Prioritization of patients at risk of heart attack using a novel full-objective ITARA based on Random Forest and Decision tree

Abstract

Heart attacks remain a major cause of morbidity and mortality, particularly among middle-aged and older adults, often aggravated by unhealthy lifestyles and limited preventive care. Early identification and prioritization of at-risk individuals are essential to avoid severe complications. While prediction models exist, they often lack robust frameworks for prioritization across multiple clinical factors. This study addresses the gap by adapting Indifference Threshold Based Attribute Ratio Analysis (ITARA), a multi-criteria decision-making (MCDM) method, into a fully objective framework enhanced by machine learning. Unlike traditional ITARA, which relies on expert-defined thresholds, the proposed approach derives thresholds from variable importance scores generated by classification models. Among the models tested, Random Forest achieved 97% accuracy and was used to produce feature importance values. These scores refined ITARA, improving prioritization accuracy. Results identified the number of major vessels (Ca) as the most critical predictor. By integrating machine learning, the ITARA framework becomes a transparent, data-driven tool that improveWs both early detection and the clinical management of heart attack risks.

Keywords: Classification; Decision tree; Heart attack; Indifference Threshold-based Attribute Ratio Analysis (ITARA); Machine learning; Multiple criteria decision making; Random Forest.