Criminal Risk and Mental Illness in Psychiatric Inpatient Units: An Opportunity to Provide Psychological Services for Unmet Criminogenic Needs

Abstract

Although the overrepresentation of people with mental illness in the criminal justice system is known, research is needed to identify the frequency of criminal justice involvement and criminogenic treatment needs in inpatient populations to improve continuity of care and access to appropriate treatments. The purpose of this study is to document the frequency of criminal justice involvement among people receiving inpatient community care, as has been done for persons with mental illness in correctional institutions, and to test the association between criminogenic risk and psychiatric symptomatology. The present study uses two samples (n = 94 and n = 142) of adults from two separate acute psychiatric inpatient hospitals in Texas. Data on psychiatric symptoms, mental health history, criminal risk, and criminal justice history were gathered from file review and self-report. Linear and negative binomial regressions were used to test associations of interest. In both samples, the frequency of prior criminal justice involvement was over 50%. The current results indicate there is a significant, positive association between measures of criminal risk and psychiatric symptoms. These findings highlight the need to address the reciprocal association between mental illness and criminal risk among people receiving inpatient psychiatric treatment with appropriate assessment and treatment.

It is well known that people with serious mental illness (SMI) are overrepresented in the criminal justice system; for example, previous estimates show 44% of the U.S. jail population reported a formal psychiatric diagnosis (see e.g., Al-Rousan et al., 2017; Bronson & Berzofsky, 2017). As a result of the large number of people entering correctional facilities with psychiatric symptoms, there has been a movement to deliver services that address co-occurring psychiatric symptoms and criminogenic risk. Criminogenic risk factors are predictors of antisocial behavior, the most robust of which (commonly referred to as the Central Eight) are antisocial attitudes, antisocial peers, antisocial personality pattern, history of antisocial behavior, family and marital functioning, lack of educational/employment achievement, lack of prosocial leisure activities, and substance abuse (Andrews & Bonta, 2017). In addition to the push to deliver services that address both mental health and these criminal risk factors, there have been calls for treatment providers to integrate co-occurring psychiatric and criminogenic risk in treatments for people with mental illness and criminal justice involvement (PMI-CJI; Draine et al., 2002; Hodgins et al., 2007; Morgan et al., 2012).

Existing research has supported the move toward integrated treatment, demonstrating that addressing both psychiatric and criminogenic risk reduces self-reported psychiatric symptoms and criminal thinking (i.e., cognitions conducive to criminal behavior; Gaspar et al., 2019; Morgan et al., 2014) and improves psychological well-being (Scanlon & Morgan, 2021). In fact, the relation between mental illness and criminogenic risk in PMI-CJI has been hypothesized to be reciprocal, such that as criminogenic risk increases, psychiatric symptomatology follows, and vice versa (Bartholomew & Morgan, 2015; Morgan et al., 2020). Thus, although addressing both psychiatric and criminogenic treatment needs appear to be important for the success of PMI-CJI, the proposed relation between mental illness and criminal risk and subsequent proposed conceptual model has not yet been empirically tested.

The interplay between criminogenic risk and psychiatric symptoms likely contributes to people with mental illness’ placement in both correctional and psychiatric facilities. Although the frequency of mental illness in the U.S. criminal justice system is well documented, less research has been conducted on the frequency of criminal justice involvement in psychiatric populations. Over 23% of 6,624 adults with SMI were arrested within 10 years in Los Angeles County (Cuellar et al., 2007). In a sample of 3,856 adults who received services from the Massachusetts Department of Mental Health, 27.9% were arrested within 10 years (Fisher et al., 2006); people who received public mental health services in this system also had increased odds of getting arrested compared to people in the general population (Fisher et al., 2011). Timmer and Nowotny (2021) found people on probation, parole, or those with recent arrests were more likely to utilize mental health care than people without criminal justice involvement. This reciprocity poses a hurdle to treatment providers in both systems, who traditionally target mental illness or criminal risk individually.

Despite the rates of criminal justice involvement in community mental health care-seeking samples, criminogenic needs go largely unaddressed in psychiatric facilities (Bonfine et al., 2020; Marquant & Torres-Gonzalez, 2018; Morgan et al., 2020). Therefore, community mental health centers and inpatient psychiatric facilities are a potentially underutilized point of opportunity to target and reduce both psychiatric and criminogenic risk, increase the impact of treatment, and ultimately improve public health and safety. However, more work needs to be done to assess how co-occurring criminogenic risk and psychiatric symptoms might be associated—if research shows these co-occurring issues are closely related, only addressing one issue in a “siloed organization of services” (Bonfine et al., 2020, p. 360), rather than addressing both as interacting problems, will likely dilute the impact of treatment that PMI-CJI receive across settings. Thus, better understanding of the relation between mental illness and criminal risk is needed to inform effective psychological services delivered to this population.

The present study aimed to document the frequency of criminal justice involvement of people receiving inpatient psychiatric community care and examine the relation between mental illness and criminogenic risk in these two samples, to improve the generalizability of the current findings. The following research questions guided this study: (a) What is the frequency of criminal justice involvement among two independent samples of people receiving inpatient psychiatric community care? and (b) What is the association between criminogenic risk and psychiatric symptomatology? Based on the conceptualization that mental illness and criminal risk are reciprocal and co-occurring in PMI-CJI (Bartholomew & Morgan, 2015; Morgan et al., 2020), we hypothesized that (a) the rates of criminal justice involvement for people with mental illness would be consistent with those of previous studies (approximately 24%–27%; Cuellar et al., 2007; Fisher et al., 2006) and (b) criminogenic risk would be significantly and positively associated with psychiatric symptoms, number of inpatient hospitalizations, and length of hospitalizations.

