summarized in Table 1 and Table 2.

eFigure 7.

Risk of depression and low job control (dichotomous: low vs. high)

Further psychosocial work-related strain

A job gratification crisis (effort-reward imbalance, ERI) is characterized by a mismatch between effort spent and rewards received at work in the form of recognition, salary and job security, among others. In the included studies, an effort-reward imbalance increased the risk of depression by 72% (EE: 1.72 [1.50; 1.97], I²= 23.9%, n = 4) (efigure 8). Our meta-analysis found that low social support at work and high job insecurity are associated with increases in risk of depression by 39% (efigure 9) and 51% (efigure 10), respectively. Lastly, high emotional work demands were found to be associated with a 42% increase in risk of depression (efigure 11).

eFigure 8.

Risk of depression and gratification (effort-reward imbalance) (dichotomous: high vs. low)

eFigure 9.

Risk of depression and social support (dichotomous: high vs. low)

eFigure 10.

Risk of depression and job insecurity (dichotomous: high vs. low)

eFigure 11.

Risk of depression and high emotional demands (dichotomous: low vs. high)

Results for anxiety disorders

Jobs characterized by high work demands and low job control (high strain)

For job strain, none of the included studies reported on the risk of anxiety disorder. The dichotomous analysis found for high psychosocial work demands, one dimension of high job strain, a 79% increase in risk of anxiety disorder (EE: 1.79 [1.44; 2.23], I²= 48.1%, n = 5, Table 1, eFigure 12). In the meta-analysis results, low job control, one dimension of high job strain, was associated with a 17% increase in risk of anxiety disorder (EE: 1.17 [1.00; 1.36], I²= 38.0%, n = 6) (efigure 13).

eFigure 13.

Risk of anxiety disorder due to low job control (dichotomous: low vs. high)

Further psychosocial work-related strain

No study on the association between effort-reward imbalance and anxiety disorder could be included in our meta-analysis. For high job insecurity, the dichotomous analysis found a 55% increase in risk of anxiety disorder (efigure 14).

eFigure 14.

Risk of anxiety disorder due to job insecurity (dichotomous: high vs. low)

Results for further mental impairments and illnesses

Only two of the included studies on burnout found at times significant risk increases, especially with high job strain, high work demands and high emotional demands at work (17, 18). In one of the available studies, somatoform disorders were associated with (quantitative) work demands, in men with low job control and in women with emotional work demands (16).

For suicidal ideation (n = 5 studies) and completed suicides (n = 3), studies found in some cases significantly increased risks for high job strain and its dimensions “high work demands” and “low job control”, for effort-reward imbalances, for threats of physical violence, excessive working hours, psychologically stressful customer contact, and work that is monotonous or involves a high level of responsibility (19– 26).

Publication bias and confidence of evidence

Confidence of evidence with regard to the association between job strain and incident depression were rated as high, based on the GRADE assessment. The confidence of evidence with regard to the association between high work demands/low job control and depression/anxiety disorder is formally low (etable 4). No evidence of publication bias was found (Egger’s test: p>0.30).

eTable 4. Determination of confidence of evidence with GRADE.

Number of studies	Overall risk of the study: ↓	Indirectness: ↓	Inconsistency: ↓	Imprecision, Range Confidence interval >2.0: ↓	Publication bias, yes or unclear: ↓	Large effect size ≥ 2.0: ↑ ≥ 5.0: ↑↑	Dose-effect relationship:↑	Residual confounding: ↑	Quality levels (high/moderate/low)
Risk: High job strain and depression
8	No (–) *1	No (–)	No (–) *2	No (–) *3	Unclear ↓	Yes ↑ *4	Yes ↑*5	No (–)	High
Risk: High work demands and depression
9	No (–) *1	No (–)	Yes ↓*6	No (–) *7	Unclear ↓	No (–)	Yes↑*8	No (–)	Low
Risk: Low job control and depression
14	No (–) *1	No (–)	Yes ↓*9	No (–) *10	No (–)	No (–)	No (–)	No (–)	Low
Risk: High work demands and anxiety disorders
5	No (–) *1	No (–)	Yes ↓*11	No (–) *12	Unclear ↓	No (–)	No (–)	No (–)	Low
Risk: Low job control and anxiety disorders
6	No (–) *1	No (–)	Yes ↓*13	No (–) *14	Unclear ↓	No (–)	No (–)	No (–)	Low

*¹ All included studies were “+“.

