Strategies for a BEA Satellite Health Care Account: Summary of a Workshop (2009)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

1 Introduction 1.1. Project Description and Report Structure In March 2008 the Committee on National Statistics of the National Acad- emies held a workshop to assist the Bureau of Economic Analysis (BEA) with next steps as it develops plans to produce a satellite health care account designed to improve its measurement of economic activity in the medical care sector and to be useful for health care policy. The ultimate objectives of the BEA program are to compile medical care spending information by type of disease that is more directly useful for measuring health care inputs, outputs, and productivity than are current estimates of spending by type of provider; produce a comprehensive set of accounts for health care–sector income, expenditure, and product; develop medical care price and real output measures that better break out changes in the delivery of health care from changes in the prices of that care; and coordinate BEA and Centers for Medicare and Medicaid Services (CMS) health expenditure statistics. BEA, at this point, still very much in the research stage of this project, has committed to produce a detailed proposal for a satellite or experimental account by the end of 2009. The agency is coordinating its work with other efforts, notably a project led by David Cutler (Harvard University) and Allison Rosen (University of Michigan), and through collaboration with colleagues at other ­ statistical agencies that has taken place during the past 1-2 years. The hope at BEA is that work associated with this project will eventually lead to methodological advances in the national income and product accounts (NIPAs) by improving the conceptual basis for accurately measuring prices and quantities of the economy’s medical care goods and services.  

 Strategies for a BEA Satellite Health CARE Account The steering committee for the workshop consisted of Barbara Fraumeni (University of Southern Maine, and formerly BEA), Joseph Newhouse (Harvard University), and Gail Wilensky (Project HOPE). The BEA-sponsored workshop included sessions covering the following topics: • Plans for a Satellite Health Care Account. Ana Aizcorbe (chief econo- mist, BEA) presented the goals of BEA’s health accounting program, progress to date on the project, and a summary of BEA’s strategies for dealing with key measurement issues and data needs. Dale Jorgenson (Har- vard University) and Matthew Shapiro (University of Michigan) served as discussants. • Constructing Nominal Expenditures by Disease. Charles Roehrig (­Altarum Institute) provided an overview of work by him and his col- leagues developing time-series estimates of national expenditures by medical condition. This type of data will be essential as health accounting projects move forward because the treatment for a specific condition or disease provides an organizing principle for defining units of service in the medical care sector. David Cutler and Allison Rosen were discussants. • Price Indexes and Volume Measures. Ralph Bradley and Bonnie ­Murphy (Bureau of Labor Statistics, Consumer Price Index and Producer Price Index programs, respectively) discussed their agency’s plans to research and generate price indexes organized by broad disease category. Price indexes, which are used to decompose changes in current dollar estimates into price and quantity components, are essential for calculating real gross domestic product (GDP) for the various sectors of the economy. The Producer Price Index (PPI) program has developed a method to quality adjust its current hospital indexes by using quality indicators contained in the CMS Hospital Compare database. The Consumer Price Index (CPI) program is generating experimental price indexes, also orga- nized by major disease category, by merging medical expenditure and utilization data from the Medical Expenditure Panel Survey with the CPI production database of the Bureau of Labor Statistics (BLS). Jack Triplett ( ­ Brookings Institution) and Patricia Danzon (University of Pennsylvania) were discussants. • Measuring Treatment Outcomes. Mark McClellan (Brookings Insti- tution) discussed the challenges of constructing measures of treatment outcomes, which are essential for monitoring the quality of medical care and, in turn, changes in the real output of the sector. He also provided an assessment of the current state of knowledge evident in the outcomes research literature and of how that information is influencing relevant policy discussions. 

INTRODUCTION  • National Accounting Issues. BEA discussed national accounting issues that must be resolved in order to produce a satellite account for health. These include how to construct measures of real expenditures for health care industries, define disease and product classes, and develop a set of deflators for medical care industries; the importance of sources of pay- ment data was also discussed. The central presentation of the session, by Brian Moyer of BEA, focused on development of an approach for coor- dinating the spending and industry sides of the national accounts. Barbara Fraumeni and Sherry Glied (Columbia University) were the discussants. • Summary comments and directions were provided by Gail Wilensky. Throughout the meeting, participants demonstrated enthusiastic support for the BEA effort and offered encouragement to the project leaders. Dale Jorgenson referred to BEA’s work on the project as landmark, noting that it could have a significant impact—and not just on measured GDP, but also on the way people think about health and the health care sector. He stated that it is promising that BEA had concentrated talent and resources on this important research topic. Matthew Shapiro agreed, observing that the statistical system appears poised to make major progress on health accounting issues, and added that the workshop occurred at a pivotal time. Both the BLS and BEA are currently engaged in inno- vative research that will enhance their ability to provide comprehensive data for measuring activity in the medical care sector of the economy. Shapiro expressed hope that participants in the workshop, and others, would be able to leverage two elements: The first is to build on the conceptual work by pioneering researchers— such as Scitovsky (1967); Cutler, McClellan, Newhouse, and ­Remler (1998); and Berndt, Busch, and Frank (1998)—that sought to measure prices and quantities of medical care using a disease treatment framework; the second is harnessing the potential of the tremendous amount of data available to implement this kind of measurement on a production-level basis. Academic researchers began the work, disease case by disease case, but the systematic production of data by a statistical agency would mark a huge step forward. Finally, CNSTAT director Constance Citro offered her encouragement and expressed hope that the workshop would help BEA push forward in the develop- ment of a satellite health care account. She also noted the complementary work by a CNSTAT panel seeking to advance foundations for a data system that would relate both medical care and other inputs to incremental changes in the health of the population. That project is funded by the National Institute on Aging and, like this workshop, is also chaired by Joseph Newhouse. The two projects are related in that development of data on expenditures, prices, and quantities for medical care—BEA’s program—is a key component of the broader health account concept as well. 

 Strategies for a BEA Satellite Health CARE Account 1.2. What Kind of “Satellite” Health CARE Account? The term “satellite account” has been used to describe different kinds of activities undertaken at statistical agencies, both in the United States and abroad. As Barbara Fraumeni put it, satellite accounts can show more detail than that present in the national accounts, or they can extend to areas not covered in the NIPAs at all. Research to improve the methodologies and data that underlie the NIPAs and GDP measurement is constantly active at BEA (and other statistical agencies as well); historically, satellite projects that entail experimental work intended to enhance the detail and accuracy of the accounts have been part of the effort. BEA’s experimental research and development (R&D) satellite account is an example—it provides a more detailed look at the composition of R&D costs and a more accurate measure of capital investment. In other instances, satellite projects have been designed to improve data on economic activities typically considered peripheral or even out of scope of the NIPAs and GDP. Environmental accounting is an example of this kind of work. Ana Aizcorbe kicked off the workshop by outlining the agency’s plans for a satellite health care account. She began by defining the purpose of the project. Compared with work by academic health economists on data systems designed to track changes in population health alongside the investment of inputs to health, she described the BEA initiative as modest. BEA is most concerned with one input to health—medical care—because it is the piece that is most relevant to the market-oriented national accounts; it is also numerically important (medical care accounts for a large and growing portion of GDP), a key policy topic, and the component of health on which BEA is most likely to be able to make progress. David Cutler supported BEA’s approach, noting that, in phasing the project, there are tasks that BEA will clearly be able to do itself and others that will require collaboration. For example, BEA is not going to hire a staff of physicians to establish what is happening with patient outcome trends in order to assess the changing quality of medical services. That will involve academic work and, per- haps someday, other government departments, such as the Department of Health and Human Services. At several points during the proceedings, participants noted the importance of differentiating aspects of the research agenda that should take place inside BEA with those that should progress elsewhere.   The satellite account also introduces a different conceptual approach to R&D spending, treating it as investment.   The BEA website provides a wealth of information on the methodologies, content, and scope of the NIPAs; “Measuring the Economy: A Primer on GDP and the National Income and Product ­Accounts” (http://bea.gov/national/pdf/nipa_primer.pdf) provides a good starting point. Information is also avail- able there about the various satellite accounts that have been developed over the years by the agency. Additional discussion of the scope and coverage of the NIPAs and the role of satellite accounts can be found in a CNSTAT report (National Research Council, 2005, pp. 14-19). 

