Survey results: priority to those who are more severely ill

If the costs of curing two ailments completely are the same, it will be more cost-effective to cure the more severe ailment. Cost-effective allocation responds in this way to severity – that is, how bad people’s health would be if they were untreated. But it does so only because of a correlation between the severity of health problems and the size of the benefit from alleviating them. Once the health improvement (effectiveness) and the cost are specified, the allocation of healthcare implied by cost-effectiveness is fixed; the severity of someone’s health state if untreated has no further relevance. There is considerable evidence that many people object to this implication of cost-effective allocation.

Let us begin with a classic Norwegian study, which Peter Ubel replicated and modified in a later study in the USA. In the Norwegian study, 150 ‘Norwegian politicians involved in healthcare decision making at a county level’ (p87)7 were presented with a scenario, which, translated into English (p896),8 reads as follows:

Imagine an illness A that gives severe health problems and an illness B that gives moderate problems. Treatment will help patients with illness A a little, while it will help patients with illness B considerably. The cost of treatment is the same in both cases. There is insufficient treatment capacity for both illnesses, and an increase in funding is suggested. Three different views are then conceivable. (1) Most of the increase should be allocated to treatment for illness B, since the effects of these are greater, (2) Most of the increase should be allocated to treatments for illness A, since these patients are more severely ill, (3) The increase should be divided evenly between the two groups. Which of these views comes closest to your own?

Eleven per cent favoured option 1, 38% option 2 and 45% option 3. Only about one out of nine favoured the most cost-effective policy. In Ubel’s replication, which was carried out in Philadelphia on prospective jurors waiting to be called to jury duty, an even smaller percentage—only 9% –favoured policy 1. Sixty four per cent chose to divide the increase equally, and 26% preferred to devote the increase entirely to those with the more severe illness A. An equal division, like a preference for treating illness A, apparently takes the greater severity of A as a reason not to distribute the additional funding in the most cost-effective way.

As Ubel notes, the earlier Norwegian study and his replication are flawed. Respondents might have chosen the equal division option out of an unwillingness to make a choice between treating A and treating B. I would add that the responses might also be explained by a concern with fair chances, rather than severity. Respondents might prefer the equal division on the grounds that health policy should not ignore the claims of any portion of the population.

In addition to replicating the Norwegian survey, Ubel investigated how people respond if the questions are modified. In a follow-up, Ubel et al 9 clarified the wording of the surveys and eliminated the option of dividing the increase in healthcare resources between illnesses A and B. Of those completing these revised surveys, about 60% favoured the cost-effective alternative of bringing the greater health improvement to the moderately ill, and about 40% favoured prioritising treating those who are more severely ill, even though the health improvement is smaller. The results of the American replication of the Norwegian study appear to be fragile: the preferences of American respondents for prioritising severity are not as widespread as the original study suggests. On the other hand, it is still the case that nearly half of Ubel’s sample were willing to sacrifice improvements in health in order to prioritise severity.

As the results of these studies  illustrate, survey respondents in various countries favour prioritising severity – that is, providing healthcare first to those who are more severely ill, even if the health benefit to them is somewhat smaller than the benefit that could be provided to those who are less severely ill. In his 2009 literature review, Koonal Shah concludes, ‘Overall, the majority of studies suggest that people are, on the whole, willing to sacrifice at least some aggregate health in order to give priority to the severely ill’ (p82).10 Gu et al maintain that ‘the empirical evidence is consistent with 16 (out of 19) studies suggesting that members of the general public are in general willing to give priority to a patient with more severe disease’ . (p47)6 (See also reference 11.) Shah10 points out, however, that a sizeable minority in most studies disagree, and Gu et al note three studies that found ‘small or no effects of severity on priority setting with the small effect favouring the less severely ill’ (p47).6

