Two principles form the basis for much priority setting in health. According to the greater benefit principle, resources should be directed toward the intervention with the greater health benefit. According to the worse off principle, resources should be directed toward the intervention benefiting those initially worse off. Jointly, these principles accord with so-called prioritarianism. Crucial for its operationalisation is the specification of the worse off. In this paper, we examine how the worse off can be defined as those with the fewer lifetime Quality-Adjusted Life Years (QALYs). We contrast this proposal with several alternative specifications.
- Resource Allocation
- Distributive Justice
- Health Economics
Statistics from Altmetric.com
Available resources typically fall short of healthcare needs, making priority setting inevitable. Many priority setting principles, prescribing how resources should be allocated, have been proposed. Of these, two general principles have been particularly prominent. According to the greater benefit principle, resources should be directed toward the intervention with the greater health benefit. In this paper, we take health benefits to be appropriately described in terms of Quality-Adjusted Life Years (QALYs). According to the worse off principle, resources should be directed toward the intervention benefiting those initially worse off. Together, the two principles accord with so-called prioritarianism.
Prioritarianism is typically defined in terms of a certain principle and a certain reason for that principle. According to Derek Parfit, the principle asserts that ‘benefiting people matters more the worse off these people are’.1 In his view, prioritarianism can be characterised as a single, complete principle, but he also describes how prioritarianism ‘contains the idea that benefits are good’ and ‘merely adds that benefits matter more the worse off the people are who receive them’.1 This latter description indicates how prioritarianism also can be seen as the combination of a greater benefit principle and a worse off principle.i Some, including Parfit, also deem necessary for prioritarianism that the reason why benefiting the worse off matters more is that the worse off are at a lower absolute level. However, many other, yet partly overlapping, reasons exist. These include reasons related to equality, compassion, humanitarianism and a concern for greater relative improvements.4 Some broader conceptions of prioritarianism allow for a variety of reasons to motivate its central principle, and such a conception is particularly useful when prioritarianism is contrasted with the sole operation of the greater benefit principle. While the latter can motivate standard Cost-Effectiveness Analysis (CEA) with QALYs, prioritarianism can motivate CEA with the so-called equity-weighted QALYs.5–7 As for divisions within prioritarianism, there is an ongoing debate over the relative virtues of ex ante and ex post prioritarianism.8 ,9 To sidestep this complex issue, we will assume certainty about outcomes so as to eliminate the distinction between the worse off ex ante and ex post.
In the field of health, a concern for the worse off is often made explicit by reference to ‘need’, ‘severity’ or ‘urgency’, concepts whose attributed meanings vary greatly. In this field, the worse off are also typically defined in terms of health, as opposed to advantage more generally.ii So defined, support for priority to the worse off—together with a concern for the greater benefit—pervades both public opinion and official health policy.7 ,11–15 Prioritarianism may thus be seen as a nexus for priority setting in health. Its broad support, however, conceals considerable disagreement over its specification. A crucial question for the operationalisation of prioritarianism is this: how should the worse off be defined? In this paper, we describe one attractive specification and contrast it with the chief alternatives.
We here consider the worse off in the context of prioritarianism in macro-level priority setting. We propose the following specification: the worse off are those with the fewer lifetime QALYs given all interventions currently implemented. In other words, the worse off are those that will have the fewer QALYs over their entire lifespan or will have had the fewer QALYs when they die. It is worth mentioning that, in practice, the lifetime QALY prospect of an individual generally varies with age and only occasionally corresponds to that at birth.
The basic idea underlying the proposal is not new. In the field of distributive justice, for example, one is almost invariably concerned with very general metrics of individual advantage, incorporating quantity and quality of life over the lifetime.iii In the fields of health economics, health policy and clinical medicine, however, the emphasis is very different, as will be illustrated below. One exception here—though not explicitly offered as a specification of the worse off—is Alan Williams's account.5 Generally wanting, in any case, is a detailed review of specifications of the worse off and a sustained analysis of the lifetime QALY view in priority setting.
The proposal can, to some extent, be assessed in terms of its direct appeal. For a more thorough and discriminating evaluation, key properties of the proposal should be contrasted with their rival specifications. Three such specifications are particularly relevant: primacy to loss, quantity only and exclusion of past health. The recurring question is this: if we want to incorporate a concern for the worse off with that for the greater benefit, which specification of the property is—ignoring intermediate positions—the more appropriate? As the general meaning of the ‘worse off’ allows for many different specifications, their implications for priority setting are one important consideration.
