Skip main navigation

Challenges for Patients Dying of Heart Failure and Cancer

Originally publishedhttps://doi.org/10.1161/CIRCHEARTFAILURE.122.009922Circulation: Heart Failure. 2022;15

Abstract

Background:

Hospice and palliative care were originally implemented for patients dying of cancer, both of which continue to be underused in patients with heart failure (HF). The objective of this study was to understand the unique challenges faced by patients dying of HF compared with cancer.

Methods:

We assessed differences in demographics, health status, and financial burden between patients dying of HF and cancer from the Health and Retirement Study.

Results:

The analysis included 3203 individuals who died of cancer and 3555 individuals who died of HF between 1994 and 2014. Compared with patients dying of cancer, patients dying of HF were older (80 years versus 76 years), had poorer self-reported health, and had greater difficulty with all activities of daily living while receiving less informal help. Their death was far more likely to be considered unexpected (39% versus 70%) and they were much more likely to have died without warning or within 1 to 2 hours (20% versus 1%). They were more likely to die in a hospital or nursing home than at home or in hospice. Both groups faced similarly high total healthcare out-of-pockets costs ($9988 versus $9595, P=0.6) though patients dying of HF had less wealth ($29 895 versus $39 008), thereby experiencing greater financial burden.

Conclusions:

Compared with patients dying of cancer, those dying from HF are older, have greater difficulty with activities of daily living, are more likely to die suddenly, in a hospital or nursing home rather than home or hospice, and had worse financial burden.

What is New?

  • Patients with heart failure at the end of life experience worse disability, caregiver support, and financial burden, compared with patients with cancer. Their deaths were also much likely to be considered unexpected by their caregivers, and were more often likely to occur in a hospital or nursing home versus home or hospice.

What are the Clinical Implications?

  • Clinicians should be aware that their patients with heart failure may face greater financial burden, disability, and have lesser caregiver support at end of life. Many patients may not be aware of the seriousness of their heart failure condition, leading to a death from heart failure being unexpected. Improved prognostic communication and disease awareness is critical to patients receiving goal-concordant care at end of life.

Heart failure (HF), a chronic, progressive, and debilitating condition, is a leading cause of hospital admissions among older Americans.1 Palliative care (PC) is a specialty and a philosophy of care that prioritizes patients’ quality of life and care concordant with patients’ goals and values. PC was historically developed and implemented for patients dying of cancer, and patients with cancer are most likely to use services such as hospice at the end of life (EOL).2,3 While PC is appropriate for patients with serious illness regardless of prognosis, particularly patients with HF,4 PC continues to be underutilized in patients with HF even at end of life. Some factors that lead to the underuse of PC includes unpredictable disease trajectory, difficulty in prognostication, discordance between clinician-predicted and patient-predicted prognosis, lack of cardiologist training in primary palliative care, poor care coordination, and lack of programs specifically designed to meet the needs of patients with HF.2

Understanding differences between patients dying of HF and cancer is important, since these differences may point to distinct needs and challenges, which in turn may influence access to palliative and hospice care. For example, in a national registry, patients with cardiovascular disease were older than cancer patients at the time of initial referral to PC.5 Therefore, to better understand the specific challenges faced by patients dying of HF, which could inform the design of curriculums and programs tailored to provide high-quality EOL care to these patients, we assessed differences in demographics, health status, ability to perform activities of daily living (ADLs)/instrumental activities of daily living (IADLs), and financial burden between patients dying of HF and cancer from the nationally representative Health and Retirement Study (HRS).

Methods

The study population included individuals surveyed in the nationally representative HRS who died from cancer or HF between 1994 and 2014. The HRS is a longitudinal cohort study of a representative sample of US adults, over the age of 50, who were interviewed biennially since 1992. The survey was conducted by the University of Michigan with funding support from the National Institute on Aging. The data are publicly available and freely available for researchers to analyze. It contains >37 000 individuals with an average response rate of ≈87% across the study period. The survey collects data on demographics, income, assets, employment, retirement, health, disability, cognition, health insurance, health expenditure, and end of life. The study oversamples African American and Latino populations and the cohort includes community dwelling individuals. Due to its in-depth and unique interviews, the HRS provides a valuable multidisciplinary data source that can be used to study different aspects of aging. The data were publicly available and deidentified, and therefore institutional review was not sought per Health and Human Services regulation 45 CFR 46.101(c).

The study population was identified through exit interviews. This interview was conducted after a participant’s death with the respondent identified from a close social network of the deceased and contains information about the participant’s end-of-life circumstances and death. The majority of respondents in exit interviews (88.3%) were closely related to the deceased participant. The exit interview was completed for almost all survey participants who died, ranging from 74.6% in 1994 to 92.1% in 2014. It contains information on participants’ place of death, cause of death, healthcare utilization in the final year of life, health status, disability, out-of-pocket (OOP) expenses, insurance status, and social support.

Data from exit interviews were merged with core interviews to better characterize participants’ circumstances. Health conditions indicated if the participant had the disease before death. Difficulty with ADLs and IADLs and help received were analyzed. The ADLs include dressing, walking across a room, bathing, eating, getting in and out of bed, and using the toilet. The IADLs include preparing hot meals, shopping for groceries, using the telephone, and taking medications.

