Background. National HIV serologic surveys increasingly rely on home-based testing and counseling (HBTC) in lieu of central laboratory testing to ensure immediate return of results.

Methods. The 2012 Kenya AIDS Indicator Survey used a two-stage stratified cluster design to obtain a representative sample of persons aged 15–64 years. Demographic information was collected, and both HBTC and central laboratory (enzyme immunoassays, EIA) testing were offered. If HIV-positive, testing for antiretroviral (ARV) drug metabolites was conducted. Prevalence was estimated for EIA, HBTC, and an adjusted method accounting for HBTC refusers that self-reported positive and/or had ARV metabolites. Models predicting HBTC refusal were developed, and estimates were weighted to account for sampling design and nonresponse.

Results. Of 11,626 EIA participants, 1,947 (16.7%) refused HBTC. HIV prevalence for EIA participants was 5.6% (95% confidence interval [CI] 4.9–6.3), versus 4.1% (CI 3.3–4.9) for HBTC. After accounting for HBTC refusers who self-reported positive, tested for ARVs, or both, HIV prevalence was 5.3% (CI 4.5–6.0), 5.3% (CI 4.5–6.1), and 5.9% (CI 5.1–6.7), respectively. HBTC refusers had significantly higher adjusted odds of being older (>24 years, AOR 1.2, CI 1.0–1.6); being wealthier (AOR 1.3, CI 1.0–1.6); having completed primary education (AOR 1.6, CI 1.3–1.9); having previously tested for HIV (AOR 1.2, CI 1.0–1.3); and being HIV positive (AOR 3.0, CI 2.3–3.9).

Conclusion. In Kenya, prevalence estimates based on HBTC is underestimated due to high refusal among HIV-infected individuals. Future studies may need to investigate and adjust for this bias.