We sought to derive contemporary population-level diagnosis rates of mild cognitive impairment (MCI) and dementia from US Medicare data and compare those rates to expected rates based on a predictive model.
Cognitive impairment is common in elderly populations but remains underdiagnosed.
We analyzed data from 2017-2019 100% samples for Medicare fee-for-service and Medicare Advantage; diagnoses were identified based on ICD-10 codes. To estimate the expected prevalence of MCI and dementia, we used the Health and Retirement Study, a nationally representative, longitudinal survey of older US adults, which includes formal cognitive assessments. We predicted MCI, dementia, and any cognitive impairment based on age, sex, race/ethnicity, dual eligibility status (ie, individuals covered by both Medicare and Medicaid), and a continuous linear trend to account for the secular decline in dementia incidence with a probit model. The model was calibrated using 2000-2014 data, validated using 2016 data, and applied to 2017-2019 Medicare data to generate expected diagnosis rates.
The prediction model performed well, with areas under the curve of 0.7128 (MCI), 0.8156 (dementia), and 0.7449 (any cognitive impairment). Differences between model-predicted rates and observed diagnosis rates were 0.2013 (MCI), 0.015 (dementia), and 0.1487 (MCI or dementia). A total of 7,291,008 MCI cases and 483,649 dementia cases were undiagnosed.
Dementia is diagnosed in the US Medicare population at approximately the expected rate; however, MCI remains substantially underdiagnosed. If failure to diagnose is not addressed, it will have negative implications for timely access to a disease-modifying treatment for Alzheimer’s disease.