Police officer occupational health: a model of organizational constraints, trauma exposure, perceived resources, and agency Authors: Royce Anders, Angélique Frapsauce, Celine Sauvezon, Daniel Gilibert Published: 2024 ABSTRACT BackgroundPolice officers constitute a work force at high risk due to their highly demanding work conditions. In a realistic paradigm, these conditions, and other determinants of their psychological health, can be linked to a multitude of variables that interplay altogether. However, current literature that simultaneously models–quantitatively from observed data–such a multitude of variables is sparse. This study took upon this objective to further theoretical and applied understandings through a measurement framework on empirical data, and allow the data to drive some features of model development, such as variable groupings into factors, and paths between factors.MethodsA total of 1312 officers from various police bureaus fully responded to a questionnaire composed of validated instruments for assessing factors related to psychological and occupational health, consisting of more than 25 variables. Statistical analyses were performed in progressing complexity, namely t-tests, correlations, multiple regression, factor analysis, and path analysis with latent factors.ResultsThe regression analysis identified 10 significant variables, in which decision latitude, organizational justice, and work recognition/meaning were the most protective, and these 10 variables coincided with those found significant in the t-test and correlational results. In higher complexity, the latent path analysis resulted in a model of 6 factors: Psychological Health, Organizational Constraints, Trauma Exposure, Perceived Resources, Sense of Agency, and Esteem. Organizational Constraints (β = -0.32, inferred by psychological demands and role conflict), Perceived Resources (β = 0.31, social support, a self/work-esteem subfactor, and organizational justice), Sense of Agency (β = 0.30, decision latitude, hierarchical position, right to carry a firearm), and Trauma Exposure (β = -0.14, frequency/time since event, used a firearm, years of service) were found significantly associated with Psychological Health. Within each factor, specific variables could be identified as the most associated, such as role conflict for constraints, self/work-esteem for resources, decision latitude for agency, and frequency of and time since trauma for trauma exposure. Our results therefore encourage us to take into account not only agency, but also past professional experiences in models for managing well-being.ConclusionsProviding police officers with social support at work, recognition, work meaning, fair proceedings and pay (organizational justice, especially for female and young officers), decision-making power (decision latitude), and minimizing conflictual information and procedures (role conflict) is of utmost importance. Officers with higher years of service, working in lower population cities, and who recently used their firearm, should be considered for trauma counseling. The degree of psychological demands of police officers should be regularly assessed, and reduced if possible. Reminders of support and integration in the force for officers with variables linked to a perceived lack of agency may be useful in their facing work challenges. Future integrative modeling research may be crucial to better understanding the relative contribution of each variable and their interplay in realistic settings, providing also a framework for measurement. Introduction Research has demonstrated that the profession of police officers is demanding, involving regular exposure to stressful events [1] and work conditions [2, 3], leading to a heightened risk of mental health issues [4–6], such as major anxiety, depressive, and mood disorders. The missions of police officers may be particularly stressful due to their typically unpredictable, dangerous, and time pressing nature [7]. These missions include handling verbal and physical threats/attacks, witnessing death/suicides, having to use firearms, and handling thefts and accidents [8, 9]. The former three have been identified as the most common [10, 11]. It is not surprising that officers risk psychological distress. For example, a recent study involving 2057 officers found at least one in four officers (28%) to suffer high levels of psychological distress [12]. The challenging work conditions of police officers not only lead to a negative impact on their mental and physical health [13, 14], but can also impact their professional commitment [15, 16], and their interactions with citizens [17, 18]. Psychological distress is often a precursor or covariate of more serious psychopathological conditions such as Post-Traumatic Stress Disorder (PTSD, [19]), burnout [20], anxiety or depression disorders [21, 22], and suicide ideation [23]. In police research, these conditions are regularly found at higher rates than in the general population [24]. For example, while PTSD is generally found in the population between 0.7–3.9%, research on police officers has found it between 7–9% in New Zealand, the Netherlands and Brazil [25–27], 8–35% in Canada [28–30], and 9.9% in France [31]. Psychopathological conditions are well known to be linked to a higher risk of physical health problems [32, 33], alcohol dependence (Gersons et al., 2000), insomnia [34, 35], and job turnover [36, 37]. These problems are associated with performance impairments in officers’ capacity to respond to calls [38, 39] and staff shortages, compromising not only the impacted officer’s own safety, but that of their colleagues and civilians [40]. Psychological distress in police officers is therefore a multifaceted problem with critical consequences, deserving expert attention. There is some heterogeneity in the practices for assessing the mental health of police officers. Most studies focus on measuring correlates of psychological distress [12, 21], while others focus on those of psychological wellness [41], and in lower quantity, both [42]. While frequently found as negatively correlated [43–45], mental wellness and distress have been long established as two distinct subdimensions that constitute an overall underlying factor of psychological health [46, 47]. Univariate analyses may find some comparable conclusions of the determinants of each one separately, whereas factor analyses making use of both to measure well-being may provide for a more robust psychometry of overall officer mental health and its determinants. Research has noted that the presence of psychological wellness does not imply the absence of distress and vice versa [48], and involves interactions with resiliency and coping mechanisms. In line with this framework, the modeling and factor analyses in the present study take into account both distress and wellness to obtain a more informative outcome variable of well-being, herein named Psychological Health at Work (PHW), and aimed to identify its determinants. Based on the predominant theories of police officers’ PHW, at least three principal groups of determining variables may be detailed, and have guided the choice of variables that can be explored as predictors of mental health in this population. The first two are extrinsic: i. organizational factors and ii. operational factors (specific to the nature of the missions), and the third group iii. intrinsic factors (specific to each police officer). Each of these three groups of factors can be broken down into variables that play out either as resources or constraints. Our aim is to show how the numerous variables identified in the literature, as linked to these factors, may be structured altogether in a predictive model of mental health, which to our knowledge, has not been attempted before. A comprehensive literature review is provided on these three factors, as well as findings highlighting specific variables composing them, and related theories. Readers who are already familiar may consider proceeding to the Research Aims section. Organizational factors: resources and constraints The Job Demands Control Support (JDCS) model by Karasek and Theorell [49] is a pioneering model in the field of PHW, often utilized as the primary measure of workplace stress [50–53]. The JDCS model established the notion that PHW can be understood through the interaction of an employee’s capacity, or difficulty thereof, to use potential resources to handle job-specific demands. In this framework, the combination of high job demands, low job control and low support gives rise to stress at work. The JDCS model is referenced in police officer studies [54, 55]. Namely, when police officers experience high job demands, these tend to coincide with high levels of psychological distress, job dissatisfaction, and physical health problems, which are reinforced by a finding that they tend to perceive low job control and low social support [56]. In line with the JDCS model, the protective element of feeling more in control can be induced through an improved sense of autonomy, leading to improved PHW. It is acknowledged that a slightly increased sense of control beneficially contributes to positive perceptions about oneself, which in turn promote well-being and health, particularly in Western cultures [57–59]. A previous meta-analysis [60] showed that autonomy was linked to greater job satisfaction, higher organizational commitment, and greater task involvement. Similarly, the results of studies have demonstrated that employees with high autonomy felt more engaged in their work [61]. In a study of social workers [62] it was found that autonomy reduced the impact of stress on burnout and intention to quit. The other protective element, the link between higher perceived social support and PHW, is also supported by numerous studies in police officers [63–65]. However, beyond the traditional JDCS perimeter that defines demands, control, and social support, there are other notable determinants of PHW that are less obvious to understand in this framework. For example, stress factors of organizational origin have been highlighted since the 1980s in police officers [66–69]. These include role conflict, autocratic management style, insufficient material resources (equipment), excessive paperwork, poor training and restricted career prospects, lack of human resources (lack of staff, poor communication, and the slowness of legal procedures) [70]. There are a plethora of other organizational factors to consider: such as internal policies and their implementation, politics, administrative procedures, staffing and hiring, prerogatives, career development, pay, and transfers. Disagreement in some of these variables could lead to feelings of organizational injustice [71]. Moreover, municipal forces may greatly vary in these variables, which may exacerbate feelings of injustice when officers compare. In conclusion, the JDCS, even if it is widely used and which provides a foundational framework, cannot fully account for the multiplicity of organizational factors that influence PHW, especially for police officers. Research and models of PHW for police officers would do well to open up to more factors, including operational ones. Operational factors: resources and constraints Common operational variables include physical demands, work schedules, and material usage [72]. For example, in police officers, irregular work cycles (alternating periods of inactivity and overactivity, team rotation) are a commonplace operational stressor [73]. But most importantly, the operation of law enforcement, involving regular exposure to verbal threats and aggression, powerful emotional expression (e.g., anger, despair), violence, road accidents, suicide, and confronting the unknown, require officers to mobilize a great deal of