This study aimed to investigate whether psychosocial working conditions are predictors of sickness absence days in a sample of employees reporting symptoms of CMDs and participating in a randomized controlled trial investigating effectiveness of psychotherapeutic consultation at work (PT-A). Results showed that higher influence at work at baseline predicted a lower risk of sickness absence days at nine (i.e., intervention period) and 15 months later (i.e., intervention and follow-up period). Dissolution was negatively associated with sickness absence days during the follow-up period (six months after intervention). In sensitivity analyses, complete case analyses showed positive associations between social support and sickness absence days nine and 15 months later (ESM_3).

We have shown that employees who report higher influence at work, report less sickness absence days nine months later (intervention period) and 15 months later (intervention period plus follow-up period). Thus, we can suggest that out of several investigated psychosocial working conditions, for employees with symptoms of CMDs, influence at work may be relevant for reduction of sickness absence days. This supports findings in the literature, stating that high job control predicted lower risk of long-term sickness absence due to depressive disorders among male Japanese workers (Inoue et al. 2010). However, these workers had no history of previous mental disorders at baseline (Inoue et al. 2010). Other research showed that low job control predicted long-term sickness absence (more than 28 days) due to depression in both men and women, also after adjustment for baseline depressive symptoms (Clumeck et al. 2009). In Swedish employees seeking primary health care, it was also shown that low influence at work was associated with subsequent register-based sickness absence days of more than 14 days (Hultén et al. 2022). Investigation of employees with diagnosed CMDs showed as well that low control at baseline was predictive of increased risks of sick leave (more than 14 days) three years later (Helgesson et al. 2023). Thus, our findings are in line with the suggestion of the job-demand-control-support model, that high control at work can reduce or even buffer negative consequences (Karasek 1979).

In analyses covering the follow-up period (i.e., six months after intervention), dissolution was predictive of lower sickness absence days. This rather counterintuitive finding may be explained. The scale dissolution of work and private life is a newer edition to the third version of the COPSOQ, which showed questionable Cronbach’s alpha values (Lincke et al. 2021). Dissolution was associated with the inability to relax and with presenteeism (Lincke et al. 2021). There is the possibility that the scale measures slightly different aspects of dissolution. The items are worded in a way that they ask about working outside of normal working hours and being available for colleagues during free time, but it may also be possible that individuals understand these items in a way that it would reflect flexible work arrangements and thus the possibility to work at their own pace or time. If this would be true, employees reporting symptoms of CMDs can potentially manage their work better, if they are given more flexibility by their employer. This was for example shown by Helgesson et al. (2023) who found that no flexibility regarding work hours or no possibility to work from home were predictive of increased sickness absence days. In contrast to this explanation, it could also show presenteeism, indicating that the employees are still working while being ill (Bernstrøm and Houkes 2018). However, as we could not find similar results for this association in the two other study periods, the finding at hand could also have occurred by chance and should be considered with caution. In addition to this, it may be worth mentioning, that all participants reported on average less symptoms of CMDs after the intervention period compared to baseline (data not shown). This may indicate, that when symptoms are less severe, employees may be more likely to avoid taking sick leave, when they have more flexible work arrangements (e.g., working at own time or pace). Further investigations are warranted to support this suggestion. In addition, the analyses showed that the assessed symptom severity was not predictive of sickness absence days at any time point (only exception was depression severity at nine months). Other research findings showed more consistently that a high CMD symptom severity is a predictor of sickness absence days (de Vries et al. 2018). As the present study population had, for example, on average a high somatic burden or moderate depression severity, this may suggest that some of the psychosocial working conditions, such as influence at work, are relevant predictors for sickness absence days in this specific study sample.

