Health Data and Health Policy
THE PURPOSE AND VALUE OF AN EXPANDED SYSTEM OF NATIONAL HEALTH ACCOUNTS
According to the Centers for Medicare & Medicaid Services (CMS), in 2007 the United States spent $2.2 trillion—or $7,421 per person—on medical care, which accounted for 16.2 percent of the gross domestic product (GDP) for the year. As a proportion of GDP, medical care spending in the United States substantially exceeds that of other industrialized countries: among all Organisation for Economic Co-operation and Development (OECD) countries, the median proportion is 8.5 percent. Are the benefits from this magnitude of medical care spending worth it? This question is one of today’s major economic issues.
Medical care spending in the United States is not only large, but it is also growing rapidly—up by 56 percent since 2000, it has roughly doubled since 1996. Increasing health care costs pose political questions that are heightened because the U.S. government pays a substantial portion of those costs in the form of Medicare, Medicaid, and other publicly funded programs. Indeed, growing health care costs are a public policy issue in all industrialized countries and, in most of them, governments finance more of the nation’s health care than is the case in the United States.
What are the sources of increased medical care spending? What are the consequences of reigning in its rate of growth (if it can be done)? These vital questions must be answered in order to forge a consensus on public policy toward health care costs; however, they demand data to enlighten debate on them. The United States has aggregate data that chart its total medical care expenditures and rate of growth. The most widely cited estimate of total health care costs comes from the National Health Expenditure Accounts (NHEAs), compiled by CMS. An alterna-
tive estimate of the total can be extracted from the National Income and Product Accounts (NIPAs) that yield information on GDP and its growth and composition. The national aggregate tells us little that is useful for addressing the major questions listed above. Knowing that total medical care spending is growing is only part of the story. We also need to know if this growth is producing benefits that are keeping pace and are, in some sense, worth it. Knowing that expenditures are growing is not the same thing as knowing why they are growing.
At a highly disaggregated level, a great amount of survey and administrative data also exist that are useful for analyses of medical care, health status, and medical expenditure trends. Missing, however, is a comprehensive analytical database lying between the two extremes—one that permits analysis at a national level of the forces driving medical care costs, of growth in medical services, and of the benefits received from expenditures on health. The panel believes that a combination of improved economic accounts for medical care and a new database for analyzing the health care sector and its impacts on population health is essential for informing policies in these areas; microdata—survey and administrative records—will also always be necessary for understanding of medical care spending, health status, and costs.
Recommendation 1.1: Work should proceed on two projects that are distinct but complementary in nature. One accounts for inputs and outputs in the medical care sector; the other involves developing a data system designed to track current population health and coordinate information on the determinants of health (including but not limited to medical care).
Groundbreaking work that will create a foundation for the first project—the medical care account—is indeed already under way at the statistical agencies. For the second project, measures of current population health could be developed quickly, and information on the determinants of health could be collected more aggressively by the statistical system immediately. However, using the health data system to construct a national health account, as outlined in Beyond the Market (National Research Council, 2005), that systematically tracks population health and its determinants requires much more research on the impact of health care on outcomes, which determines its value. The two kinds of data systems are summarized in the sections that follow and elaborated in greater detail in Chapter 2.
An Economic Account for the Medical Care Sector
An economic account integrates data from diverse sources within an overall analytic framework, so that each individual entry is consistent with all the others and with the total. The best known examples of economic accounts are the NIPAs, compiled by the Bureau of Economic Analysis (BEA), and those produced by the statistical agencies of countries around the world. In the NIPAs, consump-
tion, investment, government spending, and net foreign trade aggregate to GDP; at the same time, labor compensation, corporate profits, and rental and other incomes aggregate to a total flow of income, which balances with the flows on the expenditures side. Any component of the accounts can be used for analysis in conjunction with any other component and with any total or subtotal in the system (linking consumer spending on home electronics to household disposable income, for example). An economic account provides an integrated system in which the data can be arrayed according to the questions asked.
The existing NHEAs provide integrated information on health care spending. Although they serve a number of program and research needs, they are not designed to permit analysis of critical questions concerning medical care costs. The NHEAs provide data on the sources of financing of the medical care system (government, insurance companies, etc.) and on the institutional units—or industries—that receive funds (such as hospitals, doctors’ offices, pharmaceutical providers). The NHEAs do not, however, provide any information on what is purchased with those expenditures. They are not disaggregated to show the amounts spent for treatments of diseases or other health impairments, nor are the trends in spending disaggregated along treatment, or cost-of-disease, lines. By analogy, it is as if the NIPAs presented information on consumer expenditures at department store and grocery store industries, without any information about spending patterns for clothing and furniture or for meat and fresh vegetables. Consumers do not go to grocery stores to buy grocery store visits—they buy meat and potatoes; similarly, medical insurers and others who fund medical care do not buy hospital visits, they buy treatments for heart attacks and broken legs.
