Correction workers work under psychologically and physically demanding conditions [1
]. They are frequently exposed to unexpected hold-over shift work that may result in circadian disruption, sleep loss, and increased fatigue. Also, correction workers experience constant threat of inmate assaults and riots, which can be emotionally draining due to hyper-alertness and anxiety [4
]. Moreover, working conditions in the correctional setting are known to be associated with numerous types of health and performance outcomes, including immediate outcomes, like exhaustion and psychological distress, as well as more distant outcomes, like job dissatisfaction and impaired work ability [5
]. According to the U.S. Department of Justice, correction workers, including correctional officers and administrative and support staff, are at higher risk of suicide, substance abuse, and divorce, while their mortality rate is the second highest of any occupation [6
]. High turnover rates have also been reported [7
The negative impact of demanding correctional work on correction workers’ wellbeing can be explained by the classic stress-burnout model [8
]. Adverse working conditions of correctional work act as stressors causing stress. These stressors lead to psychophysical strain symptoms, such as various forms of psychophysical burnout. Echoing this, empirical studies on correction workers have shown significant relationships between overwork and stress [8
]. Also, studies have revealed that correction workers’ stress is associated with burnout [8
] and somatic symptoms [7
These phenomena can also be understood within theoretical frameworks of the Job Demands-Resources model [12
]. Exposure to the physically and psychologically demanding correctional working conditions with a lack of adequate resources is likely to deplete correction workers’ psychological resources for socially adaptive and physically healthy self-regulatory behaviors [13
]. Thus, correction workers’ demanding work conditions can lead to exhaustion, which may then negatively influence healthy behaviors as well as socially adaptive behaviors. Additionally, Wright and Cropanzano showed that emotional aspects of exhaustion, which can be manifested in different forms, such as depression and stress, can undermine job performance [14
In general, the negative impacts of demanding work conditions and subsequent psychophysical strain can be both extensive and derivative. Recent studies have shown various complications of stress. For example, stress can contribute to increased risk of cardiovascular diseases [15
], while burnout, as an outcome of extended exposure to stress, can be associated with impairment of psychological resources, such as emotional intelligence and self-efficacy, as well as loss of social support [16
]. Moreover, it was shown that burnout can negatively affect well-being at work, resources for coping, ability to work, and engagement [17
]. Considering these findings, it can be inferred that specific adverse effects on workers’ health in general (e.g., illnesses, psychological problems, obesity, etc.) are further compounded by compromised family health (e.g., marriage dysfunction, problematic children’s behavior) and employer dissatisfaction (e.g., due to declined performance, increased health care costs). In corrections, work-related stress can lead to increased absentee rates, internal conflict, and suboptimal employee performance [18
], all which may have reciprocal adverse influences on the work environment.
Our study attempts to elaborate the Job Demand-Resource model, which primarily focuses on stressors’ impact on stress outcomes (strains), by examining the interrelations among the potential stress outcomes that are typical for correction workers. Our view is that the various stress outcomes do not simply occur in parallel, but instead are able to have synergistic effects when they occur in combination, in line with the view that stressor-strain relationships can be reciprocal [19
]. Chronic exposure to the occupational context of correctional facilities, with its high demand that also lacks control or support, is likely to lead to various psychosocial and behavioral stress outcomes. In fact, Tennant found that enduring structural occupational stress can contribute to psychological disorders, like depression [21
]. When confronted with different types of strain symptoms, one’s resources, which are always limited in the case of correction workers, need to be used to cope with the extant symptoms first, whilst one’s resources may not be adequately restored in a timely manner in order to respond to the emergence of additional strain symptoms. Subsequently, one’s strain symptoms are likely to be exacerbated as more problems arise and as resources become more and more depleted.
The process of resource depletion is particularly concerning in situations in which strain symptoms are not simply end products but operate as derivative or secondary demands (or stressors). As various strain symptoms negatively affect a person in different ways when they are either within or outside the workplace, this suggests a person needs resources, including increased control and support, both within and outside the workplace to handle multiple strain symptoms. This inference aligns with the perspectives of Spillover Theory [22
], as well as Conservation of Resources Theory [23
]. Furthermore, when there is a lack of adequate resources, the stress-burnout process would accelerate because of the synergistic impact of multiple strain symptoms on resource depletion.
Our study explores the possible derivative or secondary role of strain symptoms as additional stressors. By doing so, the domains of stress intervention are also broadened from focusing exclusively on traditional stressors to also include strain symptoms themselves. This approach is not unlike secondary (i.e., reducing the impact of an injury and disease that has already occurred) or even tertiary (i.e., mitigating the impact of an ongoing injury and illness which has lasting effect) intervention approaches found in medical science and public health [24
]. The dire implications of doing nothing to prevent any worsening or exacerbation of the health situation for correction workers was also a motivating factor.
