The BEA is conducting research to support the eventual implementation of a Satellite Account for Health that will contain two key pieces of information necessary to assess the benefits of spending on health care services: 1) expenditures broken down by disease, and 2) price measures that accurately reflect changes in the quality of treatments. The research will involve an empirical assessment of competing approaches that may be used to obtain these estimates as well as theoretical work to consider purely measurement or national-accounting issues that arise in implementing a health account as part of the National Income and Product Accounts. We have four specific goals in mind in this project: Harmonize national data on nominal expenditures for health care Develop Satellite National Health Accounts for Costs of Illness Improve measures on real expenditures for treatment of diseases Crosswalk between disease- and product-based deflatorsOur review of the existing literature suggests that tackling the following issues would be a useful first step to support the eventual adoption of new measures in a satellite account for health: Empirically Assess Methods to Deal with Comorbidities: There are three broad approaches to dealing with comorbidities. The first, na ve view, assigns a primary diagnosis to each medical encounter and assigns any spending in that encounter to the primary diagnosis. This is consistent with the manner in which revenues for business establishments that produce more than one good are allocated: revenues for all goods produced in the establishment are allocated to whatever good the establishment declared as its primary good. The second approach involves the use of episode groupers, algorithms that sort through medical encounters and use patients' histories to allocate the spending from each encounter to a unique medical condition. This represents an improvement over the primary diagnosis approach in that it uses all available information on the patient. The final approach developed by David Cutler and Allison Rosen (Cutler and Rosen, 2005) allows for the possibility that existing comorbidities affect both the intensity and cost for treatment of other conditions. They use a regression approach that estimates the cost of treating a particular illness conditional on the patients' other illnesses. The key difference in this and the other approaches is that this does not require that dollars for each encounter map one-to-one with medical conditions.
