Abstract
Changing Lives and Changing Outcomes (CLCO) was developed to address the unique treatment needs of individuals with co-occurring mental illness and criminogenic risk. Previous evaluations of CLCO demonstrated effectiveness for male participants, but did not examine treatment effectiveness across participant sex. Therefore, the aim of this study was to examine differences in treatment outcomes among male and female probationers receiving CLCO. Participants were assessed pre- and posttreatment. Results indicated positive treatment effects across both psychiatric and criminogenic domains, including psychiatric symptomology and global mental health functioning, medication adherence, attitudes toward mental illness recovery, and antisocial cognitions. Female participants demonstrated enhanced treatment responsiveness relative to males across several domains. This study underscores the effectiveness of the program in addressing both mental health and criminogenic needs in correctional populations, while highlighting the differential impact of the program across participant sex. Implications, future directions, and limitations of the present study are discussed.
It is well known that rates of mental illness in the United States criminal justice system are disproportionately higher than those of the general population. Rates of mental illness in probation and parole samples are estimated to be two to four times that of the general population (Prins & Draper, 2009). By year-end 2016, approximately 6,613,500 people were under the supervision of the United States criminal justice system (jail, prison, probation, or parole), over 4 million of whom were under community supervision (Kaeble & Cowhig, 2018). These large numbers and the population’s high rates of mental illness highlight the importance of preparing criminal justice-involved persons with mental illness (CJ-PMI) for desistance from crime and, for those who are incarcerated, successful community re-entry.
Although treatment programs for CJ-PMI need to address co-occurring mental health and criminogenic needs (see Bartholomew, Morgan, Mitchell, & Van Horn, 2018; Gross & Morgan, 2013; Morgan, Fisher, Duan, Mandracchia, & Murray, 2010; Skeem, Manchak, & Peterson, 2011; Wilson et al., 2014; Wolff, Morgan, & Shi, 2013), such programs are sparse. One program specifically developed to address the co-occurring issues of mental health and criminogenic needs for CJ-PMI is Changing Lives and Changing Outcomes (CLCO; Morgan, Kroner, & Mills, 2018). CLCO is a comprehensive nine-module program that integrates psychotherapeutic and psychoeducational strategies for CJ-PMI. The program consists of 77 sessions with treatment plans for each session. The treatment plans include structured in-group activities, handouts, discussion, and homework exercises to teach and reinforce participants’ learning about their mental illness and criminal risk (referred to as criminalness in the treatment program, including behavior that breaks laws and violates social conventions and/or the rights and wellbeing of others, but is not restricted to illegal behavior; Morgan et al., 2018) and to develop effective strategies for coping with these co-occurring issues. The overarching goal of CLCO is “to enhance quality of life, reflected through improved mental health status and reduced criminal and psychiatric recidivism” (Morgan et al., 2018; p. xiv).
Prior examinations of CLCO have demonstrated reductions in male participants’ mental health symptomology, distress, and aspects of criminal thinking (Morgan et al., 2012), as well as male and female participants’ demonstrated learning of treatment content (Van Horn et al., 2018). Despite evidence that males and females have different learning styles (Honigsfeld & Dunn, 2003; Reese & Dunn, 2007), and the need for gender-responsive treatment approaches to improve outcomes for justice-involved women (see Bloom, Owen, & Covington, 2003; Van Voorhis, 2012), previous analyses of CLCO data did not explore potential sex differences in participant treatment outcomes. This is particularly important given findings from a recent meta-analysis indicating that both gender-responsive and gender-neutral programming reduced recidivism in justice-involved women, although the more methodologically rigorous studies showed gender-responsive programs were more effective than gender-neutral ones (Gobeil, Blanchette, & Stewart, 2016). In response to the evidence in favor of gender-responsive programs for justice-involved women and calls for gender-responsive program design and delivery (Covington & Bloom, 2007), the current study aims to explore if males and females similarly benefit from CLCO.
Specifically, the current study examines the effectiveness of CLCO for male and female CJ-PMI on probation and court-committed to a residential treatment facility for moderate- to high-risk justice-involved persons with mental health and substance abuse problems. Importantly, this is the first study to explore potential sex differences in outcomes for male and female CLCO treatment recipients. We hypothesized that because CLCO was developed from best practices from correctional and psychiatric rehabilitation literature, there would be no significant treatment differences between males and females across outcomes of interest. More specifically, we hypothesized that both male and female participants would demonstrate significant treatment improvement on measures of mental health outcomes, outcomes associated with criminal risk, and treatment satisfaction, but that there would not be significant sex differences in treatment responsiveness.
Method
Treatment Program
Residents at the 60-bed Dual Diagnosis Residential Program (DDRP) treatment facility were referred by Harris County, Houston, Texas probation. Residents are classified as moderate- to high-risk inmates with dual diagnosis to include a severe mental illness as defined by the Harris County mental health authority and the Texas Council on Offenders with Medical and Mental Impairments (TCOOMMI). These residents are committed with a Global Assessment of Functioning (GAF) score of less than 50. As a dual diagnosis facility, residents also present with significant substance abuse disorders. The DDRP is a 150-180 day treatment program with the primary treatment program being CLCO.
