OBJECTIVES. To assess (1) the differences in attitudes toward disabilities of 3 groups of subjects involved in neonatal care decision-making: health care workers, mothers of term infants, and parents of preterm infants, and (2) the impact of subject characteristics on these attitudes including parental education level, religion, and severity of disability.
METHODS. Five hundred seventy-eight subjects were interviewed, of whom 135 were health care workers, 155 were mothers of term infants, and 288 were parents of preterm infants. We assessed between-group differences of the (1) health state ranking permutations, (2) proportion of subjects willing to save the infant at all costs, (3) pivotal risks of each group to the 5 health states comprising varying combinations of mental, physical, and social disabilities, and (4) impact of personal characteristics on decision-making.
RESULTS. Most health care workers ranked the most severe disability health state as worse than death, whereas most parents of preterm infants ranked death as the worst outcome. Significantly more parents of preterm infants chose to save the infant at all costs, compared with mothers of term infants and health care workers, and the pivotal risks of the parents of preterm infants group were highest for all health states. Religious worship was associated with increased probability of saving at all costs and pivotal risks. Increased severity of neonatal complications was associated with decreased probability of saving at all costs and pivotal risks in parents of preterm infants.
CONCLUSIONS. Our findings suggest that parents of preterm infants as a group were most likely to save the infant at all costs and prepared to tolerate more severe disability health states. However, personal characteristics, in particular religious belief and severity of neonatal complications, have overriding influence on these attitudes. We suggest inclusion of experienced parents of preterm infants for more effective counseling of parents in making life-and-death decisions.
Advances in neonatal intensive care have led to substantial improvements in morbidity and mortality.1–3 Such benefits, however, have also increased the chance of survival of critically ill infants who could subsequently develop serious adverse neurodevelopmental outcomes.3,4 Whether intensive care support of all preterm infants, especially those with clinical and radiologic evidence of severe central nervous system complications, is justified is highly controversial.5–7 Decisions on whether to aggressively resuscitate extremely preterm infants who are likely to survive the neonatal period depends not only on the predicted long-term prognosis, but also on the prevailing values of society and personal beliefs of parents as well as health care workers (HCWs).2,8,9 Although there are many studies that provide the neonatologist with an evidence base on which to predict the long-term prognosis of preterm infants,1–4,8–11 few studies have investigated the attitudes of parents toward long-term adverse physical and neurocognitive health states.12,13 Parents' long-term attitudes toward disability states can be very different from their preferences in the acute situation. With increasing parental participation in routine daily care and, more importantly, life-and-death decision-making,14,15 reliable information on attitudes of parents faced with severely disabled children is essential. The first objective of our study was to use pivotal risk and health state ranking methods12 to assess the differences in attitudes between HCWs, mothers of newborn term infants (MTs) and parents of preterm infants (PPs) of <32 weeks' gestation and birth weight of <1500 g in life-and-death decision-making. The second objective was to study the impact of their personal characteristics on these decisions including parental education level, religion, and severity of disability. Understanding the values of the different groups of subjects and how personal characteristics could influence the life-and-death decisions would permit reevaluation of the counseling process and assist in redefining the ideal composition of the counseling team.
SUBJECTS AND METHODS
Three groups of subjects were interviewed: HCWs (pediatric doctors and nurses), MTs, and PPs (see the detailed subject inclusion criteria and group characteristics, which are published as supporting information at http://pediatrics.aappublications.org/cgi/content/full/123/6/1501). Subjects were excluded if they: (1) did not understand the purpose of the interview (eg, because of language barrier); (2) did not understand the use of simple proportions and probability; or (3) had a history of psychiatric illness or were currently taking psychoactive medications. For MTs, those who were already mothers of preterm infants (MPs) or had an infant who had been admitted to the NICU were also excluded. All subjects were recruited between April 2005 and July 2007 at the Prince of Wales Hospital, Hong Kong, a university-affiliated teaching hospital with a level III NICU.
Each subject was interviewed by a trained research nurse (Ms Liu) using a standard questionnaire. The research nurse rehearsed with 2 senior investigators (Professor Ng and Dr Wong) until her technique was standardized and consistent, and she could conduct the interview in layman's terms. Demographic data collected included information on the subject's gender, age, marital status, number of children, level of education, occupation, religious belief, frequency of religious worship, and for pediatric staff, the level of experience in caring for newborns. The full interview took 45 to 60 minutes to complete and comprised 4 sections: (1) testing the understanding of probability and risk; (2) ranking of the 5 different health states comprising varying degrees of disabilities (Table 1) and death in ascending order of perceived severity; (3) asking whether the subject would save the infant at all costs regardless of the health state; and (4) determining the pivotal risk12 of those who did not opt to save at all costs (see “Questionnaire Details” and “Details on the Pivotal Risk Analysis Results,” which are published as supporting information at http://pediatrics.aappublications.org/cgi/content/full/123/6/1501).