Much of the evidence showing that people support prioritising severity comes from comparing the values individuals place on health improvements to the comparative values inferred from ‘person tradeoff’ questions. Individual values—the HRQoL of health states—are derived from questions such as, ‘Suppose you had 10 years to live in health state H. How many years in full health would be equally good?’ If an individual, George, answers, ‘6 years’, and the value of full health is one and death is zero, then 10V(H)=6 and V(H)=0.6. Restoring George to full health would improve his HRQoL by 0.4 quality units, and if he were to live for 20 years, it would provide a benefit of 8 quality-adjusted life years (QALYs). In a person tradeoff, in contrast, people are asked questions, such as, ‘If you faced a choice between improving the health state of 100 people from 0.6 to 0.8 or improving the health of n other people from 0.8 to 1, for what value of n would you be indifferent?’ Someone who seeks only to maximise QALYs would answer that n=100. In fact, many people regard the improvement in health from 0.6 to 0.8 as more valuable than the improvement in the health from 0.8 to 1 (p87–9).7 Consider for example these results that Ubel et al (9, 894) report:

For example, in one study, people estimated the HRQoL [health-related quality of life] of moderate knee pain as being 0.94, a value that would imply that curing this knee condition (thereby bringing an HRQoL improvement of 0.06) is about 1/16th as valuable as saving a person’s life. Yet these people felt that greater than 1000 people would need to be cured of moderate knee pain to equal the value of saving 1 person’s life.

Before one takes this finding as establishing that people place an enormous priority on severity, one should recognise that the difference between the value implied by the time tradeoff and the value implied by the person tradeoff might have other sources, such as a special concern for saving lives or a rejection of aggregation, rather than or in addition to a general concern to prioritise severity (p41).12 Nevertheless, the contrast between the ratios implied by the quality weights and the responses to person tradeoff questions is evidence that many people prioritise severity.

The contrast between the ratios implied by the quality weights and the responses to person tradeoff questions might seem paradoxical, but the preferences from which the values of health states to individuals are inferred do not concern distribution. They express no view concerning tradeoffs – that is, social choices – among different health improvements for different individuals. Nord puts the point as follows:

Some may argue that the extent to which people have a concern for severity will be captured in their initial personal utility assessments of health states… In utility assessments,… subjects are not asked about distributive concerns or societal value. They are asked to quantify the disutility they personally would feel with different states of illness… On the basis of responses to such questions it is possible only to say which of different programs provides the greater sum of individual utility gains. The respondents have not expressed any opinion about priority setting (p28).13

Nord’s description is misleading because expressing quantitative preferences among health states is not necessarily the same thing as quantifying ‘the disutility they personally would feel with different states of illness’. Whether preferences among health states match the feelings individuals expect to have in those health states requires argument. The crucial point is that there is a difference between measuring how health states bear on individuals and setting priorities for the care of different individuals. Brock14 expresses the contrast with the following question, ‘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?’

What should one conclude concerning popular attitudes toward prioritising severity? Despite the explicit information the surveys provide, respondents in Ubel’s studies might mistakenly believe that because those with illness A are sicker, treating them will provide a larger health improvement. Another possibility is that individuals might believe that failing to prioritise severity constitutes a form of discrimination. However, these grounds for doubt are weak. It appears that a great many people prioritise severity and take the strength of an individual’s claims to healthcare resources to depend in part on how badly off the individual would be if not treated. When Ubel asked respondents to give their reasons, by far the most common response among those who favoured providing the smaller health benefit to those who are more severely ill, was that those who are more severely ill ‘deserve priority’ (8, 898).

Implementing priority for severity

Before considering whether there is an ethical justification for giving the claims of those who are severely ill greater weight than cost-effective allocation gives them, let us consider how to implement a policy of prioritising severity. Questions about the significance of severity are often described as concerned with ‘the tradeoff between health maximisation and pre-treatment severity’ (p82).10 This formulation is problematic. First, I am taking severity to be HRQoL if untreated, rather than ‘pretreatment severity’. Second, health measurement systems assign values or ‘quality weights’ to health states rather than measuring quantities or magnitudes of health.iii These values, which are described as measures of ‘HRQoL’, are in reality measures of preferences among health states. ‘Health maximisation’ consists of maximising the extent to which people’s health states satisfy their (health-related) preferences.

