Comparison of HIV Prevalence Estimates From Sentinel Surveillance and a National Population-Based Survey in Uganda, 2004-2005

Abstract

HIV programs in generalized epidemics have traditionally relied on antenatal clinic (ANC) sentinel surveillance data to guide prevention and to model epidemic trends. ANC data, however, come from a subset of the population, and their representativeness of the population has been debated. Data from a national population-based Uganda HIV/AIDS Sero-Behavioral Survey (UHSBS) were compared with those from ANC sentinel surveillance. Using geographic information system, UHSBS clusters within a 30 km radius of the ANC sites were mapped. Estimates of HIV prevalence from ANC surveillance were compared with those from UHSBS. The ANC-based HIV prevalence, 6.0% [confidence interval (CI) 5.5% to 6.5%], was similar to that from UHSBS, 5.9% (CI 5.4% to 6.4%). The ANC-based estimate correlated with that of UHSBS catchment area women who were pregnant and those who had given birth in the 2 years preceding the survey. ANC data overestimated prevalence in the 15-year to 19-year age group, were similar to UHSBS for ages 20-29 years, and underestimated prevalence in older respondents. ANC data underestimated HIV prevalence among women (6.0% vs. 7.4%; CI 6.6% to 8.2%) and urban women (7.6% vs. 12.7%) but was similar for rural women (5.3% vs. 4.9%). ANC-based surveillance remains an important tool for monitoring HIV/AIDS programs. ANC and UHSBS data were similar overall and for 15-year to 29-year olds, women who were pregnant, and women who had a birth in the 2 years before the survey. ANC estimates were lower in those > or = 30 years and in urban areas. Periodic serosurveys to adjust ANC-based estimates are needed.