Location of Death and End-of-Life Characteristics of Young Adults with Cancer Treated at a Pediatric Hospital

Abstract

Background:

Location of death (LOD) is an important aspect of end-of-life (EOL) care. Adolescents and young adults (YAs) with pediatric malignancies are increasingly treated in pediatric institutions. YAs, generally defined as 18–39 years old, deserve specific attention because adults have unique developmental and social considerations compared with younger patients.

Objective:

The goal of this retrospective cohort study was to understand the effect of treatment by a pediatric oncology program on EOL experiences for YAs. Specifically, we examined LOD, hospice, and palliative care (PC) involvement in a cohort of YAs who died of cancer in a large, quaternary care pediatric hospital.

Methods:

This was a retrospective cohort study of patients ≥18 years of age, who died of cancer between January 1, 2010, and December 31, 2017. Standardized data were abstracted from the institutional cancer registry and the electronic medical record.

Results:

YAs in this cohort more commonly died in the hospital (54.9%). Lack of hospice involvement and the presence of a documented do-not-resuscitate (DNR) order were significantly associated with inpatient death. The majority of patients had long-standing PC involvement (95.8%, median 318 days), a DNR order (78.9%), and had enrolled in hospice care (60.6%) before death.

Conclusions:

These results suggest that a significant proportion of YAs with cancer remain inpatient for EOL care. Pediatric oncologists and PC teams may benefit from additional training in the unique psychosocial needs of YAs to optimize EOL care for these older patients.

Introduction

Cancer is the leading cause of nonaccidental death in adolescents and young adults (AYAs), commonly defined as patients aged 15–39 years,¹ killing ∼9000 AYAs annually in the United States.² An important aspect of end-of-life (EOL) care in oncology is death in the patient's preferred location.^3,4 A large longitudinal study evaluating location of death (LOD) in AYAs with cancer in California found that between 1989 and 1994, hospital death rates declined from 68.3% to 53.6% and home death rates increased from 16.8% to 35.5%, remaining stable through 2011.⁵ Mack et al.⁶ examined EOL care in a cohort of AYAs with cancer (15–29 years) with Medicaid in New York state and found that 65% of patients died in an acute care setting. Only 23% of patients in this cohort had hospice care involved before death.

While adult patients who are dying tend to prefer less aggressive EOL measures,⁷ it has been suggested that young people may feel differently about the trade-offs that are worthwhile for another day.⁸ One retrospective study of AYAs with cancer in Australia found that AYAs were more than twice as likely than adults (51.9% vs. 24.3%) to express a desire to pursue all available oncological therapies as they approached EOL.⁹ Snaman et al.¹⁰ reviewed 69 deaths of AYAs (15–26 years) with cancer at St. Jude Children's Research Hospital and found that patients frequently received intensive interventions in the last month of life, including mechanical ventilation (42%), medical procedures (71%), and total parenteral nutrition (52%). Importantly, this study was limited to AYAs who died in the hospital, not AYAs who died of cancer overall. The largest study evaluating care intensity at EOL in AYAs with cancer reviewed electronic health records of 663 patients in the Kaiser Permanente Southern California registry.⁸ During the last 30 days of life, 22% of patients in this cohort were admitted to an intensive care unit (ICU), 22% had more than one ED visit, and 62% were hospitalized.⁸ It remains unclear whether the high frequency of intensive EOL care in AYAs with cancer is driven by goals and values unique to this age group or if it instead results from system issues surrounding EOL care for AYAs.

While multiple studies have cited a parental preference for children with cancer to die at home,^11–14 the most common LOD in children with cancer is less clear, with available literature largely limited to single-institution studies. A review of the National Center for Health Statistics found that 37.6% of children who died of cancer in 2011 died at home, with rates of home death relatively stable between 1999 and 2011, but did not detail other LODs.¹⁵ Three single-institution studies of 880 total patients treated in pediatric oncology programs^16–18 have found frequencies of home death ranging from 33.4% to 56.4%, with ICU deaths ∼20% and 12%–16.8% of patients dying in a non-ICU hospital environment. Thus, while LOD in children with cancer may depend somewhat on the site of cancer treatment, these studies are heterogeneous with regard to inclusion criteria, including patient age and cause of death, making it difficult to draw definitive conclusions overall.

