Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
6 Overall Conclusions and Recommendations The committeeâs statement of task (see Chapter 1) presents specific ques- tions that were addressed by the committee in Chapters 3-5. This chapter dis- cusses the overall conclusions and recommendations developed in the earlier chapters in the context of the regulatory benefits-assessment process. It also discusses implications for future Environmental Protection Agency (EPA) regu- latory impact analyses (RIAs) that include changes in ambient ozone concentra- tion.1 Part of the committeeâs charge was to evaluate recent analyses of epide- miologic studies that found a modest but consistent relationship between short- term ozone exposure and premature mortality. During its deliberations, the committee was mindful of the information needs and framework for benefits assessment. There is a fundamental difference between information needed for regulatory benefits assessment and information needed to set a protective health standard. Selection of a health-based standard focuses on the lowest ambient concentration that poses a risk of adverse health effects in the most sensitive population, whereas benefits assessment uses information to estimate all the health-risk reductions in the entire population that is expected to experience a change in ambient concentrations. OZONE MORTALITY EFFECT In carrying out its charge, the committee considered not only the recent epidemiologic evidence, but also toxicologic and pathophysiologic evidence that points to mechanisms by which ozone may contribute to premature mortality. As a gas, ozone is highly reactive, and once inhaled it is immediately engaged in the respiratory tract by the epithelial fluids and cellular membranes that it contacts. 1 In this report, ozone is used to refer to the broad array of photochemical oxidants in ambient air, of which ozone is the primary component. 160
Overall Conclusions and Recommendations 161 Because ozone does not penetrate cells but leads to pulmonary and nonpulmon- ary events, a cascade mechanism has often been proposed to account for its tox- icity. Human chamber and toxicologic studies have yielded strong evidence in- dicating that short-term exposure to ozone can exacerbate lung conditions, causing illness and hospitalization, and potentially lead to death. The available evidence on ozone exposure and exacerbation of heart conditions, which is less abundant, points to another concern. Epidemiologic studies also have found that exposure to ozone (as an indi- cator of the broader mix of photochemical oxidants) is associated with those effects. Although methodologically somewhat different, the studies have been consistent in their use of large datasets with consistent diagnostic codes for health end points, nationally available ambient air measures, and data on ad- justment for some potential confounders. The committee found that the four recent time-series analyses and meta-analyses of the relationship between expo- sure to ozone and premature mortality add to that evidence by providing robust statistical evidence of an association (Bell et al. 2004, 2005; Ito et al. 2005; Levy et al. 2005). On the basis of the additional insights obtained from its review of the new time-series studies and its review of the broader evidence, the commit- tee concludes that short-term exposure to ambient ozone is likely to con- tribute to premature deaths. Despite some continuing questions about the evi- dence, the committee concludes that it is strong enough to be used in the estimation of the expected mortality risk reduction that would result from reduc- tion in exposure to ozone or the photochemical-oxidant mixture. In its RIA for the finalized ozone national ambient air quality standards (NAAQS), EPA (2008b) analyzed a variety of assumptions about the association between ozone exposure and premature mortality. They included the assumption that the association is not causal (see Appendix B). Although it is rarely possible to exclude the possibility of zero effect in such analyses, the committee con- cludes that an absence of any effect is unlikely. INTERPRETATION OF RESULTS OF HEALTH STUDIES Those who evaluate regulatory benefits seek information from health re- searchers, for example, â¢ To what extent is the relationship between ambient ozone concentration and mortality response due to ozone as opposed to copollutants that are quanti- fied separately, such as airborne particulate matter (PM)? â¢ Given that age and health status are important in estimating the quan- tity and quality of the remaining life expectancy, how dependent is the ozone- mortality relationship on those and other personal characteristics, such as socio- economic status?
162 Ambient Ozone and Mortality: Estimating Risk-Reduction Benefits Studies relating measurements of ambient ozone to health end points must deal with formidable obstacles because the dynamics of ozone and associated pollutants are complex, particularly the effects of temporal and spatial variations on individual human exposure. The committee reviews below the major factors that it considered in its review of the evidence. These factors can affect esti- mates of risk of ozone mortality in various ways. In some cases, the factors would cause an underestimation of risk; in other cases, an overestimation. On balance, the committee considers the evidence from the studies to be strong enough for making risk estimates, but the various factors and their potential ef- fects on the estimates should be fully acknowledged. Estimating Exposure to Ambient Ozone Time-series epidemiologic studies of ozone typically characterize expo- sures in terms of ambient concentrations at a centrally located monitoring site or at several sites in the study area. That use of ambient concentrations to indicate ozone exposures is a source of considerable uncertainty related to how well they reflect actual ozone exposures and how well ozone effects can be separated from effects of other pollutants or weather conditions. The magnitude of exposure errors probably varies with several factors, in- cluding season, home ventilation characteristics, and exposure averaging time. Ozoneâs mortality effects, for example, were shown to be stronger in the warm season than in winter or the entire year. Over 24 h, personal ozone exposures are weakly associated with corre- sponding ambient concentrations, and the association is stronger in summer than in winter. The seasonal variability probably reflects increased home ventilation in the hotter summer months. Even in well-ventilated conditions, however, the ozone association is such that only minor changes in indoor ozone exposures are expected to occur in response to moderate changes in outdoor concentrations. Together, low personal exposures and weak relationships between personal exposure concentrations and ambient concentrations suggest that 24-h am- bient ozone concentrations are poor proxies for personal exposures. For shorter averaging periods, such as the afternoon (when both personal outdoor activity and ozone concentration can be at their highest), results from a scripted exposure study suggest that hourly or daily peak ambient ozone concen- tration may be an appropriate proxy for corresponding hourly or peak personal exposure. Whether observations from this study are relevant for individuals at risk of ozone-related death warrants further examination. Personal ozone expo- sure is a major source of uncertainty in ozone-mortality risk estimates; ad- ditional studies are needed to determine whether using the shorter averag- ing times reduces this uncertainty. These studies will require the development of new measurement methods that have sufficient sensitivity to measure these likely low, short-term exposures. Additional studies assess-