Method

Participants

Sample 1

Participants were 94 adults hospitalized in an acute psychiatric inpatient unit in Lubbock, Texas. Individuals were eligible to participate if they had been hospitalized for at least 5 days, were English-speaking, were at least 18 years old, were not admitted after being adjudicated incompetent to stand trial and not restorable, and were not receiving competency restoration services. Of the potential participants approached, 66 (58.75%) decided not to participate in the study. Data from Sample 1 were gathered over approximately 6 months. See Table 1 for the complete demographic data for Sample 1.

Table 1.

Demographic Information for Samples 1 and 2

Demographic Variable	M (SD)	n (%)
Sample 1 (n = 94)
Age	38.55 (11.35)
Education length	12.46 (2.13)
Number of prior hospitalizations	7.01 (12.51)
Sex
Male		53 (56.4)
Female		41 (43.6)
Prior criminal justice involvement
Yes		51 (54.3)
No		43 (45.7)
Race
White/Caucasian		54 (57.4)
Other		9 (9.6)
Black/African American		8 (8.5)
American Indian/Native American		1 (1.1)
Asian/Asian American		1 (1.1)
Ethnicity
Non-Hispanic		74 (78.7)
Hispanic		20 (21.3)
Relationship status
Single		68 (72.3)
Not single		12 (12.8)
Mental illness per records
None		5 (5.3)
Serious mental illness only		62 (66.0)
Personality disorder only		1 (1.1)
Serious mental illness and personality disorder		26 (27.7)
Psychiatric diagnosis per records
Bipolar		38 (40.4)
Major depressive disorder		29 (30.9)
Schizophrenia		21 (22.3)
Other clinical disorders (formerly “Axis 1”)		6 (6.4)
Borderline PD		12 (50.0)
PD not otherwise specified		5 (20.8)
Antisocial PD		3 (12.5)
Dependent PD		3 (12.5)
Sample 2 (n = 142)
Age	37.76 (13.67)
Education length	12.97 (2.67)
Length of hospitalization	7.74 (4.67)
Sex
Male		79 (55.6)
Female		63 (44.4)
Prior criminal justice involvement
Yes		74 (52.1)
No		68 (47.9)
Race
White/Caucasian		87 (61.3)
Black/African American		11 (7.7)
Native American		6 (4.2)
Other		2 (1.4)
Ethnicity
Non-Hispanic		100 (74.6)
Hispanic		36 (25.4)
Relationship status
Single		95 (66.9)
Not single		46 (32.4)
Mental illness per records
None		21 (14.9)
Serious mental illness only		102 (72.3)
Personality disorder only		2 (1.4)
Serious mental illness and personality disorder		16 (11.3)
Psychiatric diagnoses per records
Mood disorder		100
Anxiety disorder		39
Psychotic disorder		14
Borderline personality disorder		12
Antisocial personality disorder		5
Histrionic personality disorder		1
Schizoid personality disorder		1

Note. Due to missing demographic data and comorbid diagnoses, not all percentages totaled 100. PD = Personality Disorder.

Sample 2

Participants were 142 adults hospitalized in an acute psychiatric inpatient unit in Lubbock, Texas (a different facility from that of Sample 1, approximately 8 miles apart). Individuals were eligible to participate if they did not evidence active psychosis that reduced their capacity to provide consent or complete the assessments accurately, were English-speaking, and were at least 18 years old. Of the potential participants approached, 62 (29.25%) decided not to participate, 7 (3.30%) did not complete the assessments, and 1 was not competent (.01%); this participation rate (66%) is roughly consistent with those seen in other studies with inpatient samples (53%–55%; see Berry et al., 2017; Horowitz et al., 2018; Schröder et al., 2016). Data from Sample 2 were gathered over the course of 10 months. See Table 1 for the complete demographic data for Sample 2.

Measures

Of note, the data were collected by two different members of this research team using different assessment measures, to increase the generalizability of our findings, while still comprehensively assessing the underlying constructs of interest (mental health and criminogenic risk).

Demographic History Questionnaire

The Demographic History Questionnaire (DHQ) is a self-report measure that was administered to Samples 1 and 2 and was designed for the collection of the data used in this study. Participants self-reported their basic demographic information (e.g., age, ethnicity/race, relationship status, employment status, and income), history of criminal justice involvement, psychiatric diagnoses, and psychiatric treatment history. We operationalized SMI, for purposes of the present study, as a psychotic disorder, and/or any mood or anxiety disorder. We operationalized prior criminal justice involvement as a history of a misdemeanor and/or felony conviction, which is consistent with previous studies (Bolaños et al., 2020; Gross & Morgan, 2013). The number of prior hospitalizations was self-reported by Sample 1. The current length of hospitalization (in days) was obtained from file review for Sample 2. For both samples, we obtained psychiatric diagnoses from a medical record review.

