Keywords: Atrial Fibrillation, Emergency Medicine, Primary Health Care, Clinical Decision-Making, Artificial Intelligence

Background:

Atrial fibrillation (AF) is a common arrhythmia encountered in both emergency departments and primary care, associated with significant morbidity and thromboembolic risk. Despite the availability of up-to-date clinical practice guidelines (CPGs), such as the 2024 ESC recommendations, substantial variation in clinical management persists. Concurrently, artificial intelligence (AI) tools are emerging as potential aids in clinical decision-making. However, the degree to which real-world management aligns with guideline recommendations —and how human decisions compare to those suggested by AI models trained specifically for this context— remains unclear.

Research questions:

To what extent do actual clinical decisions and AI-based suggestions align with current clinical practice guidelines in patients with AF treated in Emergency and Primary Care settings?

Method:

This is an ongoing, retrospective observational and descriptive study. Cases of AF managed at the Hospital Clínico Universitario of Zaragoza and affiliated primary care centers will be selected. Clinical, demographic, and therapeutic variables will be collected. Each case will be assessed from three perspectives: actual clinical decisions, current guideline-based recommendations and simulated decisions produced by a clinically trained conversational AI model. The degree of concordance among these sources will be analyzed using descriptive statistics and agreement coefficients.

Results:

The study is currently in progress. It is expected to identify patterns of adherence and deviation from guideline-based management in both actual clinical decisions and AI-generated recommendations. Additionally, the potential of the AI model as an educational or clinical decision-support tool will be evaluated.

Conclusions:

This project aims to assess the appropriateness of real-world AF management and the potential role of AI in enhancing clinical decision-making, with implications for practice in both Emergency and Primary Care settings.

Points for discussion:

Can AI improve real-world adherence to clinical practice guidelines in both Emergency and Primary Care settings for atrial fibrillation management?

What are the ethical and practical implications of integrating AI-based decision support tools into everyday clinical workflows?

Could conversational AI models serve as effective educational tools for healthcare professionals in training or practice?