Click for next page ( 78


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



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 77
CHAPTER 4 COSTS OF HEALTH EFFECTS The Health Services Research, Health Statistics, and Health Care Technology Act of 1978 (Public Law 95-623) lists eight categories of costs to be evaluated in an ongoing study of the present and projected future health costs of pollution and other environmental conditions resulting from human activity. These categories are: (1) costs of prevention, treatment, cure, convalescence, and rehabilitation (2) loss of income or future earnings (3) adverse effects on productivity (4) other related direct and indirect costs (5) costs attributable to pain and suffering (6) overall increases in costs throughout the economy (7) loss of tax revenues resulting from decreases in earnings and productivity (8) costs to the welfare and unemployment compensation systems and the programs of health benefits under Titles 18 and 19 of the Social Security Act. In addition to quantifying costs, the study is to ascertain the reduction in costs that would result from "incremental reductions" in envi ronmental hazards . Cost measurements are needed to help assess the desirability of increased protection against environmental hazards. Such protection generally incurs losses in production. Resources that could be devoted to producing goods are diverted into environmental protection. At the extreme instance, it could be necessary to reduce or even eliminate the production of some desired commodities. To achieve a rational balance betwe en health e f fe c t s and produc t i on 77

OCR for page 77
losses, it is helpful and even necessary to have a measure of the health costs imposed by not taking steps against environmental hazards. This measure has to be in the same units as the production losses. Most conveniently, these units are dollars; more precisely, they are dollars of a constant purchasing power, "real dollars." Calculations of health costs do not provide a mechanical substitute for a complete process of decision making. They do provide important i nformat ion that should be taken into consideration to avoid gross errors in balancing the many benefits and losses of any environmental policy decision. There is no substitute for the value judgments that do or should emerge from political and administrative processes. Any procedure for costing health effects requires two steps 1. an estimation of the amount of each kind of health effect (death, illness, loss of work days, pain and suffering) associated wi th each kind of pollution or other envi ronment a 1 cond i t i on s tud i ed 2. an estimate of the cost of each unit of health effect (for example, cost per death or cost per day of i llness ~ . The final estimate of costs must inherit the uncertainties in the relation of health effects to environmental factors. As data on environment-related health effects improve, the cost estimates will also improve. Various methodologies of evaluating the costs of illness have evolved in recent years. Chief among these are output accounting and willingness-to-pay, which seek to measure very different kinds of costs. The output-accounting methods measures the value of actual resources forgone and output lost because of illness and disease--such as lost earnings, costs of medical treatment, and costs of prevention. This method, as developed by Rice,1 Cooper and Rice,2 Mushkin,3~4 and others, has been widely used in cost-of-illness studies and undoubtedly influenced the specification of costs in P.L. 95-623. One difficulty with this method is that it may greatly underestimate costs of illness, especially for those, Also called the "human-capital" method. We use the term, "output accounting," to emphasize its basis in the measurement of lost output and its similarities to national income accounting. (National income accounting is the description of the sources and uses of the outputs of the economic system. It is carried out by the U.S. Department of Commerce and presented in the Annual Economic Report of the President and elsewhere in reports of the Government and private economic analysts. The most familiar single number in the national income accounting system is the gross national product (GNP).) 78

OCR for page 77
such as retirees, who are not in the labor force, because they do not incur an earnings loss when they are ill or die. The bill's sponsor, however, was aware of shortcomings in the output-accounting method and noted that a willingness-to-pay approach might also be used.5 Willingness-to-pay assesses the resources individuals would forgo to reduce a risk of death or illness if they had the opportunity to make the exchange. Because of the conceptual appeal of this definition of costs, many economists favor willingness-to-pay, despite considerable difficulties in implementing it. Willingness-to-pay has been developed by Mishan,6 Thaler and Rosen,7 and a number of others. This planning committee proposes that the ongoing study use these two types of costing. The output-accounting method in effect measures health costs as the loss of goods to society as a whole, while the willingness-to-pay method seeks to measure the value set by each individual on health effects in terms of goods, and then adds up these measures over all individuals. Both of these methods imply particular ways of setting real dollar values on health effects (mortality, morbidity, and others). Neither completely exhausts the social meaning of health effects, and it is imperative that these effects be displayed separately. This would be true even if the costing methodologies were without error in their own terms, but uncertainties in costing add to the need for presentation of the health effects data. These are, in any case, an essential step in the development of cost estimates. For those unwilling to accept the equation of mortality and other health losses to dollar equivalents, the individual health effects are important information in the discussion of environmental policy. The exposition of the two approaches to measuring costs of health effects can be simplified if we imagine all the varied outputs of the economy can be replaced by a homogeneous flow of one commodity. The output-accounting approach measures the effects of illness and other health losses on this flow: some of the single output is diverted to various forms of medical care and' therefore, is lost (direct costs), and some of it is lost because persons capable of producing the output die of a disease or are impaired by it (indirect costs). In the case of death from a particular cause, it must, of course, be supposed that the individual will sooner or later die of some other illness. Hence, the product lost by the individual's death is a flow over time up to the expected longevity of that individual (more precisely, of the relevant class of individuals to which the individual belongs, usually defined by age and sex, but additional classifiers can also be used). A stream of output extending into the future can be given a value in the year of the individual's death. The future components of this stream should not be given the same value today as they will have in the future but a lesser value the further off they are. A bank when lending money has to be repaid with interest; this fact implies that the future repayments are discounted to obtain their equivalent current value in 79

