Are risk-need-responsivity principles golden? A meta-analysis of randomized controlled trials of community correction programs

Abstract

Objectives

Using meta-analysis to determine the effect size of the recidivism rate of participants in community correction programs that are conducted entirely in community settings.

Methods

Following the Preferred Reported Items for Systematic Reviews and Meta-analyses (PRISMA), 25 qualified studies contributed 35 independent effect sizes.

Results

Full participation in a program significantly reduced the recidivism rate. Participant age was a significant moderator of heterogeneity. Those aged over 18 have lower recidivism rates. Interventions that fully follow the Risk-Need-Responsivity (RNR) design principles achieved similar results to those that did not. Recidivism rates increase more than 12 months after the program ends.

Conclusions

The effectiveness of community correctional programs varies depending on the participant’s age. The RNR principles are not golden. The above factors should be carefully considered when conducting intervention design in the future. Results should be interpreted with caution due to the literature’s high heterogeneity and low quality.

Supplementary Information

The online version contains supplementary material available at 10.1007/s11292-022-09550-w.

Introduction

Community corrections are implemented in many countries worldwide. By the end of 2020, approximately 7 out of every ten people under correctional supervision in the USA were supervised in the community, while nearly 3 in 10 were incarcerated in various prisons (Rich Kluckow et al., 2022) . Since the introduction of community corrections in China in 2003, it has received a total of 4.78 million offenders, discharged a cumulative total of 4.11 million, and received more than 500,000 new yearly offenders (Bai, 2019).

No firm consensus exists on the definition of community corrections in the current works of literature, apart from the general aims of rehabilitating offenders and preventing recidivism. The differences are mainly reflected in the judicial systems, the competence of authorities, the applicable population, and its locations. Some countries, such as the USA, the UK, and France, have allowed private agencies to partner with the government for a portion of community corrections tasks. By contrast, countries such as Russia have yet to make such reforms (Byrne et al., 2019). No uniform results were found across countries on the type of offenders sentenced to community corrections. The US community corrections population includes probationers and parolees (Bureau of Justice Statistics, 2020), while the UK adds community sentences (GOV. UK, 2022), and China adds other offenders sentenced to control and provisional release from prison (Ministry of Justice of the People’s Republic of China, 2020). In Canada’s community corrections, offenders serve their sentences in both institutions and the community (Correctional Service Canada, 2019). However, in South Africa, community corrections are only in the community (Correctional Services Republic of South Africa, 2022). Despite these differences, community corrections can be generally recognized as a series of measures to prevent recidivism by placing offenders who meet certain conditions (e.g., probation or parole) in fixed community residences under the supervision of specialized agencies.

Numerous studies have demonstrated the significance of community corrections. These intersectional type studies have effectively reduced recidivism rates among community corrections offenders. For example, a sustained 20-session group cognitive behavioral intervention program effectively reduced recidivism rates among high- and medium-risk probationers in the USA with varying effects by ethnicity (Kosson et al., 2019). Intensive personal support for high-risk juvenile probationers in Germany for six consecutive months reduced recidivism rates with a more pronounced downward trend for lesser offenders (Engel et al., 2022). Recidivism rates could also be reduced by training Australian probation and parole officers to provide more scientific services to probationers (Schaefer & Little, 2020). Additionally, providing financial assistance can improve the living conditions of female probationers and reduce recidivism rates (Wilfong et al., 2021) Community-based, family-focused individual, and group treatment can effectively reduce recidivism rates among juvenile probationers while providing cost-effective alternative to residential care (Ryon et al., 2017). However, a few confounding factors also yield slightly different results. Hyatt and Barnes (2014) found that high-risk probationers are more likely to be re-incarcerated on intensive supervision probation conditions. Female probationers in England and Wales also did not have significantly lower recidivism rates than their control group counterparts after completing two cognitive skills programs (Palmer et al., 2015). High-risk juvenile probationers did not differ from controls in their risk of recidivism during the 3-year follow-up after participating in experimental case management (Ashford & Gallagher, 2019).

