DEVELOPMENT AND VALIDATION OF AN EXPLAINABLE AI-BASED MATERNAL RISK PREDICTION MODEL FOR EARLY DETECTION OF OBSTETRIC COMPLICATIONS IN PAKISTAN

Abstract

Maternal mortality and obstetric complications remain critical public health concerns in Pakistan, largely due to delayed diagnosis and inadequate early risk detection systems. Traditional clinical assessment methods are often limited in their ability to analyze complex, multidimensional maternal health data, resulting in preventable adverse outcomes. This study developed and validated an Explainable Artificial Intelligence (XAI)-based maternal risk prediction model to enhance early detection of obstetric complications in Pakistan. A quantitative, machine learning-based approach was applied using maternal health records comprising clinical, demographic, and socio-economic variables. Multiple classification algorithms, including Logistic Regression, Decision Tree, Random Forest, Support Vector Machine, and XGBoost, were evaluated. Model interpretability was enhanced using SHAP-based Explainable AI techniques to ensure transparency in predictions. The results indicated that XGBoost outperformed all other models, achieving the highest predictive accuracy and AUC score. Key predictors of maternal risk included blood pressure, hemoglobin levels, and antenatal care visits. The integration of Explainable AI significantly improved model transparency and clinical interpretability, supporting its potential adoption in real-world healthcare settings. The study concludes that explainable machine learning models can effectively support early identification of high-risk pregnancies and improve maternal healthcare outcomes in Pakistan.