Due to evidence supporting improved outcomes when AYAs with pediatric malignancies are treated on pediatric protocols,³ AYAs with pediatric malignancies are increasingly treated in pediatric institutions. EOL care for these patients therefore is also provided by pediatric oncologists and palliative care (PC) physicians. Young adults (YAs) are a subset of the AYA group, generally defined as 18–39 years old. Although most studies in EOL care in AYAs include teens aged 15–17, YAs with cancer form a unique population with specific developmental and social considerations.^19,20 Some YAs live independently or are married with children, while others live with their parents or guardians. In addition, many YAs lose financial and social independence after the cancer diagnosis and once again become dependent on their family members for care. This YA group deserves specific attention because adult patients who are treated at pediatric institutions may have limited access to age-appropriate EOL care when compared with children. There are also few studies to date examining EOL care patterns specifically in YAs and none, to our knowledge, evaluating EOL care in YAs with cancer treated in pediatric hospitals.

The goal of this retrospective cohort study was to better understand the effect of treatment by a pediatric oncology program on EOL experiences for YAs. Specifically, we examined LOD and hospice and PC involvement in a cohort of YAs who died of cancer in a large, quaternary care pediatric hospital. PC in our institution is well developed and well integrated with oncology, with stable resources over the study period. Based on the literature and our anecdotal experience, our primary hypothesis was that YAs treated in our institution and who died of relapsed/refractory cancer between 2010 and 2017 were more likely to die in the hospital than at home. Secondary hypotheses were that patients without hospice involvement were more likely to die in the hospital and that diagnosis (leukemia/lymphoma) was predictive of death in the hospital.^10,16,21,22

Methods

Study design

This study was conducted at Cincinnati Children's Hospital Medical Center (CCHMC), a large, quaternary care pediatric hospital and phase I institution. The CCHMC Institutional Review Board deemed the study exempt from review as it was a retrospective study of a deceased cohort. The institutional cancer registry was queried for all patients with a primary cancer diagnosis, who died between January 1, 2010, and December 31, 2017, and who were ≥18 years old at the time of death. An initial review of data was conducted in tandem by the study team comprising two physicians dual-trained in pediatric PC and pediatric hematology–oncology to determine exclusions. Further data collection was then conducted by a clinical research coordinator with experience in this field. The research team then met to review and resolve any inconsistencies in data collection.

Inclusion criteria were as follows: patients treated primarily in our institution, who died of relapsed/refractory cancer between January 1, 2010, and December 31, 2017, and who were ≥18 years old at the time of death. The start date was selected based on the date the electronic medical record (EMR) was available hospital-wide, allowing for a more detailed chart review and abstraction. Exclusion criteria were as follows: (1) death secondary to chemotherapy toxicity, (2) death secondary to complications of bone marrow transplant (BMT), and (3) the patient was primarily treated at another institution, seen for a second opinion or phase I chemotherapy, and care was transferred back to the primary oncologist before death.

Data were collected from the local cancer registry database and the EMR. Specific data collected included patient name, medical record number, date of birth, gender, diagnosis, diagnosis date, date of death, history of stem cell transplant, race/ethnicity, zip code, religion, payor status, cause of death, LOD, do-not-resuscitate (DNR) status, date of DNR, date of hospice referral and hospice enrollment, history of PC consult, and date of first PC consult. Notes by multiple providers (oncologists, PC providers, social workers, and chaplains) written near the time of death were reviewed in detail to obtain more information regarding events surrounding the patient death, for example, LOD if not in the hospital, discussions regarding code status, and whether patient's care had been transferred back to the primary site. Specific diagnosis was recategorized as leukemia/lymphoma, solid tumor, or brain tumor. Date of first PC consult was abstracted from the first note documented by any member of the PC team, inpatient, or outpatient.

