Automatic Holter electrocardiogram analysis in ischaemic stroke patients to detect paroxysmal atrial fibrillation:...

Area of Science:

• Cardiology
• Neurology
• Medical Informatics

Background:

• Paroxysmal atrial fibrillation (pAF) detection in ischemic stroke patients is crucial for secondary stroke prevention, often necessitating oral anticoagulation.
• Manual analysis of prolonged electrocardiogram (ECG) monitoring data for pAF detection is time-consuming and resource-intensive.

Purpose of the Study:

• To evaluate the effectiveness of an automated algorithm (AA) for pAF detection compared to manual analysis with software support (SA).
• To assess the diagnostic performance of the AA in identifying pAF in patients with acute ischemic stroke/transient ischemic attack.

Main Methods:

• Utilized data from the prospective IDEAS cohort, including 72-h Holter ECG recordings of patients presenting in sinus rhythm.
• Compared a commercially available automated algorithm (AA) against central adjudication (SA) for pAF diagnosis.
• Resolved discordant results through cardiological reference confirmation.

Main Results:

• Paroxysmal AF was diagnosed in 5.9% of the cohort (n=62).
• The AA diagnosed pAF more frequently (5.8%) than SA (4.5%).
• AA demonstrated high sensitivity (96.8%) and negative predictive value (99.8%) for identifying patients without pAF, while SA had lower sensitivity (75.8%) and NPV (98.5%). AA's specificity was 96% with a PPV of 60.6%, whereas SA achieved 100% specificity and PPV. Moderate agreement (kappa=0.591) was observed between AA and SA.

Conclusions:

• Automated determination of the absence of pAF can significantly reduce the manual review workload of prolonged Holter ECG recordings.
• Automated algorithms offer a viable and efficient alternative for pAF screening in stroke patients, aiding in timely initiation of anticoagulation therapy.