Abstract

Research objective: Understanding the magnitude of the direct medical care costs attributable to dementia is important for public and private decision makers, but estimating these costs has been difficult. Even when focused on the direct costs to the traditional Medicare program, estimates have ranged widely, with the differences driven by the definition used to identify dementia, the time period and population studied, and the study design and analytic techniques employed. This study examines the effect of different definitions of dementia and onset on estimates of the incremental cost of dementia to the traditional Medicare program.

Methods: Using survey responses from the Health and Retirement Study (HRS) linked to Medicare claims from 1991-2012, we identified two cohorts of older adults with dementia. The first cohort was identified using claims-based dementia diagnosis codes, with onset defined as the date of the first qualifying diagnosis code. The second cohort used cognitive assessments administered via survey, with onset defined as the midpoint between the two survey dates surrounding the first qualifying dementia score. Comparison groups were randomly selected for each cohort after matching on sex, birth year, and HRS-survey entry year. Controls were given the diagnosis date of their assigned case to allow for a comparison of equivalent time periods. We calculated monthly Medicare expenditures for the 12 months prior and up to 60 months following a diagnosis of dementia. To evaluate the marginal effect of dementia on Medicare expenditures, we used the estimator described by Basu and Manning (2010) for estimating costs under censoring.

Results: 4,010 dementia cases were included in our claims-based cohort and 2,359 in our survey-based cohort, with 1,297 participants included in both. For these cases, the survey-based approach estimated an onset date that was 12 months earlier, on average, than in the claims-based approach.

Using the claims-based approach, the five-year incremental cost of dementia is estimated at $14,825. Notably, 46% of the incremental cost is incurred in the first year after diagnosis since the code often first appears during an inpatient admission. Incremental costs decrease as time from diagnosis increases, such that the cost differential in year five approaches zero. In contrast, the five-year incremental cost of dementia is estimated at $8,570 using the survey-based approach, with the incremental costs distributed more uniformly over time.

Conclusions: Two commonly employed strategies for identifying dementia produce significantly different estimates of the incremental cost of dementia, and the distribution of this cost over time, to the traditional Medicare program. Both approaches indicate that increased costs among dementia cases are driven by part A covered services, including inpatient, skilled nursing and hospice care. Both approaches also suggest that estimated expenditures in the month of death are lower in the presence of a dementia diagnosis.