Sign In
Individual Sign In
Create an Account
Institutional Sign In
OpenAthens Shibboleth
May 2004

Provider Perceptions of Child Deaths

Author Affiliations

From the Department of Community Health, Salus Center, School of Public Health (Dr Andresen and Ms Seecharan), and Department of Pediatrics, Saint Louis University (Dr Toce), and SSM Cardinal Glennon Children's Hospital (Dr Toce), St Louis, Mo.


Copyright 2004 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2004

Arch Pediatr Adolesc Med. 2004;158(5):430-435. doi:10.1001/archpedi.158.5.430

Background  The quality and context of end-of-life care for children are not as well studied as they are for adults. The components of quality care are less clear, and differences between providers' perception of the quality of death are also not well understood.

Objective  To compare nurse and physician perceptions of the quality of care and events of death of the same children.

Design  Self-administered surveys following child deaths.

Setting  Academic, tertiary care, faith-based children's hospital.

Participants  Seventy-one matched sets of physicians and nurses who had cared for the same child at the time of death.

Main Outcome Measures  Self-administered survey comprising closed-ended questions on the quality of care based on focus groups.

Results  Physicians and nurses usually agreed on what had happened for events surrounding the child's care. In general, physicians tended to have more positive views of the death; however, positive reports of the quality of death of these patients were common for all providers. Agreement was high (>75%) for the matched sets with a few exceptions, including questions that asked for a provider's own behavior or expectation (eg, expecting further contact with a patient's family). However, chance-corrected agreement (measured as κ) was low.

Conclusions  There was a good quality of care at the time of death for most patients, with minimal pain and suffering reported by providers. There were differences in perception explained by roles (physicians vs nurses). Future research should examine differences by provider experience and in other care settings.

Child deaths are not common in developed countries. However, more than 53 000 children die each year in the United States,1 with most deaths occurring in the hospital.2 As noted in recent studies,37 health care providers are inadequately trained to deal with deaths among their child patients. In addition, there is little understanding of what constitutes quality for end-of-life and palliative care for children and their families and whether it differs from the experience of adults.8 However, there is substantial professional interest in promoting higher standards of care and additional research on palliative care for children.9 Although an interdisciplinary team approach is regarded as the best palliative care model, few studies10,11 provide insight into potential differences in the perception of death by different members of the health care team. Literature regarding provider perceptions is sometimes limited to a single provider type, eg, physicians1215 and nurses,16,17 or involves specific types of deaths, like cancer.12 Although hospice care settings have been used to report provider perceptions,8,17 studies1012 on provider perceptions of the quality of death of adults for general hospital settings tend to be about generic attitudes and practices and not linked to experiences of specific deaths. Finally, there is little research that includes provider perceptions of quality of care and circumstances of child deaths.12,14 When parents are queried concerning their perspectives on quality of care, they note problems in communication, inadequate support for siblings, concerns about pain management, lack of community pediatric palliative care expertise, and inadequate bereavement support.18

Although health care education needs and palliative care planning for children comprise a developing literature, we were not aware of any available provider instruments to describe and quantify the quality of care in deaths of children. The present study used a 2-stage process to examine what health care providers thought about their palliative care skills and how they viewed the quality of care around the time of death for children. Focus groups were conducted in the first phase with health care providers who had experience caring for critically ill children.19 This phase provided the key domains for the content of the subsequent quantitative survey administered to care providers for children who died. In this article, the results of the survey in which nurses and physicians responded to questions about quality of care of children who died in the hospital are described.


The pediatric palliative care program of our tertiary care children's hospital participated in a 4-year project on end-of-life care for children. Our hospital is a private, faith-based, teaching, freestanding children's hospital with 162 staffed beds. Approximately 100 children die each year in our hospital, and 20 additional children who are our patients die at home. During 2000 and the first 6 months of 2001, children who died at the hospital or who were discharged and died at home were included in this study, in which hospital medical care providers were asked about their deaths. During 1999 and 2000, all children who died in the hospital were part of an overlapping study of cost of care, and their medical records were reviewed for demographic and diagnostic information.20