The HRS collected information on various out-of-pocket (OOP) costs: hospital, nursing home, hospice, special facility, home care, doctor visits, dentist, and prescription drug costs. Beginning in wave 10 (2010), an additional category captured any additional OOP medical expenditures that could not be assigned to any of the above-mentioned categories. The costs reflected OOP expenditures, excluding expenditures for prescription drugs, since the last interview (on average 2 years). Prescription OOP costs reflected expenses in the month prior to the interview. Total major medical expenses were the sum of OOP expenses for hospital stays, nursing home stays, hospice, doctor visits, prescription drugs, special facilities or services, in-home medical care, and other medical expenses not covered by insurance. Information regarding OOP expenses came from a harmonized HRS file, where information was imputed and harmonized by the data provider. If participants could not provide the exact amount, they were presented with a series of unfolding brackets to approximate the amount. This information was then used to derive imputed expenditure values. Using this method, the HRS minimizes the nonresponse rate. All costs were adjusted to 2010 dollars based on the consumer price index for the year of death.

Baseline sociodemographic and health status characteristics were compared between participants that died of cancer or HF. Pearson χ2 test was used to compare categorical variables and ANOVA for continuous variables. The outcome variables of interest included (1) having a disability, defined as difficulty with any ADL or IADL, and (2) dying in hospital. We used multivariable logistic regression analysis to assess factors associated with having a disability and dying in hospital for both cancer and HF patients. Two models were tested to assess predictors of the outcome variables in both groups. The first model included sex, age, racial background (White, Black/African American‚ and other), marital status, education level (lower than high school level, high school level, graduate level), body mass index category (<18.5 kg/m2, <25 kg/m2, <30 kg/m2‚ and ≥30 kg/m2), whether the patient smoked before death, census region (Northeast, Midwest, South, East), and used dummy time variable to control for the trends in time in the dataset for each wave or visit. The second model included variables from the first model and controlled for the presence of the following comorbid conditions: hypertension, lung disease, diabetes, and arthritis. We were unable to include ethnicity in the regression models as there was insufficient representation of non-White Hispanic individuals (<50 across both patient groups) leading to unbalanced representation across race and ethnic backgrounds. Memory problem was also not included in the regression models as it had a significantly higher proportion of missing values (17%) compared with other conditions, leading to a reduction in the statistical power of the models.

All analyses were performed using STATA version 15.0 (STATA Corp, TX).

Results

The analysis included 3203 individuals who died of cancer and 3555 individuals who died of HF between 1994 and 2014. Baseline characteristics of the 2 groups are shown in Tables 1 and 2. Compared with patients dying of cancer, patients dying of HF were older (mean age 80.0 years versus 75.7 years), were less likely to smoke (13.2% versus 19.7%), be married/partnered (40.2% versus 50.1%) and college educated (26.3% versus 32.6%) and were less wealthy ($29 895 versus $39 008). HF decedents were less likely to have private insurance (38.9% versus 44.1%) but more likely to have Medicare (89.5% versus 81.2%), Medicaid (17.7% versus 13.7%) and receive social security benefits (92.1% versus 87.5%; all P<0.05; Table 1, Figure). No difference was noted in sex, race, ethnicity, and body mass index. Patients dying of HF had a slightly greater sum of comorbid conditions (2.9 versus 2.8) and were more likely to have hypertension (68.8% versus 57.6%), diabetes (31.4% versus 22.0%)‚ and arthritis (64.4% versus 58.6%), with no significant differences in lung disease or memory problems (Table 2).

Table 1. Baseline Characteristics and Differences Between Patients Dying of Cancer or Heart Failure (HF)

Cancer (N=3203)HF (N=3555)P valueHF with history of cancer (N=670)P value
Age‚ y
 Mean (SD)75.7 (10.1)80.0 (10.6)<0.00181.7 (9.01)0.078
Sex
 Men1670 (52.1%)1774 (49.9%)0.066376 (56.1%)<0.001
 Women1533 (47.9%)1781 (50.1%)294 (43.9%)
Race
 White2537 (79.2%)2870 (80.7%)0.249577 (86.2%)0.001
 Black559 (17.5%)567 (15.6%)79 (11.8%)
 Other103 (3.2%)113 (3.2%)13 (1.9%)
 Missing0.1%5 (0.1%)1 (0.2%)
Ethnicity
 Non-Hispanic3008 (93.9%)3314 (93.2%)0.254635 (94.8%)0.075
 Hispanic191 (6.0%)236 (6.6%)34 (5.1%)
 Missing4 (0.1%)5 (0.1%)1 (0.2%)
Body mass index, kg/m2
 <18.5204 (6.4%)257 (7.2%)0.16554 (8.1%)0.796
 18–251374 (42.9%)1495 (42.1%)294 (43.9%)
 25–301049 (32.6%)1077 (30.3%)193 (28.8%)
 >30576 (18.0%)726 (20.4%)129 (19.3%)
Smoked before death
 Yes630 (19.7%)468 (13.2%)<0.00181 (12.1%)0.325
 No2555 (79.8%)3060 (86.1%)586 (87.5%)
 Missing18 (0.56%)27 (0.8%)3 (0.5%)
Education
 Below HS level1219 (38.1%)1590 (44.73%)<0.001286 (42.7%)0.187
 HS level940 (29.4%)1029 (29.0%)205 (30.6%)
 Some college1044 (32.6%)936 (26.3%)170 (26.7%)
 Missing0 (0.0%)0 (0.0%)0 (0.0%)
Married/partnered at death
 Yes1604 (50.1%)1429 (40.2%)<0.001365 (54.5%)<0.001
 No1324 (41.3%)1850 (52.0%)305 (45.5%)
 Missing275 (8.6%)276 (7.8%)0 (0.0%)
Private insurance
 Yes1412 (44.1%)1384 (38.9%)<0.001284 (42.4%)0.001
 No1501 (46.9%)1815 (51.1%)312 (46.6%)
 Missing290 (9.05%)356 (10.0%)74 (11.0%)
Medicare
 No570 (17.8%)343 (9.7%)<0.00141 (6.1%)<0.001
 Yes2602 (81.2%)3181 (89.5%)627 (93.6%)
 Missing31 (1.0%)31 (0.9%)2 (0.3%)
Medicaid
 No2706 (84.5%)2853 (80.3%)<0.001556 (83.0%)<0.001
 Yes440 (13.7%)629 (17.7%)100 (14.9%)
 Missing57 (1.8%)73 (2.1%)14 (2.1%)
Social Security benefits
 No399 (12.5%)281 (7.9%)<0.00165 (9.7%)<0.001
 Yes2804 (87.5%)3274 (92.1%)605 (90.3%)
 Missing0 (0.0%)0 (0.0%)0 (0.0%)
Wealth
 Mean (SD)$39 008 ($55 916)$29 895 ($40 798)<0.001$34 060 ($41 654)<0.001
Death expected
 Expected2254 (70.4%)1392 (39.2%)<0.001322 (48.1%)0.003
 Unexpected586 (18.3%)1783 (50.2%)326 (48.7%)
 Other80 (2.5%)96 (2.7%)20 (3.0%)
 Missing283 (8.8%)284 (8.0%)2 (0.3%)
Place of death
 Home1176 (36.7%)1071 (30.1%)<0.001212 (31.6%)0.237
 Hospital904 (28.2%)1278 (36.0%)251 (37.5%)
 Nursing home410 (12.8%)657 (18.5%)138 (20.6%)
 Hospice372 (11.6%)131 (3.7%)36 (5.4%)
 Other64 (2.0%)143 (4.0%)33 (4.9%)
 Missing277 (8.7%)275 (7.7%)0 (0.0%)

