1. A Price Deflator for Medical Care Ana Aizcorbe (BEA) Nicole Nestoriak (BLS) 2008 World Congress on National Accounts and Economic Performance Measures for Nations May 17, 2008

2. www.bea.gov 2 The views expressed in this paper are solely those of the authors and not necessarily those of the U.S. Bureau of Economic Analysis or the U.S. Bureau of Labor Statistics.

3. www.bea.gov 3 Our research on price indexes for health spending is part of a broader initiative. BEA plans to construct a satellite account for medical care that will include:  Expenditures that are reconciled with CMS estimates.  New measures of spending by disease that will facilitate assessments of the returns on health expenditures.  Improved price deflators that will better-account for changes in quantity vs changes in prices.

4. www.bea.gov 4 Price deflators are the subject of today’s talk.  Our research uses health claims data as a laboratory to explore:  Different ways to construct price deflators,  How best to allocate spending by disease,  The usefulness of claims data (relative to other data sources),  In particular, do we really need to use claims data?  This project is very much at the research stage.  Today’s results are for HMO patients only.

5. www.bea.gov 5 Starting point for price deflator: prescription from health economists.  “Output” of medical care as the “treatment of disease,” not the individual treatments  Previous case studies suggest that this issue could be numerically important:  Psychiatric conditions (Berndt et. al.)  Heart attacks (Cutler and McClellan)  Cataract (Shapiro, Shapiro, and Wilcox)  40 conditions in four cities (Bradley et. al.)  Two National Academies panels have recommended that statistical agencies construct price indexes on this basis, even if one cannot account for changes in outcomes of care (i.e., quality)

6. www.bea.gov 6 Why is it important to define the “good” as the treatment of a homogeneous medical condition? Example: drug therapy introduced, no price change for either treatment, number of patients the same Problem:  As consumers switch to drug therapy, nominal expenditures fall  Usual price index shows no price change  Real expenditures fall even if quantities did not Solution: redefine the “good” as the “treatment of depression” and the price as P depression,t = expenditures on all types of treatments -------------------------------------------------- number of patients treated This is really just a reincarnation of the “Reinsdorf” problem.

7. www.bea.gov 7 Our data show large differences in cost per patient across treatment types… Shifts across treatments types will likely generate changes in costs.

8. www.bea.gov 8 …and our indexes show substantial cost savings generated by changes in treatment types. Source: Aizcorbe and Nestoriak (2008) Features of our index: Constructed using large claims database for HMO patients. Price = revenue from all sources Price is defined as price per patient treated for a homogeneous condition Dollars are allocated to conditions using “groupers” “Treatments” are identified using “place of service” variable.

9. www.bea.gov 9 The differences are large and pervasive. Source: Aizcorbe and Nestoriak (2008) For example, the cost of treating infectious diseases rose, on average, 36 percent from 2003:1 to 2005:4, while the costs of the underlying treatments rose 37 percent. Conclusion: viewing the bundle of treatments as the “good” implies slower increases in price (and faster increases in quantity). Health economists view these differences as productivity. Caveat: these indexes do not account for changes in “quality” of treatment. Price Indexes: P(2005:4)/P(2003:1)

10. www.bea.gov 10 Can we say more about the differences in the indexes? Is it substitution? We developed an algebraic mapping to further explore sources of these differences. The unit value index for treating disease d can be interpreted as a Laspeyres index of treatments that is adjusted for shifts in utilization of treatments. c d,i 0 x d,i 0 x d,i 1 / P d 1 I d (P) = c d 1 / c d 0 =  i ---------------- --------------- ( c d,i 1 / c d,i 0 )  i c d,i 0 x d,i 0 x d,i 0 / P d 0 The difference in the two indexes is then: c d,i 0 x d,i 0 x d,i 1 / P d 1 I d (P)- I d (T)=  i ---------------- [ ----------- - 1 ] ( c d,i 1 / c d,i 0 )  i c d,i 0 x d,i 0 x d,i 0 / P d 0

11. www.bea.gov 11 We found that reduced use of hospitals and office visits generated cost savings not offset by increased use of drugs and other, lower- cost treatments. This is consistent with long-run trends in the CMS data. Indexes Contributions to differences (percentage points)

12. www.bea.gov 12 Some Examples Ambulatory Surgical Centers: Small, growing fast, particularly in the treatment of gastrointestinal and eye conditions. Home Care: There is anecdotal evidence of shifting medical equipment from hospitals to the home in the treatment of lung conditions.

13. www.bea.gov 13 Other lessons from this research:  In theory, reweighting treatment indexes will not fix the problem.  Empirically, reweighting does narrow some of the gap.  In theory, differences in the indexes can go either way.  Empirically, the net effect is dominated by substitution towards lower cost treatments.

14. www.bea.gov 14 We will continue to explore:  An economic interpretation for this deflator.  Volume measures (as an alternative to deflation).  Alternative types of data sources for prices.  Alternative ways to allocate spending by disease.