Criminogenic risk and suicidality among justice-involved homeless veterans with comorbid mental health and substance use disorders

Abstract

Background

Veterans involved in the criminal legal system (CLS) exhibit disproportionately high rates of comorbid mental health and substance use disorders (COD) and homelessness—conditions that increase both the risk of recidivism and suicide. Despite elevated risk, limited research has examined the relationship between criminogenic risk factors and suicidality in this population.

Objectives

This study examined differences in criminogenic risk, psychiatric burden, and social determinants of health (SDOH) among CLS-involved veterans with COD, comparing those with suicidality risk to those without.

Methods

Baseline data were analyzed from 127 veterans admitted to three Department of Veterans Affairs (VA) Mental Health Residential Rehabilitation Treatment Programs who were enrolled in a randomized controlled trial of the MISSION-CJ intervention (ClinicalTrials.gov Identifier: NCT04523337). Eligible veterans had recent CLS involvement and documented COD diagnoses. Assessments included the Level of Service Inventory–Revised (LSI-R), the Measures of Criminal Attitudes and Associates (MCAA), structured psychiatric interviews, standardized symptom inventories, and measures of social determinants of health. Suicidality risk was defined as experiencing recent thoughts of suicide and/or attempted suicide in their lifetime.

Results

Most veterans (80%) were classified as having moderate-high to high criminogenic risk based on total LSI-R scores. Compared to those without suicidality risk, veterans with suicidality risk demonstrated significantly higher criminogenic risk as evidenced by total LSI-R scores (p = .01), more antisocial attitudes (p = .04), particularly entitlement and violence beliefs (p = .02, p = .01, respectively), greater emotional and functional impairment (p < .001), and a higher prevalence of probable posttraumatic stress disorder (PTSD; 64% vs. 41%, p = .009). SDOH, including unstable housing and limited community integration, were highly prevalent but did not significantly differ between those with and without risk of suicide.

Conclusions

Among CLS-involved homeless veterans with COD, suicidality was associated with elevated criminogenic risk and greater psychiatric burden, particularly PTSD. These findings underscore the need for integrated interventions that address both criminogenic and behavioral health factors to support recovery, reduce suicide risk and mental health symptoms, and mitigate CLS recidivism.

Trial registration

This study was preregistered at https://ClinicalTrials.gov with registration number NCT04523337.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12889-026-26250-6.

Background

A substantial number of Veterans in the United States (US) are involved in the criminal legal system (CLS), including the approximately 146,000 Veterans released from correctional facilities each year [1] and additional Veterans who are diverted from incarceration during trial or sentencing. Relative to non-veterans, veterans involved in the criminal legal system (CLS) have high rates of substance use and mental health disorders [2–4],with 60% having a comorbid mental health and substance use disorder (COD) [5]. CODs are often associated with poor treatment engagement, homelessness, and higher rates of arrests and re-incarceration as compared to those with a single behavioral health disorder [6–10].

In addition to being disproportionately represented in the CLS, veterans also die by suicide at significantly higher rates than non-veterans. According to the US Department of Veterans Affairs (VA), veterans accounted for 13.7% of all suicides in 2022, despite comprising only about 6.5% of the US population [11]. This disparity makes suicide prevention among veterans a top priority for the VA [12]. Identifying social determinants of health (SDOH) that elevate suicide risk is essential to developing effective prevention and early intervention strategies. According to a recent review [13], 46 meta-analyses have identified CLS involvement as a robust SDOH factor associated with increased suicide risk. In fact, veterans with CLS invol12vement are over three times more likely to have attempted suicide in their lifetime compared to those without such involvement [14]. This association applies across the CLS continuum, including arrest, arraignment, adjudication, incarceration, and community re-entry. Despite these findings, examination of suicide risk among veterans with CLS involvement has predominately focused on clinical and social correlates [15–17] such as depressive and psychotic disorders [18], as well as loneliness [19]. These factors are also known as non-criminogenic risk factors (i.e., factors that have not been shown to increase risk of criminal recidivism). Suicidality among veterans with COD is also associated with combat stress and is more prevalent among veterans with post-traumatic stress disorder (PTSD) and depression [14].

Although clinical factors for suicide, such as depression and PTSD, are generally non-criminogenic risk factors, the offender rehabilitation literature has identified several “criminogenic needs” that are robust predictors of CLS recidivism. These needs are viewed as dynamic and amenable to treatment, thereby reducing CLS recidivism risk [20]. The Risk-Need-Responsivity (RNR) model is a well-established framework for understanding both dynamic and static risk factors that contribute to CLS recidivism. The three core principles of the RNR model include: (1) risk (matching the intensity of services and interventions to an individual’s level of risk for recidivism); (2) need (focusing rehabilitation efforts on risk factors that are robust predictors of CLS recidivism, and are modifiable); and (3) responsivity (tailoring services and interventions to characteristics of an individual that may impact engagement or response to treatment) [21]. Research shows that adherence to the core principles of the RNR model is linked to reduced CLS recidivism and improved behavioral health outcomes. Criminogenic needs include factors such as antisocial personality patterns, antisocial attitudes, antisocial peers, substance use, family or marital dysfunction, and employment instability [22–24].

