Appendix Cost-Effectiveness Analysis and Fairness

F. M. Kamm

Pages: 297–310



This appendix considers some different views of fairness and whether they conflict with the use of a version of Cost-Effectiveness Analysis (CEA) that calls for maximizing health benefits per dollar spent. Among the concerns addressed are whether this version of CEA ignores the concerns of the worst off and inappropriately aggregates small benefits to many people. I critically examine the views of Daniel Hausman and Peter Singer who defend this version of CEA, and of Eric Nord who criticizes it. I come to focus in particular on the use of CEA in allocating scarce resources to the disabled.1

Cost-Effectiveness Analysis (CEA) in medical care tries to maximize health benefits produced per dollar spent. Its use is recommended when society cannot afford every form of health care and must choose what to provide. Yet it is often taken as a truism that there can be deep conflicts between maximizing benefits and distributing fairly, in general. For example, the philosopher Robert Nozick imagined a “Utility Monster” (where utility is [roughly] experiential well-being) such that for any resource up for distribution, one always produces more additional benefit at less cost if one gives the resource to the Monster rather to others even though he is already much better off than they are. This would result in one person getting all the additional benefits while others get none. This seems unfair.

However, CEA cannot result in this most extreme form of unfairness because of limits that result from how it calculates benefits. Each additional year of very healthy life is given a value of ɪ; no one can get more than ɪ per year. Still, it is possible that only those who are already very healthy can achieve many additional years at a value of ɪ at low cost if they are saved from an otherwise fatal bacteria. Maximizing health benefits per cost would imply helping them rather than people who are not as healthy and can achieve only fewer additional years at a value of less than ɪ and at higher cost if they are saved from the same threat. This too seems unfair.

Why does it seem unfair to help only the Utility Monster and the healthy people? Fairness is about how one person is treated relative to another. This is by contrast with a notion such as justice which need not be comparative; that is, we could decide what justice requires us to give a person in virtue of his characteristics independently of considering what anyone else is owed. Hence, we could treat someone justly in giving him what he is owed and yet increase unfairness if we treat him justly while not treating anyone else justly. (For example, we might punish some who deserve this even if we cannot punish all who do. This case also shows that fairness is only one moral dimension on which we can evaluate how we treat people or states of affairs we produce; it is possible that we should sometimes override fairness to be just or to achieve some other moral value.)

According to what measure shall we compare people to see if each is being treated fairly relative to others? Suppose we think that all that fairness requires in allocating benefits (via allocating resources) is that a certain amount of benefit be given the same value regardless of the person who will be benefited; no extra value should be assigned to a certain amount of benefit in person A rather than in person B. We could call this the Simple Standard. According to this standard, Nozick’s Utility Monster and our Bacteria Case need not involve unfairness, for we could give the same value to a certain amount of benefit in the lives of everyone but it happens that more benefits can be produced in the Utility Monster and in the already healthy. These examples suggest that there may be more to fairness than the Simple Standard.

Indeed, there are different views about what fairness requires. I will consider whether, according to some of these views, problems of fairness arise in the medical context if we use a version of CEA that always emphasizes maximizing health benefits per dollar spent. Without pretending to settle the matter, I will raise some issues to consider. However, it is important to realize that problems with this version of CEA need not imply that it is never consistent with fairness to use some form of cost-effectiveness evaluation. For example, it seems fair and right to treat one hundred people equally well with a cheap drug rather than with an expensive one, other things equal. It would also be fair and right to use a drug with which we can save two hundred people rather than use an equally costly one with which we can save only one hundred of these people.