For doctors and nurses, there were no statistically significant differences in distributions between either the best 3 or the worst 3 states, the saving-at-all-costs proportions, or the pivotal risks for each of the 5 health states, and so these 2 subgroups were combined into a single HCW group. Similarly for fathers of preterm infants (FPs) and MPs, the distributions of health state rankings for both the best 3 and worst 3 states, the saving-at-all-costs proportions, and the pivotal risks across the 5 states were not significantly dissimilar, and so these subjects were also combined into a single PP group.
Because this study used 6 health states (including death), there were a total of 720 combinations of possible rankings. However, it was expected that only a limited number of combinations would be chosen. The proportions of subjects in each group that chose the most common ranking combinations in terms of the best 3 and the worst 3 states (including death) were compared between groups by using the χ2 test (with correlation-adjustment for paired couples in PPs).
For each of the 3 groups, pairwise comparisons of the health state preferences in terms of pivotal risk were performed. Because each of the 5 health states was collected from the same individual, the Wilcoxon test was performed for each pair difference to determine the order of mean pivotal risks for each group. The linear mixed-effects model was fitted for the group of preterm parents to handle the correlation between paired couples.
Subject characteristics of each group were analyzed as variables in stepwise logistic regression analysis by using a mixed-effects model for PPs to handle the correlation between paired couples and to delineate the effects of each variable on the decision to save at all costs. The same variables were then fitted into stepwise linear regression models, with correlation-adjustment for paired PPs, to analyze the effects of these covariates on the pivotal risks of each group in each health state. All tests were 2-tailed and performed by using SAS 9.1 (SAS Institute, Inc, Cary, NC) and S-Plus 7.0 (Insightful Corporation, Seattle, WA). A P value of <.05 was taken as statistically significant.
Written informed consent was obtained from all study subjects. The study was approved by the Chinese University of Hong Kong Clinical Research Ethics Committee.
A total of 623 subjects were approached. Two HCWs (2 nurses refused participation), 9 PPs (6 refused consent, 3 had language barriers), and 34 MTs (21 refused consent, 2 had language barriers, 11 failed the probability test) were excluded. The characteristics of the 578 recruited subjects, 137 HCWs (34 pediatricians and 103 pediatric nurses), 189 MTs, and 297 PPs (210 mothers and 87 fathers), are summarized in Table 2.
Health State Ranking Analysis
The 3 most acceptable health states were predominantly 1 (mild global disability), 3 (predominant physical disability), and 4 (predominant cognitive disability), whereas the 3 least tolerable states were 2 (severe global disability), 5 (moderate global disability), and death. Figure 1 shows the distributions of ranking permutations of HCWs, MTs, and PPs. Regarding the worst 3 states (2, 5, and death), the majority of rankings were concentrated over 2 permutations: “death < 2 < 5” and “2 < death < 5.” PPs more frequently rated health state 2 to be better than death (death < 2 < 5 = 35.8% vs 2 < death < 5 = 31.6%), whereas most HCWs chose death as a better outcome than health state 2 (death < 2 < 5 = 13.3% vs 2 < death < 5 = 54.8%; Fig 2A). The correlation-adjusted χ2 test demonstrated that the groups ranked the worst states significantly differently (PPs versus HCWs, P < .01; PPs versus MTs, P < .01; HCWs versus MTs, P = .02). For the best 3 states, the majority of rankings were also concentrated over 2 permutations: “4 < 1 < 3” and “4 < 3 < 1.” Figure 2B illustrates that the 3 groups ranked the favorable health states less differently than the poor health states, with only HCWs and PPs showing a statistically significant difference (P = .03).
The proportion of participants who decided to save the infant at all costs was significantly different between groups with PPs being most likely to decide to save at all costs, followed by MTs, and then HCWs (Fig 3).
Demographic characteristics that could influence subject health state preferences were analyzed in a stepwise regression fashion (correlation-adjusted for paired PPs), and the parameters included age of mother, age of father, religious worship, father's education level, mother's education level, and number of children. For PPs, the severity-of-illness score (SIS) of the child was also assessed (see “Severity of Illness Score,” which is published as supporting information at http://pediatrics.aappublications.org/cgi/content/full/123/6/1501). Religious worship was divided into 3 categories: regular religious worship, infrequent religious worship, and no religion. However, additional subdivision into the different religions (ie, Buddhism, Christianity, and Islam) was not feasible because of insufficient sample size. Stepwise linear mixed-effects model fitting identified 4 variables that potentially influenced the decision of PPs to save at all costs (Table 3). Regular parental religious worship was positively associated with a significantly increased probability of deciding to save the infant at all costs, whereas secondary and tertiary education and increasing SIS were significantly, but inversely, associated with that decision. The only variable that influenced MTs was religious worship. Those who worshiped regularly had a probability of 0.61, which was higher than those who worshiped infrequently (0.33; P = .18) or had no religious beliefs (0.31; P = .02). Among HCWs, the only variable affecting this decision was also religious worship. In contrast, the probability to save was higher among infrequent worshipers (0.28), compared with regular worshipers (0.17; P = .36) or those with no religion (0.08; P = .04).
