The Process Is Not the Point: Moral Uncertainty and Substantive Justice in Health Care Priority Setting

ABSTRACT

Health actors must ration care due to resource constraints. But even when decision makers know the relevant empirical facts about an allocation decision—including how many patients will benefit from a service, which patients will benefit, how much each patient will benefit, and so on—these facts may not provide a clear answer about how to make allocation decisions. This is because competing moral principles recommend different courses of action: for instance, maximizing health benefits may involve deprioritizing more disadvantaged groups. Facing this moral uncertainty, many ethicists and policymakers have embraced using “fair processes” that emphasize procedural justice over substantively just outcomes. This paper challenges the proceduralist consensus on two grounds. First, I argue that designing fair processes generates as much moral uncertainty as making allocation decisions directly, while also failing to address the substantive allocation question that must ultimately be answered. Second, both theory and empirical evidence suggest that fair processes favor less cost‐effective interventions, risk compounding health inequities, and divert resources that might otherwise be used for health care. As an alternative approach, I argue that health actors should pursue substantive justice through Maximizing Expected Choiceworthiness (MEC)—a decision framework that aims to choose options based on their expected value across different possible moral principles, weighted by the confidence decision makers have in each principle. I demonstrate that MEC can yield practical, intuitive answers to some resource allocation questions, while providing a helpful framework for approaching others. I conclude that health actors can and should pursue substantive—rather than procedural—justice in setting health priorities.

1. Introduction

Organizations that provide health care (“health actors”) must limit access to this care. These allocation decisions have ethical implications, forcing decision makers to navigate moral uncertainty. Even when decision makers know the relevant empirical facts about an allocation decision—including how many patients will benefit from a service, which patients will benefit, how much each patient will benefit, and so on—these facts often fail to provide clear answers about how to set priorities. For instance, when limiting access to care, should organizations aim to provide services to more disadvantaged patients or to patients who stand to benefit the most? Should they give everyone a fair shot at accessing care, even if this means the benefits of care will be smaller? Should they prioritize providing services to those who bear no responsibility for their illnesses over those who plausibly do? It is often not clear what, morally, these actors ought to do.

Nonetheless, health actors must make allocation decisions. Given ambiguity about what would constitute substantively just decisions (i.e., just final allocations of health resources), some ethicists have endorsed instead making procedurally just ones (i.e., decisions made via the use of “fair processes”). ¹ This is the argument put forward by Sofia Gruskin and Norman Daniels in their straightforwardly titled article, “Process Is the Point” [1].

Health actors have found this line of reasoning persuasive. Rather than relying entirely on substantive principles—for example, an intervention's cost‐effectiveness or its distributive effects—to allocate health resources, many ministries of health, non‐governmental organizations, and private and public hospitals now aim to use fair processes to make priority‐setting decisions [2]. To date, such decisions have potentially affected more than a billion people worldwide [3, 4].

In this paper, I argue that the process is not the point; rather, health actors should aspire to make substantively just decisions. The processes they use to set priorities should be a means by which they achieve this end; they should pursue procedural justice only if doing so more reliably produces allocations in compliance with substantive justice.

2. Background

The United States is one of the most affluent countries in the world, and although it devotes 18 percent of its Gross Domestic Product (GDP)—more than $5 trillion per year—to health care, it cannot meet many of its citizens' health needs [5, 6]. Americans regularly forgo medications because they cannot afford them, wait months to see medical specialists, and find that routine medical care is not covered by their insurance [7]. These issues arise, in part, because the U.S. health care system is inefficient. But countries like Australia, Norway, and the Netherlands, which have well‐functioning health care systems, face similar challenges [5]. These problems are even more pronounced in countries with fewer resources, and have grown more acute as populations have aged and chronic illnesses have become more prevalent.

Many high‐income countries have responded to rising health care needs and costs by simply spending more. In OECD countries, real per capita health expenditures have risen consistently, and health care costs as a share of GDP have doubled in many places [8]. The relentless growth in health spending has forced governments to reduce spending in other important areas, including education and transportation; in New Jersey, one government official lamented: “We're talking about 300 teachers being laid off to be able to afford our health care bill” [9, 10]. Thus, while increasing spending may serve as a temporizing measure in some places, it has adverse consequences and does not constitute a sustainable solution. Moreover, many low‐ and middle‐income countries (LMICs) cannot afford to simply spend more on health care. As a result, around the world, countries limit access to care. ² For instance, patients may be asked to buy their own medications, pay out‐of‐pocket for care, wait months for procedures, or forgo certain services altogether [11, 12].

Health actors are thus increasingly faced with decisions about how to allocate the services they provide. Moreover, access to necessary treatments is not just constrained by money. Many patients who need organ transplants will die on transplant waiting lists [13]. And during the COVID pandemic, hospitals were forced to allocate ventilators, extracorporeal membrane oxygenation (ECMO) machines, antibody treatments, coronavirus vaccines, hospital beds, operating rooms, specialist appointments, and other health resources that were limited by supply [14].

Questions about the appropriate allocation of health resources are moral questions because their answers influence which people live, suffer, or die. These are hard questions because there are often compelling—but conflicting—moral reasons to prioritize different patient populations. In this paper, I consider the following kinds of cases:

A health actor must provide Service A or Service B. The organization is unable to provide both services due to either financial constraints or limited availability of the resource itself, nor can it partly provide both services. A Service may refer either to a given type of medical care (e.g., one or another hepatitis treatment) or medical care for a given subset of patients (e.g., hepatitis treatment for people who do not use injection drugs but not for those who do). I thus consider situations in which an organization must provide care to one set of patients or another, either because the two groups of patients have different illnesses and would benefit from different treatments, or because the two groups have the same illness, but treatment can only be provided to some. ³

Suppose, too, that there are compelling moral reasons to provide either service; perhaps the patients who need Service A are more socially disadvantaged, younger, and more likely to benefit, but those who need Service B are sicker and unambiguously bear no responsibility for their condition.

Finally, I assume that the relevant decision maker exists in a context that is broadly democratic; in other words, health actors have the authority to make priority‐setting decisions because this power has been conferred on them by laws and policies created by a democratically elected government. ⁴

2.1. Moral Disagreement and Moral Uncertainty in Health Care Priority Setting

In the health care priority‐setting literature, the kinds of cases just described have typically been discussed in terms of moral disagreement: reasonable people often disagree about what actions we should take when moral principles recommend different courses of action; for instance, one person might prioritize Service A in the above case, while another might prioritize Service B. This tension was exemplified by discussions about whether a person's coronavirus vaccination status should influence their receipt of limited coronavirus treatments. One ethicist tweeted: “Ethicist here, reporting that denying care to the unvaccinated is not ‘complicated’ or ‘nuanced’—it's just obscene, morally wrong, we shouldn't do it, full stop” [15]. To which another ethicist replied: “Whatever answer one reaches, I disagree that ‘it's not hard’, or that [arguments] like this are ‘obscene’.” The latter ethicist proceeded to write an op‐ed in the Washington Post arguing that “At a moment when COVID‐19 hospitalizations are high and treatments for the disease scarce… people who have refused coronavirus vaccines should [not be allowed] to compete for treatments with people who are immunocompromised” [16]. Clearly, people disagree about the ethical allocation of health resources when multiple relevant principles recommend different courses of action.