CLCO is a comprehensive modular treatment program that utilizes a bi-adaptive model of intervention by targeting mental illness and criminalness to improve functional outcomes for CJ-PMI. The aim of this treatment model is to help participants maximize adaptive behaviors to optimize functioning while reducing psychiatric relapse and criminal reoffending. CLCO includes a comprehensive and structured treatment manual that provides clinicians with a guide for treating CJ-PMI. The manual includes a treatment plan for each session with specific structured exercises (for both in-group and out-of-group work) designed to teach learning objectives each session. The program incorporates a psychosocial rehabilitation model, social learning paradigm, and cognitive-behavioral model for change, although cognitive-behavioral theory is more prevalent and apparent throughout the manual. The accompanying participant workbook includes worksheets, handouts, and homework assignments to be used in each session.
CLCO consists of 77 sessions (68 of which include the use of homework assignments) across a total of 9 thematic modules, with a target session length of 1 and 1/2 to 2 hours (estimated treatment dosage of 150 hours). Number of treatment sessions per week is determined by setting logistics. For example, in outpatient settings, the authors recommend no more than one to two sessions per week; however, for intensive residential (institutional) settings, the authors recommend three to five sessions per week. For the present investigation, five modules were completed at the treatment facility (approximately 3 months) and the remaining four modules in the community postrelease. Sessions were run at the DDRP 5 days per week, Monday to Friday, except holidays or on very rare occasions that therapists were unavailable (e.g., training, workshops, etc.). The program utilized an open enrollment policy, such that residents entered the program at any module; however, residents could not enter once a module was already underway. Residents were required to complete Module 1: Preparing for Change prior to entering a group. Treatment providers (generally master’s-level clinicians) were trained by the lead author and participated in bi-weekly consultation calls (typically of 30 minutes) with the lead author during the first year of program implementation. The purpose of these calls was to address CLCO programmatic and implementation/fidelity issues, as well as to consult on resident issues using the CLCO framework. 1
Participants
Residents in the present study included adult felony justice-involved persons (N = 186) from the DDRP program including 117 males (60.2%) and 74 females (39.8%). The average age of male participants was 33, while the average age of female participants was 37. The majority of male participants identified as White (42.9%), and they reported an average sentence length of 40 months. The majority of female participants identified as Black (45.9%), and they reported an average sentence length of 41.26 months. Approximately half of both male and female participants reported seeking mental health services, and more than 70% of both male and female participants reported taking psychotropic medication. Table 1 provides a detailed summary of participant demographic information.
Participant Demographics by Sex (N = 186)
Note. GED = general educational development; N/A = not available.
Measures
Demographic Form
Residents completed a demographic form when completing pretreatment measures. This form was used as an efficient tool for obtaining basic demographic information such as age, race and ethnicity, relationship status, educational information, index offense and sentencing information, as well as mental health treatment history.
The Diagnostic and Statistical Manual of Mental Disorders (fifth edition; DSM-5) Self-Rated Level 1 Cross-Cutting Symptom Measure—Adult
The DSM-5 Self-Rated Level 1 Cross-Cutting Symptom Measure—Adult (referred to here as DSM-5; American Psychiatric Association, 2013) is a 23-item self-report measure assessing mental health concerns that cut across specific psychiatric disorders and that may serve as a prelude to more intensive assessment. Items on the DSM-5 assess 13 mental health domains, including (1) Depression, (2) Anger, (3) Mania, (4) Anxiety, (5) Somatic Issues, (6) Suicidal Ideation, (7) Psychosis, (8) Sleep Issues, (9) Memory, (10) Repetitive Thoughts/Behaviors, (11) Dissociation, (12) Personality, and (13) Substance Use. Individuals indicate the extent to which each item has been bothersome in the previous 2 weeks using a 5-point Likert-type scale ranging from 0 (none/not at all) to 4 (severe/nearly every day). For the present investigation, items comprising each domain were summed to create 13 domain total scores. The DSM-5 has adequate reliability as test–retest reliability coefficients (intraclass correlation coefficients [ICCs]) ranged from .53 to .97 across the 23 items, with all but two items demonstrating “good” or better reliability (Narrow et al., 2013). The measure has further demonstrated clinical utility in assessing mental health concerns (Moscicki et al., 2013) and in the negative prediction of anxious, manic, and psychotic symptoms in a correctional sample, though its utility in positive prediction demands further research (Bastiaens & Galus, 2018). Cronbach’s alpha for the present study were as follows: Depression (.52–.75), Mania (.51–.55), Anxiety (.74–.76), Somatic Issues (.57–.68), Psychosis (.75–.83), Repetitive Thoughts/Behaviors (.70–.80), Personality (.70–.77), and Substance Use (.76–.77) across administrations. Internal consistency could not be calculated for Anger, Suicidal Ideation, Sleep Issues, Memory, and Dissociation due to only one item comprising each of these scales.
Medication Adherence Rating Scale (MARS)
The MARS (Thompson, Kulkarni, & Sergejew, 2000) is a 10-item self-report scale assessing medication compliance. Derived from two existing medication adherence scales, the MARS utilizes Yes/No response options to gauge the degree to which individuals are engaging in appropriate medicating behavior. Factor analyses indicate three underlying factors: (1) Adherence Behavior, (2) Attitudes Toward Medications, and (3) Negative Side Effects of Medication, with a total global score ranging from 0 (poor adherence) to 10 (high adherence). Original reliability estimates were found to be .75 for internal consistency (Cronbach’s alpha) and .72 for test–retest reliability, suggestive of appropriate reliability (Thompson et al., 2000). Moreover, the MARS has demonstrated good internal, convergent, and construct validity, with strong correlations with similar measures of medication adherence (Thompson et al., 2000). The adequate internal consistency of the MARS as well as its three-factor structure has been confirmed by an additional investigation (Fialko et al., 2008). Cronbach’s alpha for the present study ranged from .65 to .72 across administrations.
Recovery Assessment Scale (RAS)
The RAS (Giffort, Schmook, Woody, Vollendorf, & Gervain, 1995) is a self-report measure assessing individuals’ experience and perceptions as they relate to recovery from mental health-related concerns, and is available in both 24- and 41-item formats. The RAS items coalesce into five primary subscale domains: (1) Personal Confidence and Hope, (2) Willingness to Ask for Help, (3) Goal and Success Orientation, (4) Reliance on Others, and (5) Not Dominated by Symptoms. Individuals indicate the degree to which they agree with each item statement using a 5-point Likert-type scale ranging from 1 (strongly disagree) to 5 (strongly agree). Domain scores are calculated by summing the items comprising each subscale, with greater scores indicating more positive experiences and perceptions of themselves within the recovery process. Internal consistency, reliability, and validity of the RAS have been deemed good (Burgess, Pirkis, Coombs, & Rosen, 2011; Law, Morrison, Byrne, & Hodson, 2012; Salzer & Brusilovskiy, 2014), with a Cronbach’s alpha of .93 for the 41-tem measure (Corrigan, Giffort, Rashid, Leary, & Okeke, 1999). Further studies have attested to the internal consistency of the 24-item RAS, with consistency estimates (Cronbach’s alpha) ranging from .74 to .87 (Corrigan, Salzer, Ralph, Sangster, & Keck, 2004). A shortened 22-item version of the RAS was utilized for the purposes of the present study. Cronbach’s alpha for the present study were as follows across administrations: Personal Confidence and Hope (.83–.86), Willingness to Ask for Help (.83–.94), Goal and Success Orientation (.80–.86), Reliance on Others (.77–.80), and Not Dominated by Symptoms (.70–.73).
Positive and Negative Affect Schedule (PANAS)
The PANAS (Watson, Clark, & Tellegen, 1988) is a 20-item self-report measure assessing positive and negative affect that yields two 10-item subscale scores: Positive Affect Score and Negative Affect Score. Examples of positive affect items include attentive, determined, and strong; conversely, examples of negative affect items include hostile, guilty, and upset; individuals indicate the extent to which each word item describes their current or recent affective experience using a 5-point Likert-type scale ranging from 1 (very slightly or not at all) to 5 (extremely). Positive and Negative Affect subscale scores are calculated by summing across subscale items, with possible scores ranging from 10 to 50. Higher scores indicate greater positive and negative affects, respectively. The PANAS has demonstrated “good” internal consistency (Serafini, Malin-Mayor, Nich, Hunkele, & Carroll, 2016), with coefficients (Cronbach’s alpha) ranging from .85 to .90 for the Positive Affect scale and .84 to .91 for the Negative Affect scale across different timeframes and samples (Watson et al., 1988). Test–retest reliability has further been found to be adequate, ranging from .47 to .68 for Positive Affect and .39 to .71 for Negative Affect across timeframes (Watson et al., 1988). The PANAS has been shown to be consistent with measures of similar constructs and inversely related to dissimilar measures, attesting to its convergent and discriminant validity, respectively (Serafini et al., 2016). Original data indicated no difference in PANAS scores across gender (Watson et al., 1988); however, more recent studies have suggested higher negative affect among women, and lower negative and higher positive affect among those who are criminal justice involved (Serafini et al., 2016). Cronbach’s alpha for the present study ranged from .87 to .89 for the Negative Affect scale and was .90 for the Positive Affect scale across administrations, suggestive of good internal consistency.
Psychological Inventory of Criminal Thinking Styles—Short Form (PICTS-SF)
The PICTS-SF (Walters, 2006) is a 35-item self-report measure assessing various facets of criminal thinking (i.e., thinking processes that support and maintain a criminal lifestyle). Derived from the 80-item PICTS (Walters, 1995), the PICTS-SF is comprised of seven thinking style scales (Mollification, Cutoff, Entitlement, Sentimentality, Superoptimism, Cognitive Indolence, Discontinuity), two content scales (Current, Historical), two factor scales (Problem Avoidance, Self-Assertion/Deception), and two composite scales (Proactive Criminal Thinking, Reactive Criminal Thinking). Individuals indicate the degree to which they agree with each item using a 4-point Likert-type scale ranging from 1 (disagree) to 4 (strongly agree), with greater scores suggesting greater criminal thinking. Only the Proactive and Reactive Criminal Thinking scales were utilized in the present study. Internal consistency estimates (Cronbach’s alpha) indicate that the scales of the full PICTS are reliable for both male (range: .61-.94) and female (range: .54-.93) justice-involved persons (Walters, 2013). Test–retest stability coefficients for each of the scales further reflect sound reliability, with ranges of .73 to .96 and .47 to .92 after 2 and 12 weeks, respectively (see Walters, 2013, for review). Validity of the PICTS scales has been demonstrated via adequate correlations with other crime-related measures and outcomes (see Walters, 2013 for full review), including correlations between the Proactive scale and previous arrest for proactive crimes (e.g., burglary) as well as between the Reactive scale and arrests for reactive crime (e.g., assault). Female justice-involved persons have been shown to demonstrate greater Reactive Criminal Thinking relative to males; however, this may be constrained to White individuals (Walters, 2018). Cronbach’s alpha for the present study ranged from .89 to .90 for the Proactive scale and from .90 to .92 for the Reactive scale across administrations.
