Longitudinal Patterns of HIV Medication Adherence at an Urban HIV Clinic

Background: Medication adherence is one of the most important factors for HIV treatment success. Considering the heterogeneity of adherence patterns, HIV population would consist of groups of people with distinct adherence patterns. Thus far, only a few studies have examined the long-term adherence patterns of HIV-infected people. Methods: 473 HIV-infected patients without HAART who visited the Johns Hopkins HIV Clinic between January 2004 and December 2015 were enrolled. Daily adherence was defined as pills taken divided by pills prescribed one day. One-year adherence was calculated via average daily adherence during the period; it was assessed every 180 days from HAART initiation between January 2004 and December 2014. Latent class growth analysis was performed to identify distinct adherence trajectories in the sample. Two analyses of joint latent class growth were used to examine the link between adherence trajectories and mortality or virologic failure. Finally, the association between HAART duration and adherence was assessed via general estimating equation (GEE) analysis. Results: Five adherence patterns were identified: constantly high (17% of sample), decreasing (6%), constantly low (62%), fluctuating (8%), and increasing (8%). Three latent classes of adherence patterns were identified through the joint model with mortality: high (24%), fluctuating (6%), and low (70%). Low adherence class and fluctuating adherence class was associated with the increased mortality. Five latent classes of adherence patterns were identified through the joint model with virologic failure: high (14%), middle (46%), decreasing (5%), increasing (13%) and low (22%). Only low adherence class was associated with the increased virologic failure rate. HAART duration was not significantly associated with adherence change in the GEE analysis (adjusted OR: 1.00 95% CI: [0.95, 1.06]). Conclusion: This study describes distinct longitudinal patterns of medication adherence among people living with HIV (PLWH). Joint latent growth analysis shows that adherence patterns were associated with mortality and virologic failure. Further research is needed to find what factors determine adherence patterns in order to improve medication adherence