Univariable differences in baseline demographic data were assessed with the Pearson χ2 test for categorical variables, and ANOVA was used for continuous variables. Last column (P value) denotes difference between group that died of heart failure with and without prior cancer history. HS indicates high school.

Table 2. Health Status of Individuals Dying of Cancer and Heart Failure (HF)

Cancer (N=3203)HF (N=3555)P valueHF with history of cancer (N=670)P value
No. of conditions before death
 Mean (SD)2.75 (1.45)2.93 (1.39)<0.0013.94 (1.54)<0.001
 Missing433 (13.5%)437 (12.3%)0 (0.0%)
High blood pressure
 No1330 (41.5%)1071 (30.1%)<0.001185 (27.6%)0.132
 Yes1844 (57.6%)2445 (68.8%)465 (69.4%)
 Missing29 (0.9%)39 (1.1%)20 (3.0%)
Diabetes
 No2477 (77.3%)2417 (68.0%)<0.001461 (68.8%)0.611
 Yes705 (22.0%)1116 (31.4%)205 (30.6%)
 Missing21 (0.7%)22 (0.6%)4 (0.6%)
Lung disease
 No2585 (80.7%)2869 (80.7%)0.881498 (74.3%)0.001
 Yes603 (18.8%)663 (18.7%)158 (23.6%)
 Missing15 (0.5%)23 (0.7%)14 (2.1%)
Arthritis
 No1290 (40.3%)1241 (34.9%)<0.001220 (32.8%)0.052
 Yes1877 (58.6%)2289 (64.4%)437 (65.2%)
 Missing36 (1.1%)25 (0.7%)13 (1.9%)
Memory problem
 No1721 (53.7%)1854 (52.2%)0.233279 (41.6%)0.698
 Yes938 (29.3%)1114 (31.3%)209 (31.2%)
 Missing544 (17.0%)587 (16.5%)182 (27.2%)
Self-reported health before death
 Excellent142 (4.4%)128 (3.6%)<0.00124 (3.6%)0.008
 Very good458 (14.3%)386 (10.9%)57 (8.5%)
 Good758 (23.7%)837 (23.5%)132 (19.7%)
 Fair918 (28.7%)1131 (31.8%)226 (33.7%)
 Poor922 (28.8%)1068 (30.0%)230 (34.3%)
 Missing5 (0.2%)5 (0.1%)1 (0.2%)
Duration of final illness before death
 No warning, 1–2 h28 (0.9%)692 (19.5%)<0.001119 (17.8%)0.123
 <1 d63 (2.0%)340 (9.6%)63 (9.4%)
 <1 wk228 (7.1%)537 (15.1%)123 (18.4%)
 <1 mo438 (13.7%)540 (15.2%)120 (17.9%)
 <1 y1247 (38.9%)561 (15.8%)118 (17.6%)
 >1 y899 (28.1%)554 (15.6%)116 (17.3%)
 Missing300 (9.4%)331 (9.3%)11 (1.6%)

Univariable differences in baseline demographic data were assessed with the Pearson χ2 test. Last column (P value) denotes difference between group that died of heart failure with and without prior cancer history.

Figure.

Figure. Differences in patients dying of cancer versus heart failure. Clip art from Free Pik. ADLs indicates activities of daily living; and IADLs‚ instrumental activities of daily living.

