Key points are not available for this paper at this time.
Background: Cardiovascular diseases remain the leading cause of morbidity and mortality world-wide. Traditional risk scores, although widely used, rely on a limited number of linearly weighted variables and may perform poorly in heterogeneous populations. The growing availability of multimodal data, including electronic health records, imaging, wearable-device data, and omics, has increased interest in artificial intelligence (AI) for more individualized cardiovascular risk prediction. Objective: This narrative review examines the current role of AI in cardiovascular disease prediction, with particular attention to model performance, methodological limitations, validation standards, and barriers to clinical implementation. Methods: A structured narrative review of peer‑reviewed literature published up to December 2025 was conducted. Targeted searches in PubMed, IEEE Xplore, and Scopus were supplemented by queries of clinical trial registries and regulatory/health‑technology assessment sources. The synthesis prioritizes external validation, calibration, implementation science, health‑economic evaluation, and equity considerations. Results: Recent evidence suggests that machine learning and deep learning models may achieve better discrimination than conventional clinical risk scores, with pooled area under the curve values often reported in the range of 0.85 to 0.87. However, the literature is marked by substantial heterogeneity, inconsistent calibration reporting, limited external validation, and underuse of time-to-event modeling approaches. These weaknesses reduce reproducibility and hinder safe translation into routine practice. Conclusions: AI has considerable potential to improve cardiovascular risk prediction, but clinical adoption requires stronger methodological rigor, prospective validation, recalibration strategies, trans-parent reporting, and evaluation of ethical, organizational, and economic consequences.
Wieczorek et al. (Fri,) studied this question.