INTRODUCTION  A Broad Population Health Account The scope of the national income and product accounts has traditionally been limited, with some exceptions, to activities that take place within the market. But, because it has long been recognized that a quantitatively significant amount of economic activity occurs beyond the market, efforts have been made to expand accounting structures to allow broader measurement of the economy. The history of economic accounting includes forays by statistical agencies into such projects with broader scope. One example is the satellite accounts that have been devel- oped in various countries, including the United States, to estimate the economic contributions to society of the environment (or parts thereof). Another example is a household production account, with which several statistical agencies around the world have begun experimenting. The range of potential “GDP-expanding” areas of nonmarket accounting is catalogued in Beyond the Market: Designing Nonmarket Accounts for the United States (National Research Council, 2005). For the topic at hand, the broad-scope satellite account would define popula- tion health as the output and would encompass nonmarket components beyond those that are included in the national economic accounts. Sherry Glied described the idea clearly: Medical care and an array of other variables would be organized into a kind of health production function, thereby providing a framework for measuring changes in health and the factors affecting that change. This output concept—designed to track the value of incremental changes in health, as well as inputs in the production of health, such as medical care, the environment, life- style choices, and other factors—is clearly broader in scope than that which now guides measurement of the medical care industry of the economy in the national accounts. Natural resource and environmental accounting have the same basic characteristic; there are additions (or subtractions) to GDP driven by inclusion of nonmarket elements. During open discussion, Jorgenson identified the linkage between the medi- cal care component of the NIPAs and this broader kind of population health account: Output of the medical sector is very important in GDP measurement but, for many aspects of health policy, medical care outcomes are of the greatest concern. The inputs of the broader account would be precisely the treatments that are the outputs of the market account. In other words, the medical sector produces treatments that are, in turn, inputs into the production of incremental improve- ments in health. So there is a conceptual framework that links industries and their inputs to treatments, and treatments to outcomes; in both cases, ­Jorgenson noted, it is useful to focus on treatments, because that is what would enter into the GDP. He went on to describe measurement of medical care outcomes as the $64 tril- lion question. Here, one must deal with the issue of how medical care produces improved health, if it does, which is very difficult. The objective of the broader kind of account is, like standard market-­oriented economic approaches, to accurately estimate prices and quantities. The quantity estimate is linked directly to changes in health outcomes. When there is improve- 

 Strategies for a BEA Satellite Health CARE Account ment, people are able to enjoy longer, healthier lives, so there is an incremental change in individuals’ lifetime “income,” broadly defined, as a result of medical treatment. Finding the part of that improvement in health that is attributable to medical care is a very important research task. Jorgenson noted that workshop attendees had made numerous important contributions in this area; the most prominent being the disease-specific studies, such as the one on heart attacks by Cutler, McClellan, Newhouse, and Remler (1998). Jorgenson concluded that the nonmarket component of health accounting is presently at the research frontier. It is not something that BEA could, given the current state of the art, fruitfully think of as part of its satellite system now under development; but, he advised, these ideas and goals—particularly as they relate to measurement of the changing quality of medical care—should be kept in mind at all times as part of the long-run objective. The BEA Medical Care Account: Experimental Work to Improve the Market-Oriented NIPAs Jorgenson suggested that perhaps the satellite account concept was not quite the way to think about BEA’s project. He considers the broader health accounts (described above) a “satellite” to the NIPAs mainly because, in such a sys- tem, a nonmarket component of output is present—the incremental value of b ­ etter health—that supplements the treatment of output in the national accounts. F ­ raumeni expressed the view that, independent of whether or not it is considered satellite work, the key substantive point is that BEA’s objective initially should be to improve measurement of the medical care component of the national accounts rather than to construct a new set of accounts defining a different boundary of goods and services. She characterized BEA’s research as focusing on a different way of measuring the prices of medical care. This is critical because new price indexes mean different deflators, which, in turn, affect estimates of real medical care GDP. Aizcorbe agreed with this characterization—that the agency is trying to fix the price deflator, which requires supporting work that is experimental. Because price measurement work is central to the BEA program, other statis- tical agencies are involved. Triplett noted that, for in-scope market items, any new pricing methods that are developed can be taken back to BLS as a better way of doing things for certain purposes. He cited price adjustment methods for quality change in the computer index as an example. During the 1980s, BEA calculated the computer index using new methods that were then brought over to BLS; sometimes the methodological improvements have flowed the other way. As part of this new approach to the medical care component of the NIPAs, BEA would need to develop a technique so that the revised prices associated with expenditures are reflected in the industry accounts as well. As Triplett put it, the measurement issues can be viewed either from the demand side (the CPI, for example, is constructed from the viewpoint of the consumer) or from 

INTRODUCTION  the production or supply side, at the industry or subsector level. In price index theory, those two sides sometimes yield different measurement implications, and he asserted that this will also be the case for health care. Newhouse stated that what BEA’s work is really doing is changing the good that is being priced. That affects the price index for consumer expenditures, which carries over to the prod- uct accounts. Both sides of the account—expenditures on outputs and costs of inputs—must, by definition, balance. Fraumeni made this point: An input-output framework underlies the NIPAs and, to make the new expenditure structure agree with the industry side, a new industry piece has to be created (how this piece is to be created is the topic of discussion in Section 3.2.). The main point here is that BEA’s upgraded account will not involve a revi- sion of scope—it will be a system that stays within the market-oriented boundar- ies of the national accounts and that does not explicitly extend to a nonmarket component. Jorgenson agreed with this strategy, stating that BEA’s new work will provide an alternative way of measuring medical care prices and it is (and should be) limited to that. BEA director Steven Landefeld provided additional clarity about the objectives of the program. He reiterated the point that, historically, satellite accounts have been pursued for different purposes—one being to expand the scope of the accounts, another to provide more incremental detail or to modify the structure within scope. It is the latter that the agency is pursuing, initially in an experimental context, with the new disease-based organizing approach. BEA’s motivation is to improve measurement within the existing NIPAs, and Landefeld noted that this work could eventually affect overall measured rates of inflation and real GDP. Every time a better price index is introduced, it changes the defla- tor used to calculate real GDP, even if it does not change the categories or the scope of the accounts. Landefeld cited BEA’s travel and transportation satellite account as an exam- ple of work that takes place within the scope of the existing accounts but that breaks out additional detail for what is already covered. As in this work, which produces some fairly experimental numbers, he agreed that the agency needs to be more explicit in distinguishing between what will (or could) and will not affect official GDP estimates. He reiterated that BEA’s work on medical care is, at this point, not geared toward expanding the scope of the accounts, and that the agency would leave it to others to conduct the cutting-edge research on that front and on the issues that directly affect the public health policy debate. Indeed, the reason that the agency has explicitly partnered with Cutler and Rosen is because it is not in a position to handle the expansion of scope side of the research agenda. 1.3. Objectives of the BEA Project—A Staged Strategy for Developing a National Health CARE Account The remaining chapters of this report summarize the day’s proceedings. The views expressed by workshop participants are organized topically, more or less 

