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 130
Valuing Health for Regulatory Cost-Effectiveness Analysis 4 Beyond Ratios: Ethical and Nonquantifiable Aspects of Regulatory Decisions Benefit–cost analysis (BCA) and cost-effectiveness analysis (CEA) provide summary measures of the economic efficiency of health and safety regulations—of the net benefits they deliver and of the cost per life year or quality-adjusted life year (QALY) saved. In measuring net impacts on social welfare, both BCA and CEA implicitly contain normative assumptions regarding the relative value of different contributors to well-being. It is important that we understand the ethical assumptions implicit in BCA and CEA and consider the implications of these assumptions when using the resulting information to make decisions about regulating risks. In this chapter the Committee discusses ethical and other unquantified considerations in regulatory analysis and decisions, particularly with respect to QALY-based CEA. Normative assumptions are implicit in how the benefits measures are constructed in BCA and CEA. In CEA, for example, effectiveness measures such as life years or QALYs weight lives saved by considering remaining life expectancy. They thereby assign a greater weight to saving the life of a younger person than an older one, if other factors are equal. In BCA, willingness-to-pay measures in theory can be designed to address a wide range of factors, but in practice may not address important dimensions of the population affected or the nature of the risks. Other normative issues arise due to factors that are excluded because of the focus on efficiency or because of data or methodological limitations. Such considerations relate to:
OCR for page 131
Valuing Health for Regulatory Cost-Effectiveness Analysis The distribution of the effects across different population subgroups; Impacts that cannot be quantified easily; and Features of the risks that are not easily captured in the benefits measures. Economic analysis is only one of many kinds of information that contribute to decisions related to regulation of health and safety risks. Regulatory decisions should and do reflect a number of considerations in addition to aggregate estimates of costs and benefits.1 The goal of this chapter is to discuss the assumptions and methodological limitations inherent in such analysis that, from the Committee’s perspective, are most important to consider in making decisions. More specifically, the aggregate nature of net benefits or a cost-effectiveness ratio means that these measures by themselves cannot capture the distribution of benefits or of costs in a population. Aggregate estimates of QALY gains or cost-effectiveness ratios do not indicate the distribution of impacts over time or the magnitude of individual gains. Summed QALYs do not distinguish between health gains made within the course of a single life or across generations. Nor do they indicate whether the QALY gains represent small health improvements widely dispersed throughout a population or larger gains allocated among relatively few people. Summary benefit–cost and cost-effectiveness measures also omit benefits that are difficult to quantify. Such benefits include health and nonhealth effects for which numerical estimates are not available, for example, because the scientific research base is inadequate to support quantified estimates of impacts or because relevant monetary values or effectiveness measures have not been developed. Policy makers and the general public may also care about characteristics that are not captured in QALYs nor by many of the other valuation measures used in CEA or BCA, such as the degree to which the risk is observable or controllable, and whether the risk is especially dreaded. In this chapter, we first examine the ethical and normative2 assumptions implicit in the calculation of the cost-effectiveness ratio, including the 1 This position is reflected in the recommendations of previous expert panels and current Executive Office of the President guidance (U.S. Panel on Cost-Effectiveness in Health and Medicine, Gold et al., 1996b; Institute of Medicine Committee on Summary Measures of Population Health, IOM, 1998; Executive Order 12866, EOP, 1993; and Circular A-4, OMB, 2003a). 2 In this chapter, “normative” refers to a variety of value-based judgments and beliefs, including some that are not ethical or moral in nature. For example, the fact that some people consider death from cancer worse than death from other causes is by itself not an ethical concern, though it involves a value-based judgment that guides their behavior.
OCR for page 132
Valuing Health for Regulatory Cost-Effectiveness Analysis construction of the metric most commonly used in CEA to value health outcomes, the QALY. In the second and third sections of the chapter, we identify important normative and distributional concerns that are not reflected in cost-effectiveness ratios and that should be taken into account in regulatory decisions. In the fourth section, we discuss the importance of an accountable, public deliberative process to bring together information from economic analyses with information about the ethical, qualitative, and distributive aspects to craft regulatory policies. The final section summarizes the Committee’s conclusions. ETHICAL ASSUMPTIONS IN QALY-BASED CEA The QALY’s strengths as the effectiveness metric in CEA are largely practical ones. The QALY is well established; it has been applied in hundreds of studies over several decades, is supported by generic health assessment survey instruments, and allows morbidity and mortality information to be readily combined. However, the QALY incorporates certain ethical commitments and ignores others. This section identifies some of the ethical implications of using the QALY as a unit of measurement for valuing health outcomes in CEA. As discussed in Chapter 3, a QALY can be thought of in several different ways. Most simply, it is an index with an intuitive meaning, namely, an index that relates a particular state of impaired health (of a given duration) to some number of years in optimal health. More complex interpretations of the QALY, such as an index derived from utility theory or even as a direct measure of utility, are debatable, and are valid only under certain restrictive assumptions. When used in CEA for regulatory analysis, the QALY is probably best interpreted in its intuitive sense, as a measure of health improvement or production that facilitates comparisons with other opportunities for health gains. This pragmatic interpretation of the measure avoids the need to demonstrate that the QALY has particular properties consistent with the utility theory that underpins BCA and welfare economics. Valuing Life Years Compared with Valuing Lives Perhaps the most basic normative commitment entailed by using QALYs as an outcome measure in CEA is valuing some form of life years, rather than whole lives or preventable deaths. This move from treating all deaths equivalently to denominating losses and gains in terms of the extent of changes in longevity is illustrated by Table 4-1. Using lives as an impact measure assigns the same value to a preventable death regardless of whether the person is young, middle aged, or elderly, while the use of life years
OCR for page 133