Compared with patients dying of cancer, HF decedents were more likely to report poor health (30.0% versus 28.8%) and less likely to report excellent (3.6% versus 4.4%) or very good health (10.9% versus 14.3%; P<0.001) before dying. In the last year of life, patients dying of HF had greater difficulty with walking, using the toilet, getting in and out of bed, taking medications, eating, dressing, preparing meals, bathing, shopping, and using the phone (all P<0.001; Table 3). However, despite having greater deficits in both ADLs and IADLs, patients dying of HF were less likely to have someone help with ADLs (34.1% versus 46.7%) or IADLs (34.8% versus 58.6%) in the last year of life. However, they were more likely to report professional help with ADLs/IADLs in the last 3 months of life (24.2% versus 18.6%; all P<0.001; Table 3).

Table 3. Help Received With ADLs and IADLs for Individuals Dying of Cancer or Heart Failure (HF)

Difficulty in the previous year with:Cancer (N=3203)Heart failure (N=3555)P valueHF with history of cancer (N=670)P value
Walking across a room
 Yes553 (17.3%)1076 (30.3%)<0.001200 (29.9%)0.868
 No2622 (81.9%)2435 (68.5%)465 (69.4%)
 Missing28 (0.9%)44 (1.2%)5 (0.8%)
Toilet
 Yes390 (12.2%)673 (18.9%)<0.001131 (19.6%)0.619
 No2770 (86.5%)2832 (79.7%)530 (79.1%)
 Missing43 (1.3%)50 (1.4%)9 (1.3%)
Getting in and out of bed
 Yes405 (12.6%)748 (21.0%)<0.001146 (21.7%)0.724
 No2781 (86.8%)2780 (78.2%)520 (77.6%)
 Missing17 (0.5%)27 (0.8%)4 (0.6%)
Taking medications
 Yes258 (8.1%)602 (16.9%)<0.001114 (17.0%)0.884
 No2630 (82.1%)2605 (73.3%)500 (74.6%)
 Missing315 (9.8%)348 (9.8%)56 (8.4%)
Eating
 Yes311 (9.7%)530 (14.9%)<0.001102 (15.2%)0.963
 No2874 (89.7%)3000 (84.4%)565 (84.3%)
 Missing18 (0.6%)25 (0.7%)3 (0.5%)
Dressing
 Yes613 (19.2%)1095 (30.8%)<0.001193 (28.8%)0.252
 No2571 (80.3%)2434 (68.5%)474 (70.8%)
 Missing19 (0.6%)26 (0.7%)3 (0.5%)
Preparing hot meals
 Yes570 (17.8%)1075 (30.2%)<0.001194 (29.0%)0.383
 No2587 (80.8%)2439 (68.6%)469 (70.0%)
 Missing46 (1.4%)41 (1.2%)7 (1.0%)
Bathing
 Yes577 (18.0%)1112 (31.3%)<0.001207 (30.9%)0.791
 No2609 (81.5%)2418 (68.0%)460 (68.7%)
 Missing17 (0.5%)25 (0.7%)3 (0.5%)
Shopping for groceries
 Yes758 (23.7%)1331 (37.4%)<0.001243 (36.3%)0.339
 No2396 (74.8%)2183 (61.4%)420 (62.7%)
 Missing49 (1.5%)41 (1.2%)7 (1.0%)
Using telephone
 Yes333 (10.4%)733 (20.6%)<0.001136 (20.3%)0.698
 No2621 (81.8%)2509 (70.6%)480 (71.6%)
 Missing249 (7.8%)313 (8.8%)54 (8.1%)
Someone helped with ADLs in the last year
 Yes1495 (46.7%)1211 (34.1%)<0.001312 (46.6%)0.002
 No43 (1.3%)145 (4.1%)212 (31.6%)
 Missing1665 (52.0%)2199 (61.9%)146 (21.8%)
Someone helped with IADLs in the last year
 Yes1877 (58.6%)1238 (34.8%)<0.001344 (51.3%)0.008
 No144 (4.5%)440 (12.4%)267 (39.9%)
 Missing1182 (36.9%)1877 (52.8%)59 (8.8%)
Professional helped with IADLs/ADLs in the final 3 mo
 Yes597 (18.6%)860 (24.2%)<0.001189 (28.2%)0.826
 No1986 (62.0%)1513 (42.6%)326 (48.7%)
 Missing620 (19.4%)1182 (33.3%)155 (23.1%)

Last column (P value) denotes difference between group that died of heart failure with and without prior cancer history. ADLs indicates activities of daily living; and IADLs‚ instrumental activities of daily living.

Compared with patients dying of cancer, HF decadents were more likely to die in the hospital (36.0% versus 28.2%) or nursing home (18.5% versus 12.8%) and less likely to die at home (30.1% versus 36.7%) or in hospice (3.7% versus 11.6%; P<0.001). Patients dying of HF were much more likely to have died with no warning or within 1 to 2 hours of warning (19.5% versus 0.9%), or to have a duration of final illness of <1 day (9.6% versus 2.0%), <1 week (15.1% versus 7.1%) or 1 month (15.2% versus 13.7%) and were much less likely to have an illness duration longer than 1 year (15.6% versus 28.1%; P<0.001). Death was expected by the family among 39.2% of patients who died of HF as opposed to 70.4% of those dying of cancer (P<0.001).

Total healthcare OOP costs were similar between patients dying of cancer and HF ($9988.40 versus $9594.50, P=0.6) though there were some differences in the sources of OOP costs (Table 4). Patients dying of HF had greater OOP costs from nursing homes ($4347.80 versus $1492.60) while they had lower OOP costs from hospice ($34.60 versus $117.70), doctor visits ($799.10 versus $1213.90), and outpatient surgery ($26.80 versus $99.80; all P<0.05).

