Does an AI-driven T2G-Former model utilizing electronic health records accurately predict progression to acute coronary syndrome within 12 months in patients with stable coronary heart disease?
An AI-driven model utilizing longitudinal EHR data can accurately predict the 12-month risk of progression from stable coronary heart disease to acute coronary syndrome, offering a scalable tool for early clinical decision support.
Coronary heart disease (CHD), including its acute manifestation, acute coronary syndrome (ACS), remains a leading cause of morbidity and mortality worldwide. Early identification of ACS is critical for reducing the global burden of CHD, yet traditional diagnostic methods are often invasive, costly, and time consuming. In this study, we introduce an AI-driven approach that utilizes electronic health records (EHR) to identify transitions from stable CHD to ACS, using a comprehensive dataset of 12,336 patient records from 131 medical institutions in Jiangsu Province, China. The proposed model, applying the T2G-Former to predict ACS in 12 months, demonstrated superior performance, with an area under the curve of 0.953 and a sensitivity of 0.814 on the test set. Model interpretability was supported by SHAP values, which clarified the contribution of individual clinical variables. Our findings highlight the potential of AI-enabled EHR analysis for scalable early ACS detection and clinical decision support.
Hui et al. (Tue,) studied this question.