Background. Neonatologists are criticized for overtreating extremely premature infants who die despite invasive and costly care. Withholding resuscitation at delivery has been recommended as a way to minimize overtreatment. It is not known how decisions to forgo initiating aggressive care are made, or whether this strategy effectively decreases overtreatment.
Objective. To identify whether physicians' or parents' preferences primarily determine the amount of treatment provided at delivery, to examine factors associated with the provision of resuscitation, and to assess whether resuscitation at delivery significantly postpones death in nonsurvivors.
Methods. We evaluated delivery room resuscitation decisions and mortality for all infants born at 23 to 26 weeks gestation at the University of North Carolina Hospitals from November 1994 to October 1995. On the day of delivery, the attending neonatologist completed a questionnaire regarding discussion with the parents before delivery, the prognosis for survival estimated before delivery, the degree of certainty about the prognosis, parents' preference for the amount of treatment at delivery, and the degree of influence exerted by parents and physicians on the amount of delivery room treatment provided. Medical records were reviewed for demographics and hospital course.
Results. Thirty-one of 41 infants were resuscitated (intubation and/or cardiopulmonary resuscitation) at delivery. Resuscitation correlated with increasing gestational age, higher birth weight, estimated prognosis for survival ≥10%, and uncertainty about prognostic accuracy. Physicians saw themselves as primarily responsible for delivery room resuscitation decisions when the parents' wishes about initiating care were unknown, and as equal partners with parents when they agreed on the level of care. When disagreement existed, doctors always thought parents preferred more aggressive resuscitation, and identified parents as responsible for the increased amount of treatment at delivery. Twenty-four infants died before hospital discharge. The median age at death was 2 days when physicians primarily determined the amount of treatment at delivery, 1 day when parents primarily determined the amount of treatment, and <1 day when responsibility was shared equally. The median age at death was <1 day when physicians and parents agreed about the preferred amount of treatment at delivery and 1.5 days when they disagreed. The median age at death was <1 day when parents' preferences were known before delivery and 4 days when parents' preferences were unknown.
Conclusions. Physicians resuscitated extremely premature infants at delivery when they were very uncertain about an infant's prognosis or when the parents' desires about treatment were unknown. When parents' preferences were known, parents usually determined the amount of treatment provided at delivery. Resuscitation at delivery usually postponed death by only a few days, decreasing prognostic uncertainty and honoring what physicians perceived were parents' wishes for care, without substantially contributing to overtreatment.
- extremely low birth weight infants
- neonatal intensive care
- treatment decisions
Extremely premature infants often die, or survive with severe handicaps, despite receiving invasive and costly medical care.1–6 Neonatologists have been criticized for overtreating these infants, generating huge expenses and causing unnecessary suffering while merely postponing death or prolonging lives with devastating disabilities.5,7–11 Making decisions about treating extremely premature infants, however, is a complex process, with components attributable to physicians and parents. Frequently, physicians blame parents7,,12,13 and parents blame physicians7,9–11,14 for the overtreatment of these infants.
Some commentators have suggested that withholding resuscitation from extremely premature infants at delivery is an effective and humane way to minimize such overtreatment.2,5–815–17 Underlying this suggestion is the belief that the prognosis for extremely premature infants as a group is bad enough to warrant withholding life support from all of them, and the fear that the initiation of treatment is the point of no return, after which parents and medical staff become so attached to an infant that they cannot recognize or withdraw ineffective treatment.18 Such fears are backed by graphic stories and anecdotes,7,9–11 painful personal experiences with individual infants, and reports that intensive neonatal care merely postpones death in extremely premature infants without altering the ultimate outcome.19–21
Although numerous outcome studies have reported that some extremely premature infants are allowed to die without resuscitation in the delivery room,2,,3,621–25 we are told little about the rationale for these decisions to withhold treatment. Because there has been almost no systematic exploration about how delivery room resuscitation decisions for extremely premature infants are made, we do not know whose preferences (the obstetrician's, pediatrician's/neonatologist's, parents', or others') take precedence in determining the amount of treatment an infant receives at delivery, and little about what factors other than birth weight or gestational age are associated with the provision of resuscitation. It is also not clear whether or not resuscitation at delivery significantly postpones death in nonsurvivors and leads to overtreatment.
