A number of countries have faced (or are facing) the regional triage question in reality. For instance, in the UK, a report in the Health Service Journal suggests that NHS trust chiefs were told in March 2020 by NHS England that additional ventilators will be deployed to “areas with the most immediate need, on a fair share basis relative to patient ventilation need” [5]. If ventilators were deployed to areas in the UK with the most immediate need, then there would likely be significant inter-regional inequality with regards to both the number of additional ventilators allocated to each region, and their post-allocation critical care capacity. Indeed, the report suggested that London would likely be prioritised in the first wave under the apparent NHS England approach, and that some medical professionals had thus called for the adoption of an alternative per capita approach to allocation [5].
To simplify our investigation of these principles, and because of limited publicly available data, we shall return to our hypothetical example to guide our philosophical assessment of potential regional distribution principles.
Egalitarianism
In the individual triage question, the most straightforward egalitarian procedure is to allocate on a first come first serve basis or via a lottery; applying these procedures can mean that any patient has an equal chance of receiving care. The first thing to note in the regional triage question is that a first come first served approach will very seldom be functionally equivalent to a lottery. Allocating on the former basis is only equivalent to a lottery on the assumption that each region has an equal chance of ‘being first in the queue’, of needing the scarce resource before others.Footnote 10 However, it is unlikely that different regions will all have an equal chance of being ‘first in the queue’ in this sense, given demographic variation between regions and different levels of population density.
With a lottery allocation, we could expect each region to receive the same number of additional ventilators. Although a lottery would ensure that each region had an equal chance of receiving additional ventilators, it not clear that this is the sort of equality that should really matter for the egalitarian. What should really matter for egalitarians is rather whether individuals in different regions have equal access to care; does an individual in region A have the same chance of accessing a ventilator as an individual in region B? However, if that is the case, then egalitarianism lends greater support to a regional allocation of ventilators on a per capita basis (rather than a lottery), as illustrated in Table 2.
Table 2 Simple egalitarian allocation on pure per capita basis Yet, even this simple per capita approach does not capture all that should matter for the egalitarian; it fails to acknowledge that regions do not have equal existing per capita critical care capacity. Accordingly, if the aim of additional ventilator allocation is to (solely) ensure that allocation enables equal treatment access across regions, then allocation should be performed on the basis of the projected per capita demand of each region, in excess of their current capacity. As the table below illustrates, this approach generates a quite different prioritisation of regions (Table 3).
Table 3 Modified egalitarian allocation on projected excess per capita basis We suggest that allocating on the basis of projected per capita excess demand would constitute the most plausible egalitarian approach to allocation. To illustrate how different approaches may roughly play out on the level of individual cities, estimates suggest that Bristol and Liverpool in the UK have a comparable population size (roughly between 450'000 and 500,000) [2, 16], and roughly comparable levels of adult critical care capacity (according to NHS statistics from January 2020) [19].Footnote 11 Yet, at the start of May 2020 (when this paper was initially written), Liverpool had more than twice the number of cases of COVID-19 as Bristol (292 vs 125 cases per 100,000) [6]. Suppose we had to decide where to allocate additional ventilators on this data. The simple prospective egalitarian allocation based on a purely per capita basis would send equal numbers of ventilators to Bristol and Liverpool. We suggest that this would have been a profound mistake. The data outlined above suggests that the latter might have twice the excess per capita demand of the former. A more plausible modified egalitarian allocation at this point would thus potentially send more ventilators to Liverpool.
However, this answer, too, might be problematic. Depending on other variables that we analyse below, such an allocation may mean that fewer lives are saved.
The Save The Most Lives Principle
Perhaps paradoxically, a policy of prioritising immediate need in the regional distribution of additional ventilators may be one way of operationalising the save the most lives principle. Prima facie, such a policy might appear to be a version of the egalitarian ‘first-come first served’ approach; it may entail allocating extra ventilators first to those areas of the country affected most severely earliest in the pandemic—potentially at the expense of those affected later. We shall now explain why regional prioritisation according to immediate need may be at least partly justified, but that we must also factor in other considerations in order to fully operationalise the save the most lives principle.
