Use of undetectable viral load to improve population-based survey estimates of known HIV-positive status and antiretroviral treatment coverage in Kenya

Introduction: Underreporting of prior HIV diagnosis and antiretroviral therapy (ART) use based on self-report is well-documented in national surveys. Antiretroviral (ARV) testing has been used to improve survey estimates, by reclassifying respondents with ARVs detected in blood as previously-diagnosed and on ART. Viral load testing, which is more affordable and more routinely available than ARV testing, is also an indicator of ART use. We examined the impact of adjusting estimated knowledge of HIV-positive status and antiretroviral therapy (ART) use based on self-report with biomarkers for antiretroviral drug detection and undetectable viral load (UVL). Methods: We reclassified HIV-positive participants aged 15-64 years in the 2012 Kenya AIDS Indicator Survey (KAIS) that were unaware of their HIV-positive status by self-report as aware and on ART if either ARVs were detected or viral load was undetectable (<550 copies/mL) on dried blood spots. We compared self-report to adjustments for ARV measurement, UVL, or both. We calculated measures of accuracy for UVL and UVL & ARV-adjusted versions of knowledge of status and ART use versus ARV-adjusted self-report as a reference standard. Results: Among 235 of 648 HIV-positive respondents with UVL, self-reported status was: 65 unaware (28.7%), 25 aware, not on ART (9.9%) and 145 aware, on ART (61.3%). Treatment coverage among all HIV-positive respondents increased from 31.8% for self-report to 42.5% [95% confidence interval (CI) 37.4-47.8] based on ARV detection alone, to 42.8% (95% CI 37.9-47.8) when ARV-adjusted, 46.2% (95% CI 41.3-51.1) when UVL-adjusted and 48.8% (95% CI 43.9-53.8) when adjusted for ARV and UVL. Awareness of positive status increased from 46.9% for self-report to 56.2% (95% CI 50.7-61.6) when ARV-adjusted, 57.5% (95% CI 51.9-63.0) when UVL-adjusted, and 59.8% (95% CI 54.2-65.1) when adjusted for ARV and UVL. Sensitivity and specificity of UVL-adjusted known HIV-positive status were 95.8% and 91.3%, and of UVL-adjusted ART use were 93.0% and 88.8% respectively, versus ARV-adjusted self-report. Conclusions: Undetectable viral load may be a useful adjunct or alternative to ARV detection for adjusting knowledge of status and ART use indicators in population-based surveys.