INTRODUCTION: Women in HIV prevention trials often must typically agree to avoid pregnancy. Regardless, some become pregnant. Screening tools predicting pregnancy risk could maximize trial safety and efficiency. OBJECTIVES: We assessed incidence and correlates of pregnancy among women at high HIV risk. METHODS: We enrolled sexually-active, HIV-negative women into an observational cohort (2008-2011). At enrollment demographic, contraceptive, reproductive, pregnancy intention and behavioural data were collected. Women reported if one or both partners wanted or intended for the couple to become pregnant. We measured gender role beliefs using a locally validated eight-point index. We tested HIV and pregnancy, and inquired about sexually transmitted infection symptoms (STIs) at enrollment and monthly. HIV testing included behavioural counselling and condom provision, but did not specifically counsel women to avoid pregnancy. Cox proportional hazard modelling evaluated the associations with pregnancy. The multivariate model included the following variables "Recent pregnancy attempts", "Gender Roles Beliefs", "Self-reported STIs" and "Age". RESULTS: We screened 1068 women and excluded (24.6%, 263/1068) who did not report risk behaviour. Non-pregnant, non-sterilized women aged 18-35 (median = 21 years) enrolled (n = 438). Most women reported one partner (74.7%) and a prior live birth (84.6%). Median follow-up time was 6 months (range 0.7-15.5). Pregnancy incidence was 25.1 per 100 women-years (n = 57 pregnancies). Conservative beliefs on gender roles (Adjusted Hazard Ratio (aHR) 1.8; 95% confidence interval [CI] 1.1-2.9), recent pregnancy attempts (aHR 1.9; 95% CI 1.1-3.4) and baseline self-reported STI (aHR 2.5; 95% CI 1.4-4.4) were associated with increased incident pregnancy. Report of no pregnancy intention was associated with lowered pregnancy risk (aHR 0.3; 95% CI 0.1-0.7). CONCLUSIONS: We identified new and confirmed existing factors that can facilitate screening for pregnancy risk.