Despite the high rates of suicide and COD among those involved in the CLS, research to date has not examined criminogenic risk and suicidality among veterans with a COD and CLS involvement. To address this gap, this study examined whether there were differences in criminogenic risk factors, risk level, psychiatric burden, and SDOH between veterans who report suicidality and those who do not. The findings of this study aim to inform treatment and discharge planning across various settings for this vulnerable population.

Methods

We conducted secondary analyses on a sample of 127 veterans who were admitted to one of the three VA Mental Health Residential Rehabilitation Treatment Programs (MH RRTPs) and enrolled in a randomized controlled trial of a wraparound treatment intervention for COD called “Maintaining Independence and Sobriety through Systems Integration, Outreach, and Networking–Criminal Justice” (MISSION-CJ). The study design, analyses, and outcomes were preregistered at https://ClinicalTrials.gov (NCT04523337) and described in a study protocol article [25]. Veterans were eligible for the trial if they (a) were entering an MH RRTP, at either of the three VA healthcare systems; (b) were arrested and charged and/or released from incarceration in the past five years; and (c) have a COD diagnosis in their VA medical record. Substance use and mental health diagnoses were confirmed during eligibility screening using the Structured Clinical Interview for DSM-5 (Diagnostic and Statistical Manual of Mental Disorders 5th Edition) [26]. Veterans were only excluded if they had cognitive impairment as assessed using the Montreal Cognitive Assessment’s [27] section on Orientation; veterans were excluded if they were unable to correctly answer the date and location items. All data reported here were drawn from a baseline interview administered by phone after consent and collected prior to randomization. Study procedures were approved by VA’s Central Institutional Review Board (20 − 01).

Study setting

MH RRTPs provide treatment and rehabilitation services to veterans to address their complex behavioral health needs (i.e., substance use and/or other mental health). Services are designed for improving functional status, sustaining treatment and rehabilitation gains, recovery, and community integration by providing 24/7 care using a structured residential environment. At each VA site, we recruited from an MH RRTP that is specifically for veterans experiencing homelessness. MH RRTPs are comparable across VA sites in terms of program structure and services delivered, program length (an average of three months), individual and group-based clinical approaches such as cognitive behavioral programming, and staffing (psychiatrists, psychologists, social workers, nurses, addiction therapists, vocational therapists, and homelessness coordinators).

Measures

Suicidality

Risk of suicidality was captured from two measures: (1) the Level of Service Inventory-Revised (LSI-R), a validated and reliable scale consisting of 54 items used to measure criminogenic risk and needs, which asks if participants had attempted suicide in their lifetime (yes/no) [28, 29], and (2) the Patient Health Questionnaire (PHQ-9) [30], a questionnaire used to assess the 9 DSM-5 criteria for major depression in the past two weeks. Risk of suicide was categorized dichotomously (1 = Suicidality Risk, 0 = No Suicidality Risk). Veterans were grouped into the Suicidality Risk category if they endorsed thoughts of suicide at least “several days” in the last two weeks on the PHQ-9 and/or indicated they had attempted suicide in their lifetime on the LSI-R. Of note, this combined suicidality indicator does not distinguish between current suicidal ideation and lifetime suicide attempts, which may reflect distinct risk profiles, and lethality or method of prior attempts was not available in the dataset.

Demographics

Demographics were measured via the Addiction Severity Index (ASI), a semi structured interview for patients in substance use treatment programs [26]. These factors include age, sex, average years of education, and race/ethnicity.

Non-criminogenic needs

Social Determinants of Health (SDOH)

The ASI [31] was also used to collect multiple SDOH factors including employment, marital status, whether the veteran lived in a controlled environment in the past 30 days (i.e., a hospital, jail, or other place intended to limit access to drugs/alcohol), and housing status (i.e., own residence, renting, living with friends/family, transitional housing, homeless/no stable living arrangement; or jail/incarcerated). A dichotomous variable representing housing stability was created using reported housing status where “unstable housing” was defined as experiencing homelessness or at risk of losing one’s housing. A Timeline Follow-Back Residential interview, a retrospective calendar-based measure, identified the duration and frequency of homelessness in the past 90 days (e.g., percent of days homeless). The Community Integration Measure (CIM) [32] is a 10-item valid and reliable measure of an individual’s integration into their home and community in terms of social support, productivity (vocational & leisure), and independent living. The CIM has a maximum score of 50, with higher scores indicating greater community integration. Similarly, the Community Assessment Inventory (CAI) [33] is a valid and reliable measure of social and community support and comprises a total score (maximum score 148) and several subscales; support from partner/family inside the home (maximum score 24), support from family outside the home (maximum score 40), support from friends (maximum score 32), and support from the community (maximum score 52), with greater scores indicating more social and community support.

