PROTOCOL: The effects of resettlement/re-entry services on crime and violence in children and youth: A systematic review

Custodial phase

The pre-release stage of aftercare programs provides young people in custody the opportunity to build a strong foundation of skills and behaviors necessary for easing the transition post-release. Research suggests that successful reintegration is more likely if release planning and intervention programming begin when the individual is admitted into custody, before the start of the transition (Byrnes et al., 2002; Petersilia, 2004; Travis & Petersilia, 2001). In accordance with this principle, the case management component of the aftercare model should ideally commence at entry into custody and continue throughout the supervision period. Important in the case management process is an element of co-creation; more specifically, by allowing young people to have some self-determination in their plans for the future, the development of a pro-social identity may be strengthened (Bateman & Hazel, 2015b). Eligible offenders will generally undergo risk, vulnerabilities, and needs assessments and engage in discussions surrounding their personal goals and motivations, which will inform their case planning with respect to level of supervision and the specific interventions provided while in custody. These interventions will typically focus on preparing the offender to live pro-socially post-release, and may include psychological or behavioral interventions (e.g., cognitive behavioral therapy, anger management) or substance use treatment, mental and physical health services, education services, and discharge planning, such as identifying a safe living situation and/or employment preparation and training (e.g., Hampson & Tracey, 2016; James et al., 2013). Further, case planning engages informal networks of support before release, in order to provide young people with involved and coordinated support throughout the entire resettlement process (Bateman & Hazel, 2015b).

Transition phase

The immediate period of transition is a critical point in the overall reintegration process. Research suggests that re-offense is most likely to occur shortly after release, indicating the need for intensive and highly structured transition planning, and immediate access to matched services (Altschuler & Armstrong, 1998; Goodstein & Sontheimer, 1997; Griffin, 2004). As such, aftercare models typically involve intensive case management and supervision immediately after release, with frequent contact between case managers and community correctional staff (Altschuler & Armstrong, 1998; Greenwood & Turner, 1993; Griffin, 2004; James et al., 2016). This can include check-in meetings, drug testing, family conferencing, and various mentorship activities (Bergseth & McDonald, 2007; Bouffard & Bergseth, 2008; Sontheimer & Goodstein, 1993; Wiebush, 1993). Agencies from all stages of the resettlement process are encouraged to attend meetings during this transitional phase, in order to provide coordinated and seamless care which is unaffected by the spatial changes involved in release (Bateman & Hazel, 2015b; Hampson & Tracey, 2016).

Post-release phase

After the initial transition to the community has occurred, young custody-leavers are generally required to maintain contact with their case manager to ensure continued supervision and service participation, including reassessment of needs as necessary (Altschuler & Armstrong, 1998; Griffin, 2004; James et al., 2013). The intensity of this care gradually decreases over time as the young person becomes more established within the community (Griffin, 2004). To successfully implement this phase, it is critical that the resettlement team has established strong connections with community-based services and resources to ensure that youth receive the appropriate continuity of care to prevent any lapses or delays in service delivery and subsequent skill and behavioral regression (Altschuler & Armstrong, 2001; Mathur & Clark, 2014). The scope of the available services should be broad in an effort to meet the various needs of the adolescents and ensure that services provided in the community parallel those provided in custody (Altschuler & Armstrong, 1998; Byrnes et al., 2002).

Types of studies

Types of participants

Eligible program participants for this review include children and young people who have served or are currently serving a custodial sentence, with an age range of 10–18 years. Studies will also be eligible for inclusion if the participant age range exceeds 18 years (up to 21 years), provided that the average age of study participants is at or below 18. There are no restrictions based on gender, ethnicity, or offense.

Excluded participants include offenders being offered a specialized program based on their exceptional/unique characteristics; for example, those with serious mental health diagnoses (e.g., intellectual disability). Findings from such studies would not be generalizable to broader groups of young people participating in aftercare/resettlement programs.

