Abstract 18479: Cohort Under-Ascertainment of Prevalent Atrial Fibrillation Using the Baseline 12-Lead ECG: Implications for Atrial Fibrillation “Prediction”
Introduction: Many major cohort studies base their ascertainment of prevalent AF on the enrollment electrocardiogram (ECG). The sensitivity of this approach remains unknown. It is critical to exclude all incident AF analyses because AF itself can have important anatomic and physiologic effects.
Hypothesis: The sensitivity of AF ascertainment by baseline ECG alone is poor.
Methods: The Heart and Soul Study (H&S) enrolled 1024 ambulatory men and women with established coronary artery disease between September 2000 and December 2002 from four sites within the San Francisco Bay Area. We identified all H&S participants receiving care at the Veterans Administration (VA) (n=425), a single payer system with comprehensive electronic medical records. A baseline study ECG was performed on all subjects. A thorough review of electronic medical records was performed by an electronic key word query. All available ECGs were directly reviewed and overread by a board certified electrophysiologist. Paper charts for all non-VA hospital admissions after study enrollment were also directly reviewed.
Results: The 425 participants were followed for a median 10.5 years (interquartile rage [IQR] 5.7 [[Unable to Display Character: –]] 11.8). Prevalent AF was identified in 23 participants (5.4%) by baseline study ECG, and 77 participants (18%) by thorough chart review (p<0.001). Compared to chart review, ascertainment of prevalent AF by baseline ECG had a sensitivity of 30% and a specificity of 100% (ROC area 0.649). If only participants with prevalent AF on baseline ECG were excluded (hence misclassifying 54, or 13% of all participants, as having no history of AF), 139 (35%) of participants would have been classified as developing incident AF. In contrast, after excluding all participants with prevalent AF determined by chart review, 84 (24%) are categorized as developing incident AF.
Conclusion: Despite common practices in major cohort studies, the sensitivity of the baseline ECG for prevalent AF is poor. This may lead to common misclassification of prevalent AF participants as developing incident AF, resulting in the false identification of “predictors” of AF that are in fact due to extant AF.
Author Disclosures: J.W. Dukes: None. M.A. Whooley: None. G.M. Marcus: Research Grant; Significant; SentreHeart, Medtronic, Pfizer, NIH, AHA.
- © 2014 by American Heart Association, Inc.