Price indexes for hospitals, constructed by the Producer Price Index program of the Bureau of Labor Statistics (BLS), are already collected and published using a cost-of-disease classification, the only workable framework for obtaining price information about the cost of medical care (see Chapter 4). What has so far been missing from the NHEAs (and from the NIPAs) is a disaggregation of health care expenditures along the same lines. It is a major shortcoming in national statistics that health care, which absorbs 16 percent of GDP, is the only sector of the economy in which price indexes exist for deflation (at least for hospitals) but no comparable expenditure data exist to deflate.
Moreover, even though the NHEAs contain an investment component and data on spending on pharmaceuticals, they do not present a systematic accounting for all the inputs into the production of medical care—skilled and unskilled labor, services of capital equipment, such as scanners and so forth, pharmaceuticals and other inputs, including energy—arranged in such a way that they can be linked to the output of the sector. That is because the NHEAs are conceived as only an accounting for the sources of medical funds and the recipients of funds, not as an account that links the inputs to medical care with the output of the sector.
Thus, the first need for analysis of medical care spending is to construct an account for the medical care sector that matches what is available for other com-
parably sized sectors of the economy. It would link medical inputs to the output of medical care and would be parallel to the account for any other industry in BEA’s industry accounts (e.g., electronics and computer production, legal services). BEA’s industry accounts are an adjunct to the main accounts for GDP; they disaggregate the total economy into expenditures by industry, and they also link the outputs of industries with their respective uses of inputs. The medical care account’s major requirements are comprehensive measures of the inputs to the production of medical care (including pharmaceuticals, skilled professionals, and high-tech capital equipment) and a measure of the output of the sector. It would resemble an enlarged and improved version of the account for the health and social services industry that is already in the current BEA industry accounts.
Ultimately, as described in Chapter 2, a number of improvements in BEA’s industry accounts will be necessary. Perhaps the major change from the current statistical data on medical care is the disaggregation of total medical expenditures along cost-of-treating-disease lines. The formidable problems in doing so are subjects of Chapter 3. However, we emphasize that a long history of cost-of-disease accounts, dating at least to Rice (1966) and extending to Hodgson and Cohen (1999) and others, indicates overwhelmingly that construction of such accounts is feasible. Such accounts have also been estimated in a growing number of OECD countries, including Australia, Germany, the Netherlands, and the United Kingdom (Organisation for Economic Co-operation and Development, 2009). In Canada, they are now a part of the country’s medical care information system. Methods for allocating expenditures to the International Classification of Diseases categories have improved over time to the point that, in recent studies, over 90 percent of spending has typically been accounted for (see http://www.rti.org/files/COI_Reviews.pdf for a review of recent cost-of-illness studies and estimates). This is despite allocation difficulties involving diagnostic expenditures, for dealing with activities associated but not clearly located in a particular disease (like ambulance services or long-term care), items that affect many diseases (like the sale of aspirin), and various overhead costs for facilities and equipment. That cost-of-disease accounts have been estimated so frequently does not mean that improvements are not needed (see Chapter 3). But they are not untried exercises, as they are sometimes mistakenly perceived to be. The challenge is to improve on what has been done before and to integrate that work into an account that provides information on cost trends over time.
Disaggregating health care expenditures along a cost-of-disease framework would be useful for addressing a range of questions: Is spending increasing at a more or less constant rate across all diseases? If not, which disease categories have the fastest increases and which the slowest? Does evidence corroborate that disease categories whose treatments involve the most rapid changes in technologies have the fastest rates of growth in expenditures, as is often alleged (but infrequently backed up with solid data)? Do changes in expenditures by disease categories correlate to changes in incidences of diseases, or are trends in the population’s
underlying medical status not the primary drivers of health care spending? It is not possible to design effective policies toward health care costs without being able to address such questions, yet at present no aggregate national estimates can be brought to bear on them, and much of what is known consists of partial and incomplete information and a great deal of more or less informed anecdotes.
Current information is also inadequate for ascertaining whether expenditure growth in treating diseases comes from an increase in treatments (that is, an increase in medical services), from an increase in the prices of those services, or from both. The science and technology of medical research is producing an ever-expanding and increasingly sophisticated array of life-extending and life-enhancing treatments for the range of conditions and diseases that afflict the population. Do the prices of treatments rise with increasing use of high-tech medical procedures and equipment, as sometimes alleged, or are expenditures rising because those advances produce increases in the quantity of medical services demanded and used? The expenditures in the medical care account need to be partitioned into price (medical inflation) and quantity (growth in medical services) terms, as is done routinely for accounts covering other parts of the economy.
