A Systematic Review of Diversion Measures for First Time Entrants to the Youth Justice System

Abstract

Diversion is defined as the practice of providing an alternative outcome for children who have engaged in offending behaviour, which keeps them away from the formal criminal justice system. Diversion is a significant element of the Child First approach, and is widely implemented across England and Wales. The increase in use of diversion in recent years has contributed to reductions in First Time Entrant rates; however, evidence regarding whether these measures reduce reoffending is limited. Diversion practices vary significantly both nationally and internationally. Hence, much could be learnt from reviewing the international literature to establish what could be implemented in England and Wales. Therefore, this systematic review aims to explore the effect of different types of diversion measures on reoffending in First Time Entrants. To do this, a comprehensive search of 10 electronic databases was undertaken to systematically identify literature relating to youth justice diversion. A systematic screening procedure was followed to identify studies that met inclusion criteria, data on key information was extracted and studies were assessed using a quality assessment tool. Results were synthesised using the EMMIE framework, to increase applicability to practice. A total of 12 distinct studies were included in the review, with both experimental and quasi-experimental design. One study was conducted in Norway, one in the Netherlands and the rest in the United States of America. Findings were varied but generally suggested diversion leads to lower rates of recorded reoffending than formal processing. When comparing between types of diversion, often there was no difference in reoffending but generally, the least intensive intervention showed the best results. However, findings from one study that considered self-reported offending prompts consideration of whether diversion genuinely affects behaviour or simply influences the detection of offences. This review indicates promising results for diversion interventions but highlights the need for further research to provide conclusive evidence regarding which measures are most effective.

1. Introduction

Contact with the criminal justice system has been described as “a violent disruption of a child’s life” which risks their well-being and development (Anderson & Ratanadilok, 2019, p. 200). Formal system contact is widely understood to be harmful for children, and research suggests it may increase the risk of future offending behaviour (McAra & McVie, 2007). This is underpinned by labelling theory, which highlights the impact of stigmatisation on identity development and future behaviour (H. S. Becker, 1963). To minimise these negative effects, diversion measures intend to keep children away from formal justice processes, particularly for those with relatively minor or first offences. Diversion measures have been used within the justice system in England and Wales for decades. For example, the use of police warnings or cautions was first documented in 1858 (Steer, 1970 as cited by Cushing, 2014), though their use has fluctuated and been significantly impacted by political trends. At present, diversion measures are widely implemented across Youth Justice Services (YJSs) in England and Wales (Robin-D’Cruz & Tibbs, 2019) and diversion forms a significant part of the Child First approach (Youth Justice Board [YJB], 2021b). However, despite enthusiastic uptake in recent years, the theoretical underpinnings of diversion have received little scrutiny, and the evidence base is far from conclusive (Haines et al., 2020). It seems apparent that increasing diversion will reduce first time entrant (FTE) rates, a key performance indicator measured by the youth justice board (Youth Justice Board [YJB], 2025), as children who have been diverted are simply excluded from this statistic. However, whether it has a positive impact on their subsequent behaviour is less clear.

Smith (2024) describes areas of “uncertainty, incomplete knowledge and controversy” (p. 140) within this field and research appears to be significantly impacted by the ambiguity in both definition and practice (Hoge, 2016). Broadly, diversion aims to direct children away from the criminal justice system in order to reduce the intrusion and stigmatisation they experience. This is understood to have a positive impact on wellbeing but also on future offending behaviour, as children are less likely to have internalised a ‘criminal’ identity and are more able to desist from offending (Bramley et al., 2019). However, exactly what children are intended to be diverted from and to is unclear, and often variable in practice (Richards, 2014). Diversion can take place at different points in the criminal justice system and be determined by different organisations such as the police, YJS or court intake officers (Hoge, 2016). International differences in diversion practice are significant, but practice also varies from one US state to another (Niedzwiecki et al., 2015) and from one city to another in England and Wales (Robin-D’Cruz & Tibbs, 2019). Different interventions are often informed by different theoretical models and approaches (Smith, 2018) such as restorative justice interventions where the focus is on repairing harm (Braithwaite, 1989; Tyler, 2006) or welfare interventions which aim to promote the wellbeing of the child and their family (Pratt, 1986, as cited in Smith, 2018).

Within the field of diversion, there has historically been a tension between what Cohen (1979, p. 349) called “true diversion” which involves removing the child from the system entirely and “new diversion” which involves offering intervention in place of a formal criminal justice response. On one hand, some consider diversion as an approach that involves minimal, or even no intervention, to minimise the child’s contact with the system as much as possible and thus have the greatest effect in terms of reducing risk of labelling and stigma (Schur, 1973). Meanwhile, others consider it a proactive, intervention-focused practice, where significant resources are used to intervene with the individual to meet their needs and provide them with the support they need to be able to desist from offending behaviour, which Kelly and Armitage (2015, p. 117) call “interventionist diversion”. These two approaches are opposing, meaning the field lacks clarity regarding what effective practice should look like (Hoge, 2016). Furthermore, due to court backlogs since the COVID-19 pandemic (Crest Advisory, 2022) and the financial implications of court processing, there is often a focus on reducing the number of children who are processed through court, without evidence of whether the alternative is more effective.

Out-of-court disposals are used in England and Wales to divert children away from formal court processes (Youth Justice Board [YJB], 2024). These can be informal outcomes given by the police, in conjunction with the YJS, to avoid a child having to be formally processed. For example, Outcome 22 can be used when a child engages in some sort of diversionary intervention and therefore no further action is taken. This outcome can be used at any stage of an investigation, including in the early stages when the evidential threshold is not met (e.g., because a victim does not wish to continue), or at a later stage when the decision can be made to defer prosecution until the child has been given opportunity to engage in diversionary intervention. In this case, if they complete the identified intervention there will be no further action but if they do not, the investigation may continue, and a charge may be given. Alternatively, a Community Resolution can be given for ‘low-level’ offences where a course of action is set, such as engaging in intervention or apologising to the victim, but there are no consequences if the child does not complete this. Out-of-court disposals can also be a more formal process, in the form of a Youth Caution or Youth Conditional Caution. For this outcome a child is still interviewed by the police under caution; however, if they provide a full admission of guilt, a caution may be deemed appropriate. An assessment is completed by the YJS to identify necessary work, lasting no longer than 16 weeks. The difference between the two is that a conditional caution involves compulsory conditions which the child must complete to avoid being sent to court, while the intervention associated with standard cautions is voluntary. Whilst these processes provide some structure to the diversion methods in England and Wales, there is still variability as to how this looks in practice in different areas of England and Wales. As diversion is not a statutory requirement of YJSs, formal data is not collected on exactly how diversion is implemented in each service/area for example, who is diverted, for what offences and what impact this has on subsequent offending or desistance (Robin-D’Cruz & Tibbs, 2019). A review by Keenan et al. (2023) provides a comprehensive overview of the research in England and Wales which enhances understanding of current practice but also highlights the lack of ‘rigorous empirical research in the field’ (p. 199).

Furthermore, out-of-court disposals, as used in England and Wales, are often what Hoge (2016) describes as a “quasi-judicial program” in that the child is not entirely “divorced from the system” (p. 993) as they still meet with the police, attend a YJS building and can be requested to complete a range of interventions, in some cases with the threat of being formally charged if these are not completed. Their intervention could look very similar to a child who has been processed formally and given a short community order, the only difference being that they have not attended court. Furthermore, even informal out-of-court disposals can be disclosed on enhanced Disclosure and Barring Service checks (Kelly & Armitage, 2015), meaning it could be argued that this contact can still be oppressive and stigmatising and perhaps even leads to question whether this is aligned with the true meaning of diversion (Cushing, 2016).

There are also restrictions of diversion in different areas, which can impact who is eligible to be diverted. For example, in England and Wales children are required to admit their offence, or at least accept responsibility, to be eligible for some diversion outcomes (Youth Justice Board [YJB], 2024), which Cushing (2014) describes as “regrettable” (p. 41) as it can put pressure on children to admit guilt so that they can avoid formal processing, regardless of whether this is true. Furthermore, if decisions regarding diversion depend on factors such as prior system contact, diversion may be at risk of amplifying inequalities, as Black and mixed heritage boys are more likely to have already had contact due to this group being disproportionately criminalised (Her Majesty’s Inspectorate of Probation [HMIP], 2021).

While diversion intends to reduce offending by keeping children from the system, there is a risk that it may increase the number of children in contact with the justice system, by including children in diversion schemes who would otherwise not have had any system contact (i.e., would have been released by the police with no further action) (Hoge, 2016). Cohen (1979) described the “irony” (p. 346) of alternative diversion interventions in fact becoming “programs which supplement the existing system … attracting new populations”, describing this as “widening the net” (p. 347). The term ‘net-widening’ is now commonly used, and increased awareness means efforts are made for this to be avoided. However, it can often still occur unintentionally, for example when police or schools are tempted to use an available diversion intervention because of its reported effectiveness, with a child whose offending would have otherwise been dealt with an informal warning or typical school discipline procedures (Hoge, 2016). Being involved in diversion may have stigmatising effects, particularly when compared to having no contact with the system, and it can increase the risk of future formal processing if conditions of the diversion are breached, hence potentially criminalising children unnecessarily. Debates regarding diversion have been ongoing for decades (Bullington et al., 1978) and this tension between ‘doing nothing’ and ‘doing something’ is yet to be resolved.

