Psychometric Examination of Treatment Change Among Mentally Disordered Offenders: A Risk

Abstract

The present study examined the association of psychiatric symptomatology, criminal attitudes, and treatment changes within these domains to violent and general recidivism in a sample of 614 mentally disordered offenders. Significant pre–post changes were found on multiple measures of criminal attitudes, symptomatology, and readiness for change. Antisocial Intentions and Attitudes Toward Associates (from the Measure of Criminal Attitudes and Associates [MCAA]) predicted general recidivism and covaried with the Big Four criminogenic need domains on the Level of Service Inventory–Ontario Revision; none of the remaining psychometric measures significantly predicted violent or general recidivism. Although pre–post changes were seldom linked to changes in recidivism, positive changes in Antisocial Intentions (MCAA) significantly predicted reductions in general recidivism via Cox regression survival analysis, controlling for baseline risk and pretreatment attitudes score. Risk and need implications of psychometric assessments of treatment change in mentally disordered offender populations are discussed.

Keywords

psychometric mentally disordered offender risk need treatment change recidivism

It is widely accepted that mental illness is highly prevalent and significantly overrepresented in the criminal justice system. Prevalence estimates vary considerably, however, depending on myriad factors such as sample composition, type of mental disorder, and differences in assessment and diagnostic methodology. Several studies have reported an approximate 4% prevalence rate of psychosis among male offenders (Beaudette, Power, & Stewart, 2015; Fazel & Danesh, 2002; Fazel & Seewald, 2012), which increases to about 15% when the definition of serious mental illness incorporates major mood and bipolar disorders (Beaudette et al., 2015; Steadman, Osher, Robbins, Case, & Samuels, 2009). There is also a high level of psychiatric comorbidity among male offenders, particularly with substance misuse and personality disorders (Diamond, Wang, Holzer, Thomas, & des Anges Cruser, 2001; Fazel & Seewald, 2012). Such rates have undoubtedly contributed to a growing interest in establishing best practices in rehabilitation with mentally disordered offenders (MDOs).

Interventions with MDOs have traditionally been guided by the clinical or psychopathological model of criminal behavior (see Bonta, Law, & Hanson, 1998). This perspective underscores the importance of untreated serious mental illness, particularly schizophrenia and other psychotic disorders, as a salient risk factor for violence (Douglas, Guy, & Hart, 2009; Fazel, Gulati, Linsell, Geddes, & Grann, 2009; Hodgins, 2008). Interventions guided by the clinical model incorporate a number of techniques that are ultimately directed at improving mental health functioning as the primary means with which to reduce the likelihood of recidivism (Skeem, Manchak, & Peterson, 2011).

Despite the widespread implementation of the clinical model with MDOs (Morgan et al., 2012; Skeem et al., 2011), some have questioned the utility of this approach for recidivism reduction. Although serious mental illness, namely schizophrenia and other psychotic disorders, are associated with recidivism in the general population (Brennan, Mednick, & Hodgins, 2000; Douglas et al., 2009; Hodgins, 2008), the effect typically fails to generalize to offender samples. Indeed, the vast majority of studies with offender samples have shown psychiatric diagnoses and related clinical constructs, such as previous psychiatric hospitalizations, to be weakly (and in some cases inversely) related to recidivism (Kingston et al., 2016; Kingston, Olver, Harris, Wong, & Bradford, 2015; Rezansoff, Moniruzzaman, Gress, & Somers, 2013; Skeem, Winter, Kennealy, Louden, & Tatar, 2014).

Andrews and Bonta (1994, 2010) presented the general personality and cognitive social learning (GPCSL) model of criminal behavior, which underscored eight robust predictors of criminal behavior that were found to reside within the individual or their immediate social learning environment. The GPCSL model is differentiated from the clinical model due to the type of variables identified as salient predictors (Bonta, Blais, & Wilson, 2013). The specific predictors identified in the GPCSL model include criminal history, procriminal companions, procriminal attitudes, antisocial personality pattern, education/employment, family/marital, substance abuse, and leisure/recreation. The variables identified in the clinical model, such as psychiatric diagnoses, are considered minor or negligible risk factors.

In their most recent meta-analysis with MDOs, Bonta et al. (2013) compared the predictive utility of the central eight risk/need factors of the GPCSL model with variables subsumed within the clinical model, such as psychosis, mood disorder, and past psychiatric treatment. Results showed that the central eight risk/need factors were better predictors of general and violent recidivism when compared with clinical variables. A notable exception was that a diagnosis of personality disorder, namely, antisocial personality disorder, was a reliable predictor of recidivism, although the authors correctly noted that this variable is consistent with a central theme of the GPCSL model. However, this underscores an important point that there is considerable overlap between certain mental health and GPCSL constructs, such as substance misuse and various antisocial traits.

Based on the research noted earlier, some researchers and policy makers have suggested embedding principles of effective correctional intervention into treatment protocols used with MDOs. The risk–need responsivity (RNR) model (Andrews & Bonta, 1994, 2010) is the predominant offender rehabilitation model, and posits that treatment is most effective when it is matched to the risk posed by the offender (risk principle); when it specifically targets criminogenic needs, that is, the seven dynamic factors of the GPCSL (need principle); and when it is based on empirically supported modalities, but maintains sufficient flexibility so that it is applicable to diverse learning styles and abilities in offending populations (responsivity principle).

There has been substantial support for the RNR model in non-MDOs, such that programs adhering to these three principles are more effective than programs that do not (see Andrews & Bonta, 2010, for a review). The RNR approach has also demonstrated some support with MDOs (Skeem et al., 2014; also see Skeem, Steadman, & Manchak, 2015, for a review). Although studies have repeatedly shown that evidence-based mental health treatment has little effect on criminal recidivism, targeting criminogenic needs, such as antisocial attitudes, demonstrates the strongest effect in reducing recidivism (Morgan et al., 2012; Skeem et al., 2011). In a randomized controlled trial, Cullen et al. (2012a) found that MDOs assigned to a cognitive skills program targeting thinking styles and criminal attitudes showed significant improvements in social problem solving relative to the treatment-as-usual comparison group and that such changes were maintained up to 12 months, postrelease. In a follow-up study, Cullen et al. (2012b) reported that the treatment group engaged in fewer incidents of verbal aggression and leave violations during treatment compared with the treatment-as-usual comparison group. Despite the accumulation of evidence supporting the RNR approach in both MDOs and non-MDOs, caution is warranted in fully adapting this model with MDOs, given the limited evidence base with this group (Skeem et al., 2015).

Although considerable progress has been made in identifying criminogenic needs in MDOs that inform risk assessment, intervention, and management, the risk-change literature is far less developed. As noted earlier, the need principle of effective correctional intervention posits that intervention should target causal risk factors that when changed are associated with reductions in recidivism.

Relatively few investigations have examined the association between within-treatment change and recidivism (e.g., Beggs & Grace, 2011; Olver, Stockdale, & Wormith, 2014; Olver, Kingston, Nicholaichuk, & Wong, 2014). Ashford, Wong, and Sternbach (2008) examined MDOs diagnosed with psychotic or bipolar disorders who participated in a cognitive skills program targeting antisocial attitudes. Results showed that reductions in the Identification With Criminal Others subscale of the Criminal Sentiments Scale–Modified (CSS-M) were associated with fewer arrests and technical violations over a 12-month follow-up period. More recently, Kroner and Yessine (2013) examined the effects of a community-based, cognitive-behavioral program targeting antisocial attitudes in a large sample (n = 662) of offenders on parole. Among the various scales measuring attitude change, only the Attitudes Toward Associates subscale of the Measure of Criminal Attitudes and Associates (MCAA) scale was associated with reductions in recidivism.

The present study sought to build on the extant correctional treatment change literature, through examining the extent to which positive pre–posttreatment changes in mental health needs consistent with the clinical model and criminal attitudes as outlined in the GPCSL model are linked to recidivism among MDOs. We hypothesized the following:

Hypothesis 1: MDOs participating in a comprehensive treatment program will evidence positive change in both mental health functioning and antisocial attitudes but only positive change in the latter targets will be associated with subsequent reductions in recidivism.