Types of interventions

Eligible types of interventions include any aftercare/resettlement program that takes place while the young person is in custody, during their transition to the community, and/or when they have returned to the community. The overarching intervention objective should be to promote successful community reintegration and reduce reoffending. Program components may vary across interventions (e.g., intensive case management, housing support, education, counseling), and interventions are not required to contain any particular type of component so long as the intervention is framed as an aftercare/reentry program as opposed to, for example, a more general skills training program or substance abuse recovery program.

Boot camp interventions, wilderness therapy interventions, and intensive supervision programs are excluded from eligibility, as are probation programs that do not meet the definition of resettlement approaches (e.g., those that include only supervision without additional supports and resources). In order to increase the generalizability of findings, programs limited to very specific types of offenders will be excluded, such as specialized programs targeting re-entry for juvenile sex offenders.

Types of outcomes

Duration of follow-up

No restrictions will be placed on the duration of participant follow-up; however, the timing of measurement of quantitative outcomes for objectives 1–3 will be kept as homogenous as possible to maximize between-study commensurability. For example, if the large majority of studies use a 6-month follow-up period, the main outcome time point across all studies will be six months when available (even if some studies include longer periods).

To the extent that some studies include multiple follow-up points, such as 6-month, 1-year, and 2-year, we will also code outcomes for longer follow-ups and attempt analyses of smaller pools of studies with different outcome measurement time periods (longer and/or shorter).

Types of settings

No restrictions will be placed on intervention settings; however, depending on the different types of settings uncovered in the search, we may choose to limit analyses to groups of interventions with commensurate settings only. In particular, we anticipate a group of studies based on interventions delivered only in a community setting, and another group based on interventions which begin during custody and continue post-offender release. Careful consideration will be given to the commensurability of interventions offered in these different categories of settings, and a determination will be made regarding the appropriateness of pooling study findings in analyses for objectives 1–3.

A similar approach will be used for objectives 4 and 6. For example, implementation factors may differ based on location of the intervention; for example, attendance may be a larger barrier in the community than in a custodial setting. As such, qualitative analyses may be limited to summarizing findings from evaluations of interventions from comparable locations. Similarly, intervention costs may differ substantially for interventions delivered in a hybrid prison/community setting than solely in the community, and results will be pooled accordingly.

In addition, as social/political systems and approaches to criminal justice vary substantially across nations, to enhance commensurability and practicality of the findings the study must have been conducted in Canada, the United States, Australia, New Zealand, the U.K., or a Western European country.

Other criteria

Studies will be limited to research published since 1992 (i.e., 30 years; to focus on more contemporary types of aftercare/resettlement approaches), and to research published in English, French, or German (given the evaluation team's existing expertise). If studies in other languages are identified through the search processes, they will be retrieved and documented and a decision concerning the necessity for translation services will be made. As noted previously, studies must have been conducted in Canada, the United States, Australia, New Zealand, the U.K., or a Western European country. No restrictions will be implemented based on publication status or type (i.e., peer-reviewed; government report versus journal article).

Electronic searches

To apply a comprehensive search strategy, the following set of key terms will be used in a Boolean search, reflecting four primary constructs: 1) youth/young person, 2) offending, 3) aftercare/resettlement intervention, and 4) evaluation, theory, or cost:

Construct 1 : (youth* OR juvenile* OR adolesc* OR teen* OR “young offender*” OR “young people” OR “young person*” OR child* OR “early offender”)

Construct 2 : (crime* OR criminal* OR devian* OR violen* OR delinquen* OR offend* OR offense* OR offence* OR recidiv* OR reoffen* OR breach* OR “technical violation*” OR arrest* OR convict* OR charge* OR incarcer* OR petition* OR adjudicat* OR caution* OR “compliance during supervision” OR “return to custody”)

Construct 3 : (diversion* OR divert* OR probat* OR parole OR aftercare OR resettlement OR reentry OR “re-entry” OR “after custod*” OR supervis* OR “graduated sanction*” OR “intermediate sanction*” OR “early release” OR “pretrial release” OR “supervised release” OR wraparound OR reintegrat* OR throughcare OR “local authority care” OR “security training cent*” OR “care leaver*” OR “detention and training order” OR “youth offending team”)

