Identifying the Prevalence and Correlates of Caregiver-Reported Symptoms in Home Hospice Patients at the End of Life

Abstract

Background: Symptoms (e.g., pain, shortness of breath, and fatigue) at the end of life (EoL) are common. Although symptoms can contribute to poor quality of life at the EoL, much remains unknown regarding their prevalence and correlates in home hospice care.

Objectives: To determine the prevalence and correlates of caregiver-reported symptoms in home hospice patients during the last week before discharge using the Edmonton Symptom Assessment Scale (ESAS).

Design: This is a cross-sectional study measuring perceived patient symptoms using caregiver proxy data. Bivariate and multivariate analyses were conducted to examine patient and caregiver characteristics associated with ESAS scores.

Setting/Subjects: Subjects were from an urban nonprofit home hospice organization.

Measurements: Symptoms were measured using the ESAS.

Results: The mean ESAS score was 51.2 (SD ±17.4). In bivariate analyses, higher perceived symptom score was associated with younger patient age (p < 0.001), younger caregiver age (p < 0.001), having a cancer diagnosis (p = 0.006), and lower caregiver comfort level managing symptoms (p < 0.001). Regression model analyses showed that younger patient age (p = 0.0009, p = 0.0036) and lower caregiver comfort level managing symptoms (p = 0.0047, p < 0.0001) were associated uniquely with higher symptom scores.

Conclusions: Multiple symptoms of high severity were perceived by caregivers in the last week on home hospice. Patient age and caregiver comfort level in managing symptoms were associated with higher symptom scores. Further work is needed to improve management and treatment of symptoms in this care setting.

Introduction

Symptoms at the end of life (EoL) are common among terminally ill adults, with as many as 90% of patients experiencing one or more physical or mental health symptoms.^1,2 Although symptoms are prevalent and can lead to distress and suffering, many symptoms can be effectively managed.^1,3–9 In the United States, the Medicare Hospice Benefit allows for services directed at reducing suffering and associated symptoms among terminally ill patients with life expectancies of six months or less. This benefit encompasses a multidisciplinary team of providers (e.g., nurses, doctors, social workers, and spiritual counselors) that provides medical, emotional, and spiritual support in a home or facility setting.

Much remains unknown about symptom prevalence and correlates in hospice. A large cross-sectional descriptive study in 2001 by Kutner et al.¹ examined the prevalence of 33 symptoms among 348 hospice patients using proxy data provided by hospice staff and found significant symptom prevalence in their sample. The most commonly reported symptoms were pain, lack of energy, and lack of appetite. In a subsequent study, the same investigators analyzed data provided by 11 hospice/palliative care organizations and found no association between symptom distress and patient characteristics or physical symptoms and quality-of-life scores. In another study examining cancer patients newly admitted to home hospice care, McMillan and Small⁶ found lack of energy, pain, and shortness of breath to be the most frequently reported symptoms by patients. The authors demonstrated an association between Memorial Symptom Assessment Scale (MSAS) symptom distress and lower quality-of-life (Hospice Quality-of-Life Index) scores.

Ascertaining the prevalence and intensity of symptoms along with identifying potentially modifiable factors associated with high symptom scores constitutes important steps toward improving home hospice and EoL care. Previous studies have not examined caregiver variables and how they may impact symptom scores in the home hospice setting. Caregivers often play a critical role providing care to patients in the home hospice setting and are often responsible for performing health care activities such as administering medications when patients are unable to.

The objective of our study is to ascertain the prevalence and correlates of caregiver-reported symptom severity (using the Edmonton Symptom Assessment Scale [ESAS]) in home hospice patients during the last week before death or discharge using reports from caregiver proxy. We hypothesize that caregiver's level of comfort managing symptoms will be directly correlated with ESAS scores.

Methods

Study design and site

This cross-sectional study examined patient symptoms using caregiver proxy data in the home hospice setting. The study was approved by the Institutional Review Boards (IRB) of Weill Cornell Medicine and the Visiting Nurse Service of New York.

The Visiting Nurse Service of New York Hospice and Palliative Care (VNSNYHPC) organization is a nonprofit hospice that serves >1000 patients daily and delivers hospice care to patients in the New York City area. In addition to providing home visits by an interdisciplinary team of physicians, nurses, social workers, and spiritual care counselors, VNSNYHPC provides enrolled patients a medication kit for pain and symptom management and educational materials that describe available support services. A round-the-clock phone service is staffed by a hospice on-call team or a hospice nurse supervisor, and a hospice nurse, nurse practitioner, or physician may be dispatched to the home based on the needs of the patient and family.