Valuing Health for Regulatory Cost-Effectiveness Analysis TABLE 4-1 Lives, Life Years (LYs), and Quality-Adjusted Life Years (QALYs) Preventable Deaths LYsa LYs Discounted at 3%a LYs Discounted at 7%a QALYsb QALYs Discounted at 3%b QALYs Discounted at 7%b Age (in years) 5 1 73 29 14 65 27 13 35 1 44 24 14 37 21 12 75 1 12 10 7.9 9.1 7.6 6.1 Ratio of values by age 5/35 1 1.7 1.2 1.0 1.8 1.3 1.1 5/75 1 6.1 2.9 1.8 7.1 3.6 2.1 35/75 1 3.7 2.4 1.8 4.1 2.7 2.0 aBased on age-specific life expectancy for 2002 (NCHS, 2005). bBased on EQ-5D norms for the adult U.S. population (Hanmer et al., 2006); assumes HRQL is 1.0 for individuals through age 9, and midway between 1.0 and the value for persons age 20 for individuals ages 10 through 19. shows the difference in impact of preventable mortality on younger and older individuals. Furthermore, adjusting life years for health-related quality of life (HRQL) in the QALY calculation increases the difference between the impact estimates for a younger as compared with older person. Using QALYs gained, rather than deaths averted, as the measure of effectiveness in CEA has analogous implications in the BCA context. As discussed in Box 4-1, analyses in which monetized estimates of the value of preventable deaths vary by age have been highly controversial. Similarly, CEA using life years or QALYs gained as the effectiveness measure also appears to disadvantage older people, who have shorter average remaining life expectancies during which they can benefit from interventions. Some argue that shifting from valuing whole lives to any form of life years unfairly discriminates against older people, who may value the remainder of their lives as highly as do younger people. Similar arguments can be put forward for people with life expectancies shortened by their socioeconomic status or preexisting health conditions or disabilities. A countering perspective is provided by those who hold that everyone should be given an equal chance to have a “normal” or full lifespan and that averting deaths among those who have achieved a normal lifespan should not count for as much as averting deaths among much younger persons. One implication of this view is that measuring life-preserving gains in terms of years of life does not unfairly disadvantage older people relative to younger people (Harris, 1987; Williams, 1997). In addition to this nor-
OCR for page 134
Valuing Health for Regulatory Cost-Effectiveness Analysis BOX 4-1 The “Senior Discount” Controversy Analytic approaches that assign a lower value to premature mortality among the elderly have been the subject of heated debate in the context of BCA. In particular, the Environmental Protection Agency’s use of estimates that reflected the value of remaining life years in a sensitivity analysis of air pollution-related policies led to a significant public outcry (Skrzycki, 2003). Based on a survey of older adults’ willingness to pay for remaining life years, the analysis placed a lower value on premature mortality among the elderly (referred to as the “senior discount” in the subsequent policy debates). The controversy led the Office of Management and Budget (OMB) to issue a memorandum requiring agencies to avoid age adjustments (Graham, 2003a). OMB Circular A-4, which extends the guidance to CEA, amends the instructions regarding age adjustments (OMB, 2003a). The Circular notes that population averages, rather than values reflecting differences among subgroups, should be used in assessing both health-related quality of life and life expectancy to support the perceived fairness of the analytic approach. mative argument, survey results suggest that, in the abstract, people judge saving the lives of younger as compared with older persons more important (Cropper et al., 1994). Presenting the results of a CEA in several forms, using deaths averted, life years extended, and QALYs gained as alternatives, is one way to broaden perspectives on gains in life expectancy. In addition, reporting disaggregated estimates of regulatory impacts by key age and population characteristics—such as income, race, gender, or other factors relevant to the particular intervention—also increases the transparency of the justification for and implications of the regulatory action. Both strategies facilitate ethical deliberation. A QALY Is a QALY Is a QALY In contrast with the willingness-to-pay measures used in BCA, which may be affected by wealth and hence may vary depending on individual financial resources, by construction the value of a QALY is assumed not to vary with income. Although the relationship between willingness to pay and wealth is complex and depends on the characteristics of the good or service (see Freeman, 2003), the fact of this relationship can lead to concerns that the BCA results will be weighted toward the interests of wealthier members of society. In practice, regulatory analysts generally use willingness-to-pay values or ranges of values that reflect averages from the
OCR for page 135
Valuing Health for Regulatory Cost-Effectiveness Analysis relevant research, rather than assigning different values to health risk reductions affecting richer and poorer people. QALYs value an improvement in health of a given magnitude the same regardless of the characteristics of the person experiencing the improvement. They ignore the relative economic standing of affected populations in representing the value of health gains and reflect an ethical commitment to the income-independence of health as a societal value. In addition, each QALY unit is of equal value in all contexts. Regardless of the individual to whom a QALY accrues, the states of health preceding or following the period in question, how widely health gains are distributed within a population, how impaired one beneficiary of a health gain is relative to another, or how a given impairment affects the lives of different persons, a QALY always carries the same value. Other health metrics, such as the healthy year equivalent, the saved young life equivalent, and the age-weighted disability-adjusted life year, each convey a different aspect of the distribution of health gains that the QALY omits.3 However, none of these alternative HALY metrics are superior to the QALY in practice. No single metric can reflect all significant aspects of particular sorts of gains in health and longevity. The QALY, like any construct, imperfectly captures all aspects of what we value in good health. We know of no weighting scheme that is able to accurately reflect the full range of societal values relevant to regulatory decisions; in reality, these weights may vary depending on the particular decision-making context. Presenting supplementary information about the size and characteristics of the exposed population, the per capita magnitude of the risk, and the distribution of expected benefits will help respond to these concerns. The Source of HRQL Values The question of perspective in valuing HRQL has several dimensions. One has to do with the source of relative health state values, that is, whether they come from the general population, patients or persons experienced with the health state in question, or experts such as clinicians. The other dimension concerns the method for eliciting HRQL values and, more specifically, whether those values reflect individual preferences for one’s own health or preferences for societal investments in health more broadly. Whose Values Count? The appropriate evaluative standpoint from which to determine the relative values of different health states, conditions, and disabilities in CEA 3 These HALY measures are discussed in Chapter 3.