Construct 4 : (evaluat* OR effect* OR impact* OR outcome* OR trial* OR treat* OR program* OR randomi* OR experiment* OR assess* OR process OR implement* OR fidelity OR “proof of concept” OR “case study” OR “focus group” OR “pilot study” OR “qualitative” OR “formative evaluation” OR “cost benefit” OR “cost effectiveness” OR “cost analysis” OR “benefit cost” OR “theory of change” OR “program* theory” OR “program model” or “logic model” OR “action theory” OR “causal map”)

The complete search strategy will be applied to a series of 26 electronic databases, including various social sciences databases and those more specifically focused on criminology and criminal justice. Terms will be searched in the “abstract,” “title,” and “subject” search fields. The following databases will be searched; those on the same search platform will be combined:

EBSCO: ○

Academic Search Premier

The complete search string example is provided below for databases in ProQuest.

((abstract(youth* OR juvenile* OR adolesc* OR teen* OR “young offender*” OR “young people” OR “young person*” OR child* OR “early offender”) AND abstract(crime* OR criminal* OR devian* OR violen* OR delinquen* OR offend* OR offense* OR offence* OR recidiv* OR reoffen* OR breach* OR “technical violation*” OR arrest* OR convict* OR charge* OR incarcer* OR petition* OR adjudicat* OR caution* OR “compliance during supervision” OR “return to custody”) AND abstract(diversion* OR divert* OR probat* OR parole OR aftercare OR resettlement OR reentry OR “re-entry” OR “after custod*” OR supervis* OR “graduated sanction*” OR “intermediate sanction*” OR “early release” OR “pretrial release” OR “supervised release” OR wraparound OR reintegrat* OR throughcare OR “local authority care” OR “security training cent*” OR “care leaver*” OR “detention and training order” OR “youth offending team”) AND abstract(evaluat* OR effect* OR impact* OR outcome* OR trial* OR treat* OR program* OR randomi* OR experiment* OR assess* OR process OR implement* OR fidelity OR “proof of concept” OR “case study” OR “focus group” OR “pilot study” OR “qualitative” OR “formative evaluation” OR “cost benefit” OR “cost effectiveness” OR “cost analysis” OR “benefit cost” OR “theory of change” OR “program* theory” OR “program model” OR “logic model” OR “action theory” OR “causal map”))

OR (title(youth* OR juvenile* OR adolesc* OR teen* OR “young offender*” OR “young people” OR “young person*“ OR child* OR “early offender”) AND title(crime* OR criminal* OR devian* OR violen* OR delinquen* OR offend* OR offense* OR offence* OR recidiv* OR reoffen* OR breach* OR “technical violation*” OR arrest* OR convict* OR charge* OR incarcer* OR petition* OR adjudicat* OR caution* OR “compliance during supervision” OR “return to custody”) AND title(diversion* OR divert* OR probat* OR parole OR aftercare OR resettlement OR reentry OR “re-entry” OR “after custod*” OR supervis* OR “graduated sanction*” OR “intermediate sanction*” OR “early release” OR “pretrial release” OR “supervised release” OR wraparound OR reintegrat* OR throughcare OR “local authority care” OR “security training cent*” OR “care leaver*” OR “detention and training order” OR “youth offending team”) AND title(evaluat* OR effect* OR impact* OR outcome* OR trial* OR treat* OR program* OR randomi* OR experiment* OR assess* OR process OR implement* OR fidelity OR “proof of concept” OR “case study” OR “focus group” OR “pilot study” OR “qualitative” OR “formative evaluation” OR “cost benefit” OR “cost effectiveness” OR “cost analysis” OR “benefit cost” OR “theory of change” OR “program* theory” OR “program model” OR “logic model” OR “action theory” OR “causal map”))

OR (subject (youth* OR juvenile* OR adolesc* OR teen* OR “young offender*” OR “young people” OR “young person*” OR child* OR “early offender”) AND subject (crime* OR criminal* OR devian* OR violen* OR delinquen* OR offend* OR offense* OR offence* OR recidiv* OR reoffen* OR breach* OR “technical violation*” OR arrest* OR convict* OR charge* OR incarcer* OR petition* OR adjudicat* OR caution* OR “compliance during supervision” OR “return to custody”) AND subject (diversion* OR divert* OR probat* OR parole OR aftercare OR resettlement OR reentry OR “re-entry” OR “after custod*” OR supervis* OR “graduated sanction*” OR “intermediate sanction*” OR “early release” OR “pretrial release” OR “supervised release” OR wraparound OR reintegrat* OR throughcare OR “local authority care” OR “security training cent*” OR “care leaver*” OR “detention and training order” OR “youth offending team”) AND subject (evaluat* OR effect* OR impact* OR outcome* OR trial* OR treat* OR program* OR randomi* OR experiment* OR assess* OR process OR implement* OR fidelity OR “proof of concept” OR “case study” OR “focus group” OR “pilot study” OR “qualitative” OR “formative evaluation” OR “cost benefit” OR “cost effectiveness” OR “cost analysis” OR “benefit cost” OR “theory of change” OR “program* theory” OR “program model” OR “logic model” OR “action theory” OR “causal map”))

Additional web sources will also be searched to identify relevant unpublished works not indexed in academic databases (e.g., technical reports, conference papers, and independent research projects) in an effort to minimize the risk of publication bias. Specifically, we will identify relevant conference and meeting abstracts via the meeting archives of the American Society of Criminology, the British Society of Criminology, the European Society of Criminology, and the Conference Proceedings Citation Index. In addition, the websites of several relevant organizations will be searched, including the Australian Institute of Criminology, Confederation of European Probation, Department of Justice Canada, European Crime Prevention Network, Home Office UK, Ministry of Justice UK, National Institute of Justice, New South Wales Bureau of Crime Statistics and Research, Office of Juvenile Justice and Delinquency Prevention, US Department of Justice, and the Youth Justice Board. Last, a number of open access search engines will be searched, including the Directory of Open Access Journals (DOAJ), Bielefeld Academic Search Engine (BASE), and CrimRxiv.

Search terms on websites, meeting archives, and open access engines will include a reduced set of the terms from the database search strategy and will be adjusted across different sites and search restrictions accordingly. If the site allows for advanced search strategies, we will use multiple combined terms; if the site allows for only basic search strategies, we will attempt a series of searches such as (“young person” and resettlement) or (youth and aftercare).

Searching other resources

Additional search methods will include hand-searching the reference lists of existing review literature in the field (e.g., Weaver & Campbell, 2015), and the reference lists of all studies meeting inclusion criteria. We will also review the curricula vitae of known researchers in the field (e.g., Tim Bateman; Jeffrey Bouffard; Anne-Marie Day; Barry Goldson; Kate Gooch; Pippa Goodfellow; Paul gray; Neal Hazel; John Pitts), and we will contact the first authors of included reports to inquire if they are aware of any additional studies. Further, websites specifically affiliated with an aftercare program will be searched for evaluations. The search will also include the tables of contents of the following 25 journals, backdated to 24 months prior to the date of implementation of the electronic database search:

British Journal of Criminology

Description of methods used in primary research

Selection of studies

The initial database and gray literature search will be split into two sections, with Reviewer 1 conducting the search for the first section, and Reviewer 2 conducting the search for the second section. Specifically, each reviewer will read through the titles and abstracts of identified hits in their designated set of databases/sources to determine studies that appear potentially relevant to the review objectives and to be retrieved in full for further assessment. At this point in the selection process, we will only remove hits that are obviously irrelevant (e.g., book reviews, non-human studies). The lists from both reviewers will then be combined into a single database, and copies of all studies will be retrieved. Using the combined list, two reviewers will independently apply the inclusion and exclusion criteria to every article in the list to determine whether it meets the full set of selection criteria for objectives 1–3 and/or objectives 4–6. Discrepancies between reviewers will be resolved by a third reviewer.

Data extraction and management

The included studies will be split in half, with one coder conducting primary data extraction for each study in their set. Data will be entered into a Microsoft Excel database. A third coder will independently validate all coding for all studies by cross-checking the database coding with each research report. Discrepancies between coders will be discussed until consensus is reached.