1. Chances in proportion to need. Some think that when we cannot help everyone, fairness requires that people get a chance for medical care in proportion to their need for it, regardless of outcome in terms of CEA. If this view were correct, someone who has a weak need for a scarce resource should get a small chance to get it. But does fairness really require giving a small chance to someone who needs the resource to cure his sore throat so that if against great odds he wins, then someone else who needs the resource to save her life dies? Would we be overriding fairness merely in order to achieve a better outcome if we did not give the person with a sore throat a chance? I suspect not for in other cases achieving a better outcome would not lead us to override what fairness really requires. For example, suppose a doctor and a janitor both equally need a scarce life-saving medicine. If the doctor survives, he can save someone else’s life from another illness while the janitor cannot. Although we could achieve a better outcome if we save the doctor (two people saved rather than one), this may not be sufficient reason to deny equal chances to the doctor and janitor who both need the scarce medicine. If we would not override fairness in this case to achieve the better outcome, this suggests that in denying the person with the sore throat a proportional chance we are also not overriding fairness for the sake of a better outcome but rather we do not think fairness requires his having a chance.

What about chances in proportion to need of a group of people? Imagine that we can produce more cures per dollar if we treat six people who have one fatal disease rather than only treating five who have a different fatal disease. In this case the route to maximizing health benefits involves giving a life-saving benefit to more rather than fewer equally needy people. Hence, this case raises the question of whether it is unfair to save a greater number of people rather than to give each group a chance to be saved in proportion to the need of each member multiplied by the number of people in the group. Some think that fairness does not demand giving chances in proportion to need in the group but requires counting numbers of people, balancing one person of similar need (and, perhaps, expected outcome) against another and allowing the greater number to get the resource. On this view fairness does not require giving some chance to be helped to fewer people, when all suffer from equally serious problems.

Nevertheless, even on this view of fairness it may be right to give equal chances to be saved to the two groups if they each contain the same number of equally sick people when we only have enough resources to treat one type of fatal disease. But proponents of CEA should see no reason to give equal chances if outcome per dollar would be the same. It is only if we take seriously the personal perspectives of each person, and so recognize that each person is not indifferent to whether he or someone else survives, that we see why fairness could sometimes require giving equal chances to different people even when their need and outcomes are the same. (If we take seriously the perspectives of different people, we might even think it is wrong to deprive one person of his 50 percent chance to be treated merely because we would get a slightly better outcome if another person were treated.)

In sum, I have argued that while CEA does not necessarily contravene fairness in not giving chances to individuals in proportion to their need, it may fail to recognize an appropriate role for giving equal chances when need and outcome would be the same.

2. Priority to the worse off. Another possible fairness concern about CEA is that it is indifferent to whether an equally cost-effective benefit, such as relief from a certain amount of pain, goes to someone moderately ill or to someone severely ill. Some think fairness requires that the worse off be preferred. Indeed, it might be thought that fairness requires providing even a somewhat smaller benefit to those who are severely ill rather than a larger benefit to those who are moderately ill, holding cost constant. Such a “prioritarian” view of what fairness requires implies, roughly, that it is reasonable that the claim to benefits of those who are worse off should be weighed more heavily than the claim of those who are better off because it is right to give priority to improving the condition of a worse-off person before improving someone who is already better off than he is.2 This view of fairness implies, contrary to the Simple Standard, that a given benefit in one person sometimes has greater moral value than the same benefit in another person. (However, giving priority to the worse off is not the same as always taking care of the worse off regardless of benefit that can be achieved.)

3. Anti-Additive Aggregation.3 A third fairness concern about CEA is that providing minor, inexpensive health benefits (such as teeth fillings) in many people may be more cost-effective than providing bigger, more expensive benefits (such as treating appendicitis) in a few, but fairness may require giving greater weight to the latter. This issue arises because CEA permits adding small benefits to each of many people to produce a large aggregate benefit that is then weighed against a smaller aggregate benefit composed of adding bigger benefits to a few people. The question is when it is fair to additively aggregate and weigh smaller benefits to some people against bigger benefits to others to decide how to allocate scarce resources.4 This question about aggregation is sometimes related to the issue of giving priority to the worse off when the small benefits would go to many people already better off and the bigger benefits would to go to a few people more severely ill.

4. More on 2 and 3. Now consider the second and third concerns in greater detail by examining some responses to them. With regard to the second, some think that willingness to help the severely ill even when this produces fewer benefits per dollar need not depend on a prioritarian conception of fairness. Rather it can reflect compassion for those in dire straits. For example, Dan Hausman argues that CEA is the reasonable, rational, and not unfair way to decide how to allocate medical resources but we sometimes override it because of compassion for the severely ill. On this view, compassion can conflict with reason and it is compassion rather than a reasonable view of fairness that can lead us to help the severely ill when doing so conflicts with CEA.