*² Heterogeneity (I 2): 24.7 %

*³ 95% confidence interval (CI): [1.68; 2.35]

*⁴ Relative risk (RR): 1.99 [1.68; 2.35]

*⁵ From high to low job strain: RR: 1.48 [1.11; 1.98];

From low to high job strain: RR: 2.15 [1.61; 2.88];

Continuously high job strain: RR: 2.73 [1.96; 3.80]

*⁶ I² = 69.0% (moderate heterogeneity)

*⁷ 95% CI: [1.19; 1.61]

*⁸ From highest to lowest work demands: RR: 1.53 [1.09; 2.14];

From the next highest to the lowest work demands: RR: 1.27 [1.01; 1.60]

*⁹ I² = 54.8% (moderate heterogeneity)

*¹⁰ 95% CI: [1.04; 1.24]

*¹¹ I² = 48.1% (moderate heterogeneity)

*¹² 95% CI: [1.44; 2.23]

*¹³ I² = 38.0 % (moderate heterogeneity)

*¹⁴ 95% CI: [1.00; 1.36]

Discussion

Consistent with previous reviews (1– 5), our systematic review with meta-analysis indicates that jobs with high psychosocial work demands and low job control are associated with a significantly increased risk of depression (high strain jobs) (1, 3). This high risk appears to be primarily attributable to the high work demands. In contrast, the increased risk of depression in the analysis of Madsen et al. (3) is driven by low job control, not high work demands. Consistent with Kivimäki et al. (27), the risks of depression associated with high strain jobs, which are significantly higher than those for high work demands and low job control alone, suggest a combined effect of these two components of job strain. A Finnish study (28), evaluating the impact of combined psychosocial work-related strain on the risk of depression-related “disability pension” found evidence of a particularly high risk of depression (hazard ratio: 4.40 [2.43; 7.96]) associated with high strain and effort–reward imbalance (28). In addition, if organizational injustice is perceived by employees, their risk of depression increases even further.

In our meta-analysis, the association between high strain and depression reached the level of a doubling of the risk. Other reviews report slightly lower (1, 3) or significantly lower effect estimates ([3] in a subanalysis that includes individual participant data [IPD] [29]).The higher risk estimates in our meta-analysis compared to the reviews presented above are largely due to the inclusion of more recent follow-up data from ongoing cohort studies or newer primary studies and to different inclusion and exclusion criteria (exclusion of studies with self-reported diagnosis of depression and of studies using treatment with antidepressants as the sole outcome definition, among others, in our review). For anxiety disorders, similar stressful factors as for depression were identified.

Exposure-risk relationship: psychosocial occupational exposures and mental illness

A major strength of this systematic review is that it evaluated not only the risk of mental illness associated with specific psychosocial exposures, but also its magnitude (the “dose”). The depression risk comparison across exposure categories (inclusion of the highest and the second highest exposure categories in separate meta-analyses) indicates a positive dose-response relationship with respect to depression risk for several psychosocial work-related exposures (inclusion of the highest and the second highest exposure categories in separate meta-analyses). There is some evidence that prolonged exposure also increases the risk of mental illness. However, in most cases it was only possible to perform a dichotomous analysis. Consequently, the risk estimates calculated in the meta-analysis may be significantly underestimated.

Strengths and limitations of our systematic review

A strength of our systematic review is that it included only epidemiological studies with a follow-up duration of at least one year. In this way, reverse causality (i.e., the comparatively unfavorable rating of working conditions by individuals with preexisting mental illness) was avoided to the greatest extent possible.