emotional and cognitive resources [8, 72, 74–76], and triggers high stress [68]. These operational stressors have been associated with anxiety, depression, and PTSD symptoms in officers [4, 19, 77]. The latter link has been further consolidated in studies that especially focused on their exposure to traumatic events [78]. Given this complexity of factors driving PHW, more tools than the JDCS model and its measurement instrument, the Job Content Questionnaire (JCQ) [50], quantifying perceptions of work environment, are needed. For example, using this tool [79], it was observed that only 24.4% of men that find their job stressful are considered on job strain and 16.0% on iso-strain, and respectively 37.0% and 23.4% for women. Job strain is defined as high demands and low control and iso-strain as high demands, low control, and low social support. This observation inspired the premise [46] that it is important to distinguish between perceptions of environmental factors and health outcomes (well-being and distress) [80], which results from the adaptation to said environment. Subsequent theoretical models characterized this adaptation as what is known as a transactional process. Namely, the transactional model of Lazarus and Folkman [81] describes the cognitive process of evaluating the environment impacting psychological health as a double transaction between the subject and their environment: how the subject perceives the situation (loss, threat, or challenge) as well as their own resources (e.g., material and social means, possibilities of action, probability of success) to face it. Therefore, in this approach, stress is principally rooted in perception. Namely, stress arises when an individual recognizes an imbalance between the demands of the environment that they perceive and the resources they perceive as available to cope with said constraints [82–84]. Subsequent models continued along this path, emphasizing the importance of such subjective variables. For example, the Job Demands-Resources (JDR) model [85, 86], highlights the role of autonomy as a protective factor against stress at work, by allowing employees to adjust their work according to their abilities and preferences). Afterward, the Demands Resources and Individual Effects (DRIVE) model [87, 88] that further focuses on additional individual-level variables. The JDR framework was originally proposed as a model for burnout, and DRIVE for job stress, for which both their scopes became larger as these outcomes were increasingly recognized as being tied to other primary mental health outcomes. Complementary theories [89, 90] relax the premise that stress originates from an entirely intra-individual process. Nonetheless overall, employee perceptions can be considered just as important (if not more) than objective conditions [91–93]. Highly compatible with survey methods, perceptions are directly measurable from empirically-validated questionnaire instruments [94]. In the following subsections, we distinguish between several objective individual variables (seniority, sex, etc.) and subjective variables that are relevant to police officer research on PHW. Objective intra-individual variables: resources and constraints There are reasons to examine the role of sociodemographic variables in officers’ capacity to meet job demands. For example, expertise derived from years of experience in the field (when disentangled from accumulation of trauma exposure or stress) and level of education may be imagined as resources. In turn, the gender [28] and skin color [95, 96] of an officer may either play out as resources or constraints depending on the bureau’s socio-professional culture/district of practice, and which coping mechanisms are the most effective [97]. Other variables may include marks of socio-professional status and one’s capacity to control or exert demands, such as their right to carry a firearm at work. Though, limited research is available that brings to light the consequent role that these variables may play. Subjective inter-individual variables: resources and constraints This section considers variables that relate to an individual’s subjective appraisal of organizational elements concerning him/her. In line with major models of PHW previously presented, how an individual perceives and/or utilizes the resources available to them to meet job demands and constraints will drive their degree of wellness; but some individual variables also turn out to be added constraints in satisfying these demands. In the existing literature on police officers, subjective factors that have been associated with a positive impact on PHW include the meaning of work [41], the perception of recognition from hierarchical superiors and a participative management style [98]; interpersonal relationships with work colleagues, and social support provided by peers and supervisors [63–65], low role conflict [99], and the perception of organizational justice [100]. Developing further, based on several previous studies [41, 101], the meaning that an officer attributes to their profession may be a key resource that he/she may call upon to overcome job demands. Inversely, the absence of meaning may equate to a risk or constraint factor. Several authors have recognized that the increasing shift towards preventive actions and consequently repetitive missions (after national/regional security incidents) have led to reduced meaning and greater levels of work stress [102], in turn linked to higher levels of police burnout and suicide [103]. Beyond this loss of meaning, municipal police officers often suffer from a lack of recognition, and may risk, when considering themselves as occupying a minor role in these preventative duties, that they are perceived by citizens as lower-grade policemen, compared to their colleagues assigned in active roles [71]. Recognition at work has been linked to a multitude of beneficial outcomes: positive individual attitudes (e.g. motivation, work meaning) [104], reduced absenteeism [105], positive interpersonal relationships, autonomy, and organizational loyalty [106, 107], improved emotional state and professional well-being [108], reduced distress and burnout [109, 110], and greater work performance [111, 112]. Lack of recognition, conversely, can maintain employees' emotional exhaustion [113], encourage them to be absent or leave the company [114], promote depression or suicide at work [115] and increase PTSD rates and symptomatology [116]. Indeed for police officers, research has confirmed that strong recognition from hierarchical superiors is linked to reduced stress [98] and improved PHW [70, 117]. Furthermore, to have obtained the authorization to carry a weapon, can be perceived indirectly as recognition of the institution or of one’s status as an officer, and a guarantee of personal safety. But this variable can also have paradoxical effects: carrying a firearm exposes the officer to using it or being placed in additional situations thereto. In this way, a resource/protective factor can doubly play as a constraint leading to stress or risk of PTSD. Research has also established links between organizational justice and health [118], especially mental health [119, 120]. According to Morissette and Wemmers [121], in the general population, perceived interpersonal justice: defined as the perceived quality and fairness of interpersonal interactions in the workplace, appears to have a therapeutic influence. Their results demonstrate lower PTSD scores for individuals that did not perceive problems in interpersonal justice. Conversely, perceptions of injustice are associated with increased job stress [122], anger and self-preserving behaviors [123], police misconduct [124], and deviant behaviors [125]. With regard to police officers, much research has focused on the links between perceptions of organizational justice and the quality of police officers' relationships with constituents: in police populations, perceptions of fair treatment may help them cope with uncertainties [126], and reduce the negative impact of adverse events [123, 127]. Especially, procedural justice could be a resource for officers feeling vulnerable in the face of unpredictable situations [128]. Procedural justice is defined as the perceived fairness of the processes by which decisions affecting employees are made within the organization. Role conflict occurs when two or more role pressures exist such that compliance with one role hinders the accomplishment of another [129]. A high prevalence of role-conflict situations in the police officer population has been observed [130]. This is considered due to the nature of their work in combination with institutional requirements and characteristics that may be conflictual in nature, especially when specific roles, obligations, and responsibilities are not clearly defined or explained [131]. As a result, counterproductive work behaviors, emotional exhaustion, and role conflict are observed [2, 132, 133]. In the face of role conflict, research has highlighted the protective potential of self-efficacy and perceived justice in the interpersonal interactions of employees [133]. Research aims In summary, the main theories of psychological health at work postulate that it is the product of a person's ability to adapt to his or her work environment [80] and therefore: (i) subjective processes, more than objective elements, play a central role in assessing a person's resources and abilities and directly predict psychological health [81]; (ii) psychological health depends primarily on the perceived demands of the environment and the perceived resources to cope [83, 134]; and (iii) there is a need to integrate more variables to see how they organize themselves in theoretical models and quantitative analyses [135–137]. For example, recent research highlights the complexity of variables where they may play out as beneficial in some cases and adverse in others [138, 139]. Also, independent analyses or those with a small number of variables are known to risk higher Type I (false positive) and Type II (false negative) error rates, as opposed to conjoint/multivariate analyses where one can appropriately take into account or control for other variables also involved in a dependent variable’s outcome [140, 141]. A large proportion of police mental health studies assess a restrained number of variables or determinants [21, 36, 41, 142, 143]. In contrast, studies that simultaneously account for a large number of variables, also known as high-dimensional multivariate research, to quantitatively model police officer psychological health in the workplace, is arguably sparse yet crucial to theory/model development and testing [4, 19, 30, 144, 145]. Moreover, although these frameworks require larger sample sizes, correlated predictors may be more instrumentally modeled as composing a common factor [146], which can be more closely tied to theory. The present study provides such an approach on a consequent sample of more than 1300 officers. Therefore, the present study sought to address an identified gap in the literature concerning measurement and quantitative modeling, especially with high-dimensional, multivariate data, allowing for data-driven features together with theoretical considerations. First, as opposed to prior restrained analyses with a small set of variables that may lack coverage of variables to represent a constraint/resource at each of the organizational, operational, and intrinsic levels (or factors), this study aimed to do so and provide a more integrative model-based measurement of the relative importance of each variable compared to others, as linked to psychological health. Secondly, while previous literature has identified variables that may be associated with each factor, the relative importance of each variable as driving