Determinants of adherence to antiretroviral therapy among patients in Busia, Kenya

Antiretroviral therapy (ART) requires sustained adherence of at least 95% to achieve durable viral suppression. Kenya has made substantial national progress toward the UNAIDS 95-95-95 targets, yet locally disaggregated, peer-reviewed evidence on ART adherence remains limited, particularly in high-mobility border regions. This study examined the prevalence of ART non-adherence and its determinants among adult patients in Busia County, Kenya. This multi-facility, cross-sectional study interviewed 300 adult (≥ 18 years) patients enrolled in ART care at Level 4 and 5 public health facilities in Busia County, selected using probability-proportional-to-size and systematic random sampling. Structured interviews (in Kiswahili) captured socio-demographic, cultural, psychosocial, health-service, and socio-economic factors, supplemented with Electronic Medical Record (EMR) data. The primary outcome, adherence status, was defined from EMR pharmacy pick-up records (non-adherent if a scheduled pick-up was missed by more than 28 days in the preceding 6 months); viral load, and pill count were abstracted as secondary corroborating indicators. Bivariate associations were assessed using chi-square tests. Independent predictors of adherence were identified using multivariable logistic regression, cross-checked against a mixed-effects model accounting for clustering by facility, with statistical significance set at p <.05. Of 300 participants (40.0% male, 60.0% female), 272 (90.7%) were classified as adherent and 28 (9.3%) as non-adherents. In the final multivariable model, female sex was significantly associated with lower odds of non-adherence (logistic OR = 0.48, 95% CI [0.23, 1.00], p =.049), as was higher educational attainment (OR = 0.56, 95% CI [0.35, 0.89], p =.015). Both associations were confirmed in a facility-clustered mixed-effects model (sex: β = −0.091, p =.035; education: β = 0.064, p =.012). No other factor examined including marital status, social support, cultural beliefs, religious beliefs, stigma, health-service rating, or difficulty balancing ART with a prescribed diet was independently associated with adherence in either model. In this Busia County cohort, sex and educational attainment were the only factors independently and consistently associated with ART adherence across two different modelling approaches. Several commonly cited barriers were not independently significant, most plausibly reflecting limited statistical power given the small non-adherent subgroup rather than a true absence of effect. Sex-responsive, literacy-appropriate adherence support is recommended, alongside adequately powered follow-up research to clarify the remaining determinants.