Psychiatric symptom severity

The PTSD Checklist for DSM-5 (PCL-5) [34], a 20-item self-report measure that assesses symptoms [35–39] of PTSD from the Diagnostic and Statistical Manual of Mental Disorders, 5th Edition [40]. Items were rated on a 5-point scale (0 = not at all, 4 = extremely) and summed to create a total score (α = 0.98). PCL-5 total scores of 33 or more have been shown to be associated with a probable diagnosis of PTSD [41]. This cutoff was used to categorize veterans with and without probable PTSD into a binary variable in the present study. Although PCL-5 symptom severity can be evaluated continuously, we used the recommended cutoff score of 33 to indicate probable PTSD in order to provide a clinically interpretable baseline characterization of the sample. This approach is consistent with national guidelines and prior validation studies [35–39]. Depression severity was measured via the PHQ-9, a widely-used, reliable and valid self-report questionnaire to assess the 9 DSM-5 criteria for depression [30]. The Behavior and Symptom Identification Scale (BASIS-24) assessed functioning and psychiatric symptom severity across various mental health domains [42]. The BASIS-24 includes 24 items which are rated on a scale of 0 (“no difficulty”) to 4 (“extreme difficulty”). Total scores as well as each of the six subscale scores (Depression and Daily Functioning; Interpersonal Problems; Self-harm; Emotional Lability; Psychosis; and Substance use) are computed by averaging each response to the measure overall and respective to each subscale [42].

Psychiatric diagnoses

The Structured Clinical Interview for DSM-5 (SCID-5) [26] is a reliable and valid semi-structured interview used to diagnose mental health and substance use disorders and was administered at the outset of the baseline assessment to establish study eligibility for the presence of COD. In the current study, the SCID-5 was used to assess a broad range of mental health and substance use disorders, with a focus on the following diagnoses that tend to be common among patients in MH RRTPs: General Substance Use Disorder, Alcohol Use Disorder, Major Depressive Disorder, Generalized Anxiety Disorder, Panic Disorder, Social Anxiety Disorder, Agoraphobia, Psychotic Disorder, Specific Phobia, Gambling Disorder, Manic Episode, and Hypomanic Episode. The SCID-5 interview in this study did not include the full PTSD module; therefore, a structured interview–based PTSD diagnosis could not be derived.

Criminal history and criminogenic needs

Criminal history

All veterans were asked to report total lifetime convictions at the start of MH RRTP entry using the Level of Service Inventory–Revised (LSI-R)^29,30 The LSI-R is a widely used, validated, and reliable tool designed to assess these dynamic risk factors [28, 29]. The LSI-R generates an overall risk score by summing individual domain scores (described below), providing a comprehensive index of recidivism risk [43].

Criminogenic risk and needs

The LSI-R was also used to measure ten criminogenic domains comprising both static (criminal history) and dynamic risk factors for criminal recidivism: criminal history (nature and frequency of previous offenses), education/employment (employment history, employable skills, and educational background), financial (income, stability and economic distress), familial relationships (quality of familial relationships, and family members with CLS involvement), accommodations (living situation such as living with criminal associates and housing instability), leisure and recreation (time spent in pro-social activities), companions (tending to have close associates engaged in criminal activity), alcohol and drug use (measure of problematic substance use), emotional health (measure of mental health struggles), and attitudes/orientations (antisocial attitudes, beliefs, and values) [28, 29]. Scores across these domains are summed to create an overall index of recidivism risk; higher scores on all domains and the overall index are associated with a higher propensity to commit future criminal behavior [43]. LSI-R total scores can be categorized into different risk levels of recidivism (low (0–13), low-moderate (14–23), moderate (24–33), moderate-high (34–40), and high (41–54)) [44]. In addition to the LSI-R, criminogenic risk was also measured with the Measures of Criminal Attitudes & Associates (MCAA) [45]. The MCAA can be used alongside the LSI-R to gain greater insight into criminal thinking patterns [45] and provides a valid and reliable assessment of an individual’s (i) antisocial attitudes (Scale A) and (ii) antisocial associates (Scale B), which have been shown in meta-analyses to be among the best predictors of criminal recidivism [46, 47]. MCAA responses are summed to create a total score (maximum score 46), as well as four separate subscales; measure of attitudes towards violence (maximum score 12), antisocial intents (maximum score 12), attitudes towards entitlement (maximum score 12), and antisocial associates (maximum score 10).

Data analysis

Univariate descriptive statistics were calculated for the total sample (N = 127 veterans). In addition, bivariate analyses were conducted to examine differences between veterans for whom suicidality risk was and was not indicated. Independent samples t tests were computed for continuous variables, and chi-square tests were computed for categorical variables. We did not statistically correct for multiple comparisons (i.e., Bonferroni corrections). It should be noted that this approach increases the risk of Type I errors but reduces the risk of Type II errors [48]. Given the lack of research on associations between suicidality, criminogenic risk, and CLS involvement among veterans, we chose to protect against Type II errors and considered the current analyses as exploratory.

Results

Demographics

Table 1 includes demographic characteristics overall and by suicide risk category. Overall, veteran participants were predominantly male (97.64%), White/Caucasian non-Hispanic (52.38%), with an average age of 46 years old (M = 46.11, SD = 11.59). The average years of education for veterans at study enrollment was 13 years (M = 13.18, SD = 1.43), indicating most veterans had obtained at least a high-school level education in their lifetime. In total, 44.09% of veterans were categorized as having suicide risk and 55.91% were classified as not having suicide risk. Bivariate comparisons revealed a significant difference in the average age at enrollment based on suicide risk category. Veterans with suicide risk were significantly younger than veterans without suicide risk (p = .02, d = 0.41, 95% CI [0.06, 0.77]). No other significant differences on baseline demographics were noted between the suicide risk groups.

