Abstract MP75: Geographic Variation in Stroke Rates and Oral Anticoagulation Use among Medicare Beneficiaries with Atrial Fibrillation

Abstract

Background: Recent studies have shown strong geographic variation in oral anticoagulation (OAC) use in atrial fibrillation (AF); however, it remains unknown how this contributes to the geographic variation in ischemic stroke observed across the US. The objective of the present study was to evaluate the relationship between the geographic variation in the initiation of OAC and the incidence of ischemic stroke in a cohort of Medicare beneficiaries newly diagnosed with AF. Methods: Using 2013-2014 claims data from a 5% random sample of Medicare beneficiaries, we identified patients newly diagnosed with AF in 2013-2014 and categorized them according to their initiation of OAC. Our sample included 21,226 OAC initiators and 20,068 patients who did not initiate OAC therapy. We assigned each patient to one of the 9 US Census Divisions using the zip code, and collected their medical claims with a diagnosis of ischemic stroke. We constructed logistic regression models to estimate the average adjusted probability of OAC initiation and Poisson models to estimate the average adjusted rate of ischemic stroke, in each Census Division. Both estimates were adjusted for demographics, eligibility for Medicaid coverage and for low-income subsidy, enrollment in a Medicare Advantage Part D plan, and a comprehensive list of clinical characteristics. We computed the correlation between the average adjusted probability of OAC initiation and the average adjusted rate of ischemic stroke at the Census Division level. Results: The probability of OAC initiation was lowest in the West South Central (0.47) and highest in the West North Central (0.54) and New England (0.54). The average adjusted rate of ischemic stroke was lowest in the West North Central (0.09) and highest in the South Atlantic (0.14) and South West Central (0.14). The average adjusted probability of OAC initiation at the Census Division level and the average adjusted rate of ischemic stroke were inversely correlated, with R=-0.576, p-value=0.10. This suggests that variation in OAC initiation likely explains at least a third of geographic variation in ischemic stroke [R 2 =(-0.576) 2 =0.332]. Conclusions: Our results suggest that geographic variations in OAC initiation within the U.S. explain, in part, variations in the incidence of ischemic stroke among AF patients. Further mechanistic research using advanced causal mediation models is warranted.