Primacy to loss
The proposed specification refers directly to what people have, and the simplicity of this idea has some immediate appeal. In contrast, many other specifications of the worse off refer directly to a concept of loss, where loss is the difference between what one has and some reference level. The worse off, then, are those with the greater loss. One clear example is the official guidance on economic analyses in the health sector, currently being developed in Norway. In the preliminary report, the more severe condition is defined as that representing the greater absolute loss in QALYs, the reference being remaining quality-adjusted life expectancy in the population, adjusted for age and sex.18 Another example is found in the Netherlands, where the criterion of ‘necessity’ has been operationalised in terms of ‘proportional’ shortfall.19 According to the principle of proportional shortfall, ‘priority should be given to those patients who lose the greatest proportion of their remaining health expectancy due to some illness if the illness remains untreated’ (original emphasis).19 The reference here is remaining QALY expectation in absence of disease.
Giving primacy to a concept of loss is logically independent from utilising reference levels in the very estimation of health expectancies. So why be primarily concerned with loss per se? That question has received little or no systematic attention. Immediately, the two most plausible sources of motivation appear related to relativisation of reference level and so-called sufficientarianism.
Many seem to think that the specification of the worse off itself should be made relative to certain individual characteristics and that direct reference to loss is an intuitive way of doing so because the reference level can be easily relativised. Several specifications of the worse off involve reference levels adjusted for age, sex or both.18 ,19 Little, if any, justification has been offered for the relativisation of the reference level itself. The burden of proof also seems to reside with the proponents of such relativisation, as the presumption of equality appears quite robust in this context. In other words, the idea that everyone should have the same reference level appears to be a plausible starting point. Moreover, relativisation of reference level can have various counterintuitive effects, for example, with respect to age. For an individual or group, the reference level—when relativised as in the proposals described above—increases with higher expected age of death, which increases with actual age. Thus, one effect of adjusting for age is that those that have had more in the past are ‘entitled’ to more health over the lifetime. One needs not be committed to a lifetime view of the worse off to find this problematic.
Certainly, giving primacy to loss is not wedded to relativisation of reference level; one can simply use the same reference level for all. If the worse off are defined as those whose health is falling the farther below this level, at least the ordering of people below that level will be the same as if the worse off were defined directly as those with the lower health. The extent to which one person is worse off than another will depend on the more precise function over lifetime QALYs or QALY loss. With the same reference level for all, the strongest reason for giving primacy to loss seems related to so-called sufficientarianism.20 ,21 According to a classic formulation, the basic idea is that ‘what is morally important with respect to money is for everyone to have enough’.20 This is typically taken to imply that resources should be allocated so that as many people as possible are above some specified threshold of sufficiency. Most versions of sufficientarianism also explicitly deny priority to the worse off among those above that threshold, but some allow for priority to the worse off among those below the threshold.21 Those attracted to such a composite view may favour primacy to loss in the specification of the worse off. This is because everyone above a certain reference level will effectively be considered equally well off and will to that extent have equal priority.
Sufficientarianism in the context of lifetime QALYs has, however, its problems. Beyond the difficulty of justifying one specific threshold, the very idea of a threshold appears at odds with some of our intuitions about priority setting. This seems to be the case even when the threshold is set high, for example, at 80 lifetime QALYs. Given such a threshold, one may consider two interventions, targeting groups A and B, respectively.iv Without intervention, people in both groups will die immediately. With intervention, people in both groups will gain 2 years in perfect health. The only difference between the interventions is that without intervention groups A and B will have 80 and 90 lifetime QALYs, respectively (Figure 1a). We think group A is relevantly worse off and that there should be some priority to intervention A, something that conflicts with sufficientarianism with a threshold of 80. It is crucial that we here are considering QALYs, not money as in the original formulation. To assign priority to benefiting those with 80 lifetime QALYs over those with 90 is perfectly compatible with not assigning priority to benefiting the rich over the super rich. Given the intuitions outlined, a dilemma arises for those who want to specify the worse off with primacy to loss. Either one can accommodate those intuitions by setting a very high threshold, possibly higher than the biological maximum, or one can introduce a tripartite definition of the worse off. According to such a definition, the worse off among two people below the threshold is the one farther below, the worse off among two people at opposite sides of the threshold is the one below it, and the worse off among two people above the threshold is the one closer to it. Either way, the relevance of a threshold pales, as does the relevance of loss in the specification of the worse off. Thus, if relativisation and sufficientarianism are the strongest reasons in favour of primacy to loss, the basis for introducing the complicating factor of a reference level is rather weak. We would therefore go with the more straightforward concept until the alternative is more clearly supported.
