Article Text

Download PDFPDF

End-of-life decisions as bedside rationing. An ethical analysis of life support restrictions in an Indian neonatal unit
  1. I Miljeteig1,
  2. K A Johansson1,
  3. S A Sayeed2,
  4. O F Norheim1
  1. 1Global Health; Ethics, Economics and Culture, Department of Public Health and Primary Health Care and Centre for International Health, University of Bergen, Bergen, Norway
  2. 2Division of Medical Ethics, Harvard Medical School, Program in Ethics and Health, Harvard University, Division of Newborn Medicine, Children's Hospital Boston, Boston, Massachusetts, USA
  1. Correspondence to Ingrid Miljeteig, Global Health: Ethics, Economics and Culture, Department of Public Health and Primary Health Care, University of Bergen, Kalfarveien 31, 5020 Bergen, Norway; ingrid.miljeteig{at}


Introduction Hundreds of thousands of premature neonates born in low-income countries are implicitly denied treatment each year. Studies from India show that treatment is rationed even for neonates born at 32 gestational age weeks (GAW), and multiple external factors influence treatment decisions. Is withholding of life-saving treatment for children born between 28 and 32 GAW acceptable from an ethical perspective?

Method A seven-step impartial ethical analysis, including outcome analysis of four accepted priority criteria: severity of disease, treatment effect, cost effectiveness and evidence for neonates born at 28 and 32 GAW.

Results The ethical analysis sketches out two possibilities: (a) It is not ethically permissible to limit treatment to neonates below 32 GAW when assigning high weight to health maximisation and overall health equality. Neonates below 32 GAW score high on severity of disease and efficiency and cost-effectiveness of treatment if one gives full weight to early years of a newborn life. It is in the child's best interest to be treated. (b) It can be considered ethically permissible if high weight is assigned to reducing inequality of welfare and maximising overall welfare and/or not granting full weight to early years of newborns is considered acceptable. From an equity-motivated health and welfare perspective, we would not accept (b), as it relies on accepting the lack of proper welfare policies for the poor and disabled in India.

Conclusion Explicit priority processes in India for financing neonatal care are needed. If premature neonates are perceived as worth less than other patient groups, the reasons should be explored among a broad range of stakeholders.

  • Ethics
  • infant
  • newborn
  • end-of-life
  • priority setting
  • India
  • allocation of healthcare resources
  • healthcare economics
  • newborns and minors
  • prolongation of life and euthanasia
  • cultural pluralism

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Clinicians working under conditions of resource scarcity are often required to prioritise which patients will receive limited available medical services. When beneficial services are withheld from a patient to promote health or welfare interests of others, it is commonly understood to be a form of rationing (definition modified from Ubel1). In many countries, this practice is often implicit and ad hoc.2 One striking example involves the care of sick neonates in India. Approximately 850 000 infants among the 26 million born yearly in India are born before gestational age week 32,3 i and almost all need more than basic parental care to survive.4 For these babies, the WHO manual for treatment of newborns recommends urgent transfer from home deliveries or small health facilities to hospital.5 In order to survive, many premature babies need intensive care that includes close monitoring for irregular breathing patterns, respiratory support for surfactant deficiency and antibiotics to prevent or treat common bacterial infections. India currently lacks the capacity to provide such comprehensive services to all neonates in need.6Currently, only 2000 such beds are in place for an estimated 70 000 potential annual patients.7 ii

In previous studies of Indian neonatal medical units, we documented that doctors operate under conditions of severe resource scarcity8 and that they employ a variety of non-standardised criteria for choosing which neonates will or will not receive treatment.9 In one of our studies, we found that if a neonate was born before 28 gestational age weeks (GAW), the child almost never received treatment. In the few cases of observed treatment, the family was willing to pay the whole bill themselves and was made aware and seemed prepared to assume the risks of long-term child disability. Conversely, those neonates born after week 32 typically received initial treatment regardless of a family's ability to pay for services, partly because providers perceived the risk of long-term disability as much lower.

Between 28 and 32 weeks, we observed that decisions were approached on a case-by-case basis; we found that multiple external factors beyond the immediate clinical welfare of the newborn influenced treatment decisions in this roughly 4-week gestational window.9 For example, health providers felt a strong responsibility to avoid burdening already disenfranchised mothers and families. They sought to reduce the risk of introducing permanently disabled children into poor communities with essentially no social security or support services. They also actively sought to prohibit sex discrimination—however, they frequently acceded to family wishes on the matter, having little legal alternative in terms of the availability of protective custodial services for abandoned children. At least in one specific Indian context, neonates were seen less as individual persons with distinct moral and legal rights and more as potential additions to a complex and established family structure.9

To date, we are unaware of a systematic evaluation of the bedside rationing practice we observed in India with regard to neonates in need of intensive care services. Using our empirical data as a starting point for analysis, we ask how withholding of life-saving treatment for children born between weeks 28 and 32 might be ethically justified under conditions of significant resource scarcity. Importantly, the purpose of this paper is not to compare the utility of alternative policy strategies to reduce overall neonatal mortality in India. That is, our aim is not to assess the most cost-effective approach to dealing with the problem of premature births (eg, by considering the impact of improved antenatal services) but rather to highlight and systematise the ethical considerations at stake in India today given that clinicians are daily engaged in priority-setting.


We use a modified version of Kymlicka's ethical case analysis as a methodological framework (box 1). This pragmatic analysis was first developed to analyse new reproductive technologies and does not presuppose consensus on any particular ethical theory.10 The modified version consists of seven steps for the identification of key ethical concerns in dilemmas of this type.

Box 1 Seven steps in ethical case analysis

Ethical case analysis

  1. Statement of the problem and alternative actions/rules

  2. What is the evidence concerning the outcomes of the different alternatives?

  3. Are there guidelines or legal acts that regulate the issue at hand?

  4. Who are the affected parties?

  5. What are the benefits and burdens for the affected parties?

  6. Are substantial interests in conflict?

  7. Are fundamental principles in conflict?

Ethical analysis of the case

Statement of the problem and alternative actions/rules

Is it permissible to withhold or withdraw treatment of neonates >28 and <32 gestational age weeks?

  1. No, it is never permissible to withhold or withdraw treatment of neonates >28 and <32 gestational age weeks.

  2. Yes, it is sometimes permissible to withhold or withdraw treatment of neonates >28 and <32 gestational age weeks.

What is the evidence concerning the outcomes of the different alternatives?

When evaluating the outcomes, we use four criteria that are discussed in the literature on rationing and regarding which we assume that fair-minded people will agree they are of central relevance. They are: the severity of disease if untreated, effectiveness of the intervention, cost effectiveness and strength of evidence concerning the first three criteria.11

In table 1, we compare infants born between GA 28 and 32 weeks and list: (1) the cost and cost–effectiveness ratios of available treatments, (2) the expected individual benefit from available treatment in India. Data were estimated by a decision tree developed with the software Treeage. The decision tree, data sources and estimates are available in the Appendix. Our cost–effectiveness ratios are comparable and relatively similar to findings in a health economic evaluation of the 15 most cost-effective interventions to reduce the neonatal mortality in low-income countries by Darmstadt et al: for emergency neonatal care for all GA groups, the estimate was 17.59 USD/disability-adjusted life year (DALY) with 95% coverage and 39.37 USD/DALY with 50% coverage.3 This is within our disaggregated range of 12–73 USD/quality-adjusted life year (QALY). The threefold use of resources for a neonate at 28 GAW compared with one at 32 weeks is confirmed in studies from Western countries.12–14

Table 1

Authors' estimated outcomes based on Indian data for neonates admitted to NICU <32 weeks gestational age (GA)*

Estimates regarding severity of disease refer to the prognosis without treatment in the two GA groups. We could not find any studies on this outcome and rely on expert opinions from independent neonatologists, three working in India and three working in the US/Norway to arrive at the estimates (table 2). We report effectiveness as survival rate from live births in hospitals that provide neonatal intensive care until discharge and the 5-year survival rate from live births at a hospital with neonatal intensive care units (NICU). We estimate the mean effect per child that is admitted to hospital with NICU in QALYs.

Table 2

Estimates of disease severity, effectiveness, cost effectiveness and strength of evidence

Are there guidelines or legal acts that regulate the issue at hand?

Legal aspects

In India, there is a federal Child Protection Act,15 but its practical effect is negligible as local systems of enforcement remain lacking. Euthanasia is illegal.16


There is no formal lower limit for treatment either nationally or in the institution studied. At the National Neonatal Forum in India, the international resuscitation guidelines are regularly presented, but these have not been locally tailored to the Indian context.17 There are no formal Indian paediatric professional guidelines available to inform the decision-making process. Also, there are no ethical guidelines that attempt to assist providers in handling clinical resource allocation problems due to scarcity.

Who are the affected parties?

Our list is not exhaustive, but we view the most important affected parties as: the neonate born between 28 and 32 weeks, the parents, the extended family, other children in the unit, other patients in the hospital, other children in the community, doctors, nurses, other providers in the hospital, the local community and society at large, hospital funders and international donors.

Burdens and benefits for affected parties

We first consider burdens and benefits under alternative A: “It is never permissible to withhold or withdraw treatment of neonates >28 and <32 GAW.” For the sake of simplicity, we explore the case only when the child is 28 weeks and assume that if it is never permissible to withhold treatment for children born at 28 GAW, it cannot be permissible at older ages.

Table 3 gives an overview of the benefits and burdens alternative A would impose on affected parties. Our estimates indicate that the child has a relatively high chance of living a life without premature death and disability. Calculating the relative benefit of child survival is ethically and sociologically complex if one is willing to entertain the moral assumption that an individual human being's life may not be of inestimable value.18 We make a common-sense assumption that for most Indian families, any child survivor is given special, if not entirely equal, moral import as other members of an immediate family. On the other hand, we also acknowledge that many children with serious functional disabilities experience substantial individual and social neglect in Indian society, and this neglect is pervasive and insidious.19 As such, it is possible that for some Indian families, children with significant function disabilities may be regarded more as burdens on a family, particularly at early stages of development. Almost certain to exacerbate this is the fact that poor and rural families have essentially no access to already extremely limited early intervention (eg, developmental care), rehabilitation (eg, physical therapy) or chronic care services (eg, for chronic lung disease).20

Table 3

Affected parties' perceived benefits and burdens of treating children at 28 GAW

A more tangible burden in India is the financial cost of healthcare services, where 82% of health expenses are out of pocket and the total costs of treatment of a premature neonate in hospital extends beyond most families' paltry savings.21 Studies show that among families who have a member admitted to hospital, about 40% have to borrow money or acquire debt, most often with extremely high interest rates.22 This can lead to future familial economic burdens and obviously increases the risk of falling further into poverty.23 In the worst case, indebtedness can push families into acute nutritional crises and plays a causal role in the death of siblings or other family members.24 If the child survives with a disability and needs special care, this will bring extra expenses in terms of hospital visits, medication and loss of income for the person taking care of the child.25

Sex discrimination also remains a significant issue in India.26 Here, we make a controversial assumption and treat sex as a potential burden, given the widely documented, unfortunate socio-cultural reality still experienced by so many families. For example, in families that adhere to certain traditional cultural norms and pay no attention to unenforced laws, girls command dowries at the time of marriage.27 Additionally, mothers can be blamed for the sex of the child and for the increased costs if the child needs expensive care after birth and, as such, are in especially vulnerable social positions. Tragically, in such situations, stigmatising conditions that have remained resistant to sociological correction can increase the risk of home eviction to mother–child dyads.20

Other patients in the unit, as well as the hospital and the community, are also affected by the decision. We perceive no direct benefit to them from treating neonates born at 28 weeks. However, under conditions of resource scarcity where financial, material and human resources are pulled away from possible healthcare interventions for other in-hospital patient populations, potentially avoidable burdens are introduced.28

Are substantial interests in conflict?

Among the potential parties, we perceive that the child, the family and other hospital patients who might benefit from different resource allocations are the stakeholders whose interests are most directly affected. If it is never permissible to withhold or withdraw treatment of neonates at 28 GAW, the following legitimate interests may be in conflict. On the one hand, it is in the child's best interest to get treatment, given the high chance of survival (90%) and significant chance (69%) of no disability. On the other hand, it may be in the family's best interest to avoid treatment for the child because of high hospital expenditures, risk of future catastrophic health expenditures if the child is disabled, risk of falling into poverty and the risk of reduced health for other children in the family. In addition, the mother's interest could conflict with the child's: she may want to avoid treatment because of the risk of income loss, because she will be occupied with home-based care of the disabled child, and risk of eviction from extended family. Finally, the interest of other patients with unmet needs might be in conflict with the child's, as they are affected and could lose care services/resources for their specific health conditions and needs.

Are fundamental principles in conflict?

We have identified six possible principles derived from general theories of distributive justice that may be in conflict29 30: (1) maximising health, (2) maximising welfare, (3) reducing inequalities in health outcomes, (4) reducing inequalities in welfare, (5) non-discrimination and (6) age weighting.

The weak evidence about outcomes (tables 1 and 2) make it difficult to apply principles directly, but we attempt a brief discussion of each.

Maximising health

Quite predictably, intensive care for neonates, even at the age of 28 weeks, yields a very favourable cost–effectiveness ratio because long-term survival alone dominates the effectiveness calculation. Based on our own estimates, the cost per QALY gained varies from 73 to 12 USD for the various GA cut-offs. Clearly, early intervention and chronic care services after discharge, including costs for future hospitalisations, would increase total cost and reduce cost–effectiveness ratios; however, studies from the UK indicate that the first hospital stay constitutes the major cost.31 If the aim of the treatment policy in question is to maximise health with emphasis on long-term survival for an individual patient, the general rule should be to treat all children in the range of 28–32 weeks. If there are strict resource constraints, the child of 32 weeks should be treated before the child of 28 weeks (a better cost–effectiveness ratio).

Maximising welfare

If we assume that an assessment of welfare includes not only that of the individual child, but also that of his or her immediate and extended family, it is possible that a maximising rule could permit non-treatment of many neonates born between 28 and 32 weeks. Individual cases will vary in terms of initial hospital costs for treatment, possible catastrophic future health expenditures and risk of pushing families into extreme poverty.

Reducing inequalities in health outcomes

Reducing inequalities in health outcomes typically implies adding some extra weight to the worst off. One plausible (though not the only) definition of being “worst off” is the patient with the worst health outcome in the absence of treatment. According to this principle, the child of 28 weeks should be treated before the child of 32 weeks because the likelihood of death without intervention is higher at this earlier gestational age. Of course, other children in a family may be at increased health risk if the family falls into poverty because of indebtedness from new hospital costs, but their health risks are not likely to be as severe as the predicted immediate death of the premature neonate.

Reducing inequalities in welfare

The aim of reducing inequalities in welfare would, as above, imply some extra weighting to the worst off. Welfare is a broad notion, and we define the worst off in this context as the person with the least lifetime expectations of welfare. A neonate at 28 weeks—with an increased risk of other diseases later in life, a risk of life with disability in a poor family and a society without an adequate welfare system—will be among those worst off in terms of welfare. The principle of equity in welfare would imply high priority to all types of measures that could reduce this inequality.

However, as above, other children in the family of saved neonates may also experience a substantial loss of welfare if the family falls into poverty because of catastrophic health expenditures. If the overall pattern of inequality in expected welfare loss for the whole family is in aggregate greater by letting the neonate live, this principle would imply the acceptability of letting the child die.


In the Indian context, the most important issue in terms of non-discrimination is sex selection (although other types may also be found). The principle of non-discrimination states that if two persons are equal in all other respects than their sex (such as medical and family characteristics), they should be treated equally. If it is considered acceptable to withhold or withdraw treatment for some neonates, and the difficult choice is left to the family, it is likely that some sex discrimination may occur. Non-discrimination does not imply that it is never permissible to withhold or withdraw treatment, only that similar cases should be treated alike.

Age weighting

In our model, we give equal weight to all age groups in order to avoid methodological complexities. Figure 1 illustrates four alternative age-weighting options which are discussed in the philosophical literature.32 Granting equal weight to all age groups is referred to as “equality”. The “prudential life span” approach illustrates how prudent allocators interested in fair allocation of resources would spend a limited budget of healthcare resources over a lifetime. They would typically spend more during the first years of life as compared with the last period of life. This is in contrast with “DALY age weighting”, where the most productive age groups are favoured. Age weighting in DALY relies on mean estimates of peoples' preferences.33 The “biological lifespan” approach would argue that people should only be given weight when they have developed a personality and others have invested in their yet-unfulfilled life project.30 The “DALY” and “biological lifespan” approaches are the only perspectives that give minimal or no weight to the first years of newborns. Cost–effectiveness ratios would raise the lower weights early years are given; the measure would not be cost effective at all if early years are given zero weight.

Figure 1

Schematic figure of four different age weightings.

There is a fair amount of academic disagreement about the general concept of age weighting, even under conditions of resource scarcity.30 We do not attempt to reconcile the various positions that critically depend on controversial moral and philosophical assumptions about what might make a biological life valuable. However, it is notable that, although some commentators have recently suggested that a “biological lifespan” approach most closely corresponds with common-sense intuition, this claim lacks robust empirical validation.30 Indeed, the exponential growth in neonatal intensive care in modern healthcare economies and its attendant associated costs at least suggest that our communal intuitions about the value of human life at various stages of development are sociologically complex. As a simple example, it may be that most people would, on reflection, agree that the death of an 18-year-old is, all other things being equal, morally worse than the death of a 1-day-old. However, it is not at all clear that most would agree that extending the life of an 18-year-old for 40 years, all other things being equal, is morally preferable to extending the life of a 1-day-old by the same amount.

Concluding remarks

Our ethical analysis sketches two alternative answers to the question of whether it is justifiable to withhold or withdraw treatment of neonates >28 and <32 GAW. If the goals of prioritisation are to maximise individual health and reduce health outcome inequalities, the conclusion should be that it is not ethically permissible to limit treatment to neonates younger than 32 weeks. Neonates between 28 and 32 weeks of age score high on severity of disease, efficiency and cost effectiveness of treatment. This conclusion assumes that one gives full and equal moral weight to the value of early years of a newborn's life as compared with later ages (ie, that age weighting is not applied). If the goals of prioritisation are to maximise social welfare and reduce welfare inequalities across individuals, the conclusion might be that it is acceptable to limit treatment for these precariously positioned neonates. This second conclusion also requires acceptance of contestable conclusions: namely, that full and equal moral weight is not accorded to newborns, and that the burdens on others (eg, worsening indebtedness, denial of other healthcare services) can reasonably be anticipated and measured.

Both of our alternatives depend on the validity of our preliminary outcome analysis. We acknowledge that there are several limitations in our analysis. In the absences of any relevant, reliable published studies, the outcome data had to be estimated and modelled. A sensitivity analysis would have compensated for some of the uncertainty, but it is beyond the scope of the ethical analysis presented in this paper. We hope that the transparency and explicitness of the method used is sufficient to set forth our central concerns and promote further discussion. We advocate for more follow-up studies concerning outcomes for neonates, costs to families, communities, hospitals, in low-income countries, and from India in particular. Clinicians who are currently forced to practice bedside rationing and policymakers who might consider population-level priority-setting decisions ought to have better data to help rationally inform future decisions.

We acknowledge that there are likely more cost-effective approaches to reducing the burden of neonatal mortality in India than increasing intensive care services to neonates born below 32 weeks. Context-specific comparisons of different interventions including those programs aimed at improving access to quality antenatal care are vital to any discussion that seeks to set a nation-wide health policy agenda.34 35 Our goal in offering this analysis is not to discount the critical importance of such larger considerations. Nevertheless, it is important to recognise that no country, no matter how well-developed and routinely accessed are its antenatal service by pregnant women, has completely abolished the problem of very early premature birth.

We find that non-initiation or early withdrawal of neonatal intensive care can be used as a means to improve the welfare of others. However, it is important to recognise that a major ethical cost of this practice seems to include a willingness to tolerate socially problematic background conditions in India. There can be little doubt that the poor, marginalised, disabled and even women as a class suffer unjustifiably there. As such, we worry about any practice or policy that appears to perpetuate or reinforce normative indifference towards root sources of social injustice. This appears to be a seemingly intractable problem for those interested in priority setting under conditions of resource scarcity that has yet to be adequately addressed. We challenge economists, epidemiologists and ethicists alike to more seriously consider the ramifications of somewhat arbitrary line drawing when it comes to establishing what parts of a global socio-economic status quo must be assumed as a bow to the “reality on the ground.”

The limited available neonatal intensive care for the hundreds of thousands of premature neonates in low-income countries presents a major and unmet challenge in healthcare prioritisation schemes. If premature neonates are perceived as of less worth than older persons or other patient groups, this should be explicitly argued for, and the reasons for supporting such an intuition ought to be explored among a much broader range of stakeholders than bioethicists, policy-makers, economists and philosophers.



  • Funding University of Bergen, Norway and the Young Investigator Grant, Norwegian Research Council (Norheim). The researchers are independent of funders.

  • Competing interests None.

  • Ethics approval The project was accepted by the research ethics committee at the hospital and was approved by the Norwegian Social Science Data Services.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • i Estimate based on prevalence of prematurity in South East Asia and annual live births in India.

  • ii The estimate of 70 000 NICU beds also includes services for asphyxiated, infectious and malformed neonates.

Other content recommended for you