University of Southern California Dissertations and Theses

Racial/ethnic variation in care preferences and care outcomes among United States hospice enrollees

(USC Thesis Other)

Racial/ethnic variation in care preferences and care outcomes among United States hospice enrollees

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Title page
Racial/Ethnic Variation in Care Preferences and Care Outcomes
among United States Hospice Enrollees
By

Jeff Laguna

A Dissertation Presented to the
FACULTY OF THE USC SCHOOL OF GERONTOLOGY
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(GERONTOLOGY)

August 2014

Copyright 2014 Jeff Laguna

i
Dedication
This work, and all that led to it, is dedicated to the two most amazing people I
have ever known: my father, Richard J. Laguna, and my mother, Jodie M. Laguna. You
have always been there to love me, support me, challenge me, guide me, and motivate
me. I would not be who I am today if it were not for you two.
I am, because of you.

ii
Table of Contents
Dedication ........................................................................................................................... i
List of Tables .................................................................................................................. viii
List of Figures .................................................................................................................... x
List of Appendices ........................................................................................................... xii
Abstract ........................................................................................................................... xiii
Chapter 1: The Problem and Its Underlying Framework ............................................ 1
Background of the Problem ......................................................................................... 1
Purpose of the Study ..................................................................................................... 2
Research Questions ....................................................................................................... 3
Significance of the Problem .......................................................................................... 3
Definition of Terms ....................................................................................................... 4
Organization of the Study ............................................................................................ 6
Chapter 2: Review of the Literature ............................................................................... 8
Documentation .............................................................................................................. 9
Hospice: A Model for End-of-Life Care ..................................................................... 9
Length of Hospice Stay ............................................................................................... 12
Racial/Ethnic variation in hospice length of stay. .................................................... 14
Advance Care Planning .............................................................................................. 15
Non-hospice population. ........................................................................................... 16
Hospice enrollees. ..................................................................................................... 17
Emergent Care Utilization ......................................................................................... 18
iii
Racial/Ethnic differences in emergent care utilization at end of life. ....................... 19
Hospice enrollees. ..................................................................................................... 20
Site of Death ................................................................................................................. 20
Racial/ethnic variation in site of death. ..................................................................... 21
Hospice enrollees. ..................................................................................................... 22
Determinants of Racial/Ethnic Variation in Hospice Use ....................................... 23
Cultural beliefs. ......................................................................................................... 23
Religious values. ....................................................................................................... 25
Knowledge and education. ........................................................................................ 26
Communication and trust. ......................................................................................... 27
Importance of the Topic ............................................................................................. 29
Chapter 3: Research Methodology ................................................................................ 31
Research Questions and Hypotheses ......................................................................... 31
Research questions. ................................................................................................... 31
Hypotheses. ............................................................................................................... 31
Research Design .......................................................................................................... 32
Data Sources ................................................................................................................ 33
Agency selection. ...................................................................................................... 33
Patient selection. ....................................................................................................... 34
Data collection. ......................................................................................................... 34
Data use. .................................................................................................................... 35
Human Subjects Protection ....................................................................................... 35
Sampling Plan ............................................................................................................. 35
iv
Operationalization of the Variables and Measures ................................................. 35
Demographic and health indicator variables. ............................................................ 35
Advance directive completion. ................................................................................. 36
Do not resuscitate order election. .............................................................................. 36
Healthcare proxy designation. .................................................................................. 37
Emergent care utilization. ......................................................................................... 37
Hospice length of stay. .............................................................................................. 38
Site of death. ............................................................................................................. 38
Conceptual Model ....................................................................................................... 39
Andersen’s Behavior Model of Health Services Use. .............................................. 39
Revised model. .......................................................................................................... 40
Analysis Plan ............................................................................................................... 42
Logistic regressions. ................................................................................................. 45
Cox proportional hazards regression. ....................................................................... 46
Chapter 4: Description of Analytic Sample .................................................................. 48
Full Sample .................................................................................................................. 49
Discharge reasons. .................................................................................................... 50
Decedent Sample ......................................................................................................... 51
Chapter 5: Advance Care Planning Decisions ............................................................. 53
Bivariate Comparisons ............................................................................................... 53
Advance directive completion. ................................................................................. 53
Do not resuscitate order election. .............................................................................. 54
Healthcare proxy designation. .................................................................................. 56
v
Multivariable Models .................................................................................................. 58
Advance directive completion. ................................................................................. 59
Do not resuscitate order election. .............................................................................. 61
Healthcare proxy designation. .................................................................................. 64
Chapter 6: Emergent Care Utilization .......................................................................... 67
Bivariate Comparisons ............................................................................................... 67
Multivariable Models .................................................................................................. 68
Step 1: Demographics and health indicator variables. .............................................. 69
Step 2: Inclusion of advance care planning decisions. ............................................. 70
Chapter 7: Hospice Length of Stay ............................................................................... 73
Bivariate Comparisons ............................................................................................... 73
Log-rank tests. ........................................................................................................... 75
Multivariable Models .................................................................................................. 78
Step 1: Demographics and health indicator variables. .............................................. 78
Step 2: Inclusion of advance care planning decisions. ............................................. 79
Ancillary Bivariate Comparisons .............................................................................. 80
Hospice length of stay: 0-7 Days. ............................................................................. 80
Hospice length of stay: 0-30 Days. ........................................................................... 82
Ancillary Multivariable Models ................................................................................. 84
Hospice length of stay: 0-7 days. .............................................................................. 85
Step 1: Demographics and health indicator variables. .......................................... 85
Step 2: Inclusion of advance care planning decisions. ......................................... 87
Hospice length of stay: 0-30 days. ............................................................................ 88
vi
Step 1: Demographics and health indicator variables. .......................................... 88
Step 2: Inclusion of advance care planning decisions. ......................................... 90
Chapter 8: Site of Death ................................................................................................. 93
Overall Site of Death ................................................................................................... 93
Bivariate Comparisons ............................................................................................... 94
Site of death: Home-like setting. .............................................................................. 94
Site of death: Hospital. .............................................................................................. 96
Multivariable Analyses ............................................................................................... 98
Site of death: Home-like setting. .............................................................................. 99
Step 1: Demographics and health indicator variables. .......................................... 99
Step 2: Inclusion of advance care planning decisions. ....................................... 100
Site of death: Hospital. ............................................................................................ 102
Step 1: Demographics and health indicator variables. ........................................ 102
Step 2: Inclusion of advance care planning decisions. ....................................... 104
Chapter 9: Discussion ................................................................................................... 107
Hypothesis 1 Findings ............................................................................................... 107
Hypothesis 1A Findings ............................................................................................ 108
Hypothesis 1B Findings ............................................................................................ 109
Hypothesis 2 Findings ............................................................................................... 110
Hypothesis 2A Findings ............................................................................................ 111
Hypothesis 3 Findings ............................................................................................... 112
Hypothesis 4 Findings ............................................................................................... 113
Hypothesis 4A Findings ............................................................................................ 114
vii
Contributions to the Literature ............................................................................... 115
Advance care planning. ........................................................................................... 115
Advance directive completion. ........................................................................... 115
Do not resuscitate order election. ........................................................................ 117
Healthcare proxy designation. ............................................................................ 119
Emergent care utilization. ....................................................................................... 121
Hospice length of stay. ............................................................................................ 124
Site of death. ........................................................................................................... 126
Limitations and Research Recommendations ........................................................ 128
Conclusion ................................................................................................................. 131
References ...................................................................................................................... 133
Appendices ..................................................................................................................... 165
Appendix

viii
List of Tables
Table 1. Full Sample Size ............................................................................49
Table 2. Weighted Full Sample Description ................................................50
Table 3. Reasons for Hospice Discharge .....................................................51
Table 4. Weighted Decedent Sample Description .......................................51
Table 5. Advance Directive Completion Bivariate Comparisons ................54
Table 6. Do Not Resuscitate Order Election Bivariate Comparisons ..........56
Table 7. Healthcare Proxy Designation Bivariate Comparisons ..................58
Table 8. Predictors of Advance Directive Completion ................................59
Table 9. Predictors of Do Not Resuscitate Order Election ..........................62
Table 10. Predictors of Healthcare Proxy Designation ..................................65
Table 11. Emergent Care Utilization Bivariate Comparisons ........................68
Table 12. Predictors of Emergent Care Utilization ........................................70
Table 13. Decedent Hospice Length of Stay (Full Care Episode) Bivariate
Comparisons ..................................................................................75
Table 14. Predictors of Decedent Hospice Length of Stay: Full Care
Episode ...........................................................................................79
Table 15. Decedent Hospice Length of Stay (0-7 Days) Bivariate
Comparisons ..................................................................................82
Table 16. Decedent Hospice Length of Stay (0-30 Days) Bivariate
Comparisons ..................................................................................84
Table 17. Predictors of Decedent Hospice Length of Stay: 0-7 Days ...........86
Table 18. Predictors of Decedent Hospice Length of Stay: 0-30 Days .........90
ix
Table 19. Overall Site of Death .....................................................................94
Table 20. Site of Death (Home-Like Setting) Bivariate Comparisons ..........96
Table 21. Site of Death (Hospital) Bivariate Comparisons ...........................98
Table 22. Predictors of Site of Death: Home-Like Setting ..........................100
Table 23. Predictors of Site of Death: Hospital ...........................................104
x
List of Figures
Figure 1. Proportion of Total Days of Hospice Care in the United States ....13
Figure 2. Andersen’s Behavior Model of Health Services Use ....................40
Figure 3. Revised Conceptual Model ............................................................41
Figure 4. Multilevel Regression Models .......................................................44
Figure 5. Reduced Form Multilevel Regression Model ................................45
Figure 6. Relative Effects of Predictors for Advance Directive
Completion .....................................................................................61
Figure 7. Relative Effects of Predictors for Do Not Resuscitate Order
Election ..........................................................................................64
Figure 8. Relative Effects of Predictors for Healthcare Proxy
Designation ....................................................................................66
Figure 9. Relative Effects of Predictors for Emergent Care Utilization .......72
Figure 10. Unadjusted Survival Curves for White and Black Hospice
Decedents .......................................................................................76
Figure 11. Unadjusted Survival Curves for White and Hispanic Hospice
Decedents .......................................................................................77
Figure 12. Unadjusted Survival Curves for Black and Hispanic Hospice
Decedents .......................................................................................77
Figure 13. Relative Effects of Predictors for Decedent Hospice
Length of Stay: 0-7 Days ...............................................................88
Figure 14. Relative Effects of Predictors for Decedent Hospice
Length of Stay: 0-30 Days .............................................................92
xi

Figure 15. Relative Effects of Predictors for Site of Death: Home-Like
Setting ..........................................................................................102
Figure 16. Relative Effects of Predictors for Site of Death: Hospital ...........106
xii
List of Appendices
Appendix A. Weighted and Unweighted Example Description ........................165
Appendix B. Unweighted Full Sample Description ..........................................166
Appendix C. Unweighted Decedent Sample Description .................................167
Appendix D. Advance Directive Completion Sample Descriptives ..................168
Appendix E. Do Not Resuscitate Order Election Sample Descriptives ............169
Appendix F. Healthcare Proxy Designation Sample Descriptives ...................170
Appendix G. Emergent Care Utilization Sample Descriptives .........................171
Appendix H. Decedent Hospice Length of Stay (Full Care Episode) Sample
Descriptives ..................................................................................172
Appendix I. Decedent Hospice Length of Stay (Full Care Episode) Median
Comparisons ................................................................................173
Appendix J. Decedent Hospice Length of Stay (0-7 Days) Sample
Descriptives ..................................................................................174
Appendix K. Decedent Hospice Length of Stay (0-30 Days) Sample
Descriptives ..................................................................................175
Appendix L. Site of Death (Home-Like Setting) Sample Descriptives ............176
Appendix M. Site of Death (Hospital) Sample Descriptives .............................177
Appendix N. Original Site of Death (Full Responses) ......................................178
xiii
Abstract
Despite the rapid growth of hospice care in the United States over the past several
decades, racial/ethnic minorities continue to utilize higher levels of aggressive life-
prolonging interventions at end of life, often resulting in poorer care experiences. While
previous research has expanded understanding of racial/ethnic end-of-life disparities
outside of hospice, an in-depth analysis of the relationship between patient care
preferences and key end-of-life outcomes within a racially/ethnically diverse hospice
population remains to be conducted.
Using the 2007 wave of the National Home Health and Hospice Care Survey
(NHHCS), a retrospective analysis of clinical and service use outcomes was conducted to
test for racial/ethnic variation following hospice enrollment. Key outcomes of interest
included advance care planning, emergent care utilization, hospice length of stay, and site
of death. In total, 3,661 White, Black, and Hispanic Medicare hospice patients were
analyzed, representing approximately 788,872 older Americans. Results indicated that
advance care planning varied by race/ethnicity, with Blacks less likely to complete an
advance directive, Hispanics more likely to elect a do not resuscitate order, and both
Blacks and Hispanics less likely to designate a healthcare proxy. Findings also indicated
that Blacks were less likely to utilize emergent care following adjustment for advance
care planning. While Hispanics were more likely to die in the first week of hospice care,
Blacks were more likely to die in the first month of hospice care. Concerning site of
death, Blacks and Hispanics were more likely to die in a home-like setting, and Hispanics
were also more likely to die in a hospital. Results also indicated that advance care
planning reduced the likelihood of emergent care utilization, death in the first week of
xiv
hospice care, death in the first month of hospice care, and death in a hospital. Lastly,
patients engaging in advance care planning were also more likely to die in a home-like
setting.
Findings support racial/ethnic variation following hospice enrollment, but suggest
that differences within hospice contrast with those in the larger healthcare system.
Furthermore, results support the protective effect of advance care planning among
hospice enrollees. The data presented have substantial clinical and policy implications for
improving the care of all patients at end of life. Additional research is needed to better
understand and address reported racial/ethnic differences following hospice enrollment.
1
Chapter 1: The Problem and Its Underlying Framework
Background of the Problem
Although high-quality care for the dying is available, racial/ethnic minorities
continue to have poor end-of-life (EOL) care experiences. During the last year of life,
racial/ethnic minorities utilize less palliative-based care and more aggressive, acute
interventions that do little to extend life while negatively impacting patient symptom
management, quality of life, emotional support, and care satisfaction (Barnato, Chang,
Saynina, & Garber, 2007; Earle et al., 2008; Goldstein, Elliott, Lehrman,
Hambarsoomian, & Giordano, 2010; Hanchate, Kronman, Young-Xu, Ash, & Emanuel,
2009; Mazanec, Daly, & Townsend, 2010; A. K. Smith, Earle, & McCarthy, 2009; Teno
et al., 2004; Zhang et al., 2009). Although differences are due in part to a decreased
willingness among racial/ethnic minorities to forgo curative treatments (Barnato,
Anthony, Skinner, Gallagher, & Fisher, 2009; Casarett, Van Ness, O'Leary, & Fried,
2006; Duffy, Jackson, Schim, Ronis, & Fowler, 2006), care disparities exist (Hanchate et
al., 2009). To address these disparities, research has focused on improving racial/ethnic
minority patient knowledge of and access to EOL care programs; however, research
investigating minority outcomes within hospice, the most widely used form of EOL care
in the United States, is scarce.
Hospice is designed to meet the unique needs of dying patients in the last six
months of life. Studies of hospice have documented effective pain management (Hanlon,
Perera, Sevick, Rodriguez, & Jaffe, 2010), higher care satisfaction among patients and
families (Dy et al., 2008; Teno et al., 2004), and an increased likelihood of dying in-place
(Hogan, Lunney, Gabel, & Lynn, 2001; Teno et al., 2004). Despite these benefits,
2
hospice remains widely underutilized by racial/ethnic minorities (Enguidanos, Yip, &
Wilber, 2005; Givens, Tjia, Zhou, Emanuel, & Ash, 2010; K. S. Johnson, Kuchibhatla,
Tanis, & Tulsky, 2008; Kwak, Haley, & Chiriboga, 2008; Lepore, Miller, & Gozalo,
2011; Ngo-Metzger, Phillips, & McCarthy, 2008). Among the minorities who elect
hospice, care experiences following enrollment remain unclear largely due to outdated
studies and sample limitations. For example, research suggests that hospice length of stay
(LOS), an important care quality indicator, is longer for Blacks and Hispanics compared
to Whites, but findings are limited due to non-representative samples (Cólon & Lyke,
2003; Hardy et al., 2012; K. S. Johnson, Kuchibhatla, & Tulsky, 2011). Similarly,
hospice decedent racial/ethnic variation in site of death (SOD), another key indicator of
EOL care quality, has been examined only once since the Medicare Hospice Benefit was
established in 1983, and significant sample limitations constrain the generalizability of
findings (K. S. Johnson et al., 2005). Thus, while racial/ethnic differences in EOL care
preferences and medical service use have been documented outside of hospice, and
although research suggests that that hospice patient care experiences currently differ by
race/ethnicity, an in-depth analysis of racial/ethnic variation in key EOL outcomes
following hospice enrollment remains to be conducted.
Purpose of the Study
This purpose of this study was to investigate racial/ethnic variation in clinical and
service use outcomes among U.S. hospice patients enrolled in Medicare, the largest
payment provider of U.S. hospice care. Specifically, the study examined racial/ethnic
differences in advance care planning and its influence on emergent care utilization,
hospice LOS, and SOD. Research was conducted using secondary data analysis of the
3
2007 wave of the National Home Health and Hospice Care Survey (NHHCS), a
nationally representative survey of 1,036 agencies. A total of 3,661 hospice patients over
the age of 65 were analyzed who when weighted, represent 788,872 older adults enrolled
in the Medicare Hospice Benefit. This study offers a novel approach to examining
racial/ethnic minority EOL care experiences by investigating care preferences, utilization,
and outcomes within hospice, one of the fastest growing sectors of the U.S. health care
system. Given the rapidly growing segment of U.S. minority elders who require high-
quality EOL care, and increasing national focus on eliminating disparities in the U.S.
healthcare system (USDHHS, 2012), findings are both timely and critical.
Research Questions
The primary research questions guiding the study were as follows:
1. Is there a difference between White, Black, and Hispanic hospice enrollees
in the completion of advance care plans and care choices?
2. Does race/ethnicity influence emergent care utilization among hospice
enrollees? Does engaging in advance care planning affect this
relationship?
3. Among patients who die under the care of hospice, how does
race/ethnicity influence hospice length of stay and site of death?
Significance of the Problem
Numerous studies have documented disparities in access to hospice (Connors et
al., 1995; Greiner, Perera, & Ahluwalia, 2003; Haber, 1999; Laguna, Enguídanos,
Siciliano, & Coulourides-Kogan, 2012; O'Mahony et al., 2008), poorer provider-level
care (Enguidanos et al., 2005; Huskamp et al., 2009; Loggers et al., 2009; Mack, Paulk,
4
Viswanath, & Prigerson, 2010; Muni, Engelberg, Treece, Dotolo, & Curtis, 2011), and
worse patient-level outcomes (Goldstein et al., 2010; Hanchate et al., 2009; Mazanec et
al., 2010; A. K. Smith, Earle, et al., 2009; Young, Meterko, & Desai, 2000)

for
racial/ethnic minorities at EOL. Among racial/ethnic minority hospice enrollees, little is
known about their experience under the care of hospice. Of the few studies examining
hospice disenrollment rates, racial/ethnic minorities have been found to be more likely
than Whites to disenroll from hospice (K. S. Johnson, Kuchibhatla, Tanis, et al., 2008;
Unroe, Greiner, Johnson, Curtis, & Setoguchi, 2012), suggesting that hospice patient care
experiences may differ by race/ethnicity. Identifying differences in care planning, use of
medical services, and care outcomes within a diverse hospice population is needed to
improve understanding of minority care at EOL, and critical to reducing disparities and
improving the quality of care among seriously ill populations. Information gathered can
also inform policy (e.g., Medicare Hospice Benefit) to improve equitable access to and
continuity of hospice care for an increasingly diverse group of older adults.
Definition of Terms
To promote clarity in the diffusion of study findings, several pre-defined
constructs were employed throughout this investigation. Those constructs, along with
commonly abbreviated terms referenced in this manuscript, are listed below.
• ANOVA is an abbreviation of “analysis of variance.”
• Blacks represents patients who are non-Hispanic Black.
• CAPI is an abbreviation of “computer-assisted personal interviewing”
system, one of the primary data collection tools utilized for the 2007 wave
of the National Home and Hospice Care Survey.
5
• DNR is an abbreviation of “do not resuscitate.”
• Emergent care represents unplanned emergency medical care, including
hospital emergency department use, doctor’s office emergency visits, and
outpatient department/clinic use (including urgicenter sites).
• EOL is an abbreviation of “end of life.”
• Hispanics represents patients who identify with the Hispanic ethnicity (all
races included).
• Home-like setting represents two potential care settings: 1) a private home
or apartment, and 2) a residential care place. Moreover, a residential care
place was defined as an assisted living facility, a board and care home, or
a life care/continuing care retirement community.
• ICC is an abbreviation of “Intraclass Correlation Coefficient.”
• LOS is an abbreviation of “length of stay.”
• Minorities represents racial and ethnic minorities.
• NCHS is an abbreviation of the “National Center for Health Statistics,” the
center within the Centers for Disease Control and Prevention that
conducted the survey analyzed in this study.
• NHHCS is an abbreviation of “National Home Health and Hospice Care
Survey,” the national survey utilized in this study.
• QOL is an abbreviation of “quality of life.”
• SOD is an abbreviation of “site of death.” For the purpose of this study,
site of death was operationalized as the location of where the decedent was
staying on the last day of hospice care.
6
• Whites represents patients who are non-Hispanic White.
Organization of the Study
Chapter 1 provides an overall background of the problem, reviews the research
questions that guided the study, discusses the significance of addressing the identified
problem, introduces key definitions that were operationalized in the study, and presents a
general overview of the manuscript’s structure.
Chapter 2 examines the research literature addressing the identified problem. Key
topics covered include 1) hospice care in the United States, 2) hospice LOS, 3) advance
care planning, 4) emergent care utilization, 5) SOD, and 6) determinants of racial/ethnic
variation in hospice use. Research is presented for both hospice enrollees as well as the
general population. Lastly, the overall importance of the study topic is discussed.
Chapter 3 presents the research methodology of the study. The hypotheses driving
the study are proposed, data sources are discussed, and variable operationalization is
reviewed. In addition, the conceptual model that guided the analysis plan is proposed, and
the bivariate and multivariable statistical analyses utilized in the study are discussed.
Chapter 4 provides a general overview of the analytic samples that were utilized
in the study. Univariate descriptives are presented, and bivariate examination of
racial/ethnic variation in demographic characteristics and health indicator variables is
discussed.
Chapter 5 presents an in-depth examination into the advance care planning
hypotheses. Results that are discussed include bivariate and multivariable analyses testing
for racial/ethnic variation in 1) advance directive completion, 2) do not resuscitate (DNR)
order election, and 3) healthcare proxy designation.
7
Chapter 6 examines variation in emergent care utilization. Bivariate findings are
presented first, followed by multivariable tests for 1) racial/ethnic differences, and 2)
variation by advance care planning decisions.
Chapter 7 provides an in-depth examination into racial/ethnic differences in
hospice LOS. Bivariate and multivariable models are utilized to test for relative variation
in 1) the length of the full care episode, 2) death within the first week of hospice care, and
3) death within the first month of hospice care.
Chapter 8 investigates racial/ethnic variation in SOD. Specifically, racial/ethnic
differences in dying in a 1) home-like setting, and 2) hospital are reviewed. As with
Chapters 5-7, bivariate findings are discussed first, followed by multivariable results.
Chapter 9 presents an integrated discussion of study findings in relation to
existing literature, examines clinical implications, reviews study limitations, and proposes
recommendations for future research.
8
Chapter 2: Review of the Literature
Developed to meet the unique medical, palliative, and socioemotional needs of
patients with life-limiting illness, hospice care is both the most widely utilized, as well as
the highest regarded, model of end-of-life (EOL) care in the United States (National
Hospice and Palliative Care Organization, 2013). Since the introduction of the Medicare
hospice benefit in 1982, U.S. hospice programs have expanded rapidly, with current
estimates indicating that approximately 45% of all U.S. deaths each year are under the
care of hospice (National Hospice and Palliative Care Organization, 2012).
Notwithstanding, hospice remains widely underutilized by racial/ethnic minorities, and
care disparities have been identified as a significant contributor (L. L. Cohen, 2008).
Given recent calls by the U.S. Department of Health and Human Services for the
elimination of healthcare disparities by 2020 (USDHHS, 2012), comprehensive
examination of care disparities within the hospice system is both timely and critical.
Research investigating racial/ethnic disparities in access to and utilization of U.S.
hospice care has focused on several key topics, including advance care planning, care
utilization, and care outcomes. This literature review presents a general overview of the
U.S. hospice care model, followed by a background on racial/ethnic variation in hospice
length of stay (LOS), advance care planning, emergent care utilization, and site of death
(SOD). Next, proposed empirically-based determinants for racial/ethnic variation are
discussed. The review concludes with a discussion on the importance of addressing
current gaps in research to improve the lives of seriously ill vulnerable populations at
EOL.

9
Documentation
Research findings reported in this comprehensive literature review were obtained
by searching several research databases, including PubMed, Medline, and ProQuest.
Database search terms included hospice, end-of-life, race/ethnicity,
disparities/differences/variation, palliative care, advance care planning, advance
directives, do not resuscitate/dnr, healthcare proxy/durable power of an attorney/dpoa,
aggressive/acute/emergent care, length of stay, and site/place of death. Additional content
was also obtained by searching U.S. government and non-profit reports.
Hospice: A Model for End-of-Life Care
Originating from calls to improve the care of those life-limiting illness, as well as
reduce costs at EOL, formal U.S. hospice care was established as an official Medicare
benefit in 1982 under the Tax Equity and Fiscal Responsibility Act. Over thirty years
later, it remains the only model of EOL care covered as an official Medicare benefit. To
be eligible for the Medicare Hospice Benefit, patients must be enrolled in Medicare Part
A (Hospital Insurance), and certified by a physician as having a terminal illness with a
prognosis of six months or less to live, given expected illness progression (CMS, 2014).
Hospice care is unique in that it represents a transition from curative, life-prolonging
treatments to care that focuses on patient comfort and quality of life (QOL). Patients
enrolled in hospice services are entitled to several key benefits. First, pain and symptoms
are managed by palliative specialists, and short-term inpatient care is made available
when pain or symptoms are exacerbated. Second, patients and their families receive
supplementary psychological, socioemotional, and spiritual support (as well as other
support, as needed) from an interdisciplinary care team. Third, costs associated with
10
prescribed medication, supplies, and durable medical equipment (DME) are fully
covered. Lastly, families and caregivers are provided with additional respite support as
well as bereavement counseling following the patient’s death (National Hospice and
Palliative Care Organization, 2013).
Hospice has grown substantially in the United States over the past two decades
(Christakis & Escarce, 1996; Connor, 2007; Huskamp, Buntin, Wang, & Newhouse,
2001; National Hospice and Palliative Care Organization, 2013; U.S. General Accounting
Office, 2000). Between 1990 and 2005, the number of active hospice programs in the
United States grew from 1,604 to 4,160 (Connor, 2007). Since 2005, the number of
active hospice programs has risen above 5,500, and estimates indicate that hospice care
will become one of the fastest-growing sectors of the U.S. health care system for years to
come (National Hospice and Palliative Care Organization, 2013; C. Smith, Cowan,
Heffler, & Catlin, 2006). As of 2012, hospice programs provided care to approximately
1.5 million people in the United States, and represented approximately 45% of all U.S.
deaths (National Hospice and Palliative Care Organization, 2013). Recent data also
suggests that nearly a third of all Medicare decedents access hospice for three or more
days in the months preceding death, an 11% increase over the past decade (National
Hospice and Palliative Care Organization, 2013).
Research has linked hospice with improved pain control (Hanlon et al., 2010;
Miller, Mor, & Teno, 2003), improved QOL (Greer et al., 1986; Wallston, Burger, Smith,
& Baugher, 1988), higher patient and family satisfaction with care (Dy et al., 2008; Kane,
Bernstein, Wales, Leibowitz, & Kaplan, 1984; Teno et al., 2004), increased likelihood of
dying in place (Hogan et al., 2001; Teno et al., 2004), and in some cases increased
11
survival rates (Connor, Pyenson, Fitch, Spence, & Iwasaki, 2007; Pyenson, Connor,
Fitch, & Kinzbrunner, 2004; Taylor, Ostermann, Van Houtven, Tulsky, & Steinhauser,
2007). Hospice is also associated with reduced healthcare costs in the last months of life
(Pyenson et al., 2004; Stevenson & Bramson, 2009; Taylor et al., 2007; Teno et al.,
2004), specifically through reductions in hospitalizations (Gozalo & Miller, 2007).
Despite these benefits, studies have consistently documented the under-use of
hospice by racial/ethnic minorities (L. L. Cohen, 2008; Cólon & Lyke, 2003; Connor,
Elwert, Spence, & Christakis, 2008; Enguidanos et al., 2005; Givens et al., 2010; Greiner
et al., 2003; Hackbarth, Reischauer, & Miller, 2009; Han, Remsburg, & Iwashyna, 2006;
K. S. Johnson, Kuchibhatla, Tanis, et al., 2008; Kwak et al., 2008; Lepore et al., 2011;
Ngo-Metzger et al., 2003; Ngo-Metzger et al., 2008; A. K. Smith, Earle, et al., 2009). In a
recent study of over 98,000 Medicare beneficiaries, Givens and colleagues (2010) found
that Black and Hispanic heart failure patients were less likely than Whites to enroll in
hospice following diagnosis. Similarly, a 2011 study of over 115,000 older adults
reported that urban-dwelling Blacks and Hispanics were approximately 20% less likely to
receive hospice services than Whites (Hardy et al., 2012). Factors associated with
racial/ethnic variation in hospice utilization include patient unwillingness to forgo
curative measures (Barnato et al., 2009; Borum, Lynn, & Zhong, 2000; Casarett,
Crowley, & Hirschman, 2004; Casarett et al., 2005; Casarett et al., 2006; Duffy et al.,
2006; Earle et al., 2004; Prigerson, 1991; Weeks et al., 1998; Weggel, 1999), discomfort
discussing death with health care providers (K. S. Johnson, Kuchibhatla, & Tulsky, 2008)
contributing to fewer conversations with physicians about hospice care (Huskamp et al.,
2009; McGorty & Bornstein, 2003), an assumed lack of minority hospice care providers
12
(Washington, Bickel-Swenson, & Stephens, 2008; Yancu, Farmer, & Leahman, 2010), a
lack of awareness of advance directives and/or hospice programs, misunderstanding the
severity of one’s prognosis, and other culturally-related issues including beliefs about
death and familial expectations of caring for the dying (Burrs, 1995; Gordon, 1995;
Neubauer & Hamilton, 1990; Talamantes, Lawler, & Espino, 1995). While much of the
research concerning racial/ethnic minority healthcare at EOL has focused on issues
pertaining to access to hospice, and outcomes associated with reduced access to hospice,
examination of racial/ethnic minority patient care preferences, utilization, and outcomes
following hospice enrollment is lacking.
Length of Hospice Stay
Although the Medicare Hospice Benefit provides care for patients in the last six
months of life, research indicates that enrollment often occurs late in the disease
trajectory, resulting in short hospice care episodes (Kapo, Harrold, Carroll, Rickerson, &
Casarett, 2005). While an established benchmark does not exist for the number days of
hospice care that elicit the most favorable outcomes, researchers (Byock, Forman, &
Appleton, 1996; Daugherty & Steensma, 2003; McGorty & Bornstein, 2003), clinicians
(Christakis & Iwashyna, 1998), families (Rickerson, Harrold, Kapo, Carroll, & Casarett,
2005), and government agencies (Haupt, 2003) agree that fewer days under hospice care
limit patients from receiving the full benefits of hospice. Notwithstanding, over one-third
(36%) of hospice patients continue to die within the first week of enrollment (National
Hospice and Palliative Care Organization, 2013). For a proportional illustration of the
total days of hospice care provided in the United States, see Figure 1.

