A Method for Estimating Incidence from Annual Antenatal Prevalence and Occasional Household Prevalence Surveys

Rob Dorrington, University of Cape Town

Increasingly, it is becoming necessary to track the course of the HIV/AIDS epidemic by measuring incidence. Unfortunately, there is no cheap and simple way of doing this (national longitudinal surveys are implausible, likely to be expensive, and will be complicated by loss to follow up; laboratory tests are of doubtful accuracy; and national household prevalence surveys only provide estimates that are an average of a number of years in the past). This paper considers making use of the relationship between the ratio of the population prevalence estimated from household prevalence surveys and models to that derived from antenatal surveys to provide annual estimates of the incidence of HIV. The results show that despite very wide confidence intervals the estimates are consistent with those from other sources and suggest that incidence in South African women 15-49 was about 1.5% in 2007 and has fallen over at least the previous five years.