One problem with this view is that it conflicts with the possibility that it is fair and reasonable to give priority to treating those who are only moderately ill rather than to those slightly ill even though compassion is not triggered for the former in the way it is for those in dire straits. Similarly, we do not now feel great compassion for 20-year-olds who we know will die in 30 years at age 50 by comparison with 20-year-olds who we know will die at age 65. Yet we might still think it is morally right to invest in research that will buy five more good years for those who would otherwise die at 50 rather than in research that will buy 10 more good years for those who will otherwise die at 65, even if this conflicts with CEA. Presumably, this is because it seems reasonable to help people who would be worse off when they die (in having had shorter lives) instead of people who would be better off (in having had longer lives) even without help. This reflects a prioritarian conception of fairness.

Another problem with the view that it is compassion rather than a reasonable conception of fairness that sometimes conflicts with CEA is that we often override compassion to do what is morally reasonable. For example, we may feel greater compassion for an incurably blind person who will also have to deal with a second problem if his arthritis is not treated than for a sighted person who will become (only) nearly blind if his eye problem is not treated. In this case, holding other factors constant, the blind arthritic will be the worst-off person if he is not treated. Yet it seems morally acceptable and reasonable to cure the more severe condition of near blindness rather than (what is here assumed to be) the less disabling condition of arthritis when we cannot do both. Suppose it is morally right to resist the call of compassion (as well as the call to help the person who would be worst off) in this case. Then perhaps in other cases when we do not give up on helping the worst-off person even though helping him conflicts with CEA it is because giving up would be contrary to reason and fairness rather than to compassion. This would imply that sometimes CEA is not the reasonable and fair approach.

Further support for the view that CEA does not necessarily coincide with what is reasonable and fair comes from the third concern about additive aggregation. That is, is it always fair to additively aggregate small health benefits to many people and weigh that aggregate against the smaller aggregate of bigger benefits to fewer people, when the benefits to each group cost the same? Suppose that each of many people has a mild headache and is otherwise already much better off health-wise than someone whom we can save from appendicitis. Suppose that none of the many people is a compassionate person and each would give up no more than the money for an aspirin that could cure his headache in order to help a dying person. But there are so many of these people with mild headaches that the additively aggregated harm of many headaches that would occur if each sacrificed his aspirin money is greater than the harm prevented in using the money to save the person with appendicitis. Though none of the people is rescuing the one person because of great compassion for him, presumably they would not refuse to give up the aspirin money for him on the grounds that the sum of losses to all of them is so enormous by comparison to one person’s loss of life. No one has to be compassionate in order to realize that it would be a bizarre mistake of reason to treat the very large sum of small losses to each of the many people as if it had the same moral significance as a very large loss to a single person and allow him to suffer it to prevent a smaller loss to each of many other people.

Some suggest that the fair way to decide what to do in such cases, when the small harm (such as a headache) is occurring to each of many separate persons, is to compare in a pairwise fashion how much harm would be suffered and avoided by a severely ill person depending on whether he is helped with how much harm would be suffered and avoided by each of the many depending on whether they are helped. Fairness is comparative but, on this view, it requires comparing how individual persons fare one person at a time depending on what we do. Suppose that no one of the many will suffer anywhere near as great a loss as the single person would and they are already better off than he is. Then if our view of fairness combines pairwise comparison with priority to the worse off, curing a headache in each of many people would never take precedence over curing a much more serious condition in even one much worse-off person. A conception of fairness that involves these two components—pairwise comparison and prioritarianism—would support concerns about the fairness of CEA.