13

Median trends are also particularly useful in understanding hospice LOS, as they
are not influenced by extreme cases (i.e., outliers). Data on median trends indicate that
despite the rapid growth of hospice care in the United States, LOS has been declining for
the past several years (Christakis & Escarce, 1996; Head, Ritchie, & Smoot, 2005;
National Hospice and Palliative Care Organization, 2008, 2010, 2012, 2013), suggesting
that late referral to hospice may be an increasingly important issue (Rickerson et al.,
2005). Most recently, median hospice LOS has declined from 21.1 days in 2009 to just
18.7 days in 2012 (National Hospice and Palliative Care Organization, 2010, 2013), far
below the thirty days of care that many experts believe are necessary for hospice
providers to fully address patient and family needs (Christakis & Escarce, 1996;
Christakis & Iwashyna, 2000; Han, Remsburg, McAuley, Keay, & Travis, 2007; Haupt,
0-7 Days
35.5%
8-29 Days
27.0%
30-89 Days
17.4%
90-179 Days
8.8%
180+ Days
11.5%
Figure 1. Proportion of Total Days of Hospice Care in the United States
(NHPCO, 2013)
14
2003; Huskamp et al., 2001; McCarthy, Burns, Ngo-Metzger, Davis, & Phillips, 2003;
Quill, 2007; Stillman & Syrjala, 1999; U.S. General Accounting Office, 2000).
Racial/Ethnic variation in hospice length of stay. Current understanding of
racial/ethnic differences in hospice LOS is primarily based on a limited number of
studies. Since 2000, only four identified studies have examined racial/ethnic variation in
hospice LOS (Cólon & Lyke, 2003; Hardy et al., 2012; K. S. Johnson et al., 2011;
Rhodes, Teno, & Connor, 2007). Results from this more recent body of research suggest
longer LOS for Blacks and Hispanics, compared to Whites (Cólon & Lyke, 2003; Hardy
et al., 2012; K. S. Johnson et al., 2011), with the exception of one study, a family
questionnaire, which reported no racial/ethnic variation in hospice LOS (Rhodes et al.,
2007). Other factors that have been associated with a shorter hospice LOS include gender
(Christakis & Iwashyna, 2000; Somova, Somov, Lawrence, & Frantz, 2000), primary
diagnosis (Christakis & Escarce, 1996; Miller, Weitzen, & Kinzbrunner, 2003), Medicare
enrollment (Somova et al., 2000), nursing home placement (Somova et al., 2000), hospice
agency characteristics (Carlson et al., 2009; Wachterman, Marcantonio, Davis, &
McCarthy, 2011), and referral source (i.e., referral during hospitalization and subsequent
hospice enrollment; Han et al., 2007; K. S. Johnson et al., 2011; McCarthy, Burns, Davis,
& Phillips, 2003; Miller, Kinzbrunner, Pettit, & Williams, 2003; Miller, Weitzen, et al.,
2003; Somova et al., 2000). Importantly, a complete analysis hospice LOS among a
diverse patient population has not been conducted using a nationally representative
sample in over a decade (Han et al., 2006). Given that Medicare is the largest payer of
hospice care in the United States (Lubitz & Riley, 1993), it is critical that these data be
15
re-analyzed at a national level in order to better inform policy guiding interventions and
cost-savings plans.
Advance Care Planning
Advance care planning is the process by which patients can specify, usually
through a document called an advance directive, their preferred care plans should they
become incapacitated. Often involving family, significant others, and care providers,
advance care planning provides a medium by which patients can state general EOL
preferences, indicate the level of care aggressiveness that they would like to receive, give
specific orders concerning resuscitation, and designate others as proxy decision-makers
in the event that they are no longer able to communicate their preferences. Equally
important, advance care planning offers providers with a forum to ensure that their
patients are fully aware of their prognosis and treatment options, and gain improved
understanding of their patients’ core values, beliefs, and goals of care (Detering,
Hancock, Reade, & Silvester, 2010).
Advance care planning is fundamental to the delivery of high-quality EOL care
(Lynn et al., 1997), as clarification of goals improves patient satisfaction (Kumar,
Markert, & Patel, 2011) and ensures that patients receive care that is consistent with their
wishes (Mack, Weeks, Wright, Block, & Prigerson, 2010). Furthermore, research
indicates that advance care planning reduces family stress, anxiety, and depressive
symptoms (Detering et al., 2010). However, substantial challenges must first be
overcome in order to ensure the benefits advance care planning. First, patients should
have a solid understand their illness and prognosis in order to properly represent their
wishes in an advance directive (Fischer, Tulsky, Rose, Siminoff, & Arnold, 1998;
16
Hoffmann, Zimmerman, & Tompkins, 1996; Thorevska et al., 2005). Similarly,
healthcare proxies should also understand the illness and prognosis, as well as patient
preferences, in order to adequately fulfill their role (Fagerlin & Schneider, 2004; Lo &
Steinbrook, 2004; Teno, Stevens, Spernak, & Lynn, 1998; Volicer et al., 2002). Lastly,
care providers have an obligation to ensure that care adheres to documented preferences,
as research has indicated that advance directives are not always followed due to several
reasons, including document availability, perceived conflict with family preferences, and
organizational policies (Galambos, 1998; Perkins, 2007; Volicer et al., 2002). Identified
challenges aside, advance care planning remains an essential aspect of high-quality EOL
care, with Medicare-certified hospices mandated to inform patients of their right to
complete an advance directive (Omnibus Reconciliation Act of 1990), although patients
make the final decision as to whether or not they document their preferences in a formal
advance directive.
Non-hospice-specific population. Most research examining racial/ethnic
differences in advance care planning is among non-hospice-specific populations. These
studies indicate that racial/ethnic minorities are less likely than Whites to have advance
directives (Eleazer et al., 1996; Greiner et al., 2003; K. S. Johnson, Kuchibhatla, &
Tulsky, 2008; Kwak & Haley, 2005; McKinley, Garrett, Evans, & Danis, 1996), and
often prefer more aggressive interventions, compared to Whites. In a comprehensive
review of the literature, Kwak and Haley (2005) reported that racial/ethnic minorities
often lacked knowledge about advance directives, and were less likely to support advance
directives. Furthermore, while Blacks have consistently been found to prefer more
aggressive, acute interventions (e.g., ICU treatment, resuscitation, mechanical ventilation;
17
Barnato et al., 2009; Borum et al., 2000; Duffy et al., 2006; K. S. Johnson, Kuchibhatla,
& Tulsky, 2008; Mitchell & Mitchell, 2009), Hispanics often prefer less aggressive
interventions at EOL, but few document this preference in advance directives (Kelley,
Wenger, & Sarkisian, 2010; Morrison, Zayas, Mulvihill, Baskin, & Meier, 1998). Studies
of non-hospice-specific populations also indicate that Blacks and Hispanics are less likely
than Whites to know about or designate a healthcare proxy, frequently citing beliefs that
a healthcare proxy is unnecessary, difficulty identifying a potential proxy, and discomfort
discussing such issues with their healthcare providers (Blackhall, Murphy, Frank, Michel,
& Azen, 1995; Hopp & Duffy, 2000; Kwak & Haley, 2005; Morrison et al., 1998).
Hospice enrollees. Few identified studies have examined advance care planning
among racial/ethnic minority hospice users (Jones, Moss, & Harris-Kojetin, 2011;
Resnick, Hickman, & Foster, 2012). A recent study by Resnick and colleagues (2012)
found that Black hospice patients were less likely than White hospice patients to
complete advance directives, but the exclusion of theoretically-driven predictors limit
study interpretation (e.g., age, gender, primary diagnosis, and comorbidity count were all
excluded from multivariable analyses). Concerning hospice patient care preferences
among those who document advance directives, only one article, a descriptive data brief,
has been published (Jones et al., 2011); however, its lack of analytic testing greatly
restricts the applicability of reported findings. Thus, while hospice minority care
preferences remain widely understudied, existing research suggests that Blacks and
Hispanics complete advance directives less often than Whites, Blacks prefer aggressive
interventions more often than Whites and Hispanics, and Blacks and Hispanics designate
healthcare proxies less often than Whites.
18
Emergent Care Utilization
Emergent care utilization, or unplanned emergency medical service use, is an
important quality indicator of EOL care. Although it most commonly involves emergency
department use, emergent care can also involve an emergency doctor’s visit, and
utilization of acute outpatient services (e.g., urgicenters). Unplanned utilization of
emergent care services is contrary to the philosophy of hospice because it jeopardizes
care quality and patient experiences, often at a critical point in the dying process (Olsen,
Bartlett, & Moynihan, 2011). Not surprisingly, research of non-hospice-specific
populations has found that patients with advance directives utilize less emergent care at
EOL (Degenholtz, Rhee, & Arnold, 2004; Gozalo et al., 2011; Silveira, Kim, & Langa,
2010; Teno, Gruneir, Schwartz, Nanda, & Wetle, 2007). In a 2011 study of 474,829
nursing home decedents, Gozalo and colleagues (2011) reported that decedents without
an advance directive were at an increased risk of experiencing a burdensome transition
(i.e., multiple hospitalizations or late-stage hospitalization). The same study also found
that patients who experienced a burdensome transition before death were also more likely
to experience other markers of poor EOL care quality (e.g, feeding tube, ICU visit in the
last month of life; Gozalo et al., 2011). While enrollment in hospice has been found to
decrease hospitalization rates (Hughes et al., 1997; McCusker & Stoddard, 1987; Miller,
Gozalo, & Mor, 2001), some hospice users still seek acute care following hospice
enrollment (Legler, Bradley, & Carlson, 2011; Olsen et al., 2011), and risk compromising
hospice benefits, namely quality of death. In a 2007 study of patients enrolled in the
Mayo Hospice Program, Olsen and colleagues (2011) found that the majority of
19
hospitalized hospice patients received moderately-intense care, escalating costs and
increasing the likelihood of dying in a hospital.
Racial/Ethnic differences in emergent care utilization at end of life. Most
research examining racial/ethnic differences in emergent care utilization does not
differentiate between hospice and non-hospice users, leaving significant gaps in the
understanding of care use patterns among hospice enrollees. Regardless, these studies
indicate that Blacks and Hispanics often utilize emergent EOL care more often than
Whites (Barnato et al., 2007; Goldstein et al., 2010; Gozalo et al., 2011; Hanchate et al.,
2009; A. K. Smith, Earle, et al., 2009). A 2009 study (A. K. Smith, Earle, et al.) of over
40,000 Medicare beneficiaries with advanced stage cancer found that Blacks were more
likely than Whites to be 1) hospitalized two or more times, 2) hospitalized for two or
more weeks, and 3) admitted to the intensive care unit (ICU) in the last month of life.
Similarly, Hanchate and colleagues (2009) reported that among 158,780 Medicare
decedents, Blacks and Hispanics were more likely than Whites to utilize intensive life-
sustaining interventions at EOL, resulting in significantly higher costs. Increased
utilization of more aggressive interventions at EOL also places racial/ethnic minorities at
a greater risk of multiple poor EOL outcomes, including reduced care quality (Casarett et
al., 2005; Earle et al., 2008; Patrick, Curtis, Engelberg, Nielsen, & McCown, 2003;
Steinhauser et al., 2000), lower satisfaction with care (Baker et al., 2000; Billings &
Kolton, 1999; Teno et al., 2004), decreased QOL (Mazanec et al., 2010; Wright et al.,
2008; Zhang et al., 2009), debilitating procedures that do little to prolong life (Engle,
1998; Mezey, Dubler, Mitty, & Brody, 2002), and increased healthcare costs (Hanchate
20
et al., 2009; Hogan et al., 2001; Levinsky et al., 2001; Shugarman et al., 2004; Wennberg,
Fisher, Goodman, & Skinner, 2008; Wennberg et al., 2004).
Hospice enrollees. Few studies have examined racial/ethnic variation in emergent
care utilization following hospice enrollment. Of this body of research, findings suggest
that hospice-electing racial/ethnic minorities also utilize emergent care at higher rates
than Whites (Cintron et al., 2003; Loggers et al., 2013; Schonwetter et al., 2008; Unroe et
al., 2012). In a 2012 study conducted by Unroe and colleagues (2012), non-White
Medicare beneficiaries were more likely than White beneficiaries to visit the emergency
department, be hospitalized, and be admitted into an ICU. Moreover, the authors also
reported that among hospice decedents, racial/ethnic minorities remained significantly
more likely than Whites to utilize emergent care at higher rates. Similarly, a review of
Medicare records between 1988-1998 found that Blacks were more likely than Whites to
seek hospitalization following hospice enrollment (Cintron et al., 2003). Thus, while
emergent care utilization among hospice-electing racial/ethnic minorities remains widely
understudied, existing research suggests that Blacks and Hispanics are at an increased
risk of utilizing emergent care following hospice enrollment.
Site of Death
Numerous studies have demonstrated that the majority of seriously ill patients
prefer to die at home (Higginson & Sen-Gupta, 2000; Tang, 2003; Townsend et al.,
1990); however, most do not achieve this aim and instead, die in other settings (Burge,
Lawson, Johnston, & Cummings, 2003; Higginson, Astin, & Dolan, 1998). Current
statistics indicate that while approximately seven out of ten Americans prefer to die in
their home (Cloud, 2000), only 25% do so, with the majority dying in hospitals (45%),
21
and nursing home/long-term care facilities (22%; CDC, 2008). Research investigating
in-home death has established links with fewer medical complications (Leff et al., 2005),
reduced costs (Leff et al., 2005), less physical and emotional distress (Wright et al.,
2010), improved QOL (Wright et al., 2010), greater satisfaction with care (Leff et al.,
2006), and lower levels of complicated grief and other bereavement-related distress for
surviving family members (Kramer, Kavanaugh, Trentham-Dietz, Walsh, & Yonker,
2010; Wright et al., 2010). Conversely, dying in a hospital has been associated with
increased pain and symptoms (Nelson et al., 2001), reduced care quality (Meier, 2003),
poorer patient QOL (Wright et al., 2010), and increased psychological disorders among
survivors (Wright et al., 2010).
Racial/ethnic variation in site of death. Studies of non-hospice-specific
populations have consistently reported an increased likelihood among racial/ethnic
minorities to die acute hospital settings more often than Whites (Gruneir et al., 2007;
Hanchate et al., 2009; Hansen, Tolle, & Martin, 2002; National Center for Health
Statistics, 2011; A. K. Smith, Earle, et al., 2009; Weitzen, Teno, Fennell, & Mor, 2003;
Zheng, Mukamel, Caprio, Cai, & Temkin-Greener, 2011). A 2009 study of terminally ill
Medicare beneficiaries (A. K. Smith, Earle, et al.) found that non-Whites were
significantly more likely than Whites to die in a hospital setting. While this increased
likelihood may be due in part to a greater preference for in-hospital death among
minorities (Neubauer & Hamilton, 1990), research also suggests that other factors such as
geographic location, socioeconomic status, and morbidity differences may also be in
operation (Hanchate et al., 2009).
22
Hospice enrollees. In-home death is both supported and fostered by hospice (K.
S. Johnson et al., 2005). Research on SOD among hospice patients has reported that
hospice users are more likely than non-hospice users to die in their homes rather than in
an acute setting (Moinpour & Polissar, 1989; Pritchard et al., 1998). Among those
enrolled in hospice in 2010, 67% died in their home and 11% died in an acute setting
(National Hospice and Palliative Care Organization, 2012), compared to approximately
24% dying in-home and 40% dying in acute settings among non-hospice users over 65
(National Center for Health Statistics, 2011). Hospice patients who die in their homes
have lower rates of unmet needs, fewer concerns about being treated with respect, higher
levels of emotional support, and greater satisfaction with care, compared to non-hospice
users (Teno et al., 2004). Since the introduction of the Medicare Hospice Benefit in 1982,
only one identified study has examined racial/ethnic differences in SOD among hospice
patients (K. S. Johnson et al., 2005). Findings indicated that Blacks and Hispanics were
more likely than Whites to die in an inpatient hospice setting; however, nearly three
quarters of the sample utilized inpatient hospice services, and thus the sample represented
only a small subsample (22%; National Hospice and Palliative Care Organization, 2012)
of the larger hospice population. The study also reported an increased likelihood among
Hispanics to die in a home versus an inpatient hospice setting; however, the study did not
differentiate between death in an acute hospital setting and death in a hospital-based
hospice unit, greatly restricting finding implications. Thus, while hospice minority SOD
remains widely understudied, EOL research in general suggests that Blacks and
Hispanics may be more likely than Whites to die in an acute setting following hospice
enrollment (Gruneir et al., 2007; Hanchate et al., 2009; Hansen et al., 2002; National
23
Center for Health Statistics, 2011; A. K. Smith, Earle, et al., 2009; Weitzen et al., 2003;
Zheng et al., 2011).
Determinants of Racial/Ethnic Variation in Hospice Use
In addition to identifying existing racial/ethnic disparities in the U.S. healthcare
system, research has also focused on determining the contributing mechanisms to
enrollment disparities. Key patient-level factors that have been identified include patient
1) cultural beliefs, 2) religious values, 3) knowledge and education, and 4)
communication and trust.
Cultural beliefs. For the purposes of this manuscript, culture is defined as the
complete cultural background (including previous and current cultural identities) that
shape one’s lens through which they see the world. As such, while ethnicity is an inherent
component of one’s cultural beliefs, culture is proposed to encompass the larger historical
cultural landscape of each individual’s unique identity. Under this framework, 1) race is
proposed to account for the biological, genetically-determined, characteristics or traits of
an individual, 2) ethnicity is proposed to encompass the non-biological factors, largely
representative of one’s culture of origin, that differentiate people within racial groups
(Egede, 2006), and 3) cultural beliefs are proposed to be the product of the full cultural
history of individuals (i.e., ethnic background, previous cultures, current culture).
Numerous studies have identified cultural factors that contribute to racial/ethnic
differences in healthcare utilization at EOL (Blackhall et al., 1999; Del Gaudio et al.,
2013; Searight & Gafford, 2005; A. K. Smith, Sudore, & Pérez-Stable, 2009). First,
fundamental differences in beliefs about truth-telling and decision-making appear to
contribute significantly to racial/ethnic variation in advance care planning. While the U.S.
24
healthcare system emphasizes patient autonomy, some cultures value more collectivist
approaches, often looking to the family to share in the care decision-making process, or
in some cases even make care decisions on behalf of patients while concealing diagnoses
from patients out of respect (Searight & Gafford, 2005). Research has found that
compared to African American and White elderly patients, half as many Korean-
Americans and Mexican-Americans report wanting to be told the truth about their
diagnosis, with many Korean-Americans and Mexican-Americans preferring the family,
not the patient, to be the key decision-maker concerning EOL choices (Blackhall et al.,
1999; Blackhall et al., 1995). These cultural beliefs contrast sharply with the prevailing
U.S. preference for patient autonomy, and present significant challenges with advance
care planning for racial/ethnic minorities. Additionally, substantial within-culture
variation, often resulting from acculturation to the United States, further contributes the
complexity of the issue.
In addition to cultural differences in preferences for truth-telling and decision-
making, culture-specific beliefs, such as the Latino values of Machismo, Fatalismo, and
Marianismo, also shape the attitudes and care decisions of many patients (Del Gaudio et
al., 2013; A. K. Smith, Sudore, et al., 2009). For example, it is conceivable that Latino
patients valuing Fatalismo, or the belief that one’s fate is predetermined, may be less
likely to engage in the care planning process, and ultimately receive more aggressive life-
prolonging interventions. Similarly, men valuing machismo, or the belief that men are the
primary protectors and decision-makers of the household, may be more likely to perceive
election of hospice (and the forgoing of curative efforts) as a sign of weakness or defeat.
Moreover, even for those Latino men who do elect hospice, there still may be a latent
25
cultural propensity to not show weakness to family or care providers. As a result, care
decisions and utilization patterns of these men may differ from other non-Latino groups.
Although still in its infancy, this growing body of research suggests that culture-specific
values significantly contribute to racial/ethnic variation in healthcare decision-making,
and that these differences are magnified at EOL (Del Gaudio et al., 2013; A. K. Smith,
Sudore, et al., 2009).
Religious values. For the purposes of this manuscript, religion is defined as either
individually-identified (e.g., patients who identify themselves as such) or the
participation in formal religious activities (e.g., attending church, prayer).
Patient religious beliefs have a strong influence on EOL decision-making, with
more religious individuals often preferring more aggressive interventions at EOL
(Balboni et al., 2007; Phelps et al., 2009). In a rare study examining the relationship
between religion and views of patient autonomy, Blackhall and colleagues (1995)
reported that Protestants were more likely to believe that patients should be told the truth
about a terminal prognosis and that patients should be the primary decision-maker for
their EOL care choices. Conversely, Buddhist and Jewish participants did not support
telling patients the truth about a terminal prognosis (Blackhall et al., 1995). Interestingly,
patient race/ethnicity appears to partially moderate the relationship between religiosity
and EOL care planning and behaviors. That is, among racial/ethnic minorities, high
religiosity has been associated with a preference against advance care planning (K. S.
Johnson, Kuchibhatla, & Tulsky, 2008) and decisions not to enroll in hospice, with
highly religious African Americans maintaining more negative views toward hospice
(Kagawa-Singer & Blackhall, 2001; Reese, Ahern, Nair, O'Faire, & Warren, 1999).
26
Strong religious beliefs among non-terminal African American men have also been
associated with preferences for more aggressive interventions, regardless of prognosis, as
a means of fighting for god’s gift of life (Blocker et al., 2006). A recent study found that
many Blacks choose not to engage in advance care because “…a higher power controls
the nature and timing of death” (Carr, 2011, p. 15). Research suggests there is a
preference among some Latinos to endure suffering at EOL, wherein suffering is viewed
as a necessary and even fundamental aspect of dying (Krause & Bastida, 2011). When
understood in the context of religion/spirituality, these Latinos conceive suffering as a
means to atone for one’s sins, and ultimately deepen one’s faith in god (Braun, Beyth,
Ford, & McCullough, 2008; A. K. Smith, Sudore, et al., 2009). Collectively, this body of
literature suggests that the influence of religiosity on healthcare decision-making differs
across racial/ethnic groups and may contribute to racial/ethnic variation in care decisions
and utilization patterns at EOL.
Knowledge and education. Racial/Ethnic differences in patient knowledge and
education are also associated with variation in healthcare utilization patterns. Throughout
the past decade, multiple studies have demonstrated the effectiveness of brief educational
interventions in shifting racial/ethnic minority patient knowledge and understanding of
healthcare options (Casarett et al., 2005; Chung, Essex, & Samson, 2009; Enguidanos,
Kogan, Lorenz, & Taylor, 2011; Volandes, Ariza, Abbo, & Paasche-Orlow, 2008;
Volandes, Barry, Chang, & Paasche-Orlow, 2010; Volandes et al., 2007; Volandes,
Paasche-Orlow, et al., 2008). In a study of Spanish-speaking Latino patients, Volandes
and colleagues (2008) found that after viewing a two-minute video of an individual with
advanced dementia, preference for comfort care nearly doubled, and desire for life-
27
prolonging care reduced from 40% of patients to just 8%. Similarly, African Americans
have demonstrated improved knowledge of hospice and an increased willingness to enroll
in hospice after reviewing a targeted educational brochure describing positive patient
experiences of hospice care (Enguidanos et al., 2011). Improved racial/ethnic minority
caregiver knowledge has also been associated with patient willingness to engage in
advance care planning, preferences for comfort care over aggressive care at EOL,
enrollment in hospice earlier in the disease trajectory, and more active caregiver
involvement in monitoring the quality of hospice services (Chung et al., 2009).
Interestingly, in a study of over 30,000 dual-eligible nursing home residents, Kwak and
colleagues (2008) found that while formal years of education was associated with patient
willingness to enroll in hospice, this relationship was moderated by race/ethnicity. That
is, for Whites, formal education appears to affect patient willingness to enroll in hospice;
however, this relationship does not appear to be true for Blacks (i.e., formal education has
no association; Kwak et al., 2008).
Communication and trust. While some minorities may prefer higher levels of
aggressive care (Barnato et al., 2009; Borum et al., 2000; Duffy et al., 2006), others
report challenges communicating comfort care preferences to providers (Kelley et al.,
2010; Perkins, Geppert, Gonzales, Cortez, & Hazuda, 2002), potentially resulting in care
that is not consistent with the patient’s wishes (Barnato et al., 2009). Studies of non-
terminal patients have reported that African Americans are less likely to be actively
involved in the decision-making process and more likely to experience difficulty in
physician-patient communication (Cene, Roter, Carson, Miller, & Cooper, 2009; Cooper-
Patrick et al., 1999; Ghods et al., 2008; R. L. Johnson, Roter, Powe, & Cooper, 2004).
28
Huskamp and colleagues (2009) found that healthcare providers were less likely to
discuss hospice care with seriously ill Black and Hispanic patients, compared to White
patients. Similarly, in a study of 981 physicians, Modi and colleagues (2007) found Black
physicians were nearly twice as likely to recommend percutaneous endoscopic
gastrostomy (PEG) tube placement to their Black patients, compared to White patients,
suggesting that physician-patient race concordance is also a factor affecting
communication between patients and healthcare providers at EOL.
Research has demonstrated that historical prejudices and discriminatory acts, such
as the Tuskegee study and the U.S. sterilization campaign targeting Puerto Ricans, have
contributed to racial/ethnic minority distrust of medical professionals as well as the larger
healthcare system (Braunstein, Sherber, Schulman, Ding, & Powe, 2008; Gamble, 1997;
Harris, Gorelick, Samuels, & Bempong, 1996; Payne, 2001). In a comprehensive focus
group study of 73 Arab Muslim, Arab Christian, Hispanic, Black, and White older adults,
Duffy and colleagues (2006) found that Blacks often cited past inequities (e.g., medical
mistreatment, provider abandonment) as a barrier to hospice care. In a survey of 236
physicians, 88% of U.S.-born African American physicians, compared to 35% of White
physicians, believed that the Tuskegee study has negatively impacted medical decision-
making among African Americans (M. P. Wallace et al., 2007). These findings suggest
that even at the provider-level, the ramification of these past events has had a significant
impact on racial/ethnic minorities. Moreover, ongoing societal policies (e.g., Medicaid’s
race-neutral long-term care policies) that limit racial/ethnic minority access to care have
also likely played a significant role in promoting continued distrust, with African
Americans significantly more likely than Whites to report higher levels of racism and
29
distrust of the healthcare system (Fowler-Brown, Ashkin, Corbie-Smith, Thaker, &
Pathman, 2006). Overall, this body of research indicates that race/ethnicity-specific
factors related to communication and healthcare provider trust further contribute to
racial/ethnic variation in healthcare care preferences and utilization.
Importance of the Topic
Demographic projections indicate that the U.S. population will age significantly
throughout the next fifty years (Lutz, Sanderson, & Scherbov, 2008), with a sharp
increase in the proportion of racial/ethnic minorities (Ortman & Guarneri, 2009). As the
population shifts, a widely diverse group of older adults will require high-quality EOL
care. Since the landmark SUPPORT study found that dying patients often receive
inadequate or overly aggressive care that is inconsistent with preferences (Connors et al.,
1995), patients and their families have continued to report reduced participation in the
decision-making process (Azoulay et al., 2004; Selman et al., 2007; White, Braddock,
Bereknyei, & Curtis, 2007; Winzelberg, Hanson, & Tulsky, 2005), insufficient emotional
support (Kunik et al., 2005; Teno et al., 2004; Thornton, Pham, Engelberg, Jackson, &
Curtis, 2009; Wenrich et al., 2003), low satisfaction with care (Baker et al., 2000;
Billings & Kolton, 1999; Teno et al., 2004), poor QOL (Blinderman, Homel, Billings,
Portenoy, & Tennstedt, 2008; Blinderman, Homel, Billings, Tennstedt, & Portenoy,
2009; Lackan, Eschbach, Stimpson, Freeman, & Goodwin, 2009; Wright et al., 2008),
increased aggressive care (Mack, Weeks, et al., 2010; Parr et al., 2010; Teno, Fisher,
Hamel, Coppola, & Dawson, 2002), in-hospital death despite patient preferences (Lackan
et al., 2009; Mularski et al., 2009), poorly treated pain and dyspnea (Goodlin,
Winzelberg, Teno, Whedon, & Lynn, 1998; Lynn et al., 1997; Teno et al., 2004; Tolle,
30
Tilden, Hickman, & Rosenfeld, 2000), and inadequate bereavement support (Billings &
Kolton, 1999; Wright et al., 2008). Currently, Medicare spending during the last year of
life accounts for nearly a quarter of total annual expenditures (Hogan et al., 2001).
Research suggests that as much of 78% of the costs incurred during the last year of life
result from care received during the final thirty days (Yu, 2008), with racial/ethnic
minority utilization of aggressive interventions and underuse of hospice serving as
significant contributors (Hanchate et al., 2009; Hogan et al., 2001; Levinsky et al., 2001;
Shugarman et al., 2004; Teno et al., 2002; Wennberg et al., 2008; Wennberg et al., 2004).
Existing research focuses almost exclusively on increasing racial/ethnic minority access
to hospice care; however, it remains unclear if racial/ethnic differences in care
preferences, utilization, and outcomes persist following hospice enrollment. With
increasing national focus on improving care quality and cost-containment ("The Patient
Protection and Affordable Care Act, Pub. L. No. 111-148," 2010) as well as eliminating
healthcare disparities (USDHHS, 2012), research investigating racial/ethnic minority
healthcare at EOL, a time when such disparities are particularly significant, is critical.

31
Chapter 3: Research Methodology
In this chapter, the research questions and hypotheses are proposed, and the
overall research approach is described.
Research Questions and Hypotheses
Research questions. The study was guided by the following key research
questions:
1. Is there a difference between White, Black, and Hispanic hospice enrollees in
the completion of advance care plans and care choices?
2. Does race/ethnicity influence emergent care utilization among hospice
enrollees? Does engaging in advance care planning affect this relationship?
3. Among patients who die under the care of hospice, how does race/ethnicity
influence hospice length of stay and site of death?
Hypotheses. The primary purpose of this investigation was to provide an in-depth
examination into hospice care preferences, utilization, and outcomes among racial/ethnic
minorities in the United States. In testing for differences in care utilization and outcomes,
patient care preferences and Medicaid enrollment (as a proxy for socioeconomic status)
were included to account for potential explanatory variability between racial/ethnic
groups. Hypotheses in this study tested for racial/ethnic variation following hospice
enrollment in advance care planning decisions, emergent care utilization, hospice length
of stay (LOS), and site of death (SOD). Tested hypotheses are as follows.
1. White hospice patients will be more likely than Black and Hispanic hospice
patients to have a documented advance directive (H1).
32
a. Among those with documented advance directives, White and
Hispanic hospice patients will be more likely than Black hospice
patients to document a do not resuscitate order (H1A).
b. Among those with documented advance directives, White and
Hispanic hospice patients will be more likely than Black hospice
patients to document a healthcare proxy (H1B).
2. Black and Hispanic hospice patients will be more likely than White hospice
patients to utilize emergent care (i.e., unplanned emergency medical care;
H2).
a. Hospice patients without documented advance care plans (i.e., advance
directive, do not resuscitate order) will be more likely than those with
documented advance care plans to utilize emergent care (H2A).
3. Hospice length of stay will be longer for Black and Hispanic decedents,
compared to White decedents (H3).
4. White hospice decedents will be more likely than Black and Hispanic hospice
decedents to die in a home-like setting (H4).
a. Black and Hispanic hospice decedents will be more likely than White
hospice decedents to die in a hospital (H4A).
Research Design
This study is a retrospective analysis of clinical and service use outcomes among
Medicare-enrolled U.S. hospice patients using secondary data from the 2007 wave of the
National Home Health and Hospice Care Survey (NNHCS). Using a complex two-stage
sampling design, NHHCS 2007 data are representative of the U.S. hospice patient
33
population. Although six previous waves of the NHHCS were conducted from 1992-
2000, NHHCS 2007 represents a significant shift in the sampling frame, study design,
and data collected within the NHHCS survey family. Accordingly, NHHCS 2007 is the
first study in its family to provide detailed data on racial/ethnic minority hospice
enrollees, and much of its patient-level data are not available in previous waves.
Data Sources
NHHCS 2007 data were collected between August 2007 and February 2008 using
a two-stage probability sampling design of agencies providing home health and/or
hospice care in the United States. The sampling frame of 15,488 U.S. home health and
hospice agencies was derived from three sources: 1) the National Hospice and Palliative
Care Organization (NHPCO), 2) Verispan, L.L.C., and 3) the Centers for Medicare &
Medicaid Services Provider of Services file.
Agency selection. In the first sampling stage of NHHCS 2007, all agencies in the
sampling frame were grouped into strata based on agency-type (i.e., home health care
only, hospice care only, both home health care and hospice care) and metropolitan
statistical area (MSA; i.e., metropolitan, micropolitan, neither metropolitan nor
micropolitan), and then sorted by census region (i.e., Northeast, Midwest, South, West),
ownership type (i.e., proprietary, nonprofit, government, unknown), certification status
(i.e., Medicare, Medicaid, both Medicare and Medicaid), state, county, and ZIP code.
Following agency stratification and sorting, 1,545 agencies were randomly selected with
probability proportional to agency size (i.e., number of employees). Of these, 84 were
excluded for being out of survey scope, and 425 refused to participate, yielding a final
agency sample of 1,036.
34
Patient selection. In the second sampling stage, interviewers visited each of the
selected agencies and collected a census of all home health and/or hospice patients. The
patient sampling pool consisted of 1) all home health patients serviced by the agency the
day before the interview, and 2) all hospice discharges during a 3-month period four
months before the interview. Patients discharged more than once during the 3-month
period of data collection were treated as distinct episodes of care (i.e., separated hospice
discharges). As such, hospice patients could be represented more than once in NHHCS
2007 data. Following agency censuses, a final sample of ten patients from each agency
was randomly selected using the computer-assisted personal interviewing (CAPI) system.
If agencies serviced less than ten patients, all patients were included, and if an agency
serviced both home health and hospice patients, two samples of five patients were
included (i.e., five home health and five hospice patients). In all, 4,733 hospice
discharges and 4,683 home health patients were selected for participation in the study.
Data collection. Patient-level and agency-level data were collected from in-
person interviews with agency directors, and self-administered staffing questionnaires.
Study personnel first contacted agencies to setup in-person data collection visits. Next,
agency directors were asked to complete a paper questionnaire describing agency
characteristics two-weeks prior to the scheduled in-person visit. Finally, on the day of the
scheduled in-person visit, trained interviewers collected additional data using the CAPI
system on agency characteristics, as well as data on patient characteristics and care
received. Several steps (e.g., audio recording, in-person observation, debriefing calls)
were utilized to ensure data were collected consistently and reliably across agencies.
35
Data use. The data utilized in this study were cleaned by the Inter-university
Consortium for Political and Social Research (USDHHS, 2010). Data were accessed on
March 16, 2013, and subsequently maintained by the principal investigator.
Human Subjects Protection
An application for Exempt Review was submitted to the University of Southern
California’s Institutional Review Board on June 22, 2011. Study approval was obtained
on June 29, 2011.
Sampling Plan
The study sample consisted of 3,661 White, Black, and Hispanic Medicare-
enrolled hospice patients 65 years of age and older at the time of death/discharge.
Patients were drawn from nine strata and 657 hospice-providing agencies. When
weighted for national representation, the data represent 788,872 hospice patients in the
United States. Patients from other racial/ethnic groups were excluded due to low sample
sizes (unweighted n = 53; Pacific Islanders, Native Americans, Asians). Since Medicare
covers 84% of hospice care in the United States, this study examined the discrete
Medicare population to maximize research translatability and add national policy
implications to the analyses.
Operationalization of the Variables and Measures
All patient-level data were collected from hospice agency administrators during
in-person interviews.
Demographic and health indicator variables. Patient demographic and health
indicator variables included age (continuous), gender (binary), race/ethnicity (White,
Black, Hispanic), marital status (married/living with partner, widowed,
36
divorced/separated, never married), primary diagnosis (Cancer, Congestive Heart
Failure/Heart Disease, Lung Disease, neurological diseases, other), and the total number
of comorbidities (continuous). In care utilization (i.e., emergent care utilization) and
outcome models (i.e., LOS, SOD), Medicaid enrollment (binary) was included as a proxy
for patient/decedent socioeconomic status (Bach, Guadagnoli, Schrag, Schussler, &
Warren, 2002; Gross, Filardo, Mayne, & Krumholz, 2005).
Outcome variables included 1) advance directive completion, 2) do not resuscitate
(DNR) order election, 3) healthcare proxy designation, 4) emergent care utilization, 5)
hospice LOS, and 6) SOD. Outcome data coded as missing were excluded from all
bivariate and multivariable analyses.
Advance directive completion. Completion of an advance directive was
measured using agency binary data on the documented preference of any of the following
pre-defined advance directive categories: 1) living will, 2) DNR order, 3) do not
hospitalize order, 4) preferences for comfort measures only, 5) feeding restrictions, 6)
medication restrictions, 7) durable power of attorney, 8) healthcare proxy/surrogate, and
9) other treatment restrictions. Each of the above pre-defined categories were collected
separately and recoded into the following binary variable: 1) patients with any pre-
defined documented preference (positive binary indicator outcome), and 2) patients with
no pre-defined documented preferences (negative binary indicator outcome).
Do not resuscitate order election. Preference against resuscitation was measured
using agency binary data on the documented election of a DNR order. Responses were
recorded into two categories: 1) patient has elected to document a DNR order (positive
binary indicator outcome), and 2) patient has not elected to document a DNR order
37
(negative binary indicator outcome). Only patients with a documented advance directive
(i.e., valid advance directive completion response) were included in bivariate and
multivariable analyses of DNR election.
Healthcare proxy designation. Healthcare proxy designation was measured
using agency binary data on the documented designation of 1) a durable power of
attorney (DPOA), or 2) healthcare proxy or surrogate decision-maker. Using data from
these two recorded variables, responses were recoded into two categories: 1) patient has
elected to designate a healthcare proxy (positive binary indicator outcome), and 2) patient
has not elected to designate a healthcare proxy (negative binary indicator outcome).
Given the duplicative nature of a designated DPOA and healthcare proxy (Ouslander,
Tymchuk, & Rahbar, 1989), responses to both of these originally-recorded variables were
treated as equivalent during recoding. Only patients with a documented advance directive
(i.e., valid advance directive completion response) were included in bivariate and
multivariable analyses of healthcare proxy designation.
Emergent care utilization. Emergent care utilization was measured using agency
binary data on patient unplanned emergency medical care use while enrolled in hospice.
Emergent care was selected over other potential outcomes (e.g., emergency department
use) to model more comprehensively any unplanned emergency service use by hospice
patients. Data were originally recorded as three separate binary outcomes: 1) hospital
emergency department use (including 23-hour holding), 2) doctor’s office emergency
visit/house call, and 3) outpatient department/clinic use (including urgicenter sites). For
each of the above binary outcomes, positive responses were only recorded if care
utilization occurred during the 60 days prior to data collection, per survey data collection
38
guidelines. Using the above recorded variables, data were recoded into the following
emergent care utilization binary outcome: 1) patient utilized any of the three originally
recorded unplanned emergency care outcomes (positive binary indicator outcome), and 2)
patient did not utilize any of the originally recorded unplanned emergency care outcomes
(negative binary indicator outcome).
Hospice length of stay. Given previous research documenting racial/ethnic
variation in hospice revocation (K. S. Johnson, Kuchibhatla, Tanis, et al., 2008), hospice
LOS was examined using the subsample of patients who died under the care of hospice.
This was done to further understanding of potential racial/ethnic differences in the total
length of the care episode specifically among those who elected to continue hospice
services throughout the dying process. Using agency documentation, hospice LOS was
calculated as the difference between patient discharge (i.e., death) and enrollment dates.
As such, the variable level was continuous.
Site of death. Patient SOD was measured using agency data on where the patient
was staying on the last day of care. Pre-defined locations included: 1) the agency’s
inpatient or residential facility, 2) private home or apartment, 3) residential care place, 4)
nursing home, 5) skilled nursing facility, 5) hospital, and 6) other. Residential care place
was operationalized as an assisted living facility, a board and care home, or a life
care/continuing care retirement community. Although data on the final site of care are
available for all discharged patients, only patients who died under the care of hospice
were included in models of hospice SOD, and thus the originally recorded final site of
care was re-operationalized as SOD.
39
Two separate outcomes were tested to determine racial/ethnic variation in death in
a 1) home-like setting, and 2) hospital. Death in a home-like setting was analyzed by
recoding the above pre-defined locations into two categories: 1) death in a private home
or apartment or residential care place (positive binary indicator outcome), and 2) death in
all other locations (negative binary indicator outcome). Death in a hospital was examined
by recoding the above pre-defined locations to 1) death in a hospital (positive binary
indicator outcome), and 2) death in all other locations (negative binary indicator
outcome).
Conceptual Model
Andersen’s Behavior Model of Health Services Use. The Behavior Model of
Health Services Use proposed by Andersen (1995) provided a solid initial framework to
guide the study. Initially developed to explain health service use, Andersen’s model
(Figure 2) posits that patient health outcomes result from a dynamic interaction between
population characteristics, health behavior, and the larger environment. Specifically,
Andersen proposes that aspects of the greater healthcare system (e.g., policy, practices,
care settings) influence patient predisposing characteristics (e.g., demographics, social
structure, health beliefs), enabling resources (e.g., personal, family, medical insurance),
and overall need. These in turn affect patient health behaviors (i.e., utilization of health
services), which ultimately influence health outcomes. The model also includes feedback
loops to represent the dynamic relationship between outcomes and patient predisposing
factors, overall need, and health behaviors. Prior research has drawn from Andersen’s
model to examine and explain racial/ethnic variation in 1) healthcare utilization in the
larger healthcare system (LaVeist, Nuru-Jeter, & Jones, 2003), 2) overall care at end of
40
life (EOL; Bradley et al., 2002; S. P. Wallace, Levy-Storms, Kington, & Andersen,
1998), and 3) hospice access and utilization (Conner, 2012; Miller, Kinzbrunner, et al.,
2003).

Revised model. This study proposes an adaptation of Andersen’s model (Figure
3) to explain within-hospice racial/ethnic variation in care planning, utilization, and
outcomes. Specifically, hospice patient healthcare behaviors (i.e., advance directive
completion, resuscitation preferences, healthcare proxy designation, emergent care
utilization) and outcomes (i.e., LOS, SOD) are proposed to be a function of population
characteristics (predisposing characteristics, enabling resources, healthcare needs) and the
greater health environment (i.e., care agency). In the proposed model, patient
predisposing characteristics include age, gender, and race/ethnicity (social structure).
Furthermore, to emphasize the unique sociocultural contribution of patient race/ethnicity
Healthcare System!
!
External Environment
Predisposing
Characteristics
Enabling Resources
Need
Personal
Health Practices!
!
Use of Health
Services
Perceived
Health Status!
!
Evaluated Health
Status!
!
Consumer
Satisfaction
Environment
Outcomes
Figure 2. Andersen’s Behavior Model of Health Services Use (Andersen, 1995)
Population
Characteristics
Healthcare
Behavior
41
on health behaviors and outcomes, it is also included in the model as an enabling
resource. In doing so, the model more accurately explains the multifaceted nature of
race/ethnicity (beyond social structure) on the identified outcomes of interest. Marital
status and socioeconomic status are also operationalized as enabling resources, and
patient healthcare needs are explained using data on primary diagnosis, and number of
chronic conditions (i.e., comorbidity count). Agency random effects on endogenous
factors are denoted by blue double-arrow feedback loops. Lastly, race/ethnicity is noted
in red text to identify it as the key predictor variable of interest.

Care Agency Predisposing
Characteristics!
!
Demographics!
• Age!
• Gender!
!
Social Structure!
• Race/Ethnicity
Enabling Resources!
!
Race/Ethnicity!
!
Personal/Family!
• Marital Status!
• Socioeconomic
Status
Healthcare Needs!
!
Health Status!
• Primary Diagnosis!
• Number of Chronic
Conditions
Advance Care
Planning!
!
Advance Directives!
• Resuscitation
Preferences!
• Healthcare Proxy
Designation!
!
Emergent Care
Utilization
Length of
Hospice Stay!
!
Site of Death
Environment
Outcomes
Figure 3. Revised Conceptual Model (Adapted from Andersen, 1995)
Population
Characteristics
Healthcare
Behaviors
42
To emphasize the importance of EOL care choices and behaviors, an important
distinction is made between the original Andersen model and the proposed model: patient
health behaviors in the proposed model (e.g., advance care planning, emergent care
utilization) are considered both an outcome as well as a factor contributing to other
outcomes (e.g., LOS, SOD). This relationship between EOL health behaviors and
outcomes is represented in the proposed model by a bolded arrow. To represent the EOL
care philosophy of patient self-determination (Center to Advance Palliative Care, 2007),
advance care planning is proposed to directly affect within-hospice outcomes in the
adapted model. Building on previous research, this study investigated the independent
effect of patient race/ethnicity on advance care planning, emergent care utilization,
hospice LOS, and SOD, following adjustment for the other variables identified in the
adapted model.
Analysis Plan
Due to the complex sampling design of NHHCS 2007, additional steps were taken
to ensure accurate data representation in study analyses. Sample recruitment employed a
two-stage probability sampling design in which 1) agencies were first randomly selected
from the U.S. home health and hospice provider network, and 2) home health and hospice
patients were randomly selected from within the sampled agencies. The National Center
for Health Statistics (NCHS) recommends (Dwyer, Harris-Kojetin, Branden, & Shimizu,
2010) that data analyses include appropriate survey modeling techniques to account for
the clustered nature of the data, and that sampling weights be applied for national
representativeness. Added effort was made throughout the study to ensure that all
analyses accounted for sampling design, survey weights were correctly applied, and
43
results from hypothesis tests incorporated robust standard errors. Descriptive analyses of
all study variables were initially conducted. Fixation indices (i.e., F-statistics) estimated
from survey-adjusted chi-square tests (discrete variables) and one-way analysis of
variance (ANOVA) tests (continuous variables) were employed to investigate bivariate
relationships between non-missing outcome variables and patient demographic variables.
To improve variable transparency in bivariate analyses, continuous variables (i.e., age,
comorbidity count) were cut into balanced, mutually-exclusive groups based on the
number of valid responses.
Multivariable logistic models were tested using unconditional (i.e., random
second-level intercept) multilevel regression models. Unconditional multilevel regression
was selected to correctly model the observation dependence of patient-level data within
the hospice agencies. The model assumes fixed effects on the hypothesized relationships,
and random effects on agency intercepts. The approach can be illustrated as a two-level
procedure. The first level consists of patient-level regressions that test the hypothesized
relationships within agencies. The second level estimates agency-level regressions that
account for between-agency variation in patient-level outcomes. Given the exclusion of
level-two predictors in analyses, this model is considered an intercept-only approach
where possible interactions between variables in different levels are not modeled. An
unreduced example of the multivariable models employed to test study hypotheses,
accounting for full patient-level predictor variation between agencies, is illustrated in
Figure 4.

44

In this illustrated model formula, the level 1 regression resembles a standard
regression model where the link function of patient-level outcome y
ij
is related to a linear
combination of an estimated intercept (ß
0j
), patient-level predictors x
1ij
, x
2ij
, x
3ij
, x
4ij
, x
5ij
,
x
6ij
, x
7ij
, x
8ij
, x
9ij
, x
10ij,
, and a degree of patient-level error (e
ij
). In the level 1 regression,
subscript i is included to model fixed patient-level variation, and subscript j is included to
model random agency-level variation. Level 2 regressions are intercept-only models that
estimate coefficients to account for between-agency differences in level 1 factors.
Specifically, between-agency variation estimates are produced for the level 1 regression
intercept, and each level 1 predictor. In these series of regressions, predicted coefficient
outcomes ß
xj
are related to a linear combination of a level 2 intercept (γ
x0
), and an
,..., ( )
,..., ( )
~(, )
~,
(: )
(: )
(: )
(: / )
// (: /
)
/(:)
(: )
(: )
(
)
Level Regression
Level Regressions
ModelReducedForm IncorporatesRegressions
forboth Levels
Model Key
yx x x x x
xxxx x e
u
uu
uu
uu
uu
uu
yxxxxxx
xxx xuux
ux ux ux ux ux ux
ux ux u x e
i n level cases
jnlevelcases
eN
u
u
u
u
u
u
u
u
u
u
u
N
yPatientOutcomeofInterest
xAge
xFemalerefMale
x Black ref White
xHispanicrefWhite
xWidowedrefMarriedLivingwithPartner
x Single Divorced Separated ref Married
Living with Partner
xCongestiveHeartFailureHeartDiseaserefCancer
xLungDiseaserefCancer
xNeurologicalDiseaserefCancer
xComorbidityCount
11
12
0
0
0
0
0
0
0
0
0
0
0
0
1
2
ij j j ij j ij j ij j ij j ij
jij jij jij jij j ij ij
jj
jj j j
jj jj
jj jj
jj jj
jj j j
ij ij ij ij ij ij ij
ij ij ij ij j j ij
jij jij jij jij jij jij
jij jij jij ij
ij e
j
j
j
j
j
j
j
j
j
j
j
011 22 33 44 55
6 6 7 7 8 8 9 9 10 10
0 00 0
1 10 1 10 100 10
2 20 2 7 70 7
3 30 3 8 80 8
4 40 4 9 90 9
5 50 5 10 100 10
00 10 20 30 40 50 60
70 80 90 100 0 1
234567
8 9 10
2
0
1
2
3
4
5
6
7
8
9
10
00
10
20
30
40
50
60
70
80
90
100
11
21
31
41
51
61
71
81
91
101
22
32
42
52
62
72
82
92
102
33
43
53
63
73
83
93
103
44
54
64
74
84
94
104
55
65
75
85
95
105
66
76
86
96
106
77
87
97
107
88
98
108
99
109 1010
1
2
3
4
5
6
7
8
9
10
bb b b b b
bbbbb
bc
bc b c
bc bc
bc bc
bc bc
bc b c
cc c c c c c
cccc
v
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
xx
=+ + + + +
+++++ +
=+
=+ = +
=+ =+
=+ =+
=+ =+
=+ = +
=+ + + + + +
++++ ++
++++++
+++ +
=
=
=
=
=
=
=
=
=
=
=
=
=
J
L
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K K
N
P
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O O
R
T
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
R
T
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
R
T
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
V
X
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
V
X
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
V
X
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
,..., ( )
,..., ( )
~(, )
~,
(: )
(: )
(: )
(: / )
// (: /
)
/(:)
(: )
(: )
(
)
Level Regression
Level Regressions
ModelReducedForm IncorporatesRegressions
forboth Levels
Model Key
yx x x x x
xxxx x e
u
uu
uu
uu
uu
uu
yxxxxxx
xxx xuux
ux ux ux ux ux ux
ux ux u x e
i n level cases
jnlevelcases
eN
u
u
u
u
u
u
u
u
u
u
u
N
yPatientOutcomeofInterest
xAge
xFemalerefMale
x Black ref White
xHispanicrefWhite
xWidowedrefMarriedLivingwithPartner
x Single Divorced Separated ref Married
Living with Partner
xCongestiveHeartFailureHeartDiseaserefCancer
xLungDiseaserefCancer
xNeurologicalDiseaserefCancer
xComorbidityCount
11
12
0
0
0
0
0
0
0
0
0
0
0
0
1
2
ij j j ij j ij j ij j ij j ij
jij jij jij jij j ij ij
jj
jj j j
jj jj
jj jj
jj jj
jj j j
ij ij ij ij ij ij ij
ij ij ij ij j j ij
jij jij jij jij jij jij
jij jij jij ij
ij e
j
j
j
j
j
j
j
j
j
j
j
011 22 33 44 55
6 6 7 7 8 8 9 9 10 10
0 00 0
1 10 1 10 100 10
2 20 2 7 70 7
3 30 3 8 80 8
4 40 4 9 90 9
5 50 5 10 100 10
00 10 20 30 40 50 60
70 80 90 100 0 1
234567
8 9 10
2
0
1
2
3
4
5
6
7
8
9
10
00
10
20
30
40
50
60
70
80
90
100
11
21
31
41
51
61
71
81
91
101
22
32
42
52
62
72
82
92
102
33
43
53
63
73
83
93
103
44
54
64
74
84
94
104
55
65
75
85
95
105
66
76
86
96
106
77
87
97
107
88
98
108
99
109 1010
1
2
3
4
5
6
7
8
9
10
bb b b b b
bbbbb
bc
bc b c
bc bc
bc bc
bc bc
bc b c
cc c c c c c
cccc
v
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
xx
=+ + + + +
+++++ +
=+
=+ = +
=+ =+
=+ =+
=+ =+
=+ = +
=+ + + + + +
++++ ++
++++++
+++ +
=
=
=
=
=
=
=
=
=
=
=
=
=
J
L
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K K
N
P
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O O
R
T
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
R
T
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
R
T
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
V
X
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
V
X
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
V
X
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
Figure 4. Multilevel Regression Models
45
agency-level error term (u
xj
). Agency-level residuals of patient-level predictors (u
xj
) are
included to model the potential level 2 variation for all level 1 predictors; however, such
residuals were excluded in the final models due a lack of theoretical support in the
existing literature. Notwithstanding, level 1 intercepts included agency-level residuals to
model the nested nature of the data. In their reduced forms, estimated multilevel
regression models substitute level 2 regressions in the level 1 equation (Figure 5). The
resulting equation estimates a patient-level predicted outcome from an intercept (random
at level 2), a series of level 1 predictors (in the illustrated example, random at level 2),
and a degree of estimated error at levels 1 and 2.

Logistic regressions. Adjusting for patient-level predictors and agency-level
variation, unconditional multilevel generalized linear models with logit link functions
,..., ( )
,..., ( )
~(, )
~,
(: )
(: )
(: )
(: / )
// (: /
)
/(:)
(: )
(: )
(
)
Level Regression
Level Regressions
ModelReducedForm IncorporatesRegressions
forboth Levels
Model Key
yx x x x x
xxxx x e
u
uu
uu
uu
uu
uu
yxxxxxx
xxx xuux
ux ux ux ux ux ux
ux ux u x e
i n level cases
jnlevelcases
eN
u
u
u
u
u
u
u
u
u
u
u
N
yPatientOutcomeofInterest
xAge
xFemalerefMale
x Black ref White
xHispanicrefWhite
xWidowedrefMarriedLivingwithPartner
x Single Divorced Separated ref Married
Living with Partner
xCongestiveHeartFailureHeartDiseaserefCancer
xLungDiseaserefCancer
xNeurologicalDiseaserefCancer
xComorbidityCount
11
12
0
0
0
0
0
0
0
0
0
0
0
0
1
2
ij j j ij j ij j ij j ij j ij
jij jij jij jij j ij ij
jj
jj j j
jj jj
jj jj
jj jj
jj j j
ij ij ij ij ij ij ij
ij ij ij ij j j ij
jij jij jij jij jij jij
jij jij jij ij
ij e
j
j
j
j
j
j
j
j
j
j
j
011 22 33 44 55
6 6 7 7 8 8 9 9 10 10
0 00 0
1 10 1 10 100 10
2 20 2 7 70 7
3 30 3 8 80 8
4 40 4 9 90 9
5 50 5 10 100 10
00 10 20 30 40 50 60
70 80 90 100 0 1
234567
8 9 10
2
0
1
2
3
4
5
6
7
8
9
10
00
10
20
30
40
50
60
70
80
90
100
11
21
31
41
51
61
71
81
91
101
22
32
42
52
62
72
82
92
102
33
43
53
63
73
83
93
103
44
54
64
74
84
94
104
55
65
75
85
95
105
66
76
86
96
106
77
87
97
107
88
98
108
99
109 1010
1
2
3
4
5
6
7
8
9
10
bb b b b b
bbbbb
bc
bc b c
bc bc
bc bc
bc bc
bc b c
cc c c c c c
cccc
v
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
xx
=+ + + + +
+++++ +
=+
=+ = +
=+ =+
=+ =+
=+ =+
=+ = +
=+ + + + + +
++++ ++
++++++
+++ +
=
=
=
=
=
=
=
=
=
=
=
=
=
J
L
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K
K K
N
P
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O O
R
T
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
R
T
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
R
T
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
V
X
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
V
X
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
V
X
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
Figure 5. Reduced Form Multilevel Regression Model
46
were applied for study outcomes to test hypotheses 1, 1A, 1B, 2, 2A, and 4. Intraclass
Correlation Coefficients were calculated to assess the proportion of observed variance
between agencies (Singer, 1987), and given the large number level 2 units, between-
group variance significance was determined using the asymptotic chi-squared Wald test
(Maas & Snijders, 2003). Compared to the likelihood ratio test, the more conservative
Wald test draws from the approximate Satterthwaite degrees of freedom (Satterthwaite,
1946; Welch, 1947), which are proportionally-adjusted to provide a more accurate p-
value estimate for the test. Only patients who died while receiving hospice care (n =
3,043) were included in SOD analyses.
Cox proportional hazards regression. Hospice LOS models employed
censoring to account for patients who withdrew from hospice prior to death. Although
NHHCS 2007 data were collected at one time point, data provided by the hospice
agencies, namely the total number of days under hospice care, spanned the entire hospice
stay, allowing for analysis of patient survival following hospice enrollment. Preliminary
patient-level analyses were conducted using mean bivariate comparisons, and the Kaplan-
Meier estimator with Log-rank statistics to determine unadjusted racial/ethnic differences
in hospice LOS. Adjusting for patient-level predictors and agency-level variation
(random intercept approach), survival analysis was conducted to determine racial/ethnic
differences in hospice LOS using Cox Proportional Hazards with the Breslow estimator.
By testing for relative risk differences (among all decedents), the Cox Proportional
Hazards model permitted examination of whether one racial/ethnic group was at a
significant risk of dying earlier in the hospice care trajectory, relative to other decedents
(Hypothesis 3).
47
All data management was performed using R v.3.0.1, and all analyses were
conducted and Stata v.13 (StataCorp, 2013) with appropriate survey weights applied, and
robust standard errors employed, to account for complex survey design (Rabe-Hesketh,
Skrondal, & Pickles, 2005). Significant alpha levels were set at .05. Previous NCHS
publications using NHHCS 2007 data (Jones et al., 2011) guided appropriate alpha-level
selection. Statistical power to detect a small effect size (< 0.2) in the full and decedent
samples was greater than 0.95 for all hypothesized relationships (J. Cohen, 1977).

48
Chapter 4: Description of Analytic Sample
In this chapter, a descriptive overview of the analytic samples is presented, and
bivariate comparisons by race/ethnicity are reported.
The 2007 wave of the National Home Health and Hospice Care Survey (NNHCS)
was analyzed to test the proposed hypotheses. Two primary analytic samples were
utilized for this study. The first sample (Full Sample; n = 3,661; weighted n = 788,872)
consisted of all cases specified in the Sampling Plan, and was used to test Hypotheses 1,
1A, 1B, 2, 2A, and 3. The second analytic sample (Decedent Sample; n = 3,006; weighted
n = 667,820) was a sub-sample of patients who died during the hospice care episode. The
second analytic sample was utilized to test Hypotheses 4, and 4A.
Given the complex nature of the sampling design, problems with estimate
precision arise when presenting both unweighted and survey-adjusted weighted
descriptives and statistics. An example of the substantial percentage shifts that can occur
when accounting for the complex survey design is illustrated in Appendix A.
Accordingly, when reporting descriptive and bivariate findings from complex data such
as NHHCS 2007 data, researchers regularly present only percentages and test statistics
(Sengupta, Park-Lee, Valverde, Caffrey, & Jones, 2013); however, for complete
transparency, results from this study are also presented alongside the weighted sample
size that is represented for each univariate and bivariate finding. For unweighted
descriptions of the Full Sample and Decedent Sample, see Appendices B and C,
respectively.

49
Full Sample
The full analytic sample consisted of 3,341 White, 219 Black, and 101 Hispanic
hospice patients. When weighted for national representation, these patients represent
711,284 White, 51,368 Black, and 26,220 Hispanic hospice patients in the United States
(Table 1).

Table 1. Full Sample Size (n = 788,872)
White Black Hispanic Total
Unweighted Sample 3,341 219 101 3,661
Weighted Sample 711,284 51,368 26,220 788,872

A descriptive summary of the full analytic sample can be found in Table 2. The
mean age of the sample was 82.8 ± 8.3. Forty-three percent of the sample was female.
Forty percent of the sample was married or living with a partner, 46% widowed, 5%
divorced or separated, and 4% never married. Thirty-six percent of the sample had a
primary diagnosis of Cancer, 13% Congestive Heart Failure or Heart Disease , 16% Lung
Disease, 21% a neurological disease, and 13% another primary diagnosis. The mean
comorbidity count for the sample was 3.4 ± 2.4. Sixteen-percent of the sample was
enrolled in Medicaid. Bivariate analysis indicated significant racial/ethnic differences in
gender (F(1.88, 1219.15) = 3.85; p = .024), and Medicaid enrollment (F(1.93, 1247.18) =
22.26; p < .001). Specifically, relative to Whites, Blacks had a higher percentage of
females (58%), Hispanics had a lower percentage of females (34%), and both Blacks
(37%) and Hispanics (36%) had higher Medicaid enrollment rates. No significant
racial/ethnic differences were observed for age (F(2.00, 647.00) = 2.75; p = .065), marital
50
status (F(4.98, 3213.92) = 1.82; p = .106), primary diagnosis (F(6.51, 4220.89) = 0.85; p
= .542), or comorbidity count (F(2.00, 647.00) = 1.60; p = .202).

Table 2. Weighted Full Sample Description (n = 788,872)
Total White Black Hispanic
(n = 788,872) (n = 711,284) (n = 51,368) (n = 26,220) Sig.
Age, Mean ± SD 82.82 ± 8.30 83.00 ± 8.33 80.72 ± 8.36 82.03 ± 6.83 .065
Female, No. (%) 338,468 (42.91) 299,860 (42.16) 29,752 (57.92) 8,856 (33.78) .024
Marital Status, No. (%) .106
Married/Living with Partner 313,820 (39.78) 287,064 (40.36) 20,756 (40.41) 6,000 (22.88)
Widowed 362,512 (45.95) 325,040 (45.70) 21,328 (41.52) 16,144 (61.57)
Divorced/Separated 42,144 (5.34) 36,392 (5.12) 3,972 (7.73) 1,780 (6.79)
Never Married 29,524 (3.74) 25,888 (3.64) 3,188 (6.21) 448 (1.71)
Missing 40,872 (5.18) 36,900 (5.19) 2,124 (4.14) 1,848 (7.05)
Primary Diagnosis, No. (%) .542
Cancer 287,308 (36.42) 258,308 (36.32) 19,732 (38.41) 9,268 (35.35)
CHF/Heart Disease 100,376 (12.72) 91,180 (12.82) 5,764 (11.22) 3,432 (13.09)
Lung Disease 128,028 (16.23) 120,336 (16.92) 5,260 (10.24) 2,432 (9.28)
Neurological Diseases 168,156 (21.32) 147,512 (20.74) 12,696 (24.72) 7,948 (30.31)
Other 104,140 (13.20) 93,084 (13.09) 7,916 (15.41) 3,140 (11.98)
Missing 864 (0.11) 864 (0.12) 0 (0.00) 0 (0.00)
Comorbidity Count, Mean ± SD 3.39 ± 2.40 3.40 ± 2.42 3.07 ± 2.03 3.65 ± 2.33 .202
Medicaid Enrollee, No. (%) 125,224 (15.87) 96,912 (13.62) 18,960 (36.91) 9,352 (35.67) <.001
Missing 8,744 (1.11) 8,744 (1.23) 0 (0.00) 0 (0.00)
Footnotes
SD: Standard Deviation
CHF: Congestive Heart Failure
Percentages are presented by column
Characteristics

Discharge reasons. Patients in the full analytic sample were discharged from
hospice for a variety of reasons (Table 3), including death (85%), improved health (6%),
revocation for more aggressive treatment (5%), unspecified revocation (2%), and
relocation (2%). Racial/Ethnic differences were found in reasons for discharge (F(6.07,
3935.11) = 2.15; p = .045), with the most substantial differences among death (White
86%; Black 72%; Hispanic 76%), revocation for more aggressive treatment (White 4%;
Black 11%; Hispanic 10%), and unspecified revocation (White 1%; Black 4%; Hispanic
5%).

51
Table 3. Reasons for Hospice Discharge (n = 788,872)
Total White Black Hispanic Sig.
Documented Reason, No. (%) .044
Died 667,820 (84.66) 610,820 (85.88) 37,076 (72.18) 19,924 (75.99)
Stabilized or Improved 46,148 (5.85) 39,688 (5.58) 4,672 (9.10) 1,788 (6.82)
More Aggressive Treatment 39,664 (5.03) 31,356 (4.41) 5,704 (11.10) 2,604 (9.93)
Moved 17,196 (2.18) 15,388 (2.16) 1,688 (3.29) 120 (0.48)
Revoked (Unspecified) 13,820 (1.75) 10,284 (1.45) 2,116 (4.12) 1,420 (5.42)
Other 2,764 (0.35) 2,672 (0.38) 92 (0.18) 0 (0.00)
Missing 1,460 (0.19) 1,076 (0.15) 20 (0.04) 364 (1.39)
Footnotes
Percentages are presented by column

Decedent Sample
Derived from the full analytic sample, the decedent sample consisted of 2,787
White, 139 Black, and 80 Hispanic hospice patients. When weighted for national
representation, these patients represent 610,820 White, 37,087 Black, and 19,924
Hispanic U.S. hospice decedents.

Table 4. Weighted Decedent Sample Description (n = 667,820)
Total White Black Hispanic
(n = 667,820) (n = 610,820) (n = 37,087) (n = 19,924) Sig.
Age, Mean ± SD 82.58 ± 8.37 82.77 ± 8.40 79.91 ± 7.95 81.55 ± 7.45 .068
Female, No. (%) 291,012 (43.58) 262,660 (43.00) 21,216 (57.22) 7,136 (35.82) .126
Marital Status, No. (%) .166
Married/Living with Partner 273,588 (40.97) 252,992 (41.42) 16,064 (43.33) 4,532 (22.75)
Widowed 273,589 (45.29) 275,356 (45.08) 15,260 (41.16) 11,860 (59.53)
Divorced/Separated 273,590 (4.88) 28,340 (4.64) 2,908 (7.84) 1,356 (6.81)
Never Married 273,591 (3.83) 23,100 (3.78) 2,124 (5.73) 328 (1.65)
Missing 273,592 (5.03) 31,032 (5.08) 720 (1.94) 1,848 (9.28)
Primary Diagnosis, No. (%) .167
Cancer 255,580 (38.27) 231,428 (37.89) 16,336 (44.06) 7,816 (39.23)
CHF/Heart Disease 83,996 (12.58) 81,440 (13.33) 1,444 (3.90) 1,112 (5.58)
Lung Disease 105,364 (15.78) 98,720 (16.16) 4,336 (11.69) 2,308 (11.58)
Neurological Diseases 138,632 (20.76) 122,256 (20.02) 9,508 (25.64) 6,868 (34.47)
Other 83,384 (12.49) 76,112 (12.46) 5,452 (14.70) 1,820 (9.14)
Missing 864 (0.13) 864 (0.14) 0 (0.00) 0 (0.00)
Comorbidity Count, Mean ± SD 3.41 ± 2.42 3.43 ± 2.45 3.01 ± 1.96 3.64 ± 2.14 .223
Medicaid Enrollee, No. (%) 98,680 (14.78) 78,268 (12.81) 13,660 (36.84) 6,752 (33.89) <.001
Missing 7,556 (1.13) 7,556 (1.23) 0 (0.00) 0 (0.00)
Footnotes
SD: Standard Deviation
CHF: Congestive Heart Failure
Percentages are presented by column
Characteristics

52
A descriptive summary of the decedent analytic sample is presented in Table 4.
The mean age of the decedent sample was 82.6 ± 8.4. Forty-four percent of decedents
were female. Forty-one percent were married or living with a partner, 45% widowed, 5%
divorced or separated, and 4% never married. Thirty-eight percent of the decedent sample
had a primary diagnosis of Cancer, 13% Congestive Heart Failure or Heart Disease , 16%
Lung Disease, 21% a neurological disease, and 13% another primary diagnosis. The
mean comorbidity count was 3.4 ± 2.4. Fifteen-percent of decedents were enrolled in
Medicaid. As with the full analytic sample, bivariate analysis indicated significant
racial/ethnic differences in Medicaid enrollment among decedents (F(1.74, 1090.50) =
15.16; p < .001). Specifically, relative to Whites (13%), Blacks (37%) and Hispanics
(34%) had higher rates of Medicaid enrollment. No significant racial/ethnic differences
were observed for age (F(2.00, 629.00) = 2.70; p = .068), gender (F(1.83, 1154.96) =
2.11; p = .126), marital status (F(6.30, 3971.58) = 1.51; p = .166), primary diagnosis
(F(6.67, 4203.87) = 1.50; p = .167), or comorbidity count (F(2.00, 629) = 1.51; p = .223).
53
Chapter 5: Advance Care Planning Decisions
In this chapter, bivariate comparisons by advance care planning decisions are
presented, and multivariable results are reported. The full analytic sample was utilized to
investigate racial/ethnic variation in advance directive completion. Subsequent analyses
examining 1) do not resuscitate (DNR) order election, and 2) healthcare proxy
designation were conducted using the subsample of patients with a documented advance
directive. For frequencies and relative percentages of missing and excluded data for all
advance care planning analyses, see Appendices D (advance directive completion), E
(DNR election), and F (healthcare proxy designation).
Bivariate Comparisons
Advance care planning decisions were examined using three dependent variables:
1) advance directive completion, 2) DNR election, and 3) healthcare proxy designation.
As such, bivariate comparisons are presented separately below for each dependent
variable.
Advance directive completion. A summary of patient characteristics for those
with and without an advance directive is presented in Table 5. Bivariate analyses
indicated significant differences in advance directive completion by race/ethnicity
(F(1.67, 1078.21) = 8.98; p < .001). As hypothesized, advance directive completion rates
were significantly lower among Black patients (80%), relative to White patients (93%);
however, advance directive completion rates were not significantly lower among
Hispanic patients (96%), relative to White patients. No significant differences in advance
directive completion were observed by age (F(2.69, 1735.08) = 1.36; p = .256), gender
(F(1.00, 645) = 0.60; p = .438), marital status (F(2.51, 1612.88) = 2.14; p = .105),
54
primary diagnosis (F(3.65, 2354.37) = 0.72; p = .566), or comorbidity count (F(1.60,
1033.78) = 1.14; p = .310).