that factor’s strength of association to psychological health at work is less known, or rarely measured/quantified in the context of a global model (especially for police officers). Thirdly, related to the second aim, building on the foundation of previous theories, which often arise from separate analyses of smaller variable sets, the development of an integrative, multi-factor model was herein sought on the basis of satisfying quantitative goodness of fit measures, a model which quantifies not only the associations between factors but the representativeness of their constituent variables. Crucial to police research, we also aimed to integrate traumatic event exposure variables in the model. Indeed, this work culminates the study from an initial pilot [110] in which previously only a small subset of this data (5 variables) were considered with a focus on PTSD, and with simple analyses. Therefore more generally, the main objective of this work was to identify, in the context of an integrative model, the strongest determinants of psychological health at work (PHW), and their organization, in order to better understand the complex dynamics driving police officer well-being. An empirical modeling approach was undertaken in which the highest performing models (based on fit indices or predictive power of PHW) were selected. Furthermore in this way, confirmation biases are limited. The variables measured along the participants were selected based on the previous literature and models discussed (e.g., JDCS, JD-R, Conservation of Resources), based on their being frequently identified or central. In total, over 25 variables are taken into consideration for the modeling analyses, which include, for example: psychological distress, wellness, and PTSD symptoms, psychological work demands, decision latitude, social support, organizational justice, role conflict, work meaning and recognition, trauma exposure and firearms, years of service and other demographic variables. As noted in the beginning of this subsection, three general hypotheses may be specified: namely (i) variables most directly tied to subjective perceptions will play the most central role in predicting psychological health (i.e. injustice, demands), (ii) psychological health is significantly predicted by the combination of variables regulating constraints (i.e. role conflict, trauma exposure), and perceived resources (i.e. social support at work, recognition) to respond to thereof and one’s capacity to exert control (decision latitude, agency), and iii) accounting for these variables conjointly through a multi-factor, path analysis model can provide a model-based measurement approach that is closer to current theory (e.g., through modeling psychological constructs), controls for many variables appropriately organized, and satisfies the standards of model validity checks. Methods Participants All participants were sworn municipal police officers in France, currently in active service and at least 21 years of age. The participants were recruited by an online call published in the professional police press and by communication to different city halls. Participation was fully anonymous, on a voluntary basis, and the survey was described as aimed at ‘understanding the environment and the working conditions of municipal police officers’, and part of a doctoral thesis. After verification, it was observed that the respondents belonged to a variety of municipal forces evenly spread across the country and its major regions. The survey software recorded only full completions of the questionnaire, and hence did not necessitate removing incomplete data. Procedure The data for this cross-sectional study were collected digitally using a fully anonymized online questionnaire, hosted by a Public Service Management Center. The invitation to participate anonymously was published and disseminated in a specialized press aimed at municipal agents, which announced a study on the mental health of police officers as part of a doctoral thesis project. Although none of the respondents did so, participants had the option to contact the psychologist conducting the study via email. This survey was also shared on discussion forums dedicated to municipal police officers. Officers from various regions across France anonymously responded to this questionnaire. The first part of the questionnaire included standardized questionnaires detailed in the following subsection. The final part of the questionnaire covered socio-demographic information and work-related circumstances. The study and its analysis plan were not pre-registered, the materials detailed comprehensively below are not stored in a public repository. Measures – outcomes composing mental health Well-being and distress are often seen as two opposites on the same continuum of mental health. Similar to [80], we approached well-being and distress as two different components that can influence each other (r = –0.66 in their validation study) but do not completely overlap, and whose variations can account for the ambivalence of psychological feelings at work. Finally, these relatively comprehensive scales (25 and 23 items) address well-being and distress through three different and important aspects: the relationship with oneself (serenity vs. anxiety/depression at work), with others (social harmony vs. irritability/aggressiveness), and with work (engagement vs. disengagement). Psychological Well-being at Work Scale (PWWS) The PWWS [80] was first developed and formally validated in a population of French-speaking North American employees (Canada). It is composed of 25 items that are divided into three subscales: serenity (7 items, α = 0,78), social harmony (8 items, α = 0,78), and engagement (10 items, α = 0,83) in respect to work. All questions are on a 5-point Likert scale from ‘Almost Never’ to ‘Almost Always’, here assigned values from 0 to 4. The total score of the PWWS was herein used for which a Cronbach’s α of 0.91 was observed. Psychological Distress at Work Scale (PDWS) The PDWS [80] is composed of 23 items that are divided into three subscales that respectively measure Irritability/Aggressiveness (6 items, α = 0.83), Anxiety/Depression (9 items, α = 0.92), and Disengagement (7 items, α = 0.92) with respect to work. All questions are on a 5-point Likert scale from ‘Almost Never’ to ‘Almost Always’, here assigned values from 0 to 4. The total score of the PDWS was herein used for which a solid Cronbach’s α of 0.95 was observed. Post-traumatic Stress Disorder Checklist for the DSM‐5 (PCL-5) The PCL-5 [147] aims to measure the presence and severity of PTSD symptoms as outlined in the DSM-5. The validated French adaptation [148] was utilized which consists of 20 items divided into four subscales that respectively measure Re-experiencing (5 items), Avoidance (2 items), Negative alterations in cognition and mood (7 items), and Hyper-arousal (6 items). The respondent rates each item on a scale from 0 (not at all) to 4 (extremely), indicating how much he/she has been bothered by that particular symptom in the past month. In police officers, a total score of 31 has been proposed as efficient for recommending PTSD risk [149]. The total score of the PCL-5 was herein used for which a Cronbach’s α of 0.94 was observed, α’s of 0.91, 0.81, 0.87, and 0.81 were found respectively for the four subscales. Measures – Predictors Job Content Questionnaire (JCQ): psychological demands, decision latitude, social support The JCQ [50, 150] aims to measure the social and psychological characteristics of jobs. The validated French adaptation [151] was utilized, consisting of 26 items divided into three subscales that respectively measure Psychological Demands (9 items), Decision Latitude (9 items), and Social Support (8 items). All questions are on a 4-point Likert scale, where the respondent assesses the validity of the statement about their work environment. The three subscales of the JCQ were herein utilized for which Cronbach’s α values of 0.81, 0.75, and 0.85 were observed (after appropriately reversing the items where necessary). Meaning of Work (MOW) inventory The MOW inventory [152] has emerged as a predominant tool used in occupational psychology research and by professionals. It has been formally validated in French [153]. With about thirty items, it prompts respondents to complete the sentence ‘I do a job…’ based on various aspects such as social usefulness (e.g., '…that serves a purpose'), social contribution (e.g., 'that gives me the opportunity to be of service to others'), work rationality (e.g., 'that has clear objectives'), workload (e.g., 'with a workload adjusted to my abilities'), and cooperation (e.g., 'that is done in a spirit of teamwork'). Participants were required to rate each statement on a 5-point Likert scale, ranging from 1 ('strongly disagree') to 5 ('strongly agree'). The total score of the MOW inventory was herein used, for which a Cronbach’s α of 0.95 was observed. Organizational Justice Scale (OJS) The OJS [154, 155] aims to measure perceived justice in the workplace. The French adaptation [156, 157] was utilized, consisting of 20 items divided into four subscales that respectively measure Distributive (4 items), Procedural (7 items), Interpersonal (4 items), and Informational (5 items) justice. All questions are on a 7-point Likert scale. The total score of the OJS was herein used, for which a Cronbach’s α of 0.93 was observed, α’s of 0.93, 0.79, 0.91, and 0.91 were respectively found for the four subscales. Role Conflict Scale (RCS) To measure role conflict, we used the role conflict subscale items from the role conflict and ambiguity scale of [158], particularly its French adaptation [159]. This subscale consists of 8 items. All questions are on a 5-point Likert scale. A Cronbach’s α of 0.78 was found for the role conflict subscale total score. Traumatic event exposure, distance, and number of traumatic events Three items were used in this study concerning traumatic experience. First, an item asked if the police officer considered themselves to have been exposed to a traumatic event: “Have you been confronted with a traumatic event during your professional activities? (Yes/No)”. If the response was positive, two additional questions were presented: “If yes, which one(s)?” (with a free response space) and “How long ago?” (with a free response space to specify the number of months). If the respondent was unable to fill in these two response spaces, they could technically select "No" to continue with the questionnaire, causing the two additional questions to disappear. Upon reading the event descriptions provided, which were generally very brief (50% of respondents used less than 30 characters, including spaces), a coder checked whether the police officer mentioned a single event or felt the need to refer to multiple traumatic events (using the plural or mentioning different kinds of events). For time since the last traumatic event occurred, hereafter referred to as Trauma Distance, the continuous measure of months was converted into 6 ordinal levels: 0–0.5 years, 0.5–1.5 years, 1.5–3 years, 3–5 years, 5–10 years, and > 10 years. Officers who responded with "No" were assigned to the > 10 years category. Work Recognition Scale (WRS) The WRS was initially developed in French [160]. Twenty-seven items measure perceptions of recognition from three sources: i) superiors (nine items, e.g., 'At my workplace, I perceive that my hierarchical superiors know how to thank me when I am available'); ii) colleagues (eleven items, e.g., 'At my workplace, I perceive that in case of need, my colleagues are available to lend me a hand'), and iii) organizationally (six items, e.g., 'At my workplace, I perceive that there are meetings (ceremonies, gatherings, etc.) set up by the company to value personal