Measures of Criminal Attitudes and Associates (MCAA)
The MCAA (Mills & Kroner, 2001) is a two-part self-report questionnaire assessing antisocial attitudes, as well as relationships with criminal associates that contribute to risk for criminal activity. Part A requires that individuals provide information regarding four adults with whom they frequently associate, and assesses for the presence of both criminal activity among those adults in addition to the amount of time spent with each associate. Scores for Part A are derived by multiplying the number of Yes responses to each of the four associates (range: 0-4) by the time spent with each associate (range: 0-4), and summing these values together to form an overall Criminal Friend Index. Scores for Part A therefore range from 0 to 64, with higher scores indicative of greater involvement with criminal associates. Part B consists of four subscales aimed as assessing individuals’ degree of criminal attitudes across four domains: (1) Attitudes toward Violence, (2) Attitudes toward Entitlement, (3) Attitudes toward Associates, and (4) Antisocial Intent. Individuals are asked to mark Agree or Disagree for each of the 46 items in Part B, with global score and subscale scores calculated by summing across items. As such, Part B total scores range from 0 to 46, with higher scores suggestive of greater criminal attitudes. The MCAA has demonstrated sound reliability, with internal consistency estimates (Cronbach’s alpha) of .89 for total score and .63 to .84 for subscales among an offender sample, and test–retest reliabilities of .82 and .66 to .79, respectively (Mills & Kroner, 2001). In addition, the MCAA has demonstrated strong convergent validity with similar measures and is predictive of new charges (area under the ROC curve [AUC] range: .56-.73) as well as suspension/revocation (AUC range: .55-.64) following release (Mills & Kroner, 2001). Cronbach’s alpha for the present study were as follows: Full measure (.85–.87), Attitudes toward Violence (.78–.82), Attitudes toward Entitlement (.66–.68), Attitudes toward Associates (.73–.74), and Antisocial Intent (.73–.75) across administrations.
Brief Situational Confidence Questionnaire (BSCQ)
The BSCQ (from Breslin, Sobell, Sobell, & Agrawal, 2000) is an eight-item self-report measure designed to assess the degree to which individuals are confident in being able to refrain from substance use in eight types of situations: (1) Unpleasant Emotions, (2) Physical Discomfort, (3) Pleasant Emotions, (4) Testing Control over Use (e.g., thinking drugs are no longer a problem), (5) Urges and Temptations, (6) Conflict with Others, (7) Social Pressure to Use, and (8) Pleasant Times with Others. Individuals indicate their confidence using a sliding scale ranging from 0% (not at all confident) to 100% (totally confident). Pre- and posttreatment scores for the present study were calculated by rounding residents’ answers to the nearest 10% for each of the eight situations and taking the average, thus creating a single unitary score at each time point. Psychometric analyses have indicated strong correspondence between the BSCQ and similar measures, underscoring the questionnaire’s construct and concurrent validity (Breslin et al., 2000). Internal consistency estimates ranging from .85 to .89 are suggestive of strong reliability (Breslin et al., 2000; O’Sullivan, Watts, Xiao, & Bates-Maves, 2016). Cronbach’s alpha for the present study ranged from .89 to .91 across administrations, which is indicative of good internal consistency.
Working Alliance Inventory (WAI)
The WAI (Horvath & Greenberg, 1989) consists of 36 self-report items assessing three dimensions of the working alliance and clients’ perceptions of their therapist and therapy more broadly. Dimensions assessed include (1) Congruence Between Client and Therapist Views on the Therapeutic Process and What Activities Therapy Will Entail, (2) Congruence Between Client and Therapist Goals for Therapy, and (3) The Strength of the Bond, Trust, Acceptance, and Comfort Between the Two Parties. Clients indicate the degree to which each item is true of their alliance experience using a 7-point Likert-type scale ranging from 1 (not at all true) to 7 (very true). Greater scores indicate greater strength in the therapeutic relationship, with domain scores ranging from 12 to 84 and total scores from 36 to 252. Internal consistency estimates (Cronbach’s alpha) were found to be high, ranging from .89 to .92 across the three subscales and .93 for the full measure (Horvath & Greenberg, 1989). Adequate convergent and predictive validity of the WAI has further been ascertained (Horvath & Greenberg, 1989). Though both client and therapist forms are available, only the client version of the WAI was utilized in the present study. Cronbach’s alpha for the present study was .94 for the full measure and ranged from .83 to .85 across the three subscales.
The Client Satisfaction Questionnaire (CSQ-8)
The CSQ-8 (Larsen, Attkisson, Hargreaves, & Nguyen, 1979) consists of eight self-report items assessing client satisfaction with psychological services and clinical care. Clients utilize a 4-point Likert-type scale to indicate the degree to which they are satisfied with current or prior services. The CSQ-8 provides a single score of client satisfaction ranging from 8 to 32, with greater scores suggestive of greater satisfaction. Internal consistency of the CSQ-8 has been calculated to be .93, indicative of good reliability, while moderate correlations with mental health outcome measures suggest good validity as a measure of client satisfaction (Attkisson & Zwick, 1982). Additional studies have attested to the strong concurrent validity and have confirmed the single-factor structure of the CSQ-8 (Kelly et al., 2018). Cronbach’s alpha for the present study was .92, indicative of strong internal consistency.