However, the types of moral conflicts health actors encounter when allocating health resources can also be described in terms of moral uncertainty. A decision maker is morally uncertain when she is not sure what she all‐things‐considered morally ought to do [17]. The types of priority‐setting challenges I have described can incite disagreement within or between health actors and the public, but they can also lead to uncertainty for those tasked with making them. Moreover, even when individual decision makers are not privately uncertain, disagreements like those described above should cause health actors to act on an attitude of moral uncertainty when making priority‐setting decisions. ⁵ In short, when making the kinds of priority‐setting decisions that tend to incite moral disagreement, health actors should ensure that the moral arguments for and against a given course of action receive appropriate consideration.

In the remainder of this paper, I consider what it means to act on an attitude of moral uncertainty in these kinds of cases. I will argue that doing so does not necessitate relying on so‐called fair processes, or forgoing substantive justice in favor of procedural justice.

2.2. Priority‐Setting and Fair Processes

In light of the moral uncertainty and disagreement many priority‐setting decisions incite, some ethicists have endorsed using fair processes to make them. For instance, Daniels writes: “the outcomes of fair procedures should count as fair when we cannot agree on principles for resolving the disputes” [18, p. 25]. ⁶ This view has proven influential: it has both spawned a large literature and shaped priority‐setting practices around the world [19, 20, 21]. Nir Eyal critiques this view, but eloquently characterizes it:

“The Move: When there is no single best compromise between (allocation) principles or values, such that the best health policy (e.g. the best health resource allocation) is rationally underdetermined, a democratic deliberative process… is needed. Indeed, that rational underdeterminacy is a part of what warrants that deliberative process. How? Through two “links” between the two: (a) democratic deliberation can reach the determinacy needed; and (b) democratic deliberation treats justly and respectfully even those patients whom the (distributive) compromise selected disadvantages (e.g. deprives of important or life‐saving health resources), thereby keeping that compromise's legitimacy” [22].

There has been considerable debate about what makes a process fair, and how we can ensure real‐world processes meet whatever standards are established. My aim in this paper is not to reject any specific formulation of a fair process, but rather argue that we have erred in concluding that the best we can do under conditions of moral uncertainty is to make decisions that are procedurally, rather than substantively, just. Nonetheless, it is worth spelling out more explicitly the criteria that proceduralists believe fair processes need to satisfy, as well as how such processes tend to play out, so we can more concretely grasp the object of this criticism.

There are several fair processes that have been proposed for setting health care priorities; however, among the most influential of these is the process developed by Daniels and Sabin, Accountability for Reasonableness (A4R), and defended in Daniels' book, Just Health [18, 23]. Three specific aspects of A4R are worth highlighting because most proceduralists endorse them.

First, Daniels acknowledges the relevance of substantive justice. Daniels thinks that allocation schemes must be based on reasonable rationales, or those that appeal to “evidence, reasons, and principles that are accepted as relevant by (‘fair minded’) people who are disposed to finding mutually justifiable terms of cooperation" [18, p. 118]. In other words, just because Daniels believes we should aspire to procedural justice, this does not mean that anything goes; certain procedural elements must be satisfied, and one such element is the requirement that decisions be made on grounds that reasonable people agree are relevant. The “relevant reasons” condition seems to rule out making decisions based on justifications that are flagrantly biased or discriminatory, although Daniels does not offer explicit guidance on how or where to draw these lines. For instance, Daniels suggests that “reasons that rest on matters of religious faith” fall on the unreasonable side of the reasonable‐unreasonable dichotomy, while discriminating against overweight people falls on the reasonable side, noting that: “Reasonable people will most likely disagree both about assigning responsibility… We may all agree that Billy chose the fries over the fruit and in that sense is responsible for what he ate, but little else follows about what we continue to owe Billy if he gets fat and sick” [18, p. 157].

Second, to ensure that decisions are made on the basis of reasons and principles reasonable people see as relevant, Daniels recommends involving stakeholders in decision‐making processes. Specifically, Daniels notes that “[stakeholder participation] would increase the likelihood that a broader range of relevant reasons and rationales will be aired in the decision‐making process” [18, p. 130]. Thus, a fair process is conceived of as one in which stakeholders actually decide on the principles that are relevant, rather than one in which the principles are theoretically justifiable to reasonable people (including stakeholders). While Daniels appears to primarily justify the inclusion of stakeholders on instrumental grounds, others suggest that inclusion is intrinsically valuable in priority‐setting, lending legitimacy to the resultant decisions [24]. For instance, Bruni et al. note that: “Because the public funds and uses the health care system, citizens are the most important stakeholders of the health care system. Thus, legitimacy and fairness demand that they be at the priority‐setting table” [25].

Finally, for many proceduralists, deliberation is a critical component of priority‐setting processes. Daniels notes that “In moral disagreements, we want people to bring reasons to bear in an effort to persuade each other… A dispute that is resolved by democratic procedures after careful deliberation about the various reasons put forward on both sides has in its favor the fact that even losers will know that their beliefs about what is right were taken seriously by others” [18, p. 116]. Similarly, Baltussen et al. suggest that deliberation ensures that participants in a priority‐setting process are given the opportunity “to identify, reflect and learn about the meaning and importance of relevant social values” [26].

To summarize: proceduralists argue that priority‐setting decisions incite moral uncertainty and moral disagreement; accordingly, health actors should use fair processes to make these decisions. Fair processes, in turn, require basing these decisions on principles that reasonable people agree are relevant and involving stakeholders (who, in the case of health resources, include members of the public) directly in deliberative decision‐making processes. In the end, proceduralists believe that we should assess the fairness of priority‐setting decisions by assessing the processes used to make them.

In recent years, health actors have increasingly appealed to procedural justice in setting priorities, and have aimed to use processes that satisfy the above criteria. Some governments have adopted A4R, but many have used alternate processes [27]. For instance, in 2013, the Colombian Ministry of Health “undertook a systematic review to identify criteria, from which a shortlist was selected by relevant stakeholders… Weights were obtained from a survey of 200 people from the Colombian general population” [28]. The Ministry of Health then assessed how well each intervention would perform on the chosen criteria, multiplied these scores by the weights assigned to each, and funded the highest‐ranking interventions. In Spain, an analogous priority‐setting process asked 28 stakeholders—including patients, pharmacists, physicians, economists, and policymakers—to deliberate about and then assign weights to different ethical criteria [29]. The proliferation of efforts to use fair processes to make priority‐setting decisions reflects broad support for this enterprise, but the heterogeneity of these efforts suggests underlying ambiguity about the point of such processes.

3. Substantive and Procedural Justice in Health Care Priority Setting

Proceduralists agree that moral uncertainty necessitates a turn from substantive to procedural justice when making health priority‐setting decisions because only a fair process can yield legitimate verdicts in cases where moral principles conflict. Their views can be divided into roughly two camps: pure proceduralists believe fair processes represent the sole source of justification for a just outcome, while hybrid proceduralists believe substantive criteria limit the range of just outcomes, and that fair processes should be used to select among these. I reject both sets of views, for reasons I will return to. But it is first necessary to clarify the distinction between substantive and procedural justice.

