Attribution of Health Care Costs to Diseases: Does the Method Matter?
Allison B. Rosen, Ana Aizcorbe, Tina Highfill, Michael E. Chernew, Eli Liebman, Kaushik Ghosh, David M. Cutler
This chapter is a preliminary draft unless otherwise noted. It may not have been subjected to the formal review process of the NBER. This page will be updated as the chapter is revised.
Chapter in forthcoming NBER book Measuring and Modeling Health Care Costs, Ana Aizcorbe, Colin Baker, Ernst Berndt, and David Cutler, editors
Cost of illness (COI) studies focus on allocating health expenditures to a comprehensive set of diseases. A variety of techniques have been used to allocate spending to diseases. In this paper, we compare spending attributed to diseases using three approaches: one based on the principal diagnosis listed on each encounter’s claim, a second based on all diagnoses listed on the encounter, and a third based on decomposing a person’s total annual spending to their conditions. The study sample is large: 2.3 million commercially insured individuals under age 65. Results indicate significant differences in the allocations from the different approaches. The two claims-based encounter approaches allocate 78% of overall spending to diseases, while the person approach allocates 95% of spending to diseases. The large unallocated spending in the claims-based approach is due largely to lack of diagnosis codes for prescription medications. Spending was concentrated in a small number of conditions; the 10 most expensive diseases account for 40% of total spending with the person approach and about 18% of spending with the primary-diagnosis and all-diagnoses encounter approaches. Future research needs to pay careful attention to the choice of method in allocating spending to diseases, especially when research uses prescription medication claims data.
This paper was revised on December 15, 2016
Users who downloaded this chapter also downloaded* these: