New commentary reevaluates the effectiveness of mentoring in reducing juvenile criminal recidivism

DuBois, D. L. (2022). Reconsidering the effectiveness of mentoring for prevention of juvenile criminal recidivism: A brief comment on systematic review and meta-analysis of noninstitutional psychosocial interventions to prevent juvenile criminal recidivism (Olsson et al, 2021). Journal of Consulting and Clinical Psychology, 90(8), 647–651.

Summarized by Ariel Ervin

Notes of Interest: 

  • Interventions conducted outside the juvenile justice system are often used to reduce justice system involvement and future offenses among youth offenders.
  • Olsson et al. (2021) recently conducted a meta-analysis and systemic review of non-institutional interventions for preventing juvenile criminal recidivism. However, they couldn’t find evidence that the examined interventions were notably more effective than the control group.
  • This commentary searches for further evidence on mentoring (one of the interventions brought up by Olsson et al. (2021)) by conducting an expanded literature search.
  • The reanalysis indicates that mentoring is more effective than control treatments.
  • These findings suggest that mentoring is a promising approach to lower juvenile justice system involvement and juvenile crime.
  • It’s possible that the results found by Olsson et al. (2021) were sensitive to missed studies with positive findings, as well as the “file-drawer problem.”
  • Potential negative consequences of having missed studies can extend beyond false indications that there is an association or effect present.
  • It’s also concerning if a meta-analysis couldn’t provide evidence for a present effect due to missed studies.
  • Running multifaceted and extensive literature research searches is essential in conducting a quality meta-analysis or systemic review.

Introduction (Reprinted from the Abstract)

Interventions implemented outside of the juvenile justice system are widely utilized with youth offenders to decrease the likelihood of future offending and justice system involvement, both of which are well-documented as being costly to youth and society at large. Olsson et al. (2021) conducted a systematic review and meta-analysis of these types of interventions for youth aged 12-17 and failed to find any of the interventions examined to be more effective than control treatments in reducing future criminality. This commentary further examines the evidence for one of these interventions, mentoring, based on an expanded search of the literature that identified several additional studies of mentoring for recidivism prevention that meet the eligibility criteria utilized by Olsson et al. A meta-analysis of these studies and those identified by Olsson et al. finds mentoring to be more effective than control treatments (typically services as usual within the juvenile justice system): risk difference between groups of 0.09 (random effect model; 95% confidence interval [.03-.15]). This reanalysis thus indicates greater promise for mentoring as a tool in reducing juvenile crime and juvenile justice system involvement than was suggested by the results of Olsson et al. Importantly, it also underscores the potential for the results of meta-analyses to be sensitive to not only unidentified studies with null or negative results (the so-called “file-drawer problem”) but also missed studies with positive findings. Recommendations for literature search procedures in systematic reviews and sensitivity analyses in meta-analyses are provided with this concern in mind.

Implications (Reprinted from the first two paragraphs )

Interventions implemented outside of the juvenile justice system are widely utilized with youth offenders to decrease the likelihood of recidivism. To the extent they are effective in preventing future offending, these types of interventions can be expected to improve long-term outcomes for the youth involved and, if implemented at sufficient scale, the overall quality of life within communities. The costliness of juvenile institutionalization (e.g., incarceration), in combination with evidence that it can prove harmful to the youth involved (e.g., Strijbosch et al., 2015), underscores the importance in particular of lowering the likelihood of future contact with the juvenile justice system (e.g., arrest). Olsson et al. (2021) recently reported findings of a systematic review and meta-analysis of noninstitutional interventions for prevention of juvenile criminal recidivism. A total of 35 studies evaluating 17 unique interventions were included in their review. Five of the interventions had been evaluated in more than one study, thus permitting a meta-analysis of the average estimated effect of each interventions across multiple investigations: functional family therapy, group-based cognitive behavioral therapy, mentorship, multidimensional family therapy, and multisystemic therapy. Separate meta-analyses were conducted for randomized control trials and nonrandomized control trials (i.e., quasi-experimental studies) when multiple studies of each were available. For the primary outcome of recidivism based on officially reported delinquency (e.g., arrest as indicated in police records), none of the interventions was found to be superior to control conditions. More specifically, average effect size in the form of risk difference (RD; or, in the case of mentoring, an odds ratio because of unavailability of RD information for one of the studies) did not differ significantly (p < .05) from 0 (or 1 in the case of odds ratio).

Olsson et al. (2021) thoughtfully discuss a number of possible explanations for these findings. From a methodological standpoint, they point out that officially reported crimes may be relatively insensitive as a measure of recidivism due to underestimation of actual delinquent behavior. Limitations in the quality of the study evidence (e.g., risk of bias) are likewise discussed. Olsson et al. further note that whereas some previous meta-analyses suggest that the interventions involved may have favorable effects on juvenile behavior problems (see, e.g., Tolan et al., 2014, in the case of mentoring), these meta-analyses have included broader populations of at-risk youth (e.g., those exhibiting risk behaviors such as substance use but not necessarily prior criminal behavior), different groups of juvenile offenders (e.g., sex offenders), and were published prior to the time frame of the current review and thus potentially more subject to methodological limitations (e.g., lack of blinding in data collection, less effective control alternatives). It is difficult, of course, to assess the extent to which these factors, individually or in combination, did or not influence the primary findings of the review.

To access this article, click here.