Treatment Attrition of Probationers With Mental Illness From an Enhanced Day Reporting Center

Abstract

Treatment completion is an important outcome for both mental health and criminal justice agencies tasked with managing offenders with mental illness in the community. Previous research has shown that greater degrees of criminogenic risk factors (e.g., specific criminal history variables) predict treatment non-completion among legally mandated populations. However, most studies were conducted with offenders without mental illness. In this study, demographic (e.g., age, gender), clinical (e.g., psychiatric diagnosis), and criminogenic risk factors (measured using the Level of Service Inventory–Revised [LSI-R]) were compared by treatment completion status using 167 probationers with mental illness treated at an enhanced day reporting center. Bivariate and multivariate (i.e., forward entry logistic regression) analyses revealed that while the LSI-R total score was unrelated to treatment completion, higher scores on the LSI-R Alcohol and Drug use subscale (odds ratio [OR] = 1.25, 95% confidence interval [CI] = [1.01, 1.54]) and older age (OR = 1.04, 95% CI = [1.00, 1.09]) were significantly predictive of non-completion.

Keywords

treatment attrition treatment completion mentally ill offender day reporting program Level of Service Inventorys

Of the approximately 4 million probationers in the United States (Maruschak & Parks, 2012), 16% are estimated to suffer from mental illnesses (Ditton, 1999). This statistic appears to underestimate the service needs of probationers, as 23% report mental health service usage and report significantly more symptoms of psychiatric disorder compared with non-probationers (Crilly, Caine, Lamberti, Brown, & Friedman, 2009). These unmet needs may affect their legal supervision status as parolees with mental illness are more likely to have this status revoked as compared with their non-mentally ill peers, likely due to compliance failures (Porporino & Motiuk, 1995). Indeed, one of the most difficult problems which mental health and community corrections agencies face is supervising participants through to program completion. In this context, treatment success and failure not only impact the probationer’s mental health symptoms, but also their criminal justice status, and ultimately their risk of criminal recidivism (Olver, Stockdale, & Wormith, 2011). The present study seeks to assess the relative importance of factors that impact treatment completion among a sample of probationers with mental illness. These include criminogenic risk/need and psychiatric variables.

After many years of competing theories (Skeem & Manchak, 2008), a philosophy of community corrections (i.e., probation, parole, and various problem-solving courts) grounded in Risk Need Responsivity (RNR) is currently being integrated into practice. This principle holds that criminal behavior can be largely explained (and predicted) by deficits in specific social/environmental and psychological Risk/Needs (Andrews & Kiessling, 1980; Andrews & Robinson, 1984), such as the “Big Four” of antisocial history, antisocial personality, criminal thinking, and antisocial peers (Andrews, Bonta, & Wormith, 2006). Consequently, correctional program interventions will be most successful at reducing recidivism if they target moderate and high Risk individuals, focus on criminogenic Needs, and are delivered Responsively to the learning style of the individual (e.g., tailored to the mentally ill; Andrews & Bonta, 2010). Within this framework, research focused on treatment completion and retention can inform theory with regard to the identification of factors, which require particularly responsive accommodation.

There is a wealth of research on the factors that affect early termination from mental health treatment for non-offenders. Wierzbicki and Pekarik (1993), in their often-cited meta-analysis of 125 psychotherapy outcome studies, identified race (minority status), income, and education as significant correlates of dropout and calculated a mean dropout rate of 47% of participants in psychotherapy. Psychiatric symptom and psychological variables were not able to be analyzed due to methodological limitations in this study. However, another psychotherapy outcome meta-analysis of patients with severe mental illness found no association of diagnosis with dropout rates (Nos, Barbui, & Tansella, 2003). While the evidence for the importance of acute psychiatric symptoms to attrition is limited, several studies have found that the presence of personality disorders can predict treatment attrition for select populations and treatment modalities such as psychodynamic therapy for depression (Diguer, Barber, & Luborsky, 1993), cognitive therapy for depression (Persons, Burns, & Perloff, 1988), and residential treatment for substance abuse (Meier & Barrowclough, 2009). In addition, younger age and verbal fluency impairments have also been noted as factors influencing dropout among chronic, psychotic outpatients (Kurtz, Rose, & Wexler, 2011).