Psychological Inventory of Criminal Thinking Styles

We administered the Psychological Inventory of Criminal Thinking Styles (PICTS) to Sample 1. The PICTS is an 80-item self-report measure of thought patterns associated with criminal behavior where participants respond on a 4-point ordinal response metric ranging from 1 = Disagree to 4 = Strongly agree (Walters, 1995, 2006). Higher scores indicate higher criminal thinking. The PICTS yields a total, General Criminal Thinking score, two content scores (i.e., Proactive and Reactive Criminal Thinking), eight thinking style scores (i.e., Mollification, Cutoff, Entitlement, Power Orientation, Sentimentality, Superoptimism, Cognitive Indolence, and Discontinuity), and five Factor and Special Scales (i.e., Problem Avoidance, Interpersonal Hostility, Self-Assertion, Denial of Harm, and Fear of Change; Walters, 2006). The PICTS has demonstrated strong psychometric properties (see Walters, 2006 for review). In Sample 1, the validity indices indicated all of the profiles were valid, so no PICTS data were excluded due to invalid responding. In the present study, we only utilized the General Criminal Thinking score (Cronbach’s α = .94), given that we were most interested in global criminal thinking patterns.

The Millon Clinical Multiaxial Inventory-Third Edition

We administered the Millon Clinical Multiaxial Inventory-Third Edition (MCMI-III) to Sample 1. The MCMI-III is a 175-item true or false self-report assessment of clinical and personality symptoms where higher scores indicate more severe psychiatric symptoms (Millon & Davis, 1994). The MCMI-III includes 14 personality disorder scales and 10 clinical syndrome scales that are congruent with the Diagnostic and Statistical Manual, 4th edition (DSM-IV; American Psychiatric Association, 1994; Millon & Davis, 1994). The MCMI-III validity indices include the Correction Scale (detection of careless or random responding) and the Modifying Indices and Validity Index (assessment of validity and response style; Millon & Davis, 1994). In Sample 1, the authors determined 13 profiles were invalid after reviewing the Validity Index for all participants and excluded those profiles from the present study analyses. The MCMI-III has demonstrated strong psychometric properties in clinical samples (Millon & Davis, 1994; Millon et al., 1997). In the present study, the MCMI-III items were highly internally consistent (Cronbach’s α = .87). The MCMI-III was used in the present study to measure clusters of psychiatric symptoms.

Self-Appraisal Questionnaire

We administered the Self-Appraisal Questionnaire (SAQ) to Sample 2. The SAQ is a 72-item true or false self-report measure of risk factors associated with criminal behavior (Loza, 1996). The SAQ produces seven subscales (i.e., Criminal Tendencies, Conduct Problems, Alcohol/Drug Abuse, Anger, Antisocial Personality, Criminal History, and Antisocial Associates), and a total score, which is the sum of responses from all subscales except Anger (Loza, 1996). Higher scores indicate greater criminal risk. The SAQ has demonstrated strong psychometric properties (Loza et al., 2004, 2005; Mitchell et al., 2013). In the present study, we utilized the SAQ total score (Cronbach’s α = .94) as a measure of general criminal risk.

Measures of Criminal Attitudes and Associates

We administered the Measures of Criminal Attitudes and Associates (MCAA) to Sample 2. The MCAA is a self-report assessment of one’s number of criminal associates and criminal attitudes, as assessed by Part A and Part B (Mills & Kroner, 1999). Part A assesses up to four adult associates with whom one spends time, and the amount of time spent with these associates, and the degree to which these associates are involved in crime. Part B is 46 items with an Agree or Disagree response format that assesses 4 criminal attitudes scales: Attitudes Towards Violence (12 items), Antisocial Intent (12 items), Attitudes Towards Entitlement (12 items), and Attitudes Towards Criminal Others (10 items; Mills & Kroner, 1999). Higher scores indicate greater criminal attitudes (Mills & Kroner, 1999). The MCAA has yielded adequate to strong psychometric characteristics (Mills et al., 2002; Mills & Kroner, 1999). Since there is not an overall score available from the MCAA, we utilized all criminal attitudes scales from Part B in the present study, that is, Attitudes Towards Violence (Cronbach’s α = .85), Antisocial Intent (Cronbach’s α = .83), Attitudes Towards Entitlement (Cronbach’s α = .66), and Attitudes Towards Criminal Others (Cronbach’s α = .83).

Brief Symptom Inventory

We administered the Brief Symptom Inventory (BSI) to Sample 2. The BSI is a 53-item self-report assessment of past-week psychiatric symptom distress where participants rate each item on a 5-point ordinal response metric (i.e., 0 = Not at all, 1 = A little bit, 2 = Moderately, 3 = Quite a bit, and 4 = Extremely; Derogatis & Melisaratos, 1983). The BSI is an abbreviated version of the Symptom Checklist-90 (SCL-90) and yields nine symptom scales and three global indices (Derogatis, 1993). Generally, the BSI has demonstrated good psychometric characteristics (e.g., Boulet & Boss, 1991; Derogatis, 1993; Hoe & Brekke, 2008); however, there has been mixed support for a nine-factor structure when compared to a one-factor structure for the BSI (e.g., Hayes, 1997; Loutsiou-Ladd & Kokkinos, 2008; Piersma et al., 1994). Therefore, in the present study, we only used the BSI Global Severity Index (BSI GSI; Cronbach’s α = .96) as an indicator of global psychiatric distress.

