Design
This study used a cross-sectional descriptive survey design to construct a model of the factors influencing the mental health of police officers based on the PRECEDE model (1980). The model was validated and shown to explain the relationships among each of the factors through data collection.
Participants
The target population of this study was police officers working in the metropolitan in South Korea. In path model analysis, the sample size should be from 10 to 20 times the number of free variables based on the maximum likelihood method [33]. The number of free variables in this study was 20, and when the free variables were multiplied by 17, a total sample size of 340 was calculated. The final sample size for this study was 500, reflecting a 45% nonresponse rate.
Police stations were selected for sampling potential participants. Non-probability quota sampling was done based on the number of police reports involving police stations, which is known as a key policing demand and burden indicator in the policing profession. The number of reports was collected by requesting access to the Korean National Police Agency. The data on the number of reported cases are not disclosed to the public, but data were requested and received for research purposes. Based on the total number of police reports (1,543,885) in 96 districts in the region in 2019. We divided the group of districts into 4 groups: upper, middle, lower, and lowest groups. Each group comprised 24 districts. The number of participants in each group was assigned taking into account the number of 112 reports, a number representing the workload of each group. The proportion of cases handled per district group was 38.9% (601,141) of the total number of cases in the upper group, 25.4% (391,450) of the total number of cases in the middle group, 20.3% (314,159) of the total number of cases in the lower group, and 15.4% (237,135). Therefore, from a target of 500 participants (100%), 195 (38.9%) in the upper group, 125 (25.4%) in the middle group, 102 (20.3%) in the lower group, and 78 (15.4%) in the lowest group were conveniently represented in the allocation table.
Measures
This study utilized measures of mental health, resilience, social support, resource availability, health behaviors, job stress, and traumatic event experiences.
Mental health
The Psychosocial Well-being Index Short Form (PWI-SF) [34] was used to measure mental health. The PWI-SF was developed based on Goldenberg’s General health questionnaire-60 (GHQ-60) and modified to fit the Korean context. The instrument consists of 18 questions on social performance and self-confidence, depression, general well-being and vitality, and sleeping disturbance and anxiety. Each question was answered on a 4-point Likert scale ranging from 0 (very much) to 3 (not at all), and the total score was calculated by summation, with a total score ranging from 0 to 54. Higher scores indicated poorer mental health. When developing the tool, cut-off points for mental health levels were proposed, with a score of 27 or higher categorized as “high risk,” 9 to 26 as “potentially stressed,” and 8 or lower as “healthy.” Cronbach’s α was.92 in this study.
Resilience
The Korean version of the self-resilience scale was used to measure resilience [35, 36]. The tool consists of 25 items in five domains: hardiness, persistence, optimism, support, and spirituality. Each item is scored on a 5-point Likert scale ranging from 0 (not at all) to 4 (very much). Total scores range from 0 to 100 points; higher scores indicate higher resilience. The Cronbach’s α was .93 in this study.
Social support
The Korean version of the support scale was used to measure social support [37, 38]. The tool consists of eight questions, including concerning support from supervisors and regarding support from coworkers. Each question was answered on a 5-point Likert scale ranging from 1 (not at all) to 5 (very much). The total score was calculated by averaging and ranged from 1 to 5. Higher scores indicated higher levels of social support. Cronbach’s α was .83 in this study.
Resource availability
To measure mental health resource availability, we used the tool of attitudes regarding the use of community mental health services [39]. This tool is a translation and revision of the help-seeking availability scale barriers [40] and the help-seeking attitudes scale [41, 42]. The instrument consists of 17 items in five domains: fear of mental health services, trust in mental health services, self-exposure, internal control, and barriers. Each question was answered on a 4-point Likert scale ranging from 1 (not at all) to 4 (very much). The total score ranged from 17 to 68. Cronbach’s α was.87 in this study.
Health behavior
To measure health behavior, we used the tool of Health Behavior Questionnaire [43]. This tool consists of a total of 16 questions on smoking, alcohol consumption, diet, exercise, sleep duration, weight control, and perceived health. For the 8 habits of smoking, drinking alcohol, diet, exercise, average sleep time, regular meals, snacking, and breakfast, 1 point was given for action, and 0 points were calculated for no action. Total scores were calculated from 0 to 8 points. Cronbach’s a of the Kuder-Richardson-20 (KR-20) in this study was .40.
Job Stress
To measure job stress, we used the Korean Local Police Job Stress Tool [44]. The tool consists of 17 questions in five domains: regarding organizational culture, field actions, complaint handling, training and supervision, and the working environment. Each question was answered on a 5-point Likert scale ranging from 1 (not at all) to 5 (very much). Scores were calculated by averaging and ranged from one to five. Higher scores indicated higher levels of job stress. Cronbach’s α was .87 in this study.
Traumatic experiences
To measure the traumatic experience, we used the Traumatic Event Experience Scale [45]. This tool was developed by Thomas-Riddle and translated and modified in Korea [46]. The instrument consists of 23 items in two domains: traumatic events and secondary traumatic events. The scale measures participants to check whether they have been exposed to each event and rate their traumatization on a scale from 0 to 10. The score was calculated by dividing the traumatization by the number of traumatic events based on the previous study [46]. In this study, traumatic event exposure was constructed, and the degree of traumatic event impact was modified on a 5-point Likert scale from 1 (not at all) to 5 (very much). The scores were calculated by dividing the sum of the traumatizing events by the total number of traumatic events. The resulting scores range from 0 to 5, with higher scores indicating more traumatic experiences. Cronbach’s α was.93 in this study.
Demographic characteristics
To identify demographic characteristics, the questionnaire was composed of questions on gender, age, marital status, domestic family status, education level, height/weight, fatigue, disease diagnosed by a doctor, and treatment. To identify working characteristics, station, rank, responsibility, work period, shift work period, and job satisfaction were surveyed.
Data collection and procedure
We obtained permission for data collection from the captain of each police station and obtained informed consent from all participants. We visited 15 police stations in the metropolitan area between July and August 2020. Data were collected using a self-administered questionnaire. After the participants completed the survey, they submitted it to the researcher and received a small gift. The study was approved by the Institutional Review Board and Ethics Committee of the Yonsei University Health System (approval number: Y-2020-0045). Each participant provided their written informed consent.
Data analysis
The data were analyzed using the STATA 17.0 program. Participants’ mental health, resilience, social support, resource availability related to mental health, health behavior, job stress, traumatic experiences, physical environment, demographics, and job characteristics were analyzed using descriptive statistics. The variables for each factor were analyzed using Pearson’s correlation coefficients. In the path analysis, to match the direction of the factors, mental health level was inversely converted and analyzed. Therefore, in the descriptive statistical analysis, a higher mental health score indicated poorer mental health, whereas, in the correlation and path analyses, a higher score indicated better mental health.
Path analysis was performed to examine the direct and indirect paths of mental health based on a hypothesized path model. To verify the fitness of the hypothesized model, standardized chi-square index (χ2 /df), goodness of fit index (GFI), standardized root mean residual (SRMR), root mean squared error of approximation (RMSEA), Turker-Lewis index (TLI), and comparative fit index (CFI) were used. χ2(CMIN) is the most basic measure of model fit, with a low value and p ≥ 0.90 suggesting acceptable fit. Model fit indices goodness of fit index (GFI), Tucker–Lewis index (TLI) with indices ≥ 0.95 indicating good fitting models, whereas indices ≥ 0.90 suggest acceptable fit. SRMR and RMSEA indices ≤ 0.08 indicate acceptable model fit [33]. The 95% confidence interval (CI) did not include 0, indicating a significant difference.