Table 4. Healthcare Out-of-Pocket (OOP) Costs of Individuals Dying of Cancer and Heart Failure (HF)

Cost categoryCancer (N=3203)Heart failure (N=3555)HF with history of cancer (N=670)
N (%)Mean (SD)N (%)Mean (SD)P valueN (%)Mean (SD)P value
Hospital OOP2116 (66%)$3151.60 ($26 303.70)2217 (62%)$1875.30 ($21 717.20)0.081492 (73%)$985.00 ($3345.30)0.273
Nursing Home OOP2116 (66%)$1492.60 ($9509.20)2217 (62%)$4347.80 ($18 913.70)<0.001492 (73%)$4001.00 ($17 457.90)0.363
Hospice OOP2928 (91%)$117.70 ($1082.60)3281 (92%)$34.60 ($608.50)<0.001670 (100%)$48.87 ($730.50)0.393
Doctor visits OOP2928 (91%)$1213.90 ($8111.20)3281 (92%)$799.10 ($7476.40)0.037670 (100%)$656.40 ($2362.80)0.220
Drugs OOP2928 (91%)$2893.90 ($7780.80)3281 (92%)$2998.60 ($7795.50)0.597670 (100%)$3240.60 ($7404.00)0.292
Special facility and home care OOP812 (25%)$228.70 ($1535.90)1064 (30%)$545.60 ($6375.20)0.166178 (27%)$1290.70 ($12 661.30)0.789
Special facility OOP2116 (66%)$425.20 ($3184.30)2217 (62%)$365.00 ($1840.80)0.443492 (73%)$526.80 ($2323.80)0.110
Home care OOP2116 (66%)$530.80 ($7598.70)2217 (62%)$304.30 ($3135.50)0.196492 (73%)$570.00 ($4733.20)0.071
Other medical expenses OOP2928 (91%)$428.60 ($2806.80)3281 (92%)$329.80 ($3081.60)0.189670 (100%)$373.80 ($2333.70)0.064
Outpatient surgery OOP898 (28%)$99.80 ($648.00)812 (23%)$26.80 ($297.70)0.003209 (31%)$14.50 ($85.50)0.317
Dental OOP898 (28%)$333.20 ($1145.50)812 (23%)$283.30 ($1077.70)0.355209 (31%)$455.30 ($1632.20)0.661
Total major medical expenses OOP2928 (91%)$9594.50 ($30 280.90)3281 (92%)$9988.40 ($30 852.60)0.612670 (100%)$9804.60 ($20 238.40)0.564
Help from others OOP2233 (70%)$212.70 ($1426.80)2626 (74%)$385.10 ($3851.30)0.050567 (85%)$252.60 ($1840.10)0.579

From waves 3 to 5, respondents were asked about OOP spending in 4 categories: (1) hospital and nursing home costs; (2) doctor, dentist, and outpatient surgery costs; (3) average monthly prescription drug costs; and (4) home health care and special facilities or services costs. Beginning in wave 6, the number of categories expands to 8: (1) hospital costs; (2) nursing home costs; (3) doctor visits costs; (4) dentist costs; (5) outpatient surgery costs; (6) average monthly prescription drug costs; (7) home healthcare; and (8) special facilities costs. Beginning in wave 10, a ninth category seeks to capture any additional out-of-pocket medical expenditures that cannot be assigned to any of the other categories. Information regarding OOPs comes from harmonized Health and Retirement Study files, so information was imputed and harmonized by the data provider. Total major medical expense was the sum of reported or imputed out-of-pocket expenses for hospital stays, nursing home stays, hospice, doctor visits, drug expenses, special facilities or services, in-home medical care, and other medical expenses not covered by insurance. Last column (P value) denotes difference between group that died of heart failure with and without prior cancer history.

Multivariable regression analysis revealed that being older than 70, unmarried, underweight, not having graduated high school, and having comorbidities such as lung disease, diabetes, and arthritis were independently associated with having a disability in both patients with cancer or HF (Table 5). Results were consistent across both models, with the effect being more pronounced in HF patients. Especially the impact of other comorbidities, such as lung disease, diabetes, or arthritis, was highly associated with the probability of having a disability in HF patients.

Table 5. Multivariable Predictors of Having a Disability or Dying in the Hospital for Patients Dying of Cancer or Heart Failure

Independent variableProbability of having disabilityProbability of dying in hospital
Cancer patientsHeart failure patientsCancer patientsHeart failure patients
N=2617N=2691N=2919N=3001N=2051N=2897N=3149N=3235
RRRP valueRRRP valueRRRP valueRRRP valueRRRP valueRRRP valueRRRP valueRRRP value
Women1.170.0771.170.0661.54<0.0011.60<0.0010.870.1480.870.1121.190.0361.200.022
Age >70 y1.73<0.0011.65<0.0012.03<0.0011.96<0.0010.67<0.0010.69<0.0011.230.0621.190.099
Race (Ref=White)
 Black/African American1.150.2341.040.7180.930.5540.920.4872.07<0.0012.12<0.0011.070.5441.120.259
 Other1.400.1671.300.2630.830.4290.790.3081.550.0631.560.0531.190.4251.160.488
Married
 Yes0.830.0330.820.0240.63<0.0010.66<0.0011.240.0221.150.1131.300.0021.31<0.001
Education (Ref=below high school)
 High school0.58<0.0010.56<0.0010.63<0.0010.62<0.0011.020.8610.990.8911.160.0991.130.179
 College0.52<0.0010.51<0.0010.69<0.0010.68<0.0011.130.2661.100.3721.080.4181.030.750
BMI (Ref=BMI <18.5), kg/m2
 <250.41<0.0010.36<0.0010.51<0.0010.540.0011.260.2361.280.1831.050.7291.100.512
 <300.34<0.0010.31<0.0010.37<0.0010.42<0.0011.470.0511.560.0191.180.3031.280.105
 >300.41<0.0010.39<0.0010.530.0020.630.0181.680.0141.830.0031.150.4111.340.070
Smoking before death0.900.3340.900.3350.770.0330.730.0101.200.0951.270.0271.120.3451.080.476
Hypertension0.940.5201.050.5931.190.0601.240.010
Lung disease1.390.0021.53<0.0011.040.7061.210.048
Diabetes1.240.0421.40<0.0011.010.9121.300.002
Arthritis1.46<0.0011.47<0.0010.990.8861.080.339
Region (Ref=Northeast)
 Midwest0.880.3230.840.1651.040.7461.060.6300.56<0.0010.57<0.0010.66<0.0010.680.001
 South0.900.3850.870.2371.030.8151.050.6560.690.0020.710.0040.940.5880.950.628
 West0.750.0520.720.0190.970.8460.940.6850.610.0010.620.0010.670.0030.660.002

The presented results are from multinomial logistic regression analysis. Results are presented as relative risk ratios (RRR), indicating percentage relative risk change for a unit increase in the observed variable compared with the referent group, holding other variables constant. For categorical variables, reference category is stated in the row label; otherwise the reference is the complementary category. BMI indicates body mass index; and Ref, reference.

Among patients with cancer, those who were younger than 70, Black/African American, overweight, who smoked before death, who died more recently, and who lived in the Northeast were all more likely to die in the hospital. This was consistent even after controlling for other comorbidities. On the other hand, patients with HF who were women, older than 70, married, died more recently, and who lived in the Northeast were all more likely to die in the hospital. Among HF patients, comorbidities such as hypertension, lung disease, and diabetes increased the likelihood of dying in hospital. Racial background was not a significant predictor of in-hospital death among HF patients in either model.

Among patients dying of HF, there were 670 (18.8%) who also had a history of cancer. Compared with those dying of HF, these patients with a history of cancer were of similar age, but were more likely to be men, White, married, have private insurance or Medicare, had greater wealth, and were more likely to have a death that was expected (Table 1). They were less likely to be on Medicaid, and receive social security benefits. These patients dying of HF with a history of cancer were also more likely to have more comorbidities and report poor health (Table 2). While there were no differences in ADLs or IADLs, patients dying of HF with a history of cancer were more likely to have someone help with ADLs and IADLs in their last year of life (Table 3). No differences in OOP costs were noted (Table 4).

Discussion

This analysis provides novel insights into the different challenges that patients dying of HF and cancer face in the last year of life. Compared with patients dying of cancer, patients dying of HF were older, poorer, had worse health status, had more comorbidities, and had greater difficulty with ADLs/IADLs. However, they reported receiving less help in the last year of life. Their death was more likely to be considered unexpected and they were much more likely to die in a hospital or nursing facility versus at home or in hospice. Although both groups of patients faced significant financial burdens in the last year of life, HF patients had less wealth and were more likely to require government assistance. These findings have important implications for research and care delivery at the EOL for patients with HF.

These data build on previous studies that demonstrate the differences between patients dying of cancer and heart failure. Our study confirms previous findings that patients with HF are more likely to die in a medical facility such as a hospital or nursing home rather than at home or in hospice,6,7 are more likely to have functional limitations and are less likely to use hospice.8,9 This complexity of disease and function has direct implications for prognostication and care coordination. In the last year of life, patients with HF are more likely to be admitted to the hospital for noncardiovascular reasons than for HF exacerbation, with only 12% of hospitalizations being primarily related to HF.10 Therefore, ideally, PC interventions for HF patients should be able to account for the multimorbid nature of the disease as well as higher prevalence of geriatric impairments such as frailty within this population. Partnering with geriatrics might be crucial for the development and implementation of PC programs for HF patients.

There is a body of literature to support the fact that in patients with cancer access to early palliative care can improve outcomes including earlier hospice referral which can in turn improve caregiver experience at and after EOL, decrease inpatient costs, and provide care where the patient/family prefer.11,12 A major reason for the underutilization of PC in HF patients might be difficulties in prognostication in these patients. In cancer populations, accurate prognostic awareness correlates with less aggressive medical care at the end of life and improved caregiver outcomes.13 The significant findings in this current analysis show that most family members of patients who died of HF were not expecting death, indicating a need for improved communication about disease expectations and highlighting clinicians’ own uncertainty about prognostication in HF since the rates of actual sudden cardiac death have plummeted in recent years.14

While cancer patients are often seen in specialized centers and practices focused on 1 primary diagnosis, patients with HF have multiple care providers and locations, and therefore often experience fractured care. This can make prognostication in HF more challenging. Although both generalists and cardiologists tend to be inaccurate in their prognostic predictions, data are available to assist with HF prognosis.15 The number of HF hospitalizations is a strong predictor of mortality in community HF patients, with 1 study showing median survival after the first, second, third, and fourth hospitalization of 2.4, 1.4, 1.0, and 0.6 years, respectively.16 A Danish study of HF patients older than 70 years showed an absolute 1 year mortality of 18% associated with outpatient diuretic intensification and a 22.6% mortality for patients with a HF hospitalization.17 One simple approach that can be used in the clinical setting is the surprise question. In response to clinicians being asked if they would be surprised if a HF patient were to pass away within the next year, the response was significantly associated with all-cause mortality and had a sensitivity of 0.85 and specificity of 0.59 with a positive predictive value of 52% and a negative predictive value of 88%.18