As a point of criticism, it should be noted that only a small number of primary studies met the relatively stringent inclusion criteria, one of which was adequate study quality. Since the included primary studies did not take into account the patients’ entire lifetime, results may be biased. New-onset mental illness cannot be clearly distinguished from further episodes of preexisting illnesses which occur at least once in the lifetime of at least half of patients with depression (31). However, our review included only studies in which the target illnesses (depression, anxiety disorder) were excluded at the time of the baseline survey; alternatively, the statistical models used had to have at least included adjustments for the target illness at baseline. As a further potential source of bias, mental illness with onset early in life could have led to selection into specific job groups.The impact of preclinical depression or anxiety disorders on the mostly self-reported psychosocial work demands could have resulted in an overestimation of the assessed associations, especially when self-report data were used in combination with survey-based outcome measurement (common method bias [29]). All included studies take at least age and gender as confounders into account; the fact that socioeconomic status was not considered in some studies may, in principle, have led to an overestimation of the associations. In contrast, long follow-up periods can lead to overestimation of the associations if changes in working conditions that have occurred in the meantime are not taken into account. None of the primary studies met the requirements for very high study quality (“++”)—equivalent to a very low risk of bias—thus, there is a need for further research. The question to what extent job-related risks of disease (in terms of effect modification) may be influenced by social/cultural background or history of migration (32), is another aspect requiring further research.

Conclusions

This systematic review shows a clear association between psychosocial occupational exposures and mental illness. A particularly high risk of depression is found among workers exposed to high job strain, i.e. high work demands combined with low job control.

For everyday clinical practice, the following conclusions can be drawn from our systematic review:

• The occupational history should include psychosocial working conditions, such as time-related and emotional work demands, job control, recognition, and job security, among others.

• Avoiding excessive work demands and ensuring job control can contribute to the prevention of mental illness.

• If there are signs of impending depression or anxiety disorder due to high levels of occupational stress, it is advisable to consult with the occupational physician responsible for the employee (in agreement with the patient).

Supplementary Material

eMethodenteil

Review questions

For more information on the PECOS-framework and the inclusion and exclusion criteria see eTable 5.

Search strings of electronic database search (see also 7, 8)

Scientific publications in Latin with a German or English title and abstract were included. Search strings are presented below:

MEDLINE

Population 1*:

occupational diseases [MH] OR occupational exposure [MH] OR occupational exposure* [TW] OR “occupational health” OR “occupational medicine” OR work-related OR working environment [TW] OR at work [TW] OR work environment [TW] OR occupations [MH] OR work [MH] OR workplace* [TW] OR workload OR occupation* OR worke* OR work place* [TW] OR work site* [TW] OR job* [TW] OR occupational groups [MH] OR employment OR worksite* OR industry

Exposure 1*:

job strain OR mental strain OR occupational strain OR work strain or mental load OR workload OR work load OR organisational justice OR work stress OR job stress OR psychosocial work* OR organizational justice OR organisational injustice OR organizational injustice OR time pressure OR pressure of time OR harassment OR (effort AND reward) OR demand* OR shift work OR mobbing OR bullying OR leadership OR social relations OR social support OR job insecurity OR downsiz*

Population 2*:

occupational diseases [MH] OR occupational exposure [MH] OR occupational medicine [MH] OR occupational risk [TW] OR occupational hazard [TW] OR (industry [MeSH Terms] mortality [SH]) OR occupational group* [TW] OR work-related OR occupational air pollutants [MH] OR working environment [TW]

Exposure 2*:

stress OR conflict OR support OR climate

Outcome*:

mental disorders [MeSH] OR burnout OR mental health OR anxiety OR depress* OR emotional disorder* OR exhaust* OR psychosomatic OR somatoform* OR suicid*

Study Design*:

randomized controlled trial OR RCT OR intervention OR Epidemiologic studies [MeSH] OR Case control [tw] OR cohort study [tw] OR cohort studies [tw] OR Cohort analy*[tw] OR follow up study [tw] or follow up studies [tw] OR observational study [tw] or observational studies [tw] OR prospective study OR Longitudinal [tw] OR Retrospective [tw] NOT therapy NOT ((animals [Mesh:noexp]) NOT (humans [Mesh]))

*Suchterme wurden wie folgt verbunden: ((Population 1 AND Exposure 1) OR (Population 2 AND Exposure 2)) AND Outcome AND Study Design

