LONGITUDINAL AGREEMENT OF SELF-REPORTED CHRONIC DISEASE WITH MEDICARE CLAIMS HISTORY. (11th November 2018)

• Record Type: Journal Article
• Authors: Nagel, C; Cigolle, C; Quinones, A
• Is Part Of: Innovation in aging. Volume 2(2018)Supplement 1
• Page Start: 710
• Page End: 710
• Publication Date: 2018-11-11
• Subjects: Aging -- Periodicals; Gerontology -- Periodicals; 612.67
• DOI: 10.1093/geroni/igy023.2632
• Languages: English
• ISSNs: 2399-5300

Abstract

Abstract: Chronic disease data from longitudinal health interview surveys are used often in epidemiologic studies. Yet, these data may be limited by inaccurate self-reporting of previously diagnosed disease. We examined the agreement between respondent self-report of diabetes, cancer, stroke, heart disease, and lung disease in the Health and Retirement Study (HRS) with diagnostic codes extracted from linked Medicare claims files. Our sample consisted of respondents in the HRS waves 1998–2010 with available Medicare claims (N=21, 426). Disease status was derived from Medicare claims using Centers for Medicare and Medicaid Services (CMS) algorithms. We calculated bias and prevalence adjusted kappas to quantify the agreement between CMS-defined disease status and subsequent self-report. We then examined only respondents with CMS-defined disease and calculated the prevalence of negative self-report. Lastly, we modeled respondent factors associated with the probability of not reporting a previously diagnosed disease. The average kappa value across all diseases and waves was .71, indicating good overall agreement. However, among respondents with a history of Medicare claims indicating the presence of chronic disease, 37.4% (diabetes) to 71.3% (lung disease) reported not having that disease at a subsequent survey wave. Characteristics associated with the probability of not reporting previously diagnosed disease included Latino ethnicity, age, cognitive impairment, household wealth, and number Medicare claims. A substantial proportion of individuals with Medicare claims indicative of chronic disease reported having no history of the disease. These findings suggest gaps in patients' understanding of their chronic disease status and present challenges for researchers designing and using health interview surveys.