Best practices in reducing recidivism
Reporting on a systematic review of reviews (R004-2026)

Executive summary

Recidivism, defined as any recontact between justice-involved clients and the criminal justice system, is a global issue with widespread social, psychological, and economic consequences. To mitigate these consequences, an expansive body of research has been dedicated to identifying effective correctional interventions within institutions and community correctional settings. However, due to such a large repository of evaluations assessing numerous types of correctional interventions and employing different methodologies, there have been some inconsistent findings. Therefore, a systematic synthesis of the literature is required to integrate the findings of individual studies into a pooled metric to both quantify and qualify the effectiveness of these interventions. This is accomplished with a systematic review.

The current study consisted of a systematic review of individual systematic reviews (simply termed a review of reviews [RoR]) that investigated the effects of various intervention types on different forms of recidivism. The systematic reviews included in the current study typically presented results on the following:

• General recidivism: The commission of any new criminal offence following prior involvement with the justice system, irrespective of the nature or category of the offence. This category includes results explicitly classified under the general recidivism category, as well as those in which the type of recidivism was not specified. In instances where systematic review findings were specified to a particular offence subtype – such as sexual, violent, intra-partner/domestic violence, drug-related offences – those results were classified exclusively in the corresponding subtype categories below
• Sexual recidivism: The commission of a new sexual offence following prior justice-involvement, which include both contact offences (e.g., sexual assault) and non-contact offences (e.g., voyeurism)
• Violent recidivism: The commission of a new violent offence following prior justice-involvement, generally involving the use or threat of force against another person (e.g. assault, robbery, homicide)
• Intimate partner/domestic violent recidivism (IPV/DV): The commission of a new violent or sexual offence within a romantic or sexual relationship (i.e., intimate partner violence) or against a family/household member (i.e., domestic violence)
• Drug-related recidivism: The commission of a new offence involving possession or trafficking of drugs, impaired driving due to alcohol or other drugs, and violations of release conditions related to drug use

To this end, the authors of this study conducted a comprehensive and bilingual systematic search through several online databases, such as MEDLINE, CINAHL, AMED, PROSPERO, the Cochrane Database of Systematic Reviews, PsycINFO, and Criminal Justice Abstracts – adhering to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines. The search identified 141 reports with 513 individual intervention effects.

Overall, 60% of individual intervention effects indicated that correctional interventions were associated with a reduction in recidivism. In particular, interventions that incorporated aspects of cognitive theory and psychotherapy (i.e., cognitive behavioural therapy, cognitive skills training, other forms of psychotherapy) were found to be consistently effective for general recidivism, as well as sexual, IPV/DV, violent, and drug-related recidivism. These approaches may be particularly effective because, in their implementation, practitioners adopt a cognitive behavioural and social learning method and directly target pro-criminal attitudes and ways of thinking.

Beyond cognitive and psychotherapeutic interventions, other types of interventions were associated with reductions in each specific subtype of recidivism. Those that were associated with reductions of at least a moderate magnitude are listed below:

• General recidivism: Anger management, offence-specific programming, and pharmacology
• Sexual recidivism: Behavioural therapy
• Violent recidivism: Anger management, behavioural therapy, group therapy
• IPV/DV recidivism: Counselling, group therapy, psychoeducation

For drug-related recidivism, only "other" types of interventions (notably ignition interlock systems) yielded at least a moderate reduction in recidivism. However, this finding should be interpreted with caution as it could reflect a smaller body of literature specifically dedicated to drug-related recidivism, rather than the true reflection of overall intervention effectiveness.

In addition, the results of this study demonstrated that the interventions were associated with reductions in the five types of recidivism across various settings, including custodial settings, community settings, courts, and hospitals. Hence, to the extent that it does not compromise public safety (see risk principle), this research suggests that treatment could be delivered outside of a custodial setting and yield effective reductions in recidivism. Moreover, the current study found that interventions delivered in hospital settings were associated with reductions in general recidivism of at least a moderate magnitude, supporting the possibility that interventions delivered in treatment-focused settings may be particularly effective at reducing recidivism.

This study does have important methodological considerations and caveats. First, some primary studies were repeated across the sampled systematic reviews, which may have led to inflated or deflated results. Second, and most notably, the authors could not reliably evaluate the quality of interventions – specifically, adherence to principles of effective correctional intervention/treatment (i.e., Risk-Need-Responsivity [RNR]; Bonta & Andrews, 2024), or the fidelity with which an intervention was delivered. As a result, the findings reflect the overall effect of correctional interventions on recidivism, without consideration of intervention/ treatment quality.

Notably, the study still found that interventions were effective, suggesting that the observed effects may be conservative estimates. This implies that higher-quality interventions – those more closely aligned with evidence-based principles and delivered with greater fidelity – could potentially yield even greater reductions in recidivism. Therefore, it is recommended that future research reliably code for aspects of intervention quality to better access the impact of program characteristics when evaluating intervention effectiveness.

Third, due to the nature of the RoR, the authors were unable to compare the efficacy of different interventions types and settings, and therefore, no inferences can be made regarding the efficacy of one type or setting of intervention over another. Finally, the categorization of crimes was dependent on the specification of crimes from the authors of systematic reviews sampled in this RoR. Hence, the specific effect of each intervention and setting on specific crimes (e.g., import or export of illegal substances, impaired driving) is only inferred based on the efficacy of interventions on more general groups of crimes (e.g., drug-related recidivism).

Background and purpose

Criminal recidivism, or repeat offending, is a global issue with serious economic, policy, and public safety implications. Among federally sentenced individuals, approximately 35% were reconvicted within 3 years of being released into the community (Stewart et al., 2019). A recent study also found that among nearly 70,000 adults released from provincial custody or starting a community sentence across five Canadian jurisdictions (i.e., Ontario, Alberta, British Columbia, Saskatchewan, and Nova Scotia), 50% were reconvicted within 3 years (Pedneault et al., 2024)^{Footnote 1}. Comparable 2- to 3-year reconviction rates have been reported by other western countries, including Australia (53%), Denmark (63%), England (48%), Germany (46%), Netherlands (46%), New Zealand (60%), Sweden (61%), and the United States (60%) (Fazel & Wolf, 2015; Yukhnenko et al., 2020).

High recidivism rates pose a significant threat to public safety. Notably, this indicates a consistency in criminal behaviour that, despite the correctional system, may jeopardize the well-being of other Canadians. As a large proportion of crime is committed by a relatively small number of recidivists (Department of Justice Canada, 2019; Falk et al., 2014), identifying effective interventions which work to reduce recidivism is instrumental in increasing public safety.