Overall Conclusions and Recommendations 163 ing personal exposure-ambient concentration relationships may not only reduce uncertainties, but might help to explain variability in ozone-mortality findings across studies (that is, by determining how the strength of the correlation and the extent to which personal exposures change in response to changes in ambient concentrations varies across cities or subpopulations). In a given location, various ozone metrics (averaging periods) tend to be highly correlated, so it may seem somewhat unimportant to distinguish between them, and the degree of correlation may make it difficult for an epidemiologic study to do so. However, the choice of metric can have a larger effect on esti- mates of expected benefits of control programs. For example, a program that lowers emissions of oxides of nitrogen (NOx) could reduce peak ozone concen- trations but raise average concentrations. In that case, a cost-benefit analysis based on an association between premature mortality and average ozone could appear to have a negative effect, whereas an analysis based on an association with peak summertime ozone could show a benefit. It is unknown which is more accurate, so evaluating the benefits of urban NOx reductions is uncertain both in magnitude and in direction. The choice of exposure metric can also be important in designing a strategy to control ozone-precursor emissions (see Chapter 3). Although the recent time-series studies have to the extent possible in- cluded analyses of alternative ozone metrics (such as 1-h maximum or 8-h maximum ), it is important to examine the relationships further to examine dose- response relationships, test potential confounding more effectively, and inform future regulatory choices among different actions that might have different ef- fects on peak and multihour averages. Potential Confounding by Other Pollutants The committee found that short-term ozone exposure is likely to contrib- ute to premature mortality in addition to the risks posed by weather and PM, but studies to date have not been sufficient to control for potential confounding by or interactions with condensed-phase constituents of airborne PM, such as sul- fates, acids, elemental carbon, and metals. Colinearity among ambient pollut- ants raises concerns about possible confounding of ozone mortality effects by correlated copollutants. The concerns center on the possibility that effects associated with ambient ozone may be the consequences not only of ozone ex- posure but of correlated pollutants not included in the health-effects model. Of the possible confounders, ambient PM2.5 (PM with a diameter no greater than 2.5 Âµm) and weather have raised the most concerns about confounding. Strong summer correlations between ozone and PM2.5 in some locations may be attributed to similarities in their formation. However, associations be- tween ambient ozone and PM2.5 differ substantially in the western and eastern United States, between summer and winter, and for different PM components. Also, correlations among ambient ozone and other pollutants have been shown to vary by averaging period. These results suggest that the potential for con-
164 Ambient Ozone and Mortality: Estimating Risk-Reduction Benefits founding of ozone-health effects also varies by these factors, as discussed in Chapter 4. The potential for confounding by ambient PM2.5 probably depends also on particle composition, which varies by location and season. Studies have suggested that health effects of PM2.5 may differ by component, so con- founding might be an effect of specific PM2.5 components, such as sulfate, ele- mental and organic carbon, metals, and secondary organics; this matter should be examined. It will be difficult to address such confounding with currently available data, because PM2.5 component data have only recently been collected routinely at many sites and because winter time ozone is not often measured. In the near future, the Speciation Trends Network (STN) may provide data on new measurements by species. However, unlike data on ozone, such data are gener- ally available only once every 3-6 d (see Chapter 3). More-frequent measure- ments may be needed to improve understanding of how short-term variation in PM2.5 components might confound ozone-mortality associations. Assessing Ozone-Mortality Relationships During Winter Months There is a lack of observed association between ozone and mortality during periods when ozone is low, such as winter. Reasons for this lack of association are not well understood in part because of the decrease in moni- toring during such periods. Better understanding of the association is impor- tant for a full exploration of (1) seasonal differences in risk, (2) how these sea- sonal risk differences vary spatially between communities with warmer and cooler winters, and (3) ozone-mortality relationships at lower ozone concentra- tions. Ambient ozone is one of the most well-characterized pollutants in the United States, but ozone monitors in many locations are operated only during the ozone season, which varies from city to city. During that time, ozone moni- tors provide nearly continuous measurement, although concentrations are typi- cally reported hourly. Threshold To characterize the association between daily variations in ambient ozone concentrations and daily variations in deaths, a linear model is used in which it is assumed that the change in mortality risk is constant across pollution concen- trations. It is unlikely that the association between exposure and response at the individual level follows that simple mathematical formulation. Individuals have their own susceptibility, characterized by a unique exposure-response associa- tion. That association may be characterized by a particular threshold, a concen- tration of exposure to ozone below which there is no added risk of death. A per- sonâs threshold will vary, depending on the personâs âfrailtyâ at any given
Overall Conclusions and Recommendations 165 moment and thresholds may depend on the averaging period used to assess ex- posure. The time-series design, however, relates individual exposure not to indi- vidual risk but to the average of ambient concentrations for the âat riskâ popula- tion on any given day. Thus, the appropriate concentration-response function for the time-series studies is based on an aggregation of individual exposure- response curves. Aggregation of a large number of complex functions can yield smoother and more nearly linear curves at the population level, so we would expect the concentration-response function based on time-series studies to have a form that is relatively simple. Those results suggest a near-linear association between ambient concen- trations of ozone and daily mortality in the United States. Exposure misclassifi- cation caused by use of an average of ambient fixed-site monitoring data to es- timate population-average personal exposure makes it more difficult to distinguish between linear and threshold models. Estimates of the concentration- response curve based on epidemiologic studies with imprecisely measured expo- sure should be viewed with caution. A sensitivity analysis of the shape of the curve may be required to capture the uncertainties in this procedure. Moreover, approaches based on 24-h averaging may cloud thresholds related to actual ex- posures which may be better represented with shorter averaging times, such as by using an 8-h maximum or 1-h maximum. On the basis of its review of the evidence, the committee concludes that the association between short-term changes in ozone concentrations and mortality is generally linear throughout most of the concentration range, although uncertainties make it difficult to determine whether there is a threshold for the association at the lower end of the range. If there is a threshold, it is probably at a concentration below the current ambient air quality standard. Susceptibility and the Interpretation of Mortality Studies The evidence presented in Chapter 4 leads to the preliminary conclu- sion that the effects of ozone on acute death rates are likely to be larger among those with pre-existing disease and the list of plausible effect modifi- ers is rather long, although insufficiently investigated at this time. The role of genes in ozone mortality has not been investigated, but they undoubtedly play an important role as modifiers of various pathways involved in the effect. Inter- action of the various factors has not been investigated. One can infer that ozone mortality depends on subjectsâ susceptibility pro- file, which consists of a wide range and combination of factors. Susceptibility is most likely not a dichotomous trait but a dynamic characteristic that follows a wide distribution in the population, from no to high susceptibility. The level of susceptibility of any subject at any time may depend on the distribution of vari-
