Background: HIV rapid tests (RT) are a quick and non-technically demanding means to perform HIV voluntary counselling and testing (VCT) but understanding their limitations is vital to delivering quality VCT.
Objective: To determine the sensitivity and specificity of HIV rapid tests used for research and voluntary counselling and testing at four sites in East Africa.
Design: Cross-sectional study.
Setting: Masaka District, Uganda; a sugar plantation in Kakira, Uganda; Coastal Villages in the Kilifi District of Kenya; and the Urban slum of Kangemi located West of Nairobi, Kenya.
Subjects: Six thousands two hundred and fifty five consenting volunteers were enrolled into the study, and 675 prevalent HIV infections were identified.
Results: The RT sensitivity tended to be high for all assays at all sites (97.63-100%) with the exception of the Uni-Gold assay (90.24% in Kangemi, 96.58% in Kilifi). Twenty four RT results were recorded as 'weak positives', 22 (92%) of which were negative by ELISA. There was a high rate of RT false positives in Uganda (positive predictive values ranging from 45.70% to 86.62%).
Conclusions: The sensitivity and specificity of the RT varied significantly across sites. The rate of RT misclassification in Uganda suggests that a multiple test algorithm may be preferable to a single test as screener for HIV VCT.