Objective. To determine the frequency of selective nontreatment of extremely premature, critically ill, or malformed infants among all infant deaths in a level III intensive care nursery (ICN) and to determine the reasons documented by neonatologists for their decisions to withdraw or withhold life support.
Methods. This was a descriptive study based on review of the medical records of all 165 infants who died at a university-based level III ICN during 3 years. We determined whether each death had occurred despite the use of all available technologies to keep the infant alive or whether these were withheld or withdrawn, thereby leading to the infant's death. We also determined whether neonatologists documented either “futility” or “quality of life” as a reason to limit medical interventions.
Results. One hundred sixty-five infants died among the 1609 infants admitted during the study period. One hundred eight infant deaths followed the withdrawal of life support, 13 deaths followed the withholding of treatment, and 44 deaths occurred while infants continued to receive maximal life-sustaining treatment. For 90 (74%) of the 121 deaths attributable to withholding of withdrawal of treatment, physicians cited that death was imminent and treatment was futile. Quality-of-life concerns were cited by the neonatologists as reasons to limit treatment in 62 (51%). Quality of life was the only reason cited for limiting treatment for 28 (23%) of the 121 deaths attributable to withholding or withdrawal of treatment.
Conclusions. The majority of deaths in the ICN occurred as a result of selective nontreatment by neonatologists, with few infants receiving maximal support until the actual time of death. Neonatologists often documented that quality-of-life concerns were considered in decisions to limit treatment; however, the majority of these decisions were based on their belief that treatment was futile. Prospective studies are needed to elucidate the determinants of neonatologists' practice decisions of selective nontreatment for marginally viable or damaged infants.
During the past two decades, technological advances have greatly expanded treatment options for critical care of newborn infants. Neonatologists are now able to treat many infants who, in previous decades, would not have been resuscitated because of presumed nonviability. Application of neonatal intensive care has become so routine that some neonatologists feel culpable when they elect to limit the treatment of marginally viable infants. Many neonatologists also believe that application or continuation of intensive care may at times be either inappropriate or inhumane, such as when death is imminent or when further medical interventions would only prolong suffering.
The practice of withdrawal or withholding life support for certain critically ill infants in neonatal intensive care units was reported in 1973 in a landmark article by Duff and Campbell.1 The ensuing decades have witnessed remarkable progress in extending the limits of viability for severely ill infants. Concurrently, questions about the ethics and legality of limiting treatment for infants with malformations or handicaps have arisen in the debate that culminated in the 1985 Baby Doe regulations. (The 1985 Department of Health and Human Services “Baby Doe” regulations stated that physicians must treat all infants with life-threatening conditions unless, in the physician's reasonable medical judgment, one of three conditions is met: “[i] the infant is chronically and irreversibly comatose; [ii] the provision of such treatment would merely prolong dying, not be effective in ameliorating or correcting all of the infant's life-threatening conditions, or otherwise be futile in terms of the survival of the infant; or [iii] the provision of such treatment would be virtually futile in terms of the survival of the infant and the treatment itself under such circumstances would be inhumane.”2) These changes notwithstanding, neonatologists' current practices of withholding and withdrawing life-sustaining therapy have not been closely reexamined.
Recent surveys examining physician attitudes or their responses to hypothetical cases have documented that neonatologists frequently are willing to limit treatment for certain infants who are malformed, severely premature, or hopelessly ill.3-7 Despite such reports, however, there is little information about neonatologists' actual practices of withdrawing or withholding life support in infants.
Furthermore, two recent homicide cases illustrate the continuing legal uncertainties and potential culpability of parents and care givers in withholding or withdrawing life support for severely damaged or marginally viable infants. One case debated the rights of parents to terminate treatment of uncertain efficacy for an extremely premature infant.8 In the second case, a physician was alleged to have hastened an infant's death as futile life support was discontinued.9
We recognize that physicians' practices may not always agree with their reported attitudes and hypothetical responses as they consider the potential legal consequences of their clinical decisions. Therefore, in addition to studies of physician attitudes and treatment preferences, we believe that knowledge of current clinical practices is an essential element in the discussion about selective nontreatment of damaged or dying infants.
Our study objective was, therefore, to characterize the practices of life support discontinuation by neonatologists. We reviewed all deaths in a 3-year period at a university-based intensive care nursery (ICN) to determine the extent that neonatologists withhold or withdraw treatment for damaged or dying infants. To ascertain concurrence with the Baby Doe guidelines, we additionally examined the physicians' reasons for selective nontreatment.
We reviewed the medical records of all infants who died in the ICN at the University of California, San Francisco (UCSF) between June 1, 1989, and May 31, 1992. We abstracted information from the attending physicians' and nurses' notes. The study was approved by the Committee on Human Research at the UCSF.
The UCSF ICN is a university-based, tertiary nursery providing comprehensive neonatal care for both inborn and outborn infants from northern California. There were 1609 admissions to the ICN during the study period, 843 inborn and 766 outborn, with an average daily census of 38 patients in the ICN.
The UCSF ICN uses a multidisciplinary team led by an attending neonatologist responsible for all medical decisions. Parents are informed of their infant's status and proposed treatment plans during regular discussions with physicians and nurses; they also receive psychosocial support from social workers throughout their infant's hospitalization. Discussions regarding limitations in care may be initiated by either parents or the physician. Consensus among all team members is routinely sought when decisions to limit life support are considered. With no clear medical guidelines in most cases, the physicians' recommendations are based on their best clinical judgment. Although the attending neonatologist must either recommend or agree to withdrawing or withholding life support, the infant's parents make the ultimate decision.
Classification of Deaths
We classified each death as having occurred in one of three ways: (1) death attributable to the withholding of additional therapies, (2) death attributable to the elective discontinuation of current life support, or (3) death occurring despite maximal support. We categorized deaths cause by withholding as those in which additional medical or surgical interventions needed by the infant for immediate survival were withheld. Similarly, we categorized deaths attributable to withdrawal as those in which interventions already sustaining an infant's life were electively discontinued, representing a decision to terminate ongoing treatment. All deaths that occurred during or immediately after lifesaving surgical procedures, despite cardiopulmonary resuscitation, or during preparations for extracorporeal membrane oxygenation (ECMO) were categorized as deaths despite maximal support.
We determined diagnoses contributing to death based on the attending neonatologist's notes and death summary. (Attending neonatologists dictate detailed daily notes and discharge summaries for all infants in the ICN.) Diagnostic categories were defined as follows: (1) extremely low birth weight (ELBW) for all infants with birth weights of less than 800 g; (2) intracranial hemorrhage (ICH) of grades III and IV (by the grading of Papile et al10), periventricular leukomalacia, or other parenchymal hemorrhage; (3) necrotizing enterocolitis (NEC) of stage IIb or higher by the classification of Bell et al11; (4) hypoxic-ischemic encephalopathy (HIE); (5) respiratory failure (ie, severe hypoxemia or hypercapnea secondary to conditions such as respiratory distress syndrome, bronchopulmonary dysplasia, air leak syndromes, meconium aspiration, pneumonia, and pulmonary hypoplasia); and (6) major congenital anomalies or chromosomal abnormalities.
When more than one condition contributed to death, multiple diagnoses were assigned (eg, an infant weighing 600 g with severe respiratory distress syndrome and grade IV intraventricular hemorrhage would have been assigned three diagnoses contributing to death: ELBW, respiratory failure, and ICH).
Mortality rates in the UCSF ICN for infants with each of these diagnoses were calculated by dividing the number of infants dying in each diagnostic category by the total number of admissions to the ICN with that diagnosis for the entire study period. Diagnostic information was maintained prospectively in a computerized database in which all diagnoses were recorded on each infant's admission to the ICN, with these diagnoses verified and updated on a weekly basis by attending neonatologists.
The attending neonatologists' daily notes and death summaries were reviewed to determine their reasons for limiting treatment in those cases in which life support was either withdrawn or withheld. We categorized the following reasons physicians documented for restricting treatment: (1) futility of treatment in the face of limited life expectancy; (2) poor developmental prognosis; and (3) suffering resulting from continuing treatment. Physicians' notes were also reviewed to determine whether discussions of withdrawal of support were initiated by physicians or parents. We ascertained whether any or all of the following therapies were withheld or withdrawn: ventilatory support, ECMO, vasopressor support, and fluids. Physician orders were reviewed to determine whether “do not resuscitate” (DNR) orders were written before an infant's death.
Our definition of treatment includes medical interventions with accepted potential biological efficacy, irrespective of the appropriateness of such interventions in a broader context. We acknowledge that the terms treatment and therapy should connote interventions intended for the well-being of the patient; however, for the purpose of clarity in this study, treatment refers to a potentially efficacious medical intervention.
All data were categorical. Descriptive statistics consisted of frequency distributions of the variables.
Of 1609 admissions during the 36-month study period, 165 infants died during their ICN courses. One hundred eight of these deaths were by withdrawal of ongoing life support, 13 deaths were attributable to withholding of additional therapy, and 44 deaths occurred despite maximal treatment. Overall, 73% of deaths were attributable to withdrawal or withholding of life-sustaining treatment. In-hospital mortality was greatest for infants with the diagnosis of ICH (39%), followed by NEC (29%), ELBW (28%), congenital anomalies (24%), and respiratory failure (17%) (Fig 1).
Figure 2 shows a flow diagram of outcomes of all ICN admissions for each diagnostic category. In each diagnostic category, death by withdrawal or withholding of life support was more common than death despite maximal support (Fig 3). Conditions associated with extreme prematurity and/or severe neurologic damage had the highest proportion of deaths attributable to withdrawal or withholding of support (ICH, 88%; HIE, 86%; ELBW, 83%; and NEC, 80%). In the other diagnostic categories examined, withdrawal or withholding support was slightly less common but was nevertheless the more common mode of death (eg, 72% of deaths with respiratory failure and 67% of deaths with major congenital anomalies were attributable to withdrawal or withholding support).
There were no differences between birth weight groups in the proportion of deaths attributable to withdrawal or withholding support when compared with deaths despite maximal treatment (Table1). The majority of deaths in each birth weight category were attributable to withdrawal or withholding of life support.
Reasons for Withdrawing or Withholding Treatment
The most frequently documented reason for limiting life support was the neonatologist's belief that continued treatment was futile in the face of imminent death, noted in 74% of these deaths. One or more quality-of-life concerns was documented as a reason to limit treatment in 51% of these deaths. Quality-of-life concerns included the prognosis for severe disabilities (in 46% of withdrawal or withholding deaths) and the belief that the infant would unnecessarily suffer as a result of continued treatment (in 15% of deaths attributable to withdrawal or withholding).
Furthermore, we examined whether the physicians' documented reasons for withdrawing or withholding treatment were based only on quality-of-life concerns. Quality-of-life concerns, exclusive of any reference to the futility of treatment, were noted for 23% of deaths attributable to withdrawal or withholding of treatment compared with futility concerns in 74%. (In 4 of the 121 deaths attributable to withdrawal or withholding of support, physicians' charting cited no clear reason for the decision to limit treatment.)
We determined the proportion of infants in each diagnostic group for whom quality of life was the sole reason for decision to limit treatment resulting in death. Quality of life was the sole reason cited for withholding or withdrawing life support in 53% of the deaths with HIE and in 43% of the deaths with ICH but in only 28% of the deaths with other diagnoses (P < .05).
Process of Withdrawal or Withholding
There was eventual agreement in all cases among physicians, hospital team members, and parents that withholding or withdrawing support represented the most appropriate and humane option for the infant. Medical record notations were insufficient to determine the frequency and depth of discussions between physicians and families to reach consensus about treatment limitation. Nevertheless, we noted that the total duration of these discussions ranged from hours to several weeks, indicating the extreme variability in the time needed for physicians and parents to agree that withholding or withdrawing life-sustaining treatment represented the infant's best interests. Physician notes indicated that parents initiated the discussion of limiting treatment in only a few (13%) of these deaths.
DNR orders were written for 27 (22%) of the infants who eventually died as a result of the withdrawal or withholding of life support. DNR orders were more commonly written for infants whose deaths resulted from withholding treatment (53%) than for infants whose deaths resulted from withdrawal of life support (18%).
Intubation and mechanical ventilation were the interventions withheld for all 13 infants who died after the withholding of treatment. Similarly, in all 108 cases in which death resulted from treatment being withdrawn, mechanical ventilation or ECMO was the therapy that was discontinued. Mechanical ventilation alone was withdrawn in 97 (90% of deaths attributable to withdrawal); ECMO and mechanical ventilation were simultaneously withdrawn in 10 (9% of deaths attributable to withdrawal); and ECMO alone was withdrawn in 1 infant who thereafter received ventilatory support until death. Other therapies were often discontinued simultaneously with or after the discontinuation of respiratory support. These included vasopressors (discontinued in 11% of deaths attributable to withdrawal), fluids (19%), or both (57%). Most often such therapies were discontinued to allow parents to hold their dying infants with minimal interference from intravenous lines. There were no deaths attributable to the withdrawal or withholding of fluids.
The majority (73%) of infant deaths in our tertiary, university-based ICN resulted from neonatologists' decisions to limit life-sustaining treatment. Withdrawal of ongoing therapies was far more common than withholding treatments that would have been necessary to prolong life (65% vs 8%). Thus, for those infants who died at our institution, neonatologists limited the implementation or the continuation of life-sustaining treatment more commonly than they maintained all such treatments until the time of death. These findings contrast with those of Duff and Campbell,1 who reported in 1973 that only 14% of deaths in the ICN were attributable to withdrawing or withholding treatment. Changes in the limits of viability, and thus in the implications of “maximal” treatment, certainly prevent the conclusion that limiting life support for sick infants has dramatically increased in the last 20 years. Nevertheless, we believe that the widespread application of neonatal intensive care has likely increased the proportion of infants for whom aggressive treatment is attempted but for whom it is subsequently determined to be ineffective or inappropriate.
In our ICN, those diagnoses associated with extreme prematurity (eg, ICH, ELBW, and NEC) had the highest proportions of withdrawing or withholding treatment as the mode of the infants' deaths. Furthermore, among the deaths attributable to treatment limitation in these extremely premature infants, withdrawal of ongoing treatment was much more common than withholding of additional interventions. For instance, initial delivery room resuscitation of marginally viable, extremely premature infants was rarely withheld. Of 70 liveborn infants weighing between 500 and 799 g, 64 (91%) received intubation and resuscitation in the delivery room. At our institution, therefore, the initiation of aggressive delivery room care did not preclude subsequent withdrawal of that treatment when it was considered either futile or inhumane.
There were proportionately fewer deaths attributable to withdrawal or withholding of support among heavier infants compared with the lowest birth weight category; nevertheless, the majority of deaths in each birth weight category occurred as a result of withdrawal or withholding of treatment. Among larger critically ill or severely malformed infants, life-sustaining interventions were almost always initiated and continued until a poor prognosis was established by diagnostic studies or subspecialty consultants. A recently published review of deaths among surgical neonates reported similar findings: deaths as a result of withdrawal or withholding treatment most often occurred after studies confirmed the severity of congenital anomalies or the existence of chromosomal abnormalities.12
The generalizability of this study may be limited, in that only the practice of one institution is reported; nevertheless, we believe that the practice of selective nontreatment described at our institution may be typical of other tertiary care nurseries in the United States. To estimate the comparability of ICN practice at our institution to other US neonatal intensive care units, we compared survival of the tiniest infants in our ICN with that in other tertiary care nurseries recently reporting such statistics. Table 2 shows survival-to-discharge rates among inborn infants weighing 500 to 799 g. This birth weight group was selected because diagnoses and illness severity should show less institutional variation than in other birth weight or diagnostic groups. For each 100-g birth weight category, the survival-to-discharge rates for UCSF are at least as high as those in all other tertiary-care nurseries reporting survival rates for the same period.
Although we cannot exclude potential confounding because of unmeasured differences in population and case mix, we believe that the acuity of admissions to our ICN for ELBW is at least as high as in these other centers.13 Nevertheless, survival rates for 500- to 799-g infants at our tertiary care nursery are comparable with or higher than those in these other tertiary centers, suggesting that neonatologists' practices of withdrawing or withholding treatment for certain severely ill or malformed infants at our institution have not resulted in higher mortality rates.
Whether practices of withdrawing and withholding life support at our institution were consistent with the Baby Doe regulations was further considered by examining the reasons documented by neonatologists for discontinuing treatment. Inasmuch as the Baby Doe regulations would require physicians to provide treatment to disabled infants unless that treatment were considered futile, we examined whether neonatologists who discontinued life support documented an assessment of futility. In most cases (ie, 74% of deaths attributable to withdrawal or withholding), neonatologists explicitly documented that further care was believed to be futile in terms of the infant's survival. However, in 23% of such deaths, the only reasons physicians cited for discontinuing treatment were concerns about the infants' quality of life.
Since the Baby Doe controversy of the last decade, considerable debate about passive euthanasia in extremely premature, severely ill, or malformed infants has focused on whether quality-of-life concerns by physicians and parents should be ethically and legally acceptable considerations in these decisions. Critics of the 1985 Baby Doe regulations argued that decisions to terminate life support in infants should be based on the principle of beneficence, thereby allowing consideration of quality of life in deciding an infant's “best interests.”17-27 In circumstances in which severe incapacitation and suffering would result from extending life, such critics have argued that further treatment may not be in the infant's best interest. Some have even asserted that continued life support in such circumstances would represent a violation of the Hippocratic principle, “Primum non nocere” (“first, do no harm”).21 Thus, critics of the Baby Doe regulations have objected to the implicit proscription against quality-of-life considerations as conflicting with the ethical principles of beneficence and nonmaleficence.
Since the Baby Doe controversy, surveys of attitudes and hypothetical treatment preferences have found that neonatologists may be willing to discontinue treatment of certain infants based on their quality-of-life concerns. Recent studies have found that neonatologists are willing to limit the treatment of marginally viable premature infants in certain circumstances7 and that neonatologists frequently consider withdrawing support when continued treatment would likely cause the infant to suffer.3 A questionnaire survey of neonatologists by Kopelman and coworkers5 found that a majority of respondents believed that the Baby Doe regulations did not allow adequate consideration of infant suffering in deciding whether to discontinue life support.
Nevertheless, in the aftermath of the Baby Doe debate, no US studies as yet have reported on actual physician practices of limiting neonatal life support, ie, when, how frequently, and for whom. The only recently published studies of physician practices of withdrawing or withholding treatment in infants are from Canada,28,29 the United Kingdom,30 and Japan.31 The proportion of deaths attributable to discontinuing therapy in these reports ranged from 30% to 55% of all infant deaths.
Our study of physician practices, therefore, supports previous findings that neonatologists are willing to discontinue life support for certain infants, and that they are willing to consider the infant's quality of life in such decisions. If physicians were to have understated their quality-of-life concerns in their medical record notations, then our study would underestimate the actual importance of quality-of-life considerations in decisions to terminate life support for infants. Quality of life was the only documented reason for withdrawal or withholding treatment among 23% of deaths; however, if documentation, but not actual decision making, were influenced by Baby Doe proscriptions, then this estimate might be low. Our findings suggest that, at times, neonatologists follow the ethical principles of beneficence and nonmaleficence when these principles are at variance with the Baby Doe regulations. Future studies of physician practice should seek to clarify how neonatologists' ethical principles influence the practice of discontinuing life support.
A recommendation of the Baby Doe regulations was that infant care review committees or hospital ethics committees be developed to assure appropriate decision making and treatment when life support termination was to be considered.32 During the study period in our institution, consultation of the hospital ethics committee was rare among the cases of withdrawal or withholding of support, with only 2 documented consultations of 121 cases of treatment limitation. Because our study only examined deaths and not survivals, we cannot report the overall frequency of ethics committee consultations. The apparently infrequent use of ethics committee consultations in our institution, however, should be considered in light of the common practice of consulting subspecialists or other neonatologists. We previously reported that 75% of withdrawing or withholding deaths followed consultation from a subspecialist regarding the decision to continue or discontinue treatment.33 Others have similarly reported the lack of infant care review committees and the infrequent consultations of these committees.33-35 Thus the selective referral to our hospital ethics committee of only the most complex cases may not be atypical. The safeguards of Baby Doe ethics committees may have been obviated by the frequent use of medical consultations and second opinions at our institution.
In summary, selective nontreatment of marginally viable or severely handicapped infants continues to be a vexing ethical dilemma for neonatologists 20 years after first being described by Duff and Campbell.1 Although the intervening decades have seen significant improvements in neonatal therapies, they have also witnessed the continuation of many questions about the appropriateness of treatment in certain cases. Neither governmental regulations (ie, Baby Doe regulations) nor court decisions have resolved these moral and ethical questions.
Future debate about selective nontreatment of damaged or disabled infants should consider current physician attitudes, decision making, and actual clinical practices. Our study, therefore, provides important information suggesting that most deaths among hospitalized newborns result from neonatologists' decisions to limit treatment. Our study further suggests that such decisions to limit treatment are not infrequently based on the physician's concern about the infant's quality of life. Further studies should prospectively examine the relative importance of family attitudes, physician beliefs, professional guidelines, and governmental regulations in shaping neonatologists' decisions and practices of limiting treatment for sick or damaged infants.
This work (S.N.W.) was supported in part by Perinatology Training Grant HD-07162 from the National Institutes of Health.
We greatly acknowledge the following individuals for their invaluable contributions to this article: Drs Roderic Phibbs, William Taeusch, and John Luce for their advice and review of the manuscript; Mureen Schlueter for assistance with the computerized database; Cory Fergus for support with computer-generated graphics; and Dr Joseph Kitterman for encouragement in the development of this study.
- Received November 10, 1995.
- Accepted February 22, 1996.
Reprint requests to (S.N.W.) 333 E Superior St, Suite 404E, Chicago, IL 60611.
- ICN =
- intensive care nursery •
- UCSF =
- University of California, San Francisco •
- ECMO =
- extracorporeal membrane oxygenation •
- ELBW =
- extremely low birth weight •
- ICH =
- intracranial hemorrhage •
- NEC =
- necrotizing enterocolitis •
- HIE =
- hypoxic-ischemic encephalopathy •
- DNR =
- do not resuscitate
- Department of Health and Human Services
- Corpuz M,
- Lee K,
- Smith L,
- et al.
- Corpuz M,
- Lee K,
- Khooshnood B
- Sanders M
- ↵People of the State of Michigan v Gregory Messenger, Ingham County Circuit Court, Lansing, MI, Docket 94-67694FH, February 2, 1995
- ↵State of Georgia v Eva D Carrizales, Clayton County Superior Court, Jonesboro, GA, Docket 93-CR-01707-06, November 1, 1994
- ↵Ameli N, Williams RL. 1985–89 Maternal and Child Health Database. Statistical Appendix. Santa Barbara, CA: Community and Organization Research Institute, University of California-Santa Barbara; 1993
- Pearlman SA,
- Bristow K
- Reale MR,
- Dweck H
- Mendoza JC,
- Campbell K,
- Chance G
- Goldworth A,
- Stevenson DK
- Robertson JA
- Byrne P,
- Laing L,
- Tyebkhan J
- Kuhn S,
- Byrne P,
- Tyebkhan J,
- et al.
- Whitelaw A
- Wall SN,
- Partridge JC
- Copyright © 1997 American Academy of Pediatrics