AI-aided dynamic prediction of bleeding and ischemic risk after coronary stenting and subsequent DAPT

BackgroundContemporary risk scores for ischemic or bleeding event prediction after drug-eluting stent (DES) implantation are limited to the determination of a single time duration for dual antiplatelet therapy (DAPT) and lack flexibility in providing dynamic risk stratification. ObjectivesThis study sought to develop artificial intelligence (AI) models to dynamically predict the ischemic and bleeding risks at different time intervals for patients with DES implantation for personalized decision support for antiplatelet therapy. MethodsWe identified 81,594 adult patients who received DES implantation in the United States from the Cerner HealthFacts(R) dataset. The total prediction window covered 12-30 months after DES implantation. We designed eight prediction scenarios with four prediction intervals (3, 6, 12, and 18 months). Five AI models were developed for the ischemic and bleeding risk stratification. Model performance was assessed using the area under the receiver operating characteristic curve (AUROC). ResultsOur proposed AI model outperformed the clinical guideline-recommended tool-the DAPT score- for 12m-30m prediction (with AUROC 0.82 vs. 0.79 for ischemia, 0.77 vs 0.72 for bleeding). In the scenarios that are not covered by the DAPT score, our models demonstrated robust performance (AUROC ranges were 0.79-0.80 for ischemia and 0.75-0.76 for bleeding). ConclusionsAs the first effort dedicated to dynamically forecasting adverse endpoints after DES implantation given DAPT continuation or discontinuation, our AI-empowered approach demonstrates superior capabilities for risk stratification, holding value as a novel clinical tool that can refine the prognostic judgments of clinicians and achieve optimal DAPT management. Condensed abstractWe proposed an innovative AI-based dynamic prediction system that forecasts the ischemic and bleeding events after coronary stenting in varying time intervals given DAPT continuation or discontinuation. Our AI model not only demonstrated superiority compared with the clinical guideline-recommended tool-the DAPT score in the 12-30 months prediction, but also achieved robust performance in other scenarios that were not covered by the DAPT score. Our AI-driven approach holds value as a novel clinical tool that can refine the prognostic judgments of clinicians, enable better informed clinical decisions, and facilitate optimal DAPT management in the context of precision cardiovascular medicine.