Table 1.

Demographics overall & by suicide risk category

Demographics	Overall (N = 127)	Suicide Risk (N = 56)	No Suicide Risk (N = 71)	p	ES	95% CI
Age M (SD)	46.11 (11.59)	43.44 (11.24)	48.20 (11.51)	0.02	0.41	[0.06, 0.77]
Sex N (%)				0.71	0.03
Male	124 (97.64)	55 (98.21)	69 (97.18)
Female	3 (2.36)	1 (1.79)	2 (2.82)
Lifetime Education Years M (SD)	13.18 (1.43)	13.12 (1.42)	13.23 (1.45)	0.71	-0.02	[-0.36, 0.33]
Race/Ethnicity N (%)				0.08	0.33
White non-Hispanic	66 (52.38)	25 (44.64)	41 (58.57)
Black non-Hispanic	22 (17.46)	9 (16.07)	13 (18.57)
American Indian	2 (1.58)	0 (0.00)	2 (2.85)
Alaskan Native	1 (0.79)	1 (1.79)	0 (0.00)
Asian/Pacific Islander	5 (3.96)	4 (7.14)	1 (1.42)
Hispanic	19 (15.07)	13 (23.22)	6 (8.59)
Other	11 (8.76)	4 (7.14)	7 (10.00)

For variables with any cell less than a count of 5, Fisher’s exact test was used. For effect size (ES), Cohen’s d is reported for continuous variables, and Cramer’s V is reported for categorical variables. CI confidence interval

Non-criminogenic needs

Table 2 documents non-criminogenic risk factors for our sample of veterans. In terms of SDOH, a majority (82.68%) of veterans were not paid for a job in the 30-days prior to enrollment. Most veterans were unstably housed at baseline, with over 80% of veterans residing in transitional housing. Moreover, veterans indicated spending on average 72 out of the last 90 days homeless (M = 72.38, SD = 31.37). Overall, veterans reported poor to moderate community integration (M = 20.60, SD = 8.45) and rated their social and community support as low to moderate (M = 87.85, SD = 15.81), particularly in terms of support from partner/family inside the home (M = 16.67, SD = 4.61), support from family outside the home (M = 20.89, SD = 10.67), support from friends (M = 20.51, SD = 5.94), and support from community (M = 29.76, SD = 5.66). Bivariate analyses did not reveal any significant differences between suicide risk groups on SDOH measures.

Table 2.