Pivotal Risk Analysis
Figure 4 and Table 4 summarize the comparison of pivotal risks for each group in each health state (see “Interpretation of Pivotal Risk and Comparison With the Utility Score Used by Saigal et al,” which is published as supporting information at http://pediatrics.aappublications.org/cgi/content/full/123/6/1501). The pivotal risks of PPs are highest in all states, indicating that they were more willing to tolerate a higher probability of disability before deciding for treatment withdrawal. The subject characteristics that were influential were different depending on the group analyzed (Table 5), but importantly, gestational age of the infant did not significantly affect the pivotal risks (P = .14–.88).
Our findings suggest that PPs, as a group, are more tolerant to adverse health states than both MTs and HCWs. The PP group also has the largest proportion of subjects who wish to save the infant at all costs. The pivotal risks for each health state are significantly higher in PPs than in MTs. Importantly, our analysis shows that subject characteristics, in particular SIS, can significantly reduce the expected pivotal risks of PPs for poor health states (between 2.4% and 4.8% decrease per SIS point; Table 5), whereas religious belief was associated with higher pivotal risks and thus better acceptance of potential disabilities.
The pivotal risks were consistently lower in the less favorable health states (health states 2 and 5, ie, severe and moderate global disability) across all groups. As expected, the pivotal risk variation was also found to be greater in adverse health states. Rank permutation analysis also shows that the different groups rank favorable states (ie, the best 3 states) less differently than poor health states (ie, the worst 3 states). This can be explained by the much wider variation in attitudes toward poor outcomes and the controversial nature of life-and-death decision-making for severely disabled infants. Our results also illustrate how different groups vary in their attitudes toward specific health states (Fig 3). In general, PPs have higher pivotal risks than the other groups across all health states. These results are comparable with a Canadian study where parents of ex-premature adolescents had higher utility scores than health care professionals for adverse health care states.13 PPs also had a significantly greater tendency to choose the severe and global disability (health state 2) over death, in contrast to HCWs where the reverse was the case. As attitudes of MTs were taken to be reflective of parents with newly born premature infants, comparisons between MTs and PPs provide valuable insight into how parents' attitudes may change as they gain the experience of providing long-term care to a disabled premature infant. The results suggest that pivotal risks of PPs are influenced by more factors in the unfavorable health states than those of MTs. In particular, SIS is an important factor, decreasing the expected pivotal risk by 4.8% and 3.1% per additional complication in severe and moderate global disability patterns (ie, states 2 and 5), respectively. Although PPs have higher pivotal risks as a group, those who had a child suffering from severe complications of prematurity were significantly less inclined to tolerate disability. The relatively high pivotal risks associated with PPs might have reflected the relatively low prevalence of complications or disability among patients in our neonatal unit.
Religious belief seems to be an important factor in the attitudes of subjects to different health states. Interestingly, active worshipers have significantly higher pivotal risks in all 3 groups, whereas infrequent worshipers tend to make similar decisions compared with nonreligious subjects. Religious belief is the only factor that can be demonstrated to significantly affect HCW attitudes. This is consistent with the view that most personal characteristics of HCWs, such as age, gender, and number of children, do not substantially influence their attitudes toward disability health states. Maternal education level seems to have opposing effects in PPs and MTs. Better educated PPs tend to have lower pivotal risks, whereas the reverse is true for MTs. This suggests that better educated mothers may believe that they are better prepared in looking after a disabled child, but with adverse child care experience become less tolerant of health states associated with severe disability. The relative uniformity of decision-making among HCWs and parents with infants of high SIS suggests that the experience of working with preterm infants and understanding the consequences of such disabilities modulate their attitudes, rendering them less prepared to accept adverse neurodevelopmental outcomes.
Among those who chose to save at all costs, there was a proportion of individuals who did not rank death as the worst possible state. Of the 19 HCWs who elected to save at all costs, 5 (26%) ranked health state 2 as the worst possible state. In PPs and MTs, the corresponding percentages were 2% and 0%, respectively. This strongly suggests that for some individuals the declaration to save at all costs is an intellectual decision and may not reflect the true attitudes toward the health state. For frontline neonatologists, this raises the important issue that although the parents involved in the life-and-death decision may state that they may wish to treat the infant at all costs, their long-term preferences to disability health states that result from treatment may not be compatible with this predominantly intellectual decision. More importantly, a substantial proportion of HCWs may also be suffering from the same intellectual dilemma. Therefore, recognizing this and rendering appropriate counseling and support is essential.