Procedure

The present study uses two datasets previously collected by this research team. The data in this study were initially collected for two distinct primary studies with differing research questions and measures, but the same constructs of interest: mental health functioning and criminogenic risk. The data collection procedures for both samples included in this study were similar (see Table 2 for a description of both studies’ procedures). Participants could discontinue participation without penalty. Participants were not compensated for their participation. All participant data were confidential and anonymous. Both the hospitals’ and the university’s institutional review boards involved in the initial collection of the data for these samples approved the procedures.

Table 2.

Samples 1 and 2 Methods

Participants/procedures	Sample 1	Sample 2
• Data collected from adults in an acute psychiatric inpatient unit	✓	✓
• A roster of people admitted to the hospital was presented to research team to identify potential participants	✓	✓
• Hospital staff confirmed which patients met the inclusion criteria prior to recruitment	✓	✓
• Participants were at least 18 years old and English-speaking	✓	✓
• Participants had been hospitalized for at least 5 days and were not admitted after being adjudicated incompetent to stand trial and not restorable	✓
• Participants were not receiving competency restoration services	✓	✓
• Participants did not evidence active psychosis that reduced their capacity to provide consent or complete the assessments accurately		✓
• The research team approached potential participants about participating	✓	✓
• In a private space, the research team informed potential participants about the study’s purpose, procedures, research participants’ rights, and risks and benefits of participating	✓	✓
• The research team asked potential participants to read an informed consent and gave potential participants an opportunity to ask any questions about participating	✓	✓
• If needed, a member of the research team read the consent form (and subsequent assessments) aloud to the participant	✓	✓
• The research team assessed potential participant’s understanding by asking them to summarize the study	✓	✓
• Potential participants who were interested in participating and demonstrated an understanding of the study signed an informed consent; participants were given a copy of an unsigned consent	✓	✓
• Participants were given the assessment battery, with a member of the research team present to address questions or concerns	✓	✓
• Following completion of the assessment battery, the research team conducted a review of the participant’ s facility file	✓	✓
Measures
• Demographic History Questionnaire (DHQ)	✓	✓
• Psychopathology
Millon Clinical Multiaxial Inventory Third Edition (MCMI-III)	✓
Brief Symptom Inventory (BSI)		✓
• Criminogenic risk
Psychological Inventory of Criminal Thinking Styles (PICTS)	✓
Measures of Criminal Attitudes and Associates (MCAA)		✓
Self-Appraisal Questionnaire (SAQ)		✓

Data Analysis Plan

In order to test our hypotheses, we conducted a series of analyses in the two samples. For Research Question 1 (What is the frequency of criminal justice involvement among two samples of people receiving inpatient psychiatric community care?), we rely on basic descriptive statistics (e.g., percentages). For Research Question 2 (What is the association between criminogenic risk and psychiatric symptomatology in these two samples of people receiving inpatient psychiatric community care?), we used multiple indices of criminogenic risk and psychiatric symptoms. In the first sample, the relation between psychiatric symptoms and criminogenic risk was tested using psychiatric symptom scales and the number of prior hospitalizations as separate independent variables for the dependent variables of criminal thinking and antisocial personality disorder scales. In the second sample, factors associated with criminogenic risk (i.e., prior criminal justice involvement, criminal attitudes, and criminal associates) were examined as independent variables, with dependent variables of global severity of psychiatric symptoms and length of current hospitalization. Given this study’s primary goal was to provide ecologically valid information that might be directly applicable to clinical samples, each as independent variable was tested as a separate predictor of the outcome variable, bivariately, without adjusting for any other independent variables of interest. All analyses were conducted in IBM SPSS 25 (SPSS Inc., Chicago, USA). The two samples were analyzed separately because different measures were used between the samples.

Sample 1

Linear regressions were used to test the association between criminal risk variables and continuous psychiatric symptom criterion variables in Sample 1; the Central Eight criminogenic risk factor captured by each criminal risk variable is presented in italics and parenthesis. The independent variables were PICTS General Criminal Thinking (antisocial cognitions), the MCMI Antisocial Personality Disorder scale (antisocial personality pattern), and prior criminal justice involvement (history of antisocial behavior; coded 1 = criminal justice involvement, 0 = no criminal justice involvement). Continuous independent variables were mean centered for improved interpretability. The criterion variables in these models were the MCMI Clinical Syndromes, MCMI Severe Clinical Syndromes, MCMI Borderline Personality Disorder (BPD) scales. The MCMI BPD scale was included because BPD is consistently one of the two most common personality disorders in samples of people receiving inpatient psychiatric care (Zimmerman et al., 2008). The MCMI criterion variables approximated normal distributions. No covariates were included in Sample 1 linear regressions as no demographic variables (i.e., sex, age, race, ethnicity, relationship status, and self-reported mental health diagnosis) were significantly associated with the MCMI criterion variables. R² values are presented as measures of effect size of linear regression models.

