Does Diagnosis Make a Difference? Comparing Hospice Care Satisfaction in Matched Cohorts of Heart Failure and Cancer Caregivers

Abstract

Background:

Half of heart failure patients will die within five years of diagnosis, making them an ideal population for hospice to reach. Yet hospice originated in oncology, and problems have been noted with the enrollment of heart failure patients. Whether caregiver satisfaction, a key quality measure in hospice, differs between heart failure and cancer caregivers is unknown.

Objective:

We aimed to determine whether diagnosis makes a difference in satisfaction with hospice care in matched cohorts of heart failure caregivers and cancer caregivers.

Methods:

This was a national cohort study, using caregiver responses to the Family Evaluation of Hospice Care (FEHC) survey. Heart failure and cancer caregivers were matched via propensity scoring. The relationship between diagnosis and caregiver satisfaction was examined across the domains of symptom management, emotional support, caregiver teaching, coordination of care, and global satisfaction, both before and after matching via logistic regression.

Results:

One-to-one matching with calipers yielded 7730 matched pairs out of an original sample of 8175 heart failure caregivers and 24,972 cancer caregivers. Significant differences were found in caregiver teaching, emotional support, coordination of care, and global satisfaction prior to matching, but the effect sizes were small. All differences disappeared after matching. High rates of dissatisfaction with caregiver teaching (42%) and emotional support (30%) were found in both cohorts.

Conclusions:

The diagnosis of heart failure, in and of itself, does not appear to make a difference in informal caregiver satisfaction with hospice care. Hospice provides high-quality care for patients, but improvements are needed in caring for the caregiver.

Introduction

Half of heart failure patients will die within five years of diagnosis, leaving behind family caregivers who will experience long-term effects from end-of-life caregiving experiences.^1–4 Once synonymous with oncology, hospice is increasingly being accessed by patients with chronic conditions like heart failure.⁵ Currently, cancer accounts for just 36.5% of hospice admissions and this percentage has been relatively flat over the last few years.⁶ In comparison, the number of noncancer diagnoses is increasing, accounting for 63.5% of hospice admissions in 2013, with cardiovascular admissions (primarily heart failure) the second largest group in this category after dementia.⁶

The majority of studies on hospice outcomes have been done in the cancer population, in which hospice care has been shown to yield high family satisfaction with care.⁷ Caregivers of patients who received hospice care have been shown to have lower risk for premature death and major depressive disorder than nonhospice caregivers.² It has been assumed that these positive outcomes found in the terminal cancer population will occur in all hospice populations. However, differences in quality outcomes between hospice patients with heart failure and those with terminal cancer have been reported, undermining this assumption.^8–10

While there are clear clinical differences between heart failure and cancer, there are also demographic differences between the end-stage heart failure and cancer populations. Heart failure patients who enroll in hospice are generally older, more functionally dependent, and more likely to be female than their counterparts with cancer.¹⁰ Hospice outcomes also differ between the two diagnosis populations: patients with end-stage heart failure are less likely to enroll in hospice; when enrolled, they are more likely to have very short (<3 day) lengths of stay, be hospitalized or use emergency services, and be discharged alive than cancer hospice patients.^10–12 All of these demographic and clinical differences have been previously linked to caregiver satisfaction with hospice care.^13–15 In addition, these characteristics have been demonstrated to contribute to caregiver satisfaction when examined separately in the heart failure and cancer populations.¹⁶ Currently unknown is whether these differences impact heart failure caregiver satisfaction with hospice care when compared to cancer caregivers. This begs the question, How large a role does the type of diagnosis play in determining caregiver satisfaction? The patient with a heart failure diagnosis presents with different symptom etiologies and a higher number of comorbidities compared with a patient with a cancer diagnosis.^10,17 Are hospices equally capable of caring for these two diagnoses? Are any differences in caregiver satisfaction related to these population characteristics rather than to the diagnosis?