Embase

Population

occupational disease$.mp. or occupational disease/ or occupational exposure$.mp. or occupational exposure/ or occupational health.mp. or occupational health/ or occupational medicine.mp. or occupational medicine/ or work-related.mp. or working environment$.mp. or work environment$.mp. or work environment/ or at work.mp. or occupation$.mp. or occupation/ or work.mp. or work/ or workplace/ or workplace$.mp. or workload.mp. or workload/ or worke$.mp. or work place$.mp. or work site$.mp. or job$.mp. or occupational group$.mp. or named groups by occupation/ or employment/ or employment.mp. or worksite$.mp. or industry/ or industry.mp.

Exposure

job strain.mp. or mental strain.mp. or occupational strain.mp. or work strain.mp. or mental load.mp. or mental load/ or workload.mp. or workload/ or work load.mp. or organisational justice.mp. or work stress.mp. or job stress.mp. or job stress/ or psychosocial work$.mp. or organizational justice.mp. or organisational injustice.mp. or organizational injustice.mp. or time pressure.mp. or pressure of time.mp. or harassment.mp. or harassment/ or effort reward.mp. or demand$.mp. or shift work.mp. or shift worker/ or shift work/ or mobbing.mp. or bullying.mp. or bullying/ or leadership.mp. or leadership/ or social relation$.mp. or social interaction/ or social support.mp. or social support/ or job insecurity.mp. or job satisfaction/ or downsiz$.mp.

Outcome

mental disorder$.mp. or mental disease/ or mental disease$.mp. or burnout.mp. or burnout/ or mental health.mp. or mental health/ or anxiety.mp. or anxiety disorder/ or anxiety/ or depress$.mp. or depression/ or emotional disorder$.mp. or emotional disorder/ or exhaust$.mp. or psychosomatic.mp. or psychosomatics/ or somatoform disorder/ or somatoform$.mp. or adjustment disorder$.mp. or adjustment disorder/ or affective disorder$.mp. or mood disorder/ or occupational disease/ or work related illnesse$.mp. or suicide attempt/ or suicide/ or suicide.mp.

Study design

randomized controlled trial.mp. or randomized controlled trial/ or RCT.mp. or intervention.mp. or intervention study/ or epidemiologic studies.mp. or epidemiologic study.mp. or case control study/ or case control.mp. or cohort study.mp. or cohort studies.mp. or cohort analysis/ or cohort analy$.mp. or follow up study.mp. or follow up/ or follow up studies.mp. or observational study.mp. or observational study/ or observational studies.mp. or prospective study.mp. or prospective study/ or prospective studies.mp. or longitudinal study/ or longitudinal.mp. or retrospective study/ or retrospective.mp.

Terms are connected with “AND”

PsycINFO

Population:

occupation* OR employ* OR work* OR job

Exposure:

(TX “job strain”) OR (TX “mental strain”) OR (TX “occupational strain”) OR (TX “work strain”) or (TX “mental load”) OR (TX workload) OR (TX “work load”) OR (TX (work N3 stress*)) OR (TX “job stress”) OR (TX “organi?ational justice”) OR (TX “organi?ational injustice”) OR (TX “job insecurity”) OR (TX (time N4 pressure)) OR (TX conflict*) OR (TX support) OR (TX harassment) OR (TX climate) OR (TX downsiz*) OR (TX (effort AND reward)) OR (TX demand*) OR (TX (shift N3 work*)) OR (TX mobbing) OR (TX bullying) OR (TX leadership) OR (TX “social relations”) OR (TX “social support”)

Outcome:

(SU “mental disorders”) OR (SU “adjustment disorders”) OR (SU “affective disorders”) OR (SU “anxiety disorders”) OR (SU “work related illnesses”) OR (SU “occupational stress”) OR (emotional N2 disorder*) OR (SU “somatoform disorder”) OR burnout OR (SU “suicide”)

Study Design:

((TX “randomized controlled trial”) OR (TX RCT) OR (TX intervention) OR (TX “case control”) OR (TX (cohort W2 stud*)) OR (TX (cohort W2 analy*)) OR (TX (follow W3 stud*)) OR (TX followup) OR (TX (observational W2 stud*)) OR (TX “prospective study”) OR (TX longitudinal) OR (TX “retrospective”)) NOT (“cross-sectional” OR “prevalence study” OR therapy))

* Terms were connected as follows: Population AND Exposure AND Outcome AND Study design.

