Summarizing Population Health: Directions for the Development and Application of Population Metrics (1998)

Dan W. Brock

Brown University

This paper will discuss briefly some of the main ethical issues in the development or construction of summary measures of population health status and of the health benefits of interventions designed to improve the health status of a population. ¹ In a companion paper Norman Daniels will address issues of equity and distributive justice in the use of such measures for prioritization of health resources. Typical measures of the health status of a population at a point in time include the Health Utilities Index (HUI), ² and the Quality of Well-Being Scale (QWB). ³ Some such measure will be needed as well to calculate the burdens of various diseases as well as the benefits of interventions to reduce the burdens of disease. Typical summary measures of the benefits over time of health interventions include Quality-Adjusted Life Years (QALYs) and Disability-Adjusted Life Years (DALYs), each of which will have to make use of some point-in-time measure of health status like the HUI. Both QALYs and DALYs are intended to be comprehensive measures of the overall benefits of health interventions because they each combine changes in length of life and changes in HRQL, the two general forms of benefits of health interventions. ⁴

Both QALYs and DALYs are often employed in cost-effectiveness analyses (CEAs) that compare the aggregate health benefits secured from a given resource expenditure with different health interventions. The assumption underlying much of the use of CEAs is that limited resources (and, of course, as economists correctly remind us resources are always limited) should be used to maximize the aggregate health status of a population, or to minimize the burdens of disease for a population. Natural, even self-evident, as this assumption may appear to many health policy analysts and economists, both Daniels and I will argue that it assumes a utilitarian or consequentialist moral standard, or, more specifically, a standard of distributive justice, and that the utilitarian account of distributive justice is widely and correctly taken to be utilitarianism’s most problematic feature.

Early summary measures of the health status of populations were often measures of a single variable which stood as a more or less crude measure of the health of a population, or as a surrogate for a measure of the health of a population, such as life expectancy or infant mortality. For some purposes, and in the absence of more fine-grained data about population health, these single variable measures can sometimes provide useful information. Health interventions can also be evaluated for the impact they have on increasing life expectancy or reducing infant mortality. Since virtually no one disagrees that it is desirable to reduce infant mortality rates, we can evaluate interventions for their effects in doing so without raising the problem of how to assign relative values to different health outcomes.

The usefulness of life expectancy or infant mortality rates is clearly very limited, however, because they give us information about only one of the aims of health interventions— extending life or preventing premature loss of life—and they provide only limited information about that aim. They give us no information about another aim, at least as important, that of health interventions to improve or protect the quality of life by treating or preventing suffering and disability. Multi-attribute measures like the Sickness Impact Profile ⁵ and the MOS 36 ⁶ provide measures of different aspects of overall HRQL on which a particular population can be mapped, and an intervention assessed for its impact on these different components of health, or HRQL. Because these measures do not assign different relative value or importance to the different aspects or attributes of HRQL, they do not provide a single overall summary measure of HRQL. Thus, if one of two populations or health interventions scores higher in some respect(s) but lower in others, no conclusion can be drawn about whether the overall HRQL of one population, or from one intervention, is better than the other.

This limitation may not be serious in some contexts. For example, when evaluating some alternative interventions or pharmaceuticals in clinical trials, the impacts of the different interventions may be clustered in a limited domain of HRQL, and the different impacts in that domain of one intervention may be uniformly, or nearly uniformly, better than those of alternative interventions. Nevertheless, even in many clinical trials the outcomes of different interventions may be more multidimensional and conflicting in the sense that no one alternative dominates or is better than all the others in all of its effects on HRQL. For comparing overall population health between and among countries, states, or regions, and for resource prioritization or allocation at national, state, or regional levels, assignments of different relative value or importance to different health-related outcomes or effects are necessary.

Measures like QALYs and DALYs have the important advantage that they address both length of life and health-related quality of life, and they provide a basis for assigning relative value to length versus quality of life, as well as to different impacts on quality of life. The construction of any measure like the QALY or DALY requires a two-step process: first, different states of disability or conditions limiting HRQL are described; second, different relative values are assigned to those different conditions. Instruments like the HUI and the QWB have been developed to play this role. The determination of people’s different health-related conditions both before and after a particular health intervention is an empirical question that should be answered by appeal to relevant data regarding the burden of a particular disease and the reduction in that burden that a particular health intervention can be expected to produce; the overall HRQL

of a particular population can likewise be determined by empirical data regarding that population’s condition on the different attributes of a measure like the HUI. Needless to say, often the relevant data are highly imperfect, but that is a problem to be addressed largely by generating better data, not by ethical analysis.

The second step in developing measures like the HUI involves assigning relative values or utilities to the different conditions that reduce people’s HRQL. The developers of the DALY used expert health professionals to make these value judgments. This may have been convenient, but value judgments about the degree to which various conditions reduce quality of life is not a matter to be settled by professional expertise. Moreover, health professionals may have systematic biases that skew their value judgments about quality of life compared to those of ordinary persons. Other measures like the HUI and QWB use the value judgments of a random group of ordinary citizens to evaluate different states of disability or limitations in function. The utilities for different attributes and their levels in the HUI are shown in Table 1 .

One problem concerning which evaluations to use in examining different states of disability or functional limitation arises from the ability of individuals to adjust to their disabilities. This results in the disabled often reporting less distress and limitation of opportunity and quality of life from their disability than the nondisabled believe would occur. If the evaluations of disability states by the nondisabled are used for ranking different states of health and disability, then disabilities will be ranked as more serious health needs, but these rankings are open to the charge that they are biased by the ignorance of the evaluators of what it is like to live with the conditions in question. If the evaluations of the disabled themselves are used, however, the rankings are open to the charge of bias for the opposite reason—that is that the burden of disability has been unjustifiably underestimated because of the adjustment process that the disabled person has undergone. The dilemma here is in determining the appropriate evaluative standpoint, for ranking the importance of different disabilities, to avoid the potential for bias inherent in these differing perspectives. ⁷

Since the preferences for different states of disability or HRQL used to determine their relative values should be informed preferences, it is natural to think that the preferences of those who actually experience the disabilities should be used. The disabled should have a more informed understanding of what it is actually like to live with the particular disability in question than a person who has never experienced the disability. Nondisabled persons will often have false beliefs about what it is like to live with particular disabilities, beliefs which should not influence the relative values assigned to the HRQL of living with those disabilities. One way of avoiding that influence is to use the preferences of persons who have the disabilities in question. But this is to miss the deeper nature of the problem caused by adaptation to disabilities.

Why should we expect nondisabled persons to assign a lower quality of life to living with various disabilities than do persons who have those disabilities, even after correcting for the prejudices and false beliefs of the nondisabled about the disabled? In evaluating how bad life

with a particular disability would be, for example, on a scale on which zero represents death and one represents a full quality of life with no impairments, nondisabled persons would ask how seriously that disability would impair their pursuit of their life plans, that is the projects and activities that give their lives value and meaning. People who have lived with that same disability for a significant period of time usually will have adjusted and adapted. They will have given up the activities and projects that are no longer possible because of their disability, and substituted other new activities and projects compatible with it. Typically, both their objective ability to pursue their plans of life, as well as their subjective satisfaction with their lives, will have been improved upon by adaptation to their disability from what nondisabled persons correctly estimate their quality of life would be with that disability.

Fundamental to understanding the difficulty of adjustment to disabilities for preference evaluation of HRQL with various disabilities is that neither the nondisabled nor the disabled need have made any mistake in their different evaluations of the quality of life with that disability. They arrive at different evaluations of the quality of life with that disability because they make those evaluations from the standpoint of different plans of life, plans of life which differ specifically in the degree to which they are limited by the disability in question. People who have, and have adapted to, the disability can look back and see that before they became disabled they too would have evaluated the quality of life with that disability as nondisabled people now do. But this provides no basis for concluding that their pre-disability evaluation of the quality of life with that disability was mistaken. The problem that I call the perspectives problem is that the nondisabled and the disabled evaluate the quality of life with the disability from two different evaluative perspectives, or plans of life. They are different because of the adaptive change in plan of life and evaluative perspective that the disabled have made to their disability, but neither evaluative perspective can be shown to be mistaken as a result of that adaptation.

