This study examined the views of the UK public on which and how strongly patient features should be considered in ventilator triage due to COVID-19 (and how strongly they should factor in favour or against patients). As a reference point, we compare the views expressed by our participants to the guidance issued by the British Medical Association. The BMA guidance was first published in April 2020, just a month after the COVID-19 outbreak was recognized by the World Health Organization as a pandemic [14]. The BMA guidance was specifically chosen because it was one of the few guidelines that explicitly addresses triage from an ethical perspective [12]. It was also discussed from the legal perspective [13]. The guidance states that when resources are severely constrained, admission to intensive care should be determined by the maximisation of medical benefits, and who should receive intensive care should be largely a function of the factors that will influence the patient’s probability of surviving and recovering from the intensive treatment.
Before we discuss how our data line up with the guideline, we must keep in mind how the timing of the surveys could have affected the opinions of our participants. Our second survey was conducted in November and early December, just between the first and second big wave in the UK. According to the data provided by the British government [11], the United Kingdom was experiencing a downward trend in COVID-19 cases. At this point, much of the anxiety about overwhelmed health systems had somewhat abated, and our participants might have believed that there was ample ICU capacity. The BMA guidance, on the other hand, was published to address direr situations in which medical resources are severely constrained. Even though we asked our participants to answer the survey under the hypothetical scenarios in which there is not enough ICU beds, it is possible that the then current state of affair in the world had implicitly influenced their answers.
With that acknowledged, our results suggest that the views of our participants support, but also diverge from some of the guidelines set by professional communities, such as the British Medical Association. As mentioned earlier, maximising medical benefits to the patient is the guiding principle that underlies the triage guidance set by British Medical Association, and we saw a similar concern expressed by our participants in our surveys, particularly the first open-ended survey. However, in the second survey, the factors rated most strongly in favour of patient selection do not necessarily coincide with BMA guidance. Respondents cited urgency of needing treatment, extreme discomfort due to severity of symptoms, and longer life expectancy. Urgency of needing treatment might coincide with maximizing benefit, since giving a ventilator to a patient who could comfortably wait might lead to preventable death. However, those with the highest medical need and most severe symptoms may also have a very low chance of survival. Longer life expectancy could potentially be justified within the utilitarian framework, assuming a notion of utility that includes years of life gained, but this might penalize older patients.
It is interesting that in the second survey there was relatively low support for the idea that a patient’s low chance of survival should count against allocation, with only 51% responding that this should count against the patient, and 28% responding that it should count in favour, with 21% neutral. One interpretation is that this finding is a result of a clash between different moral intuitions. The disapproving participants could be motivated by the utilitarian intuition that giving the ventilator to a patient with a low chance of survival will lead to lower medical benefits. The approving participants might have interpreted the low chance of survival as reflecting someone who is worse off and therefore should be prioritised.
Age and frailty might seem related to who is worse off or to their value of survival. The British Medical Association, along with other triage guidance documents, advised against penalizing older patients without considering comorbidities and prognosis [2]. Our participants seemed to agree, as only 21% thought being older should count against a patient, compared to 33% when a patient was very frail. This is consistent with a previous finding, in which UK survey participants showed a preference for an older but less frail patient than one who is younger but more fail [4]. In the present study, a considerable percentage (43%) of participants thought that frailty should count in favour of the patient being allocated the ventilator, but these participants might still prefer a less frail patient when forced to choose between two equally old patients but with different degrees of frailty.
Some factors concern effects on others, including foetuses. Even though pregnancy was not explicitly mentioned by the British Medical Association guideline, in view of the finding that COVID-19 infection substantially raised the risks of complication and mortality in pregnant women [9], it is likely that the public perception of pregnancy as morally relevant and positive is in line with expert opinions.
One feature that clearly deviated from the guideline but was found to be morally relevant and favourable by our participants was having several dependents. In Study 2, we asked the participants whether having young children or elderly family members in need of care should count in favour or against of the patient. 57% of participants thought this feature should count in favour of a patient, while 41% were neutral. In Study 1, 35.7% of participants also suggested that having dependents is a morally relevant feature. Some participants might have based this answer on the straightforward utilitarian calculation that the loss of the provider would cause additional losses of utility for their dependents. A compatible interpretation of these results is that our participants perceived childcare and eldercare as essential services. The BMA did acknowledge that once the distribution of resources could no longer be decided based on medical utility, priority should be given to workers in essential services so that social disruption is limited [7]. However, BMA limited essential services to physical utilities such as transportation, electricity, and water. Considering our results, an argument could be made that the notion of essential service ought to include childcare and eldercare, and that those who are responsibility for providing childcare and eldercare ought to receive priority. The persuasiveness of such an argument depends, however, on the extent to which one thinks the public’s views ought to influence policy making.
The British Medical Association appealed to the social good to justify prioritizing essential workers, which includes all hospital workers, regardless of whether they work with COVID-19 patients. This intuition was not shared by as many participants as might have been expected. Only 44% responded positively when the patient is a healthcare professional but does not work with COVID patients, with 53% being neutral. In contrast, 69% of participants appeared to be in favour of prioritizing those who have an active role in helping others with COVID-19. The preference for these workers could be motivated by a utilitarian intuition to maximize expected utility (by lowering the probability of infection of people associated with high degrees of medical utility) or a sense of reciprocity toward workers who have engaged in COVID-related care.
