Predictive Analysis of Determinants of Treatment Outcomes in DR-TB/HIV Coinfection in a High TB Setting of Eastern Cape

Background: Drug-resistant tuberculosis (DR-TB) remains a major threat to TB control, with success rates below global targets. While clinical determinants such as resistance type are well established, the role of social and behavioral factors is less clearly defined. This study examined both clinical and socioeconomic predictors of DR-TB outcomes in the Eastern Cape, South Africa. Methods: A retrospective cohort analysis was conducted using routinely collected data. Outcomes were collapsed into successful (cured/completed) and unsuccessful (failure, death, loss to follow-up). Descriptive statistics and cross-tabulations compared outcome distributions across demographic, socioeconomic, and clinical variables. Logistic regression estimated adjusted odds ratios (ORs) with 95% confidence intervals (CIs). Random forest modelling assessed predictive performance and ranked feature importance. Results: Cross-tabulations showed significant associations between treatment outcome and gender (p=0.046), income (p=0.0037), and DR-TB type (p=0.0355). Logistic regression confirmed that males had higher odds of success than females (OR=2.11, 95% CI: 1.05–4.21), while salaried patients performed better than those without income (OR=3.46, 95% CI: 0.39–30.96). Pre-XDR TB was associated with reduced odds of success compared to RR-TB (OR=0.25, 95% CI: 0.05–1.19). The logistic model showed modest discrimination (AUC≈0.55). Random forest modelling achieved superior performance and identified age as the most important predictor, followed by patient category, income, social history, education, and DR-TB type. Conclusion: Both clinical and social factors shape DR-TB outcomes. Gender, income, and resistance patterns were consistently influential, while machine-learning analysis highlighted age and socioeconomic determinants. Integrated strategies addressing biomedical and social drivers are essential to improve treatment success in high-burden settings.