Spatial clustering of intimate partner violence and female employment in South Africa without correspondence to provincial HIV burden: a geolocated analysis of the 2016 demographic and health survey

Background South Africa has the world’s largest HIV epidemic, with women’s HIV prevalence varying nearly threefold across provinces, from 13.5% in the Northern Cape to 37.3% in KwaZulu-Natal. Intimate partner violence (IPV) and women’s economic empowerment are both associated with HIV risk, but whether their geography corresponds to HIV burden geography has not been tested using individual-level, geolocated data. Methods Secondary data from the 2016 South African Demographic and Health Survey were linked to enumeration area GPS coordinates and analyzed with correct survey weights. IPV was assessed among 2,469 ever-partnered women (69.6% aged 50 or older) across 674 clusters: female employment among 8,514 women aged 15–49 across 729 clusters. Weighted generalized additive models with a bivariate spatial smooth term tested clustering of both indicators, crude and covariate-adjusted, with a unified, 300-replication permutation test. Spearman correlations, with false discovery rate correction for multiple comparisons, examined correspondence between provincial IPV/employment estimates and HIV prevalence. Results 23.8% of ever-partnered women reported any form of IPV: emotional (18.5%), physical (16.2%), and sexual (4.6%). Spatial models found significant crude clustering for female employment (4.3% deviance explained, p = 0.017) but only borderline evidence for IPV (5.1%, p = 0.067); after adjustment for age, residence, education, and wealth, employment remained clearly clustered ( p < 0.004), and IPV showed modest evidence ( p = 0.038, close to the threshold). Neither indicator, nor any IPV subtype, correlated significantly with provincial HIV prevalence after correction (any IPV: r = 0.433, adjusted p = 0.610; employment: r = −0.050, adjusted p = 0.898); power at n = 9 was low (21%). KwaZulu-Natal, the highest-burden province for HIV (37.3%), had the second-lowest levels of both IPV (18.8%) and employment (28.6%). Conclusion Intimate partner violence and female employment show spatial clustering; employment’s is significant before and after adjustment, while IPV’s lies close to the significance threshold in both models, and neither corresponds to HIV burden geography. This does not contradict individual-level evidence linking IPV and empowerment to HIV risk; it indicates that provincial HIV programming should not assume that IPV, employment, and HIV burden are concentrated in the same provinces.