Application of machine learning methods to perform mediation analyzes and interaction analyzes in the study of fetal mortality

• Grant number: 23/17594-2
• Support Opportunities: Scholarships in Brazil - Post-Doctoral
• Start date: May 01, 2024
• End date: April 30, 2026
• Field of knowledge: Health Sciences - Collective Health - Epidemiology
• Principal Investigator: Gizelton Pereira Alencar
• Grantee: Dario Ezequiel Elias
• Host Institution: Faculdade de Saúde Pública (FSP). Universidade de São Paulo (USP). São Paulo , SP, Brazil
• Associated research grant: 16/07765-0 - Fetal deaths: challenges for knowledge and intervention, AP.TEM

Abstract

The worldwide stillbirth rate has been estimated at 18.4 per 1,000 births, therefore, fetal mortality constitutes an important public health problem. However, there are still few epidemiological studies of fetal mortality that integrate biological, social and environmental data; and that analyze interactions and mediations between characteristics. Nevertheless, given the number of characteristics, and relationships among them, that can affect the fetal mortality risk, it is necessary to apply methods that guide the construction of models. In this sense, in the area of machine learning, methods have been developed for selecting features with greater predictive capacity, for example penalized regressions; and for the analysis of dependencies among characteristics, for example Bayesian networks. For those reasons, the main aim of this project is to generate knowledge about the characteristics associated with stillbirth by applying machine learning methods to identify possible causal pathways and interactions among explanatory variables. This project will be carried out based on biological, obstetric, social and environmental data collected in a prospective case-control study that will include 415 cases (stillbirths) and 415 controls (randomly selected live births) in municipal and state hospitals in the city of São Paulo. The results will be disseminated mainly in scientific communities.