PTSD in prison settings: A systematic review and meta-analysis of comorbid mental disorders and problematic behaviours

Purpose Prevalence rates of PTSD are higher in the prison population than in the community. We sought to systematically review the extent to which this disorder is associated with other mental health disorders and problematic suicidal or aggressive behaviours in the prison population. Methods Studies reporting a relationship between PTSD and comorbid mental disorders and/or problematic behaviours in imprisoned adolescent and adult populations were identified from four bibliographic indexes. Primary studies involving clinical interviews, validated instruments leading to DSM or ICD diagnoses, or validated self-report questionnaires such as the PTSD checklist were included. Random-effects meta-analysis was conducted where possible. Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines were followed. Results This review identified 36 studies, with a combined sample of 9594 participants, (6478 male and 2847 female prisoners) from 11 countries. Thirty-four of the identified studies employed a cross-sectional design. We identified significant associations between PTSD and comorbid mental disorders including depression (OR = 3.4, 95% confidence interval (CI): 2.3–4.9), anxiety (OR = 2.9, 95% confidence interval (CI): 1.8–4.7) and substance use (OR = 1.9, 95% confidence interval (CI): 1.5–2.4). We also identified significant associations between PTSD and suicidality (OR = 3, 95% confidence interval (CI): 2.4–3.8) and aggressive behaviours (this latter finding was not subject to meta-analysis). Significant methodological heterogeneity was identified between studies. Conclusions High rates of psychiatric comorbidity among prisoners with PTSD, and links to suicidal behaviour, self-harm and aggressive behaviour, provide further support for the need for trauma-informed treatment approaches in prisons. However, significant gaps in the current evidence were apparent. In particular, a lack of large, longitudinal studies meant that the temporal relationships between PTSD and relevant outcomes cannot currently be determined.


Introduction
High levels of lifetime traumatic exposures have been reported in studies of prison populations [1,2].A recent international meta-analysis confirmed that the prevalence of PTSD in prison populations, like other mental disorders, is higher than in community populations, with a pooled point prevalence of 6% in male prisoners and 21% in female prisoners [3].
High rates of other mental disorders, and problematic behaviours such as suicidal and aggressive behaviours have also been extensively documented in prison populations [4][5][6].The relationship between PTSD and these outcomes is poorly understood and this may be perpetuating under-diagnosis and under-treatment of PTSD in prisons [7].Community and military population studies have suggested that PTSD is a disorder which is highly comorbid with other mental health disorders [8], such as depression [9] and substance misuse [10,11].PTSD has also been linked to suicidality [12], self-harm [13], criminality [14], violence and aggressive behaviour [15][16][17][18] in community, clinical and military population studies.
In the present study, we examined the associations between PTSD and comorbid mental disorders or problematic behaviours in 9594 imprisoned individuals.To the authors' knowledge, this study is the first meta-analysis that examines such associations in the adolescent and adult prison population.

Methods
This systematic review protocol was pre-registered in PROSPERO (CRD42017068958) and PRSIMA guidelines were followed [19].

Search strategies
We conducted a systematic search of the PTSD literature in prison populations, last updated on February 9 th , 2019.The search included four online databases (Embase, MEDLINE, Psy-cINFO, Web of Science) and reference lists of identified papers and relevant systematic reviews [3].For the online database searches, we used an identical combined strategy of free-text strings and subject headings (see S1 Text).