To summarise, diversion is currently a youth justice priority in England and Wales (Youth Justice Board [YJB], 2021b) and whilst the theoretical basis for this is sound—and there is significant research evidence to suggest formal processing is harmful for children and increases risk of reoffending (McAra & McVie, 2007)—there is a gap in our understanding of what good diversion looks like. Below, existing reviews of the literature are discussed and the purpose of the current review is outlined, aiming to go some way filling this gap by reviewing existing studies from around the world to inform diversion practices in England and Wales.

1.1. Previous Reviews

There are several existing reviews in the field of youth justice diversion. Whilst their findings vary, they generally suggest that diversion measures are effective at reducing recidivism when compared to formal processing, particularly when these occur pre-charge, but have differing findings related to what types of diversion measures are most effective. A review by Petrosino et al. (2010) concluded that formal processing had a negative impact on reoffending. They found the biggest differences in studies that compared formal processing to diversion with services, rather than minimal intervention, suggesting that diversion into supportive services is most effective. Meanwhile, two meta-analyses have explored the effectiveness of youth diversion measures (Schwalbe et al., 2012; H. A. Wilson & Hoge, 2013), with differing findings. The results of H. A. Wilson and Hoge’s (2013) review indicated that diversion is more effective at reducing reoffending than formal justice processing. Specifically, they found pre-charge diversion to be more effective than post-charge. Meanwhile, Schwalbe et al. (2012) found a non-significant relationship between diversion and reoffending, with only family-based interventions being found to be effective. Within these and other similar reviews, it is difficult to identify commonality in the population targeted, the point of intervention, the content of the programme and the intended outcomes. Later reviews have therefore tended to have a more specific focus, for example, J. S. Wong et al. (2016) examined diversion programmes informed by restorative justice, finding them to be generally effective. Meanwhile, D. B. Wilson et al. (2018) examined the effectiveness of specific diversion measures which are initiated pre-charge by the police. They also found that these measures significantly reduce future “delinquent behaviour” when compared to traditional processing.

While these more specific reviews have been helpful, they do not represent the types of diversion commonly used across England and Wales today (Keenan et al., 2023). Furthermore, even the latest of these reviews (D. B. Wilson et al., 2018) only includes papers until January 2017, meaning more recent research is not represented. Most recently, an evidence review has been completed by the Youth Endowment Fund and National Children’s Bureau (Keenan et al., 2023) which provides a comprehensive summary of the evidence in this field. However, it is limited to only research in England and Wales, meaning there continues to be a gap in the field for a thorough review of international research, which could help inform practice.

Furthermore, in previous reviews, there is significant heterogeneity within the populations studied, particularly concerning prior justice system involvement. This is a problem when considering that diversion intends to minimise any negative impacts of system involvement, meaning those who have previously had formal system involvement are unlikely to benefit from the effects of diversion in the same way that FTEs would. There is therefore a need for further review which is more specific about exactly who is being diverted and from what.

1.2. The Current Review

This systematic review aims to synthesise current evidence related to the effectiveness of diversion interventions in reducing reoffending in FTEs. Given the varying definitions of diversion, as discussed above, it was important to clearly define the concept of diversion within this review. Diversion is defined as any intervention or measure by which children receive an alternative outcome that avoids escalation into the formal youth justice system (Youth Justice Board [YJB], 2021a), specifically diversion which occurs prior to the child attending court. In this review, the diverted population are defined as children with no previous recorded offending, due to the need for specificity regarding the different types of diversion carried out with different groups. An up-to-date international review of the literature is also required given the increased focus on diversion internationally in the past decade and hence the potential for a large research output measuring effectiveness of new interventions which would not have been included in previous reviews. In order to increase the utility of this review for practitioners and policymakers, this review is informed by principles of realist evaluation (Pawson & Tilley, 1997) and uses the EMMIE framework (Johnson et al., 2015) to synthesise the evidence. As well as considering effectiveness, this method of synthesis considers how a given intervention might produce the desired effects and what conditions might make them more or less likely (Tompson et al., 2021). It also considers the challenges associated with implementing diversion measures and their cost-effectiveness given these are important considerations for policymakers when considering whether an effective intervention will be implemented. An overview of the EMMIE framework is provided in Table 1.

Table 1. EMMIE framework elements and definitions (Johnson et al., 2015).

EMMIE Element	Definition
Effect	The size and direction of the effect of the intervention on the outcome.
Mechanisms	Explanation of how the intervention is understood to bring about the effects.
Moderators	Conditions that make it more and less likely for the intervention to operate effectively.
Implementation	The process of implementing and sustaining the intervention and any sources of success or failure.
Economics	The cost of implementing the intervention in relation to the benefits.

1.3. Study Selection Criteria

The inclusion criteria used for selecting studies are provided in Box 1. Any study that did not fulfil the criteria was therefore excluded from the review. Publication date was restricted to 2008 onwards, due to the changes in diversion policy that happened in the UK during this year and the change in diversion practices internationally during this period, in addition to the fact that earlier research has been robustly reviewed by previous authors and therefore inclusion would lead to repetition of previous findings.

Box 1. Definitions of inclusion and exclusion criteria.

Population: Children aged 10–17 years who are suspected to have perpetrated their first offence. Studies that include individuals outside this range will be included only if more than 80% of participants are within the desired age range. Studies that describe any prior offending by the child are excluded. Intervention: Any programme, intervention, measure or scheme which diverts children away from formal contact with the youth justice system. Must take place before the child has been involved in any formal justice procedures (e.g., attending court). Comparison: Control group who have not been offered the diversion being measured. Could be formally processed through the system (e.g., attend court) or be offered an alternative form of diversion (i.e., no intervention or a control intervention). Studies where the control group are children who refused to participate or did not complete the intervention will be excluded. There must also be some attempt to ensure the comparison group is comparable to the intervention group in terms of the severity of offending (e.g., same eligibility criteria, participant matching or statistical controls). Outcome: Quantitative measures of offending behaviour including arrests, convictions, reports of antisocial behaviour, scores on self-report measures of offending behaviour. Qualitative measures will be excluded. Study Design: Experimental and quasi-experimental design with a control group. Language: No restrictions. Effort will be made to translate papers into English as accurately as possible to include in the review. Where this is not possible, publications may have to be excluded. Date: Published between January 2008 and August 2024.

2. Methods

A protocol was developed outlining the objectives and methodology for the review, which was then followed precisely to reduce the risk of bias. The steps of a systematic review were followed, as outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins et al., 2019). Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (Page et al., 2021) were also referenced to ensure all elements of the review are appropriately reported.

2.1. Search Strategy

A search strategy was developed to access all research, published and unpublished, related to the effectiveness of diversion interventions within youth justice.

2.1.1. Sources of Literature

An electronic literature search was conducted in August 2024 within the following sources:

• OVID PsycInfo;
• OVID Medline;
• Scopus;
• ProQuest National Criminal Justice Reference Service (NJCRS) Abstracts;
• ProQuest Criminal Justice Database;
• ProQuest Applied Social Sciences Index and Abstract (ASSIA);
• ProQuest Policy File Index;
• Campbell Collaboration Library;
• ProQuest Dissertations & Theses A&I;
• Open Science Framework;
• Bibliographies of existing reviews in this field;
• Forward and backward citation searches of included studies (using Scopus or Google Scholar);
• Contacting field experts regarding their knowledge of any additional studies that meet the inclusion criteria.

2.1.2. Search Terms

Search terms used within the electronic databases are outlined in Box 2. These were created following scoping searches of the literature which helped establish the types of words commonly used to describe the key terms. Scoping searches also revealed terms which needed to be more specific, for example, excluding reports related to “cardiac arrest”. This meant search terms were able to effectively capture all of the relevant literature, without creating unnecessary workload of screening large numbers of entirely irrelevant reports. Search terms were intentionally broad, to ensure inclusion of all the relevant literature, given the inconsistent definitions of terms within this field. Where necessary, the strategy was adapted to fit the relevant database.

Box 2. Search terms for online databases.

youth* OR child* OR adolescen* OR juvenile* OR student* OR pupil* OR teen* OR kids OR minors OR boy* OR girl* OR “under 18” OR “under-18” OR (young ADJ (person OR people OR offender OR offenders OR men OR women)) AND crime OR criminal* OR offend* OR offence* OR offense* OR reoffend* OR re-offend* OR antisocial OR anti-social OR illegal OR delinquen* OR devian* OR aggress* OR violen* OR recidivis* OR (arrest* NOT cardiac) OR rearrest* OR re-arrest* OR convict* OR reconvict* OR re-convict* OR (charge* ADJ3 (offence* OR offense* OR crim*)) OR adjudicat* OR justice OR “first time entrant” OR police OR policing OR court OR “law enforcement” OR corrections OR correctional OR misdemean* OR misconduct AND divert OR diverting OR diverted OR diversion* OR caution* OR (warning ADJ4 police) OR reprimand OR pre-trial OR pretrial OR pre-charge OR pre-caution OR pre-court OR pre-arrest OR prearrest OR pre-booking OR “out-of-court” OR “out of court” OR “community resolution” OR “no further action” OR “outcome 21” OR “outcome 22” OR “mental health court” OR “teen court” OR “children’s court” OR “drug court” OR “restorative justice” OR restitution OR “early intervention” OR “non-prosecution” OR “deferred prosecution” OR (triage ADJ3 (model OR scheme OR programme OR project)) OR (alternative ADJ3 (outcome OR disposal OR disposition OR program* OR intervention OR prosecution OR sentence))

2.2. Study Selection

All publications retrieved from the search were stored within Endnote 20 and duplicates were removed by the software. Publications were then exported into Rayyan (Ouzzani et al., 2016), an online screening tool. The titles and abstracts of all publications were screened for eligibility according to the above inclusion criteria (Box 1). Once all publications had been screened, the included studies were then read in full to determine eligibility. Studies were included only if the majority of participants were aged 10–17 and had not previously engaged in offending behaviour. Diversion measures had to be implemented prior to any formal justice system procedures and the study had to have an experimental or quasi-experimental design and include a comparison group. The comparison could either have been formally processed or offered an alternative form of diversion and there had to be some attempt to ensure the two groups were comparable. The outcome for all studies had to be a quantitative measure of offending.