Table 5. Advance Directive Completion Bivariate Comparisons (n = 778,520)
No Yes Sig.
Age .256
65-75 15,244 (9.54) 144,556 (90.46)
76-82 16,456 (9.03) 165,684 (90.97)
83-87 13,856 (7.46) 171,992 (92.54)
88+ 14,928 (5.95) 235,804 (94.05)
Gender .438
Female 27,852 (8.32) 307,088 (91.68)
Male 32,632 (7.36) 410,948 (92.64)
Race/Ethnicity .004
White 49,164 (7.01) 652,144 (92.99)
Black 10,312 (20.08) 41,044 (79.92)
Hispanic 1,008 (3.90) 24,848 (96.10)
Marital Status .105
Married/Living Together 26,056 (8.41) 283,880 (91.59)
Widowed 20,536 (5.69) 340,156 (94.31)
Divorced/Separated 5,136 (12.19) 37,008 (87.81)
Never Married 2,360 (8.00) 27,152 (92.00)
Missing 6,396 (17.65) 29,840 (82.35)
Primary Diagnosis .566
Cancer 23,128 (8.11) 262,060 (91.89)
CHF/Heart Disease 10,528 (10.68) 88,016 (89.32)
Lung Disease 7,916 (6.27) 118,380 (93.73)
Neurological Diseases 10,692 (6.43) 155,616 (93.57)
Other 7,960 (7.85) 93,440 (92.15)
Missing 260 (33.16) 524 (66.84)
Comorbidity Count .310
0-1 17,572 (8.45) 190,464 (91.55)
2-3 23,060 (9.31) 224,692 (90.69)
4+ 19,852 (6.15) 302,880 (93.85)
Footnotes
CHF: Congestive Heart Failure
Percentages are presented by row
Characteristics, No. (%)
Documented Advance Directive

Do not resuscitate order election. A summary of patient characteristics for those
with and without a DNR order is presented in Table 6. Reported frequencies and
percentages comprise the subsample of patients who completed an advance directive.
Bivariate analyses indicated no significant differences in DNR order election by
race/ethnicity (F(1.99, 1274.64) = 0.78; p = .459). Contrary to what was hypothesized,
55
DNR order election rates were not significantly lower among Black patients (91%),
relative to White (91%) and Hispanic patients (96%). Significant differences in DNR
order election were observed by primary diagnosis (F(3.63, 2317.87) = 3.28; p = .014).
Specifically, those with neurological diseases (95%), or other primary diagnoses (96%)
were more likely to elect a DNR order than patients with Cancer (89%), Congestive
Heart Failure or Heart Disease (90%), or Lung Disease (89%). No significant differences
in DNR order election were observed by age (F(2.77, 1771.98) = 0.58; p = .614), gender
(F(1.00, 639) = 0.53; p = .467), marital status (F(2.91, 1851.35) = 1.00; p = .390), or
comorbidity count (F(1.87, 1196.49) = 0.83; p = .429).

56
Table 6. Do Not Resuscitate Order Election Bivariate Comparisons (n = 718,036)
No Yes Sig.
Age .614
65-75 13,888 (9.61) 130,668 (90.39)
76-82 12,820 (7.74) 152,864 (92.26)
83-87 11,872 (6.90) 160,120 (93.10)
88+ 23,516 (9.97) 212,288 (90.03)
Gender .467
Female 24,212 (7.88) 282,876 (92.12)
Male 37,884 (9.22) 373,064 (90.78)
Race/Ethnicity .459
White 57,472 (8.81) 594,672 (91.19)
Black 3,748 (9.13) 37,296 (90.87)
Hispanic 876 (3.53) 23,972 (96.47)
Marital Status .390
Married/Living Together 23,224 (8.18) 260,656 (91.82)
Widowed 28,048 (8.25) 312,108 (91.75)
Divorced/Separated 4,356 (11.77) 32,652 (88.23)
Never Married 3,944 (14.53) 23,208 (85.47)
Missing 2,524 (33.16) 27,316 (66.84)
Primary Diagnosis .014
Cancer 27,764 (10.59) 234,296 (89.41)
CHF/Heart Disease 9,244 (10.50) 78,772 (89.50)
Lung Disease 13,468 (11.38) 104,912 (88.62)
Neurological Diseases 8,236 (5.29) 147,380 (94.71)
Other 3,364 (3.60) 90,076 (96.40)
Missing 20 (3.82) 504 (96.18)
Comorbidity Count .429
0-1 19,852 (10.42) 170,612 (89.58)
2-3 16,520 (7.35) 208,172 (92.65)
4+ 25,724 (8.49) 277,156 (91.51)
Footnotes
CHF: Congestive Heart Failure
Percentages are presented by row
Characteristics, No. (%)
Documented DNR Order

Healthcare proxy designation. A summary of patient characteristics for those
with and without a designated healthcare proxy is presented in Table 7. As with the
bivariate analyses examining DNR order election, reported frequencies and percentages
comprise the subsample of patients who completed an advance directive. Bivariate
analyses indicated significant differences in healthcare proxy designation by
race/ethnicity (F(1.83, 1167.93) = 5.28; p = .007). As hypothesized, healthcare proxy
designation rates were significantly lower among Black patients (39%), relative to White
patients (55%); however, healthcare proxy designation rates were also significantly lower
57
among Hispanic patients (31%), relative to White patients. Significant differences in
healthcare proxy designation were also observed by age (F(2.92, 1864.84) = 3.75; p =
.011), gender (F(1.00, 639) = 4.13; p = .043), marital status (F(2.94, 1874.01) = 6.44; p <
.001), and comorbidity count (F(1.87, 1196.27) = 4.15; p = .018). Specifically, healthcare
proxy designation rates were higher among 1) patients in the oldest age group (60%),
relative to the youngest age group (47%), and 2) males (56%), relative to females (50%).
Conversely, healthcare proxy designation rates were lower among 1) patients who were
married or living with a partner (45%), relative to widowed (58%), divorced or separated
(56%), and never married patients (59%), and 2) patients with 0-1 comorbidities (45%),
relative to patients with 2-3 comorbidities (56%), and four or more comorbidities (56%).
No significant differences in healthcare proxy designation were observed by primary
diagnosis (F(3.84, 2454.64) = 1.40; p = .234).

58
Table 7. Healthcare Proxy Designation Bivariate Comparisons (n = 718,036)
No Yes Sig.
Age .011
65-75 77,220 (53.42) 67,336 (46.58)
76-82 82,376 (49.72) 83,308 (50.28)
83-87 82,752 (48.11) 89,240 (51.89)
88+ 94,880 (40.24) 140,924 (59.76)
Gender .043
Female 154,624 (50.35) 152,464 (49.65)
Male 182,604 (44.43) 228,344 (55.57)
Race/Ethnicity .007
White 294,744 (45.20) 357,400 (54.80)
Black 25,244 (61.50) 15,800 (38.50)
Hispanic 17,240 (69.38) 7,608 (30.62)
Marital Status <.001
Married/Living Together 155,304 (54.71) 128,576 (45.29)
Widowed 141,268 (41.53) 198,888 (58.47)
Divorced/Separated 16,376 (44.25) 20,632 (55.75)
Never Married 11,228 (41.35) 15,924 (58.65)
Missing 13,052 (43.74) 16,788 (56.26)
Primary Diagnosis .234
Cancer 135,172 (51.58) 126,888 (48.42)
CHF/Heart Disease 38,336 (43.56) 49,680 (56.44)
Lung Disease 54,064 (45.67) 64,316 (54.33)
Neurological Diseases 69,344 (44.56) 86,272 (55.44)
Other 39,788 (42.58) 53,652 (57.42)
Missing 524 (100.00) 0 (0.00)
Comorbidity Count .018
0-1 105,612 (55.45) 84,852 (44.55)
2-3 99,276 (44.18) 125,416 (55.82)
4+ 132,340 (43.69) 170,540 (56.31)
Footnotes
CHF: Congestive Heart Failure
Percentages are presented by row
Characteristics, No. (%)
Documented Healthcare Proxy

Multivariable Models
Racial/Ethnic differences in advance care planning were tested using
multivariable models that were conducted in three waves: 1) multilevel model predictions
of advance directive completion (Hypothesis 1), 2) multilevel model predictions of DNR
order election (Hypothesis 1A), and 3) multilevel model predictions of healthcare proxy
designation (Hypothesis 1B). All models included patient-level (level 1) predictors to
adjust for key demographic and health characteristics, and random agency-level (level 2)
intercepts to account for between-agency variation.
59
Advance directive completion. A multilevel logistic regression was conducted to
test if White hospice patients were more likely than Black and Hispanic hospice patients
to have an advance directive. Results of this model are presented in Table 8.

Table 8. Predictors of Advance Directive Completion
OR
(Std. Err.) 95% CI Sig.
Model Predictors
Race/Ethnicity
Black 0.20 0.14, 0.30 <.001
(0.20)
Hispanic 1.04 0.87, 1.24 .703
(0.09)
Age 1.01 0.99, 1.03 .504
(0.01)
Female 0.44 0.35, 0.55 <.001
(0.12)
Marital Status
Widowed 1.60 1.42, 1.80 <.001
(0.06)
Single/Divorced/Separated 1.75 1.08, 2.85 .024
(0.25)
Primary Diagnosis
CHF/Heart Disease 0.34 0.22, 0.53 <.001
(0.22)
Lung Disease 1.06 0.33, 3.38 .920
(0.59)
Neurological Disease 2.40 1.15, 5.01 .020
(0.38)
Other Primary Diagnosis 0.71 0.46, 1.10 .127
(0.23)
Comorbidity Count 1.05 1.01, 1.10 .015
(0.02)
Random Intercept 14.67 3.19, 67.49 .001
(0.78)
Model Summary 1.01 (0.14)
Level 2 Variance 1.29 (0.17)
Akaike Information Criterion
Bayesian Information Criterion
Patient Observations 3,517
Agency Observations 651
Footnotes
CHF: Congestive Heart Failure
Reference Variables
White (Race)
Married/Living with Partner (Marital Status)
Cancer (Primary Diagnosis)
217664.92
217714.24

60
Preliminary multilevel analysis of the null model (agency-level n = 654; patient-
level n = 3,639) revealed an Intraclass Correlation Coefficient of 0.423 (95% CI: 0.346-
0.503), indicating that approximately 42% of the observed variation in advance directive
completion was due to differences between agencies. Subsequent analysis further
confirmed significant between-group variance (χ
2
= 7.59, df = 1; p = .006), suggesting
significant variation across agencies. Therefore the multilevel random intercept
methodological approach was employed to account for unexplained level 2 variance. The
adjusted model (Level 1 n = 3,517; Level 2 n = 651) revealed an AIC fit criteria of
217664.92, and a BIC fit criteria of 217714.24. As hypothesized, Black patients (OR =
0.20; p < .001) were less likely than White patients to have an advance directive;
however, Hispanic patients (OR = 1.04; p = .703) were not less likely than White patients
to have an advance directive. Female patients (OR = 0.44; p < .001), and those with
Congestive Heart Failure/Heart Disease (OR = 0.34; p < .001) were significantly less
likely than the associated referents to have an advance directive. Conversely, widowed
patients (OR = 1.60; p < .001), single/divorced/separated patients (OR = 1.75; p = .024),
and those with a neurological disease (OR = 2.40; p = .020) were significantly more
likely than the associated referents to have an advance directive. Finally, comorbidity
count was positively associated with having an advance directive (OR = 1.05; p = .015).
See Figure 6 for a blobbogram illustrating the relative effect of each predictor for
advance directive completion.

61

Do not resuscitate order election. A multilevel logistic regression was employed
to test if White and Hispanic hospice patients were more likely than Black hospice
patients to elect a DNR order. Model results can be found in Table 9.

Footnotes
CHF: Congestive Heart Failure
HD: Heart Disease
DNR: Do Not Resuscitate Order
Reference Variables
White (Race)
Married/Living with Partner (Marital Status)
Cancer (Primary Diagnosis)
No Advance Directive (Advance Care Planning)
Figure 6. Relative Effects of Predictors for Advance Directive Completion
0.1
1
10
* Comorbidity Count
Other Primary Diagnosis
* Neurological Diseases
Lung Disease
*** CHF/HD
* Single/Divorced/Separated
*** Widowed
Hispanic
*** Black
*** Female
Age
Odds Ratio
Relative Effect Size
62
Table 9. Predictors of Do Not Resuscitate Order Election
OR
(Std. Err.) 95% CI Sig.
Model Predictors
Race/Ethnicity
Black 1.09 0.83, 1.43 .553
(0.14)
Hispanic 3.68 3.52, 3.86 <.001
(0.02)
Age 0.99 0.96, 1.01 .287
(0.01)
Female 0.66 0.52, 0.84 .001
(0.12)
Marital Status
Widowed 1.12 0.98, 1.27 .090
(0.07)
Single/Divorced/Separated 0.29 0.23, 0.38 <.001
(0.13)
Primary Diagnosis
CHF/Heart Disease 1.13 0.79, 1.62 .492
(0.18)
Lung Disease 1.86 1.29, 2.69 .001
(0.19)
Neurological Disease 4.73 3.75, 5.96 <.001
(0.12)
Other Primary Diagnosis 6.82 3.32, 14.00 <.001
(0.37)
Comorbidity Count 0.90 0.83, 0.98 .016
(0.04)
Random Intercept 62.45 5.89, 662.01 .001
(1.21)
Model Summary
Level 2 Variance 1.77 (0.23)
Akaike Information Criterion
Bayesian Information Criterion
Patient Observations 3,187
Agency Observations 646
Footnotes
CHF: Congestive Heart Failure
Reference Variables
White (Race)
Married/Living with Partner (Marital Status)
Cancer (Primary Diagnosis)
196256.58
196305.12

Preliminary analysis of the null model (agency-level n = 648; patient-level n =
3,293) indicated that approximately 53% of the variation in DNR order election was due
to between-agency differences (ICC = 0.530, 95% CI: 0.448-0.610; χ
2
= 7.61, df = 1; p =
.006), thus justifying the multilevel approach. The adjusted model (Level 1 n = 3,187;
Level 2 n = 646), revealed AIC and BIC model fit criteria were 196256.58, and
63
196305.12, respectively. Contrary to what was hypothesized, Black patients (OR = 1.09;
p = .553) were equally as likely as White patients to elect a DNR order. Additionally,
Hispanic patients (OR = 3.68; p < .001) were 3.68 times as likely as White patients to
elect a DNR order. Subsequent analyses (not shown), further indicated that Hispanic
patients (OR = 3.67; p < .001) were also more likely than Black patients to elect a DNR
order. Patients with Lung Disease (OR = 1.86; p < .001), a neurological disease (OR =
4.73; p < .001), or other primary diagnoses (OR = 6.82; p < .001) were significantly more
likely than patients with Cancer to elect a DNR order. Conversely, female patients (OR =
0.66; p < .001), and those single/divorced/separated (OR = 0.29; p < .001) were
significantly less likely than the associated referents to elect a DNR order. Lastly,
patients with more comorbidities (OR = 0.90; p = .016) were less likely to elect a DNR
order than patients with fewer comorbidities. See Figure 7 for a blobbogram illustrating
the relative effect of each predictor for a documented DNR order.

64

Healthcare proxy designation. To test if White and Hispanic hospice patients
were more likely than Black hospice patients to designate a healthcare proxy, a multilevel
logistic regression was conducted (Table 10).

Footnotes
CHF: Congestive Heart Failure
HD: Heart Disease
DNR: Do Not Resuscitate Order
Reference Variables
White (Race)
Married/Living with Partner (Marital Status)
Cancer (Primary Diagnosis)
No Advance Directive (Advance Care Planning)
0.1
1
10
100
* Comorbidity Count
*** Other Primary Diagnosis
*** Neurological Diseases
** Lung Disease
CHF/HD
*** Single/Divorced/Separated
* Widowed
*** Hispanic
Black
** Female
Age
Odds Ratio
Relative Effect Size
Figure 7. Relative Effects of Predictors for Do Not Resuscitate Order Election
65
Table 10. Predictors of Healthcare Proxy Designation
OR
(Std. Err.) 95% CI Sig.
Model Predictors
Race/Ethnicity
Black 0.58 0.46, 0.75 <.001
(0.13)
Hispanic 0.16 0.11, 0.24 <.001
(0.22)
Age 1.01 1.01, 1.01 <.001
(0.00)
Female 1.04 0.90, 1.19 .597
(0.07)
Marital Status
Widowed 1.05 0.93, 1.18 .463
(0.06)
Single/Divorced/Separated 1.71 1.08, 2.72 .022
(0.24)
Primary Diagnosis
CHF/Heart Disease 1.67 1.26, 2.04 <.001
(0.10)
Lung Disease 1.10 0.74, 1.64 .646
(0.21)
Neurological Disease 2.70 1.60, 4.55 <.001
(0.27)
Other Primary Diagnosis 2.20 1.49, 3.24 <.001
(0.20)
Comorbidity Count 1.08 1.06, 1.10 <.001
(0.01)
Random Intercept 0.18 0.13, 0.26 <.001
(0.18)
Model Summary
Level 2 Variance 1.15 (0.17)
Akaike Information Criterion
Bayesian Information Criterion
Patient Observations 3,187
Agency Observations 646
Footnotes
CHF: Congestive Heart Failure
Reference Variables
White (Race)
Married/Living with Partner (Marital Status)
Cancer (Primary Diagnosis)
552471.57
552520.10

Analysis of the null model (agency-level n = 648; patient-level n = 3,293)
indicated that 49% of the variance in healthcare proxy designation was due to differences
between agencies (ICC = 0.494, 95% CI: 0.438-0.550; χ
2
= 19.62, df = 1; p <.001). The
AIC and BIC fit criteria for the adjusted model (Level 1 n = 3,187; Level 2 n = 646), were
552471.57, and 552520.10, respectively. As hypothesized, Black patients (OR = 0.58; p <
66
.001) were less likely than White patients to designate a healthcare proxy; however,
contrary to what was hypothesized, Hispanic patients (OR = 0.16; p < .001) were also
significantly less likely than White patients to designate a healthcare proxy.
Single/Divorced/Separated patients (OR = 1.71; p = .022), and those with Congestive
Heart Failure/Heart Disease (OR = 1.67; p < .001), a neurological disease (OR = 2.70; p <
.001), or other primary diagnoses (OR = 2.20; p < .001) were significantly more likely
than the associated referents to designate a healthcare proxy. Both age (OR = 1.01; p <
.001) and comorbidity count (OR = 1.08; p < .001) were positively associated with
designating a healthcare proxy. For a blobbogram illustrating the relative effect of each
predictor for healthcare proxy designation, see Figure 8.

Footnotes
CHF: Congestive Heart Failure
HD: Heart Disease
DNR: Do Not Resuscitate Order
Reference Variables
White (Race)
Married/Living with Partner (Marital Status)
Cancer (Primary Diagnosis)
No Advance Directive (Advance Care Planning)
Figure 8. Relative Effects of Predictors for Healthcare Proxy Designation
0.01
0.1
1
10
*** Comorbidity Count
*** Other Primary Diagnosis
*** Neurological Diseases
Lung Disease
*** CHF/HD
* Single/Divorced/Separated
Widowed
*** Hispanic
*** Black
Female
*** Age
Odds Ratio
Relative Effect Size
67
Chapter 6: Emergent Care Utilization
In this chapter, bivariate comparisons by emergent care utilization are presented,
and multivariable results are reported. To test all hypotheses associated with emergent
care utilization, the full analytic sample was utilized. Frequencies and relative
percentages of missing and excluded data for emergent care planning analyses can be
found in Appendix G.
Bivariate Comparisons
Emergent care utilization following hospice enrollment was examined using
binary outcome data. A summary of patient characteristics for those who did and did not
utilize emergent care utilization is presented in Table 11. Contrary to the hypothesis,
bivariate analyses indicated that emergent care utilization rates were not significantly
higher among Black patients (10%), relative to White patients (6%) and Hispanic patients
(7%; F(1.86, 1194.72) = 0.89; p = .403). Significant differences in emergent care
utilization were observed by Medicaid enrollment (F(1.00, 642) = 6.22; p = .013), and
advance care planning (F(1.79, 1149.73) = 5.02; p = .009). Specifically, emergent care
utilization rates were higher among 1) patients enrolled in Medicaid (10%), relative to
patients not enrolled in Medicaid (5%), and 2) patients without an advance directive
(14%), relative to patients with an advance directive (DNR order not elected; 7%), and
patients with a do not resuscitate (DNR) order (5%). No significant differences in
emergent care utilization were observed by age (F(2.90, 1864.80) = 1.33; p = .265),
gender (F(1.00, 644) = 0.16; p = .691), marital status (F(2.71, 1737.64) = 0.76; p = .503),
primary diagnosis (F(3.84, 2475.29) = 1.89; p = .113), or comorbidity count (F(1.88,
1212.50) = 1.80; p = .168).
68
Table 11. Emergent Care Utilization Bivariate Comparisons (n = 773,572)
No Yes Sig.
Age .265
65-75 147,996 (93.05) 11,056 (6.95)
76-82 168,232 (93.23) 12,220 (6.77)
83-87 179,260 (96.11) 7,260 (3.89)
88+ 230,892 (93.27) 16,656 (6.73)
Gender .691
Female 315,420 (94.17) 19,536 (5.83)
Male 410,960 (93.69) 27,656 (6.31)
Race/Ethnicity .403
White 656,592 (94.18) 40,568 (5.82)
Black 45,328 (90.31) 4,864 (9.69)
Hispanic 24,460 (93.29) 1,760 (6.71)
Marital Status .503
Married/Living Together 292,140 (94.41) 17,292 (5.59)
Widowed 334,636 (93.40) 23,636 (6.60)
Divorced/Separated 38,816 (92.12) 3,320 (7.88)
Never Married 28,636 (97.16) 836 (2.84)
Missing 32,152 (93.85) 2,108 (6.15)
Primary Diagnosis .113
Cancer 265,140 (94.31) 15,996 (5.69)
CHF/Heart Disease 88,776 (90.00) 9,864 (10.00)
Lung Disease 117,868 (93.00) 8,876 (7.00)
Neurological Diseases 158,956 (96.19) 6,292 (3.81)
Other 95,516 (93.94) 6,164 (6.06)
Missing 124 (100.00) 0 (0.00)
Comorbidity Count .168
0-1 197,024 (95.49) 9,316 (4.51)
2-3 226,048 (92.22) 19,076 (7.78)
4+ 303,308 (94.16) 18,800 (5.84)
Enrolled in Medicaid .013
Yes 112,236 (90.04) 12,416 (9.96)
No 606,904 (94.58) 34,776 (5.42)
Missing 7,240 (100.00) 0 (0.00)
Advance Care Planning .009
No AD, No DNR order 51,832 (86.46) 8,116 (13.54)
Yes AD, No DNR order 56,756 (92.74) 4,440 (7.26)
Yes AD, Yes DNR order 614,244 (94.66) 34,624 (5.34)
Missing 3,548 (99.66) 12 (0.34)
Footnotes
CHF: Congestive Heart Failure
AD: Advance Directive
DNR: Do Not Resuscitate
Percentages are presented by row
Characteristics, No. (%)
Emergent Care Utilization

Multivariable Models
Racial/Ethnic variation in emergent care utilization was analyzed using a two-step
multilevel modeling approach. The first step investigated emergent care utilization,
adjusting for key demographic and health indicators (Hypothesis 2), whereas the second
69
step introduced additional predictors pertaining to care planning behaviors (Hypothesis
2A). This approach was selected to determine the overall effect of race/ethnicity, and the
relative effect of advance care planning, on the hypothesized relationships. Both steps
included patient-level (level 1) predictors with random agency-level (level 2) intercepts
to account for between-agency variation. Finally, preliminary multilevel analysis of the
null model (agency-level n = 653; patient-level n = 3,624) indicated that approximately
34% of the observed variation in emergent care utilization was due to between-agency
differences (ICC = 0.344, 95% CI: 0.268-0.430; χ
2
= 11.68, df = 1; p <.001).
Step 1: Demographics and health indicator variables. To test if Black and
Hispanic hospice patients were more likely than White hospice patients to utilize
emergent care, a multilevel logistic regression was conducted. Results are presented
under Step 1 in Table 12. The adjusted (Level 1 n = 3,480; Level 2 n = 649), revealed an
AIC of 205206.42, and a BIC of 205255.66. Contrary to the hypothesis, Black patients
(OR = 0.81; p = .150) were equally as likely as White patients to utilize emergent care,
and Hispanic patients (OR = 0.56; p < .001) were less likely than White patients to utilize
emergent care. Widowed patients (OR = 1.24; p = .004), those with Congestive Heart
Failure/Heart Disease (OR = 2.49; p < .001), and patients enrolled in Medicaid (OR =
2.16; p = .036) were significantly more likely than the associated referents to utilize
emergent care. Age (OR = 0.98; p < .001), and comorbidity count (OR = 0.94; p < .001)
were negatively associated with emergent care utilization.

70
Table 12. Predictors of Emergent Care Utilization
Step 1 Step 2
OR OR
(Std. Err.) 95% CI Sig. (Std. Err.) 95% CI Sig.
Model Predictors
Race/Ethnicity
Black 0.81 0.61, 1.08 .150 0.50 0.38, 0.66 <.001
(0.14) (0.14)
Hispanic 0.56 0.42, 0.76 <.001 0.70 0.45, 1.09 .117
(0.15) (0.23)
Age 0.98 0.98, 0.99 <.001 0.99 0.98, 0.99 <.001
(0.00) (0.00)
Female 0.86 0.69, 1.07 .178 0.78 0.54, 1.13 .188
(0.11) (0.19)
Marital Status
Widowed 1.24 1.07, 1.43 .004 1.23 1.04, 1.46 .018
(0.07) (0.09)
Single/Divorced/Separated 0.82 0.38, 1.78 .619 0.68 0.33, 1.41 .298
(0.40) (0.37)
Primary Diagnosis
CHF/Heart Disease 2.49 1.73, 3.59 <.001 1.80 1.53, 2.13 <.001
(0.19) (0.09)
Lung Disease 0.98 0.42, 2.29 .954 0.89 0.42, 1.86 .751
(0.43) (0.38)
Neurological Disease 0.72 0.29, 1.77 .475 0.67 0.26, 1.75 .410
(0.46) (0.49)
Other Primary Diagnosis 1.78 0.92, 3.43 .088 1.67 0.87, 3.19 .123
(0.34) (0.33)
Comorbidity Count 0.94 0.91, 0.98 .001 1.01 1.00, 1.02 .131
(0.02) (0.01)
Medicaid Enrollee 2.16 1.05, 4.44 .036 1.72 0.78, 3.82 .182
(0.37) (0.41)
Advance Care Planning
Yes AD, No DNR order 0.47 0.30, 0.75 .002
(0.24)
Yes AD, YES DNR order 0.36 0.16, 0.78 .010
(0.40)
Random Intercept 0.04 0.02, 0.07 <.001 0.13 0.06, 0.30 <.001
(0.31) (0.42)
Model Summary
Level 2 Variance 1.18 (0.09) 1.46 (0.11)
Akaike Information Criterion
Bayesian Information Criterion
Patient Observations 3,480 3,475
Agency Observations 649 648
Footnotes Reference Variables
CHF: Congestive Heart Failure White (Race)
AD: Advance Directive Married/Living with Partner (Marital Status)
DNR: Do Not Resuscitate Cancer (Primary Diagnosis)
No Advance Directive (Advance Care Planning)
201953.59
202002.82 205255.66
205206.42

Step 2: Inclusion of advance care planning decisions. To determine the relative
effect of advance care planning on emergent care utilization, a second model was
estimated. Results are presented under Step 2 in Table 12. The second adjusted model
71
(Level 1 n = 3,475; Level 2 n = 648) was an improved fit over the first model (AIC =
201953.59; BIC = 202002.82). Consistent with what was hypothesized, patients with an
advance directive (OR = 0.47; p = .002) or DNR order (OR = 0.36; p = .010) were
significantly less likely than those without an advance directive to utilize emergent care.
Black patients (OR = 0.68; p = .031) were also found to be less likely than White patients
to utilize emergent care, and Hispanic patients (OR = 0.70; p = .117) were equally as
likely as White patients to utilize emergent care. Additionally, widowed patients (OR =
1.23; p = .018), and those with Congestive Heart Failure/Heart Disease (OR = 1.80; p <
.001) were significantly more likely than the associated referents to utilize emergent care.
Older patients (OR = 0.99; p < .001) were also less likely to utilize emergent care,
compared to younger patients. Finally, following inclusion of care planning covariates,
Black race/ethnicity, comorbidity count, and Medicaid enrollment all lost significance,
whereas Hispanic race/ethnicity gained significance. See Figure 9 for a blobbogram
illustrating the relative effect of each model 2 predictor for emergent care utilization.

72

Footnotes
CHF: Congestive Heart Failure
HD: Heart Disease
DNR: Do Not Resuscitate Order
Reference Variables
White (Race)
Married/Living with Partner (Marital Status)
Cancer (Primary Diagnosis)
No Advance Directive (Advance Care Planning)
Figure 9. Relative Effects of Predictors for Emergent Care Utilization
0.1
1
10
* Documented DNR
** No Documented DNR
Medicaid Enrollee
Comorbidity Count
Other Primary Diagnosis
Neurological Diseases
Lung Disease
*** CHF/HD
Single/Divorced/Separated
* Widowed
Hispanic
*** Black
Female
*** Age
Odds Ratio
Relative Effect Size
73
Chapter 7: Hospice Length of Stay
In this chapter, bivariate comparisons by hospice length of stay (LOS) are
presented, multivariable results are reported, and posthoc ancillary bivariate comparisons
and multivariable models are also presented. Bivariate tests of the LOS hypothesis were
conducted using the decedent analytic sample, whereas multivariable LOS survival
models utilized the full sample, censoring patients who did not die under the care of
hospice. Results of the initial analyses testing for differences in the total days of care
before death are presented first, followed two ancillary binary outcomes that were
subsequently investigated: 1) hospice LOS: 0-7 days (1 week), and 2) hospice LOS: 0-30
days (1 month). Recent data indicates that approximately 36% of U.S. hospice patients
die within the first week of care, and 63% die within the first month (National Hospice
and Palliative Care Organization, 2013). Given that shorter hospice stays limit the full
benefits of care (Byock et al., 1996; Christakis & Iwashyna, 1998; Rickerson et al.,
2005), time widows for ancillary analyses were selected to test for racial/ethnic variation
in shorter lengths of stay among decedents. Previous studies investigating key hospice
LOS outcomes have also investigated similar hospice LOS time windows (Sengupta et
al., 2013). Frequencies and relative percentages of missing and excluded data for LOS
analyses can be found in Appendix H.
Bivariate Comparisons
Hospice LOS was examined using continuous outcome data on the total days or
hospice care received prior to death. A summary of decedent characteristics by LOS is
presented in Table 13. Additional data on median differences in LOS among the decedent
sample is also available in Appendix I.
74
Bivariate analyses indicated no significant differences in decedent LOS by
race/ethnicity (F(1.00, 630.00) = 0.03; p= .862). Contrary to the study hypothesis, days of
care prior to death were not significantly greater for Black (59.53 ± 148.48) and Hispanic
decedents (60.99 ± 92.90), relative to White decedents (58.43 ± 120.52). Significant
differences in LOS were observed by age (F(3.00, 628.00) = 3.91; p = .009), gender
(F(1.00, 630.00) = 6.50; p = .011), primary diagnosis (F(1.00, 630.00) = 6.06; p = .014),
and Medicaid enrollment (F(1.00, 628.00) = 4.66; p = .031). Specifically, LOS before
death was longer for 1) decedents in the oldest age group (77.27 ± 141.71), relative to the
youngest age group (47.49 ± 107.32), 2) females (66.26 ± 131.95), relative to males
(48.53 ± 105.57), 3) decedents with Lung Disease (71.90 ± 140.11), or a neurological
disease (88.93 ± 164.91), relative to decedents with Cancer (44.07 ± 94.61), or
Congestive Heart Failure/Heart Disease (58.59 ± 109.57), and 4) decedents enrolled in
Medicaid (75.60 ± 125.73), relative to decedents not enrolled in Medicaid (54.59 ±
117.30). No significant differences in hospice LOS were observed by marital status
(F(1.00, 627.00) = 0.37; p= .545), comorbidity count (F(1.00, 630.00) = 0.50; p = .480),
or advance care planning (F(1.00, 628.00) = 3.70; p = .055).

75
n Mean ± SD Sig.
Age .009
65-75 144,880 47.49 ± 107.32
76-82 150,860 54.63 ± 110.68
83-87 147,548 48.76 ± 113.43
88+ 195,124 77.27 ± 141.71
Gender .011
Female 361,604 66.26 ± 131.95
Male 276,808 48.53 ± 105.57
Race/Ethnicity .862
White 583,124 58.43 ± 120.52
Black 36,304 59.53 ± 148.48
Hispanic 18,984 60.99 ± 92.90
Marital Status .545
Married/Living Together 262,216 50.08 ± 106.77
Widowed 286,640 66.43 ± 130.72
Divorced/Separated 31,956 32.73 ± 81.99
Never Married 24,920 51.25 ± 101.72
Missing 32,680 88.65 ± 174.40
Primary Diagnosis .014
Cancer 247,300 44.07 ± 94.61
CHF/Heart Disease 80,536 58.59 ± 109.57
Lung Disease 96,788 71.90 ± 140.11
Neurological Diseases 133,176 88.93 ± 164.91
Other 79,748 36.55 ± 76.83
Missing 864 69.41 ± 227.24
Comorbidity Count .480
0-1 174,004 57.11 ± 135.21
2-3 194,320 53.03 ±108.36
4+ 270,088 63.50 ± 121.05
Enrolled in Medicaid .031
Yes 95,952 75.60 ± 125.73
No 534,904 54.59 ± 117.30
Missing 7,556 124.39 ± 258.14
Advance Care Planning .055
No AD, No DNR order 39,456 47.85 ± 94.78
Yes AD, No DNR order 47,744 42.96 ± 105.51
Yes AD, Yes DNR order 542,360 59.29 ± 121.12
Missing 8,852 146.67 ± 246.69
Footnotes
CHF: Congestive Heart Failure
AD: Advance Directive
DNR: Do Not Resuscitate
Characteristics
Length of Stay: Full Episode
Table 13. Decedent Hospice Length of Stay (Full Care Episode)
Bivariate Comparisons (n =638,412)

Log-rank tests. Differences in time-to-death between 1) White and Black
decedents (n = 2,825), 2) White and Hispanic decedents (n = 2,765), and 3) Black and
Hispanic decedents (n = 212), were also tested using the log-rank test for equality of
76
survivor functions. Results of bivariate mean comparisons further indicated no significant
difference in time-to-death between White and Black decedents (χ
2
= 0.90; p=.344;
Figure 10), White and Hispanic decedents (χ
2
= 0.32; p=.574; Figure 11), or Black and
Hispanic decedents (χ
2
= 0.03; p=.864; Figure 12),

0.00 0.25 0.50 0.75 1.00
Percent Survival
0 100 200 300 400 500 600 700 800 900 1000 1100
Days of Hospice Care
White Decedents Black Decedents
Kaplan-Meier Survival Estimates
Figure 10. Unadjusted Survival Curves for White and Black Hospice
Decedents
77

0.00 0.25 0.50 0.75 1.00
Percent Survival
0 100 200 300 400 500 600 700 800 900 1000 1100
Days of Hospice Care
White Decedents Hispanic Decedents
Kaplan-Meier survival estimates
Figure 11. Unadjusted Survival Curves for White and Hispanic Hospice
Decedents
0.00 0.25 0.50 0.75 1.00
Percent Survival
0 100 200 300 400 500 600 700 800 900 1000 1100
Days of Hospice Care
Hispanic Decedents Black Decedents
Kaplan-Meier survival estimates
Figure 12. Unadjusted Survival Curves for Black and Hispanic Hospice
Decedents
78
Multivariable Models
Racial/Ethnic differences in hospice LOS were analyzed using a two-step Cox
Proportional Hazards multivariable model. As with emergent care utilization
multivariable analyses, the first step adjusted for key demographic and health indicators,
whereas the second step introduced additional predictors pertaining to care planning.
Both steps were employed to test relationships proposed in Hypothesis 3, and patients
who were discharged for any reason other than death were censored in survival analysis
models.
Step 1: Demographics and health indicator variables. To test if hospice LOS
was longer for Black and Hispanic decedents, compared to White decedents, a Cox
Proportional Hazards regression was conducted. Results of this model can be found under
Step 1 in Table 14. The adjusted model was significant (Level 1 n = 3,400; Level 2 n =
652; F(12.00, 632.00) = 6.13; p < .001). Contrary to study hypotheses, neither Black (HR
= 0.82; p = .216) nor Hispanic (HR = 0.86; p = .318) decedents were more likely to
experience a longer LOS than White decedents prior to death. Decedents with Lung
Disease (HR = 0.75; p = .003), a neurological disease (HR = 0.72; p < .001), or those
enrolled in Medicaid (HR = 0.79; p = .008) were at significantly reduced risk of dying
before decedents with Cancer. Similarly, older decedents (HR = 0.99; p = .001) were at a
reduced risk of an earlier death during the hospice stay, compared to younger decedents.