1.1. Socio-Technical Systems Framework and Total Worker Health® Paradigm
The present study is based on a socio-technical systems framework which recognizes the interactions between behavior and the design of work system components and their potential to promote workers’ health and safety [25
]. Specifically, key principles of joint causation within the socio-technical systems framework [26
] provide the rationale for examining the interplay among the organizational (i.e., workability, disengagement), social (i.e., work-family balance), psychological (i.e., stress, exhaustion, depression), and behavioral strain symptoms (i.e., limited physical exercises) in relation to the unique occupational context of correctional work. Understanding the interrelatedness of these various attributes can help in efforts to achieve compatibility of the work system’s elements and goals in order to promote better organizational performance as well as employee safety, health, and wellbeing [27
Socio-technical systems approaches emphasize contextual factors whenever relationships are examined among a specific set of variables used to represent the inherent complexity of the workplace. In this regard, the socio-technical systems approach represents a holistic approach that is consistent with the Total Worker Health®
framework being advanced by the U.S. National Institute for Occupational Safety and Health (NIOSH) [29
]. NIOSH defines Total Worker Health as “policies, programs, and practices that integrate protection from work-related safety and health hazards with promotion of injury and illness prevention efforts to advance worker well-being” [30
]. Researchers and practitioners seeking to advance the Total Worker Health agenda are moving beyond conventional health protection and health promotion approaches by undertaking more comprehensive assessments of the health, safety, and wellbeing of workers, and also to design and test more integrated workplace interventions [31
According to the socio-technical systems framework, how correction workers deal with the unusually demanding occupational conditions in corrections will be a key determinant of how this adversely impacts them. For example, working conditions in corrections may make it harder for correction workers to balance work and personal life. Subsequently, unique patterns of negative spillover effects [32
] can be anticipated to impact both the psychosocial and behavioral states of correction workers, such that a disrupted work-family balance may exacerbate one’s depression, stress, healthy eating, regular exercises, engagement to work, and workability. Meanwhile, the Total Worker Health paradigm speaks to the importance of integrated efforts both within (e.g., management commitment, working environment improvement) and outside the workplace (e.g., support from home and community) in order to effectively manage the adverse impact of the stressful work environment of corrections.
1.2. Analytic Approach: Bayesian Network Analysis
Bayesian Network analysis examines the relationships among variables based on their joint probability [33
]. It utilizes machine learning algorithms that can efficiently cope with the uncertainty and complexity of component interactions within a system as a whole [34
]. Bayesian Network analysis offers a network diagram (directed acyclic graph), which consists of nodes and arrows. The nodes are random variables that may consist of observed continuous or categorical quantities, or even latent variables. The arrows (i.e., edges or arcs) indicate probabilistic relationships. If two nodes are connected with an arrow, this suggests that the two nodes are conditionally dependent.
Bayesian Network analysis is efficient in examining the interactions of all system components included in the model, when individual, organizational, and physical factors co-exist and can jointly contribute, and this has the potential to provide valuable insights on the determinants of employee safety and health outcomes. This approach is remarkably suitable to socio-technical systems approaches, which emphasize the importance of examining the interplay among various system components in order to optimize their functioning for a healthy and sustainable workplace. The structural learning algorithm of Bayesian Network analysis is a type of greedy search algorithm [35
], and this approach enables investigation of every possible structural association among selected variables to estimate the most probable model that satisfies the model selection criteria of a particular learning algorithm.
Moreover, Bayesian Network analysis is less susceptible to multicollinearity problems [36
]. By leveraging the inter-correlations among variables, Bayesian Network analysis provides conditional probability distributions for the dependent relationships of study variables. A complex system can thus be viewed in a modular way by “breaking down the discovery process into the search for the specific components of a complex system, and thus avoiding…multicollinearity” (p. 59) [37
Bayesian Network analysis has been successfully applied in various settings that require quantitative modeling of complicated relationships among many variables, such as in genetic modeling and disease diagnosis [37
]. Also, the Bayesian Network approach has already been used to unveil the mutual relationships among various organizational and psychosocial factors regarding stressors, stress, and strain in workplace. Specifically, more task demands were associated with more stress at work [38
], while social support from both supervisors and co-workers was critical in workplace stress prevention [39
]. All things considered, Bayesian Network analysis is therefore well suited to investigate complicated interrelationships among multiple psychosocial and behavioral outcomes of stress.
In summary, it is worth noting that there has been no scientific study which integrates the conceptual framework of Total Worker Health and the machine learning analytic framework as a means to promote workplace health and well-being. No publication was found under the keywords search of “Total Worker Health” and “machine learning” in PubMed and PsycINFO. To address this gap, the present study adopted the Total Worker Health framework and Bayesian Network analysis approach (1) to reveal the most probable ways that psychosocial and behavioral outcome variables in corrections work are interrelated, and (2) to identify the key contributing factors of this interdependency relationship within the unique occupational context of correctional work.