In a meta-analysis of attrition rates of offender treatment programs, Olver et al. (2011) reviewed 114 studies of offenders in psychotherapeutic interventions geared toward generic and specific offender populations (i.e., sex offender and domestic violence), and in a variety of settings (e.g., custodial and community-based). An average attrition rate of 27.1% was calculated across studies. Among the findings relevant for the present study, younger age (r_w = −.10), unemployment status (r_w = .14), and higher scores on actuarial measures of criminogenic risk (r_w = .11) significantly predicted dropout from treatment. While the presence of psychosis or any personality disorder were unrelated to dropout, the presence of borderline personality disorder (r_w = .07) and antisocial personality disorder (r_w = 0.14) were significantly predictive of dropout, as was the diagnosis of a substance abuse problem (r_w = .07), and depression (with the addition of an outlier study; r_w = 0.01). Furthermore, across all types of programs, dropout from treatment was significantly associated with increased general, violent and nonviolent recidivism, with average recidivism rates 10% to 23% higher for non-completers.

A closer review of individual studies of offender subgroups reveals some interesting and pertinent findings regarding predictors of treatment failure. Several studies investigated treatment non-completion for in-prison substance abusers and treatment engagement after discharge. Hiller, Knight, and Simpson (1999) found that early dropout from a corrections-based, modified substance abuse therapeutic community was related to having a history of mental health treatment, cocaine dependence, and having higher scores on a measure of anxiety and depression. Using a sample of 339 felony probationers in this intensive, 6-month form of treatment, the authors found that the strongest predictor of dropout (regardless of reason) using multivariate analyses was higher scores on a criminality index risk assessment scale. Furthermore, among a community-based sample of 259 substance abusers released from prison, Houser, Salvatore, and Welsh (2012) identified individual-level factors associated with post-release treatment engagement. In addition to demographic variables, they included a commonly used risk assessment instrument, the Level of Service Inventory–Revised (LSI-R), as well as a measure of severity of substance abuse. Lower total scores on the LSI-R were found to be the strongest predictor of success in treatment using logistic regression analyses. Similar results were found in a study of retention among a final sample of 179 dually diagnosed (i.e., mentally ill and substance abusing) male inmates residing in a prison-based therapeutic community (Van Stelle, Blumer, & Moberg, 2004). The 25% of admissions who completed the program scored significantly lower on a measure of psychopathy, general criminal risk (as assessed by a Department of Corrections criminality index), and psychiatric symptoms (assessed by the Brief Symptom Inventory [BSI]).

While several studies have examined in-prison and community-based offender substance abuse treatment, few studies have examined predictors of success among samples of offenders with mental illness. Cullen, Soria, Clarke, Dean, and Fahy (2011) reported on dropouts from an intensive, cognitive-behavioral group intervention created to reduce criminogenic thinking styles among offenders with mental illness. Of their sample of 42 eligible severely mentally disordered offenders (most often having a primary diagnosis of a Diagnostic and Statistical Manual of Mental Disorders—4th ed.; DSM-IV; American Psychiatric Association [APA], 1994—Psychotic disorder) housed in a medium secure hospital setting, half completed the 35 to 36 two-hour sessions. Using stepwise logistic regressions including age, education, a risk assessment measure the Historical Clinical Risk -20 item (HCR-20) (Webster, Douglas, Eaves, & Hart, 1997) and a measure of psychopathy Psychopathy Checklist -Screening Version (PCL-SV) (Hart, Cox, & Hare, 1995), they found that the antisocial lifestyle component of psychopathy, a diagnosis of antisocial personality disorder, and recent violence significantly and uniquely predicted dropout. A measure of the positive and negative symptoms of schizophrenia was unrelated to treatment dropout (regardless of reason). Notably, scores on the Clinical and Risk, but not the Historical, components of the HCR-20 were significantly higher for non-completers versus completers. The Clinical and Risk parts of this instrument measure such features as active symptoms of mental illness, exposure to destabilizers, and poor insight. The Historical part of this scale, however, measures static risk factors like previous violence and prior supervision failure. This finding suggests the importance of treatment-related variables, as opposed to general criminal risk variables, in predicting treatment attrition.

From this review, it appears that few studies have focused on treatment completion among offenders with mental illness. However, research on treatment completion within general psychotherapeutic, general offender, and substance abusing offenders can inform our expectations about important predictors of treatment completion among offenders with mental illness in community settings. This is an important and timely topic as community corrections programs are utilized more often by State governments in their efforts to find community solutions to criminal behavior. Day reporting programs (DRCs) are one such form of community corrections program. DRCs are considered a form of intermediate sanction for offenders, which provide enhanced supervision and links to treatment services, and target those who would otherwise be incarcerated. Offenders typically spend their time during the day, supervised and receiving a variety of vocational and psycho-educational support services (Parent, Bryne, Tsarfaty, Valade, & Esselman, 1995). There is some evidence of the effectiveness of DRCs in reducing criminal recidivism. Ostermann (2009) found significantly lower rates of re-arrest and re-conviction among parolees assigned to DRCs, versus parolees not assigned to a program, and versus released inmates without parole supervision.