Non-criminogenic risk factors overall & by suicide risk category

Social Determinants of Health	Overall (N = 127)	Suicide Risk (N = 56)	No Suicide Risk (N = 71)	p	ES	95% CI
Paid Job in the Past 30-days N (%)				0.12	0.14
Yes	22 (17.32)	13 (23.21)	9 (12.68)
No	105 (82.68)	43 (76.79)	62 (87.32)
Housing Placement Past 30-days N (%)				0.19	0.22
Transitional Housing	104 (81.89)	45 (80.36)	59 (83.10)
Jail/incarcerated	12 (9.45)	3 (5.36)	9 (12.68)
Homeless/No Stable Environment	6 (4.72)	4 (7.14)	2 (2.82)
Living with Family/Friends	3 (2.36)	2 (3.57)	1 (1.41)
Own Residence	2 (1.57)	2 (3.57)	0 (0.00)
Days Homeless Past 90-days M (SD)	72.38 (31.37)	71.98 (31.41)	72.70 (31.55)	0.89	0.02	[-0.32, 0.37]
Community Integration Measure (CIM) M (SD)	20.60 (8.45)	21.37 (9.54)	20.00 (7.49)	0.36	-0.16	[-0.51, 0.19]
Community Assessment Inventory (CAI) M (SD)
Total	87.85 (15.81)	88.94 (17.82)	87.00 (14.08)	0.49	-0.12	[-0.47, 0.23]
Support Partner/Family Inside the Home	16.67 (4.61)	17.30 (4.34)	16.18 (4.78)	0.17	-0.24	[-0.59, 0.11]
Support Family Outside the Home	20.89 (10.67)	21.67 (10.51)	20.28 (10.82)	0.46	-0.13	[-0.48, 0.22]
Support from Friends	20.51 (5.94)	20.87 (6.09)	20.23 (5.86)	0.55	-0.11	[-0.45, 0.24]
Support from Community	29.76 (5.66)	29.08 (7.25)	30.29 (3.95)	0.23	0.21	[-0.14, 0.56]
Psychiatric Symptom Severity
Post-Traumatic Disorder Checklist (PCL-5) M (SD)	34.02 (19.07)	37.07 (19.89)	31.61 (18.19)	0.11	-0.28	[-0.63, 0.07]
Probable PTSD based on PCL-5 Score N (%)	65 (51.18)	36 (64.29)	29 (40.85)	0.009	0.23
Patient Health Questionnaire (PHQ-9) M (SD)	9.70 (6.01)	10.62 (6.09)	8.97 (5.88)	0.12	-0.27	[-0.62, 0.08]
BASIS-24
Total M (SD)	1.44 (0.36)	1.52 (0.37)	1.37 (0.34)	0.02	-0.41	[-0.77, -0.06]
Depression/Daily Functioning M (SD)	1.74 (0.72)	1.89 (0.76)	1.63 (0.67)	0.04	-0.32	[-0.72, -0.02]
Interpersonal Problems M (SD)	2.68 (0.79)	2.73 (0.75)	2.64 (0.83)	0.53	-0.11	[-0.46, 0.24]
Self-harm M (SD)	0.11 (0.40)	0.17 (0.52)	0.06 (0.27)	0.13	-0.27	[-0.62, 0.08]
Emotional Lability M (SD)	1.81 (0.95)	1.93 (0.93)	1.71 (0.95)	0.19	-0.23	[-0.58, 0.12]
Psychosis M (SD)	0.49 (0.73)	0.57 (0.82)	0.43 (0.64)	0.28	-0.19	[-0.54, 0.16]
Substance Use M (SD)	0.82 (0.68)	0.83 (0.65)	0.80 (0.71)	0.81	-0.04	[-0.39, 031]
Psychiatric Diagnoses (SCID-5)
Diagnoses N (%)
General Substance Use Disorder	84 (68.29)	38 (71.10)	46 (65.71)	0.48	0.06
Alcohol Use Disorder	69 (56.10)	26 (49.06)	43 (61.43)	0.17	0.12
Major Depressive Disorder	60 (49.18)	29 (55.77)	31 (44.29)	0.21	0.11
Generalized Anxiety Disorder	60 (48.78)	25 (47.17)	35 (50.00)	0.75	0.03
Panic Disorder	47 (38.21)	24 (45.28)	23 (32.86)	0.16	0.13
Social Anxiety Disorder	30 (24.39)	14 (26.42)	16 (22.86)	0.65	0.04
Agoraphobia	20 (16.26)	7 (13.21)	13 (18.57)	0.42	0.07
Psychotic	18 (14.63)	7 (13.21)	11 (15.71)	0.69	0.03
Specific Phobia	14 (11.38)	6 (11.32)	8 (11.43)	0.98	0.01
Gambling Disorder	8 (6.50)	6 (11.32)	2 (2.86)	0.06	0.17
Manic Episode	6 (4.88)	3 (5.66)	3 (4.29)	0.73	0.03
Hypomanic Episode	4 (3.25)	1 (1.89)	3 (4.29)	0.46	0.06

For variables with any cell less than a count of 5, Fisher’s exact test was used. For effect size (ES), Cohen’s d is reported for continuous variables, and Cramer’s V is reported for categorical variables. CI confidence interval

Overall, veterans reported an average PCL-5 score of 34.02 (SD = 19.07), and over half (51.18%) of veterans met provisional criteria for a PTSD diagnosis based on their PCL-5 score. Moreover, significantly more veterans with suicide risk (64.29%) met provisional criteria for PTSD at intake than veterans without suicide risk (40.85%, p = .009, Cramer’s V = 0.23). In terms of mental health functioning, veterans’ average total BASIS-24 score was 1.44 (SD = 0.36), indicating moderate difficulty in mental health functioning, with difficulties greatest for Interpersonal Problems (M = 2.68, SD = 0.79). Bivariate analyses revealed that veterans with suicide risk had significantly higher total BASIS-24 scores on average compared to veterans without suicide risk (p = .02, d = -0.41, 95% CI [-0.77, -0.06]). Veterans with suicide risk were also observed to have significantly higher problems with Depression and Daily Functioning compared to veterans without suicide risk (p = .04, d = -0.32, 95% CI [-0.72, -0.02]).

On average, veterans reported mild to moderate depression via the PHQ-9 (M = 9.70, SD = 6.01), with no observed difference among suicide risk groups (p < .05). Overall, SCID-5 data indicated that the most common mental health diagnoses at baseline were General Substance Use Disorder (68.29%), Alcohol Use Disorder (56.10%), Major Depressive Disorder (49.18%), Generalized Anxiety Disorder (48.78%), Panic Disorder (38.21%), Social Anxiety Disorder (24.39%), Agoraphobia (16.26%), Psychosis (14.63%), Specific Phobia (11.38%), Gambling Disorder (6.50%), Manic Episode (4.88%), and Hypomanic Episode (3.25%). No significant differences were noted in mental health diagnoses between suicide risk groups.

