Reappraising the Public Health Benefits Conferred by Environmental Policy: Considerations for an Aging Society

Abstract

Although addressing environmental pollution and unprecedented societal aging are concurrent public health challenges, rarely is the relationship between the two considered. Current approaches to valuing the public health benefits conferred by environmental policy may be ill-suited to aging populations. We describe the limitations of the age-invariant approach used by the United States Environmental Protection Agency to estimate the public health benefits corresponding to environmental regulation. These include the poor age-representativeness of the samples informing the valuation of mortality risk reduction, the exclusion of age-related outcomes from valuation, and the omission of age-related third-party expenditures. We offer an empirical framework that could address these limitations. Our recommendations may improve the calibration of environmental regulatory analysis to the changing age distribution of the United States population.

Background

Age-calibrated measures of the expected health benefits of environmental policy are scarce despite population aging and older adults’ disproportionate susceptibility to environmental pollution.¹ Because the costs of environmental regulation are often passed on to the public—in the form of decreased wages or increased taxation and prices—accurate estimates of the societal value assigned to corresponding risk reductions are essential.^2,3 The poor generalizability of current valuation methods to older persons may distort regulatory cost-benefit analyses and bias subsequent policy decisions.⁴ Addressing factors that constrain the applicability of valuation frameworks to aging societies may enable administrative agencies to strengthen policy appraisals and public health.

The United States Environmental Protection Agency (EPA) follows a set of procedures to value the health benefits of environmental policy, which can be summarized in three major steps (Figure 1).^4-6 First, analysts use the existing epidemiological literature to estimate the health benefits (often mortality risk reduction) attributable to the magnitude of pollution abatement for the population of interest. These estimates require sound epidemiological evidence to support a causal relationship between changes in the concentration of a pollutant and the health outcomes of interest. This is accomplished by calculating the number of statistical cases averted as compared to a counterfactual scenario without regulation, using epidemiological estimates of the reduction in individual risk attributable to pollution abatement and the number of persons affected by the policy.^4,6 Second, analysts use an official estimate of the monetary value that individuals assign to incremental changes in their own mortality risk reduction.⁷ This value is referred to as the individual value per statistical life (VSL) and is often empirically-derived from studies of observable behavior. Typically, the EPA uses the same VSL for all individuals. The population-average VSL reflects the dollar amount that a sample of adults would be willing to forgo to incrementally reduce mortality risk so that one fewer among them would die in the current period.^8,9 To illustrate, a population-average VSL of $10 million is consistent with an average individual willingness to trade $1000 of wealth for a 0.0001 decrease in current-year mortality risk.⁴ Third, analysts multiply the statistical lives saved by the population-average VSL; this is equivalent to calculating the value to each person affected by the policy and summing these to estimate the total population value of risk reduction conferred by regulation.⁴

Figure 1: An Overview of the Valuation of Health Risk Reduction.

We provide a schematic overview of the main procedures used by federal agencies in valuing mortality and health risk reduction.

Defining the Problem

The EPA applies the population-average VSL irrespective of the age-composition of the populations affected by the policy or to whom risk reduction or implementation costs might accrue.^5,8 Accurate and valid estimates of the relationship between the VSL and age are essential because differing assumptions can generate widely disparate estimates of the public health benefits conferred by regulatory decisions.⁴ Although the VSL can be modeled using different assumptions regarding its relationship to age, the value assigned to mortality risk reduction is typically inferred from observable behavior within labor market cohorts. As a result, the sampling frame upon which VSL estimates are predicated generally excludes older persons.^4,9 Calibrating valuation frameworks to the observed health-related needs and stated preferences of older persons could reduce uncertainty regarding the applicability of the population-average VSL. The salience of such uncertainty will increase as the average older person gains additional years of active life expectancy and benefits from innovations that may enhance health-related quality of life during disabled life-years.^10,11

Limitations to the Current Approach in an Aging Society

Below, we describe three central limitations of applying the population-average VSL and excluding age-related outcomes in cost-benefit analyses of environmental policies within aging populations. Key terms are defined in Table 1.

Table 1:

Key Terms

Term	Brief Definition
Value per Statistical Life (VSL)	The marginal rate of substitution between wealth and incremental changes in mortality risk at a point in time.
Statistical Cases Averted	The expected decrease in cases of an adverse health outcome that can be attributed to reduced exposure to pollutants due to policy enactment, as compared to a counterfactual scenario in which policy remains unchanged.
Universal Aging Outcomes	In the absence of a consensus definition, universal outcomes share several properties, as follows: (1) transcend diagnostic categories, enabling the measurement of different health outcomes across diverse health conditions, (2) incorporate information regarding health-related quality life that is easily understood, such as function or symptom burden, (3) apply a common measurement scale across multiple health-related domains.
Long-Term Services and Supports	The spectrum of formal and informal supportive healthcare services provided to persons with functional and or cognitive limitations that impair their ability to function independently. Formal long-term services and supports include postacute rehabilitative care following hospitalization (either in a skilled nursing facility or within the home) and long-term residential care. Informal long-term services and supports include the provision of custodial care, often by unpaid family members.

The poor generalizability of mortality valuation frameworks to older persons

Conventional approaches to estimating the VSL may not be applicable to older persons for several reasons. First, most VSL estimates are inferred from the wage premium that labor force participants require to accept jobs with a greater risk of accidental workplace death.¹² The poor age-representativeness of labor market cohorts, generally comprised of adults 18 to 65 years of age, diminishes generalizability to older persons.⁴ Second, the value that adults assign to mortality risk reduction is age-sensitive and does not directly correlate with remaining life expectancy.^9,12 The VSL is estimated to peak in midlife, decrease up to the oldest age studied—often less than 65 years—and continue to decline thereafter.^9,13 Whether monotonic decline in the VSL is a valid assumption for older persons requires further assessment in age-representative cohorts. Third, VSL estimates based on available stated preference surveys predate documented increases in active life expectancy¹⁰ and may, therefore, fail to capture time-varying changes in older persons’ valuation of an incremental life-year. Moreover, improved opportunities to forestall institutionalization^11,14 may influence older adults’ valuation of disabled life-years during which they require comprehensive formal and informal long-term services and supports. Lastly, the conventional valuation of mortality risk reduction neglects variation in the value that older persons might ascribe to different causes of death. Although the EPA has considered adopting a ‘cancer premium’ to accommodate individuals’ differential valuation of cancer-related mortality, such an approach does not account for differences in antecedent morbidity (e.g., the natural history and symptom burden associated with a cancer subtype) or inherent heterogeneity in the valuation of mortality due to alternative causes.^8,15

The restrictive set of nonfatal health outcomes used in the valuation of risk reduction

The EPA may include nonfatal health outcomes in its valuation of environmental policy when there is sufficient toxicological and epidemiological evidence to support a relationship between the regulated pollutant class(es) and an outcome.⁶ It is common for the EPA to value averted nonfatal outcomes through a process that is analogous to the valuation of mortality risk reduction (Figure 1). Although the valuation of nonfatal outcomes is predicated upon evidence linking specific diagnoses (e.g., nonfatal myocardial infarction) and corresponding healthcare encounters (i.e., hospitalization) to regulated pollutants,^5,6 there are various reasons why it may not be feasible to develop individual valuation frameworks for an expanding set of discrete, pollution-related outcomes. First, the factors constraining the generalizability of mortality valuation frameworks to older persons (described above) should be considered when similar approaches are used to value averted nonfatal outcomes. Second, there are potentially limitless combinations of diagnoses and variations in scope, severity, and duration therein.¹⁵ Third, disease-centered approaches to the valuation of patient preferences are poorly suited to older persons who often have multiple chronic conditions and concurrent functional limitations.¹⁶ Lastly, disease states may not be a valid surrogate for the disutility (e.g., functional decline, pain) and opportunity costs that accompany a diagnosis.¹⁶ This is especially pertinent to older persons who often value universal, disease-agnostic outcomes, such as the maintenance of functional independence, over avoided mortality.¹⁶

