Abstract
Using data from the 2016 Survey of Prison Inmates, this study used latent class analysis to examine patterns of mental health comorbidity within a large, nationally-representative sample of incarcerated adults (N = 24,848), including 7.6% with prior military service. Classes were compared on Veteran status, military service-related variables, and treatment-related variables. Results suggest four latent mental health patterns—“Low Psychopathology” (70% of the total sample), “Internalizing + Thought Disorder” (8%), “Internalizing” (14%), and “High Psychopathology” (8%). The High Psychopathology class had the highest rates of prior psychiatric/psychological treatment. Incarcerated Veterans were more likely to be in the Internalizing class, and rates of combat exposure, military service-related injury, and less-than-honorable military discharge were highest in Internalizing and High Psychopathology classes. Results attest to the importance of person-centered mental health care within correctional settings and suggest a “treatment track” or “step-based” approach may best address the needs of individuals in these settings.
The U.S. correctional system is the largest mental health institution in the nation, responsible for the care of approximately 2.2 million persons in 2018, approximately half of whom are diagnosed with a mental health condition (Al-Rousan et al., 2017; Bureau of Justice Statistics, 2021a). Possibly due to limited treatment access in the community (Wilper et al., 2009), a recent analysis of incarcerated adults in Iowa suggested almost all mental health diagnoses were first made during incarceration, and many incarcerated adults suffered complex mental health presentations, with up to 25% facing three or more co-occurring conditions (Al-Rousan et al., 2017). Mental health is an essential consideration in correctional settings, because epidemiological data suggest incarcerated adults with mental health conditions are more likely to require potentially costly treatment, to be involved in physical or verbal altercations with correctional staff and other incarcerated persons, and to be charged with institutional misconduct (James & Glaze, 2006; Semenza & Grosholz, 2019). Given this substantial burden of mental health needs among incarcerated persons, understanding common patterns of mental health comorbidity and correlates of these patterns is vital to informing prevention and ongoing treatment efforts within correctional settings.
Within this context of correctional mental health, military Veterans are commonly considered a special interest group. Commensurate with their presence in the general population, military Veterans comprise 5% to 8% of people who are incarcerated (approximately 107,400 of 1,421,700 incarcerated persons in 2016; Maruschak et al., 2021). Veterans have unique experiences of serving in the military, placing them at increased risk of exposure to potentially traumatic events, such as service-related injuries and deployment to combat zones and disaster relief efforts. Veterans are also more likely than civilians to have experienced a range of nonmilitary traumas, such as adverse childhood experiences (Blosnich et al., 2021; Katon et al., 2015) and interpersonal violence both inside and outside the home (Lehavot et al., 2018; Marshall et al., 2005).
These exposures to trauma could have notable implications for Veteran mental health, particularly in correctional settings (Aronson et al., 2020; Ross et al., 2018), where estimates using convenience samples suggest approximately 90% of Veterans report at least one form of prior trauma exposure (Hartwell et al., 2014; Saxon et al., 2001). Some evidence suggests Veterans’ experiences of trauma, both within and outside of military service, may be associated with greater risk of criminal justice involvement (Edwards et al., 2022). Furthermore, elevated rates of psychopathology are generally observed among justice-involved Veterans; across studies, approximately 4% to 39% of justice-involved Veterans are affected by posttraumatic stress disorder (PTSD), 21% to 71% by substance use disorders, 10% to 51% by depressive disorders, 14% to 51% by anxiety disorders, 3% to 11% by bipolar disorders, 8% to 61% by adjustment disorders, 4% to 14% by psychotic disorders, and 3% to 11% by personality disorders (Blodgett et al., 2015). Attesting to their unique needs, incarcerated Veterans are also more likely than their civilian counterparts to meet criteria for PTSD, substance use disorders, and traumatic brain injury (Blodgett et al., 2015; Taylor et al., 2020), to be older, White, with more formal education, and to be convicted of violent and driving under the influence offenses (Marshall et al., 2005; Mumola, 2000; Noonan & Mumola, 2007; White et al., 2012).
Numerous Veteran-specific programs aim to better support Veterans involved with the criminal justice system. For example, some correctional institutions have created “Veterans Service Units,” specialized housing units informed by military culture that offer Veteran-specific mental health services, peer support, reentry counseling, and other services (Goggin et al., 2018). Outside of correctional settings, Veterans treatment courts provide eligible Veterans a treatment alternative to incarceration (Slattery et al., 2013). Also, services based in the Department of Veterans Affairs (VA), such as Veterans Justice Outreach and Healthcare for Reentry Veterans, are specifically designed to engage VA-eligible justice-involved Veterans with mental health, substance use, and case management services (Finlay et al., 2016; 2017). Despite the immense burden of mental health within these programs (Finlay et al., 2016; 2017; Goggin et al., 2018), however, there is an extreme dearth of evidence-based treatments specifically tailored to the distinctive characteristics of justice-involved Veterans (Blonigen et al., 2018).