Procedure
Data for the present investigation were obtained during the regular course of service provision at the treatment facility. Residents completed the measures used in this program evaluation during the regular course of treatment and in accordance with the CLCO recommended assessment plan. Assessment followed a pre–post treatment paradigm, with residents completing measures at two time points during their participation in the CLCO program.
Initial assessment took place early in residents’ treatment engagement (typically just prior to beginning the program; pre-intervention period) as well as approximately 4 months following this initial assessment (4-month assessment period). Residents completed the BSCQ, MARS, PANAS, RAS, MCAA, DSM-5, and PICTS-SF at both the pre-intervention and 4-month assessment periods. Conversely, the CSQ-8 and WAI were only completed at the 4-month follow-up, thus providing a single time point for these two measures. Assessments were administered by treatment providers, who encouraged residents to complete the assessments to the best of their ability. Completed assessments were subsequently collected, scanned, and sent to the lead author for clinical interpretation. Assessment results were returned to DDRP treatment providers to be used in session with residents and to monitor resident progress, in accordance with the CLCO program.
Data Analysis
Data analysis for the present investigation followed a multi-step approach. Prior to statistical analysis, all pre- and posttreatment data were screened for entry errors and corrected on a case-by-case basis (i.e., with the correct values for a particular participant). Second, given the utilization of a pre–post research design to examine treatment responsivity following CLCO programming, only those participants with assessment data at both the pretreatment and 4-month posttreatment time points were considered for inclusion in subsequent analyses. This process yielded the final sample of participants (N = 186). Third, diagnostic analyses, including frequency estimations, were used to check for missingness among participant data. To mitigate sample bias, participants with missingness exceeding 20% for a given measure were excluded from analyses including that measure. Moreover, all statistical analyses were conducted separately for each individual measure to maximize sample size. As such, sample size varies per analysis. For those participants with missingness < 20%, multiple imputation methods (see Rubin, 1987 for review) conducted using SPSS Version 24.0 were used to compute five imputed data sets (five iterations) for each measure across pre- and posttreatment time points. These data sets were subsequently amalgamated and pooled estimates used for subsequent analyses for each measure. All measures were imputed with exception of the CSQ-8, DSM-5, and WAI due to lack of missingness.
Following correction of entry errors and imputation, primary data analysis included a series of mixed-design analysis of variance (ANOVA) to examine the impact of sex and time point on the outcome of interest (i.e., postassessment scores). Participant sex was therefore entered as a between-subjects factor and time point entered as a within-subjects factor. Post hoc paired samples t tests were subsequently used to examine pre- and posttreatment assessment scores among male and female participants following a significant ANOVA omnibus test. In light of the number of comparisons made and as an alternative to a Bonferroni adjusted alpha level, a binomial probability was calculated to examine the probability of finding 52 or more significant results out of the 79 total analyses. This calculation indicated that the likelihood of finding 52 or more significant results was less than 0.003; thus, no adjustments were made to the alpha level. All primary assumptions for the employed statistical methods were met. Outlying data points were further examined for influence; these outliers were found to be noninfluential, and as such no outliers were removed. Cohen’s d served as an indicator of treatment effect size and were interpreted consistent with standard conventions such that .20 is considered a small treatment effect, .50 is considered a moderate treatment effect, and .80 is considered a large treatment effect (Cohen, 1988).
Results
Psychopathology
Depression
A repeated measures ANOVA was conducted to examine the effect of sex and time point on depressive symptoms. There was a statistically significant interaction between the effects of sex and time point on depressive symptoms, F(1, 169) = 9.43, p = .002. Paired samples t tests were conducted to examine post hoc comparisons from pre- to posttreatment between sexes, and both females’ and males’ symptoms significantly decreased from pre- to posttreatment with a larger treatment effect for females than males. See Figure 1 for Cohen’s d effect sizes for depression, anger, mania, anxiety, dissociation, and somatic illnesses. See Table 2 for pre- and postmean and standard deviation values and t test results for depression, anger, mania, anxiety, somatic illnesses, and anxiety.

Differences in Psychopathology Per Sex
Pre–Post CLCO DSM Scores for Males and Females
Note. CLCO = Changing Lives and Changing Outcomes; DSM = Diagnostic and Statistical Manual of Mental Disorders.
p < .05. **p < .01. ***p < .001.
Anger
A repeated-measures ANOVA was conducted to examine the effect of sex and time point on anger resulting in a statistically significant interaction between the effects of sex and time point on anger, F(1, 165) = 16.59, p < .001, with significantly larger treatment effect sizes for females than males. Paired samples t tests were conducted to examine post hoc comparisons from pre- to posttreatment between sexes. As shown in Table 2, females’ symptoms significantly decreased from pre- to posttreatment, whereas males’ symptoms did not statistically significantly change from pre- to posttreatment, although a small treatment effect was detected (see Figure 1).
Mania
Results of a repeated measures ANOVA examining the effect of sex and time point on symptoms of mania indicated a statistically significant interaction, F(1, 160) = 8.51, p = .004. Paired sample t tests indicated that although both females’ and males’ symptoms of mania significantly decreased from pre- to posttreatment (see Table 2), treatment effects were larger for females than males (see Figure 1).
