Instruments and variable definitions

The paper–pencil questionnaire consisted of 65 closed questions (16 pages) and took approximately 30 min to complete. Teachers completed the questionnaire during their free time at the schools or at home. Teachers did not receive any type of direct incentive for their participation in the study.

Sociodemographic variables and working conditions were explored through the VI. National Survey of Working Conditions and Health (Spain)31 and the Unesco study on working conditions and health in teachers.19 They included: age (assessed in four categories: ≤29, 30–39, 40–49 and ≥50 years, but subsequently recategorised into two groups: <40 years, ≥40 years, based on the distribution of the variable), gender (male and female) and school location (rural and urban). Teaching level was grouped into exclusive primary, primary and secondary, and exclusive secondary. Hours per week spent in extracurricular activities (carried out outside the school) included: preparing school lessons, preparing teaching material, preparing extraprogramme activities, meeting with parents or pupils, planning and team working, administrative duties, participating in courses and correcting class tests. Teachers were asked to report the number of hours per week they regularly spend for each of these activities. Later, this variable was dichotomised at the median value (<10 and ≥10 hours/week). Finally, monthly income was explored taking as reference the minimum wage for Bolivia. It was grouped in: low (<US$238: less than the minimum wage), medium (US$239–US$476: between 100% and 200% of the minimum wages) and high (>US$476: more than double the minimum wage).

Data analysis

Questionnaires were double entered into EpiInfo V.7 for Windows. After a congruence check, data were exported to IBM SPSS Statistics V.24 for further analysis. All the numerical variables included in the study (Effort–Reward Scales, extracurricular activities and GHQ-12 score), did not follow a normal distribution, so the median and range were reported. For categorical variables, absolute and relative frequencies were calculated for the total population. Working conditions and health indicators for teachers who worked in urban and rural areas were compared using a χ²  test (categorical variables).

After an analysis of the pattern of incomplete data, missing at random mechanism was assumed, and missing values were replaced via multiple imputation by fully conditional specifications.39 In total, 10 imputation data sets were created using logistic regression equations as the imputation method (categorical variables). Combining those, logistic regression models were used to calculate crude and adjusted ORs, as well as their corresponding 95% CIs for the four health outcomes. Based on previous studies3 7 17 and the bivariate analyses, models were adjusted for age, gender, school location, teaching level, type of school, hours per week spent in extracurricular activities and income. Interactions between ERI and OC were tested. As a sensitivity analysis, we performed complete cases analysis.

Ethics approval and consent to participate

In order to protect the rights and well-being of participants, international ethical recommendations for research with human beings40 were followed. An information letter, as well as a written informed consent form, was provided to each participant. The questionnaire was completely anonymous and voluntary participation was respected at all times.

Patient and public involvement

No patients/public were involved in the development of the research question, design or implementation of the study. Dissemination of the results of the study was made available to the health and education authorities and a summary of the study was sent to each participating school.

Sociodemographic and work-related characteristics of urban and rural teachers

Teachers working in urban areas were statistically significantly older, were more likely to be female, work at the primary level, have high income and always walk or stand while working compared with teachers working in rural areas. There were no statistically significant differences between teachers working in rural and urban areas for extracurricular activities (table 1).

Median values for effort–reward Ratio showed higher reward in comparison to the effort (median 0.91; range: 0.3–2.3) in both groups, even though the median value for teachers working in rural areas was less favourable than those in urban areas (0.96 vs 0.82; p<0.01) (table 1). Likewise, the prevalence of teachers reporting ERI >1 was higher in teachers working in rural areas (35%) compared with those employed in urban areas (22%; p<0.01).

Health indicators

In general, a high percentage of teachers perceived their health as being fair or poor with higher prevalences among teachers working in rural areas (50%) compared with urban areas (28%; p<0.001). Likewise, the median value of the sum of the GHQ-12 score was significantly higher in teachers working in rural areas compared with their counterparts (4.0 vs 5.0; p<0.01). Applying a GHQ-12 score of 5 or above as a cut-off point to consider mental distress, more than half of the teachers in rural areas (51%) and 38% of those in urban areas were affected. Musculoskeletal pain during the last 7 days was reported by 29% of the teachers for neck or shoulder pain and by 17% for low back pain, without statistically significant differences between teachers employed at rural or urban schools (table 2).

Logistic regression models for health indicators

In the univariate model, teachers with ERI scores >1 compared with those with lower scores were statistically significantly more likely to self-rate their overall health as fair or poor (crude OR 1.9; 95% CI 1.2 to 2.9), to suffer from mental distress (2.0; 95% CI 1.3 to 3.0), to report low back pain (2.4; 95% CI 1.4 to 4.0) and neck or shoulder pain (1.7; 95% CI 1.1 to 2.7). After adjustment, the results remained consistent. However, they were no longer statistically significant for neck or shoulder pain. Independently of the other variables, women (as opposed to men) and those working in rural areas (as opposed to urban areas) were more likely to report worse health indicators (tables 3 and 4). The results were consistent with the case complete analysis (see online supplementary additional tables). The ERI and OC interaction did not show association with any health indicator, therefore, the interaction term was not included in the final models.

