Accounting for Health and Health Care: Approaches to Measuring the Sources and Costs of Their Improvement (2010)

Health and health care in the 21st-century United States can be partially characterized by a few indisputable trends. The country is devoting a large and growing share of its resources to medical care, spending $2.3 trillion, or 16 percent of gross domestic product (GDP), on it in 2007. That figure is projected to rise to perhaps $4.2 trillion, or 20 percent of GDP, by 2016 (Poisal et al., 2007). Not only is this spending massive in an absolute sense, it is also high comparatively, as the median expenditure of Organisation for Economic Co-operation and Development countries is around 8.5 percent of GDP. The budgetary pressures being created by this allocation—at individual, organizational, and national levels—have led to growing criticism of the productivity of U.S. health care. As the population continues to age, the strain that will be put on publicly funded programs such as Medicare will only elevate the debate about how best to meet the nation’s health care needs.

One factor contributing to the trend of rising expenditures is the ever-expanding and increasingly sophisticated array of life-extending and life-enhancing treatments produced by medical research for the range of conditions and diseases that afflict the population.² In addition, the body of knowledge about the effect of

nonmedical factors—such things as diet, exercise, and environment—on people’s health continues to grow.

For all these advances, a gaping hole in information still exists. Relative to knowledge about health care expenditure and medical science, much less is known about the return that individuals, and society in general, receive for the investment in health. Given the massive amount of resources that are spent publicly and privately, it is astonishing that so little effort has been made to systematically assess what we are buying for this investment.

At the heart of this information chasm is the need for data on how inputs into medical care translate into outputs—completed treatments and procedures—that, in combination with other factors, ultimately affect the population’s health. While some studies have suggested that productivity growth in medical care is reasonable in aggregate (Cutler and McClellan, 2001; Cutler, Rosen, and Vijan, 2006), others argue that there is substantial waste at the margin (Fisher et al., 2003; Skinner, Staiger, and Fisher, 2006). In this report, we provide guidance about what data are needed to measure the outputs produced by the medical care sector. Without this kind of information, it is impossible to credibly assess whether the nation spends too much or too little on medical care relative to, say, public health measures, and, perhaps more importantly, whether we purchase something close to the right mix of medical care goods and services for a given level of resources expended. In order for policy makers to pursue actions that reduce costs sensibly, improve performance, and, in general, enhance the efficiency of the national approach to health and medical care, a more systematic approach to compiling data for the purpose of tracking productivity in the sector is needed.

The National Income and Product Accounts (NIPAs) produced by the Bureau of Economic Analysis (BEA) and the National Health Expenditure Accounts (NHEAs) produced by the Centers for Medicare & Medicaid Services (CMS) are the foundational components of the U.S. health care data infrastructure. While the virtues and utility of these data sources are well known, they are not sufficient on their own to inform policy. The national accounts have particular difficulty decomposing medical spending increases into price, quantity, and quality change elements. For example, an increase (or decrease) in the observed price of treating a disease may reflect a change in the price of unchanged treatment inputs, a change in the amount of inputs (e.g., a surgeon’s time) required, the development and use of new drugs or procedures that alter outcomes, or simultaneous changes in more than one of these factors (National Research Council, 2005, pp. 117-118).

In this report, we offer guidance for extending these important data sources, at first in an experimental or satellite account setting, to better inform national decisions about resource allocations to health care. A health data system built on

economic accounting principles has the potential to provide a comprehensive picture of population health in relation to health care spending within an integrated framework in which consistent definitions, measurement tools, and analytic conventions are used (Rosen and Cutler, 2007). Economic accounts provide the key data infrastructure for measuring productivity, which, in turn, can be used to identify high-return spending and investment areas.

Given the wide range of statistical and research measurement needs, creating health data systems that will adequately inform policy requires a multipronged effort. Economic accounting (as practiced by BEA), price and productivity measurement (Bureau of Labor Statistics [BLS]), medical and health-oriented research (e.g., funded by the National Institute on Aging [NIA], National Institutes of Health), and health monitoring (e.g., National Center for Health Statistics, [NCHS], Centers for Disease Control and Prevention [CDC]) are all data driven, but requirements vary in terms of needed characteristics and complexity. BEA’s top priority is to measure medical care inputs and outputs correctly, since this is a major component of the market economy; the needs of BLS are similar. However, many health policy researchers are interested in establishing links between population health and its determinants—medical and otherwise—in order to develop strategies for improving care in a cost-effective manner. These efforts demand data on the multitude of factors affecting population health and on health status itself.