According to the proposal, both quantity and quality of life matter when identifying the worse off. This parallels with several other specifications.13 ,14 ,18 ,19 Some, however, specify the worse off solely in terms of quantity. The allocation of heart, lung and liver transplants in the USA, for example, depends on ‘urgency’, specified as probability of death.15 As for theoretical proposals, Persad and colleagues have advanced a ‘complete lives system’ that combines four priority setting principles, none of which are directly sensitive to quality: youngest first, prognosis in terms of life years, lottery and saving the most lives.22 Accounts with a restricted role for quality also include that of Frances Kamm, according to which the worse off are those who will have had less ‘adequate conscious time alive’ when they die,23 and that of Erik Nord, discussed below. Ignoring quality altogether can have profound implications with respect to the wide range of conditions affecting both quantity and quality, such as heart failure, and the many diseases primarily affecting quality, such as arthrosis and depression.
Several general concerns suggest that considering only quantity is insufficient. While some time alive truly is an all-purpose means, necessary whatever goal one has in life, so are also certain components of quality. Basic levels of mobility and cognitive skills are just two examples. Functions such as these can be considered ‘maximally flexible assets’,24 crucial for the ‘higher-order’ interest in being able to revise life plans or conceptions of the good,25 and essential in protecting the range of opportunities open to people.26 Moreover, pain, another determinant of quality and a key concern in healthcare, is typically considered relevant also beyond any functional limitation that may come with it. The inadequacy of considering quantity alone is further indicated by the so-called conditions worse than death, as well as by the fact that people, for themselves, trade off quantity for quality all the time. For example, we may knowingly increase the risk of death for the sake of improving quality of life when we take on dangerous travels or undergo surgery that is purely quality improving.
Also in the specific context of health, quality intuitively matters when deciding who are the worse off. Of two people living exactly the same numbers of years, one in perfect health and the other with paraplegia and great pain throughout, the latter is plausibly considered the worse off. This suggests that such a person also is relevantly worse off for priority setting and that trading off quantity for quality may be appropriate. As for whether quality can be relevant for priority setting, also beyond its role in the assessment of benefits, one may consider the following case. People in groups A and B will both gain five QALYs from a quality-improving intervention. The groups are equal in all respects but one. Without intervention, people in group A will in the future live for 10 years with severe disability and 30 years in perfect health, while people in group B will live for 40 years with an equally severe disability (Figure 1b). In our opinion, group B is relevantly worse off, and intervention B should have some priority.
Still, quality needs not be relevant in all contexts. Two contextual factors appear particularly important: timing of initial difference and type of intervention. Erik Nord, for example, is critical of the inclusion of past quality in the specification of the worse off, while he endorses the inclusion of future quality and accepts that past quantity can be relevant.6 Nord's criticism partly draws on a specific case that is refined here so that the groups differ only in terms of past quality. While the cross-cutting issue of time frame is addressed below, such cases can illuminate the general issue of quality to the extent that the past is deemed relevant. People in groups A and B will both gain one QALY from a quality-improving intervention. The groups are equal in all respects but one. People in group A have lived for 70 years in perfect health, while those in B have lived for 70 years with disability (Figure 1c). If past quality is ignored, the two groups will be considered equally badly off. Contrary to Nord, we think that group B is relevantly worse off and should have some priority in cases like this.
Nord, however, may not have concentrated on the structurally strongest case against the inclusion of past quality. Such a case is probably one in which the intervention is purely life extending and the difference in quality is one of the past only. Consider groups A and B. Without intervention, people in both groups will in the future live for 10 years in perfect health. With intervention, they will gain one additional year in perfect health, that is, one QALY. The groups are equal in all respects but one. People in group A have lived for 60 of 60 years in perfect health, while those in group B have lived for 50 of 60 years with some severe disability characterised by pain and restricted mobility (Figure 1d). This case may be more difficult than case (c) because the deficit is one of quality whereas the benefit is one of quantity. Both this asymmetry itself and the fact that many intuitively privilege quantity may explain the difference between the two cases. Nevertheless, we do find group B relevantly worse off and that intervention B should have some priority. This can be supported by considering temporal consistency. In the context of benefits from intervention, we care greatly about quality improvements, and we seem ready to trade off quantity for quality repeatedly, for example, in the construction of QALYs. In addition, many, like Nord, seem ready to trade off quantity for quality with respect to the future also in the specification of the worse off.6 There is some tension in embracing major trade-offs between quantity and quality in the future and ascribing relevance to past quantity, while denying any such trade-offs in the past. This tension is further increased if quality, and not only life years, can represent all-purpose means.