Method

Participants

Participants were 511 seriously mentally ill adult male offenders, drawn from an original pool of 614 men, with psychometric treatment change information on one or more of the measures. There were no differences between the larger group and the selected sample of 511 participants with regard to risk/need score, length of stay, or recidivism rates. Participants were admitted to the St. Lawrence Valley Correctional and Treatment Centre, an interministerial treatment facility for MDOs serving provincial (i.e., less than 2 years) jail sentences. Offenders are admitted based on the presence of a serious mental illness. In this context, the definition of serious mental illness is consistent with the Ontario Ministry of Health and Long-Term Care, that is, a major mood, psychotic, anxiety, and/or trauma-related disorder with an associated disability involving safety, instrumental living skills, and/or social functioning. Our definition of serious mental illness does not include intellectual disability, personality disorders, or substance use disorders without associated comorbidity with the aforementioned mental disorders. Full ethical approval was obtained from the St. Lawrence Valley Correctional and Treatment Centre.

Offenders served an aggregate average sentence of 1.39 years (SD = 1.04 years) and had an average length of stay of 5.70 months (SD = 2.78 months). Participants had an average age of 34.52 years (SD = 11.65 years) at admission. Most of the sample was single (n = 225, 44%), followed by those who had a previous relationship (n = 86, 16.8%), currently married or equivalent (n = 48, 9.4%), and a small portion widowed (n = 4, 0.8%), and the remainder of the sample unknown (n = 148, 28.9%). Most of the sample was White (n = 392, 76.7%), followed by individuals who were Black (n = 43, 8.4%), Indigenous Canadian (n = 40, 7.8%), with the remainder being of Other (n = 22, 4.3%) or unknown (n = 8, 1.6%) racial ethnic descent.

Treatment Program

Treatment provided at the St. Lawrence Valley Correctional and Treatment Centre targets both mental health functioning and criminogenic needs, as outlined in the GPCSL model. Psychopharmacology is the primary mental health intervention. In addition, individuals participate in a number of predominantly psychoeducational groups related to the understanding of mental health and the importance of adherence to medication. Mental health interventions also provide specific coping skills, such as relaxation, to better manage symptoms.

Programs targeting criminal behavior operate according to the RNR perspective (Andrews & Bonta, 2010). In other words, high-risk offenders receive more intensive programming than lower risk individuals, specific programs are selected for an individual based on presenting need (e.g., prior offense history), and emphasis is placed on cognitive-behavioral/social learning interventions that are tailored to the specific needs of the individual. Considerable emphasis is placed both in the program manuals and in team discussions, on core correctional practices (see Dowden & Andrews, 2004), and on related treatment processes (e.g., regular homework assignments) that have shown to be instrumental in facilitating change (Morgan et al., 2012).

All programs at the St. Lawrence Valley Correctional and Treatment Centre are manualized and highly structured. Specifically, groups target a variety of needs and presenting issues, such as substance abuse (Velasquez, Crouch, Stephens, & DiClemente, 2015), intimate partner violence (Wexler, 2013), anger management (Winogron, Van Dieten, & Gauzas, 1997), sexual aggression (Marshall, Marshall, Serran, & O’Brien, 2011), and cognitive skills (Ross, Hilborn, & Liddle, 2007). These groups were selected based on several factors, such as demonstrated empirical support and adherence to best practices in offender rehabilitation. Another important factor in selecting these programs was the length of the intervention. Given the relatively short length of stay (5.7 months) at our institution, we were restricted to selecting programs that were approximately 3 to 4 months in duration.

Measures

Marlowe–Crowne Social Desirability Scale (MCSDS)

The MCSDS (Crowne & Marlowe, 1960) is a 33-item self-report measure used to identify social desirability in responding with forced-choice, true–false questions. These questions regard common behaviors and situations under two factors: attribution and denial. In a sample of adult sexual offenders, Tatman, Swogger, Love, and Cook (2009) confirmed the two-factor structure proposed by the developers and also reported the scale to have good internal consistency estimates for the total scale (r = .85) as well as the attribution (r = .76) and denial (r = .78) factors, test–retest reliability over a 3-week time period (r = .89), and good convergent and discriminant validity with the Minnesota Multiphasic Personality Inventory validity scales. It has limited use in MDO samples (McGilloway & Donnelly, 2004).

Level of Service Inventory–Ontario Revision (LSI-OR)

The LSI-OR (Andrews, Bonta, & Wormith, 1995) is an actuarial risk–need assessment tool that was designed to appraise recidivism risk, identify criminogenic needs, and inform recommendations for treatment and case management. The tool is part of a broader family of tools that fall under the rubric of the Level of Service (LS) scales, with a large number of variations in existence. Wormith (2011) noted that the LS scales are the most widely used group of risk assessment tools on the planet. The LSI-OR includes 43 items organized around the “central eight” risk factors identified as the strongest correlates of criminal conduct, and that are embedded within the GPCSL model described earlier.

A number of studies have been conducted demonstrating the LS scales sound psychometric properties, including its reliability and predictive accuracy for a range of outcomes (e.g., Brews, 2009; Rettinger & Andrews, 2010; Wormith, Hogg, & Guzzo, 2012) and with MDOs specifically (Girard & Wormith, 2004; Skeem et al., 2014). In a recent meta-analysis of 128 studies comprising 137,931 offenders, Olver, Stockdale, et al. (2014) reported the LS scales to significantly predict general and violent recidivism (r = .29 and .23, respectively); these effects were upheld across gender and ethnicity. The LSI-OR was completed by trained corrections staff just prior to entry into our facility and, consequently, prior to completion of the self-report attitude scales. In the present sample, the LSI-OR produced an alpha coefficient of .73.

MCAA

The MCAA (Mills, Kroner, & Forth, 2002) is a measure of criminal attitudes and associates, which are both considered primary criminogenic needs and essential predictors of recidivism. The instrument consists of six subscales (Number of Criminal Friends, Criminal Friend Index, Attitudes Toward Violence, Sentiments of Entitlement, Antisocial Intent, and Attitudes Toward Associates) as well as a total score. Research shows that the MCAA is a reliable and valid measure of antisocial attitudes and associates (Bäckström & Björklund, 2008; Mandracchia & Morgan, 2012; Mills, Kroner, & Forth, 2002). Kroner and Yessine (2013) reported that male offenders demonstrated a reduction in MCAA scores following an antisocial attitude intervention program, and that among the various measures, only the MCAA antisocial associates change score predicted recidivism. The MCAA has also shown predictive validity for both general and violent recidivism in adult male offenders (Mills, Kroner, & Hemmati, 2004).

The CSS-M

The CSS-M (Simourd, 1997) is a 41-item self-report tool that measures antisocial attitudes, values, and beliefs that are directly related to criminal behavior (Simourd, 1997). The scale includes both prosocial and antisocial statements, such as “a judge is a good person” and “the law is rotten to the core,” and response options include agree, undecided, and disagree. The CSS-M consists of five subscales: Attitudes Toward the Law, Attitudes Toward the Court, Attitudes Toward Police, Tolerance for Law Violations, and Identification With Criminal Others. Higher scores on each of the subscales indicate greater antisocial attitudes, values, and beliefs. A number of investigations have been conducted on the scales’ reliability and validity (Andrews, Wormith, & Kiessling, 1985; Roy & Wormith, 1985; Simourd, 1997; Vaske, Gehring, & Lovins, 2017). The CSS-M has been used with MDOs (Morgan et al., 2012) and has shown acceptable internal consistency (α = .78-.87) and sensitivity to change in this type of offender (Ashford et al., 2008). The CSS-M has further shown significant predictive validity in a sample of probationers (Wormith & Andrews, 1995) and in a recent meta-analysis of diverse offender samples (Walters, 2016).

The University of Rhode Island Change Assessment (URICA) Scale

The URICA (DiClemente & Hughes, 1990) is a 32-item self-report measure that examines readiness for change and includes four subscales: Precontemplation, Contemplation, Action, and Maintenance. Responses are given on a 5-point Likert-type scale ranging from 1 (strong disagreement) to 5 (strong agreement). The subscales can be combined to yield a second-order continuous readiness to change score that can be used to assess readiness to change prior to commencing treatment. The URICA has been used in a number of studies with offender samples (Lewis, 2004; Polaschek, Anstiss, & Wilson, 2010; Yong, Williams, Provan, Clarke, & Sinclair, 2015) and with MDOs (DiClemente, Nidecker, & Bellack, 2008) showing acceptable levels of internal consistency and sensitivity to change.