achievements'). All questions are on a 5-point Likert scale. The total score of the WRS was herein used, for which a Cronbach’s α of 0.94 was observed, α’s of 0.95, 0.91, and 0.81 were found for the three subscales. Work Information Variables After the psychometric scales, the participant provided the following concrete work information: Years of Experience in the Position (ordinal 4 levels: 0–0.5, 0.5–8, 8–15.5, and 15.5–30) and Total Experience as a police officer (ordinal 4 levels coded likewise), works in a Hierarchical Position (yes/no, coded as "no" for standard police officers and warden, and "yes" for officers with supervisory roles), Carries a Firearm (yes/no), has Used their Firearm (yes/no), Works in a Team/Squad (yes/no), Team Size (continuous converted to ordinal 6 levels: 1, 2–3, 4–6, 7–11, 12–20, > 20), City Population (ordinal 6 levels: < 1 k, 1–10 k, 10–50 k, 50–100 k, 100–300 k, > 300 k), and Region Isolation as kilometer distance to the next city (continuous converted to ordinal 4 levels: 0–6, 6–13, 13–16, 16–30, > 30). Sociodemographic variables At the end of the questionnaire, the participant provided the following sociodemographic information: Age (continuous), Sex (binary), and Education Level (ordinal 3 levels: high school or less, undergraduate diploma, graduate diploma). Data transparency and openness statement The data which contain government employee information are not posted in a public repository due to legal and ethical constraints, however may be requested for research purposes via a data sharing agreement through contacting the corresponding author. The data were analyzed using Python, version 3.12 [161] and the packages scipy, statsmodels, factor-analyzer, and pingouin: versions 1.0 [162], 0.14 [163], 0.5.1 [164, 165], and 0.5 [166] respectively, as well as the lavaan package version 0.6 [167] in R version 4.3.3 [168]. Data analyses and pre-processing The analytical approach here utilized a similar methodology as used in [4]. First, to overview the general sample, descriptive statistics were calculated along the relevant variables and questionnaires. Then statistical tests were considered for the three dependent variables (DVs): Psychological Well-being by the PWWS, Psychological Distress by the PDWS, and PTSD symptoms by the PCL-5, plus a fourth composite variable of the three, termed Psychological Health at Work (PHW). This fourth composite variable was founded on corroborating observations of strong absolute-value correlations between the former three DVs (between 0.52 to 0.80) and grouping along the same factor in exploratory factor analyses. As a result, this composite variable was calculated through the factor weightings obtained through a confirmatory factor analysis in which the PWWS, PDWS, and PCL-5 were modeled to compose the same factor. An overview of the study variables and their Pearson correlation with these four outcome variables is provided in Table 1, in which their p-values were corrected with the Holm-Bonferroni method. Table 1 Correlations between the four dependent variables and the study variables with the Holm-Bonferroni correction (n = 1312) Well-being (PWWS)Distress(PDWS)PTSD Symptoms (PCL-5)PHW (Composite Factor)Distress (PDWS)−0.79***PTSD Symptoms (PCL-5)−0.52***0.63***PHW0.89***−0.94***−0.78***Psych Demands−0.27***0.34***0.37***−0.37***Decision Latitude0.46***−0.41***−0.21***0.42***Social Support0.49***−0.48***−0.30***0.49***Organizational Justice0.51***−0.54***−0.40***0.56***Role Conflict−0.36***0.42***0.38***−0.44***Work Recognition0.53***−0.54***−0.33***0.54***Work Meaning0.67***−0.67***−0.42***0.68***Sex−0.12***0.11***−0.12**Hierarchical0.08*−0.09*0.08*Education Level0.10**−0.08*Works Full TimeAgeHas Firearm0.09*−0.09*0.10**Used Firearm0.07†Years of ServiceExperience in PositionCity PopulationRegion IsolationWorks in TeamTeam SizeTraumatic Event0.09*0.32***−0.16***Num Events Trauma0.09*0.31***−0.15***Trauma Distance0.09*−0.12***−0.33***0.19***† p < 0.10, * p < .05, ** p < .01 , *** p < .001. PHW: Psychological Health at Work, a composite variable of the first three outcome variables (PWWS, PDWS, PCL-5) which were weighted according to the coefficients derived from a confirmatory factor analysis. In order to satisfy normality assumptions, relevant variables were normalized prior to calculating their Pearson correlation; point-biserial correlations were calculated for binary variables (e.g., Sex). Correlations with p-values > 0.10 after the Holm-Bonferroni correction are masked from the table. Table 1 Correlations between the four dependent variables and the study variables with the Holm-Bonferroni correction (n = 1312) Well-being (PWWS)Distress(PDWS)PTSD Symptoms (PCL-5)PHW (Composite Factor)Distress (PDWS)−0.79***PTSD Symptoms (PCL-5)−0.52***0.63***PHW0.89***−0.94***−0.78***Psych Demands−0.27***0.34***0.37***−0.37***Decision Latitude0.46***−0.41***−0.21***0.42***Social Support0.49***−0.48***−0.30***0.49***Organizational Justice0.51***−0.54***−0.40***0.56***Role Conflict−0.36***0.42***0.38***−0.44***Work Recognition0.53***−0.54***−0.33***0.54***Work Meaning0.67***−0.67***−0.42***0.68***Sex−0.12***0.11***−0.12**Hierarchical0.08*−0.09*0.08*Education Level0.10**−0.08*Works Full TimeAgeHas Firearm0.09*−0.09*0.10**Used Firearm0.07†Years of ServiceExperience in PositionCity PopulationRegion IsolationWorks in TeamTeam SizeTraumatic Event0.09*0.32***−0.16***Num Events Trauma0.09*0.31***−0.15***Trauma Distance0.09*−0.12***−0.33***0.19***† p < 0.10, * p < .05, ** p < .01 , *** p < .001. PHW: Psychological Health at Work, a composite variable of the first three outcome variables (PWWS, PDWS, PCL-5) which were weighted according to the coefficients derived from a confirmatory factor analysis. In order to satisfy normality assumptions, relevant variables were normalized prior to calculating their Pearson correlation; point-biserial correlations were calculated for binary variables (e.g., Sex). Correlations with p-values > 0.10 after the Holm-Bonferroni correction are masked from the table. Correlations between the four dependent variables and the study variables with the Holm-Bonferroni correction (n = 1312) † p < 0.10, * p < .05, ** p < .01 , *** p < .001. PHW: Psychological Health at Work, a composite variable of the first three outcome variables (PWWS, PDWS, PCL-5) which were weighted according to the coefficients derived from a confirmatory factor analysis. In order to satisfy normality assumptions, relevant variables were normalized prior to calculating their Pearson correlation; point-biserial correlations were calculated for binary variables (e.g., Sex). Correlations with p-values > 0.10 after the Holm-Bonferroni correction are masked from the table. Next, t-tests, regression, and latent-variable path modeling were conducted. This allowed for an increasing complexity of analyses, each offering a different scale of detail, and in which potential corroborating results, or their reproduction, could be observed. Prior to their realization, in order to satisfy modeling and statistical criteria (e.g., normality, homoscedasticity, linearity for regression), the ordinal and continuous variables were normalized via the Yeo-Johnson transformation [169] and scaled. The two-sample Welch t-tests were performed for the four dependent variables, where the samples were defined by median splits for PWWS, PDWS, and PHW, since no pathology thresholds are currently established in the literature, and for the PCL-5, a cutoff of 31, as established in a previous study on police officers [149]. The p-values were Holm-Bonferroni corrected [170] for multiple comparisons. Based on the similar results between the four DVs in the two-sample t-tests, the regression modeling was performed on the composite variable, PHW, and implemented via the statsmodels package in Python, version 0.14. The predictor variables for the model were determined on the basis of correlations that remained significant after the Holm-Bonferroni correction, which are presented in Table 1. In the regression analysis, up to 8% of the participants were filtered as outliers based on Cook’s Distance [171, 172] values that exceeded three times the mean value. Note that this percentage is coherent with the regression modeling literature where up to 10% of a sample may be acceptably identified, depending also on the characteristics of the data set and the outlier detection statistic used (for example, see [173–175]. Multicollinearity was assessed through the Variance Inflation Factor (VIF) [176] where values of less than 10 were used as the inclusion criterion; this led to the omission of one variable (Traumatic Event) and then the model being refit. Another variable, Work Meaning, was omitted due to it being strongly correlated with other variables and observing a very strong suppressive effect on the other predictors, this is developed upon in the Discussion. Finally, for the resultant model, the necessary assumptions and diagnostics were rigorously evaluated, and these are provided in the Results section. We also note that this resultant model showed nearly equivalent results (set of significant variables, their order in magnitude, R2 values) to an alternative approach we explored, a backwards stepwise regression, which involves recursive elimination of predictors based on a model likelihood criterion, known as the AIC [177, 178], and in which 23 predictor variables were provided to the algorithm. Finally, in order to obtain a more integrative and explanatory model, a structural equation modeling (SEM) framework [179, 180] utilizing the lavaan package in R was performed, and in which all participants were retained. The SEM framework combines a measurement model approach for inferring latent factors (e.g., psychological constructs) from observed variables, and a structural model, or regressions, between latent factors and/or observed variables. This approach therefore also circumvents the limitations of regression that can only model one dependent variable at a time, and for which predictor variables that are highly correlated should not be modeled simultaneously (multicollinearity violation). Two ways of conceptualizing latent factors are possible: reflective, which is the classical approach based on a shared covariance structure and parameterizing error in prediction, and formative. We utilized the reflective approach, as it was in line with our hypotheses and is recommended by an in-depth review of the two approaches, see [181]. Furthermore, as recommended by the results of a large simulation study [182] for its accuracy and suitability for larger data sets, the unweighted least squares (ULS) estimator was utilized, and robust standard errors calculated. The resultant path model with latent factors was obtained through a data-driven approach, that is, the observation of strong correlations for the development of latent factors, the previous t-test and regression results, and maximizing the goodness of fit indices, while also taking into account theoretical considerations. The SEM that appropriately satisfied the standard reference SEM diagnostics [183] (e.g., Comparative Fit Index (CFI), Normed Fit Index (NFI), Tucker-Lewis Index (TLI), Bollen’s Incremental Fit Index (IFI), Root Mean Square Error of Approximation (RMSEA) and Standardized Root Mean Error (SRMR)), with the most parsimonious factor structure that also allowed for the conjoint measurement of a large number of variables, was retained. Variants of this model in which individual variables were tested as belonging to a different factor or pathed differently were verified as leading to poorer goodness of fit indices. Results Sample A total of 1312 police officers fully completed the questionnaire. As for age, 4% of officers were between 18 to 29 years old, 24% were 30 to 39 years old, 41% were 40 to 49 years old and 31% were 50 to 65 years old. The youngest officer was 25 years old. For education, 23% of officers completed high school, 43% undergraduate and 34% graduate or a specialized degree. Seventeen percent of officers were female. For years of service, 12% worked for up to 3 years, 16% between 3–8 years, 38% between 8–15 years, and 34% for more than 15 years. A total of 86% of officers work in a team, namely, 15% worked in a team size of 2–3, 25% 3–6, 20% 6–11, 13% 11–20, and 13% > 20 members. Next, 29.5% of officers reported they had a hierarchical function, and 70.5% reported as not. As for city and municipal characteristics, 1% of officers reported working in a city with < 1 k inhabitants, 32% for 1–10 k, 37% for 10–50 k, 12% for 10–50 k, 14% for 100–300 k, and 4% for > 300 k inhabitants. 