Outcome Evaluation of a Family-Based Jail Reentry Program for Substance Abusing Offenders

Abstract

Offender reentry programs have proliferated since the passage of the Second Chance Act in 2008. This study examines the effectiveness of one such jail-based reentry program for male inmates diagnosed with substance dependency and who have minor children, the Delaware County (OH) Jail Substance Abuse Treatment program. This program served 34 offenders and their families over 2 years and was based on the Community Reinforcement and Family Training model, a treatment modality for substance abuse involving both operant conditioning and family-based therapy for behavioral modification. Results from a quasi-experimental design indicate that program participants were significantly less likely to be rearrested within 1 year after release relative to a comparison group of similarly situated offenders and more likely to comply with child support orders following release. Findings also revealed the treatment group had significantly more days to failure for those who did recidivate.

Keywords

jail reentry substance abuse treatment family-based treatment CRAFT

Introduction

Reentry research has overwhelmingly focused on programs delivered in prisons rather than jails, primarily because they offer a broader range of rehabilitation modalities and an environment better suited for longer and more intensive treatment (Kellar & Wang, 2005). While prison-based research has addressed the effectiveness of rehabilitation programs (Andrews et al., 1990; Hiller, Knight, & Simpson, 1999; Knight, Simpson, Chatham, & Camacho, 1997), questions remain as to whether these findings can be translated to jail populations. Jails typically receive a substantially smaller proportion of funding for correctional programming, and time served is brief relative to prison sentences, resulting in jails offer limited programming options.

Despite these challenges, jails offer some advantages for reentry as they deal with many offenders earlier in their offending trajectory as well as serve a considerably larger number of inmates annually, relative to prisons. The characteristics that distinguish jails from prisons may offer a unique research setting wherein programming may be examined within the context of localized administrative and operational structure and the ability to address the needs of diverse sets of inmates within and across communities. Contrary to prisons which function according to structured regimens, jail operations vary widely according to the level of management and resource availability (H. V. Miller & Miller, 2010; White, Saunders, Fisher, & Mellow, 2012). A majority of jail inmates remain in local communities after release, which offers more robust outcome measures through extended follow-up periods, reduced attrition among treatment group participants, and the opportunity to identify latent or secondary benefits of programming (J. M. Miller, 2014).

Background

Jail-Based Reentry

As jail-based reentry initiatives have increased, so too has the extant literature related to program effectiveness (see, for example, Lurigio, Fallon, & Dincin, 2000; H. V. Miller & Miller, 2010, 2015; Osher, 2006, 2007; Osher, Steadman, & Barr, 2002; Soloman, Osborne, LoBuglio, Mellow, & Mukamal, 2008). Much like the larger reentry literature, prior jail-based research has found variable levels of effectiveness across studies and program approaches. Freudenberg, Wilets, Greene, and Richie (1998), for example, reported that women in the New York City (NYC) jail who participated in pre- and post-release substance abuse treatment had lower rates of offending compared with nonparticipants. In a more recent study of reentry programming at the NYC jail (Rikers Island), White and colleagues (2012) found that participants had no better outcomes than those in the comparison group. Their analysis did indicate, however, that those who completed at least 90 days of post-release services experienced significantly fewer returns to jail and a greater number of days in the community following release. Other research in the NYC jail system indicated that cognitive-behavioral therapy for inmates diagnosed with mental health disorders was able to significantly reduce recidivism compared with a matched comparison group (Glowa-Kollisch et al., 2014).

In addition to recidivism-related outcomes (e.g., rearrest, reincarceration, probation violation), prior research has also focused on mental health and drug using behavior as a means of assessing reentry program effectiveness. Gordon, Barnes, and VanBenschoten (2006), for example, examined a jail diversion program for mentally ill offenders and found that participants showed signs of mental health improvement as well as reductions in substance use and serious offending. Similarly, Lamberti et al. (2001) evaluated a program for justice-involved mentally ill patients in Monroe County, New York, and found the program to be effective in reducing recidivism and improving community adjustments among participants. Spjeldnes, Jung, Maguire, and Yamatani (2012) reported that other post-release factors such as family support was linked with lower levels of self-reported substance use among offenders released from jail, although this study did not use an experimental or quasi-experimental design. Prior work also suggests that jail reentry programs can significantly improve the likelihood of entering treatment, returning to treatment after release, and remain abstinent (Scott & Dennis, 2012).

Other research has utilized mixed-methods research designs to determine additional aspects of program effectiveness or benefits not accurately measured through strictly quantitative outcome designs. Bahr, Harris, Strobell, and Taylor (2012), using qualitative data drawn from a sample of Utah jail inmates, found that treatment group participants reported that programming elevated recognition of the consequences of their behavior and altered their perspective regarding drug use. Prior research has also shown that participation in programming was also associated with fewer inmate altercations, improved facility climate, and improved race relations (J. M. Miller, 2014). Studies incorporating qualitative methods offer the added advantages of (a) identifying the specific elements of programming perceived as most effective for participants, (b) contextualizing quantitative findings, (c) identifying latent benefits of programming, (d) assessing offender engagement, (e) specifying causal mechanisms of behavioral change, and (f) establishing program fidelity (J. M. Miller, 2014; J. M. Miller & Miller, 2015a, 2015b; H. V. Miller, Tillyer, & Miller, 2012).

Second Chance Act and Offender Programming

Offender reentry programs, particularly in jails, have proliferated since the passage of the Second Chance Act in 2008. This legislation authorized federal grants to support programs designed to assist offenders in the process of reentry; this law also required these programs to maintain and analyze performance measures toward the goal of establishing program effectiveness. Broadly, the Second Chance Act has several goals: expunging criminal records, providing services to those offenders most in need, enhancing public safety while reducing costs, and offering opportunities for the empirical study of reentry and rehabilitation—all toward improving criminal justice practice (Council of State Governments, 2013).

Since its inception, the Second Chance Grant Program has awarded more than 500 awards totaling approximately one quarter of a billion dollars. Tens of thousands of offenders have been served by these programs, most of whom are medium- to high-risk prisoners. This is significant because the greatest potential for overall recidivism reduction is found in interventions aimed at medium- to high-risk offenders. This contrasts with traditional approaches to correctional programming which have been long plagued by “cherry-picking” offenders during selection processes (see, for example, Barnes, Miller, Miller, & Gibson, 2008). The Second Chance Grant Program funds eight separate types of projects, one of which is the family-based substance abuse treatment program for incarcerated parents administered through the U.S. Bureau of Justice Assistance.