The incomplete valuation or exclusion of third-party expenditures

The EPA may also value the averted medical costs corresponding to both fatal and nonfatal outcomes, depending on their chosen measurement approach and whether there are empirical estimates of these costs.⁶ The total value conferred by health risk reduction includes both private valuation (measured by the VSL) and the net reduction in costs incurred by third parties.¹⁷ In practice, valuation often excludes the latter, including medical and long-term services and supports (LTSS), due to the traditional simplifying assumption that third party costs are either negligible or offset by premature mortality.¹⁸ There are several reasons why these assumptions may be inapt or violated in aging populations. First, the differential concentration of morbidity and functional impairment among older persons will increase as their absolute disabled-life expectancy lengthens.^10,19 Second, disabled life-years often necessitate intensive utilization of healthcare, as well as formal and informal LTSS. For example, the median survival for persons who are diagnosed with dementia after age 75 is approximately 4 years, over which time the annual costs (in 2010 dollars) for medical and LTSS expenditures are estimated to approach $56,290 per person.^20,21 Third, the expenditures supporting older persons’ utilization of LTSS may increase as the supply of informal caregivers wanes and these responsibilities are shifted to the formal healthcare sector.^14,22 Therefore, quantifying anticipated offsets in third party expenditures may become increasingly relevant to the valuation of expected health benefits.

An illustrative example

To illustrate these three limitations, we discuss the EPA cost-benefit analysis of the 1990 Clean Air Act (CAA) amendments.⁶ The results of the EPA analysis demonstrate that the benefits of this policy far exceeded its implementation costs, with mortality risk reduction accounting for greater than 90-percent of quantified benefits.^6,23 The EPA noted multiple sources of uncertainty affecting their valuation, several of which are relevant to population aging. Addressing these uncertainties should improve confidence in the future valuation of health benefits attributable to environmental policy. Importantly, we are not disputing the validity of the EPA analysis. Rather, we are highlighting uncertainties in the valuation of health benefits that may become increasingly relevant as the population ages.

The EPA analysis of the 1990 CAA amendments applies the population-average (age-invariant) VSL (7.4 million in 2006 dollars) to estimate the value of mortality risk reduction.⁶ The official population-average VSL value used by the EPA is informed by 26 studies published between 1974 and 1991, of which 21 were conducted within labor market cohorts.⁶ Of the remaining studies that elicited respondents’ stated preferences, the most recent was published in 1991. In addition to the poor age-representativeness of labor market cohorts, these dated studies may fail to capture time-varying determinants of the value that respondents assign to mortality risk reduction across the lifespan, such as changes in active life expectancy.

In its valuation of nonfatal outcomes, the EPA considers a restrictive set of diagnoses or healthcare episodes that the epidemiological literature has linked to the pollutant classes regulated under the 1990 CAA amendments.⁶ The manner in which the EPA values nonfatal health outcomes varies based on the available epidemiological data. For example, the valuation of pollution-related chronic bronchitis is predicated upon empirical evidence of respondents’ willingness to trade their own wealth to avoid this outcome.⁶ In contrast, the EPA valued nonfatal myocardial infarction using a combination of age-stratified lost earnings and two studies that estimate attributable direct medical expenditures.^6,24,25 Despite differences in these approaches to valuing nonfatal outcomes, neither explicitly accounts for age-related factors that might influence valuation, including multimorbidity, functional status, and other determinants of health-related quality of life. Estimates of lost earnings, moreover, would not be expected to generalize to retired older persons.

The approach used by the EPA to value averted medical costs also merits further consideration. These estimates included avoided hospital and emergency department encounters for specific diagnoses, including cardiovascular and respiratory diseases.⁶ Also included were a broader set of direct medical costs accompanying nonfatal myocardial infarction, namely, prescription drug and outpatient encounters.⁶ In the decades since the studies used to value averted medical costs were conducted, the magnitude and distribution of healthcare expenditures have changed considerably. For example, societal aging has contributed to continued growth in national healthcare expenditures as a percent of gross domestic product.²⁶ These studies also predate increased utilization of rehabilitation services and reductions in the average hospital length of stay.^27,28 In addition, the inclusion of only billable healthcare episodes precludes the valuation of considerable informal care costs, which are projected to increase as the population ages.²⁹ Estimating the averted healthcare costs for specific diseases will become more complex as additional age-related outcomes that have been linked to pollution, such as dementia,⁷ are included in valuation.

Recent Applications

The EPA Benefits and Mapping Analysis Program (BenMAP) is a tool used to estimate the economic value of health impacts attributable to changes in the concentration of regulated pollutant classes.⁷ This framework incorporates more recent epidemiological literature but has similar limitations to those of the 1990 CAA EPA Analysis. The population-average VSL informing the BenMAP valuation of premature mortality relies upon the same literature informing the EPA analysis of the 1990 CAA amendments, heavily weighting estimates from labor market cohorts.⁷ BenMAP also primarily values nonfatal outcomes that are discrete, such as hospitalizations for cardiovascular and respiratory illnesses. The estimated costs of these episodes are attributed to hospital admissions and lost wages,⁷ thereby excluding LTSS pursuant to functional decline and valuing the opportunity cost of illness using estimates that may not generalize to older persons. While BenMAP does incorporate age-related conditions, such as dementia, its accompanying cost estimates also exclude sizable LTSS expenditures. Moreover, its concentration-response functions for dementia are predicated upon case ascertainment using administrative claims from hospital encounters, which may not discern dementia better than chance.^7,30

Empirical Framework to Calibrate Valuation to an Aging Population

Explicit consideration of population aging and temporal trends that affect older persons’ longevity and health-related quality of life should improve the calibration of environmental policy valuation to changes in the age-composition of the population. We first suggest estimating the stated value that representative samples of older persons assign to hypothetical reductions in their risk of adverse health states and comparing these values to those inferred from labor market and other revealed preference data. While prior studies of older persons’ stated preferences have been conducted,⁹ few have been sufficiently age-representative or recent to capture changes in active- and disabled life expectancy and corresponding health-related quality of life. The addition of standardized risk-valuation survey models to longitudinal aging and population health surveys could provide a valuable mechanism by which to longitudinally capture older persons’ stated and, potentially, revealed preferences.

Second, in adapting stated preference frameworks to older persons, we recommend comparing the performance of approaches that elicit preferences regarding nonfatal universal outcomes, as opposed to discrete (i.e., incident diagnosis) health states or healthcare encounters (i.e., hospitalization). For example, functional decline is a universal outcome that transcends diagnostic categories as a common pathway in aging and disease. Functional decline may also be assessed using common units,³¹ thereby avoiding the challenges inherent to valuing an incremental diagnosis in persons with multimorbidity.¹⁶ Beyond its universality, the successive stages of functional decline are well-defined and often used to elicit patient preferences in clinical decision-making.¹⁶ In addition, common functional scales that assess disability with basic, instrumental, and mobility activities are strong indicators of prognosis and healthcare utilization.^16,32 The direct valuation of functional profiles may also attenuate information asymmetry that could otherwise bias respondents’ expectations of a particular diagnosis.

Because functional status is dynamic, we recommend testing the performance of risk-valuation frameworks that accommodate time-varying determinants of risk sensitivity within older cohorts.³³ This could involve adapting a validated approach that elicits respondents’ stated willingness to trade wealth for a hypothetical reduction in their risk of experiencing a specified illness profile and its corresponding pattern of health states.³³ A modified version of this approach could vary hypothetical functional profiles to capture differences in risk preferences across distinct durations of latency, impairment, potential recovery, and lost life-years.^15,33 This approach overcomes limitations of the current VSL framework by accommodating respondents’ sensitivity to preceding and successive health-related quality of life.^15,33 It is also better-aligned with the natural history of functional loss in older persons, which often involves multiple episodes of disability with interspersed periods of recovery.³⁴

Third, to better estimate the total value conferred by environmental risk reduction, we suggest measurement of the average third-party expenditures that correspond to defined functional profiles. This could be accomplished by adopting a modified societal perspective to aggregate costs to the public across the healthcare and social sectors.³⁵ Measurement would necessarily be constrained to domains that can be ascertained using longitudinal health and administrative claims data. For example, administrative claims could be used to determine the expenditures associated with healthcare encounters and utilization of formal LTSS. The costs of informal supportive services could be indirectly estimated using utilization patterns within longitudinal population health surveys and region-specific market wages for formal providers (i.e., home health aides) or a forgone wage cost approach.²⁰ Within the social sector, the cost of retrofitting homes to support accessibility needs could be measured using pertinent administrative claims for durable medical equipment, and region-specific market wages for costs corresponding to structural modifications. Utilization across each category of expenditures could then be averaged and linked to discrete functional profiles from the longitudinal health surveys.