Our findings seem to have interesting implications regarding bias against disabled people, who face unique barriers in seeking medical help during the pandemic [15]. In a previous discrete choice experiment, people had shown small but statistically significant preference to people who are not physically or mentally disabled when deciding on how ventilators should be allocated [16]. In contrast, we did not see a clear indication of a bias against disabled patients in our studies. For example, 43% of survey 2 participants thought that being disabled should count neither in favour or against a patient, followed by those who thought it should count in favour (37%), and finally those who thought it should count against (20%). In our study, then, only a small minority thought disabled patients ought to be deprioritised. Nevertheless, our finding is consistent with the possibility that our participants who responded neutrally in our survey would have decided against disabled patients if they had to choose between a disabled patient and one that was not, so our findings did not completely rule out such bias.
While many of our findings conform to traditional utilitarian principles, they are also consistent with participants following certain non-utilitarian ethical principles. For instance, consider the “fair innings” argument [17, 18] which states that a member of a society is entitled to a certain length of life years, so anyone who has yet to receive this entitlement should be prioritised. The implication is that younger people in general should receive priority over older people when it comes to resource allocation. Participants in study 2 could have been thinking in terms of “fair innings” when they answered that being older should count against a patient (21%). Another non-utilitarian position is the severity approach [19], which states that, other things being equal, priority should be given to those with more severe medical conditions, because they are medically worse off. This approach is consistent with the behaviours of participants who prioritised, for instance, patients who are dying soon (83%) and those who were experiencing extreme discomfort (77%). Some responses were also consistent with the desert-based approach, which prioritise distribution based on their contribution to society [20]. Some participants favoured patients who had made more past contributions (28%) or have more potential to make future contributions (31%).
Another factor that might be related to desert is the circumstances under which the patient contracted the virus. For instance, a considerable number (40%) of participants thought that being responsible for contracting the virus should count against a patient, and 44% expressed similar attitude toward engaging in needlessly risky behaviours. The current access to vaccines, which was not available when data was collected, might increase these numbers further. These responses are especially pertinent in view of the recent phenomenon of hospitals running out of ICU beds due to unvaccinated patients [10]. Our participants’ negative reaction to certain features deviates substantially not only from the official guidelines but also potentially suggest a divergence from a wider utilitarian perspective, if participants believed that patients who willingly and unnecessarily engaged in risky behaviours should be deprioritised solely by virtue of this past behaviour. However, a utilitarian reading could construe those who responded negatively as treating past risky behaviours as a predictor for risk of future problems. If so, these participants were motivated by the lower expected utility of treating these patients.
On the other hand, participants’ attitude toward patients with a history of violent crime also cannot be straightforwardly reconciled with a general utilitarian perspective, since having committed violent crimes in the past does not directly influence the medical utility generated from treating the patient, such as chance of survival and life expectancy. 46% of our participants thought having a history of violent crimes should count against a patient. The numbers of negative responses were higher than either positive or neutral ones (though not together). Also of interest is the gap between the negative responses received for violent crimes and non-violent crimes (46% vs 21%). This gap cannot be explained by medical benefits, at least not in terms of direct medical benefits derived from treating these patients.
The divergence between the public view and the guideline raises an important question: What, if anything, do these popular opinions tell us about who should get a ventilator? We do not suggest that we should draw normative inferences directly from these surveys. The fact that many people found certain features to be morally relevant and positive does not prove that it is ethical to use that feature in allocation. Nevertheless, our findings become normatively and practically relevant on the assumption that hospital policies should conform somewhat to the moral standards of the public. Such an assumption could be argued on the ground that publicly funded hospitals, for instance, should be responsive to the values of the community. Separately, a degree of conformity between expert guidelines and moral views of the public is needed in order to maintain trust in the medical community. For such reasons, public opinion may be considered in designing and assessing public policies.
Limitations
While Study 1 reduced framing bias by having participants spontaneously determine features that they believed would be relevant to allocation, this method could potentially miss relevant features that did not spontaneously come to participants’ minds, but that they nonetheless think are important. Study 2 found evidence of such omissions when features, such as pregnancy, were perceived as relevant and favourable, although they had been mentioned relatively infrequently in Study 1. Future studies should try to include more potentially relevant features that participants might overlook unless prompted.
Second, our study did not incorporate certain features that might be seen as relevant, including the hypothetical patient’s vaccination status (at the time of the initial survey, vaccines were not yet available). As vaccines become increasingly available and vaccination hesitancy becomes a core issue, vaccine status should be included in future studies.
Third, we only recruited opinions from a lay population. A potential future study could directly compare medical experts and the lay public in a two-sample study.
Fourth, while we suggested various underlying moral intuitions as hypotheses to explain the behaviours of the participants, these are not and cannot be substantiated by the current results, as the survey were designed to probe their views on resource allocation, and not the underlying motivations or justifications.
Fifth, our study did not involve a deliberative process, with discussion and information provided about the arguments for and against treating a feature as relevant. If such a process were presented to participants beforehand, they would potentially give different responses.
Conclusions
Survey respondents perceived patients more favourably in hypothetical ventilator triage situations when the patients will die soon, are pregnant, have waited for a long time, are currently experiencing extreme discomfort, are expected to live longer, are COVID health workers, have dependents, wish to be put on ventilator, and are already on the only available ventilator. However, a considerable portion of the participants was also willing to potentially go beyond general utilitarian concerns by prioritizing ventilator allocation according to circumstances in which the patient contracted the virus, such as when the patient contracted the virus by needlessly engaging in activities with a high risk of COVID-19, and whether or not they have committed crime in the past. These results can be interpreted as illustrating that members of the UK public appear to broadly support the existing UK triage guideline in its focus on medical features and imperative to avoid discrimination based non-medical features, but also thought additional factors should be taken into account. It will be important for ethicists and policy makers to evaluate whether and how these additional features should be addressed in future iterations of triage policy.