In certain situations, we know, or can figure out, what the just outcome would be (i.e., what would be substantively just). A clear example of this kind of situation is one in which one person needs a health resource—say, a kidney—while another does not. In this case, we know that the just outcome is to give the kidney to the person who needs it, and we can achieve substantive justice in this sort of case, independent of any process we use. However, when relevant moral principles conflict, substantive justice may be rationally underdetermined: there may either be no truth of the matter about what constitutes a just outcome (as some ethicists concerned with indeterminacy have claimed), or there may be some truth of the matter, but we cannot devise a procedure that reliably reaches the right conclusion. ⁷

John Rawls describes cases of pure procedural justice as those “when there is no independent criterion for the right result” [30]. For instance: flipping a coin to determine which player gets to serve first in a tennis match constitutes a case of pure procedural justice, because there is no substantive truth about who ought to serve first, independent of the results of this coin toss [31]. By contrast, cases of imperfect procedural justice are characterized by the fact that “while there is an independent criterion for the correct outcome, there is no feasible procedure which is sure to lead to it” [30]. Rawls provides the example of a criminal trial: although there is some truth of the matter about whether a defendant is innocent or guilty, and we view verdicts that reach the wrong conclusions about innocence or guilt as unjust ones, the best we can do is design procedures that most reliably reach the substantively right results.

Daniels and some other proceduralists writing in this vein suggest that when we face moral uncertainty in setting priorities, we find ourselves in the first kind of case. Specifically, Daniels writes: “The fair process I seek will turn out to be pure as long as we have no consensus on fine‐grained distributive principles,” and notes elsewhere that “the general principles of justice for health… are simply too general and too indeterminate to resolve many reasonable disputes about how to allocate resources fairly to meet health needs, and we lack a consensus on more fine‐grained principles” [18, 32]. Thus, for Daniels, there is no single independent criterion for the right result, because “No principle enjoys our endorsement independently of the fair procedure we are employing” [18, p. 136]. For this reason, Daniels considers himself to be a pure proceduralist. ⁸

However, most proceduralists are hybrid proceduralists. For instance, Baltussen et al. propose using evidence‐informed deliberative processes [26]. They cite the example of the Netherlands Health Care Institute (ZINL), which assesses the effectiveness, cost‐effectiveness, necessity, and feasibility (among other criteria) of various health interventions to decide what care to include in the benefits package. But they argue that ZINL should expand stakeholder involvement “in order to further foster the legitimacy of their decisions” [26]. Like Baltussen et al., most proceduralists agree that substantive criteria (like effectiveness) should limit the range of outcomes under consideration, but they think it is better to additionally incorporate procedural elements, like stakeholder engagement and deliberation.

As Eyal notes, and as is evident from the examples of Daniels and Baltussen, the justification for the incorporation of procedural elements typically boils down to legitimacy and determinacy. I will return to the point of legitimacy later. But the primary problem with so‐called fair processes is that they introduce arbitrary determinacy, in that they yield decisions that are less likely to do what is most important: ensure the optimal distribution of health resources.

I will suggest that we can avoid arbitrariness by using an independent criterion to assess the justice or injustice of any given priority‐setting decision. Specifically, I will argue that decision makers should aim to maximize expected choiceworthiness. We should evaluate priority setting decisions based on whether they maximize expected choiceworthiness and should design decision‐making processes with this aim in mind. In other words, processes should only be deliberative, participatory, and so on, to the extent that incorporating these procedural elements facilitates the maximization of expected choiceworthiness.

4. Why Moral Uncertainty Does Not Necessitate a Turn to Procedural Justice

Although proceduralists have differing views on what a fair process should look like, they share a commitment to the idea that there is no substantive truth about what makes a priority‐setting decision just, independent of the process that is used. But there are two fundamental problems with this view. First, there is no more substantive truth about what makes a priority‐setting process just than there is about what makes a priority‐setting decision just. Accordingly, the presence of moral uncertainty regarding substantively just allocations of health resources cannot be used to justify a turn to procedural approaches. Second, so‐called fair processes produce allocations that more often fall outside of the range of those that are plausibly just. This is an intractable problem because, as I will show, there tends to be an inverse relationship between health care: (1) being cost‐effective and equitable and (2) having ready‐made stakeholders and advocates. Importantly, this latter critique applies even to hybrid proceduralist views. Even if we constrain the range of allocations via substantive principles, and then use a representative, deliberative process to choose among them, we still have to contend with the fact that such processes are likely to favor the less defensible of the remaining options. As a result, hybrid proceduralist views may constitute the worst of both worlds: they accept that substantive principles can identify and exclude unjust allocations (the bullet I argue we should bite) while still introducing decision‐making procedures that bias decisions towards less cost‐effective, more inequitable care.

4.1. Moral Uncertainty in Designing Fair Processes

Relying on a procedural approach introduces three important kinds of moral uncertainty: uncertainty about who should be involved in a fair process, uncertainty about the decision procedure a fair process should employ, and uncertainty about the role a fair process is meant to play in resolving moral uncertainty. As a result, health actors stand to face just as much moral uncertainty in designing a fair process as they do in simply making an allocation decision.

First, and most importantly, it is not clear who should be involved in making priority‐setting decisions. When involving the public in priority‐setting processes, governments have tended to define “the public” differently, with some aiming to involve citizens, patients or service users, or representatives of organized groups [33]. Similarly, defining “stakeholders” as “individuals or groups involved in or affected by health‐ and health care‐related decisions, programs, or policies” [34], requires making substantive commitments about who will be affected by an allocation decision before that decision is made. For instance, the UK's Joint Committee on Vaccination and Immunisation (JCVI) was tasked with prioritizing populations for COVID vaccination [35]. The JCVI included 16 members, each of whom lived in the UK. But it is not obvious why the JCVI should have only prioritized UK‐based populations for access to COVID vaccines, as opposed to, for instance, high‐risk populations outside the UK, short‐term visitors to the UK, UK citizens living internationally, or residents of former UK colonies. Moreover, even if one thinks it is appropriate for the British government to prioritize vaccinating UK residents (for moral or practical reasons), people living outside the UK (for instance, in areas UK residents frequently travel to) are affected by the UK's vaccination policies, rendering them stakeholders. Given this, why should they have been excluded from participating in the priority‐setting process? Even if we concede that stakeholders outside the UK are generally less affected by vaccine prioritization decisions than stakeholders within the UK, we must still determine how to develop processes where stakeholders are represented in a balanced or proportional way.

Moreover, limiting participation to only those residing in a given nation will predictably affect the outcomes of the prioritization decisions that are made, skewing them towards allocations that favor co‐nationals. The policymakers who opted to place 16 UK residents on the JCVI might have anticipated that doing so would increase the committee's likelihood of prioritizing UK residents. It is not clear why these policymakers would be licensed to predictably tip the scales towards certain decisions, but not make those decisions outright. And although many would contend that excluding the rest of the world from consideration in allocating scarce vaccines was unjust, by relying on a so‐called fair process, decision makers insulated this decision under a veneer of procedural legitimacy [36].

