Predicting Child Mortality: A Comparative Study of Discriminant Analysis and Logistic Regression Models

Child mortality remains a critical challenge in conflict affected Borno State, Nigeria, where insecurity, displacement, and poverty have heightened vulnerabilities. This study aimed to model under five child mortality using a Weibull accelerated failure time (AFT) model and identify predictors of survival in Maiduguri. A secondary dataset of 141 children was analyzed. The Weibull AFT model included gender, birth weight, immunization status, mother’s age, antenatal care (ANC) visits, place of delivery, residence, water source, malaria, and diarrhea history. The model showed excellent fit (likelihood ratio χ² = 40.56, p = 1.3×10; AIC full model = 171.9 vs. null = 192.5). The scale parameter (σ = 0.414) indicated highest mortality risk in early infancy. Significant protective factors were higher birth weight (HR = 0.55, p = 0.034), complete immunization (HR = 0.19, p = 0.002), and older maternal age (HR = 0.90, p = 0.027). ANC visits showed a paradoxical increased hazard (HR = 1.74, p = 0.006), suggesting reverse causality. Other variables were not significant. The Weibull model effectively identified key determinants of child survival. The ANC finding warrants further investigation. Future research should compare discriminant analysis and logistic regression against the Weibull baseline in humanitarian contexts.