Each criminal incident also yields a substantial economic burden. For example, the average cost estimate^{Footnote 2} per incident in 2014 ranged from $45,030 CAD for fraud to $5.9 million CAD for homicide (Gabor, 2015). Moreover, according to a cost of crime study based on a high-risk cohort of men in Ontario (Day et al., 2015), the average cost associated with a high-rate, chronic individual over a 15-year-period was estimated to range between $12 and $17 million CAD, compared to 3.5 million CAD for an individual who commits a single crime or engages in very low frequency criminal behaviour. Therefore, reducing recidivism could yield significant economic savings by reducing the number of costly criminal incidents.

Identifying effective interventions has been an important area of inquiry in correctional research for several decades. Beginning the 1990s, empirical research on the psychology of criminal conduct has culminated in an articulation of the principles of effective correctional intervention, operationalized through the principles of risk, need, and responsivity (RNR; Bonta & Andrews, 2004). The RNR principles aim to provide a framework in which justice-involved clients are matched with the treatment modality best suited to their specific needs. Simply put: the risk principle identifies "who" to treat (i.e., high risk cases), using empirically-based assessment tools; the need principle identifies "what" to treat, based on one's specific risk factors (i.e., criminogenic needs); and lastly, the responsivity principle is used to determine "how" to treat, suggesting that one's treatment should be tailored to their unique abilities (For more details on the RNR principles, please refer to Andrews & Bonta, 2010; Bonta & Andrews, 2024).

Beyond adherence to the aforementioned guiding principles, selecting an appropriate intervention is a complex process due to the sheer volume of available intervention types and the different settings in which such interventions may be delivered. There is a rich literature consisting of individual evaluations of correctional program effectiveness according to intervention type (e.g., cognitive behavioural therapy, counselling, case management, bootcamps, education/vocational programs; Beaudry et al., 2021; Bozick et al., 2018; Newton et al., 2018) and recidivism type (e.g., general, sexual, violent, intimate partner violence/domestic violence, and drug-related; Clarke et al., 2015; de Andrade et al., 2018; Zarling & Russell, 2022).

The most robust primary reports consist of randomized controlled trials, in which individuals are randomly assigned to either an intervention group, or a control group that does not receive the intervention (Blonigen et al., 2022; Eddy et al., 2022; Zarling & Russell, 2022). Recidivism is then compared between these groups, assuming that all factors outside of treatment are equivalent. Changes in criminogenic factors are measured at multiple time points (e.g., before treatment, during treatment, after treatment), allowing for inferences regarding the relationship between short- and long-term changes in criminogenic factors and recidivism (Brown et al., 2009; Serin et al., 2013).

The large number of individual studies conducted on the effectiveness of different types of correctional interventions, each with varying sample characteristics and parameters, can yield some contradictory and/or inconclusive results. In this scenario, it is advantageous to conduct a systematic review to integrate the findings of individual studies into a pooled aggregated metric reflecting the overall effectiveness of particular types of correctional interventions (Belur et al., 2020; Edwards et al., 2022). These reviews involve a detailed and comprehensive search strategy to identify, appraise, and synthesize findings from all relevant studies on a given topic (Uman, 2011). Systematic reviews may also consist of a meta-analysis, which is a quantitative statistical analysis that generates an aggregate summary measure to represent a particular effect (e.g., the effect of a correctional intervention). By minimizing risk of bias, systematic reviews have an advantage over typical narrative reviews, which are primarily descriptive pieces about a subset of studies on a given topic (e.g., best practices for reducing recidivism). To date, several systematic reviews have offered some indication of what interventions do and do not work, laying the foundation for widely adopted correctional models, including the principles of effective correctional interventions articulated above (Bonta & Andrews, 2024; Chadwick et al., 2015; Paparozzi & Gendreau, 2005).

After decades of research on principles and practices of effective correctional intervention, researchers and practitioners alike are faced not only with thousands of primary studies on the topic, but also with hundreds of systematic reviews on the topic that vary in scope, in methodology, and sometimes in their findings and conclusions (e.g., Bastian et al., 2010; Ioannidis, 2016; Smith et al., 2011). Accordingly, a particular method, termed a review of reviews (RoR), can synthesize these results in order to provide the most effective applied strategies for policy and practice. The strength of a review of reviews is its ability to synthesize results from the most rigorous reviews of scientific literature, and reduce the risk of bias by additionally incorporating research outside of scientific journals (Petrosino & Lavenberg, 2007; Petticrew & Roberts, 2006; Welsh & Rocque, 2014).

The RoR methodology has already been applied on the topic of correctional interventions, though such reviews are becoming increasingly outdated. As the most recent RoR on correctional interventions was published in 2018 (Barnett & Howard), an up-to-date review is necessary to account for subsequent research. Moreover, previous reviews have not necessarily taken full advantage of the strengths of the RoR method. For example, Welsh and Rocque (2014) published an RoR of 15 systematic reviews from the Campbell Collaboration's electronic library. While results found that nearly all interventions evaluated yielded primarily null effects on recidivism, a key limitation of this study was its reliance on a non-exhaustive search through a single repository of systematic reviews.

Similarly narrow in its scope, Barnett and Howard (2018) conducted an RoR that evaluated 21 systematic reviews from multiple databases of systematic reviews published between 2012 and 2016. This stringent inclusion criterion resulted in the analysis of a relatively small number of sources, precluding reliable conclusions. These limitations notwithstanding, results of this review did suggest that cognitive behavioural therapy, counselling, and education were promising avenues for correctional intervention.

The most comprehensive and rigorous RoR to date synthesized findings from a sample of 118 systematic reviews published up to 2015 (Weisburd et al., 2017). Despite substantial variability between individual systematic reviews in the reported effects of correctional interventions, the RoR found that psychotherapeutic interventions (e.g., cognitive behavioural therapy), treatment for substance use disorders, and specific deterrence practices (i.e., ignition interlocks to prevent driving while intoxicated) were associated with reductions in recidivism. Notably, Weisburd and colleagues only assessed the efficacy of interventions in the reduction of general recidivism, without considering effectiveness across recidivism subtypes (e.g., sexual recidivism, violent recidivism). Therefore, it is unclear whether these results reflect more specific reductions in a subtype of recidivism, or a more consistent reduction in all types of recidivism.

Purpose of the current study

The impetus behind the current RoR was to synthesize the results of this vast repository of research, and increase understanding of the efficacy of different types of correctional interventions. This RoR also assessed whether the interventions identified in the literature were differentially effective across recidivism subtype (i.e., general, sexual, violent, intimate partner violence/domestic violence, and drug-related). Contrary to previous RoR, the current study also consists of a more comprehensive systematic search for reviews across a several scholarly databases for interventions delivered in multiple settings internationally.

