OBJECTIVES: Accurately measuring the cost of illness is critical to understanding the financial burden of Medicare beneficiaries with concurrent diabetes and cancer. This study aimed to assess the congruency of the estimated cost of diabetes among Medicare beneficiaries with cancer from survey versus from administrative claims of the Medicare Current Beneficiary Survey (MCBS) data.

METHODS: This serial cross-sectional study utilized the MCBS survey and claims components from 2006-2012. Eligibility criteria required beneficiaries to be ≥65 years old, non-institutionalized, diagnosed with cancer, and enrolled in Medicare Parts A, B, and D. To identify diabetes and cancer diagnoses, we collectively used survey self-reports and claims for ICD-9 diagnosis codes following algorithms from the CMS Chronic Conditions Data Warehouse. The cost of diabetes was operationalized as the difference in total annual direct costs among beneficiaries with cancer and diabetes compared to those with cancer only. Costs from each year were converted to 2012 U.S. dollars. The congruency of total annual direct cost estimates from survey versus claims data was compared using the Wilcoxon signed-rank test.

RESULTS: Among 4,918 Medicare beneficiaries with cancer from 2006-2012, 1,275 (26%) were diagnosed with diabetes. Significant disagreements in the estimated total annual cost of diabetes from survey versus claims data were present in 2006-2009 and 2012 for beneficiaries with cancer and diabetes (p<0.001 for all) and all years for beneficiaries with cancer only (p<0.001 for all). The mean total annual cost of diabetes from survey data was $7,575 per capita in 2012 dollars. Majority (74%) of the total annual cost of diabetes was spent on medical costs, and the remaining 26% was spent on Medicare Part D prescriptions.

CONCLUSIONS: The decision to use the MCBS survey or claims components for economic evaluations will likely have implications on research findings, and caution is recommended when using either component alone.