OCR for page 136
Valuing Health for Regulatory Cost-Effectiveness Analysis depends on the context of the decisions that the analysis is intended to inform. In clinical studies to compare the effectiveness of alternative treatments, for example, the values of patients may be most appropriate. In contrast, for societal decisions about resource allocation in health care settings, community values (i.e., the aggregated and averaged judgments of a representative sample of individuals in the general population) regarding the relative desirability of different states of health should be used (Gold et al., 1996b). Because most economically significant regulations for which CEA is required affect all segments of society in terms of their costs and/or benefits, valuations (both for directly elicited values for specific health outcomes and valuation surveys underlying generic HRQL indexes) generally should be based on representative samples of the general U.S. population. Analysts will need to consider this issue in the context of individual regulations, however, because some rules disproportionately affect certain subgroups. For example, a regulation might impose costs and provide benefits primarily for elderly people or for residents of a single geographic area. In this case, values based on the affected subpopulation would be preferable. If obtaining subpopulation values is not feasible, it would be important to conduct uncertainty analysis on the possible differences in valuation. Little is known about the differences in health state valuation across sociodemographic subpopulations in the United States with the notable exception of the recent U.S. EuroQoL-5D (EQ-5D) valuation survey (Shaw et al., 2005). This survey oversampled the two largest minority populations, Hispanics and non-Hispanic blacks, so that reliable subpopulation estimates could be calculated. A related analysis that compared U.K. and U.S. results for these two large and methodologically consistent EQ-5D valuation surveys found significant differences in the values for particular health states (Johnson et al., 2005). These differences were not constant or systematic across health states; those characterized by severe problems had the largest discrepancies, with U.S. valuations exceeding the U.K. valuations. Although these findings suggest that it is important that the health state index values be derived from a population comparable to the one of interest, it may turn out to be less of an issue in practice than in the abstract. As illustrated by the case studies, regulatory analysis involves comparing health status with and without the condition of interest. More research would be needed to determine whether such estimates of changes in health status are as dependent on the population surveyed as are the estimates for particular conditions (Franks et al., 2006). Some research shows that preferences for particular health states are quite similar for different groups of people, when patient valuations are compared with those of nonpatients, and when the valuations of different
OCR for page 137
Valuing Health for Regulatory Cost-Effectiveness Analysis socioeconomic and ethnic groups are compared (Kaplan and Bush, 1982; Llewellyn-Thomas et al., 1982, 1993; Balaban et al., 1986). Other studies, however, suggest that people who have experienced a disease or disabling condition will tend to value that state more highly than those with no experience (Sackett and Torrance, 1978; Najman and Levine, 1981; Slevin et al., 1990).4 One possible explanation for this latter finding is that those with the condition or impairment have made adjustments or adaptations that result in lesser losses in HRQL than are anticipated prior to any illness or disability.5 Discrepancies in valuations of impaired or disabled health states between a general population and those with experience of the condition have led some to challenge the validity of general population valuations, arguing that they are uninformed and potentially discriminatory. Recent research findings in psychology and behavioral economics suggest that people incorrectly predict the impact of changes in their circumstances on their sense of well-being (Kahneman et al., 1997; Wilson et al., 2001; Gilbert and Ebert, 2002; Riis et al., 2005). If health state index values are intended to represent the relative effects of different conditions on people’s lives rather than reflecting apprehensions and prejudices about those conditions, then values elicited from people lacking knowledge about the conditions may be biased. Furthermore, people with disabilities and disability advocacy groups have objected to being “assigned” lower HRQL values than are those without disability, on the grounds that the practice leads to the devaluation of persons living with disabilities and to perpetuation of stigma associated with particular conditions (Wang, 1992; Silvers, 1996). If a health condition or disability—human immunodeficiency virus disease or paraplegia, for example—is valued lower by the general public because it is stigmatizing, using this valuation in a public policy analysis may reinforce, and be taken as condoning, such prejudice. Wasserman and Asch (2004) have suggested an alternative way to establish the relative values of preventing particular injuries and impairments. They propose that relative values should be based on the costs of restoring capacities and improving various aspects of quality of life for those who are impaired. For example, facilitative and adaptive technologies for persons with speech or writing disabilities can restore communications 4 This valuation question must be distinguished from the task of characterizing the experience of a particular health condition or disability according to a multidimensional generic survey instrument. The descriptive task is always best carried out by those who are familiar with the condition, as discussed in Chapter 3. 5 Such adaptation is not only psychological; it can involve substantial investments in rehabilitation, personal assistance, assistive technologies, and physical accommodations in the built environment (IOM, 1997).
OCR for page 138
Valuing Health for Regulatory Cost-Effectiveness Analysis abilities. The cost of providing someone who has lost capabilities through injury with the equipment and services that restore functioning could be used to establish an upper bound on the cost of that injury or impairment, relative to unimpaired health. Such investments could well exceed what is now spent on rehabilitative services for some conditions. Thus this approach to valuation is more demanding than a cost-of-illness estimate based on historical spending. Initial estimates (for the United Kingdom) of the costs of restoring capacities have been made for some conditions (Smith et al., 2004). We recognize that this strategy is not practicable in the near term, and would likely be feasible only for some kinds of disabling conditions. In concept, however, it offers an alternative to preference-based measures that avoids the apparent devaluing of lives spent with impairments. The use of general population valuations does not necessarily result in material disadvantage for those with impairments or disabilities (Gold et al., 1996b). Because the perspective is ex ante in regulatory analysis and it cannot be known with certainty who ultimately will be affected, the values of those potentially affected (represented by the general population for most economically significant health and safety regulations) are appropriate in this context. This perspective takes account of the loss of capacities and opportunities that attend illness and injury. It reflects the societal value accorded having greater rather than lesser capacities and more rather than fewer opportunities. If a subset of the general population receives the benefits and pays the costs of the regulation, then the values of this subpopulation should be used. In particular, there may be instances where the costs and benefits of a rule predominantly affect people with a preexisting illness or disability addressed by the regulation. In this case, the affected population is the same as the patient population, and patient values would be appropriate. Later in the chapter we consider the circumstances under which the lesser values placed on health improvements among those with impaired health or disabilities can be ethically problematic. Individual Preferences and Societal Values As noted in Chapter 3, empirical research suggests that each elicitation technique—standard gamble, time trade-off, category rating, and person trade-off (PTO)—produces somewhat different relative values for states of health. One distinction among these four elicitation methods is that the first three techniques query individual preferences, while the PTO method is socially oriented. PTO exercises ask for judgments about the equivalence of health improvements and life extensions for groups of people who differ in their states of health, age, and other relevant characteristics (Richardson and Nord, 1997; Nord, 1999). Respondents are asked “to compare the
OCR for page 139
Valuing Health for Regulatory Cost-Effectiveness Analysis relative benefits of treating different conditions in a context of comparing equivalent numbers needing to be treated to produce equal social benefits” (Ubel et al., 1998, p. 43). The results of PTO exercises suggest that values other than the maximization of potential aggregate health benefits, as measured by conventional QALYs, affect decisions to allocate health improvements among groups. Allocation choices using PTO tend to give greater weight to improving the health of more severely impaired groups relative to those with lesser problems (Nord, 1999). PTO elicitations also tend to distribute potential health gains more widely among groups able to benefit (Menzel et al., 1999). Because of the broad and essentially social nature of health and safety regulation, the PTO method may be particularly appropriate for valuing health outcomes in this context. As noted in Chapter 3, much work remains to be done, however, to develop the PTO technique and the empirical base from which to estimate values. The Committee concludes that research to develop better approaches to societal valuation for regulatory CEA is warranted. Combining Morbidity and Mortality in a Single Measure As detailed in Chapter 3, QALYs combine information about changes in survival and morbidity in a way that reflects individuals’ preferences for trade-offs between longevity and quality of life. This fundamental property of the QALY mirrors the actual situation of patients who face medical treatment choices that involve a risk of death. However, when QALYs are also used to compare very different health-related interventions, this framework may not mirror the actual choice as closely. Analyses of regulatory interventions are likely to encompass broader arrays of health-related effects for large population groups. CEA will at times involve the aggregation and synthesis of many health outcomes and of impacts across diverse groups of people, as illustrated in the examples in Chapter 2. In addition, evaluating the health impacts of regulatory interventions requires combining the benefits of increased length of life and improved quality of life for a population, rather than trading off longevity against improved quality of life for a given individual. When health outcomes are aggregated and averaged across diverse conditions and populations, a single summary measure will mask disparities in impacts among age or other population groups. For example, in the Environmental Protection Agency (EPA) analysis that was the subject of the air quality case study, the summary measure masks the range of impacts on the very old, the very young, and those with preexisting conditions. Thus EPA presents disaggregated results for mortality and morbidity and for different age groups (see Table 2-4). The Committee’s recommendations,
OCR for page 140
Valuing Health for Regulatory Cost-Effectiveness Analysis presented in the final chapter, address the need for reporting disaggregated analytic results in regulatory analyses. Summary This section briefly reviewed the ethical assumptions embedded in QALY-based CEA. Using QALYs in regulatory analysis widens the application of this analytic tool and introduces a complex normative construct to a new audience. Understanding what the QALY does and does not reflect in the measurement of health effects should help regulatory analysts and policy makers interpret and communicate their analytic results. In particular, analysts should keep in mind the need to present information on the nature of the individual health effects and the characteristics of affected population groups because QALYs subsume these distinctions. The societal perspective of regulatory analysis is best reflected by valuation of health states and conditions by people affected by the regulation. At the same time, it is important to keep in mind the potential biases in the valuation of some health states due to unfamiliarity, lack of experience, or because the states carry stigma. The rationale for using health state values elicited from community-based sample surveys in regulatory analysis is to reflect the preferences and values of the population likely to receive the benefits and/or bear the costs of the intervention. ETHICAL AND OTHER IMPLICATIONS OF RISKS AND OF INTERVENTIONS TO ADDRESS RISKS In this section we consider the ethical, distributional, and other factors relevant to decisions about regulating health and safety risks that are not captured in CEA. Dimensions of Value Affecting the Acceptability of Risks Not all kinds of risks are the same. Risks may differ in ways that can affect their acceptability for individuals and for society as a whole, as well as their assessment from an ethical point of view. How government agencies should address risks that differ in kind and in acceptability (both to individuals and the larger community) is a question that can be addressed only as part of broad, public, and deliberative discussions. Regardless of whether the value of risk reductions is measured by cases averted, willingness to pay for health improvements, or QALY gains, the measure is likely to exclude some aspects of the risk reductions that are valued by society. For example, a 1-in-100,000 reduction in the risk of death may be valued differently depending on the source of the risk; society