22% of officers reported the next city being 0–6 km (km) away, 24% 6–13 km, 31% 13–16 km, 19% 16–30 km, and 4% > 30 km. A proportion of 61% of officers reported carrying a firearm of which only 8% reported having used it. 56% of officers reported having been exposed to a traumatic event, in which 19% reported it being within the last 0–0.5 years, 23% 0.5–1.5 years, 15% 1.5–3 years, 14% 3–5 years, 19% 5–10 years, and 10% > 10 years. Variables associated with police officer psychological health: independent analyses The average (SD) and interquartile ranges for the three dependent variables were as follows: total psychological wellness at work (PWWS) out of 100 possible, 70.74 (13.6, [63 to 80]), psychological distress at work (PDWS) out of 92 possible was 19.90 (16.2, [7.0 to 28.3]) and PTSD symptoms (PCL-5) out of 80 possible was 13.03 (13.8, [2.8 to 19.0]). Using the suggested cut-off of 31 for PTSD in police officers [85], 12.1% satisfy the criteria for the disorder, and employing the more conservative, classical criteria of 37, leads to 8.3%. Next, as explained in the Methods, a composite variable of Psychological Health at Work (PHW) was formed using the factor weights obtained from a confirmatory factor analysis (CFA) where these three outcome variables: PWWS, PDWS, and PCL-5 were modeled to compose the same factor. As shown in Table 1., the absolute correlations observed between these three variables were all between 0.52 to 0.80, and the coefficients derived from the CFA to compose PHW were respectively 0.84, –1, and –0.66. Alongside the former three, this composite variable, PHW, is also analyzed in the t-tests, in the regression (which is limited to one outcome variable), and remodeled in the path analysis in the context of a multi-factor model. Two-sample Welch t-tests were performed on the basis of the three DVs and the composite psychological health variable (PHW), where the groups were defined by a median split for PWWS, PDWS and PHW, and for the PCL-5 variable: by the recommended cutoff of 31. The t-values are provided in Table 2, for which the p-values (Holm-Bonferroni corrected for multiple comparisons) are reflected by asterisks and the Cohen’s d effect size is provided for PHW. The variables in the table are organized based on the most significantly beneficial for PHW (or wellness) in descending magnitude (more present in groups associated with better well-being and less distress). Then the most significantly negative variables (less present in the groups with better PHW results) follow. Both sets are ordered by the PHW Cohen’s d. For brevity, variables where the maximum absolute t-value across the four outcome variables was < 1.90 (far from exhibiting a potential significant difference after Holm-Bonferroni correction) are not included in the table. Table 2Two-sample Welch t-test values on each measured variable on each dependent variableWell-being (PWWS)Distress (PDWS)PTSD Symptoms (PCL-5)PHW (Composite Factor)PHW Cohen's dWork Meaning24.34***−23.92***−13.47***25.09***1.39Work Recognition17.52***−18.14***−9.62***18.32***1.01Organizational Justice17.01***−17.92***−10.98***18.12***1.00Social Support15.49***−15.40***−8.39***15.9***0.88Decision Latitude14.0***−12.61***−5.45***13.39***0.74Trauma Distance2.88*−3.81**−6.92***6.41***0.35Has Firearm2.17−3.03*−1.733.3**0.18Hierarchical3.52**−3.16*−0.932.85*0.16Age0.19−2.06−0.351.450.08City Population−0.41−1.53−1.911.070.06Role Conflict−11.31***12.99***11.07***−14.59***0.81Psych Demands−7.13***9.92***10.88***−11.09***0.61Traumatic Event−1.912.88*7.09***−5.23***0.29Sex−3.86**2.451.9−3.36**0.19Education Level−0.962.74*1.26−3.05*0.17Years of Service−2.081.061.56−1.710.09Used Firearm−0.02−1.312.310.350.02* p < .05, ** p < .01, *** p < .001 after the Holm-Bonferroni correction for multiple comparisons. The two samples for each variable (rows) are based on a median split of the dependent variable (column), except for the PCL-5 in which the recommended cut-off score of 31 was used. Values without asterisks are non-significant. PHW: Psychological Health at Work variable, is a composite of the first three outcome variables (PWWS, PDWS, PCL-5) which were weighted according to the coefficients derived from a confirmatory factor analysis. Positive t-values reflect higher values for the variable in the group associated with more well-being, distress, PTSD symptoms, or health at work (depending on column). Cohen’s d measures effect size for which values of 0.2, 0.5, and 0.8 are respectively associated with small, medium and large effect sizes Table 2Two-sample Welch t-test values on each measured variable on each dependent variableWell-being (PWWS)Distress (PDWS)PTSD Symptoms (PCL-5)PHW (Composite Factor)PHW Cohen's dWork Meaning24.34***−23.92***−13.47***25.09***1.39Work Recognition17.52***−18.14***−9.62***18.32***1.01Organizational Justice17.01***−17.92***−10.98***18.12***1.00Social Support15.49***−15.40***−8.39***15.9***0.88Decision Latitude14.0***−12.61***−5.45***13.39***0.74Trauma Distance2.88*−3.81**−6.92***6.41***0.35Has Firearm2.17−3.03*−1.733.3**0.18Hierarchical3.52**−3.16*−0.932.85*0.16Age0.19−2.06−0.351.450.08City Population−0.41−1.53−1.911.070.06Role Conflict−11.31***12.99***11.07***−14.59***0.81Psych Demands−7.13***9.92***10.88***−11.09***0.61Traumatic Event−1.912.88*7.09***−5.23***0.29Sex−3.86**2.451.9−3.36**0.19Education Level−0.962.74*1.26−3.05*0.17Years of Service−2.081.061.56−1.710.09Used Firearm−0.02−1.312.310.350.02* p < .05, ** p < .01, *** p < .001 after the Holm-Bonferroni correction for multiple comparisons. The two samples for each variable (rows) are based on a median split of the dependent variable (column), except for the PCL-5 in which the recommended cut-off score of 31 was used. Values without asterisks are non-significant. PHW: Psychological Health at Work variable, is a composite of the first three outcome variables (PWWS, PDWS, PCL-5) which were weighted according to the coefficients derived from a confirmatory factor analysis. Positive t-values reflect higher values for the variable in the group associated with more well-being, distress, PTSD symptoms, or health at work (depending on column). Cohen’s d measures effect size for which values of 0.2, 0.5, and 0.8 are respectively associated with small, medium and large effect sizes Two-sample Welch t-test values on each measured variable on each dependent variable * p < .05, ** p < .01, *** p < .001 after the Holm-Bonferroni correction for multiple comparisons. The two samples for each variable (rows) are based on a median split of the dependent variable (column), except for the PCL-5 in which the recommended cut-off score of 31 was used. Values without asterisks are non-significant. PHW: Psychological Health at Work variable, is a composite of the first three outcome variables (PWWS, PDWS, PCL-5) which were weighted according to the coefficients derived from a confirmatory factor analysis. Positive t-values reflect higher values for the variable in the group associated with more well-being, distress, PTSD symptoms, or health at work (depending on column). Cohen’s d measures effect size for which values of 0.2, 0.5, and 0.8 are respectively associated with small, medium and large effect sizes Based on these results, in the order of Cohen’s d effect size for positive t-values, one can observe that the variables most associated with better PHW, in descending order, are Work Meaning, Work Recognition, Organizational Justice, Social Support, Decision Latitude, Trauma Distance, Has Firearm, and occupying a Hierarchical Position. Though more advanced Age and bigger City Populations suggest improved PHW, they were not significant in these analyses. Conversely, the variables associated with less PHW, in descending magnitude, are Role Conflict, Psychological Demands, Trauma Exposure, Sex and Education Level. Though more Years of Service and having Used a Firearm may suggest some worsened mental health variables, they were not significant in these analyses. Education Level was an unexpected result. Correlation analyses showed weak but significant correlations between Education Level and younger Age (r = −0.13, p < 0.001), and female Sex (r = 0.11, p < 0.001), for which the t-tests suggest higher distress and lower well-being respectively for these variables. This variable is noted further in the Discussion. Overall, it is appreciable to note the correspondence in the results between Table 1 and Table 2 despite that t-tests evaluate a potential difference in the central tendency between two sample groups (here for example, defined by low–high well-being participants through a median split), and Pearson correlations evaluate a linear relationship between two variables. Predictors of psychological health at work for police officers: conjoint analysis of variables The previous results are based on many independent comparisons by which trends may be overly simplified and control variables are not accounted for, and furthermore, in which the dependent variables are treated as binary (median-split groups) instead of analyses that make use of their full continuous variance in order to obtain more precise information. Therefore, in this section a multiple regression modeling is performed in order to correct for the aforementioned limitations. Since regression models can assess only one outcome variable at a time, and the three original outcome variables showed highly similar results in the t-tests, we modeled PHW, the composite variable, with the regression approach. The resultant model demonstrated strong indicators of goodness of fit. The significant predictive equation (F(12,1197) = 71.10, p < 0.001) yielded an R2 and adjusted R2 of 0.581 and 0.577 respectively. The diagnostics for appropriate model fit were also satisfied (e.g., see [184, 185]) in which residual normality and homoscedasticity were corroborated by the Kolmogorov–Smirnov and Breusch-Pagan tests respectively (both p > 0.05). The absence of multicollinearity by the observation of all variance inflation factor (VIF) values below 10 (maximum value observed = 3.7 for Work Recognition). Moreover, linearity and the absence of autocorrelation were confirmed by the Rainbow and Ljung-Box tests respectively (both p > 0.05). The results in Table 3 demonstrate the strongest predictors of greater PHW are, in order: Decision Latitude and Organizational Justice, Work Recognition, greater Trauma Distance (time since last traumatic event), Social Support, and Has Firearm. Conversely, the strongest predictors of lower PHW are Psychological Demands, Role Conflict, Sex, and Hierarchical Level. A number of these results mirror those found in the independent t-tests, where some variables bordering significance are no longer in the model (Education Level). Hierarchical Position is a contrary result, which may tie into previous considerations that some variables may be beneficial in some contexts (exerting control, sense of agency) while stressful in others (greater demands and responsibility). We also note that while Work Meaning was omitted from the model due to strong