When measures like the HUI or QWB are applied across different economic, ethnic, cultural, and social groups, the meaningful states of health and disability important in different groups may vary; different groups may place significantly different relative importance or value on the same states of health and disability. For example, in a setting in which most labor is manual, limitations in physical functioning will have greater importance than they would in a setting in which most individuals are engaged in non-physical, knowledge-based occupations, where certain cognitive disabilities are of greater importance. Different evaluations of health conditions and disabilities are necessary for groups with significantly different relative needs for different functional abilities, but then cross-group comparisons of health and disability, and of the relative value of health interventions in those different groups will not be possible. The health program benefits will have been measured on two different and apparently incommensurable valuational scales. These differences will be magnified when summary measures of population health are employed for international comparisons among very disparate countries.

The ethical evaluation of health differences is complicated further when great differences between and among groups or countries, such as in the level of economic development or the treatment of disadvantaged groups, are themselves unjust. For example, many measures of HRQL include some measure of subjective satisfaction or distress, which will be importantly influenced by people’s expectations. In a society which has long practiced systematic discrimination against women, for example, women may not be dissatisfied with their unjustly disadvantaged state, including the health differences that result from that discrimination. The fact

that victims are sufficiently oppressed that they accept an injustice as natural should not make its effects less serious as measures of HRQL with a subjective satisfaction or distress component would imply.

A different problem with any use of QALYs or DALYs for assessing the benefits of health interventions is that they appear to discriminate against the disabled. ⁸ For example, an intervention that extends for 10 years the life of patients with a disability that reduces their HRQL to 0.75 would produce 7.5 QALYs for each patient receiving the intervention. By an intervention that extends for 10 years the lives of patients with an otherwise unimpaired HRQL of 1.0 would produce 10 QALYs for each patient receiving the intervention. The use of QALYs for evaluating and prioritizing life-saving interventions appears to discriminate against the disabled and place less value on their lives by assigning less value to extending their lives simply because of their disability. For this reason, some have argued that quality weights should not be used in the evaluation of saving lives, as long as the individuals in question consider their lives worth living. ⁹ The problem of whether QALYs unjustly discriminate against the disabled remains unresolved.

What does it mean to say that the evaluative perspective for determining relative quality of life of different states of disability is individual or social? Consider again the HUI scale in Table 1 . It contains seven attributes, or areas of functioning, with from three to five levels of function within each attribute or area. How can the relative quality of life, or health utility, be determined for the different levels of function within the seven attributes? One natural approach is to ask individuals what their relative quality of life would be, on a scale in which one represented full, unimpaired function and zero represented death, for each level of function within each attribute. There are several different technical measures for obtaining the relative values, such as standard gambles and time trade-offs, but their details for the most part do not bear on the point of concern now.

Suppose individuals can tell us that their quality of life would be reduced from 1.00 to 0.86 at level three of attribute one if they were able to see, hear, or speak with limitations, even with equipment; they also give us utility levels for each of the other attributes and levels. Because different individuals are likely to assign somewhat different utilities or quality of life to the various attribute levels, let us suppose that we take the mean (or perhaps the median) numbers assigned by a randomly selected group of individuals. The actual health utility levels for the various attributes and levels of the HUI are shown in Table 1 . Although this approach takes account of differences between individuals within a society in the relative quality of life they assign to different levels of impaired function, many of which will arise from their different social roles, the perspective for assigning relative quality of life weights or utilities remains individual in the sense with which I am concerned here. It is individual because it asks individuals how much an individual’s (or their own) quality of life would be reduced if he or she suffered the various impairments of function.

This individual perspective is appropriate for a number of uses to which a summary measure of population health status might be put. It generally would be appropriate for

monitoring the health status or overall burden of disease in a population, as well as across different populations, as long as the mean (or median) utility weights did not vary substantially in the different populations. It would be appropriate also for comparing alternative health interventions for a given group of patients with a particular medical condition. More generally, the individual perspective is correct for evaluating alternative interventions, or for monitoring changes over time in health status or the burdens of disease, for the same person or group of persons.

In other contexts, the summary measure of health may be used for a quite different purpose—resource prioritization of alternative health interventions each of which would benefit not the same persons, but different persons; for example, a cost-effectiveness analysis of treatment interventions for different diseases. Employed in this context, here is an example of what I believe is one of the most ethically implausible implications of the individual perspective of the HUI. According to the HUI, the benefit of saving the life of a person who will live at a full quality of life level of 1.0 for a year, with no impairment of function in any of the seven attributes or areas of function, is equivalent in the aggregate benefit produced to keeping 20 different otherwise healthy individuals from having to use eyeglasses or a hearing aid to see or hear for a year (which has a utility level of 0.95); each produce 1 QALY. Here, the HUI is being employed in a cost-effectiveness analysis that weighs the trade-off between using limited resources to meet the different health needs of different groups of individuals. Do people who assign the utility level of 0.95 to requiring equipment to see or hear or speak mean by that assignment that saving one healthy person’s life is of equal importance to keeping 20 persons from having to use eyeglasses or a hearing aid? It is highly doubtful that people are thinking of such trade-offs between or among different persons or groups when they assign utility levels to the different attribute levels in the HUI. They do not understand or intend their assignments to have those implications, and they reject in their explicit trade-offs these inferred trade-offs from their utility assignments in measures like the HUI.

This individual perspective of aggregating QALYs employing the HUI displays two distinct difficulties. The first is that it does not reflect the relative ethical importance people give to saving life in comparison to health benefits that improve the quality of life of other individuals or groups. As the example of eyeglasses and hearing aids shows, the HUI allows inferences about what people’s trade-offs are between saving life and improving the quality of life, but the inferred trade-offs do not match people’s explicit trade-offs between saving life and improving quality of life. QALYs calculated with measures like the HUI do not give the relative weight to saving life versus improving HRQL that people give when asked to make explicit trade-offs of that sort between different individuals or groups. I consider a different measure below that attempts to remedy this defect.

The second difficulty is that calculations of aggregate QALYs from different health interventions fail to reflect the ethical importance people place on the fact that health benefits to different individuals or groups are being traded off or prioritized. When summary measures of population health status like the HUI are used to evaluate alternative interventions that will benefit different individuals, as opposed to alternative interventions that will benefit the same individuals, issues of equity and distributive justice are raised. There are at least two broad approaches for taking account of this difference. One is to restrict the use of summary measures like QALYs calculated with instruments like the HUI to the evaluation of health interventions that serve the same individuals; for example, alternative treatments for the same patients. In this

approach, we explicitly recognize that people do not believe that the social value or importance of different health interventions serving different people or patients can be determined simply by comparing the aggregate health benefits in QALYs to the different individuals served by the two interventions. These determinations involve issues of distributive justice between persons in which it matters who gets what benefits, not just what the aggregate benefits of the different interventions are. We might use QALYs to determine the aggregate health benefits of interventions serving different individuals, but should do so only with the clear recognition that this will not tell us the overall social value people would place on these interventions. That can only be determined by also addressing concerns for equity, such as the unsolved rationing problems that Daniels discusses in his paper.

