Identified and Selected Studies
The primary search resulted in 828 hits (PsycInfo = 302, Pubmed = 242, Web of Science = 284). Next, relevant articles were selected according to their titles and abstracts. After this selection, 162 articles remained (PsycInfo = 68, Pubmed = 52, Web of Science = 42). A further selection was made based on the content of the full text. This gave a total of 51 potentially relevant articles published until February 2017. Of 51 studies, we finally only selected empirical cross-sectional and longitudinal studies, and selected studies that treated resilience (or described related terms) as a predictor in these analyses, resulting a final set of 22 articles. At this stage, we excluded literature reviews (e.g., Honig and Sultan 2006; Shochet et al. 2011), studies evaluating interventions aimed at enhancing resilience (e.g., Andersen et al. 2016; Ramey et al. 2016), and studies assuming that not having PTSD symptoms is being resilient (e.g., Bowler et al. 2012; Galatzer-Levy et al. 2011) because these studies do not focus on investigating the relationships between resilience and officers’ functioning or problems in functioning (see Fig. 1).
General Study Characteristics
Table 1 provides a description of each study included in this review, e.g., authors; year of publication; sampling and design including response, demographics of respondents, the independent, and dependent measures; and the main results and conclusions. The main results and conclusions discuss the significant bivariate and multivariate associations.
Most of the included studies were published since 2006, with a small majority of the studies published in recent years (2013–2017: N = 12, 55%). The large majority (N = 17, 77%) had a cross-sectional design. Five studies (Greene and Nowack 1995; Marchand et al. 2015; Tang and Hammontree 1992; Walumbwa et al. 2010; Yuan et al. 2011) had a longitudinal design with a minimum of 8 weeks and a maximum of 3 years between the baseline and final follow-up. The cross-sectional studies had larger sample sizes (between N = 52 and N = 2226) than the longitudinal studies (between N = 60 and N = 343). Most of the studies (N = 18) relied only on self-report measurements. The studies of Marchand et al. (2015) and Yuan et al. (2011) conducted clinical interviews (SCID), combined with self-reported measures. Greene and Nowack (1995) and Violanti et al. (2014) used absenteeism data out of electronic databases. About 50% (N = 11) was conducted in the USA. In all studies, more men than women participated, which corresponds with the male-female ratio among police officers. The mean age of the samples varied between 25.3 and 54.4 years. The mean years of service varied between 4.7 and 20.1 years. All studies focused on (general) police officers in patrol assignment besides the study of Fyhn et al. (2015), which focused on police investigators.
Concepts and Instruments Capturing Resilience
Table 2 provides an overview of the definitions used for resilience, hardiness, and psychological capital in each of the included studies. It includes some more or less standard definitions available in the wider literature, as well as some definitions created by the authors itself and conceptualizations of the definitions. Of the identified papers, eight studies (36%) used the term resilience, nine studies (41%) used the term hardiness, and five studies (23%) used the term psychological capital.
The term resilience was considered as a uniform, unidimensional concept (N = 4) and as a combined multidimensional concept, consisting of several partial facets (N = 4). Although studies using the term resilience do differ in wordings of definitions, they all refer to resilience as an ability to deal with stressful situations.
Furthermore, Table 1 shows that resilience was measured in quite different ways, e.g., resilience was assessed by different instruments covering different topics. For example, de Terte et al. (2014) mentioned resilience as a combination of optimism, adaptive coping, emotional competence, adaptive health practices, and social support, and measured these concepts respectively with the Life Orientation Test, Brief Resilient Coping Scale, Mayer-Salovey-Caruso Emotional Intelligence Test, Health Practices Index, and Social Support Scale.
Hardiness, consisting of the facets of challenge, control, and commitment, was considered to be a personality state, trait, or style in the identified studies. Hardiness was most often measured with scales designed by Bartone: for example, the Short Hardiness Scale (SHS; Bartone 1995) or the Dispositional Resilience Scale-15 (DRS-15; Bartone 2007). These measures are both based on the longer Dispositional Resilience Scale (Bartone 2007).
