Factors associated with physician modifications to automated ECG interpretations.

Area of Science:

• Cardiology
• Medical Informatics
• Artificial Intelligence in Medicine

Background:

• Automated electrocardiogram (ECG) interpretation systems are widely used but require physician review.
• Accurate diagnoses are crucial for improving clinical workflows and patient care.

Purpose of the Study:

• To investigate the frequency and characteristics of physician modifications to automated ECG reports in routine clinical practice.
• To identify patterns and limitations in current automated ECG interpretation systems.

Main Methods:

• Retrospective analysis of 159,630 ECGs from 2011 to 2023.
• Comparison of automated preliminary ECG reports (GE Marquette™ 12SL) with finalized physician reports.
• Analysis of term-level modifications (unchanged, deleted, added) for 69 ECG-related terms.

Main Results:

• 31.3% of all ECG reports were modified by physicians.
• Modifications were more frequent for off-hours ECGs, higher ventricular rates, and longer QRS durations.
• Physicians frequently added 'prolonged QT interval' and 'electronic ventricular pacemaker' diagnoses, while deleting 'inferior infarct' and 'anterior infarct' diagnoses.

Conclusions:

• Physician modification of automated ECG interpretations is frequent in real-world clinical practice.
• Current rule-based automated systems have limitations in handling complex and nuanced ECG findings.
• Physician expertise remains essential for accurate ECG interpretation and patient management.