Health outcomes

While there is little information available on the prevalence of self-rated overall health in teachers, data for the general adult population in Latin America indicate a high variability between and within countries. For example, one study carried out in multiple Latin American countries found a prevalence of self-reported health rated as poor among 22% of people in Uruguay and 49% in Brazil.41 In comparison, the prevalence among our study population could be considered relatively high (43%).

With respect to mental distress, the prevalence in our study (45.4%) was higher than the one reported previously in German teachers (26.2%),3 42 and quite similar to the one reported in Japanese teachers (47.3%). However, the latter study applied a lower GHQ-12 cut-off point (≥4).43 Applying this cut-off point to our study population, the prevalence of distress would have been 58% and thus higher than among Japanese teachers.

Comparison between teachers working in urban and rural areas

In our study, teachers working in rural areas were more likely to perceive their health as fair or poor, to report a higher percentage of mental distress and to suffer from musculoskeletal pain than teachers from urban areas. These differences could be partially explained by demographic variables, socioeconomic status and availability of healthcare resources across countries and cultures.44 45

In Bolivia, a teaching career in public schools frequently starts in rural areas. With a job promotion, teachers may then change to an urban school after their first years on duty. This explains the younger age and lower income of the teachers working at rural schools observed in our study. For many young teachers, this means living away from their families and friends at the beginning of their career in a cultural setting where family and friends are more important than in many high-income countries. Given the infrastructure in Bolivian rural areas, the travel time from school to home town might be long, only permitting rare visits home during the school year. In addition, the low salary of school teachers in Bolivia (every fifth teacher in our study only received the minimum wage) was even lower for rural teachers. Therefore, it is not surprising that rural teachers in our study perceived the relationship between effort and reward of their work as worse than their urban counterparts.

Association between ERI and health indicators

Our results confirm previous studies indicating that ERI might be a risk factor for health outcomes including self-rated overall health,46 low back pain47 and mental distress3 in different working populations. Self-rated health was described to be a good indicator of health status and mortality risk48 especially for women.49 Work-stress could lead to changes in the nervous system that condition the perception of one’s own health or the start of a health problem. Falconer and Quesnel-Vallée50 demonstrated with a nationally representative longitudinal study in Canada that health information and acknowledgement of their disease diagnoses influence subjective health in an important way.

Similarly, our results indicate an association between ERI and mental distress,7 42 independent of socioeconomic and working characteristics. Although it was proposed that an interaction between ERI and OC could affect the association between ERI and physical and mental health,18 our results are consistent with other studies which do not confirm such effect modification by OC.51 This means that the association between psychosocial work environment and health in our population of rural and urban teachers is not modified by the intrinsic factor of OC to work.

Several studies have previously reported a relationship between a poor psychosocial work environment and MSDs in school teachers.52 This relationship could potentially be explained by the variety of job functions they perform inside and outside the classroom apart from the demands of family life. According to Bugajska et al 53 ‘adverse psychosocial work factors increase physical load’. In this sense, high job demands could be related to long working hours, few breaks at work and infrequent changes in posture. As reported in other studies, our results showed a weak association between psychosocial factors at work and neck or shoulder pain during the last 7 days,47 54 probably due to the effect of other variables not considered in the study.

Strengths and limitations of the study

This is the first study in Bolivia in which employment and working conditions were studied as potential risk factors for physical and mental health conditions in school teachers working in rural and urban areas. For this, we used validated and standardised questionnaires to measure our main variables in order to compare our findings to other studies around the world. The ERI questionnaire was recognised as a good tool to assess psychosocial risk at work51 and to predict burn-out as the consequence of incompatibilities between professionals and their job contexts.55 Self-rated health is reported as a validated indicator of a ‘wide spectrum’ of health conditions.56 In the same way, the GHQ-12 showed a good performance as a screening instrument for common mental disorders and was not influenced by gender, age or educational level.36 Finally, the Nordic Questionnaire is recommended as a screening tool in occupational settings with a high sensitivity and specificity, especially for chronic or recurring low back pain.57 However, as a limitation, there were no diagnostic evaluations for any of the health outcomes included.

As in other cross-sectional studies, causal inference is not allowed, because we cannot assume that exposure occurred prior to the outcome. Additionally, the presence of the ‘healthy worker’ bias is common in many occupational studies and it may be present in our study as well. It is possible that teachers over 40 with more health complaints were retired or on sick leave.

Response in our study was moderate. One reason for this was that in some schools, even after several reminders and personal visits to the schools, the head teacher forgot to distribute the questionnaires to the teachers. One may speculate that these were schools had worse conditions compared with the participating schools. Therefore, the selection of teachers from schools with better conditions might have resulted in an underestimation of the true exposure and outcome prevalence. On the other hand, it is also possible that teachers with dissatisfaction about working or health conditions may have been more likely to answer the questionnaire, thereby overestimating the prevalences.

Due to difficulties in geographical access, our study excluded a priori schools with less than five teachers, most of which were located in remote and dispersed rural areas. Therefore, the generalisability of our study is limited to teachers working at schools with at least five teachers.