Because of the nation’s need for more accurate information on its health care system costs and benefits, it is essential to improve the data infrastructure.

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 type of account is indeed already under way at the statistical agencies. Research programs that will contribute toward fulfilling the second objective are still in their infancy (and taking place predominantly in academic settings) and will take much longer to mature. Ideally, though, the national health data system should eventually be maintained by the statistical system.

Constructing a fully developed set of national health accounts involves three broad steps, which define the topics of interest in this report: (1) compiling detailed and comprehensive data covering the nation’s expenditures on and utilization of health care, organized in such a way that prices and quantities of

meaningfully defined units of goods and services can be measured; (2) assessing summary measures that allow the changing state of the population’s health to be tracked along a number of dimensions, and identifying those that are most appropriate for use in a health accounting framework; and (3) developing an integrated data system that allows researchers to investigate links between medical interventions and other health-related activities on one hand and population health on the other.

At the outset, we recognize (and consider it a positive) that work to improve health data within a national accounting framework is moving forward along several fronts and is motivated by multiple objectives. One strain of research is focusing on the inputs and outputs of the medical sector. Here, as would be consistent with the market-oriented NIPAs, units of medical care produced (however defined) can rightly be considered the output. This work is of immediate concern to statistical agencies such as BEA, BLS, and CMS that are responsible for producing expenditure, price, and productivity statistics.

A second strain of research is oriented toward developing statistical data for relating the population’s health status to an array of factors that affect that status in a kind of health production function. Inputs include medical care, time spent by people investing in their own health (e.g., exercise), consumption items (e.g., food), research and development, and the quality of the environment. The output—better health—produced from investments in these inputs includes both length and quality of life dimensions, which can be conceptualized jointly in summary measures of population health. In the broad-concept account, medical care is an input in the production of “health”—treatments are pursued with the objective of extending or enhancing a patient’s years of life.

The two projects are complementary; the analytic pieces for which the statistical agencies are responsible—e.g., medical care expenditure accounts and price indexes—are intermediate building blocks for a comprehensive health data system. Such a system would in turn provide a foundation for developing a national health account, such as that envisioned in the report Beyond the Market: Designing Nonmarket Accounts for the United States (National Research Council, 2005). An initial task for both lines of work is to accurately categorize nominal expenditure estimates into meaningful production units to which prices and quantities can be attached. Even those topical areas for which the statistical agency and research community objectives appear to diverge may do so for only a period. For example, for some time, researchers have investigated the health impact on the population of various disease treatments produced by the medical sector; at the moment, this is not an active area for BEA. However, ultimately, both BEA (NIPAs) and BLS (price index work) should be interested in integrating medical outcome data into their analytic apparatus in order to track changes in the quality of service, which, along with the quantity, determines the value of the industry’s output. The most methodologically advanced components of the NIPAs, and of the various price indexes and productivity measures, already track the changing

quality of goods and services and facilitate the ability to break out nominal price changes into real price and quality change components.

In thinking about next steps of these research programs, we argue that, among the inputs to health, the medical care piece should initially be given priority. For policy purposes, the most pressing need is to measure medical care expenditures and related outputs and outcomes as well as possible. In addition, accurate expenditure and price data on medical care are essential both for developing broader health accounts and for improving the medical care component of the NIPAs. Thus, the panel supports the idea that, at least initially, BEA’s satellite health accounting program should focus on input and output data relevant to the provision of the nation’s medical care.

The satellite account should include all inputs to the production of medical care, whether they are purchased by the medical “industry” or by households and regardless of how directly they affect health. In addition to clear cases of life-saving medical interventions, treatments that improve quality of life (e.g., knee repair), that are preventive in nature (e.g., a routine physical), or even that have ambiguous links to health (e.g., some over-the-counter medications) should be included so long as they are part of the medical care sector. For setting account scope, we deemphasize concerns about the extent to which the procedures improve health (physical or mental). Ideally, however, outcomes research will indicate quantitatively—in an accounting framework that values system output in terms of improvements in quality or quantity of life—which medical services are more and less productive. Arguments in favor of this approach are that it creates a relatively clear line of demarcation and that it maintains historical comparability. Also, for policy, it is important to know expenditure totals for the sector and to have an idea what people are getting for that spending.