OCR for page 77
the dollars lent. In this way, future output should be discounted to find its value today. In practice, the output attributable to an employed individual is measured by his or her earnings. The willingness-to-pay approach, by contrast, implicitly asks how much of the one commodity would each individual be willing to give up for (a small reduction in) each adverse health effect. Willingness-to-pay is measured in the same units as costs in output accounting, that is, in teems of the one commodity or "real dollars," but measures a different object. Thus, suppose a given person, faced with some probabilities of death within the year, is given the opportunity to reduce his or her probability of death by .001 (for instance, from .025 to .024~. This change being in general a benefit, the individual would be willing to give up some of the general commodity to which he or she is otherwise entitled in order to purchase this reduction. For example, if the person has an income of $15,000 a year, that is, a right to that amount of the general commodity, the person might be willing to pay up to $500 to secure the promised reduction, but no more (these figures are purely illustrative). Now suppose, as is generally true of environment-related health effects, that the reduction in probability of death within the year will affect everyone if it affects anyone. Let ~ be the number of people involved, and for simplicity, suppose all have the same willingness-to-pay, that is, $500. Then, in total, individuals are willing to pay $500~. For N large, the number of lives saved will be .001 N. which is N times the probability that anyone's life will be saved. The willingness-to-pay per life saved is then $500N/.001N = $500,000. This does not mean that any individual would be willing to pay $500,000 to avoid certain death. It is a summation of individual willingness-to-pay for small reductions in the probability of death. Individuals will in general have different willingnesses-to-pay for a fixed reduction in probability of death, depending on income, age, and probability of death. In output accounting, an average loss of wages by age and sex is assumed. In the same way that these average losses are used in output accounting, we may hope to find an average willingness-to-pay for some suitably defined group of similar individuals, for instance, those of a given age and sex. Can we find some objective measurement of the willingness-to-pay for avoiding death? If individuals could purchase such immunity on some market, there would be no problem of measurement. However, precisely because such reductions in health hazard are obtainable only by collective action, such markets are not likely to exist. In the absence of such a market, indirect inferences can still be made, although with limited reliability. One technique is to measure market decisions in which risks of mortality and morbidity enter, particularly the choice of an occupation. Under certain assumptions, in order that there be persons entering occupations of differing health risks, the wage levels must be higher for the riskier -80-

OCR for page 77
industries; individuals who choose less risky occupations at lower wages can be thought of as buying reduction in health risks. The difficulty in practice is that occupations differ in attractiveness for many reasons of which risk differences are only one. This is not an insuperable obstacle, but it is a significant one. Another kind of indirect inference as to willingness-to-pay for health is the survey; questions provide a surrogate for actual behavior. Both methods of costing and, indeed, national income measures in general raise a value judgment about income distribution. Adding up dollar values for different individuals presupposes that the dollars are of equal value. Clearly, this postulate is a value judgment and one that many would not be prepared to accept. Most especially, it is frequently held that a dollar to a poor individual is worth more (or represents higher needs) than a dollar to a rich individual. Hence, it can be argued, the dollar health costs to individuals, however defined, should be weighted differentially, with higher weights to the poor, before being added. However, detailed economic analysis under certain assumptions shows that this is less relevant for pollution, which affects persons of all incomes, than for other circumstances.* As will be seen from the following sections , an aggregate measure of health costs should include only categories (1) to (5) of those listed in P.L. 95-623; the remaining categories are included in these, although they may be interesting in their own right. The measurements of these costs according to the output-accounting and willingness-to-pay approaches are then discussed. It should be remembered throughout the discussion that neither of these methods at present can accurately measure all the relevant costs that P.L. 95-623 requests. *The use of distributional weights is controversial among economists. Some argue that income redistribution is not one of the appropriate functions of the government; others hold that allocation decisions in particular areas, such as health, should be made on pure efficiency grounds, while value judgments about income distribution should be implemented through direct redistribution methods (such as tax rates and welfare payments). However,~in the present context, these disputes are not especially germane to the issues of environmentally induced health damages. It can be argued that any reasonable reweighting of the dollar values of different individuals would make little difference to the calculated health costs. The reason is that air- and water-borne pollution affect individuals of all incomes. Hence, any measures to reduce pollution will be evaluated by adding dollar values over individuals of all possible incomes, and the result will be largely invariant to di fferent choices of weights. -81-

OCR for page 77
Costs Listed in Public Law 95-623* Direct and Indirect Costs (1), (2) - The principal measures of costs according to the output-accounting methodology are the value of resources used, say for medical care, or forgone, such as time lost from work because of illness or disease. The direct costs are simply the expenditures on health services, which is cost item (1) above. The indirect costs are the sum of the costs of time lost from work by those too ill to work, and the present discounted value of earnings lost to mortality. More specifically, the last item finds average earnings for the identifiable group (usually defined by age and sex) for the years beginning with that of actual death and extending over the expected longevity for that group. The indirect costs are cost item (2) above. Adverse Effects on Productivity (3) It has long been recognized that the direct and indirect costs noted above account only for a part of the total burden of illness. Additional adverse effects on productivity may occur in several ways. Illness may decrease productivity of people on the job, and absenteeism may increase costs of production with the end result that the value of output per unit of input declines. These indirect costs could be considerable for certain diseases that may be influenced by environmental conditions, such as diseases of the respiratory system, but are difficult to measure. Other Related Direct and Indirect Costs (4) Other related direct and indirect costs may be borne by patients and other individuals. These include non-health sector direct costs for transportation to health providers, certain household expenditures, costs of relocating such as moving expenses, and certain property losses. Illness can force a family to incur expenses in caring and providing for the sick member. These include extra expenditures for household help, special diets, special clothing, items for rehabilitation and comfort, and vocational, social, and family counseling services. Property losses include destruction of property (resulting, for example, from alcoholism and drug abuse) or the depressed value of property because of environmental conditions such as air and water pollution. Indirect costs include the time spent visiting physicians, other health professionals, and hospitalized persons by patients and/or family *The following sections discuss the costs that are listed in P.~. 95-623 and do not include the details of all possible costs associated with the environment and health. -82-