166 Ambient Ozone and Mortality: Estimating Risk-Reduction Benefits ous exogenous and endogenous determinants of susceptibility (and their interac- tions) and their change over time. The observable association between ozone and mortality reflects the weighted average of all true (but unobservable) individual concentration- response functions. The latter cannot be established, but the risk of dying be- cause of a 10-ppb increase in ozone concentration is likely to be substantially larger in the most susceptible persons than in the general population. Although susceptibility factors certainly matter, the distribution of the ozone-mortality effect estimates among groups with different categories of sus- ceptibility is not known; that is, the quantitative details of the heterogeneity of effects are not readily available. Consequently, the overall (population-weighted average) effect in the total population is the only currently scientifically sup- portable approach for use in risk assessment. Its use is appropriate, even though the population-weighted mean effect may not be a valid estimate for any specific sub-population, because the actual effect may be much larger among the suscep- tible individuals but much smaller or zero among the less susceptible. However, that is a source of an unknown amount of uncertainty when one is calculating the benefits of a reduction in ozone. If the hazard and risk distributions2 in the population are not independent, estimates of excess deaths and of life years lost in the entire population that are attributable to changes in ozone exposure based on time-series studies will not equal the averages of these quantities in all risk groups. The size of the difference is not known but could be estimated if risks for susceptible groups are obtained (see Chapter 4). Short-Term Mortality Displacement Is there evidence to say what share of the mortality associated with ozone in the daily time-series studies may be very short-term displacement of deaths that would have occurred within days or weeks in the absence of ozone expo- sure?3 Current economic valuations of mortality are based on study samples that have average remaining life expectancies. It would be problematic to extrapolate these valuations to a circumstance in which the remaining life expectancy of the population most at risk is very short. When a person is aware that death is immi- nent and quality of life is seriously compromised (for example, the person is bed-ridden, is in great pain, or has extremely diminished cognitive function), extending life by a small amount of time may not have a high value. But if the person can be treated and the condition improved and an acceptable quality of 2 In this context, the distribution of hazards among the population is the likelihood of death at any age; the distribution of risk is the association between ozone and death at any age. 3 The term harvesting is sometimes used instead of short-term mortality displacement to refer to the concept that air pollution leads to the death of people who are highly sus- ceptible and near death (and die a few days earlier than they would have without air- pollution exposure) rather than the death of people who are not otherwise near death.
Overall Conclusions and Recommendations 167 life can be restored, extending life may have a high value even if the person has other health impairments or is quite elderly. Given the design of acute-effects studies, the time lost or the prematurity of deaths is not directly estimable without possibly unrealistic assumptions. In the absence of knowledge about the amount of time lost because of acute effects, one may hypothesize that air pollutants were able to trigger death only in a pool of very frail people, those already in very bad health. Assuming that the remain- ing life expectancy of those frail people was very short even in the absence of pollution, the effect of air pollution would consist of only a minor shift of the time of death, namely, a short advancement of death. On the basis of available evidence, the committee concludes that deaths related to exposure to ozone (and other photochemical oxidants) are not restricted to people who are at very high risk of death within a few days. The evidence comes from the recent analysis of time-series in several U.S. cities that focused on identifying a mortality-displacement pattern in the time course of exposure and death (Zanobetti and Schwartz 2008). In that analysis, it was clear that short-term mortality displacement could not fully explain the ob- served increase in death; short-term ozone exposure was likely to have contrib- uted to shortening the lives of people who had compromised health, not neces- sarily just those near death. That evidence is based on results of only one study, however, and warrants confirmation by other studies. Distributed Lag Deaths related to short-term ozone exposure may not occur until several days after exposure or may be associated with multiple short-term exposures. Many studies of short-term effects investigate the change in death rates for only one or a few days, but distributed-lag models look further ahead to capture de- layed acute effects, often referred to as subacute effects. Such analyses are use- ful in understanding the statistical distribution of time between an increased concentration of ambient ozone and the time pattern of occurrence of death. On a population level, it is extremely unlikely that all ozone-related deaths occur either immediately or within 1-2 d of exposure. A more plausible model as- sumes susceptibilities to death (or frailty) and the (competing) success of inter- vention strategies to follow a distribution in which some people die immediately or within 1-2 d whereas others first suffer acute ailments (such as a myocardial infarction or pneumonia) and death may be delayed by partially successful treatment or occur as a result of the decompensation of defense mechanisms. In contrast with the PM health-effects literature, few data are available that are based on using distributed-lag models of ozone mortality. Those available suggest that effect estimates for both ozone and PM steadily increase with in- creasing time of the investigated effects. Specifically, subacute (longer-term) effects that are combinations of effects of several days or weeks of exposure are larger than immediate short-term effects, and estimates based on cohort studies
168 Ambient Ozone and Mortality: Estimating Risk-Reduction Benefits are the largest. However, the increase in the estimates is far larger for PM than for ozone; this may be a consequence of attenuation due to large errors in char- acterizing exposure to ozone, or it may reflect a more dominant role of acute disease related to ozone exposure whereas the role of chronic disease is more important in the effects of PM. As has been the case with PM, analyses con- ducted with distributed-lag models over several days appear to capture the overall effects of ozone better than same-day data, but there have been rela- tively few of them and further confirmation is warranted. Chronic Exposure Long-term effects of ozone on mortality are considered to be the result of cumulative effects on people who have chronic disease caused by repeated ex- posure to ozone. Although some inconsistencies remain to be clarified, the ob- served associations between ozone exposure and decreased small-airway lung function during childhood and adolescence suggest that ozone-related mortality is at least partially attributable to exposures across a period of more than a few days. Ozone tissue dose is highest in the small airways, so the findings are in line with expected ozone-related conditions that result in reduced lung function. The evidence of an effect of long-term ozone exposure on lung-function growth increases the plausibility of its effect on mortality. The association between poor lung function and life expectancy is strong and well established. The standard approach to investigating effects of cumulative ozone expo- sure on life expectancy is the cohort study, in which large numbers of subjects are followed for several years. After taking into account all other factors that are likely to affect mortality, cohort studies can test the null hypothesis that mortal- ity is the same among populations that have different ozone-exposure histories. However, none of the cohort studies available at this time were designed to in- vestigate chronic effects of ozone, and differences in ozone exposure among subjects in each study tended to be rather small. Several North American and European air-pollution cohort studies have focused on ambient PM or markers of local traffic to characterize exposure of study cohorts. Assessments of the association between life expectancy and am- bient ozone have been far less extensive. Analyses of the most extensive of the cohorts (that of the American Cancer Society) have found small positive effects in warm seasons, but other cohort studies have not found positive associations between long-term average ozone concentrations and cardiopulmonary mortality after controlling for PM2.5. The weak current evidence from cohort studies of an association of premature mortality with chronic exposure to ozone suggests that risks are larger than those observed in acute effects studies alone. If further con- firmed, the evidence from cohort studies of an association of premature mortality with longer-term exposure would tend to support the notion that effects seen in time-series studies reflect only a portion of the total effect.
Overall Conclusions and Recommendations 169 The use of large cohorts with long followup periods may be required because the long-term ozone-exposure mortality risk appears to be much smaller than that associated with PM2.5 It may be necessary to pool information from several existing cohorts to obtain sufficient statistical power. HEALTH-BASED INFORMATION FOR BENEFITS ASSESSMENTS Mortality Time-Series Results The committee recommends that ozone-related mortality be included in future estimates of health benefits of ozone reduction. The committee further recommends that the greatest emphasis be placed on estimates based on systematic new multicity analyses using national databases of air pollution and mortality, such as was done in the National Morbidity, Mor- tality and Air Pollution study (NMMAPS), without excluding consideration of meta-analyses of previously published studies. Emphasis should also be placed on risk estimates obtained from analyzing data on multiple days so as to include delayed acute effects. (see Chapter 4). Such health-benefits estimates should be accompanied by a broad array of analyses of uncer- tainty, while at the same time understanding that a zero value is unlikely. In light of this recommended approach, future RIAs should give little or no weight to the assumption that there is no causal association between estimated reduc- tions in the incidence of premature mortality and reduced ozone exposure. Effect of Ozone Exposure on Life Expectancy The impact of pollution on mortality may be quantified in terms of changes in mortality rates, number of deaths in a given period, or the months (or years) of life lost (or saved). EPA has calculated numbers of deaths prevented in a given year for each incremental reduction in PM or ozone concentration by applying the relative risk from epidemiologic studies to a baseline mortality rate. There is also interest in assessing the number of life-years saved, especially for cost-effectiveness analysis. The committee was asked whether there is an ade- quate basis for quantitatively characterizing the likely impact of reductions in short-term daily exposures to ozone on life expectancy. Questions about the population at risk for mortality from short-term ozone exposure underlie most of the issues raised about interpreting epidemiologic time-series results in benefits assessment, including issues of susceptibility and short-term mortality displacement. Those issues are important for economic valuation. It is important to know that the willingness-to-pay (WTP)4 values used are appropriate to the at-risk population, and this remains one of the key 4 WTP is an estimate of the amount that a person is willing to pay for changing his or her mortality risk in a given period by a small amount (see Chapters 2 and 5).
170 Ambient Ozone and Mortality: Estimating Risk-Reduction Benefits uncertainties that the committee has highlighted in its report. For example, if we know from the cause of death associated with ozone exposure that the mortality risk is primarily for respiratory causes of death, we know that the age distribu- tion of the at-risk population is skewed more toward the elderly, especially those experiencing cardiopulmonary compromise, than is total mortality in the general population. Thus, the WTP estimates should be relevant for a population with a higher average age. The currently available information lacks that level of detail. Although cardiorespiratory mortality stays at the center of the discussion, it is not known whether the age and risk profile of the people whose deaths are at- tributable to ozone corresponds to the typical person who dies of cardiorespira- tory problems. The only information available is the population-average change in mortality rates, which translates into the population-level change in life ex- pectancy. A unified survival model of a dynamic cohort study that combines long- term and short-term air-pollution exposure has been developed (see Chapter 4). By incorporating life-table methods based on time-series risk estimates, the model can be used to estimate both life-years lost and number of additional deaths expected to occur in a specified period because of changes in air- pollution concentrations under restrictive assumptions. Further complexity can be gained by incorporating risks specific to age-sex groups. It is not clear, how- ever, how long a period of long-term increased exposure is required to shorten life expectancy; it may be years. The time distribution of effect on longevity associated with immediate reductions in pollution also is not known. The risks based on time-series studies assume (by design) that reductions in pollution have an immediate benefit for longevity. As mentioned above, if the hazard and risk distributions in the population are not independent, estimates of excess deaths and life-years lost in the entire population attributable to changes in ozone expo- sure based on time-series studies will not equal the average of these quantities over all risk groups. The most complete assessment of the association between acute expo- sure and death originates in distributed-lag models that integrate the dis- tribution of the time between exposure and death. However, both the usual time-series model and the distributed-lag models focus on a short window of exposure. Effects of long-term cumulated exposure are, by design, ignored in those studies. Reductions in life expectancy based on the time-series re- sults can be calculated by using life-table methods, but lack of information about the at-risk population necessitates making the assumption that re- maining life expectancy is comparable with of others in the same age-sex cohort. Use of Information from Acute-Effects and Chronic-Effects Studies Evidence of long-term effects of ozone on mortality (or survival time) is presently weak; thus, the derivation of a relative risk to describe the association