For individual triage, we noted that priority may be given to individual patients based on the RAPRi. This essentially helps us to estimate how much good a ventilator is likely to achieve for an individual patient in the numerator (assessed in terms of that individual’s probability of survival) and how efficiently it is likely to achieve that good in the denominator (assessed in terms of that patient’s relative expected resource demand). How can we similarly inform decisions about how to save the greatest number of lives via the regional distribution of additional ventilators?
The Numerator: How much good will additional ventilators achieve?
Consider first how much good additional ventilators will achieve. If our main concern is to save the most lives possible, then the first relevant assessment of regional need should concern the raw number of projected regional cases. If we are interested in maximizing the number of lives we save, then a region that is expected to have a higher number of individuals requiring ventilation has a greater need than a region with a lower number, even if that lower number nonetheless constitutes a higher proportion of its population.Footnote 12 In turn, regional projected need can be modelled on the basis of population variables and the number of confirmed cases.Footnote 13
However, the first point of complication here is that assessing a region’s demand is a more complex matter than assessing a presenting patient’s probability of survival in the individual triage question. Such assessments will often make projections about how the public health emergency may progress, and such projections must unavoidably rely on various assumptions.
This complication suggests one virtue of prioritising areas with the most immediate need. In our discussion of the individual triage question, we noted that the principle of temporal neutrality states that the time at which a harm occurs should not make a moral difference, other things being equal. However, this principle is compatible with the claim that we can have a greater degree of certainty at different points of time about whether or not a harm will occur. Moreover, this greater degree of certainty can make a moral difference about what we ought to do. Indeed, it may be that there is a predictable difference in projected need between regions that reach their ventilator capacity earlier or later in the course of the pandemic. If lockdown measures are effective in helping to ‘flatten the curve’, then those regions that are hit hard at an earlier stage of the outbreak may end up having a higher relative surge than those regions that are affected at a later stage in the outbreak. If that is right, we may save more lives by providing additional ventilators to those hospitals affected early, compared to those affected later.
The strength of our moral reason to prevent a harm is a function of both the magnitude of the harm, and its probability. Accordingly, for the reasons outlined above, we will often have stronger moral reasons to provide additional ventilators to a region that is further along its infection trajectory than the modelled projection of regions at earlier stages. Prioritising most immediate need thus need not violate the principle of temporal neutrality.
However, in considering the good that additional ventilators can do at a regional level, a further important question is whether there might be differences in regional survival rates. In our example, suppose that Region A has a lower demand for critical care than rural area Region C because the former has a younger population. However, the same factor might also mean that patients admitted to intensive care in Region A would have a higher chance of survival than those admitted to intensive care in the Region C. Imagine, for example, that 100 additional ventilators were made available for 1 week, enabling 100 extra patients to be admitted to intensive care. If those ventilators were provided to a region where the survival rate from intensive care was 50%, the allocation would result in 50 additional survivors. However, if those ventilators were provided to a different region where the patients were older/had a higher rate of co-morbidity, (and an average survival rate of 40%), only 40 lives would be saved.
That said, there are a number of difficulties with using assessments of regional survival rates to determine distribution. One difficulty is epistemic: it may not be possible to know what the relative survival rates are for different regions until it is too late to take that into account.
