Machine learning models trained on clinical data improve prediction of coronary artery disease on CT scans compared to traditional risk scores, potentially guiding better resource utilization in cardiovascular care.