correlations and high suppressive effects with other variables, when included, a very strong standardized coefficient (B surpassing 1.0) is obtained in favor of higher PHW. Table 3Linear Regression results for the prediction of Psychological Health at Work (PHW)BCItpDecision Latitude0.22[0.18, 0.26]10.13 < .001Organizational Justice0.23[0.18, 0.29]8.31 < .001Work Recognition0.14[0.08, 0.21]4.17 < .001Trauma Distance0.11[0.06, 0.17]4.06 < .001Social Support0.09[0.03, 0.15]2.980.003Has Firearm0.04[0.01, 0.08]2.340.02Psych Demands−0.16[−0.20, −0.12]−7.72 < .001Role Conflict−0.12[−0.15, −0.05]−4.28 < .001Sex−0.07[−0.10, −0.02]−3.200.001Hierarchical−0.04[−0.08, −0.01]−2.180.03R2 = 0.581, adj. R2 = 0.577 and F(12,1197) = 71.10, p < .001. In the interest of brevity, variables with p-values above 0.05 in the model are not presented above in the table. These are: Education (B = −0.03, p = .08) and Num Events Trauma (B = −0.04, p = 0.13) Table 3Linear Regression results for the prediction of Psychological Health at Work (PHW)BCItpDecision Latitude0.22[0.18, 0.26]10.13 < .001Organizational Justice0.23[0.18, 0.29]8.31 < .001Work Recognition0.14[0.08, 0.21]4.17 < .001Trauma Distance0.11[0.06, 0.17]4.06 < .001Social Support0.09[0.03, 0.15]2.980.003Has Firearm0.04[0.01, 0.08]2.340.02Psych Demands−0.16[−0.20, −0.12]−7.72 < .001Role Conflict−0.12[−0.15, −0.05]−4.28 < .001Sex−0.07[−0.10, −0.02]−3.200.001Hierarchical−0.04[−0.08, −0.01]−2.180.03R2 = 0.581, adj. R2 = 0.577 and F(12,1197) = 71.10, p < .001. In the interest of brevity, variables with p-values above 0.05 in the model are not presented above in the table. These are: Education (B = −0.03, p = .08) and Num Events Trauma (B = −0.04, p = 0.13) Linear Regression results for the prediction of Psychological Health at Work (PHW) R2 = 0.581, adj. R2 = 0.577 and F(12,1197) = 71.10, p < .001. In the interest of brevity, variables with p-values above 0.05 in the model are not presented above in the table. These are: Education (B = −0.03, p = .08) and Num Events Trauma (B = −0.04, p = 0.13) Factor and mediation analysis of police officer psychological health Classical regression models have some limitations in that they cannot model paths and mediating relationships between variables (for example, model a predictor variable that is predicted by another variable), nor group correlated variables into (latent) factors for inference and measurement. As presented in the Introduction, research has identified that the trends that influence police officer health are complex, and so regression models alone are probably too simplistic as the model of choice for an integrative, empirical model that more realistically structures the trends and major factors associated with PHW of officers, and the quantification of said structure. Consequently, this section presents the path analysis and latent factor modeling through the Structural Equation Model (SEM) framework that was carried out in a data-driven manner, while also taking into account theoretical considerations. First, the validity of the SEM was verified, adhering to the standard goodness of fit indices and their benchmarks. These results are provided in Table 4. As shown in the first row, the resultant model satisfied a large majority of the indices that support an appropriate fit (9 out of 10). Table 4Goodness of fit indices for the structural equation model (SEM)IndicesModel χ2/dfCFIaIFIbNFIcTLIdRMSEeRMSEA 90% CIfSRMRgGFIhAGFIiObserved Value6.410.960.960.950.950.065[0.06, 0.07]0.0590.980.97Acceptable Threshold < 5 > 0.90 > 0.90 > 0.90 > 0.90 < 0.10[0.00; close to RMSEA] < 0.08 > 0.90 > 0.90aComparative Fit Index; bIncremental Fit Index; c Normed Fit Index; d Tucker-Lewis Fit Index (also known as Non-Normed Fit Index); e Root Mean Square Error of Approximation; f RMSEA Confidence Interval; g Standardized Root Mean Residual; h Goodness of Fit Index; i Adjusted Goodness of Fit Index Table 4Goodness of fit indices for the structural equation model (SEM)IndicesModel χ2/dfCFIaIFIbNFIcTLIdRMSEeRMSEA 90% CIfSRMRgGFIhAGFIiObserved Value6.410.960.960.950.950.065[0.06, 0.07]0.0590.980.97Acceptable Threshold < 5 > 0.90 > 0.90 > 0.90 > 0.90 < 0.10[0.00; close to RMSEA] < 0.08 > 0.90 > 0.90aComparative Fit Index; bIncremental Fit Index; c Normed Fit Index; d Tucker-Lewis Fit Index (also known as Non-Normed Fit Index); e Root Mean Square Error of Approximation; f RMSEA Confidence Interval; g Standardized Root Mean Residual; h Goodness of Fit Index; i Adjusted Goodness of Fit Index Goodness of fit indices for the structural equation model (SEM) aComparative Fit Index; bIncremental Fit Index; c Normed Fit Index; d Tucker-Lewis Fit Index (also known as Non-Normed Fit Index); e Root Mean Square Error of Approximation; f RMSEA Confidence Interval; g Standardized Root Mean Residual; h Goodness of Fit Index; i Adjusted Goodness of Fit Index The resultant SEM, as shown in Fig. 1, yielded six main latent factors: Psychological Health at Work, Organizational Constraints, Trauma Exposure, Perceived Resources, Sense of Agency, and Esteem. Note that the SEM framework combines two modeling approaches: measurement models for each latent factor by observed (also known as manifest) variables, and a predictive modeling structure, or regressions, between factors and/or observed variables. For the measurement models, one of the observed variables (rectangular nodes) that composes each factor (circular nodes), which is often among the variables most associated with said construct, serves as the baseline and its coefficient is fixed at either 1 or –1 (the sign determines the direction of interpretation for the factor). Therefore, the coefficients of the other observed variables composing the factor are interpreted based on their relative magnitude. The coefficients of each observed variable that receive a path from a factor represent the variable's association to the factor, which are influenced also by the latent factor's potential association (e.g., regression models) to other factors or manifest variables. This brings to note the regression structure in SEMs, in which either latent factors or observed variables may be modeled to predict others, represented by outgoing paths from the predictor (e.g., the four factors that predict PHW). By modeling convention, these coefficients are standardized around 0 (rather than 1). In regard to the PHW variable previously analyzed, when deriving this factor in the context of an integral model, accounting for all other variables, the coefficients here obtained are 0.93, –1, and –0.72, which are very comparable to the PHW previously derived (from an isolated CFA) used in the correlation, t-tests, and regression analyses, which possessed coefficients of 0.84, –1, and –0.66 respectively.Fig. 1Model of Psychological Health at WorkNote. Structural equation model (SEM) derived from a data-driven approach, with also theoretical considerations, that allows for a measurement modeling of six latent factors through observed (manifest variables), as well as predictive (or regression, structural) modeling between factors and/or observed variables, notably for the prediction of latent factor, Psychological Health at Work (PHW). The measurement model of PHW is reflective, consisting of three manifest variables: Psychological Wellness at Work (PWWS), Psychological Distress at Work (PDWS), and PTSD Symptoms (PCL-5). Latent modeled factors are depicted as circles, and observed variables as rectangles. The black arrows represent the measurement models of latent factors in which coefficient values are compared to a reference value of 1 or –1, and the dark blue arrows represent the predictive modeling in which coefficients are standardized around 0. Significance for non-reference values is represented as *** p < 0.001, ** p < 0.01, * p < 0.05. The goodness of fit indices are available in Table 4. Several mediations and six latent factors are modeled to significantly predict work wellness Model of Psychological Health at Work Note. Structural equation model (SEM) derived from a data-driven approach, with also theoretical considerations, that allows for a measurement modeling of six latent factors through observed (manifest variables), as well as predictive (or regression, structural) modeling between factors and/or observed variables, notably for the prediction of latent factor, Psychological Health at Work (PHW). The measurement model of PHW is reflective, consisting of three manifest variables: Psychological Wellness at Work (PWWS), Psychological Distress at Work (PDWS), and PTSD Symptoms (PCL-5). Latent modeled factors are depicted as circles, and observed variables as rectangles. The black arrows represent the measurement models of latent factors in which coefficient values are compared to a reference value of 1 or –1, and the dark blue arrows represent the predictive modeling in which coefficients are standardized around 0. Significance for non-reference values is represented as *** p < 0.001, ** p < 0.01, * p < 0.05. The goodness of fit indices are available in Table 4. Several mediations and six latent factors are modeled to significantly predict work wellness Beginning with Organizational Constraints: these were found to predict lower PHW with a coefficient of –0.32. Role Conflict, which was found the most associated, was modeled as the reference variable with a coefficient of 1 and was found 1.5 times stronger in the measurement model of Constraints than Psychological Demands, which resulted in a coefficient of 0.66. Next variables linked to Trauma Exposure: were half in magnitude than constraints in the prediction of lower PHW, with a coefficient of –0.14. Trauma Distance, which was found the most associated with this factor, was modeled as the reference variable with a coefficient of –1, and was followed closely by the Number of Traumatic Events, which resulted in a coefficient of 0.78. Based on the prediction model, the number of traumatic events experienced appears to increase with years of service yet be lower when working in cities with higher populations. Having Used a Firearm was also significant in the measurement model of traumatic event exposure, with a coefficient of 0.12. One’s Perceived Resources: may be useful in overcoming the aforementioned negative variables due to its coefficient of 0.31, and is similar in benefit to Sense of Agency. These resources appear similarly derived in strength from one’s Social Support, Esteem of Self or one’s Work, and Organizational Justice, with Esteem being the most associated. Esteem was hence modeled as the reference variable with a coefficient of 1, and social support and organizational justice resulted in coefficients of 0.81 and 0.85 respectively. Here, one’s said Esteem appeared most driven by their Meaning of Work, which was modeled as the reference variable with a coefficient of 1, and secondly, Work Recognition, which resulted in a coefficient of 0.92. Next, based on the prediction model, Women working in the force appear to perceive less Organizational Justice, though weakly with a coefficient of −0.06. Finally, Sense of Agency: is comparable in magnitude as Perceived Resources in predicting higher PHW, with a coefficient of 0.30. In the Introduction, as well as an example found in the regression results (Hierarchical Position), it was noted that some variables may play out as resources in some contexts or be associated with stress or constraints in others (e.g., having responsibility and decision-making power over oneself and others, too much autonomy, carrying a firearm and therefore being assigned to more dire situations where one may have to use it). This factor may more easily take into account such variables that are instrumental in exerting control but are more complex. With the variables at our disposal, Sense of Agency appeared most predominantly