Brief Psychiatric Rating Scale (BPRS)

The BPRS (Overall & Gorham, 1962) includes 18 items and is a clinician-rated tool designed to assess change in severity of psychopathology. The measure examines the following concerns: somatic concern, anxiety, emotional withdrawal, conceptual disorganization, guilt feelings, tension, mannerisms and posturing, grandiosity, depressive mood, hostility, suspiciousness, hallucinatory behaviors, motor retardation, uncooperativeness, unusual thought content, blunted affect, excitement, and disorientation. The BPRS has been shown to have adequate subscale reliabilities and appropriate differential validity (Thomas, Donnell, & Young, 2004). Reports of interrater reliability correlation coefficients have ranged from r = .67 to r = .88 (Morlan & Tan, 1998). In a MDO sample, the BPRS was a significant predictor of verbal aggression and physical violence (rs = .58-.61; Gray et al., 2003). The internal consistency for the BPRS in this sample was good at both pretreatment (α = .83) and posttreatment (α = .75).

The Brockville Quality of Life Scale (QoL)

The QoL (Long, 2004) was developed for use in community clinical settings to assess the current level of functioning in psychiatric patients. To our knowledge, this is the first study that has examined this measure in an incarcerated sample of MDOs. The version of the scale we used forms part of an electronic charting program called Decision Base. The QoL is a clinician-administered measure that includes 78 items scored on a 0 to 2 scale to reflect difficulty or symptom severity in a number of problems related to mental health. Such areas include, for example, anxiety, sadness or apathy, hyperactivity or excitement, reality testing, dependency on institutional care, and lack of insight.

Recidivism

Outcome data were retrieved on June 10, 2016, from the Offender Tracking and Information System used by the Ministry of Community Safety and Correctional Services. In this study, we used postrelease recidivism as our primary outcome measure. Recidivism was defined as a return to provincial correctional supervision on a new charge or conviction within 2 years of the completion of a provincial sentence to incarceration. Violent recidivism was defined as any charge or conviction of an offense against a person (e.g., assault, robbery), including sexual offenses. Criminal recidivism included any charge or conviction (i.e., sexual, violent, and general offending). Charges or convictions were coded in a binary manner (1 = recidivate, 0 = did not recidivate) along with charge/conviction date for the first new offense of a given category to perform survival analysis.

Procedure and Data Analytic Plan

Individuals admitted to the St. Lawrence Valley Correctional and Treatment Centre complete a standardized assessment battery that was selected for clinical and research-related purposes. Such measures were selected according to the dual mandate of our institution, that is, the treatment of mental health and identified criminogenic needs. Participants are approached by nursing staff to complete the self-report measures upon entering the institution and they complete the same battery of tests prior to discharge. Upon entering the facility, the clinician-administered measures were completed by the admitting psychiatrist (BPRS) or psychiatric nurse shortly after intake and again prior to discharge. Individuals are not included in these analyses if they refused to have their deidentified data used for research purposes.

The analyses proceeded in several phases. First, pretreatment–posttreatment differences were examined on the psychometric measures via t tests with standardized mean difference (Cohen’s d) computed to provide a measure of effect size. Cohen’s (1992) conventions of small (d = .20), medium (d = .50), and large (d = .80) were employed to interpret the magnitude of the differences. Correlations were also computed between pretreatment, posttreatment, and change scores on each of the measures with the Marlowe–Crowne to further ascertain whether the responses were associated with socially desirable responding.

Second, to examine the risk relevance of the psychometric measures, validity correlations were computed between psychometric test scores with the LSI-OR eight need domains and the total score. As LSI-OR scores were obtained at pretreatment, the pretreatment scores of the psychometric measures were examined. Positive correlations between the psychometric measures and the LSI-OR need domains and total score could indicate shared risk variance. Again, Cohen’s (1992) conventions of small (r = .10), medium (r = .30), and large (r = .50) for interpreting correlation magnitude between two continuous variables were employed.

Third, to directly examine the risk relevance of the psychometric measures, we examined the associations of pre- and posttreatment test scores with binary violent and general recidivism through receiver operating characteristic (ROC) curve analysis. ROC analyses generate an area under the curve (AUC) value ranging from 0 to 1.0, representing the probability that a randomly selected recidivist would have a higher score on a given measure than a randomly selected nonrecidivst. With values of AUC = .50 representing chance level accuracy, values of AUC = .56, AUC = .64, and AUC = .71 correspond to small, medium, and large effect sizes, respectively (Rice & Harris, 2005). Fourth, we examined the risk relevance of treatment change through examining associations of pre–posttreatment change scores with binary recidivism. If the changes were risk relevant, then positive changes should be associated with decreased recidivism. In addition, given that individuals scoring higher on the measures have not only more room for movement but also greater problem areas in a given domain, we also computed residualized change (RC) scores through regressing the change score on the pretreatment score and obtaining the residual, which would represent change not constrained by pretreatment score (Beggs & Grace, 2011). Cohen’s d was again computed to examine the standardized mean difference between recidivists and nonrecidivists on their raw change scores as well as RCs (i.e., after controlling for pretreatment score).

Finally, we conducted Cox regression survival analysis to examine the incremental predictive validity of selected RCs after controlling for baseline risk via the LSI-OR total score. Psychometric change predictors were selected purposively based on whether they demonstrated a minimum small association (d = –.20) to decreased recidivism of one of the binary criteria in the fourth set of analyses previously described. The predictors were entered simultaneously in the regression model to examine their unique contributions in the prediction of violent or general recidivism over time. Such an analysis would provide a comprehensive control for risk through inclusion of the LSI-OR as a covariate in the model and, thus, serve as a rigorous test of the potential risk relevance of the change scores captured by these measures. The use of the RC score would entail control for the pretreatment score on the psychometric predictor. Cox regression generates a hazard ratio (e^β), which represents the proportionate increase in the hazard of a given outcome occurring for every one-point change in the predictor variable. Values of e^β above 1.0 indicate a positive association between the predictor and criterion, whereas values below 1.0 indicate an inverse association.

We extended the Cox regressions by rerunning the analyses with five different measurements of the change predictors: pretreatment scores, posttreatment scores, average pre- and posttreatment scores, lowest score (of pre and post), and highest score (of pre and post). In principle, the strongest evidence for “true” change having occurred would be demonstrated by the most recent (i.e., posttreatment) measurement being most predictive. We computed Bayesian information criterion (BIC) values to evaluate prediction magnitude through the following formula: –2 log likelihood + [k × ln(n)], where k is the number of covariates entered into the model and n is the number of events (i.e., recidivists). The smallest BIC value represents the best prediction of the targeted criterion; ΔBIC units of two or greater represents positive evidence for the superiority of one model over another (Raftery, 1995).

Results

Pretreatment–Posttreatment Comparisons

Table 1 reports the pretreatment–posttreatment comparisons on the psychometric measures and associations with social desirability. Significant pre–post changes were observed across most of the measures, in the general trend of scores improving, with the exception of three out of the four subscales from the MCAA. On the CSS-M, approximately one half a standard deviation of change was observed from pre- to posttreatment on the three subscales and total score, indicating decreased criminal attitudes. On the MCAA, positive changes on the individual subscales tended to be smaller in magnitude. On the URICA precontemplation and action stages, scores increased from pre- to posttreatment, whereas contemplation and maintenance stage scores decreased. More substantive changes approaching or exceeding one standard deviation of change were observed for BPRS and QoL denoting improved symptomatology.

