Design and population
This was a longitudinal study based on two questionnaire surveys performed at an interval of 1 year. Between September and December 2018, 4935 primary and lower secondary school (grade 0-9) teachers from 105 schools in Denmark were invited to participate in the baseline survey and 2336 (47%) participated. The schools were selected randomly from a list of all municipal primary and lower secondary schools in Denmark, taking into account geographical spread and school size, so that all five main Regions in Denmark and both small and large schools were represented. The school principal gave consent for the school to participate. The teachers’ e-mail addresses were obtained from the participating schools, and questionnaires were subsequently sent to all the employed teachers by e-mail. At follow-up, 94 of the 105 schools in the baseline survey agreed to participate. In the autumn of 2019, 3883 teachers from the 94 schools were re-invited and 1830 (47%) responded. This resulted in a study population of 1198 teachers from schools of different sizes and geographical areas who responded at both survey rounds, see Fig. 1. Participation in the study was voluntary, and collected data were treated confidentially.
Pupils´ aggressive behaviour
Information about pupils´ aggressive behaviour was based on the definition of work-related violence by Wynne, Clarkin, Cox, and Griffieths [30] and measured by questionnaires regarding harassment, threats, and violence completed by the teachers. Information was obtained from the baseline questionnaire in 2018. The respondents were asked whether they had been exposed to 10 different expressions of harassment, 7 different expressions of threats, or 12 different expressions of violence within the last 12 months.
The items about harassment concerned experiences like, e.g., being verbally patronized, having private property stolen or destroyed, or having experienced harassment by parents during the last 12 months. A total sum score with a possible range of 0-40 was calculated for each participant, and the scale was then dichotomized at the 75th percentile into high level of harassment (≥6) or low level of harassment (≤5).
Threats were measured by items asking about, e.g., being threatened with objects, written threats, or being scolded or shouted at in a threatening manner during the last 12 months. A total sum score with a possible range of 0-28 was calculated for each participant, and the scale was then dichotomized at the 75th percentile into high level of threats (≥2) or low level of threats (≤1).
Violence was measured by items asking about having experienced physically violent behaviours such as being hit, spat on, or bitten or experienced violence from pupils or parents outside work during the last 12 months. A total sum score with a possible range of 0-48 was calculated for each participant, and the scale was then dichotomized at the 75th percentile into high level of violence (≥3) or low level of violence (≤2).
Because there is no validated cut-point for the three types of aggressive behaviour, decisions on dichotomization at the 75th percentile were chosen a priori based on a desire to examine the 25% most exposed part of the sample in contrast to the rest of the sample.
Burnout
Burnout was measured by a scale from the validated Copenhagen Burnout Inventory (CBI) [20]. In contrast to the widely used Maslach Burnout Inventory [31], the CBI focuses on measuring only fatigue and exhaustion using three rather similar scales (personal or general burnout, work-related burnout, and client-related burnout). To not overburden the participants with questions, we choose to use the personal burnout scale, which is the scale that measures whether a person experiences being exhausted due to private life, work life, or both.
The scale consists of six items (Cronbach’s alpha 0.88) measured at follow-up.
A total sum score between 0 and 24 was generated for each individual and dichotomized into ≤11, indicating no symptoms of burnout, and ≥ 12, indicating symptoms of burnout, as recommended by Borritz et al. [15].
Stress
To measure stress in the follow-up questionnaire, we used a validated single-item question in which stress was defined as a situation where you feel tense, restless, nervous, anxious, or have difficulty sleeping at night because you are thinking about problems [32], and added an additional two questions to get a more specific measurement: 1. “Are you experiencing this kind of stress at the moment?” [32]. This information was dichotomized into low stress level (to some degree, only a little, not at all) or high stress level (very high degree, high degree). 2. “For how long have you experienced this kind of stress?”. This information was dichotomized into short stress duration (less than 2 weeks, 2-4 weeks) or long stress duration (1-3 months, more than 3 months). 3. “What do you think is causing this kind of stress? Mainly conditions at work, mainly conditions in my private life, a combination of conditions at work and in my private life, do not know/other things.”
Based on the information about stress level and duration (items 1 and 2), a new accumulated stress variable was generated with the following four categories: 1. low stress level, short stress duration, 2. low stress level, long stress duration, 3. high stress level, short stress duration, 4. high stress level, long stress duration.
Social support
Social support from colleagues and a supervisor was measured at follow-up by the following two items from the Danish Psychosocial Questionnaire [33]: “Is there a sense of community and cohesion between you and your colleagues?” and “Can you talk with your immediate supervisor about difficulties you experience at work?” The items were measured on a 5-point Likert scale ranging from a very small extent (0) to a very large extent (100), and both scales were dichotomized into high support (≥75) and low support (< 75).
Finally, information about the covariates gender, age, and seniority was derived from the baseline questionnaire.
Statistical analysis
Covariates were selected a priori based on a review of the literature. A correlation analysis between stress and burnout at follow-up was conducted, revealing a correlation coefficient (r) of 0.49. Except for correlations between the three main exposures and between seniority and age, none of the additional correlations exceeded r = 0.31 (Additional file 1). Of the 2336 responders at baseline 1138 were non-responders at follow-up (Fig. 1). A comparison between the responders and non-responders according to gender, age, and seniority was performed.
The distribution of exposure, outcome, and covariates of the study population was presented as numbers and percentile distribution. To investigate the associations between harassment, threats, or violence and burnout, multilevel logistic regression analyses, taking into account the possible correlations between teachers from the same schools, were performed. The multilevel logistic regression was based on a two-level analysis. The first level representing the 1198 teachers and the second level representing the 94 participating schools. Crude and adjusted odds ratios (OR) with 95% confidence intervals (95% CI) were calculated. Two adjusted models were performed. Model 1 adjusting for age, gender, seniority, and baseline burnout. Model 2 further adjusting for stress. Finally, the associations between the three types of pupils´ aggressive behaviour and burnout were stratified for each of the two measures of social support (from supervisors and from colleagues). All statistical analyses were performed using the statistical software package Stata, version 16.1 (Stata Corporation, College Station, Texas, USA).
Ethical considerations
The study was approved by the Danish Data Protection Agency, which is an independent authority that supervises compliance with the rules on protection of personal data. According to Danish Law (Act on Research Ethics Review of Health Research Projects), available at: www.nvk.dk/english/act-on-research (accessed on: 8 October 2021), questionnaire- and register-based studies require neither approval by ethics or scientific committees nor written informed consent. However, the participants gave informed consent by ticking a box in the questionnaire after being informed about the purpose of the study and how their own answer could be deleted if they wanted and hereby agreed to participate in the survey.