The sample analyzed for the present study was drawn from the data set of the Gutenberg Health Study (GHS) (30). The GHS is a large-scale, prospective, and representative population study conducted since 2007. The aim of the study is to identify risk factors and causes of the most common diseases. With over 15,000 participants from the Rhine-Main region, it is one of the largest regional health studies in the world.

Data on mental health related variables were collected by questionnaires during the study days. Participation in the study was voluntary and the subjects were able to withdraw from the study at any time. In the present study data-collection took place between 2007–2017 with a total of 1,500 participants (37.5% female, age M = 48, SD = 8.2) was used. Inclusion criterion was the affiliation to one of the following occupational groups:

The group allocation was based on the occupational data recorded as part of the GHS. These were available in the form of codes for the classification of occupations (Klassifikation der Berufe: KLDB) used by the German Federal Employment Agency. The reference group “other academics” comprised of persons with educational degrees from a “university of applied sciences” or “university.” In order to avoid autocorrelations – and thus an overestimation of the strength of the associations under consideration – individuals already assigned to occupational groups (a) to (c) were not additionally assigned to reference group (d).

On the one hand, school teachers (ST) were compared to academically graduated participants who worked as teaching and training professionals (TT). This occupational group was chosen because of the occupational proximity due to the work content. In addition, school teachers were compared to another academic occupation with high levels of personal interactions, but with different work content. This occupational group for comparison was medical doctors and dentists (MD). The broader group of other academics (AC) was selected as the reference group for all intergroup comparisons because of a comparable socioeconomic status.

The independent variables of interest are listed below this paragraph. The scales to measure the work-related variables (h-l, see below) were part of the Copenhagen Psychosocial Questionnaire (COPSOQ) (31, 32). The COPSOQ is based on several theoretical models (e. g., Demand-Control Support Model or Effort-Reward-Imbalance Model) and is a scientifically validated tool, commonly employed in psychosocial risk assessment at workplaces (31). This comprehensive questionnaire, originally developed in Denmark (32) evaluates several psychosocial factors including demands at work, work organization, interpersonal relations, work-individual interface, values at the workplace, health, and well-being. Through detailed inquiry, the COPSOQ identifies areas for potential psychosocial risk intervention, facilitating healthier work environments. Each item was answered on a Likert scale. The scales used for the present study are described below.

The authors confirm that the study was approved by the local ethics committee [ethics committee of the state medical association RLP; ethics committee vote: 837.020.07(5555)] and was performed in accordance with the ethical standards as laid down in the 1964 declaration of Helsinki and its later amendments. Written informed consent was obtained from all participants.

Descriptive data (means, relative frequencies, and standard deviations) were calculated for the variables used in the analyses, both for the total sample and for each individual occupational group. In the case of particularly skewed distributions of scales (skewness >1), the median and interquartile range were calculated rather than the scale mean. In the data analyses, the three occupational groups (a) school teachers, (b) teaching and training professions, and (c) medical doctors and dentists were considered separately.

The dependent and independent variables (COPSOQ scales, 5-level Likert format) were transformed to values between 0 and 100 for data analyses (transformation for, e.g., 5-level Likert-scales was 1 = 100, 2 = 75, 3 = 50, 4 = 25, and 5 = 0 and accordingly for different numbers of Likert-scale levels). High values represent high expressions of the construct examined in the respective scale.

Binary logistic regression models (Poisson regression) were calculated in order to estimate the predictive value of sociodemographic variables, the three compared occupational groups, and the aforementioned independent variables on the dependent variables. The reference group for logistic regression analyses regarding the occupational groups was the occupational group of other academic professions.

For each dependent variable, three separate regression models were set up with different levels of adjustment; the variables considered in a previous level were retained in the next level (see Table 1). In the logistic regression models (Poisson regression), a median split was performed for each of the strains (dependent variables) to dichotomize values below or above 50. In the fully specified regression model (third regression model) the dependent variables of the different occupational groups were adjusted for sex, age, socio-economic status (SES¹), working time model, and for the scales satisfaction with work, work-privacy conflicts, role clarity, meaning of work, and bullying. To enable a more precise differentiation, all analyses were carried out separately for female and male participants.

