Preeclampsia Prediction Using Machine Learning with Electronic Medical Records in Low-Resource Settings

Preeclampsia is a serious pregnancy complication characterized by hypertension and proteinuria that can result in severe maternal and fetal morbidity. Early prediction and timely intervention are essential to improve outcomes. This study aimed to evaluate the effectiveness of various machine learning models in predicting preeclampsia, focusing on their ability to minimize false negatives and enhance clinical decision-making. A publicly available dataset comprising patient records was used for model evaluation. Several machine learning models, including Logistic Regression (LR),Passive Aggressive Classifier (PAC), Extra Trees Classifier (ETC), Support Vector Machines (SVM), and CatBoost, were implemented. Performance metrics, such as precision, recall, F1-score, sensitivity, and specificity, were calculated. Among the models tested, CatBoost outperformed the others, achieving the highest AUC of 0.9992 and an F1-score of 0.9812, demonstrating its strong predictive capability. SVM, LR, ETC, and PAC also showed competitive performance, while Random Forest exhibited the weakest classification ability. These findings highlight the potential of machine learning models, particularly CatBoost, in predicting preeclampsia conditions. The integration of such models into clinical practice could improve early prediction and intervention, ultimately enhancing maternal and fetal health outcomes. Future research should focus on validating these models with diverse datasets and integrating them into clinical decision support systems to further optimize their applicability in real-world settings.