In building a broadly useful national health accounting system, the question of how to define the unit of measurement must be tackled. Ideally, medical sector goods and services would be defined in such a way that (1) expenditures could be estimated each period for every good or service produced by the industry, (2) meaningful quantities and prices (nominal and real) could be tracked, and (3) quality change of the goods and services could be monitored. One way to proceed that embodies these three goals is to identify the output of the medical care sector as completed treatments or procedures.

A treatments-based organizing framework coordinates logically with a broader health data system because, in principle, it creates a unit of analysis for which changes in the effectiveness of various medical services may be monitored. It provides a mechanism whereby prices can be adjusted to reflect changing quality, the substitution of inputs can be handled better than they are currently, and the introduction of new treatments can be dealt with on a disease-by-disease

basis. At the present time, however, health expenditure and outcomes data are not organized in a way that feeds naturally into this accounting framework, and so their structure needs to be modified.

Although the logic of a treatment-of-disease approach is clear, it will not be easy to execute in a comprehensive national data system. It requires defining episode groups and then dealing with population heterogeneity within condition categories. A first step is to agree on a crosswalk between existing classification systems and a version that is at the right level of detail for a national health account—one logical option would be an aggregation of International Classification of Diseases-Tenth Revision categories although the final version might be arranged quite differently. Even then, problematic conceptual issues associated with using a treatment-of-disease approach remain, such as those created by the presence of comorbidities. A reasonable starting point is to treat common clusters of comorbidities (e.g., diabetes and vascular complications) as distinct disease categories; in addition, methods exist for allocating expenditures on patients with multiple conditions based on ratios of costs estimated from patients with single diagnoses. A special advisory committee could be convened by BEA to provide guidance on these conceptual and definitional tasks.

Another hurdle to implementing a disease treatment approach is the problem of defining a measurement unit for chronic episodes or any kind of service that involves a long treatment period. In the literature, episode-based price indexes have defined the good as a completed episode (e.g., the price of treating a heart attack). At the end of that episode, expenditures on the treatment are collected and that is the price of the unit. For the purpose of pricing the treatment of chronic illnesses, which may span several statistical reporting periods, the unit of measurement should include all spending on the disease for some designated and consistent period of time, most likely a year.

A large portion of medical care activities, in terms of costs, can be accounted for by looking at treatments of specific diseases. Roehrig et al. (2009) produced estimates of national health spending by medical condition using 260 categories defined in the Agency for Healthcare Research and Quality (AHRQ) Clinical Classification Software, which groups the numerous International Classification of Diseases-Ninth Revision codes into broader categories that are “clinically meaningful.” Reallocating the NHEA totals using data from the Medical Expenditure Panel Survey (MEPS), the research team found that the largest eight expenditure categories (circulatory system, mental disorders, musculoskeletal, injury and poisoning, digestive, neoplasms, respiratory, and nervous system) accounted for almost 70 percent of total expenditures. By contrast, prevention, exams, and dental—which would not fit cleanly into disease-specific categories—accounted for around 6 percent of personal health

spending; another 6 percent of total personal health expenditures from the NHEA went unallocated.

Although disease treatment categories are clearly important, it is obvious that not all medical services (and certainly not all health-related services) can be captured in them. Due to the presence of these other aspects of care, including those that may take place before and after treatments, it would not be desirable to shoehorn all medical care spending (and associated health effects) into disease categories. These areas of care might be better considered separately. We envision that, ultimately, some catch-all categories will need to be created for nondisease-specific health care and also for episodes of management—such as screening, diagnosis, and prevention.

Recommendation 2.11: Although starting with medical care on a disease-by-disease basis is a realistic way to proceed in order to begin accounting for a very significant share of the medical care economy, work should also begin on estimating the costs of, and eventually the health return from, interventions other than treating specific diseases (e.g., management, preventive, diagnostic, screening) and long-term medical services.

The idea would be to track quantities and prices of items in these categories over time, just as one would track direct disease treatment. In some cases, it may be possible to link diagnostic tests and the other forms of care backward from an eventual resulting treatment. The task of setting up the nondisease treatment categories could again be assisted by an advisory committee convened under the auspices of some combination of CMS, AHRQ, and NCHS, although one office will need to take the lead.