Sample assembly

During the study period (April 2017 through February 2018), VNSNYHPC staff generated a weekly list that contained the names of all patients discharged from home hospice in the preceding week. Additional information provided by the hospice agency included patient demographic data (age, gender, race/ethnicity, hospice diagnosis—cancer vs. noncancer, length of stay, and reason for discharge), a home hospice utilization variable (use of continuous home hospice care during the last week on hospice), as well as caregiver contact information (name, address, and phone number). Caregivers received a mailed letter introducing the study and informing them to expect a call in two weeks from a member of the research team. Eligible caregiver participants had to be 18 years or older, English speaking, and listed as a primary caregiver (e.g., family or friend) at the time the patient was admitted to the VNSNYHPC service. A trained research assistant (A.S.) called potential participants described the study, and obtained verbal consent from interested individuals. The research assistant was trained in administering surveys over the phone and in qualitative interviewing. Caregivers were allowed to skip questions if they were uncomfortable with answering or if they thought the question did not pertain to the patient or themselves. Given the sensitive nature of these calls, a physician was on call to help address any adverse events that occurred. No adverse events were reported during this study. Of the 1848 caregivers called, 804 (44%) did not answer the phone after three attempts to contact them, 653 (35%) declined participation, and 391 (21%) completed the phone survey interview.

Data collection

A semistructured interview guide was administered during the phone call after consent was obtained. Caregivers received a $25 gift card for their participation.

Dependent variable

Caregivers served as proxy respondents to measure patient symptoms using the ESAS. Caregivers were asked to recall whether the patient experienced any of the nine symptoms included in the ESAS (i.e., pain, shortness of breath, nausea, tiredness, drowsiness, lack of appetite, depression, anxiety, and well-being) and to rate their intensity on a 0 to 10 scale during the patient's last week on hospice. For descriptive purposes, we dichotomized individual symptom scores of 0 to 6 to indicate low-to-moderate symptom severity and scores of 7 to 10 to indicate high symptom severity. The ESAS has evidence of good psychometric properties,¹⁰ and been used in numerous studies of patients with terminal illnesses and those at the EoL.^11–13 Although obtaining assessments from patients would be the gold standard, given the retrospective nature of the study, this was not feasible. There has been established support for the validity of using proxy respondents' to assess symptoms.^14–16

Patient covariates

Patient-level data included age, gender, race/ethnicity, hospice diagnosis (cancer vs. noncancer), reason for discharge, use of continuous home hospice care during the last week on hospice, and length of hospice stay.

Caregiver covariates

The following caregiver demographic data were collected during phone interviews: age, gender, race/ethnicity, relationship with the patient, education level, and average hours of caregiving provided per day during the last seven days on home hospice. Since there was no validated measure of caregiver comfort in managing patient's symptoms, we asked caregivers, “how would you rate your level of comfort managing (patient's name) symptoms during the last week on home hospice care?” Response items ranged from 1 (very comfortable) to 5 (very uncomfortable).

All data obtained from the medical record and through survey questions were entered into Research Electronic Data Capture (REDCap), a secure web application for building and managing databases.

Statistical approach

For the study objective, bivariate analyses were conducted to examine patient and caregiver characteristics associated with ESAS scores. Point-biserial correlation coefficients were used for binary variables and Pearson correlation coefficients were employed for ordinal and continuous variables. Multivariate regression analyses were conducted to evaluate the unique association between the ESAS and independent variables. Two separate methods for treating missing data were used. The first model did not use imputation, examined only cases without missing data; whereas the second model used in sensitivity analyses imputed the median value for missing data. Linear regression was used to test the association of the ESAS and the potential covariates. Bivariate analyses were performed using IBM SPSS Statistics version 25 (IBM Corp, 2016) and multivariate analyses were performed using SAS (SAS Institute, Inc., SAS Version 9.4., Cary, NC: SAS Institute, Inc.; 2015).

Results

Patient and caregiver demographic data are presented in Table 1. Study patients had a mean age of 83 years, most had a noncancer diagnosis (n = 235, 60.1%), were primarily female (n = 250, 63.9%), and non-Hispanic white (n = 210, 55.7%). A minority of patients (n = 35, 9.0%) received continuous home hospice care services. The average length of stay in hospice was 98 days. Caregivers had a mean age of 59 years and most were children of the patients (n = 233, 59.7%), female (n = 297, 76.0%), non-Hispanic white (n = 180, 51.6%), and had a college education or greater (n = 271, 76.1%). Caregivers reported providing an average of 14 hours of patient care per day during the patient's last week on hospice.

Table 1.

Patient and Caregiver Characteristics

	N (%)	Mean (SD)
Patient age	391	83 (14)
Patient gender
Male	141 (36)	—
Female	250 (64)	—
Patient race/ethnicity
White	210 (54)	—
Black	63 (16)	—
Hispanic	75 (19)	—
Asian	29 (7)	—
Other/undisclosed	14 (4)	—
Hospice diagnosis
Cancer	156 (40)	—
Noncancer	235 (60)	—
Length of stay (days)	391	98 (178)
Reason for discharge from hospice
Death	351 (90)	—
Hospitalization	24 (6)	—
Other	16 (4)	—
Received continuous home care during the last week on hospice	35 (9)	—
Caregiver age	351	59 (13)
Caregiver gender
Male	94 (24)	—
Female	297 (76)	—
Caregiver race/ethnicity
White	180 (46)	—
Black	57 (15)	—
Hispanic	75 (19)	—
Asian	28 (7)	—
Other	9 (2)	—
Not specified	42 (11)	—
Caregiver relationship with patient
Child	233 (60)	—
Spouse	69 (18)	—
Relative	59 (15)	—
Grandchild	11 (3)	—
Friend	11 (3)	—
Parent	3 (1)	—
Caregiver education level
High school	84 (24)	—
College	168 (47)	—
Graduate school	103 (29)	—
Average hours per day spent caregiving during the last week on hospice	363	14.4 (9.5)
Comfort managing symptoms (1—very comfortable, 5—very uncomfortable)	349	2.3 (1.4)

Death was the major reason for discharge (n = 351, 89.8%), followed by hospitalization (n = 24, 6.1%) then other (n = 16, 4.1%). Approximately 40% (n = 155) of caregivers stated they were very comfortable managing the patient's symptoms during the last week they received home hospice care.

ESAS scores

Table 2 gives descriptive statistics for the nine symptoms that make up the ESAS. The overall mean ESAS score was 51.2 (SD = 17.4). Tiredness was the symptom with the largest number of scores of seven or greater (n = 300, 82.4%), indicating high symptom severity, followed by lack of appetite (n = 291, 80.4%), drowsiness (n = 236, 67.6%), pain (n = 164, 45.8%), and shortness of breath (n = 149, 40.1%). High psychological symptom severity, including both anxiety (n = 129, 38.1%) and depression (n = 121, 39.0%), was reported by a large minority of caregivers.

Table 2.

Edmonton Symptom Assessment Scale Scores Reported by Caregivers

	Mean (SD)	High symptom severity—score ≥7 (n)	%
Pain (N = 358)	5.35 (3.75)	164	46
Shortness of breath (N = 372)	4.99 (3.53)	149	40
Nausea (N = 366)	1.92 (2.96)	42	11
Tiredness (N = 364)	8.30 (2.50)	300	82
Drowsiness (N = 349)	7.22 (3.33)	236	68
Lack of appetite (N = 362)	8.17 (2.78)	291	80
Depression (N = 310)	4.52 (3.89)	121	39
Anxiety (N = 339)	4.52 (3.77)	129	38
Well-being (N = 351)	6.09 (3.49)	180	51

Correlates of ESAS score

Table 3 gives bivariate correlation coefficients and p-values between total ESAS score and patient, caregiver, and hospice utilization variables. Higher ESAS score was associated with younger patient age (p < 0.001), having a hospice diagnosis of cancer (p = 0.006), younger caregiver age (p < 0.001), and caregivers who were less comfortable managing patient symptoms (p < 0.001). Of note, we found a significant difference on ESAS scores between cancer (mean = 54.19, SD = 17.36) and noncancer (mean = 49.12, SD = 17.14) hospice patients with a p-value of 0.006.

Table 3.

Patient and Caregiver Bivariate Associations with the Edmonton Symptom Assessment Scale

	r	p
Patient age	−0.271	<0.001
Patient gender
Male
Female	−0.058	0.269
Patient race/ethnicitya
White	0.017	0.758
Black	0.037	0.491
Hispanic	0.028	0.606
Asian	−0.120	0.025
Other/undisclosed	0.067	0.200
Hospice diagnosis
Cancer
Noncancer	−0.144	0.006
Length of stay (days)	−0.090	0.089
Reason for discharge from hospicea
Death	−0.057	0.276
Hospitalization	0.046	0.387
Other	0.033	0.535
Received continuous home care during the last week on hospice	−0.064	0.225
Caregiver age	−0.237	<0.001
Caregiver gender
Male
Female	−0.033	0.533
Caregiver race/ethnicitya
White	−0.011	0.835
Black	0.011	0.845
Hispanic	0.025	0.652
Asian	−0.073	0.182
Other	0.072	0.185
Not specified	−0.036	0.493
Caregiver relationship with patienta
Child	−0.022	0.679
Spouse	0.047	0.371
Relative	−0.057	0.276
Grandchild	0.085	0.108
Friend	−0.011	0.832
Parentb
Caregiver education levela
High school	−0.019	0.732
College	0.068	0.210
Graduate school	−0.062	0.251
Average hours per day spent caregiving during the last week on hospice	0.002	0.972
Comfort managing symptoms (1—very comfortable, 5—very uncomfortable)	0.206	<0.001

^aDummy variables were used to code each category (as 1 vs. all else) for race/ethnicity, reason for discharge, relationship with patient, and education.

^bNot computed due to sparse data.

Table 4 gives the results of linear regression models predicting total ESAS score with and without imputation for missing covariate data. Results were similar for both models with the exception of caregiver age. In Model 1, younger patient age (p = 0.0009) and caregivers who were less comfortable managing symptoms (p = 0.0047) were significantly associated with higher ESAS score. In Model 2, younger patient age (p = 0.0036), younger caregiver age (p = 0.0077), and caregivers who were less comfortable managing symptoms (p < 0.0001) were significantly associated with higher ESAS score.

Table 4.

Results of Linear Regression Predicting the Edmonton Symptom Assessment Scale

	Model 1 (n = 286)			Model 2 (n = 362)
	Estimate	Standard error	p	Estimate	Standard error	p
Intercept	94.1392	9.1816	<0.0001	88.3832	8.4478	<0.0001
Patient age	−0.3441	0.1022	0.0009	−0.2705	0.0924	0.0036
Patient female	−1.8708	2.0722	0.3673	−1.4944	1.9382	0.4412
Patient white	2.3926	2.0160	0.2362	1.5173	1.9041	0.4261
Cancer hospice diagnosis	0.1809	2.2304	0.9354	−0.5141	2.0934	0.8062
Length of stay	−0.0045	0.0052	0.3914	−0.0040	0.0051	0.4359
Death discharge reason	−2.3528	3.2494	0.4696	−0.7166	2.9984	0.8112
Received continuous home care during the last week on hospice	1.8016	3.2158	0.5757	2.8792	3.0503	0.3459
Caregiver age	−0.1487	0.1023	0.1472	−0.2533	0.0945	0.0077
Caregiver female	−0.5214	2.3257	0.8228	−1.3386	2.1491	0.5338
Caregiver spouse	−2.6375	3.3992	0.4384	0.5321	3.1292	0.8651
Caregiver education	−2.1009	1.3055	0.1086	−0.9644	1.2430	0.4384
No. of hours with patient per day	−0.1704	0.1147	0.1383	−0.1167	0.1050	0.2673
Comfort level managing patients' symptoms during last week on hospice care	1.9954	0.7014	0.0047	2.6210	0.6556	<0.0001

Model 1 has a reduced sample size because imputation was not used to replace missing covariate data. In Model 2, missing values in the covariates were replaced by the median of that variable.

Discussion

Our study sheds further light on symptoms experienced by terminally ill patients receiving home hospice care. We found that many patients had high caregiver perceived symptom severity scores. Patients' age and caregivers' level of comfort managing symptoms were associated with higher caregiver-reported ESAS score, such that patients of younger age experienced higher symptom burden, and perceived patient symptom burden was higher for caregivers who reported less comfort managing symptoms.

This study confirms and adds to the data from previous work on patient symptoms at the EoL. Our symptom profile compares similarly with those reported in studies by Kutner et al. and McMillan and Small, although those studies used the MSAS to measure symptoms while we used the ESAS.^1,6 Tiredness, lack of appetite, and drowsiness were symptoms that received the highest proportion of scores 7 or greater. Given that ∼90% of patients in the sample died at the time of hospice discharge, we were not surprised with these results. However, we found that pain and shortness of breath, symptoms that can often be medically managed, to be rated by almost half of caregivers with scores of 7 or greater. Perceived mental health (i.e., depression and anxiety) scores of home hospice patients were reported to be in the severe range (scores of 7 or greater) by ∼40% of caregivers. The fact that multiple physical and psychological symptoms were rated in the high range by caregivers during the last week on hospice indicates the challenges of managing symptoms effectively in this target population and supports the need for future efforts to better understand, evaluate, and improve symptom management in this care setting.

Our sample had an overall mean ESAS score of 51.2 (SD = 17.4). This score appears to be on the higher end of what has been reported in the literature. One study looking at patients with metastatic cancer in the outpatient palliative care setting reported an average total ESAS score of 39.5 (SD = 17.8).¹⁷ In another study examining symptoms of patients admitted to an inpatient palliative care unit in Italy, the mean ESAS score on the day of admission was reported to be 33.9 (SD = 16.2). Although these examples provide comparative data, it is important to note that the two studies mentioned obtained patient-reported symptoms and were not conducted in the home hospice setting or in the United States.

Our linear regression analyses identified interesting findings, in particular the associations between ESAS score and caregiver age, patient age, and level of caregiver comfort managing symptoms. With regard to the association between younger caregiver age and higher ESAS scores, two international studies have shown that younger caregivers have worse perceptions of caregiving and feel less comfortable in their role compared with their older counterparts, which may impact on their symptom reporting.^18,19 We also found that younger patients had higher caregiver-reported ESAS scores, which we postulate could be related to the disease profile that younger terminally ill patients experience compared with older patients. Although we only examined patients with cancer to noncancer diagnoses, future studies are needed to compare disease-specific diagnoses (specific cancer diagnoses, congestive heart failure [CHF], chronic obstructive pulmonary disease [COPD], dementia, etc.) with sufficient sample sizes to identify symptom profiles/clusters at the EoL in the home setting. Examining specific symptom (pain, nausea, etc.) scores and how differences in score vary by age will be a future step that may provide further insight into our finding.

Level of caregiver comfort managing symptoms was also correlated with caregiver-reported ESAS scores. Our sample consisted of unpaid caregivers (family or friends) who provided many hours of care (on average 14 hours a day). Although we did not ask caregivers about their specific caregiving responsibilities, we suspect that many may be overseeing medication administration or directly administering medication to the patient in an effort to alleviate distressing symptoms. We believe that further research is needed to better explore this issue and find ways to improve caregivers' abilities to feel more comfortable and capable of managing symptoms at the EoL. Some work exists in this area. Lau et al. conducted interviews with hospice providers and family caregivers and developed a construct of effective medication management that involves teamwork, organization, symptom knowledge, medication knowledge, and personhood.²⁰ The EMPOWER⁸ and Campbell and McErlane²¹ studies, which are educational interventions designed to help family caregivers deal with pain and dyspnea in the home hospice setting, respectively, have shown promise in educating and guiding caregivers in better managing symptoms.

Our study has several limitations. Given the design of the study, we interviewed caregivers instead of patients to measure ESAS scores. Although it would be ideal to call patients to obtain ESAS information, we had concerns about recruitment (e.g., being able to obtain patient consent and adequate sample size) along with the potential burden of administering surveys to patients at this phase of life. Therefore, as part of the design of the study, we elected to collect proxy data retrospectively.^14–16 Other limitations include the lack of established validity of a retrospectively completed proxy ESAS along with poor to fair agreement in concordance between patient and caregiver assessment of symptoms of drowsiness, lack of energy, feeling sad, and feeling nervous.¹⁴ In addition, caregivers' level of comfort managing the patient's symptoms along with recall bias may impact reporting of symptoms. Our refusal and nonresponse rates were high that could have biased the sample as well. This relatively low yield reveals the recruitment challenge faced by researchers who conduct EoL/hospice research.²² Of note, our patients' average length of stay was 98 days, which is longer than the national average of 71 days and a majority of participants were either college or graduate school educated. Lastly, we only recruited from one nonprofit urban hospice organization that may not reflect the national makeup of caregivers and patients receiving home hospice care.

In conclusion, our study details the prevalence and degree of symptoms during the patient's last week on home hospice care based on caregiver proxy data. We found that multiple symptoms were present and many were perceived to be of high severity. Younger patients and caregiver level of comfort managing symptoms were associated with higher caregiver-reported ESAS scores. Further work is needed to advance the field of symptom management in the home hospice population to better support patients and caregivers while reducing suffering at the EoL.

Funding Information

The author(s) disclose receipt of the following financial support for the research, authorship, and/or publication of this article: Dr. Phongtankuel is supported by a grant from the National Institute of Aging (R03 AG053284-01) and Weill Cornell Medical College Clinical and Translational Science Center. Dr. Reid is supported by the following grants from the National Institute on Aging: Grant Nos. P30AG022845 and K24AGO53462. Dr. Prigerson is supported by a grant from the National Cancer Institute (Grant No. R35 CA197730).

Author Disclosure Statement

No competing financial interests exist.