The challenge of who to include in priority‐setting processes is a version of the democratic boundary problem, wherein we do not know who has the right to participate in collective decisions [37]. Clearly, it would be a mistake to abandon democratic processes writ large because we struggle to define the boundaries of the demos. However, this is because there are many arguments for using democratic processes to elect leaders (rather than make technical policy decisions), and a lack of reasonable alternatives. In addition, the kinds of priority‐setting decisions with which I am concerned are being made in broadly democratic contexts; thus, dispensing with a democratic process in this domain does not entail dispensing with democratic processes writ large. But those who argue that moral uncertainty justifies a turn to procedural solutions face a fundamental challenge: if there is significant, unresolved moral uncertainty about who ought to be involved in democratic priority‐setting procedures, then moral uncertainty about the principles that ought to be used can no longer serve as an argument in favor of procedural solutions.

Second, there is moral uncertainty about the decision procedure that should be used in a fair process. Daniels and Sabin propose one approach, A4R, but even scholars who support the use of A4R disagree on its appropriate scope [38]. More generally, governments have relied on a variety of public engagement processes in setting priorities: one review found that 405 techniques have been employed, with methods ranging from citizen juries, to opinion polls and surveys, to focus groups, to town meetings with voting, among others [33]. The diversity of the procedures used suggests that disagreement does not just occur at the margins; rather, health actors have fundamentally different views on the correct decision‐making process. Indeed, there may be as much uncertainty about the appropriate process to use as there is about the correct weighting of principles, since ethicists tend to favor considering only a few—rather than hundreds—of principles [39]. The wide range of potential approaches a health actor needs to choose among undermines the argument that moral uncertainty necessitates a turn to fair processes for reasons of determinacy. Moreover, this suggests that the determinacy that results from whatever process is selected (from the hundreds on offer) is liable to be arbitrary, in that a different, similarly justifiable procedure would have produced a different outcome.

Finally, people disagree on the appropriate role of a fair process in resolving moral uncertainty. Some hybrid proceduralists suggest that the appropriate role of a fair process is to select among allocation schemes that have already been deemed substantively just, while pure proceduralists contend that fair processes are meant to do more than this [38]. Whichever line one takes here creates problems. If one thinks that we should only use fair processes to select among allocation schemes that have already been deemed substantively just, as MacDougall et al. do, this forces us to determine which allocations are just or unjust before the fair process commences. It is not clear what principles would enable policymakers to draw fine lines between barely unjust and barely just allocations, as concerns about incommensurability presumably apply here, too. It is also not clear what would license policymakers to limit the scope of options to only those they consider substantively just, but why it would be illegitimate for them to take a single step further (i.e., by eliminating an option they view as likely unjust). Presumably, the tools they would use to eliminate obviously unjust schemes—evidence and reason—would not go from providing definitive answers about what options should be on the table to offering no guidance at all about how to rule out additional options. Conversely, if one thinks we should not use fair processes to rule out potentially unjust allocations, this seems to leave open the possibility that the use of a fair process will produce substantively unjust allocations of health resources. In short, if a process is substantively constrained, then substantive principles are doing heavy lifting. If it is not constrained, then fair processes can legitimate unjust outcomes. ⁹

Two final points are worth making: I have argued that the moral uncertainty health actors face in deciding how to structure fair processes is as significant as the uncertainty they face in simply making allocation decisions. But importantly, this uncertainty is additive, so not only do procedural approaches not solve the original problem—how to make decisions when important moral principles recommend different actions—they compound it. While fair processes provide a tool by which to make decisions, they do not help the people involved in those processes actually make them; fair processes simply kick the can down the road.

Moreover, by kicking the can down the road, proceduralists have shifted our focus to resolving the wrong kind of moral uncertainty. Rather than figuring out how to optimally distribute health resources, we are instead left asking questions about who should be in the room to make the process fair, and what specific decision procedure they should use. And while answering these questions may move us closer to having an allocation, it does not move us closer to having the correct allocation. After all, if the thing that matters most is distributing health resources well, we should be asking substantive questions, like whether to prioritize averting smaller harms for more badly off groups or larger harms for more advantaged ones. The processes we use should provide the best possible answers to these kinds of substantive questions. When we fail to prioritize the right questions, we end up with allocations that are determinate, but arbitrary.

4.2. Conflicts Between Substantive and Procedural Justice

As stated, my goal is not to advocate for any particular kind of process, but instead for the general proposition that we should aspire to substantive—rather than procedural—justice in setting priorities. This proposal leaves open the possibility that any given process (including A4R) could be the right one, so long as it reliably produces substantively just outcomes. Of course, proceduralists contend that part of the problem is that we do not know exactly what would constitute a substantively just allocation of resources, since people may weigh different values differently. That said, proceduralists acknowledge both that some allocations of health resources are unjust and that these allocations ought to be avoided. My contention is as follows: in our pursuit of the procedural elements that many proceduralists consider intrinsic to procedural justice—for example, stakeholder engagement and deliberation—we increase the risk of stumbling upon allocations that are outside the window of those that are plausibly just (e.g., by forgoing efficiency and equity without achieving commensurate benefits). The connection between these two claims—that proceduralists concede some allocations are unjust, and that fair processes tend to produce such unjust allocations—is not a conceptual one, but an empirical one. In the remainder of this section, I defend this empirical point, arguing that so‐called fair processes tend to produce unjust allocations. I will introduce empirical evidence suggesting that fair processes have a dubious track record, and will then argue that there are fundamental reasons to expect that so‐called fair processes will systematically lead to unjust allocations.

Because many priority‐setting processes occur in contexts with severely constrained health budgets—hence the need for priority setting in the first place—there are trade‐offs between ensuring priority‐setting processes run smoothly and achieving other, important health system goals. In Tanzania, the government spends just $39 per capita on health expenses each year [40]. (At Brigham and Women's Hospital in Boston, a simple urine pregnancy test can cost twice this [41]). One paper describing the use of A4R in Tanzania includes a concrete example of such trade‐offs: “The actual implementation of the [A4R] intervention fell short of the initial plan. A delay in funding disbursements delayed part of the implementation process. With time, and as circumstances dictated, the plan to monitor and evaluate service domains such as malaria, HIV/AIDS, emergency obstetric care and generalised care were dropped, and the focus remained merely on monitoring the priority‐setting process and management changes within the [Council Health Management Team] and at the district hospital” [42]. In short, investing limited resources in a priority‐setting process places us in the red: resources that might otherwise have been used to provide health care have instead been used to make resource allocation decisions. The process must thus provide commensurate benefits (e.g., a tendency to prioritize more cost‐effective care, or broader public acceptance of health priorities) to justify these costs.