The second approach seeks to develop a measurement tool appropriate to the evaluation of interventions that serve different groups of individuals. The most prominent example is Erik Nord’s “person trade-off” approach which explicitly asks people how many outcomes of one kind they consider equivalent in social value to “X” outcomes of another kind, where the outcomes can be for different groups of individuals. For example, people can be asked, for two diseases of equal initial severity, how many patients with the disease whose treatment is fully effective would be equivalent in social value to treating 100 patients with the other disease whose treatment is only partially effective and so results in a lesser health improvement for each patient treated; in this example, Nord found that people tended to give more weight to initial severity of illness than to the degree of health improvement, “saying that they would prefer the ‘less effective’ program even if it treated only one person more than the ‘more effective’ program.” ¹⁰

The person trade-off approach is designed to permit people to incorporate concerns for equity or distributive justice into their judgments about the social value of alternative health programs. There has been relatively little exploration and use of this methodology in comparison with the studies and methodological work on measures of aggregate QALYs, in part because many health policy analysts and health economists assume, often with little or no argument, that the social value of health programs is the sum of the individual utilities produced by the program. That utilitarian assumption is rejected by most philosophical work on distributive justice, as well as by the preferences ordinary people express for different health outcomes and programs. Because the main specific issues of equity, or “rationing problems,” together with the possibility and difficulties of incorporating equity within the health evaluation measure, are explored in Daniels’ paper for this workshop, I will not pursue the person trade-off approach further. But I do emphasize that for purposes of resource prioritization and allocation, the social approach is the proper perspective, whether by means of a methodology like the person trade-off that seeks to incorporate people’s concern for equity within the measure of the social value of health programs, or by separate attention to issues of equity outside the summary measure of population health.

Within moral and political philosophy, as well as among ordinary people, debate and disagreement continue about utilitarian and nonutilitarian accounts of distributive justice. Individual approaches to the value of health programs, which measure QALYs using an instrument like the HUI, and a social approach using the person trade-off method differ in their sensitivity to whether people hold utilitarian or nonutilitarian views. The individual approach which assesses aggregate QALYs. produced by different health interventions is insensitive to, and unable to reflect, people’s concerns for equity and the distribution of benefits. The social

approach using person trade-offs, by contrast, reflects whether different people’s evaluations are utilitarian or nonutilitarian; utilitarians and nonutilitarians will make different person trade-offs, to which the person trade-off method will be sensitive.

QALYs have been widely used in health care and other contexts to compare the outcomes of different resource allocations and health programs. QALYs can be used to measure the difference that a health care program makes in the expected years of life and in the quality of those years of life for the persons affected by the intervention. The QALY measure assumes that an additional year of life has the same value regardless of the age of the person who receives it, assuming that the different life years are of comparable quality. A year of life extension for an infant, a 30-year-old, and a 75-year-old all have the same value in QALYs, and, in turn, in a cost-effectiveness analysis using QALYs, assuming no difference in the quality of the year of life extension. This is compatible, of course, with using age-based quality adjustments to reflect differences in average quality of life of a population at different ages. For example, if average quality of life in a population of persons at age 85 is less than that of persons at age 45, a year of life extension for 45-year-olds would have greater value in QALYs than would a year of life extension for 85-year-olds.

In the World Bank Study, World Development Report 1993; Investing in Health, ¹¹ the alternative DALY measure was developed to measure the burden of disease. Probably the most important ethical difference between QALYs and DALYs is that DALYs assign different value to a year of life extension at the same level of quality, depending on the age at which an individual receives it; in particular, life extension for individuals during their adult productive work years is given greater value than a similar period of life extension for infants and young children or the elderly. The principal justifications offered for this feature of DALYs were the different social roles that individuals typically occupy at different ages and the typical emotional, physical, and financial dependence of the very young and the elderly on individuals in their productive work years. ¹²

I believe this justification of age-based differences in the value of life extension adopts an ethically problematic social (in a different sense of “social” than that used in the preceding section) perspective on the value of health care interventions that extend life, or maintain or restore functionó that is, an evaluation of the benefits to others of extending an individual’s life, or maintaining or restoring his or her function, in addition to the benefit to that individual of doing so. This social perspective is in conflict with the usual focus in clinical contexts only on the benefits to the individuals who receive the health care interventions in question. Typical practice in health policy and public health contexts is more ambiguous on this point, since there benefits to others besides the direct recipient of the intervention are sometimes given substantial weight in the evaluation and justification of health programs. For example, with treatment programs for substance abuse, the benefits of reductions in lost workdays and in harmful effects on family members of the substance abuser. Using this social perspective is ethically problematic because it gives weight to differences between individuals in their social and economic value to others; it can be argued that this discriminates against persons with fewer dependencies and social ties and so is not ethically relevant in health care resource allocation. The social

perspective justifying the DALY measure is ethically problematic, in a way the alternative QALY measure is not, if the value of health benefits for individuals should focus on the value to those individuals of the health benefits, not on the social value for others of those health benefits.

Placing different weight on life extension at different ages, however, might be justified ethically if done for different reasons. For example, Daniels has argued that because everyone can expect to pass through the different stages of the life span, giving different value to a year of life extension at different stages in the life span need not unjustly discriminate against individuals in the way giving different weight to life extension for members of different racial, ethnic, or gender groups would unjustly discriminate. ¹³ Each individual can expect to pass through all the life stages in which life extension is given different value, while he or she is a member of only one race, ethnic group, and gender. Thus, use of DALYs does not constitute unjust age discrimination comparable to gender, ethnic, or racial discrimination.

Moreover, individuals and their society might choose to give lesser weight to a year of life extension beyond the normal life span than to a year of life extension before one has reached the end of the normal life span. People’s plans and central long-term projects will typically be constructed to fit within the normal life span, and so the completion of these central projects will typically require reaching, but not living beyond, the normal life span. ¹⁴

There are significant differences in the life expectancies of different groups in American society, for example between genders and among racial and ethnic groups, as well as differences correlating with differences in socioeconomic status. In a broader international context the differences in life expectancies within and between different countries are much larger. Should these differences affect calculations of the life years gained by life-extending health care and public health interventions? An accurate estimation of the additional life years actually produced by specific health care or public health interventions should not ignore differences in life expectancies that are not caused by the particular condition that the health care intervention affects.

Yet the differences in life expectancy between and among different racial, ethnic, and socio-economic status groups, as well as the very large differences among life expectancies in economically developed and poor countries, are often principally a result of unjust conditions and deprivations suffered by those with lower life expectancies. It would seem only to compound those injustices to give less value to life-saving or life-extending interventions for groups with lower life expectancies caused by the unjust conditions and deprivations from which they suffer.

Differences in life expectancies between the genders, on the other hand, are believed to rest in significant part on biological differences, not on unjust social conditions. Whether the biologically based component of gender differences in life expectancies should be reflected in measures like QALYs or DALYs is more controversial. On the one hand, the lower life expectancy of men does not reflect an independent injustice, but, on the other hand, it is explicit public policy, required by law, not to take account of this gender-based difference in most calculations of pension benefits and annuity costs. The developers of the DALY explicitly chose to use a single uniform measure of life expectancy (except for the biologically based gender

difference), specifically that observed in Japan which has the highest national life expectancy, to measure gains from life-saving interventions. They justified their choice in explicitly ethical terms as conforming to a principle of “treating like events as like.” ¹⁵

I note below even more briefly two other ethical issues which are important in costeffectiveness analyses of health programs and in prioritization of health interventions and programs; the first arises directly in constructing a summary measure of increases and decreases in population health. Space limitations preclude any exploration of the details of these two issues.

It is both standard and recommended practice in cost-effectiveness analyses, in health care and elsewhere, to assume a time preference and thus to apply a discount rate to both the benefits and costs of different programs under evaluation. ¹⁶ It is important to be clear both about what precisely the ethical issue is in whether health benefits should be discounted, as well as why it is important for policy. It is not controversial that a discount rate should be applied to economic costs and economic benefits. The ethical issue is whether a discount rate should be applied directly to changes in well-being or health, and in particular to benefits in the form of increased well-being from health interventions. Is an improvement in well-being, such as a specific period of life extension, a reduction in suffering, or an improvement in function, extending, say, for one year of less value if it occurs 20 years from now than if it occurs next year?