Method

Search strategy

An RoR was conducted that adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (Page et al., 2021). Systematic literature searches were conducted through seven global online databases to identify systematic reviews completed prior to March 2022 on the effectiveness of correctional interventions in reducing recidivism among adult justice-involved clients. Several online databases, including MEDLINE, CINAHL, AMED, PROSPERO, the Cochrane Database of Systematic Reviews, PsycINFO, and Criminal Justice Abstracts were searched using Boolean operators (*) and following search terms (alternative spellings were also included): (recidivism OR re-imprisonment OR reconviction OR repeat offending OR reoffend OR rearrest OR recontact OR recidiv* OR re imprisonment OR reconviction* OR repeated offending OR re offend OR rearrest* OR recontact* OR re-offend) AND (systematic review OR review literature OR qualitative systematic review OR evidence synthesis OR realist synthesis OR qualitative AND (synthesis OR meta) OR meta*). French searches were also conducted across the same databases using French translations of the search terminology: (récidive OR récidivisme OR réincarcération OR nouvelle* incarcération* OR nouvelle* condamnation* OR réinfraction OR nouvelle* infraction* OR nouvelle* arrestation* OR réarrestation) AND (revue systématique OR recension systématique OR méta-analy* OR revue systématique qualitative OR recension systématique qualitative OR synthèse de* données OR synthèse réaliste OR synthèse qualitati* OR méta-synthèse OR méta synthèse OR métasynthèse OR méta-ethnographie OR métaethnographie OR méta ethnographie OR méta-étude OR méta étude OR métaétude).

Inclusion/exclusion criteria

The inclusion/exclusion criteria for this review were as follows:

1. The study must have been a systematic review assessing the effect of one or more correctional interventions designed for adult justice-involved individuals (18 years or older)
   1. Studies that primarily or exclusively focused on juvenile populations were excluded
   2. Studies focusing solely on legal sanction, policies, or administrative measures were excluded (e.g., sentence length, solitary confinement, prisons visitation) as the focus of the current RoR was treatment-based interventions
2. The study must have direct measures of recidivism, using a broad range of possible metrics (e.g., contact with criminal justice system, breach of conditions, rearrest, reconviction, reincarceration)
   1. Studies relying solely on self-reported recidivism were excluded to reduce subjectivity in the included recidivism metrics.
3. The study must have featured a comparison group (i.e., the vast majority of an untreated group) to increase confidence that differences in outcomes are due to the intervention
   1. Reports that included a mix of studies with and without comparisons groups were included
   2. Reports that only included studies with a comparison group that underwent a different intervention and an untreated group were included^{Footnote 3}
   3. Reports that consisted only of single-sample pre/post designs were excluded
4. The study must have been available in English or French
5. Only studies published from 1990 onward were included, aligned with the formal introduction of the Risk-Need-Responsivity model in 1990 (i.e., Andrews et al., 1990)

Study screening

Two independent reviewers performed title and abstract screening. After title and abstract screening was completed, full text screening was conducted by four reviewers (two of which had also conducted the abstract screening). Any disagreements between reviewers were resolved through consensus meetings.

Data extraction

Two broad types of variables were coded for each report meeting the inclusion criteria (N = 141).

1. Aggregate variables operationalized the methodological factors of each report, including the country of origin of the report and the criteria of the Assessment of Methodological Quality of Systematic Reviews (Shea et al., 2017). These criteria included, but were not limited to, the publication status of the report; whether an inter-rater assessment, quality assessment, a comprehensive literature search from at least two databases, and a measurement of risk were conducted and reported; and whether sample characteristics were reported
2. Disaggregate variables^{Footnote 4} operationalized the distinctive features of the intervention effects. An intervention effect was operationalized as the difference in recidivism between a group exposed to an intervention and a group without exposure (i.e., control; the exact operationalization is outlined below). Intervention type, intervention setting, recidivism type (e.g., general, violent), and measure of recidivism (e.g., rearrest, reconviction) were coded for each effect. Multiple effects from the same systematic review were extracted as separate data points from the same report if they reflected different intervention types, settings, recidivism types, or recidivism measures. For systematic reviews without a meta-analysis, if two or more effects from the same systematic review reflected the same intervention types, settings, recidivism types, or recidivism measures, the effect with which the most primary studies were incorporated was extracted. For systematic reviews with a meta-analysis, if two or more effects from the same systematic review reflected the same intervention types, settings, recidivism types, or recidivism measures, the effect with the highest magnitude (based on the effect size or number of individual primary studies) was extracted from that systematic review. By considering largest effect sizes from both published and unpublished studies, this methodology avoids the risk of a publication bias. In addition, the largest effect size was considered whether this indicated a decrease or an increase in recidivism. Therefore, the risk of a selection bias was also mitigated. However, as explored below, the majority of effects indicated that correctional interventions were associated with a reduction in recidivism. Therefore, as the largest effect sizes were extracted, these estimates may provide a best-case efficacy scenario for each type of intervention, setting, and recidivism outcome.

Analysis

Eligible reports included both systematic reviews with and without a meta-analysis. A systematic review without a meta-analysis is a qualitative summary of evidence on a clearly articulated topic, grounded in a reproducible methodology and using critical methods for evaluating the quality of the studies that are included. By contrast, a systematic review with a meta-analysis consists of both a systematic review of the literature and a quantitative statistical analysis to aggregate the results into a single pooled summary estimate of effect. From eligible reports (N = 141, see Figure 1), a total of 513 effects on recidivism were extracted, with 138 (26.9%) effects from systematic reviews without a meta-analysis and 375 (73.1%) from systematic reviews with a meta-analysis.

The current study assessed the effect of interventions compared to a control condition without an intervention, which is commonly termed "treatment as usual." All control conditions assessed consisted of a sentence without an intervention (e.g., incarceration or probation without any treatment – although a small number of effects reflected differences between an intervention group and a comparison group that underwent a different intervention and an untreated control group). To determine whether interventions were associated with a reduction, increase, or unchanged recidivism, the current RoR synthesized the difference in recidivism rates between the group of justice-involved individuals receiving the intervention(s) and the control group. If recidivism rates were lower in the intervention group, then the reduction in recidivism was attributed to the intervention.

The current study is the broadest RoR to date on the topic of recidivism. To synthesize the results of the unprecedented number of systematic reviews and their effects, the authors of the current RoR developed a novel method to quantify the results. For systematic reviews without a meta-analysis, the authors calculated the number and percentage of effects that demonstrated either a reduction, increase, no effect, or mixed effect of a particular intervention, and classified the intervention accordingly. Specifically, if over 50% of effects associated with a given intervention type demonstrated a particular direction of change in recidivism (i.e. reduction, increase, or no change), the overall effectiveness was specified as such. When no single direction accounted for more than 50% of the effects, the intervention was classified as having mixed outcome. For systematic reviews with a meta-analysis, the average (pooled) effect size and associated statistical significance test revealed both the direction and magnitude of the effect of a given intervention on recidivism. A nonsignificant pooled effect size was indicative of unchanged recidivism. To determine which interventions yielded a moderate or greater effect on recidivism, effects from systematic reviews with a meta-analysis were separately classified as small or at least moderate in magnitude based on whether the effect size coefficient (e.g., odds ratios, Cohen's d, Pearson's r) surpassed a relative threshold (according to guidelines in Rosenthal, 1996)^{Footnote 5}.

Figure 1: Image description

Flowchart titled "Identification of studies via databases and registers" showing the study selection process through the stages of identification, screening, and inclusion.

Identification of studies via databases and registers

Identification

A box on the left states:

Total records identified: 2069

• Cochrane Database of Systematic Reviews (CDSR): 85
• PROSPERO: 178
• MEDLINE: 245
• Criminal Justice Abstracts: 475
• PsycINFO: 1086

A connected box on the right states:

Records removed before screening: 691

• Duplicate records removed: 647
• Records published prior to 1990: 44

An arrow leads downward to:

Screening

Abstracts screened: 1378

A connected box on the right states:

Records excluded: 1073

• Juvenile population: 519
• No recidivism measure: 158
• Not a systematic review: 257
• No intervention: 138
• Other: 1

An arrow leads downward to:

Full texts sought for retrieval: 305

A connected box on the right states:

Reports not retrieved: 22

An arrow leads downward to:

Full texts assessed for eligibility: 284

• Full texts captured in English search: 283
• English full texts captured in French search: 1

A connected box on the right states:

Reports excluded: 138

• Juvenile population: 15
• Not a systematic review: 72
• No recidivism measure: 14
• No comparison group: 15
• No intervention: 18
• Other: 4

An arrow leads downward to the final box:

Included

Reports included in review: 146

• Data extracted: 141
• Insufficient information for data extraction: 5

The flow progresses from 2,069 records initially identified across five databases and registers, to 1,378 abstracts screened after removal of 691 records. Following screening and eligibility assessment, 146 reports were included in the review. Data were extracted from 141 reports, while 5 reports contained insufficient information for data extraction.

Note.The majority of eligible French language reports were available in English and were thus already included in English-language reports. Because only two eligible reports were written singularly in French, language-based comparisons were not undertaken.

Interrater reliability

The current study assessed the reliability of data extraction and coding by determining interrater reliability between two independent coders. To this end, 33 reports were randomly selected from the 141 eligible reports.

Interrater reliability was assessed on both aggregate and disaggregate variables. Cohen's κ statistic and percent agreement were selected as appropriate measures to assess agreement between the raters on categorical variables (i.e., countries from which primary studies were included; whether a duplicate data extraction with interrater concordance was calculated; whether a comprehensive literature search was conducted; whether the review was published (status of publication); whether scientific quality was assessed; whether a publication bias was reported; whether a conflict of interest was reported; the types of interventions used; the setting of the intervention; the metric of recidivism)^{Footnote 6}. For these variables, interrater reliability was very high, ranging from 94% to 100% agreement (median = 97%, n variable = 28) and κ ranging from.653 to 1.00 (median =.931, n variable = 28). None of the variables were excluded due to unacceptable interrater reliability.

Results

Search results

The reviews included in this study predominantly featured primary research conducted in the United States (80.1%), Canada (49.0%), the United Kingdom (31.2%), and Australia (23.4%) (see Table 1). Moreover, most reports adhered to best practices as per the Assessment of Methodological Quality of Systematic Reviews (Shea et al., 2017), including the reporting of interrater reliability between two evaluators (56.0%), comprehensive literature searches of primary studies from multiple publication formats (e.g., dissertations and government reports; 91.5%), and characteristics of primary studies included in their database (e.g., sample demographics, details of intervention, and recidivism outcomes; 83.7%; see Table 2 through Table 13).

Recidivism outcomes

The effect of correctional interventions on recidivism were first investigated by considering recidivism as an aggregated metric. Subsequently, the effect of correctional interventions were considered on five types of recidivism identified across the eligible reports: general, sexual, violent, intimate partner violence/domestic violence, and drug-related. General recidivism refers to the commission of any new offence. This is a broad category covering any criminal behaviour following prior justice-involvement. This category consists of effects specified to general recidivism category as well as effects reflecting an unspecified recidivism. Therefore, this category does include effects that may reflect more specific types of recidivism (i.e., sexual, violent, intimate partner/domestic violent, drug-related). In instances where systematic review findings were specified to a particular offence subtype, those results were classified exclusively in the corresponding subtype category. Violent recidivism is defined as the commission of a new violent offence following initial justice-involvement. A violent offence generally involves the application, or threat of application of force to a person (e.g., assault, robbery, homicide). Sexual recidivism is defined as the commission of a new contact (e.g., sexual assault) and non-contact (e.g., indecent exposure, voyeurism) sexual offence, following initial justice-involvement. Intimate partner violence/domestic violence recidivism (IPV/DV) refers to violent recidivism occurring within a romantic or sexual relationship or against a family/household member that causes physical, psychological, or sexual harm. Finally, drug-related recidivism consists of reoffending involving possession or trafficking of drugs, as well as impaired driving due to alcohol/drug use. This can also include violating terms of release that are related to drug use.

Recidivism-type aside, reports operationalized recidivism in a number of ways, including revocation of conditional releases, rearrests, new charges, reconvictions, reincarcerations, readmissions to a hospital, and self/victim-reports^{Footnote 7}. The number and percentage of effects associated with each metric of recidivism are presented in Table 14^{Footnote 8}. It is important to note that supplementary analyses revealed that recidivism outcomes did not vary based on the metric of recidivism utilized.

Intervention effect on recidivism

Overall, 60.0% (n = 308) of the total number of effects demonstrated that correctional interventions were associated with a reduction in recidivism, aggregated across types. This indicates that, overall, a smaller percentage of justice-involved clients that participated in the intervention recidivated compared to the control group. Only 4.3% of effects (n = 22) demonstrated an increase in recidivism, 32.7% (n = 168) demonstrated no relationship with recidivism, and 2.9% (n = 15) indicated mixed outcomes. According to systematic reviews with a meta-analysis, 59.0% of effects (n = 221) demonstrated that interventions significantly reduced recidivism, with approximately half of these effects – or one third of the total number (29.1%; n = 109) – indicating at least a moderate reduction in recidivism. Although 18 (4.8%) effects from systematic reviews with a meta-analysis demonstrated that interventions increased recidivism, only three of these effects (0.8%) indicated at least a moderate increase in recidivism.