Similar to Helgesson et al. (2023), we have not found associations between support at work and sickness absence days. However, we did observe positive associations between support at work and sickness absence days after nine and 15 months in sensitivity analyses when considering complete case analyses, showing that social support was associated with higher risks of subsequent sickness absence days (ESM_3). This may indicate that only among the full observations, the perception of having supportive colleagues or superiors may encourage them to take sick leave or make them feel less pressured to attend work when being sick, resulting in more sickness absence. Thus, in this case this may be a proxy for a supportive work environment. This line of thought is also supported by a study finding that high emotional support was associated with more sickness absence spells and days in a general middle aged working population in Sweden suggesting that this reflects an illness behavior (Karlsson et al. 2010). As that study did not focus on CMDs in particular, this behavior seems to rather fit to general sickness absence, caused by different diseases. However, as findings for imputed and complete data were different in this case and as single analyses for psychosocial working conditions could not show these associations, results have to be considered carefully as the complete case analyses may entail biased results due to potential selective drop-out (Bell et al. 2013). Further, there are also studies that did find different results indicating that low support from a superior was predictive of long-term sickness absence due to a psychiatric diagnosis (more than eight weeks; (Foss et al. 2010) or that find only non-significant associations between support and sickness absence, as the meta-analysis by Duchaine et al. (2020), for example. Thus, besides characteristics of study populations and sick leave duration, also the heterogeneity of assessed constructs can contribute to these mixed findings.

Only in sensitivity analyses considering complete cases, it was observed that possibilities for development predicted higher sickness absence days within the follow-up period (i.e., six months after intervention). However, we did not observe multicollinearity according to the variance inflation factor (VIF). In separate analyses controlling for confounding variables and a single psychosocial working condition scale, no significant association was found between possibilities for development and sickness absence days (ESM_1, ESM_2). This suggests, that this finding may be due to chance occurrence and would warrant further investigation to confirm its validity.

The lack of an association between demands and sickness absence days has also been seen in other studies investigating sickness absence due to depressive disorders (Clumeck et al. 2009; Inoue et al. 2010) or in a study investigating psychosocial factors and sickness absence in general (Niedhammer et al. 1998). For the present study, it seems that demands may be less relevant in terms of risks for subsequent sickness absence days among employees with symptoms of CMDs. However, there are also results from a meta-analysis indicating that high demands are prospectively associated with sickness absence due to a diagnosed mental disorder (Duchaine et al. 2020) or other results showing positive but also no associations for some studies (de Vries et al. 2018). Thus, these associations seem to change and look differently, depending on the study population, i.e., depending on whether study participants already had symptoms or whether a new development of CMD was investigated (as this was the case for associations between demands and sickness absence in (Duchaine et al. 2020). In light of the job-demand-control-support model, having control at work seemed to be more important for sickness absence in our study population, than the perception of demands at work.

We tested a number of additional scales assessing social relations and leadership quality, but did not observe any further associations with sickness absence days. It is important to mention that although we checked for multicollinearity with VIF, there may be high correlations between the variables assessing social relations. However, also separate analyses for every single scale did not show significant associations with sickness absence days in our study sample (ESM_1, ESM_2). While we have not found any association between recognition and sickness absence days, for example, other findings using the similar reward component of the effort-reward imbalance model, have shown increased risks for sickness absence due to mental health problems among men only, for low rewards (Ndjaboué et al. 2014). In accordance with our null findings for an association between sense of community and sickness absence days, another study among a representative sample of the French working population did also not find associations between sense of community and the number of sickness days (Bertrais et al. 2023). That study also showed that psychosocial working conditions were more likely to predict the cases of sickness absence than the number of sickness absence days (Bertrais et al. 2023). However, only in complete case analyses (not in single analyses), we observed a negative association between sense of community and sickness absence days for the follow-up period (ESM_3), which would thus go along with findings in the literature (Bertrais et al. 2023).

This is one of the first studies that investigated whether psychosocial working conditions predicted sickness absence days among employees reporting symptoms of CMDs in Germany. Thus, results for this specific study sample cannot be generalized onto other countries due to different labor laws, labor market situations or social insurance systems. Future research should aim to determine factors leading to potentially different associations (e.g., payment change from continued payment to sick pay, which countries have a staged return-to-work process).

It can be regarded as a strength that we have conducted prospective analyses covering more than one period and that we have chosen to investigate a range of different psychosocial working conditions. In contrast to many other studies, we have investigated any sickness absence days, and did not restrict our analyses to only long-term sickness absence. It has been shown that psychosocial work factors were associated with both, shorter and longer periods of sickness absence in Danish employees, for example (Thorsen et al. 2021).

There are several limitations that have to be considered when interpreting our results. As the analyses are secondary to the RCT (Weber et al. 2021), we did not perform additional a-priori power analyses but present rate ratios and confidence intervals. All measurements are based on self-report, which could introduce common method variance (Podsakoff et al. 2003). However, in regard of sickness absence, it has been shown that self-reported sickness absence data can be similar to employer’s register-based sickness absence data (Voss et al. 2008). Quality controls were included, in which unreasonable sickness absence days (i.e., reporting of higher days than the possible time period) were double checked with the participants and baseline sickness absence days were double checked by the study psychotherapists.