OCR for page 148
Valuing Health for Regulatory Cost-Effectiveness Analysis These guidance documents reflect the exceptional value society places on children’s lives and well-being, as well as recognizing their particular susceptibility to serious and long-lasting harm from health and safety risks. Table 4-1 illustrates how the use of life years or QALYs, instead of preventable deaths, weights the health impacts for children more heavily, relative to impacts on older persons. As discussed in Chapter 3, the difficulties of valuing children’s health outcomes are both empirical and conceptual. Setting aside the practical problems of measuring HRQL for children, the concern remains that an individually based HRQL metric does not fully capture the high value placed on children’s well-being and health by parents and society. Although some have suggested that HRQL measurement should encompass the effects of an illness, injury, or disability on the entire family in which it occurs, this demanding approach has not been implemented. Population Health Data and Subgroups Another potential problem for valuation is that the major population health surveys exclude some subpopulations of concern. As discussed in Chapter 3, the National Health Interview Survey and the Medical Expenditure Panel Survey are household surveys, limited to noninstitutionalized, civilian populations. By design, excluded populations are homeless people; those who are migrant or have no fixed residence; and persons in prisons, group homes, nursing homes, and other institutions (Meyers and Andresen, 2000). Undocumented aliens, migrant farm workers, and others with reasons to avoid contact with government officials or data collection activities are unlikely to be represented in survey samples. In addition, members of these groups may be particularly susceptible to certain kinds of risks targeted by regulations, such as pesticide controls and workplace safety practices. The exclusion of the groups just mentioned from routine population health surveys is also a problem for the valuation surveys underlying generic HRQL indexes, and calls into question the extent to which they can be assumed to represent accurately the values of the general population. The significance of this omission depends on whether the excluded groups represent a large enough percentage of the population to affect the survey results, and whether the values of excluded groups differ to a significant degree from the values held by those included in the survey. Although the recently conducted U.S. valuation survey for the EQ-5D used a stratified sample designed to include the three largest racial/ethnic groups in sufficient numbers to provide disaggregate results for whites, Hispanics, and non-Hispanic blacks, disaggregated results were not presented in the initial publications from this survey (Luo et al., 2005; Shaw et al., 2005).
OCR for page 149
Valuing Health for Regulatory Cost-Effectiveness Analysis Another issue with respect to low-income populations and some minority populations, such as blacks and Native Americans, is that their life expectancy is lower and their HRQL worse than that of others in the population (NCHS, 2004). If health gains are calculated for such subgroups based on subgroup population health status and longevity norms, then the potential to benefit from life extensions will be proportionately lower for these subgroups. Calculation of Health Gains Some health risks subject to regulation disproportionately affect those whose health is impaired already. In the case of air quality interventions, for example, the elderly and those with preexisting cardiac or pulmonary conditions are more likely to suffer adverse health effects. One of the most difficult issues to address is whether and how to disaggregate the general population in calculating gains in health due to a regulatory intervention. Both OMB guidance and the PCEHM’s recommendations for the reference case CEA direct the use of general population averages rather than health state index value estimates for subpopulations. (These requirements are discussed in more detail in Chapters 1 and 2; see Appendix C for the relevant text of Circular A-4.) The implications of these requirements for regulatory analysis depend on whether the general population is in fact representative of the population that achieves the health gains. For the types of economically significant health and safety regulations addressed by this report, the population affected will often reflect the same distribution of preexisting disabilities or health impairments as the general population. In such cases, using general population averages is analytically correct and will not disadvantage those who are disabled or in impaired health. The OMB and PCEHM requirements are more problematic in a case where the affected population does not reflect the same distribution of preexisting disabilities or health impairments as the general population. For example, if individuals with heart disease represent 10 percent of the general population but 50 percent of the population affected by the regulation, using population averages may not accurately capture the QALY gains attributable to the rule. In the air pollution example, reductions in preventable mortality may predominantly affect individuals with preexisting heart or respiratory conditions. Such deaths occur primarily among elderly people, and population averages for the affected age groups include a relatively high rate of heart and respiratory disease. EPA concluded that, because both the general population and the affected population in these age groups have comparably high rates of these preexisting conditions, the use of population averages
OCR for page 150
Valuing Health for Regulatory Cost-Effectiveness Analysis may provide a reasonable best estimate of the impacts of its rule. However, this conclusion is the subject of debate and scientific uncertainty. It is also important to remember that the population affected undergoes changes in health status over time. Rules tend to remain in place for extended periods, however. Thus people who are not affected at the outset may develop conditions later such that a regulatory intervention is especially beneficial. In cases where the health impairment or disability is related to the risk or intervention of interest, QALYs could unfairly value life-extending interventions for people with chronic illness or disability. For example, suppose decision makers are comparing two regulations that are equally costly, one of which affects only individuals with preexisting disabilities and another that affects individuals in better health. Also assume that the first intervention, which extends for 10 years the lives of 100 people with a chronic condition or disability valued at 0.75 (on a scale where 1.0 corresponds to optimal health and 0 corresponds to death), would produce 750 QALYs. If the second intervention extends for 10 years the lives of 100 people in near optimal health—0.95, for example—the gain would be 950 QALYs. In this case, focusing