For a medical care account, we argue that output can be defined as a completed treatment for a specific condition, or an episode of care.1 The quantity of medical services so defined is the total output of the medical sector, and in line with the previous discussion, this output is multidimensional. Medical services arrayed by disease treatments provide the “product detail” classification for the sector that is parallel with product detail routinely provided by the statistical system in other sectors, such as electronics and computer production.
In other sectors of the economy, the usual method for estimating growth in quantities of goods and services is to “deflate” expenditure change, at a detailed level, by a price index. However, price indexes (measures of inflation) in medical care have long been considered problematic, and although substantial improvements have been made in recent years, much remains to be done. We review the problems of estimating price indexes for medical care in Chapter 4. Alternative quantity-based methods are attractive in countries where the government provides medical care directly (so the price charged is not relevant), but they are also worth considering for the United States because the usual practice of deflation by price indexes has both conceptual and practical disadvantages in the case of medical care.
The “health” account summarized above takes as its boundary the medical care sector of the economy. In terms of a distinction usually drawn in the literature, it is an account for activities that lie within the market sector, though perhaps not
entirely so. For example, time used for the care of the ill by volunteers, friends, and relatives could well be considered within the scope and included in the account’s inputs without changing its basic character. In the medical care sector account, the output is the treatment of diseases, plus ancillary activities that contribute to control of disease, like periodic medical checkups and diagnostic services.
A Data System for Economic Analysis of the Production of Health
The medical care sector does not produce health; it produces medical services. The distinction between the production of medical services and the production of health is an essential one that not only has several implications, but that also suggests an inherent limitation in the type of account discussed above. Even if one were able with an improved medical care account to partition the growth in medical care expenditures into price change (inflation) and quantity change (growth in medical services) components, it still may not be possible to judge whether or not the quantity changes are in some sense excessive: Is too much being spent on late life-stage surgeries relative to preventive measures? Is too much being spent on knee replacements relative to weight reduction? Or, even, is too much spent on medical care and not enough on public health measures? Just knowing the rate of advance of medical services, arrayed by disease treatments, is not the end of the story. The big question remains: Is the $2.2 trillion the United States is spending on health care worth it? And that translates into a series of questions of the following type: Is the amount the United States spends on circulatory diseases, for example, too much, too little, or about right in terms of what we get for the dollars expended?2
Addressing what society and even what individuals get for their medical expenditures is complicated because the output of the medical care sector is not health. Instead, health is “produced” (in the language of economics) in conjunction with many other variables, or inputs, in addition to medical care: diet, environmental factors, lifestyle, and so forth. Changes in a nation’s health may owe as much or more to those nonmedical determinants as to medical care services, a relationship that has been well articulated in epidemiology since at least McKeown (1976).
It may seem plausible that increasing the resources put into health care should yield improvements in health at a nationwide level (so the United States,
The kind of detailed cost-benefit analysis required to answer such a question has permeated policy actions in some countries more deeply than it has in the United States. For example, in the United Kingdom, the National Institute for Health and Clinical Excellence determines what surgical procedures, screening tests, and drugs the national system will pay for—a drug that is predicted to add an extra six months of good-quality life for a patient for $15,150 or less is automatically approved; one adding six months for $22,750 or less might get approved; and more expensive medicines have only rarely been approved (see http://www.nytimes.com/2008/12/03/health/03nice.html?pagewanted=1&th&emc=th).
which has the world’s largest medical care expenditures per capita, should have the world’s best health); however, a tight relation between expenditures on medical care and the level of health does not necessarily exist, in part because of the influence of nonmedical determinants of health.
The nation therefore also needs a broader data system than that provided by the economic accounts for medical care, one that could be used to relate changes in the nation’s health to the factors determining health. As with economic accounts, this data system would be designed to facilitate investigations of the link between output (health status) and inputs (which are all the influences on health, including medical care). On the output side, growth consists of changes in the population’s health itself, and completed medical treatments are inputs to this production process. On the input side are statistical data on the array of factors—including, but not limited to, medical care—that affect the population’s health status. This second kind of economic data system is a long-term project that is much more on the research frontier.
The medical care account and the broader health data system are linked, and work on the former is an integral part of, and logical first stage project in the production of, the latter. Moreover, these data and analytic systems could logically be constructed along disease lines, which underscores the vital role that cost-of-disease accounting (Chapter 3) plays in each. Accordingly, construction of the health data system requires data from the medical care account and work on the two needs to proceed in a coordinated fashion.
Constructing a comprehensive health data system is an ambitious undertaking. The first major empirical difficulty is measuring the output—health—a task that requires going well beyond standard longevity figures and death rates to consider morbidity. In Chapter 5, we review summary measures of health and the state of research on health measures, and consider the research issues that must be resolved if they are to be useful in an economic accounting framework.