There are clearly several studies that have found common demographic variables as risk factors for early termination from psychotherapeutic treatment such as younger age and ethnicity. Psychiatric diagnoses, such as the presence of psychosis, have not been shown to be strong predictors of treatment outcome, particularly among the mentally ill. Finally, there is evidence that general criminal risk assessment measures are strong predictors of treatment termination.

The present study seeks to further research into treatment completion among offenders with mental illness. Several variables that were previously identified as important to this outcome will be assessed for their importance to predicting attrition from a DRC, enhanced to accommodate probationers with mental illness. Younger participants, and those with personality disorders and substance abuse disorders are expected to be found significantly more often among treatment non-completers. Given the consistent finding of the importance of overall criminal risk to non-completion of mandated treatment, those with higher total scores on the LSI-R are also expected to be significantly overrepresented among those who do not complete treatment.

Method

Program and Participants

The program in which this study was conducted is an enhanced DRC termed the Community Reporting Engagement Support and Training (CREST). This enhanced DRC is engaged in a collaborative arrangement with the local community mental health clinic to provide psychiatric assessment and treatment services and utilizes a variety of evidence-based strategies designed to reduce criminality. These strategies include use of a day reporting format, case management, as well as providing psychosocial programs, which utilize a cognitive-behavioral approach, including Illness Management and Recovery (Mueser et al., 2006), and Thinking for a Change (Golden, Gatchel, & Cahill, 2006). In addition, several manual-based, cognitive-behavioral and psycho-educational groups are offered, which target relapse prevention, health and wellness, and life skills. The DRC serves only offenders with psychiatric or co-occurring conditions referred from a variety of criminal justice agencies, but the majority is received from the state’s probation department.

Because of their low number, limited availability of criminal history records, and the potentially different circumstances of non-probationers, only probationers were included in this study. Some probationers are referred to the program as a technical condition of their probation, and others are not. Thus, failure to complete may result in a technical violation for some and other penalties for others. Program schedules for the DRC clients are arranged based on their clinical needs as assessed by case managers. Therefore, some clients may be required to attend daily and remain on site for most of the day, while others will be placed on a less intensive schedule. In addition to their direct clinical activities, case managers coordinate and facilitate client appointments at social welfare, clinical, and correctional entities. Progress updates are also provided to correctional referral agencies on a regular basis. Program completion is determined by DRC staff when clients have met the program goals outlined during the assessment process or they have been assessed and referred to a higher level of care (e.g., residential substance abuse treatment). Approximately 80% to 85% of clients who complete the DRC remain under probation supervision for many months after.

The sample comprised all admitted probationers to the DRC between July 1, 2007 (the program’s inception date) and whose discharge occurred on or prior to July 1, 2012, N = 211. Twenty of these individuals were admitted twice and one was admitted on three occasions. These duplicate admissions were removed from the analysis and their first encounter with the DRC was included, leaving a unique sample of n = 190. Consistent with previous studies of treatment attrition among offenders (Olver et al., 2011), those participants who were arrested or remanded while in the program were removed to separate risk of criminality (i.e., criminogenic risk/need) from the risk of treatment attrition (i.e., responsivity). These n = 23 individuals did not differ from the other participants on age, risk level, Global Assessment of Functioning (GAF) score, prior arrests or convictions, gender, or ethnicity. Their removal left a final sample of n = 167. Of this group, 132 successfully completed (79%) and spent a mean of 174 (range = 17-712; SD = 137) days in the program, as compared with 35 non-completers who spent an average of 134 days in the program (range = 11-605; SD = 134). There were several potential reasons for a classification of non-completion by the DRC including dropout, failure to comply, and absconding. Information that could distinguish between program-or participant-initiated discharge was not available at the time of this analysis.

The average participant (excluding missing data points) of this final sample was 35.99 (SD = 10.5; range = 18-62) years of age, male (n = 136; 81.4%), primarily English-speaking (n = 141; 84.4%), African American (n = 93; 55.7%), Caucasian American (n = 42; 25.1%), or Hispanic American (n = 27; 16.2%), with the majority’s highest level of education less than regular 12th grade (n = 81; 53.3%). While a variety of Diagnostic and Statistical Manual of Mental Disorders (4th ed., text rev.; DSM-IV-TR; APA, 2000) psychiatric diagnoses were applied to patients on admission, the majority were categorized as suffering primarily from a psychotic disorder (n = 103, 61.7%), bipolar disorder (n = 30, 18%), or depressive disorder (n = 10, 6%). Of the overall sample, Schizophrenia spectrum (n = 87; 52.1%) disorders, which included paranoid schizophrenia and schizoaffective disorder, were common. The vast majority of participants carried a diagnosis of a substance use disorder (n = 130; 77.8%), the most common being polysubstance dependence (n = 32; 22.8%). In addition, a small group (n = 28; 16.8%) suffered from a personality disorder, most often antisocial personality disorder (n = 12, 7.2%). The average number of prior arrests for the sample was M = 13.42 (SD = 12.42; range = 0-94), which led to an average of 8.52 (SD = 8.04; range = 0-54) convictions. The sample’s risk levels using LSI-R total score ranged from 14 to 48, with n = 4 (2%) considered Low Risk (LSI-R total score = 1-18), n = 25 (15%) as Medium Risk (LSI-R total score = 19-28), and n = 123 (74%) as High Risk (LSI-R total score = 29-54) according to U.S. norms (Andrews & Bonta, 2003), with 15 participant scores unavailable.