Nevertheless, despite widespread awareness of the RNR model and principles of effective correctional interventions (Bonta & Andrews, 2024), it is important to note that not all interventions are aligned with these principles – raising concerns about the quality of some of the sampled interventions. While the current study did not directly examine the quality of interventions, broader research in the field has shown that such interventions – particularly those that do not adhere to evidence-based principles – can, in some cases, lead to increases in recidivism. These risks underscore the importance of evaluating both the context and delivery quality of correctional programs.

The majority of the effects demonstrated that interventions (overall) were consistently associated with reduction across the five types of recidivism (ranging from 53.1% for sexual recidivism to 64.0% for IPV/DV recidivism; see Table 15). The results from systematic reviews with a meta-analysis (see Table 16) were also consistent (50.0% for violent recidivism to 64.0% for IPV/DV recidivism). However, only between 25.0% (for drug-related recidivism) and 48.0% of effects (for IPV/DV recidivism) indicated at least a moderate reduction in recidivism. A small percentage of the effects indicated that interventions increased general, sexual, IPV/DV, and violent recidivism (ranging from 2.0% for sexual recidivism to 10.8% for violent recidivism). Effects from systematic reviews with a meta-analysis demonstrated that nearly all of these increases were small in magnitude.

Intervention types and recidivism

Beyond estimating the overall effectiveness of correctional interventions, an important goal of the current study was to disaggregate data by intervention type in order to determine their relative recidivism reduction capacity. Each type of intervention identified from eligible reports is defined below. Interventions with similar underlying theoretical or practical roots were collapsed into a single intervention category (e.g., mental health courts and drug courts were collapsed into a single "Diversion" category).

Intervention types identified from systematic reviews and meta-analyses

Aftercare

Applied to justice-involved clients post-release from correctional institutions, these programs are designed to monitor and provide individuals with skills and resources for successful reintegration.

Anger management

Delivered in individual or group settings to teach methods to identify triggers of anger, regulate anger, and identify strategies to mitigate feelings of anger.

Behavioural therapy

Consists of behavioural and environmental manipulations to help the justice-involved client relearn societal norms (e.g., token economies).

Bootcamp

Consists of an intense, militarized regime for incarcerated justice-involved clients.

Cognitive Behavioural Therapy (CBT)

Targets the relationship between dysfunctional cognitions and behaviours in order to disrupt maladaptive behavioural patterns and establish healthier and more prosocial behavioural patterns.

Cognitive skills training

Consists of training programs that aim to improve problem-solving capacity, attention and focus, memory, risk and reward processing, and cost-benefit analyses.

Counselling

Consists of listening to individuals with empathy and encouraging them to express themselves and achieve their goals. A particularly prominent humanistic approach is "motivational interviewing": an interview-based counselling technique that enhances intrinsic motivation by resolving ambivalence towards change and empowering the individual to effect positive change.

Diversion

Strategies that divert justice-involved clients from traditional court systems into court settings that mandate treatment programs targeting specific factors underlying the criminal offence. This can include substance use (i.e., drug treatment courts), mental illness and symptoms (i.e., mental health courts), or courts that focus on specific types of crimes (i.e., IPV/DV courts).

Duluth model

Specific to perpetrators of intimate partner violence or domestic violence. This model aims to explore and address the root of aggression within the perpetrator, particularly in relation to their partner, and teaching methods to manage and mitigate this aggression.

Education/vocation

Consists of coursework, institutional education programs, hands-on training, work release, or other programs that aim to increase educational and vocational qualifications among justice-involved clients.

Group therapy

Adopts a group-based approach where peer support is an integral part of the treatment process (e.g., group therapy, Alcoholics/Narcotics Anonymous, other 12-step programs).

Medical/surgical

Consists of surgical and chemical castration. This is usually only provided to clients convicted of a sexual offence.

Multisystemic therapy

A community and home-based therapeutic approach that incorporates features of family therapy and CBT to address risk factors for recidivism, as well as provide social networks with the tools required to support the successful reintegration of the justice-involved client post-conviction. This intervention is typically provided to youth aged 12-17, but its effects can also be assessed in terms of long-term effects from juvenile delinquency into adult criminality.

Offence-specific programming

Tailored to specific types of recidivism. Typically, there was no specification of what types of interventions were incorporated in these programs.

Other psychotherapy

Collapsed collection of psychotherapies outside of behavioural therapy, multisystemic therapy, group therapy, CBT, and psychoeducation that were either specified as psychotherapies (without further specification), or were mentioned by fewer than 10 eligible systematic reviews. These therapies include psychodynamic therapy, insight-based therapy, rational emotive therapy, and motivational enhancement therapy.

Pharmacology

Involves the administration of an exogenous substance, including medicine and hormones. Pharmacological therapies include opioid agonist and partial-agonist therapy and the use of psychiatric prescription medicines for other mental disorders. "Chemical castration", however, was included in the medical/surgical category.

Psychoeducation

Provides knowledge and insight to justice-involved clients about a mental illness or substance use disorder they may exhibit. This includes information about etiology, symptomology, risk for relapse, coping strategies, available treatment strategies, need for adherence to treatment, prognosis, and destigmatization.

Psychosocial skills training

Consists of training programs that teach coping, resilience, and supportiveness.

Restorative justice

Consists of practices that serve as alternative means of engaging with the client and victim of the crime than the traditional criminal justice system. These practices seek to repair harm by facilitating communication between perpetrators and victims to help the justice-involved client understand and take responsibility for the harm they have caused, and to help victims reduce anxiety and feelings of powerlessness in response to their victimization.

Social skills training

Consists of training programs that teach skills, such as communication, interpersonal interaction, and interpersonal problem-solving.

Supervision

Consists of moderate to intensive supervision or careful monitoring of justice-involved clients within their community. Supervision can be undertaken by probation officers or family members, the latter being the case in the context of home confinement/house arrest. Monitoring can also involve the use of electronic monitoring equipment, such as an ankle bracelet.

Other

Interventions mentioned by fewer than 10 eligible systematic reviews and did not fit into any other categories. These included schema therapy, risk management training, detoxification, and self-help.

Unspecified

"Intervention" used as an umbrella term, without specification of type.

An important goal of this RoR was to determine whether each intervention type was primarily associated with reduced recidivism, increased recidivism, unchanged recidivism, or mixed recidivism outcomes. To this end, the number and percentage of effects were calculated for each direction of association – reduction, increase, no effect or mixed. Notably, although there were individual cases in this study in which interventions were associated with increases in recidivism, there were no cases in which the majority of effects (i.e., more than 50%) for any single intervention type indicated an increase in recidivism.

Table 17 denotes which intervention types were primarily associated with a reduction in recidivism, or unchanged recidivism based on the direction of the relationship between intervention type and recidivism captured by the majority of effects (i.e., more than 50%). If no single direction – reduction or no change - accounted for more than 50% of the effects, the intervention was classified as having a mixed outcome.

