Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 117
Beyond the Market: Designing Nonmarket Accounts for the United States 6 Health Health has improved dramatically since the industrial revolution. This improvement has resulted from changes in work, life-style, and environmental factors; from advances in public health and medical care; and from increases in wealth, which allow individuals to spend more on goods such as food and shelter that contribute to health. Nordhaus (2003, p. 9) concludes that “accounting for improvements in the health status of the population would make a substantial difference to our measures of economic welfare over the twentieth century in the United States.” This chapter considers the potential for measuring the population’s health, and changes in health, within an accounting framework. A population’s health is reflected by the average length of its members’ lives, and by the quality of those years, as affected by incidence and severity of disease. In contrast to medical care, health itself cannot be purchased directly and, therefore, is not measured in the national income and product accounts (NIPAs). Moreover, with health, there is nothing close to a market equivalent to help us answer valuation questions, so one must turn to other methods. A fully developed health account would enable researchers to estimate the effect of income and the flow of many other inputs on the stock of health and the value of changes in it. Measuring health is also an important prerequisite for better estimating productivity in medical care. Through these mechanisms, health accounts would facilitate more credible assessment of the desirability of policies that affect provision of health-related goods and services. Medical care involves substantial expenditures in markets and, to that extent, it is measured in the NIPAs. Where the national accounts have particular diffi-
OCR for page 118
Beyond the Market: Designing Nonmarket Accounts for the United States culty is in decomposing medical spending increases into price and real output. To put it another way, it can be difficult to differentiate among price, quantity, and quality changes. For example, a change 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 (e.g., survival rates, patient quality of life), or simultaneous changes in more than one of these factors. To measure changes in real output, changing quality must be taken into account. For indexes constructed from quantity data, quality adjustments can be made directly. If a new drug is introduced that is equally effective as the old one but only has to be taken once a week instead of every day, the quality (and, in a real sense, the per unit quantity) has increased. If the account is built from price and expenditure data, as in most NIPA components, quality adjustment is made by deflating prices. In the drug example, if the observed unit price increases by anything less than seven-fold, the real price (as dictated by how much has to be spent on the drug per week) actually decreases. Whether a direct quantity-based index or a price-deflated quantity index can be constructed depends on data availability. This quality measurement problem is prevalent in many industries, not just medical care. When a car costs more because the brakes are better, or when airline tickets cost less because the quality of service is poorer, the change in cost should be attributed to the change in real output, not a change in price. For some manufactured goods, it is possible to estimate the portion of an observed price change attributable to a change in quality in a relatively straightforward way—using direct measurement or hedonic adjustment techniques. Indeed, the Bureau of Labor Statistics (BLS) does just this for a number of goods in construction of its price indexes.1 The hedonic approach can be problematic for the medical care case. Because most people have health insurance, and because patients are not completely in charge of what they buy, it is not clear that observed prices reflect willingness to pay for particular medical sevices. Since there is not as yet a direct measure of health, or of the contribution of medical care to health, BLS is not able to estimate the productivity of medical care in the usual way. As a result, it is widely believed that some of what is captured as growth in the price of medical care in fact reflects improvements in the quality of care, and that medical care inflation is therefore overstated (see Berndt et al., 2000; Boskin et al., 1996; U.S. Bureau of Labor Statistics, 1997; Shapiro and Wilcox, 1996). Better estimates 1 Hedonic models are regression equations designed to capture the relationship between a good’s characteristics and its price. The estimated coefficients are used to decompose observed changes in price into a portion attributable to changes in items’ characteristics and a portion that represents true price change. See National Research Council (2002a: Ch. 4) for a discussion of hedonic methods and National Research Council (2002a: Ch. 6) for recommendations specifically on the pricing of medical care.
OCR for page 119
Beyond the Market: Designing Nonmarket Accounts for the United States of quality change and health output will enable construction of improved measures of medical care productivity.2 Finally, health is important to measure because it is an input to other outputs. Healthier people earn more than less healthy people and stay in the labor force longer (see Costa, 2003; Case et al., 2002). In understanding the sources of growth in national income, we need to account for changes in health. In all these ways, a more systematic monitoring of the population’s health states, and the factors that have an impact on those states, can advance research and policy. Recommendation 6.1: A health satellite account should be produced by the Bureau of Economic Analysis in collaboration with the Centers for Medicare and Medicaid Services of the U.S. Department of Health and Human Services. In this chapter we discuss a framework for developing health accounts and highlight the central choices to be confronted and the work to be done before such accounts can become a reality. While answers are not provided for every possible question, we illustrate what choices are possible and how empirical evidence can inform those choices. In the sections below, we discuss some conceptual issues underlying the idea of health accounts and describe the input and output components of a health account. In the process, we attempt to lay out the major research areas that need to be developed to inform production of such accounts. CONCEPTUAL FRAMEWORK As with several of the other areas of nonmarket activity we have examined, we recommend the formation of a satellite account for health—one that does not replace the current NIPAs or components therein, but exists alongside them. In addition to augmenting the picture provided by the NIPAs, development of a satellite health account could uncover approaches and encourage the development of data that improve the way medical care is measured in the conventional accounts. As discussed earlier in this report, the key to national income accounting is the delineation of inputs and outputs. The total value of output reflects the dollar value of sales on the production side and the dollar value of payments to factors of production on the income side. Since nearly all transactions captured in the NIPAs are market transactions, prices and quantities exist for most items. Money paid by one person to purchase a good or service becomes realized as income to another so, aside from issues relating to measurement of the return capital, the total value of output should be equal to the total value of the inputs used in its production. 2 BLS, adopting recommendations from the National Research Council (2002a), is pursuing new experimental health care price indexes. For a full description of how BLS calculates the medical care component of the Consumer Price Index (CPI), see http://www.bls.gov/cpi/cpifact4.htm.