OCR for page 77
members, and time lost from work by family members when someone in the family is ill. Unwanted job changes and loss of opportunities for promotion and education due to illness may reduce productivity and result in indirect costs. _osts related to nonmarket activities Costs of illness discussed so far are those that can be measured, for the most part by expenditures or lost earnings and output. It Is important to account as much as possible for those impacts of disease that are not reflected in market activity or the value of the gross national product (GNP). These include the value of losses in time from, and change in quality of, housekeeping, various voluntary activities, and leisure. Housekeeping services have been valued by various comparisons with marketed services, but other nonmarket activities, such as volunteer work or leisure lost to illness, have not been evaluated. Costs Attributable to Pain and Suffering (5) Illness and disease are responsible for a wide variety of psychosocial deteriorations, most of whose costs are not reflected in cost items (1) to (4) above. Victims of disease, as well as their children, spouses, siblings, friends, and co-workers, may be affected. A victim may suffer loss of a body part or speech, disfigurement, disability, impending death, pain, and grief. The victim, and those around the victim, may be forced into economic dependence and social isolation, and may suffer from anxiety, reduced self-esteem and feeling of well-being, and resentment and emotional problems that may require psychotherapy. Problems of living may develop, leading to family conflict, antisocial behavior, or in extreme cases, suicide. Disrupted development and delinquency may occur among children. The quality of life may be reduced beyond the restorative capability of current rehabilitation programs. The combination of financial strain and psychosocial problems is especially devastating. A particular disease may bring about personal catastrophes that are not reflected in the direct and indirect economic costs usually estimated for that disease, although some of these costs may increase direct and indirect costs classified under some other disease category. For example, the costs of mental illness induced by a spouse's cancer will be allocated to mental illness rather than to cancer. Overall Increases in Costs Throughout the Economy (6) Both in the short-term, and perhaps even more important, in the long-term, illness and disease have significant economic impacts throughout the economy. In 1979 the nation spent $212 billion or nine percent of GNP for health care.8 Resources And for health care are diverted from other uses. The allocat r' of resources -83-

OCR for page 77
affects employment, investment, productivity, prices, and additions to GNP in the health care sector and other sectors of the economy. However, the costs of these impacts will be fully reflected in the other categories of costs listed here, if the other costs are properly calculated. If so, adding a category of overall increases in costs could result in double-counting. Taxes and Transfer Payments (7), (8) With respect to the two remaining items specified as health costs in P.L. 95-623, taxes and transfer payments, such as public aid, disability payments, and similar items, are not costs of disease and should not be added to direct and indirect economic costs. Indirect costs are based on forgone future before-tax earnings, so lost tax payments to the government will already have been counted. Transfer payments simply reallocate income from one individual (for example, the wage earner) to another (the disabled). Although these transfers represent a cost to the wage earner in the form of a reduction in disposable income, one person's loss is another's gain, and the net cost to society resulting from this transaction in terms of resources used (and thus unavailable for other alternatives) is zero, except for costs incurred in operating the system that effects the transfers. Other examples of transfers are interest and capital gains forgone because of forced sale of assets, loss of property for failure to meet mortgage payments, and stolen property. Costs of illness, in the framework set forth so far, are the value of resources used, resulting in forgone alternatives, and resources lost because of morbidity and mortality. On the other hand, transfers, as the name implies, are shifts of control over the use of resources. Direct and indirect costs are unambiguous losses that would not occur if illness were reduced, while transfers take resources from one segment of society and give them to another. Transfers may alter the allocation of resources among competing ends, but are not a use of resources in and of themselves. For example, consider the case of drug addiction. In addition to the usual direct and indirect economic costs, society may choose to incur expenses in the criminal justice system and provide welfare payments that would not be considered necessary in the absence of drug addiction. Additions to the criminal justice system are a use of resources and therefore a cost of drug addiction that falls into the category of "other related direct and indirect costs." Welfare payments are a simple transfer. If criminal activity resulting from drug addiction causes property losses, the value of property destroyed is one of the related direct costs, while property stolen is a transfer. -84-