In conjunction with the routine mortality review of all hospital deaths, we conducted an assessment of care around the time of death by surveying health care providers caring for the child and family. The children were heterogeneous and included those with diverse conditions, for example, newborns and children with end-stage cancer, debilitating neurological conditions, or who had been injured. During the 18 months of the study, 142 children who had been cared for at our hospital died. We reviewed mortality and medical records to identify the primary physician and nurse who were (1) responsible for ongoing care of the child during the hospitalization and (2) present at the time of death. In some cases, we could not identify the providers who met both of these definitions and sent a survey to more than 1 physician or nurse. The packets consisted of a general informative letter about the study and a survey. In the same manner as the routine mortality review, we mailed these to providers within 1 month of the child's death. Some physicians and nurses sent us back the survey describing their limited knowledge of the child. In all cases, we continued to seek a primary provider. We sent reminder letters and a second survey to providers if we had not received a reply within a month. When we received multiple surveys for a patient, we selected the most relevant and complete survey from 1 physician and 1 nurse (present at death and self-identified primary role). In no instance were there more than 2 potential knowledgeable physician or nurse respondents for a child. For most children, a physician and nurse were available; however, for some patients, we did not receive responses from both types of providers. We present analyses for matched provider teams (by patient), and we conducted analyses for the total group of care providers who answered questions.

This study was reviewed and accepted as exempt by our university's human subjects committee (institutional review board). Although the survey was part of the review of deaths in the hospital, it was connected to the research phases of the project. To overcome concerns about potential negative consequences of self-reports on quality of care, all responses were voluntary and anonymous. We removed all health care provider identifiers from the data and did not ask about provider characteristics that might identify them (eg, age, gender, and specialty), except for their profession (nurse or physician) and years in practice. However, by definition, all providers were tertiary hospital–based health care providers and generally intensivists, as reflective of the location of death. Nursing respondents reported a mean ± SD of 13.3 ± 8.3 years (range, <6 months to 33 years) in practice, and physicians reported a mean of 13.4 ± 8.8 years (range, 1-30 years).


Surveys sent to providers were titled "Improving End-of-Life Care for Children" and introduced as part of the ongoing quality of care for the institution. Forms included multipart closed-ended questions. Draft question items were constructed based on health care provider focus group feedback. Fifteen questions were pilot tested with hospital health care providers, and the survey was refined based on their feedback. Content focused on aspects of the child's quality of care around the time of death (eg, addressing pain, physical, psychosocial, and spiritual needs and complying with the wishes of the child or family), as well as the provider's thoughts about his or her involvement and perception (eg, comfort with care of the child and whether he or she talked to the family about religious beliefs). Sixty-two variables were constructed from survey questions. Response categories for some questions were yes or no categories (eg, "Were you comfortable guiding the care of this child or family?"). Other questions asked for ranked Likert-type responses, for example, "Overall, how difficult were the physical problems for the child?" offered 5 responses ranging from "very difficult" to "not at all difficult." The full survey on quality of care is available from the authors and is on the Web.21 Because of the diversity of children dying in the hospital, some questions, such as those seeking the providers' perceptions of the child's opinion or emotional symptoms, were not relevant to all children because of age or cognitive development. In addition, some questions were not relevant for children dying suddenly (previously healthy children dying of injury). We dropped items from analysis when 5 or fewer physicians or nurses answered those items.


Primary analyses were conducted using SPSS software.22 Descriptive statistics include frequencies of responses among providers. We computed 95% confidence intervals for proportions based on methods recommended by Fleiss.23 We also compared physician and nurse responses using simple χ2 (dichotomous and nominal categories) and χ2 for trend (ranked categories) statistics. To examine the agreement on aspects of the child's death and care, we used percentage agreement. We also used κ statistics, which correct for agreement of responses that occur by chance. Because some categories were skewed (eg, most responses were yes), these κ values could be misleading and result in a value that is biased too low compared with overall agreement; these figures are provided as a general indicator of chance-corrected agreement.2426 The descriptive results of proportions for each question (eg, percentage of yes answers for each provider type) provide information about potential systematic differences in how nurses and physicians viewed the aspects of children's deaths and their role in the child's care. Finally, we compared all respondents by profession, adding physicians and nurses who were single respondents, to see if these additional providers changed the patterns of the matched pairs other than adding sample sizes to each provider type.