Table 1:

Pretreatment–Posttreatment Comparisons on Psychometric Measures

Measure	n	Pretreatment		Posttreatment		t	d change	Marlowe–Crowne
Measure	n	M	SD	M	SD	t	d change	Pre	Post	Change
CSS-M
Law court police	99	21.00	10.86	16.63	9.36	5.45***	.43	−.48***	−.24*	−.39***
Tolerance for law violations	99	8.36	5.21	5.98	4.35	4.37***	.50	−.40***	−.27*	−.21
Identification with criminal others	99	5.14	2.80	4.03	2.35	5.54***	.43	−.41***	−.39***	−.09
Total	99	34.51	16.76	26.64	13.92	6.69***	.51	−.51***	−.31**	−.37***
MCAA (Part B)
Attitudes toward violence	273	3.54	3.26	3.26	3.37	1.43	.08	−.18**	−.15*	−.02
Antisocial intent	273	4.50	3.28	3.90	3.20	3.45***	.19	−.51***	−.34***	−.21***
Entitlement	272	5.06	2.69	4.77	2.70	1.84	.11	−.06	−.01	−.05
Attitudes toward associates	272	6.36	2.89	6.13	2.74	1.65	.08	−.43***	−.38***	−.08
Total	271	19.56	8.83	18.04	9.31	3.44***	–.18	−.41***	−.28***	−.13*
URICA
Precontemplation	300	1.96	0.73	2.08	0.77	2.66**	.16	.24***	.21***	.01
Contemplation	300	4.29	0.59	4.12	0.69	4.06***	.27	−.26***	−.09	−.12
Action	300	3.96	0.56	4.16	0.67	4.48***	.32	−.01	−.05	−.04
Maintenance	300	3.77	0.61	3.63	0.73	3.17**	.21	−.15*	−.12	−.01
BPRS
Total	341	40.08	11.61	31.37	7.95	15.01***	.88	.02	−.02	.03
QoL
Total	410	56.69	23.60	32.39	16.68	23.10***	1.19	−.17**	−.11	−.12*

Note. CSS-M = Criminal Sentiments Scale–Modified; MCAA = Measure of Criminal Attitudes and Associates; URICA = University of Rhode Island Change Assessment; BPRS = Brief Psychiatric Rating Scale; QoL = quality of life.

p < .05. **p < .01. ***p < .001.

Both criminal attitude measures were significantly correlated with MCSD scores, with the correlations being largest for pretreatment scores, somewhat smaller in magnitude for the posttreatment scores, and only a few of the associations attaining significance with the change scores. Correlations with social desirability for the CSS-M and MCAA were negative in valence, indicating that lower endorsement of criminal attitudes was associated with higher social desirability; the inverse associations with change scores would be interpreted to mean that larger amounts of change were actually associated with decreased social desirability.

Convergent Validity Associations

Table 2 reports convergent validity correlations between the psychometric measures (pretreatment scores) with the eight criminogenic need domains and the LSI-OR total score. The largest volume of significant associations was seen between the MCAA antisocial intent, attitudes toward associates, and total score with seven out of the eight need domains and LSI-OR total score; moderate to large associations were observed with criminal history, education/employment, peers, alcohol and drug problems, antisocial pattern, and the total score. CSS-M had fewer significant associations with the LSI-OR, although significant associations were found for four out of the eight domains that were generally small to moderate in magnitude. QoL scores were significantly associated with all domains of the LSI-OR with associations ranging from small to moderate in magnitude, whereas the BPRS demonstrated generally smaller magnitude correlations with most of the LSI-OR domains, including the total score. Finally, the URICA subscales generally evinced weak and nonsignificant associations with the LSI-OR, although higher scores on three of the four subscales were associated with more serious criminal history, whereas high maintenance stage scores had modest but significant associations with peers, alcohol and drug problems, and the total score.

Table 2:

Correlations Between Psychometric Measures and LSI-OR Criminogenic Need

Measure	Criminal history	Education/employment	Family/marital	Peers	Alcohol/drug	Leisure/recreation	Attitudes	Antisocial pattern	Total
CSS-M
Law court police	.03	.33***	.00	.22*	.20*	.10	−.06	.21*	.28**
Tolerance for law violations	.06	.23*	−.03	.25*	.16	.08	−.06	.13	.22*
Identification with criminal others	.16	.10	.14	.30*	.21*	.08	−.10	.19	.26**
Total	.06	.30**	.01	.27*	.21*	.10	−.07	.21*	.29**
MCAA (Part B)
Attitudes toward violence	.14*	.09	.06	.14*	.09	.08	.08	.12*	.16**
Antisocial intent	.36***	.29***	.14*	.41***	.30***	.12	.14*	.32***	.45***
Entitlement	.09	−.03	.04	.12	.05	.10	.01	.11	.08
Attitudes toward associates	.42***	.27***	.16*	.41***	.30***	.08	.13*	.30***	.45***
Total	.35***	.22***	.13*	.37***	.26***	.13*	.12*	.29***	.39***
URICA
Precontemplation	−.08	−.03	.00	−.11	−.09	.05	.08	−.04	−.07
Contemplation	.12*	.04	.03	.08	.09	−.06	−.10	.06	.08
Action	.13*	−.02	.01	.09	.10	−.06	−.05	.09	.08
Maintenance	.15*	.02	.02	.14*	.13*	.04	−.08	.09	.12*
BPRS
Total	.12*	.11	.16**	.12*	.08	.17**	.19***	.16**	.20***
QoL
Total	.35***	.22***	.15**	.27***	.26***	.13*	.29***	.30***	.40***

Note. LSI-OR = Level of Service Inventory–Ontario Revision; CSS-M = Criminal Sentiments Scale–Modified; MCAA = Measure of Criminal Attitudes and Associates; URICA = University of Rhode Island Change Assessment; BPRS = Brief Psychiatric Rating Scale; QoL = quality of life.

p < .05. **p < .01. ***p < .001.

Predictive Accuracy

The sample was followed up for 1.79 years (SD = 0.83 years) postrelease, during which 15.5% of the men (n = 79) were convicted for a new violent offense and 46% (n = 235) were convicted for any new offense. The results of univariate predictive accuracy analyses of the psychometric pre- and posttreatment scores for violent and general recidivism are presented in Table 3. Significant prediction of either outcome would demonstrate the risk relevance of the psychometric measures. In addition, higher prediction by posttreatment scores, which may reflect changed risk and are in closer proximity to the outcome, may also indicate the dynamism of the measures. Very few of the measures significantly predicted violent recidivism, with the exception of QoL posttreatment total scores; the Tolerance for Law Violations (TLV) subscale (CSS-M) demonstrated AUC magnitudes that were reasonable in magnitude, but did not attain significance, likely owing to limited power from the sample size. Several of the MCAA scales, including the total score, however, significantly predicted general recidivism as did QoL pretreatment and posttreatment scores. High posttreatment scores on precontemplation predicted decreased general recidivism, whereas high scores on contemplation predicted the same outcome. As in previous analyses, CSS-M and BPRS scores did not significantly predict general recidivism.

Table 3:

Predictive Accuracy of Pretreatment–Posttreatment Psychometric Measures for Violent and General Recidivism

Measure	n	Violent				General
		Pre		Post		Pre		Post
		AUC	95% CI	AUC	95% CI	AUC	95% CI	AUC	95% CI
CSS-M
Law court police	99	.58	[.43, .73]	.56	[.42, .70]	.51	[.40, .63	.55	[.43, .66]
Tolerance for law violations	99	.61	[.47, .76]	.63	[.49, .77]	.54	[.43, .66	.54	[.43, .66]
Identification with criminal others	99	.59	[.45, .73]	.59	[.44, .75]	.56	[.45, .68	.58	[.47, .70]
Total	99	.60	[.45, .75]	.59	[.45, .73]	.53	[.41, .64	.57	[.45, .68]
MCAA (Part B)
Attitudes toward violence	273	.56	[.47, .66]	.49	[.40, .58]	.57*	[.50, .64]	.56	[.49, .63]
Antisocial intent	273	.49	[.40, .58]	.55	[.47, .64]	.56	[.49, .63]	.62***	[.55, .68]
Entitlement	272	.48	[.39, .57]	.51	[.42, .60]	.51	[.44, .58]	.52	[.45, .59]
Attitudes toward associates	272	.54	[.46, .63]	.58	[.50, .67]	.61**	[.54, .67]	.62***	[.55, .68]
Total	271	.53	[.45, .62]	.55	[.46, .63]	.58*	[.51, .65]	.60**	[.53, .67]
URICA
Precontemplation	310	.57	[.49, .65]	.46	[.38, .54]	.45	[.39, .52]	.42*	[.35, .48]
Contemplation	310	.48	[.40, .57]	.50	[.42, .58]	.55	[.48, .61]	.58*	[.51, .64]
Action	310	.49	[.41, .57]	.47	[.39, .55]	.54	[.48, .61]	.53	[.46, .59]
Maintenance	310	.49	[.41, .58]	.54	[.46, .61]	.53	[.47, .60]	.55	[.49, .62]
BPRS
Total	341	.50	[.41, .59]	.50	[.41, .60]	.47	[.41, .53]	.50	[.43, .56]
QoL
Total	410	.55	[.47, .63]	.61*	[.53, .69]	.56*	[.50, .62]	.58**	[.52, .63]

Note. AUC = area under the curve; CI = confidence interval; CSS-M = Criminal Sentiments Scale–Modified; MCAA = Measure of Criminal Attitudes and Associates; URICA = University of Rhode Island Change Assessment; BPRS = Brief Psychiatric Rating Scale; QoL = quality of life.

p < .05. **p < .01. ***p < .001.