Risk-Need-Responsivity (RNR) principles were widely used in community corrections interventions (Schmucker & Lösel, 2015). Meta-analyses suggest that interventions designed following RNR principles generally achieve expected outcomes (Wormith & Zidenberg, 2018). The Risk principle suggests providing different services to inmates at different risk levels based on the results obtained from scientifically validated assessment tools; the Need principle suggests designing activities that address the needs of crime or factors directly related to changing criminal behavior; and the Responsivity principle means that activities should be tailored to individuals' learning styles and abilities and promptly adjusted to their physical, social, and psychological characteristics (Viglione, 2019).

Although the validity and practical utility of RNR principles have been widely demonstrated (Polaschek, 2012), the explanation of their effectivity was unclear due to the large variation between studies. Analytical studies of recidivism rates among probation offenders have shown that recidivism rates are lower for offenders who have participated in some programs designed based on RNR principles. However, due to high heterogeneity, no credible conclusions exist on the structure of what occurs other than that factors such as type and frequency of supervision, and offender characteristics may influence recidivism rates (Smith et al., 2018). A more detailed systematic review suggests that RNR-based rehabilitation effectively reduces recidivism rates, but considerable variations occurred in program type, implementation, and so on (Lipsey & Cullen, 2007). Some studies have also concluded that interventions designed based on RNR principles were not significantly more effective in reducing recidivism rates than controls (Seewald et al., 2018).

In addition, the RNR principle has faced some criticisms and controversies. For example, rigid application of the principles in practice can be detrimental to building good relationships with clients and undermine their motivation to participate in correctional programs (Raynor & Robinson, 2005). At the same time, completely following the RNR principles is not easy in practice. A survey of frontline community corrections staff showed that while most staff would use tools related to RNR principles to conduct surveys, their decisions were not always based on the results (Miller & Maloney, 2013).

Emphasizing community and family factors is essential in community corrections. Family members can provide care and supervision for community corrections offenders, which is an entirely different experience compared with visitation and prison item delivery (Phelps, 2020). Other elements, such as community attitudes toward delinquency, tolerance for misbehavior, support resources for at-risk individuals, and community structure, can impact recidivism in adult supervision (Byrne, 1989). Community corrections offenders rely on community resources such as food banks and employment agencies due to lower resource ownership (Wallace, 2015). The presence of organizations in and near the community may strengthen their connection to the community, reducing recidivism or increasing the opportunity for delinquency (Thompson-Dyck, 2021). Adolescents have higher recidivism due to a lower frequency of family care and the presence of gangs in the community (Lockwood & Harris, 2015), and parolees living in areas with a high number of returning offenders have increased odds of recidivism (Houser et al., 2018). Halfway houses, however, are a counterexample, as community-based sanctions or transition programs are inseparable from the necessary connection to community resources (Miller, 2014). People in the community has criticized the potential risk to public safety posed by these types of halfway houses divorced from community and family factors, which is affecting their success (Wong et al., 2018).

Therefore, carefully reviewing the effectiveness of community correction programs is needed, especially the portion of the program conducted in the family and community. Gender, age, implementation of RNR principles, and follow-up period should be considered potential influencing factors. Systematic reviews and meta-analyses with rigorous study designs and high-quality literature can be used to respond to these questions (Spector & Thompson, 1991). Furthermore, meta-analysis has been used to determine whether an intervention affects an outcome by calculating the magnitude of the effect size and identifying potential moderators by meta-regression and subgroup analysis (Leucht et al., 2009). In summary, this study aimed to gain insight into the effectiveness of community corrections interventions conducted throughout the family and community. The outcome variable was the recidivism of community corrections offenders assessed by re-arrest or reconviction. The meta-analysis method was used to determine the overall effectiveness of the program and to examine what factors influenced recidivism outcomes, providing a referable basis for improving the effectiveness of such community corrections programs.

Methods

Preferred Reported Items for Systematic Reviews and Meta-analyses (PRISMA) was used to conduct the following search and analysis (Moher et al., 2009).

Search strategy

Three widely recognizable digital journal article databases in China and overseas were searched: China National Knowledge Infrastructure (CNKI), Web of Science, and ProQuest (last updated on January 10, 2022). It included synonyms of three broad categories of search terms: “randomized controlled trial,” “community correction,” and “recidivism” (see Appendix for the entire search strategy). Additionally, the reference lists of the identified papers were searched again to locate additional articles.