Inclusion and exclusion criteria
We identified studies in which associations between PTSD and relevant correlates were reported.The following inclusion criteria were applied: 1) youth or adult prison samples (where the proportion of under 18 year olds represented less than 10% of the entire sample, the study was considered representative of adult prisoners); 2) probable PTSD diagnoses were established with validated diagnostic instruments such as the Structured Clinical Interview for DSM-5 [20] or validated self-report questionnaires such as the PTSD checklist [21]; 3) the relationship between PTSD and at least one mental health comorbidity, or problematic behaviour was examined; and 4) studies of male, female or mixed samples (in cases where the proportion of females represented less than 10% of the entire sample, the study was considered representative of male sex).Studies meeting the following criteria were excluded: 1) investigations in prisoners of war, or other criminal justice settings e.g.probation, court; 2) no measure of association or effect; or 3) outcomes outside the scope of this review (i.e.physical health problems).
Titles and abstracts were screened against the inclusion and exclusion criteria.The remaining full texts of potentially eligible studies were then evaluated.Quality appraisals of included studies were performed by two independent postgraduate-level reviewers (EFI, FAC) using a checklist adapted from validated tools (see S1 Table; [22][23][24][25]).The total possible quality rating ranges from 0 to 42 points.Samples with a score of 32 and above were considered high quality; those with scores between 22 and 32 were considered medium quality; and those with scores of 21 or below were considered low quality.Disagreements between these two reviewers were resolved by consensus with a third senior reviewer (DM).

Statistical analyses
While use of meta-analyses was precluded for most relationships due to an insufficient number of studies, meta-analyses were conducted for three comorbid mental disorders-Depression, Generalised Anxiety Disorder, and Substance Use Disorder and one behavioural association, namely suicidality.Meta-analyses were conducted with Stata version 15.1 using pooled random effects odds ratios.
The significance and the magnitude of heterogeneity across studies were calculated using the Q statistic and I 2 statistic; significantly high levels of heterogeneity were indicated for Anxiety and Depression, but not for SUD or suicidality.Subgroup analyses were performed to examine differences according to gender, age, and timing of PTSD diagnosis (current or lifetime).For the purposes of analysis, "Current" timing included studies measuring PTSD at one month, 6 month and 12-month diagnostic periods.One study of incarcerated youth [26] did not provide a breakdown by gender and was therefore excluded from the gender subgroup analyses.One study [27] measured two diagnoses relating to SUD (substance abuse and substance dependence), and so both of these were included in analysis.Measures of suicidality included in the meta-analysis included lifetime measures of suicidal behaviour or ideation (n = 5) and current suicide risk (n = 2).There were insufficient studies to provide adequate statistical power for meta-regression, precluding further examination of effect size moderators.

Key features of included studies
The 36 studies  reported on a combined sample of 9594 participants, 6478 male and 2847 female prisoners.Stratified by age, identified studies reported on a combined sample of 4139 incarcerated young offenders (age range 10-19) (M = 2766; F = 1324) and 5455 adult prisoners (M = 3712; F = 1523).The key characteristics of included studies are summarised in Table 1.
Half of the studies utilised a validated structured diagnostic interview to assess participants for PTSD (n = 18) [26-28, 30, 32, 34, 35, 38, 41-43, 46, 49, 51, 54, 55, 57, 60], with the remainder using validated self-report screening questionnaires.Although the reported prevalence of PTSD was typically higher among studies which used a questionnaire with a cut-off range, magnitudes of associations did not appear to differ substantially between such studies and those which used a diagnostic interview, although this could not be examined quantitatively due to an insufficient number of studies.
There was some heterogeneity between studies in the time-period within which PTSD was measured.The majority (n = 22) assessed current PTSD (i.e.past month, past week), but several studies [30,34,38,41,42,51,55,57] measured lifetime prevalence and four measured 6-or 12-month prevalence [28,43,46,49].The overwhelming majority (94%) were cross-sectional studies, with only two studies [38,51] employing any prospective/ longitudinal element.18 studies measured associations between PTSD and other mental disorders, and 25 studies examined at least one behavioural problem.Of the 36 studies included in this review, 11% were considered high quality, 72% medium-quality and 17% low quality.Many studies [26, 27, 31-37, 39, 40, 43, 45-57, 61] reported associations between PTSD and other mental disorders or with problematic behaviours using simple group comparisons without further reporting analyses adjusted for potential confounding factors that may have contributed to the association.

PTSD and comorbidity with other mental disorders
PTSD was found to be highly comorbid with other psychiatric disorders (Table 2).Significantly higher rates of psychiatric comorbidity were found amongst those with PTSD compared to those without such a diagnosis although analyses were not always adjusted for potential     confounding.Affective disorders, most notably depression (n = 13), and anxiety disorders (n = 9) were the most frequently researched comorbidities in studies of both youth and adult prisoners.Depression.Across the eight studies included in the depression domain, the randomeffects pooled OR of a comorbid depressive disorder was 3.4 (95% CI [2.34, 4.

89]) in individuals with PTSD (Fig 2).
There was substantial heterogeneity between studies (I 2 = 60.0%).We therefore analysed results by subgroup, to explore how estimates were affected by gender, age and timing of PTSD diagnosis.
Generalized anxiety disorder.Across the eight studies included in the GAD domain, the random-effects pooled OR of a comorbid anxiety disorder was 2.43 (95% CI [1.19,4.96]) in individuals with PTSD.Substantial heterogeneity was observed (I 2 = 79.8%).After removing one study from the meta-analysis due to an effect size that was an outlier to the group [27], the observed Odds Ratio for comorbidity between PTSD and Anxiety was 2.95 (95% CI [1.83-4.74])(Fig 3).Heterogeneity estimates reduced but remained significant (I 2 = 59.1%).We therefore analysed results by subgroup, to explore how estimates were affected by gender, age and timing of PTSD diagnosis.Gender.Risk estimates of comorbidity with anxiety were higher for male (OR = 3.27, CI [1.91-5.62])than female (OR = 2.58, CI [0.95-9.58])prisoners.Significant heterogeneity was observed within the female (I 2 = 77.3%),but not the male subgroup (I 2 = 9.9%, p = 0.  Psychosis.Five studies examined psychotic illnesses, and three found statistically significant associations.However, studies reporting positive associations included two small samples of female youth [27,30] and one highly selected sample of adult Aboriginal prisoners [46].Of the five studies examining associations between PTSD and ADHD, one found a statistically significant association [59]. Personality disorder.PTSD and Personality Disorder (PD), were found to be comorbid in four of five adult studies.Among samples of incarcerated females (n = 3), PTSD was found to be most consistently and strongly associated with Borderline PD.One medium quality study of male prisoners found lifetime PTSD to be associated with ASPD [43].No study found statistically significantly elevated rates of PTSD among prisoners with primary psychopathy compared to those without; one medium-quality study found that these two disorders were negatively associated.Studies of the developmental precursors of these disorders (conduct disorder, callous unemotional traits) suggested mixed associations with PTSD.While one medium-quality study [34] found no association between PTSD and CU trait scores, another similar quality study, also of detained male youth [36], found that those with PTSD had significantly higher callousness trait scores than those without.

PTSD and its relationship with problematic behaviours
Suicidality.Across the seven studies included in the suicidality domain (which included lifetime suicide attempts and current risk of suicidal behaviour), the random-effects pooled OR was 3.03 (CI 2.45-3.76)-seeFig 5 .No significant heterogeneity was detected in the overall model.Minimal differences in risk estimates were observed for youth (OR = 2.91, CI[1.85-4.57])and adult (OR = 3.07, CI[2.40-3.92)participants; some differences in risk estimates were observed for male (OR = 3.34, CI[2.28-4.88])and female (OR = 2.90, CI[2.12-3.97])participants.
In total, twelve studies investigated the association between PTSD and problems relating to suicidality, which included suicide attempts, suicidal ideation, measures of suicide risk, or self-injurious behaviour, and nine found statistically significant associations (Table 3).Three studies investigated associations between PTSD and non-suicidal self-injury (NSSI), with all three also reporting positive main effects.However, most studies examining suicidality made simple group comparisons (e.g.PTSD vs no PTSD) and did not account statistically for other covariates which may have accounted for all or part of the observed association.Four studies (29,37,58,59) conducted multivariate analyses, and found that positive univariate associations between PTSD and NSSI or suicidality were rendered non-significant once added to a multivariate model with other significant correlates of PTSD such as childhood maltreatment or psychiatric comorbidity.
Aggressive behaviours.PTSD and violent or aggressive behaviour was assessed in nine studies using predominately male (n = 6) or adult samples (n = 5).Several (n = 6) studies reported positive associations between PTSD and aggression or violence, although few adjusted for potential confounders, and significant heterogeneity in measurement was also evident.Of note, one study of incarcerated youth reported that reactive, but not proactive, aggression was associated with PTSD symptoms [37].Evidence supporting a relationship between PTSD and violent behaviour in adult prisoners was stronger among male compared to female samples, although one adult study which found significant associations between PTSD and aggression utilised a selected sample of male prisoners with comorbid substance use problems [53].Five studies also examined the relationship between PTSD and self-reported anger or hostility, all of which found statistically significant associations (Table 3).
Offending behaviours.The role of PTSD in understanding offending and criminal behaviour was investigated in five studies, with limited evidence of an association (See  Table 3).None were considered high quality.Studies were predominately comprised of adult (n = 4), male (n = 5) samples.Three studies explored links with reoffending, and three investigated the association between PTSD and the type and severity of prisoners' offending.

