Starting in 2007 a total of 15,010 participants were recruited into the GHS cohort. Residents of Mainz and the surrounding area (Mainz-Bingen) between the age of 35 and 74 years were randomly selected (stratified 1:1 by sex and rural vs urban) and invited to participate in the study. People unable to communicate in German or who were physically unable to attend the baseline examinations at the study center were excluded. All participants gave their written informed consent to participation. The Medical Ethics Commission of Rhineland-Palatinate and local and Gutenberg-University of Mainz data protection officials reviewed and approved the study (ethics committee review number 837.020.07(5555)).
Between 2007 and 2012, five-hour baseline examinations were conducted at the University Medical Center of the Johannes Gutenberg University of Mainz . The baseline response among all successfully contacted residents was 70% . Of the participants between the ages of 35 and 64 years and in paid employment at baseline (n = 8306), 66 (0.8%) died before the 5-year follow-up and a further 980 (11.9%) were lost to follow-up (C. Drossard, personal communication, March 25, 2021).
Self-ratings of psychosocial working conditions were obtained at baseline (t0) among the working population using either the COPSOQ or the Effort-Reward-Imbalance (ERI) questionnaires. At baseline, the distribution of the questionnaires alternated weekly. At the five-year follow-up (t1) the subjects still working received the same questionnaire filled out at baseline. The general characteristics of the participants filling out either the COPSOQ or ERI questionnaire at each time point, including information on age, sex, education, and occupational sector, were analyzed descriptively.
The COPSOQ is a validated and comprehensive questionnaire used internationally to measure a wide variety of psychosocial working conditions . As Kristensen, et al.  write, the COPSOQ is intended to be as broad and comprehensive as possible by covering possibly all relevant psychosocial workplace factors. The COPSOQ is thus, “theory-based without being based on one specific theory” . The questionnaire includes a wide variety of aspects on working conditions taken from different theories and models in order to assess psychosocial working conditions as comprehensively as possible. Approximately half of the GHS participants who were working at baseline were assessed with the German standard version of the COPSOQ that comprised 25 scales in the dimensions: demands, influence and development, support and leadership, further parameters, and outcomes (strain/effects) [26, 30, 31]. The COPSOQ scaled to values between 0 and 100 (minimum and maximum value, respectively). Staying with the standard procedure for COPSOQ, scale values are calculated based on the available answers if at least half of the scale items are answered; if less than half of the items are answered, the scale value is set to missing.
Depending on the content of the scales, higher scores can represent either favorable or unfavorable conditions. Higher scores represent favorable conditions for the scales measuring influence at work, degree of freedom at work, possibilities for development, meaning of work and workplace commitment, predictability, role clarity, quality of leadership, social support, feedback, social relations, and sense of community. In the case of the scales for quantitative demands, emotional demands, demands for hiding emotions, work-privacy conflict, job insecurity, role-conflicts, and mobbing, higher scores represent unfavorable conditions.
The demand domain of the COPSOQ included four scales:
quantitative demands were estimated with four items (e.g. Do you have to work very fast? always; often; sometimes; seldom; never/hardly ever),
emotional demands with three items (e.g. Is your work emotionally to a very large extent; to a large extent; somewhat; to a small extent; to a very small extent),
demands for hiding emotions with two items (e.g. Does your work require that you hide your feelings? to a very large extent; to a large extent; somewhat; to a small extent; to a very small extent), and
work-privacy-conflict with five items (e.g. The demands of my work interfere with my private and family life. Agree fully; agree somewhat; undecided; disagree somewhat; disagree).
Five scales belonged to the thematic domain of influence and development: influence at work (4 items; e.g. Do you have a large degree of influence on the decisions concerning your work? always; often; sometimes; seldom; never/ hardly ever),
degree of freedom of work (4 items; e.g. Can you decide when to take a break? always; often; sometimes; seldom; never/hardly ever),
possibilities for development at work (4 items; Do you have the possibility of learning new things through your work to a very large extent; to a large extent; somewhat; to a small extent; to a very small extent),
meaning of work (3 items; e.g. Is your work meaningful? to a very large extent; to a large extent; somewhat; to a small extent; to a very small extent), and
workplace commitment (4 items; Are you proud of being part of this company? to a very large extent; to a large extent; somewhat; to a small extent; to a very small extent).
The interpersonal relations and leadership domain comprised nine scales:
predictability (2 items; e.g. At your place of work, are you informed well in advance concerning for example, important decisions, changes, or plans for the future? to a very large extent; to a large extent; somewhat; to a small extent; to a very small extent),
role-clarity (4 items; e.g. Do you know exactly which areas are your responsibility? to a very large extent; to a large extent; somewhat; to a small extent; to a very small extent),
role conflicts (4 items; e.g. Are contradictory demands placed on you at work? to a very large extent; to a large extent; somewhat; to a small extent; to a very small extent),
quality of leadership (4 items; e.g. To what extent would you say your immediate superior makes sure that the members of staff have good development opportunities? to a very large extent; to a large extent; somewhat; to a small extent; to a very small extent; I don’t have a superior/colleagues),
social support (4 items; e.g. How often do you get help and support from your colleagues, if needed? always; often; sometimes; seldom; never/hardly ever; I don’t have a superior/colleagues),
feedback (2 items; How often does your immediate superior talk with you about how well you carry out your work? always; often; sometimes; seldom; never/hardly ever; I don’t have a superior/colleagues),
social relations (2 items; e.g. Is it possible for you to talk to your colleagues while you are working? always; often; sometimes; seldom; never/hardly ever; I don’t have a superior/colleagues),
sense of community (3 items; e.g. Is there a good atmosphere between you and your colleagues at work? always; often; sometimes; seldom; never/hardly ever; I don’t have a superior/colleagues), and
mobbing (1 item; How often do you feel unjustly criticized, bullied or shown up in front of others by your colleagues and your superior? always; often; sometimes; seldom; never/hardly ever; I don’t have a superior/colleagues).