OCR for page 120
Beyond the Market: Designing Nonmarket Accounts for the United States Because one cannot buy health in a market and because many health science innovations are non-rival in character, there is no guarantee that the value of incremental improvements in health will be equal to their cost. The value of health output can be greater (or less) than the sum of the input values; in a growth context, the change in the value of output can differ from the change in the value of the inputs. To illustrate, imagine that people would be willing to pay $100,000 (in present value) to live a year longer than they currently expect to live. With existing knowledge, no one knows how to guarantee that outcome; hence, their desires go unfulfilled. Now suppose that a new drug is developed that guarantees 1 (and only 1) additional year of life, at a cost of only $1,000 per person, including research and development, manufacturing, and distribution. After the drug is developed, a measure of output based on the cost of the inputs used in its production will increase by much less than the output as valued by consumers. Since consumers are constrained from buying more (years of life), the marginal value of health improvement may well exceed the cost of attaining it.3 The possibility of this sort of discrepancy is one of the fundamental reasons for constructing health accounts. A traditional type of measure—such as gross domestic product (GDP)—would include only the money spent improving health. In our example, the contribution of the health sector to GDP would increase by $1,000 per person. In the “knowledge sector,” however, at small cost, something was learned that had a large value. If health knowledge is included in the health account, the correct inference is that the health sector has become more productive. Measuring the output “health” and the input of dollars (and time) spent to improve health, including dollars spent on research and development, allows calculation of how much more productive. Market-Oriented Approaches The market-transacted goods and services relating to medical care appear in various household consumption, government, and investment components of GDP. Examples are direct medical care spending, public medical care, and pharmaceutical research and development. Since 1964, the Department of Health and Human Services has published National Health Accounts that include time-series data on national health expenditures. The health accounts seek to “identify all goods and services that can be characterized as relating to health care in the nation, and determine the amount of money used for the purchase of these goods and services …” (Rice et al., 1982). 3 The point here is that the limited (and discrete) effectiveness of the pill eliminates the possibility of purchasing pills up to the point where the “the marginal value of health” equals the price of the pill. If additional pills could add additional years of life at the same cost, consumers might buy many more, until the value of the additional year of life was as low as $1,000. In this case, valuing the pills at their $1,000 cost would be as good as similar valuations for shirts or shoes.
OCR for page 121
Beyond the Market: Designing Nonmarket Accounts for the United States Medical care and health research expenditures account for a large and growing percentage of GDP (14.9 percent in 2002, up from about 9 percent in 1981 and from just over 5 percent in 1961). Table 6-1 shows the sources and uses of those expenditures. Note that the data are, by and large, organized by payer group or by type of institution making or receiving payments. From a national accounts perspective, the essential problem with health expenditure data, as currently collected, is that they do not provide price and quantity information about anything that might reasonably be considered an output. The expenditure data answer only the questions, “Where does the money come from?” and “Whom does it go to?” They do not answer the central question, “What does it buy?” (Triplett, 2001, p. 3). In this sense, the current medical care accounts are undeveloped, even by the market-oriented NIPA standards. What is the output of medical care? The Eurostat Handbook (Triplett, 2001) identifies “completed treatments” as the output of the health care sector of the economy. Ultimately, what people want from medical care is improved health, and one way they pursue this is through treatment of ailments and diseases. Doctors’ time, hospital patient days, drugs—things typically treated as outputs—are more accurately viewed as inputs used in the production of treatments. In practice, one would measure this at the disease level—how much has health improved for people with a particular disease. Though research on cost-of-disease or cost-of-treatment approaches is embryonic, prices of these service sets could TABLE 6-1 The Nation’s Health Dollar, 2002 (percent of expenditures) Sources Uses Private insurance 35 Hospital care 31 Medicare 17 Physician and clinical services 22 Medicaid and SCHIPa 16 Prescription drugs 11 Out-of-pocket 14 Nursing home care 7 Other publicb 12 Program administration and net cost 7 Other privatec 5 Other spendingd 22 aState Children’s Health Insurance Program. bIncludes programs such as workers’ compensation, public health activity, Department of Defense, Department of Veterans Affairs, Indian Health Service, and state and local hospital subsidies and school health programs. cIncludes industrial in-plant, privately funded construction, and nonpatient revenues, including philanthropy. dIncludes dental services, other professional services, home health care, durable medical products, over-the-counter medicines and sundries, public health activities, research, and construction. SOURCE: Centers for Medicare and Medicaid Services, Office of the Actuary, National Health Statistics Group. The website for the Department of Health and Human Services has detailed tables of national health expenditures, by source of funds and type and amount of expenditure (http://www.cms.hhs.gov/statistics/nhe/historical/chart.asp).