A second difficulty is more challenging—it is the relationship between demand, decision-making and outcome. For example, consider region X (Openshire) that decides not to restrict intensive care admissions, but attempts to provide intensive care to every patient with respiratory failure who would benefit, without any attempt to triage. Compare X with region Y (Closedbury), which anticipates a severe surge in demand, and decides to restrict admissions to intensive care to those patients with the highest chance of survival. Because Openshire has a less restrictive admission policy it will reach its capacity in terms of ventilators much sooner than Closedbury. It will also have a lower survival rate than Closedbury, since Openshire admitted patients who would have been excluded from intensive care in the other region. In this situation, it would appear to be problematic to provide extra ventilators preferentially to Closedbury, even if on paper they appear to have a higher survival rate.Footnote 14
A third problem is more difficult still. It appears that there are differences in the rates of severe illness from COVID-19 between different ethnic groups [13, 21]. If that is correct, it might contribute to regional differences in survival rates following admission to intensive care. However, many will feel disquieted by the suggestion that this should be factored into a decision about regional ventilator allocation, even if it would mean saving more lives. Furthermore, data from the Office of National Statistics in the UK also suggests that COVID-19 is having a proportionally higher impact in more deprived areas [20]. Accordingly, underlying socio-economic inequalities may plausibly influence differences in regional survival rates. This potentially adds further issues regarding individual responsibility for illness in lifestyle diseases, like diabetes and hypertension. In any case, it may be doubly unjust to deprive those who already suffer from structural disadvantages from receiving ventilation.
There is one final factor that will significantly influence the good that additional ventilators will do in a given region. Ventilators are complex pieces of medical equipment that require trained staff; additional ventilators will not save more lives if they are sent to hospitals that do not have enough staff to operate them effectively. Accordingly, assessment of the good that additional ventilators will do must take into account the extent to which regions will be able to staff them.
The Denominator: How Efficient Would It Be To Send A Ventilator To A Given Region?
In the individual triage question, efficiency is largely a question of the amount of time that a particular patient will use a ventilator. The less time that the resource is needed for that patient, the more additional lives the resource can be used to save. It is less clear how that might be understood in terms of regional allocation. One possibility is that regions that have a shorter average intensive care length of stay might be prioritised over those with a longer length of stay. Providing 100 ventilators for a month to a region that has an average length of stay of 2 weeks (and 50% mortality) would save 100 lives. Providing the same number of ventilators to a region with an average length of stay of 1 week (and 50% mortality) could save 200 lives.
However, given the potential for regional differences in the time of peak ventilator demand, the efficiency of regional distribution may also be affected by the timing of sending additional ventilators to particular regions. In the first wave, Edge Health, an independent agency which advises NHS Trusts in the UK, developed a ‘pressure index’ which aimed to identify regions that are facing particularly severe pressure at a given time, calculated on the basis of reported deaths after controlling for demographics and pre-COVID-19 critical care capacity [7]. Such data could be used to form an estimate for the proportion of each region’s current critical care need that it would be able to address at a given time. Since it is likely that ventilators will become available progressively over time, it may be possible to allocate a tranche of new ventilators now to meet the needs of regions that are already reaching ventilator capacity, and then distribute a later tranche based on prevailing conditions when further ventilators become available (Table 4).
Table 4 Key considerations for the regional application of the ‘save the most lives principle One interesting question this raises is whether ventilators should be re-distributed. For example, suppose that the virus surges first in region C; 500 ventilators are made available to region C, and it is soon able to meet patient needs. Suppose now that there was then a surge in need in region D; should those ventilators be taken from region C and re-allocated? It would be hard to see why that shouldn’t occur, assuming the ventilators are not in use. If they were currently being used by patients, then this strategy of re-allocation would raise similar questions to the ethics of withdrawing treatment in the individual triage question [3, 17]. Some might feel that only spare ventilators should be redistributed to another region. However, if patients currently being treated in region C have a significantly worse prognosis than patients in region D (who will miss out if the ventilators are not redistributed), there is an ethical argument for mass reallocation. An alternative strategy in this scenario would be to redistribute patients to available supply by transferring them to hospitals with available capacity. Indeed, this has occurred in the international context, with Germany accepting patients from both Italy and France into their hospitals [1].
Operationalising The Save The Most Lives Principle For Regional Allocation Questions
The above considerations suggest that we should operationalise the save the most lives principle for regional allocation questions as follows. First, we must model the projected demand of each region in excess of existing capacity; doing so allows us to rank regions in terms of how many lives will likely be lost if critical care capacity does not increase. Recall that although an egalitarian approach would support a per capita assessment of need (which would adjust this figure for regional population), the save the most lives principle does not support such an approach.
In our example of country X, allocation purely on the basis of projected demand in excess of existing capacity would yield the following ranking of regional priority (Table 5).
Table 5 Allocation on the basis of Projected Excess Demand Allocating in accordance with the projected demand of each region in excess of existing capacity would be the most straightforward way to operationalise the save the most lives principle. However, it would also be incomplete, as it would fail to accommodate a number of important factors that will influence whether sending additional ventilators to a particular region will save the most lives. Rather than adjusting excess peak demand for population (as an egalitarian per capita approach requires), our discussion above suggests that the excess peak demand could instead be adjusted for survival rates, and efficiency of distribution.
To adjust for survival rates, one could calculate each region’s Adjusted Survival Rate (ASR) as follows:Footnote 15
$$Total\;survivals/Total\;number\;of\;eligible\;patients\;treated = ASR$$
The ASR would aim to adjust for different admission criteria by calculating survival for a common cohort of patients—for example survival probability of patients aged 60–70 without severe co-morbidities.
Each region’s project peak excess demand could then be multiplied by its ASR to provide an estimate of how much good additional ventilators will do in that region. However, incorporating the regional ASR into this assessment introduces a number of controversial factors. Some regions may have lower ASRs as a result of having a large number of patients who are worse off, (for example, because they are more vulnerable to severe disease due to socio-economic deprivation or beacuse they are members of an at risk ethnic minority group). Others may have less effective critical care units, potentially as a result of past unjust allocation decisions.
The extent to which the regional ASR ought to feature in our assessment of how much good additional ventilators will do depends on whether we think it is acceptable for such inequalities to influence resource allocation decisions. We turn to this question in the next section. To conclude this part of the discussion though, we note that in order to operationalise the save the most lives principle in a comprehensive manner, we should also accommodate considerations of efficiency. Accordingly, excess projected need (potentially adjusted for regional ASR) could also be adjusted by each region’s average length of intensive care stay, and the current ICU pressure in the region. The former can be understood in terms of the length of the average ICU stay in the region, divided by the national average; the lower the figure, the more efficient the ICU.Footnote 16 The current ICU pressure in a region can be understood in terms of the proportion of the region’s current critical care need that can be addressed with existing capacity; the lower the proportion, the greater the ICU pressure. This yields the following rough formula that could be used to comprehensively operationalise the save the most lives principle in the regional triage question into a Regional Allocation Index (RAI):
Regional Allocation Index
$$\frac{{\left[ {{\mathbf{Expected}}\;{\mathbf{Benefit}}} \right] = {\text{ Projected}}\;{\text{excess}}\;{\text{demand}} \times {\text{ASR}}}}{{\left[ {{\mathbf{Expected}} \, {\mathbf{Efficiency}}} \right] = \, \left[ {{\text{average}}\;{\text{length}}\;{\text{of}}\;{\text{intensive}}\;{\text{care}}\;{\text{stay}} \times {\text{current}}\;{\text{ICU}}\;{\text{pressure}}} \right]}}$$
Finally, if distribution on the basis of the RAI is to ensure that allocation saves the most lives possible, the number of ventilators that we allocate to a region on the basis of its RAI would be subject to a limit determined by the region’s capacity to safely staff additional beds.
The RAI formula could be used to determine which regional distribution of an additional life-saving resource would save the most lives; the higher a region’s RAI, the more lives that will be saved by sending additional ventilators to that region. In principle, the formula could also be applied to support the redistribution of critical care resources that were in place prior to an emergency. In some cases, saving the most lives may require the redistribution of existing resources, and not merely the allocation of additional resources.
Of course, it might be argued that the save the most lives principle is not all that matters here, and that other moral reasons would speak against a redistributive policy. We shall now consider the weight the save the most lives principle should bear compared to other principles in the regional triage question.