Worryingly, there is evidence that widely used priority‐setting processes systematically fail to achieve these kinds of benefits, in part because they cannot reliably be implemented well, especially in underresourced settings. One review found that: “In evaluating A4R in Kenya, Tanzania and Zambia researchers even found that it could not be implemented according to its four key conditions. They and others in [high‐income countries] found A4R to be conceptually abstract, difficult to implement and susceptible to contextual limitations” [27]. The authors of the Tanzania study also note that “limited local autonomy, low level of public awareness, unreliable and untimely funding, inadequate accountability mechanisms, and limited local resources were the major contextual factors that hampered the full implementation.” In a separate study in Uganda—which has a constitution mandating public participation in policymaking, as well as political structures designed to facilitate this—researchers similarly found that it was difficult to achieve meaningful public engagement in priority‐setting, due to “a lack of both individual and community empowerment, poverty, a lack of interest, and a lack of mobilization, as well as a failure to effectively implement policies mandating public participation in health‐sector planning and priority‐setting” [43]. Finally, the authors of the review note that “We found three comprehensive [priority‐setting processes] which acknowledge the role of context (political, social, financial, cultural etc.), and which have been put forward as candidates to guide priority‐setting in a wide range of settings, including LMICs… While two are in regular use, it is evident from the following summaries that none provides a process yet able to be recommended at any level in LMICs” [27]. ¹⁰

My point is not that we should abandon the pursuit of democratic processes in LMICs because they do not always live up to their aspirational goals. ¹¹ Rather, my argument is narrow: more involved priority‐setting processes deprive already strained health care systems of badly needed health resources. This cost can only be justified if these processes achieve some kind of benefit—ideally, the benefit that justified implementing a given process in the first place. But there is little evidence that so‐called fair processes consistently satisfy their own procedural criteria, let alone produce substantively just allocations of health resources.

It is reasonable to object that the aforementioned data should not cause us to condemn fair processes writ large, just as an unfair election should not lead us to broadly condemn democracy. Proceduralists routinely make this point, and studies documenting the ongoing challenges we face in effectively implementing priority‐setting processes typically also call for more time, energy, and resources to be invested into improving these processes. But crucially, while investing additional resources into these processes could result in them more reliably meeting their own procedural criteria (e.g., achieving meaningful deliberation), there are two reasons to think that even optimally designed and implemented processes will fail to reliably produce substantively just allocations of health resources. One of these reasons pertains to efficiency, and the other pertains to equity. Importantly, both challenges reflect a structural problem rather than a failure of execution: they would persist even for a process designed and implemented as well as one realistically could be.

As discussed, proceduralists almost universally recommend engaging stakeholders in priority‐setting processes to ensure that people who are affected by decisions have a say in them. However, some kinds of health care have ready‐made advocates, while other kinds do not. Treatments that specific people know they will need—for instance, gene therapies for identifiable genetic conditions—or that certain actors tend to profit from providing—for instance, expensive screening tests—tend to have vocal stakeholders; by contrast, treatments that people cannot anticipate they will need, or that provide small benefits to many people, have fewer natural advocates. Unfortunately, many of the most cost‐effective health interventions fall into this latter category [44]. As a result, engaging stakeholders in setting priorities is liable to skew decisions towards less cost‐effective care.

In 2022, the United States passed legislation enabling Medicare to negotiate drug prices with pharmaceutical companies, a measure that is expected to save hundreds of billions of dollars, and that has been almost universally lauded by health policy scholars [45]. However, perhaps surprisingly, some of the people most wary of this policy are patients taking expensive drugs and the patient advocacy groups that represent them [46]. For instance, one patient with cystic fibrosis expressed reluctance that a state drug affordability board was considering negotiating the price of Trikafta, a drug that costs $200,000 per year, due to concerns that this could chill future drug development. “It just puts Trikafta as a whole at risk… It would start here, but it could create a ripple effect” [47]. While the patients who are on Trikafta may or may not benefit from the government negotiating the drug's price, many more people stand to lose if the government does not, because their co‐pays may increase or they may lose access to needed services as the state strives to control rising health care costs.

Critically, these other people often do not know they stand to lose, and thus are less likely to participate in priority‐setting processes, or engage as deeply if they do. Stakeholder engagement is thus inherently asymmetric: because many stakeholders may not realize they are stakeholders, there is at best a weak relationship—and more likely an inverse relationship—between what care most warrants advocacy, and what care has natural advocates. Thus, while involving patients directly in priority‐setting processes seems like it would lead to priorities that better align with the public's values, in practice, doing so is likely to instead bias these processes towards the interests of small, powerful patient groups and the private companies that back them [48]. As a result, the kinds of cost‐effective public health interventions that should be no‐brainers for a health system are more likely to end up on the chopping block when priorities are set in accordance with the procedural criteria proceduralists tend to endorse.

Second, as discussed, the way decision‐making processes are structured predictably influences the decisions they yield. For instance, it is unsurprising that a process that included only UK residents would generate an allocation that benefited UK residents. The problem is that it is virtually impossible to structure a priority‐setting process such that a representative sample of the public can meaningfully participate. This means that decisions will often be biased in favor of the advantaged people who are able to attend and advocate for their positions, and biased against people who—through no fault of their own—cannot attend or meaningfully participate.

Unfortunately, the procedural elements many have claimed are intrinsic components of procedural justice are also liable to produce allocations that compound health inequities. This challenge cannot be solved by simply optimizing process design or throwing resources at the problem (e.g., by compensating participants, providing free transportation to the priority‐setting process, and so on), in part because medical conditions—the very things stakeholders are meant to be deliberating about—pose fundamental barriers to participation.

At baseline, people with disabilities or low self‐reported health are significantly less likely to vote [49, 50]. Although these disparities can be mitigated somewhat by decreasing barriers to participation, the more involved a priority‐setting process is, the harder it will be for people with significant disabilities to participate [50]. Thus, the stakeholders with the most significant health issues, and who are most likely to be affected by the decisions that are made, are often least able to participate in making them. This is an inherent challenge in democratic decision‐making, but in other contexts it may be surmountable, while in health care priority setting, it cannot be overcome: after all, there is no treatment that would allow a patient with advanced dementia to participate in a deliberative priority‐setting process, but such processes determine whether advanced dementia treatments will ever become available. Indeed, the very clinical tools that are used to assess the severity of dementia—which evaluate qualities like judgment, problem solving, memory, and functioning in social groups—might be used to assess one's ability to participate in deliberative processes [51]. Nor is dementia a cherry‐picked example: many illnesses impair functioning by sapping people of energy, attention, cognition, and time. These same resources are prerequisites to engaging in deliberative processes. As a result, many of the people most in need of health resources will be unable to participate in even the most carefully designed, well‐implemented deliberative processes. Processes that, by their very design, tend to exclude the people most in need of health resources are unlikely to produce allocations that favor those people.

5. Substantive Justice Under Moral Uncertainty

Proceduralists suggest that in cases where different moral theories recommend different courses of action, we must pursue procedural, rather than substantive, justice because we lack an independent criterion for assessing whether allocations are substantively just. I have argued that there are significant conceptual and practical problems with the proceduralists' view; however, I have not suggested an alternate approach. In the remainder of this paper, I describe what substantive justice in health care priority setting might look like, even under conditions of moral uncertainty.

5.1. Patients With Asthma or Obesity

During the COVID‐19 pandemic, the following kinds of cases often arose: ministries of health had limited supplies of COVID‐19 vaccines and needed to decide whether to first vaccinate patients with asthma or obesity. Data showed that vaccinating patients with obesity would save many more lives, because patients with obesity were at much higher risk [52, 53]. Using lives saved as a proxy for utility, prioritizing the patients with obesity could be expected to produce greater utility. But this did not answer the question of who should be vaccinated first. Principles aside from utility matter too. For instance, another principle Daniels suggests it is reasonable to care about—whether someone is personally responsible for their health condition—seems to recommend prioritizing the patients with asthma, because they clearly bear no responsibility for their medical condition, whereas some contend that at least some patients with obesity do [54]. To the extent that saving lives and holding people personally responsible are conflicting goals, it is intuitive that the former is more important, but many reasonable people (including Daniels) assign some weight to the latter. Thus, the decision about who to prioritize involves moral uncertainty. Given that health actors could not simultaneously vaccinate both groups, they needed to choose which to prioritize first.