Effectively tailoring evidence-based treatments to the needs of justice-involved Veterans requires a firm understanding of the mental health needs of this population. Substantial research has investigated diagnostic prevalence in justice-involved Veterans (for a review, see Blodgett et al., 2015). However, comparatively little is known about patterns of diagnostic comorbidity within this population. For example, preliminary evidence suggests high rates of comorbidity between substance use and other mental health conditions (Blodgett et al., 2015); however, patterns of comorbidity between non-substance use disorders in incarcerated Veteran samples (e.g., between PTSD and depression) has not been thoroughly characterized.
Relatively stable patterns of mental health comorbidity have been observed among community civilians (El-Gabalawy et al., 2013; Weich et al., 2011) and Veterans (Edwards et al., 2021). Across several large-scale samples, investigations of comorbidity structure have yielded strong evidence for broad, overarching dimensions (i.e., spectra) of psychopathology: internalizing, externalizing, and thought disorder (for reviews, see Kotov et al., 2021; Krueger et al., 2018). The internalizing spectrum is characterized by depressive, anxiety, and trauma-related disorders; externalizing by substance use and disruptive behavior disorders; and thought disorder by bipolar and depressive disorders with psychosis, cluster A personality disorders, and schizophrenia spectrum conditions (Kotov et al., 2020; 2021). Likewise, in studies utilizing latent class analysis, four patterns of psychopathology comorbidity are typically observed—a “healthy” pattern characterized by low rates of psychopathology; an “internalizing” pattern characterized by depressive and anxiety disorders; an “externalizing” pattern characterized by substance use and impulse-control disorders; and a “high pathology” pattern characterized by elevated rates across various mental health conditions (Edwards et al., 2021; El-Gabalawy et al., 2013; Weich et al., 2011). Some evidence suggests Veteran incarceration is most closely associated with the “high pathology” pattern (Edwards et al., 2021). However, generalization of these patterns to correctional settings has not yet been investigated.
Tailoring evidence-based treatments to the needs of justice-involved Veterans may also greatly benefit from understanding relations between mental health and military-specific experiences, such as combat exposure, service-related injury, and discharge type. A notable body of literature has linked combat exposure to the mental health of incarcerated Veterans, with results highlighting the need for trauma-informed care in this population (Finlay et al., 2019). A comparably smaller body of literature has examined other aspects of military experience, such as branch of service, length of service, and discharge type, to overall rates of justice involvement (e.g., Brooke & Gau, 2018). Unfortunately, these latter results are often inconclusive or inconsistent across samples (Finlay et al., 2019). Some military experiences—particularly the presence of a service-related injury and discharge type—are closely associated with eligibility for Veteran benefits (Veterans Benefits Administration [VBA], 2021). If these experiences are related to justice-involved Veterans’ mental health needs, it could have notable implications on the availability and/or accessibility of needed treatments and services for this population.
Current Research
To further understanding of the mental health needs of incarcerated Veterans and thereby inform ongoing efforts to support this population, the current research used latent class analysis to characterize patterns of mental health comorbidity in a large sample of adults incarcerated in state and federal prison to examine the relations of these patterns to Veteran status, military service-related variables (i.e., combat exposure, presence of a service-connected injury, discharge type), and treatment-related variables (i.e., history of psychiatric hospitalization and receipt of psychiatric medication, psychological treatment, and anger management while incarcerated). Informed by previous research, we hypothesized four latent patterns of mental health comorbidity would be present: (a) a pattern characterized by low psychopathology, (b) an internalizing pattern characterized by high rates of depressive and anxiety disorders, (c) a disinhibited or externalizing pattern characterized by high rates of substance use, and (d) a high psychopathology class characterized by high comorbidity across disorders. We also hypothesized the latter three classes would be associated with more extensive treatment histories relative to the first class. Finally, we hypothesized that a disproportionately higher rate of Veterans would be present in classes characterized by PTSD and/or substance use.
Method
Data for the current analyses were drawn from the 2016 Survey of Prison Inmates (Bureau of Justice Statistics, 2021b), a large, nationally representative sample of adults incarcerated in U.S. state and federal correctional facilities in 2016. The Survey of Prison Inmates serves to provide national statistics on various characteristics of incarcerated adults, including, among other things, demographics and mental health history and treatment.
Participants
The sample for the 2016 Survey of Prison Inmates was drawn from 2,001 unique prisons (1,808 state and 193 federal). A total of 24,848 adults, including 20,064 incarcerated in state facilities and 4,784 incarcerated in federal facilities, participated in the survey, comprising a response rate of 70% (69.3% among those in state facilities and 72.8% among those in federal facilities). Weighting procedures to account for sampling design allow the data to represent the total 1,502,671 individuals estimated to be housed in the United States during the data collection period. Further information about sampling and weighting procedures can be found elsewhere (see Bureau of Justice Statistics, 2021b).