As a further parameter, job insecurity was measured with 4 items (e.g. Are you worried about becoming unemployed? to a very large extent; to a large extent; somewhat; to a small extent; to a very small extent).
More information on the German standard version of COPSOQ can be found at www.copsoq.de, more information on the international development of the questionnaire at www.copsoq.network.org.
The ERI questionnaire (2006 version when the survey was started) included 6 effort items, 11 reward items, and 6 items regarding overcommitment. German versions of the ERI questionnaire have been validated and tested in the general population [32, 33]. The ERI-items first asked if a work condition was experienced, and then a follow-up question asked for a rating of the strain related to this working condition. For example, the presence of physical demands at work was assessed with a yes or no response to the statement, “My work is/was physically demanding”. If yes was selected, “And how much does that strain you?” was asked with four answer possibilities (i.e. not at all, moderately, strongly, very strongly). The ERI-ratio was calculated by dividing the sum of the effort items by the reward and correcting for the difference in number of items. Ratio values diverging from 1 indicate an imbalance between effort expended and reward received: values over one indicate the effort experienced exceeded the reward, and values under one indicate the reward experienced exceed the effort. To ease comparison with the COPSOQ items, the ERI results were also rescaled to values between 0 and 100.
Among both groups of participants, four health and work strain outcomes were also assessed. These outcomes are part of the standard version of the COPSOQ, and we included them into the ERI to enable comparisons between the COPSOQ and ERI instruments and analyze criterion validity . The internal outcomes measured were job satisfaction, general health status, burnout symptoms, and the satisfaction with life scale (SWLS). Job satisfaction was assessed using seven items (e.g. If you look at your work situation as a whole, how satisfied are you with your career prospects?) with four answer possibilities (i.e. very satisfied, satisfied, unsatisfied, or very unsatisfied). General health status was assessed with one question from the EQ-5D (EuroQol) where health was rated on a scale from 0 (worst health imaginable) to 10 (best health imaginable) using the question: “Your state of health: If you rate the best conceivable state of health with 10 points and the worst conceivable with 0 points: How many points would you award your current state of health?”. Burnout was assessed with six questions from the Copenhagen Burnout Inventory (CBI) (e.g. How often do you feel tired? always, often, sometimes rarely, never). Satisfaction with life was assessed with five items from the Satisfaction with Life Scale (SWLS) (e.g. In most areas my life corresponds with my ideal expectations. Agree completely; agree; agree somewhat; neither agree nor disagree; disagree somewhat; disagree; disagree completely) . All of the outcomes were rescaled to values between 0 and 100. Increasing values represent more desirable levels of job satisfaction, general health, and satisfaction with life. However, for the adverse outcome of burnout, rising values are less desirable.
We analyzed the results of the COPSOQ and ERI questionnaires at both study time points descriptively using means and standard deviations. In correlation analyzes, all bivariate relations between workplace factors and outcome factors were assessed. In a final step, linear regression models were selected using stepwise selection to consider the associations between psychosocial working conditions and health or strain outcomes measured contemporaneously. The same method was used in several international COPSOQ validation studies (i.e. [30, 35]) as well as in the previous analyses with the data of the GHS . The prospective relationships between the psychosocial conditions at baseline and the outcomes at follow-up were also analyzed with multiple linear regression models. For each outcome, one COPSOQ model and one ERI model were selected with stepwise selection; models presented always include only the selected predictors, thus maximizing case numbers available.
Model selection for the ERI subgroup included the variables for the rescaled (0–100) effort, reward and overcommitment scales, as well as the ERI ratio. For the COPSOQ linear regression models, we report the model selected in the fifth step of model selection, unless the next model selection step resulted in an improvement in explained variance (R2) by two or more percentage points. This was done because the number of factors that are statistically significant is associated with the sample size, which is, in this case quite large. Based on our analyses at baseline, after the first five scales were selected, additional model variables, although statistically significant, explained little additional variance in all of the dependent outcome variables. Therefore, we also stopped the selection of prospective models after five variables were selected. This also ensured the comparability of results with the results reported in the German COPSOQ validation  and at baseline . Collinearity of the models was assessed with the variance inflation factor (VIF) using a cut-off of 5. We examined the outcome variance explained by the models (R2) to estimate and compare the predictive performance of the questionnaires. To give a complete picture of all relations between (candidate) predictors and outcomes correlation matrixes of all workplace factors with the outcome factors (Pearson correlation coefficients) are shown in the supplementary material (Supplementary Tables S3-S6).
All of the analyses were conducted in SPSS 23. Like the initial baseline study described by Nübling, et al. , we did not adjust for age and sex since the age and sex distribution of the groups receiving the ERI or COPSOQ were comparable. In a sensitivity analysis, we adjusted for age, sex, and socio-economic status. The socio-economic status index comprised information on education, income, and occupational status and ranged from 3 to 21, with higher values indicating higher status.