THE ACCURACY OF DEMENTIA DIAGNOSIS IN MEDICARE CLAIMS

Abstract

Listen

Translate article

Medicare claims are commonly used to identify patients with dementia. But diagnoses recorded in claims may not capture mild dementia cases. Such mis-identification may result in biased estimates of disease prevalence and costs. This study estimates the sensitivity and specificity of Medicare claims to identify dementia in a cohort of older adults with clinically diagnosed dementia. The sample was drawn from participants in the Washington Heights-Inwood Columbia Aging Project (WHICAP), a multiethnic, population-based, prospective study of cognitive aging in which dementia status was assessed using a rigorous clinical protocol. The current study included 455 subjects who were diagnosed with dementia at enrollment and 1,689 subjects who were never diagnosed with dementia any time during the study. ICD-9-CM diagnosis codes in all available Medicare claims were used to determine claims-based identification of dementia. Sensitivity and specificity of claims-identified dementia compared to clinically diagnosed dementia were computed, using clinical diagnosis as the gold standard. Logistic regression was used to estimate the relationship between patient clinical and demographic characteristics and predictive value in claims identification. Medicare claims correctly identified 1,808 cases (agreement rate with clinical diagnosis=84.7%), resulting in a kappa value of 0.43 (moderate agreement). The sensitivity and specificity of dementia identification in Medicare claims was 0.38 and 0.97, respectively. Among subjects clinically diagnosed with dementia, 281 (61.8%) were not identified in Medicare claims as demented. Individuals with a clinical diagnosis of dementia but not identified as having dementia in the claims data were younger, with lower education, eligible for Medicaid, had fewer comorbidities, and better function and cognition than those for whom there was agreement between claims data and clinical diagnosis. Among those who were clinically diagnosed as non-demented, 55 (3.2%) were identified in the claims as demented. Disagreement in this group was associated with older age, unmarried status, more comorbidities, worse function and cognition. Mis-identification of dementia in Medicare claims is quite common, although there are both false positive and negative assessments of disease. Both clinical and sociodemographic characteristics affect predictive value of claims-based diagnoses. Mis-identification may be associated with biased estimation of disease prevalence and cost of AD.