Procedure
Data were collected from January through October of 2016 via face-to-face interviews using computer-assisted personal interviewing (CAPI). In CAPI interviews, interviewers read questions aloud and enter respondent answers directly into a laptop computer, allowing for automation of skip patterns and routing criteria throughout the interview. Interviews lasted approximately 50 minutes on average, including 2 minutes for consent and 48 minutes for the survey. Interviews were conducted in English (94%) and Spanish (6%) in accordance with respondent language needs. Current analyses represent a secondary analysis of the 2016 Survey of Prison Inmates, approved by the Institutional Review Board at Teachers College, Columbia University.
Measures
Mental Health History
Respondents’ histories of mental health disorders (including bipolar, depressive, psychotic, anxiety, posttraumatic stress, and personality disorders) were assessed via self-report. Participants were asked, “Have you ever been told by a medical doctor or mental health professional, such as a psychiatrist or psychologist, that you had. . .” and responses were coded dichotomously to reflect the presence or absence of each disorder. History of alcohol use disorder was assessed through a series of self-report questions corresponding to Diagnostic and Statistical Manual of Mental Disorders (5th ed.; DSM-5; American Psychiatric Association, 2013) diagnostic criteria for alcohol use disorder. Consistent with DSM-5 diagnostic criteria, endorsement of three or more items was interpreted as reflecting a history of alcohol use disorder.
Service-Related Variables
Respondents’ Veteran status was assessed via self-report (“Have you ever served in the United States Armed Forces?”). Respondents reporting prior military service also completed follow-up questions about their branch of service, combat exposure (coded dichotomously), history of injury resulting from training or combat (coded dichotomously), and discharge type. To aid in ease of interpretation, Veterans’ discharge types were coded as “honorable” (including both honorable and general under honorable conditions), “less-than-honorable” (including general without honorable conditions, other than honorable, bad conduct, and dishonorable), and “other.”
Treatment-Related Variables
Respondents’ history of mental health treatment was assessed via self-report, including any history of psychiatric hospitalization and receipt of psychiatric medication, counseling or psychotherapy, and anger management or conflict resolution classes since admission to prison. Each was coded dichotomously to reflect the presence or absence of history. Presence was coded as “1,” and absence was coded as “0.”
Analysis Plan
Descriptive statistics were first calculated to characterize the sample in terms of demographic categories, military service-related variables, mental health history, and treatment-related variables. Chi-square analyses also examined differences in demographic composition, mental health history, and treatment-related variables in Veterans versus non-Veterans, and correlation analyses examined bivariate associations between each mental health disorder, treatment-related variables, and military-service variables. Next, the three-step approach to latent variable modeling was used (Asparouhov & Muthén, 2014). For the first of these steps, latent class analysis using MPlus Version 8.5 (Muthén, & Muthén, 1998–2020) was used to determine the number and nature of latent mental health classes among all incarcerated persons (Veterans and non-Veterans). Seven predictors were included for these analyses, including bipolar, depressive, psychotic, anxiety, posttraumatic stress, personality, and alcohol use disorders. Analyses compared two to eight classes according to the Akaike information criterion, Bayesian information criterion (BIC), entropy statistics, and proportion of participants comprising the smallest class. Traditionally, best fitting models have comparatively lower BIC values, higher entropy values, and a significant number of members included within each class. For the second step, after the best-fitting model was selected, respondents were assigned to the latent class with the highest probability of fitting their mental health presentation. Finally, for the third step, Pearson’s Chi-square tests for independence were used to compare latent class membership to Veteran status, treatment-related variables, and military service-related variables.
Where applicable, cases were weighted in accordance with 2016 Survey of Prison Inmates weighting procedures to promote generalizability to incarcerated individuals within the United States as a whole. Consistent with previous research (e.g., Greenberg & Rosenheck, 2009), proportional down-weighting procedures were created by dividing the existing weight by the average number of incarcerated adults represented by each case (1,502,671/24,848 = 60.47). This allowed weighted sample sizes to differ little from the original sample size, adjusted only for sampling design, thereby ensuring statistical tests were not overly sensitive. Given the large sample size and multiple analyses, a more conservative alpha of .01 was also used to determine the statistical significance of statistical results.
Results
Descriptive Statistics
Demographically, respondents identified as predominantly male, heterosexual, White, and never married, with an average weighted age of 39.19 (SDwtd = 12.14). A weighted total of 1,775 (7.6%) respondents reported having served in the U.S. Armed Forces. Compared with their non-Veteran counterparts, these Veterans were more likely to be older, male, currently or previously married, White or Native American, and to have received diagnoses of mood, posttraumatic stress, or anxiety disorders. Most incarcerated Veterans reported prior service in the Army, Navy, or Marine Corps. Approximately 74% had a military discharge that was “honorable” or “general under honorable conditions,” 31% reported at least some combat exposure, and 29% reported experiencing a service-related injury. Comprehensive demographic, mental health, and service-related information is listed in Table 1.
