Use of a Clinical Electrocardiographic Database to Enhance Atrial Fibrillation/Atrial Flutter Identification...

Area of Science:

• Cardiology
• Medical Informatics
• Public Health

Background:

• Administrative data possess limited sensitivity for identifying atrial fibrillation/atrial flutter (AF/AFL) cases.
• Integrating clinical electrocardiogram (ECG) data may improve the diagnostic yield for AF/AFL detection.

Purpose of the Study:

• To evaluate the diagnostic yield of AF/AFL case finding by linking administrative data with ECG interpretations from the Marquette Universal System for Electrocardiography (MUSE) repository.
• To assess the incremental diagnostic value of incorporating ECG data into existing administrative data-based algorithms for AF/AFL identification.

Main Methods:

• Retrieved 369 ECGs from the MUSE repository for validation, comparing MUSE coding definitions against a blinded, duplicate review reference standard.
• Calculated agreement (Cohen κ), sensitivity, and specificity for various MUSE coding definitions.
• Assessed the incremental diagnostic yield by linking clinical registries, administrative data, and the MUSE repository (n=11,662) for preexisting and incident AF/AFL.

Main Results:

• The agreement between MUSE diagnosis and reference comparison varied (Cohen κ: 0.57–0.75), with sensitivity ranging from 60.6% to 79.1% and specificity from 93.2% to 98.0%.
• A coding definition prioritizing AF/AFL in the first three ECG statements achieved the highest sensitivity (79.1%) with minimal specificity loss (94.5%).
• Incorporating ECG data increased the diagnostic yield of preexisting AF/AFL by 14.5% and incident AF/AFL by 7.5%–16.1% compared to administrative data alone.

Conclusions:

• Routine ECG interpretation using MUSE coding demonstrates high specificity and moderate sensitivity for AF/AFL detection.
• Inclusion of MUSE ECG data in case identification algorithms significantly enhances the detection of AF/AFL cases missed by administrative data alone.