The model achieved a very good classification performance, and the derived risk groups adequately defined distinct risk trajectories of time-to-event data. These risk clusters outperformed the classification performance of the HEART score, with a significant gain in recall, precision, and negative predictive value. This performance generalized well to data from an external clinical site. ML, machine learning; ECG, electrocardiogram; HEART, History, ECG, Age, Risk factors, Troponin; NPV, negative predictive value; OR, odds ratio. Created with BioRender.com (credit to Z.B.)¹⁹