Per Table 17, nearly all intervention types were associated with reductions in general recidivism, such that the group that received the intervention was less likely to recidivate compared to the control group. Moreover, as mentioned above, there was no unique intervention type for which the majority of its effects indicated an increase in recidivism. The only intervention types that did not produce reductions in general recidivism, according to the majority of effects, were bootcamps, the Duluth method, restorative justice (despite restorative justice effectively reducing sexual and violent recidivism), and intensive supervision. This suggests that most types of correctional intervention are more effective than correctional interventions with limited rehabilitative focus (e.g., bootcamps and intensive supervision) as well as correctional sentences without a treatment component.

Table 18 denotes whether the effect size of an intervention reached a magnitude that indicates at least a moderate reduction in recidivism based on whether the majority of effects from systematic reviews with a meta-analysis (>50%) indicated that the difference in recidivism between the treatment and control group was at least of a moderate magnitude.

Per Table 18, the strongest reductions in general recidivism (i.e., at least a moderate reduction) were associated with participation in anger management programs, offence-specific programs, and pharmacological interventions. It should be noted that while medical/surgical interventions yielded reductions in general recidivism of at least a moderate magnitude, these effects were specific to individuals who committed a sexual offence.

Notably, Tables 17 and 18 illustrate that justice-involved clients participating in cognitively-focused and psychotherapeutic interventions – including cognitive behavioural therapy (CBT), cognitive skills training, and other forms of psychotherapy – consistently experienced recidivism-reduction effects for violent, sexual, IPV/DV, or drug-related offending. Moreover, all reductions in recidivism linked to psychotherapeutic interventions were at least of a moderate magnitude, with the exception of the reduction in drug-related recidivism.

The other types of interventions were not consistently associated with reductions in all recidivism types. The robust effectiveness of CBT, cognitive skills training, and related psychotherapies notwithstanding, some intervention types were particularly effective in reducing certain forms of recidivism, but not others. For sexual recidivism, the most effective interventions – yielding at least a moderate reduction in recidivism – included behavioural therapy. For violent recidivism, the most effective intervention types included anger management, behavioural therapy, and group therapy. For IPV/DV recidivism, the most consistently effective interventions included counselling, group therapy, psychoeducation, and social skills training. Finally for drug-related recidivism, only "other types of interventions" (i.e., ignition interlocks and unspecified programs for relapse and driving while intoxicated prevention) yielded moderate effects.

Intervention settings

The current study also assessed whether the effect of interventions varied as a function of the particular setting in which they were delivered. Five settings were identified in the sampled systematic reviews, which we defined in below. Table 19 presents a summary of the intervention settings associated with reduced, unchanged, or mixed recidivism outcomes based on the direction of the majority of the effects. Similar to effects associated with intervention types, none of the settings showed a majority of effects (i.e., more than 50%) indicating an overall increase in recidivism. Table 20 presents a summary of the intervention settings associated with at least a moderate reduction in recidivism, according to effects from systematic reviews with a meta-analysis.

Settings identified from systematic reviews and meta-analyses

Custody

Consists of incarceration in provincial/state and federal prison systems.

Community

Consists of interventions delivered in a non-custodial and non-residential setting in the general community.

Court

Consists of interventions delivered in a judicial court setting. These are most typically court-diversion practices (e.g., mental health courts).

Residential

Consists of interventions delivered in the clients own residence or fixed address.

Hospital

Consists of interventions delivered in general or psychiatric hospitals.

Table 19 and 20 showed that interventions delivered in all settings were associated with reductions in general recidivism, although effects were predominantly small in magnitude; only interventions delivered in hospital settings evidenced reductions in general recidivism of at least a moderate magnitude. Similarly, interventions delivered in all settings were associated with reductions in violent recidivism, but these reductions were at least moderate in magnitude across all settings except courts. For sexual recidivism, reductions of at least moderate magnitude were evident in custody, community, and residential settings. IPV/DV recidivism evidenced reductions in all settings but court settings, with at least moderate effects. Finally, drug-related recidivism was reduced following interventions applied in community and court settings, with only the latter being associated with at least moderate reductions.

Discussion

The ultimate goal of this study was to assess the effectiveness of correctional interventions in producing reductions in recidivism using a broad, comprehensive approach that had not yet been utilized in previous RoRs. Intervention effectiveness was also assessed through disaggregation by (1) intervention type and (2) the setting in which the intervention was delivered. An exhaustive search of systematic reviews evaluating the effectiveness of correctional interventions identified a breadth of intervention types stemming from theories of biology, chemistry, and medicine (e.g., pharmacology, medical/surgical); psychotherapy (e.g., CBT, Multisystemic therapy); schematization and socialization (e.g., Duluth model; social and psychosocial skills training); behaviourism (e.g., behavioural therapy); restorative justice; court diversion (i.e., drug treatment courts, mental health courts), education and training (e.g., including psychoeducation, cognitive skills training, anger management), counselling, "scared-straight" (e.g., bootcamps), and aftercare. The broad range of interventions evaluated in the empirical literature provides a wealth of data from which to synthesize findings and identify interventions and characteristics thereof that yield the strongest recidivism reduction effects.

Overall effects of correctional interventions on recidivism

The overall effect of correctional interventions was determined by synthesizing the difference in recidivism rates between justice-involved clients who participated in correctional interventions (regardless of intervention type) and control groups that did not. From this synthesis, it was inferred that just over half of intervention effects (60%) demonstrated that correctional interventions were associated with a reduction in recidivism. This was true for each type of recidivism when their disaggregated changes were assessed. This suggests that correctional interventions could reduce recidivism more effectively than incarceration or community sentences without a treatment component.

These findings also tentatively support the body of evidence that contradicts earlier positions that "nothing works" in the treatment of justice-involved clients (Farabee, 2005; Kirby, 1954; Martinson, 1974), by identifying a repertoire of intervention strategies that have the potential to effectively reduce recidivism. "What works," according to the current study, is consistent with a more contemporary forensic approach of evidence-based therapeutic, rehabilitative, and educational practices (Bonta & Andrews, 2024; Cullen & Gendreau, 2007; MacKenzie & Lattimore, 2018; Riley, 2011).

Despite the overall positive trends, over 70% of effects from meta-analyses reflected primarily small reductions in recidivism. Approximately 30% of effects observed in this study indicated no change in recidivism, and a small portion (4-10%) still showed increased recidivism. These null or negative outcomes may reflect limitations in how interventions adhered to evidence-based principles. As intervention quality was not systematically assessed in this study, further research is recommended to evaluate the extent to which these programs align with established evidence-based practices and to better understand their impact. Further research should also explore how the effectiveness of correctional interventions can be improved to further reduce recidivism.