The first major task in developing a national health account—whether it is BEA’s satellite version or the broader health type—is to devise a method for allocating economy-wide spending on medical care into the treatment-of-disease categories described above. Because they serve as building blocks for many kinds of health data systems, improving the methodology for organizing and tracking health care expenditures is an immediate priority.³

Given the number of different disease classification schemas currently in use in the U.S. health care system, it is essential for all players, in both public and private sectors, to participate in and come to a consensus on the development of a single unified version.

Recommendation 3.1: A concerted effort is needed to reach consensus on how to classify diseases and about what the criteria are by which diseases are disaggregated from the very broad International Classification of Diseases chapters. The National Center for Health Statistics should lead the effort, working with the Agency for Healthcare Research and Quality, the Centers for Medicare & Medicaid Services, the Bureau of Economic Analysis, and other relevant statistical agencies. As part of this effort, U.S. agencies should participate in ongoing standardization efforts (such as those sponsored by the Organisation for Economic Co-operation and Development or the World Health Organization) to benefit from international expertise, to consider these as the basis for a national system, and to facilitate international comparisons.

Once an agreed-on classification system has been established, several options exist for attributing spending across treatment episodes for the range of disease categories. One is an encounter-based method in which spending is attributed to one or to several diagnoses as reflected by data extracted from patient claims. A second, broader approach involves constructing episodes of treatment—which may include numerous encounters over a predefined period—then adding up dollars spent nationally on each of the range of diseases and conditions. A third possibility is a person-based approach, in which spending on various treatments is tracked on a person-by-person basis over a set period of time.

BEA, for its part, is working both internally and with outside researchers to establish what the allocations look like under the different methods and whether it matters for estimating aggregate expenditures and prices. These researchers (Rosen and Cutler, 2007) have already begun producing episodes-of-treatment cost estimates for the different methods, demonstrating that it can be done; spending could also be further broken down into additional subcategories such as disease prevention, diagnosis, and screening activities. Whichever method of allocating expenditures is chosen, it has to offer a solution to the comorbidity problem.

At this point, the panel cannot definitively recommend one method for tracking expenditures and output (and eventually outcomes) associated with medical treatments over all others. And the appropriate concept depends a great deal on the specific application. For example, if the goal is to compare costs and health effects for a given disease, as is done in cost-effectiveness analyses, a person-based approach is likely to be most appropriate. In contrast, if price index construction is the goal, federal agencies may find an episode-of-treatment approach more meaningful. For managers of a health plan trying to understand why emergency room spending patterns have changed, real-time answers may be possible only with an encounter-based approach. What is needed is more empirical work to compare different approaches and to determine more definitively which is best under different conditions.

Recommendation 3.3: The Bureau of Economic Analysis (working with academic researchers and with the Bureau of Labor Statistics) should continue to investigate the impact of different expenditure allocation approaches— particularly the episode- and person-based methods—on price index construction and performance. Research is needed to determine which method is best under different circumstances.

As part of this effort, BEA (perhaps in coordination with AHRQ and NIA) should sponsor a workshop during which private-sector vendors that produce various disease-grouping systems could present their products, identify how they are used in the marketplace, and describe the underlying rules and logic.

At this point, a cautious approach is warranted, as it is too early for BEA to buy into a particular method for aggregating treatment costs. This means that there may need to be parallel sets of accounts going on, at least on a research basis, for some time. The statistical agencies should continue to experiment with competing methods, and research should be designed to test variation in results from different data sources.

Recommendation 3.4: The Bureau of Economic Analysis, working with academic researchers (and perhaps other agencies, such as the Centers for Medicare & Medicaid Services and other parts of the Department of Health and Human Services), should collaborate on work to move incrementally toward the goal of creating disease-based expenditure accounts by attempting a “proof of concept” prototype. Using a subgroup of the population with good data coverage, the prototype would attempt to demonstrate that dollars spent in the economy on medical care can be allocated into disease categories in a fashion that yields meaningful information.

The Medicare population, the military, or veterans—groups for which there is available spending and health data (and for whom a lot of the medical care action takes place)—would be logical choices. Alternatively, a disease-costing pilot could be done for a well-defined, geographically (and administratively) complete group, such as found in parts of Intermountain Healthcare, the Geisinger Health Care System, or one of the Hawaiian islands before attempting it on a national basis.

Much of what is required to develop health and health care accounts has to do with medical care price deflation. Once the nominal expenditure flows for the sector have been figured out correctly, a daunting problem in itself, the next task is to begin estimating disease-based price indexes. In this report, we take the

position, expressed by many other health economists, that price indexes organized by treatments of disease or episodes of illness should be used as deflators for the medical sector components of the national accounts.