We received at least 1 physician response for 124 (87%) of 142 children who died during the study, and 91 questionnaires (64%) from at least 1 nurse for these children. Usable questionnaires were returned from 178 physicians and 139 nurses. For 71 children, we had physician and nurse respondents. In addition, to compare all nurses with all physicians, we included a single physician or nurse for the children who had only 1 medical care respondent.

We conducted a medical records review of the 71 children in our study. Two thirds were male and white. Twelve (17%) children were aged 13 years and older, and 24 (34%) were younger than 2 months at the time they died. Twenty (28%) children had a neonatal medical diagnosis, 13 (18%) had a neurological diagnosis, 9 (13%) had a cardiac diagnosis, and 8 (11%) died suddenly of injury. Table 1 gives descriptive results of the children in more detail. Based on the larger medical records review of all hospital deaths, the only difference in our patients for the present analysis was that fewer patients were classified as dying suddenly of injury (11% for our patient group vs 25% overall).

Table 1. 
Image not available
Characteristics Among 71 Children

There was general agreement between physicians and nurses concerning the circumstances of death. For example, there was a high degree of agreement concerning the amount of ventilatory support, use of hospice, and place of death. The responding health care providers believed overwhelmingly that pain and physical symptoms were well managed and that there was discussion about the possibility of death and treatment options. Providers skipped some questions if the question was not relevant; for example, when a death occurred at home, providers did not usually respond to questions about symptoms in the hospital, and when a death was sudden, questions about family and provider interactions were skipped. In general, physician respondents tended to answer questions about care with more positive views of the quality of care and with more assurance that the care had been appropriate and consistent with the child's and family's preferences. For example, 91% of physicians said they were comfortable guiding the care of the child compared with 87% of nurses, and 94% of physicians said the child was "at peace" during his or her last days of life compared with only 72% of nurses. Slightly more physicians than nurses thought symptom management had been successful in the hospital and thought the health care team and family agreed on the plan for the child. However, there were several aspects of the child's care in which nurses were more involved than physicians or gave more positive responses about the care and death of the child. Nurses were much more likely than physicians to talk to parents about informing siblings about the possibility of the patient's death (86% vs 49%) and somewhat more likely to think that the family's and child's wishes had been followed in the care plan (48% vs 34%). They also reported that they talked more to families about religious or spiritual beliefs (83% vs 59%). Agreement was high (>75%) for the matched sets with a few exceptions. For questions in which nurses and physicians were answering for their own conduct (eg, further contact with a patient's family), these lower levels of agreements were understandable. Questions with multiple response categories tended to have much less exact agreement but were not significantly different in trend analysis. κ Statistics were low except for a few instances (eg, κ = 0.66 for reporting where the child died and κ = 0.62 for agreement on if the child had a peaceful death). Some of the low κ agreements could be explained by skewed responses (most answering yes or no) or by questions in which the 2 respondents were answering about their own behavior. Most nurses and physicians reported that they were comfortable guiding care; the agreement was 78%, although κ = −0.12. Table 2 gives these results in more detail. None of the differences between physician and nurse respondents were statistically significantly different. When we included all nurses and physicians (regardless of patient matching), these patterns were virtually identical (data not shown). The only difference was that there were slightly smaller differences between provider types. For example, on the question asking providers if the child was at peace during his or her last days, the responses were affirmative for 94% of physicians and 72% of nurses in the matched analysis, and 94% and 85% for the total groups of health care providers.

Table 2. 
Image not available
Perceptions of Care and Quality of Death Among Physicians and Nurses Matched by Child*

Most health care providers in this study reported that they were comfortable in guiding the care of children at the end of life, including providing anticipatory guidance. They also had positive reports of the adequate pain and physical symptoms management in the hospital. In case of differences in the perception of the care, physicians generally reported a more positive quality of care around the time of death than did nurses. We are unaware of other reports on the difference in care providers' perception of the aspects of quality of care and experiences at the time of specific child deaths. However, most research shows that physicians and nurses are concerned about symptoms management, especially pain, as was the case with our study.1012,16,17 In addition, we found there were differences in the role of providers at the end of life. Nurses were more likely to be involved in talking with families about sibling, psychosocial, and religious issues, while physicians were more often involved in talking to families about decision making. These differences are likely to reflect usual team roles for hospital providers. These role differences and perceptions persisted, but were smaller, when we considered the perceptions of providers who were not matched by describing the same patient. This suggests that some random error or "noise" is introduced when gathering provider perceptions in general that are not based on a single patient. Although addressing psychosocial, emotional, and spiritual needs has traditionally been the role of nurses and social workers, there is increasing recognition of the importance of physicians' skills in providing such support for the family.14 The American Academy of Pediatrics, in their recent statement on palliative care for children,9 and other pediatric palliative care specialists7 have recommended that all physicians caring for children be taught palliative care skills, including management of pain, psychosocial, emotional, and spiritual symptoms at the end of life for children and their families.