With regard to the job-related variables, the following differences were found between the occupational groups (see Table 2). On the scale satisfaction with work the mean value of school teachers was below that of the two comparison groups; however, all three occupational groups were above the mean value of the reference group of other academics.

On the work-privacy conflicts scale, the three occupational groups had higher mean values than the reference group of other academics. This was most pronounced in the occupational group of medical doctors and dentists, followed by school teachers.

The highest mean value on the role clarity scale was also shown by medical doctors and dentists, while the other groups were at a comparable and lower level.

Regarding the meaning of work, the mean value of the school teachers was below that of the two comparison groups, but it was also above the mean value of the reference group of academics.

With regard to bullying, all occupational groups and the reference group showed left-skewed (towards zero) distributions. The value of the median for the school teachers was 0, the same value as in the occupational group teaching and training professions. The median for the occupational group of medical doctors and dentists, and the reference group of other academics was only slightly higher. Considering the scaling from 0–100, it can be stated that low values for bullying were found across all occupational groups.

With regard to satisfaction with work, all three compared occupational groups (including school teachers) were above the level of the reference group of other academics. School teachers were slightly less satisfied with work compared to teaching and training professions, as well as medical doctors. However, in another study primary teachers showed higher scores on job satisfaction compared to social workers. We further analyzed differences in terms of work privacy conflicts. Again, all three occupational groups showed higher values than the reference group. The overlap of work and private life at two workplaces has been discussed in terms of higher conflicts in teachers (3, 24). However, we did not see pronounced differences in teachers. Instead, the biggest differences were observed in medical doctors. These had the highest and correspondingly least favorable values in this regard. In contrast, role clarity was highest and thus most pronounced among medical doctors and dentists; the other two occupational groups were at comparable levels slightly above the mean of the reference group. In contrast to our results, in another study (24) school teachers rated role clarity worse than individuals of other occupational groups. Differences between the occupational groups were rather minimal in terms of role clarity overall, and can be seen as well-defined. The same was evident in terms of how those occupational groups rated the meaning of their work. Differences were small, with the highest values observed in medical doctors. We also noted that bullying was very little to not at all prevalent within the occupational groups compared.

These results give a first hint towards the possibility that the occupational groups analyzed might have been more similar than expected. Along the same line, it is plausible to assume that the compared occupational groups might be more heterogeneous within each group than expected. We think that the variance for job-related factors in each group outweighed the variance between groups, which led to non-significant predictive values for occupational groups. One explanation might be the narrow selection (academics only) of occupational groups in our study. Studies with more accentuated between-groups differences compared a much wider range of occupational groups (23).

In contrast to similar studies (23, 41), none of the considered occupational groups was a significant the predictor of job-related strains. Therefore, simply belonging to the occupational group of teachers was not significantly associated with higher levels of burnout, lower levels of self-rated general health, or lower levels of life satisfaction. It is possible that none of the occupational groups predicted the above outcome variables due to the high homogeneity between the groups. A Belgian study showed similar results to ours, differences in job demands and strains were only small when comparing teachers with other occupational groups (14).

Specific sociodemographic and job-related variables showed significant associations with the outcomes considered, but not the occupational group itself. Therefore, comparisons between occupational groups seem to fall short regarding the variables considered in the present study. Clear sex differences were found several times, especially with regard to higher burnout symptoms in women. Part-time employment, for example, was a significant risk factor for burnout only for female teachers, but not for males. This could be due to the fact that the backgrounds and motivations for taking up part-time employment might be different between the sexes. The fact that women in our study worked part-time more often than men could also be interpreted as an early attempt to reduce anticipated or existing multiple burdens, respectively, role expectations. Given these sex differences regarding burnout, it was unexpected to not find such differences in self-rated health as well as satisfaction. The higher scores for burnout in female teachers might partly be explained by the tendency that females tend to be more interested in inter- and intrapersonal processes than males (42) and is also in line with a role stereotypes perspective, from which it might seem to be more acceptable to admit mental burdens for females than for males.

Consistently across all three outcomes, work-privacy conflict was a significant predictor. Particularly in the case of school teachers, it seems worth taking a closer look at work-privacy conflicts, since part of the work takes place in the home environment and not only at the school workplace, which is systematically inherent in the profession of school teachers in general. An increased potential for a conflict between work and private life is accordingly expected to be found among school teachers. Corresponding results were reported in other studies with large sample sizes of school teachers where significantly higher values for work-privacy conflicts compared to the general population have been reported (24, 27). Teachers showed a 1.5 times higher level of work-privacy conflicts compared to a composition of other occupations (24).