solely on the QALY gains would lead decision makers to select the second intervention, even though it extends the same number of lives as the first. Hence the use of QALYs for evaluating and prioritizing life-saving interventions appears to discriminate against people with impaired health or disabilities by assigning less value to extending their lives simply because of their disability. The reduction of average HRQL that occurs with increasing age produces the same general effect in comparisons between life extensions among 20-year-olds and 70-year-olds. An alternative to assessing QALY gains based on comparison to actual health status is an approach that assumes that affected individuals would be in optimal health as a result of the intervention. As discussed in Chapter 2 (see Box 2-5), EPA recently presented QALY-based results that do not adjust life years gained to reflect the less-than-optimal HRQL that would be expected during those additional years of life (EPA, 2005a, Appendix G). Instead, EPA calculated health gains due to averted mortality as life years spent in optimal health. In our case study of air quality improvements, we followed a different practice. We estimated the gain in QALYs due to increased life expectancy based on average health state values for the general population in each age group assessed. This approach assumes that, in the absence of the regulation-related risks, individuals would face the same degree of impairment as the average member of the U.S. population of the same age. The Committee concludes that EPA’s practice—which essentially gives greater weight to QALY gains from life extensions than from HRQL improvements—was less transparent than the alternative, namely to calculate all
OCR for page 151
Valuing Health for Regulatory Cost-Effectiveness Analysis QALY gains the same way regardless of whether morbidity or mortality is affected and to present those gains in disaggregated form so that the differences in the types of impacts are apparent. Appendix A provides additional discussion and presents the results of the Committee’s case study analysis. The Treatment of Future Generations in CEA Although many regulations have the potential to affect future generations, those where the costs are incurred primarily in the near term but the benefits occur largely in the future (or vice versa) pose particular ethical issues, especially if the effects of the policy are not easily reversible. Such is the case, for example, with regulations governing the construction of nuclear waste repositories. Construction may impose costs on the current generation, whereas future generations may be affected by the release of radiation if safeguards fail. Contaminants with reproductive or developmental effects provide other examples; controlling exposures among members of the current generation will benefit the subsequent generation. Assessing the impacts of these types of regulations poses analytic as well as ethical challenges regardless of whether BCA or CEA is used to estimate costs and benefits. Future Effects Perhaps the biggest issue associated with rulemakings relates to the ability to predict future conditions with and without the regulation. This problem pervades all aspects of the analysis. For example, Harrington et al. (2000) compared the predicted and actual costs of several regulations, and found that one of the key factors leading to overestimates of future costs was the difficulty inherent in predicting technological innovation. Such innovations may affect the benefits of a rule as well as its costs. The regulatory analyses reviewed by the Committee applied varying approaches to addressing this problem (see Robinson, 2004, and Appendix A). For example, in the Food and Drug Administration’s (FDA’s) analysis of its juice processing rule, the agency assumed that current conditions remained constant, so that both costs and benefits were the same in each future year (FDA, 1998, 2001). In addition to presenting this annual value, FDA calculated the present value of costs and benefits over an infinite time horizon. In contrast, in its analysis of emission controls for nonroad diesel engines, EPA limited the time period addressed to 20 years and presented costs and benefits on an annual basis as well as in present value terms (EPA, 2004b). These estimates took into account the phase-in of regulatory requirements as well as predicted changes over time in pollutant emissions and in the demographic characteristics of the affected population. As re-
OCR for page 152
Valuing Health for Regulatory Cost-Effectiveness Analysis quired by OMB (2003a), regulatory analyses generally also present the timing of the undiscounted impacts. This presentation of both discounted and undiscounted results is particularly important when costs and benefits are widely separated in time. If there is a lag between the costs and the effects, the estimates of cost-effectiveness will vary depending on the length of time that elapses as well as the discount rate used, regardless of whether the benefits are measured in dollars, QALYs, life years, or cases avoided.7 For example, if both costs and benefits are discounted, two regulatory options that differ only in the year in which benefits occur will have different present values if a positive discount rate is used. The regulatory option with the nearer term benefits will appear to be more cost effective. This outcome is derived from the underlying rationale for discounting, which reflects a general preference to receive benefits soon and delay costs.8 Table 4-2 presents this issue in simplified form. As indicated in the table, the option without a lag between costs and benefits will be more cost-effective in present value terms when compared to another option with equivalent, but more delayed, benefits.9 This difference in present values increases as the discount rate increases.10 Future Generations When risks are imposed or benefits accrue in the distant future, the ethical concerns and issues related to discounting are more difficult and less satisfactorily addressed. Moral obligations to future generations should be considered separately from the question of discounting practices.11 Present- 7 In the majority of rules considered in the Committee’s review of current practices (Robinson, 2004), costs and reduced incidence of illness, injury, or death occur in the same or relatively proximate time periods. FDA’s analysis indicates that the reduction in the incidence of illness is likely to occur in the same year as the reduction in juice contamination. EPA makes the same assumption for changes in the incidence of the nonfatal effects of nonroad engine diesel emissions, while indicating that preventable mortality is distributed over a 5-year period after exposure. 8 For example, most individuals generally would prefer to receive money today rather than at a later date because they can invest the money and earn interest. The present (discounted) value today of $100 received in a future year (t) is the amount that one would need to invest today to yield $100 in year (t). 9 For simplicity, Table 4-2 assumes that the QALY losses all occur in a single year. However, for most chronic illnesses and for preventable mortality, a change in incidence in the current year will have future year effects, and these future year effects will also be discounted. 10 See Portney and Weyant (1999), especially the essay by Weitzman (1999) for discussions of the interaction of the discount rate and the time period over which the discounting occurs. 11 See, for example, “On Discounting Regulatory Benefits: Risk, Money, and Intergenerational Equity” (Sunstein and Rowell, 2005) for a discussion of this issue.