Although there are ambiguities in preference surveys and good reasons to doubt the accuracy of people’s predictions about what they would prefer or would choose if, for example, they actually faced a time tradeoff between longer life and a higher quality of life, it is reasonable to regard the numbers that one infers from those surveys as rough measures of what one might call the ‘personal’ or ‘individual’ value of health – that is, the value of health to a particular individual. Those who prioritise severity and also seek to enhance population health must find some way to tradeoff enhancing the total of individual value against the considerations that support prioritising severity.

If the health authorities seek both to enhance health and to prioritise severity, then there are several ways to proceed. The best known relatively specific scheme for prioritising severity while enhancing health is Erik Nord’s proposal 13 15–17 to encompass the priority of severity in an assignment of a ‘social value’ V(H_i ) to a health state H _i. If the social values properly capture the priority that more severe health states should have, then maximising the sum of the social value of each person’s health will implement the proper tradeoff between cost-effectiveness and prioritising severity. This approach can be summarised as follows.

The value encompassing maximisation strategy:

Maximise = ∑_jV(H_j), where V(H_j)=f[U(H_j),S(H_j)]. 

H_j is the individual’s j’s health state, U(H_j) is the individual or personal value of H_j, S(H_j) is the severity of H_j, V(H_j) is the social value of H_j, and f is some function that enables health economists to infer the social values of health states from measures of the individual values and severities of health states. In lieu of specifying f and deriving V(H_j) from U(H_j), it may be possible to measure V(H_j) directly by means of person tradeoffs. This scheme retains maximisation and consequentialism but argues that QALYs are the wrong maximand.

The value-encompassing maximisation strategy is attractive because it is algorithmic, at least in principle. Once the social value of health states is known, then, ceteris paribus, policy is determined.iv The tradeoff between prioritising severity and maximising the individual value of health is implicit in the function f that transforms individual values into social values (or, less explicitly, in person tradeoffs). By insisting that f be concave, Nord ensures that the same increase in individual value will be a larger increase in social value when an individual is in worse health. To determine how concave members of the population think that f should be, Nord relies on surveys of population preferences among different distributions of health improvements. As the results discussed above illustrate, there is no popular consensus about whether f should be concave, let alone how concave it should be. The best health economists can do is to minimise the extent of disagreement about the weight placed on severity.

To operationalise the value encompassing maximisation strategy, Nord has laid out a tentative set of severity weights. He suggests social values like those shown in table 1, on the assumption that the difference in individual value from every row to the row below is roughly the same:

It is questionable whether the best way to respond to the diversity of views concerning the weight that should be placed on severity is to build some compromise into the function f or to average the values elicited from person tradeoffs. Should deontic questions concerning how those who are in different health states ought to be treated be transformed into telic questions concerning the social value of different health states?v If the right way to encompass ethical considerations such as severity is to treat them as constraints on policies rather than as factors influencing the values of policies, then concerns about severity cannot be comprehended by maximising the social value of health.

Suppose we set aside these doubts and assume that severity of health deficiencies influences the value of remedying them. Does Nord’s operationalisation capture what it is intended to? Severity is, in Nord’s words, a matter of ‘current impairments and expected future loss of quality of life and/or length of life due to the illness’ (italics added) (p67).3 Although this does not rule out consideration of impairments or future QALY losses that are independent of ‘the illness’, it suggests that they have a different relevance for healthcare allocation. Yet the social values in table 1 measure impairments without regard to their source.

For example, suppose that the health authorities are considering policies to allocate physical therapy for back pain, when resources are too limited to treat everyone in need. As table 2 illustrates, there are three groups in need.

