Practice Profile: Dropout Prevention Programs

Klima, Miller, and Nunlist (2009) conducted a meta-analysis on the impact of evidence-based intervention and prevention programs for truancy and school dropout.

To be included in the meta-analysis, evaluations needed to adhere to the requirements of Washington’s truancy laws and be implemented by schools, courts, or law enforcement; include a comparison group equivalent on key variables (such as attendance patterns and academic achievement); and assess at least one of three outcomes: school attendance, high school graduation, or dropout status. Studies with high attrition or a single group pre–posttest design were excluded from the analysis. Random assignment studies were preferred for inclusion; but non-randomly assigned control group studies were also included if sufficient information was provided to demonstrate comparability between the treatment and comparison groups on important preexisting conditions such as school attendance, achievement, and grade retention.

The search for program studies identified 877 possible candidates; only 22 studies met the criteria for methodology and relevant outcomes. The 22 studies included data on 30 independent samples (4,014 treatment participants) for the dropout outcome, 29 independent samples (2,712 treatment participants) for the academic achievement/school performance outcome, and 6 (635 treatment participants) independent samples for the graduation outcome. The number of participants in the control groups was not reported. The studies included a number of program types—alternative educational, mentoring, behavioral, youth development, and academic remediation programs—as well as alternative schools. No information on race/ethnicity or other demographic information was reported for the studies included in the meta-analysis.

Authors calculated an overall weighted mean effect size for all studies using a random effects model. The authors also reported an overall weighted mean effect size by program class (e.g., academic remediation programs, alternative educational programs, and so forth).

Wilson and colleagues (2011) conducted a meta-analysis on the effects of dropout prevention and intervention programs. Their analysis of general dropout programs included 152 studies with 317 independent samples.

To be included in the meta-analysis, evaluations needed to have centered on school-based or school- affiliated psychological, educational, or behavioral prevention or intervention programs for students with school-aged youth (specifically ages 4–22); used an experimental or quasi-experimental design with at least 10 subjects in each study group; assessed a qualifying measure of school completion or dropout; and been published or have the study reported in 1985 or later.

Studies included in the analysis were technical reports (77 percent) or journal articles (20 percent), which were reported in the United States (97 percent). The mean reporting year was 1994. The studies comprised a number of program types including school or class restructuring, vocational training, academic services, community services, mentoring, counseling, and alternative schools. Many of the evaluated programs implemented multiple program types. The mean age of the sample was 15 years and the average grade level was 9th. The sample was 50 percent male and 50 percent female; and the race/ethnicity of the sample was Black (39 percent), white (33 percent), Hispanic (22 percent), or another minority (9 percent). No information on the number of students in the treatment and control groups was reported.  

Odds ratios were used as the effect size metric. Authors reported an inverse variance weighted analysis that incorporated both the sampling variance and between-studies variance into study weight levels. Random effects weighted mean effect sizes were calculated. The authors also reported mean effect size for different program types on school dropout.

Building on the data from Wilson and colleagues (2011), and using the same eligibility criteria, Tanner–Smith and Wilson (2013) analyzed 74 studies on dropout prevention programs reported between 1985 and 2009. Eligible studies had to specifically measure program effects on school absenteeism, in addition to the inclusion criterion from the previous work.

Twenty-four studies used randomized controlled trial (RCT) designs and 50 studies used quasi-experimental designs (QEDs). Authors reported effects of both study designs separately. Almost all of the RCTs were conducted in the United States (96 percent); just over half of the participants were male (52 percent), and 23 percent were white. The mean age for participants in the RCTs was 14 years old and the average grade level was 8th. The mean reporting year of the studies was 1993. Approximately 50 percent of the interventions were delivered in school classroom settings. No information on treatment and control group size was reported. 

Authors reported results of RCTs and QEDs separately due to the considerably different average effect sizes and differences in student characteristics. Odds ratios were reported as the effect size metric and adjusted with Hedges’ g. Authors reported an inverse variance weighted mean effect size for all studies using a random effects model. Only one effect size per sample was included in the analysis.