Community Reinforcement and Family Training (CRAFT)

This study evaluates a jail-based reentry program that utilized the CRAFT intervention (Meyers & Squires, 2001; Meyers, Villanueva, & Smith, 2005) to reduce recidivism and improve family cohesion. The community reinforcement approach (CRA) to substance abuse treatment is based on operant conditioning and aims to assist individuals in rearranging their lifestyle so that productive, drug-free living offers greater benefit than does the use and abuse of licit or illicit substances. First developed in the 1970s as an alcohol intervention, CRA emphasizes the importance of a social environment that enhances positive reinforcement for sober behavior (Azrin, 1976; Hunt & Azrin, 1973). Consistent with the principles of operant conditioning, CRA views punishment as an ineffective method for inducing behavioral change. Instead, CRA focuses on assisting the addict in identifying new enjoyable substance free activities. Individuals with substance abuse disorders engage in less pleasant activities than do those without such problems (e.g., Roozen et al., 2008), and non-substance-related rewards encourage people to reduce their substance use (e.g., Vuchinich & Tucker, 1996). Although CRA was developed as an alcohol intervention, it has since been applied in the treatment of other substance abuse disorders, including cocaine and opioid addiction (for a review, see Roozen et al., 2004). CRA may integrate cognitive-behavioral therapy with pharmacological interventions (Roozen et al., 2004) or include voucher-based incentive programs to promote abstinence (Budney & Higgins, 1998). Other variants include the Adolescent Community Reinforcement Approach (A-CRA) and CRAFT, the intervention used in this evaluation.

As noted above, the CRAFT model is a variant of CRA which involves friends and family members, or concerned significant others (CSOs), in the treatment intervention process. Substance use disorders profoundly affect the lives of both family members and those afflicted. CSOs experience a range of negative consequences involving damaged relationships, financial problems, interpersonal violence, anxiety, and depression (Fals-Stewart & Kennedy, 2005; Kirby, Dugosh, Benishek, & Harrington, 2005; Orford et al., 1992). CSOs often desire for substance users to seek treatment, although research suggests that most are resistant (Meyers, Miller, Hill, & Tonigan, 1999; Snow, Prochaska, & Rossi, 1992). Prior work also suggests, however, that once in treatment, the inclusion of CSOs shows promise in reducing drug abuse (Kaufman & Kaufman, 1992; O’Farrell, 1993). Family and social support also appear to play an important role in the management of substance abuse over time (Moos, Finney, & Cronkite, 1990; Spjeldnes, Jung, Maguire, & Yamatani, 2012). From an operant conditioning standpoint, then, evidence suggests that CSOs are capable of exerting significant influence over the behavior of substance using loved ones. The CRAFT approach uses the influence of CSOs to encourage substance users into treatment and assists them in maintaining sobriety (Meyers et al., 1999; Meyers, Roozen, & Smith, 2011).

Specifically, CRAFT teaches CSOs how to alter the home environment of the user to reward behaviors that promote sobriety and withhold rewards when the individual is using drugs. CRAFT does not pressure or demand individuals to seek treatment but uses the operant strategies described above. Therapists guide CSOs in learning specific skills and strategies including (a) raising user’s awareness of possible negative consequences of drug abuse, (b) strategies for preventing dangerous situations, (c) contingency management training, (d) social skills training, (e) planning activities to interfere with actual and potential drug use, (f) strategies to interfere with drug use, and (g) preparing to initiate treatment and provide support once the user is ready (Meyers et al., 1999).

Evidence for the CRAFT model has generally been positive, and the approach is considered effective relative to other intervention strategies (for reviews of CRAFT, see Meyers et al., 2011; Roozen, Blaauw, & Meyers, 2009; Roozen et al., 2004; Roozen, de Waart, & van der Kroft, 2010). Overall, empirical evidence suggests that the CRAFT model produces favorable outcomes more so than programs that use only one treatment component (Meyers et al., 1999; Szapocznik, Kurtines, Foote, Perez-Vidal, & Hervis, 1986, 1983). The modality has been shown as particularly effective for alcohol and opiate-addicted populations (Abbott, Weller, Delaney, & Moore, 1998), especially for enhancing an addict’s motivation to embrace rehabilitation and change (Meyers et al., 1999; W. R. Miller, Meyers, & Hiller-Sturmhofel, 1999).

Prior research has examined a number of outcomes including drug use (Meyers et al., 1999), treatment engagement (Roozen et al., 2010), and CSO well-being (Meyers et al., 1999). A meta-analysis conducted by Roozen and his colleagues (2010) indicated that CRAFT produced 3 times more patient engagement than Al-Anon approaches and twice the engagement of the Johnson Institute intervention (i.e., confrontational intervention by CSOs). Kirby, Marlowe, Festinger, Garvey, and LaMonaca (1999) also found CRAFT to be superior to a 12-step self-help group in inducing treatment entry of drug users and retaining CSOs in treatment. Meyers et al. (1999) reported that CSOs showed significant reductions in depression, anxiety, anger, and physical symptoms following participation in CRAFT. More recently, CRAFT has been combined with a buprenorphine detoxification program for opioid-addicted individuals (Brigham et al., 2014). Results indicated that the combined intervention exerted a significant effect on opioid and other drug use among program participants. Although many of these studies have been conducted in clinical settings, research has also shown CRAFT to be successful in community-based treatment settings (Dutcher et al., 2009). CRAFT has not been empirically examined, however, within the criminal justice system.

Delaware County Jail Substance Abuse Treatment (DCJSAT) Program Overview

The DCJSAT program targeted serious drug-abusing offenders who met the facility’s diagnostic criteria for alcohol and/or substance dependence and who had dependent children. Admission to the program was contingent upon two criteria: (a) diagnostic criteria for substance dependence and (b) voluntary participation in treatment. The modality addresses substance abuse through both operant conditioning and family-based therapy for behavioral modification (Meyers & Squires, 2001; Meyers et al., 2005). CRAFT is designed to eliminate positive reinforcement for drug use and enhance family-based positive reinforcement for sobriety (W. R. Miller et al., 1999). CRAFT relies on both in-patient cognitive-behavioral therapy delivered in individual and group settings, as well as family-assisted therapy during the transition from confinement into the community (i.e., reentry).

