Objectives: To explore which clinic-level factors influence treatment outcomes in a multisite antiretroviral therapy (ART) programme in South Africa.
Design: Retrospective cohort study using 36 clinics.
Methods: We used random effects modelling to investigate clinic-level factors influencing ART outcomes, adjusting for patient-level factors and accounting for clustering at clinic level. Outcomes were unsuppressed viral load (>400 copies/ml) at 24 months after ART start and time to loss to follow-up.
Results: At clinic level, the mean proportion of patients with unsuppressed viral load at 24 months was 16% (range 8-33%). Loss to follow-up was also highly variable across clinics ranging from 3.5 to 23.4/100 person-years. Unsuppressed viral load was associated with a lower doctor-patient ratio [for every 500 patients, compared with >2.6 doctors: <0.7 doctors: adjusted odds ratio (OR) 1.52, 95% confidence interval (CI) 1.04-2.21; 0.7-2.6 doctors, OR 1.33, CI 0.91-1.93, P trend 0.04] after adjustment for patient factors. Combinations of psychosocial support interventions were weakly associated with reduced loss to follow-up [>6 interventions vs. <4 interventions: hazard ratio 0.39 (CI 0.15 - 1.04), P = 0.11]. Flexibility of services, integration of services, staff motivation, staff leadership and location of clinic were not consistently associated with improved outcomes.
Conclusion: The dominant clinic-level influences on patient outcomes were doctor : patient ratio, and combination interventions to reduce loss to follow-up. Further research is needed to define optimum staffing levels that are required to roll out ART and the combination intervention that is most effective to reduce loss to follow-up.