The results are limited somewhat in their generalizability because of the setting and because these surveys occurred during the early phase of a project intended to increase the quality of end-of-life care for children. However, tertiary care teaching hospitals are common locations for children receiving the specialty care required for life-limiting conditions. In addition, the high level of comfort with providing pain and symptom management by these health care providers may be better than that in prior reports,27 in part because of the ongoing in-service education provided by the program. Our hospital also is faith-based, and the high percentage of health care providers who dealt with spiritual care issues for patients and their families may reflect this setting. Finally, not all survey questions were completed by all providers. Although we are aware of several case-specific reasons for their nonresponse (sudden deaths or deaths at home and age or cognitive limits of children), providers did not explain their lack of response and often did not check a response box as "not applicable." More in-depth questions, and possibly personal interviews, may be needed in the future to fully understand provider experiences with the heterogeneous patients represented among children who die.

Future research should examine if differences are linked to other provider characteristics, for example, training and gender. However, in their qualitative study concerning adult deaths, Steinhauser and colleagues11 concluded that professional role is more important than provider gender or ethnicity. Confirmation of provider characteristics, other than profession, should also be subjected to quantitative investigation. If professional training and role continue to predict perception of death, it may suggest that professional training need to pay closer attention to these differences. A positive report on the quality of care at the time of pediatric patient deaths is a goal, but these positive perceptions must be valid. Finally, parent perceptions need to be incorporated into research on the quality of care for children who die. Other investigations have found discrepancies between parents' perceptions and physicians' written reports of symptoms, with parents being more likely to report symptoms of fatigue, poor appetite, constipation, and diarrhea.27 However, there is inherent difficulty in interviewing parents in a similar time frame and with as high a response compared with providers,10 which adds to the complexity of this kind of research on the quality of death. Our research included a sample of parent interviews, but these were scheduled no earlier than 6 months following the death, and there were few cases (n = 6) that overlapped with surveys of provider perceptions. Increasing the research questions to involve families is important, but the pragmatic aspects of high burden of research and potential low response among bereaved family members and incomplete response of providers themselves may require a large multicenter effort for adequate sample size. In addition, the resulting multirespondent sets would not represent all cases. Nonetheless, the mandate to provide the highest standard of care to dying children and their families is clear and deserves continued attention.

What This Study Adds

Although quality of end-of-life care has been studied in adults, there has been little research on children's deaths. In addition, differences in the health care providers' perceptions of the quality of care at the time of a child's death are not clear. We found that most of the providers reported a good quality of care, with some differences explained by the roles of the physicians and nurses. Additional research, including the experience of the provider, may provide further insight into the providers' perceptions.

Back to top
Article Information

Corresponding author and reprints: Elena M. Andresen, PhD, Department of Community Health, Salus Center, School of Public Health, Saint Louis University, Suite 300, 3545 Lafayette Ave, St Louis, MO 63104 (e-mail:

Accepted for publication September 17, 2003.

This study was funded, in part, by grant 035489 from the Robert Wood Johnson program in Promoting Excellence in End-of-Life Care, Missoula, Mont; by administrative support of the SSM Cardinal Glennon Children's Hospital; and by grant U48CCU710806 from the Centers for Disease Control and Prevention Research Center at Saint Louis University.

We thank the many health care providers who gave their valuable time and experiences to this work. Our special thanks to our FOOTPRINTS team members: Mary Ann Collins, MHA, Judy Carron, RN, Dennis O'Connor, MD, Patricia Codden, RN, PNP, Upasana Nanda, MPH, Kim Pacatte, RN, Lisa Griebel, Drew Gerald, and Jenna Putzel.

Hoyert  DLArias  ESmith  BLMurphy  SLKochanek  KD Deaths: final data for 1999. Natl Vital Stat Rep. 2001;491- 113
Feudtner  CSilveira  MJChristakis  DA Where do children with complex chronic conditions die? patterns in Washington State, 1980-1998. Pediatrics. 2002;109656- 660
Charlton  R Medical education: addressing the needs of a dying child. Palliat Med. 1996;10240- 246
Frager  G Pediatric palliative care: building the model, bridging the gaps. J Palliat Care. 1996;129- 12
Khaneja  SMilrod  B Educational needs among pediatricians regarding caring for terminally ill children. Arch Pediatr Adolesc Med. 1998;152909- 914
Papadatou  D Training health professionals in caring for dying children and grieving families. Death Stud. 1997;21575- 600
Sahler  OJFrager  GLevetown  MCohn  FGLipson  MA Medical education about end-of-life care in the pediatric setting: principles, challenges, and opportunities. Pediatrics. 2000;105575- 584
Low  JTSPayne  S The good and bad death perceptions of health professionals working in palliative care. Eur J Cancer Care (Engl). 1996;5237- 241
Committee on Bioethics and Committee on Hospital Care, American Academy of Pediatrics, Palliative care for children. Pediatrics. 2000;106351- 357
Steinhauser  KEChristakis  NAClipp  ECMcNeilly  MMcIntyre  LTulsky  JA Factors considered important at the end of life by patients, family, physicians, and other care providers. JAMA. 2000;2842476- 2482
Steinhauser  KEClipp  ECMcNeilly  MChristakis  NAMcIntyre  LTulsky  JA In search of a good death: observations of patients, families and providers. Ann Intern Med. 2000;132825- 832
Hilden  JMEmanuel  EJFairclough  DL  et al.  Attitudes and practices among pediatric oncologists regarding end-of-life care: results of the 1998 American Society of Clinical Oncology survey. J Clin Oncol. 2001;19205- 212
Saunderson  EMRidsdale  L General practitioners' beliefs and attitudes about how to respond to death and bereavement: qualitative study. BMJ. 1999;319293- 296
Bartel  DAEngler  AJNatale  JEMisra  VLewin  ABJoseph  JG Working with families of suddenly and critically ill children: physician experiences. Arch Pediatr Adolesc Med. 2000;1541127- 1133
Redinbaugh  EMSullivan  AMBlock  SD  et al.  Doctors' emotional reactions to recent deaths of a patient: cross sectional study of hospital doctors. BMJ. 2003;327185- 190
Kristjanson  LJMcPhee  IPickstock  SWilson  DOldham  LMartin  K Palliative care nurses' perceptions of good and bad deaths and care expectations: a qualitative analysis. Int J Palliat Nurs. 2001;7129- 139
Petrosino  BM Characteristics of hospice patients, primary caregivers and nursing care problems: foundation for future research. Hosp J. 1985;13- 19
Contro  NLarson  JScofield  SSourkes  BCohen  H Family perspectives on the quality of pediatric palliative care. Arch Pediatr Adolesc Med. 2002;15614- 19
Andresen  EMNanda  UToce  S Research on the quality of death of children. Qual Life Newslett. 2002;2817- 18
Toce  SSAndresen  EM The roles of data collection, evaluation, and research in pediatric palliative care. Support Voice. 2002;811- 13
Not Available, Promoting Excellence. Instruments. Available at: September 3, 2002
Not Available, SPSS, Release 10.0.5.  Chicago, Ill SPSS Inc1989-1999;
Fleiss  JL Statistical Methods for Rates and Proportions. 2nd ed.  New York, NY Wiley & Sons1981;14- 17
Agresti  AGhosh  ABini  M Ranking κ: describing potential impact of marginal distributions on measures of agreement. Biom J. 1995;37811- 820Article
Byrt  TBishop  JCarlin  RB Bias, prevalence and κ. J Clin Epidemiol. 1993;46423- 429
Feinstein  ARCicchetti  DV High agreement but low κ, I: the problems of two paradoxes. J Clin Epidemiol. 1990;43543- 549
Wolfe  JGrier  HEKlar  N  et al.  Symptoms and suffering at the end of life in children with cancer. N Engl J Med. 2000;342326- 333