OCR for page 122
Beyond the Market: Designing Nonmarket Accounts for the United States in principle be tracked by type, then quality adjusted and weighted appropriately (Triplett, 2001, p. 1). The current lack of pricing by treatment is, to some extent, a data collection issue. As shown in Table 6-1, data are currently collected by institution and not by treatment. Quality adjustment at the treatment level is another difficult issue. Yet research by Cutler et al. (1998), Shapiro and Wilcox (1996), Triplett (1999), and others has been moving in this direction over the past decade, and solutions to difficult productivity and pricing issues are emerging. The treatment-based cost-of-disease approach is more consistent with the way the rest of the market-oriented NIPAs are designed: “if the medical care sector is the industry, then the treatments are the products the industry produces” (Triplett, 2001, pp. 3-4). While recognition is growing that treatments and outcomes—not time with the doctor, days in the hospital, and so on—are the conceptually relevant units, measurement of the appropriate quantities and prices is, at this point, very incomplete. The Aging-Related Diseases (ARD) Study of the Organisation of Economic Co-operation and Development (OECD) has begun cataloging information on the cost of treatments in a few areas such as heart disease, stroke, and breast cancer (Triplett, 2002). Research on cost-of-disease accounts is progressing in several countries, including the United States, Canada, the United Kingdom, Australia, and the Netherlands. This kind of work begins with a reorganization of data from the national economic accounts into expenditures by international disease classification codes. Much of this work in the United States is being done by the National Center for Health Statistics (NCHS). A Broader Approach Research on disease and treatment-cost frameworks has great potential to improve the usefulness of national health care accounts. For a health (not health care) account that would fit into a set of satellite nonmarket accounts, however, one would want to go much further. Specifically, in accord with the rest of this report, we advocate an account that (1) includes both market and nonmarket inputs and outputs, (2) estimates input and output values independently, and (3) defines outputs that are linked to utility as directly as possible. While the treatment-based approach described above redefines medical sector output in a way that is appropriate for the NIPAs, it does not necessarily advance the objective of measuring inputs and outputs independently. For example, the technique for pricing heart disease treatment may simply sum up the costs of inputs, such as those associated with angioplasty, hospital time, and pharmaceuticals. Most of the data in the ARD Study capture input costs since prices are generally not charged directly on the basis of complete medical treatments. In a prospective payment system, one might be able to obtain reasonably good cost estimates for most one-time procedures that do not result in complications (e.g., a bypass operation, or the delivery of a baby), but one does not get a quote for,
OCR for page 123
Beyond the Market: Designing Nonmarket Accounts for the United States say, full treatment of heart disease in the way one might for repairing a broken transmission on an automobile (Triplett, 2002, p. 9). More important is the question of how output is defined: utility is generated by improved health, which is a nonmarket good, not by the treatment itself. Unlike, say, a vacation or trip to a restaurant, most people do not subject themselves to medical treatment for the fun of it—quite the contrary. And medical care is not the only, and possibly not even the largest, input to the production of health. There are in fact a large number of factors—both medical and nonmedical, and market and nonmarket—that influence health. In this context, health condition is the output and medical treatment is one of many inputs. It is this concept of health that we are primarily concerned with for nonmarket accounts. The conceptual framework for this strain of health accounting has its antecedents in the literature on the determinants of health and life expectancy. Just as national health expenditures have changed, so too has the nation’s health, perhaps even more dramatically (see Nordhaus, 2003; Murphy and Topel, 2003). By most measures, improvements in the health of the population have outpaced the increase in spending on medical care; since improvements in health are related to a broad range of interrelated factors, however, one cannot say for certain what factors are most important (Cutler and Richardson, 1997; Cutler, 2004). The literature—which includes McKeown (1976), Fogel, (1986, 2004), Preston (1993), Riley (2001) and many others—attempts to attribute health improvements, typically as measured by life expectancy, to determinants such as medical care, diet and nutrition, and environmental and life-style factors. McKeown (1976), for example, demonstrated that, in England and Wales, declines in mortality between 1900 and 1970 resulted primarily from reductions in the incidence of infectious diseases such as tuberculosis and smallpox. Furthermore, the bulk of this health improvement occurred prior to the development of effective medical therapies. He attributes much of the reduction in disease prevalence to improvements in basic living standards, particularly nutrition. Fogel also focused on diet, using height and body mass as indicators of nutritional status. His research suggests that nutrition played an especially large role during the eighteenth and nineteenth centuries, explaining perhaps 90 percent of the decline in mortality for England and France in that period. For the United States, where diets had significantly improved by the turn of the century, Preston and Haines (1991) argues that improvements in public health—ranging from better hygiene to swamp drainage to mosquito control—were central to reducing waterborne diseases, which reduced mortality rates both directly and indirectly through nutritional consequences. Cutler (2004) argues that major advances in medicine constitute the dominant explanation for improved health in the United States since the mid-twentieth century. Indeed, a primary reason for developing an experimental health account is to add rigor to research aimed at attributing improvements in health to specific types of expenditures, behavioral changes, and increases in basic medical science
OCR for page 124
Beyond the Market: Designing Nonmarket Accounts for the United States TABLE 6-2 Health Inputs and Outputs, Market and Nonmarket Inputs Outputs Medical care Health status Market labor/capital Longevity Volunteer labor Quality of life Time invested in individual’s own health Other consumption items Research and development Quality of environment knowledge. Statistics on causes of death are extremely helpful to policy; consider, for example the success of campaigns to increase seat belt use. Still, mortality rates alone present an incomplete picture of a population’s health, and the analyses noted above focus on very long-term trends. An ideal, and more sensitive, health metric—and one that would be most useful to health policy—would incorporate both fatal and nonfatal health outcomes. For example, in planning how to target public health expenditures, it would be extremely helpful to know with accuracy the extent to which smoking, obesity, and other factors contribute to disease burden.4 It is this concern that has led to development of measures of health status such as quality-adjusted life years (QALYs). We return to these related measures in our discussion of outputs below. Forming nonmarket health accounts begins with identifying the key inputs and outputs. One possible structure is suggested in Table 6-2. Many of the inputs and outputs of the health sector are stocks. An early developer of the framework for thinking about health in this way was Michael Grossman (1972). His model portrays health as “a durable capital stock that yields an output of healthy time…. Individuals inherit an initial amount of this stock that depreciates with age and can be increased by investment” (p. 1). This interpretation is most obvious with health, but it also applies to other nonmarket areas—the stock of research and development and the quality of the environment, for example. And the value of these stocks may be quite large: as we discuss below, health capital has been estimated to be roughly $7 million per person. In practice, though, we frequently want to measure changes in health capital, rather than the level. For example, to determine the productivity of the medical system, one would want to know how much is spent on medical care in a year and compare that to the value of the improvement in health resulting from that intervention. For this question, measuring the endowment of health is less important 4 The major ideas underlying burden-of-disease measures can be found in Murray and Lopez (1996).
