A Machine Learning Model for Predicting the Risk of Perinatal Mortality in Low-and-Middle-Income Countries: A Case Study

Perinatal mortality is the death that happens between 22 weeks of gestation and the first seven days of birth. This has become an essential indicator for measuring the quality of maternal and childcare in Low-and-Middle-Income Countries (LMICs). Tools based on Artificial Intelligence (AI) have emerged with immediate relevance in medical contexts, more precisely with Machine Learning (ML) tools due to the ability to learn from past and present observations and be able to generate future predictions, promising positive results in maternal and childcare processes. This paper presents a Random Forest (RF) model for predicting the risk of perinatal mortality in LMICs. We initially characterized the prenatal control process in LMICs. Second, potentially predictive features of perinatal mortality were identified considering the literature review and medical expertise. Subsequently, a data pre-processing procedure was executed to improve the data quality. The RF algorithm was employed to model the risk of perinatal mortality based on social and clinical variables. A case study in a Colombian healthcare institution was used to validate the proposed approach. The results show an RF model with an accuracy of 99.16%, sensibility = 87.50%, and specificity = 100%.