The DCJSAT program targeted 34 medium- to high-risk male substance abusing offenders who were housed in a unit dedicated to this initiative and physically separated from the general jail population. The program operated on a rolling admission basis so as to operate at full capacity and a Reentry/Recovery Coordinator maintained a waiting list for services to further ensure optimal utilization. The program was delivered across three interrelated phases (medical stabilization, primary treatment, and aftercare maintenance).

Once clients were medically cleared for assessment, a certified addiction specialist conducted a comprehensive medical and psychosocial screening assessment using the “Solutions for Ohio’s Quality Improvement and Compliance (SOQIC) Adult Diagnostic Assessment” developed by the Ohio Departments of Mental Health and Alcohol and Drug Addiction Services. This instrument objectively assesses criminogenic risk and needs through examination of a wide range of psychosocial functioning, including the offender’s living situation; social, educational, and employment information; mental health treatment history; substance abuse history; criminal record; and physical and sexual abuse history. Data gathered during these screenings were then utilized to customize individualized treatment plans (ITPs).

Phase II involved the development of ITPs, distinct treatment plans designed for each offender based on identified and anticipated needs. Phase II entails in-patient counseling and, consistent with the treatment protocol, was estimated at 90 days which aligns well for the targeted jail population’s approximate average length of incarceration. This phase was delivered over 20 hours per week and involved individual and group counseling, case management, and crisis intervention services. Over the course of the 90-day program, offenders received approximately 270 hours of cognitive-based intervention therapy, and the reentry accountability plan was modified as needed to assist offenders with post-release challenges. The development of the reentry plan specifically targeted offender criminogenic needs and guided services delivery in the community following release.

Near the end of the 90 days, the offender’s family met with counselors to receive education and training regarding their involvement during the post-release phase of programming. The CRAFT model is rooted in the belief that environmental contingencies can exert significant influence in encouraging or discouraging drug use. As a result, this approach relies heavily on family members to provide various forms of positive reinforcement which encourage clients to maintain sobriety. This model expects the family to serve as an integral element of the aftercare plan and assist the offender in embracing a sober lifestyle that is more rewarding than recreational drug use.

Following successful completion of Phase II, offenders were transitioned to aftercare maintenance upon release (Phase III). During this phase, relapse was monitored through the use of urinalysis testing. Clients received group counseling and also participated in family counseling sessions with those members designated as “concerned significant others” (see Meyers et al., 1999). Family members were taught the skills and strategies for altering a loved one’s behavior and motivating the individual to change. This third and final phase of the program typically lasted 60 days following release but was modified per individual needs.

Method

Sample

A three-step process was followed to draw a quasi-probability sample of offenders to be included in the program as either a participant or a comparison group member. First, every person entering the jail between April 1, 2012, and December 31, 2013, was given a brief survey that covered several life domains such as housing and employment status, whether the offender was married, and whether the offender owed child support. Upon receipt of the completed survey, a program manager identified whether the person had minor children, whether they indicated use of alcohol or an illegal substance, and whether they had a mental health concern. Once clients were medically cleared for assessment, a certified addiction specialist conducted a comprehensive medical and psychosocial screening and assessment using the “SOQIC Adult Diagnostic Assessment” developed by the Ohio Departments of Mental Health and Alcohol and Drug Addiction Services. This instrument objectively assesses criminogenic risk and needs and covers a wide range of psychosocial functioning including the offender’s living situation; social, educational, and employment information; mental health treatment history; substance abuse history; criminal record; and physical and sexual abuse history.

The second step to identifying potential program participants required each offender’s criminal history to be manually reviewed. Program eligibility was determined by a set of inclusionary criteria. To be eligible for the program, the offender must have been male, have children, be dually diagnosed with a drug use/abuse problem and a mental health concern, and no criminal history of sexual offending.

Finally, the third step of the process was to invite all eligible offenders to participate in the program. Each time an eligible offender was identified, a project staff member contacted the offender and his probation officer. Offenders who agreed to participate were placed into the treatment group. Offenders who did not agree to participate, those who were excluded based on receiving a sentencing length that did not align with the program, and those who could not participate for another reason (other than being deemed ineligible prior to this phase) were included in the comparison group. Most of the comparison group participants would have been placed in the treatment group had their sentence been long enough or had they chosen participation.

Measures

Responsibilities

All of the variables provided for the analysis can be “grouped” into four broad categories. The first category, labeled Responsibilities Variables, measured various outcomes related to the offender’s overall well-being and whether he was fulfilling his obligations such as paying child support prior to entering the jail.

Recidivism

The second group of variables measures whether a recidivism event occurred post-release. In other words, did the offender re-offend after graduation from the program? Four recidivism variables were collected and are presented for analysis: whether the offender recidivated with a probation violation (no = 0, yes = 1); whether the offender was charged with a new crime (no = 0, yes = 1); whether the offender was found to have recidivated with either a probation violation or a new crime (no = 0, yes = 1); and, if recidivism did occur, the time (in days) that lapsed in between the release date and the date of the recidivism event. The latter variable—time to recidivism—was calculated by subtracting the release date from the recidivism date.

Key independent/predictor variable

The key independent variable for the study was a dichotomous variable distinguishing between the program participants assigned to the comparison group (coded 0) and those assigned to the treatment group (coded 1).

Controls

A range of background variables were included to help rule out alternative explanations for any differences that may emerge between the comparison group and the treatment group. The background variables included measures of each offender’s criminal history (number of charges), whether the offender had ever been charged with a violent offense (no = 0, yes = 1), whether the offender had completed high school or earned an equivalent diploma (no = 0, yes = 1), whether the offender was married (no = 0, yes = 1), whether the offender was employed prior to incarceration (no = 0, yes = 1), the offender’s age in years, and whether the offender was Black (no = 0, yes = 1).

Analysis

The analysis unfolded in two interrelated parts: (a) a bivariate analysis and (b) a multiple regression analysis. The analysis began with bivariate estimation of the relationships among the key variables in the analysis. This step consisted primarily of cross-tabulations, chi-square analyses (χ²), and difference-in-means estimation. The multiple regression analyses were estimated next. Two methods were used during this phase of the analysis: logistic regression and survival analysis. Given that most of the outcome variables were coded dichotomously (e.g., Recidivism: no = 0, yes = 1), the logistic regression model formed the backbone of the multiple regression analyses. Briefly, logistic regression is one of the most commonly used generalized linear models (GLMs; Long, 1997) and can be expressed as

\ln (\frac{P_{i} [y_{i} = 1]}{1 - P_{i} [y_{i} = 1]}) = b_{0} + b_{1} (t r e a t m e n t_{i}) + \sum_{q = 1}^{Q} b_{q} (C_{q i}),

where b₀ is the constant and provides information about the average logged odds for cases with a zero on all of the x variables, b₁ is the regression parameter estimate of the impact of the treatment group indicator (coded 0 for nonparticipants and 1 for treatment group participants) on the logged odds of having a 1 on y (e.g., Recidivism), and $\sum_{q = 1}^{Q} b_{q} (C_{q i})$ represents the impact of the various control variables on the logged odds of having a 1 on y.

An important point to note about the logistic regression model is that the coefficients can be difficult to interpret. As a result, scholars often transform the results into probability values. In this way, the results from the logistic regression model can be made comparable with a simple calculation of the proportion of cases with a 1 on the dependent variable given some value on the x’s (i.e., P[y = 1 | x]). To transform the results of the logistic regression model into probability values, both sides of the equation must be exponentiated and converted using the following calculation:

P_{i} (y_{i} = 1 | t r e a t m e n t_{i}, c_{1 i}, \dots, c_{q i}) = \frac{e^{b_{0} + b_{1} (t r e a t m e n t_{i}) + \sum_{q = 1}^{Q} b_{q} (C_{q i})}}{1 + e^{b_{0} + b_{1} (t r e a t m e n t_{i}) + \sum_{q = 1}^{Q} b_{q} (C_{q i})}} .

Plugging in values for the independent variables into the above equation allows one to solve for the probability of recidivism (or some other outcome variable) for a treatment group participant after controlling for the influence of the other covariates.

The logistic regression model is the central estimation technique used in the analysis below. Yet, an important element of program evaluation and policy analysis concerns the timing of events. Specifically, most evaluators seek to answer two questions. The first is whether the program/policy appeared to have an effect (i.e., does the program work?). For the most part, the logistic regression analysis will provide answers to this question, with the noted limitation that causality cannot be fully addressed with quasi-experimental data. The second question considers whether the program affected offenders’ time to recidivism? This question recognizes that a perfect program—meaning one that reduces recidivism to zero—does not exist. Research has shown, however, programs that lengthen the time to recidivism for offenders may be more successful in the long term as compared with programs that do not affect recidivism timing (see Kurlychek, Brame, & Bushway, 2006). To address this question, the multiple regression technique known as survival analysis will be carried out. Briefly, survival analysis uses the variables on the “right-hand” side of the equation to predict timing to recidivism rather than a recidivism event. For those offenders who recidivated, their value for the dependent variable is the amount of time (in days) that had lapsed between their release date and the date of their new offense. For offenders who did not experience a recidivism event, their value for the dependent variable is the time between their release date and the project end date. Offenders who did not recidivate are considered “right censored.”

Results

Descriptive statistics for the focal program are provided in Table 1. A total of 66 offenders (all were male) provided data to the analysis with 34 offenders being enrolled in the program (i.e., the treatment group) and the remaining 32 being utilized as comparison cases (see Figure 1). Note that the case count for each of the variables is included in Table 1. These values reveal that certain cases were missing information on specific variables. The most important point gleaned from this portion of the analysis is that one offender—a member of the comparison group—remained incarcerated at the time of analysis. Listwise deletion was used for all missing data, meaning this case was included in the sample but was removed from all analyses of post-incarceration outcomes. As can be seen in Table 1, the available data were “grouped” into the four broad categories listed in the “Measures” section of this report. Most variables were coded dichotomously, meaning the value presented for the mean reflects the percentage of cases coded as 1 for that variable. For example, the mean for the any recidivism variable was .508, meaning that roughly 50% of the sample experienced a recidivism event.

Table 1.

Descriptive Statistics for DCJSAT Variables.

	n	M	SD	Minimum	Maximum
Responsibilities variables
Child support paid	26	0.46	—	0	1
Housing	66	0.91	—	0	1
Substance abuse program participation	33	0.70	—	0	1
Mental health service participation	33	0.36	—	0	1
Employment	34	0.38	—	0	1
Recidivism variables
Probation revocation recidivism	65	0.246	—	0	1
New charge recidivism	65	0.262	—	0	1
Any recidivism	65	0.508	—	0	1
Time to recidivism	65	174.65	144.91	1	566
Key independent variable
Treatment group indicator	66	0.515	—	0	1
Background variables
Criminal history	66	8.727	11.424	0	78
Violence	66	0.409	—	0	1
Education	66	0.773	—	0	1
Marital status	66	0.166	—	0	1
Employment prior	66	0.348	—	0	1
Age	66	30.909	7.003	19	53
Black	66	0.106	—	0	1

Note. DCJSAT = Delaware County Jail Substance Abuse Treatment.

Figure 1.

Mosaic chart of recidivism outcomes by DCJSAT treatment group indicator.

Bivariate chi-square analyses (χ²) are presented in Table 2. Due in part to the structure of data collection and the focus of certain variables, post-release information for comparison group members was only available on measures that could be tracked via automated systems available to Delaware County Sheriff Office (DCSO) staff. As shown in Table 2, a higher percentage of treatment group members (57.14%) reported paying their child support as compared with comparison group members (0%), and this difference was statistically significant, but the presence of a zero in the bottom left cell of the table prevents us from attaching much substantive meaning to the results of the statistical test (χ² = 5.31, p < .05). In addition, there did not appear to be a statistically significant difference across the two groups in terms of housing security (χ² = .61, p > .05).

Table 2.

Bivariate Analysis of DCJSAT Treatment Status on Responsibilities Variables.

Responsibilities variables	DCJSAT treatment status		χ²
Responsibilities variables	Comparison group	Treatment group	χ²
Child support paid
No	5 (100%)	9 (42.86%)	5.31
Yes	0 (0%)	12 (57.14%)	p = .021
Housing
No	2 (6.25%)	4 (11.76%)	.61
Yes	30 (93.75%)	30 (88.24%)	p = .436
Substance abuse program participation
No	—	10 (30.30%)	—
Yes	—	23 (69.70%)	—
Mental health service participation
No	—	21 (63.64%)	—
Yes	—	12 (36.36%)	—
Employment
No	—	21 (61.76%)	—
Yes	—	13 (38.24%)	—

Note. DCJSAT = Delaware County Jail Substance Abuse Treatment.

