Multimorbidity Index as a Tool for Projection of Health and Mortality among U.S. Older Adults: Medicare-Based Analysis

Igor Akushevich, Duke University
Julia Kravchenko, Duke University
Heather Whitson, Duke University
Harvey J. Cohen, Duke University

Medicare-based data allow for developing a tool for measuring individual multimorbidity in US elderly population. Such a tool can improve prediction of mortality and health in the elderly population compared to the standard Charlson Comorbidity Index (CCI). A new Adjusted (for US elderly population) MultiMorbidity Index (AMMI) was developed and applied to Medicare-based datasets such as the National Long Term care Survey (NLTCS-Medicare) and the Surveillance, Epidemiology and End Results Registry data (SEER-Medicare) by estimating the presence (or absence) of a disease in an individual, calculating disease-specific weights, and elaborating the list of contributing diseases. Properties of the index were studied empirically, then validated using population models of health and mortality, and, finally, tested in a sensitivity analysis. The AMMI demonstrated better predictive power (AUC=0.90––i.e., approximately 90% of cases of next month survival status were predicted correctly) than CCI (AUC=0.84).