Anticipated Challenges, Potential Responses, and Steps Forward

Addressing the empirical framework that we have outlined will require the collaboration of multiple disciplines over many years, making it well-suited to emerging team-science funding mechanisms.³⁶ Prior public opposition to cost-benefit analysis and use of the VSL, when misconstrued as a means of government agencies assigning a value to individual human lives, poses another challenge.³ Age-adjustment has led to similar public opposition because of concerns that it discounts the value of older lives.⁴ These challenges may be addressed by emphasizing that the VSL is an instrument that captures the value that individuals themselves assign to incremental changes in their own risks.^3,4 Emerging evidence, moreover, suggests that the benefits of pollution abatement disproportionately accrue to older persons, even after accounting for differences in age-related factors.¹

As work advances on tailoring environmental policy valuation to an aging society, additional logistical and empirical challenges should be considered. First, the longitudinal attributable fraction of lifetime functional loss or age-related comorbidities due to environmental exposures require improved assessment across multiple exposure categories and mechanisms. Understanding this relationship will require disentangling the timescale of anticipated health effects and determining whether these effects are regionalized or global. Adapting longitudinal population health surveys to enhance both temporal and geographic resolution should enhance researchers’ ability to examine the relationship between local environmental exposures and universal outcomes. Second, the use of universal outcomes (including functional status) should enhance ascertainment of third-party social sector and long-term healthcare expenditures but may not account for variation related to treatment strategies for specific diseases. The accuracy with which functional profiles predict third-party expenditures merits further evaluation across a broad set of diseases and age-strata to identify exceptions. Third, stated and revealed preference studies can be complementary because of their different strengths and limitations.³⁷ Understanding how to best contrast and combine stated and revealed preference studies may, therefore, improve the validity and accuracy of official VSL estimates. For example, estimates of older persons’ stated preferences should be compared, when possible, to their revealed preferences outside of the labor market context.³⁸

In the interim, it may be useful for the EPA to present bounded estimates of the value conferred by specific regulatory policies for different age-strata. To do so, researchers should consider potential differential risk reduction within older age groups and adjust the VSL for factors that affect its value over the life course, such as age-specific survival probabilities, while bearing in mind the current uncertainty regarding the best approach to age-adjustment.¹

Summary

We have identified and described central, albeit non-exhaustive, limitations of the valuation framework used by the EPA to measure the public health benefits conferred by environmental policy in the context of population aging. We have also articulated an empirical framework that could inform progress towards age-calibrated policy valuation. While there are clear tradeoffs between parsimony and granularity, the exclusion of age-related preferences and outcomes may distort the expected benefits of environmental policy. These considerations are especially relevant to aging societies, which include a substantial proportion of older persons who will incur prolonged periods of functional decline and disability with corresponding, intensive healthcare and LTSS utilization. Moreover, valuation that is predicated upon cohorts that were studied several decades ago may not capture changes in the appraisal of morbidity and mortality risk among older persons, especially as active life expectancy increases and quality of life during disabled life-years improves. Addressing these considerations may enhance the ability of regulators and policymakers to align environmental policy valuation with considerations that are germane to an aging society.

Key Points

• The approach used by the Environmental Protection Agency (EPA) to value the health benefits conferred by environmental regulatory policy is not well-calibrated to older adults due to their general exclusion from the samples upon which valuation estimates are predicated.

• The omission of universal aging outcomes and the corresponding health-related needs and preferences of older adults may also bias the valuation of environmental regulatory policy.

Why does this paper matter?

The relationship between unprecedented population aging and the valuation of environmental policy is poorly understood. Although older adults represent a rapidly expanding demographic and are differentially susceptible to environmental pollution, they are not often well-considered in the valuation of environmental policy. Because environmental policy valuation informs consequential regulatory decisions, calibration of these measures to the population age composition will be increasingly important as older adults comprise a larger share of the United States population.