Where publications were not in English, an online translation tool was used to establish eligibility. All of these publications were excluded based on their title/abstract and therefore no further translation was required.

2.3. Quality Assessment

The QualSyst checklist for assessing the quality of quantitative studies (Kmet et al., 2004) was used to assess the included studies. This validated tool was selected due to its capacity to be used with any quantitative study design, meaning it was suitable for all included studies. The tool includes a checklist of 14 questions which are rated as “yes”, “partial”, “no” or “not applicable” and a summary score is then calculated. The manual provides guidance regarding how each item should be scored (Kmet et al., 2004). Quality assessment was completed for each paper by the first author. An impartial second reviewer then assessed a random selection of four papers (25%) to ensure inter-rater reliability. Discrepancies in ratings between the two reviewers were discussed and an agreement reached. The quality assessment was intended to provide more detailed information for the review, rather than identify publications to be excluded.

2.4. Data Extraction

Data extraction was completed for all included studies using a pre-defined proforma, which is considered best practice to reduce risk of error and bias (Taylor et al., 2021). Before use, the proforma was piloted on three studies included in D. B. Wilson et al.’s (2018) review and improvements were made. The proforma was completed for each study and saved in preparation for data synthesis. Data extracted included general information related to the report, characteristics of the study, details of the intervention and control group, outcomes measures, results observed, any mechanisms and moderators discussed, any information regarding implementation or cost/benefit analysis and finally the conclusions drawn.

3. Results

The above database search resulted in a total of 14,785 results which were downloaded into EndNote 20 (see Figure 1). The reference lists of previous reviews in this field (Keenan et al., 2023; Schwalbe et al., 2012; D. B. Wilson et al., 2018; H. A. Wilson & Hoge, 2013) were also downloaded, restricted by publication date, and added to the folder in EndNote 20. This resulted in addition of 163 records. Next, 4163 duplicates were removed, leaving 10,785 results. After titles and abstracts were screened, a further 10,653 were removed, meaning only 132 remained. These reports were then used for forward and backward citation searching using Scopus and Google Scholar and a further 25 new reports were added whose title and abstract fit the screening criteria. Field experts were contacted, and one additional paper was identified (Nadel et al., 2019).

Figure 1. PRISMA diagram of systematic study selection process.

There was therefore a total of 153 records for potential inclusion. These reports were then read in full by the first author and compared to the inclusion criteria to confirm their eligibility, consulting the second author when required to discuss eligibility queries. Ninety-eight studies were excluded due to not fitting the inclusion criteria. Reasons for exclusion are provided within Figure 1.

This resulted in 21 reports being included in the final review: 14 journal articles, six PhD or MSc theses, and one technical report. Three of the theses and the technical report reported the same studies included within the journal articles. However, these were included, in line with Cochrane guidance (Higgins et al., 2019), to provide the most complete information possible about the studies. Within the remaining 17 reports, there were three large scale, long-term studies that are explored within multiple reports—The Crossroads Study (Beardslee et al., 2019; Cauffman et al., 2021; Fine et al., 2017; Robertson et al., 2021), The Indianapolis Juvenile Restorative Justice Experiment (Hipple & McGarrell, 2008; Jeong et al., 2012) and The Philadelphia Police School Diversion Program (Goldstein et al., 2019; NeMoyer et al., 2023a). So as not to inflate their findings, the groups of reports are recorded under one ‘study’, but it should be noted that each report has its own methodology.

3.1. Characteristics of Included Studies

The characteristics and main findings of the included studies are presented in Table 2. Ten studies were completed in the USA, one in Norway and one in the Netherlands. Studies were published between 2008 and 2024. None of the studies were included in previous reviews (Petrosino et al., 2010; Schwalbe et al., 2012; D. B. Wilson et al., 2018; H. A. Wilson & Hoge, 2013). Three studies used a randomised experimental design, while nine were quasi-experimental, all using quantitative data.

Table 2. Characteristics and results of included studies.