OCR for page 125
Beyond the Market: Designing Nonmarket Accounts for the United States than determining how health changes. This is fortunate because empirical work generally has estimated the value of incremental health improvement, not the value of the health endowment itself. We envision a data system designed to monitor changes in health—as measured by life expectancy, disease prevalence, and activity impairments—and as many of the determinants of that change as possible. Measuring productivity in the health sector requires keeping track of the flow of inputs and isolating the contribution of each to better health. If a policy change extends a treatment using existing medical therapies to more people in an afflicted group, we would like to know how that policy affects health outcomes. It is essential, within this framework, to track separately the various inputs to health. When the relationships among the health inputs, their interactions, and the health output are not fully understood, it makes little sense to try to aggregate the inputs. A satellite health account should, at least initially, be oriented toward providing data that could be used to estimate productivity at the disease level. Clearly, identifying the effect of medical care on health, separate from the effects of other inputs, is difficult. But the problem does not need to be solved in order to begin work to construct a health account. The immediate goal is to devise an approach to measuring health and the various inputs to health, leaving questions of causal relationships to others. Because policy demands more complete information about health sector productivity, however, the need for solutions to these questions is the primary reason for pursuing the health accounts. The long-range goal is to be able to identify the influence of such inputs as diet, treatments, and life-style on health status, at least on a disease-by-disease basis. As research reveals new information about these relationships, the value of the data collected in a health account will compound. MEASURING AND VALUING INPUTS There are at least six major inputs to the production of health: medical care provided in market settings; medical care services provided without payment; time that individuals invest in their own health; consumption of nonmedical goods and services, some of which may improve health and others of which are harmful, and nonmedical technology and safety devices; research and development that may lead to improvements in medical technology and knowledge, and environmental and “disease state” factors (and shocks). The first major input—medical care provided in market settings—is readily measurable. When sick, people may visit doctors, undergo tests, and take medica-
OCR for page 126
Beyond the Market: Designing Nonmarket Accounts for the United States tions. The national income accounts already measure the monetary flows associated with payments for these goods and services. As noted, however, there are a variety of other inputs to health that also should be measured. Recommendation 6.2: Market inputs to health such as expenditures on medical care—already measured in the NIPAs—should be included in the health satellite account. The account should go further, however, and measure both the quantity and quality of medical care and include nonmarket inputs relating to time, diet, exercise, and other factors. The second major health input—care service provided without payment—is missed in the national accounts. There are two types of noncompensated care services—volunteer services and services provided by family members. Many hospitals utilize volunteer labor, for example, to perform tasks ranging from fund raising to providing ambulance service. The more important category of noncompensated services is that of family members who provide care for sick or injured relatives. For an elderly person with infirmities who is cared for by a spouse or child, no monetary transaction is involved, and so the services are not reflected in the national income accounts. This exclusion is particularly important in the case of long-term care services. Recent estimates suggest that the value of nonmonetary long-term care services may be even greater than the value of market-provided services. LaPlante et al., (2002)—using data from the National Health Interview Survey on Disability, a nationally representative household survey conducted between 1994 and 1997—show that more people receive unpaid personal assistance services than paid services, and that the average weekly amount of unpaid help per noninstitutionalized adult is also higher. It is easy to see the relevance of measuring unpaid care time to health care policy. For example, apparent cost savings associated with recent reductions in the length of hospital stays have been partly offset by an increased nonmarket burden on families who have to care for patients who are discharged “quicker and sicker” (Pamuk et al., 1998). In such a case, policy makers should be aware of both market and nonmarket costs. Until recently, the data available to measure volunteer and family time devoted to health care activities have been limited. For the past several years, a supplement to the Current Population Survey has collected information on volunteerism, which it defines as persons who do unpaid work for or through an organization. There are real questions, however, about how accurately people can report time devoted to such activities over a period as long as a year and the coding of the type of volunteer work performed is very broad. The information on time use being collected in the American Time Use Survey (ATUS) should be both more accurate and more detailed. Many of the relevant activities will be reported within the “caring for and helping household members” top-level category; some entries in the “caring for and helping non-household members” category also will be relevant.
OCR for page 127
Beyond the Market: Designing Nonmarket Accounts for the United States Recommendation 6.3: Unpaid time spent providing health-related services—primarily volunteer work and intrafamily care—should be measured and included in the health satellite account. The amount of time spent in these activities that improve or maintain health should be calculated using data from the new BLS time-use survey and other relevant sources. As with any nonmarket transaction, there is a question about what wage to use in valuing these unpaid services. Consider a middle-aged woman who earns $25 per hour in the labor force and also spends time caring for elderly parents. Should the implicit price of the caring services be $25 an hour (or some function thereof), or $10 per hour, the amount that such services might (hypothetically) cost in the market? Our general answer, as discussed earlier in this report, is to value time devoted to activities that someone else could have been hired to perform at the market replacement cost. The example given earlier was that time spent by a homeowner putting a roof on the house should be valued at a roofer’s wage, adjusted for the relative productivity of the homeowner as compared to the professional roofer. For a person taking care of elderly parents, the market alternative is to hire a caretaker at the prevailing wage for that occupation. Recommendation 6.4: Unpaid time spent providing health-related services should be valued based on a replacement labor cost approach, adjusted for productivity and quality differences between the services provided by the market and the unpaid provider. In other applications discussed in this report, there is reason to think of the replacement wage as an upper bound for the purpose of valuing nonmarket time, since market service providers often offer more specialized skills than those possessed by a home producer. This may not be the case, however, for health care, elder care, or child care, where family members may offer the more nurturing or caring option. It is interesting that, in many cases, individuals choose to care for family members even when their market wage is above that paid to hired caregivers—that is, even when the market value of their time input is less than they could have earned by devoting the time to labor-market activity. As in earlier applications, we interpret the difference between the individual’s market wage and the replacement cost of the caregiver services provided to a family member as an estimate of the consumption value received from supplying the service personally rather than through the market. We do not recommend, however, that this consumption value be included in the output column of the health account. The satisfaction associated with caring for a family member does not contribute directly to the production of health, which is what the health account is trying to measure. The third major input to the production of health is time that individuals invest in their own health. People exercise to prevent heart disease, sleep to refresh their bodies, and spend time out of the labor force to recover from illness.