79
Table 14. Predictors of Decedent Hospice Length of Stay: Full Care Episode
Step 1 Step 2
HR HR
(Std. Err.) 95% CI Sig. (Std. Err.) 95% CI Sig.
Model Predictors
Race/Ethnicity
Black 0.82 0.60, 1.12 .216 0.86 0.63, 1.16 .325
(0.16) (0.16)
Hispanic 0.86 0.64, 1.16 .318 0.85 0.63, 1.15 .279
(0.15) (0.15)
Age 0.99 0.98, 1.00 .001 0.99 0.98, 0.99 <.001
(0.00) (0.00)
Female 0.91 0.80, 1.03 .145 0.92 0.80, 1.04 .187
(0.07) (0.07)
Marital Status
Widowed 0.98 0.92, 1.05 .563 0.97 0.91, 1.04 .409
(0.03) (0.03)
Single/Divorced/Separated 1.00 0.79, 1.27 .996 1.01 0.80, 1.27 .959
(0.12) (0.12)
Primary Diagnosis
CHF/Heart Disease 0.82 0.66, 1.01 .065 0.84 0.68, 1.04 .106
(0.11) (0.11)
Lung Disease 0.75 0.62, 0.90 .003 0.74 0.61, 0.90 .002
(0.10) (0.10)
Neurological Disease 0.72 0.60, 0.86 <.001 0.73 0.61, 0.88 .001
(0.09) (0.10)
Other Primary Diagnosis 0.92 0.73, 1.15 .458 0.92 0.73, 1.16 .481
(0.11) (0.12)
Comorbidity Count 1.00 0.98, 1.03 .812 1.00 0.98, 1.03 .881
(0.01) (0.01)
Medicaid Enrollee 0.79 0.66, 0.94 .008 0.78 0.66, 0.93 .007
(0.09) (0.09)
Advance Care Planning
Yes AD, No DNR order 1.38 0.92, 2.09 .122
(0.21)
Yes AD, YES DNR order 1.35 0.98, 1.85 .063
(0.16)
Model Summary
Survey-Adjusted F-Statistic 6.13 5.61
Design Degrees of Freedom 643.00 641
Model Significance <.001
Patient Observations 3,400 3,392
Agency Observations 652 650
Footnotes Reference Variables
CHF: Congestive Heart Failure White (Race)
AD: Advance Directive Married/Living with Partner (Marital Status)
DNR: Do Not Resuscitate Cancer (Primary Diagnosis)
No Advance Directive (Advance Care Planning)
<.001

Step 2: Inclusion of advance care planning decisions. To determine the relative
effect of advance care planning on decedent hospice LOS, a second model was estimated
(see Step 2 in Table 14). The adjusted model was significant (Level 1 n = 3,392; Level 2
80
n = 650; F(14.00, 628.00) = 5.61; p < .001). As with the first model, neither Black (HR =
0.86; p = .325) nor Hispanic decedents (HR = 0.85; p = .279) were more likely to
experience a longer LOS prior to death, compared to White decedents. Those with Lung
Disease (HR = 0.74; p = .002), a neurological disease (HR = 0.73; p = .001), or those
enrolled in Medicaid (HR = 0.78; p = .007) were at significantly reduced risk of dying
before decedents with Cancer. Age (HR = 0.99; p < .001) was also found to be negatively
associated with an earlier death. Lastly, following inclusion of care planning covariates,
none of the variables lost or gained significance.
Ancillary Bivariate Comparisons
To further investigate racial/ethnic variation in hospice LOS, two ancillary binary
outcomes were examined within the original LOS variable: 1) hospice LOS: 0-7 days (1
week; coded as positive/negative indicator), and 2) hospice LOS: 0-30 days (1 month;
coded as positive/negative indicator). The ancillary outcomes were analyzed to determine
if there were racial/ethnic differences in shorter hospice stays among decedents. As with
prior LOS analyses, the decedent sample was utilized for ancillary analyses. For
frequencies and relative percentages of missing and excluded data for ancillary LOS
analyses, see Appendices J (0-7 days) and K (0-30 days).
Hospice length of stay: 0-7 Days. A summary of decedent characteristics by a
hospice LOS of seven days or less is presented in Table 15. Bivariate analyses indicated
no significant differences in death within the first week by race/ethnicity (F(1.89,
1193.79) = 0.74; p = .472), with 39% of White decedents, 31% of Black decedents, and
33% of Hispanic decedents enrolled in hospice for seven days or less prior to death. No
significant differences in decedent LOS of 0-7 days were observed by age (F(2.98,
81
1876.31) = 2.42; p = .065), gender (F(1.00, 630.00) = 0.37; p = .542), marital status
(F(2.95, 1849.34) = 1.27; p = .283), primary diagnosis (F(3.73, 2351.23) = 0.68; p =
.599), comorbidity count (F(1.96, 1235.92) = 0.64; p = .523), Medicaid enrollment
(F(1.00, 628.00) = 3.82; p = .051), or advance care planning (F(1.94, 1220.85) = 0.30; p
= .733).

82
No Yes Sig.
Age .065
65-75 86,480 (59.69) 58,400 (40.31)
76-82 94,544 (62.67) 56,316 (37.33)
83-87 81,284 (55.09) 66,264 (44.91)
88+ 128,756 (65.99) 66,368 (34.01)
Gender .542
Female 166,252 (60.06) 110,556 (39.94)
Male 224,812 (62.17) 136,792 (37.83)
Race/Ethnicity .472
White 353,252 (60.58) 229,872 (39.42)
Black 25,084 (69.09) 11,220 (30.91)
Hispanic 12,728 (67.05) 6,256 (32.95)
Marital Status .283
Married/Living Together 156,908 (59.84) 105,308 (40.16)
Widowed 181,028 (63.16) 105,612 (36.84)
Divorced/Separated 16,288 (50.97) 15,668 (49.03)
Never Married 13,792 (55.35) 11,128 (44.65)
Missing 23,048 (70.53) 9,632 (29.47)
Primary Diagnosis .599
Cancer 156,852 (63.43) 90,448 (36.57)
CHF/Heart Disease 46,412 (57.63) 34,124 (42.37)
Lung Disease 60,076 (62.07) 36,712 (37.93)
Neurological Diseases 81,892 (61.49) 51,284 (38.51)
Other 45,016 (56.45) 34,732 (43.55)
Missing 816 (94.44) 48 (5.56)
Comorbidity Count .523
0-1 105,876 (60.85) 68,128 (39.15)
2-3 114,652 (59.00) 79,668 (41.00)
4+ 170,536 (63.14) 99,552 (36.86)
Enrolled in Medicaid .051
Yes 65,908 (68.69) 30,044 (31.31)
No 320,420 (59.90) 214,484 (40.10)
Missing 4,736 (62.68) 2,820 (37.32)
Advance Care Planning .733
No AD, No DNR order 22,288 (56.49) 17,168 (43.51)
Yes AD, No DNR order 29,392 (61.56) 18,352 (38.44)
Yes AD, Yes DNR order 332,020 (61.22) 210,340 (38.78)
Missing 7,364 (83.19) 1,488 (16.81)
Footnotes
CHF: Congestive Heart Failure
AD: Advance Directive
DNR: Do Not Resuscitate
Percentages are presented by row
Characteristics, No. (%)
Length of Stay: 0-7 Days
Table 15. Decedent Hospice Length of Stay (0-7 Days) Bivariate Comparisons
(n = 638,412)

Hospice length of stay: 0-30 Days. A summary of decedent characteristics by a
hospice LOS one month or less can be found in Table 16. Bivariate analyses indicated no
significant differences in a hospice LOS of 0-30 days by race/ethnicity (F(1.96, 1234.62)
= 2.63; p = .074), with 66% of White decedents, 76% of Black decedents, and 65% of
83
Hispanic decedents dying within the first thirty days of hospice care. Significant
differences in a hospice LOS of 0-30 days were observed by age (F(2.89, 1822.58) =
3.03; p = .030), marital status (F(2.92, 1829.66) = 2.64; p = .049), primary diagnosis
(F(3.78, 2383.51) = 2.70; p = .032), and Medicaid enrollment (F(1.00, 628.00) = 3.87; p
= .049). Specifically, rates of dying within the first thirty days of hospice care were lower
among 1) decedents in the oldest age group (60%), relative to the youngest age group
(68%), 2) widowed decedents (64%), relative to decedents who were married or living
with a partner (69%), divorced or separated (79%), or never married (73%), and 3)
decedents enrolled in Medicaid (60%), relative to decedents who were not enrolled in
Medicaid (67%). Additionally, rates of dying within the first thirty days of hospice care
were higher among decedents with Cancer (69%) and other primary diagnoses (75%),
relative to decedents with Congestive Heart Failure/Heart Disease (63%), Lung Disease
(60%), or a neurological disease (61%). No significant differences in a hospice LOS of 0-
30 days were observed by gender (F(1.00, 630.00) = 2.57; p = .109), comorbidity count
(F(1.98, 1246.76) = 2.61; p = .075), or advance care planning (F(1.90, 1190.44) = 0.26; p
= .763).

84
No Yes Sig.
Age 0.030
65-75 45,828 (31.63) 99,052 (68.37)
76-82 50,548 (33.51) 100,312 (66.49)
83-87 41,284 (27.89) 106,264 (72.02)
88+ 78,420 (40.19) 116,704 (59.81)
Gender 0.109
Female 85,952 (31.05) 190,856 (68.95)
Male 130,128 (35.99) 231,476 (64.01)
Race/Ethnicity 0.074
White 200,956 (34.46) 382,168 (65.54)
Black 8,540 (23.52) 27,764 (76.48)
Hispanic 6,584 (34.68) 12,400 (65.32)
Marital Status 0.049
Married/Living Together 81,348 (31.02) 180,868 (68.98)
Widowed 104,348 (36.40) 182,292 (63.60)
Divorced/Separated 6,800 (21.28) 25,156 (78.72)
Never Married 6,736 (27.03) 18,184 (72.97)
Missing 16,848 (51.55) 15,832 (48.45)
Primary Diagnosis 0.032
Cancer 76,204 (30.81) 171,096 (69.19)
CHF/Heart Disease 30,112 (37.39) 50,424 (62.61)
Lung Disease 38,300 (39.57) 58,488 (60.43)
Neurological Diseases 51,748 (38.86) 81,428 (61.14)
Other 19,592 (24.57) 60,156 (75.43)
Missing 124 (14.35) 740 (85.65)
Comorbidity Count 0.075
0-1 53,936 (31.00) 120,068 (69.00)
2-3 58,812 (30.27) 135,508 (69.73)
4+ 103,332 (38.26) 166,756 (61.74)
Enrolled in Medicaid 0.049
Yes 38,844 (40.48) 57,108 (59.52)
No 174,328 (32.59) 360,576 (67.41)
Missing 2,908 (38.49) 4,648 (61.51)
Advance Care Planning 0.763
No AD, No DNR order 12,136 (30.76) 27,320 (69.24)
Yes AD, No DNR order 14,712 (30.81) 33,032 (69.19)
Yes AD, Yes DNR order 182,596 (33.67) 359,764 (66.33)
Missing 6,636 (74.97) 2,216 (25.03)
Footnotes
CHF: Congestive Heart Failure
AD: Advance Directive
DNR: Do Not Resuscitate
Percentages are presented by row
Characteristics, No. (%)
Length of Stay: 0-30 Days
Table 16. Decedent Hospice Length of Stay (0-30 Days) Bivariate Comparisons
(n = 638,412)

Ancillary Multivariable Models
Racial/Ethnic variation in binary hospice LOS outcomes were analyzed using
multivariable models conducted in two waves: 1) death in the first week (0-7 days), and
85
2) death in the first month (0-30 days). As with previous study analyses, two-step
multilevel models were estimated to first adjust for key demographic and health
indicators, followed by the inclusion of additional care planning predictors. All models
included patient-level (level 1) predictors with random agency-level (level 2) intercepts
to account for between-agency variation.
Hospice length of stay: 0-7 days. Preliminary analyses of the null model
(agency-level n = 639; patient-level n = 2,901) found that 5% of the observed variation in
death within the first week was due to between-agency differences (ICC = 0.047, 95%
CI: 0.021, 0.100; χ
2
= 27.21, df = 1; p <.001). Despite the low Intraclass Correlation
Coefficient, previous research has suggested that between-level 2 unit variation should
still be accounted for in study design (Nezlek, 2008) and thus, the hierarchical analytical
design was modeled to the data.
Step 1: Demographics and health indicator variables. To test for racial/ethnic
variation in dying within the first seven days of hospice care, a multilevel logistic
regression was conducted. Results of this model are presented under Step 1 in Table 17.
The adjusted model (Level 1 n = 2,776; Level 2 n = 634), revealed an AIC fit criteria of
616693.63, and a BIC fit criteria of 616746.99. Hispanic decedents (OR = 1.15; p = .036)
were significantly more likely than White decedents to die within the first seven days of
hospice care. Conversely, Black decedents (OR = 0.71; p = .019) were significantly less
likely than White decedents to die within the first week. Single/Divorced/Separated
decedents (OR = 1.37; p = .008), and those with Congestive Heart Failure/Heart Disease
(OR = 1.35; p < .001), a neurological disease (OR = 1.53; p < .001), or other primary
diagnoses (OR = 1.97; p = .007) were significantly more likely than the associated
86
referents to die within the first seven days of hospice care. Lastly, age was negatively
associated with dying within the first seven days of hospice care (OR = 0.98; p < .001).

Table 17. Predictors of Decedent Hospice Length of Stay: 0-7 Days
Step 1 Step 2
OR OR
(Std. Err.) 95% CI Sig. (Std. Err.) 95% CI Sig.
Predictors
Race/Ethnicity
Black 0.71 0.54, 0.95 .019 0.78 0.62, 0.97 .027
(0.14) (0.12)
Hispanic 1.15 1.01, 1.32 .036 1.18 1.07, 1.30 .001
(0.07) (0.05)
Age 0.98 0.97, 0.98 <.001 0.98 0.98, 0.99 <.001
(0.00) (0.00)
Female 0.96 0.85, 1.09 .552 0.99 0.85, 1.15 .886
(0.06) (0.08)
Marital Status
Widowed 0.93 0.69, 1.26 .635 0.93 0.69, 1.27 .658
(0.15) (0.16)
Single/Divorced/Separated 1.37 1.09, 1.73 .008 1.37 1.17, 1.59 <.001
(0.12) (0.08)
Primary Diagnosis
CHF/Heart Disease 1.35 1.14, 1.60 <.001 1.37 1.10, 1.71 .005
(0.09) (0.11)
Lung Disease 1.41 0.90, 2.20 .138 1.38 1.00, 1.90 .050
(0.23) (0.16)
Neurological Disease 1.53 1.43, 1.64 <.001 1.53 1.41, 1.66 <.001
(0.03) (0.04)
Other Primary Diagnosis 1.97 1.20, 3.22 .007 2.06 1.24, 3.43 .005
(0.25) (0.26)
Comorbidity Count 1.01 0.99, 1.03 .531 0.98 0.96, 1.01 .126
(0.01) (0.01)
Medicaid Enrollee 0.82 0.66, 1.00 .054 0.75 0.58, 0.97 .031
(0.11) (0.13)
Advance Care Planning
Yes AD, No DNR order 0.61 0.43, 0.87 .007
(0.18)
Yes AD, YES DNR order 0.61 0.51, 0.73 <.001
(0.09)
Random Intercept 2.71 1.47, 4.98 .001 4.04 1.94, 8.42 <.001
(0.31) (0.38)
Model Summary
Level 2 Variance 0.80 (0.11) 0.79 (0.08)
Akaike Information Criterion
Bayesian Information Criterion
Patient Observations 2,776 2,771
Agency Observations 634 633
Footnotes Reference Variables
CHF: Congestive Heart Failure White (Race)
AD: Advance Directive Married/Living with Partner (Marital Status)
DNR: Do Not Resuscitate Cancer (Primary Diagnosis)
No Advance Directive (Advance Care Planning)
616693.63
616746.99
614667.00
614714.42

87
Step 2: Inclusion of advance care planning decisions. The second multilevel
model predicting death within the first seven days of hospice care added advance care
planning decisions to the previous model (see Step 2 in Table 17). The second adjusted
model (Level 1 n = 2,771; Level 2 n = 633), indicated an improved fit over the first
model (AIC = 614667.00; BIC = 614714.42). As with the first model, Hispanic decedents
(OR = 1.18; p < .001) were significantly more likely than White decedents to die within
the first week of hospice care, and Black decedents (OR = 0.78; p = .027) were
significantly less likely than White decedents to die within the first week. The change in
effect size between the first and second models was 0.71 vs. 0.78 for Black decedents,
and 1.15 vs. 1.18 for Hispanic decedents. Similarly, single/divorced/separated decedents
(OR = 1.37; p < .001), and those with Congestive Heart Failure/Heart Disease (OR =
1.37; p = .005), a neurological disease (OR = 1.53; p < .001), or other primary diagnoses
(OR = 2.06; p = .005) were significantly more likely than the associated referents to die
within the first seven days of hospice care. Conversely, decedents enrolled in Medicaid
(OR = 0.75; p = .031), and those with an advance directive (OR = 0.61; p = .007), or do
not resuscitate (DNR) order (OR = 0.61; p < .001) were significantly less likely than the
associated referents to die within the first seven days of hospice care. Age (OR = 0.98; p
< .001) was also negatively associated with dying within the first seven days of hospice
care. Finally, Medicaid enrollment was the only predictor that changed significance
between the two models. See Figure 13 for a blobbogram illustrating the relative effect of
each Step 2 predictor for dying within the first seven days of hospice care.

88

Hospice length of stay: 0-30 days. Analysis of the null model (agency-level n =
639; patient-level n = 2,901) suggested that 3% of the variance in dying within the first
thirty days of care was due to between-agency variation (ICC = 0.026 (95% CI: 0.007,
0.093; χ
2
= 13.93, df = 1; p <.001). As with the previous model, random effects modeling
was maintained.
Step 1: Demographics and health indicator variables. To test for racial/ethnic
variation in dying within the first thirty days of hospice care, a multilevel logistic
regression was conducted. Results of this model are presented under Step 1 in Table 18.
The adjusted model (Level 1 n = 2,776; Level 2 n = 634), revealed AIC and BIC model fit
criteria were 578290.89, and 578338.32, respectively. While Black decedents (OR =
2.55; p < .001) were significantly more likely than White decedents to die in the first
Footnotes
CHF: Congestive Heart Failure
HD: Heart Disease
DNR: Do Not Resuscitate Order
Reference Variables
White (Race)
Married/Living with Partner (Marital Status)
Cancer (Primary Diagnosis)
No Advance Directive (Advance Care Planning)
Figure 13. Relative Effects of Predictors for Decedent Length of Stay: 0-7 Days
0.1
1
10
*** Documented DNR
** No Documented DNR
* Medicaid Enrollee
Comorbidity Count
** Other Primary Diagnosis
*** Neurological Diseases
Lung Disease
** CHF/HD
*** Single/Divorced/Separated
Widowed
** Hispanic
* Black
Female
*** Age
Odds Ratio
Relative Effect Size
89
thirty days of hospice care, Hispanic decedents (OR = 1.19; p = .278) were equally likely
as White decedents to die during the first month. Decedents with other primary diagnoses
(OR = 1.70; p = .002) were significantly more likely than those with Cancer to die in the
first thirty days of hospice care. Conversely, female decedents (OR = 0.84; p = .041), and
those with Lung Disease (OR = 0.73; p = .024) were significantly less likely than the
associated referents to die within the first month of hospice care. Lastly, older decedents
(OR = 0.98; p = .034), and those with more comorbidities (OR = 0.90; p < .001) were less
likely to die during the first month of hospice care.

90
Table 18. Predictors of Decedent Hospice Length of Stay: 0-30 Days
Step 1 Step 2
OR OR
(Std. Err.) 95% CI Sig. (Std. Err.) 95% CI Sig.
Predictors
Race/Ethnicity
Black 2.55 1.64, 3.96 <.001 2.54 1.62, 3.99 <.001
(0.23) (0.23)
Hispanic 1.19 0.87, 1.62 .278 1.20 0.94, 1.52 .148
(0.16) (0.12)
Age 0.98 0.96, 1.00 .034 0.98 0.96, 0.99 .007
(0.01) (0.01)
Female 0.84 0.71, 0.99 .041 0.80 0.66, 0.97 .021
(0.09) (0.10)
Marital Status
Widowed 0.99 0.77, 1.27 .911 1.00 0.76, 1.31 .981
(0.13) (0.14)
Single/Divorced/Separated 1.56 1.00, 2.44 .050 1.55 1.04, 2.32 .032
(0.23) (0.21)
Primary Diagnosis
CHF/Heart Disease 0.65 0.40, 1.08 .094 0.69 0.40, 1.18 .175
(0.25) (0.28)
Lung Disease 0.73 0.56, 0.96 .024 0.86 0.74, 1.01 .067
(0.14) (0.08)
Neurological Disease 0.84 0.62, 1.13 .253 0.91 0.70, 1.18 .476
(0.15) (0.13)
Other Primary Diagnosis 1.70 1.22, 2.35 .002 1.71 1.21, 2.40 .002
(0.17) (0.17)
Comorbidity Count 0.90 0.88, 0.93 <.001 0.89 0.85, 0.92 <.001
(0.02) (0.02)
Medicaid Enrollee 0.97 0.70, 1.34 .838 0.96 0.72, 1.29 .800
(0.17) (0.15)
Advance Care Planning
Yes AD, No DNR order 0.99 0.68, 1.45 .958
(0.19)
Yes AD, YES DNR order 0.51 0.31, 0.83 .007
(0.25)
Random Intercept 31.45 6.49, 152.43 <.001 74.48 24.67, 224.87 <.001
(0.81) (0.56)
Model Summary
Level 2 Variance 0.82 (0.09) 0.73 (0.08)
Akaike Information Criterion 578290.89 576066.63
Bayesian Information Criterion 578338.32 576114.04
Patient Observations 2,776 2,771
Agency Observations 634 633
Footnotes Reference Variables
CHF: Congestive Heart Failure White (Race)
AD: Advance Directive Married/Living with Partner (Marital Status)
DNR: Do Not Resuscitate Cancer (Primary Diagnosis)
No Advance Directive (Advance Care Planning)

Step 2: Inclusion of advance care planning decisions. The second multilevel
model predicting death within the first thirty days of hospice care included advance care
planning covariates. Results from this model are presented under Step 2 in Table 18. The
91
second adjusted model (Level 1 n = 2,771; Level 2 n = 633) demonstrated improved fit
over the first model (AIC = 576066.63; BIC = 576114.04). Black decedents (OR = 2.54; p
< .001) were still found to be more likely than White decedents to die in the first thirty
days of hospice care, and Hispanic decedents (OR = 1.20; p = .148) remained as likely as
White decedents to die in the first month of care. The relative effect size difference
between the first and second models for Black decedents was 2.55 vs. 2.54.
Single/Divorced/Separated decedents (OR = 1.55; p = .032), and those with other primary
diagnoses (OR = 1.71; p = .002) were significantly more likely than the associated
referents to die within the first thirty days of hospice care. Female decedents (OR = 0.80;
p = .021), and those with a DNR order (OR = 0.51; p = .007) were significantly less
likely than the associated referents to die within the first month of hospice care. As with
the first model, age (OR = 0.98; p = .007), and comorbidity count (OR = 0.89; p < .001)
remained negatively associated with dying within the first thirty days of hospice care.
Lastly, the primary diagnosis of Lung Disease lost significance following inclusion of
care planning covariates, and the marital status of single/divorced/separated gained
significance. For a blobbogram illustrating the relative effect of each Step 2 predictor for
dying within the first thirty days of hospice care, see Figure 14.

92

Footnotes
CHF: Congestive Heart Failure
HD: Heart Disease
DNR: Do Not Resuscitate Order
Reference Variables
White (Race)
Married/Living with Partner (Marital Status)
Cancer (Primary Diagnosis)
No Advance Directive (Advance Care Planning)
Figure 14. Relative Effects of Predictors for Decedent Length of Stay: 0-30 Days
0.1
1
10
** Documented DNR
No Documented DNR
Medicaid Enrollee
*** Comorbidity Count
** Other Primary Diagnosis
Neurological Diseases
Lung Disease
CHF/HD
* Single/Divorced/Separated
Widowed
Hispanic
*** Black
* Female
** Age
Odds Ratio
Relative Effect Size
93
Chapter 8: Site of Death
In this chapter, bivariate comparisons by site of death (SOD) are presented, and
multivariable results are reported. To test the hypothesis associated with SOD, the
decedent analytic sample was utilized. For frequencies and relative percentages of
missing and excluded data for SOD analyses, see Appendices L (home-like setting) and
M (hospital).
Overall Site of Death
A descriptive summary of decedent SOD by race/ethnicity is presented in Table
19. Inpatient hospice agencies were excluded from study analyses due to
operationalization concerns with the survey questionnaire; however, a full summary of
frequencies and associated percentages of decedent SOD by race/ethnicity, including
inpatient hospice agencies, is provided in Appendix N. Forty-three percent of decedents
died in a private home or apartment, 26% died in a nursing home or skilled nursing
facility, 9% died in a hospital, 5% died in a residential care place (i.e., assisted living
facility, board and care home, continuing care retirement community), and 1% died in
other locations. No significant differences in SOD by race/ethnicity were observed
(F(5.40, 3340.02) = 0.55; p = .749).

94
Table 19. Overall Site of Death (n = 667,820)
Total White Black Hispanic Sig.
.749
Private Home/Apartment 289,900 (43.41) 262,260 (42.94) 18,392 (49.61) 9,248 (46.42)
Residental Care Place 36,596 (5.48) 36,008 (5.90) 588 (1.59) 0 (0.00)
Nursing Home/SNF 176,572 (26.44) 162,232 (26.56) 9,316 (25.13) 5,024 (25.22)
Hospital 57,236 (8.57) 52,744 (8.63) 2,280 (6.15) 2,212 (11.10)
Other 492 (0.07) 492 (0.08) 0 (0.00) 0 (0.00)
Missing 107,024 (16.03) 97,084 (15.89) 6,500 (17.53) 3,440 (17.27)
Footnotes
SNF: Skilled Nursing Facility
Percentages are presented by column

Bivariate Comparisons
Hospice decedent SOD was examined using two dependent variables: 1) death in
a home-like setting (i.e., private home or apartment, and residential care place), and 2)
death in a hospital. As such, bivariate comparisons are presented below separately for
each dependent variable.
Site of death: Home-like setting. A summary of patient characteristics by death
in a home-like setting is presented in Table 20. Bivariate findings were discordant with
the hypothesis, as rates of death in an home-like setting were not significantly higher for
White decedents (58%), relative to Black (62%) or Hispanic decedents (56%; F(1.57,
972.13) = 0.15; p = .814). Analyses indicated significant differences in death in a home-
like setting by age (F(2.83, 1751.78) = 6.91; p < .001), gender (F(1.00, 618.00) = 13.97;
p < .001), marital status (F(2.93, 1799.80) = 12.13; p < .001), primary diagnosis (F(3.77,
2329.87) = 17.97; p < .001), Medicaid enrollment (F(1.00, 616.00) = 57.95; p < .001),
and advance care planning (F(1.93, 1191.30) = 4.15; p = .017). Specifically, rates of
dying in a home-life setting were lower for 1) decedents in the oldest age group (49%),
relative to the youngest age group (69%), 2) males (52%), relative to females (66%), 3)
95
widowed (51%), or never married decedents (39%), relative to decedents who were
married or living with a partner (69%), or divorced or separated (67%), 4) decedents with
Congestive Heart Failure/Heart Disease (55%), or Lung Disease (55%), relative to
decedents with Cancer (76%), a neurological disease (41%), or other primary diagnoses
(43%), 5) decedents enrolled in Medicaid (27%), relative to decedents not enrolled in
Medicaid (64%), and 6) decedents with an advance directive (74%), relative to decedents
without an advance directive (56%), or do not resuscitate (DNR) order (57%). No
significant differences in death in a home-like setting were observed by comorbidity
count (F(2.00, 1233.01) = 0.90; p = .406).

96
Table 20. Site of Death (Home-Like Setting) Bivariate Comparisons (n = 560,592)
No Yes Sig.
Age <.001
65-75 35,648 (30.70) 80,480 (69.30)
76-82 51,060 (38.67) 80,996 (61.33)
83-87 55,936 (41.88) 77,624 (58.12)
88+ 91,452 (51.13) 87,396 (48.87)
Gender <.001
Female 82,536 (34.06) 159,756 (65.94)
Male 151,560 (47.62) 166,740 (52.38)
Race/Ethnicity .814
White 215,264 (41.92) 298,268 (58.08)
Black 11,596 (37.93) 18,980 (62.07)
Hispanic 7,236 (43.90) 9,248 (56.10)
Marital Status <.001
Married/Living Together 68,996 (30.56) 156,780 (69.44)
Widowed 130,128 (49.30) 133,836 (50.70)
Divorced/Separated 8,508 (33.45) 16,928 (66.55)
Never Married 11,800 (61.26) 7,460 (38.73)
Missing 14,664 (55.12) 11,492 (44.88)
Primary Diagnosis <.001
Cancer 51,080 (24.38) 158,424 (75.62)
CHF/Heart Disease 32,188 (44.61) 39,972 (55.39)
Lung Disease 38,888 (44.80) 47,916 (55.20)
Neurological Diseases 72,988 (59.05) 50,624 (40.95)
Other 38,932 (57.16) 29,176 (42.84)
Missing 20 (4.95) 384 (95.05)
Comorbidity Count .406
0-1 66,376 (43.19) 87,304 (56.81)
2-3 64,116 (37.65) 106,192 (62.35)
4+ 103,604 (43.79) 133,000 (56.21)
Enrolled in Medicaid <.001
Yes 64,008 (72.85) 23,852 (27.15)
No 169,340 (36.03) 300,712 (63.97)
Missing 748 (27.91) 1,932 (72.09)
Advance Care Planning .017
No AD, No DNR order 15,960 (43.90) 20,392 (56.10)
Yes AD, No DNR order 11,300 (25.91) 32,320 (74.09)
Yes AD, Yes DNR order 206,808 (43.12) 272,848 (56.88)
Missing 28 (2.90) 936 (97.10)
Footnotes
CHF: Congestive Heart Failure
AD: Advance Directive
DNR: Do Not Resuscitate
Percentages are presented by row
Characteristics, No. (%)
Site of Death: Home

Site of death: Hospital. A summary of patient characteristics by death in a
hospital is presented in Table 21. As with death in a home-like setting, bivariate findings
were incongruent with the hypothesis, as no racial/ethnic variation in hospital death was
observed (F(1.85, 1142.84) = 0.41; p = .648). Specifically, rates of hospital death were
97
10% for White decedents, 7% for Black decedents, and 13% for Hispanic decedents.
Significant differences in hospital death were only observed by comorbidity count
(F(1.98, 1223.51) = 5.24; p = .006), with higher rates for decedents with 0-1
comorbidities (15%), relative to decedents with 2-3 comorbidities (9%), or four or more
comorbidities (8%). No significant differences in death in a hospital were observed by
age (F(2.92, 1803.74) = 1.91; p = .128), gender (F(1.00, 618.00) = 0.25; p = .615),
marital status (F(2.48, 1522.27) = 1.39; p = .248), primary diagnosis (F(3.76, 2326.06) =
1.80; p = .131), Medicaid enrollment (F(1.00, 616.00) = 3.23; p = .073), or advance care
planning (F(1.98, 1224.57) = 1.35; p = .259).

98
Table 21. Site of Death (Hospital) Bivariate Comparisons (n = 560,592)
No Yes Sig.
Age .128
65-75 100,184 (86.27) 15,944 (13.73)
76-82 116,236 (88.02) 15,820 (11.98)
83-87 121,508 (90.98) 12,052 (9.02)
88+ 165,428 (92.50) 13,420 (7.50)
Gender .615
Female 218,928 (90.36) 23,364 (9.64)
Male 284,428 (89.36) 33,872 (10.64)
Race/Ethnicity .648
White 460,788 (89.73) 52,744 (10.27)
Black 28,296 (92.54) 2,280 (7.46)
Hispanic 14,272 (86.58) 2,212 (13.42)
Marital Status .248
Married/Living Together 200,460 (88.79) 25,316 (11.21)
Widowed 240,212 (91.00) 23,752 (9.00)
Divorced/Separated 21,632 (85.04) 3,804 (14.96)
Never Married 18,288 (94.95) 972 (5.05)
Missing 22,764 (87.03) 3,392 (12.97)
Primary Diagnosis .131
Cancer 191,356 (91.34) 18,148 (8.66)
CHF/Heart Disease 65,012 (90.09) 7,148 (9.91)
Lung Disease 74,464 (85.78) 12,340 (14.22)
Neurological Diseases 114,104 (92.31) 9,508 (7.69)
Other 58,036 (85.21) 10,072 (14.79)
Missing 384 (95.05) 20 (4.95)
Comorbidity Count .006
0-1 129,988 (84.58) 23,692 (15.42)
2-3 155,252 (91.16) 15,056 (8.84)
4+ 218,116 (92.19) 18,488 (7.81)
Enrolled in Medicaid .073
Yes 82,412 (93.80) 5,448 (6.20)
No 418,332 (89.00) 51,720 (11.00)
Missing 2,612 (97.46) 68 (2.54)
Advance Care Planning .259
No AD, No DNR order 30,492 (83.88) 5,860 (16.12)
Yes AD, No DNR order 40,316 (92.43) 3,304 (7.57)
Yes AD, Yes DNR order 431,584 (89.98) 48,072 (10.02)
Missing 964 (100.00) 0 (0.00)
Footnotes
CHF: Congestive Heart Failure
AD: Advance Directive
DNR: Do Not Resuscitate
Percentages are presented by row
Characteristics, No. (%)
Site of Death: Hospital

Multivariable Analyses
Racial/Ethnic differences in SOD were investigated using multivariable analyses
conducted in two waves: 1) multilevel model predictions of dying in a home-like setting
(Hypothesis 4), and 2) multilevel model predictions of dying in a hospital (Hypothesis
99
4A). Both waves were conducted using two-steps multilevel models, with the first step
adjusting for key demographic and health indicators, and the second step further
adjusting for advance care planning decisions. All models included patient-level (level 1)
predictors with random agency-level (level 2) intercepts to account for between-agency
variation.
Site of death: Home-like setting. Preliminary multilevel analysis of the null
model (agency-level n = 627; patient-level n = 2,680) revealed that approximately 26%
of the observed variation in dying in a home-like setting was due to differences between
agencies (ICC = 0.259, 95% CI: 0.205, 0.321; χ
2
= 17.32, df = 1; p <.001).
Step 1: Demographics and health indicator variables. To test for racial/ethnic
variation in dying in a home-like setting, a multilevel logistic regression was conducted.
Results of this model are presented under Step 1 in Table 22. The adjusted model (Level
1 n = 2,570; Level 2 n = 622), revealed an AIC fit criteria of 409665.85, and a BIC fit
criteria of 409712.66. Model findings were inconsistent with what was hypothesized, as
Black (OR = 1.90; p = .019), and Hispanic decedents (OR = 1.21; p = .026) were both
significantly more likely than White decedents to die in a home-like setting. Conversely,
widowed decedents (OR = 0.69; p < .001), single/divorced/separated decedents (OR =
0.60; p = .020), those with Congestive Heart Failure/Heart Disease (OR = 0.40; p < .001),
Lung Disease, (OR = 0.55; p = .024), a neurological disease (OR = 0.14; p < .001), or
other primary diagnoses (OR = 0.25; p = .003), and decedents enrolled in Medicaid (OR
= 0.18; p < .001) were significantly less likely than the associated referents to die in a
home-like setting.