We cannot exclude that the perception of psychosocial working conditions, assessed at baseline, has changed over time for the participants. Especially, as half of the employees took part in an intervention which was aimed at reducing symptoms of CMDs by understanding and dealing with the impact of psychosocial working conditions on mental health, it may be possible that the perception of working conditions has changed. If perceptions of working conditions have changed over the investigated period, the results at hand may be misleading. Future analyses would need to clarify this. In addition, this may indicate that the participants were more interested in improving their mental health status (as they participated in an RCT in which they received psychotherapeutic consultations), than a general working population with CMD symptoms. Results of analyses covering only the follow-up period (i.e., T2) have to be considered carefully as we have observed convergence issues when including the independent variables into the analyses. Thus, the addition of further variables did not result in increased variance but rather showed that the model became unstable, probably due to the complexity and broad range of psychosocial working conditions. Further, all our models show generally low model fit values, indicating that there may be other more important predictors for sickness absence days among our study population (Montano 2020).

As we have observed some additional associations in sensitivity analyses considering complete cases (ESM_3), it is important to mention that the complete case analyses may be biased due to selective dropout and should be considered carefully. For example, we observed that, on average, sickness absence days at baseline were lower among the participants who reported sickness absence days for all three time periods compared to participants with missing data for any sickness absence days (difference of about four days, data not shown), indicating that there may be a selective drop-out of participants with higher sickness absence days at baseline. Something similar was checked and observed for some of the psychosocial working conditions: More favorable working conditions (for recognition, dissolution, unfair treatment and degrees of freedom) were found among the participants who reported sickness absence days at all three time points, indicating that participants with less favorable working conditions have dropped out of the study.

As we have investigated which psychosocial working conditions are predictive of subsequent sickness absence days at different time points, it would also be of interest to investigate in the future, if changes in the perception of working conditions (possibly due to the intervention) and in objective working conditions may contribute to different risks of sickness absence days. However, as we could not observe that the investigated broad range of psychosocial working conditions had a great effect on sickness absence days, it seems rather unlikely that calculated changes would show very different results. In line with this thought, there are studies available, which showed for example, that adverse changes in psychosocial working conditions increased the risk for sickness absence only slightly, and that favorable changes in working conditions were not associated with sickness absence at all (Saastamoinen et al. 2014). Other findings suggested that changes in sickness absence are not affected by changes in perception of psychosocial working conditions in the Finnish Food Industry (Siukola et al. 2011). Thus, there may be other factors (such as previous sickness absence or age in our case), which are more important in regard of sickness absence days (Montano 2020). In light of these, Montano (2020) also considers the relevance of the effects of psychosocial working conditions, i.e., how individuals react to and cope with these, as well as the subjective health status and a range of moderators or mediators for the likelihood of sickness absence. Among our study population, factors outside of the working environment, such as private conflicts or family conflicts may be more relevant for sickness absence than psychosocial working conditions. For example, in our sample, 71% of the participants reported to participate in the PT-A due to work-related stress and personal problems (Rothermund-Nassir et al. 2025). This may explain that for sickness absence other factors were more relevant next to the psychosocial working conditions. Nevertheless, improving influence for employees (regarding amount of work and decisions affecting work) may be a good opportunity to reduce sickness absence days, as by this a greater flexibility may help to deal with the work tasks according to one’s own workability, and has to be tested in future research (Helgesson et al. 2023). Repeating the analyses within a representative sample may also be helpful for interpretation of our results. As we performed the study with employees who took part in a RCT mostly via the company, it may be that the participants had better working conditions that allowed them to participate than a general working population with CMD symptoms. It may also be possible that they have a more active health behavior as they were actively seeking for support to cope with their symptoms. Another promising future research idea would be to investigate the effects for different occupational sectors. As working conditions may differ in certain occupations (for example in healthcare, in crafts or in administration the most stressful conditions may differ), there may be different magnitudes of effects for working conditions on sickness absence days among employees with CMD. Results may be relevant to find suitable starting points for interventions to reduce sickness absence days for these occupational groups, as well.