Criminogenic risk and needs

Table 3 reports the criminal history and criminogenic risk profile of veterans overall and by suicide risk category. Overall, veterans experienced an average of 7 lifetime convictions (M = 7.74, SD = 14.04), with no significant differences between suicide risk groups. Table 3 also outlines the criminogenic risk profile according to the LSI-R, both raw domain scores and risk level based on the total score. In total, 79.53% of veterans were categorized as moderate to high risk in terms of their criminogenic risk at baseline. Bivariate analyses revealed that 87.5% of veterans with suicide risk were categorized as moderate to high criminogenic risk, while 73.24% of veterans without suicide risk were categorized as moderate to high criminogenic risk. Despite approaching significance, bivariate analyses did not reveal a significant difference in criminogenic risk levels (p = .11, Cramer’s V = 0.24). Overall, total LSI-R scores were consistent with the risk level categories, with veterans scoring moderate to high in their average score (M = 28.19, SD = 6.87). Despite bivariate analyses not demonstrating a significant difference between LSI-R risk levels, there was a significant difference in total criminogenic risk LSI-R scores such that veterans with suicide risk scored significantly higher compared to those without suicide risk (p = .01, d = -0.45, 95% CI [-0.81, -0.09]).

Table 3.

Criminogenic risk and need overall & by suicide risk category

Criminal History and Criminogenic Risk and Needs	Overall (N = 127)	Suicide Risk (N = 56)	No Suicide Risk (N = 71)	p	ES	95% CI
Criminal History (ASI)
Lifetime Convictions M (SD)	7.74 (14.04)	8.69 (17.18)	6.98 (11.03)	0.49	-0.12	[-0.47, 0.22]
Criminogenic Risk Level (LSI-R) N (%)				0.11	0.24
Low Risk	3 (2.36%)	0 (0.00%)	3 (4.23%)
Low-Moderate Risk	23 (18.11%)	7 (12.50%)	16 (22.54%)
Moderate Risk	73 (57.48%)	32 (57.14%)	41 (57.75%)
Moderate-High Risk	19 (14.96%)	11 (19.64%)	8 (11.27%)
High Risk	9 (7.09%)	6 (10.72%)	3 (4.21%)
Criminogenic Risk and Needs (LSI-R) M (SD)
Total	28.19 (6.87)	29.96 (5.72)	26.88 (7.38)	0.01	-0.45	[-0.81, -0.09]
Criminal History	4.63 (2.35)	4.73 (2.32)	4.57 (2.38)	0.71	-0.06	[-0.42, 0.29]
Education/Employment	5.36 (1.92)	5.52 (1.81)	5.24 (2.01)	0.43	-0.14	[-0.50, 0.21]
Financial	1.11 (0.75)	1.26 (0.76)	1.00 (0.72)	0.05	-0.36	[-0.72, 0.01]
Familial Relationships	2.02 (1.08)	2.03 (1.10)	2.01 (1.08)	0.91	-0.02	[-0.38, 0.33]
Accommodations	1.61 (0.93)	1.67 (0.98)	1.57 (0.89)	0.55	-0.11	[-0.46, 0.25]
Leisure and Recreation	1.14 (0.77)	1.21 (0.75)	1.08 (0.79)	0.37	-0.16	[-0.52, 0.19]
Companions	2.88 (1.46)	3.19 (1.41)	2.65 (1.46)	0.04	-0.37	[-0.73, -0.01]
Alcohol and Drug Use	5.32 (2.41)	5.46 (2.42)	5.22 (2.41)	0.59	-0.09	[-0.45, 0.26]
Emotional Health	2.59 (0.96)	3.26 (0.82)	2.10 (0.74)	< 0.001	-1.5	[-1.90, -1.09]
Attitudes and Orientations	1.49 (1.28)	1.59 (1.38)	1.41 (1.19)	0.44	-0.14	[-0.51, 0.21]
Measure of Criminal Attitudes & Associates (MCAA) M (SD)
Total	20.35 (8.51)	22.05 (9.32)	19.01 (7.60)	0.04	-0.36	[-0.71, -0.01]
Attitudes towards Violence	4.07 (2.98)	4.80 (3.19)	3.49 (2.69)	0.01	-0.44	[-0.80, -0.09]
Antisocial Intent	4.44 (3.22)	4.67 (3.42)	4.26 (3.07)	0.47	-0.13	[-0.47, 0.22]
Attitudes towards Entitlement	4.88 (2.15)	5.35 (2.48)	4.52 (1.78)	0.02	-0.39	[-0.75, -0.04]
Attitudes towards Associates	6.94 (2.52)	7.21 (2.48)	6.73 (2.55)	0.28	-0.19	[-0.54, 0.16]

For variables with any cell less than a count of 5, Fisher’s exact test was used. For effect size (ES), Cohen’s d is reported for continuous variables, and Cramer’s V is reported for categorical variables. CI confidence interval

Bivariate analyses revealed several significant differences on LSI-R domain scores between suicide risk groups. Veterans with suicide risk scored significantly higher than veterans without suicide risk on measures of criminal companions (p = .04, d = -0.37, 95% CI [-0.73, -0.01]). Furthermore, veterans with suicide risk were also significantly higher risk on measures of emotional health problems than veterans without suicide risk (p < .001, d = -1.5, 95% CI [-1.90, -1.09]). Veterans with suicide risk were observed to have greater financial risk and reliance on social assistance programs compared to veterans without suicide risk (p = .05, d = -0.36, 95% CI [-0.72, 0.01]). No other significant differences were observed between veteran suicide risk groups on LSI-R domain scores.