Five studies measured psychological capital, e.g., a combination of resilience, self-efficacy, optimism, and hope (Farr-Wharton et al. 2016; Lu et al. 2015; Ojedokun and Balogun 2015; Siu et al. 2015; Walumbwa et al. 2010). All of these studies used the Psychological Capital Questionnaire to measure psychological capital, indicating that psychological capital was defined and assessed in a (much) more uniform way than resilience and hardiness.
All definitions in some way refer to abilities, strengths, styles, or traits enabling police officers to cope successfully with difficult, stressful, or adverse events (either as a moderator or mediator), thus enabling police officers to (keep) carry(ing) out their duties. However, none of the instruments assessed concrete behaviors showing that resilient police officers were indeed able to cope with adversity and stressful situations more successfully, e.g., to perform better in their law enforcement and related policing tasks.
Measurements of Other Variables
In the included studies, (mental) health problems were treated mostly as dependent variables, such as PTSD symptoms (e.g., Andrew et al. 2008; Andrew et al. 2013)), psychological distress (e.g., Andrew et al. 2008; de Terte et al. 2014), physical health (e.g., de Terte et al. 2014; Fyhn et al. 2015; Greene and Nowack 1995), and burnout (e.g., Fyhn et al. 2015; Gupta et al. 2012). There was much less attention for the relationship between resilience and non-health variables, like stressors (Farr-Wharton et al. 2016; Velichkovsky 2009), personality (Gupta et al. 2012; Velichkovsky 2009), and job-related variables (e.g., Hills and Norvell 1991; see Table 1).
Table 1 furthermore shows that the identified studies vary widely in terms of the measures used. For example, PTSD symptoms were measured with the Impact of Event Scale(-R, -K), Structured Clinical Interview for DSM-IV Axis I Disorders, Modified PTSD Symptom Scale, PTSD checklist, and Civilian Mississippi Scale.
Results Predictive Value of Resilience
As shown in Table 1, dependent variables studied in relationship to resilience were predominantly (mental) health related.
For Physical Health
Five studies examined the predictive value of resilience for physical health of police officers. Greene and Nowack (1995) found in a longitudinal study of 3 years that hardiness was very weakly and negatively associated with hospitalization and not associated with absenteeism, based on the adjusted R-squared of hardiness and controlled for psychological well-being, age, and ethnicity. Cross-sectional studies examined various aspects of physical health. De Terte et al. (2014) found weak to moderate positive associations between resilience and physical health, based on the R-squared of resilience and controlled for traumatic event exposure. Velichkovsky (2009) found very weak to moderate negative associations between resilience and chronic illness, smoking, and alcohol use, using correlations and logistic regression analyses. Hills and Norvell (1991) found a very weak positive association between hardiness and physical symptoms, based on the adjusted R-squared. Violanti et al. (2014) found with rate ratios a relationship between hardiness commitment and the total score of stressors for 1-day work absences in a binomial regression analysis.
For General Mental Health
With respect to mental health, cross-sectional studies found a moderate positive association between psychological capital and psychological well-being (Farr-Wharton et al. 2016) and weak to moderate negative associations between psychological capital, anxiety, and depressive symptoms (Ojedokun and Balogun 2015), using SEM analyses. In another SEM analysis, Siu et al. (2015) found a weak negative association between psychological capital and stress symptoms. In contrast, Velichkovsky (2009) found in a correlational analysis a strong negative association between resilience and stress symptoms. Resilience was moderate negatively associated with psychological distress, based on the adjusted R-squared of resilience and controlled for traumatic event exposure (de Terte et al. 2014). Andrew et al. (2008) and Andrew et al. (2013) conducted multiple regression analyses and controlled for age, education, and marital status. Based on the standardized regression coefficients, Andrew et al. (2008) and Andrew et al. (2013) found weak to moderate negative associations between hardiness (2008: men: control, women: control and commitment; 2013: men: challenge, control, and commitment, women: control and commitment) and depressive symptoms. Besides that, in men, hardiness was weakly and negatively associated with psychological symptoms (Andrew et al. 2008) and there were weak negative associations between hardiness (men: challenge, control, and commitment; women: commitment) and anxiety symptoms (Andrew et al. 2013).
