Integrating feature importance techniques and causal inference to enhance early detection of heart disease

Abstract

Heart disease remains a leading cause of mortality worldwide, necessitating robust methods for its early detection and intervention. This study employs a comprehensive approach to identify and analyze critical features contributing to heart disease. Using a dataset of 270 patients, three well-known feature importance techniques--Boruta, Information Gain, and Lasso Regression--are applied to determine the top five features for heart disease detection. Following the identification of these key features, the g-computation method, a causal inference technique, is utilized to explore the causal relationships between these features and the presence of heart disease. The findings provide valuable insights into not only the features that are highly correlated with chronic heart disease but also those that have a direct causal impact on the classification of patients. This integrated approach enhances the understanding of heart disease etiology and can inform more effective diagnostic and therapeutic strategies.