Study selection

All titles and abstracts as well as included full texts were screened for inclusion and exclusion criteria independently by two reviewers (SD, AF, FH and MS). Publications without an abstract were initially included if the title did not clearly indicate whether the inclusion and exclusion criteria were not met. Disagreements regarding the inclusion of publications were discussed and solved. The title-abstract and full text screening was done according to a study-specific guideline. Furthermore, a pilot phase where all reviewers screened the same title-abstracts and full texts was initially done. The results were discussed in working group meeting moderated by AS.

Rating of methodological study quality

The methodological quality of studies published after 2013 was assessed according to the previous review (7). Therefore a hybrid tool was used combing the SIGN (Scottish Intercollegiate Guideline Network 2008) and CASP tool (Critical Appraisal Skills Programme 2008). The domains consider selection bias and information bias, as well as confounding. For avoiding selection bias, studies were required to have at least a 50% response and a loss-to-follow-up of less than 50%. Furthermore, individuals characterized by the presence of target disease symptoms at baseline should have been excluded from the study population. Alternatively, adjustments of target disease symptoms at baseline should have been made in the statistical models. Age and gender were specified as “core confounders”. In addition, adjustment of socioeconomic status was considered important. However, a lack of adjustment was still compatible with a “+”-rating and did not lead to downgrading and an exclusion of the study from the systematic review.

In the study by Svane-Petersen et al. (2020, e8) risk estimates for cumulative job control were adjusted for job control in the last year. This was regarded as having an over adjustment. Thus, the senior author Ida Madsen was contacted, and she carried out a re-analysis (which revealed minor changes in the risk estimates). The risk estimates from the reanalysis (Madsen 2021, e9) were included in the meta-analysis for job control (eFigure 6).

The included publications were quality-controlled in accordance with the strategy described (etable 6). The rating of study quality and extraction of study results was done by two reviewers independently (SD, AF, FH und MS). Disagreements in quality rating and uncertainties in data extraction were discussed and solved in work group meetings which were moderated by AS. As a result, 41 publications were rated “+”. None of the studies received a “++”-rating. Since all methodologically inadequate were excluded from this systematic review, and thus only studies with a “low risk of bias”-rating (“+”) were included, a tabular presentation of the risk of bias ratings is not necessary. Minor limitations of individual studies are shown in the right-hand column of the extraction table (etable 4).

Data extraction

The data of included studies were summarized in extraction tables. These tables contain information on the study population (study region, sample size, age, number of cases and controls, time of the baseline and follow-up, and information on response), measurement of exposure (questionnaire, exposure measurement, exposure level), the methods of outcome assessment, the type of data analysis used, and the relevant results. The data were extracted in English.

Statistical Analysis

Pooled risk estimates were calculated if at least three studies were available that considered the same exposure source and outcome. If possible, risk estimates were calculated separately for women and men.

Some studies published results of the same study population in different publications. In this case, the publication with the higher quality or the first published was used. Exposure assessment differed between studies. For the analysis, exposure was often divided into two categories (i.e. high and low, yes or no), or more rarely, in more than two categories (low, medium, high). It was sometimes considered as a continuous variable. In order to summarize the studies for meta-analysis, three different approaches were used:

Meta-analysis 1 is meaningful for assessing the dose-response relationship. Additionally, a dose-response relationship can be derived from the comparison of the pooled risk estimates from meta-analysis 3a. and 3b.: a positive dose-response relationship can be assumed if the pooled risk estimate according to 3a. is higher than the pooled risk estimate according to 3b. The risk estimates were pooled with random-effects meta-analyses using the Stata metan package (version 14.0). The heterogeneity is shown as I2.

