The Intensive Care Lifeboat: a survey of lay attitudes to rationing dilemmas in neonatal intensive care

Willingness to admit patients in the absence of scarcity

Limited resources and allocation

When forced to choose between patients because of limited resources, the majority of respondents chose to direct treatment to the infant with the better predicted outcome or lower cost of treatment (Fig. 2). The only exception was a scenario in which one infant was predicted to survive for 40 years, while the second was predicted to survive for 41 years. Two thirds of respondents elected to toss a coin to decide between the two infants in that scenario.

There was a significantly higher rate of utilitarian responses in scenarios with greater difference in predicted chance of survival, predicted duration of survival, or cost of treatment (Figure 2a, c and d). For example, in relation to predicted chance of survival, there was a significant asymmetry in responses to questions 10 and 9 (Q10 49 vs. 51 %, Q9 10 vs. 20 %, X ² = 24.154, P < 0.000, Cramer’s V = 0.406) and questions 9 and 7 (Q9 10 % vs. 20 %, Q7 20 % vs. 60 %, X ² = 25.000, P < 0.000, Cramer’s V = 0.424). There was also asymmetric distribution of responses for questions related to duration of survival (Fig. 2c) and cost of treatment (Fig. 2d). There were, however, no significant differences in the proportion of utilitarian responses between different degrees of disability (Fig. 2b).

In the lifeboat question, 98 % of respondents (88/90) elected to save five drowning people rather than one person: only 2 % elected to toss a coin to decide.

Personality traits did not appear to influence responses. There was no significant difference between the Empathic Concern scores of those who gave utilitarian responses and those who gave egalitarian responses to key indicator questions. There was also no significant difference between Need for Cognition scores or Social Economic Conservatism scale scores for utilitarian and egalitarian respondents.

Utilitarian responses to scenarios involving scarcity (i.e. where only one intensive care bed was available) were correlated with responses to questions in the absence of scarcity. The greater a respondent’s utilitarian propensity, the greater the effect of cost of treatment (F(29.5, 183.2) = 1.56, p < 0.05), chance of survival (F(18.1, 112.4) = 1.88, p < 0.05) and life expectancy (F(20.2, 125.0) = 1.73, p < 0.05) on inclination to admit the patient (in the non-scarce scenarios). Conversely, the willingness of respondents who were less utilitarian to admit patients remained similar regardless of prognostic information.

Some demographic characteristics influenced responses. Female respondents were significantly more inclined than male respondents to choose the egalitarian option of tossing a coin to determine which patient to admit when the trade-off was between a newborn infant with a mild disability and a newborn infant with no disability (p < 0.01). Parents were significantly more likely than non-parents to give an egalitarian response when asked to choose between treating one newborn infant or two (p = 0.01). Religious respondents were significantly more inclined to choose the utilitarian option of admitting a newborn infant with a 51 % chance of survival, over an infant with a 49 % chance (p = 0.02). Age, marital status, highest level of education, and country of origin had no significant association with allocation preferences.

Implications for ethical debate

Empirical findings like these cannot be used deductively to yield normative conclusions [33]. However, they may contribute to a dynamic process of reflective equilibrium [23]. Our study results provide valuable data on the relationship between the intuitions of the general public and ethical arguments.

The strikingly utilitarian tendency of the general public in this study is consistent with previous studies on healthcare practitioner resource allocation preferences [16, 18, 19, 22], however, contrasts with previous studies of the general public, which have indicated a preference for a more egalitarian approach [35–38].

One difference between this and previous rationing surveys is the requirement for respondents to choose between two individual patients in need of life-saving treatment, rather than choosing between two groups of patients, or different types of treatment. For example, in a survey conducted in Norway and repeated in the US, participants were asked to distribute an increase in health funding between the treatment of two illnesses of different severity [36, 38]. Participants gave priority to patients with more severe illness, even if they would benefit less from treatment. Life-or-death rationing at the cot-side in intensive care may appear closer to an emergency triage situation than prioritizing funding or health policy decisions [39]. It may be that utilitarian intuitions are stronger where the outcome for individual patients is more explicit.

One significant factor influencing NICU resource allocation decisions in our survey was the degree of difference between predicted outcomes in competing patients. This finding was consistent with a previous survey by Ubel et al., which investigated whether the public prioritizes equity or efficiency in distributing scarce organs to children needing a liver transplant [40]. That study found that respondents were less likely to give patients an equal opportunity for receiving treatment where there was a larger difference in prognosis between transplant candidates [35]. In our study, the same trend was seen in relation to three of the variables investigated. In a choice between infants with a 49 and 51 % chance of survival, a bare majority of respondents (52.7 %) chose the utilitarian option of treating the newborn infant with a higher chance of survival. In contrast, where there was a 10 % difference in chance of survival, three quarters of respondents elected to admit the infant with better prognosis.

Our study did not investigate the reasons why participants gave different responses. The largest number of egalitarian responses were seen when the discrepancy between outcomes was small; perhaps the difference in lifespan was considered by respondents to have negligible normative value. Alternatively, it may be that respondents were skeptical about the ability of clinicians to accurately predict the length of survival in adult life. Uncertainty about predictions might support a more egalitarian approach [41]. Finally, our results could be consistent with an ethical approach that balances a number of different ethical principles including equality and utility [40, 42]. Equality might be thought to outweigh small gains in utility, and lead to a different response in marginal cases.

Interestingly, three of the prognostic factors (in the absence of scarcity) showed a significant interaction with respondents’ utilitarian propensity to scenarios where resources were limited. The exception was for information about severity of future disability, which suggests that such information may affect a respondents’ inclination to admit regardless of utilitarian propensity.

Implications for ethical theory

The results of this paper may also be of value for ethical theory. We gave our respondents a version of a much-discussed philosophical example, where they must choose between sending a lifeboat to save five people or one person [30]. Egalitarian philosophers have claimed that in such a situation we ought to toss a coin to decide (though acknowledge that this is counterintuitive) [30]. However, to our knowledge, the views of the general public about this have never been elicited. The overwhelming majority of respondents in our survey chose to send the lifeboat to save the larger number.

Political philosopher John Rawls famously described a procedure for developing fair and just public policy [26]. He imagined a group of hypothetical rational decision-makers who would have to decide how society should be structured without knowing their place in that society – whether they would be rich or poor, healthy or unhealthy, and so on. Rawls’ ‘veil of ignorance’ is designed to overcome prejudice or bias and is often thought to favour those who are worst off. Although Rawls did not apply this decision-making procedure to health care resource allocation, philosophers have taken different views about the sort of allocation policy that would be chosen behind the veil. For example, Singer et al. propose that in a scenario where two people need a life-saving treatment, but only one can receive it, a rational egoist would assume they have equal chance of being either person, and that they would maximize their own chances by directing treatment to the patient with better prognosis [43]. In contrast, Harris suggests that decision-makers behind the veil are likely to be risk-averse and would focus on reducing the chance for the individual at the time of allocation of the worst outcome (death) [44]. Harris argues that this would be accomplished by an egalitarian approach of random allocation (for example, tossing a coin) [44].

There have been few empirical studies of the impact of the veil of ignorance on public views about resource allocation [45]. We asked our respondents to imagine making policy about resource allocation in intensive care that would affect their own children (but without knowing their child’s prognosis). Respondents were predominantly utilitarian in their responses behind the veil. This appears to support Singer et al’s predictions [43]. 89.3 and 77.3 % of respondents chose the infant with a greater chance of survival and infant with a less severe disability, respectively. There were no statistically significant differences between responses to the policy questions and their equivalent versions without the veil. This finding does not resolve the debate between utilitarianism and egalitarianism, but does imply that the veil of ignorance thought experiment would favour the former (at least in situations where patients would be equally badly off without treatment).

Demographic characteristics

There were some associations between demographic characteristics and responses to the survey. Parents were more egalitarian than non-parents in allocating resources based on cost of treatment. This is consistent with a survey of mothers of NICU infants from South Africa, in which the majority rejected rationing of resources entirely [22]. Females were more inclined to be egalitarian than males in relation to allocation of resources based on severity of disability. Religiosity influenced resource allocation preference based on chance of survival. Interestingly, respondents who described themselves as religious were more likely to give a utilitarian response than non-religious people. This finding is in contrast with previous studies, where a high proportion of people who chose egalitarian allocation were religious [46].

There was no relationship between age, marital status, or highest level of education and resource allocation preference. In a previous Australian study of the general public’s resource allocation preferences, all of these three factors were seen to influence treatment decisions [47].

Personality tests

The mean scores of respondents on the Empathic Concern portion of the Interpersonal Reactivity Index, the Need for Cognition scale and the 12-item Social and Economic Conservatism Scale (SECS), had no relationship with their resource allocation inclinations. This might reflect that resource allocation preferences are not influenced by level of empathy or a greater level of deep thinking, but rather depend upon context-dependent moral evaluations.

Strengths and limitations

This study is distinctive in assessing the views of a sample of the general public, rather than healthcare practitioners on resource allocation in the NICU. It provides valuable comparative data on responses to philosophical examples that have been widely discussed, but not previously studied empirically. The results of the survey should be taken cautiously, however. Our small online sample may not be representative of the wider general public. Mechanical Turk samples have been shown to be more diverse than convenience samples of students or the adult population, but less representative than face-to-face population sampling [47].

We were not able to determine why respondents chose particular answers, and responses may reflect uncritical initial responses that might change with further time to consider, or might be sensitive to the way in which cases were presented. Our scenarios specified equal starting points and equal clinical need for infants needing intensive care, therefore did not test prioritarian intuitions. Furthermore, in order to isolate factors that influenced decisions, scenarios were necessarily somewhat unrealistic and compared single variables sequentially. It is possible that a larger sample and factorial survey design would have allowed analysis of the interaction between variables in more complex real-world allocation scenarios [48].