To evaluate how decisions about treatment of extremely premature infants are made, this study examined factors associated with the provision or withholding of resuscitation at delivery. We attempted to identify who primarily determined the level of care provided at delivery, and assessed whether or not resuscitation at delivery altered the time of death for infants who died. It was our hypothesis that parents' preferences regarding the initiation of intensive care would correlate more strongly with the amount of treatment provided at delivery than would physicians' preferences, and that uncertainty about an extremely premature infant's prognosis would be associated with the provision of resuscitation at delivery.
We performed a longitudinal cohort study in which all infants born alive at 23 to 26 weeks gestation at the University of North Carolina Hospitals from November 1994 through October 1995 were monitored for treatment received, decisions to limit care, and outcome. Gestational age was determined by best obstetric criteria, or by modified Ballard examination26 (performed on admission to the neonatal intensive care unit [NICU]) if there was a greater than 2-week discrepancy between the two methods. The attending neonatologist (or neonatal fellow, when acting as attending) completed a questionnaire after delivery on the day of birth (see Appendix). The questionnaire included the infant's prognoses for survival and quality of life estimated before delivery, communication between the neonatologist and parents regarding preferences about resuscitation at delivery, and the amount of influence exerted by the obstetrician, neonatologist, parents, and others on the amount of delivery room treatment provided. Hospital charts were reviewed for demographics, maternal history, delivery room treatment, and subsequent hospital course.
Resuscitation at delivery was defined as intubation and/or cardiopulmonary resuscitation in the delivery room. The categories on the questionnaire of “limited resuscitation with more treatment if good response” and “full resuscitation,” were designed to further refine attitudes about the extent of treatment to be provided at delivery. Both were coded as preferences for resuscitation, but different choices for limited or full resuscitation were coded as disagreements about the desired amount of treatment. Comfort care was defined as care short of intubation, including warming, drying, stimulating, or providing blow-by oxygen. For the purposes of this study, bag-mask ventilation was not defined as resuscitation if it was not accompanied by endotracheal intubation. Nasal continuous positive airway pressure was not available in the delivery room. It was standard practice to administer exogenous surfactant to all mechanically-ventilated infants of <27 weeks gestation who were admitted to the University of North Carolina Hospital's NICU. Survival was defined as living to hospital discharge.
Categorical variables were analyzed with χ2tests (or Fisher's exact test when cell sizes were small) and continuous variables were analyzed with t tests, assuming normality. Parallel analyses using the Wilcoxon rank sum test for nonparametric samples yielded the same results. Because physicians circled their answers on the questionnaire, these responses were analyzed as discrete variables.
As this study was observational except for physician surveys, the requirement for informed consent was waived by University of North Carolina Hospital's Institutional Review Board.
Forty-two infants were born alive at 23 to 26 weeks gestation during the study period. One infant who was not enrolled within 24 hours of delivery was excluded. The remaining 41 infants constituted the study cohort, including 7 infants born at 23 weeks gestation, 12 infants at 24 weeks gestation, 14 infants at 25 weeks gestation, and 8 infants at 26 weeks gestation. Twenty-three (56%) of the mothers were black and the rest were white. Seventeen were married, 22 were single, and 2 were of unknown marital status. Eighteen (44%) had private insurance, and 39 (95%) had received some prenatal care. Sixteen (39%) of the deliveries were by cesarean section.
Forty of the 41 deliveries were attended by a neonatologist or neonatal fellow acting as attending, at the request of the obstetrician. The six neonatologists or neonatal fellows who attended deliveries of extremely premature infants during the study period had been practicing medicine from 4 to 17 years. All but one had children of their own. Four stated that they were religious. Five withheld treatment from at least 1 infant in the study and all six indicated that they preferred to provide comfort care only to some infants.
These physicians completed delivery questionnaires on 39 infants. The 2 infants without physician questionnaires were included only in the analyses for which the chart review provided data.