Assessment of the confidence of evidence

We used the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach for grading the quality of the total body of evidence (e10), following the example of e11 with modifications (e12). We used the following levels of quality: high, moderate, and low. An initial “high” level would indicate having randomized studies. If only observational studies are included (as in our case), the starting level would be “moderate”. The quality of evidence was downgraded based on five factors; study limitations (high risk of bias), indirectness of evidence, inconsistency (heterogeneity measured by I2), imprecision (range of the confidence intervals of studies >2.0), and publication bias (yes or unclear, through a funnel plot/Egger’s test). Study findings with large effect sizes (effect estimate >2.0), an observed dose-response relationship, or the presence of residual confounding, resulted in an upgrade of the quality of evidence. If an effect larger than 5.0 was present, the quality of evidence was upgraded twice.

Results

General summary

Most of the studies included various occupations (n=19). Six studies examined workload-related mental illnesses among blue collar workers in paper mills/steel mills/truck manufacturers, electric utilities, the petroleum industry, forestry operations and heavy industry. Three studies were conducted in the public sector. Two studies included doctors and two studies social services with high-risk jobs, such as employees in special schools, psychiatric wards, eldercare and the Prison and Probation Services. One study included teachers.

Mainly, women and men were included in the study population, with the exception of a Danish study (Danish Eldercare Worker Cohort Study) that included only females, and a Japanese study with male employees of a power supply company. The ratio between female and male study participants varied greatly between individual studies.

The duration of the studies ranged from 18 months to 22 years. The vast majority of the included publications were based on a single follow-up. Seven publications considered two follow-ups. Four publications included three or more follow-ups. In a Danish study, the study participants were examined annually over a period of 16 years.

1. Are psychosocial working conditions associated with an increased risk of mental disorders or burnout?

2. How can the exposure-risk relationships between psychosocial working conditions and the occurrence of a psychological illnesses and syndromes (i.e. burnout) be described?

3. Are there any specific job groups, psychosocial working conditions and/or work tasks associated with a particularly high risk of developing a psychological disorder or burnout?

1. Meta-analysis approach 1. Continuous: Summary of studies with continuous exposure

2. Meta-analysis approach 2. Dichotomous: highest versus lowest level of exposure (studies with two categories: exposed versus non-exposed; studies with more than two categories: highest versus lowest exposure category)

3. Meta-analysis approach 3. Highest (3a.) versus second highest (3b.) category (studies with more than two categories): Meta-analysis 3a.: is the same as meta-analysis 2 but restricted only to studies with more than two categories; Meta-analysis 3b.: second highest exposure versus lowest exposure category.

Figure 1.

Risk of depression for high vs. low job strain. Dichotomous analysis finds a twofold increase in the risk of depression with high job strain.

95% CI, 95% confidence interval; m, men; w, women

eTable 5. PECOS criteria for inclusion and exclusion of studies.

Code	Category	Inclusion criteria	Exclusion criteria
P	Population	Working population (Women and men aged 17–70 years)	– Children and adolescents up to 17 years – People unemployed, people with unpaid jobs, retirees, inactive people – People aged over 70 – Military
E	Exposure	Job strain *1, gratification crisis (effort-reward imbalance *2: combination of high effort spent with low rewards received at work), social support, mobbing, work climate and workplace conflict, job insecurity, organizational injustice*3, low job control, long working hours, shift and night work, experience of violence and sexual harassment	– Chemical solvents (e.g. with lead, manganese) – Physical factors (e.g. noise, electromagnetic fields) – Physical work demands (e.g. handling of loads, adverse postures) – Stressors outside the workplace (e.g. care of family members) – Vocational training or studies
C	Comparison	Employees not subjected to the respective exposure	– No related to gainful employment
O	Outcome	Burnout (inclusive emotional exhaustion); unipolar affective disorder, anxiety disorder, and somatoform disorder at the symptom, syndrome, and disorder levels, suicide and suicidal ideation	– Physical diseases/complaints – Chronic fatigue syndrome – Posttraumatic stress disorder – Psychosocial wellbeing – Sleep disorders – Schizophrenia – Pain disorders, e.g. “non-specific low back pain“ – Return-to-work
S	Study design	Only primary studies are included: – Cohort studies (prospective, retrospective) – Case-control studies – Case-cohort studies – Mortality studies	– Cross-sectional study – Qualitative study – Ecological study – Field study – Experimental study