Negative binomial regressions were used to test the associations between criminal risk variables and the number of prior hospitalizations. The number of prior hospitalizations (the criterion variable) is a non-normally distributed count variable with a dispersion greater than 1; therefore, negative binomial regressions were used instead of Poisson regression. The PICTS General Criminal Thinking, the MCMI Antisocial Personality Disorder scale, and prior criminal justice involvement were included as independent variables in the negative binomial regression models. No covariates were included in Sample 1 negative binomial regressions as no demographic variables (i.e., sex, age, race, ethnicity, relationship status, and self-reported mental health diagnosis) were significantly associated with the number of prior hospitalizations. Incident rate ratios (IRRs) and 95% confidence intervals (CIs) are presented as measures of effect size for negative binomial regression models.

Sample 2¹

Linear regressions were used to test the association between a measure of psychiatric symptom severity and criminal risk variables; the Central Eight criminogenic risk factor captured by each criminal risk variable is presented in italics and parenthesis. The independent variables in Sample 2 linear regressions included SAQ total score, which includes responses to the following subscales: Criminal Tendencies (antisocial attitudes), Conduct Problems (history of antisocial behavior), Criminal History (history of antisocial behavior), Alcohol/Drug Abuse (substance use), Antisocial Personality (antisocial personality pattern), and Antisocial Associates (antisocial associates), MCAA Attitudes toward Associates (antsocial associates), MCAA Attitudes toward Violence (antisocial attitudes), MCAA Attitudes toward Intent (antisocial attitudes), MCAA Attitudes toward Entitlement (antisocial attitudes), and prior criminal justice involvement (history of antisocial behavior). Continuous independent variables were mean centered to improve interpretability. The criterion variable in Sample 2 linear regressions was the BSI GSI; scores on the BSI GSI approximated a normal distribution. In Step 1 of the linear regression model, significant covariates (mental illness and race) were entered; the primary independent variables were entered in Step 2 of the model. R² and change in R² values are presented as measures of effect size.

Negative binomial regressions were used to test the associations between criminal risk variables and length of hospitalizations. Because the length of hospitalization (the criterion variable) is a non-normally distributed count variable with a dispersion greater than 1, negative binomial regression models were used instead of Poisson regression. In Sample 2 negative binomial regression models, the independent variables were SAQ total score, MCAA Associates, Violence, Intent, Entitlement, and prior criminal justice involvement. Psychiatric diagnosis was significantly associated with the length of hospitalization and was included as a covariate in these models. IRR and 95% CI are presented as measures of effect size for negative binomial regression models.

Results

Sample 1²

Sample 1 demographic information, including self-reported prior criminal justice involvement and psychiatric diagnoses obtained from file reviews, is presented in Table 1. Univariate outliers were identified as scores greater than 3.29 SD from the mean and were deleted (Tabachnick & Fidell, 2013). Multivariate outliers were identified as scores with Mahalanobis Distance p values < .001 and were deleted (Tabachnick & Fidell, 2013).

Preliminary Analyses in Sample 1

Categorical demographic variables, sex, race, ethnicity, relationship status, and psychiatric diagnosis (SMI/no SMI), were tested against the criterion variables (i.e., MCMI BPD scale, MCMI Clinical Syndromes, MCMI Severe Clinical Syndromes, and number of prior hospitalizations) using analysis of variance (ANOVA). Linear regressions were used to test the relations between continuous demographic (age and education) and criterion variables. Sample 1 covariate testing revealed no demographic variables were significantly associated with the criterion variables, so no covariates were included in linear regression models.

Frequency of Prior Criminal Justice Involvement in Sample 1

In Sample 1, the frequency of prior criminal justice involvement among the sample of 142 people receiving inpatient psychiatric care is 52.1%.

Criminogenic Risk as a Predictor of Mental Illness in Sample 1

Separate linear regressions were used to test the hypothesis that PICTS General Criminal Thinking, the MCMI Antisocial Personality Disorder scale, and prior criminal justice involvement would be positively associated with MCMI psychiatric symptoms. As hypothesized, the General Criminal Thinking scale was significantly positively associated with the MCMI BPD scale, Clinical Syndromes, and Severe Clinical Syndromes scales (β ranging from .41 to .60, R² ranging from .17 to .36; see Table 3). Also consistent with our hypotheses, the MCMI Antisocial Personality Disorder scale was significantly positively associated with the MCMI BPD scale, Clinical Syndromes, and Severe Clinical Syndromes scales (β ranging from .33 to .61, R² ranging from .11 to .37; see Table 3). Associations with prior criminal justice involvement were only significantly positively associated with the MCMI BPD scale, but not the MCMI Clinical Syndromes or Severe Clinical Syndromes scales (Table 3), which is contrary to our hypotheses.

Table 3.

Criminogenic Risk Variables as Predictors of Psychiatric Symptomatology in Sample 1

Independent variables	n	β	R 2
Criterion variable: MCMI BPD
PICTS GCT	60	.60***	.36
MCMI antisocial	71	.61***	.37
Prior criminal justice involvement	72	.26*	.07
Criterion variable: MCMI Clinical Syndromes
PICTS GCT	61	.41**	.17
MCMI antisocial	72	42***	.17
Prior criminal justice involvement	73	.12	.01
Criterion variable: MCMI Severe Clinical Syndromes
PICTS GCT	61	43**	.19
MCMI antisocial	72	.33**	.11
Prior criminal justice involvement	73	.07	.00

Note. Each independent variable was analyzed in a separate model. BPD = borderline personality disorder; GCT = General Criminal Thinking; MCMI = Millon Clinical Multiaxial Inventory; PICTS = Psychological Inventory of Criminal Thinking Styles.

