Computer-based rhythm diagnosis and its possible influence on nonexpert electrocardiogram readers

Area of Science:

• Cardiology
• Medical Informatics
• Artificial Intelligence in Healthcare

Background:

• Automated electrocardiogram (ECG) systems aim to enhance healthcare quality through accurate, timely cardiac rhythm diagnoses.
• The diagnostic accuracy of current automated ECG systems remains a significant concern for clinicians.

Purpose of the Study:

• To evaluate if two independent automated ECG systems outperform nonexpert clinicians in rhythm diagnosis accuracy.
• To determine the rate at which automated systems provide correct diagnoses when nonexpert clinicians make errors.

Main Methods:

• Manual analysis of 500 ECGs by senior experts and nonexpert clinicians.
• Automated analysis of the same 500 ECGs using two independent systems.
• Comparative assessment of diagnostic accuracy between clinicians and automated systems.

Main Results:

• Expert reading identified 420 sinus rhythms and 156 rhythm disturbances.
• No statistically significant difference in accuracy was found between nonexpert clinicians and automated systems (p = .45 for system A, p = .11 for system B).
• Automated systems correctly diagnosed 28% ± 10% (System A) and 25% ± 11% (System B) of cases where clinicians were incorrect (p = .09).

Conclusions:

• Automated ECG systems did not demonstrate superior average accuracy in rhythm diagnosis compared to nonexpert clinicians.
• Computer-assisted ECG diagnoses may be erroneous in situations where human clinicians also fail to achieve accurate diagnoses.