Measures

LSI-R

The LSI-R (Andrews & Bonta, 1995) is a well-established measure of criminogenic risk level that uses a structured rating scale format. It consists of 54 binary items (i.e., rated 0 if absent, 1 if present), giving a maximum score of 54. This full score is divided into 10 major areas: Criminal History, Employment/Education, Financial, Family/Marital, Accommodation, Leisure/Recreation, Companions, Alcohol/Drug, Emotional/Personal, and Attitudes/Orientation. The LSI-R’s psychometric properties have been previously described among criminal justice populations (Bonta & Motiuk, 1990). In the present study, the total score demonstrated adequate internal consistency, α = .79, with a range of subscale statistics reflecting the number of items within each scale from a low of Financial = .15 to a high of Alcohol/Drugs α = .73 (see Table 1).

Table 1.

Inter-Correlations and Alpha Coefficients for LSI-R Total and Subscales.

LSI-R scale	1	2	3	4	5	6	7	8	9	10	Total
1. Criminal History	(.68)
2. Education/Employment	.18*	(.68)
3. Financial	−.02	.08	(.15)
4. Family/Marital	.19*	.08	.16	(.50)
5. Accommodations	.13	.15	.11	.25**	(.41)
6. Leisure/Recreation	−.12	.25**	.15	.08	.17*	(.44)
7. Companions	−.08	.18*	.20*	.36***	.09	.21**	(.53)
8. Alcohol/Drug	.13	.17*	.25**	.13	.24**	.17*	.16*	(.73)
9. Emotional/Personal	.11	−.08	.18*	.23**	.02	.06	.17*	.12	(.24)
10. Attitude/Orientation	.07	.15	.09	.18*	.03	.19*	.20*	.03	.03	(.66)
LSI-R total score	.51***	.59***	.29***	.47***	.41***	.30***	.41***	.55***	.24**	.39***	(.79)

Note. Alpha coefficients for the LSI subscales are presented in parentheses. Spearman’s Rho was performed for all inter-correlation analyses. LSI-R = Level of Service Inventory–Revised.

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

Procedure and Statistical Analyses

Prior to beginning this archival research, appropriate Institutional Review Board (IRB) approvals were obtained from the University of New Haven IRB. A review of electronic and paper records of the data from the DRC study site was then conducted to collect demographic, clinical, and risk assessment variables. This information contained basic demographic data as well as psychiatric diagnostic labels according to the DSM-IV for all five axes. All psychiatric diagnoses were applied by the participants’ treating psychiatrist at the local community mental health center. Where participants had several Axis I conditions, his or her first or primary Axis I condition was analyzed. All Axis I conditions that were substance use disorders were treated as a separate disorder, and primary substance use disorders were analyzed separately. In addition, personality disorders and GAF scores were also collected and analyzed.

As part of standard intake procedure of the program, separate LSI-R forms were completed by the referring probation officer as well as the receiving program case manager. In all but a few cases of missing data, the LSI-R records from the latter source were used in our analyses as they were more likely to be on record and relied on more recent information. Both probation officers and case managers were appropriately trained in the use of the LSI as part of their position duties. Criminal history records were obtained from the state probation agency using identifiers gathered from the DRC to calculate the average number of arrests and convictions prior to program enrollment.

Reliability analyses were conducted on the LSI-R total and subscales. Bivariate analyses (i.e., t tests for continuous data, chi-square for categorical data) were used to assess the differences between treatment completers and non-completers using various demographic, clinical, and risk assessment variables. In several cases, Mann–Whitney U analyses were performed because of skewed or kurtotic data. This was particularly true for the subscales of the LSI-R, because the program targets individuals who are generally considered higher risk according to the LSI-R. This results in a negative skew toward higher LSI-R scores. For example, LSI-R Criminal History skew = −0.47 (SE = 0.20), kurtosis = −0.60 (SE = 0.39); LSI-R Financial skew = −1.53 (SE = 0.20), kurtosis = 1.46 (SE = 0.39). Multivariate binary logistic regression analyses were used to compare characteristics which were previously identified as significantly different between completion groups to determine the most important predictors of treatment failure. Given the exploratory nature of the study, a more liberal p < .10 was used in the bivariate analyses to determine inclusion in the multivariate analysis. An inter-correlation matrix of LSI-R scales was calculated to assess multicollinearity in preparation for the regression analyses, using Spearman’s Rho correlations due to variable skew.

Results

As shown in Table 1, the psychometric properties of the LSI-R for this sample were variable. The Total scale score showed adequate internal consistency (α = .79) with a range from a low of Financial (α = .15) to a high of Alcohol/Drug (α =.73). Inter-correlations among the various subscales ranged from r_s (150) = −0.08 to r_s (150) = 0.59, which indicated an acceptable level of scale co-variance.

Table 2 shows that treatment completers were not significantly different from non-completers on any of the demographic or psychiatric variables. However, non-completers had a significantly higher LSI-R Accommodation risk score U = 1,361.50, z = −2.09, p = .04, and significantly higher LSI-R Alcohol and Drug-Related risk score, U = 1,206, z = −2.73, p = .01. Although only reaching trend levels of significance, non-completers were significantly older, t(165) = 1.74, p = .08.

Table 2.

Comparison of Treatment Discharge Groups.

Variable	Non-completers (n = 35)	Completers (n = 132)	t/χ²/U	p
Variable	M (SD)/%	M (SD)/%	t/χ²/U	p
Age at admission^a	38.63 (10.56)	35.29 (9.96)	1.74	.083
Gender (male)^b	80.0	81.8	0.06	.806
Race^b			1.95	.583
African American	60.0	54.5
Caucasian	22.9	25.8
Hispanic	11.4	17.4
Other	5.7	2.3
Psychiatric diagnosis^b
Schizophrenia spectrum (Y/N)	57.1	50.8	0.45	.501
Personality disorder (Y/N)	14.3	17.4	0.64	.721
Antisocial personality (Y/N)	5.7	7.6	0.14	.705
Borderline personality (Y/N)	0.0	3.8	1.37	.242
Substance use disorder (Y/N)	82.9	76.5	0.65	.422
Polysubstance dependence (Y/N)	28.6	21.2	0.85	.356
GAF on admission	49.67 (6.15)	49.11 (6.54)	0.44	.659
Criminal history^c
Previous arrests	14.37 (10.64)	13.17 (12.88)	2,071.00	.347
Previous convictions	9.40 (7.23)	8.29 (8.25)	2,031.00	.272
LSI-R scales^c
Criminal History	5.83 (2.62)	5.70 (2.15)	1,698.50	.687
Employment/Education	6.17 (2.02)	6.03 (2.44)	1,771.50	.955
Financial	1.86 (0.35)	1.66 (0.56)	1,483.00	.068
Family/Marital	2.07 (1.19)	2.27 (1.28)	1,618.50	.424
Accommodation	2.14 (0.79)	1.72 (0.96)	1,361.50	.036
Leisure/Recreation	1.79 (0.49)	1.70 (0.57)	1,667.50	.454
Companions	3.45 (1.27)	3.63 (1.24)	1,632.50	.459
Alcohol/Drugs	6.21 (2.57)	5.08 (2.23)	1,206.00	.006
Emotional/Personal	4.34 (0.72)	4.22 (0.82)	1,658.00	.524
Attitudes/Orientation	1.86 (1.48)	1.54 (1.35)	1,559.00	.279
Total	35.34 (1.48)	33.41 (6.74)	1,438.50	.105

Note. Psychiatric diagnoses according to DSM-IV-TR. GAF = Global Assessment of Functioning; LSI-R = Level of Service Inventory–Revised; DSM-IV-TR = Diagnostic and Statistical Manual of Mental Disorders (4th ed., text rev.; American Psychiatric Association, 2000).

t tests utilized for significance testing.

χ² utilized for significance testing.

Mann–Whitney U utilized for significance testing.

Finally, the LSI-R scales of Alcohol/Drugs, Accommodation, and Financial as well as Age at Admission were entered into a forward entry (Likelihood Ratio) logistic regression equation to determine the best predictor of treatment non-completion (i.e., non-completion coded as 1, completion coded as 0) for this sample of mentally ill probationers (see Table 3 below). This resulted in a significant model, R² = .10 (Nagelkerke), χ²(2) = 10.08, p = .006, for LSI-R Alcohol/Drugs (odds ratio [OR] = 1.25, 95% confidence interval [CI] = [1.01, 1.54]) and Age at Admission (OR = 1.04, 95% CI = [1.00, 1.09]). This shows that for every unit increase in this LSI scale, the odds of dropping out of the program increase by 25%.

Table 3.

Forward Entry LR Predicting Treatment Non-Completion.

	B	SE	Wald/score	Significance	OR	95% CI for OR
	B	SE	Wald/score	Significance	OR	LL	UL
Step 1
LSI-R-alcohol/drugs	0.24	0.10	5.25	.022	1.27	1.04	1.56
Constant	−2.80	0.66	17.82	.000	0.06
Age at admission^a			4.08	.044
LSI-R accommodation^a			2.84	.092
LSI-R financial^a			1.98	.159
Step 2
Age at admission	0.04	0.02	3.95	.047	1.04	1.00	1.09
LSI-R-alcohol/drugs	0.22	0.11	4.27	.039	1.25	1.01	1.54
Constant	−4.31	1.05	16.71	.000	0.01
LSI-R accommodation^a			3.65	.056
LSI-R financial^a			1.21	.272

Note. LR = Logistic Regression; OR = odds ratio; CI = confidence interval; LL = lower limit; UL = upper limit; LSI-R = Level of Service Inventory–Revised.

Predictor excluded at this step. Score statistic utilized.

Discussion

The present study sought to extend recent research into the factors influencing treatment non-completion for legally mandated populations. Previous studies on this topic using samples of substance abusing offenders in (Hiller et al., 1999) and out (Houser et al., 2012) of prison, a meta-analysis of non-mentally ill offenders (Olver et al., 2011), and an inpatient study of offenders with mental illness (Cullen et al., 2011) found criminogenic risk levels to be an important predictor of non-completion of treatment. However, the present study did not find a relationship between the total score on a commonly utilized criminogenic risk assessment measure (i.e., the LSI-R) and treatment attrition. Other characteristics that previously predicted treatment non-completion among non-mentally ill offenders (i.e., younger age; Olver et al., 2011) were also not significant in this study. Indeed, older age was found to be significantly predictive of non-completion. Finally, psychiatric diagnostic variables, which were previously poor predictors of non-completion among offenders with mental illness (Cullen et al., 2011), but important among substance abusers (Meier & Barrowclough, 2009) and general offenders (Olver et al., 2011), were, expectedly, also not significant in distinguishing completers from non-completers.

The reason for this lack of finding with regard to overall criminogenic risk/need is possibly due to the nature of the sample. As described in the literature review, studies using the LSI-R to predict treatment outcomes have typically involved samples of general offenders or those in substance abuse treatment. Previous studies involving risk assessment in samples of psychiatric patients in criminal justice settings have utilized the HCR-20 (e.g., Cullen et al., 2011). The latter instrument contains many more items specific to the treatment of psychiatric patients such as Lack of Insight, Active Symptoms of Major Mental Illness, Impulsivity, and Unresponsive to Treatment. Indeed, the Historical subscale of the HCR-20, which Cullen et al. (2011) did not find predictive of treatment dropout, contains many items, which are similar to those on the LSI-R such as Presence of a Major Mental Illness and Prior Supervision Failure. While these items may be important and distinguishing features among offenders without mental illness, it is unclear if they are as useful among the mentally ill. Perhaps an instrument that includes more items specific to the mentally ill will improve the assessment of mentally ill probationers for treatment planning purposes. Furthermore, unlike prisoners and parolees, our sample of probationers may contain many individuals for whom a chronic criminal disposition is not characteristic. The percentage of those with antisocial personality disorder, for instance, found in this study of less than 8% is much lower than prevalence estimates of this condition in prison (e.g., 47%; Fazel & Danesh, 2002) and substance-abusing probationer (24.3%) populations (Lurigio et al., 2003). Thus, the predictive power of criminal history variables may lose their potency in our sample.

Somewhat less clear is the explanation for the finding of older age and risk of attrition. While this finding is only in the trend direction, it was one of the two most important predictors in the forward entry regression analysis. As noted in the literature review, several studies have identified younger age as an important predictor of dropout from treatment in psychotherapeutic outpatient settings (Wierzbicki & Pekarik, 1993) and in various offender settings (Olver et al., 2011). This finding extends to patients with more severe conditions such as schizophrenia (Linden et al., 2001). However, other studies have also found the absence of a relationship between age and dropout in a forensic psychiatric setting (Cullen et al., 2011) and a sample of offenders in drug aftercare treatment (Houser et al., 2012). It is possible that a unique set of variables in this sample resulted in this finding. An examination of client attitudes may help elucidate this finding.

Our finding with regard to the Alcohol/Drug risk assessment was significant and indicates that those who score higher on this particular subscale of the LSI-R are at increased risk of noncompliance. This section of the LSI-R is composed of nine items that identify previous and current alcohol and drug problems as well as marital/family, medical, legal, and employment problems that result from alcohol or drug use. This is possibly explained by the fact that within offender programs that do not specialize in substance abuse treatment, having a substance abuse problem is an important predictor of treatment non-completion (Olver et al., 2011) suggesting that enhanced substance abuse interventions may assist in reducing non-completion rates in this population.

There are several limitations to this study. The use of psychiatric diagnoses identified by practicing clinical staff introduces potential error in the form of inconsistencies between diagnosing clinicians, which would be reduced if structured interviews were utilized. The lack of structured interviews may also have resulted in an under-identification of personality disorders. Second, while psychiatric diagnostic labels provide a clinically approachable language with which to communicate mental illness, a measure of symptom severity would have provided a more nuanced variable and thus perhaps a more significant predictor of treatment outcome. Finally, the “in the field” nature of this study suggests that there may be variation in standards of the LSI-R between case managers at the facility despite the fact that they were trained in the use of the instrument. Unfortunately, there was insufficient data for a formal inter-rater reliability analysis. In addition, in some cases, the referring probation officer completed the LSI-R prior to referral to the program, and thus, for a small number of these protocols, there may be variation in educational background and training of those completing the LSI-R. However, as all parties were trained in the official scoring of the LSI-R, there should be acceptable consistency between scorers. Despite these limitations, this study represents an important step in understanding the role of criminogenic factors in predicting treatment non-completion among mentally ill offenders. Future studies on this topic could include separate analyses for subtypes of non-completion to elucidate characteristics that predict participant-versus program-initiated attrition.

Footnotes

Acknowledgements

The authors would like to thank Dana Ashby, Gregory Bivens, and the staff at CREST and The Connection, Inc.; Charles Barber, Director of The Connection Institute for Innovative Practice; Brian Hill of the Court Support Services Division of the State of Connecticut; and Taylor Scarpone, University of New Haven undergraduate student, for their assistance in data collection.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The research reported in this article was aided by a University of New Haven Summer Research Grant.

References

American Psychiatric Association. (1994). Diagnostic and statistical manual of mental disorders (4th ed.). Washington, DC: Author.

American Psychiatric Association. (2000). Diagnostic and statistical manual of mental disorders (4th ed., text rev.). Washington, DC: Author.

Andrews

D. A.

Bonta

J. L.

(1995). The Level of Service Inventory-Revised. Toronto, Ontario, Canada: Multi-Health Systems.

Andrews

D. A.

Bonta

J. L.

(2003). Level of Supervision Inventory-Revised: U.S. Norms Manual Supplement. Toronto, Ontario, Canada: Multi Health Systems.

Andrews

D. A.

Bonta

J. L.

(2010). The psychology of criminal conduct (4th ed.). New Providence, NJ: Matthew Bard.

Andrews

D. A.

Bonta

J. L.

Wormith

J. S.

(2006). The recent past and near future of risk and/or need assessment. Crime & Delinquency, 52, 7-27. doi:10.1177/0011128705281756

Andrews

D. A.

Kiessling

J. J.

(1980). Program structure and effective correctional practices: A summary of the CAVIC research. In Ross

R. R.

Gendreau

(Eds.), Effective correctional treatment (pp. 441-463). Toronto, Ontario, Canada: Butterworths.

Andrews

D. A.

Robinson

(1984). The Level of Supervision Inventory: Second report :A report to Research Services-. Toronto, Canada: Ontario Ministry of Correctional Services.

Bonta

Motiuk

L. L.

(1990). Classification to correctional halfway houses: A quasi-experimental evaluation. Criminology, 28, 497-506.

10.

Crilly

J. F.

Caine

E. D.

Lamberti

J. S.

Brown

Friedman

(2009). Mental health services use and symptom prevalence in a cohort of adults on probation. Psychiatric Services, 60, 542-544.

11.

Cullen

A. E.

Soria

Clarke

A. Y.

Dean

Fahy

(2011). Factors predicting dropout from the Reasoning and Rehabilitation program with mentally disordered offenders. Criminal Justice and Behavior, 38, 217-230. doi:10.1177/0093854810393659

12.

Diguer

Barber

J. P.

Luborsky

(1993). Three concomitants: Personality disorders, psychiatric severity, and outcome of dynamic psychotherapy of major depression. American Journal of Psychiatry, 150, 1246-1248.

13.

Ditton

P. M.

(1999). Mental health treatment of inmates and probationers (Bureau of Justice Special Reports, NCJ 174463). Washington, DC: U.S. Department of Justice.