OCR for page 130
Beyond the Market: Designing Nonmarket Accounts for the United States ent intake affect health capital—is an essential component on the input side of a health account. Data on products in the nonmedical technology category—products such as seat belts, air bags, and smoke detectors—should also be tracked. Even though the way in which these factors combine with others to affect a population’s health status is not yet exactly known, information about them will be useful for research on health determinants and medical care productivity. The fifth input to health is technology created by research and development (R&D). Research and development generates information and intellectual capital which later may produce flows of services in the form of medical advances and other innovations. Some R&D occurs in the private sector, generally by pharmaceutical and medical device companies. Other R&D is paid for by the public sector and occurs in national laboratories or at universities. In the case of the former, research costs generally are reflected in the prices of the drugs that are sold. The health account should not value this type of R&D separately when it would double count the market component of the account. The benefits from basic science research in a university or nonprofit setting are more likely to be missed in the market accounts and, hence, to need special attention. There are estimates of the flow amount of R&D that can be entered into the health account. Recommendation 6.6: New technologies created by research and development expenditures should be measured and included in the health accounts. One research area that is not well covered in the NIPAs is the development of non-rival innovations. Much of this work is likely to take place in nonprofit research settings. Initial work on valuing nonmarket R&D might reasonably focus on such organizations. Additionally, in general, the value of research and development should be capitalized to reflect depreciation and obsolescence. The sixth input category includes the various environmental factors that affect health. Air pollution and unclean water harm health, while public parks may improve health. Valuation of the environment is a topic that is dealt with elsewhere in this report and more extensively in Nature’s Numbers (National Research Council, 1999). We thus do not take it up here, other than to say that the environment enters the health production function alongside diet, life-style, and other factors that ultimately could be brought into a health satellite account. In so doing, it should be recognized that only a subset of environmental factors influence health. To take a simple example, air pollution may lead to an increased incidence of asthma and to an obscured view of the Grand Canyon. A health account would be concerned only with the former. More research is needed before the role of these “exogenous” factors—changes in the physical environment or the appearance of new diseases—can be understood. Keeping track of these factors in a set of accounts would be a useful step. While we see the above-described components of a health account as the most important, there are other factors to keep in mind. For example, the demographic make-up of a population will affect measures of “average health.” All
OCR for page 131
Beyond the Market: Designing Nonmarket Accounts for the United States else equal, one would expect a younger population to be classified as more healthy than an older one by most measures. The issue of how to separate pure age effects from changes in the disease environment and other shocks requires further study. Ideally, an account would coordinate data useful to researchers who want to estimate the effect of changing demographics on health levels. As the health account is developed, perhaps a demographics module (such as that discussed in Chapter 2) can be integrated. Not only is the demographic information necessary for estimating a population’s health stock, it also would be needed for estimating per capita effects (analogous to per capita GDP). MEASURING AND VALUING HEALTH Defining the Output Two outputs are associated with investments in health. The first is the flow of better health that results from the investment. The present discounted value of this flow is termed health capital. People enjoy being healthier just as they enjoy consuming better food or nicer clothing.5 The second output of investments in health is the additional income that a healthier population generates. A complete set of health accounts would include the present value of expected future earnings, perhaps augmented to include the value of expected future home production, that results from changed health. This is analogous to the way human capital is valued in the labor economics literature. Recommendation 6.7: In the health satellite account, output should be measured independently of inputs (and the two need not be equal). Changes in both components of output—the consumption flow of good health and the additional (or reduced) income that a healthier (or less healthy) population earns—should be measured. Health events, such as the contraction of a disease, affect the future consumption flow of health. These events, as well as interventions in response to those events, impact individuals’ quality of life across future periods, often in unpredictable ways. Money spent on a bypass surgery operation often will improve quality of life, and possibly reduce future mortality rates. Likewise, present levels of health are linked to past actions—the production of health is an intertemporal process tied to past levels of health care, diet, environmental and other factors (Triplett, 2002, p. 2). The second output, that relating to the additional income that a healthier population earns, is most starkly observed in poorer countries, where many people have substantially impaired productivity because of poor levels of health (Strauss 5 Michael Grossman (1972) cites Jeremy Bentham as including health in one of 15 “simple pleasures.”
OCR for page 132
Beyond the Market: Designing Nonmarket Accounts for the United States and Thomas, 1998). But the link between health and income exists in developed countries as well. For example, people who are depressed are less likely to finish their education, more likely to drop out of the labor force, and less productive at work than are people who are not depressed (see Sturm et al., 1999, Leroux et al., 2003). Similarly, older individuals who suffer adverse health shocks are more likely to drop out of the labor force than are people who do not suffer such shocks (Smith, 2004). A health account thus would include the present value of the expected flow of future earnings, as in the human capital account. The economic return to better health can be measured analogously to the economic return to education (see Chapter 5). Wage equations can be estimated to determine how health affects total income, hours of work, and wages. There are a host of econometric issues in specifying these equations: measurement of health is difficult; health may be endogenous to income; other factors such as pollution or smoking may influence both health and income, and so on (Farrell and Fuchs, 1982). The armamentarium of modern econometrics can be brought to bear on these problems, however, and it is reasonable to expect good results. The novel challenge in constructing health accounts is the measurement of health. Without a measure of health, one cannot monitor and value the central outputs of the medical system. Because health is a multidimensional concept, encompassing length and quality of life and involving both mental and physical health, it is very difficult to measure. Most of the research to date has sought a method of summarizing information on diverse health states using a common metric and then valuing a common increment on that scale. The scale typically is called a quality-of-life scale, and the resulting index of the stock of health a quality-adjusted life expectancy measure (for a detailed discussion, see Weinstein et al., 1996). Imagine ranking all medical conditions on a scale ranging from 0 to 1, where 0 is death and 1 is perfect health. If no state is worse than death, every health state can be placed on that interval. A person has a quality of life each year, which may vary over time. Now suppose that we know the value of some particular increment on the common scale—for example, the value of perfect health relative to death. Then, the value of an individual’s health state can be given as the product of the present value of her or his quality of life relative to perfect health, times the value of a year in perfect health. Formally, the value of health is defined as: (6.1) where V is the value of perfect health; δt is the applicable discount rate and qt is an individual’s quality of life for year t. This framework is common in the literature, but it should be noted that it is not innocuous. It implies, for example, that similar quality-of-life improvements are worth the same amount at all ages; that people are not risk averse; that the
OCR for page 133