Statistical analysis

Demographic data, disease group, PC, and EOL characteristics were summarized using descriptive statistics. Univariable analyses, including two-sample t-test, Wilcoxon rank-sum test, chi-square, and Fisher's exact tests, were run to examine the relationship between individual predictor variables and the outcome variable (LOD, hospital vs. others). Variables that reached significance of p ≤ 0.1 in univariable analysis (DNR status, no or unknown vs. yes, and hospice involved, no or unknown vs. yes) were included in the multiple logistic regression model and subjected to stepwise variable selection. Variables with adjusted p-value ≤0.05 were retained in the final model.

Results

A total of 154 patients, aged 18–49 years, who had been treated for cancer in our institution died during the time period studied. Thirty-two patients were primarily eliminated due to death unrelated to cancer, leaving a cohort of 122 patients. Fifty-one patients were excluded for the following reasons: care was transferred back to a local oncologist (N = 32), death related to BMT (N = 17), and death immediately after therapy was initiated (N = 2). The final cohort included 71 patients, aged 18–42 years. Table 1 describes demographic data of eligible participants. Table 2 describes EOL characteristics of the cohort. The location “other” included hospice units or nursing homes. Table 3 describes results of the multiple logistic regression model.

Table 1.

Demographic Information of Cohort

	N (%)
Gender
Female	36 (50.7)
Race
White	59 (83.1)
Black	8 (11.3)
Other	3 (4.2)
Hawaiian	1 (1.4)
Diagnosis
Solid tumor	33 (46.5)
Brain tumor	23 (32.3)
Leukemia or lymphoma	15 (21.1)

	Median (IQR)
Age at death, years	24.2 (20–27)

IQR, interquartile range.

Table 2.

End-of-Life Characteristics of Cohort

DNR status	N (%)
Yes	56 (78.9)
Limited	1 (1.4)
Unknown	14 (19.7)
Palliative care
Yes	68 (95.8)
No or unknown	3 (4.2)
Yes and >30 days before death	63 (88.7)
No or ≤30 days before death	8 (11.3)
Hospice enrollment
Yes	43 (60.6)
No	26 (36.6)
Unknown	2 (2.8)
Location of death
Hospital	39 (54.9)
Home	23 (32.4)
Other	8 (11.3)
Unknown	1 (1.4)
Hospice enrollment, location of death
Hospital	14 (32.6)
Home	22 (51.2)
Other	6 (13.9)
Unknown	1 (2.3)
CCHMC hospice, location of death
Home	14 (56.0)
Hospital	11 (44.0)
Time from DNR to death, days
Range	0–411
Median (IQR)	21 (4–63)
CCHMC hospice, location of death
Range	8–1468
Median (IQR)	318 (126–580)

CCHMC, Cincinnati Children's Hospital Medical Center; DNR, do-not-resuscitate.

Table 3.

Predictors of Death in the Hospital

Variable	Odds ratio (95% CI)	p
Hospice	58.8 (5.1–1000)	0.0011
Never involved vs. involved
DNR status	13.2 (1.3–125)	0.0278
No or unknown vs. yes

Diagnosis did not significantly predict LOD in this cohort: 63.0% of patients with solid tumors, 52.4% of patients with brain tumors, and 73.3% of patients with leukemia/lymphoma died in the hospital. A majority of patients had PC involvement (95.8%), evidence of a DNR order (78.9%), and were enrolled in hospice at some point (60.6%). All patients who died in the oncology unit had a DNR order at the time of death. Patients with leukemia/lymphoma were least likely to enroll in hospice care: 26.7% of these patients enrolled in hospice compared with 66.7% with solid tumors and 71.4% with brain tumors. PC was involved for a median of 318 days before death.

Thirty-nine (54.9%) eligible patients died in the hospital. Two died in the ICU and the remainder died in the oncology inpatient unit. One patient died in the ICU after a code event, having clearly expressed the desire to remain Full Code, and the other died after withdrawal of life-sustaining interventions. Therefore, only one of the 71 patients in this cohort of YAs with refractory cancer died after having received cardiopulmonary resuscitation.

Discussion

This retrospective cohort study sought to describe the EOL experiences of YAs treated in a large, quaternary care children's hospital and who died of relapsed/refractory cancer. Our primary hypothesis was that YAs treated in our institution and who died of refractory cancer between 2010 and 2017 were more likely to die in the hospital than at home. Secondary hypotheses were that patients without hospice involvement were more likely to die in the hospital and that diagnosis (leukemia/lymphoma) was predictive of death in the hospital.

We found that YAs who died of refractory cancer in our institution more often died in the hospital (54.9%), with 33.8% of patients dying at home or in a hospice unit. These findings are similar to those found in other studies examining LOD in YAs with cancer. Lack of hospice involvement and lack of documented DNR status were associated with death in the hospital. The majority of patients (78.9%) had evidence of a DNR order in our institutional EMR at the time of death. As a portion of patients were enrolled in hospice programs not affiliated with our hospital at the time of death, and hospice-level data were not accessed, this number may have in fact been higher than captured with these methods.

We believe that the high frequencies in this cohort of DNR orders and deaths outside of the ICU indicate that the majority of patients in this cohort received comfort-directed care at the time of death, regardless of LOD. This may support the notion that the care delivered was directed primarily by the patient's goals, but future prospective studies are required to answer this question. Factors that have been cited as influencing the relatively higher rate of inpatient deaths for YAs nationally include the low frequency of PC and hospice referrals in this population.^6,23,24 Yet, not only was PC involved for virtually all (95.8%) of the patients in this cohort, comparable with findings at St. Jude Children's Research Hospital,^23,24 but also the median time of PC involvement was nearly a year (318 days) before the patients died. The majority (60.6%) also had evidence of hospice enrollment. The PC team in our institution is automatically consulted for all oncology patients and is the primary pain service for oncology and bone marrow transplant services. Additionally, PC physicians serve as hospice physicians in our hospital-based hospice program. Discussion of goals of care and preferred LOD, in addition to symptom management issues, is a standard part of the PC consult. Thus, given the high frequency and length of PC involvement, lack of hospice enrollment may indicate patient preference to not enroll rather than evidence that hospice was never offered.

While conventional wisdom and early work in hospice research heralded a home death as a good death and equated home death with high-quality EOL care, death at home is not always feasible or preferred. Current work evaluating LOD as a quality metric supports that death in the preferred location and the opportunity to plan LOD are better quality metrics than death at home.^4,5,25,26 Due to limitations of a retrospective study, we were not able to abstract how commonly patients were given the opportunity to plan their LOD, what factors may have impacted that choice, or how commonly that plan was carried out. However, we believe that the low frequency of ICU death and high frequency of PC involvement support the notion that patients were given the opportunity to discuss EOL planning and suggest that the hospital was indeed the preferred LOD for many YAs with cancer.

Lack of hospice enrollment had a significant association with death in the hospital in this cohort. This is not surprising and more likely reflects that patients did not feel that hospice enrollment was in line with their goals than that hospice care was never offered. Still, nearly a third of patients who enrolled in hospice died in the hospital, contributing to the high overall frequency of inpatient death. In our experience, YAs may prefer to not die at home for multiple reasons. Some with young children have worried that their children would be impacted by the memory of their death and thus desired to die elsewhere. Others have stated that they do not want their parents, spouses, or partners to become their primary caregivers; being in the hospital provides dignity to those who prioritize their independence as they die and allows family members to just be family. Some bereaved families have cited the importance of the oncology team at the hospital as their primary source of support in the cancer journey and have been hesitant to involve a new team (e.g., hospice program) as time becomes short. Finally, some YAs with cancer have been readmitted to the hospital for more aggressive symptom control at EOL.

Hospice programs may offer EOL care in an inpatient unit, providing another option for LOD. To remain eligible for inpatient hospice care, though, the patient must be actively dying or experiencing uncontrolled symptomatology. However, in the days to weeks leading up to the time when the patient is actively dying, patients and their families do not have access to this level of care through the traditional adult hospice model. We have cared for some YAs whose families were not physically capable of caring for them at home, with or without hospice support, and historically these patients have not been transferred to nursing homes or long-term care facilities and instead are cared for in the hospital. Thus, admission to the inpatient oncology unit may offer more flexibility for YAs at EOL than the traditional model of hospice care, highlighting a limitation of hospice for patients in this age group, both in our institution and broadly.

YAs who live near our hospital are eligible for enrollment in a pediatric hospice program in the division of homecare in our institution. Although we encourage YAs to enroll in traditional adult hospice programs, many choose to continue care with our hospice as our PC team has experience managing their pain and we work closely with the oncology team. This hospice program does not have an inpatient hospice unit, though, and 44% of patients who enrolled in our program ultimately died in the hospital. The oncology unit does offer larger rooms for patients who are dying, but aside from the increased space, these rooms are designed for typical inpatient care and not EOL care. There is no outdoor access, for example, or additional privacy for family members visiting and holding vigil at the bedside.

Although patients with leukemia/lymphoma were least likely to enroll in hospice, consistent with other studies,^16,18 we did not find that diagnosis was a statistically significant predictor of LOD: 63.0% of patients with solid tumors, 52.4% with brain tumors, and 73.3% with leukemia/lymphoma died in the hospital. This may be due to the flexibility of admission to our institution at EOL for all YAs, regardless of diagnosis, and may again reflect that when available, the hospital is either the preferred LOD for many YAs or the best option when a hospice unit is not available.

A pediatric oncology unit is most often designed for care of patients through the age of 22 years. As AYA programs grow to include a greater proportion of patients in their 20s and 30s, it is critical to design pediatric oncology centers to also serve the needs of older YA patients, including EOL care. Staff caring for AYA patients may benefit from additional training in communication of sexual health and intimacy as well as substance abuse. Future multisite prospective studies are required to understand the unique desires and needs of YAs as they die and reasons for preferred LOD to enable the design of systems capable of delivering high-quality EOL care while also tailoring the plan to needs of each patient.

This study is a single-institution study and therefore findings may not be generalizable to other pediatric or adult oncology programs. We did not have adequate statistical power to detect a difference in proportions among diagnoses. This was also a retrospective cohort study, and we found limited documentation of preferred LOD or other patient-reported outcomes; these data are not standardized in the EMR and therefore could not be systematically abstracted. We are a large, phase I trial oncology institution and receive national and international referrals, which may attract patients and families seeking more aggressive care. To account for this, we limited this study to only those patients who were treated primarily at our institution and excluded patients who were discharged back to their primary team for EOL care after being seen for phase I or II trials or secondary referrals.

Conclusions

We found that YAs with cancer treated at a large, quaternary, pediatric oncology program more commonly died in the hospital. The majority of patients in this cohort had long-standing PC involvement, had documented DNR status when they died, and had enrolled in hospice care before death. Only one patient died after receiving cardiopulmonary resuscitation in the ICU, which was in line with that patient's goals. Lack of hospice involvement and documented DNR status were associated with death in the hospital. These results suggest that YAs with cancer may prefer to remain in the hospital for EOL care. Pediatric oncologists and PC teams may benefit from additional training in the unique psychosocial needs of YAs to optimize EOL care for these patients.

Footnotes

Acknowledgments

The authors gratefully acknowledge the contributions of Millicent Frimpong-Manso who assisted with data collection for this study. Dr. Thienprayoon's work on this study was supported by intramural grants, including the Place Outcomes Research Award and the Schmidlapp Women's Scholar Award.

Author Disclosure Statement

No competing financial interests exist.

References

Institute NC. Adolescents and young adults with cancer. Accessed September 26, 2018, 2018 from https://www.cancer.gov/types/aya

Society

AC.

Key Statistics for Cancers in Young Adults. Accessed September 25, 2018, 2018 from https://www.cancer.org/cancer/cancer-in-young-adults/key-statistics.html

Ram

, Wolach

, Vidal

, et al. Adolescents and young adults with acute lymphoblastic leukemia have a better outcome when treated with pediatric-inspired regimens: systematic review and meta-analysis. Am J Hematol. 2012; 87(5):472–8.

Dussel

, Kreicbergs

, Hilden

, et al. Looking beyond where children die: determinants and effects of planning a child's location of death. J Pain Symptom Manage. 2009; 37(1):33–43.

Rajeshuni

, Johnston

, Saynina

, et al. Disparities in location of death of adolescents and young adults with cancer: a longitudinal, population study in California. Cancer. 2017; 123(21):4178–84.

Mack

, Chen

, Boscoe

, et al. High intensity of end-of-life care among adolescent and young adult cancer patients in the New York State Medicaid Program. Med Care. 2015; 53(12):1018–26.

Weeks

, Cook

, O'Day

, et al. Relationship between cancer patients' predictions of prognosis and their treatment preferences. JAMA. 1998; 279(21):1709–14.

Mack

, Chen

, Cannavale

, et al. End-of-life care intensity among adolescent and young adult patients with cancer in Kaiser Permanente Southern California. JAMA Oncol. 2015; 1(5):592–600.

Fletcher

, Hughes

, Pickstock

, Auret

. Advance care planning discussions with adolescent and young adult cancer patients admitted to a community palliative care service: a retrospective case-note audit. J Adolesc Young Adult Oncol. 2018; 7(1):112–9.

10.

Snaman

, Kaye

, Lu

, et al. Palliative care involvement is associated with less intensive end-of-life care in adolescent and young adult oncology patients. J Palliat Med. 2017; 20(5):509–16.

11.

Goldman

, Beardsmore

, Hunt

. Palliative care for children with cancer—home, hospital, or hospice?. Arch Dis Child. 1990; 65(6):641–3.

12.

Kassam

, Skiadaresis

, Alexander

, Wolfe

. Parent and clinician preferences for location of end-of-life care: home, hospital or freestanding hospice?. Pediatr Blood Cancer. 2014; 61(5):859–64.

13.

Vickers

, Thompson

, Collins

, et al. Place and provision of palliative care for children with progressive cancer: a study by the Paediatric Oncology Nurses' Forum/United Kingdom Children's Cancer Study Group Palliative Care Working Group. J Clin Oncol. 2007; 25(28):4472–6.

14.

Vickers

, Carlisle

. Choices and control: parental experiences in pediatric terminal home care. J Pediatr Oncol Nurs. 2000; 17(1):12–21.

15.

Cawkwell

, Gardner

, Weitzman

. Persistent racial and ethnic differences in location of death for children with cancer. Pediatr Blood Cancer. 2015; 62(8):1403–8.

16.

Brock

, Steineck

, Twist

. Trends in end-of-life care in pediatric hematology, oncology, and stem cell transplant patients. Pediatr Blood Cancer. 2016; 63(3):516–22.

17.

Kaye

, Gushue

, DeMarsh

, et al. Illness and end-of-life experiences of children with cancer who receive palliative care. Pediatr Blood Cancer. 2018; 65(4):DOI: 10.1002/pbc.26895.

18.

Thienprayoon

, Lee

, Leonard

, Winick

. Racial and ethnic differences in hospice enrollment among children with cancer. Pediatr Blood Cancer. 2013; 60(10):1662–6.

19.

Wiener

, Weaver

, Bell

, Sansom-Daly

. Threading the cloak: palliative care education for care providers of adolescents and young adults with cancer. Clin Oncol Adolesc Young Adults. 2015; 5:1.

20.

Mack

, Chen

, Cannavale

, et al. End-of-life care intensity among adolescent and young adult patients with cancer in Kaiser Permanente Southern California. JAMA Oncol. 2015; 1(5):592–600.

21.

Bell

, Skiles

, Pradhan

, Champion

. End-of-life experiences in adolescents dying with cancer. Support Care Cancer. 2010; 18(7):827–35.

22.

Hui

, Didwaniya

, Vidal

, et al. Quality of end-of-life care in patients with hematologic malignancies: a retrospective cohort study. Cancer. 2014; 120(10):1572–8.

23.

Keim-Malpass

, Erickson

, Malpass

. End-of-life care characteristics for young adults with cancer who die in the hospital. J Palliat Med. 2014; 17(12):1359–64.

24.

Fraser

, Miller

, McKinney

, et al. Referral to a specialist paediatric palliative care service in oncology patients. Pediatr Blood Cancer. 2011; 56(4):677–80.

25.

Bluebond-Langner

, Beecham

, Candy

, et al. Preferred place of death for children and young people with life-limiting and life-threatening conditions: a systematic review of the literature and recommendations for future inquiry and policy. Palliat Med. 2013; 27(8):705–13.

26.

Grinyer

, Thomas

. The importance of place of death in young adults with terminal cancer. Mortality. 2004; 9(2):114–31.