Background: To present and compare population-based and antenatal-care (ANC) sentinel surveillance HIV prevalence estimates among women in a rural South African population where both provision of ANC services and family planning is prevalent and fertility is declining. With a need, in such settings, to understand how to appropriately adjust ANC sentinel surveillance estimates to represent HIV prevalence in general populations, and with evidence of possible biases inherent to both surveillance systems, we explore differences between the two systems. There is particular emphasis on unrepresentative selection of ANC clinics and unrepresentative testing in the population.
Methods: HIV sero-prevalence amongst blood samples collected from women consenting to test during the 2005 annual longitudinal population-based serological survey was compared to anonymous unlinked HIV sero-prevalence amongst women attending antenatal care (ANC) first visits in six clinics (January to May 2005). Both surveillance systems were conducted as part of the Africa Centre Demographic Information System.
Results: Population-based HIV prevalence estimates for all women (25.2%) and pregnant women (23.7%) were significantly lower than that for ANC attendees (37.7%). A large proportion of women attending urban or peri-urban clinics would be predicted to be resident within rural areas. Although overall estimates remained significantly different, presenting and standardising estimates by age and location (clinic for ANC-based estimates and individual-residence for population-based estimates) made some group-specific estimates from the two surveillance systems more predictive of one another.
Conclusion: It is likely that where ANC coverage and contraceptive use is widespread and fertility is low, population-based surveillance under-estimates HIV prevalence due to unrepresentative testing by age, residence and also probably by HIV status, and that ANC sentinel surveillance over-estimates prevalence due to selection bias in terms of age of sexual debut and contraceptive use. The results presented highlight the importance of accounting for unrepresentative testing, particularly by individual residence and age, through system design and statistical analyses.