OCR for page 77
Although the value of a transfer per se is not a cost in the output-accounting method, transfers undoubtedly have an impact on quality of life and welfare of both givers who lose and recipients who gain. For this reason, it may be desirable to investigate transfers, exploring for example, the redistribution of income that takes place and the impact on those who lose and those who gain. If it is desirable for certain purposes to estimate the impact of disease on taxes or various transfer payments, it should be made clear that these are not additional costs of disease but already have been included in estimates of the total cost of disease. Failure to make a distinction between transfer payments and costs will result in a serious overestimate of costs because of double counting. Estimating Costs of Illness and Disease by the Output-Accounting Method Most often, studies of cost of illness provide an estimate of the economic burden resulting from the prevalence* of disease or illness during a given time period, such as a year. In these studies, several of the health costs to be evaluated present relatively little difficulty, once the impact of the environment on health is determined, although the data and/or methodology are severely lacking for other costs. The most progress has been made with respect to categories (1) and (2) above, direct and indirect economic costs, but here, too, there are problems. Direct and Indirect Costs It is possible with existing data and methods to calculate national estimates of certain direct and indirect economic costs of disease for the broad three digit International Classification of Diseases (ICD) categories, such as neoplasms, diseases of the circulatory system, diseases of the digestive system, and so on.2 It is also possible to disaggregate some of these costs into more specific disease categories; for example, expenditures for short-term hospital care for neoplasms can be itemized by selected cancer sites 10. The costs included in Cooper and Rice2 are the direct medical care expenditures for hospital care, physicians', dentists', and professional services, nursing home care, drugs and drug sundries, eyeglasses and appliances, as well as the indirect costs, or lost earnings due to morbidity and premature mortality. There are well over 200 separate cost-of-illness studies that use output-accounting methodology.ll Some of these are national in scope, but most are limited to a selected population or geographic *prevalence the total number of cases or amount of disease existing in a population during a specified time period. -85-

OCR for page 77
area, and all but a very few are restricted to one or a few disease categories. Data sources and methods vary among these studies, and many limit their investigation to one or several of the direct and indirect economic costs outlined above. The discussion that follows is confined to data needs for calculating national estimates of disease costs. There is a need for continuing estimates of the economic costs of disease that are national in scope and cover a broad spectrum of disease categories. These can be used to estimate the economic costs of the health effects of environmental hazards as the relationships between environmental contaminants and health are specified. There will also be a need for special studies, possibly with much greater detail than provided by the ongoing study specified in P.L. 95-623, of costs for specific diseases, population groups, or geographic areas. The nature and timing of these special studies may be determined by the concerns of the ongoing study, or may be undertaken outside the ongoing study altogether. Total and net direct costs The standard methodology for estimating direct costs may be misleading if applied without modification to the health effects of pollution and other harmful environmental conditions. (See Garber.l2) Output-accounting studies have not distinguished the net direct costs of a disease from the total direct costs. Total direct costs are direct costs as conventionally measured. A disease may cause a change in the direct costs a patient bears in the future as well as during the illness. Net direct costs include these future charges, and can be defined as the changes i n the d i scounted value of lifetime medi Cal expenditures resulting from the disease. These net direct costs are the total direct costs (the medical costs directly attributable to the disease) less any reductions (or plus any increases) in future costs attributable to the disease. The distinction is likely to be important primarily for fatal illness. Those who die from a disease may suffer substantial direct costs before their deaths. However, they avoid the health-related expenses they would have borne in the future had they not died with this disease. Although the total direct costs of a disease are always zero or positive, the net direct costs may be negative. For each fatal case of a disease, then, the net direct costs would be the direct costs as conventionally measured, minus the discounted value of future direct costs that would have been expected in the absence of the fatal disease. With the simplifying assumption that conventionally measured direct costs approximate net direct costs for non-fatal cases, a correction need only be applied to the fatalities. If, in addition, a particular fatal disease accounts for only a small part of total national medical expenditures for all individuals, the change in net direct costs resulting from a change in the disease incidence can be further simplified, as follows. A lifetime profile of total direct costs for all diseases (or total medical and health-related -86-

OCR for page 77
expenditures) could be developed. Age/race/sex specific categories of discounted expenditures similar to profiles obtained for indirect costs, could be calculated, perhaps using insurance data. With a known age/race/sex distribution of mortality from the disease, estimated discounted expenditures avoided because of death could be subtracted from total direct costs of the disease to obtain the net direct costs. The distinction between net direct costs and~total direct costs can be illustrated by considering two men, one age 80 and the other age 50. Each enjoys "average" health for his age. What would be the cost if each suffered an accidental death involving no medical expenditures? Total direct costs of these two accidental deaths would be the same for each. But because the older man would be expected to incur lower direct costs over his remaining lifetime--mainly because he would be expected to die sooner--his net direct costs would be quite different from those of the 50-year-old. If, in addition, the 50-year-old suffered from chronic diseases requiring exceptionally costly treatments, the difference in net direct costs would be more pronounced. Thus, the accidental death might bring about a large reduction in medical expenditures in this case. It is thus entirely possible that elimination of particular diseases will ultimately lead to greater national health care expenditures: it would still be desirable to eliminate these diseases, if feasible when all relevant factors are considered. But the economic benefits may be reflected in the savings of indirect costs, rather than direct costs. Use of total direct costs rather than net direct costs would exaggerate the savings in medical expenditures that could be realized by elimination of some diseases, because elimination of diseases that incur low direct costs may result in people dying later of diseases that involve much greater medical expenditures. Net direct costs are thus the appropriate measure of direct cost reductions that can be expected from a reduction in the incidence of a disease. However, it is also important to present total direct costs so the figures can be compared to existing cost-of-illness studies that use the output-accounting methodology. Allocating direct costs ScitovskY13 has raised several additional caveats regarding the estimation of direct costs. The basis for national estimates of direct costs is the national health expend) ture series produced now by the Health Care Financing Administration and formerly by the Social Security Administration. Although the estimates of total~national health expenditures appear to be reasonably accurate, there is considerable evidence that the series overestimates hospital expenditures, possibly by as much as 11 percent, and underestimates expenditures for physicians' services and drugs. -87-