Demographic and Mental Health Characteristics of Incarcerated Adults
Note. Veteran versus non-Veteran univariate group differences of p < .05 noted by asterisk(*).
Correlation analyses suggested that each mental health disorder was associated with increased likelihood for all identified modes of treatment, with strongest associations typically observed for psychiatric medications and lowest associations observed for anger management. In contrast, service-related variables had low associations with treatment-related variables and all mental health disorders, except PTSD. See Table 2.
Weighted Correlation Analyses
Note. PTSD = posttraumatic stress disorder.
p ≤ .01.
Latent Class Analysis
Seven indicators—bipolar, depressive, psychotic, posttraumatic stress, anxiety, personality, and alcohol use disorders—were included in latent class analyses to characterize the mental health classes of all incarcerated persons (Veterans and non-Veterans). Due to the ambiguity and low rates of “other emotional condition” disorder diagnoses, these diagnoses were not included as an indicator. Latent class analyses with two to eight classes were completed. Of these, the four-class solution was the best fitting and included: (a) Low Psychopathology (70% of the total sample), (b) Internalizing + Thought Disorder (8%), (c) Internalizing (14%), and (d) High Psychopathology (8%). Tests of conditional independence using the Cochran–Mantel–Haenszel statistic also supported a four-class solution, suggesting the strength of associations between indicators was reduced substantially after accounting for class membership. In contrast to civilian samples (e.g., Edwards et al., 2021; El-Gabalawy et al., 2013; Weich et al., 2011), no class characterized by substance use emerged from analyses. See Tables 3 and 4.
Summary of LCA Models With Two to Eight Classes
WNote. LCA = latent class analysis; AIC = Akaike information criterion; BIC = Bayesian information criterion; italicized class identified as best fitting.
Four Class Model Summary
Note. PTSD = posttraumatic stress disorder.
The Low Psychopathology class was characterized by generally low rates of psychopathology and slightly elevated rates of alcohol use disorder. On average, incarcerated persons in this class had low rates of mental health diagnoses (M = 0.33, SD = 0.53, range = 0-2), and 70% had no mental health diagnosis. By contrast, incarcerated persons in the Internalizing + Thought Disorder class had elevated rates of bipolar, depressive, and psychotic disorders, though only marginally higher rates of personality and alcohol use disorders relative to those in the low psychopathology class. Respondents had, on average, 2.15 (SD = 0.85, range = 1-5) disorders. In the Internalizing class, incarcerated persons had higher rates of mood, anxiety, and PTSDs; again, rates of alcohol use disorder were similar to those of other classes. Members of this class had an average of 3.01 (SD = 0.86, range = 1-5) disorders, with over 99% having two to five disorders. Finally, the High Psychopathology class was characterized by elevated rates of all forms of psychopathology; members of this class had an average of 5.03 (SD = 0.87) disorders, with over 99% having four or more disorders. See Figure 1.
Incarcerated Adult Mental Health, 4 Class Solution
Comparing demographic composition across classes, results suggested members of the Low Psychopathology were more likely to be Black or Hispanic, male, and heterosexual relative to other classes. Members of the Internalizing + Thought Disorder class were more likely to be Black, sexual minority, and never married. Those of the Internalizing class were more likely to be White, female, sexual minority, and divorced. Finally, members of the High Psychopathology class were slightly more likely to be female, White, American Indian/Alaska Native, and divorced or separated. These demographic patterns were notably similar across Veteran and non-Veteran respondents, though the magnitude of between-class differences was generally smaller among Veteran respondents.
Latent Class Membership and Military Service
After assigning cases to the best-fitting latent class, chi-square analyses suggested class membership significantly differed by Veteran status, χ(3) = 62.54, p < .001, w = 0.05, combat exposure, χ(3) = 47.63, p < .001, w = 0.17, service-related injury χ(3) = 55.90, p < .001, w = 0.18, and coded military discharge type, χ(6) = 28.10, p < .001, w = 0.13. Specifically, incarcerated Veterans were notably more likely to be in the Internalizing class relative to incarcerated non-Veterans. Among Veterans, a history of combat exposure and service-related injuries were associated with greater likelihood of membership in the Internalizing or High Psychopathology classes and lower likelihood of membership in the Low Psychopathology or Internalizing + Thought Disorder classes. Veterans with less-than-honorable or medical discharges were also more likely to be in the High Psychopathology class than other classes. See Figure 2.
Class Membership by Military Service Characteristics
Latent Class Membership and Treatment History
Class membership also significantly differed by all forms of mental health treatment history, including prior psychiatric hospitalization, χ(3) = 5,549.85, p < .001, w = 0.49, and receipt of psychiatric medication, χ(3) = 7,099.53, p < .001, w = 0.55, psychotherapy, χ(3) = 5,200.75, p < .001, w = 0.47, and anger management, χ(3) = 47.01, p < .001, w = 0.04, while incarcerated. Overall, members of the High Psychopathology class were more likely than members of other classes to have a history of these treatments; members of the Internalizing + Thought Disorder and Internalizing classes had moderate rates of treatment history; and members of the Low Psychopathology class were least likely to have such history. Whereas class differences in treatment history were pronounced for histories of hospitalization, medication, and psychotherapy, rates of participation in anger management classes were largely similar across classes, ranging from 34% to 41%. See Figure 3. Notably, these patterns of association between class membership and treatment history were largely similar when examining only the subset of respondents with a history of military service (psychiatric hospitalization [χ[3) = 324.82, p < .001, w = 0.43); medication (χ(3) = 601.64, p < .001, w = 0.58); psychotherapy (χ(3) = 406.57, p < .001, w = 0.48); anger management (χ[3] = 11.57, p = .01, w = 0.08)].
Class Membership by Mental Health Treatment History
Discussion
This study examined patterns of mental health comorbidity in a large sample of adults incarcerated in state and federal prisons. Results yielded three primary findings: First, in partial support of our first hypothesis, results reflect a partial replication of findings from community samples, with four latent mental health classes among incarcerated adults: (a) Low Psychopathology, characterized by low rates of psychopathology; (b) Internalizing + Thought Disorder, characterized by elevated rates of bipolar, depressive, and psychotic disorders; (c) Internalizing, characterized by elevated rates of mood, anxiety, and PTSDs; and (d) High Psychopathology, reflected by elevated rates of psychopathology across disorders. Second, in partial support of our second hypothesis, the latter three classes were associated with significantly higher rates of prior psychiatric hospitalization, psychiatric medication while incarcerated, and psychotherapy or counseling while incarcerated, though similar rates of anger management while incarcerated, in comparison to the Low Psychopathology class. Finally, in support of our final hypothesis, incarcerated Veterans were disproportionately represented in the Internalizing class.
Mental Health in Incarcerated Samples
Descriptive statistics suggested depressive and alcohol use disorders to be most common among incarcerated adults, each affecting approximately one in four incarcerated adults, whereas psychotic and personality disorders were least common, each affecting approximately one in 10 incarcerated adults. Although prevalence rates were largely comparable to previous research, current analyses suggested a bipolar disorder prevalence of 22% compared with 1% to 16% in previous research (Prins, 2014). Further research is needed to clarify the cause of this discrepancy. Likely causes include differences in mode of assessment, stringency of diagnostic method, and validity of participant self-report.
In the field of quantitative nosology, structural investigations of diagnostic comorbidity have identified a robust set of hierarchical dimensions of psychopathology, including the internalizing, externalizing, and thought disorder spectra (Kotov et al., 2021; Krueger et al., 2018). Although structural models are derived using factor analytic methods, similar mental health patterns are observed in community samples using latent class analysis—a “healthy” pattern characterized by low rates of psychopathology; an “internalizing” pattern characterized by depressive and anxiety disorders; an “externalizing” pattern characterized by substance use and impulse-control disorders; and a “high pathology” pattern characterized by elevated rates across various mental health conditions (Edwards et al., 2021; El-Gabalawy et al., 2013; Weich et al., 2011). Despite consistency across studies using community samples, results of the current study suggest this pattern may only partially generalize to correctional populations. Like community samples, most incarcerated adults had a Low Psychopathology mental health presentation, characterized by generally low rates of psychopathology. Consistent with overall elevated rates of psychopathology typically observed in correctional settings (Blodgett et al., 2015; Prins, 2014), members of the Low Psychopathology class comprised only 70% of this incarcerated sample in comparison to 69% to 84% in community samples (Edwards et al., 2021; El-Gabalawy et al., 2013; Weich et al., 2011). Demographically, members of this class were more likely to be Black or Hispanic, male, and heterosexual relative to other classes. Although further research is necessary to determine the potential cause(s) of this demographic composition, such results may reflect the tendency to underestimate and/or under-treat the mental health needs of these demographic populations due to stigma, cultural considerations, or other barriers to care (Cook et al., 2017).
Correspondingly, members of the Low Psychopathology class were unlikely to have a history of mental health treatment, reiterating the minimal mental health needs of these incarcerated adults. However, nearly 20% of incarcerated adults in the Low Psychopathology class met the criteria for alcohol use disorder, and 34% participated in anger management programming while incarcerated. These results may suggest a pervasiveness of disinhibited, externalizing mental health concerns among incarcerated samples that were unable to emerge from latent class analyses due to limited diagnostic criteria and/or inseparability across classes. Results also highlight the prominence of substance use and anger management difficulties within correctional settings, even in the absence of other mental health needs. These findings are consistent with the Risk-Need-Responsivity Model of criminal behavior, which suggests antisocial personality and attitudes (commonly targeted in anger management programming) and substance use are associated with criminal behavior independent of other forms of mental health (Bonta & Andrews, 2007).
The next two classes, Internalizing + Thought Disorder and Internalizing, most closely correspond with “internalizing” classes in community samples. Though internalizing disorders (e.g., depression, anxiety) are typically collapsed into one mental health class in community samples (Edwards et al., 2021; El-Gabalawy et al., 2013; Weich et al., 2011), the emergence of two “internalizing”-type classes in the current analysis is consistent with patterns of comorbidity surrounding these diagnoses. For members of the Internalizing + Thought Disorder class, bipolar and depressive disorders were most prevalent, followed by psychotic disorders, and some personality disorder comorbidities. This accords with a long line of research demonstrating that manic symptoms can be conceptualized as belonging to both internalizing and thought disorder spectra (Keyes et al., 2013; Kotov et al., 2020). Correspondingly, across samples, an extremely strong relationship is routinely observed between bipolar disorder and psychotic disorders due to overlapping features, such as positive symptoms, hyperactive cognition, and disrupted sleep patterns (Cicero et al., 2022; Goldberg et al., 2009). By contrast, members of the pure Internalizing class were somewhat less likely to have a psychotic or personality disorder comorbidity, but more likely to have co-occurring anxiety or PTSDs. A large majority of members in these classes had two-to-five co-occurring mental health disorders, further highlighting the mental health nuances of incarcerated adults struggling from these disorders. Demographically, members of these classes were more likely to be female and sexual minorities, consistent with research suggesting a notable association between internalizing symptoms and minority stress (Dyar et al., 2020; Feinstein et al., 2017).
In chi-square analyses, members of the Internalizing and Internalizing + Thought Disorder classes had largely similar rates of prior mental health treatment, with members of the Internalizing + Thought Disorder class slightly more likely to have a history of inpatient treatment (i.e., hospitalization) and members of the Internalizing class slightly more likely to have a history of outpatient treatment (i.e., psychiatric medication or psychotherapy/counseling while incarcerated). In contrast to community samples, these classes cumulatively comprised approximately 22% of the current sample (vs. 7-12% in community samples; Edwards et al., 2021; El-Gabalawy et al., 2013; Weich et al., 2011). These results underscore the importance of comprehensive, specialized services for incarcerated adults struggling with internalizing disorders despite stereotypes that characterize incarcerated samples by externalizing disorders (e.g., substance use, antisocial personality).
Similar to community samples, a High Psychopathology class was also identified within this incarcerated sample. Proportionally, this class comprised slightly more members than is generally seen in community samples (8% vs. 1.6% to 8%; Edwards et al., 2021; El-Gabalawy et al., 2013; Weich et al., 2011). This class was characterized by extremely high rates of various forms of psychopathology, and more than 99% of members had four or more co-occurring disorders. Diagnostically, members of this class were most likely to report a previous diagnosis of bipolar, depressive, or anxiety disorder (each ≥ 80%). Compared with the prevalence of disorders in other classes, rates of personality and psychotic disorders were also particularly elevated in the High Psychopathology class. The presence of a personality or psychotic disorder diagnosis may, therefore, serve as a proxy for psychopathological complexity in correctional settings. Consistent with this, these disorders are typically associated with extreme elevations in comorbidity rates (Cassano et al., 1998; Tomko et al., 2014). Individuals with psychotic and personality disorders also routinely experience misdiagnosis and/or false communication of diagnostic information (e.g., communicating an anxiety disorder diagnosis rather than a personality disorder diagnosis) due to concerns around mental health stigma or client reaction (Lequesne & Hersh, 2004; Milton & Mullan, 2014), thereby potentially increasing the number of unique diagnoses provided by health care professionals. Furthermore, although all forms of psychopathology were elevated in the High Psychopathology class and class membership was closely associated with treatment history, individual diagnoses showed only small to moderate associations with treatment history. Such findings reiterate the importance of a person-centered mental health approach above diagnosis-based approaches.
These results suggest mental health programming to support justice-involved persons should not adopt a “one size fits all” approach. Rather, programs may consider offering multiple “treatment tracks” or a “stepped-based approach,” with each track or step tailored to the unique needs of each mental health class. For example, a track corresponding to the Low Psychopathology mental health class would likely include opportunities to participate in anger management and substance use programming even in the absence of other forms of mental health care. A track corresponding to the Internalizing + Thought Disorder class would represent the next step of care and would add medication and evidence-based psychotherapy that targets disruptions to mood (e.g., cognitive behavior therapy for depression; Pardini et al., 2014) and/or destabilization of circadian rhythms and cognitive impairment (e.g., social rhythm therapy; Crowe et al., 2019). Next, a track corresponding to the Internalizing class would offer all supports of the Low Psychopathology and Internalizing + Thought Disorder tracks in addition to options for evidence-based services for anxiety and PTSDs (e.g., Seeking Safety; Najavits, 2002). At the highest step of care, a track corresponding to the High Psychopathology class would provide intensive medication and psychotherapy services capable of balancing multiple treatment targets and complex comorbidities (e.g., Dialectical Behavior Therapy; Tomlinson, 2018).
Given observed demographic differences across classes, programming should also consider the potential role of racial-based trauma and minority stress as contributing to psychopathology and/or mental health history. Helms and colleagues (2012) proposed that experiencing racism might serve as a catalyst to psychopathology, such as PTSD. Correspondingly, Veterans of color often have a higher prevalence of mental health diagnoses and greater risk of incarceration relative to White Veterans (Koo et al., 2015; Tsai et al., 2022). Culturally relevant interventions (e.g., group race-based stress and trauma intervention; Carlson et al., 2018) may help ameliorate the impact of race-based stressors on mental health both within and outside of the military. Overall, practitioners of trauma-focused psychotherapy should prioritize cultural competence and cultural humility to aid in addressing broader intersecting identities (Harvey & Tummala-Nara, 2007).
Mental Health of Incarcerated Veterans
To the authors’ knowledge, this study is the first to examine the mental health of incarcerated Veterans using latent class analysis. Findings therefore offer unique insight into mental health patterns of incarcerated Veterans, particularly within the context of the larger population of incarcerated persons and within the context of service-related factors. First, results provide additional evidence to suggest incarcerated Veterans have unique mental health needs relative to their non-Veteran peers. Replicating and expanding upon previous findings (Blodgett et al., 2015), univariate analyses suggested incarcerated Veterans reported higher rates of posttraumatic stress, anxiety, and depressive disorders and lower rates of bipolar disorders than incarcerated non-Veterans. Correspondingly, incarcerated Veterans were more likely to be members of the Internalizing class relative to their non-Veteran peers. Although we did not find support for our hypothesis that Veterans would be found in classes characterized by substance use, support for observed patterns of comorbidity extends beyond incarcerated Veterans. For example, in a nationally representative sample of U.S. Veterans, one study found comparatively weaker associations between substance use and PTSD relative to associations between PTSD, mood, anxiety, and personality disorders (Smith et al., 2016), suggesting these latter disorders may either increase susceptibility to PTSD or arise as consequences of trauma-related stress (Breslau, 2009; Pietrzak et al., 2011).
Class membership was also significantly associated with military-specific experiences, including combat exposure, service-related injuries, and discharge type in this sample. Specifically, Veterans in the Internalizing, and High Psychopathology classes were notably more likely to report histories of combat exposure, service-related injury, and less-than-honorable discharges relative to the other classes. Less-than-honorable and medical discharges were particularly high among members of the High Psychopathology class, consistent with previous research suggesting personality disorder features and negative affect tend to be associated with maladjustment to the military and receipt of less than honorable discharges (Fiedler et al., 2004; Lubin et al., 1999) .
These results have significant implications for mental health efforts for justice-involved Veterans, particularly those with less-than-honorable discharge types. A less-than-honorable discharge type disqualifies Veterans for many benefits to which they may otherwise be entitled, including VA-sponsored health care, educational benefits, and financial support offered through service-related disability compensation, loan programs, and pension programs (VBA, 2021). However, results suggest justice-involved Veterans with these discharge types are more likely to have mental health patterns characterized by substantial mental health complications and associated with extensive treatment needs and higher rates of combat exposure and service-related injuries. Given risk, VA and community-based programs intended to support the mental health of justice-involved Veterans are likely requisite. To ensure Veterans with less-than-honorable discharge types are connected to care, these programs should ensure access to health and social services is not impeded and that this substantial proportion of Veterans most in need of services are capable of accessing needed services.
Given the overrepresentation of service-related injuries among incarcerated Veterans with mental health complications, programs should also consider partnerships with specialized medical programs equipped to manage common service-related injuries, such as auditory disruptions (e.g., tinnitus, hearing loss), mobility limitations, back problems, sciatica, and migraines (VBA, 2021). Again, it is vital that medical services available through such partnerships also do not discriminate on the basis of discharge type, as many justice-involved Veterans with service-related injuries have less-than-honorable discharges. Finally, as routinely suggested throughout the literature (e.g., Finlay et al., 2019), given the elevated rates of combat exposure and overall trauma exposure in the Veteran population, programming should be trauma-informed regardless of the presence or absence of a PTSD diagnosis.
Limitations and Future Directions
The current study is characterized by several methodological strengths, particularly a well-defined, nationally representative sample and use of an analysis strategy that is largely novel for this area of research. Nevertheless, results should be considered within the context of a few methodological limitations. Most notably, the assessment of mental health diagnoses was reliant upon participants’ self-reported receipt of diagnoses by previous providers. This historical approach to prior diagnoses inhibits presumptions of temporal order. For example, without further information, it remains unclear whether elevated rates of PTSD in the Veteran sample are due to aspects of military service (e.g., combat-related PTSD) or experiences outside of military service (e.g., child abuse).
Furthermore, these self-report diagnoses were not verified through chart reviews, consultation with respondents’ previous providers, or completion of validated measures. As such, validity of respondents’ reports may be limited. It is common, for example, for diagnoses to change throughout the course of an illness or treatment (e.g., Person A’s diagnosis of major depressive disorder is changed to bipolar disorder after experiencing a manic episode; Person B receives inaccurate diagnoses of anxiety and bipolar disorders before being accurately diagnosed with a personality disorder), contributing to possible over-reporting of diagnoses. Consistent with this possibility, the cumulative prevalence of bipolar and depressive disorders—which are diagnostically mutually exclusive—exceeded 100% in the Internalizing + Thought Disorder, Internalizing, and High Psychopathology classes, implying that many members of these groups reported having received both diagnoses. Furthermore, some providers may fail to share accurate diagnostic information with clients due to concerns about the client’s reaction or stigma (Perkins et al., 2018), contributing to possible under-reporting of diagnoses, particularly for highly stigmatized disorders (e.g., personality and psychotic disorders). Notably, despite these possible threats to validity, the latent class structure of psychopathology in the current study was remarkably similar to that of previous research that used well-validated assessments of psychopathology (Edwards et al., 2021; El-Gabalawy et al., 2013; Weich et al., 2011), implying good overall construct validity of diagnostic assessment.
Relatedly, respondents’ military-specific experiences were also assessed using self-report. Some Veterans may have been hesitant to disclose their Veteran status due to fears of losing benefits (Beckerman & Fontana, 1989), failed to disclose less-than-honorable discharges due to shame or stigma, or failed to recognize injuries as resulting from military service. Consequently, results likely reflect an underestimation of the true prevalence of incarcerated Veterans and adverse military-specific experiences. Future research may therefore benefit from the administration of diagnostic instruments with sound psychometric properties and/or review of official records (e.g., DD Form 214 to verify Veteran status and discharge type) to corroborate respondent self-report. Gathering further information about the nature of service-related injuries may also be helpful, because certain injuries (e.g., traumatic brain injury) could have notable implications for mental health (Brenner et al., 2013).
Third, the method of assessing alcohol use disorders was different than that of other mental health disorders, possibly resulting in an underestimation of the true prevalence of alcohol use disorders among incarcerated adults. Consistent with this possibility, rates of alcohol use disorder were notably lower than typically observed in previous research (Blodgett et al., 2015), and no latent class emerged from analyses characterized by elevated rates of substance use, as typically seen in community samples (Edwards et al., 2021; El-Gabalawy et al., 2013; Weich et al., 2011). Relatedly, non-alcohol substance use disorders were not consistently assessed in the 2016 Survey of Prison Inmates and were therefore unable to be included in analyses. Future research would therefore benefit from further exploration of alcohol and other substance use disorders within the context of mental health among incarcerated persons.
Fourth, analyses only included data from the 2016 Survey of Prison Inmates. Generalization to other correctional settings (e.g., jails) or other forms of justice involvement (e.g., probation, parole, treatment courts) therefore remains a question for future research. Future research should investigate the generalizability of results by exploring replicability in other justice-involved samples.
Conclusion
This research is the first to examine latent patterns of mental health comorbidity in a large, nationally representative sample of incarcerated adults in state and federal prisons. Results suggest these latent classes are largely similar to that of community samples and include a large, Low Psychopathology class characterized by low rates of psychopathology, a small Internalizing + Thought Disorder class characterized by high rates of bipolar and depressive disorders, a moderately-sized Internalizing class characterized by elevated rates of mood, anxiety, and PTSDs, and a small High Psychopathology class characterized by elevated rates of psychopathology across disorders. The latter three classes were associated with more extensive treatment histories in comparison to the Low Psychopathology class. Incarcerated Veterans were disproportionately represented in the Internalizing class, with Veteran class membership closely associated with military-specific experiences, including combat exposure, service-related injuries, and discharge type. Results attest to the importance of person-centered mental health care within correctional settings and suggest a “treatment track” or “step-based” approach may best address the needs of individuals in these settings.
Footnotes
Authors’ Note:
Work for this article was supported by the Department of Veterans Affairs, Office of Academic Affiliations, VA Special MIRECC Fellowship Program in Advanced Psychiatry and Psychology, by the VISN-2 MIRECC, and by the VISN-19 MIRECC. The views expressed here are the authors’ and do not necessarily represent the views of the Department of Veterans Affairs.
