Abstract P109: Consistency of Self-reported Prior Diagnosis of Atrial Fibrillation and Medicare Claims Diagnoses: The Reasons for Geographic and Racial Differences in Stroke

Introduction: Atrial fibrillation (AF) is the most common sustained cardiac arrhythmia. Although self-reported prior diagnosis is a widely used method for ascertainment of AF in population studies, recall bias may impact the accuracy of this method. Objective: To examine the consistency of self-reported AF as compared to diagnoses recorded in Medicare claims. Methods: Among 30,239 black and white adults aged 45 years and older who enrolled in the REasons for Geographic And Racial Differences in Stroke (REGARDS) study, 8,588 participants aged 65 years and older with a minimum of 6 months of Medicare fee-for-service coverage at the time of enrollment were included in this analysis. Using AF diagnosis codes from Medicare claims as the “gold standard”, we calculated sensitivity, specificity, positive predictive value (PPV), and negative predicted value (NPV) of self-reported prior diagnosis of AF by a physician or other health professional. Results: The average age of the study population was 73.4 years, 50.5% were women, 31.6% were black, and 10.1% reported a prior diagnosis of AF. The prevalence of Medicare claims diagnoses for AF ranged from 28.6% to 32.5% depending on definition ( Table ). Compared to Medicare claims, self-reported prior diagnosis of AF had 22.4-24.5% sensitivity, 95.6-95.7% specificity, 68.8-71.5% PPV, and 71.9-76.0% NPV. Results were consistent across strata defined by age, gender, race, and cognitive status. Conclusion: Compared to Medicare claims, self-reported prior diagnosis by a physician or other health professional had low sensitivity but high specificity to detect AF. This suggests that an approach utilizing only self-report to detect AF would not be ideal. However, most participants who reported a prior diagnosis of AF had Medicare claims supporting this diagnosis.