Health economists have developed the conceptual tools that are needed to construct price indexes for medical care goods and services demanded by consumers. Much of the academic literature has relied on patient claims data to provide a picture of price trends for treating specific conditions (e.g., Berndt, Busch, and Frank, 1998). In these studies, the goods and services have been defined as a completed episode; this parallels the conceptual approach described above on how to categorize nominal expenditures. For example, for a heart attack patient, this may involve time and expenditures on a series of initial treatments plus those that take place during the recovery period. At the end of that episode, data are collected to estimate all dollars spent over the entire period; this forms the basis for pricing a completed episode.

Under optimal conditions, a price index will embed a capability to pick up the substitution of inputs that takes place over time. For its satellite medical care account, BEA proposes to take the system-wide spending over some period of time on treatments of each condition (such as for depression), regardless of treatment mode (e.g., drug versus talk therapy), and divide it by the number of patients treated. The idea is to calculate a unit value that includes all spending and then allows for substitution across treatment types as well as for the introduction of new drugs or other therapies.

A number of examples can be cited from the literature of how spending by traditionally defined medical care industries is handled to form the price of a specific treatment. Pricing of cataract treatment is one of the most prominent of these (Shapiro, Shapiro, and Wilcox, 2001). If price indexes track industries in a way that segregates by type of service facility, then surgeries taking place in a hospital are sampled to generate one price index, whereas surgeries taking place in clinics are sampled in the estimation of another. A traditional industry-based index of this sort will miss the price effect that accompanies a scenario in which people switch from the more expensive hospital surgery to a less expensive ambulatory care facility or an office. Assuming quality remains constant (i.e., treatment outcomes are comparable), then a direct comparison price index capable of capturing the facility shift would more accurately estimate the decrease in price.

Even though the BLS Producer Price Index (PPI) for hospitals has been available with a cost-of-disease classification for more than 15 years, BEA does not deflate (adjust for price inflation) at the disease treatment level. Expenditure data are not available by cost of disease, so there is nothing for BEA to deflate. For that reason, BEA uses the aggregate PPI for hospitals as the deflator. Clearly, the units of analysis should be the same for price and expenditure data used in a medical care account (or any economic account for that matter). In the future, this problem could be resolved by using new data from the Census Bureau and from the PPI, and the U.S. statistical system will have gone a long way toward

providing expenditures and price indexes for medical care that are harmonized around a cost-of-disease framework.

Recommendation 4.3: The Census Bureau should give high priority to providing annual data for hospitals and other medical care industries, grouped by a cost-of-disease system that matches the one used in the 2007 Economic Census and in the Producer Price Index. The panel notes the welcome provision of new data on receipts categorized by disease that accompany publication of the 2007 Economic Census, though a considerable amount of work remains to be done to bring the quality of these data up to the standards needed for use in official statistics.

As a first step to improving price measurement for medical care, it is useful to account for the reduction in costs resulting from the substitution of inputs in treatments. However, in order to tell the full story, we must turn to the (probably more quantitatively significant) question, “Is the new treatment technology better, worse, or the same as the old one?” Although BEA’s plan is to defer some aspects of price work until a greater consensus about methods emerges, it is essential to think about quality change from the beginning. Improvement in medical procedures creates a major measurement issue, and any price index that does not confront it will ultimately be less than satisfactory. BEA is aware of the importance of quality adjustment but, at the moment, reasonably claims to not have a systematic way of dealing with it for medical care. Perhaps, however, there are some cases that can be addressed on the basis of the existing literature and that BEA could get started on sooner rather than later.

Thinking seriously about how to measure changes in the quality and, in turn, the real cost of medical care requires the monitoring of information about outcomes associated with that care. These efforts will require expertise from medical researchers. Information about medical care outcomes and the changing quality of treatments is also essential for balanced policy discussions about resource allocation, which have focused nearly exclusively on the cost side while neglecting the changes in value that accrue as a result of the expenditures. As an example, the price of treating heart attack patients has increased dramatically, but so too has the desirability of outcomes. Fifty years ago, heart attack patients were told to rest—a practically free treatment for cases in which that rest was done at home. Treatments now exist that, while much more expensive, most would consider preferable because they have greatly extended life expectancy (rest is now thought to be counterproductive). When measuring prices of treating heart attack patients, the changes in the efficacy of the treatment must be monitored along with its cost. This kind of quality adjustment is a well-established component of accurate price measurement.