Psychometric Treatment Change and Recidivism

The extent to which pre–posttreatment change scores on the psychometric measures capture changes in risk, and hence are also risk relevant, was examined through analysis of change scores with recidivism (Table 4). The associations between change scores and recidivism were most consistently in the expected direction (i.e., negative) and of greater magnitude when RC scores were employed. Some raw change scores were actually associated with increased violent recidivism contrary to expectations. Although none of the change measures significantly predicted decreased violent recidivism, some small range associations (d = –.20 or higher) were observed for the RCs for certain CSS-M and MCAA subscales, as well as the QoL total change score. Similarly, none of the raw change scores significantly predicted general recidivism, although after controlling for pretreatment score, changes on MCAA subscales—Antisocial Intent and Attitudes Toward Associates (d < .10)—were associated with decreased general recidivism, supporting the risk relevance of these criminal attitude measures. Paradoxically, positive changes on attitudes toward violence (MCAA) were associated with increased violent recidivism.

Table 4:

Changes on Psychometric Measures and Relationship (Cohen’s d) to Violent and General Recidivism

Change measure	n	Violent		General
Change measure	n	d	d RC	d	d RC
CSS-M
Law court police	99	.24	.07	−.09	−.15
Tolerance for law violations	99	.10	−.20	.00	−.11
Identification with criminal others	99	.11	−.16	−.05	−.22
Total	99	.22	−.02	−.07	−.17
MCAA (Part B)
Attitudes toward violence	273	.35*	.28*	.08	−.02
Antisocial intent	273	−.17	−.18	−.18	−.31*
Entitlement	272	−.10	−.09	−.06	−.08
Attitudes toward associates	272	−.14	−.26	−.04	−.23†
Total	271	−.01	−.05	−.06	−.17
URICA
Precontemplation	310	.29^†	.22	.06	.16
Contemplation	310	−.08	−.07	−.04	−.14
Action	310	−.03	−.03	−.08	.00
Maintenance	310	.16	.18	.04	.10
BPRS
Total	341	−.04	.02	.01	.07
QoL
Total	410	−.08	−.23	.02	−.14

Note. d RC = d value of differences between recidivists and nonrecidivists in RC score produced through controlling for pretreatment measure. Negative d values indicate that positive treatment changes are associated with decreased recidivism, whereas positive d values indicate that positive treatment changes are associated with increased recidivism. RC = residual change; CSS-M = Criminal Sentiments Scale–Modified; MCAA = Measure of Criminal Attitudes and Associates; URICA = University of Rhode Island Change Assessment; BPRS = Brief Psychiatric Rating Scale; QoL = quality of life.

†

p < .10. *p < .05.

To provide comprehensive controls for baseline risk, the incremental predictive validity of selected change predictors (i.e., demonstrating at least d = –.20 with one of the recidivism outcomes, that is, the threshold for a small effect) for violent and general recidivism was examined after controlling for LSI-OR total score through Cox regression survival analysis. As the association for MCAA attitudes toward violence exceeded this threshold, but was in the opposite direction anticipated, we also included change scores from this measure in analyses. The RC score was used as a means of controlling for pretreatment score. As seen in Table 5, LSI-OR total scores significantly uniquely predicted violent and general recidivism across all regression models; in most instances, a one-point increase in LSI-OR score was associated with a 7% increase in the hazard of recidivism. Only changes in antisocial intent from the MCAA were significantly associated with decreases in general recidivism after controlling for LSI-OR score; in this instance, each one-point increase in change was associated with an 8% decrease in the hazard of general recidivism after controlling for baseline risk. None of the other change measures was incrementally predictive of decreased recidivism after controlling for LSI-OR score.

Table 5:

Cox Regression Survival Analysis: Relationship of Standardized RC Scores on Selected Measures to Recidivism Controlling for the LSI-OR

Cox regression model (1-8)		B	SE	Wald	p	e ^β	LL	UL
Violent recidivism
1	LSI-OR	.18	.05	15.17	<.001	1.20	1.09	1.31
	Tolerance toward law violations RC	−.08	.07	1.35	.245	0.92	0.80	1.06
2	LSI-OR	.06	.02	8.52	.004	1.07	1.02	1.11
	MCAA attitudes toward violence RC	.10	.06	2.62	.106	1.10	0.98	1.23
3	LSI-OR	.06	.02	6.69	.010	1.06	1.01	1.10
	MCAA attitudes toward associates RC	−.08	.07	1.27	.260	0.92	0.80	1.06
4	LSI-OR	.07	.02	11.63	.001	1.07	1.03	1.11
	URICA precontemplation RC	.30	.22	1.92	.166	1.35	0.88	2.06
5	LSI-OR	.06	.02	9.71	.002	1.06	1.02	1.09
	Quality of life RC	−.01	.01	1.02	.313	0.99	0.98	1.01
General recidivism
6	LSI-OR	.06	.03	6.74	.009	1.07	1.02	1.12
	Identification with criminal others RC	−.05	.07	0.48	.489	0.96	0.84	1.09
7	LSI-OR	.06	.01	22.72	<.001	1.07	1.04	1.10
	MCAA antisocial intent RC	−.09	.04	5.11	.024	0.92	0.85	0.99
8	LSI-OR	.07	.01	23.95	<.001	1.07	1.04	1.10
	MCAA attitudes toward associates RC	−.03	.04	0.33	.563	0.98	0.90	1.06

Note. RC = residual change; LSI-OR = Level of Service Inventory–Ontario Revision; LL = lower limit; UL = upper limit; MCAA = Measure of Criminal Attitudes and Associates.

In the final set of analyses, we computed BIC values to examine five possible measurements of the predictor variables with change information included in the Cox regression survival analyses: pretreatment, posttreatment, average of pre and post, lowest score, and highest score. As seen in Table 6, the most recent assessment (i.e., posttreatment) was most frequently the strongest predictor of general recidivism and, hence, the reference model, as denoted by the lowest BIC value. With respect to MCAA antisocial intent, the posttreatment model was superior to the pretreatment, lowest score, and highest score models. For MCAA attitudes toward associates, however, the average score and lowest score models were each superior to the pretreatment, posttreatment, and highest score models. There was no evidence for the superiority of Identification with Criminal Others models in the prediction of general recidivism, or for any of the TLV and URICA (precontemplation) models in the prediction of violent recidivism. Finally, with respect to the prediction of violent recidivism, the MCAA attitudes toward violence pretreatment model was superior to the posttreatment and lowest score models, whereas for MCAA attitudes toward associates, the posttreatment, average, and lowest score models were superior to the pretreatment and highest score models.

Table 6:

BIC Values of Selected Change Predictors for Violent and General Recidivism

Model predictor	BIC value
Model predictor	Pre	Post	Average	Lowest	Highest
Violent recidivism
Tolerance toward law violations	144.97	144.26	143.91	144.04	145.03
MCAA attitudes toward violence	466.68^a,b	469.08	468.64	469.17	467.74
MCAA attitudes toward associates	465.11	462.69^c,d	463.07^c,d	461.84^c,d	465.18
URICA precontemplation	577.28	577.23	577.95	577.72	577.75
Quality of life	689.64	688.62	688.17^b	690.35	687.51^b,c
General recidivism
Identification with criminal others	453.46	452.17	452.55	452.22	453.15
MCAA antisocial intent	1,224.05	1,216.88^b,c,d	1,218.58^c	1,219.31^c	1,220.43^c
MCAA attitudes toward associates	1,216.73	1,217.10	1,214.41^a,c,d	1,214.36^a,c,d	1,217.22

Note. BIC = Bayesian information criterion; MCAA = Measure of Criminal Attitudes and Associates; URICA = University of Rhode Island Change Assessment.

Superior to the posttreatment model. ^bSuperior to the lowest score model. ^cSuperior to the pretreatment model. ^dSuperior to the highest score model. ^eSuperior to the average model.

Discussion

In the present study, we examined the extent to which MDOs showed change with regard to mental health variables as well as antisocial attitudes underscored by the GPCSL. In addition, we examined the risk relevancy of these constructs and the extent to which observed change was associated with subsequent reductions in violent and general recidivism.

Significant pre–post changes were observed across the bulk of the psychometric measures, corresponding to important treatment foci: There were measureable improvements in criminal attitude endorsement, decreased global psychiatric symptomatology, and improved self-reported motivation to change on certain indices of the URICA. Associations with a measure of social desirability indicated that endorsement of fewer problem areas in these domains was associated with greater impression management; however, this was most pronounced on pretreatment measures, and the linkages between social desirability and change were such that lower social desirability scores at pretreatment were associated with greater pre- to posttreatment change across several of the measures (or there was no association at all). The pattern of findings would indicate that social desirability ultimately had little bearing on self-reported changes in the treatment domains assessed.

The extent to which the domains are risk relevant, however, is another matter. Convergent validity correlations with the LSI-OR risk and need domains, including the total score, demonstrated that measures of criminal attitudes had the strongest convergence and, by implication, common risk relevance. Unexpectedly, this was observed with somewhat smaller, but nevertheless significant, associations between the psychiatric symptomatology measures with most of the LSI-OR domains. Smaller magnitudes of associations would suggest, perhaps, less shared risk variance and hence, risk relevance, compared with domains intentionally assessing criminogenic domains such as criminal attitudes.

These results generally bore out when it came to the prediction of the two recidivism outcomes, general and violent recidivism. Owing to shrinking cell sizes for some of the measures (e.g., CSS-M) coupled with a low base rate of violence, associations with this outcome seldom attained significance despite some reasonable AUC magnitudes (e.g., .60s) that would conventionally attain significance. Pre- and posttreatment measures, however, more frequently predicted general recidivism, particularly criminal attitude domains such as antisocial intentions and antisocial associates, consistent with the extant literature of the central eight (Olver, Stockdale, et al., 2014). Interestingly, more severe psychiatric symptoms (QoL) predicted recidivism; however, the QoL includes specific items that also tap into central eight risk factors, such as alcohol abuse and physical violence, which may explain a portion of the risk relevance of this measure. Moreover, high contemplation stage of change scores of the URICA at posttreatment also predicted recidivism, although paradoxically, low precontemplation scores predicted increased recidivism as well. It is possible that individuals endorsing low precontemplation reported higher scores on the other stages, and that high scores on contemplation, indicating general self-awareness of problem areas but lack of progress in skills usage to manage the problem area, understandably reflected decreased progress and increased risk posttreatment.

Evaluating the Risk Relevance of Within Treatment Change

The results of psychometric change analyses demonstrated that for measures that had risk relevance, that is, converged with risk–need domains and predicted recidivism, changes on those measures assessed at two time points were also more likely to be associated with recidivism. Positive changes on the criminal attitude domains, antisocial intent, and to a lesser degree, attitudes toward associates (both from the MCAA), were associated with decreased general recidivism; the former significantly so, even after controlling for baseline risk (LSI-OR total score), suggesting the association is unlikely to be spurious. The results are consistent with Simourd, Olver, and Brandenburg (2016), who found changes on the CSS-M following a prison-based criminal attitudes program were associated with decreased recidivism postrelease. Similarly, Olver, Kingston, et al. (2014) found that changes on a psychometric measure of aggression following sex offender treatment were linked to reductions in multiple recidivism outcomes; however, changes on all other psychometric measures of a miscellany of psychological constructs (e.g., loneliness, acceptance of responsibility, intimacy, empathy, and sex offender beliefs) were not.

Contrary to our expectations, positive changes on the MCAA’s attitude toward violence were actually associated with increased violent recidivism, even after controlling for pretreatment score (although the magnitude of the association was attenuated to some degree). Our best possible explanation is that men at the highest risk of violence and who had the most changes to make in terms of attitudes supportive of violence may be driving this unexpected association. That the association was reduced to nonsignificance after controlling for LSI-OR score in Cox regression survival analyses is consistent with this explanation. Alternatively, it is possible that it may be a spurious association given that attitudes toward violence was one of few domains without significant pre–post changes; it demonstrated weak convergence with the LSI-OR; and in univariate prediction analyses, it tended to be a weak and inconsistent predictor of outcome.

Limitations, Conclusions, and Future Directions

Several potential limitations of the present study need to be considered. This sample consisted of provincial offenders and, by definition, received a shorter sentence length (i.e., 2 years less a day) than federal offenders. As such, our results may not be generalizable to offenders who receive longer prison sentences. In addition, the follow-up time was relatively short with an average of 1.79 years postrelease. This is offset by the fact that the base rate of violent (15.5%) and criminal (46%) recidivism was relatively high to yield sufficient power for our analyses. It is important to note that recidivism outcomes were restricted to within the province of Ontario. Consequently, if an offender was released and moved to another province and reoffended, then this outcome would be lost. In addition, it is also possible that some participants may have been arrested without receiving a new criminal charge (i.e., technical violation) and, as such, would not be captured using our method of identifying recidivism.

It is important to note that our assessment battery was limited, in that we focused on specific psychiatric symptomatology and antisocial attitudes. There are a number of other relevant variables, such as additional criminogenic needs, and treatment-relevant processes (e.g., engagement in treatment) that likely have relevance and would be important for researchers to consider for future investigations. The lack of interrater reliability, which was another limitation with our assessment protocol, may have influenced the predictive accuracy of the relevant variables. Despite this, we feel this methodology is ecologically valid with the results having applicability in routine clinical practice. Finally, although we suggest that the observed change is likely a reflection of the intervention, it is impossible to say without an adequate control group. Although this does not affect the central hypotheses regarding the relationships between change and recidivism, it does affect identifying the mechanism of change.

Our findings lend further support to the utility of the GPCSL model, in general, and RNR principles, in particular, as important elements of a comprehensive treatment program for MDOs. It should be noted that the RNR model incorporates serious mental illness as a responsivity factor and, therefore, encourages including this treatment target in programming, although it would be weighted less heavily than general criminogenic factors. As an example, some individuals with active symptoms of mental illness may find it difficult to engage in treatment or attend to the treatment content. Therefore, targeting such symptoms may be an important first step in the treatment process to promote therapeutic engagement and gain.

There are additional reasons to incorporate mental health variables in treatment protocols used with MDOs. Of course, there is the obvious fact that improving mental health symptoms affects positively on mental health outcomes and overall quality of life (Morgan et al., 2012). In terms of criminal outcomes, however, studies have shown that mental illness can be a relevant predictor of recidivism, at least for a small subset of offenders (Kingston et al., 2016; Peterson, Skeem, Kennealy, Bray, & Zvonkovic, 2014) and, as noted by Peterson et al., the degree to which mental illness directly precedes criminal activity can vary within offenders, across time. A complete policy shift away from mental health service would fail to address mental illness as an identified criminogenic need in this subset of offenders; given that we are not able to reliably identify symptom-based offenses, it is prudent to continue to provide mental health services.

Finally, addressing mental health symptoms may help to promote improvement in an individual’s identified criminogenic needs. For example, managing mental health symptoms may allow one to make more adequate use of leisure time, resist urges to turn to substance use to manage symptoms, and to obtain employment, all of which are established criminogenic needs and have been shown to reduce the likelihood of recidivism. Future studies should examine how the inclusion of mental health treatment targets acts synergistically with criminogenic needs in reducing criminal recidivism among MDOs.

Footnotes

Acknowledgements

The authors thank R. Karl Hanson and Kelly M. Babchishin for their helpful consultation on change analyses.

Drew A. Kingston, PhD, is a registered psychologist in the province of Ontario and is currently the director of groups and program evaluation at the St. Lawrence Valley Correctional and Treatment Centre, a secure treatment unit for incarcerated mentally disordered offenders. He has published in the areas of serious mental illness, hypersexuality, paraphilic disorders, the impact of pornography on sexual aggression, and the sexual offense cycle.

Mark E. Olver, PhD, is professor in the Department of Psychology, University of Saskatchewan, and a registered doctoral psychologist in the province of Saskatchewan. He conducts research and provides training on the assessment and treatment of high-risk violent and sexually violent offender groups.

References

Andrews

D. A.

Bonta

(1994). The psychology of criminal conduct. Cincinnati, OH: Anderson.

Andrews

D. A.

Bonta

(2010). The psychology of criminal conduct (5th ed.). New Providence, NJ: LexisNexis.

Andrews

D. A.

Bonta

Wormith

J. S.

(1995). Level of Service Inventory–Ontario revision (LSI-OR): Interview and scoring guide. Toronto, Canada: Ontario Ministry of the Solicitor General and Correctional Services.

Andrews

D. A.

Wormith

J. S.

Kiessling

J. J.

(1985). Self-reported criminal propensity and criminal behavior: Threats to the validity of assessments and personality. Ottawa, Ontario: Solicitor General Canada, Secretariat. Retrieved from https://www.ncjrs.gov/App/Publications/abstract.aspx?ID=97992

Ashford

J. B.

Wong

K. W.

Sternbach

K. O.

(2008). Generic correctional programming for mentally ill offenders: A pilot study. Criminal Justice and Behavior, 35, 457-473. doi:10.1177/0093854807313356

Bäckström

Björklund

(2008). The Measures of Criminal Attitudes and Associates (MCAA): Further testing of structural and criterion-related validity. Criminal Justice and Behavior, 35, 1398-1410. doi:10.1177/0093854808322239

Beaudette

J. N.

Power

Stewart

L. A.

(2015). National prevalence of mental disorders among incoming federally-sentenced men offenders (Research Report, R-357). Ottawa, Ontario: Correctional Service Canada.

Beggs

S. M.

Grace

R. C.

(2011). Treatment gain for sexual offenders against children predicts reduced recidivism: A comparative validity study. Journal of Consulting and Clinical Psychology, 79, 182-192.

Bonta

Blais

Wilson

H. A.

(2013). The prediction of risk for mentally disordered offenders: A quantitative synthesis 2013-01. Ottawa, Ontario: Public Safety Canada.

10.

Bonta

Law

Hanson

(1998). The prediction of criminal and violent recidivism among mentally disordered offenders: A meta-analysis. Psychological Bulletin, 123, 123-142.

11.

Brennan

P. A.

Mednick

S. A.

Hodgins

(2000). Major mental disorders and criminal violence in a Danish birth cohort. Archives of General Psychiatry, 57, 494-500. doi:10.1001/archpsyc.57.5.494

12.

Brews

A. L.

(2009). The level of Supervision Inventory and female offenders: Addressing issues of reliability and predictive validity (Unpublished master’s thesis). Department of Psychology, University of Saskatchewan, Saskatoon, Canada.

13.

Cohen

(1992). A power primer. Psychological Bulletin, 112, 155-159.

14.

Crowne

D. P.

Marlowe

(1960). A new scale of social desirability independent of psychopathology. Journal of Consulting Psychology, 24, 349-354.

15.

Cullen

A. E.

Clarke

A. Y.

Kuipers

Hodgins

Dean

Fahy

(2012a). A multi-site randomized controlled trial of a cognitive skills programme for male mentally disordered offenders: Social–cognitive outcomes. Psychological Medicine, 42, 557-569. doi:10.1017/S0033291711001553

16.

Cullen

A. E.

Clarke

A. Y.

Kuipers

Hodgins

Dean

Fahy

(2012b). A multisite randomized trial of a cognitive skills program for male mentally disordered offenders: Violence and antisocial behavior outcomes. Journal of Consulting and Clinical Psychology, 80, 1114-1120.

17.

Diamond

P. M.

Wang

E. W.

Holzer

C. E.

Thomas

, & des Anges Cruser. (2001). The prevalence of mental illness in prison. Administration and Policy in Mental Health and Mental Health Services Research, 29, 21-40. doi:10.1023/A:1013164814732

18.

DiClemente

C. C.

Hughes

S. O.

(1990). Stages of change profiles in outpatient alcoholism treatment. Journal of Substance Abuse, 2, 217-235. doi:10.1016/S0899-3289(05)80057-4

19.

DiClemente

C. C.

Nidecker

Bellack

A. S.

(2008). Motivation and the stages of change among individuals with severe mental illness and substance abuse disorders. Journal of Substance Abuse, 34, 25-35.

20.

Douglas

K. S.

Guy

L. S.

Hart

S. D.

(2009). Psychosis as a risk factor for violence to others: A meta-analysis. Psychological Bulletin, 135, 679-706.

21.

Dowden

Andrews

D. A.

(2004). The importance of staff practice in delivering effective correctional treatment: A meta-analytic review of core correctional practice. International Journal of Offender Therapy and Comparative Criminology, 48, 203-214.

22.

Fazel

Danesh

(2002). Serious mental disorder in 23 000 prisoners: A systematic review of 62 surveys. The Lancet, 359, 545-550. doi:10.1016/S0140-6736(02)07740-1

23.

Fazel

Gulati

Linsell

Geddes

J. R.

Grann

(2009). Schizophrenia and violence: Systematic review and meta-analysis. PLoS Medicine, 6, e1000120.

24.

Fazel

Seewald

(2012). Severe mental illness in 33 588 prisoners worldwide: Systematic review and meta-regression analysis. The British Journal of Psychiatry, 200, 364-373.

25.

Girard

Wormith

J. S.

(2004). The predictive validity of the Level of Service Inventory-Ontario Revision on general and violent recidivism among various offender groups. Criminal Justice and Behavior, 31, 150-181. doi:10.1177/0093854803261335

26.

Gray

N. S.

Hill

McGleish

Timmons

MacCulloch

M. J.

Snowden

R. J.

(2003). Prediction of violence and self-harm in mentally disordered offenders: A prospective study of the efficacy of HCR-20, PCL-R, and psychiatric symptomatology. Journal of Consulting and Clinical Psychology, 71, 443-451.

27.

Hodgins

(2008). Violent behaviour among people with schizophrenia: A framework for investigations of causes, and effective treatment, and prevention. Philosophical Transactions of the Royal Society B: Biological Sciences, 363, 2505-2518. doi:10.1098/rstb.2008.0034

28.

Kingston

D. A.

Olver

M. E.

Harris

Booth

B. D.

Gulati

Cameron

(2016). The relationship between mental illness and violence in a mentally disordered offender sample: Evaluating criminogenic and psychopathological predictors. Psychology, Crime & Law, 22, 678-700.

29.

Kingston

D. A.

Olver

M. E.

Harris

Wong

S. C.

Bradford

J. M.

(2015). The relationship between mental disorder and recidivism in sexual offenders. International Journal of Forensic Mental Health, 14, 10-22. doi:10.1080/14999013.2014.974088

30.

Kroner

D. G.

Yessine

A. K.

(2013). Changing risk factors that impact recidivism: In search of mechanisms of change. Law and Human Behavior, 37, 321-336.

31.

Lewis

(2004). The relationship between the URICA and correctional treatment in a sample of violent male offenders (Unpublished doctoral dissertation). University of Saskatchewan, Saskatoon, Canada.

32.

Long

(2004). Decision-Base Brockville Quality of Life Scale: Integrated mental health software. Unpublished Integrated Mental Health Software, Washington, DC.

33.

Mandracchia

J. T.

Morgan

R. D.

(2012). Predicting offenders’ criminogenic cognitions with status variables. Criminal Justice and Behavior, 39, 5-25.

34.

Marshall

W. L.

Marshall

L. E.

Serran

G. A.

O’Brien

M. D.

(2011). Rehabilitating sexual offenders: A strength-based approach. Washington, DC: American Psychological Association. doi:10.1037/12310-000

35.

McGilloway

Donnelly

(2004). Mental illness in the UK criminal justice system: A police liaison scheme for mentally disordered offenders in Belfast. Journal of Mental Health, 13, 263-275.

36.

Mills

J. F.

Kroner

D. G.

Forth

A. E.

(2002). Measures of Criminal Attitudes and Associates (MCAA): Development, factor structure, reliability, and validity. Assessment, 9, 240-253. doi:10.1177/1073191102009003003

37.

Mills

J. F.

Kroner

D. G.

Hemmati

(2004). The Measures of Criminal Attitudes and Associates (MCAA): The prediction of general and violent recidivism. Criminal Justice and Behavior, 31, 717-733. doi:10.1177/0093854804268755

38.

Morgan

R. D.

Flora

D. B.

Kroner

D. G.

Mills

J. F.

Varghese

Steffan

J. S.

(2012). Treating offenders with mental illness: A research synthesis. Law and Human Behavior, 36, 37-50.

39.

Morlan

K. K.

Tan

S. Y.

(1998). Comparison of the Brief Psychiatric Rating Scale and the Brief Symptom Inventory. Journal of Clinical Psychology, 54, 885-894. doi:10.1002/(SICI)1097-4679(199811)54:7<885::AID-JCLP3>3.0.CO;2-E

40.

Olver

M. E.

Kingston

D. A.

Nicholaichuk

T. P.

Wong

S. C.

(2014). A psychometric examination of treatment change in a multisite sample of treated Canadian federal sexual offenders. Law and Human Behavior, 38, 544-559. doi:10.1037/lhb0000086

41.

Olver

M. E.

Stockdale

K. C.

Wormith

J. S.

(2014). Thirty years of research on the Level of Service Scales: A meta-analytic examination of predictive accuracy and sources of variability. Psychological Assessment, 26, 156-176.

42.

Overall

J. E.

Gorham

D. R.

(1962). The Brief Psychiatric Rating Scale. Psychology Reports, 10, 799-812. doi:10.2466/pr0.1962.10.3.799

43.

Peterson

J. K.

Skeem

Kennealy

Bray

Zvonkovic

(2014). How often and how consistently do symptoms directly precede criminal behavior among offenders with mental illness? Law and Human Behavior, 38, 439-449. doi:10.1037/lhb0000075

44.

Polaschek

D. L.

Anstiss

Wilson

(2010). The assessment of offending-related stage of change in offenders: Psychometric validation of the URICA with male prisoners. Psychology, Crime & Law, 16, 305-325. doi:10.1080/10683160802698766

45.

Raftery

A. E.

(1995). Bayesian model selection in social research. Sociological Methodology, 25, 111-163.

46.

Rettinger

L. J.

Andrews

D. A.

(2010). General risk and need, gender specificity, and the recidivism of female offenders. Criminal Justice and Behavior, 37, 29-46. doi:10.1177/0093854809349438

47.

Rezansoff

S. N.

Moniruzzaman

Gress

Somers

J. M.

(2013). Psychiatric diagnoses and multiyear criminal recidivism in a Canadian provincial offender population. Psychology, Public Policy, and Law, 19, 443-453.

48.

Rice

M. E.

Harris

G. T.

(2005). Comparing effect sizes in follow-up studies: ROC area, Cohen’s d, and r. Law and Human Behavior, 29, 615-620. doi:10.1007/s10979-005-6832-7

49.

Ross

R. R.

Hilborn

Liddle

(2007). Reasoning & rehabilitation 2: Short version for adults. Ottawa, Ontario: Cognitive Centre of Canada.

50.

Roy

R. E.

Wormith

J. S.

(1985). The effects of incarceration: Measuring criminal sentiments, Volume 1. Ottawa, Ontario: Ministry of the Solicitor General of Canada, Secretariat.

51.

Simourd

D. J.

(1997). The Criminal Sentiments Scale-Modified and Pride in Delinquency Scale: Psychometric properties and construct validity of two measures of criminal attitudes. Criminal Justice and Behavior, 24, 52-70. doi:10.1177/0093854897024001004

52.

Simourd

D. J.

Olver

M. E.

Brandenburg

(2016). Changing criminal attitudes among incarcerated offenders. International Journal of Offender Therapy and Comparative Criminology, 60, 1425-1445. doi:10.1177/0306624X15579257

53.

Skeem

J. L.

Manchak

Peterson

J. K.

(2011). Correctional policy for offenders with mental illness: Creating a new paradigm for recidivism reduction. Law and Human Behavior, 35, 110-126.

54.

Skeem

J. L.

Steadman

H. J.

Manchak

S. M.

(2015). Applicability of the risk-need-responsivity model to persons with mental illness involved in the criminal justice system. Psychiatric Services, 66, 916-922. doi:10.1176/appi.ps.201400448

55.

Skeem

J. L.

Winter

Kennealy

P. J.

Louden

J. E.

Tatar

J. R.

(2014). Offenders with mental illness have criminogenic needs, too: Toward recidivism reduction. Law and Human Behavior, 38, 212-224.

56.

Steadman

H. J.

Osher

F. C.

Robbins

P. C.

Case

Samuels

(2009). Prevalence of serious mental illness among jail inmates. Psychiatric Services, 60, 761-765.

57.

Tatman

A. W.

Swogger

M. T.

Love

Cook

M. D.

(2009). Psychometric properties of the Marlowe-Crowne Social Desirability Scale with adult male sexual offenders. Sexual Abuse: A Journal of Research and Treatment, 21, 21-24. doi:10.1177/1079063208325203

58.

Thomas

Donnell

A. J.

Young

T. R.

(2004). Factor structure and differential validity of the expanded Brief Psychiatric Rating Scale. Assessment, 11, 177-187. doi:10.1177/1073191103262893

59.

Vaske

J. C.

Gehring

K. S.

Lovins

(2017). Gender differences in the measurement of criminal thinking. Criminal Justice and Behavior, 44, 395-415.

60.

Velasquez

M. M.

Crouch

Stephens

N. S.

DiClemente

C. C.

(2015). Group treatment for substance abuse: A stages-of-change therapy manual. New York, NY: Guilford.

61.

Walters

G. D.

(2016). Friends, cognition, and delinquency: Proactive and reactive criminal thinking as mediators of the peer influence and peer selection effects among male delinquents. Justice Quarterly, 33, 1055-1079.

62.

Wexler

D. B.

(2013). The STOP program: Innovative skills techniques options and plans for better relationships. New York, NY: W. W. Norton & Company.

63.

Winogron

Van Dieten

Gauzas

(1997). Controlling anger and learning to manage it. Toronto, Ontario, Canada: Multi Health Systems.

64.

Wormith

J. S.

(2011). The legacy of D.A. Andrews in the field of criminal justice: How theory and research can change policy and practice. International Journal of Forensic Mental Health, 10, 78-82.

65.

Wormith

J. S.

Andrews

D. A.

(1995, June). The development and validation of three measures of criminal sentiments and their role in the assessment of offender attitudes. Paper presented at the annual convention of the Canadian Psychological Association, Charlottetown, Canada.

66.

Wormith

J. S.

Hogg

Guzzo

(2012). The predictive validity of a general risk/needs assessment inventory on sexual offender recidivism and an exploration of the professional override. Criminal Justice and Behavior, 39, 1511-1538. doi:10.1177/0093854812455741

67.

Yong

A. D.

Williams

M. W. M.

Provan

Clarke

Sinclair

(2015). How do offenders move through the stages of change? Psychology, Crime & Law, 21, 375-397. doi:10.1080/1068316X.2014.989166

Psychometric Examination of Treatment Change Among Mentally Disordered Offenders: A Risk–Needs Analysis

Abstract

Keywords

Method

Participants

Treatment Program

Measures

Marlowe–Crowne Social Desirability Scale (MCSDS)

Level of Service Inventory–Ontario Revision (LSI-OR)

MCAA

The CSS-M

The University of Rhode Island Change Assessment (URICA) Scale

Brief Psychiatric Rating Scale (BPRS)

The Brockville Quality of Life Scale (QoL)

Recidivism

Procedure and Data Analytic Plan

Results

Pretreatment–Posttreatment Comparisons

Convergent Validity Associations

Predictive Accuracy

Psychometric Treatment Change and Recidivism

Discussion

Evaluating the Risk Relevance of Within Treatment Change

Limitations, Conclusions, and Future Directions

Footnotes

Acknowledgements

References