BackgroundMany HIV voluntary testing and counselling centres in Africa use rapid antibody tests, in parallel or in sequence, to establish same-day HIV status. The interpretation of indeterminate or discrepant results between different rapid tests on one sample poses a challenge. We investigated the use of an algorithm using three serial rapid HIV tests in cohabiting couples to resolve unclear serostatuses.MethodsHeterosexual couples visited the Rwanda Zambia HIV Research Group testing centres in Kigali, Rwanda, and Lusaka, Zambia, to assess HIV infection status. Individuals with unclear HIV rapid antibody test results (indeterminate) or discrepant results were asked to return for repeat testing to resolve HIV status. If either partner of a couple tested positive or indeterminate with the screening test, both partners were tested with a confirmatory test. Individuals with indeterminate or discrepant results were further tested with a tie-breaker and monthly retesting. HIV-RNA viral load was determined when HIV status was not resolved by follow-up rapid testing. Individuals were classified based on two of three initial tests as "Positive", "Negative" or "Other". Follow-up testing and/or HIV-RNA viral load testing determined them as "Infected", "Uninfected" or "Unresolved".ResultsOf 45,820 individuals tested as couples, 2.3% (4.1% of couples) had at least one discrepant or indeterminate rapid result. A total of 65% of those individuals had follow-up testing and of those individuals initially classified as "Negative" by three initial rapid tests, less than 1% were resolved as "Infected". In contrast, of those individuals with at least one discrepant or indeterminate result who were initially classified as "Positive", only 46% were resolved as "Infected", while the remainder was resolved as "Uninfected" (46%) or "Unresolved" (8%). A positive HIV serostatus of one of the partners was a strong predictor of infection in the other partner as 48% of individuals who resolved as "Infected" had an HIV-infected spouse.ConclusionsIn more than 45,000 individuals counselled and tested as couples, only 5% of individuals with indeterminate or discrepant rapid HIV test results were HIV infected. This represented only 0.1% of all individuals tested. Thus, algorithms using screening, confirmatory and tie-breaker rapid tests are reliable with two of three tests negative, but not when two of three tests are positive. False positive antibody tests may persist. HIV-positive partner serostatus should prompt repeat testing.