5. Singer. By contrast, Peter Singer, a philosopher who supports CEA, believes it is morally correct to additively aggregate smaller individual benefits to better-off people and weigh the aggregate against a bigger individual benefit to a worse-off person. For example, in a New York Times Magazine article on rationing,5 he considered how to compare the health benefit achieved in saving one person’s life with curing a serious but non-life-threatening condition such as quadriplegia in another. He tells us to consider the trade-off each person would reasonably make in his own life between length of life and quality of life. Suppose every person (already disabled or not) would be indifferent between living ten years with quadriplegia or living five years non-disabled. This seems to indicate that people take living with quadriplegia to be half as good as living nondisabled.6 Singer thinks that such data would show that using our resources to cure two quadriplegics is just as good as saving someone else’s life when all three people would have the same life expectancy if helped (for example, ten years). His reasoning seems to be that if someone would give up five out of ten years of his own life rather than be quadriplegic, that would justify curing one person’s quadriplegia rather than saving someone else’s life for five years; the combined benefit of curing two people with quadriplegia would therefore justify not saving someone else’s life when doing so would give her an additional ten years.

Several things seem problematic about this reasoning. First, in the trade-off between quality and quantity that a person might make in his own life, it is that person who benefits from the trade-off. When we make trade-offs between different people, the people who get the improved quality of life are not the same people who suffer the loss of more years of life. Hence, trade-offs between people may raise different moral issues than trade-offs within one life.7 This is related to the point made earlier that fairness considerations arise when we take seriously that different people are not indifferent to whether benefits and losses fall in someone else’s life or their own. Second, the conclusion that curing two quadriplegics who would live for ten years anyway is equal to saving for ten years someone else who would otherwise soon die depends on adding the benefit to two people to weigh it against the loss to the single person. We can see how problematic this is by considering the following example: Suppose that the trade-off test within one person’s life showed that a small disability (e.g., a permanently damaged ankle) made life slightly less than 95 percent as good as a nondisabled life. This implies that a person would rather live nine and a half years without the small disability than ten years with it. On Singer’s view, this implies that we should cure one person’s small disability rather than save someone else who will otherwise soon die so that he can live for an additional half year. It also implies that we should cure the small disability in twenty-one people rather than save someone so that he can live for an additional ten years. This sort of problematic reasoning may have led to the rationing plan in Oregon many years ago in which resources were to be allocated to cap many people’s teeth rather than save a few people’s lives.8

A third concern about Singer’s reasoning is that someone’s imagined sacrifice of five years of life to avoid quadriplegia is imagined to leave him with five years of life instead of ten and he need not be willing to give up most or all of the remaining five years, thus dying immediately, to avoid quadriplegia. Yet if we choose to cure the two quadriplegics, the person who dies as a result of our choice will die immediately so that we may cure a condition whose victims, it is assumed, would not die immediately to avoid it.

Note that the problem with always additively aggregating small benefits can exist independently of aiming to give priority to treating those more severely ill. For suppose all patients have the same disability. We have a choice of making very small improvements in the degree of disability in each of a great many patients or providing a complete cure to one patient. It would not be morally unreasonable (and might not be unfair) to do the latter, for we then make a significant difference in this person’s life rather than a barely perceptible difference to each of many others. This is so even though additively aggregating the many barely perceptible differences across persons creates an enormous total difference.9

6. Risk. Notice that in many of the cases we have considered if a particular person is not helped, he will certainly suffer a great loss and be worse off than others who will certainly avoid only small losses and be better off than he. But we may also consider the role of uncertainty and risk in deciding what is fair. We know that it can be reasonable for each individual to take a small risk of a large loss (such as losing many years of life) if this is the price of having a high probability of getting a smaller benefit. For example, someone might run a small risk of dying from an aspirin in order to get a high probability of relief from a nonlethal headache. If everyone in a community does this, in a large enough population it is certain that someone will die from an aspirin though each person had only a very small chance of dying. It seems morally permissible to allow individuals to expose themselves to such a small risk of the large loss for the sake of the small benefit. This is so even if we know that someone who took the risk that came to fruition will certainly die because there will then be nothing we can do to save him. However, when it is still possible to save this person whose risk of dying has gone from small to certain, or when someone was always known to be the person who would die, fairness may require aiding that person if we can, for example, with all the aspirin that would otherwise be used to prevent many people from each having a headache.10

Suppose there are a few people who will certainly die unless we treat them with a scarce resource. Should we do so or rather use our resource to prevent a small risk of death to each of many others when it is certain that eventually more than a few of these will certainly die? (Notice that to make sure it is only the known probability of any given person dying whose relevance to an allocation decision we are judging, we should hold constant the time at which those already ill and those who will become ill would die. Otherwise, we may be judging the relevance to an allocation decision of sooner rather than later deaths, not greater or smaller probabilities each person has of dying.)