 Strategies for a BEA Satellite Health CARE Account following the logical flow of process steps that may take place as BEA’s work on health accounting unfolds. Essentially, BEA presented its plan, then participants provided critique, feedback, and detailed guidance on the plan. A sketch outline of BEA’s project phases (and the rest of this report) follows. Categorizing Nominal Expenditures on Medical Care: Using a Disease-Based Framework The first major topic, and the one on which BEA is currently focusing most of its effort, is to rework the way expenditures on medical care goods and ser- vices are organized. The idea is to identify units of output that are meaningful from a consumer standpoint—and to get price and quantity measures for those units right. For example, a patient is not typically interested in buying an hour of a doctor’s time or a day in a hospital; rather, a patient seeks treatment for a particular condition or ailment—the consumer wants to buy improved health. Workshop participants identified several potential units of measurement, such as a patient encounter or an episode, and considered the strengths and weaknesses of each. In Chapter 2, we describe characteristics of a system, as envisioned by workshop participants, for defining expenditure categories or “buckets” in which to allocate dollars spent in the economy on medical care. One of the most difficult issue that arises is comorbidity; when patients require medical care for multiple conditions, it becomes much more complicated to assign expenditures to predefined categories accurately. For a seemingly simple event—a visit to the doctor—even at the claim level, there can be up to four diag- noses. So the question becomes how to make the judgment call in these cases: for the depressed patient with back pain who visits the doctor, should dollars spent be apportioned to the back pain or to the depression category? As described in this report, workshop participants considered and discussed several different m ­ ethods—first listed diagnosis, grouper algorithms, person-based, etc.—for pars- ing spending in the presence of more than one condition. During the opening presentation, Aizcorbe laid out the approaches that BEA is currently studying for allocating medical care spending by disease and for han- dling the comorbidity problem. She noted that it is not yet clear which one will be most appropriate for the national accounts. So far, much of BEA’s work has relied on “grouper” programs that sort spending using claims data and thus is based on the patient’s history. These programs allocate the dollars recorded on each claim into particular disease categories. Aizcorbe stated that, even if this does not turn out to be the option that is ultimately used, BEA would want to be able to answer the question, “Why weren’t episode groupers used for parsing expenditures by disease?” The common complaint among workshop participants about grouper programs is that, at this point, they are constructed in a methodological black box; they are proprietary, and it is not clear if or when BEA would be given access to their internal workings. BEA participants expressed the view that, once the unit 

INTRODUCTION  of output is defined indicating how expenditures are to be parsed, pricing medi- cal care inputs (doctors’ time, hospital services, drugs, medical equipment, etc.) will pose less severe conceptual problems. There are exceptions, however. For example, high-tech medical equipment is poorly measured in U.S. statistics, but this is not very different from other industry accounts in the NIPAs. Another task on BEA’s agenda is to continue work to reconcile the nominal spending estimates in the NIPAs with those reported by CMS in the national health expenditure accounts. In contrast to BEA’s proposed framework, in which the focus is on organizing expenditure data by medical treatment categories for the full range of specifically defined diseases and conditions, CMS’s accounts— which are the official estimates of health care spending for the United States—are designed to track final payments by type of provider. Although the data systems at the two agencies serve different purposes and therefore will not be identically structured, expenditure totals on medical care and for equivalently defined sub- categories should align. Collaborative research is under way at CMS and BEA to further this objective. Price Indexes: Calculating Real GDP for Medical Care Many of the difficult tasks required for developing a satellite health care ccount relate to the calculation of price indexes needed to deflate nominal expen- diture estimates for the measurement of real GDP. Aizcorbe expressed the view that the treatment-oriented price indexes that have been advocated by health economists are what should be used as deflators for the national accounts. For double-entry national accounts such as the NIPAs, deflators must be created for both the spending side (consisting of outputs purchased by consumers) and the industry side (consisting of inputs provided doctors, drug producers, hospitals, etc., in the production of medical care). Brian Moyer, whose presentation is sum- marized in Section 3.2., outlined BEA’s strategy for modifying the structure of the industry accounts and for implementing the types of deflators that would be needed for the totals of each side to agree (or at least to minimize the statistical discrepancy between the two). When a good is defined as the treatment of a disease or episode of illness, it necessarily means that a data series on spending by disease must be generated. Aizcorbe stated that figuring out how to use these data series to construct improved price deflators was a high priority for BEA. Sherry Glied, speaking from the per- spective of a health economist (as opposed to an economic accountant), succinctly summarized the tasks at hand: Currently, many individual goods and services are delivered by different industries—doctors, hospitals, pharmaceutical companies, and so on. The plan is to continue to deflate those things using the PPI, or some- thing similar, then to take all those individual components and ­create something called medical care, which is not any one treatment but a set of treatments for a disease. Doing that involves integrating services from different industries, pric- 