For objectives 1–3, an extensive coding scheme involving approximately 75 variables will be used to extract data from each study. The coded variables will fall into categories of general study characteristics (e.g., publication type, program delivery year), program components (e.g., vocational skills, family involvement), information about the study sample (e.g., age, racial mix, gender mix, sample size), study characteristics (e.g., research design, baseline group differences, outcome measure, follow-up period), and outcomes (e.g., means and standard deviations, frequencies, group sample sizes, F-statistics).

For objective 4, additional variables will be used to code qualitative research. Extracted information will relate to study quality (e.g., reporting practices, ethical considerations, limitations), data characteristics (e.g., types of data presented), and process evaluation outcomes (e.g., data on program fidelity, resourcing, or staff qualification).

For objective 5, any information regarding data from intervention cost analyses (e.g., cost-benefit, cost-effectiveness, cost in relation to participants served) will be extracted from each study.

For objective 6, any information concerning the intervention's theory of change will be extracted, for example, theorized pathways for participants as they move between program activities towards desired outcomes.

See the draft coding form in Supporting Information: Appendix A for more details.

Assessment of risk of bias

Measures of treatment effect

Data from each individual study will be standardized so that results across studies can be meaningfully pooled. Effect size calculation will depend on the type of data presented in each individual study. More specifically, we anticipate the following calculations, although the final list will depend on the results of the study coding

(a)

For studies that present means and standard deviations for both groups at post-test, the basic standardized mean difference will be calculated as the mean of the treatment group (M _T ) minus the mean of the control group (M _C ) divided by the pooled standard deviation (SD_pooled):

Cohen ′ s d = M T − M C SD pooled . $\mathrm{Cohen}^{\prime} {\rm{s}}\,d=\phantom{\rule{}{0ex}}\frac{M_T-\phantom{\rule{}{0ex}}{M}_C}{{\mathrm{SD}}_{\mathrm{pooled}}}.$

Where the pooled standard deviation will be calculated by:

SD pooled = ( n T − 1 ) T 2 + ( n C − 1 ) C 2 ( n T − 1 ) + ( n C − 1 ) . ${\mathrm{SD}}_{\mathrm{pooled}}=\phantom{\rule{}{0ex}}\sqrt{\frac{{(n_T-1)}_T^2+\phantom{\rule{}{0ex}}{(n_C-1)}_C^2}{(n_T-1)+(n_C-1)}}.$

In which

SD T 2 ${\mathrm{SD}}_T^2$

is the standard deviation of the treatment group and

SD C 2 ${\mathrm{SD}}_C^2$

is the standard deviation of the control group.

The standard error will be calculated by:

SE d = n T + n C n T n C + d 2 2 ( n T + n C ) ${\mathrm{SE}}_d=\sqrt{\frac{n_T+n_C}{n_T{n}_C}+\frac{d^2}{2(n_T+n_C)}}\,$

To correct for possible small sample bias, we will apply Hedges' correction, resulting in Hedges' g:

Hedges ′ g = 1 − 3 4 n − 9 M T − M C pooled , $\mathrm{Hedges}^{\prime} g=\phantom{\rule{}{0ex}}\left[1-\frac{3}{4n-9}\right]\left[\frac{M_T-\phantom{\rule{}{0ex}}{M}_C}{{}_{{pooled}}}\right],$

where n = n _T  + n _c , n _T and n _T are sample sizes for the treatment and control groups, respectively, and SD_pooled is the pooled estimate of the standard deviation of the treatment and control groups.

(b)

If the report involves a pretest-post-test control group design, we will adapt the formula above to factor in the baseline data obtained at pretest and increase precision, as follows:

ES SMD = 1 − 3 4 ( n T + n C − 2 ) ( M T , post − M T , pre ) ( M C , post − M C , pre ) SD pre . ${\mathrm{ES}}_{\mathrm{SMD}}=\left[1-\frac{3}{4(n_T+n_C-2)}\right]\left[\frac{(M_{T,\mathrm{post}}-\phantom{\rule{}{0ex}}{M}_{T,\mathrm{pre}})(M_{C,\mathrm{post}}-\phantom{\rule{}{0ex}}{M}_{C,\mathrm{pre}})}{{\mathrm{SD}}_{\mathrm{pre}}}\right].$

Using a small samples bias corrector, M _t,post is the post-test mean of the treatment group, M _t,pre is the pretest mean for the treatment group, M _c,post is the post-test mean for the control group, M _{c, pre} is the pretest mean for the control group, and SD_pre is the pooled standard deviation at pretest (Morris, 2008), calculated as follows:

SD pre = ( n T − 1 ) T , pre 2 + ( n C − 1 ) C , pre 2 n T + n C − 2 . ${\mathrm{SD}}_{\mathrm{pre}}=\phantom{\rule{}{0ex}}\sqrt{\frac{{(n_T-1)}_{T,\mathrm{pre}}^2+\phantom{\rule{}{0ex}}{(n_C-1)}_{C,\mathrm{pre}}^2}{n_T+n_C-2}}.$

The standard error will be calculated as:

SE SMD = n T + n C n T n C 2 ( 1 − ρ ) + ( ES SMD ) 2 2 ( n T + n C ) , $\mathrm{SE}{}_{\mathrm{SMD}}=\sqrt{\frac{n_T+n_C}{n_T{n}_C}2(1-\rho )+\frac{{({\mathrm{ES}}_{\mathrm{SMD}})}^2}{2(n_T+n_C)}},$

where

ρ $\rho$

is the population correlation between the pretest and post-test measures, operationalized as the sample correlation between the baseline and outcome measures if provided in the study report. If the correlation is not available, we will follow the What Works Clearinghouse (2020) and assume a value of 0.5 for

ρ $\rho$

.

(c)

If a study presents an unstandardized beta coefficient, standard deviation of the dependent variable, and sample sizes, the effect size will be calculated as follows, where SD_pooled is calculated using pre-test n's:

ES β = 1 − 3 4 n − 9 β SD pooled . ${\mathrm{ES}}_{\beta }=\phantom{\rule{}{0ex}}\left[1-\frac{3}{4n-9}\right]\left[\frac{\beta }{{\mathrm{SD}}_{\mathrm{pooled}}}\right].$

The standard error will calculated by:

SE β = n T + n C n T n C + ( ES ) 2 2 ( n T + n C ) . ${\mathrm{SE}}_{\beta }=\sqrt{\frac{n_T+n_C}{n_T{n}_C}+\frac{{(\mathrm{ES})}^2}{2(n_T+n_C)}}.$

(d)

If a study presents an F-test with unequal sample group sizes, the effect size will be calculated as follows:

ES F = F ( n T + n C ) n T n C , ${\mathrm{ES}}_F=\phantom{\rule{}{0ex}}\sqrt{\frac{F(n_T+\phantom{\rule{}{0ex}}{n}_C)}{n_T{n}_C}},$

where F is the F-statistic, n _T is the post-test sample size for the treatment group and n _C is the post-test sample size for the control group. The standard error will be calculated as per the formula in bullet (c).

(e)

For studies that present dichotomous outcome measures (e.g., in the form of percentages or raw numbers representing how many participants were rearrested at least once), effect sizes will be computed as odds ratios. The odds will refer to the odds of recidivism (e.g., arrest, conviction, or incarceration) compared to no recidivism for an individual who participated in an aftercare intervention, relative to the odds of recidivism for an individual in the control group (Lipsey & Wilson, 2001). Specifically:

OR = ad bc = P a P d P b P c = P a ÷ P b P c ÷ P d = P a ( 1 − p c ) P c ( 1 − p a ) , $\mathrm{OR}=\phantom{\rule{}{0ex}}\frac{{ad}}{{bc}}=\phantom{\rule{}{0ex}}\frac{P_a\,P_d}{P_b\,P_c}=\phantom{\rule{}{0ex}}\frac{P_a\phantom{\rule{}{0ex}}\div\phantom{\rule{}{0ex}}{P}_b}{P_c\phantom{\rule{}{0ex}}\div\phantom{\rule{}{0ex}}{P}_d}=\phantom{\rule{}{0ex}}\frac{P_a(1-p_c)}{P_c(1-p_a)},$

For ease of interpretation, all odds ratios will be log transformed to LORs; LORs are centered around a value of zero, with zero indicating that recidivism is equally likely to occur in both groups. Data will be coded (or reverse-coded) such that an LOR below 0 indicates that the outcomes favor the control group (with the treatment group being more likely to recidivate), and a value above 0 indicates that the aftercare intervention has a beneficial impact on the treatment group (a lower rate of recidivism).

