Decoding health disparities: gender, ethnicity, & chronic diseases in Latin Americans with individual data & machine learning

Chronic diseases disproportionately affect ethnic and gender groups, yet the social determinants driving these disparities in Latin America are not fully understood. In this study, we analyzed data from 96,726,891 adults obtained from national health surveys in Brazil, Mexico, and Ecuador during 2018–2019. We employed random forest models to predict chronic disease diagnoses based on education, occupation, and access to essential services like sewage, piped water, and garbage collection. Our models performed best for indigenous and black individuals, underscoring significant inequities. Education emerged as a stronger predictor for women, while occupation was more influential for men. Specifically, removing education data reduced model performance for women by 59.6%, whereas removing occupation data reduced performance for men by 31.6%. These findings highlight the need for public policies tailored to the unique needs of different gender and ethnic groups—promoting improved employment opportunities for men, enhanced educational access for women, and better housing conditions for indigenous and black populations.