Fresh Start

Abstract

Objectives:

A meta-analysis was conducted in order to generate more understanding regarding the efficacy of aftercare programs in reducing the recidivism rates of juvenile offenders reentering their communities following a period of custody.

Method:

30 eligible primary studies were obtained through a systematic literature review and were coded. Recidivism was the outcome variable and 11 moderating variables were grouped according to either a sample, treatment, methodological, or study characteristic. A summary effect size was computed while moderator analyses and meta-regression were also conducted.

Results:

The summary effect size of aftercare programs was non-significant but subgroup univariate and multivariate analyses yielded significant treatment effects for samples of youth who averaged over 16.5 years of age and whose predominant index offense was violent. Well-implemented aftercare programs also yielded substantial treatment effects.

Conclusion:

The findings suggest that under specific conditions aftercare can reduce recidivism rates for youth involved in the juvenile justice system.

Keywords

aftercare youth meta-analysis recidivism program implementation

Introduction

It is estimated that within the United States more than 80,000 juveniles return to their communities following a period of custody each year, and evidence suggests that more than 50% of youth are rearrested within 3 years or less. Not surprisingly, these trends have prompted much interest among policy makers, program designers, and researchers in the area of youth reentering their communities, and one approach to the reentry process that is attracting more attention is aftercare (Barton, Jarjoura, & Rosay, 2012).

Aftercare consists of reintegrative services designed to prepare youth placed out of their homes for reentry into their communities. This is done by establishing collaborative arrangements with the community to ensure the delivery of services and supervision. Two factors distinguish aftercare from the traditional youth justice model. They are (a) when reentering their communities, young offenders in aftercare receive both services and supervision, whereas youth involved in the traditional juvenile justice system typically are supervised for a prescribed period and may or may not receive services; and (b) aftercare youth participate in intensive services while they are incarcerated, during their transition to the community, and while they are under supervision within the community. Thus, unlike the traditional youth justice model, aftercare includes the provision of services prior to and following the youths’ release back into the community (Gies, 2003).

A variety of aftercare programs have been created, including the Philadelphia Intensive Probation Aftercare Program, the Juvenile Aftercare in Maryland Drug Treatment Program, the Skillman Intensive Aftercare Project, and the Michigan Nokomis Challenge Program. Although these and other aftercare programs may vary in terms of origin and approach, they all share the aforementioned aftercare concept, that is, during incarceration there is a major programmatic focus on preparation for reentry and upon reentry there is a follow-up period that is characterized by both community supervision and the provision of services (Gies, 2003).

James, Stams, Asscher, De Roo, and van der Laan (2013) explain there has been a marked increase in the number of aftercare programs for juvenile offenders within the past two decades, and they also stress the importance of conducting research in order to ascertain whether “aftercare is effective for specific groups of juvenile … offenders, and what specific aspects of aftercare programs moderate their effectiveness” (p. 264). The knowledge base regarding the efficacy of aftercare programs is tenuous and replete of mixed results (Barton et al., 2012), though James et al. (2013) furthered understanding within this field when they detected a small treatment effect in their meta-analytic review of aftercare programs for both juvenile and young adult offenders. Moreover, their moderator analyses indicated that aftercare programs are particularly effective when they are well implemented and target ethnic minority groups as well as youth involved in gangs.

In this study, we also employ a meta-analytic approach in order to address the following two research questions. They are (1) what is the treatment effect of aftercare programs for young offenders and (2) how is the treatment effect of aftercare programs moderated by a variety of participant, treatment, methodological, and study characteristics? The selected participant characteristics were age, ethnicity, sex, and offender type, while the treatment characteristics were treatment intensity and implementation quality. The methodological characteristics were how recidivism was measured, time of follow-up, and study design. As for the study characteristics, they were type of publication and whether or not a study was funded by a granting agency. A more detailed explanation of these moderating variables is found subsequently.

As outlined subsequently, our study, including our selection of moderating variables, was informed by James et al. (2013), but it also differs from it in several ways. For one, our analysis of primary studies was based on an updated systematic review of the literature, and we conduct our analyses under the random-effects model, as opposed to the fixed-effect model utilized by James et al. (2013). Our review of the literature occurred during the Winter and Spring of 2013 and we included several studies (Abrams, Shannon, & Sangalang, 2008; Bottcher & Ezell, 2005; Greenwood & Turner, 1993; Iutcovich & Pratt, 1998; Pessin, 2008; Wells, Minor, Angel, & Stearman, 2006) that were not included in the review by James et al. (2013). It should also be noted that the only meta-analytic review of juvenile aftercare programs that we located was the one conducted by James et al. (2013).

Second, our eligibility criteria differed from James et al. (2013) in that we only included studies that involved youth committed to juvenile-oriented facilities, unlike James et al. (2013) who also included young adults incarcerated in adult facilities. Third, after we ascertained which moderating variables significantly influenced treatment effects, we conducted a maximum likelihood random-effects meta-regression with several influential variables. This enhanced the methodological rigor of the study as it allowed us to estimate the treatment effect of each moderator while simultaneously controlling the influence of other moderators as well as to estimate the variance within the outcome variable that could be explained by all of the moderators in the model (Borenstein, Hedges, Higgins, & Rothstein, 2009). Given this study’s aforementioned distinctiveness from James et al. (2013), it is evident that its purpose is to further contribute to the existing knowledge base regarding the efficacy of aftercare programs in reducing the recidivism rates of juvenile offenders who reenter their communities following a period of custody.