On at least one view, fairness requires helping those with a higher individual risk of dying as ascertained by a pairwise comparison of the risks each person faces at the time we must allocate resources. This is because if we engage in pairwise comparison, we could justify to each person with the low risk of death our decision to not help him and instead help the person who it is known will certainly die. By contrast, it seems we could not justify to the person who it is known will certainly die our decision to leave him and instead help each of those who have a small chance of death. On this view, it is each individual’s comparative risk at the time we must allocate, not whether in the ultimate outcome more rather than fewer people die, that should lead us to allocate the money. Hence we could have reason to favor the less-cost-effective treatment policy that saves fewer people. This is so even if fairness requires saving the greater number of people when these are all people who are known at the time we allocate to each face certain death at the same time if not helped.11

Except for the fact that time of death is not held constant, this is like the situation we may face when deciding whether to allocate scarce funds to combat AIDS by either treating those already ill or by preventing future cases. Suppose fewer people overall will die from AIDS if the money is put into prevention than if it is put into treatment, and so prevention is most cost-effective. Without prevention more people who once had only a small risk of getting AIDS will eventually face certain death. But at the time we must decide how to spend funds there is a smaller group of other people who already have a known prognosis of certain death if they are not treated at that time. It is not the case that there are already some people in the larger group who have a known prognosis of certain death if prevention measures are not taken now. Rather there are many people each of whom has a small chance of being a person who will face certain death. Hence, the treatment policy might be recommended on the grounds that we should help those who face known certain death rather than help any of those who as far as we know have only a small risk of death.12

7. Disability and discrimination. Another possible fairness concern is that CEA might involve discrimination against those who are poor or disabled. This is because it may cost more to treat these people by contrast to the rich or nondisabled, and the health benefits achieved may also be less. Peter Singer relies on CEA when he argues that if we accept that disability can make a person’s life less good health-wise, and we want to maximize the health benefits we get with our resources, we should save the life of a nondisabled person rather than of someone whose disability cannot be cured if he lives on (other things equal).13 In doing this, we maximize quality-adjusted life years (QALYs). The only alternative to this, Singer says, is to deny that disability per se makes someone’s life not as good health-wise, and then there would be no reason to allocate resources to cure or prevent disabilities, which seems wrong.

I agree that understanding the issue of disability and allocation of scarce resource should not depend on accepting the view that disabilities make little difference to the quality of life. For if we hold this view, we may see little reason to invest in curing disabilities. We should also recognize that a satisfied “mood” that may be equally present in the disabled and nondisabled is not the sole measure of the goodness of one’s life; one’s objective capacities matter as well. Consistent with all this, one proposed response to a view like Singer’s is offered by Eric Nord, Norman Daniels, and Mark Kamlet in their 2009 article.14 They think it is important to distinguish two different questions. The first is: “Is a health state one we would prefer to cure?” A second question arises if we have this health state and it cannot be cured but life is still worth living, and we also have a life-threatening treatable condition but the medicine is scarce. The second question as stated by the authors is: “Should we defer to those who can be restored to more complete health than we can because they lack the untreatable condition?” The authors say we can reasonably answer “yes” to the first question—we would prefer a cure to the health state—and “no” to the second question. They do not say what explains the reasonableness of these responses.

Given the way their second question is phrased, it might be thought that one simple explanation is that the person with the untreatable condition does not have a duty to defer because he does not have a moral duty to sacrifice what is very important to him (his life) to produce the outcome that would be considered best from an impartial point of view. The fact that this view, which is standardly held by those who reject consequentialism, might explain the consistency of the first and second answers suggests to me that the second question as phrased by Nord, Daniels, and Kamlet is the wrong one to pose if we want to get to the heart of the issue in allocating scarce resources.

