Health inequalities are globally widespread due to the regional socioeconomic inequalities. Myocardial infarction (MI) is a leading health problem causing deaths worldwide. Yet medical services for it are often inequitably distributed by region. Moreover, studies concerning MI's potential spatial risk factors generally suffer from difficulties in focusing on too few factors, inappropriate models, and coarse spatial grain of data. To address these issues, this paper integrates registered 1098 MI cases and urban multi-source spatio-temporal big data, and spatially analyses the risk factors for MI severity by applying an advanced interpretable model, the random forest algorithm (RFA)-based SHapley Additive exPlanations (SHAP) model. In addition, a community-scale model between spatio-temporal risk factors and MI cases is constructed to predict the MI severity of all communities in Wuhan, China. The results suggest that those risk factors (i.e., age of patients, medical quality, temperature changes, air pollution and urban habitat) affect the MI severity at the community scale. We found that Wuhan residents in the downtown area are at risk for high MI severity, and the surrounding suburb areas show a donut-shape pattern of risk for medium-to-high MI severity. These patterns draw our attention to the impact of spatial environmental risk factors on MI severity. Thus, this paper provides three recommendations for urban planning to reduce the risk and mortality from severe MI in the aspect of policy implication.
Keywords: GIS spatial Analysis; Interpretable machine learning; Myocardial infarction; Spatio-temporal factors; Urban environment.
Copyright © 2022 Elsevier Ltd. All rights reserved.