Methods: We conducted a rigorous search using three electronic databases. Studies were required to be published from 2002-2016 in English, analyze quantitative data, evaluate an SV prevention program delivered in U.S. middle schools, measure at least one mediator or outcome related to SV victimization or perpetration, and include a comparison or control group. Searches yielded 487 articles potentially relevant for review. After title, abstract, and full-text review, as well as reference list checks, 10 studies fully met inclusion criteria. For each study, we systematically extracted and synthesized data on sample size, intervention used, outcomes studied, research design, multivariate statistical methods used, and software package(s) used. Two coders independently reviewed each study.
Results: Studies were published from 2004-2015, and samples included sixth-ninth graders. Sample sizes ranged from 460–3616 students with 2–36 schools per study. The majority of studies reported software packages (e.g., SPSS, SAS) used to analyze study data (n = 8). Seven studies used cluster-randomized designs, with randomization occurring at the school level (n = 5) or the classroom level (n = 2). Three studies employed quasi-experimental designs. Studies included a variety of different data analytic techniques with the majority using multilevel modeling or robust standard error adjustment to account for data clustering (n = 7). No study used propensity score analysis (PSA) methods.
Implications: The majority of included studies used cluster-randomized designs with well-matched statistical analyses. These designs and analyses represent progress toward making causal claims. However, none of these studies use advanced statistical methods, in particular PSA techniques, that would further enhance researchers’ ability to more confidently make causal claims. By creating comparable treatment and control groups, these methods help researchers better approximate the treatment effect estimates that may be obtained from a successful randomized experiment when using data from quasi-experimental studies or studies in which randomization was unsuccessful. To help inform future research and practice related to SV prevention, we present information on research designs and methods currently in use, along with methodological strengths and areas for growth. We also highlight ways to augment future research in this field, including the potential for PSA techniques to enhance SV prevention research.