Bayesian Accelerated Failure Time modelling of maternal mortality during pregnancy

Maternal mortality remains a critical global health issue, particularly in resource-constrained regions. This study introduces a novel Bayesian Accelerated Failure Time (AFT) model incorporating the Alpha Power Transformed Generalized Weibull (APTGW) distribution to improve survival analysis of maternal mortality during pregnancy. The APTGW model offers superior flexibility in capturing varying hazard shapes, making it particularly suited for maternal health data where risks evolve across pregnancy stages. A dataset of 549 maternal cases, including 9 mortalities, was analyzed using key predictors such as maternal age, gestational age at admission, antenatal care visits, gravidity, history of preeclampsia, and current preeclampsia status. Bayesian inference with Elastic Net regularization effectively addressed data sparsity and rare events, ensuring robust risk estimation. Model comparison using WAIC and LOOIC confirmed the APTGW model’s superior fit over Weibull and Lognormal distributions. Findings highlight that young maternal age, fewer than four antenatal visits, later gestational age at admission, previous abortion and preeclampsia (current or historical) significantly reduce survival time. The results underscore the urgent need for targeted interventions, particularly increasing antenatal care beyond WHO’s minimum recommendations and enhancing preeclampsia surveillance. The proposed Bayesian APTGW framework advances maternal survival analysis, offering refined risk factor insights to inform data-driven maternal health interventions.