PTSD and comorbidity with other mental disorders
This systematic review of the association between PTSD and both other mental disorders and behavioural problems in youth and adult prison populations is based on 36 studies from 11 countries worldwide (Table 4).Results from meta-analyses indicated that the psychiatric disorder with the strongest association with PTSD was comorbid depression, followed by anxiety disorders.Prisoners with PTSD were also significantly more likely to have a substance use disorder, although the effect size was relatively small.A number of other disorders did not have sufficient data to permit a meta-analytic synthesis.Nevertheless, systematic review at least suggests the following.In adults, comorbidity with psychosis was less strongly evident than with neurotic disorders, and there was an association between PTSD and Cluster B personality disorders, particularly among female prisoners.A possible inverse relationship was observed between PTSD and ASPD with psychopathic features in men [55].In adolescents, PTSD was not found to be any more likely to present among those with ADHD or conduct disorder than those without.

PTSD and associated behavioural problems
Evidence for an association between PTSD and behavioural problems in prison was mixed (Table 4).Results from meta-analyses indicated a significant association between PTSD and measures of suicidality, with risk estimates slightly higher among male prisoners.Associations with measures of aggression or offending behaviours did not permit meta-analytic syntheses.However, systematic review suggests that there are significant associations between PTSD and aggressive behaviours, particularly in adult samples.Consistent with findings on psychopathy [55,56], there was some indication in the literature that PTSD was not associated with instrumental violence [37], which could suggest that aggressive behaviour in PTSD occurs primarily in the context of arousal and reaction to perceived threat, as opposed to callousness or lack of empathy.This review found limited evidence of an association between PTSD and offending type or recidivism.

Youth vs adult samples
Results from meta-analyses suggested that adult samples reported stronger associations with depression and anxiety compared to youth samples.The association between PTSD and substance misuse was stronger amongst studies of incarcerated youth.Findings must be considered in light of previous findings that rates of reported trauma and PTSD may be higher in youth samples compared to adult samples [49], and that younger age has been cited as a risk factor for outcomes including institutional violence or self-harm [5,62].Interestingly, only one identified study compared samples of both youth and adult prisoners, and reported no significant interactions between age and PTSD in the prediction of anger and hostility [49].

Impact of gender
Rates of PTSD in prison are higher amongst females compared to males [3].However, our meta-analyses found stronger effect sizes among male samples for depression, anxiety and substance use comorbidities.This finding is consistent with a previous high quality study which highlighted that males with PTSD were more likely to have comorbid disorders compared to females with PTSD [28].Gender differences in the types of mental disorders and behaviours examined by studies were also noted.ASPD, Psychopathy, ADHD and CU traits were more frequently investigated in male samples, while comorbidity between PTSD and BPD were only investigated in female studies.Similarly, problems relating to externalising behaviour (violence, aggression, offending) were investigated more amongst male prisoners, while internalising (suicidality, self-harm) behaviours were more consistently examined in female samples.