Authorship (Year), Publication Type and Location	Description of Diversion and Eligibility Criteria	Control Intervention	Total Sample Size and Group Sizes	Sample Demographic Characteristics	Offence Types	Condition Assignment	Length of Follow-Up and Outcomes Measured	Results	Quality Assessment Score
Adamson (2018) Thesis Louisiana, USA	Reacting Early and Changing Habits (REACH). A two-hour programme facilitated by re-entry adult prisoners using ‘Scared Straight’ tactics. The intervention runs bi-monthly with approximately 75 children per class. Referred for diversion by juvenile court, parents or schools and then they are screened to determine which diversion they are allocated to.	Alternative diversion. Majority did community service, but alternatives included cognitive behaviour therapy and substance abuse programmes.	N = 476 REACH diversion n = 319 Alternative diversion (control group) n = 157 Note: the REACH sample only includes those who completed the entirety of the two-hour programme.	Age range = 10–17 years Mean age = 13.9 years Gender = 62% male Ethnicity = 21% White, 79% Other	Misdemeanour only	Quasi-experimental Logistic regression used to control variables (gender, race, age and living arrangements) Historical comparison group. Prior to the development of REACH.	Follow-up: Up to 4 years for control group and up to 2 years for diversion group Outcome: Reoffending. Defined as whether a participant was re-arrested in the state of Louisiana during adolescence, after completing the diversion. Arrests that took place after the participants turned 18 were not recorded. Moderating variable: Age	Those who participated in REACH were less likely to be rearrested than those who had an alternative diversion (OR = 0.520, p = 0.004). Age was found to moderate the relationship in that REACH was only effective for older (14–17 years) children and not for younger (10–13 years).	82%
D’Amico et al. (2013) Journal California, USA	Free Talk. A six-session motivational interviewing group intervention. It is a rolling group (can join at any time) and sessions last around 55 min. Children directed to this intervention as a sanction from Teen Court, which is a diversion method in which the case is heard and sentenced by a group of their peers. As part of this intervention they are all also randomly drug tested. Completion leads to their offence being removed from their criminal record.	Standard group, based on abstinence-based Alcoholics Anonymous approach. A rolling group, with six sessions lasting 55 min each. Children also diverted to this intervention via Teen Court, random allocation determined which they received.	N = 193 Teen Court with Free Talk n = 113 Teen Court with Usual Care (control group) n = 80	Age range = 14–18 years Mean age = 16.6 years Gender = 67% male Ethnicity = 45% Hispanic, 45% White, 10% Mixed and Other	Alcohol and other drug (AOD) offences including possession of AOD and driving under the influence.	Randomisation Block randomisation procedure—for each group of five children, three were allocated Free Talk and two allocated Usual Care. This was to ensure the groups had sufficient participants to run successfully.	Follow-up: 1 year after the first offence Outcome: Reoffending. Data obtained from the Probation department. Additional outcomes: Alcohol and marijuana use, experience of negative consequences due to alcohol or drug use and, using alcohol or drugs before sex and self-reported delinquency (non-offending behaviours). Additional outcomes available for 97% of the participants. Moderating variable: Completion (95% completed all six sessions in the timeframe).	Children in the Free Talk group were less likely to reoffend. 22% of the Free Talk group reoffended compared to 28% of the control group. However, this was not significant. The reoffending rate was lower for those who had been allocated Free Talk and had completed all six sessions (19%) whereas this did not impact the control group (28%).	92%
Hartsell and Novak (2022) Journal Florida, USA	Drug Courts and Teen Court. Drug Courts are voluntary programmes that provide substance use treatment and supervision for those who have been arrested and have difficulties with substance use. Teen Courts is a voluntary diversion in which a child’s case is heard in front of a court made up of their peers, who take on the roles of judge, attorneys and jury.	Children disposed to Probation (traditional formal processing) and children whose cases were dismissed (no processing).	N = 11,567 Teen Court n = 5604 Drug Court n = 636 Dismissed n = 3910 Probation (control group) n = 1147 An additional 19 participants were excluded due to incomplete data.	Age range = 14–17 years Mean age = 15.3 years Gender = 60% male Ethnicity = 45% White, 34% Black, 20% Latinx, 1% Other	Any non-felony offences. Also excluded any child that was detained for their offence and any child with more than six separate offences on their first arrest.	Quasi-experimental Propensity score estimation methods used to reduce selection bias.	Follow-up: 1 year from the point of initial referral Outcome: Reoffending. Measured using official records of re-arrest.	Children diverted to Teen Court were less likely to be rearrested than children formally processed by probation (ATE = 0.10, p = 0.002). There was no difference in re-arrests between children diverted to Drug Court and children on probation (ATE = −0.10, p = 0.155). Children in the diversion groups (Teen Court and Drug Court) and the formal processing group (probation) were all more likely to be rearrested than the children whose cases were dismissed (p < 0.001).	95%
Jacobsen (2013) Thesis Washington, USA	Teen Court. Peer volunteers conduct a hearing in a mock court setting. They are assigned to roles including judge, prosecutor, defence, court clerk, bailiff and jury. The process mirrors real court procedure, for example with the child meeting with their defence team prior, who then explain mitigating circumstances to the court. The process is supervised by an adult staff member from the juvenile court but they do not participate.	Alternative diversion. Community Accountability Boards (CAB) which involve a panel of three adult volunteers Staff Assignment (SA) where one juvenile court staff member meets with the child. In both types, child and parents attend and the child is expected to explain the circumstance that led to their referral. Sanctions are decided and if the child refuses them, they will be referred to court.	N = 205 Teen Court n = 31 Community Accountability Boards (control group) n = 88 Staff Assignment (control group) n = 86 An additional 15 participants were excluded due to not entering a diversion agreement, either because they refused or the programme rejected them.	Age range = 8–17 years Mean age = 14.9 years Gender = 61.0% male Ethnicity = 84.4% White and 15.6% Other	Minor offences, which includes ‘misdemeanor and gross misdemeanor charges as well as the occasional Class C Felony’. The most common offences was ‘minor in possession of alcohol’ (26%) followed by ‘theft in the third degree’ (22%).	Quasi-experimental Non-random assignment to conditions but all have to meet the same general criteria for diversion.	Follow-up: Variable. Sample consists of all files destroyed in the same year. Files destroyed when a child turns 18 or 23. Outcome: Reoffending. Defined as any subsequent conviction for a criminal offence, excluding traffic violations and non-offender matters such as truancy. Deferred prosecution cases or arrests not resulting in conviction also excluded. Additional outcome: number of subsequent convictions.	There was no significant difference in reoffending rate for the three different types of diversion, (F(2, 202) = 0.17, p = 0.84). Teen Court was found to be no more effective than the two control groups.	82%
Nadel et al. (2019) Journal Florida, USA	Civil Citation. Police officers determine whether to arrest a child, or present them with a civil citation. If so, they are assessed and assigned to an intervention programme. This can vary considerably depending on location. Most common intervention is community service, also more tailored programmes such as mental health counselling and substance abuse treatment. If the child successfully completes the intervention they avoid a formal record. If they fail to complete, they are arrested for the original offences.	Formal processing (i.e., arrest)	N = 102,024 Civil citation n = 31,620 Arrest n = 70,404	Mean age = 14.9 years Gender = 62% male Ethnicity = 35% Black, 17% Hispanic, 48% Other Different counties use civil citations to different extents—from not being used at all (Bradford County), to over 90% of eligible children receiving this outcome (Miami-Dade County)	Non-serious misdemeanour offences	Quasi-experimental All participants are eligible for civil citation due to type of offence. Demographic and offence type used as control measures.	Follow-up: 6-month, 1-year and 3-year periods. Outcome: Reoffending. Defined as whether a participant was arrested during the follow-up period.	Those who received a civil citation were significantly less likely to reoffend than those who were arrested by approximately 54%. The effect remained significant across all three follow-up periods; however, the odds of reoffending increased slightly with time.	95%
Norris et al. (2011) Journal Ohio, USA	Teen Court. Children are processed and sentenced by their peers (other children who have offended). Sanctions that can be given include: community service, verbal/written apologies, education and drug/alcohol screening. Admission of guilt required for engagement with teen court. Successful completion of the diversion avoids a criminal record, while failure to complete results in a referral to juvenile court.	Alternative diversion. Children diverted prior to the inception of teen court with the standard diversion offer where sanctions were given by probation officers.	N = 821 Teen Court n = 635 Standard Diversion (control group) n = 186	Age range = 11–18 years Mean age = 14.9 years Gender = 59.4% male Ethnicity = 72.5% White and 27.5% Other	Status offences (e.g., truancy, curfew, underage drinking) or low-level misdemeanours (e.g., shoplifting, trespassing, vandalism). Excludes any drug or violent offences.	Quasi-experimental design Historical comparison group. No matching but all control group screened for teen court and only a few cases rejected.	Follow-up: Between 7–8 years for control group and 2–8 years for teen court group. Follow up depends on date of referral. Outcome: Reoffending. Excludes traffic violations. Offences committed after turning 18 were not recorded. Moderating variables: Offence category, offence severity, sanction severity (number of sanctions the child received) and programme completion. Completion was 95.0% for Teen Court and 85.5% for regular diversion. Teen Court non-completion tended to be related to committing a new crime, failing to complete sanctions, being unable to attend or non-payment of fees. Regular diversion non-completion tended to be related to not admitting guilt or breaking the conditions of the diversion contract.	No statistically significant difference between reoffending rates in Teen Court and the comparison diversion group (χ2 = 0.602, df = 1, p = 0.438). The more sanctions a child was given in both groups, the more likely they were to reoffend. However, this effect disappeared when non-completers were removed from the sample. Offence category/severity had no effect.	82%
R. Y. Owen (2020) Thesis Utah, USA	Peer Court. A voluntary intervention based on restorative justice principles. Cases heard by a panel of 4–6 trained peer mentor volunteers from local schools who take on roles of the court (e.g., bailiff, clerk, chair). The child attends with a parent/guardian and both are asked questions regarding the event and their home/school life. All panel members then deliberate and decide on assignment of dispositions. One of the mentors is allocated to the child to provide weekly check-ins to support their progress. Adult supervisors are present but rarely have contact with the child themselves. Referrals generally received by schools but can also be made by parents, community police officers or juvenile court.	Formal processing. Children referred directly to the juvenile court.	N = 2682 Peer Court n = 1348 Juvenile Court (control group) n = 1334 73.6% of those referred chose to participate and of those, 61.9% successfully completed.	Age range = 8–19 years Mean age = 13.35 years Gender = 54.4% male Ethnicity = 55.7% Latinx, 23.6% White, 20.7% Other	Mainly minor infractions (e.g., truancy, tobacco use, disorderly conduct) however also includes more serious offences (e.g., bringing a weapon to school, assault).	Quasi-experimental Propensity score matching used.	Follow-up: 3 years from end of contact with the diversion. Outcome: Reoffending. Taken from court administrative data—any referral to juvenile court in the follow up period. Moderating variables: participation in and completion of the programme.	There was no difference in reoffending between those who participated in peer court and those processed through juvenile court (p > 0.05). Those who were referred to Peer Court but chose not to participate were found to be significantly more likely to reoffend than those who were processed through juvenile court (OR = 1.326, p < 0.05) Within those who chose to participate, completing peer court had no significant effect on reoffending.	91%
Sandøy et al. (2024) Journal Norway	Conditional waiver of prosecution (CWP). Enforced with a trial period from 6 to 24 months in which time any further offence will result in the resumption of the original case. Can also be enforced with specified conditions (e.g., intervention with healthcare or social services, drug/alcohol testing, substance use intervention).	Monetary fine	N = 3276 CWP with trial period n = 524 CWP with trial period and conditions n = 1016 Fine (control group) n = 1736	Age range = 15–17 years Mean age = 16.1 Gender = 67% male Ethnicity = 89% of sample born in Norway to parents also born there. 7% classified as ‘Immigrant’ and 4% as ‘Immigrant Background’.	Minor drug offences (drug use/possession)	Quasi-experimental Selection bias controlled for using statistical methods.	Follow-up: 5 years from the date of initial decision Outcome: Reoffending. Time in months from initial decision to next recorded offence.	Children given a fine had a shorter time until their next offence than those given a CWP. Effect was only significant for those CWPs with conditions (HR = 0.637, 95%CI = 0.46–0.88, p < 0.01). In the five years after the first offence, children given either CWP had fewer offences—20% reduction for standard CWP and 52% for CWP with conditions.	95%
Vooren et al. (2023) Journal The Netherlands	Halt programme. Voluntary, based on restorative justice. Personalised programme designed by practitioners depending on the severity of the offence and the child’s needs. Includes sessions with the child and their parents, for a maximums of 20 h. Programme consists of: community work, learning assignments, financial compensation and meeting with the victim(s). Key component is that they must apologise to their victim and victim tells them how they were impacted by the offence. Completing the assignments means they avoid a criminal record.	Placebo—enrolled in Halt programme (to avoid prosecution) but given no intervention.	N = 944 Halt diversion group n = 479 No intervention (control group) n = 465 A further 120 children took part in the intervention but did not consent to take part in the study.	Age range = 12–18 years Mean age = 14.52 years Gender = 71.4% male Ethnicity = 67.9% have two Dutch parents and 32.1% do not	Various ‘minor’, non-violent offences under the categories of public safety, firework offences, demolition, reckless behaviour and property offences.	Randomisation	Follow-up: 1 year Outcome: ‘Criminal involvement’ which is understood to mean reoffending. Taken from administrative data. Moderator variables: gender, age, parents born in The Netherlands. Secondary outcome: educational attainment. 78 participants in the Halt condition did not complete the programme. 14% of the sample could not be linked to the reoffending data, meaning they were removed. Statistical analysis demonstrated that this non-linkage had a minimal effect on the results.	Participation in the programme was found to increase the probability of reoffending by 39.3% in the following year. There was no evidence of different effects based on gender. It was also found to decrease educational attainment by 29.1%. This impact was found to be bigger for females than males.	92%
The Crossroads Study
Fine et al. (2017) Journal California, USA	Informal processing where the child’s case is dealt with by the Probation department. There are two options: Sanctioned Dismissal (informal probation without regular meetings, case dismissed on completion of minimal sanction such as writing an essay) or Diversion (supervision with a probation officer for 3–6 months with sanctions such as writing an essay and completing community service).	Formal court processing. Sent to the District Attorney who files a petition with the court. The child attends a court hearing in front of a judge and is given a sentence involving probation supervision and other sanctions, which are monitored by the court.	N = 359 Sanction and Dismiss = 54 Diversion n = 161 Formal court processing (control group) n = 144	Age range = 13–17 years Mean age = 15.3 years Gender = 100% male Ethnicity = 78.3% Latinx, 17.5% White and 4.2% Other	Moderate severity offences (e.g., vandalism, theft, assault) where the processing decision is variable and relatively unconstrained by sentencing laws. All charges that had a 0.35–0.65 probability of being formally processed.	Quasi-experimental Sample matched in that they are eligible for any outcome. Analysis controlled for key demographic variables.	Follow-up: 6 months Outcome: Reoffending. Taken from official records of arrest and measured using the Self-Report of Offending measure (Huizinga et al., 1991).	When controlling for variables (age, race, offence characteristics, socioeconomic status) and controlling for self-reported reoffending, children who were processed formally were 3.27 times more likely to be rearrested than children who received diversion and over 9.35 times more likely to be rearrested than children who received a sanction dismissal (p < 0.05). When controlling for race, age and parent’s education, children in all three groups had similar levels of self-reported reoffending.	100%
Beardslee et al. (2019) Journal and Thesis California, USA	As above	As above. Also includes a group that self-report to have committed the same types of offences but have never been arrested for them (no justice involvement group)	N = 631 No justice involvement n = 99 Sanction and Dismiss n = 109 Diversion n = 163 Court-ordered Probation (control group) n = 172 Adjudicated (control group) n = 88	Age range = 13–17 years Mean age = 15.53 years Gender = 100% male Ethnicity = 74.2% Hispanic, 23.3% White, 2.4% Black, 0.2% Other	As above	As above	As above Sample was impacted by potential participants refusing, their parents refusing or being ineligible due to not being able to schedule their interview within six weeks of arrest.	Formally processed children were more likely to be rearrested than those who received sanction and dismiss (t = 2.23, p = 0.025) or those who received diversion (t = 2.03, p = 0.042). All groups were significantly more likely to be arrested in the follow up period than the group who had never had justice involvement—none of whom were arrested in the follow-up. The diverted group had the lowest rates of self-reported reoffending, lower than those formally processed but also those who had had no initial justice system involvement. When splitting the groups further, there was no significant difference in self-reported reoffending in the diversion and court-ordered probation groups (the middle severity groups).	95%
Robertson et al. (2021) Journal and Thesis California (n = 532), Louisiana (n = 151) and Pennsylvania (n = 533), USA	As above	Formal court processing as above.	N = 1216 Diversion/informal processing n = 743 Formal court processing (control group) n = 473	Age range = 13–17 years Mean age = 15.29 years Gender = 100% male Ethnicity = 45.9% Hispanic, 36.9% Black, 14.8% White, 2.5% Other	As above	As above	Follow-up: 3 years Outcomes measured every six months for 3 years. Outcome: Reoffending. Taken from official records of arrest and measured using the Self-Report of Offending measure (Huizinga et al., 1991). Moderating variable: callous-unemotional traits For self-report outcomes: retention rates ranged from 95.48% at the 6-month follow-up to 91.34% at the 36-month follow-up with an average retention rate of 93.38% across the 6 follow-up points.	Overall, the formally processed group had more self-reported and recorded reoffending. However, after controlling for key individual, family and neighbourhood characteristics, decision at time of arrest (diversion vs. formal processing) did not predict self-reported reoffending. However, it did predict re-arrests. The effects of formal processing on self-reported reoffending was moderated by level of callous-unemotional traits, such that those with high callous-unemotional traits reoffended at high rates regardless of whether they were processed formally or diverted.	86%
Cauffman et al. (2021) Journal California (n = 532), Louisiana (n = 151) and Pennsylvania (n = 533), USA	As above	As above	As above	As above	As above	As above	Follow-up: 5 years Outcomes measured every six months from initial arrest to 3 years and then two annual reviews. Outcome: Reoffending. Taken from official records of arrest and measured using the Self-Report of Offending measure (Huizinga et al., 1991). Additional outcomes: incarceration, self-reported physical aggression, school/employment data, mental health and cognitive measured using psychometrics, self-reported peer delinquency and self-reported exposure to violence. The average non-attendance at follow-up interviews was 9%, with 70% of the sample missing no interviews, 15% missing one and 14% missing two or more. Those with missing data were more likely to be Black, but otherwise did not differ.	Children who were formally processed were more likely to be re-arrested (p < 0.001). Of those formally processed, 60% were re-arrested at least once during the 5 years compared to 43% of those informally processed. Formally processed children also more likely to be incarcerated, engaged in more violence, reported a greater affiliation with delinquent peers, reported lower school enrolment, were less likely to graduate high school within 5 years, reported less ability to suppress aggression, and had lower perceptions of opportunities than informally processed children.	100%
The Indianapolis Juvenile Restorative Justice Experiment
Hipple and McGarrell (2008) Journal Indiana, USA	Family Group Conferences. A form of restorative justice which involve a meeting between the offender (accompanied by his or her family or supporters) and the victim (also accompanied by supporters). Admission of guilt required to take part in the family conference. In this condition, the conference is facilitated by civilian volunteers.	Also Family Group Conference but facilitated by police officers instead of civilians.	N = 215 Civilian co-ordinator n = 132 Police co-ordinator (control group) n = 83 The sample included all conferences that the researchers were able to observe (54% of the total sample of the original study).	Age range 7–15 years Mean age = 12.5 years Gender = 63.3% male Ethnicity = 43.0% White and 36.8% Other No statistically significant differences between the two groups on age, gender, race and initial offence type.	Battery (or assault), trespass, mischief, conversion (shoplifting) or felony D theft.	Quasi-experimental. All participants meet criteria for family group conference, assignment to conditions is based on co-ordinator availability (not random). At the start of the study, more allocated to civilians and later more to police.	Follow-up: 2 years Outcome: Reoffending. Defined as any contact with the court (i.e., re = arrest) within the 2-year period following the qualifying date of initial arrest. Secondary outcomes: time until re-arrest, offence type and number of offences. Moderating variables: length of conference, police officer in uniform vs. not in uniform.	There was no significant difference in re-arrest rates between those who attended conferences facilitated by civilians (42.4% rearrested) compared to police (42.2% rearrested). There was also no difference in the time until re-arrest or number of re-arrests. All moderating variables were also non-significant (p > 0.05).	86%
Jeong et al. (2012) Journal and Thesis Indiana, USA	Family Group Conferences, as described above. Includes those delivered by police and civilians.	Participants assigned to one of twenty-three other diversion programmes. These include Teen Court (23.6%); Shoplifting Program (23.9%); Victim-Offender Mediation (21.7%); Community Service (14.9%); and other projects (15.9%).	N = 782 Family group conference n = 400 Alternative diversion (control group) n =382	Age range = 7–14 years Mean age = 12.7 years Gender = 62.2% male Ethnicity = 39.8% White and 60.2% Other	As above	Randomisation in one-month blocks.	Follow-up: 12 years Outcome: Reoffending. Defined as any recorded re-arrest from county court records. In cases where there was multiple charges, the most serious reoffence is recorded. Additional outcomes: reoffence type, time to reoffence. Moderating variable: Programme completion. Completion rate = 80.5% for the experimental group and 61% for the control. Non-completion included child/parent refusing to participate, dropping out or being rearrested.	There were no significant differences between the family group conference group and the control diversion group in reoffending prevalence or time to reoffence (p > 0.05).	92%
The Philadelphia Police School Diversion Program
Goldstein et al. (2019) Journal Pennsylvania, USA	Diverted from arrest at school. Instead of being taken to the police station and arrested following an offence at school, the child is referred to this programme. It involves being visited by a social worker at home and then being referred to voluntary community-based services (e.g., academic support, mentoring, drug and alcohol counselling) for support away from the criminal justice system.	Arrested (in the year prior to the diversion being implemented)	N = 2302 Diversion n = 1281 Arrested (control group) n = 1021	Age range = 10–22 years Mean age = 14.82 years Gender = 67% male Ethnicity = 73.9% Black, 14.4% Latinx/Hispanic, 9.4% White, 2.3% Other	Minor offences (e.g., disorderly conduct, marijuana possession) committed on school property	Quasi-experimental Sample matched in that they would have been eligible for the diversion had it been available when they committed their offence. Propensity score matching used to reduce effect of selection bias.	Follow-up: 2 years from initial incident. Outcome: Reoffending. Defined as any recorded re-arrest. Secondary outcome: time to rearrest	Diverted children demonstrated less reoffending in the two year follow up; however, after propensity score matching, these were no longer significant (29.2% re-arrest in the diverted group and 31.5% in the arrested control group).	91%
NeMoyer et al. (2023a) Journal and Technical report Pennsylvania, USA	As above	As above	N = 958 Diversion n = 427 Arrested (control group) n = 531	Age range = 10–22 Mean age = 15.26 Gender = 65.3% male Ethnicity = 75.4% Black, 24.6% Other	As above	As above	Follow-up: 5 years from initial incident Outcome: Reoffending. Defined as any recorded re-arrest. Secondary outcomes: time to rearrest, school discipline and academic achievement	Forty-one percent of the diverted sample were rearrested, compared to 48% of the control. When controlling for covariates, children who had been diverted were significantly less likely to be rearrested in the five years following their initial incident at school (OR = 0.73, 95%CI [0.54, 0.99], p = 0.043) than children who had not been diverted. There were lower rates of reoffending in the diversion group and the time to reoffence was longer. There were no observed differences in school-related outcomes.	95%