For PTSD Symptomatology
In total, eight studies assessed the independent predictive value of resilience for PTSD symptomatology. Longitudinal studies found both no independent effect of hardiness on PTSD symptoms in a multiple regression analysis (Marchand et al. 2015) and very weak independent effects of the resilience facets of “belief in the benevolence of the world” (positive effect) and “social adjustment” (negative effect) on PTSD symptoms, based the adjusted R-squared and controlled for ethnicity and traumatic event exposure (Yuan et al. 2011). Cross-sectional studies (Andrew et al. 2008; Andrew et al. 2013; de Terte et al. 2014; Lee et al. 2016; McCanlies et al. 2014; Prati and Pietrantoni 2010) found varying associations between resilience, hardiness, and PTSD symptoms. In linear regression analyses, Andrew et al. (2008) and Andrew et al. (2013) controlled for age, education, and marital status and based on the standardized regression coefficients, they found weak to moderate negative associations between hardiness and PTSD symptoms. Andrew et al. (2008) found that, for women, the hardiness facet commitment was negatively associated with PTSD symptoms. Andrew et al. (2013) found that for men, all hardiness facets were associated with PTSD symptoms; for women, there was an association with the hardiness facets control and commitment. De Terte et al. (2014) found a moderate negative association between resilience and PTSD symptoms, based on the R-squared of resilience, while controlling for traumatic event exposure. McCanlies et al. (2014) conducted ANCOVAs to examine the relationship between resilience and PTSD symptoms and controlled for age, gender, ethnicity, education, and alcohol use. Based on the standardized regression coefficients, McCanlies et al. (2014) found a strong negative association between resilience and PTSD symptoms. Following Prati and Pietrantoni (2010), resilience was weakly negatively associated with PTSD symptoms, based on the R-squared of resilience. In a logistic regression analysis, Lee et al. (2016) found a negative association between resilience and PTSD symptoms, controlling for age, education, marital status, smoking, alcohol use, service area, duration of patrol service, job stress, and depression.
Four cross-sectional studies examined the relationship between resilience and burnout. Fyhn et al. (2015) found a very weak negative association between hardiness and burnout, based on the adjusted R-squared of hardiness and controlled for age, gender, police experience, and position experience. Gupta et al. (2012) found in a correlational analysis a weak to moderate negative association between resilience and burnout, whereas Velichkovsky (2009) found a moderate to strong negative association between resilience and burnout, based on correlations. Hills and Norvell (1991) did not find an association between hardiness and burnout in a stepwise regression analysis.
The two cross-sectional studies on the predictive value of resilience for personality showed in correlational analyses weak to moderate positive associations between resilience and the personality characteristics of conscientiousness and agreeableness (Gupta et al. 2012) and a weak negative association between resilience and type A personality (Velichkovsky 2009).
For Psychological Capital
Walumbwa et al. (2010) found in a hierarchical linear model that leaders’ psychological capital, followers’ psychological capital, and service climate were moderately associated with the supervisory-rated performance of followers, based on the R-squared. In a cross-sectional design and using SEM analyses, studies found that psychological capital was moderate positively associated with job satisfaction (Siu et al. 2015) and stressors (Farr-Wharton et al. 2016). Hills and Norvell (1991) found a weak negative association between hardiness and job satisfaction in a stepwise regression analysis.
For Interaction and Mediation Effects of Resilience
Five studies examined the interaction effect of resilience. Tang and Hammontree (1992) found in a longitudinal design a weak interaction effect between hardiness and police stress on absenteeism, based on the adjusted R-squared of the interaction effect. In two cross-sectional studies, James et al. (2006) found in a multiple regression analysis no interaction effect between hardiness and dysphoria on anger. Johnsen et al. (2017) found an interaction effect between self-efficacy and hardiness for performance satisfaction, but not for perceived strain, based on an OLS regression analysis.
Lu et al. (2015) and Siu et al. (2015) conducted SEM-analyses in cross-sectional designs. Lu et al. (2015) found that job stress and identification with the police organization were very weakly associated with job satisfaction through psychological capital. Siu et al. (2015) found an indirect effect of psychological capital via stress symptoms and job satisfaction to the turnover intention of a police officer.