In our multi-adjusted comparison study of six indicators of health behavior, teachers were less likely than other workers to smoke, to have used cannabis in the past year, to gamble regularly and to be overweight or obese. When restricting the comparison group to other occupations belonging to the same socio-professional category (the intermediate and managerial/professional occupations), these differences were attenuated. The difference in BMI was no longer significant but those regarding tobacco, cannabis and gambling remain highly significant. No differences were observed between teachers and non-teachers regarding alcohol consumption and sleep duration once important confounders such as demographic and socioeconomic factors were taken into account.

On the whole, our findings are consistent with the “healthy teacher distinction” hypothesis: as compared to employed persons with similar socioeconomic profile, teachers were more likely to have healthier behaviors, at least in certain domains, such as tobacco, cannabis, and gambling. These associations between healthier behaviors and teaching status may be related to a higher level of health awareness among teachers, which could be underpinned by two mutually enhancing mechanisms [18]. Firstly, it is possible that those choosing a teaching career may be more health-conscious than those who do not, perhaps through self-selection by shared personality traits, beliefs and values, as well as by social origins, that would influence health behaviors [34]. Secondly, once individuals have entered the teaching profession, they may be prompted (consciously or unconsciously) to conform to healthy norms and internalize positive attitudes towards health and negative attitudes towards health risks [35]. Overall, it would be interesting to assess to what extent teachers feel that they are entrusted with a responsibility for health education of the young. Such a motivation could partly explain their tendency to endorse healthy behaviors as role models.

In contrast to the situation with tobacco and cannabis, the absence of a teacher distinction regarding alcohol consumption is of note and we estimated that one in ten teachers reported at-risk alcohol consumption. The link between alcohol and the school environment, notably with respect to specific work-related stressors and culture, should be examined in more detail [36]. A hypothesis, that remains to be tested, would be that alcohol use, which generally takes place in the private sphere outside school hours and is thus not visible at school, would be less stigmatized than tobacco and cannabis, at least in the educational community. Both the substantial prevalence of at-risk drinking in teachers and the absence of a teacher distinction for alcohol suggest that providing better information on the risks associated with excessive drinking is an important target for a holistic prevention approach in the educational community.

After adjusting for potential confounders, we did not observe significant differences in sleep duration between teachers and other workers. Although sleeping enough is important to be fit, sleep needs greatly differ across individuals [37]. Moreover, sleep duration has been shown to vary by occupation, with work affecting sleep through various factors, including long/extended work hours and other job-related stressors [38, 39]. Therefore, normal sleep duration may not be the best indicator for evaluating sleep hygiene. Another point is that sleep habits, like alcohol consumption, are less visible at school than smoking habits or BMI. Thus, setting an example with respect to sleep hygiene would be less relevant for teachers.

Teachers were less likely to be obese than other workers, but the difference was not significant when restricting the comparison group to other intermediate and managerial/professional occupations. This observation points at the image issue and stigmatization risk in the work environment. In fact, teachers are particularly exposed in the classroom, and appearance, including corpulence, may be considered as crucial to limit stigmatization and achieve teaching effectiveness. However, the absence of a teacher distinction toward BMI among the intermediate and managerial/professional occupations suggests that interaction with the public and exemplarity duty are not the sole factors that contribute to relatively low obesity rates among employees. It would be interesting to examine in-depth the associations between BMI, jobstrain, job insecurity and sector type.

Classically, health behaviors are studied to identify populations at risk or more prone to be involved in risky behaviors. Here, the approach is quite different. We aimed to evaluate to what extent teachers’ health behaviors differ from those of other occupations, considering such health behaviors as valuable indicators of health awareness. Little information was previously available about this topic, although it could clearly contribute, as part of a global health education approach, to establishing more relevant and effective teacher preparation programmes [40]. Teachers are well-placed to play a role in health education due to their familiarity and proximity with the young which is a prerequisite to delivery of health promotion messages toward children and adolescents [41]. For this to be effective, teachers should first of all be well-informed themselves and also encouraged to put into practice their health knowledge. This could perhaps be implemented through training programmes aimed at enhancing the role of teachers as health educators and empowering them. This would be a necessary first step to change behavior effectively among their pupils and students [10].

The strength of our study rests on the quality and richness of the source data obtained through a nationally-representative telephone survey using validated epidemiologic instruments. In particular, we were able to take into account important demographic and socioeconomic confounders. Such adjustment was crucial to identify the specific association of being a teacher with the health-related variables analyzed [18, 42].

However, we were not able to explore certain critical areas of disease prevention and health promotion (notably, physical exercise, which is known to be an important determinant of health risk [43, 44]) nor were we able to explore heterogeneity of health behaviors between teachers. In fact, the data required for such analyses were not available in the Health Barometer survey. Another limitation of the study was that our indicators of health behaviors were based on self-report, and the reliability of the data so obtained may be affected both by selective response and by distorted response among participants. First, and due to non-response, prevalences of at-risk behaviors may have been underestimated. Indeed, some evidence supports that respondents in population-based survey are healthier than non-respondents [45–47]. Consistently, in the 2010 Health Barometer, teachers were more likely to respond than other workers. Although data were weighted taking teaching status, socio-economic and demographic characteristics into account, this is not sufficient to exclude the risk of a residual “healthy respondent” bias, where those who answer the survey have healthier behaviors than those who decline. However, it would not be a problem in this association study, unless this bias is differential according to teaching status, which we have no reason to believe. Second, we cannot exclude a social desirability bias, whereby teachers would be less prone than other adults to report unhealthy behaviors, which could confound, at least in part, the observed associations [48]. Yet, differential desirability bias, if present, would be of low magnitude, as telephone survey provides more anonymity than face-to-face interviews. Moreover, the eventual tendency of interviewed teachers to conform in their responses to the social ideal of their professional category would be limited as the job question was put at the very end of the survey, along with the other socioeconomic items. Finally, given the cross-sectional design of our study, associations between teaching status and health behavior should not be interpreted in terms of causality.

In conclusion, this study examined teachers’ health behaviors and provides a first step toward understanding teachers’ attitudes towards health. Such knowledge is a prerequisite to improve health education strategies at school through a more effective involvement of teachers. The results suggest a relatively high level of health awareness but also identify alcohol consumption as a potential area for improvement. Such data can be useful in designing health education materials that better match the needs of teachers as health educators. A challenge for future research would be to evaluate teacher-pupil relationships better and to understand how this relationship may promote healthy behaviors in pupils.