100
Table 22. Predictors of Site of Death: Home-Like Setting
Step 1 Step 2
OR OR
(Std. Err.) 95% CI Sig. (Std. Err.) 95% CI Sig.
Predictors
Race/Ethnicity
Black 1.90 1.11, 3.24 .019 2.14 1.27, 3.60 .004
(0.27) (0.27)
Hispanic 1.21 1.02, 1.43 .026 1.20 1.03, 1.39 .017
(0.09) (0.08)
Age 0.98 0.96, 1.00 .103 0.98 0.96, 1.00 .016
(0.01) (0.01)
Female 0.99 0.81, 1.21 .937 1.13 0.90, 1.41 .291
(0.10) (0.12)
Marital Status
Widowed 0.69 0.56, 0.84 <.001 0.70 0.58, 0.84 <.001
(0.10) (0.09)
Single/Divorced/Separated 0.60 0.39, 0.92 .020 0.57 0.33, 0.99 .047
(0.22) (0.29)
Primary Diagnosis
CHF/Heart Disease 0.40 0.29, 0.56 <.001 0.47 0.33, 0.67 <.001
(0.17) (0.18)
Lung Disease 0.55 0.32, 0.92 .024 0.65 0.41, 1.04 .072
(0.27) (0.24)
Neurological Disease 0.14 0.11, 0.18 <.001 0.15 0.12, 0.18 <.001
(0.12) (0.10)
Other Primary Diagnosis 0.25 0.10, 0.62 .003 0.25 0.12, 0.54 <.001
(0.47) (0.38)
Comorbidity Count 1.00 0.96, 1.05 .963 1.02 0.97, 1.07 .430
(0.02) (0.03)
Medicaid Enrollee 0.18 0.13, 0.23 <.001 0.15 0.13, 0.18 <.001
(0.14) (0.08)
Advance Care Planning
Yes AD, No DNR order 2.91 1.39, 6.07 .005
(0.38)
Yes AD, YES DNR order 1.57 1.37, 1.80 <.001
(0.07)
Random Intercept 28.77 8.20, 100.91 <.001 45.08 14.72, 138.06 <.001
(0.64) (0.57)
Model Summary
Level 2 Variance 1.20 (0.07) 1.26 (0.07)
Akaike Information Criterion
Bayesian Information Criterion
Patient Observations 2,570 2,568
Agency Observations 622 622
Footnotes Reference Variables
CHF: Congestive Heart Failure White (Race)
AD: Advance Directive Married/Living with Partner (Marital Status)
DNR: Do Not Resuscitate Cancer (Primary Diagnosis)
No Advance Directive (Advance Care Planning)
409665.85
409712.66
407270.15
407316.96

Step 2: Inclusion of advance care planning decisions. To determine the relative
effect of advance care planning on dying in a home-like setting, a second model was
estimated. Results are presented under Step 2 in Table 22. The second adjusted model
101
(Level 1 n = 2,568; Level 2 n = 622) was an improved fit over the first model (AIC =
407270.15; BIC = 407316.96). As with the first model, Black (OR = 2.14; p = .004) and
Hispanic decedents (OR = 1.20; p = .017) were more likely than White decedents to die
in a home-like setting. Relative effect size differences between the first and second
models was 1.90 vs. 2.14 for Black decedents, and 1.21 vs. 1.20 for Hispanic decedents.
Those with an advance directive (OR = 2.91; p = .005), or DNR order (OR = 1.57; p <
.001) were significantly more likely than those without an advance directive to die in a
home-like setting. Conversely, widowed decedents (OR = 0.70; p < .001),
single/divorced/separated decedents (OR = 0.57; p = .047), those with Congestive Heart
Failure/Heart Disease (OR = 0.47; p < .001), a neurological disease (OR = 0.15; p <
.001), other primary diagnoses (OR = 0.25; p < .001), and decedents enrolled in Medicaid
(OR = 0.15; p < .001) were significantly less likely than the associated referents to die in
a home-like setting. Following inclusion of care planning covariates, Lung Disease lost
significance in the model. See Figure 15 for a blobbogram illustrating the relative effect
of each Step 2 predictor for dying in a home-like setting.

102

Site of death: Hospital. Analysis of the null model (agency-level n = 627;
patient-level n = 2,680) suggested that 31% of the variance was accounted for by agency-
level differences ICC = 0.319, 95% CI: 0.234, 0.420; χ
2
= 15.699, df = 1; p <.001).
Step 1: Demographics and health indicator variables. To test for racial/ethnic
variation in dying in a hospital, a multilevel logistic regression was conducted. Results of
this model are presented under Step 1 in Table 23. The adjusted model (Level 1 n =
2,570; Level 2 n = 622), revealed AIC and BIC model fit criteria were 189611.79, and
189658.60, respectively. As hypothesized, Hispanic decedents (OR = 7.93; p < .001)
were significantly more likely than White decedents to die in a hospital; however, Black
decedents (OR = 0.54; p = .447) were not more likely than White decedents to die in a
hospital. Those with Congestive Heart Failure/Heart Disease (OR = 2.11; p = .023), Lung
Footnotes
CHF: Congestive Heart Failure
HD: Heart Disease
DNR: Do Not Resuscitate Order
Reference Variables
White (Race)
Married/Living with Partner (Marital Status)
Cancer (Primary Diagnosis)
No Advance Directive (Advance Care Planning)
Figure 15. Relative Effects of Predictors for Site of Death: Home-Like Setting
0.1
1
10
*** Documented DNR
** No Documented DNR
*** Medicaid Enrollee
Comorbidity Count
*** Other Primary Diagnosis
*** Neurological Diseases
Lung Disease
*** CHF/HD
* Single/Divorced/Separated
*** Widowed
* Hispanic
** Black
Female
* Age
Odds Ratio
Relative Effect Size
103
Disease (OR = 3.18; p < .001), or other primary diagnoses (OR = 6.24; p < .001) were
significantly more likely than those with Cancer to die in a hospital. Conversely,
single/divorced/separated decedents (OR = 0.67; p < .001), and those enrolled in
Medicaid (OR = 0.44; p < .001) were significantly less likely than the associated referents
to die in a hospital. Lastly, both age (OR = 0.97; p < .001), and comorbidity count (OR =
0.83; p < .001) were negatively associated with in-hospital death.
104
Table 23. Predictors of Site of Death: Hospital
Model 1 Model 2
OR OR
(Std. Err.) 95% CI Sig. (Std. Err.) 95% CI Sig.
Predictors
Race/Ethnicity
Black 0.54 0.11, 2.63 .447 0.46 0.09, 2.46 .364
(0.81) (0.86)
Hispanic 7.93 4.50, 13.98 <.001 6.12 3.85, 9.75 <.001
(0.29) (0.24)
Age 0.97 0.96, 0.99 <.001 0.97 0.96, 0.98 <.001
(0.01) (0.01)
Female 1.09 0.90, 1.32 .378 1.17 0.97, 1.40 .097
(0.10) (0.09)
Marital Status
Widowed 0.90 0.74, 1.10 .300 0.92 0.72, 1.18 .521
(0.10) (0.12)
Single/Divorced/Separated 0.67 0.56, 0.80 <.001 0.65 0.52, 0.82 <.001
(0.09) (0.12)
Primary Diagnosis
CHF/Heart Disease 2.11 1.11, 4.00 .023 1.82 0.89, 3.71 .101
(0.33) (0.36)
Lung Disease 3.18 1.74, 5.81 <.001 2.73 1.66, 4.50 <.001
(0.31) (0.25)
Neurological Disease 1.03 0.58, 1.84 .920 1.02 0.53, 1.94 .962
(0.30) (0.33)
Other Primary Diagnosis 6.24 4.01, 9.69 <.001 4.78 2.80, 8.15 <.001
(0.23) (0.27)
Comorbidity Count 0.83 0.78, 0.87 <.001 0.84 0.80, 0.88 <.001
(0.03) (0.03)
Medicaid Enrollee 0.44 0.28, 0.70 .001 0.38 0.20, 0.72 .003
(0.24) (0.33)
Advance Care Planning
Yes AD, No DNR order 0.47 0.28, 0.78 .003
(0.26)
Yes AD, YES DNR order 0.46 0.41, 0.52 <.001
Random Intercept 3.60 3.09, 4.18 <.001 3.43 3.14, 3.74 <.001
(0.08) (0.04)
Model Summary
Level 2 Variance 1.64 (0.20) 1.52 (0.11)
Akaike Information Criterion
Bayesian Information Criterion
Patient Observations 2,570 2,568
Agency Observations 622 622
Footnotes Reference Variables
CHF: Congestive Heart Failure White (Race)
AD: Advance Directive Married/Living with Partner Marital Status)
DNR: Do Not Resuscitate Cancer (Primary Diagnosis)
No Advance Directive (Advance Care Planning)
189611.79
189658.60
188717.65
188764.46

Step 2: Inclusion of advance care planning decisions. The second multilevel
model predicting death in a hospital included advance care planning covariates (see Step
2 in Table 23). The adjusted model (Level 1 n = 2,568; Level 2 n = 622) was found to be
105
a better fit to the data, compared to the first model (AIC = 188717.65; BIC = 188764.46).
Following inclusion of advance care planning covariates, Hispanic decedents (OR = 6.12;
p < .001) remained significantly more likely than White decedents to die in a hospital (a
effect size change from 7.93 in the first model). Furthermore, Black decedents (OR =
0.46; p = .364) also remained as likely as White decedents to die in-hospital. Decedents
with Lung Disease (OR = 2.73; p < .001), or other primary diagnoses (OR = 4.78; p <
.001) were significantly more likely than decedents with Cancer to die in a hospital.
Single/Divorced/Separated decedents (OR = 0.65; p < .001), those enrolled in Medicaid
(OR = 0.38; p = .003), and decedents with an advance directive (OR = 0.47; p = .003) or
DNR order (OR = 0.46; p < .001) were significantly less likely than the associated
referents to experience in-hospital death. Furthermore, older decedents (OR = 0.97; p <
.001), and those with more comorbidities (OR = 0.84; p < .001) were significantly less
likely to die in a hospital. Lastly, predictor significance was lost for Congestive Heart
Failure/Heart Disease following inclusion of advance care planning covariates. For a
blobbogram illustrating the relative effect of each Step 2 predictor for dying in a hospital,
see Figure 16.

106

Footnotes
CHF: Congestive Heart Failure
HD: Heart Disease
DNR: Do Not Resuscitate Order
Reference Variables
White (Race)
Married/Living with Partner (Marital Status)
Cancer (Primary Diagnosis)
No Advance Directive (Advance Care Planning)
Figure 16. Relative Effects of Predictors for Site of Death: Hospital
0.01
0.1
1
10
100
*** Documented DNR
** No Documented DNR
** Medicaid Enrollee
*** Comorbidity Count
*** Other Primary Diagnosis
Neurological Diseases
*** Lung Disease
CHF/HD
*** Single/Divorced/Separated
Widowed
*** Hispanic
Black
Female
*** Age
Odds Ratio
Relative Effect Size
107
Chapter 9: Discussion
This is the first known in-depth investigation of racial/ethnic variation in care
preferences and outcomes following hospice enrollment. Examined outcomes included
advance directive completion, do not resuscitate (DNR) order election, healthcare proxy
designation, emergent care utilization, decedent length of stay (LOS), and site of death
(SOD). In this chapter, hypotheses are revisited with respect to study findings,
contributions to the existing literature are discussed, limitations are considered, and
recommendations for future research are proposed.
Hypothesis 1 Findings
White hospice patients will be more likely than Black and Hispanic hospice
patients to have a documented advance directive.
Relatively few previous studies have examined racial/ethnic variation in advance
directive completion following hospice enrollment (Jones et al., 2011; Resnick et al.,
2012). Drawing from findings of these studies, it was hypothesized that Black and
Hispanic hospice patients would be less likely than White hospice patients to have a
documented advance directive. Results of this study were semi-consistent with this
hypothesis. Specifically, as hypothesized, Black hospice patients were estimated to be
80% less likely than White hospice patients to complete an advance directive (OR = 0.20;
p < .001); however, Hispanic patients were not found to be less likely, but instead equally
as likely as, White patients to complete an advance directive (OR = 1.04; p = .703). The
finding that Black hospice patients are less likely than White hospice patients to complete
an advance directive is consistent with both of the previously identified studies of hospice
samples, which reported that Blacks were less likely than Whites to complete advance
108
directives (Jones et al., 2011; Resnick et al., 2012). This finding is also consistent with
studies examining the non-hospice population (Eleazer et al., 1996; Greiner et al., 2003;
K. S. Johnson, Kuchibhatla, & Tulsky, 2008; Kwak & Haley, 2005; McKinley et al.,
1996). Concerning Hispanic patients, findings are not consistent with studies of the non-
hospice population, which have reported reduced advance directive completion among
Hispanic, compared to White, patients (Kelley et al., 2010; Morrison et al., 1998). No
previously identified study has examined variation in advance directive completion
among Hispanic hospice patients.
Hypothesis 1A Findings
Among those with documented advance directives, White and Hispanic hospice
patients will be more likely than Black hospice patients to document a do not resuscitate
order.
Only one identified study has examined racial/ethnic differences in DNR order
election following hospice enrollment (Jones et al., 2011). Drawing from this study, as
well as studies of the non-hospice population, it was hypothesized that Black hospice
patients would be less likely than White and Hispanic hospice patients to elect a DNR
order. Results were semi-consistent with this hypothesis. That is, Black hospice patients
were estimated to be equally as likely as White hospice patients to elect a DNR order (OR
= 1.09; p = .553). Furthermore, while Hispanic hospice patients were 3.68 times more
likely than White hospice patients to elect a DNR order (OR = 3.68; p < .001), they were
also 3.67 times more likely than Black hospice patients to elect a DNR order (OR = 3.67;
p < .001). The finding that Black hospice patients are equally as likely as White hospice
patients to elect a DNR order contrasts with previous studies of hospice (Jones et al.,
109
2011), and non-hospice populations (Barnato et al., 2009; Borum et al., 2000; Duffy et
al., 2006; K. S. Johnson, Kuchibhatla, & Tulsky, 2008; Mitchell & Mitchell, 2009),
which have reported stronger preferences among Blacks for more aggressive end-of-life
(EOL) treatments, and reduced likelihood of electing a DNR order. Conversely, the
finding that Hispanic patients were more likely than Black patients to elect a DNR order
is consistent with literature indicating that Hispanics often prefer less aggressive care,
despite lower documentation levels (Kelley et al., 2010; Morrison et al., 1998); however,
the finding that Hispanic patients were also more than White patients to elect a DNR
order has not been previously documented in the hospice- or non-hospice-electing
literature.
Hypothesis 1B Findings
Among those with documented advance directives, White and Hispanic hospice
patients will be more likely than Black hospice patients to document a healthcare proxy.
Research investigating healthcare proxy designation among racially/ethnically
diverse populations is scarce (Blackhall et al., 1995; Hopp & Duffy, 2000; Kwak &
Haley, 2005). Drawing from the few existing studies, it was hypothesized that White and
Hispanic hospice patients would designate healthcare proxies at higher rates than Black
hospice patients. Hispanic hospice patients were hypothesized to designate healthcare
proxies at similar rates as Whites given studies suggesting that environments encouraging
advance care planning would facilitate documentation of informal preferences (Kelley et
al., 2010; Morrison et al., 1998). Results of this study were semi-consistent with this
hypothesis. Although Black hospice patients were estimated to be 42% less likely (OR =
0.58; p < .001) than White hospice patients to designate a healthcare proxy, Hispanic
110
hospice patients were also estimated to be 84% less likely (OR = 0.16; p < .001) than
White hospice patients to designate a healthcare proxy. The finding that Black hospice
patients are less likely than White hospice patients to designate a healthcare proxy is
consistent with the small body of previous research of non-hospice samples (Blackhall et
al., 1995; Hopp & Duffy, 2000; Kwak & Haley, 2005). However, despite being in an
environment that encourages advance care planning, Hispanic hospice patients remained
less likely than White hospice patients to designate healthcare proxies. No prior identified
study has investigated racial/ethnic variation in healthcare proxy designation rates
following hospice enrollment.
Hypothesis 2 Findings
Black and Hispanic hospice patients will be more likely than White and hospice
patients to utilize emergent care (i.e., unplanned emergency medical care).
As with previously discussed outcomes, racial/ethnic variation in emergent care
utilization following hospice enrollment is limited to a few studies (Cintron et al., 2003;
Schonwetter et al., 2008; Unroe et al., 2012). Drawing from this existing literature, it was
hypothesized that both Black and Hispanic hospice patients would be more likely than
White hospice patients to utilize emergent care. Results of this study were inconsistent
with this hypothesis. Black hospice patients were estimated to be equally as likely as
White hospice patients to utilize emergent care (OR = 0.81; p = .150), and Hispanic
hospice patients were estimated to be 44% less likely than White hospice patients to
utilize emergent care (OR = 0.56; p < .001). The finding that neither Black nor Hispanic
hospice patients were more likely than White hospice patients to utilize emergent care is
inconsistent with prior research reporting significantly higher utilization among
111
minorities, both within hospice the system (Cintron et al., 2003; Schonwetter et al., 2008;
Unroe et al., 2012), as well as throughout the larger U.S. healthcare system (Barnato et
al., 2007; Goldstein et al., 2010; Gozalo et al., 2011; Hanchate et al., 2009; A. K. Smith,
Earle, et al., 2009). Additionally, the finding that Hispanic patients were significantly less
likely that White patients to utilize emergent care following hospice enrollment has not
been previously documented.
Hypothesis 2A Findings
Hospice patients without documented advance care plans will be more likely than
those with documented advance care plans to utilize emergent care.
No identified study has examined the association between advance care planning
and emergent care utilization following hospice enrollment. Drawing from prior studies
of non-hospice-electing patients, it was hypothesized that patients engaging in advance
care planning would be less likely than those not engaging in such planning to utilize
emergent care following hospice enrollment. Results were consistent with this
hypothesis, as hospice patients with an advance directive were 53% less likely (OR =
0.47; p = .002) to utilize emergent care following hospice enrollment. Likewise, hospice
patients with a DNR order were 64% less likely (OR = 0.36; p = .010) to utilize emergent
care following hospice enrollment. These findings match those of previous studies of
non-hospice-specific populations, which have consistently reported lower levels of
emergent care use among those who engage in advance care planning (Degenholtz et al.,
2004; Gozalo et al., 2011; Silveira et al., 2010; Teno, Gruneir, et al., 2007).
Findings also indicated that racial/ethnic variation in emergent care utilization
persisted following adjustment for advance care planning. Specifically, Black hospice
112
patients were found to be 32% less likely than White hospice patients to utilize emergent
care following hospice enrollment (OR = 0.68; p = .031), whereas Hispanic hospice
patients were estimated to be equally as likely (OR = 0.70; p = .117). As with the
previous hypothesis testing, the finding that Black and Hispanic hospice patients were not
more likely than White hospice patients to utilize emergent care following adjustment for
advance care planning is inconsistent with previous research (Barnato et al., 2007;
Cintron et al., 2003; Goldstein et al., 2010; Gozalo et al., 2011; Hanchate et al., 2009;
Schonwetter et al., 2008; A. K. Smith, Earle, et al., 2009; Unroe et al., 2012).
Hypothesis 3 Findings
Hospice length of stay will be longer for Black and Hispanic decedents, compared
to White decedents
Racial/Ethnic variation in hospice LOS is largely understudied (Cólon & Lyke,
2003; Hardy et al., 2012; K. S. Johnson et al., 2011; Rhodes et al., 2007). Drawing from
the four identified studies in the past decade, Black and Hispanic decedents were
hypothesized to report longer stays under the care of hospice, relative to White hospice
decedents. Results were not consistent with this hypothesis, as neither Black (HR = 0.06;
p = .626) nor Hispanic (HR = 0.02; p = .850) hospice decedents were estimated to be
more likely than White hospice decedents to experience longer hospice LOS. This
finding also persisted following adjustment for advance care planning. Results contrast
with the majority of existing studies, which have reported longer LOS among hospice
patients (Cólon & Lyke, 2003), and decedents (Hardy et al., 2012; K. S. Johnson et al.,
2011). However, results correspond with one previous study by Rhodes and colleagues
(2007) which reported no difference in hospice LOS among White and Black decedents.
113
Hypothesis 4 Findings
White hospice decedents will be more likely than Black and Hispanic hospice
decedents to die in a home-like setting.
While several previous studies have examined racial/ethnic variation in SOD
among non-hospice-specific populations, only one identified study has examined SOD
differences following hospice enrollment among Whites, Blacks, and Hispanics (K. S.
Johnson et al., 2005). Drawing from this body of literature, it was hypothesized that
White hospice decedents would be more likely than Black or Hispanic hospice decedents
to die in a home-like setting. Results were inconsistent with this hypothesis. Black
hospice decedents were estimated to be 2.14 times more likely that White hospice
decedents to die in a home-like setting (OR = 2.14; p = .004), and Hispanic hospice
decedents were found to be 20% more likely than White hospice decedents to die in a
home-like setting (OR = 1.20; p = .017). These differences persisted following
adjustment for advance care planning. Given that the one previously identified study of
SOD among hospice decedents utilized significantly different comparison settings (i.e.,
death in an inpatient setting, including inpatient hospice), results are difficult to compare.
Regardless, findings were inconsistent with the previous study which reported no
difference between Black and White decedents in dying in a home setting, relative to an
inpatient hospice setting (K. S. Johnson et al., 2005). Conversely, Hispanic decedent
findings were similar with those of the previous study which reported a decreased
likelihood among Hispanic, compared to White, decedents to die in an inpatient hospice
setting, relative to a home (K. S. Johnson et al., 2005). Findings from this study are
discordant with studies of the larger U.S. healthcare system, which have repeatedly found
114
an increased likelihood among Blacks and Hispanics to die in acute settings (Gruneir et
al., 2007; Hanchate et al., 2009; Hansen et al., 2002; National Center for Health
Statistics, 2011; A. K. Smith, Earle, et al., 2009; Weitzen et al., 2003; Zheng et al., 2011).
Hypothesis 4A Findings
Black and Hispanic hospice decedents will be more likely than White hospice
decedents to die in a hospital.
Although only one identified study has examined racial/ethnic variation in SOD
following hospice enrollment, death in an acute setting was not included in the reported
analyses. Therefore, drawing from research of non-hospice-specific populations, Black
and Hispanic hospice decedents were hypothesized to be more likely than White hospice
decedents to die in a hospital. Findings were semi-consistent with this hypothesis.
Specifically, Hispanic decedents were estimated to be 6.12 times more likely than White
decedents to die in a hospital (OR = 6.12; p < .001). Conversely, Black decedents were
found to be equally as likely as White decedents to die in a hospital (OR = 0.46; p =
.364). Numerous studies of the non-hospice population have documented an increased
risk among Blacks and Hispanics, relative to Whites, to die in a hospital (Gruneir et al.,
2007; Hanchate et al., 2009; Hansen et al., 2002; National Center for Health Statistics,
2011; A. K. Smith, Earle, et al., 2009; Weitzen et al., 2003; Zheng et al., 2011). While the
Hispanic findings in this study are consistent with this body of literature, the finding that
Black hospice decedents are equally as likely as White decedents to die in a hospital
contrasts with what has been previously reported.

115
Contributions to the Literature
By investigating racial/ethnic variation in care preferences, utilization, and
outcomes following hospice enrollment, this study addresses several gaps in the existing
literature. When considered alongside previous research, multiple factors can be
identified to inform understanding of racial/ethnic variation following hospice
enrollment.
Advance care planning. Findings indicate that racial/ethnic variation in advance
care planning behaviors persist following hospice enrollment. While some of the reported
differences mirror what has been documented outside of the hospice system, findings also
suggest substantial differences between minorities enrolling in hospice and their non-
hospice-utilizing counterparts.
Advance directive completion. Hispanic patients were found to be equally as
likely as White patients to complete an advance directive. This finding illustrates the
potential value of advance care planning education for Hispanic populations. Previous
studies have documented the effectiveness of educational programs in overcoming
Hispanic barriers to advance care planning (Volandes, Ariza, et al., 2008; Volandes et al.,
2010; Volandes et al., 2007). It is possible that the existing structure of hospice provides
improved access to advance care planning among Hispanics, resulting in advance
directive completion rates comparable with White populations. This would also explain
why Hispanics were equally as likely as Whites to document an advance directive, but
Blacks remained significantly less likely to document, as non-hospice-specific studies
suggest that many Hispanics prefer less aggressive EOL interventions, but that few
document the preference (Kelley et al., 2010; Morrison et al., 1998). The hospice
116
philosophy facilitates patient education of advance care planning so that well-informed
decisions can be made, and thus it is possible that existing programs within hospice
increase advance directive completion rates among Hispanics. This study also found that
Black hospice patients were less likely than White hospice patients to complete an
advance directive, suggesting that hospice-based advance care planning educational
interventions, in their current form, do not influence advance directive development
among Black hospice patients. Bullock (2006) found that following an advance care
planning educational intervention, religious beliefs still significantly affected willingness
to engage in advance care planning for many Blacks. The study found that among healthy
Black older adults, most participants refused to complete advance directives following
intervention, citing no perceived benefits to having an advance directive, valuation of
enduring suffering and fighting until the end, beliefs that god can cure illnesses beyond
medical intervention, and general distrust of service providers and the larger healthcare
system (Bullock, 2006). These findings in conjunction with those reported in this study
suggest that religious beliefs could have contributed to some of the observed racial/ethnic
variation in advance directive completion. It is also possible that patient-provider
communication, as well as trust in healthcare providers, may have also influenced
decisions to complete advance directives for Black hospice patients (Kelley 2010,
Perkins 2002; Duffy 2006; Braunstein 2008; Gamble 1997; Harris 2001). As non-
hospice-specific studies have reported, communication and trust remain a significant
healthcare issue in the United States for many patients and providers (Braunstein et al.,
2008; Gamble, 1997; Harris et al., 1996; Perkins et al., 2002). Challenges associated with
117
communication and trust may persist in the hospice system, and contribute to the
differences observed in this study.
Do not resuscitate order election. This study also found that among the
racial/ethnic minorities who engage in advance care planning, preferences for
resuscitation that are well-documented outside of hospice are not present within the
hospice system. First, Hispanic hospice patients were more likely than White hospice
patients to elect a DNR order. Although this finding could be the result of selection bias
(i.e., Hispanic patients who prefer more aggressive care do not enroll in hospice), it also
potentially illustrates the value of the hospice experience. Under the care of hospice,
patients experience many of the benefits associated with an integrated care approach that
addresses holistic needs, often including religio-cultural components. Following hospice
intervention, patients are able to focus more of their time, energy, and resources to
consider their care choices. It is conceivable that as Hispanic patients in this study
experienced the benefits of palliative patient-centered care, the perceived need for more
aggressive interventions decreased, and patients documented less aggressive preferences
at higher rates. Finally, the increased likelihood among Hispanics to elect a DNR order
could also be a function of hospices accurately documenting patient preferences. Loggers
and colleagues (2013) found that among a non-hospice-specific sample, similar
proportions of Hispanics and Whites preferred less aggressive EOL interventions;
however, Whites were more likely to document this preference through a DNR order.
Given that the current study did not find such differences in DNR election, it is possible
that among those hospice-enrollees choosing to engage in advance care planning,
hospices are effective at ensuring that DNR preferences are accurately documented.
118
This study also found that among those engaging in advance care planning, Black
hospice patients were equally as likely as White hospice patients to elect a DNR order, a
finding that contrasts with the existing literature (K. S. Johnson, Kuchibhatla, & Tulsky,
2008; Kwak & Haley, 2005; McKinley et al., 1996). Previous research suggests two
possible explanations to account for this finding. First, it is possible that this trend arises
from self-selection, with Black hospice patients who prefer less aggressive interventions
more likely to complete an advance directive and thus, elect DNR orders at a similar rate
as their White counterparts. It is also possible that for some Black hospice patients, the
timing of advance care planning education, as well as the hospice experience, may
support more favorable views for comfort-based interventions following hospice
enrollment. Although previous research has found that educational interventions often do
little to shift views of advance care planning among healthy Black patients (Bullock,
2006), a recent study has reported that seriously ill Black patients are willing to
reconsider DNR election preferences following a palliative-based intervention (Sacco,
Carr, & Viola, 2013). Sacco et al. (2013) reported that while none of the Black patients in
their study (n = 1,113) had documented a DNR order prior to palliative care intervention,
65% of Black patients (n = 724) had elected a DNR order following palliative care
consultation. When considered with findings reported in this manuscript, it is possible
that after experiencing the palliative and patient-centered benefits of hospice care, some
Black hospice patients elected to document a DNR order. However, additional time-
ordered investigation is needed to confirm this hypothesis. Finally, when considering
DNR election findings with advance directive completion findings of this study, the fact
that Black hospice patients were less likely than White hospice patients to document an
119
advance directive, but equally as likely as Whites to elect a DNR order (among those
completing advance directives) suggests that formal documentation of DNR orders may
not be important to Black patients who elect hospice care; however, data from this study
restrict examination of this possibility.
Although several factors are certainly in operation, overall DNR election findings
from this study highlight the potential value of healthcare settings that encourage advance
care planning. Namely, of the racial/ethnic minorities who engaged in advance care
planning, the stronger preferences for resuscitation that have been well-documented
outside of hospice were not observed. While this may be due in-part to patient self-
selection (i.e., only those who preferred less aggressive care completed advance
directives), results support the general hospice philosophy of encouraging all patients to
document their preferences, regardless of aggressiveness, and suggest that potential value
of advance care planning education and palliative-based patient-centered care in reducing
existing racial/ethnic disparities.
Healthcare proxy designation. Results also indicated racial/ethnic variation in
healthcare proxy designation rates following hospice enrollment. Specifically, both Black
and Hispanic hospice patients were found to be less likely than White hospice patients to
document a formal healthcare proxy. These findings are congruent with previous research
studies, which have found that many racial/ethnic minorities value informal over formal
healthcare proxy designation (Carr, 2011; Cruz-Oliver, Talamantes, & Sanchez-Reilly,
2014; Morrison et al., 1998). Morrison and colleagues (1998) reported that the prevailing
family-centered culture of many Hispanics fosters the belief that the family will work
closely together as a unit to ensure that the dying loved one will receive adequate and
120
appropriate care. Accordingly for these Hispanics, formal designation of a healthcare
proxy is often considered to be unnecessary, given existing informal family designation
(Morrison et al., 1998). Likewise, in a study of chronically ill older adults, Carr (2011)
found that for some Black patients, beliefs in a higher power often shape views
concerning the nature and timing of death, and thus ultimately affect decisions regarding
healthcare proxy designation. The author illustrates this concept with the contention, “If
God controls life and death decision-making, then legal documents specifying one’s
medical treatment preferences may be deemed irrelevant, undesirable, or as intruding
upon God’s plan” (Carr, 2011, p. 15). Furthermore, given that this study also found a
reduced likelihood among Blacks to engage in any advance care planning (through the
documentation of an advance directive), it is also possible that the lower rates of
healthcare proxy designation is a function of a wider view among Blacks in this study
that the documentation of care preferences is generally unnecessary. Considering the
strong trend toward lower rates of healthcare proxy designation among both Black and
Hispanic hospice patients, it could be reasoned that while education may help overcome
educational barriers to advance care planning, others factors, such as cultural preferences,
may not be impacted. Additional investigation into the cultural context of EOL decision-
making is needed to better understand these findings.
Finally, results suggest that advance care planning, specifically advance directive
completion and DNR order election, significantly influences care utilization and
outcomes following hospice enrollment. Patients who engaged in advance care planning
were found to have 1) lower emergent care utilization rates (patients completing advance
directives, and patients electing DNR orders), 2) reduced likelihood of dying within the
121
first week of hospice care (patients completing advance directives, and patients electing
DNR orders), 3) reduced likelihood of dying within the first month of hospice care
(patients electing DNR orders), 4) increased likelihood of dying in a home like setting
(patients completing advance directives, and patients electing DNR orders), and 5)
reduced likelihood of dying in a hospital (patients completing advance directives, and
patients electing DNR orders). Although some of these findings may relate to the timing
of advance directive completion (e.g., perhaps the advance directive was in-place earlier
in the dying process), these findings as a whole strongly support the value of advance
care planning on multiple key EOL outcomes within the hospice system.
Emergent care utilization. Results of this study suggest that the racial/ethnic
variation in emergent care utilization observed outside of hospice care differs from trends
observed within hospice care. Research of the larger healthcare system has consistently
documented higher rates of emergent care utilization for racial/ethnic minorities,
compared to Whites (Barnato et al., 2007; Goldstein et al., 2010; Gozalo et al., 2011;
Hanchate et al., 2009; A. K. Smith, Earle, et al., 2009). In contrast, the current study
found that racial/ethnic minorities enrolled in hospice care were not more likely to utilize
emergent care. In fact, Blacks were less likely, and Hispanics equally as likely, as Whites
to utilize emergent care, following adjustment for advance care planning. Similar results
were also observed when examining emergency department utilization rates within the
same sample utilized in this study (not shown). Two possible explanations to account for
this divergent finding are discussed below.
First, it could be conceived that hospice-electing racial/ethnic minorities differ
significantly in their care utilization patterns from those who do not elect hospice. It is
122
possible that those who prefer less aggressive interventions naturally elect hospice care,
and thus utilize less emergent care following hospice enrollment. As a result of this self-
section, racial/ethnic minorities would not be expected to utilize emergent care at a higher
rate than Whites, as observed in this study. Moreover, it is also possible that hospice
enrollees who prefer more aggressive care may be disenrolling from hospice in order to
pursue more aggressive interventions, and thus skewing the findings. Previous studies of
higher hospice disenrollment rates among racial/ethnic minorities further support this
possibility (K. S. Johnson, Kuchibhatla, Tanis, et al., 2008; Kapo, MacMoran, &
Casarett, 2005; Unroe et al., 2012). Second, it is possible that the hospice system elicits
less aggressive care decisions by patients and thus, contributes to comparable emergent
care utilization patterns across race/ethnicity. This protective effect of advance care
planning against utilization of emergent services was also reported in a recent systematic
literature review of 113 studies over the past decade (Brinkman-Stoppelenburg, Rietjens,
& van der Heide, 2014). As purported earlier, the value of the hospice experience may be
functioning as a key contributor to study findings, with patients shifting care preferences
and decisions after experiencing the many benefits of hospice care. The shift from
standard acute care to patient-centered hospice care could promote improved
understanding and awareness of the negative outcomes associated with aggressive, futile
care, thus reducing minority patients’ desire for aggressive interventions.
Interestingly, the effect of race/ethnicity on emergent care utilization changed
considerably following adjustment for advance care planning decisions. Specifically,
prior to adjustment, Hispanics were found to be less likely than Whites to utilize
emergent care, and Blacks were found to be equally as likely as Whites to utilize
123
emergent care. However, when advance care planning decisions were added to the
analytic model, Blacks were found to be less likely than Whites to utilize emergent care,
and Hispanics were found to be equally as likely as Whites to utilize emergent care.
Previous studies examining emergent care utilization among hospice enrollees offers a
potential explanation. In a study of 292 stage IV cancer hospice enrollees, Loggers and
colleagues (2013) found that Hispanic hospice patients were less likely than White
hospice patients to utilize intensive EOL care. The authors of this study concluded that
patient documentation against resuscitation played a critical role in study findings
(Loggers et al., 2013). Thus, it may be possible that the increased likelihood of Hispanics
in this study to elect a DNR order may have also influenced emergent care utilization
models. This hypothesis is substantiated by the fact that the Hispanic effect went away
following adjustment for DNR order election (i.e., Hispanics became equally as likely as
Whites as DNR order election became significant in predicting a reduced likelihood of
utilizing emergent care). Conversely, given that Black hospice patients were less likely
than White hospice patients to complete an advance directive, it is conceivable that the
relative effect of Black vs. White race/ethnicity on emergent care use would shift
following adjustment for advance care planning decisions. Not surprisingly, this is what
was observed in this study, with Black hospice patients becoming less likely than White
hospice patients to utilize emergent care following inclusion of care planning covariates.
This study also found that patients who preferred less aggressive care often received it,
indicating good patient-provider communication. Loggers and colleagues (2009)
previously reported that among Black non-hospice enrollees, preferences for less
aggressive EOL were not associated with the type of care received (i.e., acute
124
interventions remained prevalent), suggesting poor provider-patient communication.
However, findings from this study suggest congruency between patient preferences and
the type of care received, and support hospice providers’ effectiveness in maintaining the
central value of patient-guided care.
Hospice length of stay. Hospice LOS was examined in this study using three
distinct outcomes: 1) the full care episode, 2) survival of one week or less, and 3) survival
of one month or less. Results from models examining the full care episode indicate no
racial/ethnic variation in hospice LOS. This finding persisted following adjustment for
advance care plans. These results are inconsistent with most of the previous literature
(Cólon & Lyke, 2003; Hardy et al., 2012; K. S. Johnson et al., 2011), which has reported
longer hospice LOS for racial/ethnic minorities; however, it is consistent with one
previous study reporting no racial/ethnic variation in hospice LOS following analysis of
over 120,000 hospice decedents (Rhodes et al., 2007). Although the initial LOS finding
from this study indicates no racial/ethnic variation in hospice LOS, the divergence of this
finding from the larger empirical literature prompted a second wave of analyses
examining racial/ethnic variation in shorter care episodes (Byock et al., 1996; Christakis
& Iwashyna, 1998; Rickerson et al., 2005). Results from the additional analyses indicated
that, compared to White hospice patients, Black hospice patients were significantly less
likely to die in the first week (OR = 0.78; p = .027), but more likely to die in the first
month of hospice care (OR = 2.54; p < .001). Similarly, compared to White hospice
patients, Hispanic hospice patients were significantly more likely to die in the first week
(OR = 1.18; p < .001), and equally as likely to die in the first month of hospice care (OR
= 1.20; p = .148). All findings remained relatively unchanged following adjustment for
125
advance care planning decisions. Taken together, these findings suggest that racial/ethnic
minorities enter hospice later in the disease trajectory compared to Whites, a finding
previously documented in the empirical literature (K. S. Johnson et al., 2011).
Several implications can be drawn from the finding that racial/ethnic minorities
are at an increased risk of dying within the first month of care. First, racial/ethnic
minorities may experience late referral to hospice. It is possible that poor patient-provider
communication about EOL care options and preferences prior to hospice enrollment may
be contributing to delayed hospice enrollment for Blacks and Hispanics (Kelley et al.,
2010; Perkins et al., 2002), and ultimately resulting in shorter hospice care episodes for
minorities. Similarly, patient lack of trust in healthcare providers (Braunstein et al., 2008;
Duffy et al., 2006; Gamble, 1997) may also influence timely enrollment, with some
patients insisting on more aggressive interventions in order to ensure that they are not
under-treated by healthcare providers. Racial/Ethnic minorities may also be deferring
hospice enrollment for religio-cultural reasons. As previously discussed, cultural values
such as Machismo/Fatalismo (Del Gaudio et al., 2013; A. K. Smith, Earle, et al., 2009)
can influence racial/ethnic minority decision-making at EOL. Similarly, other factors,
such as the perceived value of suffering (Krause & Bastida, 2011), and the belief that a
higher being determines one’s health outcomes (Blocker et al., 2006; Carr, 2011), may
also be influence the timing of hospice enrollment.
Timely enrollment in hospice is important to ensure receipt of adequate hospice
benefits. Studies indicate that at least thirty days of care are necessary for hospice
providers to fully address the needs of their patients (Christakis & Iwashyna, 2000; Han
et al., 2007; Haupt, 2003; Huskamp et al., 2001; McCarthy, Burns, Ngo-Metzger, et al.,
126
2003; Quill, 2007). Although study findings suggest both Black and Hispanic hospice
patients are at-risk of not experiencing the full benefits of hospice care, additional
research is also needed to ascertain patient and family views of the hospice experience. A
growing body of research suggests that actual enrollment is more important than the
number of days enrolled in hospice (Rickerson et al., 2005; Schockett, Teno, Miller, &
Stuart, 2005; Teno, Shu, et al., 2007). These studies found that many patients who receive
care for less than one month still report positive views of hospice and describe their needs
as being met. Thus, although racial/ethnic minorities may be experiencing shorter hospice
care episodes, it is possible that these patients are receiving the preferred duration of
hospice care, and benefiting from this care. Additional studies are needed to test this
hypothesis, and improve understanding of the relative effect of late hospice enrollment on
racial/ethnic minority populations.
Site of death. This study, one of the first comprehensive examinations of
racial/ethnic differences in SOD among Medicare hospice beneficiaries, found significant
racial/ethnic variation, with both Black and Hispanic patients more likely than White
patients to die in a home-like setting. These findings are in sharp contrasts with those
found in the larger healthcare system, where racial/ethnic minorities are more likely than
Whites to die in an acute setting (Gruneir et al., 2007; Hanchate et al., 2009; A. K. Smith,
Earle, et al., 2009; Zheng et al., 2011). Given research documenting that hospice patients
who prefer to die at home often do (Jeurkar et al., 2012), it is possible that many
racial/ethnic minorities who elect hospice care favor in-home death and thus, are not
representative of the larger population. The fact that similar proportions of Whites,
Blacks, and Hispanics died in nursing homes (≈ 25%) further indicates that this
127
population of racial/ethnic minorities is unique and not necessarily representative of the
general population. Nevertheless, it is also conceivable that the value of the hospice
experience, as described previously, also influences patient preferences concerning SOD.
Accordingly, following hospice enrollment, patients who would traditionally prefer to die
in more acute settings, may revise their preferences after experiencing high quality
hospice care in the home. Notwithstanding, additional longitudinal studies that include
patients’ preferred location of death are needed to test these assertions, and further
understanding of racial/ethnic variation in hospice SOD.
This study also found that Hispanic hospice patients were more likely than White
hospice patients to die in a hospital setting, a finding that speaks to the diversity of the
U.S. Hispanic population. Studies have documented significant variation in medical
decision-making within Hispanic samples (Barnato et al., 2009; Cruz-Oliver et al., 2014),
demonstrating how a single racial/ethnic group (of multiple countries of origin) could
have significant within-group variation in healthcare preferences and usage. The finding
that Hispanic hospice patients were more likely than White hospice patients to die in both
home-like and hospital settings may serve as an illustration of this phenomenon. Similar
studies of non-hospice-specific samples have also observed this trend (Barnato et al.,
2009; Kalish & Reynolds, 1976). In a study of community-dwelling Medicare
beneficiaries, Barnato and colleagues (2009) found that Hispanics were more likely than
Whites to die in a home as well as a hospital setting. The authors concluded that although
Hispanics were more likely than Whites to die in a hospital, most preferred and
experienced in-home death. Although this manuscript was not able to examine patient
SOD preferences, it is also possible that a small portion of Hispanics in this study
128
preferred to die in a hospital setting, and the majority preferred death in a home-like
setting, as observed by Barnato and colleagues. The fact that 46% of Hispanics in this
study died in a home-like setting, and only 11% died in a hospital setting supports this
hypothesis; however, additional investigation that incorporates patient preferred location
of death is needed to test this assertion, and will likely contribute to the understanding of
potential subgroups within the Hispanic population.
Limitations and Research Recommendations
Despite the significant contributions of this study, limitations must be considered.
First, as a result of the recent redesign of the National Home Health and Hospice Care
Survey (NHHCS), data on key variables are only available for the 2007 wave and thus,
the cross-sectional nature of the study restricts time order and causality-based
conclusions. Additional studies that follow patients across the entire hospice stay will
provided added insight into causal relationships suggested by this study. Second, nominal
care preference data restrict in-depth exploration into complex constructs that guide
patient decision-making at EOL. In the future, additional measures of care preferences,
such as preferred location of death, will expand understanding of the association of these
preferences on subsequent patient outcomes. Similarly, qualitative measures beyond
simple binary responses may reveal additional aspects of patients’ preferences that are
lost in binary data collection. For example, inclusion of patient measures of religious
views, acculturation, cultural beliefs, knowledge/education, and provider
communication/trust will provide added context and explanatory power to the variation
revealed by this study. This added richness will likely improve understanding of the
relationship between patient’s preferences, care utilization, and ultimately care outcomes
129
both within and outside the hospice system. The coding structure of emergent care
utilization following hospice enrollment was extremely limited, as data was only
collected as occurred versus not occurred. Although beneficial in exploring general
differences in care use, this type of data collection prevents an in-depth examination into
the type of care that is utilized, as well as specification on the location and length of the
unplanned care episode. Emergent care utilization was also only recorded if the episode
occurred during the 60 days prior to data collection, as specified in survey data collection
guidelines. As a result, additional emergent care episodes that are beyond the 60-day
scope are lost. Furthermore, data on multiple acute care episodes are lost and thus, little is
known regarding potential cases of multiple emergent care use episodes under the care of
hospice. Future studies would benefit from data collection that accounts for these
emergent care descriptive characteristics. Finally, given the unique nature of the
emergent care use variable (i.e., atypical operationalization accounting for service use
beyond hospitalization and/or emergency department use), results from this study,
particularly comparisons between findings from this study and other studies, should be
considered cautiously. NHHCS data rely on hospice agency records, and as such patient-
data were never collected directly from the patients. Two potential complications arise
from this limitation. Given that patient-level data is based on agency medical records, it is
only as good as the record keeping system for each agency. While some agencies may
have a well-structured record system established, other agencies may rely on less
effective methods. With the large number of agencies utilized in the study (n = 657), it is
reasonable to assume that agency-level variation in record-keeping accounted for some of
the variance in missing data. Another potential complication arising from agency-level
130
record data collection is that some data may have never been collected. For example,
some patients may have formally designated a healthcare proxy, but not documented this
designation with the hospice agency, resulting in skewed data. Prospective studies would
benefit from using a combined agency- and patient-level data collection technique. Such
an approach would help account for variance in missing data. While significant
consideration was given to variable selection and analysis, it is possible that results from
death in a home-like setting are skewed due to operationalization issues. That is, it is
possible that some patients died in a setting that they considered to be their home, such as
nursing facility, but data coding prevented the correct assignment of this outcome. Future
studies investigating hospice SOD should consider utilizing patient self-identified home
settings to ensure proper representativeness. This study also did not account for agency-
level predictors that could affect outcomes of interest. For example, research suggests
that hospice characteristics, such as for-profit status, hospice size, years of certification,
and geographic location all influence patient-level outcomes (Carlson et al., 2009;
Sengupta et al., 2013; Wachterman et al., 2011). Additional studies that account for these
agency-level characteristics will likely expand understanding of the association between
patient race/ethnicity and care utilization and outcomes. As indicated above, previous
studies have documented an increased likelihood among racial/ethnic minorities to
disenroll from hospice. It is possible that this trend could be contributing to biased results
in this study, as the racial/ethnic minorities who prefer more aggressive interventions
could be disenrolling from hospice care, and thus leaving only a subset of minorities who
prefer and utilize less aggressive EOL care. Follow-up studies that longitudinally account
for disenrollment will help determine the relative effect of this trend on study outcomes.
131
While this study examined the influence of advance directive completion and DNR order
election on care utilization and outcomes, it is possible that healthcare proxy designation
may have also influenced these outcomes. Accordingly, additional examination of the
relative effect of healthcare proxy designation on racial/ethnic within-hospice variation is
needed to further understand the role of advance care planning on hospice patient care
utilization and outcomes. Although socioeconomic status was modeled through Medicaid
enrollment status in this study, a more accurate examination of socioeconomic status that
includes information on patient education and income will likely provide a more precise
estimate of the relative effect of the construct on the investigated outcomes. Lastly, as
with all national hospice data, there is limited representation of all racial/ethnic minority
groups, notably Asian/Pacific Islanders, and Native Americans/Alaska Natives. Future
investigation of potential variation across other racial/ethnic groups will increase
understanding of preferences and outcomes for other potentially vulnerable populations,
and likely lead to improved hospice care for all patients.
Conclusion
Recent calls by the United States Department of Health and Human Services for
the elimination of healthcare disparities by 2020 (USDHHS, 2012) highlights a
significant need for equitable care within the U.S. healthcare system. Studies have
repeatedly documented racial/ethnic disparities in access to and utilization of health
services across multiple settings, and despite significant improvements over the past
decade, important gaps remain. It is widely-understood that EOL decisions, and
subsequent care outcomes, substantially impact several quality indicators at EOL. As a
result, considerable effort has been made to improve racial/ethnic minority access to
132
hospice care in the United States; however, relatively few studies have investigated
racial/ethnic differences among hospice enrollees. Given known EOL disparities outside
of hospice care, further examination of disparities within the hospice system is needed to
ensure equitable care for all hospice enrollees, and address government petitions to end
disparities in the United States. This study provides the first in-depth investigation of
racial/ethnic variation in care preferences, utilization, and outcomes following hospice
enrollment. Results indicate that race/ethnicity is a significant contributor of patient-level
variation in EOL decision making and care outcomes following hospice enrollment. As
such, this study highlights an important need for future studies to advance understanding
of care differences following hospice enrollment to ultimately ensure adequate quality of
care for all patients. Equally-important, this investigation also underscores a potential
need for Medicare Hospice Benefit policy adjustments, as the current structure of the
benefit may disservice some racial/ethnic minorities who could benefit substantially from
hospice care, but do not elect it due to current restrictions (e.g., requirement to forgo
curative care). As researchers, healthcare providers, and policy makers, we have an
obligation to ensure equity throughout the healthcare system. By identifying key
differences in care preferences, utilization, and outcomes among U.S. hospice patients,
this manuscript represents a concerted step toward ensuring that that all people are cared
for equally at end of life.
133
References
Andersen, R. M. (1995). Revisiting the behavioral model and access to medical care:
Does it matter? Journal of Health and Social Behavior, 36(1), 1-10.
Azoulay, E., Pochard, F., Chevret, S., Adrie, C., Annane, D., Bleichner, G., . . .
Schlemmer, B. (2004). Half the family members of intensive care unit patients do
not want to share in the decision-making process: A study in 78 French intensive
care units. Critical Care Medicine, 32(9), 1832-1838.
Bach, P. B., Guadagnoli, E., Schrag, D., Schussler, N., & Warren, J. L. (2002). Patient
demographic and socioeconomic characteristics in the SEER-Medicare database
applications and limitations. Medical Care, 40(8 Suppl), IV-19-25.
Baker, R., Wu, A. W., Teno, J. M., Kreling, B., Damiano, A. M., Rubin, H. R., . . . Lynn,
J. (2000). Family satisfaction with end-of-life care in seriously ill hospitalized
adults. Journal of the American Geriatrics Society, 48(5 Suppl), S61-69.
Balboni, T. A., Vanderwerker, L. C., Block, S. D., Paulk, M. E., Lathan, C. S., Peteet, J.
R., & Prigerson, H. G. (2007). Religiousness and spiritual support among
advanced cancer patients and associations with end-of-life treatment preferences
and quality of life. Journal of Clinical Oncology, 25(5), 555-560.
Barnato, A. E., Anthony, D. L., Skinner, J., Gallagher, P. M., & Fisher, E. S. (2009).
Racial and ethnic differences in preferences for end-of-life treatment. Journal of
General Internal Medicine, 24(6), 695-701.
Barnato, A. E., Chang, C. C., Saynina, O., & Garber, A. M. (2007). Influence of race on
inpatient treatment intensity at the end of life. Journal of General Internal
Medicine, 22(3), 338-345.
134
Billings, J. A., & Kolton, E. (1999). Family satisfaction and bereavement care following
death in the hospital. Journal of Palliative Medicine, 2(1), 33-49.
Blackhall, L. J., Frank, G., Murphy, S. T., Michel, V., Palmer, J. M., & Azen, S. P.
(1999). Ethnicity and attitudes towards life sustaining technology. Social Science
& Medicine, 48(12), 1779-1789.
Blackhall, L. J., Murphy, S. T., Frank, G., Michel, V., & Azen, S. (1995). Ethnicity and
attitudes toward patient autonomy. Journal of the American Medical Association,
274(10), 820-825.
Blinderman, C. D., Homel, P., Billings, J. A., Portenoy, R. K., & Tennstedt, S. L. (2008).
Symptom distress and quality of life in patients with advanced congestive heart
failure. Journal of Pain & Symptom Management, 35(6), 594-603.
Blinderman, C. D., Homel, P., Billings, J. A., Tennstedt, S., & Portenoy, R. K. (2009).
Symptom distress and quality of life in patients with advanced chronic obstructive
pulmonary disease. Journal of Pain & Symptom Management, 38(1), 115-123.
Blocker, D. E., Romocki, L. S., Thomas, K. B., Jones, B. L., Jackson, E. J., Reid, L., &
Campbell, M. K. (2006). Knowledge, beliefs and barriers associated with prostate
cancer prevention and screening behaviors among African-American men.
Journal of the National Medical Association, 98(8), 1286-1295.
Borum, M. L., Lynn, J., & Zhong, Z. (2000). The effects of patient race on outcomes in
seriously ill patients in SUPPORT: An overview of economic impact, medical
intervention, and end-of-life decisions. Study to Understand Prognoses and
Preferences for Outcomes and Risks of Treatments. Journal of the American
Geriatrics Society, 48(5 Suppl), S194-198.
135
Bradley, E. H., McGraw, S. A., Curry, L., Buckser, A., King, K. L., Kasl, S. V., &
Andersen, R. (2002). Expanding the Andersen Model: The Role of Psychosocial
Factors in Long ‐Term Care Use. Health Services Research, 37(5), 1221-1242.
Braun, U. K., Beyth, R. J., Ford, M. E., & McCullough, L. B. (2008). Voices of African
American, Caucasian, and Hispanic surrogates on the burdens of end-of-life
decision making. Journal of General Internal Medicine, 23(3), 267-274.
Braunstein, J. B., Sherber, N. S., Schulman, S. P., Ding, E. L., & Powe, N. R. (2008).
Race, medical researcher distrust, perceived harm, and willingness to participate
in cardiovascular prevention trials. Medicine, 87(1), 1-9.
Brinkman-Stoppelenburg, A., Rietjens, J. A., & van der Heide, A. (2014). The effects of
advance care planning on end-of-life care: A systematic review. Palliative
Medicine, 0(0), 1-26.
Bullock, K. (2006). Promoting advance directives among African Americans: A faith-
based model. Journal of Palliative Medicine, 9(1), 183-195.
Burge, F., Lawson, B., Johnston, G., & Cummings, I. (2003). Primary care continuity and
location of death for those with cancer. Journal of Palliative Medicine, 6(6), 911-
918.
Burrs, F. A. (1995). The African American experience: breaking the barriers to hospices.
The Hospice Journal, 10(2), 15-18.
Byock, I. R., Forman, W. B., & Appleton, M. (1996). Academy of hospice physicians'
position statement on access to hospice and palliative care. Journal of Pain &
Symptom Management, 11(2), 69-70.
136
Carlson, M. D., Herrin, J., Du, Q., Epstein, A. J., Cherlin, E., Morrison, R. S., & Bradley,
E. H. (2009). Hospice characteristics and the disenrollment of patients with
cancer. Health Services Research, 44(6), 2004-2021.
Carr, D. (2011). Racial Differences in End-of-Life Planning: Why Don't Blacks and
Latinos Prepare for the Inevitable? Omega: Journal of Death and Dying, 63(1), 1-
20.
Casarett, D. J., Crowley, R. L., & Hirschman, K. B. (2004). How should clinicians
describe hospice to patients and families? Journal of the American Geriatrics
Society, 52(11), 1923-1928.
Casarett, D. J., Karlawish, J., Morales, K., Crowley, R., Mirsch, T., & Asch, D. A.
(2005). Improving the use of hospice services in nursing homes: A randomized
controlled trial. Journal of the American Medical Association, 294(2), 211-217.
Casarett, D. J., Van Ness, P. H., O'Leary, J. R., & Fried, T. R. (2006). Are patient
preferences for life-sustaining treatment really a barrier to hospice enrollment for
older adults with serious illness? Journal of the American Geriatrics Society,
54(3), 472-478.
Cene, C. W., Roter, D., Carson, K. A., Miller, E. R., 3rd, & Cooper, L. A. (2009). The
effect of patient race and blood pressure control on patient-physician
communication. Journal of General Internal Medicine, 24(9), 1057-1064.
Centers for Disease Control and Prevention, National Center for Health Statistics, &
National Vital Stastistics System - Mortality. (2008). Deaths by place of death,
age, race, and sex: United States, 2005.
137
Christakis, N. A., & Escarce, J. J. (1996). Survival of Medicare patients after enrollment
in hospice programs. New England Journal of Medicine, 335(3), 172-178.
Christakis, N. A., & Iwashyna, T. J. (1998). Attitude and self-reported practice regarding
prognostication in a national sample of internists. Archives of Internal Medicine,
158(21), 2389-2395.
Christakis, N. A., & Iwashyna, T. J. (2000). Impact of individual and market factors on
the timing of initiation of hospice terminal care. Medical Care, 38(5), 528-541.