Factors Associated With Resuscitation at Delivery
Thirty-one infants (76%) were resuscitated at delivery. Intubation was accomplished in the delivery room for 30. One infant who became apneic in transit from the delivery room was intubated immediately on entry to the NICU, and was coded as receiving resuscitation at delivery. Resuscitation was associated with increasing gestational age (29% at 23 weeks, 67% at 24 weeks, 93% at 25 weeks, 100% at 26 weeks), and higher birth weight (median birth weight, 735 g; range, 395–1015 g, for resuscitated infants vs 588 g; range, 425–715 g for nonresuscitated infants; P = .007). Also associated with resuscitation at delivery were better estimated prognoses for survival and quality of life, greater uncertainty on the part of physicians about the accuracy of their prognoses, and younger maternal age (Table 1).
The provision of resuscitation at delivery was associated with a preference by either the physician or the parents that an infant be treated. By contrast, the withholding of resuscitation at delivery was associated with parents', but not with physicians', preferences for comfort care only. Whereas treatment was invariably withheld when parents desired comfort care only, resuscitation was provided in 50% of the cases in which physicians preferred comfort care only.
Factors not associated with resuscitation at delivery were maternal race, marital status, insurance status, religion, history of infertility, route of delivery, and 1-minute Apgar score. Neonatologists in this study did not consider obstetric treatment of labor and delivery or the influences of other third parties to significantly affect the amount of delivery room treatment they provided. Our sample was not large enough to determine whether the identity of the attending physician was associated with the provision of resuscitation.
Physician/Parent Preferences and Decision-making
The amount of treatment that parents wished their infant to receive was known by the neonatologist before delivery for 27 of 39 of these extremely premature infants (69%). In general, parents preferred more aggressive treatment than did doctors for infants at gestational ages <26 weeks; at 26 weeks, preferences of physicians and parents for resuscitation were nearly identical (Fig 1).
When parents' wishes were known before delivery, physicians and parents agreed about the preferred amount of treatment for 16 of the 27 infants (59%) (Table 2). Resuscitation (full or limited) was preferred for 10 of these infants (62%). Nine received full resuscitation at delivery and 1 received positive pressure ventilation with bag and mask but was not intubated. The 6 infants for whom physicians and parents agreed that comfort care only was preferable were not resuscitated at delivery.
Parents and physicians disagreed about the preferred amount of treatment for 11 of the 27 infants (41%) for whom parents' wishes were known before delivery. Disagreement was more frequent for infants at earlier gestational ages. In all cases of disagreement, the parents preferred more treatment than did the physician. Nine of these 11 infants (82%) were resuscitated at delivery (including 6 for whom the physician would have preferred to provide comfort care only) and 2 were not.
Parents' wishes about resuscitation were unknown before delivery for 12 infants (31%). Eleven of these (92%) were resuscitated. Three of them were resuscitated although the doctor would have preferred to provide comfort care only.
Physicians attributed primary responsibility for resuscitation decisions differently depending on their communication and accord with parents. When doctors and parents agreed on the preferred amount of treatment, responsibility was most often seen as shared equally between physicians and parents. When doctors and parents disagreed, parents were usually seen as primarily responsible for the amount of treatment the infant received. When parents' wishes were unknown, physicians saw themselves as primarily responsible for the resuscitation decision.
Mortality in this cohort was 59% (24/41). Mortality by gestational age was 71% at 23 weeks, 83% at 24 weeks, 50% at 25 weeks, and 25% at 26 weeks. Resuscitation at delivery was withheld from 10 infants (24% of the cohort). All of these infants died. The other 14 nonsurvivors died after having been resuscitated at delivery (34% of the cohort). Treatment was withdrawn from 12 of these infants. The remaining 2 infants died while still receiving mechanical ventilation. In all, treatment was withheld from 42% of the nonsurvivors and withdrawn from 50%. Figure 2 shows the treatment status and outcome of these infants by gestational age.
Mortality was highest for those infants whose parents and physicians disagreed about the preferred amount of treatment at delivery (Table 2). The proportion of infants who were resuscitated at delivery but ultimately did not survive was greatest when parents were seen as primarily responsible for the decision to resuscitate, intermediate when doctors were seen as primarily responsible, and lowest when decision-making was shared equally.
The ages at which the infants died are shown in Fig 3. Twenty of the 24 nonsurvivors (83%) died in <1 week. Infants who were resuscitated at delivery died at a median age of 2 days, whereas those who were not resuscitated at delivery died at a median age of <1 day, on the day of birth (P = .059) (Fig 3A). The median age at death was 2 days when physicians identified themselves as primarily responsible for deciding whether or not to resuscitate, 1 day when parents were viewed as primarily responsible, and <1 day when decision-making responsibility was seen as shared equally (P = .89) (Fig 3B). Nonsurvivors died at a median age of <1 day when physicians and parents agreed about the preferred amount of treatment at delivery and 1.5 days when they disagreed (P= .57). When the physician knew before delivery how much treatment parents desired (no matter whether they agreed about resuscitation at delivery or not), the median age at death was <1 day, versus 4 days when the physician did not know the parents' preference before delivery (P = .29) (Fig 3C).
Three infants (12.5% of the nonsurvivors) died after 30 days of age. Primary responsibility for the decision to resuscitate these infants was the physician's in 1 case, the parents' in 1 case, and shared equally in 1 case. Treatment was withdrawn from 2 of these infants and the third died while still receiving mechanical ventilation. (The only other nonsurvivor from whom life-sustaining treatment was not withdrawn died at 2 days of age.)
Physician Prognostication and Decision-making
Overall, the study physicians' prognoses for survival were relatively accurate. No infant given a poor prognosis (<10% chance of survival) survived, 40% (6 of 15 infants) given a fair prognosis (10%–40% chance) survived, 56% (5 of 9 infants) given a moderate prognosis (40%–60%) survived, and all infants (4 of 4) given a good prognosis (60%–90%) survived. No infant was assigned an excellent prognosis (>90% survival) at birth. Self-fulfilling prophecies may have been responsible for some of the prognostic accuracy, however, because only 45% of infants given a poor prognosis were resuscitated, compared with 100% given a good prognosis.
Physicians' prognosticating was not appreciably more accurate when they felt fairly certain of their prognoses than when they did not. Fourteen of 15 infants who were given poor or fair prognoses died when the physician was relatively certain about the prognosis (certainty ranked greater than 3 on a 5-point scale), compared with 3 of 4 when the physician was relatively uncertain about the prognosis (certainty ranked below 3). All infants who were given good prognoses survived, regardless of the physician's certainty about the prognosis.
Figure 4 shows physicians' prognoses for infants' quality of life in relation to their prognoses for survival. There was a notable similarity between these two prognoses (correlation coefficient R = 0.84;P = .0001).
Twenty-nine of 39 infants (74%) received the amount of treatment at delivery that the physician believed was preferable. When the amount of resuscitation coincided with the physician's preference, fewer infants for whom treatment was initiated ultimately died (35% vs 78%;P = .05). Infants who were initially resuscitated in accordance with the physician's preference but who later died had a higher median age at death than resuscitated nonsurvivors from whom doctors would have preferred to have withheld treatment (6 days vs 2 days), although this difference was not statistically significant (P = .59).
Studies evaluating outcomes of extremely premature infants have reported mortality rates ranging from 25% to 100%, with approximately one-third to one-half of survivors exhibiting severe morbidities.2,,4,6,21,2227–31 Because the personal and economic costs of ineffective care are great, life-sustaining medical treatment is not always initiated at extremely premature births.2,,3,621–25 Although weighty judgments are being made around the time of delivery, the few studies evaluating these resuscitation decisions have been surveys that posed hypothetical questions,13,,15,32 rather than examinations of actual behavior at the births of extremely premature infants. The only report we located that extensively analyzed factors associated with the initiation of mechanical ventilation in this population was not limited to decisions made at delivery.6 That study found that infants from whom mechanical ventilation was withheld had lower gestational ages and birth weights, were less likely to have received antenatal steroids, and had lower 1-minute Apgar scores.
In our examination of decision-making about treatment at delivery, we found a pattern of resuscitation that suggests the following:
When physicians are very uncertain about an infant's prognosis they tend to resuscitate at delivery. Uncertainty must be lessened, but need not be completely resolved, for treatment to be foregone. Of note, physicians felt that the prognosis was virtually certain at birth for only 3 infants in this study.
When parents' desires are known before delivery, parents' preferences usually determine the amount of treatment provided at delivery.
When parents' desires are unknown, physicians most often resuscitate infants at delivery.
