Practice Profile: Domestic Violence Courts

Gutierrez and colleagues (2017) conducted a meta-analysis of outcome evaluations of domestic violence courts (DVCs) to examine the impact on recidivism. A comprehensive search was conducted to identify eligible studies. Search methods included a keyword search, reference list review, and contacting researchers in the specialty court field. The following inclusion criteria were applied: 1) the study had to examine a DVC, 2) the study had to report a recidivism measure as an outcome, 3) the study had to have a comparison group of some kind (randomly assigned or retrospective), and 4) the study had to provide enough information to calculate an effect size (specifically odds ratios).

The search yielded a total of 20 outcome evaluations, which included 26 unique samples. All 20 studies took place in either the United States or Canada, and publication dates ranged from 1996 to 2013. Of the 26 unique samples, 18 used a convenience/non-equivalence sample, while the remaining 8 used a matched, equivalent sample. Additionally, 14 were quasi-experimental designs and 5 were randomized controlled trials. The remaining 7 were either not reported or other study designs. Lastly, 20 samples were drawn from unpublished reports and the other 6 were published reports. The total number of aggregated participants was 26,601. The combined sample had a mean age of about 32, was mostly male, and of mixed race (across 13 of the 26 samples). The average percentage of individuals with prior criminal justice system involvement (e.g., prior arrests) was 54 percent. Regarding study quality, the majority of studies (69.2 percent) were classified as “weak” according to the outcome rating of the Collaborative Outcome Data Committee (CODC) guidelines which evaluates individual studies in order to differentiate between levels of methodological quality. The CODC guidelines consist of 21 items that represent indicators of study quality (e.g., experimenter involvement, attrition, group equivalency). Additionally, 15.4 percent were rated as “rejected,” 15 percent were rated as “good,” and none were rated as “strong.”

Odds ratio was chosen as the effect size and the most appropriate index to estimate the impact of a dichotomous variable (DVC participation versus nonparticipation) on dichotomous outcome (recidivism versus no recidivism). A weighted approach was used to allow studies with larger sample sizes to contribute more to the overall estimation of the effect size. A fixed-effect analysis was conducted.

The authors conducted an additional analysis to determine whether DVCs, according to the literature, were adhering to the risk-need-responsivity (RNR) model. The RNR model states that individuals should be matched to programming and services according to their level of risk, individual needs, and learning style in order to achieve the best possible outcomes for youths (Andres, Bonta, and Hoge 1990). Results suggested that studies which demonstrated adherence to the RNR model showed significantly greater treatment effects when compared to studies which did not adhere to the RNR model.