OCR for page 77
diagnostic categories in the International Classification of Diseases. Much of the data employed are sample estimates, and the extent to which the costs for broad categories of disease can be disaggregated, or itemized, depends on the diagnostic category and type of cost. For instance, expenditures for short-term hospital care and physicians' services for neoplasms have been disaggregated for selected cancer sitesl and expenditures for individual digestive diseases are available.27 It is clear that, given the nature of the samples, hospital care can be disaggregated while maintaining statistical reliability to a greater extent than can physicians' services. - Determining the portion of the total economic costs of disease that is attributable to environmental hazards is difficult because of the paucity of knowledge about the effects of the environment on health. The difficulty is compounded by the requirement of P.L. 95-623 not only that present but also future health costs be quantified and, further, that the reduction in health costs that would result from incremental reductions in pollutants and other hazards, as well as the level of health costs, be ascertained. The need to estimate future costs and incremental reductions in costs fosters interest in a conceptually appealing but little-used method for estimating costs of disease, the incidence approach. In order to estimate future costs and determine the reduction in health costs that would result from incremental changes in environmental conditions that lower the incidence* of disease, one can calculate the discounted lifetime costs associated with the incidence of disease from onset until cure or death. This measures the discounted total direct costs saved by preventing a new case of disease. Incremental changes in environmental conditions may also reduce the severity of a disease. For example, reducing air pollution may lessen the future severity of a case of emphysema, so that use of health services decreases. These costs are difficult to estimate, however, as they require knowledge of the likely course of a disease, medical care that will be used, amount of disability and debility, time between onset and death or -cure, and the impact of morbidity and mortality on earnings. These factors vary greatly even within a specific disease category, and for a disease such as cancer will depend on organ site, histological type of cellular change, and stage of disease development when treatment commences. Tithe incidence of a disease is the number of new cases in a given time period, such as a year; the prevalence is the number of cases existing during a specified time period, regardless of when they began. Incidence-based costs are the costs of the disease from onset until cure or death, for cases that began during the specified time period. Prevalence-based costs are the costs incurred during the time period of interest, such as one year. -94-

OCR for page 77
Further, other medical care costs besides those resulting directly from the disease in question must be taken into account, 12 because a chain of health effects may occur. One disease may bring about another disease or illness, so that some part of the cost of the latter should be included in the cost of the antecedent disease. Thus, diabetes causes other chronic conditions, such as blindness, heart disease, and other vascular disease. Similarly, arthritis may be a complication of psoriasis. Various physical ailments may result from alcohol abuse. Digestive diseases can contribute to heart, lung, or kidney failure. Premature death from one disease will mean that an individual forgoes medical care expenses of some kind that would have been incurred subsequent to death had the person not had the fatal disease. As in the prevalence approach, the appropriate measure of net direct costs in the incidence approach includes expenses of subsequent diseases originating from the antecedent disease that is of interest and excludes, in the case of a fatal disease, the direct costs of nonrelated diseases that would have been incurred over the remaining lifetime. But attempts to estimate costs per case of a disease are hampered by severe limitations of data and knowledge e Estimates of direct costs will be the same by either the prevalence or incidence approach for acute conditions having a short durations and in a theoretical steady-state situation. It is uncertain, however, how prevalence and incidence estimates of costs will compare during periods of transition between steady states, for example when public policy may be reducing or eliminating environmental hazards, thereby lowering the incidence of certain diseases. The development of incidence cost estimates for a broad spectrum of disease categories suitable for ascertaining the costs of environmental hazards would require major new data collection efforts and new surveys. A more feasible alternative, initially at least, is to rely on the existing methods and more easily obtainable data and attempt to develop models to estimate incidence costs from prevalence costs. Initial insight might be gained from a comparative analysis of prevalence-basedl and incidence-based28 studies of particular diseases, such as various types of cancers. There also are incidence studies of coronary heart disease29 and stroke30 that could be compared in a limited way with prevalence estimates of costs for diseases of the circulatory system and cerebrovascular diseases.18 For ascertaining levels of costs during a period of time, even where incidence-based costs are conceptually superior, prevalence-based costs will probably suffice and may well be superior, because of the severe data restrictions and lack of knowledge of future trends that beset attempts to estimate incidence costs. Ultimately, the ongoing study will have to address the problems of estimation of the several costs identified in P.L. 95-623. The issue may not be simply how to estimate these costs, but also whether the costs as specified can reasonably be quantified and whether there -95-

OCR for page 77
are additional burdens of disease to be considered, or some to be deleted, from among those set down in the legislation. This applies perhaps most, but not exclusively, to the mandate to determine "the overall increases in costs throughout the economy" resulting from diseases caused, contributed to, or aggravated by environmental hazards. The Wil1ingness-to-pay Approach to Costs of Disease ~ The other prominent method for assessing the costs of illness is willingness-to-pay. It is a measure of the values individuals.place on changes in the risk of adverse health outcomes. Most of 'the willingness-to-pay literature has analyzed death risks. The cost of increased incidence of disease, by this approach, is the combined amount all individuals would pay to avoid the increased risk each faces. For each individual, the cost of increased incidence is his or her willingness to pay per unit increase in probability of disease multiplied by the actual probability of disease in the population as a whole. An increase in the number of cases of disease raises the risk, even to people who never acquire the disease. Thus, willingness-to-pay costs, like output-accounting costs, are based on actual number of cases of disease. The values of willingness-to-pay may be based partly on the same considerations as output accounting, such as the medical expenditures and earnings losses that would result from the disease. But willingness-to-pay also includes lost leisure, pain, suffering, and other subjective costs of illness. At least in principle, all costs that matter to the potential victims of disease are accounted for, even those that are not fully reflected in expenditures or output losses. This approach appeals to many economists partly because it has a firm conceptual foundation in welfare economics. ~ detailed presentation of the method can be found in Thaler and Rosen,7 and Garberl2 has presented a critique and comparison with output accounting. Willingness-to-pay estimates, like those of output accounting, are likely to be related to income. The amount someone would pay to avoid the risk of an illness, or would demand for bearing excess risk, is higher for wealthier persons. However, those people who are permanently outside the labor force--such as retirees--might be willing-to pay significant amounts for risk reduction, even though they have negligible future earnings. Their demand for safety, like their demand for other goods, is unlikely to fall drastically upon retirement. The costs of disease to this population, if estimated by an output-accounting approach, would be markedly understated, because the indirect costs (discounted future earnings) of those no longer in the labor force are zero. Output-accounting costs depend explicitly on -96-