Inclusion and exclusion criteria

Studies published in both Chinese and English in peer-reviewed journals were marked. We selected those published after January 1, 2007, as similar topics of systematic reviews had been published before (Lipsey & Cullen, 2007). To obtain higher quality and more credible evidence, the current study sought to support a more robust study design by including studies on levels 4–5 of the Maryland Scientific Methodology Scale (Sherman et al., 1997), which was used to assess the criminological literature.

A study would be included if it fully met the following criteria: (a) a randomized controlled trial or quasi-experimental study, which required at least one experimental and one control group. The control group must have received an alternative intervention or no intervention compared with the experimental group (i.e., no contact or standard probation); (b) the target group is under community corrections, such as probationers and parolees, who must be living in the community at the time of service; (c) the outcome variable should be recidivism from administrative record violations, offenses, or offenses following prior convictions or arrests; and (d) the study provided sufficient data to calculate an effect size.

A study would be likewise excluded if it meets one of the following criteria: (a) the participants in the experimental and control groups were not all community corrections offenders or at least one group remained incarcerated at the start of the service program; (b) treatment using drugs such as methadone, and (c) insufficient statistics after contacting the authors or if the full text was unavailable.

Figure 1 illustrates the flowchart of the selection procedure. Titles and abstracts were screened initially, and full-text screening was conducted on studies that appeared to meet the inclusion and exclusion criteria. Data extraction (e.g., author and year of publication, correctional setting, numbers of individuals in the experimental and control groups, gender, the average age of participants, intervention length and type, control type, and follow-up period for recidivism) was performed by one author and reviewed by a second senior author (see Table 1).

Fig. 1.

Flowchart of the study selection

Table 1.