associated with one’s sense of Decision Latitude, which was modeled as the reference variable with a coefficient of 1, and was found about three times higher in magnitude as whether one works in a Hierarchical Position or not (coefficient of 0.35). Being authorized to carry a Firearm at work was also significant in the measurement model of Sense of Agency, as associated with a higher values of the factor, but weakly with a coefficient of 0.12. Discussion Due to stressful working conditions, police officers are a population at high risk for mental health issues, which in turn, can compromise their own safety, that of their colleagues, and civilians. In a realistic paradigm, these conditions, and other determinants of their psychological health, can be linked to a multitude of variables that interplay altogether. However, current literature that simultaneously models–quantitatively from observed data–such a multitude of variables is sparse. This study took upon this objective to further theoretical and applied understandings through a measurement framework on empirical data, and allow the data to drive some features of model development, such as variable groupings into latent factors (e.g., psychological constructs), paths between factors, and the possibility to quantify the relative strength of each variable’s association to officer psychological health (or other constructs), when many other variables are accounted for simultaneously, which is closer to realistic settings. Allowing for some freedom for the expression of the data, several general hypotheses were outlined: namely (i) variables most directly tied to subjective perceptions will play the most central role in predicting psychological health (e.g., injustice, esteem for one’s work), (ii) psychological health is significantly predicted by the combination of variables regulating constraints (e.g., demands, trauma) with those related to perceived resources to respond (e.g., support) and one’s capacity to exert control (decision latitude, hierarchical position, ability to defend oneself with a weapon), and iii) a viable measurement model that holistically quantifies theoretical components can be obtained through the conjoint analysis of variables into a multi-factor, path analysis model design, which satisfies model validity checks of reference. With these objectives, this large-scale study took into account more than 25 variables measured among more than 1300 police officers, where the principal psychological construct to model was Psychological Health at Work (PHW), quantified by three facets: psychological wellness, distress, and PTSD symptoms. It was sought to identify the most pertinent determinants of PHW for police officers, in a model that infers also other psychological constructs and their relation to PHW, and thus advance current theoretical understandings in how measured (manifest) variables and inferred constructs play out or are organized into an integrative model; and furthermore, quantify their relative strength of association to PHW in a high-dimension, multivariate context. Statistical analyses and modeling were carried out from the most simple to the complex, finishing with a six-factor model that fully integrated the identified significant manifest variables, and which corresponds to current theories of PHW. The t-tests (and correlation analyses) identified determinants of the dependent variables (DVs) of psychological well-being, distress, and PTSD symptoms, which in most cases, were similar along the three. The key protective variables shared among the three were Work Meaning, Recognition, Organizational Justice, Social Support, Decision Latitude, and greater Trauma Distance (time since the last traumatic event experienced). The risk variables shared among the three were Role Conflict, Psychological Demands, and Trauma Exposure. Some distinctions elucidated by these analyses included the authorization to carry a Firearm at work as associated with lowered distress rather than the other two DVs, occupying a Hierarchical Position associated with higher psychological well-being and lowered distress rather than any difference in PTSD symptoms, female Sex associated with lowered psychological well-being, and higher Education Level associated with a slightly higher distress. This final result with Education Level as a risk factor was unexpected. Looking into it, further analyses suggest Education Level as a mediating variable for the negative effects of younger Age and female Sex: namely, correlation analyses showed weak but significant correlations between higher Education Level and younger Age (r = −0.13, p < 0.001), and higher Education Level and female Sex (r = 0.11, p < 0.001), for which the t-tests also suggest higher distress and lower well-being for both Age and female Sex. As in the t-tests of Table 2, corroborated by the correlations also in Table 1., the three DVs psychological well-being, distress, and PTSD symptoms shared most of the same significant associated variables. Furthermore, these DVs were highly correlated (absolute r values between 0.52 to 0.80, high factor loadings in the same factor). Based on these observations, a composite variable of the three was derived through Confirmatory Factor Analyses (CFA) in order to provide analyses on an overall dependent variable of PHW. This variable, was analyzed in a multiple regression model in a data driven manner. The model, resulting in the main significant predictors of PHW, corroborated the results of the t-tests: the majority of variables identified as significant in the independent analyses, were also the case in this conjoint analysis, with the exception of Education Level and Num Events Trauma. The regression analysis provides additional granularity, quantifying the strength of association of each variable, relative to others, in a predictive context of PHW. Note that Work Meaning was exceptionally strong in the t-tests for higher PHW (Cohen’s d = 1.39), and showed a dominating role or strong suppression of other predictor variables, which posed a challenge to include it in the regression model. One may speculate that this occurrence is compatible with a hypothesis of a bidirectional link with PHW. That is, while Work Meaning may be a protective factor against stress, conversely, high levels of stress, distress, or traumatic experience can disrupt or lead to a strong loss of meaning [186, 187], and in this case, show a trend of comorbidity with low levels of PHW–making the variable difficult to disentangle–yet all the more pertinent for occupational research on highly stressful careers. Altogether, both the t-tests and regression analyses could be considered too simplistic or ‘linear’, and a more integrative model that allows for paths, mediations, and factor analyses could provide for a more sophisticated framework to better clarify the derivation of PHW for police officers. A path analysis with latent factors was conducted to further advance theoretical understandings of PHW in an empirical-modeling approach, using the Structural Equation Modeling (SEM) framework. As in Fig. 1, the data gave rise to a model composed of five principal factors and one nested factor (Esteem): Psychological Health at Work, Organizational Constraints, Trauma Exposure, Perceived Resources, Sense of Agency, and Esteem. The variables and their weights compared to others within the same factor, are explained in a detailed manner in the Results section. Instead of each manifest variable’s direct association with PHW, its association through a latent factor (or psychological construct), and in some cases, a mediating path is considered, in its link to PHW. For example, an officer’s Years of Service significantly predicts a greater likelihood of being exposed to more Traumatic Events, playing a role in a more global construct of lived trauma, which in-turn is associated with lower PHW. Conversely, officers working in more Highly Populated Cities seem to have a lower incidence of Traumatic Events. Also related to trauma, one observes that having the authorization to carry a Firearm at work is significantly tied to one’s Sense of Agency, while having Used a Firearm is significantly tied to Traumatic Exposure. Thirdly, female Sex is significantly associated with lower perceived Organizational Justice and thus lower perceived Resources, which in-turn is associated with lower PHW. Furthermore, an exploratory modification of the SEM was examined to better understand the unexpected link of lower Education Level on PHW in the t-tests. It was found that Education Level could significantly predict female Sex (β = 0.11, p < 0.001) which in turn predicted lower perceived Organizational Justice and thus lower PHW. Further studies would be useful to clarify if Education Level is a risk factor only when linked to lower Age and female Sex, or if it negatively interacts with the socio-cultural qualities of police bureaus or is simply directly associated with a heightened perception of demands yet lower perception of resources. However, the results confirm the central role of subjective perceptions in predicting psychological health at work (hypothesis i). They also suggest that objective factors (e.g. years of service, use or carrying of a weapon, status and gender) contribute to broader perceptions (respectively: Trauma Exposure, Sense of Agency, Perceived Organizational Resources). The resultant data-derived model is compatible and contributes to theoretical understandings such as those considered in the frameworks of the Job Demands Control (JDC) model by [188], the Job Demands-Resources (JDR) model [85, 86] and the Demands Resources and Individual Effects (DRIVE) model [87, 88]. To our knowledge, the present work is one of the few to model integratively as many variables related to constraints and resources, and elucidate their respective weight in these psychological constructs, and observe the predictive value of these factors that are altogether hypothesized to determine PHW. Therefore, previous theoretical models suppose links by arrows that link nodes, but typically do not quantify the respective contribution that each node brings to an outcome (or its strength of association), nor do they typically quantify, in a global model of multiple constructs, the respective weights of variables that are supposed to belong to the same factor. Some variables, albeit significant, may be very low in weight when accounted for in a global model (for example Fig. 1, City Population and sex Female effects respectively at standardized β = –0.08 and –0.06, p < 0.001 for each). Furthermore, beyond a simplistic two-dimensional structure of Demands and Resources (or Control), our model emphasizes the dynamics of past professional experiences, or Trauma Exposure, which are highly relevant to police officers, as well as Sense of Agency (which can be considered a distinction between Control and Resources), and the Esteem one has for his/her self and his/her work. These results thus confirmed that police officers have to regulate both constraints and perceived resources (hypothesis ii). They also show the existence of constraints and resources both at the level of the organizational environment (unsuitable work organization and social resources enabling them to value their jobs) and at an individual level (past traumatic experiences but individual power to act thanks to a strong sense of agency). Finally, the analyses confirm the value of taking into account many of the variables, simultaneously, to see how they organize themselves to determine psychological health at work (hypothesis iii). One of the likely challenges of future research is to identify constraints and resources at both the individual and interpersonal levels, as noted in [189], in order to reveal how constraints and resources compensate for