Article Text

Neonatal deaths: prospective exploration of the causes and process of end-of-life decisions
  1. Jonathan Hellmann1,2,3,
  2. Robin Knighton1,
  3. Shoo K Lee2,4,
  4. Prakesh S Shah2,4
  5. on behalf of the Canadian Neonatal Network End of Life Study Group
    1. 1Division of Neonatology, Hospital for Sick Children, Toronto, Canada
    2. 2Department of Paediatrics, University of Toronto, Toronto, Canada
    3. 3Department of Bioethics, Hospital for Sick Children, Toronto, Canada
    4. 4Department of Paediatrics, Mount Sinai Hospital, Toronto, Canada
    1. Correspondence to Dr Prakesh S Shah, Department of Pediatrics, University of Toronto, Staff Neonatologist, Mount Sinai Hospital, Rm 19-213D 600 University Avenue, Toronto, Ontario, Canada M5G 1X5; pshah{at}


    Objective To determine the causes and process of death in neonates in Canada.

    Design Prospective observational study.

    Setting Nineteen tertiary level neonatal units in Canada.

    Participants 942 neonatal deaths (215 full-term and 727 preterm).

    Exposure and outcome Explored the causes and process of death using data on: (1) the rates of withdrawal of life-sustaining treatment (WLST); (2) the reasons for raising the issue of WLST; (3) the extent of consensus with parents; (4) the consensual decision-making process both with parents and the multidisciplinary team; (5) the elements of WLST; and (6) the age at death and time between WLST and actual death.

    Results The main reasons for deaths in preterm infants were extreme immaturity, intraventricular haemorrhage and pulmonary causes; in full-term infants asphyxia, chromosomal anomalies and syndromic malformations. In 84% of deaths there was discussion regarding WLST. WLST was agreed to by parents with relative ease in the majority of cases. Physicians mainly offered WLST for the purpose of avoiding pain and suffering in imminent death or survival with a predicted poor quality of life. Consensus with multidisciplinary team members was relatively easily obtained. There was marked variation between centres in offering WLST for severe neurological injury in preterm (10%–86%) and severe hypoxic-ischaemic encephalopathy in full-term infants (5%–100%).

    Conclusions and relevance In Canada, the majority of physicians offered WLST to avoid pain and suffering or survival with a poor quality of life. Variation between units in offering WLST for similar diagnoses requires further exploration.

    • Neonatology
    • Palliative Care
    • Mortality
    • Ethics

    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.

    What is already known on this topic

    • Withdrawal of life-sustaining therapies precedes the majority of deaths in neonatal units.

    What this study adds

    • When offering to withdraw life-sustaining treatment for critically ill neonates, achieving consensus with parents and the team was not perceived as difficult by physicians.

    • Withdrawal of life-sustaining treatment was offered to prevent pain, suffering and predicted poor quality of life.


    Despite universal health insurance and a highly regionalised perinatal–neonatal care system, marked variation in the mortality among neonates admitted to neonatal intensive care units (NICU) has been reported in Canada.1–3 The sources of variation in outcome otherwise as definitive as death are likely to reflect differences in (1) medical factors: diagnosis, severity of illness,4 ,5 treatment and its effectiveness; (2) parental wishes, choices, preferences and cultural traditions; (3) strength of medical recommendations, views and values of the professionals, ‘ethical culture’ or guidelines of the unit; and (4) legal concerns and perceptions of legal precedent, among other factors.6 ,7

    Exploration of differences in mortality due to medical measures has been well-studied. Scoring systems are used to account for physiological differences and severity of illness. However, the impact of other factors on mortality is far less evaluated. Individual studies have reported on parents’ perspectives, either through the eyes of treating physicians or via limited post-death parental interviews. Similarly, legal constraints or perceptions that influence the decision-making process regarding redirection of care have been reported.8 ,9 The missing or ‘hidden’ element in such reports is the physician and the team's different approaches to, and comfort levels regarding end-of-life (EOL) decision making.

    Studies have documented that decisions to limit or forgo life-sustaining treatment are made in the majority of NICU deaths.10–15 However, the factors that play a role in the variation of decisions in the face of similar neonatal conditions are not described. Despite relatively similar rates of withdrawal of life-sustaining treatment (WLST) as the terminal event, it is not known when in the infant's clinical course this occurred, on what basis the issue was raised, who raised the issue, what degree of consensus was achieved, what elements of life-sustaining medical treatment were withdrawn, how these decisions were reached and where these infants died.

