Artificial Intelligence for Perioperative Risk Prediction and Prevention in Cardiac Surgery: A Narrative Review and...

Area of Science:

• Cardiovascular Surgery
• Medical Artificial Intelligence
• Health Informatics

Background:

• Perioperative complications in cardiac surgery significantly impact patient outcomes, institutional performance, and healthcare costs.
• Current perioperative management relies on static risk stratification and reactive quality assessment, despite advances in surgical techniques.
• There is a need for more dynamic and predictive approaches to manage risks in cardiac surgery.

Purpose of the Study:

• To synthesize evidence on artificial intelligence (AI) and machine learning for perioperative risk prediction in cardiac surgery.
• To critically appraise the limitations and challenges of translating AI/ML models into clinical practice.
• To propose a framework for integrating predictive analytics into cardiovascular care.

Main Methods:

• This narrative review synthesizes existing literature on AI and machine learning applications in cardiac surgery risk prediction.
• The review covers various complications, including acute kidney injury, mortality, prolonged mechanical ventilation, postoperative atrial fibrillation, and ICU deterioration.
• Methodological limitations, validation gaps, and fairness concerns of current AI/ML models are critically appraised.

Main Results:

• AI and machine learning models demonstrate improved predictive discrimination compared to conventional risk scores for specific perioperative endpoints.
• Existing models are often endpoint-specific, developed at single institutions, and lack prospective clinical validation or implementation.
• Significant methodological limitations, validation gaps, and fairness concerns hinder the clinical translation of these predictive models.

Conclusions:

• Preventive Cardiovascular Intelligence (PCInt) is proposed as a framework integrating predictive analytics and quality improvement for anticipatory cardiovascular care.
• PCInt requires prospective validation and operationalization across the surgical lifecycle.
• Addressing implementation barriers, regulatory issues, and ethical considerations is crucial for future research and value-based perioperative care.