Beyond the Market: Designing Nonmarket Accounts for the United States utility from health in one period is independent of health status in other periods; and that utility from longevity is independent of the utility from health-state level (for a discussion of the assumptions underlying various health valuation metrics, see Hammitt, 2003; Eeckhoudt and Hammitt, 2001). Still, some such structural representation, perhaps with the introduction of age-specific adjustments, is the way most researchers have made progress on this difficult issue. There are important distinctions between how health output would be valued in conjunction with the frameworks just discussed and how market goods and services are valued in the NIPAs. There are often gaps between individuals’ willingness to pay for a good or service and the cost of providing that good or service. Consider a serendipitous discovery of a new technique for curing a terrible disease. Suppose further that this technique costs little or nothing to implement. In the health accounting frameworks discussed above, this discovery would add significant value because it enhances health. Put another way, value is high because willingness to pay is high. In contrast, in a NIPA framework, this discovery would have little or no value because little or no resources are required in order to enjoy it. Things of value that require no resources have no value in a NIPA framework. The distinction between what one would be willing to pay for a good or service and what one has to pay is ubiquitous in the valuing of nonmarket transactions. Measurement procedures that use willingness to pay for nonmarket activities are not compatible with how market activity is measured, and overstate the value of nonmarket activities relative to conventional measures of market activities. One can envision that, for many health policy purposes, it would be helpful to use a willingness-to-pay metric to value health output. For other applications, it might be conceptually preferable to use more conventional valuations even though, in many instances, it may be impossible to estimate anything that would reasonably serve as an analog to a market price. Sometimes the nature of available data may not leave much choice in the matter; either way, though, health accounts must be transparent with respect to how increments to health are valued. Ideally, when the health account produces a valuation that includes consumer surplus, an attempt should be made to decompose total value into surplus and product to allow comparability with the NIPA measurement framework. Measuring Health Status The first step in measuring health output is to develop a scale along which different health states can be compared. The objective is to develop measures that allow changes in a population’s expected quantity and quality of life to be estimated. Such measures would be capable, for example, of reflecting an average change in the length or quality of life for a population in response to the appearance of a new disease, to an effective new treatment for an old disease, or to
OCR for page 134
Beyond the Market: Designing Nonmarket Accounts for the United States changes in environmental factors. There is a vibrant and growing literature on the measurement of health states, which we draw from in this section. The standards for what constitutes good health, as well as expectations regarding length of life, shift over time. Still, the panel believes that health should be viewed in absolute terms. That is, measures of health—such as life expectancy, morbidity, and ability to perform daily tasks without pain or other limiting factors—are cardinal measures. They can and should be measured consistently over time.6 A health measurement should show improvement in health whenever the distribution of individuals across health outcomes improves, even if expectations about health are rising along with health outcomes. With this in mind, we highlight two approaches to health status measurement. The first approach—the disease state approach—involves estimating the quality of life for people with different diseases and then multiplying those measures by disease prevalence in the population making adjustments to reflect the percentage of people with multiple conditions. A population’s stock of health capital changes with the prevalence of diseases or with changes in the impact of these diseases on human functioning. Disease-state measurement is complicated by the fact that diseases become more or less prevalent over time for reasons that are hard to explain. New diseases appear (e.g., AIDS, SARS); others fade away or are no longer diagnosed (e.g., neuralgia); and various diseases are at different stages of the proliferation or containment cycle in different regions of the world. A second, related, approach to health measurement—the health-impairment approach—involves asking people about ways that health problems interfere with their lives. Since diseases presumably have effects that interfere with normal activities, this is a more direct way to measure health. As with the disease-state approach, the idea is to quantify changes in the population’s health status that result from various health events or from investments in the inputs to health (medical care, time, research, etc.) made in response to those events. One factor that complicates this type of measurement is that characteristics such as age, gender, life-style, motivation, and preferences may modify the impact of an illness on different individuals; these characteristics also influence the efficacy of treatments. Within these two approaches, there are several ways to confront the need to quantify the relative severity of different conditions. One way is to survey experts, presumably physicians and health researchers, about the extent to which diseases affect quality of life. As with any of the available options, this would entail assigning a numerical score to various disease or impairment states. For example, 6 This does not mean that the relationship between a given health impairment and utility must remain constant. For example, the disutility of a walking impairment may be less today than it was 200 years ago when there was less infrastructure developed to accommodate people with physical limitations. The weights associated with impairments on the quality-of-life scale may have to be adjusted over time, as might the value of a statistical life.
OCR for page 135
Beyond the Market: Designing Nonmarket Accounts for the United States someone with a walking impairment might be assigned a 0.7, relative to someone in perfect health who would receive a 1.0. It is not clear, though, that experts in disease and treatment necessarily have an advantage in evaluating the relative utilities associated with different impaired conditions or disease states. A second approach, around which a significant literature has grown, involves asking individuals to assess their current health status against counterfactual alternatives. The most theoretically appealing health state measurements involve explicit utility comparisons. For example, a time tradeoff approach can be used in which a person with, say, arthritis is asked a question of the form: “Suppose you have 10 years to live in your current health state. How many years would you be willing to give up to live without arthritis?” A related approach is a standard gamble. Here, the subject is asked: “Imagine that an operation is developed that will cure your arthritis but, with some probability, you will not survive the operation. What chance of success would you need in order to go ahead with the operation?” These types of questions can be used to construct a utility assessment for arthritis. In practice, the most common way to value health states is with simple multiple choice questions: “How would you describe your health—excellent, very good, good, fair, or poor?” or with easy-to-use scales: “On a scale of 0 to 100, how would you rate your health currently?” (Torrance, 1987). The answers then can be compared across people. There is an important difference between these scales and utility-based measures that require a survey respondent to compare his or her health state against that of someone in a reference (healthy) population. In the former, one generally compares health ratings of people with a condition to the ratings of others without the condition; people are not asked to evaluate the counterfactual of living a life different from the ones they have. In the latter, people are asked to evaluate their current health state and then to evaluate an alternative with or without a particular condition. One possible advantage of this self-evaluation method is that it controls for other factors that differ across the population. The problem is the speculative nature of such surveys (the definitive volume on subjective evaluation is Kahneman et al., 1999). For instance, the evidence shows that people with an affliction (such as paralysis) report systematically different quality-of-life scores from people without the affliction who are asked to rate what their health would be with the affliction. By comparing survey results that ask people to rate their own health as it is, one may be better able to attribute differences in average satisfaction ratings to a specific condition.7 If the other important factors influencing self health assessments can be controlled for and embedded, estimates of 7 There are still complications. For example, while it is easier for respondents to provide answers using a linear analog scale, that approach also has known biases, in that people tend to choose answers closer to the middle than they do with alternative approaches.