A similar set of analyses was carried out for the recidivism outcomes. As shown in Figure 1, when the any recidivism variable was analyzed, comparison group members were much more likely to have experienced a recidivism event (75%) compared with their counterparts in the treatment group (27.27%). This discrepancy suggests that the program was successful in reducing the probability of recidivism among treatment group participants (χ² = 14.81, p < .001).

Although the program appears to have had the desired impact on any recidivism, it is important to note that the previous analysis was bivariate, meaning that it did not account for other group differences that may be impacting the results. To rule out these effects, the multiple logistic regression model was estimated, and the results are presented in Tables 3 to 5. Three different sets of results are presented to represent the three different outcome variables: probation revocation recidivism, new charge recidivism, and any recidivism. The first set of results (Table 3) reveal the DCJSAT program did not appear to affect offenders’ likelihood of experiencing a probation revocation event. Indeed, the results in Model 1 and Model 2 suggest that there were no statistically significant (i.e., p > .05) differences in the odds of probation revocation between the treatment group members and the comparison group members. Each of the odds ratios are below 1.00, which suggests that members of the treatment group had lower odds of probation revocation compared with those in the comparison group. Thus, there is substantive evidence consistent with the program being successful, but those results did not reach the threshold of statistical significance set for this study (i.e., p > .05). Finally, Model 3—which omitted cases that did not successfully complete the program—reveals a moderately statistically significant effect of DCJSAT on probation revocation recidivism (odds ratio = .243, p < .10). The coefficient suggests the odds of probation revocation are approximately 75% lower (calculated as [1 − odds ratio] × 100) for treatment group members compared with offenders in the comparison group.

Table 3.

Logistic Regression of Probation Revocation Recidivism on DCJSAT Treatment Group Indicator and Control Variables.

	Model 1	Model 2	Model 3
Treatment group = 1	0.489	0.321	0.243*
Treatment group = 1	(0.289)	(0.232)	(0.206)
Criminal history		1.041	1.027
Criminal history		(0.0284)	(0.0264)
Violence = 1		0.632	0.786
Violence = 1		(0.438)	(0.576)
Education = 1		1.362	1.014
Education = 1		(1.109)	(0.852)
Marital status = 1		1.340	1.156
Marital status = 1		(1.135)	(1.013)
Employment prior = 1		3.971*	4.487*
Employment prior = 1		(3.071)	(3.989)
Age		0.963	0.982
Age		(0.0467)	(0.0508)
Black = 1		1.363	1.483
Black = 1		(1.319)	(1.463)
Constant	0.455**	0.607	0.449
Constant	(0.173)	(0.980)	(0.794)
Observations	65	65	56

Note. Standard errors in parentheses. Odds ratios are reported; Model 3 omits treatment group participants who did not complete the program successfully. DCJSAT = Delaware County Jail Substance Abuse Treatment.

p < .1. **p < .05. ***p < .01.

Table 4.

Logistic Regression of New Charge Recidivism on DCJSAT Treatment Group Indicator and Control Variables.

	Model 1	Model 2	Model 3
Treatment group = 1	0.129***	0.144**	0.0405***
Treatment group = 1	(0.0903)	(0.109)	(0.0494)
Criminal history		0.995	0.996
Criminal history		(0.0238)	(0.0248)
Violence = 1		0.822	0.952
Violence = 1		(0.591)	(0.808)
Education = 1		0.697	0.430
Education = 1		(0.526)	(0.383)
Marital status = 1		1.282	1.286
Marital status = 1		(1.133)	(1.260)
Employment prior = 1		0.650	1.460
Employment prior = 1		(0.564)	(1.606)
Age		1.013	1.016
Age		(0.0535)	(0.0639)
Black = 1		1.180	1.700
Black = 1		(1.225)	(1.935)
Constant	0.778	0.786	0.811
Constant	(0.277)	(1.428)	(1.796)
Observations	65	65	56

p < .1. **p < .05. ***p < .01.

Table 5.

Logistic Regression of Any Recidivism on DCJSAT Treatment Group Indicator and Control Variables.

	Model 1	Model 2	Model 3
Treatment group = 1	0.125***	0.0728***	0.0182***
Treatment group = 1	(0.0706)	(0.0548)	(0.0224)
Criminal history		1.081	1.086
Criminal history		(0.0528)	(0.0708)
Violence = 1		0.606	1.026
Violence = 1		(0.391)	(0.782)
Education = 1		0.689	0.363
Education = 1		(0.511)	(0.327)
Marital status = 1		1.657	1.765
Marital status = 1		(1.407)	(1.747)
Employment prior = 1		2.722	9.527*
Employment prior = 1		(2.056)	(11.62)
Age		0.979	0.996
Age		(0.0427)	(0.0499)
Black = 1		1.418	1.643
Black = 1		(1.412)	(1.752)
Constant	3.000***	4.190	3.061
Constant	(1.225)	(6.246)	(5.344)
Observations	65	65	56

p < .1. **p < .05. ***p < .01.

Table 4 provides the results from three multiple logistic regression models. This time, the new charge recidivism variable was analyzed as the dependent variable. The findings presented in the table reveal, quite clearly, that DCJSAT treatment group participants had substantially lower odds of experiencing new charge recidivism even as compared with the comparison group members. The findings presented in Model 2 suggest that the treatment group members had approximately 86% lower odds of receiving a new charge compared with the comparison group. These results are presented as predicted probabilities in Figure 2, where treatment group members had a far lower probability of receiving a new charge (approximately 9%) when compared with those in the comparison group (approximately 42%). These results suggest that the DCJSAT program was successful in reducing the likelihood that an offender would commit a new crime post-release. Interestingly, and consistent with expectation, treatment group members who successfully completed the program showed even lower odds of recidivism (see results in Model 3).

Figure 2.

Predicted probability of new charge recidivism by DCJSAT treatment group indicator.

The final set of multiple logistic regression results is presented in Table 5. The results presented in Table 5 reveal that treatment group members had statistically lower odds of experiencing any recidivism when compared with comparison group members. This finding is not surprising given that the previous sets of logistic regression results revealed that treatment group members had the lowest odds of receiving a probation revocation (although most results were not statistically significant) and the lowest odds of receiving a new charge post-release. Also consistent with the previous two analyses is that treatment group members who successfully completed the program had the lowest odds of any recidivism as compared with members of the comparison group (Model 3 results).