Taken together these results, work-privacy conflicts seem to play a major role in teachers’ mental and somatic health and therefore should be one target for interventions to aim for.

The exact classification of occupational groups according to the KLDB definition made it possible to clearly distinguish various occupational groups and to compare them with one another within the GHS data set. The fact that the participants live in the same region means that effects resulting from substantially different regional conditions can be excluded as far as possible. Yet, the external generalization of the results is limited because our sample is representative for the region of Mainz and Mainz-Bingen, Germany.

The job-related factors examined were derived from previous survey results, but other operationalizations could have led to different results. For example, variables like quantitative demands, the degree of freedom at work, workplace commitment, quality of leadership or social support could be additional, respectively, alternative ways to composite job-related factors.

Another limitation is the fact that the present study is cross-sectional, so that the variables cannot be interpreted in terms of causality. The division into predictors and outcome variables corresponds to the stress–strain concept (9) and the usual operationalization of other authors. However, it can be assumed that there is a reciprocal interaction, so that the assumed outcome variables (e.g., burnout) also influence the predictors (e.g., work-privacy conflict).

Furthermore, due to different sample sizes between the occupational groups, differences might not have been able to become significant. We did see some differences between the considered occupational groups, which were rather minimal.

A further limitation is that the data collected is based on subjective assessments and does not consider objective data (e.g., diagnosed disorders). However, this limitation applies to all occupational groups considered, so that at least no systematic bias due to the comparison of different data sources occurred between the different occupational groups. However, if there were systematic differences in the response tendencies of the corresponding occupational groups, this would be expressed here.

What can also be seen as a limitation to the study is, that we did not have data on non-responders, therefore corresponding analyses were not performed, which could have helped to get a clearer picture.

For further studies on the present topic, it would make sense to depict job-related factors more broadly than in the present study. Job-related factors, for example, could be operationalized by quantitative demands, social relationships at or connectedness with the workplace. The same applies to strains, for example these could be: intention to quit, clinical indicators (e.g., depression) or somatic symptoms. Another implication of the present study arises from the exclusive use of quantitative data. Particularly with regard to the sex-specific differences in the results, qualitative data, e.g., from individual interviews, could be a valuable supplement for future studies in order to gain insights into the everyday reality of part-time teachers, for example.

The high relevance of work-privacy conflicts with regard to all outcomes (burnout, self-rated health status, life satisfaction) is noteworthy. Other studies also identified work-privacy conflicts as a relevant predictor for certain strains (24, 27, 39). Interventions addressing this issue therefore could have multiple positive effects for teachers. Preventive or interventive strategies on behavioral level could be in form of trainings (e.g., face-to-face trainings or E-learnings) to establish a better separation between work and private life.

Trainings on how to deal with work-privacy conflicts should also highlight the chances that go along with structural conditions of the teachers’ job. High flexibility and compatibility with private life are positive aspects of the teachers’ profession. A reflection of these positive aspects could contribute to reframe teachers’ attitudes towards their job. Highlighting positive aspects of the teaching profession could also be beneficial for the professional socialization of teachers in terms of health.

Further useful interventions can be seen (1) on school level. Regulations regarding reachability for students and their parents or the allocation of working space in order to give support in the separation of work and private life could be helpful. (2) On a political level, expanding high-quality support systems (e.g., all-day childcare) would ease the burden on teachers who are parents. It would strengthen resources to cope with multiple stresses in order to improve the situation, for especially predominantly female teachers. Also, individual differences in the phase of life – e.g., career launch, parenting with differences for instance in terms of children’s age, nursing relatives – concern various needs of the employees. Those needs should be considered, on personal, school and political level, respectively.

Based on the available data, we cannot answer the question why females do not seem to benefit from a reduction in working hours with certainty, but we can reasonably assume that a gender role model that strives more toward equality – especially when it comes to childcare – could reduce existing inequalities. In order to be able to better grasp and deal with definite everyday problems that arise among school teachers with strongly pronounced work-privacy conflicts, in-depth studies addressing these issues are indispensable.