Intimate partner violence (IPV) affects one-third of women globally and the negative health consequences of IPV are well-documented. Over the last decade, multiple interventions have been developed, adapted, and evaluated to address IPV and two separate reviews identified 95 completed randomized control trials (RCT) and quasi-experimental evaluations of interventions aimed at preventing IPV and other types of violence against women and girls. As more longitudinal data on the effectiveness of interventions becomes available, researchers can move from asking questions regarding the “average” success of an intervention, to unpacking its effect on different sub-populations, thereby refining the goal of the intervention. We use data from two interventions—Stepping-Stones/Creating Futures (SS-CF)in South Africa, and Indashyikirwain Rwanda—to explore two research questions: Are the interventions tested equally effective at preventing the onset of violence as they are for stopping or reducing the intensity of violence pre-existing in a relationship? And, how do different strategies for measuring and coding IPV as an outcome, affect estimates of intervention effectiveness?
To assess differences in treatment outcomes by baseline reporting of IPV, we created three new binary variables and one count variable: IPV cessation, IPV reduction (binary and count) and IPV prevention. Both trials used an intention-to-treat analysis approach. In the SS-CFtrial, outcomes were analyzed using generalized estimating equation models accounting for the clustered nature of the data. The analysis for the Indashyikirwa trial included generalized linear mixed effects model with a logit link function to compare the effect of the intervention between the 2 study arms for all binary variables and a negative binomial link function for count data. We include both male and female participants in analyses assessing differences in treatment effects by baseline reporting of IPV. In each study we split the sample into two groups and assessed treatment effects for ongoing IPV (reduction, cessation) among participants who had reported experiencing/perpetrating IPV at baseline and assessed new onset of IPV (prevention) among participants who had not reported experiencing/perpetrating IPV at baseline.
Results indicated that for both men and women, the Indashyikirwa intervention in Rwanda was more successful at reducing or stopping ongoing experience (women) and perpetration (men) of IPV than it was at preventing its onset. The SS-CF intervention in South Africa, by contrast, was more successful at preventing men from starting to perpetrate IPV than it was in reducing the intensity of men’s perpetration or stopping it entirely. Analyses comparing binary and count measures of trial outcomes indicates that count measures may perform better in samples with sufficient data points across the range of possible scores.
Conclusion and implications
Subgroup analyses are critical to unpacking trial results and future evaluation studies should plan for these types of analyses to ensure adequate sample sizes for all subgroups of interest. Such analyses in IPV intervention programs can help identify the extent to which existing interventions are addressing the needs of different populations and thereby inform future programmatic, funding, and policy decisions.