Overall Conclusions and Recommendations 171 is more difficult than in the case of PM (see Chapter 4). The translation of the results into numbers of attributable deaths is possible but leaves two main chal- lenges: First, the assessment of acute effects would be incomplete if based solely on the usual time-series studies, given that subacute (delayed) effects are not captured with these methods; only one study has published estimates of subacute effects of ozone based on distributed-lag models (Zanobetti and Schwartz 2008). Second, in the absence of abundant quantifiable evidence of chronic effects of ozone on mortality, total life-years lost because of both acute and chronic effects cannot be estimated with confidence. As mentioned previously, the current evi- dence from cohort studies, although it is weak, supports the notion that the ef- fects estimates seen in the-series studies reflect only a portion of the total effect. The question of how to use the epidemiologic information to translate into risk changes and benefits assessment is the subject of debate. So far, many risk assessments have derived as an intermediate step the number of deaths attribut- able to air pollution on an annual basis. That approach has several limitations that are of particular concern for chronic effects but, conceptually, apply also to acute effects. A problem in discussing lives saved is related to the misleading message that any policy may prevent death whereas the only result one can expect is a postponement of deathâa longer life. The term attributable death is at least less misleading than lives saved. Conceptually, however, attributable death still im- plies a body count rather than the life-years in premature deaths. In the case of death due to short-term ozone exposure of people with some pre-existing disease (either acute or chronic), the presentation of âannual attributable deathâ may be an acceptable approach. It should be noted, however, that in the case of quanti- fying chronic effects of pollution, assumed to complicate the underlying chronic diseases that increase the susceptibility to premature death, the attributable-death concept is less appropriate because it does not take fully into account the long- term dynamic in a population in which the underlying risk profiles change. A reduction in air pollution will lead to longer life expectancy and thus increase the number of elderly people. Age-adjusted mortality rates are expected to be lower under cleaner conditions, but the absolute number of deaths will steadily increase as the population ages. Consequently, the âattributable deathsâ will not be the same throughout the years after an improvement in air quality (Hurley et al. 2000; Miller and Hurley 2003; Rabl 2006; Brunekreff et al. 2007). Recommendation: EPA should study emerging understanding of ozone- mortality associations and evaluate their use and implications in benefits assessments, including relationships between changes in mortality rates, annual deaths prevented, and years of life saved. The alternative ap- proaches for expressing ozone mortality effects will lead to rather similar results if one is supposed to express only the most immediate (acute) ef- fects of pollution changes. However, with the integration of subacute ef- fects estimates, and particularly in the case of use of estimates of long- term chronic effects, the discrepancies between the approaches increase
172 Ambient Ozone and Mortality: Estimating Risk-Reduction Benefits and the conceptual flaws of the attributable-cases model become more pronounced. VALUATION An important question for benefits valuation is whether a personâs value for reducing risks of death should be based on the magnitude of the life exten- sion (that is, the loss of life-years that is avoided or prevented). Although there are intuitive reasons for saying yes, the question is how it should be done. Esti- mates of the value of a statistical life (VSL) are often used in cost-benefit analy- ses for programs expected to reduce mortality risks in a population. VSL esti- mates are derived from estimates of peopleâs WTP for reducing their mortality risk in a given period by a small amount (see Chapters 2 and 5). If the effects of reducing mortality risk are measured in terms of years of life extension (in- creased remaining life expectancy), then it may be more appropriate to use a monetary value per statistical life-year (VSLY). For any given risk reduction the average WTP value can be summarized as an average VSL or as an average VSLY if the remaining life expectancy and rate of time preference are known for the population from which the average WTP value was derived. Difficulties arise when either of these summary meas- ures is assumed to be constant and used to estimate values for risk reductions in other populations or other risk reduction contexts. EPAâs current primary approach to economic valuation of mortality-risk reductions is a variation on its long-standing approach of using the same VSL for all annual mortality reductions. In its primary benefits estimates, EPA ap- plies the VSL to all lives saved regardless of the age or health status of the popu- lation experiencing the change in mortality risk and regardless of the cause of the change. Because there is some expectation that WTP for mortality-risk reduction will vary with the characteristics of the population affected or with the context of the risk change, EPA asked the committee to assess scientific approaches to assigning economic values to reductions in mortality risk associated with ozone reductions and to address the questions of economic valuation for different changes in life expectancy. Valuation in the Context of Cost-Benefit Analysis The charge to this committee concerns monetary valuation of mortality risk reduction for regulatory impact analyses that are based on the basic prem- ises of cost-benefit analysis. In this context, therefore, we focus on WTP values for mortality risk reduction. Cost-benefit analysis, however, focuses on eco- nomic efficiency and many other ethical and legal factors are appropriate to con- sider in policy and regulatory decisions. In general, these issues should be con- sidered separate from the cost-benefit analysis rather than interjected into the
Overall Conclusions and Recommendations 173 valuation estimates because this endangers the neutrality of the analysis. How- ever, there may be some instances when such interjection is appropriate. For example, the decision by EPA to not adjust WTP estimates for local differences in income levels is justified because to do so creates a situation that favors greater environmental protection in wealthier locations, an outcome policy mak- ers judge to be unfair and in many cases illegal. The committee stresses that government decision-makers need informa- tion on how the WTP for mortality risk reductions varies by risk characteristics, population characteristics, and for both risk as a private good and as a public good. How they choose to use this information, however, is not strictly a techni- cal decision, but depends on ethical precepts, legal precedent, the quality of the evidence and other factors that may be beyond the analystsâ control or purview. Recommendation: We recommend that the issues contained in this find- ing of the committee be considered within the Office of Management and Budget and the agencies that use monetary values for mortality risk reduc- tions in their regulatory analyses. These agencies should develop a plan for generating the information needed to determine how WTP varies for different populations and different risk contexts. In addition, there should be an exploration to determine the appropriate uses of this information. Such an exploration should go beyond economic considerations and in- clude ethical and public policy perspectives. Willingness-to-Pay Estimates Both economic theory and the available empirical evidence are inconclu- sive about how individualsâ WTP values for reducing their own risk vary with two important individual characteristics: age as a proxy for remaining life ex- pectancy and health status. We conclude that the empirical evidence is insuf- ficient to support a specific quantitative adjustment of the WTP estimates to account for differences in remaining life expectancy, but we do not reject the concept that such adjustments may be appropriate. It is plausible that people with shorter remaining life expectancy are willing to devote less of their resources to reducing their mortality risk than would people with longer remain- ing life expectancy. Characteristics of the risk that may affect individual WTP values include the type of risk (for example, illness or accident) and its latency. The literature is inconclusive about the influence of risk characteristics (see below). The effects of latency on WTP values are straightforward conceptually (WTP for a future mortality-risk reduction should be less than that for an equivalent immediate risk reduction), and a small body of empirical literature supports such a concept. However, the epidemiologic literature is not sufficient to estimate the degree of latency of changes in mortality risk from ozone exposure.