*¹ Job-Demand-Control-Support Model and Job Strain

The job-demand-control-support model (JDC) was introduced by Robert Karasek (e43) and is today the most prominent theoretical model on the association between occupational stress and health (e44). It incorporates the two dimensions “work demands” and “job control”. The latter (also known as “decision latitude”) refers to the level of control an employee has over work tasks. Thus, it is possible to differentiate between jobs with low and high work demands or low and high job control. The JDC model explains job strain as being a combination of work demands and job control. Accordingly, four different types of jobs can be distinguished: passive jobs, active jobs, low-strain jobs, high-strain jobs. For jobs with high work demands in combination with low job control (high-strain jobs), a high risk of adverse effects on health is postulated. In contrast, jobs with low work demands combined with high job control are referred to as “low strain jobs.” “Active jobs” are characterized by high work demands combined with high job control. According to the JFC model, active jobs are potentially associated with a further development of personal skills as well as higher productivity. Passive Jobs are associated with low work demands in combination with low job control. Generally, passive jobs can promote a passive lifestyle (e.g. a low level of physical activity). Johnson and Hall (e45) added a third dimension to the original JDC model: social support at work by superiors and coworkers. This component in regarded as an essential resource for preventing excessive strain, as high social support at work can compensate for the negative effects of high work demands. Karasek and Theorell (e46) further advanced the so-called Job-Demand-Control-Support (JDCS) model. The JDCS model postulates that the highest risk to health is associated with jobs characterized by high work demands in combination with low job control and low social support. Jobs with this constellation are also referred to as “iso-strain jobs“.

*² Gratification crisis (effort-reward imbalance)

The Job Gratification Crisis model (effort-reward imbalance) (e47, e48) is also a widely used work stress model for estimating mental stress or strain. The basic assumption of the model is that effort spent is rewarded according to the principle of social reciprocity. Gratification can be in the form of pay/salary, career opportunities and job security, as well as recognition and appreciation. Any mismatch between continuously spent high effort and the reward received can trigger a marked stress reaction (job gratification crisis) and give rise to adverse health effects. In addition, lack of job alternatives and unfavorable employment contracts as well as certain coping strategies increase the risk for a job gratification crisis (e48). Numerous scientific studies from various European countries and occupational groups have found a detrimental effect of effort-reward imbalance on health with regard to cardiovascular disease, musculoskeletal conditions and mental illness.

*³ Injustice at the workplace

“Organizational justice” is the term used to describe the workplace fairness perceived by employees. The concept of organizational justice takes into account the mechanisms by which ?employees evaluate perceived fairness at the workplace and the effects on other job-related factors, such as productivity (e49).

Organizational justice is split into two dimensions (e50): distributive justice and procedural justice. Distributive justice describes the perceived fairness of the distribution of wages and profits in relation to the service rendered. Factors such as the level of education of (other) employees, the responsibility assumed by each employee and the effort spent are also taken into account when evaluating this fairness. In contrast, procedural justice takes the fairness of the procedures of profit distribution into account. It is split into two dimensions: the evaluation of the actual formal procedures and the interactional justice. Examples of formal procedures include the evaluation of the promotion or remuneration practice, as well as the evaluation of the objectivity of the procedure. Interactional justice takes into account how the communication and justification of the distribution of profits and wages is evaluated by individual employees, given that this justification can strongly influence the evaluation of these same decisions as fair or unfair. In later studies, formal procedures and interactional justice are also referred to as procedural justice und relational justice.

eTable 6. Study appraisal using SIGN (2008) and CASP (2008).

Appraisal	Description
++	All or most of the criteria are met. It is very unlikely that the criteria that were not met would change the conclusions of the study.
+	Some of the criteria are met. It is unlikely that the criteria that were not met would change the conclusions of the study.
–	Few or none of the criteria were met. It is likely or very likely that the criteria that were not met would change the conclusions of the study.

eFigure 2b.

Risk of depression: Active job vs. low job strain

eFigure 12.

Risk of anxiety disorder due to high psychosocial work demands (dichotomous: high vs. low)