TRANSFER LEARNING APPLICATIONS FOR THE PREDICTION OF MORTALITY AND SEVERE MATERNAL MORBIDITY: A MULTICENTRIC ANALYSIS IN EIGHT LOW AND MID-INCOME COUNTRIES

• Grant number: 24/07552-3
• Support Opportunities: Scholarships in Brazil - Post-Doctoral
• Start date: April 01, 2025
• End date: September 30, 2026
• Field of knowledge: Health Sciences - Collective Health - Epidemiology
• Principal Investigator: Alexandre Dias Porto Chiavegatto Filho
• Grantee: Marcela Quaresma Soares
• Host Institution: Faculdade de Saúde Pública (FSP). Universidade de São Paulo (USP). São Paulo , SP, Brazil

Abstract

Maternal mortality remains a global challenge, especially in low and middle-income countries (LMICs), where 94% of maternal deaths occur. Despite advancements in the production, quality, and availability of data on maternal health and social conditions, traditional analysis methods fail to capture the specificities and interrelations among the various factors involved in maternal adverse events. In this context, the present project aims to use machine learning (ML) techniques to predict adverse maternal events, such as mortality and severe morbidity in LMICs, and adapt the information obtained to the Brazilian reality. The analysis will utilize data from eight LMICs from the Global Network's Maternal Newborn Health, which will serve as the basis for learning transfer learning (TL) algorithms for the Brazilian reality. TL is an ML approach that uses knowledge acquired in a source domain to improve performance in a target domain by fine-tuning pre-trained models or extracting relevant features. This study is expected to generate scientific knowledge on the use of TL for the prediction of maternal adverse events, developing models that assist in the efficient allocation of resources and interventions, the individualization of care, and the optimization of the management of high-risk cases. The implementation of these strategies aims to contribute to improving health and reducing maternal mortality.