Resuscitation at delivery generally postpones death by only a few days. As such, it likely does not contribute substantially to overtreatment of extremely premature infants.
Not surprisingly, we found that infants who were thought to have better prognoses for survival were more apt to be resuscitated. We found that physicians' estimates for quality of life and for survival were remarkably similar, implying that doctors may conflate these two outcomes in prognosticating for extremely premature infants. As the proportions of infants who die and who suffer severe morbidity at various gestational ages are rarely identical in published outcome data, this conflation may be a generalization or it could indicate a mistaken impression on the part of physicians. If these two prognoses are combined when physicians counsel parents, some misleading impressions about anticipated outcomes may be conveyed. For some parents, clearly distinguishing between these two outcomes (death and suffering) may affect decisions about resuscitation.
The preferences of both physicians and parents significantly determined whether or not resuscitation was provided. We found that neonatologists and parents frequently disagree about the preferred amount of resuscitation for extremely premature infants at delivery, with parents generally wanting more intensive treatment than do physicians. Despite disagreement, doctors usually honored parents' wishes for resuscitation, even when they preferred to withhold treatment. When physicians did not know parents' preferences before delivery, infants were almost always resuscitated, including infants whom doctors would have preferred not to resuscitate. This suggests that physicians in this study believed that parental input in the decision-making process was a value worth preserving.
A recent survey of neonatologists indicates that the absence of predelivery counseling is not unusual.13 Despite a policy at our hospital to counsel all mothers who are at imminent risk of delivering extremely premature infants in advance of delivery, in nearly one-third of these births predelivery counseling did not occur. Our findings invite the speculation that simple physician-parent communication before the delivery of an extremely premature infant—whether or not there is initial disagreement about the preferred level of care—might decrease the amount of overtreatment.
In this cohort of extremely premature infants, resuscitation at delivery usually postponed death by only a few days. The median age at death did not differ significantly whether parents or physicians were primarily responsible for the decision to resuscitate or whether they agreed or disagreed about resuscitation at delivery. In most cases, death occurred after life-sustaining treatment was withdrawn. This suggests that greater certainty about prognosis and agreement between physicians and parents were usually achieved quickly and acted upon. This study did not, however, take account of deaths that occurred after hospital discharge that may have been attributable to complications of prematurity. As such, the amount of overtreatment may be underestimated.
Because parents preferred substantially more treatment initially than did doctors, and a greater proportion of infants died after being resuscitated when parents requested treatment despite doctors' disagreement, it is tempting to conclude that parents are the driving force behind most of the excessive treatment of extremely premature infants. However, several important caveats must be made in this regard.
First, parents' preferences were not surveyed directly in this study, but were relayed by the attending neonatologist. We did not administer a parent questionnaire at delivery because we believed that posing questions about limiting treatment, in the context of an impersonal research method without accompanying in-depth counseling, could be confusing or hurtful to families in emotional crisis. It is possible that the neonatologists interpreted parents as wanting more treatment than they actually did. Of note, however, are several surveys using hypothetical case scenarios to elicit parents' and physicians' attitudes which found that parents usually did prefer more aggressive treatment of extremely premature infants than did physicians.15,,33
Second, counseling by the physician could have affected parents' preferences. In a study of elderly patients' preferences for cardiopulmonary resuscitation, patients' wishes were found to change after they were counseled by a physician. This was attributed primarily to providing them with new knowledge about outcomes after cardiopulmonary resuscitation.34 Because most parents are ignorant of the outcomes of extremely premature infants, counseling by the neonatologist would likely influence parents' choices about providing life-sustaining treatment. Directed counseling could make the physician's preference seem to be the parents'. Nonetheless, physicians' preferences clearly did not fully determine parents' wishes (at least regarding initiation of treatment) because disagreement about resuscitation at delivery was so frequent. It is likely that whoever (physician or parents) prefers more treatment determines the amount of treatment provided at delivery.
Third, by evaluating only live births, we did not examine obstetricians' influence on the amount of treatment provided to extremely premature infants. Insofar as obstetric counseling and treatment of premature labor affects the survival of fetuses in utero, the influence of these physicians was underestimated.
Finally, although our study accords with several recent reports that most extremely premature infants who die in NICUs do so at <1 week of age,4,,21,22,35,36 one institution has reported a near-doubling of the time to death in the past 5 years (with most infants <1000 g now dying at 3- to 6-weeks of age).37Another NICU which aggressively resuscitates and treats all extremely premature infants had only 1 of 7 nonsurvivors die at <1 week of age.27 Because parents rarely suggest foregoing care until doctors first offer nontreatment as an option,32,,38physicians must be willing to withhold or withdraw life-sustaining treatment in order that dying not be extended.
Whether postponing the median time to death by 2 days is considered overtreatment or appropriate and humane care is a complex ethical question. When considered from the perspective of resource allocation, the relative proportion of NICU resources devoted to the care of these infants is small,35,,36 and may be justified as an appropriate response to medical uncertainty and personal values. On an individual basis, overtreatment is prolongation of care that is ineffective or does more harm than good. Most people believe that the suffering that accompanies medical treatment is justified so long as the possibility of a good outcome exists to make it worthwhile. Labeling the medical care of these infants “overtreatment” or “appropriate” then depends on what one considers to be an acceptable prognostic certainty for poor outcome.
Decisions to resuscitate fueled by parents' desires to initiate life-sustaining treatment raise the question of whether ineffective treatment provided solely for the parents' sake constitutes overtreatment. This issue is complicated, however, by the fact that an infant's prognosis is very rarely certain at delivery, and that parents have the ultimate (legal, moral, and emotional) responsibility for decisions affecting their child. Honoring parents' wishes was accepted by physicians in this study to constitute appropriate rather than excessive care in most cases.
In summary, our findings suggest that the decision to resuscitate an extremely premature infant at delivery is related to uncertainty about prognosis, the physician's perception of the parents' preferences for treatment, and a value system that upholds parental decision-making. Our results imply that initiation of treatment at delivery should not be feared as the point of no return and that delivery need not be thought of as the optimum time to withhold life support. Instead, initial resuscitation may allow time for increased prognostic certainty and opportunity for joint physician-parent decision-making.
We thank Peggy McCracken, Maryellen Lane, and David Francis for their efforts in data collection, and Wayne Price and Julie Gavozov for help in preparing figures and tables for this manuscript. We also thank the attending physicians and fellows who participated in this study for their time and effort.
- Received June 1, 1997.
- Accepted February 10, 1998.
Reprint requests to (M.W.D.) Division of Neonatal/Perinatal Medicine, CB #7596, Fourth Floor, UNC Hospitals, University of North Carolina, Chapel Hill, NC 27599-7596.
- NICU =
- neonatal intensive care unit
- ↵Muraskas JK, Weiss MG, Myers TF. Neonatal viability in the 1990's: held hostage by technology. Pediatr Res. 1996:39:234. Abstract
- ↵Kolata G. Parents of tiny infants find care choices are not theirs. New York Times. Sept 30, 1991;CXLI:1. Section A
- ↵Frey D. On the border of life. New York Times Magazine. July 9, 1995:22. Section 6
- Silverman WA
- Harrison H
- ↵Rosenthal E. As more tiny infants live, choices and burdens grow. New York Times. Sept 29, 1991;CXLI:1
- ↵Chira S. New medical quandary at heart of a trial. New York Times. Aug 3, 1994. National Section
- Lee SK,
- Penner PL,
- Cox M
- Sauve R,
- Robertson C,
- Etches P,
- Byrne P
- Rhoden NK
- Sulmasy DP,
- Sugarman J
- Hack M,
- Friedman H,
- Fanaroff AA
- deRegnier R-A,
- O'Shea TM,
- Roberts D
- Barton L,
- Hodgman JE,
- Pavlova Z
- Wall SN,
- Partridge JC
- Derleth DP
- Gunkel JH,
- Banks PLC
- LaPine T,
- Jackson JC,
- Bennett FC
- Lantos J,
- Tyson J,
- Meadow W
- Saigal S,
- Burrows EA,
- Stoskopf BL,
- Rosenbaum PL,
- Streiner DL
- Meadow W,
- Reimshisel T,
- Lantos J
- Pearson EB,
- Bose CL,
- Kraybill EN
- Wright K
- Fedson A,
- Lantos J,
- Stocking C,
- Meadow B
- Copyright © 1998 American Academy of Pediatrics