C25 PREDICTION OF ALL–CAUSE MORTALITY FOLLOWING PERCUTANEOUS CORONARY INTERVENTION IN BIFURCATION LESIONS USING MACHINE LEARNING ALGORITHMS – THE RAIN–ML PREDICTION MODEL

Abstract

Abstract Aims Stratifying prognosis following coronary bifurcation percutaneous coronary intervention (PCI) is an unmet need. Machine learning (ML) may identify patterns from multidimensional, non–linear relationships to make outcome predictions. We sought to develop a ML–based risk stratification model built on clinical, anatomical and procedural features to predict all–cause mortality following contemporary bifurcation PCI. Methods and Results Multiple ML models to predict all–cause mortality were tested on a cohort of 2,393 patients (training, n = 1,795; internal validation, n = 598) undergoing bifurcation PCI with contemporary stents from the real–world RAIN (veRy thin stents for patients with left mAIn or bifurcatioN in real life) registry. Among 38 commonly available features, 25 (13 patient–related, 12 lesion–related) were selected to train ML models. The best performing model (the RAIN–ML prediction model) was validated in an external validation cohort of 1,701 patients undergoing bifurcation PCI from the DUTCH PEERS (DUrable polymer–based sTent CHallenge of Promus ElemEnt versus ReSolute integrity: TWENTE II) trial and the BIO–RESORT trial cohorts. The area under the receiver operating characteristic curves for the prediction of 2–year mortality was 0.786 (0.74–0.83) in the overall population, 0.736 (0.72–0.847) at internal validation and 0.706 (0.6919–0.794) at external validation. Performance at risk ranking analysis, k–center cross validation, and with continual learning confirmed the generalizability of the models, available also as an online interface. Conclusions The RAIN–ML prediction model represents the first tool combining clinical, anatomical and procedural features to predict all–cause mortality among patients undergoing contemporary bifurcation PCI with a good discriminative performance.