Strengths and limitations
This is the first systematic review and meta-analysis, to our knowledge, to investigate associations between PTSD and comorbid mental disorders and problematic behaviours in prison populations.It included studies of both imprisoned youth and adults which employed validated tools to measure PTSD diagnosis and symptoms.One of the main limitations of this review was the methodological heterogeneity between the studies, such as variations in the time period of measurement of both PTSD and comorbidities (i.e.past year or lifetime), varying definitions of outcome measures, and differences in the criminal justice characteristics of the sample (i.e.short-term detainees vs sentenced prisoners).Most of these studies were cross-sectional in design, limiting any causal inferences.Only four [28,42,46,49] studies identified by this review were considered high quality, and many included studies had small sample sizes (<100).Of 36 studies examined, only 12 took account of potential confounders which may have explained any associations identified in simple group comparisons (or univariate analyses).Most domains explored using meta-analyses indicated significant heterogeneity.These variations made comparisons between studies challenging, precluding the use of meta-analyses in most cases, and meta-regression in all cases.It was also of note that only two studies identified by this review specifically investigated the construct of Complex PTSD (CPTSD) [32,42].In addition to symptoms of "simple" PTSD (i.e.reexperiencing, hypervigilance), CPTSD also requires disturbances in affect dysregulation, negative self-concept and interpersonal relationships [63,64].Findings from this review highlight the increasing need for research which differentiates between these two disorders, to examine their potentially distinct roles in adverse outcomes.
A final limitation was the lack of information on relationships with offending behaviour and recidivism.Limited to no relationship between PTSD and offending behaviour was identified by this review, however only five studies, the majority of which were cross-sectional, investigated such associations.Given that preventing recidivism remains a central task for those working in prisons, future research is needed to explore this further, and establish whether or not PTSD is prospectively linked with different forms of offending behaviour or criminal activity, as well as readmission to custody.

Implications and conclusions
PTSD is a common disorder within prison populations [3].People in prison are more likely to have experienced cumulative, multiple traumas across their lifetime [65], further increasing the risk of developing mental health problems [66], a pathway that may in part be mediated by the presence of PTSD symptoms [32,61,67].The presence of PTSD has been linked to poorer treatment outcomes including functional impairment and treatment adherence [68].However, PTSD often goes undetected by mental health services [69].Screening for this disorder is not routinely embedded in clinical services, and the disorder typically remains un-diagnosed and untreated within prison settings [5,7,70].The need for improved identification and treatment of PTSD in prison settings is further underscored by findings suggesting that spontaneous long-term remission rates of this disorder are modest [71], and that the evidence for the efficacy of short-term trauma-based therapies in this population is limited [72].While the development of trauma-informed care is a welcome recent development, there is little consensus on how it is best defined or operationalized in prison settings [73,74].We have demonstrated that prisoners with PTSD are significantly more likely to also have comorbid depressive, anxiety or substance use disorder diagnoses, and that adult male prisoners with PTSD may be at the greatest risk of having co-occurring mental health difficulties.Findings suggesting associations between PTSD and suicidality also have important implications for future research into pathways to self-harming and suicidal behaviour in prison environments [5,75].While relationships with suicidal behaviour or ideation are likely to be complex and influenced by several factors, including comorbid disorders like depression, the specific role of PTSD has, until recently, been overlooked [5].Therefore, improved screening and identification of PTSD is essential to improve access to clinical treatment and should be prioritised as an important first-step.
Finally, while this review found evidence for several cross-sectional associations between PTSD and other important mental health and behavioural problems, there was a notable lack of studies which investigated prospective outcomes-only two studies identified by this review employed a longitudinal design, both measuring readmission to custody.Thus, while there was evidence of associations between PTSD and suicidality or aggression, causal relationships between PTSD and subsequent risk of such adverse outcomes could not be assessed.This review has therefore highlighted the lack of robust research in this area and the need for future longitudinal studies utilising standardised and validated measures of both PTSD and outcomes, to explore the longer-term impact of PTSD on youth and adults in custody.

Statements
The manuscript does not contain clinical studies or patient data.The authors declare that they have no conflict of interest.
Fig 1 describes the study selection process.

Table 2 .
(Continued) a Gender distribution of sample not reported b CU traits only measured within a subsample of total study sample (N = 373) c Includes sample from Caraballo, 2013 https://doi.org/10.1371/journal.pone.0222407.t002