Predicting mother-to-child HIV transmission among mother-baby pairs in Kenya: a focused comparison of Random Forest and XGBoost models

Abstract

HIV remains one of the most pressing global public health concerns, with the latest estimates indicating that approximately 39.9 million people are living with the virus worldwide (WHO, 2024). The epidemic’s burden disproportionately affects specific populations, notably women and adolescent girls, who constitute a significant proportion of new infections. In 2023, sub-Saharan Africa (SSA) was particularly impacted, with 62% of all new HIV infections occurring among girls and women (Joint United Nations Programme, 2024). This underscores the complex social, economic, and healthcare disparities that continue to fuel the epidemic. Among those most affected by HIV are pregnant women living with HIV (PWLHIV), who face unique health challenges (USAID, 2024) risking not only their own well-being but also that of their children. Each year, approximately 1.3 million girls and women living with HIV become pregnant, highlighting the urgent need for interventions that address the risks associated with pregnancy and HIV co-management (WHO, 2024). Mother-to-child transmission (MTCT) remains the primary route for HIV transmission to infants, accounting for most paediatric HIV cases worldwide (Gill et al., 2020) and can occur at various stages - during pregnancy, labour and delivery, or breastfeeding WHO (2024). Despite significant global efforts and strides made in preventing mother-to-child transmission (PMTCT) through antiretroviral therapy (ART), breastfeeding alternatives, and other interventions, cases of HIV transmission from mother to child continue to arise (White AB et al., 2024). According to UNAIDS, new HIV infections among children have significantly reduced over the years, yet an estimated 120,000 cases still occurred in 2023 (Joint United Nations Programme, 2024).