Lack of prognostic communication throughout the disease process can lead to a disconnect between clinicians and patients and is also a significant contributor to the underutilization of PC in patients with HF. In a study from 2017 of cardiologists and patients with HF, cardiologists identified 69% of patients as high risk for left ventricular assist device, transplant or death in the next year, while only 14% of patients considered themselves at risk of these outcomes at 1 year. The actual rate of high-risk events in this population was 38% at 13 months, suggesting that patients greatly underestimate the risk of adverse outcomes from HF.19

Another novel aspect of this analysis is the presence of OOP cost data. Although cancer patients have traditionally been thought of as being particularly burdened by rising healthcare costs, this analysis shows that patients dying of HF have largely equally high burden of OOP costs. This finding is particularly notable given that patients dying of HF had much lower wealth than patients dying of cancer. Considering that financial burden is largely defined as the ratio between costs and income or wealth,20 this suggests that HF patients might be experiencing greater financial burden at the end of life. Given that previous work has shown that patients with atherosclerotic cardiovascular disease have greater financial burden than those with cancer,21 these findings for HF patients at end of life build on that observation.

The presence of wealth data is also novel, since wealth may be a more reliable marker to assess disparities among racial/ethnic groups than income.22 This is in part because even when their incomes are similar, Black and Hispanic people have much lesser wealth than White people. Financial burden among patients with HF remains greatly understudied and will be an area that will benefit from greater investigation.

Our study has several limitations. The most recent cohort of the HRS died in 2014 and therefore there is a possibility that some of these findings may have changed in the interim. However, analyses from more recent registries point to similar trends4 Furthermore, HRS offers the deepest phenotyping of the burden faced by patients at EOL and a unique aspect of the database is that it includes bereaved caregiver interviews, which offers considerable novel insight. HRS also lacked more in-depth data about comorbidities and about other aspects of end of life experience such as implantable cardioverter defibrillator shocks.

In conclusion, patients with HF face many of the same challenges faced by patients with cancer at end of life. However, patients dying of HF face many unique challenges, including a higher prevalence of disability and multimorbidity and possibly greater financial burden. Despite these challenges, HF patients are less likely to receive informal caregiver support. Furthermore, while 70% of caregivers of cancer patients felt the death was expected, only 39% of caregivers of HF patients felt the same way, suggesting both suboptimal communication as well as greater difficulty in predicting outcomes in HF patients. These findings should inform the design of PC interventions and hospice programs designed specifically to meet the challenges that HF patients face.

Article Information

Disclosures Dr Warraich is a scientific advisor for Embrace Prevention Care. Dr Nohria is a consultant for Takeda Oncology, AstraZeneca‚ and Bantam Pharmaceuticals. Dr Orkaby is supported by Veterans Affairs Clinical Science Research and Development Career Development Award 2 award IK2-CX001800 and National Institute on Aging Grant for Early Medical/Surgical Specialists' Transition to Aging Research award R03-AG060169. The remaining authors have no relevant disclosures.

Footnotes

For Sources of Funding and Disclosures, see page 1078.

Correspondence to: Haider J. Warraich, MD, 4B-132, 1400 VFW Pkwy, VA Boston Healthcare System, Boston, MA 02132. Email

