Palliative Care in Patients With High-Grade Gliomas in the Neurological Intensive Care Unit

Abstract

Background:

Palliative care improves quality of life in patients with malignancy; however, it may be underutilized in patients with high-grade gliomas (HGGs). We examined the practices regarding palliative care consultation (PCC) in treating patients with HGGs in the neurological intensive care unit (NICU) of an academic medical center.

Methods:

We conducted a retrospective cohort study of patients admitted to the NICU from 2011 to 2016 with a previously confirmed histopathological diagnosis of HGG. The primary outcome was the incidence of an inpatient PCC. We also evaluated the impact of PCC on patient care by examining its association with prespecified secondary outcomes of code status amendment to do not resuscitate (DNR), discharge disposition, 30-day mortality, and 30-day readmission rate, length of stay, and place of death.

Results:

Ninety (36% female) patients with HGGs were identified. Palliative care consultation was obtained in 16 (18%) patients. Palliative care consultation was associated with a greater odds of code status amendment to DNR (odds ratio [OR]: 18.15, 95% confidence interval [CI]: 5.01-65.73), which remained significant after adjustment for confounders (OR: 27.20, 95% CI: 5.49-134.84), a greater odds of discharge to hospice (OR: 24.93, 95% CI: 6.48-95.88), and 30-day mortality (OR: 6.40, 95% CI: 1.96-20.94).

Conclusion:

In this retrospective study of patients with HGGs admitted to a university-based NICU, PCC was seen in a minority of the sample. Palliative care consultation was associated with code status change to DNR and hospice utilization. Further study is required to determine whether these findings are generalizable and whether interventions that increase PCC utilization are associated with improved quality of life and resource allocation for patients with HGGs.

Introduction

Despite advances in medical therapy, patients with high-grade gliomas (HGGs) have an exceptionally poor prognosis. Glioblastoma (GBM), the most common HGG, increases in frequency with age and has a median survival of 15 months with standard treatment.¹ High-grade gliomas cause significant medical comorbidities that often necessitate hospitalization. Compared to patients with brain metastases, those with primary malignant brain tumors have poorer performance status, higher requirements for nursing care, and greater family overburdening.² In contrast to other malignancies, disease progression often leads to cognitive decline, seizures, aphasia, personality changes, and depressed level of consciousness.³ The progressive neurologic impairments that characterize the natural history of HGGs present unique challenges for patients and their families with respect to end-of-life decision-making.

Palliative care involves multidisciplinary specialized medical and nursing care to improve quality of life by alleviating physical and mental suffering. Many hospitals have a palliative care service that is staffed by physician and nurse consultants with specialized training. Early integration of these services improves quality of life and survival and decreases readmission rates and cost per hospitalization in patients with advanced malignancies.^4-6

Prior literature suggests that palliative care, despite its benefits, is underutilized in patients with primary malignant brain tumors. After diagnosis, it is estimated that 22% of patients with GBM spend one quarter of their remaining lives in the hospital.⁷ Many of these patients have prolonged intensive care unit (ICU) stays requiring intubation, enteral feeding, and resuscitation.⁸ A survey of providers attending the Society for Neuro-Oncology Annual Meeting in 2012 found that only 50% of providers felt comfortable dealing with end-of-life issues, while 32% felt that patients’ expectations for ongoing therapy hindered their ability to make palliative care referrals.⁹

We conducted a retrospective cohort study to examine the use of palliative care consultation (PCC) in patients with previously diagnosed HGGs who were admitted to our neurological intensive care unit (NICU). Our hypothesis was that the utilization of PCC would be low. Furthermore, we examined the association between PCC and change in code status with do not resuscitate (DNR), utilization of hospice services, odds of death outside the hospital, 30-day mortality, hospital length of stay (LOS), and 30-day hospital readmission rate.

Methods

Study Design and Participant Selection

This was a retrospective observational cohort study of consecutive patients with a previously diagnosed HGG admitted to an academic tertiary medical center NICU from January 2011 through December 2016. The NICU is comprised of 22 beds and is staffed by a multidisciplinary team of neurointensivists, fellows, advanced practice providers, and residents. There is an inpatient palliative care service that is comprised of physicians from multiple disciplines and advanced practice providers.

