Three-Year Outcomes From a Middle School Dating Violence Prevention Program

In 2018, we randomized 24 urban and suburban middle schools in southeast Texas to receive either the intervention (n = 12) or health class as usual (n = 12). We then recruited seventh grade students (n = 2768; 50.4% female, 25.6% Black, 37.0% Hispanic, 7.2% white, 14.1% Asian, 16.1% other) and assessed them at baseline (before intervention), as well as at 12-, 24-, and 36-months postintervention. Study details, including the consort diagram for this trial, were reported in this journal.⁶  The study was approved by the authors’ institutional review board.

We assessed physical ARA perpetration with 4 items from the Conflict in Adolescent Dating Relationships Inventory,⁷  which has been shown to be sensitive to measuring change over time. We focused analyses on the subgroup of students who reported engaging in physical dating violence (DV) perpetration at baseline and completed follow-up (n = 198). The primary outcome is the 36-month incidence of subsequent physical DV perpetration, calculated by looking across the 12-, 24-, and 36-month follow-up assessments using a coding scheme of 1 = perpetration at any 12 month interval and 0 = no perpetration. Retention rates were 83% (at 12-month follow-up), 82% (24 months), and 79% (36 months). Attrition was higher for males, as well as Hispanic and Black participants and lower for those living with both parents and from college educated households.

We used descriptive statistics to present a profile for the full sample and by treatment condition. We used a multilevel logistic regression model to assess intervention effectiveness in reducing the 36-month incidence of physical ARA reperpetration among students with a baseline history of physical ARA perpetration. Multilevel models provide corrected SE estimates that can arise from the nonindependence of the data in clustered sample designs, where intraclass correlation is present among students sampled from the same school. To adjust for the potential confounding that may arise from group differences, we included gender, race, ethnicity, parental education, and baseline sexual experience in the model. To be clear, gender, race, and ethnicity are not risk factors; instead, they represent the differential impact of systemic sexism and racism on minoritized populations, which increases their risk for a number of negative outcomes. We entered the treatment group assignment into the regression model as an indicator (0 = control, 1 = treatment) and used a Wald test (β/SE) to determine statistical significance. We set the Type I error rate a priori at α = 0.05. We present unadjusted and adjusted odd ratios to estimate effect size.

As shown in Table 1, students in the subsample of physical ARA perpetration are more likely to be female, Non-Hispanic Black, and sexually experienced at baseline and less likely to have college educated parents versus the randomized cohort. Additionally, there was a differential distribution of race and ethnicity between the 2 treatment conditions within the subpopulation. This social construct was entered into the adjusted models to ensure the estimate of intervention impact was not confounded by this difference. As shown in Table 2, of the students reporting baseline physical ARA perpetration, the 36-month incidence of repeat perpetration in the control group was 58.2% vs 47.4% in the intervention group, adjusted odds ratio = 0.49; 95% confidence interval, 0.25 to 0.97; P = .04.

In one of the few ARA intervention studies to measure effectiveness beyond 2 years, we found that Fourth R prevented the reperpetration of dating violence. Teens with a history of ARA who received the intervention evidenced substantially less perpetration of ARA 3 years later than their counterparts in the control group. Although primary prevention programs remain the gold standard, efforts to interrupt ARA perpetration are needed, especially given the strong link between ARA and adult intimate partner violence.³  Coupled with existing studies showing the efficacy of Fourth R as a universal primary prevention approach,^6,8  the current study lends support to using this intervention as a tertiary or targeted approach to violence prevention. The racial and ethnic differences in the subsample that were controlled for in this analysis are important and warrant further investigation, especially as a potential moderator of intervention effects. Limitations include reliance on self-report, focus on physical ARA perpetration, and limited generalizability. However, the strengths (eg, randomized controlled design, long-term follow-up, sample diversity) and importance of study findings (ie, Fourth R appears effective in preventing the reperpetration of ARA) outweigh these limitations. Efforts to prevent ARA should be broadly employed with mental health, school, and family support to address both first and repeated incidents of violence.

ARA

adolescent relationship abuse