^*p < .05.

^**p < .01.

^***p < .001.

Separate negative binomial regressions were used to test the hypothesis that PICTS General Criminal Thinking, the MCMI Antisocial Personality Disorder scale, and prior criminal justice involvement would be significantly positively associated with the number of prior hospitalizations. However, prior criminal justice involvement was the only criminal risk variable significantly associated with the number of prior hospitalizations (Table 4), such that the rate of prior hospitalizations is 1.83 times higher for those with prior criminal justice involvement than those not previously involved in the criminal justice system.³

Table 4.

Criminogenic Risk Variables as Predictors of Number of Prior Hospitalizations in Sample 1

Independent variables	n	b (SE)	IRR [95% CI]
PICTS GCT	53	.01 (.01)	1.01 [.99, 1.02]
MCMI antisocial	56	.02 (.02)	1.02 [.97, 1.07]
Prior criminal justice involvement	73	.60 (.21)**	1.83 [1.22, 2.74]

Note. Each independent variable was analyzed in a separate model. IRR = incident rate ratio; GCT = General Criminal Thinking; MCMI = Millon Clinical Multiaxial Inventory; PICTS = Psychological Inventory of Criminal Thinking Styles.

^**p < .01.

Sample 2

Preliminary Analyses in Sample 2

ANOVAs were used to test the associations between categorical demographic variables (sex, race, ethnicity, relationship status, and mental health diagnosis) and criterion variables (length of hospitalization and BSI GSI). Linear regressions were used to test the relations between continuous demographic (age and education) and criterion variables. Sample 2 covariate testing revealed no demographic variables were significantly associated with the length of current hospitalization. Race (White or non-White), F(1, 104) = 4.46, p = .037, η² = .04, and psychiatric diagnosis, F(1, 135) = 13.84, p < .001, η² = .09, were significantly associated with BSI GSI scores. Covariates were only included in models in which they were significantly associated with the criterion variable; for this reason, covariates differed between analyses. Demographic information for Sample 2 is presented in Table 1.

Frequency of Prior Criminal Justice Involvement in Sample 2

In Sample 2, the frequency of self-reported prior criminal justice involvement among the sample of 94 people receiving inpatient psychiatric care is 54.3%.

Criminogenic Risk as a Predictor of Mental Illness in Sample 2

Separate linear regressions were used to test the association between measures of criminogenic risk (i.e., SAQ total score, MCAA scales, and prior criminal justice involvement) and BSI GSI, a measure of psychiatric symptom severity. Consistent with a priori hypotheses, the SAQ total score, MCAA Associates, MCAA Intent, and MCAA Violence were significantly positively associated with BSI GSI scores (β ranging from .21 to .26, R² ranging from .04 to .07; see Table 5). Contrary to our hypotheses, MCAA Entitlement and prior criminal justice involvement were not significantly associated with the BSI GSI (β ranging from .01 to .16, R² ranging from .01 to .03; see Table 5).

Table 5.

Criminogenic Risk Variables as Predictors of BSI GSI Index in Sample 2

Independent variables	n	β	R 2
SAQ total score	104	.25**	.06
MCAA associates	104	.21*	.04
MCAA entitlement	104	.16	.03
MCAA intent	104	.21*	.04
MCAA violence	104	.26**	.07
Prior criminal justice involvement	104	.01	.01

Note. Each independent variable was analyzed in a separate model. BSI GSI = Brief Symptom Inventory Global Severity Index; SAQ = Self-Appraisal Questionnaire; MCAA = Measures of Criminal Attitudes and Associates.

^*p < .05.

^**p < .01.

Separate negative binomial regressions were used to test measures of criminogenic risk (i.e., SAQ total score, MCAA scales, and prior criminal justice involvement) as independent variables for the dependent variable of length of current hospitalization, a behavioral measure of mental illness. Contrary to our hypotheses, no variables were significantly associated with length of hospitalization, including SAQ Total Score, MCAA Associates, MCAA Violence, MCAA Intent, MCAA Entitlement, and prior criminal justice involvement (Table 6; see Footnote 1).

Table 6.

Criminogenic Risk Scales as Predictors of Length of Current Hospitalization in Sample 2

Independent variables	n	b (SE)	IRR [95% CI]
SAQ total score	134	.01 (.01)	1.01 [.99, 1.02]
MCAA associates	135	.01 (.03)	1.01 [.95, 1.08]
MCAA entitlement	136	.03 (.04)	1.03 [.95, 1.12]
MCAA intent	135	−.00 (.03)	1.00 [.94, 1.06]
MCAA violence	135	.00 (.03)	1.00 [.95, 1.07]
Prior criminal justice involvement	135	.21 (.18)	1.23 [.86, 1.77]

Note. Each independent variable was analyzed in a separate model. IRR = incident rate ratio; BSI GSI = Brief Symptom Inventory Global Severity Index; SAQ = Self-Appraisal Questionnaire; MCAA = Measures of Criminal Attitudes and Associates. No models were statistically significant (p < .05).