Nord’s scheme makes no distinction between whose condition is ‘relevantly’ more severe (as measured by their HRQoL if untreated) and those whose health problems by this measure are more severe owing to an independent disability. Refusing to distinguish relevant from irrelevant severity makes ethical sense, but diminishes the intuitive appeal of prioritising severity. When individuals think about prioritising severity, they envision giving priority to group 1 over both groups 2 and 3, rather than giving the same priority to groups 1 and 2 or giving priority to group 2 over group 3.vi To maintain that priority is called for only in the case of ‘relevant’ severity – that is, greater severity that is ‘due to the illness’ – requires some ethical defence. Why assign different priority to two groups of individuals whose health state is equally bad and who can benefit equally? The intuition that calls for prioritising only relevant severity is difficult to justify.

Philosophical rationales for prioritising severity

As previously noted, in Ubel’s study, the most common answer to the question, ‘Why give additional weight to the claims of those who are more severely ill?’ is that they ‘deserve priority’. This reason merely shifts the question: Why do those who are more severely ill ‘deserve priority’? Surveys and the reflections of philosophers and health economists have provided the following possible answers:

1. Adaptation: ‘People may think that patients and disabled people over time more readily adapt to moderate conditions than to severe ones’ (p283).17

2. Urgency: The needs of those who are more severely ill may be more urgent; a decision not to assist them is more likely to be irreparable  (p898).8

3. Distributional equality: Justice requires lessening inequalities; treating those who are more severely ill lessens health inequalities (Dolan12 17 ; Nord and Johansen, p28217; Kamm2, p302).

4. Priority: Justice requires giving priority to those who are worse off (Kamm2; Nord, p30213).

5. Compassion and solidarity: ‘One [reason] is a simple emotional one, namely that their feeling of obligation to help others increases with the intensity of other people’s suffering and/or the badness of their prognosis’ (p282).17

The first two reasons maintain that, owing to adaptation or urgency, the benefit of treating the more severely ill is larger than the benefit of treating the less severely ill. These arguments in favour of prioritising the claims of the more severely ill do not challenge cost-effective allocation. Urgency and adaptation are instead reasons why the cost-effectiveness of treating those who are more severely ill has been underestimated. If, as these first two considerations allege, there are greater total health benefits from treating those who are more severely ill than health economists have realised, then defenders of prioritising severity are objecting to a miscalculation of ‘effectiveness’ rather than to a flaw in allocating healthcare by its cost-effectiveness.

Although misdirected, the first two arguments for prioritising severity escape a serious problem facing the remaining arguments. Unlike the three remaining arguments, considerations of adaptation and urgency provide guidance concerning how much extra weight to place on addressing the health needs of those whose health is worse. In contrast, distributive egalitarianism, prioritarianism and compassion or solidarity say nothing about how much priority should be given to those who are worse off. Empirical investigations suggest that the views of members of various populations disagree, and the prioritarian, egalitarian and compassion/solidarity arguments provide no guidance concerning how to resolve those disagreements. Nor do they ground any reason to rely on the average of the differing weights people place on severity.

Prioritarians and distributional egalitariansvii may hold that prioritising severity contributes to the value of allocations, or that considerations of justice require favouring the treatment of those who are more severely ill. Either way, prioritarian views maintain that the benefits and burdens of the worse off should count more heavily,viii while egalitarians favour eliminating inequalities in some ‘currency’ (such as well-being) for which individuals are not responsible. On the assumption that those who are more severely ill are not responsible for their poor health and that they are, all things considered, worse off, both the distributional egalitarian and the prioritarian apparently support placing greater weight on the needs of those who are more severely ill.

These prioritarian and distributional egalitarian arguments for prioritising severity are problematic. Because prioritarianism says nothing about how much priority should be placed on benefits and burdens received by the worse off, it does not support any particular weighting of their claims. Egalitarianism might appear to escape this problem, because the degree of priority that would equalise well-being (to the extent to which equalising is feasible) is in principle knowable. But equality is not plausible as the sole principle governing distribution. Egalitarians maintain instead that social policies should balance the demands of equality against other moral desiderata such as making people better off. Without any definite guidelines concerning how to trade off equality and well-being, the egalitarian, like the prioritarian, leaves unspecified how much priority should be placed on the treatment of those who are severely ill.