MCAA total scores indicated moderate levels of antisocial attitudes and associations in the current sample (M = 20.35, SD = 8.51); total scores were significantly higher in the veterans with (vs. without) suicide risk (p = .04, d = -0.36, 95% CI [-0.71, -0.01]). MCAA sub-scales further demonstrate veterans had mild to moderate risk at baseline, with the most severe risk category being Antisocial Associates (M = 6.94, SD = 2.52), Attitudes towards Entitlement (M = 4.88, SD = 2.15), Antisocial Intent (M = 4.44, SD = 3.22), and Attitudes towards Violence (M = 4.07, SD = 2.98). Per the bivariate analyses, veterans with suicide risk demonstrated significantly higher scores on Attitudes towards Violence than veterans without suicide risk (p = .01, d = -0.44, 95% CI [-0.80, -0.09]). In addition, veterans with suicide risk also demonstrated significantly higher scores on Attitudes Towards Entitlement than veterans without suicide risk (p = .02, d = -0.39, 95% CI [-0.75, -0.04]). A correlation matrix of criminogenic risk variables is provided in Supplementary Table 1; although several domains were minimally to moderately correlated with one another across measures (average Pearson r = .19), this was to be expected and did not suggest redundancy.

Discussion

The present study is among the first to examine the intersection of suicidality, criminogenic needs, and psychiatric burden among veterans in the CLS with COD. Across the full sample, most participants were classified as having moderate to high criminogenic risk, and nearly half reported lifetime suicidality. Veterans with suicidality demonstrated significantly higher total criminogenic risk scores and greater antisocial attitudes compared to those without suicidality. They also reported more functional and emotional impairment and were more likely to meet criteria for probable PTSD, underscoring the cumulative psychiatric burden carried by this subgroup. Although criminogenic cognitions and psychiatric burden differentiated those with and without suicidality, SDOH did not—likely due to restricted variability. Nearly all participants experienced severe disadvantage across housing, employment, and community integration, limiting our ability to detect meaningful group differences.

The current findings complement and extend prior research on risk for suicide among veterans involved in the CLS, particularly through the lens of criminogenic needs. Much of that literature, however, has only focused on the association between suicide risk and diagnoses of antisocial personality disorder. For example, Black and colleagues [49] observed that incarcerated persons diagnosed with antisocial personality disorder had a higher risk of suicide, particularly if they had comorbid attention-deficit hyperactivity disorder. Additionally, other studies using the Psychopathy Checklist-Revised (PCL-R) [50], a validated instrument for predicting criminal recidivism, have demonstrated that the PCL-R factor measuring chronic antisocial deviance is significantly associated with suicide history, even after accounting for histories of abuse or other established risk factors [51, 52].

While the current findings align with this literature, they extend it by identifying specific criminogenic needs—such as attitudes toward violence, entitlement beliefs, and antisocial peers—that were elevated among Veterans with suicidality. This pattern raises important questions about mechanisms linking antisociality and suicide risk. A substantial body of evidence points to impulsivity as a central, cross-cutting factor. Impulsivity is a core feature of antisocial personality disorder and is associated with substance use problems, functional impairment, and maladaptive coping—all domains reflected in the RNR model. Empirical work supports this linkage: Malouf et al. [53] found that lower self-control substantially increased the likelihood of prior suicide attempts among jail inmates, and Verona et al.⁵³ demonstrated that impulsivity statistically accounted for the association between suicide attempts and antisocial deviance on the PCL-R. More broadly, recent meta-analytic evidence confirms that elevated impulsivity is a significant predictor of suicidal ideation, suicide attempts, and self-harm across both justice-involved and clinical populations [54–56]. These findings collectively strengthen the interpretation that the co-occurrence of criminogenic risk and suicidality in this veteran population may reflect a shared vulnerability characterized by impaired impulse control.

The current findings may also have clinical implications for the care for veterans involved in the CLS. For example, prior research on perceived barriers to care for CLS-involved veterans has noted that some providers may have biases towards veterans who exhibit antisocial tendencies [57], which may result in a reluctance to work with these veterans or skepticism of their reported mental health problems. However, the current findings and extant literature suggest that, among CLS-involved veterans, those with antisocial tendencies are at greatest risk for suicide and in most need of clinical care and monitoring. Similarly, it may be assumed among mental health providers that suicide risk among CLS-involved veterans is primarily a function of whether that veteran has a history of internalizing problems, such as depression or PTSD. However, epidemiological research by Verona and colleagues [58] indicates that suicidality among individuals with externalizing tendencies is not necessarily due to comorbid internalizing disorders. Consequently, there may be value in educating behavioral health providers who work with CLS-involved veterans to understand that those at highest risk for criminal recidivism, particularly those who exhibit antisocial behaviors, warrant regular and rigorous suicide risk assessments, regardless of their history of internalizing problems. Importantly, antisocial and impulsive traits confer dual risk—not only for self-directed violence but also for violence toward others [59]. Clinicians working in inpatient and other intensive settings should monitor both suicide risk and potential for interpersonal aggression when treating individuals with prominent antisocial features.