OCR for page 77
earnings, whereas earnings affect willingness-to-pay in a less direct manner. As indicated earlier, willingness-to-pay studies of death risks typically report a "value of life," which is the estimated value of a small change in the probability of death, near a probability of zero. The value of life is the value of the change in risk resealed to a probabi li ty of one. Both surveys and inferences from observed economic behavior have been used to estimate value-of-life figures (as well as the willingness-to-pay to avoid non-fatal illness or accidents). Survey-based estimates have been little used or discussed, and those that have been published are better described as illustrations of methodology than as serious attempts to derive representative values. These questionnaires directly ask people, in one form or another, how much they would pay to support a program that would reduce their probabi li ty of death in the next year. 31, 32 The observation that individuals attach an implicit value to ri sk in many aspects of their lives, because so much behavior consists of actions that may alter risk of death, is fundamental to the market-based willingness-to-pay methodology.33 Studies of labor markets have found, nearly universally, that risky jobs are rewarded with extra compensation. For instance, Thaler and Rosen7 used a risk variable defined as the excess probability of death of individuals in various occupations, adjusted for age. The variable was expressed in units of excess deaths per work-year. The sample mean risk was about .001, and according to one of their regressions a worker at an annual excess death risk of .001 was awarded about $176 more in annual wages than a worker who faced no excess risk. In this case, the value of life is $176/.001=$176,000 compensation per excess stati st ical death. This figure is used for comparison, and in no way implies that this same compensation per unit risk is valid for large changes in ri sk or for small changes around an initial probabi lity much higher than .001. Estimates for the value of li fe derived from wage differentials suitably controlled for other factors range from about $175,000 to $1,500,000 in 1978 dollars (Table 4-1~. The lower figures probably underestimate values for the general population because they are based on compensation paid to workers accepting especially risky jobs. Workers in risky jobs would tend to be those who value risk of death less highly than others, so the compensation they accept for risk is probably less than the compensation a randomly selected individual would demand for comparable risk. Table 4-1 gives value of life estimates from selected market-based studies . The comments concern di fferences among the data sets or methodologies used. Each study uses regression analysis -97-

OCR for page 77
to explore the effects of job risk (and of other pertinent variables) on earnings. Viscusi's study differs from Thaler and Rosen Is in a number of ways, the most important being the use of an industry, rather than occupational, risk variable. The variability of these results emphasizes the need for an adequate representative data base in willingness-to-pay studies. TABLE 4-1. IMPLICIT VALUE OF LIFE FROM WIttINGNESS-TO-PAY STUDIES Study Value of Life Comments Thaler and Rosen7 $200,000 (1967 $) Low income, high risk sample; annual incremental death risk=.001 Viscusi34 $1-1.5 million Uses industry risk; (1969 $) worker perception of danger raises risk 1/lOth that of Thaler and Rosen sample Smith35 $1.5 million Sample similar to (1973 A- Viscusi's; $2.5 million no control for (1976 $) injuries Bailey36 $170,000- Adjusts for $584,000 (1978 $) survivor's benefits; inflates incomes - to make sample repre sentative: corrects for injuries; adjusts for business taxes; adjusts for direct costs borne outside family Willingness-to-pay approaches to costs of morbidity raise some practical difficulties. The nature and severity of the morbidity attributable to environmental agents is highly variable. Health effects may range from minor throat irritation to permanently disabling lung disease. Although measurement of the value of these health effects is conceptually similar to the analysis of mortality by either the willingness-to-pay or output-accounting approach, a comprehensive attempt to measure costs accurately would require massive data-gathering and analytical efforts. There have been -98-

