Using Explainable Artificial Intelligence for Mapping Health Vulnerability: Interaction-Based Analysis of Multiple Sources of Data in Latin America
Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Ethnic diversity, geographic isolation, access to medical care, and genetic predisposition all interplay with one another and impact the health disparities noted in Latin America. This study presents a new, explainable AI (XAI) framework that detects and characterize regional health vulnerability using data integrated from multiple, open-source domains, such as disease risk, medical supply chains, and genetic ancestry. By employing a common custom Interaction Index, machine learning classifiers and clustering algorithms, as well as decision tree modelling, the study aims to uncover latent inequality patterns across Latin American regions. The methodology champions transparency and interpretability, prioritizing congruence with international standards for ethical AI in healthcare. The findings show that regions with high levels of ethnic diversity, higher genetic risk and lower access to medical care — and particularly mid to high altitudes — exist in clusters of extreme vulnerability. The insights gained expand upon and contribute to the emerging literature by offering interpretable boundaries and geographic insights to inform equitable resource allocation and infrastructure planning. This work provides a replicable and scalable model of how health system stakeholders can use data to inform decision-making when trying to ameliorate structural disparities.
