Miller, Nilsen Langdon. Expanded methods to model last year of life expenditures in the medicare current beneficiary survey with sensitivity analyses and new findings. Retrieved from https://doi.org/doi:10.7282/t3-th4x-z541
DescriptionTerminal year of life medical expenditures in the elderly are generally much higher than medical expenditures in other years of life. These expenditures have often been portrayed as wasteful and inefficient use of heroic procedures with little change of success at the end of life. Prior studies in this area have focused predominately on Medicare expenditures; only a handful considered overall medical expenditures. The Medicare Current Beneficiary Survey (MCBS) comprises data on all medical expenditures, not just those paid for by Medicare, but these data only include expenditures aggregated by calendar year. Thus, linear models have been applied to estimate terminal year of life medical expenditures.
This paper adds to the body of information on this topic using recent data, from the 2006-2013 MCBS. We modified previous and developed new methods to estimate end-of-life expenditures both overall and broken down by payer and service category. We also developed new sensitivity analyses to test data and model assumptions. Additionally, we compared variances of the estimates from two predominate methods for robust variance estimation in complex surveys like the MCBS, the Taylor Series Expansion (TSE) method and Fay’s modification to the Balanced Repeated Replication (BRR) method. Finally, we utilize our new method of analysis to examine univariate and multivariate covariate effects of calendar year, age, sex, race, and geographical region on selected categories of terminal year of life expenditures. This is to our knowledge, the first multivariate analysis of multivariate correlates for total end-of-life medical expenditures.