OCR for page 77
limited attempts to ascertain willingness-to-pay for the avoidance of minor health effects (based on questionnaires),37 but evaluation of the costs of all types of morbidity resulting from environmental hazards would require major new research. Labor market studies are unlikely to yield much information about the costs of specific types of morbidity, apart from occupational injuries, without substantially more detailed data than are currently available. Furthermore, applying compensating differential arguments to answer such questions would severely strain the assumption that workers act under perfect information. - Although pain and suffering aspects of morbidity are important components of health costs, perhaps even dominating the financial losses due to morbidity, existing studies yield little information about their value. Therefore, until further research is done on the non-monetary costs of morbidity, the best estimates will be those based on measures of the value of time lost due to disease. Such figures are likely to underestimate the true costs of pollution- related morbidity. The value of lost hours of leisure and work time is the opportuni ty cost of morbid) ty. A decline in quali ty of work hours (productivity) or of leisure time should also be included, but its measurement wi 11 be di fficult . For chronic, disabling di seases, the productivity loss might be measured by the difference between a person's actual earnings history and that which would have been predicted in the absence of disease. Preliminary, rough estimates of productivity losses due to pollution indicate that they may be substantial.38 The data and methods for estimating these productivity losses need further improvement before definitive results can be expected. For valuation of morbidity and mortality, surveys to elicit willingness-to-pay may yield information that alternative approaches could not. They have greater potential for valuing pain and suffering than any of the alternative methods and can be readily applied to differential costs of diseases. For instance, wage premiums may poorly reflect the costs of long-term exposure to toxic agents because workers inadequately perceive the risk. Market studies are best applied to occupational accidents, where both workers and researchers recognize the risk and its relation to work. The valuation of a large range of health effects that would be difficult to infer from market behavior can be obtained by questionnaires. Progress in assessing costs of specific health effects will be facilitated if, after identifying the diseases caused by environmental agents and symptoms caused or exacerbated by them, the ongoing study performs or supports surveys to assess wi llingness-to-pay to avoid the health effects. Health effects of concern might include various acute respiratory complaints, angina, _99_

OCR for page 77
and chronic respiratory disease, and death from various diseases. Any such survey should include important socioeconomic variables, such as age, sex, occupation, wages, education, and wealth. Health status variables, especially a listing of the respondents' chronic diseases, will also be valuable. These data will enable researchers to derive a willingness-to-pay function for each of the health effects studied, based on personal characteristics of the individuals at risk. The theory and methodology for inferring willingness-to-pay from job-risk premiums is already well-developed. (See; for example, Thaler and Rosen.7) Although applications of hedonic prices* to other activities and to product demand may yield further information on value of risk, they are unlikely to provide more accurate estimates than the labor market studies. Few methodological breakthroughs could be expected in this area, but improvements in data could yield better estimates of risk premiums. To obtain better estimates of job risk premiums, this committee suggests that new data sets containing personal and job characteristics for workers be collated. They should include the major personal characteristics relevant to earnings, accident histories by firm, occupation, and industry, disease histories of the workers, and work histories (by occupation/industry, if not by firm). Casualties should be divided into fatalities, permanently disabling injuries, and temporary disability. There should also be measures of worker's compensation coverage, because failure to correct for such insurance will bias the wage premium for injuries downward. A much needed simple improvement in the labor market data would be the use of a better defined earnings or wage variable. A worker making a decision to reject or accept risky work should rationally consider the after-tax compensation; thus, net wages should be the basis for a willingness-to-pay calculation. Gross earnings should also be determined, because another cost to society of illness and death is taxes not collected (a cost which the worker, however, probably does not consider). Some of the data sets commonly used for willingness-to-pay studies do not have clearly defined wage variables. Conclusion Both the output-accounting method of estimating costs of illness and disease by direct and indirect costs and the willingness-to-pay approach seek to quantify in monetary terms aspects of the burden of Prices imputed to product characteristics rather than to the products themselves. In this case, the hedonic prices of risk of death and other health effects are sought. -100-

OCR for page 77
illness, but from very different perspectives. Direct and indirect costs are a measure of resources used and lost due to disease and illness, some portion of which would be made available for other uses if the burden of health problems were reduced. Lessened direct and indirect costs are among the benefits to be derived from reductions in environmental hazards. Willingness-to-pay values are measures of the monetary values attached by individuals to changes in welfare that would accompany changes in the probability that a particular event, such as death from a specific disease, would occur. Willingness-to-pay could be helpful in indicating how persons value health and life and in deriving social preferences regarding public policy for control of environmental hazards. Willingness-to-pay might be especially helpful in assessing the intangible burden of pain and suffering, which is not amenable to evaluation in terms of the monetary value of resources used or forgone. Although conceptually different, serving different purposes, and measuring different aspects of disease, these two measures are not completely divorced from each other. It seems likely that a person's perception of the medical care expenses, lost earnings, and pain and suffering that would have to be endured as a result of i llness would influence willingness-to-pay for reductions in the probability of suffering morbidity and/or mortality. And Garberi2 has shown that under reasonable assumptions the sum of direct and indirect costs is a lower bound for willingness-to-pay values of life. Nevertheless, these two approaches for estimating costs of illness are not simply alternatives. They are conceptually different and, together or separately, can contribute to greater understanding of the costs of health effects of environmental hazards. 1 0 1