This is because it should be an impartial distributor who is allocating the resource, not a candidate for the resource, and the mere fact that a candidate need not defer to another candidate does not mean that the impartial distributor must give these people the same chance of treatment. Analogously, someone need not give up his medicine that will save his leg so that someone else may use it to save his own two legs even though the second person will be worse off without the medicine than its owner would be if he gives up his medicine. But if the drug is publicly owned and to be distributed by an impartial agent, that agent should prefer to help the person who would otherwise lose two legs. So a crucial issue in dealing with Singer’s CEA-inspired view is whether, if an impartial distributor says yes to the first question (“Is a health-state one we would prefer to cure”), this distributor should also decide to treat the person who can be restored to more complete health. In my own past work (first in Kamm 2004),15 I have been interested in the answer to these questions (which I shall refer to as the Impartial Questions).

Suppose I am the impartial distributor. When I imagine a case in which someone has a paralyzed finger, I can see that this can make life not as good in a small way, other things equal, and give us a reason to fix the disability. Hence, my answer to the first question is yes. But when I consider whether to save someone’s life from pneumonia when I can save only one person, the fact that one of the people I would save has a paralyzed finger and the other has all his fingers working should, I think, make no difference to whom I choose, given the important benefit that is at stake for each person and that each person desires to be the one to live. Hence, I should answer “no” to the second Impartial Question. Part of the explanation for this, I have suggested, is that a factor (such as a paralyzed finger) could give us a reason to act in one context (curing it) while it is an irrelevant consideration in another context where the action in question (saving a life) is different. This is an instance of what I call “contextual interaction.”16

It may be clear that small differences, like a paralyzed finger, should not affect who is chosen for a life-saving resource. But what is the explanation of this irrelevance? A possible explanation is that in this two-person contest for a scarce life-saving resource, either person would get the greater part of the best possible outcome that can be had by someone (i.e., a worthwhile life whether with or without a paralyzed finger). It is also the case that the alternative for each to being saved would be very bad (death) and each wants to be the one to survive. It is crucial to this explanation that we are dealing with separate persons and that as impartial allocators acting from a moral point of view, their different perspectives on an outcome (viz., each cares who survives) should influence what we should do. Otherwise, it would be clear that we should maximize QALYs as we would do if we had a choice with respect to one person of merely saving his life or saving his life and also unparalyzing his finger, holding costs constant.

But what of larger disabilities that bring down quality of life as far as 0.5 or somewhat below, so that it is not true that whomever is helped that person would get the greater part of the best possible QALY outcome that can be gotten by any candidate? I have suggested at least two grounds for why we should still give equal chances for a life-saving procedure to the disabled and nondisabled. Importantly, neither ground depends on the view that a disabled life is as good for someone as a nondisabled one, other things equal. First, each person can get what it is morally most important that people have, namely a worthwhile life, and each wants to be the one to survive. (Call this the Moral Importance Ground.) Second, when one’s only option is to have a life at 0.5 quality rating, it may be reasonable to care about keeping it as much as it would be reasonable to care about keeping a life quality-rated at 1. (Call this the Only Option Ground.) Note that this is consistent with its being reasonable to care to have the life rated at 1 rather than at 0.5 and even with its being reasonable to risk death to get it, were this possible.

But now imagine two nondisabled patients. One could live for twenty years if he had a scarce life-saving surgery and the other could live for five years. The Moral Importance and Only Option Grounds also seem to imply the seemingly mistaken view that it would be wrong to favor the person who would live much longer. If we disagree, we will need an argument that allows significant differences in length of life, but not significant differences in quality of life, to count in rationing decisions. One argument I have suggested is that we should distinguish between the “type” of person someone is, constituted by the qualitative features of his life, and how long any type of life goes on. Respect for persons might often require ignoring types when rationing but not require ignoring big differences in how long any given type will persist. (Call this the Respect Ground.)

Now consider another case about which we ask the first and second Impartial Questions. Suppose we agree paraplegia is a condition that we should prefer to cure. Now imagine two people with paraplegia who each need to be saved from fatal heart disease. The only difference between them is that in one of the people the scarce heart disease medicine will also cure his paraplegia. This is a case in which one candidate has an untreatable condition (paraplegia) and a treatable one (heart disease) and another candidate lacks untreatable conditions (because both his heart disease and his paraplegia can be cured). I suggest that it might be right for the impartial distributor of a scarce resource to choose to save the candidate whose paraplegia will also be cured rather than the other candidate. However, I think that this is neither simply because a life with paraplegia is worse than one with full mobility (other things equal) nor because CEA would rate a treatment as more effective if a person is saved to a life of higher rather than lower quality. Rather, it has something to do with both how bad paraplegia is and, crucially, with our medical procedure causing the person to no longer have paraplegia. That is, suppose it is not unfair to treat the second person if and because we can also cure his paraplegia. This does not imply that it would be fair to treat the heart condition of someone who is and will remain unparalyzed quite independently of our treatment rather than treat the heart condition of a permanently paralyzed person.

These two different heart cure cases suggest that a possible problem with CEA is that it does not distinguish (i) the case in which our treatment is more cost-effective in one candidate because it saves a life and also causes the change in disability status from (ii) the case in which our treatment is more cost-effective in one candidate because it saves someone who is already nondisabled independent of our doing anything to cure him of disability. In the latter case, we maximize QALYs by “piggybacking on” (i.e., taking advantage of his independently held) nondisabled condition. However, the two heart cure cases need not imply that causation always matters. For example, suppose each of two people has a paralyzed finger and a scarce life-saving drug that each needs will unparalyze the finger in only one of the people. The two Heart Cases need not imply that we should give that one person the medicine. A condition that we would prefer to cure may not be serious enough in itself that our being able to cure it in one person but not another should make a difference to which person we give a drug that each needs to survive.17

8. Conclusion. I have considered several views about what fairness requires and allows in conditions of certainty and uncertainty and how CEA understood in its strongest form may conflict with fairness. It has not been my aim to decide which conception of fairness is correct or to decide how important fairness is relative to other moral considerations. Nor has it been my aim to deny that CEA should sometimes play a role in allocating scarce resources. However, if the value of maximizing good outcomes relative to cost is neither a preeminent value nor necessarily consistent with fairness, there are bound to be moral questions about limits on the use of CEA that will need to be resolved.18



This appendix pertains to issues discussed in Chapters 4 and 8. It is a revised version of the article of the same name printed in the Journal of Practical Ethics 3 (2015) which was a response to Daniel Hausman’s “How Can We Ration Health Care Fairly and Humanely?” (originally presented by him at “Bioethical Reflections: A Conference in Honor of Dan Brock,” at Harvard Medical School, November 22, 2014). All references to Hausman are to that paper. Hausman focused on discussions in several of Brock’s articles (including his “Ethical Issues in the Use of Cost-Effectiveness Analysis for the Prioritization of Health Care”) about the problems of fair chances, priority to the worse off, aggregation, and discrimination. Hence, the order in which I discuss some issues in this appendix follows the order in which Hausman chose to discuss Brock’s work.


There is a noncomparative view about giving priority to the worse off according to which the moral value of giving a benefit to someone varies with how well off in absolute terms that person is—the worse off, the greater the value. This view does not require comparing how well off someone is relative to others. (It should not be interpreted to imply the mistaken view that a world in which moral value is maximized because everyone was very badly off but is completely helped is a better world than one in which no existing person was very badly off and each got only the little help he needed.) I am focusing on the comparative prioritarian view in taking it to be an interpretation of fairness which is a comparative value.


Some use the term “aggregation” to include the idea of addition. Here the term is used to connote merely putting items together in some way and adding them is just one way of doing this.


It is not unfair to additively aggregate small benefits to a few and weigh these against an aggregate of the same small benefits to many others, other things equal. But I am concerned with additively aggregating and weighing smaller benefits against bigger benefits.