Psychological and physical health problems and concerns are diverse and prevalent among correction workers. In order to reveal the interplay among potential outcomes within their unique occupational context in a manner consistent with socio-technical systems approaches, the present study adopted the Bayesian Network analytic framework because of its flexibility and efficiency in exploring all possible interrelations among the study variables. Guided by well-established frameworks for workplace stress mechanisms, including the Job Demands-Resource Model [12
], Spillover Theory [22
], and Conservation of Resources Theory [23
], study variables were selected based on assumptions about their potential interrelations. Subsequently, use of the data-driven Bayesian Network approach identified a model showing a probable scenario in which correction workers’ exhaustion had dominant influence over a chain reaction that significantly raised the risk of several other negative psychosocial and behavioral outcomes. Correction workers’ exhaustion was found to be associated with a lack of work engagement, depressed mood, and also interfered with work-family balance. Additionally, increased disengagement was found to be associated with work-related stress, while depressed mood was closely associated with less regular physical activity. At the same time, depressed mood and work-family imbalance were jointly associated with reduced work ability.
The results extend the previous finding of a close relationship between psychological distress and work-family imbalance among correction workers [54
]. The identified joint processes of how the impact of exhaustion can be enhanced or mitigated by other factors also extends to previous findings regarding the impact of exhaustion on depression, absenteeism, depersonalization, and reduced personal accomplishment that were based on a French correctional officer sample [55
4.1. Theoretical and Analytic Implications
First, the study findings complement relationships predicted by the Job Demands-Resource Model. In this model, job demands that result in exhaustion can negatively impact worker engagement, and this in turn is associated with work-related stress. Also, this model suggests that job demands that are closely associated with exhaustion will be negatively associated with psychosocial (i.e., mood and work-family imbalance) and organizational behavior outcomes (i.e., work ability). These same relationships were supported by the data-driven Bayesian Network model identified in the present study.
However contrary to relationships posited in the conventional stress models, relationships between stress to outcomes (i.e., engagement, workability) were not detected in the present study. This may be partly explained by the fact that other factors, such as depressed mood and work-family imbalance, were considered simultaneously along with the engagement and stress variables, speaking to the importance of examining psychosocial variables in combination with occupational context variables. Our findings indicate that the Job Demand-Resource Model may be a reasonable framework to first approach occupational health and safety issues, but it is also important to account for workers’ adjustment to their particular work context, as is the case in the present study of correction workers.
Second, although a stress-to-burnout directional relationship has been widely accepted in previous studies [8
], the present study showed that burnout (i.e., exhaustion) may be associated with derivative stress, which is subsequent to employee disengagement. In fact, there have been previous reports that disengagement, which may be followed by burnout, leads to more stress in workplace [56
] and also a distressed mental state [57
]. It should be noted that the finding here of a burnout-to-stress relationship doesn’t necessarily reject the well-established causal link from stress to burnout, and also that our findings were based on the unique context of correctional work for which well-established causal linkages may not apply. Nonetheless, the results do suggest the need for adopting an unbiased perspective on how stress may be exacerbated by burnout in a particular population of workers. Within an autoregressive model framework [58
], an outcome at one point in time can serve as a cause at another future point in time, and this may also be the case for the reciprocal stress and burnout relationship. Future studies could take a longitudinal approach to further examine the nature of the causal mechanisms behind the stress-burnout relationship.
4.2. Practical Implications
The results support use of the Total Worker Health framework for considering a wide range of factors impacting worker health and wellbeing. Integrated interventions that promote safety, health, and well-being among correction workers appear to be warranted, given the complex interplay of the psychosocial and behavioral factors reported on here. The results also suggest that taking steps to reduce the level of exhaustion in this working population may provide the most efficient way to reduce depressed mood, interruption of work-family balance, and a lack of work engagement. These factors were also shown to increase the risks of work-related stress, a reduction of regular physical exercise, and lowered work ability. Continued failure to manage workers’ exhaustion would be ignoring an apparent primary risk factor.
However, management of this primary risk factor may prove particularly challenging if there is short staffing and also overtime requirements in an organization that functions 24/7 and that cannot afford to be understaffed at any time. If work-related exhaustion cannot be avoided, then intervention efforts can next focus on the proper management of workers’ depressed mood because this was also shown to be strongly associated with the loss of work-family balance, increased negative attitudes toward their job, and reduced workability. Establishing a hierarchy of risk factors and then selecting the most attainable intervention is compatible with a participatory approach to intervention planning [28
] which aims at identifying and addressing workers’ needs and concerns that are most salient given their present circumstances.
4.3. Limitations and Suggestions for Future Study
Some limitations of the present study need to be addressed in future research. Generalizability of the findings is limited given the uniqueness and relatively small size of the sample. In particular, it can be noted that only half of the data was utilized for Bayesian Network model learning, while the remaining half of the data was utilized for the validation of the learned Bayesian Network model. Also, the present study was cross-sectional, which did not allow examination of the dynamic relationships among study variables across time. Moreover, the final Bayesian Network model’s accuracy level was not at an ideal level, although it was found to be meaningfully above chance. To resolve this issue, a larger sample can be used for Bayesian Network model learning to enable more robust and reliable modeling. More data points generally ensure more reliable estimation of probability, particularly when the probability is smaller. Also, future studies are needed to more clearly demonstrate the distinct roles of job demands, job control, and support in the extended mechanisms affecting stressors, stress, strain, and the exacerbation of the strain symptoms. The role of individual differences, such as gender, age, and tenure in the interrelations among stress and stress outcomes can also be examined in future studies.