OCR for page 153
Valuing Health for Regulatory Cost-Effectiveness Analysis TABLE 4-2 Discounting and Timing of Impacts Time Period Regulatory Option 1 Regulatory Option 2 UNDISCOUNTED RESULTS Year 0 Costs = $100 million; benefits = 400 QALYs Costs = $100 million Year 10 — — Year 20 — — Year 30 — Benefits = 400 QALYs Cost per QALY $100 million / 400 QALYs = $250,000 per QALY $100 million/400 QALYs = $250,000 per QALY RESULTS DISCOUNTED AT 3 PERCENT Present value in year 0 Costs = $100 million; benefits = 400 QALYs Costs = $100 million; benefits = 165 QALYs Cost per QALY $100 million/400 QALYs = $250,000 per QALY $100 million/165 QALYs = $610,000 per QALY RESULTS DISCOUNTED AT 7 PERCENT Present value in year 0 Costs = $100 million; benefits = 400 QALYs Costs = $100 million; benefits = 53 QALYs Cost per QALY $100 million/400 QALYs = $250,000 per QALY $100 million/53 QALYs = $1.9 million per QALY NOTES: For simplicity, this example assumes all the quality-adjusted life year (QALY) impacts occur in a single year and ignores the lifetime effects of chronic illness as well as the life years lost to premature mortality. It also does not provide information on the uncertainty in the estimates. All estimates are rounded to two significant digits. ing undiscounted impacts, and their timing, along with a discussion of impacts on future generations, as OMB (2003a) advises, allows decision makers to identify situations where concerns about long-term impacts suggest that decisions should not be based simply on the discounted present value of the results.12 Such presentation is necessary because otherwise, discounting may lead the present generation to impose extremely high costs on future generations, resulting in undesirable welfare losses as well as inequities between generations (Revesz, 1999). In addition, discounting 12 Chapter 2 discusses other aspects of the OMB guidance on discounting, such as the selection of the appropriate discount rate and the need for sensitivity analysis. Circular A-4 also describes the rationale for discounting nonmonetary as well as monetary measures of benefits in regulatory analysis (see Appendix C). In the context of health and medicine, Gold et al. (1996b) provide a detailed discussion of discounting, and recommend discounting both costs and benefits at 3 percent in the reference case and conducting sensitivity analyses using rates ranging from 0 to 7 percent.
OCR for page 154
Valuing Health for Regulatory Cost-Effectiveness Analysis could give an undesirable priority to programs that would produce benefits more rapidly, but with substantially less overall improvement in health, when compared to programs that produce benefits later, but with substantially more overall improvement. At the same time, a failure to discount could impose significant burdens on the present generation that regulatory interventions would alleviate. If regulators discount costs at a positive rate but value lives saved now and lives saved later equally, then the analysis paradoxically indicates that lifesaving spending should be postponed indefinitely, because the net benefit becomes increasingly favorable into the future (Keeler and Cretin, 1983). Furthermore, because future generations can reasonably be expected to inherit a richer and more technologically advanced world, there may be less reason to protect future generations from present choices (Weitzman, 1999). Others have suggested that the discounting of benefits to future generations might be thought of as part of a mutually beneficial intergenerational trade or contract (Lind, 1982). How future benefits and harms (costs) are viewed is likely to depend on the perspective adopted. For example, parents will likely take a more precautionary attitude toward protecting the world their children and grandchildren will inherit than might people unaffiliated with younger generations. Representing the interests of future generations in current policy discussions is difficult but ethically obligatory. Future generations will be affected by current decisions, particularly if the consequences are not easily reversed. As those involved in such discussions consider the future effects of their choices, they should factor in the implications of their decisions for those who will live in the future. Alternative normative frameworks—the “just savings” principle of Rawls’ social contract theory (1971, 1993), tort law, and utilitarianism—each can support a principle of compensation to guide discussion about the mix of benefits and costs that the present generation bequeaths to future ones (Sunstein and Rowell, 2005). Comparing Cost-Effectiveness Ratios The assumption underlying the use of CEA in regulation is that resources should be used to maximize the aggregate health status, or to minimize disease burdens, of a population. Some have suggested ranking regulatory programs from the lowest cost-per-QALY ratio to the highest, in order to identify better or more efficient investments in health production. Hahn (2005) has argued in favor of the use of such summary rankings, which he calls “regulatory scorecards” and which OMB has described as “league tables” (EOP, 2002). Although such scorecards enable compari-
OCR for page 155
Valuing Health for Regulatory Cost-Effectiveness Analysis sons across widely different interventions and provide useful information, they can mislead (Parker, 2003). Given the many relevant features of decisions about the regulation of health and safety risks that are not part of the quantified economic analyses, and given considerable differences in the methodologies used to generate the summary results, the rankings of cost-effectiveness ratios are ambiguous. Furthermore, as discussed in Chapter 2, the legislative mandates and requirements for regulation vary across programs and agencies, making such comparisons less meaningful. Whether or not such cross-programmatic and interagency comparisons of CEAs might be helpful to decision makers, without being misleading, remains an open question. The Committee recommends against using summary rankings as the principal basis for policy decisions because the substance and methods of economic analysis do not support unqualified comparisons across widely different contexts. IMPROVING REGULATORY DECISION MAKING An important adjunct to the sorts of improvements in regulatory analyses discussed above is to strengthen the regulatory decision-making process itself. Such strengthening would involve greater transparency and ensuring a deliberative policy process that incorporates nonquantified information, including consideration of the distributive and ethical features of a proposed regulatory action. We discuss two fundamentally different strategies for introducing societal values and equity considerations into public policy decisions. One strategy is to incorporate information about distributive priorities directly into the CEA. This could either involve weighting health state index values to reflect priorities or stipulating values in the calculation of health-related effects. The other strategy is to pair the quantified economic analysis with qualitative information presented in a transparent and open process of regulatory development. The two approaches could also be combined. Several approaches to societal weighting of health state index values have been proposed. First, standard index values could be modified with numerical factors or weights that convey priorities for age groups, severity of condition, or particularly vulnerable groups. These weights could be estimated by asking a representative sample of the general population to make PTOs between health improvements that are equal in terms of conventional index value gains, but different in terms of the characteristics of the people whose health is improved (Nord et al., 1999; Ubel et al., 2000). A variant on this approach would transform the health state index values into values that reflect societal values for giving priority to the worst off, which could be done by compressing the values of less severely impaired
