RT Journal Article
SR Electronic
T1 Comparison of adult HIV prevalence from national population-based surveys and antenatal clinic surveillance in countries with generalised epidemics: implications for calibrating surveillance data
JF Sexually Transmitted Infections
JO Sex Transm Infect
FD BMJ Publishing Group Ltd
SP i17
OP i23
DO 10.1136/sti.2008.030452
VO 84
IS Suppl 1
A1 E Gouws
A1 V Mishra
A1 T B Fowler
YR 2008
UL http://sti.bmj.com/content/84/Suppl_1/i17.abstract
AB Background: Estimates of the impact of HIV in countries with generalised epidemics are generally based on antenatal clinic surveillance data collected over time. In an attempt to obtain geographically more representative estimates of HIV prevalence, many countries are now also conducting national population-based surveys in which HIV testing is included. We compare adult HIV prevalence estimates from antenatal clinic surveillance to those from national population-based surveys to assess the implications for calibrating surveillance data.Methods: HIV prevalence estimates derived from fitting prevalence curves to antenatal clinic surveillance data are statistically compared to prevalence from national population-based surveys using data from 26 countries with generalised epidemics for the year in which the survey was conducted. Appropriate transformations are applied to inform the correction factors needed to adjust prevalence in countries where population-based surveys have not been conducted.Results: HIV prevalence derived from antenatal clinic surveillance data generally overestimate population-based survey prevalence by about 20% (95% confidence interval: 10% to 30%) in both urban and rural areas.Conclusions: In countries where national population-based HIV surveys have been conducted, survey estimates of HIV prevalence (adjusted for potential survey biases as appropriate) can be used directly to calibrate antenatal clinic surveillance data. In countries where national HIV surveys have not been conducted, HIV prevalence derived from antenatal clinic surveillance data should be multiplied by about 0.8 to adjust for overestimation.