In addition, prioritarianism and distributional egalitarianism prioritise benefits to those who are worse off all things considered regardless of whether they are worse off owing to independent health conditions or owing to features of their lives unrelated to health. Neither provides any general support for prioritising treatment of those who are more severely ill at the moment or whose future prospects are worse. As several authors point out2 18–20 and especially Nord3, a ‘special concern and priority for the worst off’ does not justify singling out for special concern those whose untreated health state is at the moment especially bad.

Consider table 3, which depicts the following subpopulations in some hypothetical society. Suppose that remaining life expectancy if not treated is in each group 10 years.

• The indigent. They have had persistent mediocre health. They suffer however more from any other deprivations such as poor living conditions, social exclusion and stigma. Their lifetime well-being is lower than any other group.

• The injured poor. Now in their retirement years, they were fine until injury or disease in their 40s left them with serious impairments and a low HRQoL. Although their lifetime well-being is better than the indigent, their lifetime HRQoL is lower than that of any other group.

• The rich disabled. They have been disabled for most of their lives. However, owing to their other resources, they have a reasonably high level of life-time well-being. Their average HRQoL up to this time is the lowest of any group.

• The newly afflicted. These are middle-aged and older individuals who have had a high level of well-being, but who now, owing to injury or disease, are seriously ill. Without treatment their circumstances are dire. Their health state, if untreated, now and in the future is the worst of any group.ix

The first two columns concern past well-being and past health, while the last three columns state the average future and lifetime expectations if there are no interventions to improve their health or well-being.

Which group is worst off? Depending on how one understands ‘worst off’, it could be any one of the four. Each is worst off in at least one column. From the perspective of prioritarianism or distributive egalitarianism, the indigent are worst off in virtue of how low their well-being has been up to now and in virtue of their low lifetime expected well-being. On the other hand, the current health problems of the newly afflicted are the most severe. Their HRQoL is worst now and in the immediate future. If members of all four groups were in need of physical therapy for their backs (and could benefit equally from it at the same cost), the greatest social value (in Nord’s sense) would lie in treating the newly afflicted, followed by the injured poor and the rich disabled.

Once one recognises the different ways in which individuals may be advantaged or disadvantaged – only some of which are matters of health – it is hard to see how prioritarianism or distributional egalitarianism supports prioritising severity, as ‘severity’ is defined here or by Nord. Those who emphasise ‘separate spheres’x and urge health economists to focus exclusively on health would counsel ignoring the first and third columns. Doing so would undermine the indigent’s claims, but it would not come close to providing a prioritarian or distributional egalitarian justification for prioritising severity.

A defender of prioritising severity might respond to this critique by arguing that the factors that diminish well-being tend to go together, so that those whose immediate untreated health expectations are worse will usually be worse off overall. But the premise in this response is questionable. Many among the worst off in society are, at any given moment in time, far healthier than those approaching death who, regardless of their wealth and privilege, are often most severely ill.

In addition, prioritarianism and distributional egalitarianism are as much in need of justification as the severity objection to cost-effective allocation they purportedly ground. The only argument in their favour rests on the intuitions that support their implications. This argument may seem forceful, because the inequalities that characterise the contemporary world are so deeply morally objectionable. However, the fact that egalitarianism and prioritarianism match the moral disapproval of contemporary inequalities counts for little, because other moral theories – even non-egalitarian views such as utilitarianism – also condemn current inequalities. If there is a case for prioritising severity (in the sense of immediate health prospects if untreated), it must lie elsewhere.

Compassion, solidarity and severity

That brings us to the last of the five arguments in defence of prioritising severity: that compassion or solidarity requires prioritising severity. Compassion is grounded in a vivid appreciation of the hardships of particular individuals and may place too little weight on the consequences of policies for unidentified individuals. Moreover, as Nord and Johansen mention, compassion tends to focus on suffering and prognosis, not on the whole range of hardships imposed by health deficiencies.

