Association Between Palliative Care and Death at Home in Adults With Heart Failure.

Background Palliative care is associated with improved symptom control and quality of life in people with heart failure. There is conflicting evidence as to whether it is associated with a greater likelihood of death at home in this population. The objective of this study was to describe the delivery of newly initiated palliative care services in adults who die with heart failure and measure the association between receipt of palliative care and death at home compared with those who did not receive palliative care. Methods and Results We performed a population-based cohort study using linked health administrative data in Ontario, Canada of 74 986 community-dwelling adults with heart failure who died between 2010 and 2015. Seventy-five percent of community-dwelling adults with heart failure died in a hospital. Patients who received any palliative care were twice as likely to die at home compared with those who did not receive it (adjusted odds ratio 2.12 [95% CI, 2.03-2.20]; P<0.01). Delivery of home-based palliative care had a higher association with death at home (adjusted odds ratio 11.88 [95% CI, 9.34-15.11]; P<0.01), as did delivery during transitions of care between inpatient and outpatient care settings (adjusted odds ratio 8.12 [95% CI, 6.41-10.27]; P<0.01). Palliative care was most commonly initiated late in the course of a person's disease (≤30 days before death, 45.2% of subjects) and led by nonspecialist palliative care physicians 61% of the time. Conclusions Most adults with heart failure die in a hospital. Providing palliative care near the end-of-life was associated with an increased likelihood of dying at home. These findings suggest that scaling existing palliative care programs to increase access may improve end-of-life care in people dying with chronic noncancer illness.

Clinical Perspective

What Is New?

This cohort study of 74 986 community‐dwelling adults who died with heart failure found that 75.4% died in a hospital and 47.1% received palliative care.

Palliative care often started ≤30 days before death and was predominantly delivered by generalist physicians during transitions of care across different settings.

Initiation of palliative care was associated with a more than 2‐fold increase in the probability of death at home compared with those who did not receive palliative care (adjusted odds ratio 2.12 [95% CI, 2.03–2.20]; P<0.01).

What Are the Clinical Implications?

A significant gap exists in access to high‐quality end‐of‐life care for adults with heart failure.

Current models of palliative care delivery are associated with death at home.

These findings suggest that scaling existing palliative care programs to increase access may improve end‐of‐life care in people dying with chronic noncancer illness.

Introduction

Heart failure (HF) is a leading cause of healthcare use and expenditure.1 Annually, there are over 1 million visits to the emergency department and more than 4 million hospital admissions attributable to HF in North America.2, 3 Despite advances in cardiovascular care, the median 5‐year survival for hospitalized patients with HF is only 25%.4 Up to 30% of people die within 1 year following hospitalization for HF,5 and the median time to death following admission to a hospital for HF is 2.4 years.6 One in 5 adults with HF are readmitted to the hospital within 30 days;7 50% are rehospitalized for cardiovascular causes within 1 year, and 30% of these admissions are for HF specifically.5 These repeated hospitalizations near the end of life have a negative effect on quality of life.8 In light of this evidence new paradigms are needed to deliver high‐quality end‐of‐life care for people with HF.

Palliative care focuses on the care of people who are suffering from serious illness with a goal of improving the quality of life for them and their caregivers.9 Cardiovascular societies recommend the integration of palliative care early in the course of a person's disease to improve symptoms, pain, and quality of life instead of at the end of life.10, 11, 12 Although palliative care is associated with improved quality of life and reduced symptom burden in people with HF, there is conflicting evidence as to whether palliative care is associated with a reduction in overall healthcare utilization or an increased likelihood of death at home.9, 11, 13, 14, 15, 16, 17, 18 Only half of the published studies demonstrated an association between the receipt of palliative care and death at home.18 Although the evidence for patient‐reported preferences on location of death in patients with HF is less well known than for those with cancer, 40% of people with serious illness report that they value the health services available to care for them in their home,19 and up to 87% of people prefer to die at home.20, 21 Death at home has been identified as a quality indicator for the delivery of palliative care.22, 23

The evidence for palliative care is presently skewed toward patients with cancer, which may limit its applicability to those with noncancer illness.9, 11 Current models for the delivery of palliative care may need to be redesigned to address the often unpredictable clinical course for patients with chronic noncancer illness such as HF in comparison to those with cancer.24 Few studies have described population‐based end‐of‐life care and the delivery of palliative services care in people dying with HF.25, 26, 27 The objective of this study was to describe healthcare use near the end of life in adults with HF and to measure the association between receipt of newly initiated palliative care and death at home.

Methods

Access to Data

The data set from this study is held securely in coded form at ICES (formerly known as the Institute for Clinical Evaluative Sciences). Although data‐sharing agreements prohibit ICES from making the data set publicly available, access may be granted to those who meet prespecified criteria for confidential access, available at http://www.ices.on.ca/DAS. The full data set creation plan and underlying analytic code are available from the authors on request, with the understanding that the programs may rely on coding templates or macros that are unique to ICES.

Study Design, Setting, and Data Sources

We conducted a population‐based cohort study in Ontario, Canada, using linked clinical and health administrative databases. Ontario is Canada's most populous province with nearly 14 million residents. All residents of Ontario have universal access to hospital care and medically necessary physicians’ services, and those aged ≥65 years of age are provided universal prescription drug insurance coverage. The administrative data sets used in this study were linked using encoded identifiers at the patient level and analyzed at ICES. These data sets are routinely used to conduct studies involving palliative care (Data S1).28, 29, 30, 31, 32 The use of data in this project was authorized under Section 45 of Ontario's Personal Health Information Protection Act and was approved through a privacy impact assessment process that does not require review by a Research Ethics Board.

Study Cohort

Our cohort included Ontario adults with an existing diagnosis of HF who were in their last 2 years of life and died between January 1, 2010 and December 31, 2015. A diagnosis of HF was identified by the presence of 1 hospital record or physician claim, followed by a second record from either source within 1 year. This method has been previously validated with a sensitivity of 84.8% and a specificity of 97.0%.33 We only included patients whose HF diagnosis was made ≥2 years before death because we are unable to measure the severity of heart failure using administrative data. We were specifically interested in end‐of‐life care as it related to people with chronic advanced HF and not, for example, end‐of‐life care for people with metastatic cancer who subsequently developed new acute HF in close proximity to death. Our study's index date (ie, the start of the observation period for our primary exposure) was, therefore, 2 years before the date of death.

People who were admitted to a nursing home in the last 2 years of life were excluded because the delivery of palliative care in this setting is not well captured using administrative data, and because the level of care in nursing homes typically allows people to die there.29 The exclusion of people residing in this care setting may influence the absolute number of patients who receive palliative care based on our method of capturing its delivery. We also excluded people who received consultation with palliative care before the index date using a 5‐year lookback to determine prior exposure. This “new‐user” design is similar to pharmacoepidemiology studies and minimizes bias by restricting analysis to people who are initiating treatment when outcome risks are likely to vary over time.34

Patient Characteristics

We measured demographic and clinical variables including duration of HF, comorbidities and chronic conditions,35 use of acute healthcare services in the year before the study's index date, cardiac‐specific procedures in the 7 years before the index date, and, in the case of adults aged ≥66 years, the use of cardiovascular medications within 6 months of the index date. Cause of death was grouped according to a modified form of Becker leading cause of death.36 We also determined the presence of functional decline in the year before the index date in a subset of adults who had completed homecare assessments (see Data S1 for details).

Receipt of Palliative Care

The main exposure was a person's first encounter with palliative care in the last 2 years of life. We identified the delivery of palliative care based on a unique set of widely used physician claims fee codes.28, 29, 30, 32, 37, 38, 39, 40, 41, 42 These codes were created to specifically indicate the delivery of palliative care by both generalist and specialist palliative care physicians and are related to therapies not intended to be curative, such as symptom management or counseling. Therefore, we captured a “palliative approach” to patient care and performed secondary analyses that measured potential differences between generalist and specialist palliative care. A physician was deemed to be a palliative care specialist based on a previously validated method with a sensitivity of 76.0% and specificity of 97.8%.37 We measured the care setting and timing of initial consultation, the physician specialty for each palliative care encounter, the total number of palliative care encounters, as well as the models and setting of palliative care delivery.

We measured the delivery of 4 different models of physician palliative care delivery that were developed from an iterative process involving 8 healthcare providers (physicians and nurses) with expertise in palliative care.31 These models are derived using the proportion of palliative care fee codes claimed by physicians, which classifies them as a palliative care specialist or palliative care generalist described above. The 4 models of palliative care were (1) no physician‐based palliative care (0% of claims are palliative fee codes), (2) generalist palliative care (care is provided from either a primary‐care physician or medical or surgical specialist such as an oncologist or general surgeon whose annual billing is comprised of ≤10% of palliative care fee codes), (3) consultation palliative care (care provided by both palliative care specialists and generalists), and (4) specialist palliative care (a physician whose annual billing is comprised of >10% palliative care fee codes). The setting of palliative care was determined using physician claims, which included the locations in which care was delivered. There were 6 possible settings where palliative services could be delivered: inpatient, outpatient, home‐based, multiple locations, third‐party case management, and other (when the location is unknown). Third‐party case‐management fee codes were deemed to be a separate palliative care visit, even though they may be billed in conjunction with care delivered in other settings (Data S1).

Outcomes

The primary outcome was the location of death, which included hospital, home, or other. Deaths that occurred in a dedicated palliative care unit or hospice were categorized as other because there are very few palliative care units and hospices in Ontario.

Secondary outcomes included the number of hospitalizations and visits to the emergency department in the past 30, 90, and 360 days of life; the cumulative number of days spent in the hospital and in the emergency department in the past 30, 90, and 360 days of life; and the proportion of patients who underwent coronary artery bypass surgery, percutaneous coronary intervention, or newly implanted pacemakers or intracardiac defibrillators within the past 2 years of life. Incident use of these devices was determined using a 7‐year lookback to ensure no prior device had been implanted.

Statistical Analyses

The association between receipt of newly initiated palliative care and the likelihood of death at home versus hospital was measured using multivariable multinomial logistic regression. The association between receipt of newly initiated palliative care and hospitalization in the last 30 days of life was measured using multivariable logistic regression. All models were adjusted for covariates with a standardized difference of ≥0.1, which indicates imbalance among the study groups.43 These variables included age, sex, rurality, neighborhood income, duration of HF, and the presence of metastatic cancer, dementia, diabetes mellitus, myocardial infarction, and stroke. We did not account for clustering by physician or facility because most people receive end‐of‐life care from many physicians in multiple facilities.

We performed a prespecified subgroup analysis in all adults who had received at least 1 palliative care visit to describe the model‐specific factors of palliative care delivery associated with dying at home using multivariable multinomial logistic regression. We used a chi‐squared test to compare categorical variables describing models of palliative care between death at home and death in hospital. We stratified the intensity of palliative care as ≤4 or >4 visits because 58.4% of patients in our cohort had ≤4 palliative care visits. We stratified the timing from initial palliative care consultation until death based on previously recognized time frames: (1) “Optimal,” between 2 years and 6 months before death; (2) “Appropriate,” between 6 months and 30 days before death; and (3) “Late,” <30 days before death.44

Because populations are aging and the burden of HF in the elderly is growing, we performed 2 sensitivity analyses measuring the association between palliative care and location of death in an additional 10 065 and 15 194 people with a more recent diagnosis of heart failure of ≥1 year and ≥6 months before death, respectively.

All analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC) and used a 2‐sided type 1 error rate of 0.05 as the threshold for statistical significance.

Results

Baseline Characteristics

There were 116 780 adults with HF who died during the study period, and 36 798 subjects were excluded from the study because they were institutionalized in a nursing home during the study period, indicating the substantial care needs of these patients. The final cohort consisted of 74 986 people (Figure 1).

Figure 1

Flow diagram for the creation of the study sample. All adults with a diagnosis of heart failure made at least 2 years before their death were assessed for inclusion in the study. Patients who received their first consultation with palliative care and who did not reside in a nursing home at any point during the study period were included and subsequently divided into 2 groups: those who received palliative care and those who did not receive palliative care. OHIP indicates Ontario Health Insurance Plan.

Subjects in the cohort were typically older men (median age 78 years) with multiple cardiovascular risk factors who had had HF for a median of 7 years. The majority were prescribed HF‐related therapies including diuretics.12, 45 Cardiovascular disease was the leading cause of death, followed by cancer and respiratory disease (Table 1).

Table 1

Characteristics by Receipt of Palliative Care of Adults Dying With Heart Failure in Ontario Between 2010 and 2015

	Receipt of Palliative Care		Weighted Standardized Differencea
	No (N=39 694)	Yes (N=35 292)
Age, median (IQR), y	79 (71–86)	82 (75–87)	0.25
Female sex, n (%)	16 999 (42.8)	17 092 (48.4)	0.11
Duration of heart failure, median (IQR), y	7.1 (4.2–11.5)	7.1 (4.1–11.6)	0
Functional decline, n (%)b	5284 (13.3)	6117 (17.3)	0.11
Chronic conditions, n (%)
Arrhythmia	18 755 (47.2)	17 221 (48.8)	0.03
Chronic kidney disease	6106 (15.4)	5083 (14.4)	0.03
Diabetes mellitus	12 753 (32.1)	10 235 (29.0)	0.07
Hypertension	35 508 (89.5)	31 748 (90.0)	0.02
Coronary artery disease	28 296 (71.3)	24 638 (69.8%)	0.03
Previous myocardial infarction	8229 (20.7)	6437 (18.2)	0.06
Stroke	3145 (7.9)	3101 (8.8)	0.03
Primary cancer	2209 (5.6)	2937 (8.3)	0.11
Metastatic cancer	269 (0.7)	669 (1.9)	0.11
Chronic obstructive pulmonary disease	8513 (21.4)	7218 (20.5)	0.02
Dementia	1215 (3.1)	1497 (4.2)	0.06
Depression/anxiety	5228 (13.2)	5215 (14.8)	0.05
Cardiovascular medications
Antiplatelet	6573 (16.6)	5867 (16.6)	0
Anticoagulant	11 917 (30.0)	11 362 (32.2)	0.05
ACEi/ARB	24 203 (61.0)	22 098 (62.6)	0.03
β blocker	20 255 (51.0)	18 543 (52.5)	0.03
MRA	4542 (11.4)	4047 (11.5)	0
Furosemide	21 292 (53.6)	19 651 (55.7)	0.04
Digoxin	6316 (15.9)	5874 (16.6)	0.02
Metolazone	963 (2.4)	899 (2.5)	0.01
Cardiovascular devices/proceduresc
CABG	614 (1.5)	471 (1.3%)	0.02
PCI	750 (1.9)	563 (1.6%)	0.02
Pacemaker	612 (1.5)	593 (1.7%)	0.02
ICD	328 (0.8)	228 (0.6%)	0.02
Cause of death, n (%)
Cardiovascular disease	19 721 (49.7)	11 684 (33.1)	…
Cancer	2623 (6.6)	10 233 (29.0)	…
Sepsis	688 (1.7)	397 (1.1)	…
Diabetes mellitus	1956 (4.9)	1085 (3.1)	…
Dementia	640 (1.6)	1186 (3.4)	…
Falls	1102 (2.8)	822 (2.3)	…
Respiratory diseases	5373 (13.5)	4053 (11.5)	…
Gastrointestinal diseases	1929 (4.9)	1259 (3.6)	…
Genitourinary diseases	1726 (4.3)	1691 (4.8)	…
All other causes	3936 (9.9)	1882 (5.3)	…

ACEi indicates angiotensin‐converting enzyme inhibitor; ARB, angiotensin receptor blocker; CABG, coronary artery bypass graft; ICD, implanted cardioverter‐defibrillator; IQR, interquartile range; MRA, mineralocorticoid receptor antagonist; PCI, percutaneous coronary intervention.

Weighted standardized differences compare baseline characteristics of the study groups. A standardized difference less than 0.1 indicates good balance between the study groups for a given covariate.43

For people with a completed homecare assessment within the last 2 years of life. Functional decline was defined as patients who required a new homecare assessment in the 1 year before the index date or as a 1‐point increase on their activities of daily living scale from prior homecare assessments.

A 7‐year lookback was used to determine the prior presence of these devices and procedures.

Receipt of Palliative Care, Death at Home, and Healthcare Use

There were 35 292 (47.1%) people who received at least 1 palliative care visit in the 2 years preceding death. Although the baseline characteristics were similar between the 2 groups, those who received palliative care were slightly older and more likely to be female and to have comorbid cancer (Table 1).

Overall, 75.4% of patients died in a hospital. The proportion of people who received palliative care and died at home was more than twice that of those who did not receive palliative care (23.0% versus 10.7%). Palliative care was associated with a more than 2‐fold likelihood of dying at home than in a hospital, compared with those who did not receive palliative care (adjusted odds ratio 2.12 [95% CI, 2.03–2.20]; P<0.01) (Table 2). The magnitude of the association was larger in those with a more recent diagnosis of heart failure of ≥1 year and ≥6 months before death (adjusted odds ratio 3.06 [95% CI, 2.93–3.2] and 2.91 [95% CI, 2.79–3.04], respectively).We also found that being female (adjusted odds ratio 1.11 [95% CI, 1.06–1.15]; P<0.01), having metastatic cancer (adjusted odds ratio 1.68 [95% CI 1.44–1.95]; P<0.01), and having dementia (adjusted odds ratio 1.37 [95% CI 1.25–1.51]; P<0.01) all were associated with an increased likelihood of death at home.

Table 2

Association Between Receipt of Palliative Care and Location of Death, Hospitalization, or Emergency Department Visits in Adults Dying With Heart Failure in Ontario Between 2010 and 2015

	Unadjusted Odds Ratio (95% CI)	Adjusteda Odds Ratio (95% CI)
Death at home (vs hospital)	2.22 (2.13–2.32)	2.12 (2.03–2.20)b
Hospitalization in the last 30 days of life	1.26 (1.22–1.3)	1.27 (1.23–1.31)b
Emergency department visitc in the last 30 days of life	0.76 (0.73–0.78)	0.77 (0.74–0.79)b

Models were adjusted for age, sex, rurality, neighborhood income, duration of heart failure, and the presence of metastatic cancer, dementia, diabetes mellitus, myocardial infarction, and stroke.

P<0.01.

Emergency department visits not resulting in hospital admission.

Palliative care was also associated with increased odds of hospitalization and a larger number of days in hospital in the last 30, 90, and 360 days of life (Table S1). Fewer than 3% of all patients underwent invasive procedures or device implantation during the study period, but palliative care was associated with lower rates of these procedures compared with those who did not receive palliative care.

Delivery of Palliative Care

Among all patients who received palliative care, 64.6% received their first palliative care consultation outside of a hospital, and 53.4% received palliative care services across multiple care settings. Across all 4 models of palliative care delivery, generalist physicians had the highest frequency of palliative care claims, at least 1 encounter with 61.0% of subjects. The delivery of palliative care was led by family physicians who were nonspecialists in palliative care 45.2% of the time. A generalist palliative care model was the most common type used and was delivered to 55.1% of people (Table 3, Table S2).

Table 3

Description of the Delivery of Palliative Care in Adults Dying With Heart Failure in Ontario Between 2010 and 2015

	Location of Death			All Patients Receiving Palliative Care n=35 292	P Valuea
	Home n=8109 (23.0%)	Hospital n=25 991 (73.6%)	Other n=1192 (3.4%)
Location of initial palliative care consultation, n (%)
Inpatient	615 (7.6)	3749 (14.4)	72 (6.0)	4436 (12.6)	<0.01
Outpatient	2316 (28.6)	11 160 (42.9)	434 (36.4)	13 910 (39.4)	…
Home‐based	3050 (37.6)	5405 (20.8)	434 (36.4)	8889 (25.2)	…
Subacute careb	10 to 16 (0.1 to 0.2)c	300 to 326 (1.2 to 1.3)c	4 to 10 (0.3 to 0.8)c	333 (0.9)	…
Third partyb	2111 (26.0)	5350 (20.6)	243 (20.4)	7704 (21.8)	…
Other	0 to 6 (0.0 to 0.1)c	0 to 6 (0.0 to 0.0)c	0 to 6 (0.0 to 0.5)c	0 to 6 (0.0 to 0.0)c	…
Locations of all Palliative Care Provided, n (%)
Inpatient	74 (0.9)	1532 (5.9)	8 (0.7)	1614 (4.6)	<0.01
Outpatient	621 (7.7)	6764 (26.0)	229 (19.2)	7614 (21.6)	…
Home‐based	1915 (23.6)	3179 (12.2)	356 (29.9)	5450 (15.4)	…
Multiple locations	5212 (64.3)	13 117 (50.5)	531 (44.5)	18 860 (53.4)	…
Third party	279 (3.4)	1197 (4.6)	61 (5.1)	1537 (4.4)	…
Other	8 (0.1)	202 (0.7)	7 (0.6)	227 (0.6)	…
Model of palliative cared, n (%)
Generalist only	4007 (49.4)	14 743 (56.7)	710 (59.6)	19 460 (55.1)	<0.01
Consultative	2486 (30.7)	5955 (22.9)	256 (21.5)	8697 (24.6)	…
Specialist only	1616 (19.9)	5293 (20.4)	226 (19.0)	7135 (20.2)	…
Number of palliative care visits, n (%)
1 to 4	3623 (44.7)	16 278 (62.6)	718 (60.2)	20 619 (58.4)	<0.01
5 to 9	1552 (19.1)	4043 (15.6)	171 (14.3)	5766 (16.3)	…
10 to 14	845 (10.4)	1695 (6.5)	67 (5.6)	2607 (7.4)	…
15+	2089 (25.8)	3975 (15.3)	236 (19.8)	6300 (17.9)	…
Timing of initial palliative care consultatione, n (%)
Optimal	1884 (23.2)	4287 (16.5)	296 (24.8)	6467 (18.3)	<0.01
Appropriate	3626 (44.7)	8732 (33.6)	502 (42.1)	12 860 (36.4)	…
Late	2599 (32.1)	12 972 (49.9)	394 (33.1)	15 965 (45.2)	…

A chi‐squared test was used to compare categorical variables describing models of palliative care between death at home vs death in hospital.

Subacute care includes both nursing homes and complex continuing‐care units. Third‐party care typically includes telephone support, weekly case management, and outpatient case conference. See Data S1 for details.

Data presented as ranges of values in accordance with ICES privacy policy to prevent a disclosure of a cell size containing 5 or fewer subjects.

The 4 models of palliative care were (1) no physician‐based palliative care, (2) generalist palliative care (eg, from a primary‐care physician or medical specialists such as internists and oncologists), (3) consultation palliative care (ie, care from both palliative care specialists and generalists), and (4) specialist palliative care.31

Timing is measured in terms of proximity to death (ie, the number of days from initial visit until death): optimal, between 2 years and 6 months before death; appropriate, between 6 months and 30 days before death; late, less than 30 days before death.

Among those who had at least 1 palliative care encounter, consultation was most frequently initiated late in a person's course of disease, occurring ≤30 days before death in 45.2% of people receiving palliative care (Table 3). Furthermore, a larger number of patients who died in a hospital received late palliative care compared with those who died at home (49.9% versus 32.1%, respectively). The main hospital admission diagnoses by receipt of palliative care are presented in Tables S3 and S4.

There was a significant and positive association between patients who received home‐based palliative care exclusively and death at home compared with those who received inpatient palliative care (Figure 2). Palliative care services that were delivered across multiple care settings were associated with increased odds of dying at home compared with those who received inpatient palliative care. Palliative care that was delivered as a consultative model involving a palliative care specialist compared with a generalist was also significantly associated with dying at home. Finally, receiving a higher intensity of palliative care, as reflected by having >4 visits, increased the odds of dying at home by more than 2‐fold.

Figure 2

Subgroup analysis of components of palliative care delivery and death at home. Association between delivery of different components of palliative care services and death at home (vs death in hospital) among adults dying with heart failure in Ontario between 2010 and 2015 who received palliative care. Models were adjusted for age, sex, rurality, neighborhood income, duration of heart failure, and the presence of metastatic cancer, dementia, diabetes mellitus, myocardial infarction, and stroke.

Discussion

Our cohort study of 74 986 community‐dwelling adults who died with HF found that almost three quarters (75.4%) died in a hospital. Overall, 35 292 (47.1%) people received palliative care in the last 2 years of life, and palliative care was associated with a 2‐fold increase in the odds of death at home compared with people who did not receive it. Among patients receiving palliative care, nonspecialist palliative care physicians led its delivery 61.0% of the time. A generalist model of care was most often used (55.1% of patients), and palliative care was provided to 53.4% of patients in multiple locations. These findings support our hypothesis that palliative care in patients with HF is associated with death at home, a recognized indicator of high‐quality end‐of‐life care.22, 23

Our findings have significant implications for the delivery of palliative care to people with noncancer illnesses such as HF. The proportion of adults dying with heart failure who receive palliative care is half of that for those dying with cancer (88%).38 Prior research showed that 20% to 45% of patients died in an acute‐care setting, a number that is substantially lower than the 75.4% of people in our study who died in a hospital with HF.28, 46 These findings identify an opportunity to improve end‐of‐life care for patients with HF because most people report a preference for death at home.21

Our study demonstrated that the current models of palliative care being delivered are associated with an increased likelihood of death at home when patients receive it. As populations age and care becomes increasingly complex due to rising rates of multimorbidity,47 continued investments in the training of both palliative generalists and specialists, along with expansion of home‐based palliative care, will be needed to help people die at home.28

Several prior studies reported similar findings to ours. In a recent systematic review measuring the effect of palliative care interventions on adults with HF, receipt of palliative care was significantly associated with death at home in 2 of the 4 studies that measured this outcome.18 Only 1 of the studies included in the systematic review was a randomized control trial, and it found a similar magnitude to our study in the likelihood of death at home (odds ratio 2.20 [95% CI, 1.3–3.7]). Of the remaining 3 lower‐quality studies, 1 did not examine home deaths,48 1 was underpowered to detect a difference,49 and the other study was a nonrandomized pilot study that demonstrated a trend toward increased likelihood of death at home but was limited in its conclusions by its study design.50 Our study provides further evidence to support the association between palliative care and death at home in general, and specifically for patients with heart failure.

The limitations of our study are the lack of information on patient and caregiver preferences for location of death, their willingness to engage with palliative care, and on their overall goals of care and advance care plans. Furthermore, some patients may not be able to die at home (despite their preferences) because their symptoms are too complex for home‐based management. We measured a “palliative approach” to care that includes both generalist and specialist palliative care physicians, which may underestimate the magnitude of our findings given that we found an associated increased likelihood of death at home with specialist palliative care in a subgroup analysis comparing the 2 different models. In other jurisdictions effective specialist palliative care may be delivered by other types of clinicians such as nurse practitioners,51 which has important implications for its scalability. Our results may also be underestimated due to misclassification bias if physicians are providing palliative care but are not using palliative care fee codes. We also captured palliative care using fee codes in administrative date, a strategy that has been less successful in health systems without universal coverage.52 Our study setting also lacks robust hospice networks like those found in many areas of the United States, which may limit the ability of patients with significant care needs to die outside of the hospital setting unless they are admitted to a nursing home.29 Our study excluded a large number of people admitted to nursing homes, which may partially account for the larger number of deaths in hospital when compared with other studies where 10% of people died in a nursing home.53 Compared with other developed countries, Canada has been found to deliver hospital‐centric end‐of‐life care, with over 50% dying in hospital compared with fewer than 30% in the United States.54 Therefore, its generalizability to other jurisdictions requires confirmation. We also intentionally studied the population of patients with chronic heart failure and found similar results when we included those with a more recent diagnosis of HF. However, we were unable to determine if they had preserved or reduced ejection fraction. Therefore, potential differences in our findings that are related to the classification of HF remain unknown. We suspect that the higher numbers of hospitalizations and days spent in the hospital in the last year of life among people receiving palliative care are related to the severity of their illness and increased care needs, which are not readily measured in administrative data. Late referral to palliative care may come as a consequence of hospitalization events that occur near death and trigger care teams to refer, although this is not wholly explained because nearly 65% of subjects in our study had their first consultation with palliative care outside the hospital setting.

Significant questions remain about which models of palliative care and the timing of its initiation will have important effects on patient‐ and policy‐oriented outcomes.55 Prior work reported a 45% relative reduction in the number of hospital admissions at 12 weeks of follow‐up using a transitional home‐based palliative care model compared with usual care.56 In a randomized trial comparing inpatient palliative care to usual care, there was no reduction in readmission to hospital at 30 days.57 Future work will need to evaluate in more detail the effects of timing and different models of palliative care on place of death and healthcare use and take into consideration the potential impact of cultural and racial differences between patients on preferences for these outcomes.58

Conclusions

Most people with HF die in a hospital and do not receive palliative care. Providing palliative care near the end of life was associated with an increased likelihood of dying at home. These findings suggest that scaling existing palliative care programs to increase access may improve end‐of‐life care in people dying with chronic noncancer illness.

Sources of Funding

This study was supported by ICES, which is funded by an annual grant from the Ontario MOHLTC (Ministry of Health and Long‐Term Care). The analysis was supported by a research grant Dr Tanuseputro and Dr Hsu received from the MOHLTC, as well as by a grant from the Canadian Institutes for Health Research (PJT‐153251), which funded some of this work and will fund the open access fee. The opinions, results and conclusions reported in this paper are those of the authors and are independent of the funding sources. No endorsement by ICES or the Ontario MOHLTC is intended or should be inferred. Parts of this material are based on data and information compiled and provided by the Canadian Institute for Health Information. However, the analyses, conclusions, opinions, and statements expressed herein are those of the authors, and not necessarily those of the Canadian Institute for Health Information. We thank IMS Brogan Inc. for use of their Drug Information Database. Dr Quinn and Dr Stall receive funding from the Canadian Institutes of Health Research Vanier Scholarship Program, the Eliot Phillipson Clinician‐Scientist Training Program, and the Clinician Investigator Program at the University of Toronto. Dr Lee is supported by a midcareer investigator award from the Heart and Stroke Foundation and is the Ted Rogers Chair in Heart Function Outcomes. Dr Kavalieratos receives research funding from the National Institutes of Health (K01HL133466), the Cystic Fibrosis Foundation, and the Milbank Foundation.

Disclosures

None.

Data S1. Association Between Palliative Care and Death at Home in Adults With Heart Failure

Table S1. Location of Death, Healthcare Utilization, and Invasive Cardiac Procedures in Patients Dying With Heart Failure in Ontario Between 2010 and 2015, by Receipt of Palliative Care

Table S2. Description of the Physician Mix in Delivery of Palliative Care to Patients Dying With Heart Failure in Ontario Between 2010 and 2015

Table S3. Admission Diagnoses by Study Group for First Hospitalization in the Last Year of Life

Table S4. Admission Diagnoses by Study Group for Hospitalizations in Which Patients Died

Click here for additional data file.