Collecting outcomes data that can be linked to spending is a big challenge, but it is not impossible. Although data are already quite good for a few specific diseases, the reality is that this research will have to draw from a patchwork of data sources, public and private. Pharmaceutical companies collect extensive data on the comparative effectiveness of new technologies versus old ones in order to make the case for reimbursement. Many of these data are collected as part of clinical trials that, while generally not done as head-to-head comparisons of alternative therapies, provide evidence about quality changes occurring in medical care. The current focus on “comparative effectiveness research” at AHRQ and other agencies is also producing further evidence on this front. Data on outcomes are going to continue to improve over time—ideally, BLS, BEA, and academic researchers would be able to take advantage of the millions of dollars spent collecting these data. Integrating emerging information from the research community is an essential enhancement that can be made to the data infrastructure.

Recommendation 6.7: Significant analytic value could be added to Medicare data compiled by the Centers for Medicare & Medicaid Services if more clinical information on outcomes and patient characteristics were included. It is possible that the addition of only a few clinical items to claim forms could greatly enhance performance measurement. The National Committee on Vital and Health Statistics, an advisory body to the secretary of health and human services, could provide guidance on this. The linking of tests and procedures to diagnoses would also be useful for confronting the comorbidity problem. Even though adding lines to claim forms involves major political hurdles, the panel recommends that it be done.

Serving as an example (and providing encouragement that the task is feasible), three decades ago, RAND developed an experimental claim form that included clinical information (Newhouse and the Insurance Experiment Group, 1993), and it proved effective in defining patient episodes.

In developing a national data system designed to track current population health and coordinate information on its determinants, the novel challenge is measuring health. In order to answer the question “What are people getting for their health care dollar?” it is necessary (though not sufficient) to be able to track changes in the health of the population accurately. Furthermore, the conceptual framework requires monitoring changes in the population’s health status using metrics that include both mortality and morbidity effects. Although such measures are already widely used for identifying unmet or underaddressed health needs, there is no undisputed definition of or method for measuring population health (Kindig and Stoddart, 2003; McDowell, Spasoff, and Kristjansson, 2004). How-

ever, in cost-effectiveness analysis of medical interventions, it is now standard to combine impacts on survival and the health of survivors into quality-adjusted life years (QALYs) (Gold, 1996).

Historically, population health has been summarized by life expectancy at various ages and other statistics based solely on deaths. Recently, however, a number of summary measures of health-related quality of life (HRQoL) have been collected in population data sets in the United States. These measures can be used to adjust life expectancy for quality to obtain a more complete measure of health. As with ordinary life expectancy, quality-adjusted life expectancy (QALE) measures do not predict future health but instead summarize health in the current year. Two inputs are needed: (1) a life table describing mortality experience in the population, and (2) the average HRQoL per year of age in a population. Age-specific death rates from NCHS would be needed as the first input. Data from population surveys using HRQoL indexes would suffice for the second. Given that multiple surveys (each with a somewhat different population scope) are required to adequately measure health across the population broadly, the statistical agencies should agree on a common quality of life instrument to use in the different surveys.

Recommendation 5.1: A committee of members from agencies responsible for collecting population health data (Agency for Healthcare Research and Quality, National Center for Health Statistics, Census Bureau, etc.) should be charged with identifying and putting in place a single standard population health measurement tool (or set of tools) to use in a wide range of surveys. The best instrument, which is situation specific, may simply be the one that can be added to enough surveys collected over time so that most of the population is covered.

Ideally, agencies would collaborate and choose at least one instrument that can be followed over time for the purpose of having a comparable measure across years. For example, it would be useful if a generic quality of life measure were added to the Medicare Current Beneficiary Survey (MCBS). MEPS and MCBS should have at least one instrument in common that both will use consistently over time. The advantage of standardization is that it avoids problematic comparisons across instruments.

All things considered, the panel thinks that QALE is the most appropriate metric for summarizing population health for a national health account. Although life expectancy is currently the most widely used metric for this purpose in the United States and other developed countries, the underlying methods needed to move to QALE have become the subject of extensive scientific investigation (and were developed in particular for the health questions typically asked in ongoing surveys).

