Study of bias in antenatal clinic HIV-1 surveillance data in a high contraceptive prevalence population in sub-Saharan Africa

Abstract

To describe patterns, sources and consequences of bias in antenatal clinic (ANC) HIV prevalence estimates in a high contraceptive prevalence population. HIV surveillance in Africa relies on data from pregnant women attending ANCs. HIV estimates from pregnant women understate female infection levels in low income, high fertility populations. Bias in high contraceptive use, delayed sexual debut populations remains undescribed. Comparison of parallel cross-sectional population and antenatal survey data from rural Zimbabwe, where 60% of women are recent contraceptive users. HIV prevalence in recently pregnant women (25.7%; n = 576) and all women (25.5%; n = 5138) is similar over the age-range 15–44 years. As in high fertility populations, HIV prevalence is higher in pregnant women at young ages and lower at older ages but the crossover point occurs later due to delayed sexual activity. HIV understatement at older ages due to HIV-associated infertility is mitigated by less HIV infection and less frequent ANC attendance in contraceptive users. The local ANC HIV prevalence estimate is lower [21.2%; n = 1215; risk ratio versus pregnant women in the general population, 0.8; 95% confidence interval (CI), 0.7–1.0], possibly because women from more remote areas are included. ANC estimates overstate the relative risk of HIV in more educated women (age-adjusted odds ratio, 1.1; 95% CI, 0.8–1.4 versus 0.7; 95% CI, 0.6–0.9). ANC estimates understate female HIV prevalence in this low fertility population but, here, the primary cause is not selection of pregnant women. ANC estimate adjustment procedures that control for contraceptive use and age at first sex are needed.