Unlike compassion, solidarity derives from a connection with a group with which one’s identity is entwined. This bond gives rise to a commitment to sharing burdens and risks.21 Unlike compassion, solidarity is in principle mutual. Though it may be possible to possess some minimal solidarity with the whole of humanity, solidarity grows weaker as groups grow larger and less proximate. Nevertheless, solidarity is of great normative importance as a ‘glue’ that stabilises social life and makes it more rewarding.

In contrast to allocating health resources by their cost-effectiveness, which reflects an impartial concern for everyone’s health, compassion is grounded in a concern for particular individuals and solidarity is grounded in group identification. Accordingly, morality demands a compromise between the demands of compassion and solidarity and the impartial concerns of justice and fairness.

Compassion seems to have its place mainly at the bedside, not at the level of policy. Those charged with treating victims of a natural disaster may need to gird themselves against their compassion toward those begging for immediate help. Unlike the impartial benevolence that grounds both allocation by cost-effectiveness and triage in the case of disasters, personalised benevolence can play only a limited role in social policy. At the bedside and in the shifting interstices of impartial policies, there is an urgent need for the moral sentiments that refuse to abandon comrades or those who are in desperate need. These sentiments should be cherished and nurtured.

Solidarity, on the other hand, might reasonably have a significant role in the design of national health policies, which are inevitably more concerned with citizens and residents than with all of humanity. However, the bearing of solidarity (and compassion, for that matter) on the different construals of severity is obscure. Why should solidarity call for a greater concern with what I called in the Implementing priority section ‘relevant’ severity than with greater severity owing to independent health (or for than matter, non-health) problems? As someone who cares about the health deficiencies of particular individuals or of those who are part of my community, why should I prioritise severity rather than, as egalitarianism and prioritarianism imply, focusing my attention on those who are worst off more generally?

It thus seems that compassion or solidarity do not justify prioritising severity. Although they demand that health policy show a human face, they support no principle of allocation that challenges cost-effectiveness. For the health system to show a human face means that from time-to-time there should be exceptions to cost-effective allocations in cases of particularly unfortunate individuals or groups. There should also be serious laments when desperate pleas are denied.

Conclusions

Assigning personal values to health states leaves open questions about how health-related resources ought to be distributed among people. Some distinction between social value, which reflects ethical concerns with distribution, and individual value is important. Survey data show that many people favour placing greater weight on treating people whose health without treatment is worse. In a representative democracy, one would expect health policy to conform to the popular consensus. Whether that policy is morally defensible is, however, a separate question.

This paper has questioned whether there is solid moral argument in defence of prioritising severity. It argues first, that it is hard to define ‘severity’. Measuring severity as HRQoL if untreated leads to absurd results unless one limits comparisons to health states that have the same duration. Measuring severity by absolute or proportional shortfalls in QALYs or life-expectancy implies age-weighting, which is foreign to the concerns about severity. Second, neither egalitarianism nor prioritarianism support prioritising severity, as it is conceived in this paper, let alone provide any guidance concerning how much priority severity ought to have. Third, popular attitudes toward severity draw a morally arbitrary distinction between, on the one hand, relevant differences in impairments, which result from salient health threats, and, on the other hand, irrelevant differences in impairments, which result from differences in pre-existing health states. Finally, although compassion and solidarity lead to dissatisfaction with cost-effective healthcare allocation, they have no clear bearing on a policy of prioritising severity.

Some might argue that health policy must be guided by popular attitudes, even if they have no moral defence and, like prioritising severity, imply fewer aggregate health improvements. As a political matter, conforming to popular attitudes may be unavoidable. But there is a role for opposition and criticism, particularly when popular attitudes are conflicted as well as unjustified.

As mentioned at the beginning, there are other grounds for qualms about relying on cost-effectiveness to govern the allocation of health-related resources. But the severity objection is in need of moral justification