The purpose of this study was to determine whether differences in caregiver satisfaction with hospice care exist between heart failure caregivers and cancer caregivers and if differences were due to diagnosis or clinical and demographic characteristics. We hypothesized that heart failure caregivers would be less satisfied with hospice care than cancer caregivers based on differences in live discharge and emergency service use found in prior studies.^9,10

Methods

Sample

This study was a retrospective analysis of the National Hospice and Palliative Care Organization's (NHPCO) national dataset, using the Family Evaluation of Care Survey (FEHC). Adult (21+) caregivers of adult hospice patients with a primary diagnosis of heart failure or cancer who provided complete responses to the FEHC were included in this study. Due to concerns around lack of randomness in missing answers, we chose to exclude questionnaires with incomplete demographic data on caregivers or patients.¹⁸

Data from the FEHC and additional hospice organizational data from the NHPCO were used. These data included basic demographic variables (age, gender, race, ethnicity, and education) for both caregivers and patients; primary diagnosis; and clinical variables (symptoms, place of care, and length of stay). Organizational data provided by the NHPCO included organizational size (average daily census) and whether or not the organization was freestanding.

Measurement

The FEHC is a 61-item questionnaire mailed one to three months after the patient's death that asks family members of hospice decedents to assess the end-of-life care provided.¹⁹ The FEHC has four domains, which examine caregiver satisfaction with (1) symptom management, (2) emotional support provided, (3) teaching provided by the hospice, and (4) coordination of care. Each question on the FEHC has multiple-answer choices, one of which is designated by NHPCO to be the “desirable” answer; all others are considered “negative” answers. For each domain, a domain score (the percentage of negative responses) is calculated. Higher numbers indicate a lower quality outcome of care, and a domain score of greater than 0.20 is considered an opportunity to improve care.^19,20 In addition to the four domains, global satisfaction with care is measured via one five-point Likert scale question, with responses ranging from excellent to poor. FEHC results have historically been highly positive, with few problems reported. As this affects the normality of the data, we chose to dichotomize the outcomes. As 0.2 is considered the cut-off for determining a domain score to be problematic (indicating need for improvement), we dichotomized domain scores into satisfied (<0.2) or dissatisfied (≥0.2). As a rating of less than excellent on the scaled question for global satisfaction is considered a defect, we also dichotomized responses into excellent (satisfied) or less than excellent (dissatisfied).

Analysis

First, descriptive analyses on demographic, clinical, and organizational characteristics for heart failure and cancer cohorts were conducted. A correlational analysis was performed to check for multicollinearity between the characteristics and between the outcomes. Logistic regression models were used to examine bivariate associations with diagnosis and how characteristics predicted cohort assignment. These data (patient, caregiver, and hospice characteristics) were used to generate propensity scores, which reflect the probability that a patient/caregiver dyad would be in the heart failure cohort versus the cancer cohort.²¹ These characteristics have been noted to impact caregiver burden and satisfaction, and our goal was to examine the impact of diagnosis apart from population characteristics.^15,22,23 Propensity scores have primarily been used to balance baseline covariate differences in observational trials seeking to establish treatment effect.²⁴ But they also have been used to estimate the true effect of one characteristic on outcomes. For example, in estimating the impact of patient gender on coronary bypass graft outcomes, Parolari and team used propensity score matching to balance out other demographic and clinical baseline covariates, which might confound the analysis.²⁵ In this study, the propensity score is being used to estimate the likelihood that any given participant falls into one diagnosis population (versus the other) based on their baseline covariates. Postpropensity score matching, it should be impossible to predict the diagnosis of any given participant by baseline covariates alone. This enables us to measure the true effect of diagnosis on caregiver satisfaction.

There are multiple methods of propensity scoring, each of which aims to reduce the distance between observations from the two cohorts.²¹ In order to select the method with the greatest bias reduction, multiple methods for propensity scoring including nearest neighbor (with/without replacement and with/without calipers), optimal matching, and full matching were evaluated for bias reduction. The method that achieved the greatest bias reduction was 1:1 matching with calipers (set at 0.2) and this was used to match the two cohorts for comparison.²⁶

After propensity scoring was complete, logistic regression with odds ratios was used to examine the difference in satisfaction between the two cohorts. Separate logistic regression equations were run for each satisfaction outcome (global satisfaction and domain scores from the four domains). Stata/MP statistical software version 12.0 (StataCorp, College Station, TX) was used for most statistical analysis, although R was used for the propensity score matching.