Peter Singer, “Why We Must Ration Health Care,” New York Times, July 19, 2009



I say “seems” because it is possible that as the absolute number of years to be lived not disabled decreases (e.g., from 5 to 2 even if the ratio of unparalyzed to paralyzed years in the choice does not fall below 1:2), people would no longer be indifferent.


In Chapter 8 I make use of an intrapersonal trade-off test to judge which of two possible outcomes is worse and then apply the result to see which of two people would suffer a worse outcome. Nothing I say there depends on ignoring the fact that if one person suffers the less bad outcome it is not so that he (rather than someone else) can avoid the worse outcome.


For further discussion of this see my

Intricate Ethics (New York: Oxford University Press, 2007)

, chapter 2.


Larry Temkin has emphasized this point. See his Rethinking the Good: Moral Ideals and the Nature of Practical Reason (Oxford University Press, 2012).


One place I discuss this issue is in

“Should You Save This Child? Gibbard on Intuitions, Contractualism, and Strains of Commitment,” in Reconciling Our Aims, ed. A. Gibbard (New York: Oxford University Press, 2008)



A limited version of this view (and its application to cases like AIDS) is argued for by Norman Daniels (2012) and Johann Frick (2013 and unpublished) though they may not hold time of the relevant deaths constant. See

Norman Daniels, “Reasonable Disagreement about Identified vs. Statistical Victims,” Hastings Center Report 42 (2012), 35–45

; Johann Frick, “Uncertainty and Justifiability to Each Person: Response to Fleurbaey and Voorhoeve,” in ed. Nir Eyal et al., Inequalities in Health: Concepts, Measures, and Ethics (New York: Oxford University Press, 2013); Johann Frick, “Treatment versus Prevention in the Fight Against HIV/AIDS and the Problem of Identified versus Statistical Lives” in Inequalities in Health, ed. Glenn Cohen et al. (New York: Oxford University Press, 2015).


Frick proposes such an argument but also considers objections to it in his “Treatment and Prevention in the Fight Against HIV/AIDS.”


See Peter Singer et al., “Double Jeopardy and the Use of QALYS in Health Care Allocation,” in Unsanctifying Human Life: Essays on Ethics, ed. Helga Kuhse (Oxford: Blackwell, 2002).



Eric Nord, Norman Daniels, and Mark Kamlet, “QALYs: Some Challenges,” Value in Health 12, Suppl 1 (2009), S10–S15

. Norman Daniels brought my attention to what was said in this article in his commentary on my

Bioethical Prescriptions: To Create, End, Choose, and Improve Lives (New York: Oxford University Press, 2013)

at a panel in February 2013.



Frances M. Kamm, “Deciding Whom to Help, Health Adjusted Life-Years, and Disabilities,” in Public Health, Ethics and Equality, eds. S. Anand, F. Peter, and A. Sen (New York: Oxford University Press, 2004)



I first discussed contextual interaction in my “Killing and Letting Die: Methodological and Substantive Issues,” Pacific Philosophical Quarterly 64, no. 4 (1983). Singer gives a counterargument that is meant to show that it is reasonable to connect the answer to the first Impartial Question to an answer to the second Impartial Question. The argument claims the morally right way for an impartial allocator to make his decision is determined by what any person would decide about his possible future treatment when he is ignorant of which particular person (disabled or not) he will be. Singer thinks such a person would want to maximize his chances of living in the better condition (e.g., with all working fingers). I do not think this argument is correct and argue against it in Bioethical Prescriptions. But it is useful to see an argument, aside from maximizing good outcomes, that has been thought to connect a “yes” answer to the first question to a “yes” answer to the second question.


I have considered the issues discussed in Section 7 in more detail in

“Aggregation, Allocating Scarce Resources, and the Disabled,” Social Philosophy and Policy 26 (2009), 148–97

, and in “Aggregation, Allocating Scarce Resources, and Discrimination Against the Disabled,” in Bioethical Prescriptions.


I am grateful to Julian Savulescu and a reader for the Journal of Practical Ethics for comments on an earlier version of this appendix.