Characteristics of the studies included in the meta-analysis

Author and year of publication	Setting	Experimental group/control group	% Male	Mean age/range	Types of intervention;	Type of control	Duration	Follow-up period
Anderson et al. (2021)	Female juvenile offenders	153 803	0%	15.94	Family-based intervention (RNR)	Not receive the intervention	Not reported	1 year
Antonio and Crossett (2017)	Offenders who were released to parole supervision	2115 2115	90.4%	34	Cognitive Life Skills (CLS) program (non-RNR)	Standard services	49 days	Not reported
Ashford and Gallagher (2019)	Older, high-risk juveniles	29 114	93.1%	16.7	Level I supervision and a team approach to case management (RNR)	Level I supervision	Not reported	6–30 months
Ayoub (2020)	Parolees released to Upper Manhattan	213 291	98%	29.4	The reentry court (non-RNR)	Traditional parole	Not reported	18 months
Barnes et al. (2010)	Offenders classified as low risk	800 759	66.5%	41	Low-intensity community supervision (RNR)	General supervision	1 year	1 year
Barnes et al. (2017)	High-risk probationer	457 447	80.5%	30.27	CBT program (RNR)	Standard, intensive probation	14 weeks	1 year
Burraston et al. (2012)	Juvenile offenders	28 31	87%	13–18	The class plus cell phone (CBT) (non-RNR)	Standard programming available to juveniles on probation	278 days	1 year
Hamilton et al. (2018)	Probationers and parolees	41 41	94.87%	33	A focused deterrence notification meeting (Non-RNR)	Not asked to attend the meeting	Not reported	17 months
Hatcher et al. (2008)	Male offenders with a community rehabilitation order	53 53	100%	27.42	The aggression replacement training program (non-RNR)	A community rehabilitation order	Not reported	10 months
Hatcher et al. (2012)	Male offenders sentenced to a community rehabilitation/Punishment Order	66 173	100%	26.85	Either the “Enhanced Thinking Skills” program or the “Think First” program (non-RNR)	A Community Rehabilitation Order but without the requirement to attend an offending behavior program	20 2-h sessions/22 2-h program sessions	Not reported
Hollin et al. (2008)	Male probationers	616 2789	100%	28.23	R & R/ETS/Think First (CBT) (non-RNR)	No treatment	38 2-h sessions/20 2-h sessions/22 2-h sessions	188 to 1,457 days
Lancaster et al. (2011)	Latino/a youth	120 120	45.8%	14.38	Psycho-educational counseling group with a life skills emphasis (non-RNR)	Subject to measures stipulated 1by the local adjudicating judge	7 group counseling sessions	3–24 months
Lowenkamp et al. (2009)	The individual on probation	121 96	71%	33.5	A cognitive behavioral therapy developed to integrate cognitive skills and cognitive restructuring modalities of offender treatment (non-RNR)	Probation during the same time period as the treatment case	2 sessions each week for a total of 22 sessions	6 months
Lussier et al. (2014)	High-risk sex offenders in Canada	31 82	100%	41.2	The CHROME program and was offered CHROME services (non-RNR)	Offenders who were not supervised by the CHROME program and were never offered or never received CHROME services	Nor reported	1 year
Manno et al. (2012)	First-time teenaged driving offenders	137 43	81.3%	16.2	The teen RIDE curriculum (non-RNR)	No exposure to this program	1 day	6 months
McGuire et al. (2008)	Adjudicated offenders under probation supervision	215 194	100%	26.87	R & R/ETS/Think First (CBT) (RNR)	Not allocated to the program	38 2-h sessions / 20 2-h sessions/ 22 2-h sessions	17 months
Mills et al. (2008)	Convicted drink drivers	1740 9667	91%	35	A standardized program that includes educational components and elements of group cognitive behavioral therapy (non-RNR)	A community control group of convicted drink drivers who received legal sanctions alone	Nine weekly sessions of 2-h duration/three weekly sessions of 6-h duration	2 years
Palmer et al. (2008)	Adult male offenders	1131 2778 2390	100%	27.63	R & R/ETS/Think First (CBT) (RNR)	No treatment	38 2-h sessions/20 2-h sessions/22 2-h program sessions	365 to 1229 days
Palmer et al. (2011)	Male offenders who had a history of substance use	41 178	100%	30.93	A structured cognitive behavioral approach (non-RNR)	No treatment	20 2-h sessions	1 year
Palmer et al. (2012)	Drink-driving offenders	144 231	100%	35.43	Offending behavior program (non-RNR)	No treatment	14 sessions of 2-h duration	1 year
Palmer et al. (2015)	Women offenders serving community sentences	281 520	0%	28.47	ETS/Think First (CBT) (non-RNR)	No treatment	20 2-h sessions/22 2-h program sessions	1 year
Pearson et al. (2011)	Offenders on community orders or licenses supervised within the UK National Probation Service	3819 2110	85.73%	29	A structured probation supervision program, based on “What Works” principles (RNR)	A “traditional” probation supervision	7 sessions	2 years
Quinn and Quinn (2015)	Defendants with multiple DWI offenses	286 200	84%	38.53	Cognitive behavioral therapy (CBT) program (non-RNR)	Standard services	16 weeks	3 years
Schaefer and Little (2020)	Supervising offenders in the community	376 308	79%	32.29	Environment corrections (RNR)	No treatment	Two days and seven group-based booster training sessions	6 months
Van Deinse et al. (2021)	Individuals with serious mental illnesses on probation	47 53	54%	35.95	Specialty mental health probation (non-RNR)	Regular probation	Not reported	6 months/12 months

Quality assessment of studies

The quality of the included studies was assessed separately by two authors using the “Quality Assessment Tool for Quantitative Studies” (Effective Public Health Practice Project, 1998), which shows acceptable reliability and validity and has been tested for assessing the quality of randomized and non-randomized studies (Thomas et al., 2004). It has a total of 21 questions covering six aspects: (1) selection bias, (2) study design, (3) confounders, (4) blinding, (5) data collection methods, and (6) withdrawals and dropouts. According to the instructions in the companion manual, each aspect was given a rating of either “strong,” “moderate,” or “weak” based on the study’s content. If a study had no “weak” ratings for any of the six aspects, it had a low risk of bias and an overall “strong” rating. Meanwhile, if a study was rated weak for any of the six aspects, it had an overall rating of “moderate.” Lastly, if a study was rated weak for two or more aspects, it had a high risk of bias and an overall “weak” rating.

Statistical analyses

The effect size for the adequate degree of community corrective measures to reduce recidivism was expressed as the odds ratio (OR), which is the ratio of the number of target events occurring to the number of events non-occurring in the experimental group divided by the ratio of the number of target events occurring to the number of events non-occurring in the control group. The smaller OR indicates a lower likelihood of recidivism in the experimental group. We used the random-effects model because it assigned a similar weight to studies with different sample sizes, and substantial heterogeneity was expected between them (e.g., intervention types and follow-up periods).

