Retention on ART and predictors of disengagement from care in several alternative community‐centred ART refill models in rural Swaziland

INTRODUCTION: A broad range of community‐centred care models for patients stable on anti‐retroviral therapy (ART) have been proposed by the World Health Organization to better respond to patient needs and alleviate pressure on health systems caused by rapidly growing patient numbers. Where available, often a single alternative care model is offered in addition to routine clinical care. We operationalized several community‐centred ART delivery care models in one public sector setting. Here, we compare retention in care and on ART and identify predictors of disengagement with care. METHODS: Patients on ART were enrolled into three community‐centred ART delivery care models in the rural Shiselweni region (Swaziland), from 02/2015 to 09/2016: Community ART Groups (CAGs), comprehensive outreach care and treatment clubs. We used Kaplan–Meier estimates to describe crude retention in care model and retention on ART (including patients who returned to clinical care). Multivariate Cox proportional hazard models were used to determine factors associated with all‐cause attrition from care model and disengagement with ART. RESULTS: A total of 918 patients were enrolled. CAGs had the most participants with 531 (57.8%). Median age was 44.7 years (IQR 36.3 to 54.4), 71.8% of patients were female, and 62.6% fulfilled eligibility criteria for community ART. The 12‐month retention in ART was 93.7% overall; it was similar between model types (p = 0.52). A considerable proportion of patients returned from community ART to clinical care, resulting in lower 12 months retention in care model (82.2% overall); retention in care model was lowest in CAGs at 70.4%, compared with 86.3% in outreach and 90.4% in treatment clubs (p < 0.001). In multivariate Cox regression models, patients in CAGs had a higher risk of disengaging from care model (aHR 3.15, 95% CI 2.01 to 4.95, p < 0.001) compared with treatment clubs. We found, however, no difference in attrition in ART between alternative model types. CONCLUSIONS: Concurrent implementation of three alternative community‐centred ART models in the same region was feasible. Although a considerable proportion of patients returned back to clinical care, overall ART retention was high and should encourage programme managers to offer community‐centred care models adapted to their specific setting.