OCR for page 136
Beyond the Market: Designing Nonmarket Accounts for the United States the impact of a condition will be unbiased (Cutler and Richardson, 1997:252). To date, there has been little comparison of these approaches; more is needed. A number of health assessment questionnaires have been developed. The names indicate the degree of experimentation in the field: SF-36 (for short form, 36 items; the longer form was developed in the 1970s); Euroqual-5D (for five domains); the QWB (quality of well-being) survey; YHL (years of healthy life expectancy); and so on. Among these instruments, common dimensions include impairment in functioning due to physical problems, impairment in functioning due to mental health issues, pain and discomfort, and vitality. Surveys may also attempt to track intervention-related factors such as side effects to medication or even the dollar cost of treatments. Questionnaires ask about a range of activities, from those that are directly economic, such as one’s ability to work, to more basic indicators of functioning, such as eating, arising, dressing, reaching, and general mobility.8 The health assessment surveys are designed to measure health-related quality of life with single numbers that represent the population’s preferences for combinations of symptoms relating to mobility, physical activity, and social activity. The scales grade symptoms by the degree to which they affect everyday activities. A lengthy comparative literature is being developed, with adherents of each of these approaches, especially for characterizing health outcomes over time for patients having such serious illnesses as cancer or AIDS. The federal government, itself a producer of health assessment measures, uses them in policy formation.9 Every decade, the U.S. Department of Health and Human Services outlines a set of “healthy people” goals: the first was for 1990; subsequent editions have been issued for 2000 and 2010 (see U.S. Department of Health and Human Services, 2000). These decennial volumes provide a thorough overview of work on these approaches—including the YHL approach noted above—and report results for a set of survey measures designed to track the population’s health. Among the measures tracked are: self-rated health and recent days of physical health, mental health, and activity limitations. Metrics such as “quality-adjusted life-years” and “disease-adjusted life-years” are already widely used in medical applications to identify unmet health needs, and to guide policies for addressing those needs. 8 Survey instruments designed to estimate the relationship between health impairments and the ability to carry out work often do so both in terms of increased absence and reduced productivity. The Health and Labor Questionnaire, the Work Limitations Questionnaire, and the Work Productivity and Activity Impairment Questionnaire are just a few of these surveys. The results from these instruments are often validated against other measures such as quality-of-life scales, and typically measure productivity in percentage terms against an unaffected group (Reilly et al., 1993). 9 The Centers for Medicare and Medicaid Services in the Department of Health and Human Services also maintain a set of “national health accounts,” but (as described above) these are spending accounts, not health accounts.
OCR for page 137
Beyond the Market: Designing Nonmarket Accounts for the United States Despite its vibrancy, this area of research is quite new, and great care needs to be taken when comparing survey measures of health status and well-being across groups, even within rich countries. Because most measures of health status are subjective, differential reporting norms in different groups may affect quality-of-life estimates for those groups. For example, women may be less inclined to describe themselves in extreme terms than are men. As a result, women may report their health as being more towards the middle of the health distribution than men do, less excellent and less poor. This inclination might lead to the misperception that adverse health conditions affecting women are less serious than those affecting men. Recent research has suggested two approaches for addressing response comparability problems. The first is to ask questions about hypothetical, researcher-designed health conditions of different groups. These “vignettes” can be used to benchmark responses that individuals in different groups give about their own health (King et al., 1998). For example, in some cases, self-reported health may differ by gender; women offer less extreme responses than do men to certain kinds of identical vignette questions. This kind of information could be used to adjust, at the group level, different response patterns about own health. A second solution, alluded to earlier, is to make comparisons of health conditions across individuals within a group and then standardize across groups using a common benchmark (Cutler and Richardson, 1997). For example, one could examine the health of women with and without a particular condition and form the health decrement of the condition as the difference between the two. The health decrement for women then can be compared to the health decrement for men. To make an explicit quality-of-life estimate, one might impose, for example, the assumption that people of the same age with no adverse health problems had the same quality of life. Measures of health should be based on objective observation of health outcomes. The field of health assessment has not settled on one approach to measuring health as superior to the others. At this point, the full range of methods, each of which attempts to place various disease or impairment states along a common scale, should be considered further for measuring changes in the population’s health status. Comparisons of competing approaches are just beginning. As research progresses, it may be possible to settle on a preferred approach. For the time being, however, experimentation with the various approaches just described should be encouraged. Valuing Increments of Health Economic analysis of health generally uses a compensating differential framework to estimate the values attached to better and worse health states. Ideally, situations are found in which people are trading health (or expectations about health) for income; monetary estimates associated with the tradeoffs are
OCR for page 138
Beyond the Market: Designing Nonmarket Accounts for the United States then used to infer the valuation of health. The voluminous economic literature on compensating differentials in labor markets provides a useful starting point. People make decisions involving risk comparisons all the time: what job to take and what safety devices to have in their car or house are two examples. In the classic case, the additional income one must pay to attract people to work in risky jobs is taken as a measure of the monetary cost associated with additional health risk. For example, to examine tradeoffs between current occupational hazard or risk and current wages, one might compare the wages paid to window washers who work at street level with the wages paid to window washers working on skyscrapers. The revealed willingness to pay in these settings is a measure of the value that people place on health risks. If the skyscraper job adds a 1-in-10,000 risk of death and pays $100 more per year than the street-level job that is otherwise similar in terms of skill requirements, intensity of the work pace, scheduling flexibility, working conditions, and so on, the implied value of the remaining life span of the person taking the riskier job is $1,000,000. Self selection makes the problem more complicated when aggregating across a population, as people taking riskier jobs will tend to be at the population margin in their valuation of life and/or level of risk aversion. The multidimensionality of health dictates the way in which the compensating differential methodology can be exploited. There are so many health states that it is impossible, in practical terms, to find associated market conditions for each. Consider the situation of extending the life of retired people who have had heart attacks. There is no ready, comparable market situation we can imagine in which people face the risk of living after a heart attack. And even if there were approximations to the longevity benefits, the exact valuation would depend on such subtle factors as the severity of mobility impairment, interaction with other conditions that people have, and so on. The number of disease state and age combinations alone makes it clear that a contingent valuation cannot be established for every health state. This is why research has focused on valuing a common health state, a year in perfect health—the V in equation 6.1. Viscusi and Aldy (2003) offer a comprehensive review of the literature that seeks to estimate the value of a statistical life. They find a great deal of variation in estimates of risk tolerance and, in turn, of the value of life. In examining more than 100 studies on mortality or injury risk premiums, the authors find the median value attached to a statistical life to be around $7 million for prime-aged workers in the United States. Reflecting the uncertainty of such estimates, however, the upper bound is higher than the lower bound by a factor of more than two at the 95 percent confidence interval (Viscusi and Aldy, 2003, p. 68). While there is high variation in these kinds of estimates, and such values may not be applicable in all cases, they are probably reasonably accurate in many market-based settings. On a per-year basis, estimates in the literature translate into values in the $75,000 to $150,000 range (Kenkel, 1990; Murphy and Topel, 2003). In these kinds of studies, $100,000 for an additional healthy year is a typical value used.