Heterogeneity was examined using a significance test of Cochran’s Q statistic (p-value), with the magnitude of statistical heterogeneity assessed using I² (Engels et al., 2000). The importance of I² was interpreted using the following indices: low (0–40%), moderate (30–60%), substantial (50–90%), and considerable (75–100%) (Higgins et al., 2003). It would then be analyzed with caution in the subgroup analysis. We conducted Egger’s test to check small sample bias more accurately (Egger et al., 1997). Heterogeneity refers to the differences between the included multiple studies with higher heterogeneity, indicating that the total effect size is unreliable. The sources of heterogeneity should be identified as they can be influenced by randomization factors and the differences between the characteristics of the subjects (Higgins et al., 2019).

Subgroup analysis is used to determine whether subgroup conditions significantly influence the presence of heterogeneity among studies’ outcomes. This process was carried out by observing whether the differences between the effect sizes of subgroups after combining their effect sizes were statistically significant.

These abovementioned statistics use Stata (version 16) (StataCorp, 2019).

Results

Characteristics of the studies

As shown in Table 1, a total of 25 studies were finally identified, reporting on a combined sample of 35,615 participants (i.e., participants in the experiment group = 12,690 and participants in the control group = 22,925). Two of the studies consisted of an all-female sample (Anderson et al., 2021; Palmer et al., 2015), while eight consisted of an all-male sample (Hatcher et al., 2008, 2012; Hollin et al., 2008; Lussier et al., 2014; McGuire et al., 2008; Palmer et al., 2008, 2011, 2012). Five of the studies examined a sample of adolescents (Anderson et al., 2021; Ashford & Gallagher, 2019; Burraston et al., 2012; Lancaster et al., 2011; Manno et al., 2012), while the rest used a sample of adults with an average age between 26.85 and 41.2 years old. Studies also employed quasi-experimental designs (n = 21) or randomized control trial designs (n = 4) with the experimental group, including various interventions that follow RNR principles (n = 8) and interventions that do not fully follow RNR principles (n = 17).

Quality assessment of studies

Eighty-four percent (n = 21) of the studies were rated “Weak” in quality, 16% (n = 4) were “Moderate,” whereas none were rated “Strong” (see Table 2). Twenty-four studies were ordered “Strong” on confounders and data collection methods because of their clear search strategy. However, 19 of them demonstrated a weak study design because they were quasi-experimental studies. Owing to the specificity of the target group, it also performs poorly in the aspect of blinding design. The quality assessment of studies in this research is consistent with those obtained in previous similar studies in the same field (Malik et al., 2021; Yoon et al., 2017).

Table 2.

Quality assessment results of the included studies

Author and year	Selection bias	Study design	Confounders	Blinding	Data collection methods	Withdrawals and dropouts	Global rating of methodological quality
Anderson et al., (2021)	Moderate	Weak	Strong	Weak	Strong	Moderate	Weak
Antonio and Crossett, (2017)	Moderate	Weak	Strong	Weak	Strong	Strong	Weak
Ashford and Gallagher, (2019)	Moderate	Weak	Strong	Weak	Strong	Strong	Weak
Ayoub, (2020)	Strong	Strong	Strong	Weak	Strong	Strong	Moderate
Barnes et al. (2010)	Strong	Strong	Strong	Weak	Strong	Strong	Moderate
Barnes et al. (2017)	Moderate	Weak	Strong	Weak	Strong	Moderate	Weak
Burraston et al. (2012)	Moderate	Weak	Strong	Weak	Strong	Strong	Weak
Hamilton et al. (2018)	Strong	Strong	Strong	Weak	Strong	Moderate	Moderate
Hatcher et al., (2008)	Moderate	Moderate	Strong	Weak	Strong	Weak	Weak
Hatcher et al., (2012)	Moderate	Moderate	Strong	Weak	Strong	Weak	Weak
Hollin et al. (2008)	Moderate	Weak	Strong	Weak	Strong	Weak	Weak
Lancaster et al. (2011)	Moderate	Weak	Strong	Weak	Strong	Strong	Weak
Lowenkamp et al., (2009)	Moderate	Weak	Strong	Weak	Strong	Moderate	Weak
Lussier et al., (2014)	Moderate	Weak	Strong	Weak	Strong	Strong	Weak
Manno et al., (2012)	Moderate	Weak	Strong	Weak	Strong	Moderate	Weak
McGuire et al. (2008)	Moderate	Weak	Strong	Weak	Strong	Weak	Weak
Mills et al., (2008)	Moderate	Weak	Strong	Weak	Strong	Weak	Weak
Palmer et al. (2008)	Moderate	Weak	Moderate	Weak	Strong	Weak	Weak
Palmer et al. (2011)	Moderate	Weak	Strong	Weak	Strong	Weak	Weak
Palmer et al. (2012)	Moderate	Weak	Strong	Weak	Strong	Weak	Weak
Palmer et al. (2015)	Moderate	Weak	Strong	Weak	Strong	Weak	Weak
Pearson et al., (2011)	Moderate	Weak	Strong	Weak	Strong	Moderate	Weak
Quinn and Quinn (2015)	Moderate	Weak	Strong	Weak	Strong	Strong	Weak
Schaefer and Little (2020)	Moderate	Weak	Strong	Weak	Strong	Strong	Weak
Van Deinse et al., (2021)	Strong	Strong	Strong	Weak	Strong	Strong	Moderate