Patients included in this analysis were identified retrospectively from our institution’s electronic medical records, which were cross-referenced with an internal registry maintained by our palliative care service. The study was approved by the institutional review board, which authorized a waiver of informed consent, as the investigation involved no more than minimal risk to participants’ privacy. Inclusion criteria included age greater than 18 years and ICU LOS greater than 48 hours, except if death occurred within 48 hours of ICU admission due to intentional limitation of life-sustaining therapies (n = 1). These criteria were chosen to select for a critically ill population and to exclude patients who were admitted for elective tumor debulking with an uncomplicated course (as it is the practice of our institution to admit all postoperative tumor resection cases to the NICU). Patients were excluded if a tissue diagnosis was either not obtained prior to admission or was not consistent with an HGG. High-grade glioma was defined as grade III or grade IV astrocytoma in accordance with histological criteria per the World Health Organization. We did not use molecular genetic data for further classification. Patient demographics including age, sex, and race, and history of prior tumor resection, and admission reason were captured as part of this retrospective cohort study.

Outcomes

The primary outcome was the incidence rate of an inpatient PCC. Furthermore, to assess the impact of PCC on patient outcomes, we evaluated PCC as an independent variable. For this analysis, we estimated the association between PCC and secondary outcomes, including change in code status to DNR (by the participant or a surrogate decision maker) during hospitalization, LOS, discharge disposition (home, inpatient rehabilitation, skilled nursing facility, hospice, death), death within 30 days of admission, death location (hospital vs out of hospital—including home or hospice facility), and 30-day readmission rate. Code status was defined as any documentation of a patient’s wishes with regard to end-of-life care (including DNR order) present in the electronic medical record.

Statistical Analysis

Normality of continuous data was assessed graphically and confirmed using the Shapiro-Wilk test. As all continuous variables were non-normally distributed, they are reported as median with interquartile range (IQR) and were compared using a Wilcoxon rank sum test. Categorical data are presented as frequencies and were compared using χ² or Fisher exact test when contingency table cell counts were less than 5. Univariate logistic regression was used to associate the independent variables age, sex, race, admission indication, and history of prior tumor resection with PCC. We subsequently generated a multivariable logistic regression model estimating the odds of PCC using the independent variables age, sex, race, and admission indication (elective tumor debulking). Prior tumor resection was not included in the model predicting PCC as we did not anticipate that it would affect whether palliative care services would be requested. When PCC was used as an independent variable to assess its clinical impact, univariate logistic regression was used for categorical outcome measures. These analyses were also subsequently adjusted for age, sex, race, and admission indication to generate multivariable logistic regression models. Ordered logistic regression was used to estimate the association between PCC and LOS (in days) as well as ICU LOS. All tests were performed at the 2-sided level with an α of .05 using STATA 15.0 (College Station, Texas). P values <.05 were considered statistically significant.

Results

Patient Characteristics and Primary Outcome

Of the 90 participants included in the analysis, 89 had GBM, 1 patient had anaplastic astrocytoma, and 78 (87%) had previously undergone surgical debulking. The median age was 61 years (IQR: 53-69), 32 (36%) were female, and 73 (81%) were Caucasian. The most common reason for admission was elective tumor debulking (39%), followed by seizures (16%), altered mental status (12%), headache (6%), and intracranial hemorrhage (6%; Table 1). A PCC was documented in 16 (18%) patients (Table 1).

Table 1.

Demographics and Baseline Characteristics.^a

Participant Characteristic	No Palliative Care Consult (n = 74)	Palliative Care Consult (n = 16)	P Value
Age, median year (IQR)	61 (53-69)	61 (46-68)	.38
Sex, female	24 (32%)	8 (50%)	.18
Race			.92
Caucasian	60 (81%)	13 (81%)
Black	6 (8%)	1 (6%)
Asian	2 (3%)	1 (6%)
Unknown	6 (8%)	1 (6%)
Prior tumor resection	63 (85%)	15 (94%)	.69
Admission reason			.04
Elective tumor debulking	33 (45%)	2 (13%)
Seizure	10 (14%)	4 (25%)
Altered mental status	6 (8%)	5 (31%)
Intracranial hemorrhage	4 (5%)	1 (6%)
Headache	3 (4%)	2 (13%)
Sepsis	4 (5%)	0
Other	14 (19%)	2 (13%)

^aAll continuous variables reported as median (interquartile range).

There was no significant difference in age, sex, race, or history of prior debulking surgery among patients who received a PCC (Table 1). Of the measured demographic factors, only admission reason was statistically significantly different between patients who received a PCC and those who did not (Table 1). Altered mental status and seizure (as admission diagnoses) were more common in patients who received a PCC than in patients who did not (31% vs 8% and 25% vs 14%, respectively), whereas elective tumor debulking was more common in patients who did not receive a PCC (45% vs 13%). In multivariable regression, adjusting for age, sex, race, and admission indication, only admission indication (elective tumor debulking) was inversely associated with PCC (adjusted odds ratio [OR]: 0.19, 95% confidence interval [CI]: 0.04-0.91; P = .04).

Palliative Care Consultation and Secondary Outcomes

Palliative care consultation was associated with a higher odds of code status change to DNR (OR: 18.15, 95% CI: 5.01-65.73, P < .01; Table 2). This association remained statistically significant after adjustment for possible confounding variables (adjusted OR: 27.20, 95% CI: 5.49-134.84; P < .01). Patients who received a PCC had a longer hospital LOS (OR: 4.78, 95% CI: 1.77-12.88; P < .01) and ICU LOS (OR: 3.42, 95% CI: 1.22-9.58; P < .01) in ordered logistic regression. These differences remained significant in a multivariable ordered logistic regression (adjusted OR: 4.60, 95% CI: 1.60-13.22; P < .01; and adjusted OR: 3.97, 95% CI: 1.22-12.90, P = .02, respectively). Those who received a PCC had greater odds of being discharged to hospice care (OR: 24.93, 95% CI: 6.48-95.88, P < .01; adjusted OR: 35.01, 95% CI: 6.29-194.89, P < .01) and to die within 30 days of admission (OR: 6.40, 95% CI: 1.96-20.94, P < .01; adjusted OR: 6.29, 95% CI: 1.75-22.64, P < .01). We did not find a statistically significant difference in place of death (outside vs inside the hospital; OR: 10, 95% CI: 0.91-110.28, P = .06) nor in 30-day readmission rate (OR: 0.38, 95% CI: 0.08-1.86; P = .24) between those who received a PCC and those who did not (Table 2). In the multivariable model estimating the association between PCC and 30-day readmission, only age was independently associated with a lower odds of readmission, with 7% lower odds for every year of life (adjusted OR: 0.93 per year, 95% CI: 0.89-0.98; P < .01).

Table 2.

Clinical Outcomes Associated With Palliative Care Consultation.^a

Clinical outcome	No Palliative Care Consult (n = 74)	Palliative Care Consult (n = 16)	P Value	Unadjusted OR (95% CI)	P Value	Adjusted OR (95% CI)b	P Value
Code status change to DNR	8 (11%)	11 (69%)	<.01	18.15 (5.01-65.73)	<.01	27.20 (5.49-134.84)	<.01
Readmission within 30 days of dischargec	20/70 (29%)	2/15 (13%)	.33	0.38 (0.08-1.86)	0.24	NA	NA
Death within 30 days of admission	10 (14%)	8 (50%)	<.01	6.40 (1.96-20.94)	<.01	6.29 (1.75-22.64)	<.01
Death outside hospital	5 (50%)	7 (88%)	.15	10.00 (0.91-110.28)	.06	NA	NA
Discharge dispositiond			<.01	24.93 (6.48-95.88)	<.01	35.01 (6.29-194.89)	<.01
Home	38 (51%)	0 (0%)
Rehab	20 (27%)	2 (13%)
Skilled nursing facility	6 (8%)	2 (13%)
Hospice	6 (8%)	11 (69%)
Died (in hospital)e	4 (5%)	1 (6%)
LOS, days (IQR)f	6 (4-11)	14 (7-21)	<.01	4.78 (1.77-12.88)	<.01	4.60 (1.60-13.22)	<.01
ICU LOS, days (IQR)f	1 (1-3)	4 (1-9)	.03	3.42 (1.22-9.58)	<.01	3.97 (1.22-12.90)	.02

Abbreviations: CI, confidence interval; DNR, do not resuscitate; ICU, intensive care unit; LOS, length of stay; OR, odds ratio.

^aAll continuous variables reported as median (interquartile range).

^bOdds ratios were adjusted for age, sex, and elective admission indication (tumor debulking).

^cOnly those who survived their initial admission were included.

^dOdds ratios shown here were generated using discharge to hospice versus all others.

^eOne patient was discharged to home, then readmitted and died in the hospital within 30 days of the initial admission and thus not included in the “Died” subcategory.

^fOrdered logistic regression used here to estimate association between LOS and palliative care consultation, with the odds ratio indicating the odds of a 1-day longer hospitalization with palliative care consultation.

Discussion

In this single-center retrospective study of critically ill patients with a previously diagnosed HGG, the 2 main findings were as follows: (1) inpatient PCC was utilized in a minority of patients and (2) PCC was associated with a change in the trajectory of patient care, most notably with a change in code status to DNR.

It is remarkable that only 18% of participants received a PCC despite having previously been diagnosed with a fatal illness. Moreover, the majority of participants were beyond the early stages of their disease, as evidenced by the fact that 87% had undergone a prior tumor resection. Our results are similar to those of a retrospective review of 530 patients with primary and metastatic brain tumors (60% with HGGs), which found that only 12% of patients received a PCC during their inpatient admission.¹⁰ The low incidence of PCC in patients with primary malignant brain tumors is particularly concerning, as tumor progression is associated with cognitive decline and neurologic impairment. Thus, as the disease progresses, patients may lack the mental and/or physical capacity to participate in discussions about their wishes, shifting the onus to frequently overburdened surrogate decision makers. This notion is supported by a recent cross-sectional study of patients with intracranial tumors, which found that nearly 25% of patients lacked the necessary mental capacity, when assessed through formal cognitive batteries, to consent for surgical resection.¹¹ These findings are consistent with the growing body of literature showing that formal palliative care services are often underutilized in patients admitted to the NICU including those with HGGs.¹²

The reasons for limited inpatient PCC in patients diagnosed with malignant brain tumors in the ICU are multifactorial and might include limited palliative care resources, competing physician and patient priorities, diffusion of responsibility among multiple caregivers, lack of standardized pathways, discomfort among physicians and patients surrounding discussion of these issues, unpredictable timing of disease progression, and lack of patients’ awareness of their prognosis.¹³ The reasons for the infrequent use of early palliative care for patients with HGGs in the NICU setting and the impact of interventions aimed at increasing PCC warrant further study.

Our findings suggest that PCC, when utilized in the ICU, can impact end-of-life decision-making and goals of care in this patient population. Participants who received PCC were more likely to have their code status changed to DNR, be discharged to hospice, and die within 30 days of admission. We interpret these findings as suggestive that a PCC may help patients and their families to clarify their goals of care. These results are similar to those of a recent retrospective single-center cohort study on patients admitted to a NICU over 8 months (including patients with malignant brain tumors), which found that a PCC was associated with an increased short-term mortality and odds of discharge to hospice.¹⁴ Of note, studies in patients with systemic cancer (including those with cerebral metastases) have found that early inclusion of palliative care services is not associated with decreased survival.⁵

In this study, PCC was associated with an increased hospital and ICU LOS. It is possible that participants who received PCC were sicker, as evidenced in part by the fact that significantly fewer participants in the PCC group were admitted for elective tumor debulking, and more were admitted for new neurologic impairments, such as altered mental status and seizures. Moreover, none of the patients who received a PCC were discharged home, which suggests that severity of illness may have been greater in this population.

Palliative care consultation in the ICU was associated with a nonsignificant trend toward an increased odds of death outside the hospital and a lower odds of 30-day readmission. Similarly, a recent study on the implementation of a comprehensive palliative care program in nearly 900 patients with primary malignant brain tumors found that this intervention significantly reduced readmission rates in the last 2 months of life.⁶ However, the palliative care intervention in that study was instituted early in the patient’s disease course, as opposed to during an ICU admission. Of note, place of death, specifically dying in a preferred location, outside of the hospital setting, has been shown to be independently associated with an improved quality of life in patients with HGGs.¹⁵ The greater odds of discharge to hospice and the trend toward death outside the hospital in patients with HGGs who received a PCC again suggest that this intervention may help shift the focus of care to improving the quality of life in these patients during the later stages of disease.

Our study is limited by its small sample size and involvement of a single neurocritical care program, which is illustrated by the wide CIs. These factors may also have rendered our population susceptible to selection bias such that patients with HGGs admitted to our NICU could have represented a population with more severe disease or could have been treated by neurointensivists who feel more comfortable leading goals-of-care conversations. Therefore, these patients and their families may have less frequently received a PCC, while those in whom palliative care was already involved (or in whom comfort measures were already in place) were less likely to be admitted to our ICU.

Our study is also limited by the unmeasured confounders that are inherent to retrospective studies. Although it is the practice of our institution to admit all postoperative tumor resection patients to the NICU, neither illness acuity nor premorbid functioning (ie, Karnofsky performance index) was measured in our study and may have confounded our results. Thus, patients who received a PCC may have been sicker and therefore more willing to accept a code status change, to transition to comfort care, and to have a prolonged LOS. Furthermore, we only followed patients for the duration of their admission and for up to 30 days from discharge and cannot comment on outcomes of interest beyond these time points. Finally, as this study involved only a single center, it may not be generalizable to other institutions. However, it is largely consistent with other literature. Given these limitations, the results of this study must be viewed as preliminary and should be used for hypothesis generation.

Conclusion

The results of this study should be viewed in context of an emerging body of literature on palliative care, which suggests that formal palliative care services, despite their positive impact on quality of life, may be underutilized in patients admitted to the NICU, including those with malignant brain tumors. Future studies might aim to elucidate barriers to palliative care and to study the long-term medical, financial, and humanistic impact of interventions that increase patient access to palliative care resources both early and late in the disease course.