Evaluating atrial fibrillation artificial intelligence for the ED: statistical and clinical implications.

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Diagnostics

Background:

• Atrial fibrillation (AF) is a common arrhythmia.
• Early detection of AF can improve patient outcomes.
• Electrocardiograms (ECG) are standard diagnostic tools.

Purpose of the Study:

• To evaluate an artificial intelligence (AI) algorithm for detecting the ECG signature of AF during normal sinus rhythm.
• To assess the AI algorithm's ability to predict incident AF in patients presenting to the emergency department (ED) with palpitations.

Main Methods:

• Retrospective study of 1403 patients aged 18+ presenting with palpitations and a 12-lead ECG.
• Exclusion of patients with prior or concurrent AF.
• Follow-up ECG or Holter monitor within one year to identify new AF diagnoses.
• Performance evaluation using Area Under the Receiver Operating Characteristics Curve (AUC), sensitivity, specificity, and predictive values.

Main Results:

• 3.1% of patients were diagnosed with new AF within one year.
• The AI-ECG algorithm achieved an AUC of 0.74 (95% CI 0.68-0.80).
• At the optimal threshold, sensitivity was 79.1% and specificity was 66.1%.

Conclusions:

• The AI-ECG algorithm demonstrated statistical significance in predicting incident AF.
• Clinical utility for screening in this ED population with low AF incidence was limited.
• Further research may explore broader applications of AI in ECG-based AF detection.