Discussion

Although it has been well documented that people with mental illness are overrepresented in the criminal justice system (Bronson & Berzofsky, 2017; Prins, 2014; Steadman et al., 2009), there has not been a reciprocal focus on the frequency of criminal justice involvement in community mental health settings. The data used in this study are from two independent groups of patients in inpatient units; the similarities seen across these two samples offers evidence of the reproducibility of our findings, enhancing the robustness of our results. Results from these two independent samples demonstrated that approximately 50% of people receiving inpatient psychiatric care had a self-reported history of misdemeanor or felony convictions, compared to an estimated 33% of adults in the United States with a misdemeanor or felony arrest (Fields & Emshwiller, 2014). The current results exceed the rates found by Cuellar et al. (2007; 23.6%) and Fisher et al. (2006; 27.9%) and further highlight that a significant portion of the community mental health population experiences criminal justice involvement.

Community mental health treatment providers may need a similar change in philosophy and treatment planning. Research supports that people with mental illness who become involved in the criminal justice system present with criminal risk factors similar to their criminal justice involved peers without mental illness: PMI-CJI are higher in criminal risk than people not involved in the criminal justice system (Bartholomew et al., 2018; Bolaños et al., 2020; Morgan et al., 2010; Skeem et al., 2014; Wilson et al., 2014; Wolff et al., 2011). That 50% of psychiatric patients have a history of a misdemeanor and/or felony convictions is noteworthy, and also disconcerting as it is likely that the rates of contact with the criminal justice system would be even higher if other types of interactions are considered (e.g., arrests). This remarkable finding has significant implications for the planning and delivery of psychological services.

In the pathoplasticity model of psychopathology, personality and experiences influence the nature and course of onset of psychopathology, and vice versa (Widiger & Smith, 2008). Per the pathoplasticity model, the compounding and interactive effect of multiple disorders and life events complicates clinical presentations and treatment models. Pathoplastic features that may be impacting the clinical presentation of some people with SMI in community mental health settings are criminal risk factors and criminal justice involvement. The rates of criminal justice involvement obtained in our two studies (54.3% and 52.1%, respectively) exceeds 23.6% found by Cuellar et al. (2007) and 27.9% found by Fisher et al. (2006). This discrepancy may be in part because both prior studies relied on large community-based (nonhospitalized) samples, whereas our studies were exclusively samples of people receiving inpatient psychiatric care from two different acute care psychiatric facilities. In other words, it is possible that with increased severity in psychopathology resulting in hospitalization, there is a reciprocal increase in criminal risk and subsequently, criminal justice involvement.

The results of this study also demonstrated that criminogenic needs (dynamic criminal risk factors that need to be the focus of treatment when providing correctional rehabilitative services; Andrews & Bonta, 2017) are positively associated with psychiatric symptomatology and functioning, across two independent samples using various indicators of mental illness and criminal risk (see Table 7 for a summary of the current results). Additionally, all statistically significant models in both samples also demonstrated marginal to strong clinical significance, with R² values ranging from .04 to .37. The relation between mental illness and criminal risk found in this study further supports a complex multidirectional model, such that untreated mental illness may result in increased criminal risk and vice versa. Although this model is supported by the current results, further research needs to examine if this is in fact a reciprocal relation, consistent with the pathoplasticity model of psychopathology. In the proposed model, mental illness and criminalness (i.e., illegal and non-illegal behaviors that violate social norms and the rights and well-being of others) are both independently tied to outcomes, such that untreated mental illness is associated with increased symptoms and hospitalization, and untreated criminalness is associated with increased criminal recidivism—a framework that underscores the importance of intervening with appropriate psychological services.

Table 7.

Results Summary for Criminogenic Risk Variables as a Predictors of Mental Illness Variables in Samples 1 and 2

	Criterion variables: Psychiatric symptoms and hospitalizations
Independent variables: Criminal risk	Sample 1	Sample 2
General Criminal Thinking (PICTS)	MCMI BPD*	—
	MCMI Clinical Syndromes*
	MCMI Severe Clinical Syndromes*
	Number of prior hospitalizations
Antisocial Personality Disorder Scale (MCMI)	MCMI BPD*	—
	MCMI Clinical Syndromes*
	MCMI Severe Clinical Syndromes*
	Number of prior hospitalizations
Prior criminal justice involvement	MCMI BPD*	BSI GSI
	MCMI Clinical Syndromes	Length of current hospitalization
	MCMI Severe Clinical Syndromes
	Number of prior hospitalizations*
General criminal risk (SAQ total score)	—	BSI GSI*
		Length of current hospitalization
Antisocial attitudes (MCAA)
Attitudes toward antisocial associates	—	BSI GSI*
		Length of current hospitalization
Attitudes toward entitlement	—	BSI GSI
		Length of current hospitalization
Attitudes toward intent	—	BSI GSI*
		Length of current hospitalization
Attitudes toward violence	—	BSI GSI*
		Length of current hospitalization

Note. PICTS = Psychological Inventory of Criminal Thinking Styles; MCMI = Millon Clinical Multiaxial Inventory; BPD = borderline personality disorder; BSI GSI = Brief Symptom Inventory Global Severity Index; SAQ = Self-Appraisal Questionnaire; MCAA = Measures of Criminal Attitudes and Associates—indicates the predictor was not available for that sample;

^*Model was statistically significant.