Because we had to exclude significant amounts of data due to incomplete responses, we compared the characteristics of those with complete answers (respondents) versus those with incomplete answers (nonrespondents). We compared respondents to nonrespondents via t-tests and chi-square tests and then examined whether respondent status predicted any of the outcomes by logistic regression. This study was approved by the institutional review board of the University of Pennsylvania.

Results

Data from 2011 on 90,548 heart failure and cancer caregiver respondents were provided by NHPCO. Of these, 34,411 were eligible for inclusion, of which 8175 were heart failure caregivers and 26,236 were cancer caregivers, reflecting the known distribution of cancer and cardiovascular diagnosis in hospice.

As Table 1 demonstrates, caregiver and patient demographic and clinical variables were significant correlates of diagnosis cohort. Caregivers from both cohorts were likely to be white females. Heart failure caregivers were more likely to be adult children, while cancer caregivers were generally female spouses/partners. Heart failure patients were more likely to suffer shortness of breath, while slightly less likely to suffer pain than cancer patients. In assessing for multicollinearity, only patient and caregiver race and ethnicity had correlations above 0.7. For matching, only caregiver race and ethnicity were included.

Table 1.

Comparison of Baseline Demographic and Clinical Characteristics of Heart Failure and Cancer Cohorts

	Heart failure (n = 8175) %	Cancer (n = 26,236) %	Odds ratio
Patient demographics
<70 years of age	5.4	35.4	1.0
70–79 years	14.5	28.3	3.4^*
80–89 years	44.4	28.3	10.37^*
≥90 years	35.8	8.1	29.2^*
Female	55.5	49.4	1.3^*
White	95.5	93.5	0.8^*
Hispanic	1.5	1.9	1.2
≤ High school diploma	67.9	58.9	0.7^*
Caregiver demographics
<50 years of age	8.2	16.5	1.0
50–69	61.4	55.2	2.2^*
≥70	30.4	28.3	2.2^*
Female	75.7	70.5	1.3^*
White	95.5	93.5	0.8^*
Hispanic	1.8	1.9	1.1
≤ High school diploma	36.3	38.0	1.08^**
Patient's spouse/partner	27.8	49.8	1.0
Patient's child	56.9	34.4	3.0^*
Other relationship	15.3	15.8	1.7^*
Patient clinical characteristics
Pain	81.7	91.7	2.0^*
Dyspnea	64.3	50.0	0.6^*
Anxiety	49.1	56.2	1.3^*
Hospice length of stay
<7 days	37.0	29.9	1.0
8–29 days	21.4	27.4	0.6^*
1–3 months	23.8	30.7	0.6^*
>4 months	17.8	12.0	1.2^*
Received hospice care in nursing home	30.2	12.6	0.3^*
Hospice organization characteristics
Average daily census >50	50.4	47.7	1.11^*
Freestanding (versus provider-based) hospice	47.0	49.7	1.12^*

p < 0.0001.

p < 0.01.

After propensity score matching was completed, 7370 caregivers from each cohort were matched for a total sample of 14,740. We were able to match 90% of the total heart failure caregivers available for matching, but only 28% of the total cancer caregivers available. Table 2 shows that although differences in demographic, clinical, and agency variables remain between the two cohorts, the differences are greatly reduced from the prematched sample.

Table 2.