A total sample of 126,587 participants contributed to the data reviewed. They ranged in age from seven to 22 years with a mean age of 15.0 years. The number of participants in each study ranged from 193 to 102,024. All studies had a majority male sample however, the size of this majority varied greatly from 54.4% to 71.4% in studies without restrictions and 100% male in the Crossroads Study which restricted inclusion to boys only. Ethnicity of participants was well reported, with all studies referencing ethnicity or heritage and most studies including a relatively detailed ethnicity split. This demonstrates an improvement in modern studies as eight studies in D. B. Wilson et al.’s (2018) review did not include any information about ethnicity. However, some studies only reported simplified groups such as “White” and “other” or “Dutch parents” and “non-Dutch parents” which likely missed relevant detail. The split of ethnicities greatly varied depending on the location of the study, for example studies in California had a large proportion of Latinx participants, whilst studies in Philadelphia had predominantly Black participants.

The most common diversion intervention measured was teen court, which was evaluated in four separate studies (Hartsell & Novak, 2022; Jacobsen, 2013; Norris et al., 2011; R. Y. Owen, 2020), one of which also measured drug court (Hartsell & Novak, 2022). All other diversion interventions were unique. Two interventions were based on restorative justice principles (Jeong et al., 2012; Norris et al., 2011), one used ‘scared straight’ principles (Adamson, 2018), one used ‘civil citation’ which included community service hours, access to counselling, treatment, and other appropriate community resources (Nadel et al., 2019), and another used motivational interviewing (D’Amico et al., 2013). Those diverted in the Crossroads study experienced probation supervision (Petrosino et al., 2010; Schwalbe et al., 2012; D. B. Wilson et al., 2018; H. A. Wilson & Hoge, 2013) and those diverted in the Philadelphia School Diversion Program were referred to voluntary community-based services by a social worker (Goldstein et al., 2019; NeMoyer et al., 2023a). One study examined a ‘conditional waiver of prosecution’ which involved no immediate punishment but a period of 6–24 months in which any further offence will result in the original case being resumed (Sandøy et al., 2024). For some this also involved specified conditions such as attending appointments with healthcare or substance-use services.

Similarly, comparison groups varied, with six comparing the diversion to formal processing and six comparing it to another type of diversion. Formal processing included being arrested (compared to being diverted pre-arrest) (e.g., Goldstein et al., 2019) and being processed through a formal court system (e.g., Hartsell & Novak, 2022). Alternative diversions tended to be the standard diversion being offered in that area, for example diversion delivered by probation (Jacobsen, 2013). One study created a placebo group in which children were diverted but did not receive the intervention being studied (Vooren et al., 2023).

All studies used a binary measure of reoffending, taken from official records. Reoffending was defined differently across the studies, some used re-arrest data (Adamson, 2018; Goldstein et al., 2019; Hartsell & Novak, 2022; Nadel et al., 2019) one used court referral data (R. Y. Owen, 2020) and another used reconviction data (Jacobsen, 2013). Some studies did not provide a definition of reoffending, for example Vooren et al. (2023) reported to use data on “criminal involvement” (p. 698) with no information on how this was defined. Two papers indicated exclusion criteria, namely excluding traffic violations (Jacobsen, 2013; Norris et al., 2011). The follow-up periods in the studies varied greatly, from as short as six months to as long as 12 years.

Time to reoffend was examined by Sandøy et al. (2024) and in studies associated with the Indianapolis Juvenile Restorative Justice Experiment (Hipple & McGarrell, 2008; Jeong et al., 2012). These studies also measured the number of subsequent offences, as did Jacobsen (2013). Self-reported offending was used by the Crossroads Study and D’Amico et al.’s (2013) study, in addition to officially recorded reoffending.

3.2. Quality of Included Studies

The quality assessment scores calculated using the QualSyst tool (Kmet et al., 2004) are provided in Table 2. Overall, the quality of included studies was high, with all scoring above Kmet et al.’s (2004) suggestion of a “conservative” cut-off of 75%. The high quality was likely impacted by the inclusion criteria, which excluded papers that did not attempt to make the comparison groups comparable, for example by using matching or statistical controls. This means some low-quality studies may have been excluded within the screening process. Furthermore, the use of official records in all the studies significantly limited the potential for bias within data collection and significantly reduced attrition, both markers of quality. Attrition was minimal in most included studies, with only one study having significant attrition related to issues with linking data between databases, resulting in 14% of the sample being removed (Vooren et al., 2023). However, authors conducted statistical analysis which demonstrated this attrition had minimal effects on the results. The main attrition issue within this review related to the three studies had variable follow-up lengths, due to collecting reoffending data on the same date, regardless of when the child’s offence was committed. This meant that some participants had longer in which to reoffend than others, hence skewing the data. In Jacobsen’s (2013) study, this appears to have affected both the intervention and control groups equally, while in Adamson’s (2018) and Norris et al.’s (2011) studies the follow-up period for the comparison group appears to be significantly longer than for the intervention group. Additionally, in these same two studies, data was only accessible up until the participants turned 18, meaning data was missing for older children in the sample, an example of differential attrition. Neither study appears to have attempted to control for these issues, meaning their results and findings are likely to have been impacted by this bias.

Most quasi-experimental studies acknowledged that randomisation would be the “gold standard” for these types of studies but this is difficult in justice settings due to access to funding and ethical issues associated with giving different outcomes to children who have engaged in the same offending behaviour. Blinding, the practice of reducing bias by preventing researchers and/or participants from knowing which group they are allocated to (National Institute for Health and Care Excellence, 2025), was not used in any of the studies due to the challenges of disguising intervention type. There is a risk that police could have changed their arresting behaviour based on the knowledge of the study or their opinions on diversion; however, there was no evidence of this within the reports.

Selection bias is a key concern for quasi-experimental research, as there is a risk that the samples differ with respect to key variables (Stuart & Rubin, 2008). Propensity score matching (PSM) was used in The Philadelphia Police School Diversion Program (Goldstein et al., 2019; NeMoyer et al., 2023a) and two other studies (Hartsell & Novak, 2022; R. Y. Owen, 2020) as a way to reduce the impact of selection bias. PSM is a statistical method used to replicate a randomised control trial by matching individuals on a set of observable characteristics to better estimate the true effect of an intervention (Austin, 2011). One study used regression methods to reduce selection bias, having a similar effect of being able to reduce the effects of confounding variables (Sandøy et al., 2024). Historical comparison was used by two studies (Adamson, 2018; Norris et al., 2011), comparing the intervention group with a group of children who would have been eligible for the intervention had it been developed when they offended. The critique of this type of sampling is that there may be inherent differences between the groups; however, in this review both studies used regression methods to control for group differences, reducing this bias. The Crossroads Study ensured comparability of groups by selecting children with the same moderate severity offences that are eligible to be processed either formally or informally (Beardslee et al., 2019; Cauffman et al., 2021; Fine et al., 2017; Robertson et al., 2021), as did Nadel et al. (2019). However, authors do acknowledge there may have been some difference between the groups, so key demographic variables were statistically controlled to reduce this potential bias. Similarly, Jacobsen (2013) compared those allocated to teen court with those allocated to the traditional diversion offer, where the general eligibility criteria was the same. However, in this study no statistical controls were used to reduce selection bias, meaning there may have been differences between who was allocated to each group, which may have impacted reoffending.

There was evidence of sampling bias in some of the studies, for example, Beardslee et al.’s (2019) sample of children who had offended but not been arrested were recruited through peers of participants in the main study. However, most studies used pre-existing data which reduced bias and meant they were able to include large sample sizes from across a state (Hartsell & Novak, 2022; Nadel et al., 2019) or country (Sandøy et al., 2024). Sample sizes were generally well matched, though some studies did have imbalanced samples. For example, Hartsell and Novak’s (2022) four sub-samples differed in size from 636 to 5604.

Statistical reporting was mixed. Often, significance of a finding was discussed but effect size was not clearly identified (i.e., could only be found within a table). Power analysis was also rarely mentioned. In order to examine publication bias, the results from journal articles were compared to thesis documents. This did not indicate evidence of bias, as there were similar ratios of significant and non-significant results in the two types of report. However, publication bias may still be relevant, as there may be studies that have never made it into the public domain due to their findings (Hopewell et al., 2009).

The description of the details of the interventions was relatively poor in several studies, meaning experiments were not reproducible. It was not easy to distinguish key features within the text and prior knowledge of the system was often assumed. Often, there was good description of the process of diversion (e.g., attending teen court) but vague information about what happened next (e.g., the sanctions given to the child).

3.3. Data Synthesis

It was not appropriate to use meta-analysis due to the heterogeneity of the interventions and comparators examined in this review, and the variation in methodological study designs used (Lensen, 2023). If completed, an effect estimate would have no meaning or practical utility. Therefore, a narrative synthesis of the included studies is provided below. This synthesis uses the EMMIE framework (Johnson et al., 2015, see Table 1), intended to provide greater context to the review findings, as discussed previously.

3.3.1. Effects

The effects of the diversion intervention on the outcomes of interest were well reported. Studies that compared one type of diversion to another tended not to find a significant difference in reoffending (D’Amico et al., 2013; Jacobsen, 2013; Jeong et al., 2012). Meanwhile, diversion compared to formal processing, tended to show more promising results (Cauffman et al., 2021; Hartsell & Novak, 2022; Nadel et al., 2019; NeMoyer et al., 2023a; Sandøy et al., 2024). This suggests that diversion may be generally more effective than formal processing but that the specific type of diversion used is less important.

Four studies measured the effectiveness of a teen court intervention, one found it was more effective than formal processing (Hartsell & Novak, 2022) while another found no significant difference between teen court and formal processing (R. Y. Owen, 2020). Hartsell and Novak (2022) found it to be more effective than drug court, meanwhile two other studies found it to be no more effective than alternative diversion measures (Jacobsen, 2013; Norris et al., 2011). This difference in findings is likely related to the different implementation of teen court in different areas and, more importantly, the sanctions available for the children to be given.