    The objective of this study was to determine the cause and process of deaths in neonates admitted to Canadian NICUs. Documenting the cause of death involved a detailed diagnostic categorisation of all neonatal deaths. To explore the process of death we captured: (1) the rate of withholding or withdrawing life-sustaining treatment, (2) the reason for raising the issue of withholding or WLST, (3) the degree of consensus reached with parents, (4) the nature of the decision-making process, (5) the specific elements withheld, (6) the age of death, and the time between WLST and death and (7) the location of the neonatal death.


    Design and setting

    This was a prospective, observational cohort study, nested within the Canadian Neonatal Network (CNN) data collection system. The CNN is a national perinatal–neonatal database that collects data according to strict guidelines.16 Data included demographic factors, key obstetric variables, neonatal illness severity, diagnoses and resource use. The present study was conducted after obtaining ethics approval from each institutional review board.

    Study population

    Infants who died after admission to participating NICUs over a 2-year period (October 2008–September 2010) were included in the study. Of the 26 NICUs in the network 19 (73%) participated in this study, representing 75% of infants admitted to Canadian NICUs during this period. Data on those infants who were admitted to a participating NICU and then died in other wards/hospitals or at home (eg, if in palliative care) were also collected. Stillbirths, infants weighing <500 g and/or born at <23 weeks gestation were excluded.

    Consent specific to this study

    This study collected additional data to the standard database platform. The additional data (see online supplementary appendix 1) in this study required direct input from each final treating physician. Consent was therefore obtained from individual physicians within each institution to provide their input on the EOL process. Data were de-identified and no input from parents was sought.

    Data collection

    Causes of death

    The cause of death was recorded in two gestational age categories: preterm (<37 weeks’ gestation) and term (≥37 weeks’ gestation). Physicians were required to categorise one cause of death as the most responsible cause and other factors as contributing causes. Cardiorespiratory arrest and death after WLST were considered a final event and not the cause of death. The list of causes of death was developed from a review of a number of proposed classification systems.2 ,17 ,18 The list was pilot-tested at two sites and modified.

    Process of death

    The EOL decision-making component, with categories of decisions reached after discussion with parents and the team, listed all the possible options that pertain to this situation and included whether or not a decision was made antenatally to provide active or comfort care. The options on the reason for withholding or WLST are in the appendix. They included: (1) to avoid pain and suffering, as death was inevitable in the short term (<24 h) even if treatment was continued; (2) to prevent prolonged pain and suffering, as death was likely within the longer term (days to weeks); and (3) to prevent survival with a likely poor quality of life. There was some flexibility to allow physicians to enter ‘other’ categories. We did not include a category used by previous Dutch and Swiss reports of intentional or active termination of the patient's life.19 ,20

    We captured the number of formal parental meetings in which withholding/withdrawal of treatment was discussed; however, we could not determine the exact number of non-formal discussions between family members and physicians.

    The intent of this study was to focus on determining whether a decision with parents was reached, and to describe the ‘ease’ of getting to that decision, as well as recording the responsible physician's degree of certainty regarding the prognosis. In view of the multidisciplinary team's involvement in the decision-making process, the responsible neonatologist was also asked to comment on the ease of deriving team consensus.

    Data verification

    Data on mortality from each centre in the CNN were cross-referenced to capture all deaths. Site abstractors were encouraged to approach the attending physician immediately after each death and at the mortality review meetings of that unit.

    Statistical analysis

    Descriptive statistical methods were used to summarise the study population. The causes of deaths were classified in various categories for term and preterm infants. The primary causes of death were analysed for preterm and term infants separately. The percentage of cases in which there was discussion regarding withholding or withdrawal of treatment measures was calculated. If withholding or withdrawal of treatment was initiated, the elements of withholding or withdrawal were analysed.

    The reasons for offering withholding or WLST were compared across NICUs in two specific conditions: grade 3 or 4 intraventricular haemorrhage or periventricular leukomalacia in preterm infants born at <32 weeks’ gestation, and perinatal asphyxia in term infants.


    The questionnaires were completed by attending physicians in most cases within days of death, almost all within a week after death and a very few forms were completed at mortality meetings of the unit. During the 2-year study period 942 deaths out of 11 738 NICU admissions (incidence rate of 8.0%) were recorded. Of these, 758 (80%) occurred in the NICU, 129 (14%) in the delivery room and 55 (6%) after discharge with redirection of care. The number of deaths ranged from 5 to 119 (mortality rate 1.9%–16.8%) between sites over the 2-year period. There were 727 deaths (77%) in preterm and 215 (23%) in term infants. The top 10 causes of death in preterm and term infants are listed in table 1.

    Table 1

    Causes of death in infants

    Process of death

    Discussions of WLST

    Discussions regarding WLST occurred in 798 (84%) deaths. The reasons given by physicians for discussion and non-discussion are highlighted in figure 1. The majority of discussions (95% of those where data were available) took place in the NICU. The characteristics of the discussions and the process are depicted in table 2.

    Table 2

    Characteristics of discussions and responses

    Figure 1

    Flow diagram of patients included in the study. WLST, withdrawal of life-sustaining treatment.

    Prognostic certainty and consensus

    Five hundred and twenty (55%) of physicians reported that they were absolutely certain, 281 (30%) very certain and 31 (3%) moderately certain of the prognosis. In two cases the physicians indicated that they were very uncertain regarding the prognosis; data were missing in 108 cases.

    Among those who had discussions regarding WLST (n=798), the process of achieving consensus with parents was described as easy by 424 (54%) physicians, somewhat easy by 197 (25%), of average difficulty by 74 (9%), somewhat difficult by 55 (7%) and very difficult by 43 (5%). In five cases data were missing. The majority of physicians (99%) reported that it was easy to achieve consensus with the healthcare team regarding offering WLST and only in four instances it was noted to be very difficult.

    Process of WLST

    In 668 (84% of those who had WLST discussions) cases WLST was agreed upon between physician and parents. For 500 preterm infants who had WLST, 446 and 218 had withdrawal of assisted ventilation and cardiac support, respectively. Among 168 term infants who had WLST, 152 and 80 had withdrawal of respiratory and cardiac support, respectively. Some patients had withdrawal of multiple life-sustaining measures. A total of 573 neonates (61%, which could be as high as 67% if the denominator is only those with available data) received analgesia prior to withdrawal of treatment, 279 (29%) did not and in 90 (10%) data were not available. The median time of death was 1 h after WLST (IQR0.50–2.58 h, range 0.02–921.1 h). The median age at death was 5 days (IQR 1–17 days).

    Variations in the EOL decision-making process

    We explored WLST in more detail in two conditions: grade 3 or higher intraventricular haemorrhage or periventricular leukomalacia in preterm infants, and asphyxia in term infants (tables 3 and 4).

    Table 3

    Reasons for withdrawal of life-sustaining treatment among very preterm infants with severe neurological injury: comparison between sites

    Table 4

    Reasons for withdrawal of life-sustaining treatment among term infants with birth asphyxia and hypoxic-ischaemic encephalopathy: comparison between sites


    In this prospective observational study of a representative national cohort, we observed that approximately 71% of deaths in Canadian NICUs occur after WLST. The major causes of death among preterm infants were extreme prematurity, intraventricular haemorrhage and pulmonary causes, and among term infants asphyxia and congenital anomalies. The majority of physicians offered WLST to avoid or prevent survival with a poor quality of life. In most cases, consensus was relatively easily achieved with parents and other healthcare team members. However, marked variations were noted between centres in the frequency of initiation of such discussions, even with relatively similar diagnostic coding of severe neurological injury for preterm and hypoxic-ischaemic encephalopathy for term infants.

    While most published studies show that death in the NICU occurs following WLST (58%,13 68%,11 69%,10 72%,14 74%,12 75%21 and 79%15), they give no data on the decision-making process. Mortality rates are commonly used as a measure of NICU performance. However, ethical decisions regarding the use of life-sustaining therapy may influence such mortality data. Variation may also reflect the degree of comfort in making EOL decisions, frequency of making such decisions and culture in which such decisions are made, institutional policy, fear of legal constraints9 ,22 or more general societal views on neonatal care.23 In this study, our examination of the criteria for physicians’ reasons for WLST revealed that their assessment of ‘quality of life’ was the major reason. It is clearly acknowledged that quality of life is a value judgement about factual qualities; it is not a technical term and we cannot attach measurement scales to its determination.24

    The proportion of physicians who withdrew life-sustaining treatment based on an assessment of poor quality of life was higher than in a previous report12 and may reflect a growing appreciation of the burdens of intensive care survivors.25 ,26 This has also been reported in studies from Europe.27–29 In the USA it has been noted that centres with the shortest length of stay had the highest mortality,6 suggesting a trade-off between aggressiveness of interventions, and mortality and neurodevelopmental impairment. It should be noted that 84% of the time physicians offered redirection of care. The physicians reported absolute certainty (55%), or were very certain (30%) for a total of 85%; however, if the denominator had been only those with available data then these numbers could be 64%, 32% and 96%, respectively. This indicates that a degree of uncertainty is accepted by physicians when they discuss redirection of care. We did not investigate the reasons for their certainty or uncertainty regarding prognosis; we believe that their actions were based largely on the prognostic certainty of the primary and secondary causes of death.

    Our data revealed that in the majority of cases physicians initiated such discussions and in most cases the discussion resulted in consensus. However, we cannot make any inference regarding parental perception as to how the issue was discussed, how much opportunity or time was available, and whether there was an element of coercion involved. Our study provides a broad view of prevailing practice in Canadian NICUs; however, we acknowledge that exploring only the physician's view may have introduced bias. In addition, these are the views and opinions of Canadian neonatologists and they could be different from those practitioners in other parts of the world.

    When exploring at a deeper level whether real differences existed in the ‘ethical culture’ of units, we found marked variation in disorders of similar severity. For the purpose of this study, ‘ethical culture’ was conceptualised as the prevailing ethical consciousness and sensitivity to moral issues within that context. The ‘culture’ may not be explicit but is understood by the relevant healthcare professionals as the readiness to engage in ethical discourse, the respect given to diverse views and values, the past experience of the unit and the awareness of the group as to the normative boundaries of deviation in ethical decision making. In preterm infants, a unit in which the condition of 6 of the 13 premature infants with severe neurological injury prompted discussion leading to seven deaths (one death occurred before WLST discussion) is more likely to have a different approach to EOL decision making than a unit where four discussions with the parents of 39 infants with a similar disorder led to nine deaths (in the remaining five deaths WLST discussion did not take place at all). Similarly, variations were observed in term infants with severe hypoxic-ischaemic encephalopathy. These variations across the country might also reflect differences in case-mix, severity or parental choice,30 rather than the ethical culture of that particular unit. This could also be due to physician-related factors such as inability to prognosticate with certainty, previous experiences with similar situations and prevailing practices in the NICU. A detailed and larger study would be required to explore the decision-making process regarding WLST where individual case dynamics might be revealed. While previous attempts to explicate this aspect of EOL decision making have been hypothetical case-based assessments,31 ,32 behaviour and attitudes in real situations could be markedly different. One of the ideas behind this study was to trigger individual units’ reflection on the ethical dimensions of EOL decision making, and how they might make this more overt, explicit and measurable in their practice.

    The strengths of our study are that it included a nationally representative sample, a prospective design, healthcare team agreement to disclose such sensitive information and detailed, meticulous data collection. A key component of this study is that information regarding offering WLST and the circumstances surrounding each infant's death was obtained through direct exploration from the physician responsible for the neonate at the time of death and not from charts which may not adequately reflect these nuances of care.

    There are some limitations to our study. Physicians were asked to provide one cause as the single most important reason for death; it is likely that more than one reason was discriminative in the decision-making process. In addition, there is no well-accepted classification of causes of death in the literature; our categorisation attempted to capture a systems-based approach. We acknowledge that despite our best efforts to obtain information in real time, on occasion information was collected a few days after death. In addition, there was no attempt to capture or quantify parental views, reactions or perceptions of pain, suffering and quality of life. Finally, legal concerns are often cited as constraints for WLST. In Canada, despite variations in practices across provinces, the paucity of medicolegal cases in these situations suggests that decisions are currently made in a shared and ultimately consensual process.

    Conclusion and future directions

    Redirection of care to facilitate provision of palliative and comfort care prior to death in the NICU is commonplace. Obtaining consensus with parents and the team is crucial and can be reached in the majority of cases in Canadian NICUs. While the absolute number of deaths in each NICU may be small, collectively a large number of deaths do occur. Our study emphasises the need to further consider how neonatal death occurs and to focus on the potentially modifiable non-medical factors that could influence these challenging situations. Exploring the reasons for the differences between the teams in various centres in the rate of recommending WLST may play a major role in identifying hidden ethical determinants of deaths in individual centres.


    We would like to thank the data abstractors of the Canadian Neonatal Network (CNN) and staff of the CNN Coordinating Centre for providing organisational support to the CNN. As employees of MiCare, Mount Sinai Hospital, Toronto, Ontario, or their respective hospitals, all these people received salaried compensation for their assistance. We would like to thank Mr Xiang Ye, MSc for statistical support and Dr Ruth Warre, PhD for editorial assistance during the preparation of this manuscript, both of whom are employees of the Mother Infant Research Centre, Mount Sinai Hospital, Toronto. Written permission has been obtained from both of them by the corresponding author.


    Supplementary materials

    • Supplementary Data

      This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.


    • Collaborators Members of the EOL Study Group of the Canadian Neonatal Network (Collaborators): Wayne Andrews (Janeway Children’s Health and Rehabilitation Centre, St John’s, Newfoundland, Canada); Antoine Payot (Sainte Justine Hospital, Montreal, Quebec, Canada); Barbara Bullied (Everett Chalmers Hospital, Fredericton, New Brunswick, Canada); Aaron Chiu (St. Boniface General Hospital, and Winnipeg Health Sciences Center, Winnipeg, Manitoba, Canada); Kimberly Dow (Kingston General Hospital, Kingston, Ontario, Canada); Shaheen Doctor (Women's College Hospital, Toronto, Ontario, Canada); Jonathan Hellmann (Hospital for Sick Children, Toronto, Ontario, Canada); Prakesh Shah (Mount Sinai Hospital, Toronto, Ontario, Canada); Kevin Coughlin (St. Joseph’s Health Centre; London, Ontario, Canada); Douglas D. McMillan (IWK Health Centre, Halifax, Nova Scotia, Canada); Cecil Ojah (St. John Regional Hospital, St. John, New Brunswick, Canada); Abraham Peliowski (Royal Alexandra Hospital and Stollery Children's Hospital, Edmonton, Alberta, Canada); Bruno Piedboeuf (Centre hospitalier universitaire de Québec, Sainte Foy, Quebec, Canada); Patricia Riley (Montreal Children's Hospital and Royal Victoria Hospital, Montreal, Quebec, Canada); Nicole Rouvinez-Bouali (Children Hospital of Eastern Ontario Children's Hospital of Eastern Ontario, and Ottawa General Hospital, Ottawa, Ontario, Canada) Thierry Dabowal (Children's Hospital of Eastern Ontario, and Ottawa General Hospital, Ottawa, Ontario, Canada); John Watts (Hamilton Health Sciences Centre, Hamilton, Ontario, Canada); Horacio Osiovich (Children’s and Women’s Health Centre of British Columbia, Vancouver, British Columbia, Canada).

    • Contributors JH: conceptualised and designed the study, developed the protocol, obtained funding, interpreted the data, reviewed and revised the manuscript, and approved the final manuscript as submitted. RK: coordinated data collection and approved the final manuscript as submitted. SKL: reviewed the protocol, interpreted the data, reviewed and revised the manuscript, and approved the final manuscript as submitted. PSS: conceptualised and designed the study, revised the protocol, interpreted the data, drafted and revised the manuscript and approved the final manuscript as submitted.

    • Funding This study was supported by an operating grant from the Canadian Institutes of Health Research (#FRN 87364). Additional funding was provided by individual participating hospitals. The funding agencies had no role in the design, collection, analyses, or interpretation of the results of this study. Prakesh Shah has full access to all of the data in the study and takes responsibility for integrity of the data and accuracy of the data analysis.

    • Competing interests None declared.

    • Ethics approval All study sites and Hospital for Sick Children Ethics Board.

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

    • Data sharing statement Completed between sites. Data are in secure password protected computers.

    Linked Articles