Availability of Volunteer-Led Home-Based Care System and Baseline Factors as Predictors of Clinical Outcomes in HIV-Infected Patients in Rural Zambia

BACKGROUND: We assessed the impact of home-based care (HBC) for HIV+ patients, comparing outcomes between two groups of Zambians receiving antiretroviral therapy (ART) who lived in villages with and without HBC teams. METHODS: We conducted a retrospective cohort study using medical charts from Macha Mission Hospital, a hospital providing HIV care in Zambia's rural Southern Province. Date of birth, date of ART initiation, place of residence, sex, body mass index (BMI), CD4+ cell count, and hemoglobin (Hgb) were abstracted. Logistic regression was used to test our hypothesis that HBC was associated with treatment outcomes. RESULTS: Of 655 patients, 523 (80%) were eligible and included in the study. There were 428 patients (82%) with favorable outcomes (alive and on ART) and 95 patients (18%) with unfavorable outcomes (died, lost to follow-up, or stopped treatment). A minority of the 523 eligible patients (n = 84, 16%) lived in villages with HBC available. Living in a village with HBC was not significantly associated with treatment outcomes; 80% of patients in a village with HBC had favorable outcomes, compared to 82% of patients in a village without HBC (P = 0.6 by χ(2)). In bivariable analysis, lower BMI (P<0.001), low CD4+ cell count (P = 0.02), low Hgb concentration (P = 0.02), and older age at ART initiation (P = 0.047) were associated with unfavorable outcomes. In multivariable analysis, low BMI remained associated with unfavorable outcomes (P<0.001). CONCLUSIONS: We did not find that living in a village with HBC available was associated with improved treatment outcomes. We speculate that the ART clinic's rigorous treatment preparation before ART initiation and continuous adherence counseling during ART create a motivated group of patients whose outcomes did not improve with additional HBC support. An alternative explanation is that the quality of the HBC program is suboptimal.