Comparison of Postmatch Characteristics of Heart Failure and Cancer Cohorts

	Heart failure % (n = 7370)	Cancer % (n = 7370)	Odds ratio
Patient demographics
<70 years of age	5.8	9.3	1.0
70–79	15.6	18.2	0.72^*
80–89	47.4	47.4	0.62^*
≥90	31.2	25.1	0.50^*
Female	53.8	53.1	0.97
White	95.3	94.1	1.14^**
Hispanic	1.5	1.7	0.87
≤ High School diploma	66.9	65.1	1.09^***
Caregiver demographics
<50 years of age	8.5	11.4	1.0
50–69	60.1	56.0	0.69^*
≥70	31.5	32.6	0.77^*
Female	75.2	73.5	0.91^***
White	95.2	94.2	1.11
Hispanic	1.8	1.9	0.93
≤ High school diploma	37.3	36.6	1.03
Patient's spouse/partner	30.2	34.5	1.0
Patient's child	55.1	49.2	0.78^*
Other relationship	14.7	16.4	0.97
Patient clinical characteristics
Pain	83.2	85.1	0.96
Dyspnea	61.8	56.3	1.15^*
Anxiety	49.6	50.1	0.99
Hospice length of stay
<7 days	35.6	28.9	1.0
8–29 days	20.1	25.5	1.5^*
1–3 months	29.8	34.8	1.4^*
>4 months	14.6	10.8	0.91
Hospice care received in nursing home	26.6	25.0	1.09
Hospice organization characteristics
Average daily census >50	50.1	50.0	0.98
Freestanding (versus provider-based) hospice	57.4	59.3	0.93^***

p < 0.0001.

p < 0.01.

***

p < 0.05.

About 75% of caregivers reported that hospice care was “excellent,” and nearly 95% were satisfied with symptom management. However, almost half were dissatisfied with the teaching provided by the hospice and a third were dissatisfied with emotional support provided. The prematched cohorts had statistically significant differences in global satisfaction (OR = 1.06, p = 0.03), emotional support (OR = 1.07, p = 0.02), caregiver teaching by hospice (OR = 0.90, p < 0.0001), and coordination of care (OR = 0.93, p = 0.04), but the effect sizes were very small. While heart failure caregivers were slightly more likely to be dissatisfied with emotional support and to report lower global satisfaction, they were less likely to be dissatisfied with caregiver teaching and coordination of care. These differences disappeared postmatching (see Table 3).

Table 3.

Outcomes Pre- and Postpropensity Score Matching for Heart Failure and Cancer Caregiver Cohorts

Satisfaction measure	Prematch heart failure % dissatisfied	Prematch cancer % dissatisfied	Odds ratio (p-value)	Postmatch heart failure % dissatisfied	Postmatch cancer % dissatisfied	Odds ratio (p-value)
Global satisfaction	25.5	24.3	1.06 (0.030)	23.4	24.3	1.05 (0.209)
Symptom management	6.8	6.6	1.02 (0.700)	6.2	6.4	1.03 (0.709)
Caregiver teaching	41.6	44.2	0.90 (<0.0001)	44.4	43.2	0.95 (0.131)
Emotional support	30.9	29.5	1.07 (0.020)	31.1	29.6	0.93 (0.055)
Coordination of care	17.9	18.9	0.93 (0.040)	18.2	18.7	1.03 (0.483)

Our missing data analysis indicated that respondents were more likely to be white, female, non-Hispanic, and hold a college degree or higher. They were also more likely to be served by a freestanding hospice agency, rather than a provider-based agency (see Table 4). There was no significant relationship between respondent status and satisfaction outcomes.

Table 4.

Family Evaluation of Care Respondent versus Nonrespondent Characteristics

Variable	Respondents % (n = 34,411)	Nonrespondents % (n = 56,137)
Diagnosis
Heart Failure	23.71	20.84
Cancer	76.29	79.16
Patient sex
Male	49.23	43.80
Female	50.77	42.23
Missing		13.96
Patient education
< High school diploma	19.93	13.90
High school diploma	40.89	29.46
Some college	19.45	14.03
Bachelor's	10.17	7.01
Graduate degree	9.56	8.48
Missing		27.13
Patient ethnicity
Non-Hispanic	98.14	85.44
Hispanic	1.86	2.96
Missing		11.60
Patient race
White	94.00	85.32
Black	3.06	4.67
Other	2.94	3.12
Missing		6.89
Patient-caregiver relationship
Spouse/partner	44.62	44.63
Child	39.82	33.72
Other	15.56	15.07
Missing		6.58
Caregiver age
>80	10.05	2.40
70–79	18.79	3.97
60–69	30.11	5.25
50–59	26.57	4.39
<50	14.48	2.45
Missing		81.55
Caregiver sex
Male	28.23	21.14
Female	71.77	54.20
Missing		24.66
Caregiver education
< High school diploma	5.31	4.24
High school diploma	32.34	22.92
Some college	29.41	20.60
Bachelor's	15.51	10.20
Graduate degree	17.43	14.09
Missing		27.95
Caregiver ethnicity
Non-Hispanic	98.06	68.21
Hispanic	1.94	2.59
Missing		29.21
Caregiver race
White	93.99	66.56
Black	3.01	3.64
Other	3.00	2.19
Missing		27.60
Place of Care
Nursing home	16.84	13.33
Other	83.16	72.00
Missing		14.66
Length of Stay
<2 days	9.34	6.86
3–7 days	22.49	20.09
8–14 days	15.00	15.15
15–29 days	10.87	15.29
1–3 months	28.94	23.67
4–6 months	7.54	7.87
>6 months	5.82	6.32
Missing		4.75
Pain	89.30	88.26
Dyspnea	53.31	50.77
Anxiety	54.51	51.32
Ownership
Freestanding	50.91	67.04
Provider-based	49.09	32.96

