Rwanda has achieved the global 95–95–95 HIV targets, but sustaining retention in HIV care remains essential for maintaining viral suppression and reducing HIV-related morbidity and mortality.
Women receiving antiretroviral therapy (ART), especially younger women and those experiencing treatment or demographic instability, remain at increased risk of loss to follow-up (LTFU).
Machine learning models using routine national HIV case-based surveillance (CBS) data can accurately identify women at high risk of disengagement before prolonged interruption of care occurs.
Random Forest models demonstrated strong predictive performance for identifying women at risk of LTFU, with recall performance of 73.9% and an AUC of 78%.
Integrating predictive analytics into Rwanda’s national HIV program could strengthen differentiated service delivery, improve targeted retention interventions, and reduce avoidable treatment interruption among women living with HIV.
