Dementia in community-dwelling elderly patients: A comparison of survey data, medicare claims, cognitive screening, reported symptoms, and activity limitations.

Background And Objective: Population-based estimates of dementia can vary widely depending on the data source and methodology used to assess cognitive impairment. The aim of this study was to assess individual and composite measures of dementia constructed using survey and administratively collected medical data for the purpose of improving insight into inherent and avoidable biases and the interpretation of these measures.

Methods: A national survey of 5089 community-dwelling elderly persons was linked to 5 years of Medicare claims. Kappa coefficients were used to assess dementia agreement between survey, cognitive screen, and claims. Odds ratios (ORs), adjusted for age, gender, and education, were computed to assess the relation of single- and multi-source dementia variables to cognitive symptoms and activity limitations.

Results: Dementia prevalence among persons aged 65 and over ranged from 4.5% to 16.8% across sources. Agreement in case identification between data sources was low, ranging from 0.15 to 0.41. Despite low agreement, ORs were consistently increased for limitations in cognitively related activities/symptoms and non-overlapping dementia sources.

Conclusion: The finding that non-overlapping dementia data sources are associated with significantly increased limitations in cognitive activities suggests that forecasts of national dementia-related resource use based on single-source data could produce serious underestimates of future need.