Discussion

Our research hypothesis that diagnosis would affect caregiver satisfaction with hospice care was not supported by our study results. Statistically significant differences in caregiver satisfaction between the prematched heart failure and cancer caregivers appear to have been due to demographic and clinical differences, but effect sizes were small. An important incidental finding is that while hospices are doing a good job of caring for the patient (high satisfaction with symptom management), there are significant gaps when it comes to caring for the caregiver.

We found little difference in caregiver satisfaction by diagnosis despite previous studies showing that heart failure patients are more likely to use emergency care services and leave hospice care than cancer patients.⁹ Mitchell and colleagues found similar results when they examined satisfaction rates in dementia and cancer hospice caregivers and found no significant difference based on diagnosis.²⁷ Caregiver perceptions of hospice care do not appear to be altered by disease process, despite differences in objective quality measures. From the perspective of family caregivers, hospice professionals currently provide equally satisfactory care to heart failure and cancer, particularly in the area of symptom management. Hospice professionals should continue to educate themselves on specific heart failure issues (e.g., fluid volume management), but symptom management in the end of life may have more similarities across the diagnosis spectrum than differences.²⁸ This suggests that future symptom management trials should include patients and caregivers from multiple diagnosis groups, rather than being diagnosis specific.

While diagnosis does not appear to make a difference in caregiver satisfaction with hospice care, there were differences between the prematched populations in terms of satisfaction. Patients and caregivers cannot be separated from their demographic and clinical differences in real life: that the average heart failure caregiver is a 70-year-old female caring for her mother does impact care planning for that caregiver. While the effect sizes were small in this study, we suggest that future research into end-of-life and hospice outcomes in the heart failure population adjust for these demographic and clinical characteristics. As hospice organizations partner with cardiology practices to create programs specifically designed for the end-stage heart failure patient, they should take these characteristics into account and consider them when evaluating program outcomes.¹¹ This should include evaluation of wider social support for the patient, attention to the caregiver's health and self-care, and the availability of teaching methods that are adapted to visual and auditory disabilities.

Taking caregiver characteristics into consideration is particularly important given that hospices are falling short in caring for caregivers. The high levels of dissatisfaction with teaching and emotional support were concerning. Caregivers' perceptions of the care provided to them have historically been an area of weakness for hospice agencies.^29,30 Csikai and colleagues documented that caregivers reported a lack of information and education on what to expect in terms of the dying process and how to properly care for their loved one.^30,31 Providing adequate caregiver teaching and emotional support is particularly challenging when the length of hospice stay is short.^13,29

Hospice organizations and professionals must examine what they provide in terms of caregiver teaching and support and how they provide it. Caregivers should be treated as second patients, as well as members of the care team. Caregiver knowledge, desire for knowledge, and learning preferences should be assessed upon hospice admission. Caregiver emotional and spiritual needs, along with desire for emotional and spiritual support, should also be assessed. A care plan for the caregiver should be established along with a care plan for the patient. In addition, further research is needed to understand the current state of caregiver teaching and emotional support, the best methods and timing for providing teaching and emotional support, and the relationship between caring for the caregiver and outcomes such as health care utilization and acute care death.

