Background In resource-limited settings, HIV infection is often diagnosed using two rapid tests. If the results are discordant, a third tie-breaker test is often used to determine HIV status. This study characterized samples with discordant rapid tests and compared different testing strategies for determining HIV status in these cases. Methods Samples were previously collected from 173 African adults in a population-based survey who had discordant rapid test results. Samples were classified as HIV positive or HIV negative using a rigorous testing algorithm that included two fourth-generation tests, a discriminatory test, and two HIV RNA tests. Tie-breaker tests were evaluated, including: rapid tests (one performed in-country), a third-generation enzyme immunoassay (EIA), and two fourth-generation tests. Selected samples were further characterized using additional assays. Results Twenty-nine (16.8%) samples were classified as HIV positive; 24 (82.8%) of those samples had undetectable HIV RNA. Antiretroviral drugs were detected in one sample. Sensitivity was 8.3%–43% for the rapid tests; 24.1% for the third-generation EIA; 95.8% and 96.6% for the fourth-generation tests. Specificity was lower for the fourth-generation tests than the other tests. Accuracy ranged from 79.5–91.3%. Conclusions In this population-based survey, most HIV-infected adults with discordant rapid tests were virally suppressed without antiretroviral drugs. Use of individual assays as tie-breaker tests was not a reliable method for determining HIV status in these individuals. More extensive testing algorithms that use a fourth-generation screening test with a discriminatory test and HIV RNA test are preferable for determining HIV status in these cases.