A limitation of our use of pivotal risk analysis is the inability to measure subject attitudes toward health states at the extremes of the scale. Worse-than-death health states would carry a 0% pivotal risk, whereas mild disability states may be associated with 100% pivotal risk, but in reality be less desirable than complete health (see “Interpretation of Pivotal Risk and Comparison With the Utility Score Used by Saigal et al,”13,16 which is published as supporting information at http://pediatrics.aappublications.org/cgi/content/full/123/6/1501). By using health state ranking permutation analysis in our study, the analysis of health states that are felt to be worse than death could be analyzed intuitively and served as a useful adjunct to the pivotal risk analysis of health state preferences. Another frequent criticism of this type of study is that the results may not be generalizable to other ethnic or religious groups. We chose to study the local population, because it is unique in 2 important aspects. Although Hong Kong citizens are strongly influenced by Eastern cultures, they enjoy a high standard of neonatal care comparable with the West. The difference in attitudes between HCWs and PPs to the adverse health states in our study is consistent with previous studies on Western subjects.12,13 This suggests that the acceptance of various types of disabilities may be similar across different ethnic groups and cultures. In addition, with a large proportion of the local population being atheists, we were presented with the rare opportunity to study the effects of religion, in particular Christianity, on attitudes of HCWs and parents to different disabilities (Table 2). We believe that our study makes an important contribution to the sparse literature on this important topic. Additional studies to explore the underlying reasons why the different subject groups made different decisions are warranted.
Our findings suggest that the groups responsible for neonatal life-and-death decision-making have significantly different attitudes toward disability health states and that this phenomenon exists in both Eastern and Western settings. However, multivariate analyses of the impact of factors, such as religious belief, education level, and severity of complications of the infant, show that these personal characteristics have over-riding influence. Although HCWs are generally well-versed in the expected long-term outcomes of sick preterm infants, their pivotal risks and health state rankings remain significantly different from PPs. The crucial changes in parental attitudes may be largely the result of religious belief and the experience gained from caring for a disabled child, as suggested by the substantial impact of SIS on pivotal risks (with SIS of 11 points, the expected pivotal risk could decrease by 26%–52%; Table 5), rather than from well-informed, good-quality evidence regarding the long-term neurodevelopmental prognosis of the child. Our results, therefore, challenge the conventional view that good-quality evidence and communication of such evidence to parents alone will lead to better decision-making. This implies that to facilitate decision-making that would result in a definitive course of action compatible with long-term attitudes of the parents, advice from parents who have undergone similar experiences and have prolonged experience of caring for disabled children may be as, if not more, valuable than simply supplying factual prognostic data. We suggest that PPs with appropriate experience should be involved in the counseling process so that a balanced view could be presented to parents of critically ill infants and increase the probability that they would make informed decisions that they would not later regret.
This study was supported by the Department of Pediatrics, The Chinese University of Hong Kong, Hong Kong, China.
- Accepted September 29, 2008.
- Address correspondence to Pak Cheung Ng, MD, Prince of Wales Hospital, Department of Pediatrics, 6/F, Clinical Sciences Building, Sha Tin, New Territories, Hong Kong, China. E-mail:
The authors have indicated they have no financial relationships relevant to this article to disclose.
What's Known on This Subject
Many studies provide prognostic data that enable HCWs to counsel parents of preterm infants of probable outcomes. Studies exploring long-term attitudes of parents caring for infants with disabilities are lacking.
What This Study Adds
HCWs and parents have significantly different attitudes toward disability health states, and their personal characteristics such as religious belief, education level, and severity of complications of the infant have overriding influence on these attitudes.
- ↵Blanco F, Suresh G, Howard D, Soll RF. Ensuring accurate knowledge of prematurity outcomes for prenatal counseling. Pediatrics.2005;115 (4). Available at: www.pediatrics.org/cgi/content/full/115/4/e478
- ↵Wilson-Costello D, Friedman H, Minich N, Fanaroff AA, Hack M. Improved survival rates with increased neurodevelopmental disability for extremely low birth weight infants in the 1990s. Pediatrics.2005;115 (4):997– 1003
- Silverman WA. Compassion or opportunism? Pediatrics.2004;113 (2):402– 403
- ↵Wall SN, Partridge JC. Death in the intensive care nursery: physician practice of withdrawing and withholding life support. Pediatrics.1997;99 (1):64– 70
- ↵Dear PRF. Attitudes to the risks of adverse neonatal outcome. Contemp Rev Obstet Gynaecol.1989;1 :207– 213
- ↵Patrick DL, Erickson P. Health Status and Health Policy: Quality of Life in Health Care Evaluation and Resource Allocation. New York, NY: Oxford University Press; 1993
- Copyright © 2009 by the American Academy of Pediatrics