each other and how individual and collective regulations can provide mutual support in the event of imbalance in one of the two fields. It is also pertinent to note that this model that emerged from the data has the advantage of considering police officers' adverse past experiences in relation to their sometimes reduced sense of agency, which can lead them to a state of learned helplessness and pessimism about their ability to act (in line also with the PERMA-V model, see [190]). This Agency factor is consistent with foundational theories such as Bandura’s [191] accounts of human agency in social cognitive theory and Seligman’s [192] learned helplessness, notably their links to well-being, victimization [193] and psychopathology [194]. It is well established that severely impaired well-being can affect self-regulation in numerous ways [195, 196]. Here, it is likely that pessimism and the feeling of learned helplessness in the face of events can cause police officers to have difficulty perceiving available resources, both in terms of their own resources and those available in their work environment, which is also compatible with Hobfoll’s [197] Conservation of Resources Theory, highlighting that a loss of resources can lead to not being able to exploit other resources. However, it cannot be ruled out that impaired well-being could also lead them to more frequently notice every objective constraint they face; in doing so, the person may not be able to overcome the difficulties encountered. Considering these observations, it can be considered promising to support police officers, especially those facing professional difficulties, severely impaired well-being, or trauma, in managing their constraints. This support could be particularly beneficial in restoring their self-esteem and seizing opportunities in their professional environment [198, 199]. In the face of encountered constraints, helping officers to develop a sense of personal control predicts increased life satisfaction [200–202]. Limitations and suggestions for future work Several limitations of the current study and analyses are important to note. First, for certain questionnaires this study only utilized the total scores. Analyzing subscales independently may lead to more insights into PHW and its determinants. Furthermore, while over 25 variables were taken into account herein, one could consider including other variables noted in the Introduction (e.g., irregular work shifts, insufficient material resources, officer skin color, coping strategies, support from family and whether other family members are also in policing). Including more variables could serve different objectives. For example in the integrative model, each factor (or construct) is estimated by 2 to 5 variables. First, it is preferred that each latent factor consist of at least three variables. Moreover, additional variables may help in more comprehensively quantifying the factor, and measuring the most relevant variables that represent said factor and its influence or association to the others in the structural model. Or, certain variables may yield new factors that can be taken into account in an expanded theoretical model. However, the addition of variables poses a significant challenge for obtaining sufficient observation numbers (recruiting many respondents) and maintaining reasonable questionnaire length. Though, with greater participant numbers and variables, more robust model selection and validation analyses (cross-validation, train/test data sets) may also be envisioned. Another limitation, difficult to overcome in police samples, is the imbalanced gender distribution (17% female of 1312 officers). In the analyses, the female Sex did appear significantly associated with lower Well-being (t-tests), lower PHW (t-tests and regression), and lower perceived Organizational Justice/Resources (SEM). Greater statistical strength in elucidating the impact of gender could be obtained with greater observations. Future work could consider tripling the sample size and then subsampling to obtain similar gender distributions. Higher Education Level being correlated with female Sex (and lower Age) should also be further examined in future work. Clustering analyses and their interactions may also be interesting, for example variables on Team Size, Works in Team, and Region were not found significant in the different analyses, but they may bring about insights when considering sub ensembles of police officers or additional variables that measure for positive or negative team dynamics. Due to the constraints of realizable computational complexity, model moderation (or interactive effects) and bidirectional links were not assessed. For example, one could consider that an even more realistic model than in Fig. 1, would possess such links, as in Fig. 2 below. Future work could consider interaction effects at perhaps a smaller scale (e.g., in the regression analyses). Finally, it is also important to note that, as with most non-longitudinal studies on police officers, the current methodology does not allow for establishing a causal relationship between factors, but rather an associative relationship. Therefore, future longitudinal studies could evaluate possible causal relationships between different factors.Fig. 2A More Complex Model of Psychological Health at Work. Note: A theoretical proposition of the data-derived model in Fig. 1, which consists of both bidirectional and moderation links, that is perhaps more realistic but too complex to estimate given current computational and data limits A More Complex Model of Psychological Health at Work. Note: A theoretical proposition of the data-derived model in Fig. 1, which consists of both bidirectional and moderation links, that is perhaps more realistic but too complex to estimate given current computational and data limits Finally, while Sense of Agency could be modeled as intertwined with regular Perceived Resources, it may be worthwhile to disentangle it, as well as other classes of variables/strategies to respond to demands and stressors. As previously described, research has noted how certain variables present a level of complexity in which they may play out as resources in some contexts or be associated with stress or constraints in others [138, 139]. In terms of Agency, for example, having responsibility, decision-making power over oneself and others, or very high autonomy (little direction), may be empowering but also stress-inducing depending on the situation. This stress may also be amplified when the officer must make decisions in strict compliance with departmental protocol and objectives. Furthermore for officers, being authorized to carry a firearm may relate to a status symbol in the force and a sense of security, but may also mean a higher probability of being assigned to more dangerous interventions. Indeed our analyses showed that using one’s firearm is well-associated with lower psychological health, and in the regression analysis, Hierarchical Position was negatively associated with psychological health. In contrast, it is more difficult to imagine common, repeated situations in which higher values of the variables grouped in Resources (e.g. Social Support, Organizational Justice, Recognition/Meaning) could drive lower PHW for a typical officer. Here Sense of Agency was disentangled from Resources for these reasons, and future work that involves more or different variables may consider other factors and how they may be optimally organized (nested, mediated, or separated) within the model. ​​For Bakker and Demerouti [85], resources and constraints seem to be defined a priori. But according to the transactional model [81, 83] resources and constraints only exist if they are perceived as such by individuals. Thus, some resources may be more related to the perception of the environment (e.g. “perceived resources”) and others more related to the ability to autonomously generate stressors (e.g. “sense of agency”). Similarly, some constraints may stem from the perception of the environment (e.g. “organizational constraints”), while others are more linked to personal history (“trauma exposure”). Some JDR studies mention the existence of resources that can be either organizational or more personal (such as employees' energy and commitment levels), but more rarely personal constraints. This distinction adds value in relation to JDR, by distinguishing between two types of resources, but also two types of constraints. For all that, the opposition between what belongs to the environment and to individual experience is complex to conceptualize, and more work is needed to clarify it conceptually. As things stand, certain environment-related predictors may be common to several professions, and are well known in organizational psychology. Predictors linked to the experience and operational management of stress and trauma, on the other hand, appear to be more specific. Conclusions and recommendations From a theoretical standpoint, this study provides a number of complementary elements that integrate and partly enrich classic theories of police occupational health, such as:the distinction between different types of constraints and different types of resources;the paradoxical nature of certain elements as resources and constraints;the existence of strong links with psychological health, suggesting the possibility of bidirectional links; andthe potential for complex interactions between these variables that could impact police officers' occupational health. the distinction between different types of constraints and different types of resources; the distinction between different types of constraints and different types of resources; the paradoxical nature of certain elements as resources and constraints; the paradoxical nature of certain elements as resources and constraints; the existence of strong links with psychological health, suggesting the possibility of bidirectional links; and the existence of strong links with psychological health, suggesting the possibility of bidirectional links; and the potential for complex interactions between these variables that could impact police officers' occupational health. the potential for complex interactions between these variables that could impact police officers' occupational health. Due to their highly demanding work conditions police officers are at high risk for psychopathology, lower well-being, and increased job turnover. On a more practical level, according to our research, providing police officers with social support at work, recognition, positive perception of work, sense of fair proceedings and pay (organizational justice, especially for female and young officers), decision-making power (decision latitude), and decreased conflictual information and procedures (role conflict) are of utmost importance. Additional attention to officers with higher years of service, those working in lower population cities, and those who have recently used their firearm, especially with trauma counseling, can mitigate psychological health consequences. Ideally, the degree of psychological demands of police officers should be regularly assessed and reduced if possible. Reminders of support and integration in the force to officers who may feel marginalized due to perceiving lower decision-making power, such as through working a subordinate role, and/or having fewer authorizations than their colleagues may be useful in their facing work challenges. Future integrative modeling research, with larger scales of data and participants, may be crucial to better understanding the relative contribution of each variable and their interplay in realistic settings, providing a framework for measurement that is more tightly-related to theories of occupational health, and enabling their advancement in novel ways.