Validation of a self‐report adherence measurement tool among a multinational cohort of children living with HIV in Kenya, South Africa and Thailand

INTRODUCTION: There are few data on adherence and low‐cost measurement tools for children living with HIV. We collected prospective data on adherence to antiretroviral therapy (ART) among a multinational cohort of children to evaluate an adherence questionnaire. METHODS: We enrolled 319 children ages 0 to 16 years on ART in Kenya (n = 110), South Africa (n = 109) or Thailand (n = 100). Children were followed up for six months of adherence monitoring between March 2015 and August 2016 using Medication Event Monitoring Systems (MEMS (®)) with at least one viral load measure. At month 3 and 6, children or their caregivers were administered a 10‐item adherence questionnaire. Repeated measures analyses were used to compare responses on questionnaire items to external adherence criteria: MEMS (®) dichotomized adherence (≥90% of doses taken vs. <90%), 48‐hour MEMS (®) treatment interruptions and viral suppression (<1000 copies/mL). Items associated with outcomes (p < 0.10) were coefficient‐weighted to calculate a total adherence score, which was tested in multivariate regression against MEMS (®) and viral suppression outcomes. Odds ratios (OR) and 95% confidence intervals (95% CI) were calculated. RESULTS: Mean child age was 11 years and 54% were female. Children from Thailand (median age 14 years) were significantly older compared to Kenya (10 years) and South Africa (10 years). Prevalence of viral suppression was 97% in Thailand, 81% in South Africa and 69% in Kenya, while the prevalence of MEMS (®) adherence ≥90% was 57% in Thailand, 58% in South Africa and 40% in Kenya. Across sites, child‐reported adherence using the questionnaire was significantly associated with dichotomized MEMS (®) adherence (OR 1.8, 95% CI 1.4 to 2.4), 48‐hour treatment interruptions (OR 0.41, 95% CI 0.3 to 0.6), and viral suppression (OR 3.4, 95% CI 1.7 to 6.7). We did find, however, that different cut‐points for the adherence score may be context‐specific. For example, MEMS (®) non‐adherent children in Kenya had a lower adherence score (0.98) compared to South Africa (1.77) or Thailand (1.58). CONCLUSIONS: We found suboptimal adherence to ART was common by multiple measures in this multi‐country cohort of children. The short‐form questionnaire demonstrated reasonable validity to screen for non‐adherence in these diverse settings.