End-of-Life Care Among Adolescent and Young Adult Cancer Patients Living in Poverty

Abstract

Background

End-of-life (EOL) care outcomes among adolescents and young adults (AYAs) with cancer living in poverty remain poorly understood. Our primary aim was to examine the effect of poverty on EOL care for AYAs with cancer.

Methods

We conducted a multisite retrospective study at three academic sites of AYA cancer decedents aged 15–39 who died between January 2013-December 2016. Medical record-based EOL care outcomes included hospice referral, palliative care (PC) consultation, cancer treatment in last month, and location of death. We applied two measures of poverty: zip code with median income ≤200% of Federal Poverty Level and public insurance or lack of insurance. Logistic regression analyses were conducted.

Results

We identified 252 AYA cancer decedents. 41% lived in a high-poverty zip code and 48% had public insurance or lacked insurance; 70% had at least one poverty indicator. Nearly 40% had a hospice referral, 60% PC consultation (76% inpatient), 38% EOL cancer treatment, and 39% died in the hospital. AYA patients living in low income ZIP codes were less likely to enroll in hospice (p=<0.01), have an early PC referral (p=<0.01), or receive EOL cancer treatment (p=.03) in bivariable analyses, although only EOL cancer treatment met statistical significance in multivariable models. No differences in location of death (P=.99) were observed.

Conclusions

AYA cancer patients experience low rates of hospice referral and high rates of in hospital death regardless of socioeconomic status. Future studies should evaluate early inpatient PC referrals as a possible method to improve EOL care.

Precis:

AYA patients with cancer living in poverty frequently experience inpatient death and chemotherapy near the end of life, raising concerns about end-of-life care quality. Early inpatient PC referrals may be one method to improve their end-of-life care which requires prospective validation.

INTRODUCTION

End-of-life (EOL) clinical outcomes for adolescents and young adults (AYAs) with cancer living in poverty remain poorly studied. Approximately 70,000 AYAs are diagnosed with cancer and 15,000 die from cancer annually in the United States.^{1, 2} AYAs, defined by the National Cancer Institute as patients between the ages of 15 to 39 years,³ are a unique cancer population facing different challenges compared to their adult counterparts. Specifically, as young patients cope with a poor prognosis, they must also face profound losses, including the inability to find partners, nurture families and careers, and grow old.^{4, 5} AYAs may also be especially vulnerable to clinical care challenges related to their EOL care.^6–9 Patients, caregivers, and even clinicians experience young people’s deaths as tragic events and departures from the natural order of life.⁵ Consequently, most of these young patients die after receiving intensive treatments in the last month of life.^{10, 11}

Poverty compounds the challenges faced by AYAs in the United States (US) as one-fifth of young adults in the US live in poverty.³ AYAs living in poverty may experience longstanding stressors related to instability in family life, education, and community; and lower health status.^{12, 13} While there is recognition that adults with cancer living in poverty face challenges around EOL care,^{12, 13} including limited access to EOL care services such as hospice,^14–16 the EOL care received by impoverished AYAs has received almost no attention.

More work is needed to establish the role of poverty in EOL care choices such as palliative and hospice care among AYA patients, and to understand the best models for EOL care delivery for AYA patients living in low resource areas. The American Society of Clinical Oncology (ASCO) has identified established measures to define high-quality EOL care in adult patients with incurable solid tumor cancer, which may or may not define optimal EOL care for AYAs: use of palliative care, hospice care, cancer-directed therapy 14 days prior to death, and death at home.^{17, 18} Palliative care offers expertise and resources for symptom control and quality of life for patients with cancer regardless of prognosis.¹⁷ In contrast, hospice care is available to patients with a prognosis of less than six months.¹⁹ In AYA patients, inpatient palliative care involvement has led to less intensive EOL care²⁰ and may offer a model for EOL care for patients who are unable or unwilling to access home hospice services. Thus, use of palliative care may be one way in which AYAs can access high-quality EOL care even without hospice.