174 Ambient Ozone and Mortality: Estimating Risk-Reduction Benefits Recommendation: Although there are many concerns about the accuracy of using the same WTP estimate (or range of estimates) for all mortality- risk reductions, we recommend that EPA use this approach, with appropri- ate scaling to the size of the risk change, as the most scientifically sup- portable approach for monetary valuation of ozone-related mortality, given the currently available information in the economics and epidemiol- ogy literatures. Empirical evidence of how WTP varies with population or risk characteristics is not sufficiently consistent to support a change in this practice that EPA has been using for many years. Estimates of Value of a Statistical Life-Year The use of a constant VSLY in the valuation of increases in life expec- tancy requires the assumption that WTP values for mortality risk reductions be consistently declining with increasing age. Available empirical evidence does not support that assumption and therefore does not support the use of a constant VSLY. The literature does not reject the use of a non-constant VSLY, however, and the epidemiological literature may favor reporting life-years saved. Recommendation: Unless future research produces empirical support for the assumptions that underlie a constant VSLY, EPA should not attempt to make adjustments for remaining life expectancy by calculating life-years saved and using a constant VSLY to value them. It may be appropriate to calculate and report life-years saved (in addition to reporting changes in annual mortality ranges and reductions in premature deaths), but it is not appropriate to use a constant VSLY as a monetary valuation of life-years saved, except, perhaps in a bounding exercise. The committee cautions against use of such an analysis in anything but a sensitivity analysis, how- ever (see below). There is likely to be good reason to use a non-constant VSLY or a non-constant VSL, once the literature is sufficient to make this transition. The committee stresses, however, that the status quo of using a uniform VSL should be continued until there is sufficient empirical evi- dence of how WTP for mortality risk reduction varies with differences in remaining life expectancy and other factors, which the committee con- cludes is not yet available. Value of a Statistical Life, Individual Characteristics, and Risk Contexts Most of the revealed-preference and stated-preference studies relied on by EPA to obtain estimates of the VSL have estimated WTP for mortality-risk re- duction in a context (such as traffic accidents and workplace accidents) and for a population that differ from the context and population relevant to the pollution- related risks that EPA is assessing. Applying the available estimates in EPAâs
Overall Conclusions and Recommendations 175 assessments in a different risk context (illness vs accident) and for different population characteristics introduces considerable uncertainty about how these factors affect average WTP values. However, the current literature is inconclusive as to how much the WTP values may vary with these factors. Recommendation: EPA should ensure that the average WTP estimates selected from the literature reflect results of both revealed-preference and stated-preference studies. The agency should take into account the strengths and weaknesses of each study approach and consider how closely the available studies match the policy context in population at risk and type of risk. EPA should give less weight to wage-risk studies in se- lecting WTP estimates than it has in the past. Given the limited studies available for different risk contexts, it is difficult to say how much the WTP values may differ, but the wage-risk studies are a poor match to the population and to the risk context for the ozone-mortality case. Sensitivity Analyses The direction of the expected error, if any, in using the average VSL in the literature for valuing changes in ozone-related mortality risk is more likely to be toward overstating the WTP to reduce the risk. That is because greater mortality risk associated with ozone appears to be in the elderly population and because latent risk may be involved. Given this populationâs substantially less than aver- age remaining life expectancy, it is possible that its WTP to reduce mortality risk would be less than the average WTP for the population as a whole. How- ever, we expect that the lower WTP as a result of the elderly populationâs lower remaining life expectancy is offset to some extent by higher WTP because of poorer health status or higher baseline risk. Although results in the empirical literature are not consistent, several studies suggest that WTP to reduce mortal- ity risk is constant or declines slightly with age. That evidence suggests that, for ozone, a proportional adjustment of the VSL for remaining life expec- tancy (that is, using a constant VSLY) would result in WTP that is too low. Recommendation: Given the uncertainty in the accuracy of available VSL estimates for ozone-related mortality, EPA should conduct sensitivity analyses with alternative estimates or assumptions. The purpose of sensi- tivity analyses is to see whether the overall conclusions of the cost-benefit comparison are changedâfor example, whether net benefits are still posi- tive under alternative economic-valuation assumptions. In general, there is less confidence in the sensitivity analyses because the alternative assump- tions are more speculative than the primary assumptions or deviate from the status quo, which the committee feels puts a burden of proof on those who would overturn it. Therefore, EPA should present results in a way that does not give equal weight to the outcomes of sensitivity analyses and
176 Ambient Ozone and Mortality: Estimating Risk-Reduction Benefits the outcomes of the use of primary assumptions. However, results of sen- sitivity analyses can be included in summary conclusions. EPA should se- lect alternative assumptions for sensitivity analyses on the basis of either theory or evidence. For example, different published empirical estimates of the relationship between WTP for mortality-risk reduction and age could be selected as illustrative of the range of results in the literature, in- cluding estimates of VSLY or VSL that vary with age. RESEARCH RECOMMENDATIONS The committee was asked to identify major gaps in knowledge about the benefits of reducing ozone exposure and the most promising research strategies to close the gaps, including additional data, analyses, or research needed to sepa- rate the contributions of ozone and other gaseous or particulate components of air pollution to the total short-term effect on premature mortality documented in the literature. The health-related recommendations (from Chapter 3 and 4) and the future research needs for valuation (from Chapter 5) briefly summarized in this section should be addressed as part of EPAâs research strategy for estimat- ing the benefits of reducing ambient ozone. The committee recognizes that many of the recommended research activi- ties are complex and will be difficult to undertake, and that sufficient resources may not be available to undertake all of them in the near term. Therefore, EPA and other agencies that might carry out the recommended research will need to set priorities and develop a strategy for addressing the various information needs. Recommendations for Future Health Research The health-related recommendations are presented in three broad catego- ries related to enhancing exposure assessment, enhancing epidemiologic studies, and reducing uncertainty. Enhanced Exposure Assessment for Epidemiology Evaluate exposure metrics to determine whether and how much daily peak exposures, such as 1h or 8h exposures, and longer-term average exposures, such as over 24h, are associated with ozone-related mortality. Identify the ap- propriate exposure metrics to relate how control programs will affect ozone con- centrations and health. Consider that control strategies may have quite different effects on 24-h average concentrations and shorter-term exposures. Also con- sider that peak short-term concentrations on lower-ozone days will respond dif- ferently from peak short-term concentrations on high-ozone days and often in the opposite direction.
