Understanding the social determinants of child mortality in Latin America over the last two decades: a machine learning approach

Abstract

Abstract The reduction of child mortality rates remains a significant global public health challenge, particularly in regions with high levels of inequality such as Latin America. We used machine learning (ML) algorithms to explore the relationship between social determinants and child under-5 mortality rates (U5MR) in Brazil, Ecuador, and Mexico over two decades. We created a municipal-level cohort from 2000 to 2019 and trained a random forest model (RF) to estimate the relative importance of social determinants in predicting U5MR. We conducted a sensitivity analysis training two more ML models and presenting the mean square error (MSE), root mean square error (RMSE), and median absolute deviation (MAD). Our findings indicate that poverty, illiteracy, and the Gini index were the most important variables for predicting U5MR according to the RF. Furthermore, non-linear relationships were found mainly for Gini index and U5MR. Our study suggests that long-term public policies to reduce U5MR in Latin America should focus on reducing poverty, illiteracy, and socioeconomic inequalities. This research provides important insights into the relationships between social determinants and child mortality rates in Latin America. The use of ML algorithms, combined with large longitudinal data, allowed us to evaluate the effects of social determinants on health more carefully than traditional models.