In studies that compared diversion to non-intervention or less intensive intervention, namely dismissal (Hartsell & Novak, 2022), ‘sanction and dismiss’ (The Crossroads Study, e.g., Fine et al., 2017) or a placebo group who avoided prosecution but were not offered intervention (Vooren et al., 2023), these participants were always least likely to be rearrested. Similarly, the 2 h REACH intervention evaluated by Adamson (2018) was much less intensive than the control group (which included interventions such as substance abuse programmes and community service) and again this was found to be more effective with respect to re-arrest. Meanwhile, in Beardslee et al.’s (2019) study, their ‘undetected’ sample who had committed offences but never been arrested and thus had no intervention at all, were again the least likely to be rearrested. This provides support for the idea of minimal intervention being the most effective form of diversion.

Opposing this, Sandøy et al. (2024) found that a conditional waiver of prosecution was more effective when there were compulsory conditions attached, making it a more intensive intervention. However, this may be because the conditions were effective in their own right, for example substance use intervention or support provided by social services.

Interestingly, the Crossroads Study found that formal processing leads to higher rates of arrest and incarceration after five years and higher rates of self-reported violence but no significant difference in self-reported total offending.

3.3.2. Mechanisms

Most of the studies considered mechanisms within their report. Several studies considered the negative consequences of formal processing and indicated that diversion works by reducing exposure to these consequences. The stress and potential trauma of police/court processes (NeMoyer et al., 2023a), the interruption caused to the child’s life (Beardslee et al., 2019), increased contact with peers who engage in delinquent behaviour (Robertson et al., 2021), exposure to unsafe settings (Beardslee et al., 2019) and stigmatisation (Goldstein et al., 2019) were all discussed as negative consequences that diversion allows children to avoid. Stigmatisation and labelling were discussed by several authors (e.g., Adamson, 2018; Cauffman et al., 2021; Nadel et al., 2019). They consider that the stigma associated with formal justice system involvement can reduce opportunity for prosocial and socially normative opportunities, leading children to become ostracised and pushing them further into criminal behaviour (Goldstein et al., 2019). Furthermore, it can lead to the development of “delinquent” self-identities, which become self-fulfilling as the child behaves in the way that is expected of them (Beardslee et al., 2019; Nadel et al., 2019). The impact of this stigma is also amplified by receiving a criminal record, the impact of which can last into adulthood (Cauffman et al., 2021). Adamson (2018) relate their findings to the fact that REACH is a very short intervention (2 h), meaning children are not embedded in the system and thus the impact of stigmatisation is limited. Similarly, Hartsell and Novak (2022) suggest teen court may be perceived as less formal than drug court or probation, making it less stigmatising and “ensnaring” (p. 1838) and thus more effective.

R. Y. Owen (2020) notes that, while their aim is to reduce this negative impact, diversion interventions can also lead to labelling, as children are often still involved in a justice system process which may have associated stigma. Jacobsen (2013) notes that, while teen court is intended to feel different to formal court, there are significant similarities, which may reduce its effectiveness. Beardslee et al. (2019) also highlight that, while there are differences between the diversion and formal processing experienced by their participants (e.g., shorter time supervised, not having to attend court, no time in police custody, smaller fines), there are also significant similarities between the two as both groups are given community supervision and asked to complete sanctions. Diversion may therefore have been experienced as similar to formal processing in some studies, hence limiting its effectiveness.

Another frequently discussed mechanism to explain why those children who are diverted or dismissed are less likely to be rearrested is the increased monitoring and surveillance experienced by formally processed children (Beardslee et al., 2019; Fine et al., 2017; Goldstein et al., 2019; Hartsell & Novak, 2022; NeMoyer et al., 2023a). Hartsell and Novak (2022) explain that those who are diverted will usually be monitored less closely and for less time than those formally processed, meanwhile those who are dismissed will experience even less monitoring and have no risk of being arrested for technical or administrative violations. They suggest that they may have found teen court to be more successful than drug court due to the different levels of monitoring involved in the two. Beardslee et al. (2019) suggest those with more surveillance may simply be more likely to have their illegal behaviour detected and reported, supported by their finding that none of the ‘undetected’ group were arrested during the follow-up period, despite self-reporting the same level of offending in the follow-up period. They suggest increased surveillance does not just come from the justice system, but also the school, neighbours, parents of peers, etc., who might be more likely to interpret their behaviour as worthy of reporting, with the knowledge that they have been involved with the system.

Deterrence theory, which suggests that children should receive punishment for their offending behaviour, to learn not to repeat it (G. Becker, 1968), was discussed as a mechanism in two studies (Sandøy et al., 2024; Vooren et al., 2023). This theory was proposed by Vooren et al. (2023) as their Halt intervention was found to be significantly less effective than a placebo, leading the authors to suggest that completing the Halt programme did not act as a sufficient deterrent to reoffending due to the nature of the intervention, which focused on restorative justice rather than punishment. They suggest that the restorative focus may have led children to feel ‘justified’ (p. 705) in their offending after completing the intervention as the intervention is considered to have cancelled out any wrongdoing. However, what the authors fail to consider is it may not be that the Halt programme increased risk of reoffending, rather that the placebo group, by not having to attend any intervention, experienced the positive effects of non-intervention. Meanwhile, Sandøy et al. (2024) suggest the success of the conditional waiver of prosecution with conditions may be related to it being experienced as more intrusive and unpleasant than the other options and thus may have a deterrent effect. However, they also recognise it may instead be because the conditions associated allow the child to access effective services.

3.3.3. Moderators

Most of the included studies explored at least one moderating variable, though often non-significant results were found. The impact that completion of the diversion intervention had on effectiveness was explored by three studies. R. Y. Owen (2020) found that the decision to participate in teen court was predictive of reoffending, but whether the child successfully completed all the associated sanctions was not. Meanwhile, Norris et al. (2011) found that those who completed teen court were significantly less likely to reoffend then non-completers. D’Amico et al. (2013) also found that completing all sessions of motivational interviewing slightly reduced likelihood of reoffending. They suggest that completing the sessions could increase the effectiveness of the programme, or non-completion could simply be an indicator of those who are not ready or willing to make positive change. Norris et al. (2011) explored whether the number of sanctions given by a teen court moderated the effectiveness, finding that the more sanctions a child was given, the more likely they were to reoffend. However, this effect disappeared when non-completers were removed, which they suggest may be because too many sanctions increases the risk of drop-out as participants feel alienated by the process. Alternatively, they also suggest it could be that those giving the sanctions had identified those at greatest risk of reoffending and given them more sanctions because they felt they needed the additional intervention.

Within the Crossroads Study, Cauffman et al. (2021) found formal processing was related to lower perceptions of opportunities, less ability to suppress aggression, a greater affiliation with delinquent peers, and slightly lower odds of graduating high school. As those formally processed were found to be more likely to reoffend, any of these factors could be considered moderators. Meanwhile, within the same longitudinal study, Robertson et al. (2021) found that callous-unemotional traits made participants more likely to reoffend, regardless of whether they were diverted or processed formally.

Adamson (2018) found that age was a significant moderator, in that REACH was only effective in reducing re-arrest for older children (age 14–17 years) and not younger (10–13 years) when compared to alternative diversion offers. They suggest that having a first offence at a young age may be indicative of more complex problems and therefore a longer intervention may be required for these children, rather than the 2 h minimal intervention. Similarly, Norris et al. (2011) found that younger children in their sample (age 11–14 years) did not benefit from either teen court or the comparison diversion. They suggest that perhaps these children are too young to ‘take it seriously’.

Vooren et al. (2023) found that the Halt programme was less suitable for those who had committed solo offences compared to group offences. They suggest there may be psychological differences between those who engage in solo, compared to group, offending behaviour. This, alongside other features of a child’s offending behaviour, is worthy of further research.

Several studies also made suggestions regarding moderating variables in their discussion sections, despite not examining these in the study. For example, Jeong et al. (2012) suggested the reason the positive impact of family group conferences did not persist long-term is because it is a short one-hour-long intervention, suggesting that if the conferences were repeated they may have had a greater impact. Similarly, Vooren et al. (2023) suggest that the time between offence and sanction may impact effectiveness, suggesting the poor performance of Halt may have been related to the wait of up to two months between offence and intervention. While Hipple and McGarrell (2008) found the type of facilitator was not a significant moderator of the relationship between diversion and reoffending, they suggest that factors such as how seriously the child takes the intervention and how much the facilitator sticks to their training might moderate effectiveness. All of these potential moderators could be worthy of future investigation.

3.3.4. Implementation

None of the included studies discussed intervention implementation in their reports. This kind of information is usually excluded from published journal articles due to limitations on word count. However, the Philadelphia Police School Diversion Program does have a separate publication, not included in this review, which provides a full implementation outcomes analysis (NeMoyer et al., 2025).

3.3.5. Economics

Financial benefits of diversion were referred to in many of the studies, though usually in very minimal detail (e.g., Cauffman et al., 2021). The relative cost savings of teen court were mentioned by two papers, which both considered the economic benefit of an intervention that relies mainly on volunteers, compared to traditional court processing that has high staffing costs (Norris et al., 2011; R. Y. Owen, 2020). The Philadelphia Police School Diversion Program papers both briefly mention economics but the authors have conducted a full cost–benefit analysis that is published separately (NeMoyer et al., 2023b). They found that the programme produced an annual cost saving of approximately $1.9 million in its fifth year of operation, when comparing the cost of running the diversion programme to the direct savings associated with arrests and the indirect savings related to victim and lost productivity costs.

