Artificial Intelligence in Cardiac Rehabilitation: Assessing ChatGPT's Knowledge and Clinical Scenario Responses

Area of Science:

• Cardiology
• Artificial Intelligence
• Digital Health

Background:

• Cardiac rehabilitation (CR) significantly improves patient outcomes but faces low participation due to access and socioeconomic barriers.
• Artificial intelligence (AI) presents a potential solution to enhance CR accessibility and effectiveness.

Purpose of the Study:

• To evaluate the potential of ChatGPT, an AI language model, in supporting cardiac rehabilitation.
• To assess ChatGPT's ability to provide guideline-aligned recommendations and patient motivation strategies for CR.

Main Methods:

• A cross-sectional study assessed ChatGPT-4's responses to 40 cardiology guideline-based questions.
• Responses were evaluated for guideline adherence by expert cardiologists, with inter-rater reliability measured using Cohen's kappa coefficient.

Main Results:

• ChatGPT-4 provided compliant responses to all questions, demonstrating strengths in risk stratification and patient safety.
• Limitations were identified in managing specific patient groups (elderly) and training protocols (high-intensity interval training).
• High inter-rater reliability (90%) confirmed the consistency of expert evaluations.

Conclusions:

• ChatGPT shows potential as a supplementary decision-support tool in CR, offering guideline-compliant information.
• Current limitations in contextual understanding and lack of clinical validation necessitate further development before independent use.
• Future research should focus on AI personalization, clinical validation, and seamless integration with healthcare professionals for optimal CR support.