A very different reason for including quality emerges if one considers the implications of the greater benefit and worse off principles combined. Alone, QALY maximisation has been charged with disability discrimination.27 In the context of life-extending interventions, for example, the chronically disabled will tend to be assigned lower priority by the greater benefit principle than those who are not. This is because with an equal number of years saved, the number of QALYs gained will be lower for the chronically disabled. In contrast, the worse off principle—when sensitive to quality—will assign higher priority to the chronically disabled, other things equal. Thus, when such a principle operates together with a greater benefit principle, the former can mitigate the allegedly discriminatory effects of the latter.
The immediate plausibility of including quality in specific situations will certainly depend on the exact quality-adjustment weights used. However, when the question is whether quality should be generally included or generally excluded in the specification of the worse off, the former option appears to have the stronger case.
Exclusion of past health
The proposal considers health over the entire lifetime, as do some other specifications of the worse off.5 This focus is controversial, however, as some consider current and future health only.v For example, the Swedish Parliamentary Priorities Commission advanced a need and solidarity principle in which need related to the severity of current condition and prognosis.14 The Danish Council of Ethics has emphasised ‘gravity of the disease’, understood as ‘point of departure in terms of health and prognosis’.13 Examples from Norway, the Netherlands and the USA were provided above.15 ,18 ,19
Several general considerations motivate an emphasis on whole lives. Many think, for example, that the individual is the unit of ultimate moral significance for public policy. Moreover, while there are critics,28 the standard account of personal identity asserts that individuals typically extend through time, from birth to death, as a single person. Together, this can motivate a primary concern for how individuals’ lives go as a whole or at least tell against excluding any particular part offhand. A similar concern can also find motivation in an idea of equality in basic human worth. While by no means implied, this fundamental idea may suggest that it is primarily whole lives, rather than parts of lives, that are of equal worth. In the specific context of health, a concern for whole lives has been supported by the so-called fair innings argument. In Alan Williams's version, the basic premise is that everyone is entitled to some ‘normal’ span of health, and one implication being that ‘anyone failing to achieve this has in some sense been cheated, whilst anyone getting more than this is ‘living on borrowed time’’.5 To be sensitive to such concerns, lifetime health must be taken into account.
Even if the whole of people's lives is what is of primary relevance, one could still argue that what has happened in the past is irrelevant for the distribution of future benefits and thus for priority setting. Given the general considerations outlined, however, the burden of proof seems to lie with those who want to argue those points. This is further suggested by the fact that we make inter-temporal trade-offs for ourselves all the time.vi For example, we often sacrifice our current well-being for the sake of more well-being in the future. Also in cases involving others, we often consider relevant how people have fared in the past when assigning priority to future benefits. For example, Dan Brock has argued that ‘...if one child has had little opportunity for travel in the past in comparison with her well-traveled sibling, fairness supports giving a travel opportunity to her now that can only go to one of them, even if the well-traveled sibling might enjoy and benefit from the trip more’.4 In priority setting in health, many kinds of cases involve differences in the past. Cases emphasising differences in past quality were discussed above. Those cases can illuminate also the issue of time frame to the extent that quality is generally deemed relevant. To decide whether past health should be excluded, however, we should also consider cases involving differences in past quantity. One such case is the following. People in groups A and B will die immediately without intervention. With a life-extending intervention, people in both groups will gain 10 additional years in perfect health, that is, 10 QALYs. The groups are equal in all respects but one. People in group A have lived for 40 years, while those in B have lived for 70 years (Figure 1e). In such a case, we think group A is relevantly worse off and that intervention A should have some priority. The reason is that, without intervention, people in group A will have had much fewer years of life overall than people in group B.
The main challenge to the relevance of past quantity is probably cases of the following kind involving intense, current pain.6 ,29 People in groups A and B currently experience intense pain of the same magnitude. The groups are equal in all respects but one. People in group A have lived for 40 years, while those in group B have lived for 60 years (Figure 1f). Our proposal privileges neither current deficits nor pain beyond its influence on the quality-adjustment weights underlying QALYs. According to the proposal, people in group A are worse off in terms of initial lifetime QALYs and have, for that reason, some priority to analgesics. Some find this implication highly objectionable, thinking that the two groups should have equal priority. This we may call the equal priority judgement. When reflecting on this case, one must keep in mind that while intense pain gives a strong reason to assist both groups, that is not possible in the present case. Equal priority here means selecting one group at random.