Economic accounts are in the business of organizing data for measuring the value of goods and services generated in the economy, so it makes sense to

think in terms of QALYs, which are a measure of the good that is being generated from medical care and other health-affecting activities. And, presumably, data should be of sufficient quality to estimate QALE in the United States using QALY methods.

Recommendation 5.3: Initially, the national health account should focus on quality-adjusted life expectancy measured in quality-adjusted life years as the best summary measure of health in each year.

Throughout this report, we focus on methods for identifying, quantifying, and valuing inputs to health—beginning with treatments of specific medical conditions—and for measuring changes in the population’s health. We cannot predict the accuracy of future efforts to quantify the causal links between medical care, health-enhancing activities, and other inputs to changes in the population’s health. However, we can say that a major issue driving this area of research will be how to gauge the productivity of the medical care system. For this purpose, it is important that investigators have data that allow them to attribute population health effects to factors that work separately or interactively with medical care. It is advisable to begin with medical care, but the elements included in a broader boundary of health goods, services, and activities become more important when trying to determine causality for outcomes.

For example, data on important risk factors that impact future health should also be collected. Risk factors are characteristics of individuals or individual behaviors that are associated with future changes in population health. A comprehensive list of these is beyond the scope of this report. However, an accounting of health gains from current medical investments should recognize that reductions in risk factors in the present can improve QALYs in the future. We suggest several criteria for choosing among which, if any, data on inputs to future health to collect: (1) there should be feasible or actual reasonable national estimates available, (2) solid scientific links should exist between changes in the risk factor and future health, (3) the impact on future health should be big enough to make data collection worthwhile, and (4) interventions to change the risk factor should be reflected in the expenditure side of the national health account. Many of the relevant variables are already collected regularly through the CDC’s Behavioral Risk Factor Surveillance System.

Integrating health expenditure and health outcome data is a long-term project that has only begun. It will require time-consuming disease modeling—estimating interaction between risk factors, specific diseases, and health outcomes. This kind of scientific endeavor is beyond the scope of traditional (or even satellite) accounting. A full accounting of nonmedical factors affecting health seems futuristic at

this point because only scant data exist that would allow modeling of the joint distribution of environmental and other factors. Nonetheless, researchers would benefit if the national health accounts program began accumulating data on time use in health-improving preventive activities, consumption trends and other risk factors, behavioral trends, and the environment in a centralized and accessible way.

In constructing a data infrastructure to underlie the national health and health care accounts, the ultimate objective is to maintain a set of minimum common data on the entire population—or as close to it as possible. The United States has a decentralized statistical system, and it is a complex task to determine the best roles for AHRQ, CMS, NCHS/CDC, and other agencies developing a coordinated health data system. The national health account envisioned by the panel would rest on a foundation linking information across many data sources—government/private, administrative/survey, national/local. Efforts are successfully under way to combine data sources (see Rosen and Cutler, 2007), showing that it can be done. For example, because NHEA data are reported as aggregate spending by payer and service category but not at the disease level, the Cutler/Rosen project team has linked them to expenditure surveys—MEPS and its precursor survey, the National Medical Expenditures Survey, both collected by AHRQ—in order to gain the level of detail needed on several aspects of health (including quality of life and the presence of diseases) and health care utilization. It is also important to consider how nonstatistical agency data (military, veterans affairs, etc.) as well as nongovernment data (e.g., those maintained by health plans and drug companies) can be exploited.

One important purpose of this report is to begin identifying key data sources and to provide an indication of how they can be combined or better organized to add value to the national data infrastructure needed to move forward in a cost-effective manner; this report does so through a number of specifically targeted recommendations. Because surveys are expensive and burdensome, cost efficiency requires linking survey data to administrative records whenever possible. One potentially rich source of case-by-case information on expenditures and treatments is patient claims data. Medicare and Medicaid would provide coverage of large portions of the population. If claim forms included the “right” information, they would go a long way in satisfying some health accounting data needs. A lot could be done with claims data for the insured population because of their enormous size and wide coverage.

For constructing a core data set to underlie national health and health care accounts, it is reasonable to use MEPS supplemented by claims information wherever population or disease coverage gaps appear. Survey and claims data

are complementary because of the trade-off that exists between sample size and representativeness. Survey data are essential to the accounting project because of the detailed patient information they can provide. However, their sample sizes are adequate only for high-prevalence conditions such as cardiovascular disease and risk factors. In contrast, insurance claims data provide a large sample, but at the expense of representativeness—no single source provides a national sample.