Predicting postpartum hemorrhage in pregnant mothers in low-income settings. A machine learning approach.

Abstract

Postpartum hemorrhage (PPH) remains a significant contributor to maternal mortality globally, particularly in low-income countries where awareness and research on its severity, risk factors, and predictive modeling are limited. This project analyzed 2094 deliveries to identify PPH risk factors and developed a machine learning model for prediction. The Extreme Gradient Boosting model demonstrated superior performance with an AUC of 97.0%, accuracy of 96.0%, precision of 96.0%, and recall of 97.0%. Key identified risk factors associated with increased PPH include: Number of ANC Visits (P-Value: 0.00);Weight of Baby (g) (P-Value: 0.00); Duration of Labour (P-Value: 0.00); Cervical Tear (P-Value: 0.00), Episiotomy (P-Value: 0.00), and Perineal Tears (P-Value: 0.00). The study successfully established avalidated machine learning model capable of predicting mothers at risk of PPH.