Method

Eligibility Criteria

Aftercare program: All of the primary studies included in our meta-analytic study consisted of evaluations of aftercare programs in which juvenile offenders were committed to a detention center or similar facility for a period of time and then released into the community following their period of commitment. Moreover, once youth were released aftercare consisted of both monitoring and supervision as well as various services (e.g., counseling, educational, etc.) intended to promote a successful and crime-free reentry into the community. This community-based aftercare component was a necessary condition for a primary study to be included in this meta-analytic study.

Control group: Only those primary studies on juvenile aftercare programs that included a control group were included in our meta-analytic study. Consequently, those studies that did not contain a control group and focused exclusively on youth who participated in aftercare were excluded from our analyses. There were no specific control group conditions, such as probation or supervisory services only, which warranted exclusion from our meta-analytic study.

Juvenile facility: Although there were no specific age or offense-related criteria, participants in both the treatment and the control groups had to be committed to a youth-oriented facility or detention center prior to their community reentry. Studies in which participants were incarcerated in adult prisons or jails were excluded.

Recidivism: All included studies had to report data pertaining to the number of youth in both the treatment and the control group who reoffended once their period of commitment had ended and they were residing in their communities.

Time frame: There was no exclusionary criterion based on time as relevant studies were included regardless of when they were conducted.

Location of studies: There was no exclusionary criterion based on the location of the studies. Despite our open-ended criterion regarding the location of studies all but one (Gray et al., 2005) of the included studies were U.S.-based studies.

Publication status: There was no exclusionary criterion based on the publication status of relevant studies. Thus, both published studies and those derived from the gray literature were included so long as they adhered to the eligibility criteria outlined earlier.

Primary Study Search

In order to locate the primary studies that met all of the eligibility criteria outlined earlier, one of the coauthors took primary responsibility for searching a wide range of electronic databases and Internet sources. These were PsychINFO; C2 SPECTR; ERIC; ProQuest Dissertations and Theses; ProQuest Social Services Abstracts; ProQuest Social Sciences; Cochrane Central Register of Controlled Trials; Cochrane Database of Systematic Reviews; ClinicalTrials.gov.; Library and Archives Canada; PubMed; PsychARTICLE; DocuBase; Google; Google Scholar; Canadian Newsstand Complete; Leddy Library Catalogue; Medline; and Proquest Historical Newspapers. To conduct this search, a series of key words were used in various combinations, including juvenile, youth, delinquent, offender, program, aftercare, aftercare program, aftercare programs, evaluation, recidivism, juvenile recidivism, youth, intensive aftercare, intensive aftercare program, offender, offenders, reintegration, transition, services, delinquent, justice, justice programs, and reentry. Please refer to Tables 1 and 2 that, respectively, outline the number of citations from each database and Internet source utilized in this study and the key word search combinations.

Table 1.

Number of Citations.

Citations From Databases
Database	Citation	Duplicate Removed	Abstract Removed	Included
Proquest Social Sciences	78	8	61	9
PsycINFOR	52	11	34	7
C2-SPECTR	282	86	195	1
Cochrane Central Registry of Controlled Trials	4	0	3	1
ERIC Online Database	36	0	32	4
ClinicalTrials.Gov	3		3	0
PubMed	118	8	109	1
PsycARTICLES	21	1	19	1
Proquest Dissertations and Thesis	122	0	115	7
DocuBase	1		1	0
Canadian Newstand Complete	1		1	0
Leddy Library Catalogue	0			0
Proquest Historical Newspapers	158		157	1
Proquestr CBCA Education	179	3	174	2
Medline	2		2	0
Library and Archives Canada	0			0
Total	1,057	117	906	34
Citations from Internet
Google	13,000 (first 10 results pages reviewed [100])	9	99	1
Google Scholar	31,900 (100 first 10 results pages were review [100])	12	74	4
Total	200	21	164	5

Table 2.

Key Words Used to Search the Literature.

Target Population		Juvenile OR Youth OR delinquent* OR Offender*
	AND
Intervention		Program* OR Aftercare Program* OR Probation OR Intensive Aftercare Program OR Aftercare Program OR Aftercare Programs* OR Aftercare OR Evaluation OR Intensive Aftercare OR Services
	AND
Outcome		Transition* OR Recidivism OR Reintegration OR reenty OR Juvenile Recidivism OR Reoffend* OR Justice OR Justice Program*

Another nine publications that met the eligibility criteria outlined earlier and were thus included in this meta-analytic study were located by hand searching the reference lists of various articles. Seven publications (Aos, 2004; Bergseth & McDonald, 2007; Cillo, 2001[doctoral dissertation]; Drake & Barnowski, 2006; Gray et al., 2005, Sealock, Gottfredson, & Gallagher, 1997; Unruh, Gau, & Waintrup, 2009) were found in James et al. (2013), one publication (Iutcovich & Pratt, 1998) was located in Bergseth (2010), and another publication (Josi & Sechrest, 1999) was detected in Kurlychek, Wheeler, Tinik, and Kempinen (2011). Please refer to Figure 1 for a flowchart that outlines the steps involved in our systematic review of the literature.

Figure 1.

Literature review flowchart.

Coding

Our coding of the variables, both the outcome variable and the moderators, was informed by James et al. (2013). Our study contained 1 outcome variable and 11 moderating variables. The authors developed a coding form to help guide the coding process and this form is available from the lead author upon request.

Outcome Variable

Recidivism

For recidivism, we combined both felonies and misdemeanors but excluded status and traffic offenses when they were listed. One study was based in the United Kingdom (Gray et al., 2005) and offenses were not divided into felony and misdemeanor categories. Recidivism was coded as a dichotomous variable (yes/no) and the number of per participant offenses was not taken into account. In one study (Cillo, 2001), however, the number of youth in the treatment and control groups who recidivated was not listed, only their respective sample sizes and the mean number of offenses per participant in each group, so raw recidivism frequencies were estimated for both groups by multiplying their sample sizes by the mean number of offenses per participant.

Moderating Variables

All 11 of the moderating variables were categorical and consisted of either 2 or 3 categories. In a manner analogous to James et al. (2013), the moderators were grouped according to their defining characteristic. These characteristics were sample, treatment, methodological, or study.

Sample Characteristics

The mean age of the participants, both in the treatment and in the control groups, was divided into three categories, that is, 16.5 years or less, more than 16.5 years, and not reported (NR). In terms of ethnicity, samples consisting of up to and including 70% ethnic minority youth were one category and those with more than 70% ethnic minority youth were another category. As for sex, samples were categorized as either mixed, which was less than or equal to 95% males, or male. This latter category pertained to samples consisting of more than 95% males. For the mixed studies, all but one (Drake & Barnowski, 2006) still consisted primarily of male participants. The final sample characteristic, offender type, referred to the predominant index offense among the sample, that is, nonviolent, violent, or NR.

Treatment Characteristics

There were two moderating variables that pertained specifically to the aftercare treatment. For one, we considered treatment intensity that was measured by computing in each study the difference between the treatment group and the control group in the monthly average number of face-to-face contacts between professional caseworkers and youth who had reentered their respective communities. The median (Mdn) difference was then computed (Mdn = 3.32 in favor of the treatment groups) and the treatment intensity for each study was categorized as at/below median, above median, or NR. The above median category indicated relatively greater treatment intensity.

Implementation quality of the treatment was measured by reviewing comments and ratings within the selected studies. The categories were well implemented, implemented with difficulty, and NR for those studies that did not rate implementation quality.

Methodological Characteristics

There were three moderators that pertained to the methodological characteristics of the selected primary studies. The first was how recidivism was established or measured. This variable had three categories. These were self-report, contact, which is described by Bergseth (2010) as subsequent contact with law enforcement authorities, such as a referral or rearrest, and adjudication. In studies in which both contact and adjudication were provided, we selected adjudication as it is a “more robust way of establishing whether a person committed a crime” (James, Stams, Asscher, De Roo, and van der Laan, 2013, p. 265). The second moderator was follow-up or the point in time in which a determination was made regarding the recidivism status of youth following their reentry into the community. For follow-up, studies were coded with one of the three categories, that is, below 12 months, 12–24 months, and above 24 months. Finally, the design of each study was coded as quasi-experimental, matched control group (MCG), or randomized clinical trial (RCT).

Study Characteristics

There were two moderating variables pertaining to study characteristics. Study type referred to whether or not the study was published in a peer-reviewed journal or whether it was in the form of a thesis/dissertation or a research report. We also recorded in a dichotomous fashion (yes/no) whether a study was funded by a granting agency. In one study (Rowland, 2007), the author acknowledged the contribution of a government agency but did not indicate whether this contribution pertained to funding, so this study was coded as NR.

Effect Size

We measured the overall treatment effect, or effect size, of the reviewed aftercare programs using the risk ratio (RR), in which a RR of 1.0 would mean that the risk of recidivism was the same for participants in both the treatment and the control groups. An RR of less than 1.0 indicates the recidivism risk is lower for treatment group participants and a RR greater than 1.0 implies the recidivism risk is lower for control group participants (Borenstein et al., 2009). As for our moderator analyses, this also involved computing RRs to ascertain the relationship between the selected moderators and the treatment effect of aftercare programs.

Random Effects

The bulk of meta-analytic studies utilize either the fixed-effect or the random-effects model when calculating treatment effects. The fixed-effect model assumes there is one true, or common, treatment effect size underlying all the primary studies within a meta-analytic study, while the random-effects model, though it has less statistical power, assumes that due to the diversity of both the participants and in the implementation of interventions there may be different effect sizes underlying various studies and therefore yields results that are more generalizable (Borenstein et al., 2009; James et al., 2013). Due to the variegated nature of the primary studies included in this meta-analytic study, we relied upon the random-effects model.

The random-effects model was used both in our RR computations of the overall treatment effect and in the moderator analyses. These computations were conducted with the software program Comprehensive Meta-Analysis (CMA) Version 2, and all of the primary studies were weighted by their respective sample sizes (Borenstein et al., 2009).

Test for Homogeneity

In order to test for homogeneity, or the null hypothesis that all of the primary studies in our meta-analysis share a common effect size, the Q statistic was computed. We also computed the I ² value to determine the extent to which the variance between the studies in terms of the effect size for the outcome variable, that is, recidivism, is due to nonrandom factors (Borenstein et al., 2009; James et al., 2013).

Publication Bias

A common concern among meta-analysts is that published studies exhibit a relatively higher rate of significant findings than the unpublished studies. If confirmed, this is known as publication bias. An assumption regarding publication bias is that when it is absent the findings from primary studies will be distributed symmetrically around the mean effect size, since the sampling error is random. Consequently, in this study we employed Duval and Tweedie’s Trim and Fill random-effects analysis to test for publication bias (Borenstein et al., 2009; James et al., 2013).

Multivariate Analysis

As noted earlier, in our study we also conducted a maximum likelihood random-effects meta-regression. This was done by utilizing meta-analysis macros for the Statistical Package for Social Sciences (SPSS; Wilson, 2002).

As delineated subsequently, four particularly influential moderating variables pertaining to sample, treatment, and methodological characteristics were represented in the meta-regression model, and we selected log odds ratio as the effect size metric. To conduct this analysis, we also selected suitable reference categories and developed dummy variables. This process is further explicated in the forthcoming results section.

Results

Description of Studies

Our sample consisted of 30 primary studies that met the eligibility criteria outlined earlier. This resulted into a total of 6,620 participants, with 3,114 being in treatment groups and 3,506 in control groups. The mean age of the participants was less than or equal to 16.5 years in 67% of the studies, and in nearly two thirds (63%) of the studies more than 95% of the participants were male. Furthermore, in 43% of our sampled studies, the predominant index offense among the participants was nonviolent. Nearly half (47%) of the studies reported implementation difficulties, while 43% of the studies employed an RCT design. Moreover, 47% of the studies were research reports, while 43% appeared as peer-reviewed journal articles. Please refer to Table 3 for the description of the primary studies.

Table 3.

Description of Primary Studies.

Primary Study Characteristics	k	Percentage
Sample
Age
≤16.5 yrs	20	67
>16.5	9	30
Not reported	1	3
Ethnicity
≤70%	14	47
>70%	16	53
Sex
Mixed	11	37
Male	19	63
Offender type
Nonviolent	13	43
Violent	8	27
Not reported	9	30
Treatment
Treatment intensity^a
≤Median	4	13
>Median	4	13
Not reported	22	73
Implementation^a
Difficulties	14	47
Well implemented	11	37
Not reported	5	17
Methodology
Recidivism established
Self	2	6
Contact	14	47
Adjudicated	14	47
Follow-up^a
<12 months	6	20
12–24 months	22	73
>24 months	2	6
Design^a
Quasi	10	33
MCG	7	23
RCT	13	43
Study
Study type
Thesis/Diss.	3	10
Report	14	47
Journal article	13	43
Study funded?^a
No	13	43
Yes	16	53
Not reported	1	3

Note. diss. = dissertation; MCG = matched control group; RCT = randomized clinical trial; k, number of primary studies the category appears in this meta-analysis.

^a Percentages do not add up to 100 due to rounding.

Test for Homogeneity

The aforementioned test for homogeneity was conducted and the following results were yielded: Q(29) = 101.37, p < .001, I ² = 71.326. From this, we may clearly reject the null hypothesis that all of the primary studies in our meta-analysis share a common effect size. Moreover, the I ² value indicates that approximately 71% of the variance between studies in terms of the effect size for the outcome variable (i.e., recidivism) are due to nonrandom factors. This substantial level of nonrandom variance clearly justifies our analyses of moderator effects within this study (Borenstein et al., 2009; James et al., 2013).

Effect Size

The summary treatment effect in our meta-analysis was in favor of the treatment group but at a nonsignificant level (RR = .931, p = .117). Please refer to Figure 2 for a forest plot that depicts the summary treatment effect of the aftercare programs.

Figure 2.

Summary treatment effect of aftercare programs.

Publication Bias

Results from the aforementioned Trim and Fill analysis yielded only three imputed studies and the adjusted point estimate was very close to the estimated treatment effect size described above, with RR = .948. Thus, there was no indication of publication bias in this study. Please refer to the funnel plot in Figure 3 in which a visual display of our Trim and Fill analysis is depicted, with the effect sizes being plotted on the x axis and the standard errors plotted on the y axis.

Figure 3.

Funnel plot of standard error by log risk ratio. The solid points represent imputed scores from the Trim and Fill analysis.

Tests for Heterogeneity

A small proportion of the tests for heterogeneity between the categories of the moderating variables yielded significant differences, as significant subcategorical differences were detected in offender type, QBetween(2) = 10.881, p = .004, and implementation quality, QBetween(2) = 13.494, p = .001, while age exhibited differences that approached statistical significance, QBetween(2) = 4.483, p = .106. Nevertheless, due to the fact that we conducted two-tailed tests with all of the moderating variables, we were able to detect meaningful effect sizes from our moderator analyses (Koehler, Lösel, Akoensi, & Humphreys, 2013).

Moderator Analyses

Below are the highlights of our moderator analyses. Please refer to Table 4 for the complete results of our moderator analyses.

Table 4.

Risk Ratios (RR) for Moderating Variables.

Moderator	k	RR	CI_95%		QBetween
Moderator	k	RR	Lower	Upper	QBetween
Sample
Age					0.218
≤16.5 yrs	20	0.985	0.876	1.108
>16.5 yrs	9	0.791*	0.663	0.944
Not reported	1	1.113	0.62	1.197
Ethnicity					1.426
≤70%	14	0.976	0.855	1.115
>70%	16	0.865	0.747	1.003
Sex					0.813
Mixed	11	0.978	0.842	1.136
Male	19	0.895	0.792	1.011
Offender type					10.881**
Nonviolent	13	1.001	0.892	1.123
Violent	8	0.666**	0.535	0.829
Not reported	9	0.978	0.838	1.142
Treatment
Intensity					0.219
≤Median	4	0.917	0.710	1.185
>Median	4	0.887	0.704	1.118
Not reported	22	0.941	0.845	1.049
Implementation					13.494**
Difficulties	14	1.105	0.981	1.244
Well	11	.816**	0.714	0.933
Not reported	5	.815*	0.671	0.989
Methodology
Recidivism Est.					3.321
Self	2	0.893	0.624	1.278
Contact	14	0.829*	0.706	0.972
Adjudicated	14	1.007	0.878	1.155
Follow-up					0.218
<12 months	6	0.901	0.697	1.167
12–24 months	22	0.912	0.798	1.043
>24 months	2	0.991	0.704	1.394
Design					2.735
Quasi	10	0.987	0.833	1.170
MCG	7	0.993	0.796	1.240
RCT	13	.825*	0.698	0.976
Study
Study type					3.431
Thesis/Diss.	3	0.887	0.648	1.214
Report	14	1.016	0.877	1.176
Journal article	13	.827*	0.703	0.974
Study funded?					0.687
No	13	0.891	0.778	1.021
Yes	16	0.957	0.823	1.112
Not reported	1	1.009	0.689	1.478

Note. diss. = dissertation; MCG = matched control group; RCT = randomized clinical trial. k, number of primary studies the category appears in this meta-analysis.

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

Sample Characteristics

Among the sample characteristics, age was a significant moderating variable, as studies in which the youth averaged over 16.5 years of age in both the treatment and control groups yielded treatment effects (RR =.791, p = .010). In other words, the treatment group youth were approximately 21% less likely to recidivate than their control group counterparts. When the sample of youth consisted of over 70% ethnic minorities, aftercare yielded a treatment effect at a near significant level (RR = .865 p =. 054). Likewise, studies with samples of more than 95% male exhibited treatment effects at a level that approached statistical significance, as RR = .895, p = .075. As for offender type, when the predominant index offense of the sample was violent, the treatment group was just over 33% less likely to recidivate (RR = .666, p < .001).

Treatment Characteristics

As for treatment intensity, no treatment effects were detected in the above median (RR =.887, p = .310), the at/below median (RR =. 917, p = .508), and the NR (RR = .941, p = .272) categories. Conversely, variation in implementation quality exhibited some noteworthy treatment effects. When aftercare programs were reportedly well implemented, treatment group youth were over 18% less likely to recidivate than their control group counterparts (RR = .816, p = .003). When difficulties with the aftercare program implementation were noted, however, youth in control groups were less likely to recidivate at a level that approached statistical significance, as RR = 1.105, p = .100.

Methodological Characteristics

When recidivism was measured through contact, aftercare exhibited a treatment effect at a significant level, as RR = .829, p = .021. When recidivism was measured through self-report, the effect size was also greater for the treatment group, albeit at a nonsignificant level (RR = .893, p = .536). Meanwhile, in studies that reported recidivism through adjudication, there was a slight treatment effect in favor of the control group but again, this was nonsignificant as RR = 1.007, p = .923. No significant treatment effects were yielded for the various follow-up categories, but variation in study design was associated with distinct effect sizes. For instance, studies utilizing the RCT design exhibited a substantial treatment effect (RR =.825, p = .025), whereas the quasi-experimental or MCG designs were not associated with treatment effects.

Study Characteristics

In terms of study characteristics, aftercare was associated with reduced recidivism rates at a significant level in journal articles (RR = .827, p = .023). Moreover, unfunded studies evinced a treatment effect at a level that approached statistical significance (RR = .891, p = .096).

Multivariate Analysis

For our meta-regression analyses, the moderating variables that were incorporated into the model as predictor variables were age, offender type, implementation quality, and study design. As discussed earlier, this translated into the selection of variables that, respectively, represented sample, treatment, and methodological characteristics.

For age, less than 16.5 years was the reference category and more than 16.5 years and NR were converted into dummy variables. In terms of offender type, nonviolent was the reference category and violent and NR were converted into dummy variables. For implementation quality, difficulty with implementation was the reference category and both well implemented and NR were converted into dummy variables while study design had quasi-experimental as the reference category, with MCG and RCT being converted into dummy variables. It should also be noted that none of the aforementioned predictors’ variance inflation factor levels had a value that exceeded four, which indicated that the standard assumption regarding multicollinearity was not violated (“Variance Inflation Factor,” n.d.).

Results of the meta-regression analyses indicate that the treatment group youth were less likely to recidivate when the samples’ predominant index offense was violent crime and their average age was more than 16.5 years, thus mirroring the results of our aforementioned moderator analyses. Programs that were reportedly well implemented yielded significant treatment effects, as did studies that employed an RCT design. Moreover, inspection of the meta-regression model’s R ² statistic indicates that the model accounted for approximately 68% of the variance within the outcome variable. Please refer to Table 5 for the results of the meta-regression.

Table 5.

Results of Maximum Likelihood Random-Effects Meta-Regression.

Moderator	B	SE	Sig	β
Constant	.7135	.1486	***	.0000
>16.5 years	−.4067	.1669	*	−.3194
Age NR	.6767	.5529		.1484
Violent offenders	−.5846	.2341	*	−.3132
Violence NR	−.3213	.2065		−.2628
Well implemented	−.6164	.1705	***	−.4905
Implemented NR	−.6190	.2281	**	−.3929
RCT	−.4464	.1700	**	−.2127
MCG	−.2363	.1964		−.1814
R ²	.6835

Note. diss. = dissertation; MCG = matched control group; NR = not reported; RCT = randomized clinical trial; SE = standard error.

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

Discussion

In response to Research Question #1, it should be noted that while the summary treatment effect reported in the results section earlier is diminutive and is not statistically significant, this finding is noteworthy since “evidence for the effectiveness of juvenile reentry programs remains scant” (Barton et al., 2012, p. 96). Furthermore, the fact that we employed a random-effects model when computing data gathered from an up-to-date systematic review of the literature enhances this study’s contribution to the topic’s nascent knowledge base.

As for Research Question #2, the findings from our univariate moderator and meta-regression analyses provide important practice implications. In terms of sample characteristics, recall that studies in which participants had a mean age of more than 16.5 years exhibited a robust treatment effect. As adolescents mature, there is a greater expectation they will exhibit the intellectual capacity to participate in and enjoy society’s activities of production, culture, and leisure so the educational, skill building, and therapeutic aspects of aftercare programs may be particularly effective for older youth (“Mentoring youth,” n.d., Steinberg, Chung, & Little, 2004).

As for younger adolescents, Latimer (2001) determined through a meta-analysis of 35 treatment programs for juvenile offenders that programs which targeted youth under the age of 15 were more successful at reducing recidivism when they included youths’ families in their interventions. This may highlight the need to increase family involvement in aftercare programs in order to decrease recidivism in younger adolescents. Overall, we suggest there be practical considerations of closely attuning programs to the age of their participants and to the dynamic factors associated with criminal risk at various age groupings (Van der Put et al., 2012).

The finding that studies yielded a positive treatment effect when the participants’ predominant index offense was violent also has notable practice implications. For instance, Hawkins et al. (1998) report that various antisocial behaviors such as stealing, destruction of property, smoking, drug selling, truancy, and dropping out of school as well as negative attitudes such as dishonesty and favorability toward aggressiveness are all linked to violent behavior. Thus, it is possible that aftercare programs are particularly effective in reducing the aforementioned problematic behaviors and attitudes in youth who are inclined toward violent criminal activities and this contributes to a reduced risk of recidivating. More research in this area is clearly warranted.

Recall also from our moderator analysis that a treatment effect for aftercare was detected at a near significant level when study samples consisted of over 70% ethnic minority youth. This finding, though it must be cautiously interpreted, suggests that the comprehensive nature of aftercare is perhaps more effective at addressing the unique circumstances faced by ethnic minority youth than are standard surveillance-oriented juvenile probation services. Hence, there is a need to further investigate whether specific aftercare programmatic factors cultivate an increased treatment effect for ethnic minority youth who may disproportionately face barriers linked to recidivating. These barriers include living in distressed neighborhoods characterized by an inordinate distance from jobs, exposure to violence and crime, and reliance upon low-quality public services (Galvez, 2010; Kubrin & Stewart, 2006).

In terms of treatment characteristics, our findings provide substantial evidence that well-implemented aftercare programs can markedly reduce the likelihood that participant youth will reoffend. Indicators of well-implemented programs include the presence of aftercare professionals who demonstrate their commitment to the well-being of the youth they are working with in a variety of ways, such as energetically engaging in supervisory and mentoring activities, providing referrals to service providers within the educational and occupational enhancement realms, and regularly participating in organizational activities (Bouffard & Bergseth, 2008; Wiebush, Wagner, McNulty, Wang, & Le, 2005).

Conversely, documented indicators of difficulties in program implementation include the level of contact between youth and professional staff during the aftercare program being less than initially anticipated, youth feeling abandoned by staff due to inadequate contact, staff turnover, and low levels of communication between relevant detention facility staff and community service providers (Abrams et al., 2008; Barton, Jarjoura, & Rosay, 2008; Sealock et al., 1997). Clearly, this finding regarding the treatment effect of well-implemented programs has important implications for program managers and practitioners, and we suggest more studies, including process evaluations, be conducted in order to generate greater insights into the substantive distinction between aftercare programs that are well implemented and those that are not. We also suggest that additional outcome studies be done in order to better understand what specific aspects of program implementation are linked to reduced recidivism rates and other dimensions of youth well-being.

Our finding that variation in treatment intensity has negligible effects on youths’ propensity to recidivate, while plagued by a small sample of eligible studies (n = 8), suggests that program planners and administrators should place primary emphasis on the quality rather than quantity of contacts between youth and professional staff. Although regular contact between professional aftercare workers and youth may be a necessary condition for an effective aftercare program, it is apparently not sufficient, and this reflects the need for comprehensive services that address the complex situations that youth face.

Some key findings also pertained to methodological characteristics. For instance, in both the univariate moderator and meta-regression analyses, the RCT study design was significantly associated with a treatment effect. Given that the RCT design is more likely to minimize selection bias and other threats to internal validity that may compromise a study’s capacity to exhibit real treatment effects (Singleton & Straits, 2010), this finding provides further evidence that aftercare programs can result in the reduced likelihood of recidivating among formerly incarcerated youth.

Although it was not incorporated into our meta-regression model, the univariate moderator analyses indicated a sizeable treatment effect when recidivism was measured through contact rather than self-report or adjudication. Blumstein and Larson (1971) observed that as measuring recidivism is taken “further into the system” (p. 125), that is, from arrests/contacts to adjudication to reincarceration, there is a decreased likelihood for errors of commission (i.e., false positives, or the erroneous counting of recidivism when no criminal activity occurred) while the likelihood of errors of omission (i.e., false negatives, or failure to detect reoffending behavior when criminal activity actually occurred) increases. This suggests that although the presence of false positives are more likely when contact is used as a recidivism measure, the heightened monitoring and service provision from justice professionals who characterize aftercare programs (Iutcovich & Pratt, 1998) may have decreased the likelihood of occurring of false positives. Consequently, this may have contributed to the reduced recidivism risk of youth in the treatment groups when contact was used to measure recidivism.

Limitations

Despite the fact that one of the merits of this meta-analytic study is that we conducted an updated systematic review of the literature, its sample size of 30 primary studies can hardly be considered ample, as the sixth edition of the Publication Manual of the American Psychological Association reports that meta-analytic studies with less than 50 are “relatively small” (American Psychological Association, 2010, p. 183). This is, of course, a shortcoming that we could not control and we anticipate that over time more primary studies on juvenile aftercare programs will be conducted and published as articles, reports, and so forth, and this will also allow for further meta-analytic studies that manifest ever-growing complexity and statistical power. This greater complexity could include the analyses of potentially significant moderating variables that were not included in this study, such as participant attrition, aftercare treatment design (e.g., individual therapy, group therapy, and a combination of the two), and the proportion of participants who reportedly involved in youth gangs (James et al., 2013). Such variables could also be incorporated into a meta-regression model in order to further explore their influence on the treatment effects of aftercare.

Recommendations for Future Studies

More aftercare studies are needed that provide multidimensional measures of recidivism, such as specific offenses committed and their frequency so as to develop insights into the nuanced effects of aftercare programs and various moderators. Also, researchers could incorporate several indicators of recidivism, such as contact, adjudication, and reincarceration, into singular studies when assessing the effects of aftercare. This would minimize the potential for studies to be reliant on a single recidivism measure that may be plagued with either the error of commission or the omission discussed earlier and instead allow for a comprehensive measure of recidivism and thus a more precise estimate of the treatment effect (Harris, Lockwood, Mengers & Stoodley, 2011).

We also see the need for more aftercare evaluative studies to include relevant measures other than recidivism as outcome variables. For instance, Cillo’s (2001) evaluation of an aftercare program incorporated youth well-being measures that captured important dimensions such as inter-/intrapersonal strength, school functioning, and affective strength. We suggest that more studies containing similar measures would substantially contribute to the knowledge base surrounding the efficacy of juvenile aftercare programs.

Summary

Steinberg, Chung, and Little (2004) wrote that human service professionals have traditionally “advertised two dismal findings about young offenders reentering the community from the justice system—that nothing works (i.e., youthful offenders cannot be rehabilitated) and that there are no success stories (i.e., delinquents are destined for failure)” (p. 21). The findings in this meta-analytic study clearly suggest otherwise. Although the summary treatment effect was very modest and nonsignificant, further subgroup univariate and multivariate analyses of various sample, treatment, methodological, and study characteristics provided solid evidence that well-implemented aftercare programs can substantially reduce the recidivism risk of juvenile offenders and that aftercare may be particularly effective with older youth whose criminal histories are more violent in nature. Moreover, the fact that studies which employed the RCT design were more apt to yield treatment effects further bolsters the likelihood that participants are benefitting from aftercare programs.

A key component of evidence-based practice is being cognizant of the contemporary research findings, which suggest that specific interventions are effective in improving the overall quality of life for social and human service clients (Yegidis, Weinbach, & Meyers, 2012). In this study, our findings indicate that aftercare programs can boost the quality of life for juvenile offenders by decreasing their chances of engaging in further criminal activities. We are hopeful that future studies are done which can further inform program developers, administrators, and practitioners the steps they should take to bolster the efficacy of aftercare programs in reducing the recidivism rates and increasing the well-being of youth who are involved in the juvenile justice system.

Footnotes

Authors’ Notes

This article is based on a paper presentation delivered by the first author at the Annual Conference of the Canadian Association of Social Work Education in June 2013 in Victoria, British Columbia, Canada.

Declaration of Conflicting Interests

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

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The research reported in this article was supported by Director’s research funds for faculty-student research projects, University of Windsor School of Social Work.

References

References marked with an “*” indicates studies included in the meta-analysis.

*Abrams

L. S.

Shannon

S. K.

Sangalang

(2008). Transition services for incarcerated youth: A mixed methods evaluation study. Children and Youth Services Review, 30, 522–535.

American Psychological Association. (2010). Publication manual of the American psychological association (6th ed.). Washington, DC: American Psychological Association.

*Aos

(2004). Washington state’s family integrated transitions program for juvenile offenders: Outcome evaluation and benefit-cost analysis. Olympia, WA: Washington State Institute for Public Policy.

Barton

W. H.

Jarjoura

G. R.

Rosay

A. B.

(2012). Applying a developmental lens to juvenile reentry and reintegration. Journal of Juvenile Justice, 1, 95–112.

*Barton

W. H.

Jajoura

G. R.

Rosay

A. B.

(2008). Evaluation of the boys and girls club of America, targeted re-entry initiative (Final Report). Unpublished manuscript.

Bergseth

K. J.

(2010). Youth characteristics, intervention, and recidivism: The case of aftercare for youth returning from placement (Unpublished doctoral dissertation). North Dakota State University, Fargo, ND.

*Bergseth

K. J.

McDonald

T. D.

(2007). Reentry services: An evaluation of a pilot project in Clay County, MN. Fargo, ND: North Dakota State University. Department of Criminal Justice and Political Science.

Blumstein

Larson

R. C.

(1971). Problems in modeling and measuring recidivism. Journal of Research in Crime and Delinquency, 8, 124–132.

10.

Borenstein

Hedges

L. V.

Higgins

J. P. T.

Rothstein

H. R.

(2009). Introduction to meta-analysis. Chichester, UK: Wiley.

11.

*Bottcher

Ezell

M. E.

(2005). Examining the effectiveness of boot camps: A randomized experiment with a long-term follow up. Journal of Research in Crime and Delinquency, 42, 309–332.

12.

*Bouffard

J. A.

Bergseth

K. J.

(2008). The impact of reentry services on juvenile offenders' recidivism. Youth Violence and Juvenile Justice, 6, 295–318.

13.

*Cillo

G. C.

(2001). Evaluation of a theory-based transitional aftercare program for court-adjudicated adolescents (Unpublished doctoral dissertation). Fordham University, New York, NY.

14.

*Drake

E. K.

Barnowski

(2006). Recidivism findings for the juvenile rehabilitation administration’s mentoring program: Final report. Olympia, WA: Washington State Institute for Public Policy.

15.

*Fagan

J. A.

(1990). Treatment and reintegration of violent juvenile offenders: Experimental results. Justice Quarterly, 7, 233–263.

16.

Galvez

M. M.

(2010). What do we know about housing choice voucher program location outcomes? A review of recent literature. Washington, DC: The Urban Institute.

17.

Gies

S. V.

(2003). Aftercare services. Juvenile Justice Bulletin. Washington, DC: U.S. Department of Justice, Office of Juvenile Justice and Delinquency Prevention.

18.

*Gray

Taylor

Roberts

Merrington

Fernandez

Moore

(2005). Intensive supervision and surveillance programme (The final report). Oxford, England: Youth Justice Board for England and Wales. Probation Studies Unit, Centre for Criminology, University of Oxford.

19.

*Greenwood

P. W.

Deschenes

Adams

J. L.

(1993). Chronic juvenile offenders: Final results from the skillman aftercare experiment. Santa Monica, CA: The RAND Corporation.

20.

*Greenwood

P. W.

Turner

(1993). Evaluation of the paint creek youth center: A residential program for serious delinquents. Criminology, 31, 263–279.

21.

Harris

P. W.

Lockwood

Mengers

Stoodley

B. H.

(2011). Measuring recidivism in juvenile corrections. Journal of Juvenile Justice, 1, 1–16.

22.

Hawkins

J. D.

Herrenkohl

Farrington

D. P.

Brewer

Catalano

R. F.

Harachi

T. W.

(1998). A review of predictors of youth violence. In Loeber

Farrington

D. P.

(Eds.), Serious & violent offenders: Risk factors and successful interventions (pp. 106–146). Thousand Oaks, CA: Sage.

23.

*Iutcovich

J. M.

Pratt

D. J.

(1998). A final report for: Assessment of aftercare services provided to delinquent youth. Erie, PA: Keystone University Research Corporation.

24.

James

Stams

G. J. J.M.

Asscher

J. J.

De Roo

A. K.

van der Laan

P. H.

(2013). Aftercare programs for reducing recidivism among juvenile and young adult offenders: A meta-analytic review. Clinical Psychology Review, 33, 263–274.

25.

*Josi

D. A.

Sechrest

D. K.

(1999). A pragmatic approach to parole aftercare: Evaluation of a community reintegration program for high-risk youthful offenders. Justice Quarterly, 16, 51–80.

26.

Koehler

J. A.

Lösel

Akoensi

T. D.

Humphreys

D. K.

(2013). A systematic review and meta-analysis on the effects of young offender treatment programs in Europe. Journal of Experimental Criminology, 9, 19–43.

27.

Kubrin

Stewart

E. A.

(2006). Predicting who reoffends: The neglected role of neighborhood context in recidivism studies. Criminology, 44, 165–197.

28.

Kurlychek

M. C.

Wheeler

A. P.

Tinik

L. A.

Kempinen

C. A.

(2011). How long after? A natural experiment assessing the impact of the length of aftercare service delivery on recidivism. Crime & Delinquency, 57, 778–800.

29.

Latimer

(2001). A meta-analytic examination of youth delinquency, family treatment, and recidivism. Canadian Journal of Criminology, 43, 237–253.

30.

Mentoring youth in reentry. (n.d.). Retrieved from http://msiy.edc.org/resources/MSIY%20Publications/MSIY_Brief_Mentoring%20Youth%20in%20Reentry.pdf

31.

*Pessin

S. R.

(2008). An examination of the effects of race on the assignment of aftercare services and the effects of aftercare services on recidivism for juvenile offenders (Unpublished master’s thesis). Georgetown University, Washington, DC.

32.

*Rowland

M. K.

(2007). Family-focused reintegration for youth on parole: Evaluation of a state-wide program. (Unpublished doctoral dissertation). Indiana University, IN.

33.

*Sealock

M. D.

Gottfredson

D. C.

Gallagher

C. A.

(1997). Drug treatment for juvenile offenders: Some good and bad news. Journal of Research in Crime and Delinquency, 34, 210–236.

34.

Singleton

R. A.

Straits

B. C.

(2010). Approaches to social research (5th ed.). New York, NY: Oxford University Press.

35.

*Sontheimer

Goodstein

(1993). An evaluation of juvenile intensive aftercare probation: Aftercare versus system response effects. Justice Quarterly, 10, 197–227.

36.

Steinberg

Chung

H. L.

Little

(2004). Reentry of young offenders from the justice system: A developmental perspective. Youth Violence and Juvenile Justice, 2, 21–38.

37.

*Unruh

D. K.

Gau

J. M.

Waintrup

M. G.

(2009). An exploration of factors reducing recidivism rates of formerly incarcerated youth with disabilities participating in a re-entry intervention. Journal of Child and Family Studies, 18, 284–293.

38.

Van der Put

C. E.

Stams

G. J. J. M.

Deković

Hoeve

Van der Laan

P. H.

Spanjaard

Barnoski

RP.

(2012). Changes in relative importance of dynamic risk factors for recidivism during adolescence. International Journal of Offender Therapy and Comparative Criminology, 56, 296–316.

39.

Variance Inflation Factor. (n.d.). Retrieved from http://how2stats.blogspot.ca/2011/09/variance-inflation-factor-vif.html

40.

*Wells

J. B.

Minor

K. I.

Angel

Stearman

K. D.

(2006). A quasi-experimental evaluation of a shock incarceration and aftercare program for juvenile offenders. Youth Violence and Juvenile Justice, 4, 219–233.

41.

*Wiebush

R. G.

Wagner

McNulty

Wang

T. J.

(2005). Implementation and outcome evaluation of the intensive aftercare program (Final report). Washington, DC: U.S. Department of Justice, Office of Justice Programs (95-JN-FX-0023).

42.

Wilson

D. B.

(2002). Meta-analysis macros for SAS, SPSS, and Stata. Retrieved June 18, 2013 from http://mason.gmu.edu/∼dwilsonb/ma.html

43.

Yegidis

B. L.

Weinbach

R. W.

Meyers

L. L.

(2012). Research methods for social workers (7th ed.). Upper Saddle River, NJ: Allyn & Bacon.