We aimed to develop a HIV risk scoring algorithm for targeted screening among women in South Africa. We used data from five biomedical intervention trials (N = 8982 Cox regression models were used to create a risk prediction algorithm and it was internally and externally validated using standard statistical measures; 7-factors were identified as significant predictors of HIV infection: <25 years old, being single/not cohabiting, parity (<3), age at sexual debut (<16), 3+ sexual partners, using injectables and diagnosis with a sexually transmitted infection(s). A score of ≥25 (out of 50) was the optimum cut point with 83% (80%) sensitivity in the development (validation) dataset. Our tool can be used in designing future HIV prevention research and guiding recruitment strategies as well as in health care settings. Identifying, targeting and prioritising women at highest risk will have significant impact on preventing new HIV infections by scaling up testing and pre-exposure prophylaxis in conjunction with other HIV prevention modalities.
Keywords: HIV infection; HIV prevention; KwaZulu-Natal; Population attributable risk; Risk prediction tool; South Africa.