General efficacy of interventions in community correction group

A meta-analysis of 25 studies with a total of 35 effect sizes showed that measures taken in the experimental group were moderately effective in reducing recidivism compared with the control group under general supervision (OR = 0.60, 95% C.I. = [0.39, 0.82], Z = 5.52, p < 0.001 as illustrated in Fig. 2). The report showed that the ratio of recidivism to the absence of any recidivism was 0.6 times in the experimental group than in the control group, and that the experimental group had a relatively lower recidivism rate and achieved a better recidivism prevention effect. Considerable heterogeneity was identified across studies (Q = 173.91, df = 34, p < 0.001, I² = 92.10%), which suggests that subgroup analysis is needed to explore its sources. Slight sample bias was found to be statistically significant (z = 2.57), suggesting that small samples of studies influence the overall effect size.

Fig. 2.

Forest plot of pooled effect sizes from the meta-analysis

Subgroup analysis with different characteristics

We performed four subgroup analyses to determine the source of heterogeneity (see Fig. 3). Apart from age, no significant differences existed between subgroup analyses in gender, type of intervention, or period of follow-up. This finding also led to the fact that we were not able to clearly explain the source of heterogeneity.

Fig. 3.

Forest plots for subgroup analysis

Specifically, all effect sizes were significant except for females in the gender subgroup. Compared with the RNR interventions (OR = 0.62, 95% C.I. = [0.33, 0.91], p < 0.001), the non-RNR intervention (OR = 0.57, 95% C.I. = [0.27, 0.87], p < 0.001) showed smaller effective size. RNR interventions refer to programs that simultaneously follow all three RNR principles, and non-RNR interventions refer to programs that fail to follow all three principles simultaneously. Follow-up periods of 12 months or below (OR = 0.53, 95%C.I. = [0.14, 0.93], p = 0.008) had smaller effect sizes than those at 12 months or above (OR = 0.68, 95% C.I. = [0.48, 0.88], p = 0.00). Effect sizes were also much larger in the 18 years old and below group (OR = 0.98, 95% C.I. = [0.45, 1.51], p = 0.00) than for those above 18 years of age (OR = 0.45, 95% C.I. = [0.27, 0.63], p = 0.00).

Discussion

Main findings

A total of 25 studies were included in this meta-analysis to investigate the effectiveness of community corrections programs in reducing recidivism rates over the past 15 years. This study produced three main findings: (a) complete participation in a community corrections program significantly reduces recidivism rates, (b) subgroup analysis revealed that participant age was a moderating variable for heterogeneity, and (c) the adult participants who completed a scientifically guided non-RNR program and had a follow-up period of 12 months or less had relatively lower recidivism rates. These findings may inform future programs designed specifically for offenders in community correctional programs.

We found that if community corrections offenders were able to complete an entire program, their recidivism rates were significantly reduced, thus confirming previous studies (Golden et al., 2006; Palmer et al., 2008). Notably, however, not all community corrections offenders complete the appropriate service program, and some programs even have high dropout rates (Palmer et al., 2015). Programs are designed to be most effective at full participation, and the positive effects are reduced when someone is dropping out (Hollin et al., 2008; McMurran & Theodosi, 2007). The recent COVID-19 outbreak has posed an unexpected numerical and formal challenge to the community corrections system. A few countries have therefore begun to reduce the size of their prison populations by transferring some prisoners to the community due to the high risk of transmissibility in prisons (Rapisarda & Byrne, 2020). Additionally, online interventions are replacing traditional face-to-face activities due to the mentioned high transmissibility risks (Braeuer et al., 2022; Fromberger et al., 2021). Whether the higher dropout rate of online interventions affects the overall effectiveness of the program is unclear.

Grouping by age should be considered when designing intervention programs. We divided the studies into two categories based on the included studies: the average age of participants above or below 18. Therein, the adolescent program is described in the study as well. The results were cross-checked with those of previous studies (Kuanliang et al., 2008; Redondo et al., 1999). Evidently, age is a significant moderator of heterogeneity. A previous study even suggested that adolescents lack decision-making capacities and are susceptible to external coercion, leading to a greater risk of delinquency than adults (Scott & Steinberg, 2008). The age-crime curve further illustrates a strong association between age and delinquency, with the latter increasing during adolescence, peaking in late adolescence, and declining rapidly afterwards (Shulman et al., 2013). Owing to the lack of sufficient data, this study did not further explore whether the group over 18 should be subdivided again. However, age, as a significant moderator of heterogeneity, indicates the need for further exploration in future studies.

In contrast to previous studies (Andrews & Bonta, 2010), we did not find that interventions implemented following RNR principles were more effective in reducing recidivism rates than non-RNR interventions, and the values of ORs for both were close. However, both have achieved a positive effect of reducing recidivism rate, which is worthy of affirmation. This finding may be because these non-RNR principles interventions also used methods and techniques that have been widely proven to be effective with community corrections offenders, such as cognitive behavioral therapy. At the same time, they are not entirely contrary to RNR principles but more often simply do not completely follow the risk principle and distinguish between high- and low-risk community corrections offenders, such as programs like Think First, which all help community corrections offenders learn new beliefs and behaviors through cognitive restructuring, social skills training, anger control, and empathy training (Landenberger & Lipsey, 2005). In contrast to previous studies (Andrews & Bonta, 2010), we did not find that interventions implemented following RNR principles were more effective in reducing recidivism rates than non-RNR interventions, and the values of ORs for both were close. This finding may be because these non-RNR principles interventions also used methods and techniques that have been widely proven to be effective with community corrections offenders, such as cognitive behavioral therapy. At the same time, they are not entirely contrary to RNR principles but more often simply do not completely follow the risk principle and distinguish between high- and low-risk community corrections offenders, such as programs like Think First, which all help community corrections offenders learn new beliefs and behaviors through cognitive restructuring, social skills training, anger control, and empathy training (Landenberger & Lipsey, 2005). In a word, although the effectiveness of the RNR principle has been widely proven, whether completely following the RNR principle is necessary to carry out community correction projects in practice should be decided according to the actual situation. Programs that follow some but not all of the RNR principles are also effective in reducing recidivism rates. In response to changing social needs, RNR principles are being refined (Taxman & Smith, 2021), and other framework guidelines for intervention programs for community corrections offenders are likewise being developed (Mackey et al., 2022; Strauss-Hughes et al., 2022). However, the effectiveness of these new frameworks remains to be tested in practice. In future community correction projects, whether completely following a certain principle is necessary still requires further empirical testing.

The results on follow-up duration suggest that sustaining the effects of the program may be difficult over time. The longest follow-up period in the included studies was 36 months, commonly categorized as 12 months and below or above. However, as the follow-up period increased, the differences between the experimental group and the control group decreased, suggesting that the effects of the intervention programs were not sustained over time (Ashford & Gallagher, 2019). These results suggest that establishing a follow-up period of more than 12 months is not effective in reducing recidivism rates among community corrections offenders Consideration should be given to restarting new projects after completion, as appropriate, to prevent recidivism. Family support is one of many factors that may reduce recidivism rates. Some scholars believe that the success of probation and cooperation of the offender’s family are closely related. A comprehensive probation framework suggests that parental support is an important factor in probation success (Schwalbe, 2012). One study showed that regular home visits to youth on probation significantly reduced their recidivism rates (Alarid & Rangel, 2018). Therefore, increasing the family support factor in the design of community correction programs is important.