As previously noted, recidivism can have substantial public safety, social, and economic implications if not addressed. Recidivism accounts for a substantial portion of crime (Department of Justice Canada, 2019; Falk et al., 2014) and millions in taxpayer funds (Day et al., 2015; Gabor, 2015). The current study investigated whether the type of intervention applied moderated whether the intervention effectively reduced recidivism. This was assessed by evaluating the relationship between each intervention type and changes in general recidivism as well as sexual, violent, IPV/DV, and drug-related recidivism. The results and implications of this evaluation is described below.

Individual effects of separate intervention types

The interventions that yielded the most robust effects were those based on cognitive theories of psychology. These interventions included cognitive skills training, which aims to improve problem-solving capacity, attention and focus, memory, risk and reward processing, and cost-benefit analyses of their actions; and CBT, which – in a correctional setting – focuses on identifying cognitive antecedents or maladaptive thinking patterns that underlie, lead to, or justify criminal conduct, and then modifying them to support prosocial behaviour (Bonta & Andrews, 2024).

These findings are consistent with the RNR's general responsivity principle stating that treatment grounded in cognitive behavioural and social learning approaches optimizes recidivism reduction (Bonta & Andrews, 2024). Hence, this offers support that intervention adhering to, at least, the responsivity principle can more often than not be associated with reduction in recidivism. Future research that also assesses for the implementation of tenets of the risk (i.e., use of a risk assessment tool prior to treatment) and need principles (e.g., targeting antisocial cognitions) could further identify the strengths and potential weaknesses of correctional interventions and RNR, as a whole.

Beyond cognitive-based and, more broadly, psychotherapeutic interventions as a whole (e.g., insight-based therapy, rational emotive therapy, motivation enhancement therapy), other types of correctional intervention also yielded meaningful reductions in specific forms of recidivism based on this RoR. For sexual recidivism, other effective intervention types included behavioural therapy. This suggests that prosocial reinforcement even in the absence of cognitive training could lead to reductions in sexual recidivism. Violent recidivism reductions were associated with anger management, behavioural therapy, and group therapy, suggesting that interventions focusing on emotion-regulation, prosocial reinforcement, and group-based support can attenuate violent reoffending.

Similarly, IPV/DV recidivism reductions were associated with group therapy, social skills training, counselling, and psychoeducation, highlighting the importance of social skills development and other educational components in the reduction of intimate partner violence and domestic violence. Finally, for drug-related recidivism, few interventions yielded at least moderate recidivism reduction, with these being limited to "other" types of interventions (notably interlocks). However, this finding should be interpreted with caution and could be an artefact of the more limited body of literature specifically dedicated to drug-related recidivism versus other forms of offending – specifically, the number of effects available for the former was 31, versus 371 for general recidivism.

The current study identified a few interventions which yielded no change in recidivism, and were therefore labelled ineffective. These approaches focused on being "tough" on the offender, including bootcamps and intensive supervision without an additional treatment component designed to affect sustained behavioural change. This finding is consistent with previous research suggesting that such strategies may be counterproductive effect in reducing recidivism (Barnett & Howard, 2018), particularly due to their negligence of the RNR principles (Bonta & Andrews, 2024). However, restorative justice practices were also found to be relatively ineffective, which counters hypotheses that restorative justice could be instrumental in reducing future criminal behaviour and recidivism (Daly, 2021; Federal-Provincial-Territorial Working Group on Restorative Justice, 2020; United Nations Office on Drugs and Crime, 2020). One reason for the null effects of restorative justice practices could be variability in the implementation of therapeutic and rehabilitative components to the practices. Therefore, it may be premature to deem restorative justice practices, as well as bootcamps and intensive supervision, ineffective until the quality of these interventions and their adherence to substantiated therapeutic principles has been systematically assessed.

The implications of treatment quality

Despite some general conclusions about the most promising forms of correctional intervention, the reader is cautioned that beyond intervention type, there are many other factors that can potentially impact program effectiveness. One reason why correctional interventions could yield suboptimal results lies in the quality of the interventions implemented. Research has supported guidelines on the "what", "who", and "how" in correctional interventions, particularly with the RNR model (Bonta & Andrews, 2024). Although the quality of individual systematic reviews was assessed during the PRISMA search process, an evaluation of the quality of each intervention was not. Notably, given inconsistent and unreliable data, this RoR could not code for the extent to which interventions adhered to the principles of effective correctional intervention – namely, those encapsulated under the RNR model. Further research should consider adherence to RNR, as well as other treatment frameworks (e.g., the Good Lives Model; Ward et al., 2007), as a potential moderator of the efficacy of correctional interventions on recidivism. This could help further explain why the vast majority of effects (~70%) demonstrated only small reductions in recidivism and why another ~10% found increased recidivism post interventions. It is likely that the effects were diluted by incorporating all interventions into the analysis regardless of quality.

Conversely, it is reasonable to posit more positive outcomes would be obtained from interventions adhering to the core RNR principles and implemented with fidelity (i.e., implemented as intended by program developers). For one, the current study offers promising results on the use of CBT and other cognitively-focused psychotherapeutic interventions and reductions in recidivism, which is directly in line with the RNR framework (Bonta & Andrews, 2024). Moreover, a long history of systematic reviews has demonstrated that adherence to RNR enhanced the efficacy of correctional interventions (Dowden & Andrews, 2000; Hanson et al., 2009; Prendergast et al., 2013; Travers et al., 2021). In contrast, systematic reviews that have reported increased recidivism post-intervention have noted that this may reflect non-adherence to RNR (Lowenhamp et al., 2006; Shaffer, 2011) as well as treatment attrition (Olver et al., 2011). Although more recent systematic reviews have found inconsistent results (Duan et al., 2023; Fazel et al., 2024), the operationalization of RNR principles in their meta-analyses was vague, which could potentially explain the null effect. Moreover, Duan and colleagues (2023) claimed that CBT is incompatible with RNR when, in fact, CBT is a key element of the general responsivity principle (see Bonta & Gendreau, 2024 for full commentary). This fundamental error further suggests that studies may have been miscoded. To the authors' knowledge, no RoR has incorporated level of RNR adherence (or other indicators of program quality) as a moderating variable. Based on the current study, however, one may posit that RNR adherence would be associated with substantial reductions in general and specific forms of recidivism.

The reader is also encouraged to consider other factors that could have moderated the results of correctional interventions, including treatment fidelity; the quality of the client-provider relationship; and the potential for clients to have been enrolled in several intervention programs, either sequentially or in parallel. None of these indicators could be included in the RoR given the lack of sufficient detail available in the primary articles. However, previous empirical research has underscored these factors as important moderators of program effectiveness (Blagden et al., 2016; Bonta & Andrews, 2024; Polaschek, 2011; Riley, 2011; Rotter & Carr, 2013).