Anxiety
A repeated-measures ANOVA was conducted to examine the effect of sex and time point on symptoms of anxiety resulting in a statistically significant interaction, F(1, 159) = 5.06, p = .03. Paired samples t tests examined post hoc comparisons from pre- to posttreatment between sexes and indicated both females’ and males’ symptoms significantly decreased from pre- to posttreatment (see Table 2), although the treatment effect was larger for females than males (see Figure 1).
Somatic Illnesses
Results of a repeated-measures ANOVA indicated a statistically significant interaction between the effects of sex and time point on somatic symptoms, F(1, 161) = 7.84, p = .006. Paired samples t tests used for post hoc comparisons from pre- to posttreatment between sexes found that both females’ and males’ somatic symptoms significantly decreased from pre- to posttreatment (see Table 2), but females obtained larger treatment effects than their male counterparts (see Figure 1).
Dissociation
Results of a repeated-measures ANOVA also indicated a statistically significant interaction between the effect of sex and intervention time point on dissociative symptoms, F(1, 166) = 5.83, p = .02. Post hoc comparisons showed that both females’ and males’ symptoms significantly decreased from pre- to posttreatment (see Table 2), while treatment effects were larger for females than males (see Figure 1).
Other Psychopathology and Distress Outcomes
Repeated-measures ANOVAs were conducted to examine the effect of sex and intervention time point on suicidal ideation, psychosis, sleep issues, memory, repetitive thoughts and behavior, personality functioning, and substance abuse. Scores significantly decreased from pre- to posttreatment for Suicidal Ideation, F(1, 162) = 10.45, p = .001, d = .51; psychosis, F(1, 165) = 21.05, p < .001, d = .70; Sleep Issues, F(1, 164) = 52.68, p < .001, d = 1.12, Memory, F(1, 164) = 10.64, p = .001, d = .51; Repetitive Thoughts and Behavior, F(1, 160) = 25.81, p < .001, d = .81; Personality Functioning, F(1, 162) = 40.97, p < .001, d = 1.00; and Substance Abuse, F(1, 160) = 39.61, p < .001, d = 1.00 (see Table 3); however, none of the interactions were significant (p > .05), meaning that there were no significant differences in scores between sexes across time points on these outcomes of interest.
Pre–Post CLCO Scores on Mental Health Measures
Note. DSM-5 = DSM-5 Level 1 Cross-Cutting Symptom Measure; MARS = Medication Adherence Rating Scale; RAS = Recovery Assessment Scale; PANAS = Positive and Negative Affect Scale; BSCQ = Brief Situation Confidence Questionnaire; WAI = Working Alliance Inventory.
p < .05. **p < .01. ***p < .001.
Medication Adherence
A repeated-measures ANOVA was conducted to examine the effect of sex and intervention time point on medication adherence. Scores significantly increased from pre- to posttreatment, F(1, 165) = 67.37, p < .001, d = 1.25 (see Table 3). The interaction, however, was not significant (p = .18) such that there were no statistically significant differences in scores between sexes across time points.
Perceptions of Individual Recovery
A repeated-measures ANOVA was conducted to examine the effect of sex and intervention time point on clients’ personal confidence and hope, willingness to ask for help, goals and success orientations, reliance on others, and perceptions of their symptoms. Scores significantly increased from pre- to posttreatment for personal confidence and hope, F(1, 127) = 13.60, p < .001, d = .67; willingness to ask for help, F(1, 111) = 5.22, p = .02, d = .46; goals and success orientations, F(1, 126) = 6.84, p = .01, d = .46; reliance on others, F(1, 111) = 9.94, p = .002, d = .59; and perceptions of their symptoms, F(1, 113) = 12.22, p = .001, d = .67 (see Table 3). None of the interactions were significant (p > .05), meaning that there were not significant differences in recovery scores between sexes across time points.
Affect
A repeated-measures ANOVA was conducted to examine the effect of sex and intervention time point on positive and negative affect. Positive affect significantly increased from pre- to posttreatment, F(1, 148) = 41.41, p < .001, d = 1.06, whereas negative affect significantly decreased from pre- to posttreatment, F(1, 148) = 18.47, p < .001, d = .70 (see Table 3). The interactions were not significant for either positive affect (p = .62) or negative affect (p = .10), meaning that there were no significant differences in scores between sexes across time points.
Criminogenic Cognitions and Antisocial Associates
A repeated-measures ANOVA was conducted to examine the effect of sex and time point on proactive and reactive criminal thinking scores. There were no statistically significant changes in proactive criminal thinking scores from pre- to posttreatment (p = .73), and the interaction between sex and time point was also not significant (p = .94). There was, however, a statistically significant decrease in posttreatment reactive criminal thinking scores (M = 86.47, SD = 29.34) compared to their pre-treatment scores, M = 99.59, SD = 32.20; F(1, 148) = 27.43, p < .001. The interaction was also not statistically significant (p = .31), meaning that there were not significant differences in scores between sexes across time points. See Figure 2 for Cohen’s d effect sizes for changes in criminogenic cognitions and antisocial associates

Changes in Criminogenic Cognitions and Associates
A repeated-measures ANOVA was conducted to examine the effect of sex and time point on attitudes toward violence, entitlement, associates and antisocial intent. Scores significantly decreased from pre- to posttreatment for attitudes toward violence, F(1, 148) = 24.07, p < .001; entitlement, F(1, 173) = 9.29, p = .003; and antisocial intent, F(1, 171) = 32.15, p < .001. Scores did not statistically significantly differ for attitudes toward associates (p = .09). The interactions were not statistically significant for attitudes toward violence (p = .88), entitlement (p = .20), or antisocial intent (p = .61), meaning that there were not statistically significant differences in scores between sexes across time points.