OCR for page 77
REFERENCES Rice, D.P. Estimating the Cost of Illness. Public Health - Services Publication No. 947-5, Washington, D.C.: U.S. Government Printing Office, May 1966. 2. Cooper, B.S., and Rice, D.P. The economic cost of illness revisited. Social Security Bulletin 39: 21-36, February 1976. Mushkin, S.J. and Callings, F.A. Economic costs of disease and injury. Public Health Reports 74 795-809, September 1959. Mushkin, S.J. Health as an investment. Journal of Political Economy, 70: Part 2 (Supplement) pp. 129-157, October 1962. 5. Congressional Record, H10588-H10597, September 25, 1978. 6. Mishan, E.J. Evaluation of life and limb a theoretical approach. Journal of Political Economy 79:687-705, 1971. Thaler, R. and Rosen, S. The value of saving a life: evidence from the labor market. In Household Production and Consumption. Studies in Income and Wealth. Number 40. Edited by N.E. Terleckyj. New York Columbia University Press (for the National Bureau of Economic Research), 1976. 8. Gibson, R.M. National health expenditures, 1979. Health Care Financing Review 2 1-36~ Summer 1980. 9. Mushkin, S.J. and J~andefeld, J.S. The economic benefits of improvements in mortality experience, 1900-1975. Report B3 Revised, Georgetown University Public Services Laboratory, Washington, D.C. January 19, 1978. 10. Rice, D.P. and Hodgson, T.A. Social and economic implications of cancer in the United States. World Health Statistics Quarterly 33:56-100, 1980. 11. Synthesis of Cost of Illness Methodology. Reports in preparation under Contract No. 233-79-3010, National Center for Health Services Research. Contractor is Public Services Laboratory, Georgetown University, Washington, D.C. 12. Garber, A.M. Costs of Health Effects. Paper prepared for this Institute of Medicine planning study. October 6, 1980. 13. Scitovsky, A.A. Estimating the direct costs of illness. In Costs of Illness Estimation. Edited by H.A. Sandifer and T. Hu, Georgetown University, Washington, D.C. To be published. 102

OCR for page 77
14. Institute of Medicine. Reliability of Hospital Discharge Abstracts. Washington, D.C. National Academy of Sciences, . . February 19 7 7 . 15. Institute of Medicine. Reliability of Medicare Hospital Di s charge Records . Washington, D. C. National Academy of Sciences, November 1977. 16. Institute of Medicine. Reliability of the National Hospital Discharge Survey Data. Washington, D.C. National Academy of Sciences, April 1980. 17. National Center for Health Statistics. Data Systems of the National Center for Health Statistics. DREW Publication No. (PUS)80-1247. Washington, D.C. U.S. Government Printing Office, 1980. 18. Paringer, L. and Berk, A. Costs of i llness and disease fiscal year 1975. Washington, D.C.: Georgetown University Public Servi ces Laboratory, January 4, 1977. 19. Walker, R.E. and Gauger, W.H. The dollar value of household work. Information Bulletin No. 60. Ithaca, New York: New York College of Human Ecology, 1973. 20. Mushkin, S.J. and Landefeld, S.J. Non-health sector costs of illness. Report A7. Washington, D.C. Georgetown University Public Services Laboratory, January 25, 1978. 21. Mortality and the life cycle of the family some implications of recent research. World Health Statistics Report 29:220-234, 1976. 22. Feichtinger, G. and HansTuwka, B. The impact of mortality on the li fe cycle of the fami ly. International Union for the Scientific Study of Population Conference, Mexico City, 1977. 23. Preston, S.H. Demographic and social consequences of various causes of death in the Uni ted States . Social Biology 21: 144-162, Surer 1974. 24. Mushkin, S.J. Biomedical Research Costs and Benefits, pp. 383-387. Cambridge, MA Ballinger Publishing Company, 1979. 25. Abt, C.C. The social costs of cancer. Social Indicators Research 2:175-190, 1975. 26. Richmond C.A., Honton, E.J., and Stacey, G.S. Methodology to estimate the cost of pain and suffering in burn victims. Draft report to U.S. Consumer Product Safety Commission, 1980. -103-

OCR for page 77
27. U. S. National Commission on Digestive Diseases. Report to the Congress of the United States of the National Commission on _ Digestive Diseases, Volume 4' Part 4. Epidemiology and Impact. DHEW Publication No. (NIH) 79-1887 , Washington, D. C.: U. S. Government Printing Of fice, 1979. 28. Hartunian, N.S., Smart , C. N., and Sizer, J.L. The incidence and economic cost s of cancers . Paper prepared for the Insurance Institute for Highway Safety by Policy Analysis, Inc., Brookline, MA, December 1978. ~ 29. Hartunian, N.S., Smart, C.N., and Gad, N. The incidence and economic costs of coronary heart disease. Paper prepared for the Insurance Institute for Highway Safety by Policy Analysis, Inc., Brookline, MA, August 1978. 30. Thompson, M. and Mills, E. The incidence and economic costs of stroke. Paper prepared for the Insurance Institute for Highway Safety by Policy Analysis, Inc., Brookline, MA, March 1978. 31. Action, J. P. Evaluating Programs to Save Lives: lhe Case of Heart Attacks. R-950-RC. Santa Monica, CA: The Rand .. Corporation' 1973. 32. Jones-Lee, M.W. The Value of Life. Chicago: The University of Chicago Press, 1976. Bailey, M.J. Safety decision and insurance. American Economic Review 68:295-298, 1978. 34. Viscusi, S.W. Labor market valuations of life and limb: empirical evidence and policy implications. Public Policy 26:359-386, 1978. 35. Smith, R.S. The feasibility of an "injury tax" approach to occupational safety. Law and Contemporary Problems 38:730-744, - 1974. 36. Bailey, M.J. Reducing Risks to Life: Measurement of the Benefits. Washington, D.C.: American Enterprise Institute, 1980. 37. Loehman, E.T., Bera, S.V., and Arroyo, A.A. Distributional analysis of regional benefits and cost of air quality control. The Journal of Environmental Economics and Management 6:222-243, 1979. 38. Cracker, T.D., Schulze, W., Ben-David, S., and Kneese, A.V. Methods Development for Air Pollution Epidemiology, Volume I: Experiments in the Economics of Air Pollution Epidemiology. U. S. ~ Environmental Protection Agency. Washington, D. C.: U. S. Government Printing Office, 1979. 104-