Thus far, the results have indicated the DCJSAT program was successful in reducing the likelihood of recidivism among treatment group participants. Whether DCJSAT affected the timing to recidivism has yet to be analyzed. Most of the offenders who experienced a recidivism event did so within the first 200 days post-release (i.e., the highest amount of density appears between 0 and 200). Figure 3 provides a box-and-whisker plot of time to recidivism across treatment group status. This plot suggests treatment group members had longer times to recidivism as compared with comparison group members.

Figure 3.

Box plot of recidivism timing by DCJSAT treatment group indicator.

These basic analyses suggest that DCJSAT may have been effective in lengthening time to recidivism. But, as with the chi-square analyses presented earlier, these figures do not account for the other background factors that may be driving the observed differences. To rule out the background influences, the Cox survival regression model was estimated, and the results are presented in Table 6. The results of the survival model are interpreted as the impact of each variable on the time to recidivism. Hazard rates are presented, so a value above 1.00 indicates the time to recidivism was reduced. Thus, if DCJSAT was successful in lengthening time to recidivism, we should anticipate seeing hazard rates that are below 1.00. Table 6 reveals this exact pattern of findings. Specifically, treatment group members, after accounting for background factors, had longer times to recidivism when compared with members of the comparison group. These results are presented graphically in Figure 4, which plots the probability of survival on the y-axis and the amount of time since release on the x-axis. Two lines are plotted, one for the treatment group (the lighter line at the top) and another for the comparison group (the darker line at the bottom). This plot shows the treatment group members experienced much higher probabilities of “survival” (i.e., not recidivating) as compared with the comparison group members, and this difference was fairly constant across the entire analysis period.

Table 6.

Cox Survival Model of Recidivism Timing on DCJSAT Treatment Group Indicator and Control Variables.

	Model 1	Model 2	Model 3
Treatment group = 1	0.302***	0.254***	0.157***
Treatment group = 1	(0.119)	(0.111)	(0.0873)
Criminal history		1.019*	1.016
Criminal history		(0.0109)	(0.0115)
Violence = 1		0.661	0.766
Violence = 1		(0.290)	(0.366)
Education = 1		0.937	0.631
Education = 1		(0.444)	(0.328)
Marital status = 1		2.097	2.048
Marital status = 1		(1.141)	(1.192)
Employment prior = 1		1.698	2.452
Employment prior = 1		(0.857)	(1.469)
Age		0.976	0.971
Age		(0.0306)	(0.0354)
Black = 1		1.646	1.924
Black = 1		(1.003)	(1.251)
Observations	65	65	56

Note. Standard errors in parentheses. Hazard rates are reported; Model 3 omits treatment group participants who did not complete the program successfully. DCJSAT = Delaware County Jail Substance Abuse Treatment.

p < .1. **p < .05. ***p < .01.

Figure 4.

Cox proportional hazards regression survival functions by DCJSAT treatment group indicator.

Discussion

Treating mental and substance abuse disorders is no longer a supplemental component of offender programming within the criminal justice system as evidenced by national-level funding initiatives such as Residential Substance Abuse Treatment, Justice and Mental Health, and, most recently, the Second Chance Act. The majority of programming prompted through these funding streams, and, consequently, evaluation research, are situated in prison rather than jail contexts. The logic for immediately implementing treatment in jails is compelling and supported by the sheer numbers of offenders that transition through jails as well as jail as the first or early system interaction for most offenders.

Unfortunately, general misunderstanding about the nature, population, and actual functions of jails has precluded the level of treatment compared with other junctures in the criminal justice system due to the brevity of jail stays and the traditional transitional dynamic of the jail role. The recent Vera report “Incarceration’s Front Door: The Misuse of Jails in America” (Subramanian, Delaney, Roberts, Fishman, & McGarry, 2015) indicates the need to rethink the role of jails from a mere processual system function to preventive and early intervention treatment opportunities that further multiple interrelated objectives and stakeholder interests. The Vera report reveals that although violent and property crime has dropped almost 50% over the last 20 years, the nation’s jail population has spiraled from 6 million in 1993 to 11.7 million in 2013. The report also relates that despite popular belief that jails serve as a holding function aligned with the legal processing of offenders, 60% of jail inmates are not criminally charged as the bulk of admissions are for behaviors associated with substance abuse and mental health disorders. With nearly three quarters of a million people across the nation in jail on an average day, jails contribute heavily to observations of the criminal justice system as the United States’s largest health care provider. To reverse the thematic misuse of jails, treatment initiatives, such as the DCJSAT program, must be implemented more broadly and intensely (i.e., feature evidence-based modalities delivered with fidelity). The evaluation role is central to normalizing jail-based treatment with optimal impact as successful replication is, ostensibly, dependent upon empirical specification of programming effectiveness.

Evaluation of the DCJSAT suggested that treated participants fared better post-release in terms of recidivism than did a matched comparison group. Analysis centered on four critical post-release outcomes: probation revocation, new charge recidivism, any recidivism (i.e., either a probation revocation or a new charge), and time to recidivism. Significant differences in the rate of recidivism were found between program participants (27.7%) and a comparison group (75%), while program participation significantly lowered the odds of both new charge recidivism and any recidivism. No significant differences were found, however, between the treatment and comparison groups for probation revocations. Participants were also more likely to comply with child support orders following release.

Replication of the CRAFT modality, based on our analyses, appears promising but should be considered cautiously and according to study limitations. While we infer that indicators of treatment success are indeed a function of treatment, additional fidelity assessment of the CRAFT modality is needed. This outcome evaluation was part of a larger two-phase mixed-methods evaluation design wherein process steps specified implementation intensity but inconsistent fidelity in terms of services delivery (see H. V. Miller & Miller, 2015). Also, the treatment group served by the DCJSAT was more homogeneous than most jail populations and additional evaluation with more participant diversity is needed. Program participation was also voluntary which may suggest a critical difference between the treatment and comparison groups in terms of commitment to change. Specifically, it is possible that the treatment group experienced more favorable outcomes because they were predisposed to behavioral change relative to nonparticipants. More fundamentally, this study represents the first evaluation of the CRAFT modality delivered in a criminal justice setting, generally, and in a jail, specifically, so additional studies are needed to observe effectiveness for evidence-based modality qualification.

Footnotes

Acknowledgements

The authors acknowledge former Delaware County (OH) Jail Director Joseph P. Lynch for initiating the researcher-practitioner partnership featured herein.

Authors’ Note

Points of view in this document are those of the authors and do not necessarily represent the official position or policies of the Delaware County (OH) Sheriff Office or the U.S. Department of Justice.