Even if the equal priority judgement was to survive thorough reflection, there are still two main reasons why that judgement about a particular case is of limited relevance when choosing whether past health should generally be included in the specification of the worse off: limited applicability and limited discriminatory power. As for limited applicability, the equal priority judgement seems to lose force if we modify the original case in any one of the following directions. First, the case can be modified so that the past difference between the groups is one of quality rather than quantity or age. This will make the situation more similar to that in case (c). Moreover, the appeal of the equal priority judgement appears further reduced if that past difference in quality is one of pain. Second, the case can be modified so that the current situation is one characterised by mild or moderate, rather than intense, pain.29 Here, the appeal of the equal priority judgement may be further reduced if the quality deficit in question was not one of pain at all. Overall, to the extent that the equal priority judgement loses force as one makes any one of these modifications, the judgement appears to have quite limited applicability to the wide range of cases involving differences in past health. As for limited discriminatory power, cases involving intense, current pain may be less indicative of the relative relevance of lifetime versus future-only specifications than may be thought. Applied to such cases, a specification of the worse off including only the present and the future seems problematic in much the same way as the lifetime view appears problematic to some. Consider a modified case (f) in which the two groups are equal with respect to past health. The only difference now is that people in groups A and B will live for another 10 and 20 years, respectively. Even if the worse off is defined in terms of present and future health only, people in group A will be deemed worse off and thus have priority to pain relief.
From the preceding, two kinds of cases appear to challenge the relevance of past health: those involving differences in past quality in the context of life extension and those involving differences in past quantity in the context of relief of intense, current pain. Even if these cases were truly problematic, the lifetime view, for its part, can adduce support both from numerous other cases and from several general considerations. Thus, on the question of whether the past should be generally included or generally excluded, the former option seems better founded. Crucially, including past health in the specification of the worse off does not imply that we care less about the size of future benefits. The latter concern is catered to by the greater benefit principle.
In this paper, the worse off have been specified as those with the fewer lifetime QALYs given all interventions currently implemented. This proposal has been put forward in the specific context of priority setting when two general principles are to be combined: a greater benefit principle and a worse off principle. The proposed specification was contrasted with its chief alternatives. To this end, we introduced both general considerations and intuitions about specific cases. Even for those disagreeing, we hope to have brought out the central issues at stake. While we find the proposal superior to the alternatives, it does have potentially problematic implications in need of further inquiry. One challenging issue pertains to interventions targeting people with only a very few lifetime QALYs without intervention. Also to be decided is how the greater benefit principle and the worse off principle are to be weighed against each other. When appropriately balanced, it can be assessed how national and global priorities may change as one advances from pure QALY maximisation to lifetime QALY prioritarianism. The potential impact is great.
I am truly grateful to Kjell Arne Johansson, Frode Lindemark, Ole Frithjof Norheim and Bjarne Robberstad for motivating the writing of this paper and for their enthusiastic support throughout. I also extend my sincere thanks to Kristine Bærøe, Reidar Lie, Erik Nord, Toby Ord, Einar Torkilseng, Alex Voorhoeve, two anonymous reviewers of this journal and the participants at seminars in Bergen and New Delhi for their helpful comments. Olav Pekeberg kindly assisted in formatting figures.
Competing interests None.
Contributors Kjell Arne Johansson, Frode Lindemark, Ole Frithjof Norheim and Bjarne Robberstad.
Provenance and peer review Not commissioned; externally peer reviewed.
↵i Larry Temkin, for example, has characterised prioritarianism as a compromise between two principles.2 Others have characterised it as a compromise among concerns, values or ideas, rather than principles. Whether the central trade-off is located within or between principles is of limited relevance in the present context. Bertil Tungodden has made a similar remark.3
↵iv In this and subsequent cases, it will be assumed that the interventions, including the people targeted, are equal in every possible respect if not stated otherwise. This includes parity in costs and size of benefits. To explore the degree of willingness to trade off size of benefit for the competing concern in question, one should modify the respective case so that the two interventions differ also in terms of benefit. Throughout, it is also assumed that available resources admit only one of the two interventions to be implemented and that the difference between groups is due to conditions for which one cannot be held responsible.
↵v Among these, some allow for the possibility that age can be a separate concern. Age may also influence a future-oriented specification of the worse off indirectly.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.