Interestingly, one paper briefly considers the economic cost of net-widening, noting that including cases in diversion who would have been better served by non-intervention is not cost effective (Norris et al., 2011). This is an important point to note, given there is often significant emphasis on the economic benefits of diversion rather than costs.

4. Discussion

The aim of this systematic review was to synthesise evidence related to the effectiveness of pre-court diversion interventions for children in conflict with the law for the first time. To the authors’ knowledge, it is the first systematic review to consider diversion measures specifically for this population. The included studies are varied with respect to the interventions evaluated, comparison groups used, and results observed, though there are also some trends within the findings.

Overall, the findings provide tentative support for the use of diversion when compared to formal processing, with several studies indicating lower levels of reoffending in these groups. In particular, the findings indicate that the least intensive intervention (including no intervention at all) is usually the most effective at reducing reoffending rates. There was no particular type of diversion measure that appeared most effective and often study design inhibited this type of comparison between studies.

Reducing exposure to the negative consequences of formal processing, namely stigmatisation along with others, was the most commonly described mechanism. This aligns with the literature discussed within the Introduction of this paper, which identifies the function of diversion in reducing stigma and subsequently reducing reoffending (Bramley et al., 2019). The risk that diversion measures can themselves be stigmatising, and perhaps not as different to formal processing as they intend to be, was also discussed. Consideration was given to whether diversion may reduce reoffending rates due to these children being less closely monitored, and thus affecting detected offending rather than behaviour. This was supported by Beardslee et al.’s (2019) findings when comparing those whose offending had gone ‘undetected’ to those who had been arrested. The results of the review also suggest moderators that should be considered include age, intervention uptake and completion, whether the offence was perpetrated as part of a group and individual differences related to personality, namely ‘callous-unemotional’ traits.

The studies were all assessed as relatively high quality and there was no observed difference in the findings of those assessed as the highest quality and those assessed as lower quality, nor between those studies reported in journals compared to thesis documents. While this is a comprehensive review, there are some limitations, outlined below, which provide guidance for areas of future research and review.

4.1. Limitations and Recommendations for Research

The review only included papers from 2008 onwards, due to the changes in diversion policy that happened in the UK during this time, and the fact that there are many existing reviews that have reviewed earlier research. However, given this review’s increased specificity, with the addition of restricting to only studies of FTEs, it may have been beneficial to have included older papers (e.g., McGarrell & Hipple, 2007) for a more complete picture of the literature base.

None of the included studies were conducted in the UK and the majority were conducted in the USA, where criminal justice practices differ significantly. However, this could also be considered a strength of the review, as it provides insight into alternative approaches which the UK could consider adopting. Given the amount of diversion activity that occurs across England and Wales (Robin-D’Cruz & Tibbs, 2019), the lack of inclusion in this review indicates a need for more evaluative research into current practice. It would also be interesting to learn from research in countries which have systems better aligned with Child First principles, such as Scotland (Hazel, 2008) and New Zealand (New Zealand Government, 2024), than those included in this review.

During this review, it was at times challenging to align the definition of “first-time entrant” used in England and Wales (see Box 1) with the population described in the papers. Whilst the definition of FTE is precise, in practice there is some ambiguity regarding this classification, and it is generally a term used within key performance indicator data rather than within services. Since the definition requires that a child must receive a caution or conviction, children may have interactions with the police long before they are formally classified as an FTE. There is also no limit on the number of police contacts or diversion outcomes a child can experience before being classified as an FTE. Because this definition is based on outcomes, any changes to national guidelines concerning decision-making related to these outcomes will affect which children are classified as FTEs. The screening of the papers also highlighted that the term “diversion” is poorly defined and used inconsistently. Other terms used within papers are also poorly defined (e.g., “post-adjudication”) and an intervention with the same name (e.g., drug court) often looks different and is used with different types of children in different areas. While this is somewhat to be expected given that the review explores international literature, future research studies should endeavour to clearly define their intervention and its purpose. Additionally, the screening process of this review highlighted that often diversion interventions are offered to children with a wide range of offending histories, including those who have previously been in custody. While it is positive for all children to be given support to find pathways out of the system, it does seem that “true” diversion may not be being measured in many of these studies. When considering how FTE rates can be reduced in England and Wales, it would be beneficial for future research to focus specifically on this group.

While it is a strength that all studies used an objective, quantitative measure of reoffending in the form of police or court data, this does limit the findings of the review as proven reoffending significantly underrepresents true offending behaviour (Ministry of Justice, 2016), and detection rates differ for different types of offences (N. Owen & Cooper, 2013) which can bias findings. Future research should include self-report measures of offending, similar to the Crossroads Study (e.g., Cauffman et al., 2021) which are not impacted by detection and sanctioning. Furthermore, this will help identify whether diversion measures truly have a positive effect on children’s offending behaviour or simply reduce the likelihood that they are detected, due to being less closely monitored.

As well as binary reoffending data, whether from records or self-report, it is important for studies to consider additional reoffending variables including time to reoffend, number of reoffences and severity of reoffending (Jeong et al., 2012). Desistance can be a lengthy, non-linear process, meaning focusing only on binary reoffending is oversimplified and may mean important markers of success are missed (Bateman & Wigzell, 2020). It is important to have realistic expectations of interventions in youth justice, where the recipients are children who often live in chaotic environments, meaning expecting no ‘slip ups’ may be unrealistic. Looking at long-term data allows for a more accurate picture of the outcomes to be captured, exemplified by studies of the Philadelphia School Diversion Program in which short-term outcomes displayed no significant difference, but long-term outcomes did (Goldstein et al., 2019; NeMoyer et al., 2023a). Additionally, non-offending outcomes are also highly important and often overlooked (K. Wong, 2019). These could include the child’s wellbeing, their educational performance, educational adjustment, use of leisure time, substance use or relationships with their parents (Hoge, 2016). Future studies should consider including these additional markers of success and future systematic reviews may consider exploring a broader range of outcomes to measure effectiveness.

Future research should further explore the mechanisms involved in the relationships between diversion and outcomes for children (Smith, 2024). Hipple et al. (2014) have explored features of Family Group Conferences that improve outcomes, and more research similar to this would be highly useful for practice. Furthermore, it would be interesting for research to explore the meaning that participants make of the intervention they engage with, in order to provide more insights into possible moderators. For example, Norris et al. (2011) suggest that perhaps feeling rejected or embarrassed in a teen court makes it less effective (Norris et al., 2011). It could be beneficial to explore what types of interventions work for children with different characteristics, so that children can be more effectively placed into the diversion interventions that are most likely to work for them (Norris et al., 2011).

This review highlighted the absence of consideration of how diversion interventions are implemented within the included studies. It would be beneficial for future research to evaluate processes as well as outcomes (Mears et al., 2016).

4.2. Conclusions and Implications for Policy and Practice

Overall, this systematic review provides support for minimal intervention and maximum diversion. It tentatively supports diversion into social/healthcare services while highlighting the need for further research into what works for diverting FTEs away from the youth justice system to ensure welfare intervention efforts are not inadvertently punitive or stigmatising. No studies found diversion to be significantly less effective than formal processing, but there is more research to be done to examine how effective diversion is for FTEs and which types of diversion work best. Generally, the findings of this review indicate that diversion practices should continue, as they have the potential to reduce reoffending, as well as being able to process cases more quickly, which is preferable for victims (Crest Advisory, 2022), and with less cost (The Centre for Justice Innovation, 2016). However, it is important to be conscious of the risk of net-widening which should be avoided at all costs.

Of concern is the indication that diversion may simply be effective in removing children from the spotlight of the criminal justice system, so that their offending goes undetected, rather than truly impacting their behaviour. This was supported by findings within the Crossroads Study, in which arrest data was associated with the type of system processing, but overall self-reported offending behaviour was not (Beardslee et al., 2019; Cauffman et al., 2021). This brings into question the findings of all previous studies of diversion interventions which have relied on officially recorded data alone, highlighting the importance of future research considering both self-reported and recorded offending to ensure diversion interventions are facilitating children to make changes to their behaviour, rather than simply allowing them to go “under the radar” (Beardslee et al., 2019, p. 342). Further longitudinal research may also provide clarity regarding whether the initial decision to divert a child, while perhaps initially only keeping their further offending from being detected, may in the long-term lead to changes in behaviour due to not becoming entrapped within the system. This is an important area for future research as there is a risk in this field that practice is driven by meeting targets, but this may not translate to policy interventions that actually benefit children.

The findings of this systematic review should be considered within the context of changes to policy and practice in England and Wales, exploring how elements of effective interventions internationally could be integrated into current diversion offers. For example, consideration may be made regarding how diversion offers could involve less direct intervention with children to reduce stigmatisation and any necessary intervention could be delivered by external services. Furthermore, better evaluation of existing diversion in England and Wales is required in order to establish how and where improvements are required. For example, the vast majority of YJSs in England and Wales have a point of arrest diversion service (Robin-D’Cruz & Tibbs, 2019) but very few have been evaluated and when they have there have often been issues of small sample sizes and unmatched groups (Haines et al., 2013). This highlights the need for funding for research in this area to establish what current practice is working well, so that good practice can be shared, rather than services working in silos.

Adopting best practice in diversion will also require collaboration with associated organisations. For example, the police, who hold significant responsibility in decision-making, tend to be driven by performance indicators related to number of charges (College of Policing, 2023), which diversion measures inherently jeopardise. Hence, a shared understanding of the benefits of diversion and clear policies regarding its implementation will reduce the risk of disagreements between police forces and YJSs. Similarly, HMIP will need to have clear evidence-based criteria for assessing services diversion measures, so that they are not inadvertently encouraging poor practice.