OCR for page 139
Beyond the Market: Designing Nonmarket Accounts for the United States Recommendation 6.8: Recognizing that there is a range of uncertainty, a satellite health account should be based on a dollar figure for the value of a year in perfect health derived from estimates in the literature. That value should be updated as further research indicates. Because of the controversial nature of valuation in this account, especially but not limited to valuation on the output side, measures quantifying physical changes in health status (e.g., quality-adjusted life-years) should be published alongside any monetary measures. The high marginal value of life raises a fundamental question of affordability. Bounds on the value of life must ultimately be constrained by household budgets—willingness to pay must be anchored by ability to pay. Since, for most people, annual income is less than the $75,000 to $150,000 per person valuation figure used in the literature, people could not afford to buy many years of life before running out of income. How is this possible? The answer comes from recognizing the importance of initial constraints. The value of life as reflected in compensating variation studies is an estimate of the slope of the indifference curve at the current allocation. It answers the question of how much people would be willing to give up to get an additional year of life, given their current income and life expectancy. As the level of cash income and life expectancy change—for example, if one gives up more income to live longer—the indifference curve tangency point would change, and so would the value of life. With lower cash incomes and longer life expectancy, the marginal value of a year of life would fall. In this context, the value assigned to health should be allowed to increase over time along with income growth (see Costa and Kahn, 2003, for a discussion of rising life valuation). Another set of issues arises in considering whether and how the value of life differs across people. These derived values of life clearly are related to income; rich people can afford to spend more on safety devices and medical treatment than poor people. One set of health accounts thus might value health to the wealthy more highly than health to the poor. Another might give equal weights. One can imagine circumstances in which weighing is appropriate, e.g., for measuring demand or for assessing the effect of health on productivity. In the latter case, one might want to weight by labor supply, or by labor income. Even in this case, if the expected incidence of disease were independent of socioeconomic status, the unweighted average value of health would be an appropriate measure to use. In other applications, such as for general measures of well-being or socioeconomic progress, equal weights are conceptually attractive. More generally, one can think about developing national health accounts from behind Rawls’s veil of ignorance (without a priori knowledge of where they are going to be in the distribution of wealth or income, people will agree to an equal distribution). Suppose one were to value health knowing only what the prevalence of disease will be and the average income in society. In that case, the value of health would be the value to the person with median income. Alternatively, one may wish to
OCR for page 140
Beyond the Market: Designing Nonmarket Accounts for the United States focus on the health of the least well-off members of society or consider an index that focuses on those with the poorest health. DATA REQUIREMENTS Some of the data required for a satellite health account are already collected, but these data are not necessarily available in the form required for such an account. Before a credible health account can be produced, new data will be needed to measure the population’s health status on a regular and continuing basis. For some years, the United States has had relatively good data on health status, such as those from the National Health and Nutrition Examination Survey (NHANES) and the National Health Interview Survey. Until 1999, NHANES—designed specifically to obtain information on the health status of the U.S. population—was conducted only once a decade. Now, however, it is an ongoing data collection program, and survey results are published more frequently; some are available annually. This is an important step toward developing a national health account. Output measurement will require more research that attaches meaningful physical health status data to monetary measures, including work to produce better data on the link between health and earnings (and vice versa). Data sources also will need to be developed that track changes in the relationship between prevalence of disease and years of healthy life and between medical interventions and health outcomes. Health accounts also will require improved measures of the inputs to health. Better organized and more accurate data on medical care spending, aggregated by disease treatment, are part of what is needed. Improved data on care time, such as will be produced in the ATUS, also are necessary for developing the input side of the health account. Measuring the quality of both the inputs to and the outputs of improved health is a further area for needed research. The statistical agencies are working to develop approaches for handling difficult-to-measure changes in the quality of health treatments. The Bureau of Labor Statistics, for example, is working on experimental medical care price indexes based on disease- and diagnosis-based units. Currently, data on medical care prices are organized primarily by institutional provider (e.g., payments to hospitals, doctors, or drug companies), not by treatment. If new treatments are developed for particular conditions that require fewer resources, this is not reflected in the form of a lower price level. Focusing on the cost of treating diseases or diagnoses allows prices to reflect changes in the mix of inputs used to treat particular conditions. Quality-corrected data on the full range of complete treatments do not yet exist—and as Triplett (2002) points out, we are still very far from having this information on per-case cost trends by disease. Such data would be extremely useful, even for conventional accounting of the medical care sector but also for the development of the health satellite account contemplated here.
Representative terms from entire chapter: