Test-retest stability of self-reported violence against women measures: results from the stepping stones and creating futures pilot

Background: Stability of measures in quantitative social science research is crucial to understand. There is very little evidence on the stability of violence against women and girls measures in the global South. Objective: To assess the test-retest stability of violence against women and girls measures, amongst young (18–30) people in South Africa. Methods: Data were collected from 124 women and 112 men at zero weeks (time 1) and two weeks (time 2), who resided in urban informal settlements in South Africa. Prevalence of each construct was assessed using chi-square contingency tables. Stability of self-report over time was assessed using Cohen’s Kappa. Bivariate logistic regression assessed factors associated with changing responses between time 1 and time 2. Results: At group level prevalence of all measures showed no significant differences. Stability of self-report: kappas for past year physical IPV were both k0.20, for ever physical IPV (women k0.58; men k0.50). Sexual IPV in past 12m (women k0.44; men k0.18), and for ever sexual IPV (women k0.56; men k0.46). Kappas for men’s perpetration of non-partner sexual violence was k0.29 for past 12m and k0.38 ever. In bivariate regression, completion of secondary education was associated with a reduced odds of changing responses over the time-period for sexual IPV ever women (OR0.16, 0.02–1.04), sexual IPV past 12 months men (OR 0.09, 0.01–0.56), past 12 month non-partner sexual violence men (OR0.19, 0.02–1.41) and lifetime non-partner sexual violence (OR0.23, 0.04–1.19). Being male, compared to being female, was associated with an increased likelihood of changing responses for past 12 month sexual IPV (OR2.10, 1.08–4.09). Conclusions: Prevalence estimates of violence against women measures are stable at group level, but stability of self-reported measures remains a concern. Individual statistical analyses must be treated with caution. Future studies are required to develop further understandings of stability of measures over time.