Overall Conclusions and Recommendations 177 Investigate potential confounding by evaluating regional and seasonal associations between ambient concentrations and exposures to ozone and PM2.5 (and its components), how the data affect researchersâ ability to control for con- founding, and correlations between the various pollutant concentrations. When possible, focus on groups of individuals who are sensitive to ozone exposures and use data on the chemical and physical components and size distribution of PM2.5. Include PM speciation data to account for seasonal and geographic vari- ability in the relationship between ozone and its potential confounders. Include the growing STN database in analyses of potential confounding of the ozone associations. Ensure that STN collects data frequently enough on the particle components most relevant to understanding the potential for confounding. Monitor ozone in winter and report the measurements. The size of the winter program should be sufficient to allow researchers to examine seasonal differences in risk, how the differences vary spatially between communities with warmer and cooler winters, and ozone-mortality relationships at lower ozone concentrations. Ozone is a regional pollutant, and winter measurements need not be collected at all the summer locations; but collect them at the frequency of summer measurements. Evaluate air-quality numerical models, such as the Community Multis- cale Air Quality (CMAQ) model for use in ozone epidemiologic studies to ex- tend the spatial scale of available data. Consider the uncertainty associated with the models before drawing inferences about mortality risk assessment associated with ozone exposure. Evaluate ozone exposure models such as the Air Pollutants Exposure (APEX) model that are used to improve characterization of ozone exposure at the population level by taking into account human activity. Assess whether the models can be used to improve epidemiologic studies or benefits assessments. Enhanced Epidemiologic Studies Explore thresholds by studying panels of individuals considered to be susceptible to premature death from ozone exposure, such as those with im- paired lung and heart function. Further explore how individual thresholds may vary and the extent to which thresholds depend on the frailty of the individual at any given moment. Because it is not clear whether ozone is associated with mortality in the cooler months, examine warmer months separately. Conduct a sensitivity analy- sis on concentration-response relationships to capture more fully the uncertain- ties contributed by reliance on average fixed-site monitoring data to estimate population-average personal exposure. Explore short-term mortality displacement and include use of alterna- tive study methods, such as investigation of subjects who have diseases (for ex-
178 Ambient Ozone and Mortality: Estimating Risk-Reduction Benefits ample, diabetes or heart disease) that are known to induce high mortality risk associated with air pollution. Identify susceptibility characteristics that may have important effects on ozone-mortality relationships and develop a distribution of the estimates of the ozone mortality effect among the categories of susceptibility. To the extent that data are not available, use models and assumptions for sensitivity analysis. Conduct distributed-lag analyses as one part of future epidemiologic in- vestigations to improve understanding of the statistical distribution of time be- tween an increase in the ambient concentration of ozone and a pattern of occur- rence of death. Conduct cohort studies to examine further the association between long- term ozone exposure and mortality. Develop long-term ozone-exposure models that can distinguish variations in exposure at the individual level and between and within cities. To the extent that new cohort studies strengthen evidence of long-term effects, consider including estimates based on that evidence in bene- fits assessments. Study cardiovascular effects of ozone exposure, both in human and ani- mal models. Design studies to identify genetic susceptibility factors. Uncertainty Characterize uncertainty of results of epidemiologic models and discuss their reliability and estimated uncertainty about which model (if any) is reasona- bly correct. Consider Bayesian approaches for uncertainty analysis, including the possibility of additional expert elicitation once the recent experience with this approach to PM risk assessments has been evaluated. Conduct sensitivity analysis intermittently as computational models and input distributions are developed to focus resources on the most important inputs or parts of the model. Identify the bases of estimates to distinguish between data-derived esti- mates of some components (such as the concentration-response function) and the expert opinions about other components that are not supported by scientific data. Recommendations for Future Research on Valuation Explore how WTP varies with mortality-risk changes and changes in life expectancy, and explore how WTP varies with population characteris- tics such as age, health status, and baseline risk levels. Although the research needs would be addressed primarily by using stated-preference methods, explore ways to address research needs with further developments of revealed- preference methods.