14.

Fazel

Danesh

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

15.

Golden

L. S.

Gatchel

R. J.

Cahill

M. A.

(2006). Evaluating the effectiveness of the National Institute of Corrections’ “Thinking for a Change” program among probationers. Journal of Offender Rehabilitation, 43(2), 55-73. doi:10.1300/J076v43n02_03

16.

Hart

Cox

Hare

(1995). The Hare PCL: SV Psychopathy Checklist: Screening Version. New York: Multi-Health Systems.

17.

Hiller

M. L.

Knight

Simpson

D. D.

(1999). Risk factors that predict dropout from corrections-based treatment for drug abuse. The Prison Journal, 79, 411-430. doi:10.1177/0032885599079004004

18.

Houser

K. A.

Salvatore

Welsh

W. N.

(2012). Individual-level predictors of community aftercare completion. The Prison Journal, 92, 106-124.

19.

Kurtz

M. M.

Rose

Wexler

B. E.

(2011). Predictors of participation in community outpatient psychosocial rehabilitation in schizophrenia. Community Mental Health Journal, 47, 622-627. doi:10.1007/s10597-010-9343-z

20.

Linden

Godemann

Gaebel

Kopke

Muller

Muller-Spann

. . .Tegeler

(2001). A prospective study of factors influencing adherence to a continuous neuroleptic treatment program in schizophrenia patients during 2 years. Schizophrenia Bulletin, 27, 585-596.

21.

Lurigio

A. J.

Cho

Y. I.

Swartz

J. A.

Johnson

T. P.

Graf

Pickup

(2003). Standardized assessment of substance-related, other psychiatric, and comorbid disorders among probationers. International Journal of Offender Therapy and Comparative Criminology, 47, 630-652. doi:10.1177/0306624X03257710

22.

Maruschak

L. M.

Parks

(2012). Probation and parole in the United States, 2011. Washington, DC: Bureau of Justice Statistics.

23.

Meier

P. S.

Barrowclough

(2009). Mental health problems: Are they or are they not a risk factor for drop out from drug treatment? A systematic review of the evidence. Drugs: Education, Prevention and Policy, 16, 7-38. doi:10.1080/09687630701741030

24.

Mueser

K. T.

Meyer

P. S.

Penn

D. L.

Clancy

D. M.

Salyers

M. P.

(2006). The Illness Management and Recovery program: Rationale, development, and preliminary findings. Schizophrenia Bulletin, 32(Suppl.), S32-S43. doi:10.1093/schbul/sbl022

25.

Nos

Barbui

Tansella

(2003). How often do patients with psychosis fail to adhere to treatment programmes? A systematic review. Psychological Medicine, 33, 1149-1160. doi:10.1017/S0033291703008328

26.

Olver

M. E.

Stockdale

K. C.

Wormith

J. S.

(2011). A meta-analysis of predictors of offender treatment attrition and its relationship to recidivism. Journal of Consulting and Clinical Psychology, 79, 6-21. doi:10.1037/a0022200

27.

Ostermann

(2009). An analysis of New Jersey’s day reporting program and halfway back programs: Embracing the rehabilitative ideal through evidence based practices. Journal of Offender Rehabilitation, 48, 139-153. doi:10.1080/10509670802640958

28.

Parent

D. G.

Bryne

Tsarfaty

Valade

Esselman

(1995). Day reporting centers (Vols. 1 & 2). Washington, DC: U.S. Department of Justice.

29.

Persons

J. B.

Burns

D. D.

Perloff

J. M.

(1988). Predictors of dropout and outcome in cognitive therapy for depression in a private practice setting. Cognitive Therapy and Research, 12, 557-575.

30.

Porporino

F. J.

Motiuk

L. L.

(1995). The prison careers of mentally disordered offenders. International Journal of Law and Psychiatry, 18, 29-44.

31.

Skeem

J. L.

Manchak

(2008). Back to the future: From Klockar’s model of effective supervision to evidence-based practice in probation. Journal of Offender Rehabilitation, 47, 220-247. doi:10.1080/10509670802134069

32.

Van Stelle

K. R.

Blumer

Moberg

D. P.

(2004). Treatment retention of dually diagnosed offenders in an institutional therapeutic community. Behavioral Sciences & the Law, 22, 585-597. doi:10.1002/bsl.602

33.

Webster

C. D.

Douglas

Eaves

Hart

(1997). HCR-20: Assessing risk for violence—Version 2. Simon Fraser University: British Columbia, Canada.

34.

Wierzbicki

Pekarik

(1993). A meta-analysis of psychotherapy dropout. Professional Psychology: Research and Practice, 24, 190-195. doi:10.1037/0735-7028.24.2.190