OCR for page 156
Valuing Health for Regulatory Cost-Effectiveness Analysis health states toward the upper end of the 0-to-1 scale (Nord, 2001). By locating moderate health states closer to the upper end of the HRQL scale, the value of improvements for the moderately ill is reduced relative to improvements for the severely ill. Another approach to building equity considerations directly into the cost-effectiveness ratio is to value all reductions in preventable mortality at 1.0, rather than at the postregulatory health state index value that is actually expected to pertain. This is the approach EPA adopted in its pilot CEA in the Clean Air Interstate rule (EPA, 2005a), as described earlier in this chapter and more fully in Box 2-5. Despite their apparent usefulness and appeal in combining distributive concerns with health production, formula-based approaches to incorporating societal values into CEA calculations are problematic. First, there is no consensus as to how equity weights should be calculated, or even whether their use is appropriate. It is also difficult to adjust health state index values for more than one dimension; should that adjustment be for age, severity of condition, or initial health status? Second, valuing all gains in longevity as life years in optimal health, as with EPA’s Morbidity-Inclusive Life Year approach, changes the conventional relationship between morbidity and mortality effects and could lead to social choices that violate individual preferences in choices between quality and quantity of life (Johannesson, 2001). Finally, building equity considerations into the quantitative analysis in any of these forms makes the cost-effectiveness ratio less transparent, and therefore potentially more confusing and ambiguous for some. In light of these concerns with adjusting health state index values to reflect distributional considerations, the Committee endorses a different strategy. In our view, standardizing the presentation of quantified analyses and their data inputs, assumptions, and methods offers the best chance for informed and transparent regulatory decision making. Presenting economic analyses in a common format and informing the deliberative process with alternative analyses helps to demonstrate how quantified results depend on value assumptions. Although we do not recommend that the CEA calculations be adjusted to incorporate distributional concerns quantitatively, we recognize that agencies might want to develop supplementary analyses using other measures and weighting schemes as sensitivity analyses. Such alternative quantifications could help to clarify the different implications of different regulatory strategies. By including distributional and normative considerations in a public, transparent, and deliberative decision process, distinct concerns can remain separate. For example, how much should the fact that an auto safety requirement affects children count in judging the acceptability of its costs? A public and deliberative policy-making process permits the airing of reasonable disagreements about various priorities, rather than em-
OCR for page 157
Valuing Health for Regulatory Cost-Effectiveness Analysis bedding one version of them in the CEA calculations. A fair and transparent process can resolve open questions of value in ways that achieve and maintain legitimacy. Daniels and Sabin (1997, 2002) have characterized a fair process for decision making about health and health care as having certain central requirements or features. In the following summary, the Committee adapts these conditions for a fair process to the regulatory context. Publicity: The regulatory development process should be transparent and involve publicly available rationales for decisions affecting health and longevity. People have a basic interest in knowing the grounds for decisions that fundamentally affect their well-being. Relevance: Those who are affected by regulatory decisions, including those who bear the costs of regulations as well as those who realize the benefits, must agree that the rationales rest on relevant reasons, principles, and evidence. Revisability and Appeals: The regulatory process should make provisions for revisiting and revising decisions in light of new evidence and arguments. Enforcement: There should be a mechanism for ensuring that the previous three conditions are met. These conditions hold decision makers accountable for the reasonableness of their choices in regulating health and safety risks. Decisions that meet these conditions provide a form of “case law” that helps make future reasoning more coherent. Many of the issues underlying regulatory interventions, both matters of fact and of values, are points of disagreement. A fair and transparent process of this sort adds legitimacy to the results. It also contributes to societal learning about the appropriate grounds for making the kinds of trade-offs involved and thus enhances broader democratic processes over time. The demand for fair process is a fundamental part of our political system. It is embedded in the statutory and administrative requirements for regulating risks, as discussed in Chapters 1 and 2. Further progress towards the goals of fair and transparent risk regulation is possible. CONCLUSIONS The Committee’s key conclusions based on the discussion in this chapter follow. CEA and BCA alike provide a useful but incomplete basis for informed societal decisions about reducing risks to human health and safety through
OCR for page 158
Valuing Health for Regulatory Cost-Effectiveness Analysis regulation. The most feasible and desirable way to account for ethical and normative considerations in regulatory policy is to include them explicitly in the deliberative policy-making process. The choice of QALYs as the basis for measuring the production of health through regulatory interventions entails certain value commitments and ignores others, and these limitations should be made explicit in regulatory analysis. While some societal values regarding the distribution of health benefits could be incorporated through quantitative modifications of health state values, such adjustments are of questionable validity and make the quantification of health improvements more difficult to interpret. However, presenting the quantitative results of such alternative measures as sensitivity analyses may help to highlight those distributive implications in a way that promotes consideration of them in the deliberative process. Presenting the components of summary economic analyses individually is an important contribution to the transparency and accountability of regulatory decisions because such disaggregated information may be easier to understand and it conveys the relative contributions of various health impacts to the summary results. Public participation in the development of regulatory priorities and specific regulations is vital to well-informed policy making. Existing administrative procedures that govern the issuance of regulations provide a framework for publicity, transparency, public involvement, and accountability. They do not guarantee adequate citizen participation in setting regulatory agendas and rulemaking, however. Greater public understanding of the environmental, health, and safety risks and the benefits and costs of strategies to mitigate such risks can be promoted by well-conducted and clearly presented regulatory impact analyses. The next and final chapter presents the Committee’s recommendations for regulatory analysis and policy development. Our recommendations reflect the conclusions above, as well as discussions and evidence that appeared earlier in this report.
Representative terms from entire chapter: