Composed of 202 federal highway police officers, working in the state of São Paulo, coming from ten regional police stations, in addition to the PRF Headquarters, located in the city of São Paulo.

We used the following research instruments: a General Questionnaire (GQ), consisting of 41 questions, for the purpose of characterizing the participant in sociodemographic and occupational terms (sex, leisure, age, educational level, career length, number of children, and having been criminally prosecuted), as well as the characterization of the stress perceived by the participant regarding variables related to the environment and work routines. This measurement was made according to the participant’s degree of agreement with statements that characterized each type of work situation, through Likert-type scales, which ranged from 0 (not true) to 10 (true) (e.g.: “ I feel that my work puts my life at risk on a daily basis”). These statements covered the following domains of the participant’s professional life: (1) self-perception of life-threatening danger (daily risk of life, risk of life at work), (2) self-perception of job satisfaction (degree of job satisfaction, low remuneration); (3) self-perception of the physical and emotional conditions involved in being a PRF officer (having daily stress, feeling physically or mentally tired after a day at work, having frequent insomnia); (4) self-perception of the working conditions offered by the corporation (availability of material and equipment resources, small teams, lack of support from the institution, lack of training as a police officer, lack of team training, long working hours, period of work, and whether having a good environment would improve the relationship at work); (5) self-perception of the conflict between work and family life (feeling aggressive after joining the corporation, conflicts between work routine and family); (6) self-perception in dealing with situations that could result in critical incidents (dealing with ill-natured people as Law offenders; feeling stress with serious accidents); feeling stress when dealing with fatalities and/or serious injuries; feeling stress with reporting accidents and deaths to family members; death resulting from the performance of their legal duties) and finally (7) reporting their suicidal ideations in the twelve months prior to the interview. Finally, the officer’s emotional stress was objectively measured using Lipp’s Stress Symptom Inventory for Adults (LSSI), which also identifies the phase and type of predominant symptom (physical or psychological) among people who meet the criteria for stress. The LSSI presents high reliability, with Cronbach’s a (Alpha) coefficient values between 0.9083 and 0.9124 (Lipp, 2000).

The sample size was calculated using the following parameters: expected proportion of stress of 45%, relative precision of 15%, and confidence level of 5%. For sample selection, we used the random and stratified sampling technique (the police station in which it operated was considered as a stratum) and, for operational convenience, we opted for twenty police officers per stratum. We chose to disclose the results on a consolidated basis, not by stratum – in order to comply with the secrecy requirements. Thus, from a reference list, delivered by the Department of Human Resources of the PRF, the sample components were drawn and then summoned to a meeting, aiming to bring everyone together on the same day and time. On the stipulated date, the selected servants were informed about the purpose of the research and the voluntary character of participation. Servants who were officers and who were allocated in the state of São Paulo were included in the sample. Four police officers who resigned their participation in the research were excluded (justified by feeling tired after 24 hours of work; leaving for a trip; family member hospitalized in the Intensive Care Unit; and, finally, one officer did not feel comfortable participating in the research). Upon consenting to participate, the police officer received the Informed Consent Form and the research instruments. Data collection took place on the premises of the police stations, in rooms or auditoriums, depending on the local infrastructure. The application time ranged from 30 to 50 minutes. If an officer chose to leave the study or could not participate, new police officers were drawn, and a new date was scheduled.

We used the R3.3.2 software for descriptive and inferential statistical analyses. Initially, we calculated the prevalence of the sociodemographic and occupational characteristics of the sample and, later, we estimated the prevalence of the outcome (stress) and the association with the independent variables (general questionnaire). In this sense, intergroup differences (bivariate analyses) for categorical (qualitative) variables were estimated using the chi-square or Fisher’s test. Integral differences (bivariate analyses) for numeric variables (quantitative) were evaluated using the Student’s t-test (for variables with normal distribution) and the Mann-Whitney test (for variables with non-normal distribution). The normal distribution (or not) of numeric variables was evaluated using the Shapiro-Wilk test.

Subsequently, the variables that reached statistical significance in the bivariate analyzes (p ≤ 0.05) were separated into eight multiple regression models (divided according to the nature of the variables, that is, whether sociodemographic, occupational, or self-perceived stress, as described in the item about the instruments used in this study, illustrated in Table 1) to identify the explanatory variables related to the outcome (stress; LSSI). At this level of analysis, the explanatory variables of the outcome were identified after controlling for the confounding effect arising from the sociodemographic variables. The association between the variables was measured using the Prevalence Ratio (PR). The adequacy of the adjustment was evaluated through the analysis of residues, as well as through the evaluation of multicollinearity between the variables through the Variance Inflation Factor coefficient. In all cases, the null hypothesis was rejected for p-value less than or equal to 0.05 (p ≤ 0.05).

To carry out this study, the ethical standards that control research involving human beings were observed; we assured participants that their personal data would be kept confidential. The Ethics Committee of the Faculdade de Medicina da Universidade de São Paulo (Faculty of Medicine of the University of São Paulo) approved the study, under n° 114/19 and by the General Manager of the PRF, process nº. 238/2012.

The predominant profile identified for the sample was: male (93.1%); median age 43; being married or in a stable union (82.2%); having children (76.3%); having a college degree (85.6%); having time for leisure activities (86.6%); and residing in the same location as their work (51.5%). Professionally, the predominant profile comprised individuals: with a median career length of 15 years; performing an operational function (62.4%); working on a shift schedule (69.4%), day and night (73.3%); and not performing any other professional activities (94%). Finally, 82.2% of the professionals reported never having been criminally prosecuted.

According to the LSSI results, we estimate that 43.1% (95% CI = 36.2–50.0) of the police officers (N = 87) were experiencing stress at the time of the interview. Of these, 2.3% (95% CI = 0.2–8.0) were in the “alert” phase; 82.7% (95% CI = 73.2–90.0) in “resistance”; 11.5% (95% CI = 5.7–20.1) in “near exhaustion” and 3.5% (95% CI = 0.7–9.7) in “exhaustion.” As for the type of symptom, 60.9% (95% CI = 49.9–71.2) of the individuals manifested a predominance of psychological stress, 33.3% of physical stress (95% CI = 23.6–44.3%), and 5.8% (95% CI = 1.9–12.9) both (*).

The results of the associations between the sociodemographic and outcome categorical variables (stress; LSSI) are shown in Table 2. The association of the outcome was identified only with the leisure variable; in this sense, having time set aside for leisure reduced the prevalence of emotional stress symptoms among police officers by 40% (PR = 0.6; 95% CI = 0.4–0.8).

Results of associations between numeric and outcome variables (stress; LSSI) are shown in Table 3. Participants who had stress symptoms were younger than their colleagues without stress (p = 0.05). Additionally, officers who did not present stress symptoms reported a higher degree of satisfaction with their professional activity (p < 0.001).

The relationship between categorical occupational variables and the outcome (stress; LSSI) is shown in Table 4. Having thought about suicide in the (12) twelve months prior to the interview increased the prevalence of stress by almost 80% (PR = 1.8; CI 95% = 1.2–2.5).

The relationship between the variables of self-perception of stress (in the face of routine work situations) and the outcome (stress; LSSI) is shown in Table 5. In this sense, the police officers categorized as “stressed,” according to the LSSI, evaluated all situations in the routine life of the police as stressful, except for the variable “good work environment improves relationship at work”.

The results of the logistic regressions performed for the stress outcome (LSSI) are described in Table 6, separated according to the model. Thus, controlling for the effects of sociodemographic variables, we identified the following results: (a) in Model 2, we identified that the variable “having had suicidal ideations in the last twelve months” (Yes/No type) emerged as a risk factor for stress among servants (p-value = 0.031, coef. +1.346); (b) in Model 3, the participant’s perception of life-threatening danger they face daily as part of their police role did not interfere with the level of stress; (c) in Model 4, the variables “degree of job satisfaction” and “remuneration incompatible with the role” interfered with the outcome as follows: the lower the “degree of job satisfaction” (p-value = 0.001, coef. -1.31) and the greater the perception of “remuneration incompatible with the role” (p-value = 0.012, coef. +0.166) the greater the probability of stress; (c) in Model 5, the variable “insomnia” interfered with the outcome as follows: the higher the level of insomnia (p-value = 0.0021, coef. +0.156), the greater the probability of stress; (d) in Model 6, the variable “lack of training as a police officer” interfered with the outcome as follows: the greater the perception of “lack of training as a police officer” (p-value = 0.011, coef. +0.288), the greater the probability of stress; (e) in Model 7, the variable “participant’s perception that being an officer interferes in family life” interfered as follows: the greater this perception of interference (p-value < 0.001, coef. +0.308) the greater the probability of stress; (f) Model 8: self-perception variables in dealing with situations that can result in critical incidents (acute stress) were not related to stress.

Regarding the sociodemographic variables, we observed that the leisure and educational level variables remained in some models, as follows: (a) the leisure variable remained in all models, and can be considered as a protective factor against stress, except when included with the variable “self-perception of the conflict between work and family life,” when it lost significance; (b) the educational background variable interfered with the outcome as follows: the higher the educational background, the less likely the officer was to developing stress (V4 – p-value = 0.018, coef. +1.698; V8 – p-value = 0.023, coef. +1.662).