Artificial Intelligence for Cardiovascular Risk Prediction: An Umbrella Review of Applications and Translational...

Area of Science:

• Cardiology
• Medical Informatics
• Artificial Intelligence

Background:

• Cardiovascular diseases (CVDs) are a leading global cause of death.
• Traditional risk models lack accuracy and generalizability.
• AI, machine learning (ML), and deep learning (DL) offer advanced methods for personalized CVD risk estimation using multimodal data.

Purpose of the Study:

• To conduct an umbrella review of AI applications in cardiovascular risk prediction.
• To assess the predictive performance of AI models.
• To identify translational limitations of AI in cardiology.

Main Methods:

• Systematic search of PubMed, Scopus, and Web of Science (Jan 2015-Oct 2025).
• Inclusion of RCTs, cohort studies, systematic reviews, and meta-analyses.
• Primary outcome: predictive performance (Area Under the Curve - AUC); secondary: methodological quality assessment.

Main Results:

• AI models exceeded 0.90 AUC in over 70% of imaging studies.
• Multimodal data integration improved detection of coronary artery disease (CAD) and heart failure (HF).
• Wearable monitoring showed 18-25% lower hospitalization rates versus usual care.

Conclusions:

• AI demonstrates strong predictive accuracy (AUC) for cardiovascular risk.
• Barriers to implementation include methodological heterogeneity, bias, and lack of external validation.
• Clinical translation requires multicenter trials, explainable AI, and standardized guidelines like TRIPOD-AI.