Limitations and strengths

The findings of this study should be considered in the light of certain limitations and strengths. Because use of and response to the FEHC is voluntary, it is difficult to know whether caregiver respondents are truly representative of the national population of informal caregivers. In comparing our patient characteristics to those from Medicare Payment Advisory Committee (MedPAC), they are fairly representative. National data on end-of-life caregivers is scarce, however. And the exclusion of incomplete responses and significant missing data on caregiver variables meant that some caregiver responses were lost, with a higher volume of younger and Hispanic caregivers lost than others. In addition, propensity scoring does not balance the two cohorts in terms of unobserved covariates such as caregivers' previous caregiving knowledge, financial strain, or patient comorbidities.

Well-known issues with both postservice surveys and the measurement of satisfaction are that responses are historically skewed toward the positive. and users' global sense of the experience influences their responses to specific questions.³² Furthermore, respondents' view of their specific health care providers (e.g., nurses) influences their responses; thus, respondents who liked their care providers tend to provide positive responses about the service in general.^32,33 While caregiver satisfaction is a key quality measurement, hospice providers should also examine objective outcomes among diagnosis groups.

Despite these limitations, this study is strengthened by being one of the first to use a large, national dataset to compare caregiver satisfaction between the heart failure and cancer populations. It provides a critical examination of the way in which diagnosis may influence caregiver satisfaction. The propensity score analysis used to compare heart failure and cancer caregivers helps to ensure that issues with sampling and observed confounders did not unduly influence the outcome.

Conclusions

Overall, hospice care was perceived in a positive light by the informal caregivers in this national study as evidenced by high satisfaction rates with care provided to the patients. A heart failure diagnosis does not affect caregivers' satisfaction with hospice care. These findings underscore the important role hospice plays in shaping the experiences of heart failure caregivers at vulnerable times in their lives. Nevertheless, our data indicates a potential need for cardiovascular and hospice care providers to examine the services provided to caregivers in the areas of teaching and emotional support and their implications in shaping caregiver outcomes.

Footnotes

Author Disclosure Statement

None of the authors have any disclosures or conflicts of interest.

References

Mozaffarian

, Benjamin

, Go

, et al.: Heart Disease and Stroke Statistics—2015 Update: A Report From the American Heart Association. Circ, 2014; 131:e329–e322.

Christakis

, Iwashyna

: The health impact of health care on families: A matched cohort study of hospice use by decedents and mortality outcomes in surviving, widowed spouses. Soc Sci Med, 2003; 57:465–475.

Garrido

, Prigerson

: The end-of-life experience: Modifiable predictors of caregivers' bereavement adjustment. Cancer, 2013; 120:918–925.

Carr

: “I don't want to die like that …”: The impact of significant others' death quality on advance care planning. Gerontologist, 2012; 52:770–781.

National Hospice & Palliative Care Organization: NHPCO's Facts and Figures: Hospice Care in America, 2013 edition. National Hospice & Palliative Care Organization. www.nhpco.org2013 (last accessed May 1, 2015).

National Hospice & Palliative Care Organization: NHPCO's Facts & Figures: Hospice Care in America, 2014 edition. National Hospice & Palliative Care Organization. www.nhpco.org2014 (last accessed May 1, 2015).

Addington-Hall

, O'Callaghan

: A comparison of the quality of care provided to cancer patients in the UK in the last three months of life in in-patient hospices compared with hospitals, from the perspective of bereaved relatives: Results from a survey using the VOICES questionnaire. Palliat Med, 2009; 23:190–197.

Blecker

, Anderson

, Herbert

, et al.: Hospice care and resource utilization in Medicare beneficiaries with heart failure. Med Care, 2011; 49:985–991.

Cheung

, Schaefer

, May

, et al.: Enrollment and events of hospice patients with heart failure vs. cancer. J Pain Symptom Manage, 2013; 45:552–560.

10.

MacKenzie

: Abstract 270: Emergency services use in heart failure patients on hospice. Circ Cardiovasc Qual Outcomes, 2013; 6:A270.