10 Strategies for a BEA Satellite Health CARE Account ing those bundles of services, then, ideally, capturing quality change associated with various medical innovations that may, among other things, allow substitution among medical care inputs. Items in this medical care box, then, will be deflated using price indexes organized along these disease treatment lines. Index construction may draw from both the BLS’s CPI and PPI programs, and BEA may introduce new price indexes specifically designed to meet the needs of the satellite account. The PPIs provided by BLS may turn out to be close to what is needed to deflate industry-side items, such as office visits, prescription drugs, and so on. However, the PPIs are not adequate for use on the spending side because the “good” demanded by consumers—treatments—does not align with this kind of delineation; rather, it combines these items. Participants from BEA and BLS speculated that the differences between estimates based on the different kinds of indexes are numerically important. Pre- liminary research by Bradley and his colleagues (see Appendix A) found clear differences between indexes based on the PPI scope and the CPI scope. BEA has also done research and commissioned outside work using a different episode grouper that found significant variation as well. Measuring Quality Change in Medical Care Ideally, medical care price indexes used to deflate the economic accounts would be capable of reflecting changes in the quality of medical goods and ser- vices. In national income and product accounting, it is customary to adjust for instances when goods and services—in this case, treatments—are getting better or worse. Thinking seriously about how to measure changes in the quality and, in turn, the real cost of medical care requires monitoring information about out- comes associated with that care. Glied noted that, for this work, it seems logical to implement an episodes-of-treatment concept that reflects the way in which the medical profession—and not economists so much—measures outcomes. One way to compare quality across treaments, for example, is to examine clinical trials; while they have limitations, it would be counterproductive to construct episodes of treatment that did not match with outcome concepts that are being measuring in other fields. BEA staff reported that they will postpone tackling the difficult quality change issue in the immediate future. The agency does not have medical exper- tise, so its strategy is to go after the things it believes can be fixed; later, research from the public health and health economics fields may advance enough to provide guidance about how to move forward on the quality measurement and outcomes pieces. Aizcorbe indicated that, even though BEA is not in a position to make major headway on the topic now—the program is still very much in the research stage—the agency realizes that accounting for quality change is an ultimate goal and that this need should be kept in mind as the satellite framework develops. 

INTRODUCTION 11 Other Issues: Source of Payment Data Throughout the course of the day, a range of other topics was covered. One of these is data needs—especially those related to expenditure accounting and to price index development. Those working on these topics have, because of their complexity, discovered the need to draw from a broad array of data sources— aggregate and micro, longitudinal and cross-sectional, survey and administrative, public and private. Notes on data issues appear throughout this report, particularly in Sections 2.4. and 3.6. Source of payment data is another important accounting topic raised by J ­ orgenson. The national accounts view GDP as an aggregate measure of eco- nomic activity that involves multiple sectors. In the case of medical care, house- holds supply the patients; the business sector includes the providers of medical care; government plays an important role as a payer of a large portion of this medical care or, in some cases, by functioning as an insurer. Jorgenson, along with participants from CMS, noted the importance of keeping track of payments for medical care by the various sources. As noted above, the national health expenditure accounts compiled by CMS are organized by sources of payment—the part paid for by the government, the part paid for by private insurance, the part paid out-of-pocket by the household sector, and so forth. Because the national accounts are used to monitor the gov- ernment budget, they have to be able to indicate precisely the level of public expenditures used to purchase or pay in part for medical services. For private sources of payment, even though they constitute a relatively small portion of total payments, it is also important to distinguish between the parts made by indi­ viduals and by businesses. Here, the concern is not with the business of providing medical service, but the role of businesses as buyers of health insurance; the data must allow users to distinguish the health insurance industry and its activities from those of the medical care sector itself. When thinking about payments for treatments generated by providers, J ­ orgenson noted that those have to be segregated, as they are in the national health expenditure accounts, by sources of payment, so that links can be made to the other sectors in the economy. These accounts support a key function, which is to document the flow of payments among households, industry, and govern- ment budgets. In developing satellite accounts, it is important not to lose sight of the tremendous heterogeneity that characterizes medical care and how that is reflected in part in the sources of payment data.