Accuracy of predicting risk of cognitive impairment using Medicare administrative data to operationalize the Lancet Commission's risk factors

Medicare administrative data can be utilized to predict dementia and MCI risks with good accuracy. Importantly, race/ethnicity remains an independent predictor when controlling for presence of chronic conditions. As diagnosis codes have limited information on disease severity and management, this finding could point to either worse disease control or social or genetic factors. Adding person-level information on social determinants of health to Medicare data would enhance their utility for detection of cognitive impairment.