11.

Goodlin

, Kutner

, Connor

, et al.: Hospice care for heart failure patients. J Pain Symptom Manage, 2005; 29:525–528.

12.

Miller

, Weitzen

, Kinzbrunner

: Factors associated with the high prevalence of short hospice stays. J Palliat Med, 2003; 6:725–736.

13.

Schockett

, Teno

, Miller

, Stuart

: Late referral to hospice and bereaved family member perception of quality of end-of-life care. J Pain Symptom Manage, 2005; 30:400–407.

14.

Teno

, Casarett

, Spence

, Connor

: It is “too late” or is it? Bereaved family member perceptions of hospice referral when their family member was on hospice for seven days or less. J Pain Symptom Manage, 2012; 43:732–738.

15.

Rhodes

, Mitchell

, Miller

, et al.: Bereaved family members' evaluation of hospice care: What factors influence overall satisfaction with services?. J Pain Symptom Manage, 2008; 35:365–371.

16.

MacKenzie

, Buck

, Meghani

, Riegel

: Abstract: Comparing predictors of heart failure and cancer caregiver satisfaction with hospice care. Circ, 2014; 130:A16291.

17.

Goodlin

: Palliative care for end-stage heart failure. Curr Heart Fail Rep, 2005; 2:155–160.

18.

Allison

: Multiple Regression: A Primer. Thousand Oaks, CA: Pine Forge Press, 1999.

19.

Connor

, Teno

, Spence

, Smith

: Family evaluation of hospice care: Results from voluntary submission of data via website. J Pain Symptom Manage, 2005; 30:9–17.

20.

Teno

, Clarridge

, Casey

, et al.: Validation of toolkit after-death bereaved family member interview. J Pain Symptom Manage, 2001; 22:752–758.

21.

Austin

: A comparison of 12 algorithms for matching on the propensity score. Stat Med, 2014; 33:1057–1069.

22.

Rhodes

, Teno

, Connor

: African American bereaved family members' perceptions of the quality of hospice care: Lessened disparities, but opportunities to improve remain. J Pain Symptom Manage, 2007; 34:472–479.

23.

Teno

, Shu

, Casarett

, et al.: Timing of referral to hospice and quality of care: Length of stay and bereaved family members' perceptions of the timing of hospice referral. J Pain Symptom Manage, 2007; 34:120–125.

24.

Austin

: An introduction to propensity score methods for reducing the effects of confounding in observational studies. Multivariate Behav Res, 2011; 46:399–424.

25.

Parolari

, Dainese

, Naliato

, et al.: Do women currently receive the same standard of care in coronary artery bypass graft procedures as men? A propensity analysis. Ann Thorac Surg, 2008; 85:885–890.

26.

Austin

: Optimal caliper widths for propensity‐score matching when estimating differences in means and differences in proportions in observational studies. Pharm Stat, 2011; 10:150–161.

27.

Mitchell

, Kiely

, Miller

, et al.: Hospice care for patients with dementia. J Pain Symptom Manage, 2007; 34:7–16.

28.

Solano

, Gomes

, Higginson

: A comparison of symptom prevalence in far advanced cancer, AIDS, heart disease, chronic obstructive pulmonary disease and renal disease. J Pain Symptom Manage, 2006; 31:58–69.

29.

Miceli

, Mylod

: Satisfaction of families using end-of-life care: Current successes and challenges in the hospice industry. Am J Hosp Palliat Care, 2003; 20:360–370.

30.

Csikai

: Bereaved hospice caregivers' perceptions of the end-of-life care communication process and the involvement of health care professionals. J Palliat Med, 2006; 9:1300–1309.

31.

Csikai

, Martin

: Bereaved hospice caregivers' views of the transition to hospice. Soc Work Health Care, 2010; 49:387–400.

32.

Williams

, Coyle

, Healy

: The meaning of patient satisfaction: An explanation of high reported levels. Soc Sci Med, 1998; 47:1351–1359.

33.

Zifko-Baliga

, Krampf

: Managing perceptions of hospital quality. Mark Health Serv, 1997; 17:28–35.