Overall Conclusions and Recommendations 179 Seek results on total age effects to understand better how age and re- maining life expectancy affect WTP for reductions in mortality risks or exten- sion of life expectancy. Ask that future (or even previous) studies report total age effects (that is, WTP by age cohort) in addition to effects of age alone on WTP for a small reduction in mortality risk in a given period. Make datasets available for meta-analysis by urging researchers who receive EPA funding to make their datasets available for meta-analyses in addi- tion to providing their published results. Explore and develop full life-cycle methods for communicating and valuing changes in mortality risk as a shift in the survival curve, which plots survival probabilities in all future periods and from which life expectancy is derived. Explore effects of mortality-risk characteristics on the valuation of re- duction in these risks. Consider different types of risk (such as accident and ill- ness) and the latency of the change in risk. Compare values of (or nonmonetary preferences for) risks in different contexts (e.g., Magat et al. 1996; DeShazo and Cameron 2004). Explore the potential usefulness of studies that analyze preferences regarding public goods by developing approaches to distinguish paternalistic from nonpaternalistic altruism to avoid double-counting of benefits. Conduct conceptual analyses to learn how such results might be appropriately used in cost-benefit analysis, in which the usual paradigm is to sum beneficiariesâ pri- vate WTP. REGULATORY IMPACT ANALYSES INVOLVING OZONE MORTALITY As discussed above and in Chapter 4, future RIAs based on currently available studies concerning ozone-related mortality should give greatest weight to the results of multicity time-series analyses. As EPA (2008b) had done for its RIA for the finalized ozone NAAQS, most of the emphasis should be placed on estimates based on systematic new multicity analyses, such as was done in the NMMAPS, without excluding consideration of meta-analyses of previously pub- lished studies. Future RIAs should incorporate research results on the mortality effects of chronic ozone exposure and research that addresses key uncertainties related to potential confounding factors, exposure measures, and susceptibility as appropriate. Health-benefits estimates should be accompanied by a broad array of analyses of uncertainty, while at the same time understanding that a zero value would be unlikely. Future RIAs should give little or no weight to the assumption that there is no causal association between estimated reductions in the incidence of premature mortality and reduced ozone exposure unless new information emerges that refutes the interpretation of this association as causal. Presentations like that included in Table 7-14 of EPAâs recent RIA for the finalized ozone
180 Ambient Ozone and Mortality: Estimating Risk-Reduction Benefits NAAQS, showing a variety of assumptions about the association between ozone exposure and mortality, should be revised in light of this recommendation (see Appendix B). Risk analyses conducted with distributed-lag models over several days ap- pear to capture the acute and the somewhat delayed subacute mortality effects of ozone exposure better than single-day data. Those models should be part of fu- ture benefits assessments to the extent that they are available. There are many concerns about its accuracy for all mortality-risk reduc- tions, but using a specific WTP value and a corresponding VSL estimate (or range of estimates) is the most scientifically supportable approach at this time to monetary valuation of ozone-related mortality, given the information in the eco- nomics literature. Before recommending a substantial change in EPAâs approach for valuation of mortality risk reductions, it is necessary that there be fairly con- clusive empirical evidence to support a specific change in the approach. It is the committeeâs judgment that the available evidence is not sufficient to support such a change at this time, but alternative approaches should be explored in sen- sitivity analyses and further research should be conducted to answer the ques- tions raised about the validity of EPAâs current approach. As new information emerges on population characteristics of those susceptible to mortality from ozone and on variations in WTP for mortality-risk reductions (or increases in life expectancy) due to different population characteristics, benefits-assessment methods may need to be revised. EPA should consider placing greater emphasis on reporting decreases in age-specfic death rates and increases in life expectancy than on reporting esti- mates of lives saved. Such a change is needed to be responsive to recent reports that it is not possible to identify specific deaths attributable to air-pollution ex- posures (e.g., Brunekreef et al. 2007). For example, if the relative risk is 1.002 for a unit change in pollution (for example, for a change in tons of a pollutant emitted over a period), it means that there is about a 0.2% higher mortality rate for every unit increase in pollution in the population group to which the esti- mated relative risk applies. The relative risk is a proportional risk estimate. It may be estimated for the general population, but it would be preferable if it were estimated for the specific population groups that may be expected to have differ- ent susceptibilities to the effects of pollution exposure. In any case, it should be applied in a benefits assessment to the population for which it was estimated. If the scenario is for a two-unit reduction in pollution with a relative risk of 1.002 per unit change in pollution, and the relevant population has a baseline annual mortality rate of 1 in 100, the reduction in the annual mortality rate attributable to the change in pollution is about 4 per 100,000. For a population of 1 million, EPA would report that as 40 lives saved. The committee recommends that EPA report preferentially the annual mortality-rate change of 4 in 100,000 for the applicable population and to develop models for consistent calculation of life- expectancy changes and changes in numbers of deaths (e.g., Rabl 2006). The WTP studies from which VSL estimates are derived match the annual mortality-rate change estimates. For example, the valuation literature reports
Overall Conclusions and Recommendations 181 WTP estimates of $40-400 for a 4-in-100,000 risk reduction (translating to a VSL of $1-10 million). Turning the annual mortality-rate changes and WTP values into saved lives and values per life saved is a mathematical transforma- tion that seems convenient for summary tables and policy assessments, but can be misleading and can obscure the underlying derivation and appropriate inter- pretation of the studies. It also ignores the effect of life-expectancy changes on the population composition over time, and this will cause the annual number of deaths to decrease at first and then to return to its previous level or even increase over time (e.g., Miller and Hurley 2003). The health-related research and valuation research recommended in this report should be addressed as part of EPAâs research strategy for estimating the mortality risk-reduction benefits of reducing exposure to ambient ozone. How- ever, the research needs should not be viewed as a basis for postponing consid- eration of ozone mortality relationships in benefits assessment until more infor- mation is obtained. Also, it would be a mistake to assume that the committeeâs discussion of uncertainties and research needs broadly applies to the current understanding of air pollution and health in general. For example, this report has indicated where there is a greater understanding of many aspects of PM- mortality relationships relative to those for ozone. Continued enhancement of knowledge and methods for valuation of ozone mortality risk reduction benefits will inform future regulatory decision making and help in understanding the relative importance and value of effects caused by various pollutants.