References

  • 1. Agarwal MA, Fonarow GC, Ziaeian B. National trends in heart failure hospitalizations and readmissions from 2010 to 2017. JAMA Cardiol. 2021; 6:952–956. doi: 10.1001/jamacardio.2020.7472CrossrefMedlineGoogle Scholar
  • 2. Warraich HJ, Meier DE. Serious-illness care 2.0 – meeting the needs of patients with heart failure. N Engl J Med. 2019; 380:2492–2494. doi: 10.1056/NEJMp1900584CrossrefMedlineGoogle Scholar
  • 3. Cross SH, Kaufman BG, Quest TE, Warraich HJ. National trends in hospice facility deaths in the united states, 2003–2017. J Pain Symptom Manage. 2021; 61:350–357. doi: 10.1016/j.jpainsymman.2020.08.026CrossrefMedlineGoogle Scholar
  • 4. Maqsood MH, Khan MS, Warraich HJ. Association of palliative care intervention with health care use, symptom burden and advance care planning in adults with heart failure and other noncancer chronic illness. J Pain Symptom Manage. 2021; 62:828–835. doi: 10.1016/j.jpainsymman.2021.02.017CrossrefMedlineGoogle Scholar
  • 5. Warraich HJ, Wolf SP, Mentz RJ, Rogers JG, Samsa G, Kamal AH. Characteristics and trends among patients with cardiovascular disease referred to palliative care. JAMA Netw Open. 2019; 2:e192375. doi: 10.1001/jamanetworkopen.2019.2375CrossrefMedlineGoogle Scholar
  • 6. Chuzi S, Molsberry R, Ogunseitan A, Warraich HJ, Wilcox JE, Grady KL, Yancy CW, Khan SS. Trends in place of death for cardiovascular mortality related to heart failure in the United States from 2003 to 2017. Circ Heart Fail. 2020; 13:e006587. doi: 10.1161/CIRCHEARTFAILURE.119.006587LinkGoogle Scholar
  • 7. Cross SH, Warraich HJ. Changes in the place of death in the United States. N Engl J Med. 2019; 381:2369–2370. doi: 10.1056/NEJMc1911892CrossrefMedlineGoogle Scholar
  • 8. Lastrucci V, D’Arienzo S, Collini F, Lorini C, Zuppiroli A, Forni S, Bonaccorsi G, Gemmi F, Vannucci A. Diagnosis-related differences in the quality of end-of-life care: a comparison between cancer and non-cancer patients. PLoS One. 2018; 13:e0204458. doi: 10.1371/journal.pone.020445CrossrefMedlineGoogle Scholar
  • 9. Warraich HJ, Wolf SP, Troy J, Swetz KM, Goldstein NE, Mentz RJ, Jain N, Desai AS, Kamal AH. Differences between patients with cardiovascular disease and cancer referred for palliative care. Am Heart J. 2021; 233:5–9. doi: 10.1016/j.ahj.2020.11.016CrossrefMedlineGoogle Scholar
  • 10. Madelaire C, Gustafsson F, Kristensen SL, D’Souza M, Stevenson LW, Kober L, Torp-Pedersen C, Gislason G, Schou M. Burden and causes of hospital admissions in heart failure during the last year of life. JACC Heart Fail. 2019; 7:561–570. doi: 10.1016/j.jchf.2019.03.018CrossrefMedlineGoogle Scholar
  • 11. Wright AA, Keating NL, Balboni TA, Matulonis UA, Block SD, Prigerson HG. Place of death: correlations with quality of life of patients with cancer and predictors of bereaved caregivers’ mental health. J Clin Oncol. 2010; 28:4457–4464.doi: 10.1200/JCO.2009.26.3863CrossrefMedlineGoogle Scholar
  • 12. Temel JS, Greer JA, Muzikansky A, Gallagher ER, Admane S, Jackson VA, Dahlin CM, Blinderman CD, Jacobsen J, Pirl WF, et al. Early palliative care for patients with metastatic non-small-cell lung cancer. N Engl J Med. 2010; 363:733–742. doi: 10.1056/NEJMoa1000678CrossrefMedlineGoogle Scholar
  • 13. Wright AA, Zhang B, Ray A, Mack JW, Trice E, Balboni T, Mitchell SL, Jackson VA, Block SD, Maciejewski PK, et al. Associations between end-of-life discussions, patient mental health, medical care near death, and caregiver bereavement adjustment.JAMA. 2008; 300:1665–1673. doi: 10.1001/jama.300.14.1665CrossrefMedlineGoogle Scholar
  • 14. Shen L, Jhund PS, Petrie MC, Claggett BL, Barlera S, Cleland JG, Dargie HJ, Granger CB, Kjekshus J, Køber L, et al. Declining risk of sudden death in heart failure. N Engl J Med. 2017; 377:41–51. doi: 10.1056/NEJMoa1609758CrossrefMedlineGoogle Scholar
  • 15. Warraich HJ, Allen LA, Mukamal KJ, Ship A, Kociol RD. Accuracy of physician prognosis in heart failure and lung cancer: comparison between physician estimates and model predicted survival. Palliat Med. 2016; 30:684–689. doi: 10.1177/0269216315626048CrossrefMedlineGoogle Scholar
  • 16. Setoguchi S, Stevenson LW, Schneeweiss S. Repeated hospitalizations predict mortality in the community population with heart failure. Am Heart J. 2007; 154:260–266. doi: 10.1016/j.ahj.2007.01.041CrossrefMedlineGoogle Scholar
  • 17. Madelaire C, Gustafsson F, Stevenson LW, Kristensen SL, Køber L, Andersen J, D'Souza M, Biering-Sørensen T, Andersson C, Torp-Pedersen C, et al. One-year mortality after intensification of outpatient diuretic therapy. J Am Heart Assoc. 2020; 9:e016010. doi: 10.1161/JAHA.119.016010LinkGoogle Scholar
  • 18. Straw S, Byrom R, Gierula J, Paton MF, Koshy A, Cubbon R, Drozd M, Kearney M, Witte KK. Predicting one-year mortality in heart failure using the “Surprise Question”: a prospective pilot study. Eur J Heart Fail. 2019; 21:227–234. doi: 10.1002/ejhf.1353CrossrefMedlineGoogle Scholar
  • 19. Ambardekar AV, Thibodeau JT, DeVore AD, Kittleson MM, Forde-McLean RC, Palardy M, Mountis MM, Cadaret L, Teuteberg JJ, Pamboukian SV, et al. Discordant perceptions of prognosis and treatment options between physicians and patients with advanced heart failure. JACC Heart Fail. 2017; 5:663–671. doi: 10.1016/j.jchf.2017.04.009CrossrefMedlineGoogle Scholar
  • 20. Slavin SD, Khera R, Zafar SY, Nasir K, Warraich HJ. Financial burden, distress, and toxicity in cardiovascular disease. Am Heart J. 2021; 238:75–84. doi: 10.1016/j.ahj.2021.04.011CrossrefMedlineGoogle Scholar
  • 21. Valero-Elizondo J, Chouairi F, Khera R, Kittleson MM, Forde-McLean RC, Palardy M, Mountis MM, Cadaret L, Teuteberg JJ, Pamboukian SV, et al. Atherosclerotic cardiovascular disease, cancer, and financial toxicity among adults in the United States. JACC CardioOncol. 2021; 3:236–246. doi: 10.1016/j.jaccao.2021.02.006CrossrefMedlineGoogle Scholar
  • 22. Pollack CE, Cubbin C, Sania A, Hayward M, Vallone D, Flaherty B, Braveman PA. Do wealth disparities contribute to health disparities within racial/ethnic groups?J Epidemiol Community Health. 2013; 67:439–445. doi: 10.1136/jech-2012-200999CrossrefMedlineGoogle Scholar