Given that LORs (based on dichotomous data) are not numerically compatible with standardized mean differences (based on continuous data), we will adjust by using a Cox transformation (Sanchez-Meca et al., 2003):

ES cox = L ( OR ) 1.65 . ${\mathrm{ES}}_{\mathrm{cox}}=\phantom{\rule{}{0ex}}\frac{{\rm{L}}(\mathrm{OR})}{1.65}.$

Where the standard error will be calculated as:

SE cox = . 367 1 a + 1 c + 1 b + 1 d . ${\mathrm{SE}}_{\mathrm{cox}}=\sqrt{.367\left[\left(\frac{1}{a}\right)+\left(\frac{1}{c}\right)+\left(\frac{1}{b}\right)+\left(\frac{1}{d}\right)\right]}.$

If the studies include both pretest and post-test data, this formula will be adapted to take this information into account. The pretest-adjusted formula is as follows:

ES cox = L OR , post 1.65 − L OR , pre 1.65 . ${\mathrm{ES}}_{\mathrm{cox}}=\phantom{\rule{}{0ex}}\left[\frac{{{\rm{L}}}_{\mathrm{OR},\mathrm{post}}}{1.65}\right]-\phantom{\rule{}{0ex}}\left[\frac{{{\rm{L}}}_{\mathrm{OR},\mathrm{pre}}}{1.65}\right].$

With the standard error calculated as:

SE cox = . 367 1 a pre + 1 c pre + 1 b pre + 1 d pre . ${\mathrm{SE}}_{\mathrm{cox}}=\sqrt{.367\left[\left(\frac{1}{a_{{pre}}}\right)+\left(\frac{1}{c_{{pre}}}\right)+\left(\frac{1}{b_{{pre}}}\right)+\left(\frac{1}{d_{{pre}}}\right)\right]}.$

Unit of analysis issues

Criteria for determination of independent findings

Dealing with missing data

We anticipate substantial “missing data” to be encountered throughout the coding process; specifically, across the set of included studies certain pieces of information may not be presented by study authors (e.g., sample size, participant age, participant gender). We will attempt to obtain this information by contacting study authors; otherwise, when possible, we will calculate the missing data using data available in the evaluation report. For example, a study may report only the total sample size, rather than specifying the sample sizes per treatment/control group. In these cases, when other information regarding the sample size is available such as degrees of freedom from a reported F-test, we will calculate the treatment and control group sample sizes using an assumption of proportional group attrition. If missing quantitative data needed for inclusion in meta-analysis cannot obtained from study authors and cannot be calculated, the study will be excluded from meta-analysis and coded as “data not available.”

Data synthesis

Assessment of heterogeneity

Assessment of reporting biases

Sensitivity analysis

Treatment of qualitative research

The study will synthesize qualitative data by drawing on thematic analysis and framework synthesis (Carroll et al., 2011; Dixon-Woods, 2011; Thomas & Harden, 2008). Following a systematic search of the literature (using electronic databases and gray literature sources), retrieved qualitative studies will be read and re-read by two independent reviewers, and findings will be coded into themes based on the initial framework of the aftercare model. Further, as new themes emerge that are not decided a priori, the framework will be updated to represent the emerging synthesis. As in previous studies utilizing framework synthesis (e.g., Abbott et al., 2019), our goal is to present an understanding of the effects of aftercare and resettlement programming as well as an explanation of the mechanisms that work to produce these effects. Triangulation between results and findings of the quantitative and qualitative evidence reviews, as well as existing literature, will form the basis of a convergent mixed-methods review (Noyes et al., 2019, p. 21).