Due to the unlikelihood that participants’ criminal associations would change within the rather short 4-month timespan, participants’ number of criminal friends and Criminal Friend Index (as measured by Part A of the MCAA) were examined at the pre-treatment time period using descriptive analyses. Male participants endorsed a mean of 1.72 (SD = 1.45) criminal associates, with a mean Criminal Friend Index of 8.30 (SD = 9.70). Similarly, female participants endorsed a mean of 1.66 (SD = 1.50) criminal associates, with a mean Criminal Friend Index of 11.91 (SD = 13.91). Collectively, these results indicate that, of the four associates reported by male and female participants in the MCAA, on average, nearly half are characterized as criminal associates (i.e., have been previously and/or currently are involved in the criminal justice system). Moreover, male and female participants reported spending a significant amount of time in the presence of such associates.
Confidence in Avoiding Drugs or Alcohol
A repeated-measures ANOVA was conducted to examine the effect of sex and time point on situational confidence in avoiding drugs or alcohol in high-risk situations and resulted in a statistically significant interaction, F(1, 153) = 5.72, p = .02. Paired samples t tests examined post hoc comparisons from pre- to posttreatment between sexes and indicated females’ and males’ scores significantly improved from pre- to posttreatment (see Table 3). Examination of effect sizes further indicated a larger treatment effect for females than males (see Figure 3).

Differences in Situational Confidence Per Sex
Working Alliance and Client Satisfaction
A one-way repeated-measures ANOVA was conducted to examine the effect of sex on overall client perceptions of their therapist and therapy, client’s agreement with the therapist’s goals for therapy, and the degree to which clients trusted their therapist. Females reported statistically significantly greater overall positive perceptions of their therapist and therapy, F(1, 141) = 4.46, p = .04, and agreement with the therapist’s goals for therapy, F(1, 141) = 7.06, p = .01, than their male counterparts (see Table 3); however, females and males did not significantly differ on method agreement (p = .08) or trust (p = .15; see Figure 4 for effect sizes differences). A one-way repeated measures ANOVA was conducted to examine the effect of sex on client satisfaction. There was a statistically significant difference between females (M = 27.44, SD = 3.95) and males (M = 25.61, SD = 5.46) on client satisfaction, F(1, 180) = 6.02, p = .02, d = .37, per the CSQ-8.

Differences in Working Alliance Per Sex
Discussion
The co-occurrence of mental health and criminogenic needs among offender populations presents unique challenges for practitioners and staff operating within correctional settings. The present study examined the effectiveness of a psychosocial treatment program, CLCO, in addressing co-occurring mental illness and criminalness in a sample of adult justice-involved persons from a residential treatment facility. This study continues prior investigations of CLCO, while simultaneously being the first to examine the efficacy of the program as it relates to potential sex differences between male and CJ-PMI.
Main Findings
Results of the current investigation provide additional support regarding the effectiveness of CLCO in treating mental illness and criminalness in justice-involved populations. With regard to mental health–related outcomes, participants demonstrated significant reductions in psychiatric symptomology across numerous domains, including (1) Depression, (2) Anger, (3) Mania, (4) Anxiety, (5) Somatic Concerns, (6) Suicidal Ideation, (7) Psychosis, and (8) Sleep Issues. Participants further demonstrated significant reductions in psychiatric symptoms associated with (9) Memory Concerns, (10) Problematic Repetitive Thoughts/Behavior, (11) Dissociative Experiences, (12) Personality Dysfunction, and (13) Substance Use. Concomitant with these reductions in mental illness symptomology were improvements in more global mental health functioning from pre- to post-intervention; namely, greater positive affect, reduced negative affect, and greater confidence in one’s ability to refrain from substance use in critical situations (e.g., when experiencing unpleasant emotions or social pressure). Medication adherence behaviors and attitudes toward recovery were similarly enhanced—participants exhibited greater hope, orientation to personal goals, willingness to seek help when needed, and refusal to be dominated by their symptoms across time periods. As such, participants displayed not only reduced psychiatric symptoms, but improved perceptions and behaviors instrumental to treatment and recovery as a result of engagement in the CLCO program. Cohen’s d effect sizes for psychiatric symptomology and mental health treatment factors ranged from 0.18 to 1.25, with many in the moderate to high range.
Additional results highlight the positive impact of CLCO upon criminalness, and in particular facets of cognition and attitudes that underlie criminal behavior and criminal justice involvement. Indeed, from pre- to post-intervention, participants displayed significantly lower attitudes toward violence, as well as entitled attitudes and antisocial intent. That is, participants demonstrated reductions in beliefs supportive and tolerant of violence (e.g., using violence for goal achievement), egocentric views (e.g., the right to fulfill their own needs), and intention to commit future antisocial acts, respectively. Moreover, programming was associated with reduced levels of reactive criminal thinking post-intervention. Participants therefore exhibited fewer cognitions associated with engagement in reactive criminal behavior (i.e., committing crime impulsively or spontaneously). The impact of CLCO on criminalness generally corresponded to medium to large effect sizes, suggesting that the program exerts a medium to large effect upon criminal thinking and attitudes.
CLCO and Sex Differences
Notwithstanding the general effectiveness of CLCO on psychiatric symptomology and criminalness, male and female justice-involved persons in the present study were found to differ on several fronts in response to programming. On one hand, though participants collectively demonstrated reduced depressive, anger, manic, anxious, somatic, and dissociative symptomology at post-intervention—and thus benefited from treatment to some degree—reductions in these domains were greater for females than for males. Although we hypothesized no sex treatment differences, findings that females benefited more than males is consistent with prior research showing that female justice-involved persons have different psychological needs than their male counterparts (see Loper, Carlson, Levitt, & Scheffel, 2009; Salisbury & Van Voorhis, 2009). Not surprisingly given prior findings (see Binswanger et al., 2010; Diamond, Wang, Holzer, Thomas, & Cruser, 2001; Gunter, Chibnall, Antoniak, McCormick, & Black, 2012), female justice-involved persons in this study exhibited greater psychiatric symptomology at pre-intervention, yet reported less symptomology at post-intervention relative to male justice-involved persons. Similarly, male and female participants both displayed significant improvement in their confidence in refraining from substance use in critical situations; however, this benefit was particularly pronounced for females. Such findings speak to an interaction between participant sex and time point on psychiatric symptomology, such that female justice-involved persons appear to benefit more from CLCO programming in treating these components of mental illness. Beyond symptomology, female justice-involved persons were further found to be more satisfied with the psychological services received, and perceived greater congruence between personal and therapist goals for treatment and a stronger alliance with their therapist overall relative to their male counterparts. Although it is promising that CLCO appears to produce small to moderate effects among males and generally moderate to large effects among females, these sex differences further highlight the need to continue to examine differential mechanisms of change in female offender treatment relative to male treatments.
CLCO and Criminal Thinking
Unfortunately, similar to the findings of Morgan, Kroner, Mills, Bauer, and Serna (2014), this study again failed to demonstrate that CLCO reduces participants’ proactive criminal thinking. It appears that this aspect of criminogenic risk is more resistant to treatment change than reactive criminal thinking, and further work will need to be done to produce reductions in this specific criminogenic cognition. On the other hand, results from this study indicated a large treatment reduction (d = .84) in reactive criminal thinking relative to a moderate treatment effect in the prior Morgan et al. (2014) study (E.S. = .59), such that program revisions designed to enhance treatment effects in this domain (i.e., increase the Problematic Thoughts module length from 10 to 13 sessions, and move the module to earlier in the therapeutic sequence to allow for enhanced exposure to cognitive restructuring) may have partially proved beneficial.
Implications
The results of this study provide a number of practical implications. First and foremost, the results show that comprehensive interventions designed to holistically treat the co-occurring issues of mental illness and criminalness can reduce psychiatric symptoms while simultaneously enhancing general mental health functioning and also reducing important indicators of criminal justice risk. Specially, CLCO is a promising program for reducing aspects of psychiatric symptomatology, enhancing mental illness recovery, and reducing important aspects of criminogenic risk. These findings are relevant to all aspects of CJ-PMI including pretrial (e.g., specialty courts) and corrections (i.e., jail and prison) including community corrections (i.e., probation and parole). In other words, recommendations to integrate mental health and criminal rehabilitation (Draine & Solomon, 1992; Hodgins, 1995; Morgan et al., 2012; Skeem, Winter, Kennealy, Louden, & Tatar, 2014) to achieve improved mental health and criminal justice outcomes are not only reasonable and efficacious, but feasible and realistic.
Limitations
This study is not without limitations. First, there is no control group to compare the treatment gains against, so it is possible that the obtained improvements could have been achieved simply by placement in the residential facility or with any appropriate intervention. Second, because outcome data were limited to self-reported symptoms and functioning, it was not possible to evaluate community outcomes. Future research needs to account for these limitations. Specifically, several studies have now demonstrated the promising effects of CLCO (Morgan et al., 2014; Van Horn et al., 2018) such that future studies should include a randomized controlled clinical trial design to include behavioral (observational) data and community outcomes to more thoroughly evaluate the effectiveness of CLCO. Future studies should also explore differential treatment outcomes based on pretreatment symptom and criminal risk severity in an effort to identify how higher risk males and females respond to CLCO (Reviewer 1, personal communication via submitted review; May 18, 2019).
Conclusion
Despite these limitations, this study presents important findings for the treatment of CJ-PMI. These results replicate findings from Morgan et al. (2014) suggesting that CLCO is an effective intervention for reducing mental health symptomatology and important aspects of criminal risk for CJ-PMI. Although this study lacked a control group, the magnitude of results suggests that the obtained treatment change are not likely to reflect the passage of time, regression toward the mean, changes in therapeutic milieu, or other dynamic covariates that contribute to treatment success. Importantly, the findings from this study demonstrate that CLCO is equally, if not slightly more, promising for female participants as it is for males. These findings suggest that despite gender differences in pathways to crime, it is possible that mechanisms of change may not be different for justice-involved males and females with mental illness. This is an important area to be specifically investigated further. Nevertheless, based on these and previous findings, CLCO remains a promising intervention for reducing psychopathology and criminal risk for CJ-PMI.
Footnotes
Authors’ Note:
We thank the Harris County Community Supervision and Corrections Department for their support in this project. The views and opinions expressed herein represent those of the authors and do not necessarily reflect the views and opinions of Texas Tech University, The Harris Center, or the Harris County Community Supervision and Corrections Department.
