The coronavirus disease 2019 (COVID-19) pandemic has repeatedly confronted health care providers with scarcity. Shortages of personal protective equipment have forced hospitals to prioritize which health care workers will be protected. Expected shortages of ventilators for patients with severe acute respiratory distress elicited widespread debate over the triage protocols by which they would be allocated.
The United States was forced to ration supplies of the first drug that was shown effective for treatment of COVID-19—remdesivir—and will likely face shortages of others once they are available outside of clinical trials. And it is widely acknowledged that once a safe, effective vaccine is developed, it will be months before enough doses can be manufactured to immunize everyone in the United States, let alone the world.
Some shortages were and are preventable. Others are not, or cannot be avoided right now by the states and health care institutions that must decide how to treat sick patients. How should these allocation decisions be made? A recent model hospital policy from a team at the University of Pittsburgh (the “Pittsburgh policy”) proposes a weighted lottery system among eligible patients when medications are scarce. There is discussion of extending this system to scarce COVID-19 vaccines, as well.
Weighted lotteries give the appearance of balancing competing ethical considerations. But, as I will show, they actually lead to unjust outcomes.
Priority Setting And The Pittsburgh Policy
In discussions of health care priority setting, including in the context of the current pandemic, two general values have broad support. First, we should allocate scarce resources in ways that produce more, rather than fewer, benefits. Second, we should allocate them in ways that reduce, rather than exacerbate inequalities. The Pittsburgh policy endorses both values. Its first two goals are:
- To safeguard the public’s health by allocating scarce treatments to maximize community benefit.
- To lessen the impact of social inequities on COVID-19 outcomes in disadvantaged communities.
The policy operationalizes these goals by giving higher priority to essential workers (defined by states) and patients from disadvantaged communities (according to Area Deprivation Index scores). It gives lower priority to patients who are expected to die within a year from an underlying end-stage condition. In addition, “If clear evidence emerges that certain clinical subgroups derive larger benefits from the treatment than others (e.g. a lower number needed to treat to save a life), these groups should receive priority.” This would be another way to maximize community benefit.
Suppose we agreed with these goals and with operationalizing them through these three (or four, if preference is given to subgroups that disproportionately benefit) characteristics. One might think that the natural next step would be to develop a scoring metric to assign points to eligible patients: more points to those in higher priority groups and fewer points to those in lower priority groups. Patients would be treated according to their score until supplies of treatment were exhausted. First, a hospital would treat all essential workers from disadvantaged communities without an underlying end-stage condition. Then, if supplies remained, the group with the next highest score—perhaps other essential workers without underlying end-stage conditions. This would continue until all eligible patients had received treatment or no medication remained. A simple lottery could be used to choose among patients with the same score, if not all could be treated. Such a system would best achieve the ethical goals of the policy.
This is not how the Pittsburgh policy works. Instead, all eligible patients are put into a lottery, where all have some chance of receiving treatment. Their chances differ depending on their characteristics—this is how the lottery is “weighted.” Someone who is expected to die within a year has half the chance of a “general community member” (someone with none of the characteristics that affect priority). An essential worker from a disadvantaged community has a 50 percent higher chance than a general community member. Were evidence to emerge that treatment is more likely to save the lives of patients in some clinical subgroups than others, those subgroups would be given a higher probability of winning the lottery.
What are the implications of a weighted lottery? In the context of a scarce, potentially life-saving treatment, it will mean that some essential workers from disadvantaged communities will die so that some patients who would anyway die soon from an underlying end-stage condition can be treated. If some clinical subgroups are known to respond better to treatment than others, the weighted lottery will save fewer lives—among otherwise similar individuals—than if treatment were to go first to patients most likely to benefit. In other words, a weighted lottery is worse at maximizing benefit and reducing inequality than directly prioritizing patients according to those goals. This is because it gives an extra chance to those who are expected to benefit less and those who are more privileged to get treated in place of a patient higher on the priority ranking.
Objection: The Value of Chances
It might be responded that a weighted lottery gives every eligible patient a chance of benefit and that having a chance is valuable. But we must consider what is lost. When a patient with a low likelihood of benefit wins the lottery, a patient with a higher likelihood of benefit goes untreated. When a patient from a socially privileged group—someone who has already had more chances in life—wins the lottery, a patient from a disadvantaged group goes untreated. These are worse outcomes according to the very ethical goals stated in the Pittsburgh policy. Those who promote weighted lotteries should, at least, explain why they accept such sub-optimal outcomes in exchange for giving everyone a chance at treatment. Without a compelling justification, the policy should be regarded as unjust.
Suppose that the defenders of weighted lotteries convinced us that everyone ought to have some chance. Every eligible patient could be given a chance while still making the outcomes more just by simply changing the odds. Instead of giving essential workers from disadvantaged communities a 50 percent higher chance, they could have three times the chance, or four times the chance. This would increase total benefits and reduce inequality more than the Pittsburgh policy. Every eligible patient would still have some chance. Supporters of weighted lotteries also owe an explanation of why they assigned the numbers they chose given the apparently unjust outcomes to which they will lead.
Objection: Lotteries As Experiments
Douglas White and Derek Angus note another advantage of allocating using lotteries. A lottery, “creates a natural experiment that achieves random allocation in which some patients receive the drug while others do not; researchers can use the lottery’s registry to assess the effectiveness of the scarce drug.” If the drug is not allocated by some type of lottery then the data gathered from patient outcomes may be less informative than it could have been. Could this justify the weighted lottery?
In clinical research, participants are frequently asked to take on risks—from medical procedures used solely to gather data or from experimental drugs—in order to generate knowledge that will ultimately benefit society. Foregoing an effective drug for which one would otherwise be prioritized can be understood as another research risk. However, there are ethical limits to how risky research participation can be. In the context of a deadly pandemic, it is a lot to ask of patients that they forgo treatment for which they would otherwise be prioritized so that scientists can conduct more rigorous research. Why, they might reasonably ask, should we be the ones making that sacrifice? Moreover, researchers are normally required to ask competent adults for their consent to be in research. Here, highly risky research would be conducted without consent. The collection of research data cannot be an independent justification for allocating scarce treatments using weighted lotteries.
When COVID -19 treatments are scarce not everyone who could benefit from treatment can receive it. States and hospitals must implement fair procedures for allocating limited resources. Weighted lotteries take into account ethically relevant differences between patients who are eligible for treatment. Patients who are expected to benefit more and patients who are more disadvantaged get a higher probability of receiving treatment.
But weighted lotteries do not maximize benefits or optimally reduce inequality. If those are our goals—as the Pittsburgh policy suggests they should be—then weighted lotteries do not serve them well. They should not be used.
The views expressed are the authors’ own. They do not represent the position or policy of the National Institutes of Health, the U.S. Public Health Service, or the Department of Health and Human Services.