Statistical and Machine Learning Analysis of Influence Factors on Maternal Health Risk

Abstract

One in four maternal deaths in low-resource countries, such as Liberia. We include a data-driven approach with statistical methods and machine learning (ML) to assess maternal health risks and policy in this study. We applied correlation analysis, multinomial logistic regression, ML algorithms (decision tree, random forest) to predict maternal health risk categories, using a dataset of 1014 patients. The results indicated that key predictors included age, blood pressure and blood sugar. Indeed, we outperform traditional model in terms of accuracy 85.3% accuracy for the random forest model. We recommend that data including ML tools be merged into national healthcare M&E systems that would suggest its beneficial allocation and prevention of chronic diseases. This study contributes to the field of maternal health analytics and informs evidence-based policy making in Liberia.