In Chapter 6, we discuss approaches for attributing the output (improved health) level achieved to the inputs or determinants of health—medical care as well as nonmedical and nonmarket inputs to health. Measurement of the inputs that determine health also poses serious challenges. Much still remains to be learned about the nonmedical determinants of health. Fortunately, the body of knowledge about the effect of these factors—things such as diet, exercise and environment—on the population’s health is growing.
REPORT AUDIENCE, REPORT STRUCTURE
The Panel to Advance a Research Program on the Design of National Health Accounts was assembled to examine how health and medical care data systems might be designed and implemented within a national accounting framework. The panel was charged to study and make recommendations about where to target research to best improve the knowledge base for developing statistical
data for relating the population’s health status to an array of factors—including, but not limited to, medical care—that affect that status. During the study, the panel confronted a number of vexing conceptual issues that are integral to both (1) improving input and output accounting of the medical care sector, and (2) developing broader health accounts and data systems. The panel’s priority was to provide guidance on how best to further develop the market-oriented medical care components of the NIPAs (and the NHEAs); this strategy was in large part a reflection of our belief that these improvements are also essential building blocks for creating a national data system that extends and organizes information about the inputs and outputs of health. The panel was also asked to assess the extent to which existing data sources—such as those housed at CMS and the National Center for Health Statistics (NCHS)—and emerging sources—such as the new BLS American Time Use Survey—are meeting data needs for a coordinated health monitoring data system.
The audience for this report includes statistical agencies, research funding agencies, academic groups working on satellite health accounts, and Congress, which will be funding improvements to health data generally. For the most part, the audience categories parallel those into which the panel’s recommendations fall. Our recommendations on constructing a medical care account are directed mainly to federal statistical agencies, including BEA and the agencies that provide source data to BEA, which include the Census Bureau, BLS, and NCHS. One extremely positive development is BEA’s initiation of a project to build an account for medical care as a “satellite” account to the NIPAs. For this effort, we (1) assess what the agency could reasonably accomplish in the short term in light of what is currently known and what data are available, and (2) make recommendations to the source agencies that compile medical and health care data for their improvement and modification, with the aim of facilitating construction of the new medical care account.
The second audience includes researchers developing health data programs and the agencies funding these initiatives. For example, major topic areas such as measuring nonmarket and nonmedical inputs to health and defining and measuring population health are, for the foreseeable future, mainly the province of researchers and funding agencies interested in establishing links among medical care, health outcomes, and population well-being. For this group of recommendations, our guidance takes a tone of “here is what is still unknown and where more research and more (or different) data collection could fruitfully be commissioned.”
Many (if not all) of the areas of emphasis by the statistical agencies—for example, medical care price indexes and expenditure accounts—are intermediate steps in the broader health data project. For example, an initial task for both is to get the spending-side categories and estimates in order and to confront the problem of how to distribute nominal expenditure estimates into meaningful production units to which prices and quantities can be attached. Indeed, this is really where
both groups—those working on medical care accounts and on the linking of health inputs and outputs—are focused at the moment. Even where the projects do not currently overlap, they eventually may. For example, academic researchers are already well into thinking about how to track outcomes of the medical sector while statistical agencies are not. Ultimately, however, both BEA (NIPAs) and BLS (price index work) will also need to integrate medical outcome data because effectiveness of treatments (or preventive or diagnostic services), in terms of their impact on patient health, is the quality dimension that needs to be monitored in order to estimate changes in real prices and quantities correctly. As the statistical agencies continue work relevant to a satellite medical care account, they need to keep the broader and more ambitious ideas in mind so that the programs progress in a way that will not impede future work to track health care outcomes and quality change. Thus, those whose primary mission is funding research should be aware that the work under way at the statistical agencies, particularly BEA, and the data needed for this work are also part of the long-term research project to provide an accounting for the determinants of health.
The remainder of this report is structured as follows: Chapter 2 outlines the purpose and design elements of medical care and broader health accounts and discusses the conceptual bases for each. Boundaries for the medical care account are considered, and inputs and outputs are disentangled and catalogued. Chapter 3 focuses on the task of allocating medical care expenditures within a treatment-of-disease organizing framework. Disease classification schemes are assessed and the strengths and weaknesses of alternative approaches to allocating expenditures are identified. Some attention is given to data needs and alternative modeling approaches. Chapter 4 lays out strategies for measuring prices and quantities of medical care—essential for an economic account—through improved price indexes. Chapter 5 considers the role of population health measures in health accounting and offers guidance about which among the competing metrics are most promising. In Chapter 6, we discuss how data systems can be used to support disease modeling research oriented toward linking changes in population health to the array of inputs that affect it. Attention is given to how the U.S. health data infrastructure could be upgraded—in part through development of economic accounts—for this purpose.