Distant benefits are appropriately discounted when they are more uncertain than proximate benefits. Proximate benefits, such as restoration of an individual’s function, also are of more value than distant benefits if they make possible a longer period of benefit by occurring sooner. But neither of these considerations require the use of a discount rate—they will be taken account of in the measurement of expected benefits of alternative interventions. The ethical question is whether an improvement in an individual’s well-being is of lesser value if it occurs in the distant future than if it occurs in the immediate future, simply and only because it occurs later in time. This is a controversial issue in the literature on social discounting and my own view is that no adequate ethical justification has been offered for applying a discount rate directly to health and well-being. The avoidance of paradoxes that arise if a discount rate is applied to costs and the same discount rate is not applied to benefits, has influenced many economists to support use of the same discount rate for costs and benefits, ¹⁷ but I believe these are properly dealt with not through discounting, but instead through directly addressing the ethical issues of equity between different generations that are raised.

The policy importance of this issue is relatively straightforward in the prioritization of health care interventions. Use of a discount rate for evaluating alternative health care programs that take significantly different lengths of time to produce their benefits leads to giving an unwarranted priority to programs producing benefits more rapidly. Put differently, a program that produces benefits in health and well-being 20 years into the future will be given lower priority than an alternative health care program that produces substantially less overall improvement in health and well-being, but produces that improvement much sooner. Many public health and

preventive interventions—for example, vaccination programs and changes in unhealthy behavioróreap their health benefits years into the future. If those benefits are inappropriately discounted, they will receive less priority and result in a health policy that produces fewer overall health benefits over time than could have been produced.

It is widely agreed that cost-effectiveness analyses in health should reflect the direct health benefits for individuals of their medical treatment, such as improving renal function or reducing joint swelling, and of public health programs, such as reducing the incidence of infectious diseases through vaccination programs. The direct costs of medical treatment and public health programs, such as the costs of health care professionals’ time and of medical equipment and supplies, should also be reflected. But medical and public health interventions typically also have indirect benefits and costs. Some disease and illness principally affect adults during their working years, thereby incurring significant economic costs in lost workdays associated with the disease or illness, whereas other disease and illness principally affect either young children or the elderly who in each case are not typically employed and so do not incur lost wages or work time from illness. Should an indirect economic burden of disease of this sort be given weight in prioritizing among different health care interventions?

From a broad utilitarian perspective encompassing all effects of disease and of efforts to treat or prevent it, indirect benefits and costs are real benefits and costs, even if not direct health benefits and direct treatment costs. They should be reflected in the overall cost-effectiveness accounting of how to use scarce health resources to produce the maximum aggregate benefits. Is there ever a moral reason to ignore these indirect costs and benefits in health resource prioritization? Giving priority to the treatment of one group of patients over another because treating the first group would produce indirect benefits for others (for example, other family members who were dependent on these patients) or would reduce indirect economic costs to others (for example, the employers of these patients who incur less lost work time) could be held to fail to treat each group of patients with the equal moral concern and respect, and in particular the equal moral concern for their health care needs, that all people deserve. Instead, giving lower priority to the second group of patients simply because they are not a means to the indirect benefits produced or indirect costs saved by treating the first group of patients gives the second group of patients and their health care needs lower priority simply because they are not a means to these indirect benefits or cost savings to others. This Kantian reason for ignoring indirect benefits and costs could serve as a moral basis of the idea of “separate spheres,” that is, that the purpose of health care and of public health is health and the reduction of disease, and so only these goals and effects should guide health care and public health programs. ¹⁸

The six ethical issues discussed very briefly above are all issues involved in developing a summary measure of population health, and of changes in population health, that permits equitable evaluation of populations’ health or health programs. In each case, there are important ethical and value choices to be made in constructing the measures; the choices are not merely technical, empirical, or economic; they are moral and value choices as well. There are further ethical issues in the use of these measures that Daniels takes up in his companion paper.