Intervention settings and recidivism

The current study also assessed how intervention efficacy varied based on the delivery setting, which included custody, community, hospital, court (e.g., diversion practices), and residential (e.g., house-arrest) settings. Meaningful reductions in recidivism were observed from interventions delivered across all settings. Notably, both community and custodial interventions were associated with reductions in recidivism. However, the current RoR was unable to directly compare the efficacy of interventions delivered in custodial versus community settings. The only conclusion that can be drawn based on these findings is that correctional interventions have the potential to be effective in both of these general settings, and that, should a client be deemed fit, community interventions should be as part of a sentence.

It should be noted that extant research has directly compared the efficacy of interventions delivered in community versus custodial settings, and has found that community-based interventions were associated with larger reductions in recidivism than custodial interventions (Andrews & Bonta, 2006; Koehler & Losel, 2024). Indeed, these results may partially reflect that those individuals receiving a community sentence are generally lower risk and have less serious index offences than those sentenced to custody. Hence, to the extent that it does not compromise public safety, treatment delivery in a community setting is recommended over custody-delivered treatment under circumstances where an offender's overall risk level safely allows for a community sentence.

Interestingly, the current study found that interventions delivered in hospital settings (e.g., CBT and other psychotherapies, pharmacological interventions, medical/surgical interventions) were associated with reductions in general recidivism of at least a moderate magnitude, and this was also true for violent and IPV/DV recidivism. A possible reason for this finding could be that hospitals and psychiatric facilities may have a more inherent treatment and care-oriented philosophy in their interaction with the client compared to programming delivered in a traditional prison (Lamb & Weinberger, 2005; Urbanoski et al., 2008). For instance, a qualitative study found that staff members at correctional facilities expressed concern that that justice-involved clients, particularly those with serious mental health concerns, have unmet needs in correctional facilities and would benefit from more comprehensive treatments in hospital settings (Segal et al., 2018). Given limited resources, however, whether a justice-involved client is receiving treatment from a traditional prison environment or a hospital setting will depend on the severity of mental health concerns and the nature of the criminal offence (Brown et al., 2024; Pozzulo et al., 2022). Moreover, research has found that justice-involved clients are increasingly being displaced from hospital-based care towards custodial care due to both stagnating or declining hospital capacities and increasing rates of mental health concerns in the general (non-correctional) population (Lamb & Weinberger, 2005; Huxter, 2012; Kim, 2016; Slate, 2017).

Limitations

There are certain caveats and limitations inherent in the current research. As previously mentioned, the inability to reliably code for RNR adherence or other indices of intervention quality precluded determining the extent to which these factors impact overall program effectiveness. It is recommended that primary studies include clear details on RNR adherence and implementation fidelity so that future systematic reviews may incorporate these metrics.

In addition, the current RoR did not encode the specific moderating effect of follow-up time. Therefore, the current results are unable to inform whether these effects reflect short- or long-term changes in behaviour and attitudes related to criminal behaviour. Typical follow-up periods range between 2 and 5 years (Pedneault et al., 2024; Stewart et al., 2019), and longer follow-up periods are typically associated with higher rates of recidivism. Therefore, future research should consider follow-up length as a moderating factor for the efficacy of treatment programs, and identifying interventions that yield short- and/or long-term reductions in recidivism.

In addition, a limitation of RoRs, in general, is the lack of independence of each systematic review included in the analysis. With a large dataset of systematic reviews on specific topics, different systematic reviews have likely pooled their results based on identical references (i.e., original/primary sources). Ideally, it is recommended that there be no overlap in primary sources between systematic reviews included in an RoR (Petticrew & Roberts, 2006; Petrosino & Lavenberg, 2007; Welsh & Rocque, 2014). An independence analysis was conducted by the authors of this RoR, which required extracting the reference for the primary studies pooled in each effect included in this RoR. This analysis found that there was substantial overlap in primary studies between systematic reviews, which has also been found in other RoRs (i.e., Weisburd et al., 2017). Consequently, it is plausible that the results of primary studies were double-counted, potentially inflating the results of the RoR.

Furthermore, another limitation with RoRs, in general, is that they typically produce an "average of averages." In turn, this raises the risk of obscuring important details and provide rather an oversimplified view of the data. For instance, the current study create categories of recidivism based on prior categorizations in the sampled systematic reviews, precluding an examination of the effect of interventions on specific crimes (e.g., import/export of illegal substances). Therefore, the results of this should not be considered the authority on what works and what does not work in correctional settings, but rather provide indices of what might work, and what may require further research.

Finally, while some indices of methodological quality of the sampled systematic reviews were extracted and evaluated, there was not enough information from the sampled systematic review to determine the moderating effect of methodological quality on intervention effects. The authors of the current RoR encourage others to report their methodological assessments according to the Assessment of Methodological Quality of Systematic Reviews (Shea et al., 2017). Future work should consider these quality indicators when determining the aggregated efficacy of various correctional interventions.

Future directions

The current RoR offers several avenues for future research. For instance, as mentioned above, future primary studies should consider RNR adherence and additional elements of treatment quality as important moderating factors in evaluations of correctional interventions. While a future RoR could include a measure of intervention quality (to the extent possible based on primary studies), a large-scale meta-analysis is also an option that would avoid the risk of overlap in primary studies.

In addition, the current RoR was unable to directly compare the efficacy of various intervention types delivered in different settings (e.g., CBT delivered in custody compared to a community setting), nor could it assess the effectiveness of intervention against each other. Future studies should explore how the delivery setting influences the effectiveness of specific interventions and include direct comparisons among different types of interventions to better understand their effects on recidivism reduction.

Furthermore, while the current study assessed the effect of interventions on various forms of recidivism, further research should consider the efficacy of interventions on various populations, perhaps with a Gender-based Analysis Plus lens. Indeed, many of the interventions were developed using a sample consisting primarily of adult Caucasian male justice-involved clients. Future RoRs should consider the efficacy of interventions within other populations of clients, including youth, members of various ethnoracial groups, and members of different biological sexes, gender identities, and sexual orientations.

Conclusion

The current study involved a comprehensive search for systematic reviews assessing the effect of correctional interventions on recidivism. The general finding suggests that correctional interventions, particularly therapeutic interventions that are cognitive behavioural in nature (e.g., CBT), are more effective at reducing recidivism than more punitive or restrictive approaches that do not incorporate a treatment component. Further research is required, particularly in regard to how adherence to effective principles of correctional intervention interact with program type. Nevertheless, the results of this study offer strong support for the rehabilitative model of correctional intervention.

References

Date modified:

2026-08-28