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: This project was sponsored by Grant 2011-RN-BX0004 awarded by the U.S. Bureau of Justice Assistance, Office of Justice Programs, U.S. Department of Justice.

Author Biographies

J. Mitchell Miller, PhD, is professor of criminology, University of North Florida, where he teaches and researches in the areas of drugs, crime, criminological theory, and program evaluation. He is currently conducting mixed-methods program evaluations of offender treatment programs for the U.S. Bureau of Justice Assistance in Louisiana and Tennessee. He is also serving as editor-in-chief of the Wiley-Blackwell Series of International Encyclopedias in Criminology & Criminal Justice.

Holly Ventura Miller, PhD, is associate professor of criminology, University of North Florida. As a former National Institute of Justice W.E.B. DuBois fellow and the immediate past president of the Southern Criminal Justice Association, her research interests include correctional policy, program evaluation, and immigration and crime. Her recent work has appeared in the Journal of Interpersonal Violence, Criminal Justice and Behavior, and Criminology & Public Policy.

J. C. Barnes, PhD, is associate professor in the School of Criminal Justice, University of Cincinnati. He is a biosocial criminologist whose research seeks to understand how genetic and environmental factors combine to affect criminological phenomena.

References

Abbott

P. J.

Weller

S. B.

Delaney

H. D.

Moore

B. A.

(1998). Community reinforcement approach in the treatment of opiate addicts. American Journal of Drug and Alcohol Abuse, 24, 17-30.

Andrews

D. A.

Zinger

Hoge

R. D.

Bonta

Gendreau

Cullen

F. T.

(1990). Does correctional treatment work? A clinically relevant and psychologically informed meta-analysis. Criminology, 28, 369-404.

Azrin

N. H.

(1976). Improvements in the community-reinforcement approach to alcoholism. Behaviour Research and Therapy, 14, 339-348.

Bahr

S. J.

Harris

P. E.

Strobell

J. H.

Taylor

B. M.

(2012). An evaluation of a short-term drug treatment for jail inmates. International Journal of Offender Therapy and Comparative Criminology, 57, 1275-1296.

Barnes

J. C.

Miller

J. M.

Miller

H. V.

Gibson

C. L.

(2008). Juvenile drug court program admission, demeanor and cherry-picking: A research note. American Journal of Criminal Justice, 33, 166-176.

Brigham

G. S.

Slesnick

Winhusen

T. M.

Lewis

D. F.

Guo

Somoza

(2014). A randomized pilot clinical trial to evaluate the efficacy of Community Reinforcement and Family Training for Treatment Retention (CRAFT-T) for improving outcomes completing opioid detoxification. Drug and Alcohol Dependence, 138, 240-243.

Budney

A. J.

Higgins

S. T.

(1998). A community reinforcement plus vouchers approach: Treating cocaine addiction (Pub. No. 98–4309). Rockville, MD: National Institute on Drug Abuse.

Council of State Governments. (2013). Second Chance Act. Washington, DC: Author. Retrieved from http://csgjusticecenter.org/nrrc/projects/second-chance-act/

Dutcher

L. W.

Anderson

Moore

Luna-Anderson

Meyers

R. J.

Delaney

H. D.

Smith

J. E.

(2009). Community Reinforcement and Family Training (CRAFT): An effectiveness study. Journal of Behavior Analysis in Health, Sports, Fitness, and Medicine, 2, 80-90.

10.

Fals-Stewart

Kennedy

(2005). Addressing intimate partner violence in substance abuse treatment. Journal of Substance Abuse Treatment, 29, 5-17.

11.

Freudenberg

Wilets

Greene

M. B.

Richie

B. E.

(1998). Linking women in jail to community services: Factors associated with rearrest and retention of drug-using women following release from jail. Journal of the American Medical Women’s Association, 53, 89-93.

12.

Glowa-Kollisch

Lim

Summers

Cohen

Selling

Venters

(2014). Beyond the bridge: Evaluating a novel mental health program in the New York City jail system. American Journal of Public Health, 104, 2212-2218.

13.

Gordon

J. A.

Barnes

VanBenschoten

(2006). The dual treatment track program: A descriptive assessment of a new “in-house” jail diversion program. Federal Probation, 27(2), 9-18.

14.

Hiller

M. L.

Knight

Simpson

D. D.

(1999). Prison-based substance abuse treatment, residential aftercare and recidivism. Addiction, 94, 833-842.

15.

Hunt

G. M.

Azrin

N. H.

(1973). A community-reinforcement approach to alcoholism. Behaviour Research and Therapy, 11, 91-104.

16.

Kaufman

(Eds.). (1992). Family therapy of drug and alcohol abuse (2nd ed.). Boston, MA: Allyn & Bacon.

17.

Kellar

Wang

(2005). Inmate assaults in Texas county jails. The Prison Journal, 85, 515-534.

18.

Kirby

K. C.

Dugosh

K. L.

Benishek

L. A.

Harrington

V. M.

(2005). The significant other checklist: Measuring the problems experienced by family members of drug users. Addictive Behaviors, 30, 29-47.

19.

Kirby

K. C.

Marlowe

D. B.

Festinger

D. S.

Garvey

K. A.

LaMonaca

(1999). Community reinforcement training for family and significant others of drug abusers: A unilateral intervention to increase treatment entry of drug users. Drug and Alcohol Dependence, 56, 85-96.

20.

Knight

Simpson

D. D.

Chatham

L. R.

Camacho

L. M.

(1997). An assessment of prison-based drug treatment: Texas’ in-prison therapeutic community program. Journal of Offender Rehabilitation, 24(3/4), 75-100.

21.

Kurlychek

M. C.

Brame

Bushway

S. D.

(2006). Scarlet letters and recidivism: Does an old criminal record predict future offending? Criminology & Public Policy, 5, 483-504.

22.

Lamberti

J. S.

Weisman

R. L.

Schwarzkopf

S. B.

Price

Ashton

R. M.

Trompeter

(2001). The mentally ill in jails and prisons: Towards an integrated model of prevention. Psychiatric Quarterly, 72, 63-77.

23.

Long

J. S.

(1997). Regression models for categorical and limited dependent variables. Thousand Oaks, CA: Sage.

24.

Lurigio

A. J.

Fallon

J. R.

Dincin

(2000). Helping the mentally ill in jails adjust to community life: A description of a post-release ACT program and its clients. International Journal of Offender Therapy and Comparative Criminology, 44, 532-548.

25.

Meyers

R. J.

Miller

W. R.

Hill

D. E.

Tonigan

J. S.

(1999). Community Reinforcement and Family Training (CRAFT): Engaging unmotivated drug users in treatment. Journal of Substance Abuse, 10, 291-308.

26.

Meyers

R. J.

Roozen

H. G.

Smith

J. E.

(2011). The community reinforcement approach: An update of the evidence. Alcohol Research & Health, 33, 380-388.

27.

Meyers

R. J.

Squires

D. D.

(2001). Community reinforcement approach: A guideline developed for the behavioral health recovery management project. Springfield: Illinois Department of Human Services.

28.

Meyers

R. J.

Villanueva

Smith

J. E.

(2005). The community reinforcement approach: History and new directions. Journal of Cognitive Psychotherapy, 19, 247-260.

29.

Miller

H. V.

Miller

J. M.

(2010). Community in-reach through jail reentry: Findings from a quasi-experimental design. Justice Quarterly, 27, 893-910.

30.

Miller

H. V.

Miller

J. M.

(2015). A promising jail reentry program revisited: Results from a quasi-experimental design. Criminal Justice Studies, 28, 211-225. doi:10.1080/1478601X.2014.1000489

31.

Miller

H. V.

Tillyer

R. T.

Miller

J. M.

(2012). Recognizing the need for prisoner input in correctional research: Observations from an in-prison DWI reduction program evaluation. The Prison Journal, 92, 274-289.

32.

Miller

J. M.

(2014). Identifying collateral effects of offender reentry programming through evaluative fieldwork. American Journal of Criminal Justice, 39, 41-58.

33.

Miller

J. M.

Miller

H. V.

(2015a). Rethinking program fidelity in criminal justice. Criminology & Public Policy, 14, 339-349.

34.

Miller

J. M.

Miller

H. V.

(2015b). Validating program fidelity: Lessons from the Delaware County Second Chance Initiatives. American Journal of Criminal Justice, 40.

35.

Miller

W. R.

Meyers

R. J.

Hiller-Sturmhofel

(1999). The community-reinforcement approach. Alcohol Research & Health, 23, 116-121.

36.

Moos

R. H.

Finney

J. W.

Cronkite

R. C.

(1990). Alcoholism treatment: Context, process, and outcome. New York, NY: Oxford University Press.

37.

O’Farrell

T. J.

(Ed.). (1993). Treating alcohol problems: Marital and family interventions. New York, NY: Guilford Press.

38.

Orford

Rigby

Miller

Tod

Bennett

Velleman

(1992). Ways of coping with excessive drug use in the family: A provisional typology based on the accounts of 50 close relatives. Journal of Community & Applied Social Psychology, 2, 163-183.

39.

Osher

F. C.

(2006). Integrating mental health and substance abuse services for justice-involved persons with co-occurring disorders. Baltimore, MD: National GAINS Center.

40.

Osher

F. C.

(2007, Jan/Feb). Short-term strategies to improve reentry of jail populations. American Jails, pp. 9-18.

41.

Osher

F. C.

Steadman

H. J.

Barr

(2002). A best practice approach to community reentry from jails for inmates with co-occurring disorders: The APIC model. Delmar, NY: The National GAINS Center.

42.

Roozen

H. G.

Blaauw

Meyers

R. J.

(2009). Advances in management of alcohol use disorders and intimate partner violence: Community reinforcement and family training. Psychiatry, Psychology and Law, 16, S74-S80.

43.

Roozen

H. G.

Boulogne

J. J.

Van Tulder

M. W.

Van Den Brink

DeJong

C. A. J.

Kerkhof

J. F. M.

(2004). A systematic review of the effectiveness of the community reinforcement approach in alcohol, cocaine, and opioid addiction. Drug and Alcohol Dependence, 74, 1-13.

44.

Roozen

H. G.

de Waart

van der Kroft

(2010). Community reinforcement and family training: An effective option to engage treatment-resistant substance-abusing individuals in treatment. Addiction, 105, 1729-1738.

45.

Roozen

H. G.

Wiersema

Strietman

Feij

J. A.

Lewisohn

P. M.

Meyers

R. J.

. . . Vingerhoets

A. J. J. M

. (2008). Development and psychometric evaluation of the pleasant activities list. American Journal on Addictions, 17, 422-435.

46.

Scott

C. K.

Dennis

M. L.

(2012). The first 90 days following release from jail: Findings from the Recovery Management Checkups for Women Offenders (RMCWO) experiment. Drug and Alcohol Dependence, 125, 110-118.

47.

Snow

M. G.

Prochaska

J. O.

Rossi

J. S.

(1992). Stages of change for smoking cessation among former problem drinkers: A cross-sectional analysis. Journal of Substance Abuse, 4, 107-116.

48.

Soloman

A. L.

Osborne

J. W. L.

LoBuglio

S. F.

Mellow

Mukamal

D. A.

(2008). Life after lockup: Improving reentry from jail to the community. Washington, DC: Urban Institute.

49.

Spjeldnes

Jung

Maguire

Yamatani

(2012). Positive family social support: Counteracting negative effects of mental illness and substance abuse to reduce jail ex-inmate recidivism rates. Journal of Human Behavior in the Social Environment, 22, 130-147.

50.

Subramanian

Delaney

Roberts

Fishman

McGarry

(2015). Incarceration’s front door: The misuse of jails in America. New York, NY: Vera Institute of Justice.

51.

Szapocznik

Kurtines

W. M.

Foote

Perez-Vidal

Hervis

(1986). Conjoint versus one-person family therapy: Further evidence for the effectiveness of conducting family therapy through one person with drug-abusing adolescents. Journal of Consulting and Clinical Psychology, 54, 395-397.

52.

Szapocznik

Kurtines

W. M.

Foote

F. H.

Perez-Vidal

Hervis

(1983). Conjoint versus one-person family therapy: Some evidence for the effectiveness of conducting family therapy through one person. Journal of Consulting and Clinical Psychology, 51, 889-899.

53.

Vuchinich

R. E.

Tucker

J. A.

(1996). Alcohol relapse, life events, and behavioral theories of choice: A prospective analysis. Experimental and Clinical Psychopharmacology, 4, 19-28.

54.

White

M. D.

Saunders

Fisher

Mellow

(2012). Exploring inmate reentry in a local jail setting: Implications for outreach, service use, and recidivism. Crime & Delinquency, 58, 124-146.