Different moral theories or moral principles assess options (e.g., prioritize people with asthma or obesity) in terms of their choiceworthiness. MacAskill et al. define the choiceworthiness of an option as “the all‐things‐considered strength of reasons in favour of that option” [17]. For instance, one version of utilitarianism would assess an allocation scheme that saves the most lives as the most choiceworthy one. The problem we face when we are morally uncertain is that different theories that we have some credence in deem different options more or less choiceworthy. In the above case, the principles of utility and responsibility deem opposing allocation schemes more choiceworthy. We need some way of making decisions when multiple ethical principles we have credence in recommend different courses of action.

Notably, the case just described is exactly the kind of case proceduralists use to justify a turn towards fair processes. Daniels writes:

“Any decision we make about how to use resources to meet health needs can create winners and losers. Yet, we lack a consensus on principles for resolving disputes among these winners and losers. We have reasonable moral disagreement about how to treat people fairly. Perhaps we can eventually arrive at philosophical analyses of these issues that produce wide acceptance of appropriate principles, and we should invest in such work, but in the real world, resources must be allocated in real time in a way that is perceived to be legitimate and fair” [18, p. 25].

And while philosophical analysis may not yet have produced widespread acceptance of the appropriate first‐order principles, it has produced a widely accepted account of how we ought to weigh those principles when they conflict.

5.2. Maximizing Expected Choiceworthiness (MEC)

MEC is the most widely endorsed theory of how decision makers facing moral uncertainty ought to make moral decisions [55]. ¹² MEC says that when we face moral uncertainty, the option we should choose “is the one with the highest expected choiceworthiness, where the expected choiceworthiness of an option is given by the sum, over all moral theories, of the decision maker's credence in that theory multiplied by the choiceworthiness of the option, if that theory is true” [17]. MEC for moral uncertainty is thus analogous to expected value theory for empirical uncertainty. A health actor aiming to implement MEC in the above case would determine the credence they had that each principle ought to be the decisive one (utility and responsibility) and which principle the stakes of erring are higher under. In this case, the stakes of erring are higher under the principle of utility, because prioritizing the patients with obesity will save far more lives, but patients with obesity are, at most, only slightly responsible for their health condition (Table 1).

Table 1.

MEC in asthma or obesity.

	Utility	Responsibility
Prioritize patients with asthma	Significant wrong	Permissible
Prioritize patients with obesity	Permissible	Mild wrong

In addition, while utility clearly matters when allocating health resources—no one would contest that we should try to save more lives—it is not obvious that theories of personal responsibility should influence our allocation decisions. As Daniels himself notes, even if one determines that a patient is responsible for their health condition, this does not necessarily mean that we should deny him resources on this basis. Thus, we should also have more credence that utility matters in this case than we should that personal responsibility does. We should prioritize the patients with obesity for COVID vaccination.

This example illustrates a few important points. First, the intuition that drives the conclusion that we should prioritize patients with obesity seems to match the logical steps taken when reasoning about this case using MEC. Specifically, “Which theory are the stakes higher under?” and “Which theory do we have more credence in?” seem like the right questions to ask to reach the conclusion we reached.

Second, using MEC, we arrived at a substantive verdict in a case where principles Daniels contends may constitute “relevant reasons” conflict. Crucially, many proceduralists have suggested that such verdicts cannot be reached. Their suggestion—that we use a fair process to make these kinds of decisions—leaves open the possibility that we might deprioritize patients with obesity in the above case, an outcome that would be substantively unjust.

I have thus shown that definitive verdicts can be reached in at least some cases where moral principles conflict. Critically, the substantive verdict we reached seems, intuitively, to be the right one: decision makers should prioritize patients with obesity in the above case. Indeed, it would be a mistake to prioritize the patients with asthma or to conclude that either outcome is fine (so long as a fair process is used), because prioritizing patients with obesity will save many more lives, and the proposition that people should be denied health resources because they may bear some responsibility for their conditions is, at best, a dubious one. Thus, MEC not only yields a definitive verdict; it yields the right verdict. The concept of substantive justice is not rendered moot by the presence of moral uncertainty.

While the case I have described is a relatively straightforward one, and MEC will not provide ready‐made conclusions in all the complex priority‐setting cases we are likely to encounter in the real world, the fact that we can reach substantively just verdicts under conditions of moral uncertainty suggests a path towards doing so in more complex ones [56].

5.3. Patients With Chronic Conditions or the Elderly

5.3.1.

Imagine a ministry of health is in the same situation as before, but is now trying to decide whether to prioritize healthy elderly people or slightly younger people with chronic conditions. Imagine we know that prioritizing the healthy elderly will save significantly more lives, since this population is at higher risk, but we also know that people with preexisting conditions are more disadvantaged, by virtue of having chronic conditions, and also being slightly younger, meaning they haven't had the opportunity to live as long (Table 2).

Table 2.

MEC in chronic conditions or the elderly.

	Utility	Prioritize the worst off
Prioritize the elderly	Permissible	Wrong
Prioritize the chronically ill	Wrong	Permissible

It is not clear under which theory the stakes are higher: is it a worse thing to prioritize a somewhat better off person under a principle that recommends prioritizing the worst off, or to save fewer lives under a principle that recommends saving the most lives? Thus, our decision under MEC would come down to what credence we have in each principle. Troublingly, when we face a given priority‐setting decision, it is not immediately clear what credences we ought to have in each principle relevant to that decision; indeed, reasonable people would disagree on this point. MEC correspondingly ceases to be (immediately) action‐guiding in this harder case. And many of the moral dilemmas we face when allocating health resources are more complex still. Indeed, most ethicists who work on issues related to priority‐setting think that at least two principles (maximizing benefits and prioritizing the worst off) are relevant, and some believe that additional principles (in particular, reciprocity and treating people equally) are as well [39]. Furthermore, there are different ways of cashing out the above principles: for instance, we might think that “maximizing benefits” means considering not just lives saved, but also life years saved, and the quality of the life years that are saved. Moreover, we might be considering prioritizing more groups than just the two described above. Thus, to use MEC in the real world, we might need to draw a table with many more cells, and then carefully determine the stakes of each principle under consideration, as well as the credences we should have in each principle. Is MEC thus only useful in straightforward cases, like Patients with Asthma or Obesity?

The answer to this question is no, because there is nothing, in principle, that makes Patients with Asthma or Obesity different from Chronic Conditions or the Elderly. The difference between these cases is a difference of degree (with respect to their complexity), not kind. Thus, if substantive justice is within our grasp in the former case, it ought to be attainable in the latter case as well, although we may have to reach farther. While MEC does not deliver the kind of obvious, ready‐made verdicts in hard cases as it does in easy ones, it still provides guidance as to how we might approach such verdicts.

5.4. Implementing MEC

I will not provide a complete account of what a decision‐making procedure that reliably achieves substantively just allocations might look like, but MEC can give us a general sense. Under such a procedure, health actors would need to: (1) Determine what choiceworthiness score each relevant theory assigns to each option and (2) Determine what credences to have in different moral viewpoints [55, 57]. For each, health actors would also have to determine who should be involved and what procedure they should use (Table 3).

Table 3.

Decision procedure recommended by MEC.

	Who should be involved in this decision (e.g., ethicists, empirical experts, the public)?	What decision procedure should we use (e.g., voting, surveys, deliberation)?
What should our credences in different moral principles be?	Credence decision makers	Credence procedure
What choiceworthiness score does each principle assign to each option?	Choiceworthiness decision makers	Choiceworthiness procedure

Importantly, while this could lead to an involved process, such a process would not be required for its own sake. The goal of any allocation procedure would be to distribute resources well (i.e., to produce substantively just allocations). Thus, more time‐consuming and expensive processes would only be undertaken when this would be expected to improve the final allocations of health resources. This would likely be the case for only the most high‐stakes decisions (i.e., those affecting many people, or affecting each in significant ways). After all, if only $50,000 is being distributed, spending $25,000 on a decision‐making process will not achieve an optimal distribution of resources. This has an intuitive implication: health actors should be guided by MEC in making resource allocation decisions, but need not undertake extensive processes in most cases. For day‐to‐day decisions, individual decision makers within organizations might use the credences established in more extensive processes, while serving as choiceworthiness decision makers themselves: researching the relevant empirical facts, making MEC decision tables, and identifying the most choiceworthy allocations. It is only when making the most significant decisions that more involved processes should be used.

MEC also has implications for what a more involved process might look like in high‐stakes cases. Our credences in different moral principles should be based on the strength of the arguments for those principles. The people best positioned to make these decisions are those equipped to assess how normative arguments are made and weigh the case for including different principles. At its most formal, such a process might necessitate reviewing the relevant ethics literature, airing competing arguments, evaluating survey data on a population's credences in various principles, and determining the appropriate weight to assign to each principle. Because the resulting credences would be grounded in reasons, rather than the bare preferences of those involved, they would be, almost definitionally, justifiable to reasonable people.

By contrast, a broader coalition of decision makers might be involved in assessing the choiceworthiness of different options under each principle. For instance, determining what score the utility principle assigns to prioritizing the chronically ill rather than the elderly first requires having empirical experts estimate relevant facts, such as how many DALYs would be averted by each option. The same decision makers involved in determining credences could then assess what each principle recommends in light of those facts by reviewing the relevant moral arguments and deliberating about them. Again, the results of such a procedure would rest on evidence and reason, rendering it publicly justifiable.

Notably, even in the most formal cases, this procedure does not assign a direct role to the public. This is not because the public's input lacks value, but instead because the two tasks at issue call for specific competencies. Priority‐setting decisions made directly by the public would only track substantive justice if the participants happened to be well‐positioned to assign credences and assess the choiceworthiness of the various options under consideration. This is not, in general, what we should expect.

Still, the public's values are not irrelevant. There are good reasons to systematically assess them through large surveys or qualitative research (rather than through the deliberation of a small, frequently unrepresentative group of stakeholders). For instance, if many members of the public reject a particular principle, this should often lower our credence in that principle. Similarly, if many members of the public support prioritizing a particular group, this gives us one, potentially overridable reason to prioritize that group. Critically, though, in a democratic society, a fair priority‐setting process need not—and indeed, generally should not—directly involve the public.

6. Three Objections

6.1. Indeterminacy

One might reasonably claim that I have dodged the issue of indeterminacy by presenting prioritization decisions as discrete choices between distinct patient populations. What some proceduralists instead seem to envision is a deliberative process that would result in participants assigning, for example, 32.7 percent weight to equity and 67.3 percent weight to efficiency in setting priorities, and then allocating resources to different populations on this basis. It is hard to imagine that MEC would render clear, substantive verdicts in this kind of case (or that such substantive verdicts even exist).

I think this is the wrong question to ask, though. The trouble for Daniels is that all policy issues are like this: a decision about whether to charge $3.45 or $3.50 for a subway ride undoubtedly introduces trade‐offs between efficiency and equity. The solution to every policy decision cannot be to dispense with ideas about what is substantively just because there is ambiguity about which policy is most just. Resolving each policy decision of this kind via the use of a fair process would make the already rusted wheels of bureaucracy grind to a screeching halt.

Rather, policymakers must be entitled to some leeway in putting options on the table, though all involve the balancing of moral principles under circumstances where no weighting is clearly morally superior. It would be fine, for instance, for policymakers to propose two different subway fare schemes—a flat scheme, like that used on the New York subway, where each ride costs $3.00, regardless of the distance traveled—or a staggered scheme, like that used on the Washington, D.C. Metro, where each ride costs between $2.25 and $6.75, depending on the time of day and distance traveled [58, 59]. MEC could then be used to provide a single, determinate answer to the question “Is it better to use a flat or staggered subway pricing scheme?” even if it could not render a verdict on whether we should charge $3.45 or $3.50 for a ride.

When setting health care priorities, policymakers are generally making the first kind of decision (i.e., one that is discrete and well‐defined), like whether a social insurance scheme should cover a given service or not [60]. This is exactly the kind of question for which MEC can provide a determinate answer.

6.2. Legitimacy

A second, related challenge pertains to legitimacy. Daniels has argued that fair processes address the problem of political legitimacy, or as he notes: “Under what conditions do decision makers have the moral authority to set the limits they impose?" [18, p. 103]. One might worry that a health actor pursuing substantive, rather than procedural, justice would lack the moral authority to deny treatment to certain people. Specifically, if one thinks certain procedural requirements—like the direct involvement of stakeholders in decisions—need to be satisfied for decisions to be legitimate, then processes that do not directly engage the public would be illegitimate. I have suggested that procedures designed to achieve substantively just outcomes do not necessarily need to involve certain procedural components (e.g., public engagement), so long as they reliably produce substantively just outcomes. Does my proposed process thus risk stripping priority‐setting decisions of their legitimacy? I think not, for several reasons.

First, it is worth noting that Daniels himself thinks that “consumer participation is not generally either a necessary or a sufficient condition for establishing legitimacy” [18, p. 129]. However, Daniels does assert that “[stakeholder participation] would increase the likelihood that a broader range of relevant reasons and rationales will be aired in the decision‐making process” [18, p. 130]. This, in turn, can “increase the legitimacy of limit‐setting decisions in both private and public organizations by increasing accountability for reasonableness.” Daniels' assertion that it is compliance with the procedure of A4R that enhances legitimacy is at odds with the views of other proceduralists, many of whom think the legitimacy of priority‐setting decisions is grounded in procedural justice, but do not share Daniels' views regarding the appropriate contents of a fair process [61]. Relatedly, given that the overwhelming majority of priority‐setting decisions are not made via the use of A4R, Daniels' view suggests that the overwhelming majority of priority‐setting decisions are not legitimate. This would mean that, on Daniels' view, many democratically elected governments, running well‐functioning health care systems, and making priority‐setting decisions using mechanisms other than A4R lack the moral authority to make these decisions. While this could be argued, it defies common sense.