The current findings are also consistent with recent work demonstrating that criminogenic needs—particularly antisocial attitudes and antisocial associates—are significantly higher among CLS-involved veterans with PTSD compared to those without PTSD. This pattern aligns with a broader body of literature indicating that trauma exposure and PTSD are well-established risk factors for suicidal ideation and behavior [60]. PTSD has been linked to heightened emotional dysregulation, impulsivity, and impaired decision-making, which may exacerbate the influence of antisocial attitudes and peer affiliations on risk-taking behavior. When considered alongside criminogenic needs, the presence of PTSD may amplify risk for both recidivism and suicide by reinforcing maladaptive coping strategies, increasing the salience of antisocial networks, and reducing engagement in prosocial activities or treatment. Consequently, the current findings suggest that CLS-involved veterans with both elevated criminogenic needs and PTSD represent a subgroup at particularly high risk for a constellation of adverse behavioral health outcomes, including trauma-related symptom exacerbation, functional impairment, and suicidality. Identifying and addressing the co-occurrence of PTSD and criminogenic risk factors may therefore be critical for developing targeted interventions that reduce suicide risk, interrupt the cycle of criminal justice involvement, and promote long-term recovery among veterans with co-occurring disorders.

Limitations

Several limitations should be considered when interpreting the present findings. First, the analyses were cross-sectional and relied exclusively on baseline data collected prior to randomization into the MISSION-CJ trial. As such, temporal or causal relationships between criminogenic risk factors and suicidality cannot be inferred. Future longitudinal analyses will be needed to determine whether elevated criminogenic needs predict subsequent suicide risk or attempts over time among CLS-involved veterans with COD. Second, the sample size was modest and consisted primarily of homeless male Veterans enrolled in MH RRTPs. Although this aligns with sex distributions in the U.S. Veteran population, it limits generalizability to female Veterans and those not experiencing homelessness. Also, veterans engaged in residential treatment may differ from CLS-involved veterans in other settings (e.g., community-based care, incarceration, or probation/parole) with respect to their clinical profiles, motivation for treatment, and access to behavioral health services. Replication in larger and more diverse samples are warranted. Third, several study variables were assessed via self-report measures, which are subject to recall and social desirability biases. This is particularly salient for sensitive constructs such as suicidal thoughts, past suicide attempts, and criminal history. Although validated measures were used, future research incorporating multi-method assessment strategies (e.g., collateral reports, administrative records) would strengthen the validity of findings. Fourth, while the study included comprehensive assessments of psychiatric symptoms, criminogenic risk, and SDOH, it did not control for multiple statistical comparisons. This analytic choice prioritized reducing Type II error given the exploratory nature of this study, but it also increases the likelihood of Type I error. Analyses were limited to bivariate comparisons, multicollinearity statistics such as VIF could not be computed; however, correlation patterns (Supplementary Table 1) indicated no problematic redundancy among variables. However, findings should therefore be interpreted with caution and considered hypothesis-generating rather than confirmatory. Finally, the study’s operationalization of suicidality grouped together veterans who endorsed lifetime suicide attempts and those who reported current suicidal ideation. Although this approach allowed for examination of a broader suicidality construct, it did not permit differentiation between past and current risk, which may have distinct correlates and implications for intervention. Future studies should examine these groups separately to clarify potential differences in criminogenic and clinical risk profiles.

Conclusions

This study is among the first to examine the relationship between criminogenic risk factors and suicidality among CLS-involved veterans with COD. Findings indicate that veterans with suicidality risk demonstrated higher overall criminogenic risk, more criminogenic attitudes—particularly related to violence and entitlement—and greater emotional and functional impairment compared to their counterparts without suicidality risk. These results underscore the importance of integrating criminogenic risk assessment and targeted interventions into behavioral health care for this population to reduce both suicide risk and risk for CLS recidivism. Future research should build on these findings by employing longitudinal designs to clarify the temporal relationship between criminogenic needs and suicide risk, testing whether targeted interventions addressing these dynamic risk factors can mitigate both outcomes, and exploring how trauma-related conditions such as PTSD interact with criminogenic needs to influence suicide vulnerability. Such efforts could inform the development of tailored, multicomponent interventions that more effectively support recovery, enhance safety, and promote successful community reintegration for veterans with COD involved in the CLS.

Authors’ contributions

PS and DB conceptualized the manuscript idea and drafted and critically revised the manuscript. MA cleaned and performed all analyses as well as contributed significantly to writing. DS provided support in conceptualizing and critically reviewed the manuscript. MC and KB critically reviewed the manuscript and provided feedback. All authors read and approved the final manuscript.

Funding

This study was funded by the Department of Veterans Affairs Office of Research and Development grant and preregistered at https://ClinicalTrials.gov with registration number NCT04523337.

Data availability

The datasets generated and/or analyzed during the current study are not publicly available due to privacy restrictions set forth by the Department of Veterans Affairs. De-identified data sets may be available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The following research is in compliance with the Declaration of Helsinki. All clients provided informed consent prior to participating and the study was approved by VA’s Central Institutional Review Board (# 20 − 01).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.