Differential impact on men in an IPV prevention intervention: A post hoc analysis using latent class analysis of the Stepping Stones and Creating Futures intervention in South Africa.

Working with men to prevent their perpetration of intimate partner violence and non-partner sexual violence is increasingly recognised as effective. However, in any given context there are a multiplicity of masculinities, each of which has a different association with violence perpetration. There remains lack of clarity about whether such interventions impact all men and masculinities equally. We undertook a post-hoc analysis of men involved in the successful Stepping Stones and Creating Futures cluster randomized control trial in Durban, South Africa, to assess: i) whether there were different groups of men, ii) the factors associated with group allocation, and iii) whether the intervention had a differential impact on these groups in terms of violence perpetration. We used Latent Class Analysis (LCA) to identify masculinity classes, based on fit statistics and theoretical plausibility, and then used descriptive statistics (numbers, percentages, means) and p-values and multinomial models (95% confidence intervals, p-values) to assess factors associated with allocation to each class. To assess intervention impact by group, we used an intention-to-treat analysis, comparing men in each masculinity class, by intervention and control arm, using generalized estimating equations reporting unadjusted and adjusted odds ratios (aORs). In total 674 were recruited at baseline, and the LCA identified three classes of men: high violence (29% of men), medium violence (50% of men) and low violence (21% of men). Multinomial models showed those in more violent classes were more supportive of violence, had more adverse experiences, more depression and had worked more. By masculinity class, the impact of SS-CF showed reductions among the most violent men, with significant reductions in past year physical IPV (aOR0.59, p = 0.014), emotional IPV (aOR0.44, p = 0.044) and economic IPV (aOR0.35, p = 0.004), with non-significant reductions among other classes of men. This analysis suggests intensive group-based interventions can have significant impacts on the most violent men in communities.