Data Science Application for Creation of Maternal Morbidity and Mortality Predictive Software

Abstract

In Mexico, the estimated Maternal Mortality Ratio is 34.6 deaths per 100,000 estimated births. Consequently, healthcare facilities and services have given precedence to prenatal care, childbirth services, and postpartum care. In Mexico, the Ministry of Health maintains an open database concerning maternal deaths, encompassing 58 variables. Among these variables is the CIE (International Statistical Classification of Diseases and Related Health Problems), which covers a total of 248 diseases linked to maternal deaths. Currently, there is no software that classifies women undergoing pregnancy check-ups (according to their socio-clinical risk of mortality), using variables selected with data science. This project is rooted in the methodology advanced by International Business Machines (IBM) for the implementation of data science. The software's utilized model was constructed through the Naïve Bayes supervised learning algorithm, yielding an accuracy of 0.7236. The overall precision stood at 0.75, with an overall recall of 0.74, and an overall F1-score of 0.71. For the eclampsia during labor class, precision reached 0.71, recall was 0.94, and the F1- score attained 0.81. As for secondary or late postpartum hemorrhage, precision scored 0.81, recall measured 0.43, and the F1-score was 0.56.