However, a bigger problem for Daniels is that his argument is circular. Specifically, he asserts that priority‐setting decisions are only legitimate if they adhere to the requirements of A4R, but he defines the requirements of A4R in terms of political legitimacy, writing: “I shall try to characterize the general conditions such a process must meet if it is to yield outcomes that are perceived as fair and legitimate.” The circular nature of Daniels' justification for A4R makes it difficult to determine exactly what it is that makes a decision‐making procedure legitimate for Daniels—whether it is consent, or respect, or something else—but his justification seems to hew closely to Rawls' [22]. For Daniels, legitimacy is secured by ensuring “relevant reasons and rationales” are considered, while for Rawls “political power is legitimate only when it is exercised in accordance with a constitution (written or unwritten) the essentials of which all citizens, as reasonable and rational, can endorse in the light of their common human reason” [62].

The problem is that Rawls' conception of political legitimacy differs from Daniels' in at least one key respect. For Rawls, political legitimacy rests on the idea of public reason. A large literature has developed around conceptions of public reason, which can be better described as a family of theories [55]. But many of these theories do not support (only) Daniels' recommended process. For instance, as Fabienne Peter notes, “On a common reading today, the Rawlsian idea of public reason is understood in terms of a hypothetical consensus on substantive reasons” (emphasis added) [63]. Thus, there is no requirement that reasonable people literally agree on the priorities that are set, but rather that a law or policy be theoretically justifiable to reasonable people. The process I have recommended satisfies this requirement, because I have defined a substantively just decision as one that maximizes expected choiceworthiness, and an option's choiceworthiness is determined by the “all‐things‐considered strength of reasons in favour of that option.” Though it may constrain the reasons that can be canvassed, it is hard to see how a process designed with the goal of maximizing the all‐things‐considered strength of reasons in favor of an option could not be made compliant with a Rawlsian public justification principle.

Indeed, it may actually be Daniels' recommended process—and others that cannot reliably produce substantively just outcomes—that risks failing the test of political legitimacy. As Danielle Allen, an ardent defender of deliberative democratic processes, writes: “Plato and Aristotle argued that the first criterion for judging how well a given regime performed was simply whether… [rulers] deliver basic well‐being to the population. This is the first requirement for political legitimacy of any kind” [64]. In short, whether a health actor's decisions are legitimate depends in large part on the substance of those decisions; health actors should be judged, first and foremost, on how well they deliver health care. This is precisely the standard I have argued priority‐setting processes should meet—they should reliably produce substantively just results. By contrast, fair processes of the kind Daniels envisions seem more likely to introduce problems pertaining to legitimacy, because they are less likely to guarantee substantively just decisions.

Finally, one might worry about the difficulty of auditing or appealing a MEC‐based process. But in this respect, a process aimed at substantive justice has an advantage over deliberative alternatives: because a MEC‐based decision is built from explicit components—the credence assigned to each principle, and the choiceworthiness each principle assigns to each option—its reasoning is transparent in a way that a deliberative verdict is not. As a result, a patient might challenge the outcome of a MEC‐based decision by targeting a specific, identifiable claim (e.g., that a particular principle was assigned too little credence, that the empirical facts informing a choiceworthiness estimate were wrong, or that a relevant option was never considered). By contrast, when deliberative processes are used, it is hard to directly contest the outcomes, since the justification for those outcomes rests on the fairness of the process. If a process satisfies certain procedural criteria but generates an unjust outcome, stakeholders have little mechanism for recourse. In this way, a process oriented towards substantive justice may not only yield more legitimate decisions, but additionally ones that might be more readily audited or appealed.

6.3. Is MEC Just Proceduralism Relocated?

Lastly, one might object that I have not escaped proceduralism so much as relocated it. After all, when MEC fails to deliver an immediate verdict—as it does in Chronic Conditions or the Elderly—I propose that we assemble a group of people to evaluate ethical arguments and deliberate about what credences to assign to them. This looks much like the proceduralist's move: faced with moral uncertainty, the proceduralist retreats to an involved process. If the proceduralist's appeal to a process merely kicks the can down the road, why does mine not do the same?

The answer lies in what the deliberation is for. For the pure proceduralist, the output of a fair process is constitutive of a just outcome; there is no independent criterion against which the result can be checked, so whatever decision an appropriate process yields simply is the just allocation. For the hybrid proceduralist, substantive criteria constrain the boundaries, but fair processes still choose among the remaining options, meaning a fair process is still constitutive of a just outcome within the defined range. By contrast, on my view, the process is only ever a mechanism for estimating a just outcome that exists independent of whatever process is used. This distinction has two consequences.

First, under MEC, an impeccable process can still reach the wrong answer. This might occur because, for instance, the choiceworthiness scores assigned to various options fail to account for important empirical facts. This result—that a good process can have a bad outcome—is what makes these decisions answerable to a substantive standard, and therefore contestable on substantive grounds, as discussed above.

Second, the kind of process I endorse changes who belongs in the room. In some cases, it may be helpful to have a broad coalition of experts; in others, individual policymakers might suffice. MEC's scalability is a strength of the view: involved processes are only required when the stakes are sufficiently high. By contrast, on a proceduralist's view, who participates in a fair process determines whether that process is fair. As a result, getting the membership wrong means getting the outcome wrong. And as I have shown, it is very hard to get the membership right.

Finally, a process for making decisions is a necessary component of priority setting: somehow, health actors need to go from not having a decision to having one. The point I have argued—that we should pursue substantive, rather than procedural, justice in setting priorities—does not obviate the need for some kind of decision‐making procedure. But any process should be judged, first and foremost, on whether it tends to produce substantively just decisions, not on whether it includes procedural elements for their own sake. And even a proceduralist who rejects this point might recognize that participants in a fair process still need guidance for deciding how to allocate resources, and that MEC may help guide their deliberations.

It might still be objected that all of this presupposes there is a substantive truth to estimate. An anti‐realist about distributive justice might deny this, but I can grant this point: after all, my argument does not require a determinate truth for every hard case; it merely requires acknowledging that some allocations are better or worse than others. Proceduralists already grant that certain allocations fall outside the range of those that are plausibly just. Once it has been conceded that a process can arrive at the wrong answer, that process is responsible for tracking something substantive beyond itself. The question then becomes how it should reliably do so. I have argued that we will do better by aiming directly at that substance than by attempting to design the fairest possible process.

7. Conclusion

Health care priority‐setting decisions matter because they have significant bearing on people's health and wellbeing; when we make these decisions badly, we cause suffering that we might have avoided, or compound inequities that we could have mitigated. My conclusion is thus a simple one: our primary aim in allocating resources should be to allocate them well, and the processes that we use to set priorities should be designed with this aim in mind. As with all issues in public policy, there is moral uncertainty about what it means to get these decisions right, and empirical uncertainty about which processes work. But given the stakes of these decisions, it would be a mistake to conclude that unjust outcomes should be accepted when certain procedural elements are satisfied. Rather, health actors should strive to get the right resources to the right patients, via processes that allocate resources in accordance with MEC.

Data Availability Statement

The author has nothing to report.