GEOGRAPHIC DISPARITIES IN INCIDENCE AND MORTALITY OF ALZHEIMER’S DISEASE

Abstract

Abstract Our estimates showed that geographic patterns of incidence rates extracted from Medicare data and mortality rates calculated using Multiple-Cause-of-death data are significantly different. The significant gap in mortality rates between the East and West parts of the U.S. present in death certificate data is not observed when Medicare data is used. In addition, high incidence rates of Alzheimer’s Disease (AD) identified in the stroke-belt states when using Medicare data is not fully reproduced when death certificates are used. Therefore, we used 5%-Medicare data linked to the National Death Index to resolve this inconsistency in findings using both types of data for the same individuals. We found that the excess rates of AD in the stroke-belt states cannot be attributed to higher proportions of vulnerable populations as the associated differences hold across race- and gender-specific subgroups: 1,717 (1,704-1,731) vs. 1,280 (1,275-1,286) for Whites; 2,052 (2,017-2,086) vs. 1,821 (1,795-1,848) for Blacks; 1,538 (1,520-1,556) vs. 1,156 (1,140-1,164) for males; and 1,878 (1,861-1,894) vs. 1,383 (1,376-1,391) for females. We evaluated the sensitivity and specificity for having AD as an underlying cause of death in all these regions and developed a predictive model that predicts occurrence of AD in death certificate based on information in individual Medicare records. Our findings show that underrepresentation of AD in death certificates is a strong contributor to geographic disparities in AD mortality. Predictive quality of occurrence of AD in death certificates can be improved by advancing the predictive model and using a broader set of predictors from Medicare records.