Although treatment programs, such as Changing Lives and Changing Outcomes (Morgan et al., 2017), have been developed for the co-occurring issues of criminogenic risk and mental illness among PMI-CJI, such programming has been implemented primarily in correctional and forensic mental health systems. In fact, unlike efforts to tailor traditional correctional rehabilitation programs such as Thinking for a Change (Bush et al., 1997) or Moral Reconation Therapy (Little & Robinson, 1988) to the needs of PMI-CJI, there has not been reciprocal efforts for relevant mental health programming to be tailored to the criminogenic needs of people with mental illness who are criminal justice involved. Future research also needs to examine how the community mental health sector is delivering evidenced-based psychological services for reducing criminal risk and recidivism, and where lacking, those working in the community mental health sector must alter the service delivery model (see e.g., Bonfine et al., 2020) and services to meet the treatment needs of patients who are currently involved in or are at risk of becoming involved in the criminal justice system. Recent literature has proposed that the criminogenic needs of PMI-CJI should be treated in tandem with mental health services “to reduce redundancy” (Bonfine et al., 2020), engage in “collaborative” care (Lamberti, 2016), and “take an interdisciplinary approach” (Kim, 2016). The sequential intercept model, for example, provides a framework for a systems-level approach to intervening with PMI-CJI (Munetz & Griffin, 2006), and a similar model is needed for intervening with atrisk people with mental illness in psychiatric or community mental health settings. The frequency of criminal justice involvement and the criminogenic needs in the current psychiatric samples provides additional support for such an integrated model. Therefore, assessment of criminal risk should be best practice when evaluating people with mental illness entering the mental health system.

Several measures exist that would facilitate assessment of both mental health and criminal risk needs including, for example, the Level of Service/Case Management Inventory (Andrews et al., 2000) and the Historical Clinical and Risk Management—20 Version 3 (HCR-20V3; Douglas et al., 2013). Assessments such as these, which integrate an actuarial-based criminal risk assessment with an assessment of mental health concerns and needs, will guide service delivery. Based on these assessments, interventions targeting identified criminal risk factors should be implemented to reduce the risk of criminal justice involvement. These assessments can also guide clinicians’ use of interventions that are designed to reduce the impact of mental illness and criminal risk’s interaction (pathoplasticity), while simultaneously providing services to reduce mental health symptoms and enhance recovery. Follow-up services should include monitoring criminal risk, and outcomes should include effectiveness toward recovery, and reduced or avoided criminal justice interaction. Mainstream use of such assessments and interventions has the potential to reduce the number of people with mental illness in the criminal justice system and inform policy impacting this population. On a systemic level, recent work by Prins (2019) demonstrated that interactions with the criminal justice system via arrests and convictions increase criminal risk factors, and therefore increase the likelihood of future criminal justice involvement; these findings further underscore the importance of primary intervention to prevent people with SMI’ initial contact with the justice system.

Although the results of these two samples on the frequency and interaction of co-occurring disorders (i.e., pathoplasticity) for PMI-CJI have important implications, this work is not without limitations. These data were cross-sectional which precludes determinations of causality; future research should continue exploring the relation between mental illness and criminal risk using longitudinal designs. Although self-reported substance use diagnosis was not significantly associated with the outcomes in Sample 2, we could not test if substance use diagnoses were significantly associated with outcomes in Sample 1, which is another limitation. Both studies were limited to patient self-report of criminal justice involvement. Although over- or underreporting of criminal justice involvement is unlikely (see e.g., Kroner et al., 2006), the present study is limited by the absence of file review to confirm participant report. The study would have also been strengthened by the inclusion of a community mental health outpatient sample as a control group from which to compare the severity of symptoms with levels of criminal risk. Because information was not gathered on the number of participants who were approached by the study team and did not agree to participate, there may have been a self-selection bias, which would limit our ability to generalize to all people receiving inpatient psychiatric care in these facilities. In addition, our samples were relatively small, demographically homogenous, collected in the same region of the United States, and required to meet our inclusion criteria; therefore, further replication is needed prior to generalizing these findings. Lastly, criminal justice involvement was operationalized as a history of a misdemeanor and/or felony conviction, which excludes other interactions with the criminal justice system (e.g., arrests). Thus, future studies should consider people with mental illness’ various interactions with the criminal justice system.

The results from our study demonstrate the importance of community mental health systems addressing not only the psychiatric treatment needs of individuals with people with mental illness involved in the criminal justice system, but also the criminogenic needs. The finding that approximately 50% of individuals housed in inpatient community mental health systems are criminal justice involved is a notable statistic and likely an underestimation of the percentage of people with mental illness who have had interactions with the criminal justice system and were not charged with a misdemeanor and/or felony offense. These findings clearly necessitate clinical attention and enhanced treatment designed to address the complicated co-occurring relation between mental illness and criminal risk. To better serve PMI-CJI, we must alter our treatment approaches by implementing programs specifically designed to treat these comorbid issues (e.g., Changing Lives and Changing Outcomes), and we must alter our service delivery model in order to best meet the client’s needs (see Bonfine et al., 2020).

Impact Statement.

This study found more than half of two psychiatric inpatient samples were previously criminal justice-involved, and criminogenic needs were significantly, positively associated with mental illness symptoms. These findings underscore the need to provide better, more interdisciplinary treatment for psychiatric inpatients with criminal risk factors.