Clinical Electrophysiology -> SVT/AVNRT/WPW/AT: -> Epidemiology of Cardiac Arrhythmias/ Epidemiology D-PO01 - Featured Poster Session (ID 11) Poster

D-PO01-023 - Validity Of Self-reported Versus CMS-identified Atrial Fibrillation In The Vital Study (ID 47)


Background: In cohort studies, atrial fibrillation (AF) is frequently ascertained by either using claims data from the Centers for Medicare and Medicaid Services (CMS) or through participant self-report.
Objective: To compare the yield and accuracy of AF diagnoses ascertained by self-report versus CMS linkage.
Methods: The Vital Rhythm Study, a randomized trial testing vitamin D and omega-3 fatty acid supplementation in the primary prevention of AF among 25,131 individuals without preexisting cardiovascular disease or cancer, has utilized a combination of participant self-report on annual questionnaires and linkage to CMS claims to ascertain AF diagnoses. Medical records are then reviewed by an endpoint committee of cardiologists to confirm or disconfirm incident AF events. The yield and positive predictive value (PPV) of these two methods of AF ascertainment were compared among 15,436 participants who were age 65 or older.
Results: Over 4.35 years of follow-up, a total of 675 participants self-reported AF and 817 had an AF diagnosis in CMS claims data. Medical records were obtained on 654 AF diagnoses that were self-reported (522, 77%) or identified by CMS linkage (467, 57%) and reviewed by the endpoint committee. Of the AF diagnoses where records were obtained, 548 cases of incident AF were confirmed (84%) by medical record review. Of these confirmed events, 490 (89%) were self-reported by participants and 393 (71%) had an AF diagnosis in CMS claims data. The PPV (confirmed/ascertained events) for confirmed AF by medical record review was 94% (490/522) for self-report, 84% (393/467) for CMS linkage, and 97% (344/355) for AF events identified by both sources.
Conclusion: The yield and accuracy of AF diagnoses identified by self-report was higher than those identified through CMS linkage. A significant proportion of confirmed incident AF cases were not identified by CMS linkage (29%) and ascertainment was optimized when both methods were utilized. When AF events were both self-reported and identified via CMS linkage, the PPV was very high.