Women in the community corrections population are also easily overlooked. The small sample size of the all-female program in this study does not allow comparing the effectiveness of male-only, female-only, or mixed-gender community correction programs in reducing recidivism rates. However, the differences between male and female community corrections offenders remain worthy of discussion. Several studies have noted that findings that hold true for men are not necessarily the same for women (Palmer et al., 2015). This finding is due to the fact that treatment, research, and rehabilitation are based on the lived experiences of men; the experiences of women are therefore often overlooked (Covington, 2018). Whether male and female offenders have the same criminogenic needs is also debated (Covington & Bloom, 2003). Thus, future studies should consider special activities for women when conducting their specific designs.

Limitations

Similar to previous research in the field of criminology (Beaudry et al., 2021; Malik et al., 2021), the specificity of community corrections offenders (in terms of factors such as offense type and risk level) makes rigorous randomized controlled trials difficult to conduct, and the effectiveness of activities is affected by staff levels. Avoiding these limitations is difficult in this paper and somewhat affects the credibility of the findings. Nonetheless, this problem is commonly associated with meta-analyses of crime-related topics (Wong & Bouchard, 2022). All studies included in this paper were conducted in developed countries, thus limiting the generalizability of the paper’s results by not foregoing community corrections in developing countries such as Asian countries. All studies were published in English, and analyzing languages other than English was difficult due to the authors’ language proficiency limitations. Moreover, the relatively small sample size of our included studies also meant that thoroughly exploring certain moderating factors, such as intervention duration, risk level, or experimental group size, was difficult in our meta-analysis. It has been suggested that some risk factors, such as poor peer groups, employment, marital status, and income, may impact offending among community corrections offenders. Hence, future high-quality studies targeting these moderating factors should be ongoing (Yukhnenko et al., 2020). Finally, measures targeting recidivism rates are also a posed limitation. Few meta-analyses have included outcomes other than recidivism to show changes in community corrections offenders in an integrated manner. Given that primary studies typically do not report other outcomes, uniformly measuring mental health is difficult. In practice, however, intervention programs for community corrections offenders should continue to be better evaluated as programs evolve over time (Pappas & Dent, 2021). Furthermore, reducing the frequency and severity of subsequent offenses should be considered a positive outcome (James et al., 2013).

Supplementary Information

Below is the link to the electronic supplementary material.

Biographies

Wenjie Duan Ph.D.

is a professor of social work at the East China University of Science and Technology. He obtained his Ph.D. degree at the City University of Hong Kong in 2013. Prof. Duan is mainly engaged in One Health and specializes in social and mental indicators and randomized control trial interventions. He has been appointed as an editor of international journals, such as Child & Family Social Work (2022–2024), Research on Social Work Practice (2018–2022), and Journal of Evidence-Based Social Work (2020–2022). He has published more than 70 papers in Social Science & Medicine, Journal of Happiness Studies, Children and Youth Services Review, and Social Indicator Research, among others.

Zichuan Wang

is currently studying for a master’s degree in the School of Sociology and Public Administration at the East China University of Science and Technology. He is engaged in evaluation-based behavioral and psychological research, including psychometric measurement and assessment, evidence-based correction, and more.

Caiyun Yang Ph.D.

is an associate professor of social work at Shanghai Normal University. She obtained her Ph.D. degree at Ecust China University of Science & Technology in 2015. Prof. Yang is mainly engaged in social work, community corrections and social policy. She has published more than 20 papers in Frontiers in Public Health, Child & Youth Care Forum, among others.

Shuting Ke

is currently pursuing her master degree in the College of Philosophy and Law & Political Science at Shanghai Normal University. She is engaged in social work, community corrections.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was partially supported by Shanghai Youth Project of Philosophy and Social Science (No. 2022ESH006): Research on Localization of Community Correctional Social Work in China in the Context of High-Quality Development.

Declarations

Conflict of interest

The authors declare no competing interests.