We conducted a multisite retrospective study to evaluate EOL care for cancer patients at the intersection of youth and poverty—a historically understudied and underserved population with a high risk for poor-quality EOL care. Specifically, we aimed to characterize the nature and determinants of palliative and hospice care among AYA patients living in poverty. We hypothesized that AYA patients living in poverty would experience fewer palliative care and hospice referrals while being more likely to receive cancer treatment in the last month of life and die in the hospital. By investigating EOL outcomes in AYA patients with cancer living in poverty, we hope to define the needs of this unique patient population to inform future interventions designed to improve EOL care delivery.

MATERIALS AND METHODS

This study was conducted with the Palliative Care Research Cooperative, which facilitated involvement of member sites, provided oversight, as well as data coordination and monitoring. Each study site obtained local IRB approval.

Sample.

The sample included AYA cancer decedents identified retrospectively at three sites: the University of Alabama at Birmingham (UAB), University of California San Diego (UCSD), and University of Iowa (UI). Site selection was based on high racial, ethnic, and geographic diversity with a goal to represent both urban and rural settings. Patients were included if they died between January 2013 and December 2016, were 15–39 years of age at death, received care at the designated study site during the last 30 days of life, and were diagnosed at least 30 days prior to death. The three-year study period was selected to ensure adequate numbers of decedents to meet our primary aim. We focused on the most recent years available at the time of study initiation. We selected a 30-day window prior to death to ensure patients had adequate time between diagnosis and death to receive palliative and hospice care. With 252 AYA decedents, we achieved greater than 80% power at alpha level of .05 to detect a 20% difference in hospice for those living in poverty versus not living in poverty. Our goal was to identify equal numbers of patients at each site, for N=84 at each site. Two sites fell short of this goal despite abstraction of charts for all eligible patients, with 61 patients eligible at UI and 83 at UAB. Additional charts, selected at random, were therefore abstracted at UCSD to achieve our full planned sample size.

Data Source and Design.

This study was a correlational design. The primary data source was medical records. Each study site performed medical record reviews to confirm eligibility and to abstract relevant data elements. Sites also used their local cancer registry data to identify information on dates of birth, cancer diagnosis, and death. Medical records were abstracted by trained reviewers at each study site using a standardized medical record abstraction form and entered into a REDCap database. Data quality was evaluated with 31% of records abstracted by a second reviewer and any variable discrepancies, such as cancer stage and/or hospice enrollment, addressed by re-review and consensus. Publicly available US Census data were also used in this study. Specifically, we used the information on Federal Poverty Level (FPL) and linked it to the zip code data derived from the medical records.

Study Measures

End-of-life care.

We selected four outcome measures of EOL care for this study: hospice referral, palliative care consultation, treatment in last month, and location of death. These four outcome measures were selected as they represent existing, widely used, and validated EOL care quality indicators.^1–6 For this study, we were especially interested in use of hospice and palliative care based on our prior work.^{7, 8}

Hospice referral was operationalized as whether or not an AYA was referred to hospice care during the study time period. While prior studies using administrative data have ascertained hospice enrollment, we found that the hospital medical records identified hospice referral more reliably. Palliative care consultation was defined as a palliative care consult, including both inpatient or outpatient settings, that was documented in the medical record. Early palliative care was defined as occurring more than seven days prior to death. Cancer-directed treatment in the last month was defined as whether or not an AYA received any cancer treatments in the last 30 days of life including chemotherapy, radiation, surgery, or an interventional clinical trial. Location of death was defined as whether an AYA died within the hospital as opposed to at home, inpatient hospice, or elsewhere.

Poverty.

We used two indicators of poverty for this study. Our area-level measure of socioeconomic status was whether or not an AYA resided in a zip code area where the median income was less than or equal to 200% of the FPL versus greater than 200% of the FPL. The threshold of less than 200% FPL was chosen to be consistent with prior work²¹ as a more inclusive measure of material hardship.^{22, 23} Insurance type was defined as whether an AYA was uninsured or enrolled in a public (e.g., Medicaid insurance) versus a private insurance plan. For analysis, Medicaid and uninsured patients were grouped together.

Covariates.

Demographic and cancer characteristics included sex (female or male), ethnicity (Hispanic or non-Hispanic), race (Caucasian, Black, Other, or Unknown), age at diagnosis (≤21, 22–29, or 30–39 years), age at death (15–21, 22–29, or 30–39 years), cancer type (solid tumor or hematological malignancy), and study site (UCSD, UAB, or UI).

Statistical Analysis

The primary goal of this study was to examine the effect of poverty on EOL care for AYA with cancer. Descriptive statistics were calculated on all variables for AYAs in the sample. The Pearson chi-square test for differences in proportions and the Wald test for difference in means were used to provide comparisons among sample variables according to our measures of poverty (socioeconomic status and insurance type). A multivariate logistic regression was conducted to examine the effect of socioeconomic status and insurance type on EOL outcomes, while controlling for demographic and cancer characteristics. Individual regressions were conducted for each EOL outcome and results were reported as adjusted odds ratios (aOR). All analyses were conducted using Stata 11.0 software (StataCorp LP, College Station, TX).

RESULTS

Overall, 252 AYAs with cancer were identified at the three sites combined (43.7% at UCSD, 33.7% at UAB, and 22.6% at UI) (Table 1). Racial and ethnic minorities included 21% of patients who were Hispanic and 18% were Black. About one-fifth of patients died before the age of 22 (21%); one-quarter died between the ages of 22 and 29; and over one-half died in their thirties (56%). Most (82%) had a solid tumor malignancy. Regarding our selected indicators of poverty, 41% of patients lived in a zip code with a median income less than or equal to 200% of the FPL, and 48% had Medicaid or lacked insurance. In these AYA patients with cancer, 50% had a single indicator of poverty and 20% had two indicators, such that 70% overall had at least one poverty indicator. Patients from UAB and UI were more likely than UCSD patients to live in high-poverty zip codes (p<.01), but UCSD had higher rates of patients with public insurance or lack of any insurance (p<.01).

Table 1.

Patient sociodemographic characteristics. (N=252)

	Full sample	Median income by ZIP code			Insurance status
	%	%, <=200% FPL	%, >200% FPL	P value, chi squared test	%, Medicaid insured or uninsured	%, all other insurance	P value, chi squared test
	N=252	N=103	N=149		N=123	N=129
Female sex	47	39	52	0.04	42	51	0.16
Hispanic ethnicity	21	16	25	0.08	30	12	<0.01
Race				<0.01			0.02
White	59	61	58		51	67
Black	18	24	13		17	18
Other	4	1	6		6	2
Unknown	19	14	24		26	13
Age at diagnosis				0.14			0.17
≤21	27	33	24		25	30
22–29	26	28	25		26	30
30–39	46	39	51		52	40
Age at death				0.02			0.17
15–21	21	28	15		16	25
22–29	24	25	23		23	25
30–39	56	47	62		61	50
Married	34	25	40	<0.01	20	47	<0.01
Cancer diagnosis				0.55			0.64
Solid Tumor	82	84	81		83	81
Hematologic Malignancy	18	17	20		17	19
Site of care				<0.01			<0.01
UCSD	43	19	59		52	34
UAB	33	49	22		23	43
UI	24	32	19		25	23

UAB = University of Alabama Birmingham, UCSD = University of California San Diego, UI = University of Iowa; Note: percentages may not add up to 100 due to rounding.

Overall, 39% of patients had documentation of hospice referral, with no difference by area-level poverty or insurance (Table 2). Fewer patients had documented enrollment (29%) than referral, with an equal number of charts lacking documentation across study sites regarding an ultimate hospice decision. Of patients with documented enrollment, more than two-thirds (69%) took place more than seven days before death, and home hospice was the most commonly used modality (84%).

Table 2.

End-of-life care received. (N=252)

	Full sample	Median income by ZIP code			Insurance status
	%	%, <=200% FPL	%, >200% FPL	P value, chi squared test	%, Medicaid insured or uninsured	%, all other insurance	P value, chi squared test
	N=252	N=103	N=149		N=123	N=129
Hospice referral	39	36	42	0.36	42	36	0.34
Hospice enrollment	29	21	35	<0.01	30	29	0.27
Unknown	29	47	17		24	33
If enrolled, >7 days before death	69	68	69	0.64	68	70	0.97
Unknown	5	9	4		5	5
If enrolled, location of hospice				0.56			0.75
Inpatient	11	14	10		8	14
Outpatient	84	77	87		87	81
Unknown	5	9	4		5	5
Receiving hospice care at time of death	78	68	83	0.13	89	68	0.06
Unknown	19	32	13		11	27
Palliative care consultation	60	64	57	0.31	60	61	0.96
If palliative care consultation, >7 days before death	75	64	84	<0.01	79	72	0.45
If palliative care consultation, location				0.11			0.82
Inpatient/Other	76	83	71		76	77
Outpatient	21	17	26		23	21
Unknown	2	0	4		1	3
Health care proxy	26	23	28	0.39	24	28	0.53
DNR/DNI order	36	42	32	0.10	32	40	0.20
Cancer treatment in last month	38	30	44	0.03	33	43	0.08
Cancer treatment in last week	14	9	17	0.07	16	11	0.21
Inpatient death	39	39	39	0.99	34	43	0.13

Note: percentages may not add up to 100 due to rounding.

Overall, most patients (60%) received a palliative care consult, with no difference by area-level poverty or insurance. Three-quarters of palliative care referrals occurred more than seven days before death, although AYA patients living in high-poverty areas were less likely to have early referrals than patients not living in poverty (p=<0.01). Most palliative care consults occurred in the hospital (76%). Only one-quarter of patients had documented health care proxies and 36% had do-not-resuscitate orders documented in medical records before death.

A minority of patients reviewed received cancer-directed therapy in the last month (38%) or last week (14%) of life. Patients who lived in high-poverty areas had lower rates of cancer-directed therapy in the last month of life (p=.03). Thirty-nine percent of patients died in the hospital, with no difference by poverty measures. There were no observed differences in the other outcomes measured.

The results of the multivariate regression analyses estimating the association between poverty and EOL care for AYAs are presented in Supplemental Table 1 and Supplemental Table 2. Hospice referral, palliative care consultation, and location of death were not associated with poverty indicators. However, socioeconomic status was associated with treatment in the last month of life; AYAs who resided in areas below 200% FPL were less likely to receive any cancer treatment in their last month of life (aOR=0.53, 95%CI=0.29–0.99), compared to their peers residing in communities above 200% FPL.

Several control variables were related to EOL care. Patients with hematologic malignancies had lower odds of hospice referral (aOR=0.29, 95%CI=0.13–0.67) and higher odds of inpatient location of death (aOR=3.95, 95%CI=1.94–8.03). UI patients had higher odds of hospice referral (aOR=2.77, 95%CI=1.23–6.24). Caucasian patients had higher odds of palliative care consultation (aOR=2.76, 95%CI=1.06–7.19) relative to patients for whom race was not documented in medical records. No other covariates were related to EOL care.

DISCUSSION

Many of the AYA patients in our study experienced poverty, with 70% having at least one poverty indicator. We identified few differences in EOL care between patients in our sample with and without poverty indicators. However, given the high rates of poverty in our sample, our findings may reflect the EOL care experiences of AYAs with limited resources, as well as opportunities to improve care for all AYAs regardless of socioeconomic status. Prior work demonstrates significant challenges in EOL care for AYAs in general,^{8, 10, 11, 24} and our study expands on this work by highlighting areas of special concern for a population experiencing poverty.

Of note, these young cancer patients with limited resources had similarly low rates of hospice referral compared to AYA patients with cancer not living in poverty. Regardless of poverty level, most AYA patients who enrolled in hospice did so in the last week of life. Late hospice referrals can come at the cost of high-quality care for the AYA population as a whole. For example, surrogate markers of advance care planning including documented health care proxy and do-not-resuscitate status were also low, which may further indicate poor integration of advance care planning that could be facilitated in hospice. Furthermore, the lack of early hospice integration may have unmeasured consequences such as complicated bereavement for surviving family members, including young children.

These patients did, however, receive high rates of palliative care consultations with three-quarters of referrals occurring greater than seven days prior to death. Notably, most AYA patients living in poverty received palliative care referrals in the hospital and were less likely to have early outpatient referrals. Prior studies demonstrate that the benefits of palliative care referral for patients with solid tumor cancers is most pronounced when completed early in the outpatient setting (i.e., within eight weeks of incurable diagnosis).^25–29 These early integrated palliative care studies demonstrate improved quality of life, enhanced prognostic awareness, less depression, and decreased caregiver distress. Not a single study has shown harm caused by palliative care integration. Yet, these studies included primarily white, English-speaking adult patients with solid tumor cancer and access to an academic tertiary medical center, which may limit their application to an AYA cancer patient population living in poverty.^25–21 To meet the unique needs of AYA patients with cancer with limited resources, we may need to tailor palliative care delivery and explore innovative integrative strategies.

For instance, one palliative care study in patients with hematologic malignancies receiving hematopoietic stem cell transplantation occurred in the hospital.³⁰ Patients were randomized to receive usual transplant care or palliative care consultations at least twice weekly during their hematopoietic stem cell transplantation. After two weeks, patients who received palliative care experienced less deterioration in quality of life and less depression with impacts extending out to three months.³⁰ Again we observe the importance of palliative care integration early in the course of illness (i.e., at transplantation) rather than reacting to an unplanned hospital admission for treatment toxicity or poorly controlled symptoms. Given the potential for limited access of AYA patients with cancer to outpatient palliative care, the experience in adult cancer patients with hematologic malignancies receiving early inpatient palliative care offers an alternative model that may be more suiting to the AYA patient population. In our study, the small proportion of AYA patients with hematologic malignancies (n=48) were less likely to receive a hospice referral and more likely to die in the hospital. Future studies may consider an early inpatient palliative care intervention for AYA patients with incurable cancer or high symptom burden in accordance with ASCO guidelines.^{17, 24}

Of interest, AYA patients with cancer living in poverty may have poor access to care leading to suboptimal EOL outcomes. A minority of these patients received cancer-directed therapy (e.g., chemotherapy, radiation, surgery, or interventional clinical trial) in the last month of life, with patients from high-poverty areas demonstrating even lower rates. Mirroring the high acuity and unique cancer biology of hematologic malignancies, it is unclear whether lower use of cancer-directed therapy in the last month of life defines quality EOL care in AYAs with incurable cancer. In fact, less cancer-directed therapy at the EOL may represent limited access to the care they need. Moreover, we did not observe any difference in AYA inpatient mortality and poverty measures, which may be a result of limited resources to receive adequate EOL care at home or a preference to die in the hospital.

Strengths of this study include the large number of AYA patients studied living in poverty. Our sample had high racial, ethnic, and geographic diversity, and sites were selected to represent different potential underpinnings of poverty (e.g., urban and rural poverty). Yet, our aim to further characterize the EOL outcomes of AYA patients living in poverty is limited by its retrospective design occurring at three academic sites. Analysis was limited to patients who had died in the study period and did not include a comparator arm. Despite high levels of poverty and efforts to include racial and ethnic minorities, most patients were white. Poverty was defined by zip code and insurance status which may not capture all AYA patients experiencing poverty, and area-level measures cannot speak to individual socioeconomic status. However, the addition of insurance status allowed confirmation of high individual poverty in this young population. We acknowledge that neither measure of poverty is perfect and chose not to evaluate a combination of both ZIP code and insurance in order to be consistent with existing literature. We also describe completion of surrogates of advance care planning such as documentation of health care proxy and code status but cannot further characterize any advance care planning conversations.

Quality metrics for EOL outcomes have been defined for adult patients with solid tumor cancer, and their application to care preferences of young people is unknown. EOL outcomes in AYA patients with incurable cancer remain understudied and undefined. Nonetheless, our findings raise concerns about EOL care quality for AYAs who live in poverty. A deeper exploration of care priorities and unmet needs of AYA patients with cancer experiencing poverty at the EOL should occur prior to targeted interventions.