Chung, K., Essex, E. L., & Samson, L. (2009). Does caregiver knowledge matter for
hospice enrollment and beyond? Pilot study of minority hospice patients.
American Journal of Hospice and Palliative Medicine, 26(3), 165-171.
Cintron, A., Hamel, M. B., Davis, R. B., Burns, R. B., Phillips, R. S., & McCarthy, E. P.
(2003). Hospitalization of hospice patients with cancer. Journal of Palliative
Medicine, 6(5), 757-768.
Cloud, J. (2000). A kinder, gentler death. Time, 156(12), 60-67.
Cohen, J. (1977). Statistical power analysis for the behavioral sciences. New York:
Academic Press.
Cohen, L. L. (2008). Racial/ethnic disparities in hospice care: A systematic review.
Journal of Palliative Medicine, 11(5), 763-768.
Cólon, M., & Lyke, J. (2003). Comparison of hospice use and demographics among
European Americans, African Americans, and Latinos. American Journal of
Hospice & Palliative Care, 20(3), 182-190.
138
Conner, N. E. (2012). Predictive Factors of Hospice Use Among Blacks Applying
Andersen's Behavioral Model. American Journal of Hospice and Palliative
Medicine, 29(5), 368-374.
Connor, S. R. (2007). Development of hospice and palliative care in the United States.
Omega: Journal of Death and Dying, 56(1), 89-99.
Connor, S. R., Elwert, F., Spence, C., & Christakis, N. A. (2008). Racial disparity in
hospice use in the United States in 2002. Palliative Medicine, 22(3), 205-213.
Connor, S. R., Pyenson, B., Fitch, K., Spence, C., & Iwasaki, K. (2007). Comparing
hospice and nonhospice patient survival among patients who die within a three-
year window. Journal of Pain & Symptom Management, 33(3), 238-246.
Connors, A. F., Dawson, N. V., Desbiens, N. A., Fulkerson, W. J., Goldman, L., Knaus,
W. A., . . . Damiano, A. (1995). A Controlled Trial to Improve Care for Seriously
III Hospitalized Patients. Journal of the American Medical Association, 274(20),
1591.
Cooper-Patrick, L., Gallo, J. J., Gonzales, J. J., Vu, H. T., Powe, N. R., Nelson, C., &
Ford, D. E. (1999). Race, gender, and partnership in the patient-physician
relationship. Journal of the American Medical Association, 282(6), 583-589.
Cruz-Oliver, D. M., Talamantes, M., & Sanchez-Reilly, S. (2014). What Evidence is
Available on End-of-life (EOL) Care and Latino Elders? A Literature Review.
American Journal of Hospice and Palliative Medicine, 31(1), 87-97.
Daugherty, C. K., & Steensma, D. P. (2003). Overcoming obstacles to hospice care: an
ethical examination of inertia and inaction. Journal of Clinical Oncology, 21(9
suppl), 42-45.
139
Degenholtz, H. B., Rhee, Y. J., & Arnold, R. M. (2004). Brief communication: The
relationship between having a living will and dying in place. Annals of Internal
Medicine, 141(2), 113.
Del Gaudio, F., Hichenberg, S., Eisenberg, M., Kerr, E., Zaider, T., & Kissane, D.
(2013). Latino Values in the Context of Palliative Care Illustrative Cases From
the Family Focused Grief Therapy Trial. American Journal of Hospice and
Palliative Medicine, 30(3), 271-278.
Detering, K. M., Hancock, A. D., Reade, M. C., & Silvester, W. (2010). The impact of
advance care planning on end of life care in elderly patients: Randomised
controlled trial. British Medical Journal, 340.
Duffy, S. A., Jackson, F. C., Schim, S. M., Ronis, D. L., & Fowler, K. E. (2006).
Racial/ethnic preferences, sex preferences, and perceived discrimination related to
end-of-life care. Journal of the American Geriatrics Society, 54(1), 150-157.
Dwyer, L. L., Harris-Kojetin, L. D., Branden, L., & Shimizu, I. M. (2010). Redesign and
operation of the National Home And Hospice Care Survey, 2007. Vital and
Health Statistics. Series 1: Programs and collection procedures(53), 1-192.
Dy, S. M., Shugarman, L. R., Lorenz, K. A., Mularski, R. A., Lynn, J., & Center, R. A.-S.
C. E.-B. P. (2008). A systematic review of satisfaction with care at the end of life.
Journal of the American Geriatrics Society, 56(1), 124-129.
Earle, C. C., Landrum, M. B., Souza, J. M., Neville, B. A., Weeks, J. C., & Ayanian, J. Z.
(2008). Aggressiveness of cancer care near the end of life: Is it a quality-of-care
issue? Journal of Clinical Oncology, 26(23), 3860-3866.
140
Earle, C. C., Neville, B. A., Landrum, M. B., Ayanian, J. Z., Block, S. D., & Weeks, J. C.
(2004). Trends in the aggressiveness of cancer care near the end of life. Journal of
Clinical Oncology, 22(2), 315-321.
Egede, L. E. (2006). Race, ethnicity, culture, and disparities in health care. Journal of
General Internal Medicine, 21(6), 667-669.
Eleazer, G. P., Hornung, C. A., Egbert, C. B., Egbert, J. R., Eng, C., Hedgepeth, J., . . .
Wilson, M. (1996). The relationship between ethnicity and advance directives in a
frail older population. Journal of the American Geriatrics Society, 44(8), 938-943.
Engle, V. F. (1998). Care of the living, care of the dying: reconceptualizing nursing home
care. Journal of the American Geriatrics Society, 46(9), 1172-1174.
Enguidanos, S., Kogan, A. C., Lorenz, K., & Taylor, G. (2011). Use of role model stories
to overcome barriers to hospice among African Americans. Journal of Palliative
Medicine, 14(2), 161-168.
Enguidanos, S., Yip, J., & Wilber, K. (2005). Ethnic variation in site of death of older
adults dually eligible for Medicaid and Medicare. Journal of the American
Geriatrics Society, 53(8), 1411-1416.
Fagerlin, A., & Schneider, C. E. (2004). Enough: The failure of the living will. Hastings
Center Report, 34(2), 30-42.
Fischer, G. S., Tulsky, J. A., Rose, M. R., Siminoff, L. A., & Arnold, R. M. (1998).
Patient knowledge and physician predictions of treatment preferences after
discussion of advance directives. Journal of General Internal Medicine, 13(7),
447-454.
141
Fowler-Brown, A., Ashkin, E., Corbie-Smith, G., Thaker, S., & Pathman, D. E. (2006).
Perception of racial barriers to health care in the rural South. Journal of Health
Care for the Poor & Underserved, 17(1), 86-100.
Galambos, C. M. (1998). Preserving end-of-life autonomy: The patient self-determination
act and the uniform health care decisions act. Health & Social Work, 23(4), 275-
281.
Gamble, V. N. (1997). Under the shadow of Tuskegee: African Americans and health
care. American Journal of Public Health, 87(11), 1773-1778.
Ghods, B. K., Roter, D. L., Ford, D. E., Larson, S., Arbelaez, J. J., & Cooper, L. A.
(2008). Patient-physician communication in the primary care visits of African
Americans and whites with depression. Journal of General Internal Medicine,
23(5), 600-606.
Givens, J. L., Tjia, J., Zhou, C., Emanuel, E., & Ash, A. S. (2010). Racial and ethnic
differences in hospice use among patients with heart failure. Archives of Internal
Medicine, 170(5), 427-432.
Goldstein, E., Elliott, M. N., Lehrman, W. G., Hambarsoomian, K., & Giordano, L. A.
(2010). Racial/Ethnic differences in patients' perceptions of inpatient care using
the HCAHPS survey. Medical Care Research and Review, 67(1), 74-92.
Goodlin, S. J., Winzelberg, G. S., Teno, J. M., Whedon, M., & Lynn, J. (1998). Death in
the hospital. Archives of Internal Medicine, 158(14), 1570-1572.
Gordon, A. K. (1995). Deterrents to access and service for blacks and Hispanics: The
Medicare Hospice Benefit, healthcare utilization, and cultural barriers. The
Hospice Journal, 10(2), 65-83.
142
Gozalo, P. L., & Miller, S. C. (2007). Hospice enrollment and evaluation of its causal
effect on hospitalization of dying nursing home patients. Health Services
Research, 42(2), 587-610.
Gozalo, P. L., Teno, J. M., Mitchell, S. L., Skinner, J., Bynum, J., Tyler, D., & Mor, V.
(2011). End-of-life transitions among nursing home residents with cognitive
issues. New England Journal of Medicine, 365(13), 1212-1221.
Greer, D. S., Mor, V., Morris, J. N., Sherwood, S., Kidder, D., & Birnbaum, H. (1986).
An alternative in terminal care: Results of the National Hospice Study. Journal of
Chronic Diseases, 39(1), 9-26.
Greiner, K. A., Perera, S., & Ahluwalia, J. S. (2003). Hospice usage by minorities in the
last year of life: Results from the National Mortality Followback Survey. Journal
of the American Geriatrics Society, 51(7), 970-978.
Gross, C. P., Filardo, G., Mayne, S. T., & Krumholz, H. M. (2005). The impact of
socioeconomic status and race on trial participation for older women with breast
cancer. Cancer, 103(3), 483-491.
Gruneir, A., Mor, V., Weitzen, S., Truchil, R., Teno, J., & Roy, J. (2007). Where people
die: A multilevel approach to understanding influences on site of death in
America. Medical Care Research and Review, 64(4), 351-378.
Haber, D. (1999). Minority access to hospice. American Journal of Hospice & Palliative
Care, 16(1), 386-389.
Hackbarth, G., Reischauer, R., & Miller, M. (2009). Report to the Congress: Medicare
payment policy. Testimony on behalf of MedPAC, March17.
143
Han, B., Remsburg, R. E., & Iwashyna, T. J. (2006). Differences in hospice use between
black and white patients during the period 1992 through 2000. Medical Care,
44(8), 731-737.
Han, B., Remsburg, R. E., McAuley, W. J., Keay, T. J., & Travis, S. S. (2007). Length of
hospice care among U.S. adults: 1992-2000. Inquiry, 44(1), 104-113.
Hanchate, A., Kronman, A. C., Young-Xu, Y., Ash, A. S., & Emanuel, E. (2009). Racial
and ethnic differences in end-of-life costs: Why do minorities cost more than
whites? Archives of Internal Medicine, 169(5), 493-501.
Hanlon, J. T., Perera, S., Sevick, M. A., Rodriguez, K. L., & Jaffe, E. J. (2010). Pain and
its treatment in older nursing home hospice/palliative care residents. Journal of
the American Medical Directors Association, 11(8), 579-583.
Hansen, S. M., Tolle, S. W., & Martin, D. P. (2002). Factors associated with lower rates
of in-hospital death. Journal of Palliative Medicine, 5(5), 677-685.
Hardy, D., Chan, W., Liu, C. C., Cormier, J. N., Xia, R., Bruera, E., & Du, X. L. (2012).
Racial disparities in length of stay in hospice care by tumor stage in a large
elderly cohort with non-small cell lung cancer. Palliative Medicine, 26(1), 61-71.
Harris, Y., Gorelick, P. B., Samuels, P., & Bempong, I. (1996). Why African Americans
may not be participating in clinical trials. Journal of the National Medical
Association, 88(10), 630-634.
Haupt, B. J. (2003). Characteristics of hospice care discharges and their length of service:
United States, 2000. Vital and health statistics. Series 13: Data from the National
Health Survey(154), 1-36.
144
Head, B., Ritchie, C. S., & Smoot, T. M. (2005). Prognostication in hospice care: can the
palliative performance scale help? Journal of Palliative Medicine, 8(3), 492-502.
doi: 10.1089/jpm.2005.8.492
Higginson, I. J., Astin, P., & Dolan, S. (1998). Where do cancer patients die? Ten-year
trends in the place of death of cancer patients in England. Palliative Medicine,
12(5), 353-363.
Higginson, I. J., & Sen-Gupta, G. (2000). Place of care in advanced cancer: A qualitative
systematic literature review of patient preferences. Journal of Palliative Medicine,
3(3), 287-300.
Hoffmann, D. E., Zimmerman, S. I., & Tompkins, C. J. (1996). The dangers of directives
or the false security of forms. The Journal of Law, Medicine & Ethics, 24(1), 5-
17.
Hogan, C., Lunney, J., Gabel, J., & Lynn, J. (2001). Medicare beneficiaries' costs of care
in the last year of life. Health Affairs, 20(4), 188-195.
Hopp, F. P., & Duffy, S. A. (2000). Racial variations in end-of-life care. Journal of the
American Geriatrics Society, 48(6), 658-663.
Hughes, S. L., Ulasevich, A., Weaver, F. M., Henderson, W., Manheim, L., Kubal, J. D.,
& Bonarigo, F. (1997). Impact of home care on hospital days: A meta analysis.
Health Services Research, 32(4), 415-432.
Huskamp, H. A., Buntin, M. B., Wang, V., & Newhouse, J. P. (2001). Providing care at
the end of life: Do Medicare rules impede good care? Health Affairs, 20(3), 204-
211.
145
Huskamp, H. A., Keating, N. L., Malin, J. L., Zaslavsky, A. M., Weeks, J. C., Earle, C.
C., . . . He, Y. (2009). Discussions with physicians about hospice among patients
with metastatic lung cancer. Archives of Internal Medicine, 169(10), 954.
Jeurkar, N., Farrington, S., Craig, T. R., Slattery, J., Harrold, J. K., Oldanie, B., . . .
Casarett, D. J. (2012). Which hospice patients with cancer are able to die in the
setting of their choice? Results of a retrospective cohort study. Journal of Clinical
Oncology, 30(22), 2783-2787.
Johnson, K. S., Kuchibhatla, M., Sloane, R. J., Tanis, D., Galanos, A. N., & Tulsky, J. A.
(2005). Ethnic differences in the place of death of elderly hospice enrollees.
Journal of the American Geriatrics Society, 53(12), 2209-2215.
Johnson, K. S., Kuchibhatla, M., Tanis, D., & Tulsky, J. A. (2008). Racial differences in
hospice revocation to pursue aggressive care. Archives of Internal Medicine,
168(2), 218-224.
Johnson, K. S., Kuchibhatla, M., & Tulsky, J. A. (2008). What explains racial differences
in the use of advance directives and attitudes toward hospice care? Journal of the
American Geriatrics Society, 56(10), 1953-1958.
Johnson, K. S., Kuchibhatla, M., & Tulsky, J. A. (2011). Racial differences in location
before hospice enrollment and association with hospice length of stay. Journal of
the American Geriatrics Society, 59(4), 732-737.
Johnson, R. L., Roter, D., Powe, N. R., & Cooper, L. A. (2004). Patient race/ethnicity and
quality of patient-physician communication during medical visits. American
Journal of Public Health, 94(12), 2084-2090.
146
Jones, A. L., Moss, A. J., & Harris-Kojetin, L. D. (2011). Use of advance directives in
long-term care populations. NCHS Data Brief(54), 1-8.
Kagawa-Singer, M., & Blackhall, L. J. (2001). Negotiating cross-cultural issues at the
end of life: "You got to go where he lives". Journal of the American Medical
Association, 286(23), 2993-3001.
Kalish, R. A., & Reynolds, D. K. (1976). Death and ethnicity: A psychocultural study:
Ethel Percy Andrus Gerontology Center, University of Southern California.
Kane, R. L., Bernstein, L., Wales, J., Leibowitz, A., & Kaplan, S. (1984). A randomised
controlled trial of hospice care. The Lancet, 323(8382), 890-894.
Kapo, J., Harrold, J., Carroll, J. T., Rickerson, E., & Casarett, D. (2005). Are we referring
patients to hospice too late? Patients' and families' opinions. Journal of Palliative
Medicine, 8(3), 521-527.
Kapo, J., MacMoran, H., & Casarett, D. (2005). "Lost to follow-up": Ethnic disparities in
continuity of hospice care at the end of life. Journal of Palliative Medicine, 8(3),
603-608.
Kelley, A. S., Wenger, N. S., & Sarkisian, C. A. (2010). Opiniones: End-of-life care
preferences and planning of older Latinos. Journal of the American Geriatrics
Society, 58(6), 1109-1116.
Kramer, B. J., Kavanaugh, M., Trentham-Dietz, A., Walsh, M., & Yonker, J. A. (2010).
Complicated grief symptoms in caregivers of persons with lung cancer: The role
of family conflict, intrapsychic strains, and hospice utilization. Omega: Journal of
Death and Dying, 62(3), 201-220.
147
Krause, N., & Bastida, E. (2011). Religion, suffering, and self-rated health among older
Mexican Americans. The Journals of Gerontology Series B: Psychological
Sciences and Social Sciences, 66(2), 207-216.
Kumar, G., Markert, R. J., & Patel, R. (2011). Assessment of hospice patients' goals of
care at the end of life. American Journal of Hospice & Palliative Care, 28(1), 31-
34.
Kunik, M. E., Roundy, K., Veazey, C., Souchek, J., Richardson, P., Wray, N. P., &
Stanley, M. A. (2005). Surprisingly high prevalence of anxiety and depression in
chronic breathing disorders. Chest, 127(4), 1205-1211.
Kwak, J., & Haley, W. E. (2005). Current research findings on end-of-life decision
making among racially or ethnically diverse groups. Gerontologist, 45(5), 634-
641.
Kwak, J., Haley, W. E., & Chiriboga, D. A. (2008). Racial differences in hospice use and
in-hospital death among Medicare and Medicaid dual-eligible nursing home
residents. Gerontologist, 48(1), 32-41.
Lackan, N. A., Eschbach, K., Stimpson, J. P., Freeman, J. L., & Goodwin, J. S. (2009).
Ethnic differences in in-hospital place of death among older adults in California:
Effects of individual and contextual characteristics and medical resource supply.
Medical Care, 47(2), 138-145.
Laguna, J., Enguídanos, S., Siciliano, M., & Coulourides-Kogan, A. (2012).
Racial/Ethnic Minority Access to End-of-Life Care: A Conceptual Framework.
Home Health Care Services Quarterly, 31(1), 60-83.
148
LaVeist, T. A., Nuru-Jeter, A., & Jones, K. E. (2003). The association of doctor-patient
race concordance with health services utilization. Journal of Public Health Policy,
312-323.
Leff, B., Burton, L., Mader, S., Naughton, B., Burl, J., Clark, R., . . . Burton, J. R. (2006).
Satisfaction with hospital at home care. Journal of the American Geriatrics
Society, 54(9), 1355-1363.
Leff, B., Burton, L., Mader, S. L., Naughton, B., Burl, J., Inouye, S. K., . . . Burton, J. R.
(2005). Hospital at home: Feasibility and outcomes of a program to provide
hospital-level care at home for acutely ill older patients. Annals of Internal
Medicine, 143(11), 798-808.
Legler, A., Bradley, E. H., & Carlson, M. D. (2011). The effect of comorbidity burden on
health care utilization for patients with cancer using hospice. Journal of Palliative
Medicine, 14(6), 751-756.
Lepore, M. J., Miller, S. C., & Gozalo, P. (2011). Hospice use among urban Black and
White U.S. nursing home decedents in 2006. Gerontologist, 51(2), 251-260.
Levinsky, N. G., Yu, W., Ash, A., Moskowitz, M., Gazelle, G., Saynina, O., & Emanuel,
E. J. (2001). Influence of age on Medicare expenditures and medical care in the
last year of life. Journal of the American Medical Association, 286(11), 1349-
1355.
Lo, B., & Steinbrook, R. (2004). Resuscitating advance directives. Archives of Internal
Medicine, 164(14), 1501-1506.
Loggers, E. T., Maciejewski, P. K., Jimenez, R., Nilsson, M., Paulk, E., Stieglitz, H., &
Prigerson, H. G. (2013). Predictors of intensive end-of-life and hospice care in
149
Latino and White advanced cancer patients. Journal of Palliative Medicine,
16(10), 1249-1254.
Loggers, E. T., Maciejewski, P. K., Paulk, E., DeSanto-Madeya, S., Nilsson, M.,
Viswanath, K., . . . Prigerson, H. G. (2009). Racial differences in predictors of
intensive end-of-life care in patients with advanced cancer. Journal of Clinical
Oncology, 27(33), 5559-5564.
Lubitz, J. D., & Riley, G. F. (1993). Trends in Medicare payments in the last year of life.
New England Journal of Medicine, 328(15), 1092-1096.
Lutz, W., Sanderson, W., & Scherbov, S. (2008). The coming acceleration of global
population ageing. Nature, 451(7179), 716-719.
Lynn, J., Teno, J. M., Phillips, R. S., Wu, A. W., Desbiens, N., Harrold, J., . . . Connors,
A. F., Jr. (1997). Perceptions by family members of the dying experience of older
and seriously ill patients. SUPPORT Investigators. Study to Understand
Prognoses and Preferences for Outcomes and Risks of Treatments. Annals of
Internal Medicine, 126(2), 97-106.
Maas, C. J., & Snijders, T. A. (2003). The multilevel approach to repeated measures for
complete and incomplete data. Quality and Quantity, 37(1), 71-89.
Mack, J. W., Paulk, M. E., Viswanath, K., & Prigerson, H. G. (2010). Racial disparities
in the outcomes of communication on medical care received near death. Annals of
Internal Medicine, 170(17), 1533-1540.
Mack, J. W., Weeks, J. C., Wright, A. A., Block, S. D., & Prigerson, H. G. (2010). End-
of-life discussions, goal attainment, and distress at the end of life: Predictors and
150
outcomes of receipt of care consistent with preferences. Journal of Clinical
Oncology, 28(7), 1203-1208.
Mazanec, P. M., Daly, B. J., & Townsend, A. (2010). Hospice utilization and end-of-life
care decision making of African Americans. American Journal of Hospice &
Palliative Care, 27(8), 560-566.
McCarthy, E. P., Burns, R. B., Davis, R. B., & Phillips, R. S. (2003). Barriers to hospice
care among older patients dying with lung and colorectal cancer. Journal of
Clinical Oncology, 21(4), 728-735.
McCarthy, E. P., Burns, R. B., Ngo-Metzger, Q., Davis, R. B., & Phillips, R. S. (2003).
Hospice use among Medicare managed care and fee-for-service patients dying
with cancer. Journal of the American Medical Association, 289(17), 2238-2245.
McCusker, J., & Stoddard, A. M. (1987). Effects of an expanding home care program for
the terminally ill. Medical Care, 25(5), 373-385.
McGorty, E. K., & Bornstein, B. H. (2003). Barriers to physicians' decisions to discuss
hospice: Insights gained from the United States hospice model. Journal of
Evaluation in Clinical Practice, 9(3), 363-372.
McKinley, E. D., Garrett, J. M., Evans, A. T., & Danis, M. (1996). Differences in end-of-
life decision making among black and white ambulatory cancer patients. Journal
of General Internal Medicine, 11(11), 651-656.
Meier, D. E. (2003). Old age and care near the end of life Geriatric Medicine (pp. 281-
285): Springer.
151
Mezey, M., Dubler, N. N., Mitty, E., & Brody, A. A. (2002). What impact do setting and
transitions have on the quality of life at the end of life and the quality of the dying
process? Gerontologist, 42 Spec No 3, 54-67.
Miller, S. C., Gozalo, P., & Mor, V. (2001). Hospice enrollment and hospitalization of
dying nursing home patients. The American Journal of Medicine, 111(1), 38-44.
Miller, S. C., Kinzbrunner, B., Pettit, P., & Williams, J. R. (2003). How does the timing
of hospice referral influence hospice care in the last days of life? Journal of the
American Geriatrics Society, 51(6), 798-806.
Miller, S. C., Mor, V., & Teno, J. M. (2003). Hospice enrollment and pain assessment
and management in nursing homes. Journal of Pain & Symptom Management,
26(3), 791-799.
Miller, S. C., Weitzen, S., & Kinzbrunner, B. (2003). Factors associated with the high
prevalence of short hospice stays. Journal of Palliative Medicine, 6(5), 725-736.
Mitchell, B. L., & Mitchell, L. C. (2009). Review of the literature on cultural competence
and end-of-life treatment decisions: The role of the hospitalist. Journal of the
National Medical Association, 101(9), 920-926.
Modi, S., Whetstone, L., & Cummings, D. (2007). Influence of patient and physician
characteristics on percutaneous endoscopic gastrostomy tube decision-making.
Journal of Palliative Medicine, 10(2), 359-366.
Moinpour, C. M., & Polissar, L. (1989). Factors affecting place of death of hospice and
non-hospice cancer patients. American Journal of Public Health, 79(11), 1549-
1551.
152
Morrison, R. S., Zayas, L. H., Mulvihill, M., Baskin, S. A., & Meier, D. E. (1998).
Barriers to completion of health care proxies: an examination of ethnic
differences. Archives of Internal Medicine, 158(22), 2493-2497.
Mularski, R. A., Puntillo, K., Varkey, B., Erstad, B. L., Grap, M. J., Gilbert, H. C., . . .
Sessler, C. N. (2009). Pain management within the palliative and end-of-life care
experience in the ICU. Chest, 135(5), 1360-1369.
Muni, S., Engelberg, R. A., Treece, P. D., Dotolo, D., & Curtis, J. R. (2011). The
influence of race/ethnicity and socioeconomic status on end-of-life care in the
ICU. Chest, 139(5), 1025-1033.
National Center for Health Statistics. (2011). Health, United States, 2010: With Special
Feature on Death and Dying. Hyattsville, MD.
National Hospice and Palliative Care Organization. (2008). NHPCO Facts and Figures:
Hospice Care in America. Alexandria, VA.
National Hospice and Palliative Care Organization. (2010). NHPCO Facts and Figures:
Hospice Care in America. Alexandria, VA.
National Hospice and Palliative Care Organization. (2012). NHPCO Facts and Figures:
Hospice Care in America. Alexandria, VA.
National Hospice and Palliative Care Organization. (2013). NHPCO Facts and Figures:
Hospice Care in America. Alexandria, VA.
Nelson, J. E., Meier, D. E., Oei, E. J., Nierman, D. M., Senzel, R. S., Manfredi, P. L., . . .
Morrison, R. S. (2001). Self-reported symptom experience of critically ill cancer
patients receiving intensive care. Critical Care Medicine, 29(2), 277-282.
153
Neubauer, B. J., & Hamilton, C. L. (1990). Racial differences in attitudes toward hospice
care. The Hospice Journal, 6(1), 37-48.
Nezlek, J. B. (2008). An introduction to multilevel modeling for social and personality
psychology. Social and Personality Psychology Compass, 2(2), 842-860.
Ngo-Metzger, Q., McCarthy, E. P., Burns, R. B., Davis, R. B., Li, F. P., & Phillips, R. S.
(2003). Older Asian Americans and Pacific Islanders dying of cancer use hospice
less frequently than older white patients. The American Journal of Medicine,
115(1), 47-53.
Ngo-Metzger, Q., Phillips, R. S., & McCarthy, E. P. (2008). Ethnic disparities in hospice
use among Asian-American and Pacific Islander patients dying with cancer.
Journal of the American Geriatrics Society, 56(1), 139-144.
O'Mahony, S., McHenry, J., Snow, D., Cassin, C., Schumacher, D., & Selwyn, P. A.
(2008). A review of barriers to utilization of the medicare hospice benefits in
urban populations and strategies for enhanced access. Journal of Urban Health,
85(2), 281-290.
Olsen, M. L., Bartlett, A. L., & Moynihan, T. J. (2011). Characterizing care of hospice
patients in the hospital setting. Journal of Palliative Medicine, 14(2), 185-189.
Omnibus Reconciliation Act of 1990, Sections 4206, 4751 C.F.R. (1990).
Ortman, J. M., & Guarneri, C. E. (2009). United States population projections: 2000 to
2050: United States Census Bureau.
Ouslander, J. G., Tymchuk, A. J., & Rahbar, B. (1989). Health care decisions among
elderly long-term care residents and their potential proxies. Archives of internal
medicine, 149(6), 1367-1372.
154
Parr, J. D., Zhang, B., Nilsson, M. E., Wright, A., Balboni, T., Duthie, E., . . . Prigerson,
H. G. (2010). The influence of age on the likelihood of receiving end-of-life care
consistent with patient treatment preferences. Journal of Palliative Medicine,
13(6), 719-726.
The Patient Protection and Affordable Care Act, Pub. L. No. 111-148, 124 Stat. 119
through 124 Stat. 1025 Stat. (2010).
Patrick, D. L., Curtis, J. R., Engelberg, R. A., Nielsen, E., & McCown, E. (2003).
Measuring and improving the quality of dying and death. Annals of Internal
Medicine, 139(5 Pt 2), 410-415.
Payne, R. (2001). Palliative Care for African Americans and Other Vulnerable
Populations: Access and Quality Issue. In K. M. Foley & H. Gelband (Eds.),
Improving Palliative care for Cancer (pp. 153-160). Washington, D.C.: National
Academy Press.
Perkins, H. S. (2007). Controlling death: The false promise of advance directives. Annals
of Internal Medicine, 147(1), 51-57.
Perkins, H. S., Geppert, C. M., Gonzales, A., Cortez, J. D., & Hazuda, H. P. (2002).
Cross-cultural similarities and differences in attitudes about advance care
planning. Journal of General Internal Medicine, 17(1), 48-57.
Phelps, A. C., Maciejewski, P. K., Nilsson, M., Balboni, T. A., Wright, A. A., Paulk, M.
E., . . . Prigerson, H. G. (2009). Religious coping and use of intensive life-
prolonging care near death in patients with advanced cancer. Journal of the
American Medical Association, 301(11), 1140-1147.
155
Prigerson, H. G. (1991). Determinants of hospice utilization among terminally ill
geriatric patients. Home Health Care Services Quarterly, 12(4), 81-112.
Pritchard, R. S., Fisher, E. S., Teno, J. M., Sharp, S. M., Reding, D. J., Knaus, W. A., . . .
Lynn, J. (1998). Influence of patient preferences and local health system
characteristics on the place of death. SUPPORT Investigators. Study to
Understand Prognoses and Preferences for Risks and Outcomes of Treatment.
Journal of the American Geriatrics Society, 46(10), 1242-1250.
United States Department of Health and Human Services; Office of Disease Prevention
and Health Promotion (2012). Healthy People 2020. Washington, DC.
Pyenson, B., Connor, S., Fitch, K., & Kinzbrunner, B. (2004). Medicare cost in matched
hospice and non-hospice cohorts. Journal of Pain & Symptom Management,
28(3), 200-210.
Quill, T. E. (2007). Is length of stay on hospice a critical quality of care indicator?
Journal of Palliative Medicine, 10(2), 290-292.
Rabe-Hesketh, S., Skrondal, A., & Pickles, A. (2005). Maximum likelihood estimation of
limited and discrete dependent variable models with nested random effects.
Journal of Econometrics, 128(2), 301-323.
Reese, D. J., Ahern, R. E., Nair, S., O'Faire, J. D., & Warren, C. (1999). Hospice access
and use by African Americans: Addressing cultural and institutional barriers
through participatory action research. Social Work, 44(6), 549-559.
Resnick, H. E., Hickman, S., & Foster, G. L. (2012). Documentation of Advance
Directives Among Home Health and Hospice Patients: United States, 2007.
American Journal of Hospice and Palliative Medicine, 29(2), 26-35.
156
Rhodes, R. L., Teno, J. M., & Connor, S. R. (2007). African American bereaved family
members' perceptions of the quality of hospice care: Lessened disparities, but
opportunities to improve remain. Journal of Pain & Symptom Management,
34(5), 472-479.
Rickerson, E., Harrold, J., Kapo, J., Carroll, J. T., & Casarett, D. (2005). Timing of
hospice referral and families' perceptions of services: are earlier hospice referrals
better? Journal of the American Geriatrics Society, 53(5), 819-823.
Sacco, J., Carr, D. R. D., & Viola, D. (2013). The effects of the palliative medicine
consultation on the DNR status of African Americans in a safety-net hospital.
American Journal of Hospice and Palliative Medicine, 30(4), 363-369.
Satterthwaite, F. (1946). Of variance components. Biometrics, 2(6), 110-114.
Schockett, E. R., Teno, J. M., Miller, S. C., & Stuart, B. (2005). Late referral to hospice
and bereaved family member perception of quality of end-of-life care. Journal of
Pain & Symptom Management, 30(5), 400-407.
Schonwetter, R. S., Clark, L. D., Leedy, S. A., Quinn, M. J., Azer, M., & Kim, S. (2008).
Predicting emergency room visits and hospitalizations among hospice patients
with cardiac disease. Journal of Palliative Medicine, 11(8), 1142-1150.
Searight, H. R., & Gafford, J. (2005). Cultural diversity at the end of life: Issues and
guidelines for family physicians. American Family Physician, 71(3), 515-522.
Selman, L., Harding, R., Beynon, T., Hodson, F., Coady, E., Hazeldine, C., . . .
Higginson, I. J. (2007). Improving end-of-life care for patients with chronic heart
failure: "Let's hope it'll get better, when I know in my heart of hearts it won't".
Heart, 93(8), 963-967.
157
Sengupta, M., Park-Lee, E., Valverde, R., Caffrey, C., & Jones, A. (2013). Trends in
Length of Hospice Care From 1996 to 2007 and the Factors Associated With
Length of Hospice Care in 2007: Findings From the National Home and Hospice
Care Surveys. American Journal of Hospice and Palliative Medicine, 0(0), 1-9.
Services, C. f. M. a. M. (2014). Your Medicare Benefits. Available at:
http://www.medicare.gov/publications/pubs/pdf/10116.pdf. Accessed
April 28, 2014.
Shugarman, L. R., Campbell, D. E., Bird, C. E., Gabel, J., T, A. L., & Lynn, J. (2004).
Differences in Medicare expenditures during the last 3 years of life. Journal of
General Internal Medicine, 19(2), 127-135.
Silveira, M. J., Kim, S. Y. H., & Langa, K. M. (2010). Advance directives and outcomes
of surrogate decision making before death. New England Journal of Medicine,
362(13), 1211-1218.
Singer, J. D. (1987). An intraclass correlation model for analyzing multilevel data.
Journal of Experimental Education.
Smith, A. K., Earle, C. C., & McCarthy, E. P. (2009). Racial and ethnic differences in
end-of-life care in fee-for-service Medicare beneficiaries with advanced cancer.
Journal of the American Geriatrics Society, 57(1), 153-158.
Smith, A. K., Sudore, R. L., & Pérez-Stable, E. J. (2009). Palliative care for Latino
patients and their families: Whenever we prayed, she wept. Journal of the
American Medical Association, 301(10), 1047-1057.
158
Smith, C., Cowan, C., Heffler, S., & Catlin, A. (2006). National health spending in 2004:
Recent slowdown led by prescription drug spending. Health Affairs, 25(1), 186-
196.
Somova, M. J., Somov, P. G., Lawrence, J. C., & Frantz, T. T. (2000). Factors associated
with length of stay in a mid-sized, urban hospice. American Journal of Hospice &
Palliative Care, 17(2), 99-105.
StataCorp. (2013). Stata Statistical Software/MP: Release 13. College Station, TX:
StataCorp LP.
Steinhauser, K. E., Clipp, E. C., McNeilly, M., Christakis, N. A., McIntyre, L. M., &
Tulsky, J. A. (2000). In search of a good death: observations of patients, families,
and providers. Annals of Internal Medicine, 132(10), 825-832.
Stevenson, D. G., & Bramson, J. S. (2009). Hospice care in the nursing home setting: A
review of the literature. Journal of Pain & Symptom Management, 38(3), 440-
451.
Stillman, M. J., & Syrjala, K. L. (1999). Differences in physician access patterns to
hospice care. Journal of Pain & Symptom Management, 17(3), 157-163.
Talamantes, M. A., Lawler, W. R., & Espino, D. V. (1995). Hispanic American elders:
Caregiving norms surrounding dying and the use of hospice services. The Hospice
Journal, 10(2), 35-49.
Tang, S. T. (2003). When death is imminent: Where terminally ill patients with cancer
prefer to die and why. Cancer Nursing, 26(3), 245-251.
Taylor, D. H., Jr., Ostermann, J., Van Houtven, C. H., Tulsky, J. A., & Steinhauser, K.
(2007). What length of hospice use maximizes reduction in medical expenditures
159
near death in the US Medicare program? Social Science & Medicine, 65(7), 1466-
1478.
Teno, J. M., Clarridge, B. R., Casey, V., Welch, L. C., Wetle, T., Shield, R., & Mor, V.
(2004). Family perspectives on end-of-life care at the last place of care. Journal of
the American Medical Association, 291(1), 88-93.
Teno, J. M., Fisher, E. S., Hamel, M. B., Coppola, K., & Dawson, N. V. (2002). Medical
care inconsistent with patients' treatment goals: Association with 1-year Medicare
resource use and survival. Journal of the American Geriatrics Society, 50(3), 496-
500.
Teno, J. M., Gruneir, A., Schwartz, Z., Nanda, A., & Wetle, T. (2007). Association
between advance directives and quality of end-of-life care: A national study.
Journal of the American Geriatrics Society, 55(2), 189-194.
Teno, J. M., Shu, J. E., Casarett, D., Spence, C., Rhodes, R., & Connor, S. (2007).
Timing of referral to hospice and quality of care: Length of stay and bereaved
family members' perceptions of the timing of hospice referral. Journal of Pain &
Symptom Management, 34(2), 120-125.
Teno, J. M., Stevens, M., Spernak, S., & Lynn, J. (1998). Role of written advance
directives in decision making. Journal of General Internal Medicine, 13(7), 439-
446.
Thorevska, N., Tilluckdharry, L., Tickoo, S., Havasi, A., Amoateng-Adjepong, Y., &
Manthous, C. A. (2005). Patients' understanding of advance directives and
cardiopulmonary resuscitation. Journal of Critical Care, 20(1), 26-34.
160
Thornton, J. D., Pham, K., Engelberg, R. A., Jackson, J. C., & Curtis, J. R. (2009).
Families with limited English proficiency receive less information and support in
interpreted intensive care unit family conferences. Critical Care Medicine, 37(1),
89-95.
Tolle, S. W., Tilden, V. P., Hickman, S. E., & Rosenfeld, A. G. (2000). Family reports of
pain in dying hospitalized patients: A structured telephone survey. Western
Journal of Medicine, 172(6), 374-377.
Townsend, J., Frank, A., Fermont, D., Dyer, S., Karran, O., Walgrove, A., & Piper, M.
(1990). Terminal cancer care and patients' preference for place of death: A
prospective study. British Medical Journal, 301(6749), 415.
U.S. General Accounting Office. (2000). Medicare: More Beneficiaries Use Hospice;
Many Factors Contribute to Shorter Periods of Use (Publication no. GAO/HEHS-
00-182). Washington, DC: U.S. General Accounting Office.
United States Department of Health and Human Services. Centers for Disease Control
Prevention. National Center for Health Statistics. (2010). National Home and
Hospice Care Survey, 2007: Inter-university Consortium for Political and Social
Research (ICPSR) [distributor]. Retrieved from
http://doi.org/10.3886/ICPSR28961.v1
Unroe, K. T., Greiner, M. A., Johnson, K. S., Curtis, L. H., & Setoguchi, S. (2012).
Racial differences in hospice use and patterns of care after enrollment in hospice
among Medicare beneficiaries with heart failure. American Heart Journal, 163(6),
987-993. e983.
161
Volandes, A., Ariza, M., Abbo, E., & Paasche-Orlow, M. (2008). Overcoming
educational barriers for advance care planning in Latinos with video images.
Journal of Palliative Medicine, 11(5), 700-706.
Volandes, A., Barry, M., Chang, Y., & Paasche-Orlow, M. (2010). Improving decision
making at the end of life with video images. Medical Decision Making, 30(1), 29.
Volandes, A., Lehmann, L., Cook, E., Shaykevich, S., Abbo, E., & Gillick, M. (2007).
Using video images of dementia in advance care planning. Archives of internal
medicine, 167(8), 828.
Volandes, A., Paasche-Orlow, M., Gillick, M., Cook, E., Shaykevich, S., Abbo, E., &
Lehmann, L. (2008). Health literacy not race predicts end-of-life care preferences.
Journal of Palliative Medicine, 11(5), 754-762.
Volicer, L., Cantor, M. D., Derse, A. R., Edwards, D. M., Prudhomme, A. M., Gregory,
D. C. R., . . . Fox, E. (2002). Advance Care Planning by Proxy for Residents of
Long ‐Term Care Facilities Who Lack Decision ‐Making Capacity. Journal of
the American Geriatrics Society, 50(4), 761-767.
Wachterman, M. W., Marcantonio, E. R., Davis, R. B., & McCarthy, E. P. (2011).
Association of hospice agency profit status with patient diagnosis, location of
care, and length of stay. Journal of the American Medical Association, 305(5),
472-479.
Wallace, M. P., Weiner, J. S., Pekmezaris, R., Almendral, A., Cosiquien, R., Auerbach,
C., & Wolf-Klein, G. (2007). Physician cultural sensitivity in African American
advance care planning: A pilot study. Journal of Palliative Medicine, 10(3), 721-
727.
162
Wallace, S. P., Levy-Storms, L., Kington, R. S., & Andersen, R. M. (1998). The
persistence of race and ethnicity in the use of long-term care. The Journals of
Gerontology Series B: Psychological Sciences and Social Sciences, 53(2), S104-
S112.
Wallston, K. A., Burger, C., Smith, R. A., & Baugher, R. J. (1988). Comparing the
quality of death for hospice and non-hospice cancer patients. Medical Care, 26(2),
177-182.
Washington, K. T., Bickel-Swenson, D., & Stephens, N. (2008). Barriers to hospice use
among African Americans: A systematic review. Health & Social Work, 33(4),
267-274.
Weeks, J. C., Cook, E. F., O'Day, S. J., Peterson, L. M., Wenger, N., Reding, D., . . .
Dawson, N. V. (1998). Relationship between cancer patients' predictions of
prognosis and their treatment preferences. Journal of the American Medical
Association, 279(21), 1709-1714.
Weggel, J. M. (1999). Barriers to the physician decision to offer hospice as an option for
terminal care. Western Journal of Medicine, 98(3), 49-53.
Weitzen, S., Teno, J. M., Fennell, M., & Mor, V. (2003). Factors associated with site of
death: A national study of where people die. Medical Care, 41(2), 323-335.
Welch, B. L. (1947). The generalization ofstudent's' problem when several different
population variances are involved. Biometrika, 28-35.
Wennberg, J. E., Fisher, E. S., Goodman, D. C., & Skinner, J. S. (2008). Tracking the
Care of Patients with Severe Chronic Illness-The Dartmouth Atlas of Health Care
2008.
163
Wennberg, J. E., Fisher, E. S., Stukel, T. A., Skinner, J. S., Sharp, S. M., & Bronner, K.
K. (2004). Use of hospitals, physician visits, and hospice care during last six
months of life among cohorts loyal to highly respected hospitals in the United
States. British Medical Journal, 328(7440), 607.
Wenrich, M. D., Curtis, J. R., Ambrozy, D. A., Carline, J. D., Shannon, S. E., & Ramsey,
P. G. (2003). Dying patients' need for emotional support and personalized care
from physicians: perspectives of patients with terminal illness, families, and
health care providers. Journal of Pain & Symptom Management, 25(3), 236-246.
White, D. B., Braddock, C. H., 3rd, Bereknyei, S., & Curtis, J. R. (2007). Toward shared
decision making at the end of life in intensive care units: Opportunities for
improvement. Archives of Internal Medicine, 167(5), 461-467.
Winzelberg, G. S., Hanson, L. C., & Tulsky, J. A. (2005). Beyond autonomy:
diversifying end-of-life decision-making approaches to serve patients and
families. Journal of the American Geriatrics Society, 53(6), 1046-1050.
Wright, A. A., Keating, N. L., Balboni, T. A., Matulonis, U. A., Block, S. D., &
Prigerson, H. G. (2010). Place of death: correlations with quality of life of
patients with cancer and predictors of bereaved caregivers' mental health. Journal
of Clinical Oncology, 28(29), 4457-4464.
Wright, A. A., Zhang, B., Ray, A., Mack, J. W., Trice, E., Balboni, T., . . . Prigerson, H.
G. (2008). Associations between end-of-life discussions, patient mental health,
medical care near death, and caregiver bereavement adjustment. Journal of the
American Medical Association, 300(14), 1665-1673.
164
Yancu, C. N., Farmer, D. F., & Leahman, D. (2010). Barriers to hospice use and
palliative care services use by African American adults. American Journal of
Hospice & Palliative Care, 27(4), 248-253.
Young, G. J., Meterko, M., & Desai, K. R. (2000). Patient satisfaction with hospital care:
effects of demographic and institutional characteristics. Medical Care, 38(3), 325-
334.
Yu, W. (2008). End of life care: medical treatments and costs by age, race, and region.
Health Services Research and Development Study IIR 02-189. Available at:
http://www.hsrd.research.va.gov/research/abstracts/IIR_02-‐189.htm.
Accessed on February 11, 2012.
Zhang, B., Wright, A. A., Huskamp, H. A., Nilsson, M. E., Maciejewski, M. L., Earle, C.
C., . . . Prigerson, H. G. (2009). Health care costs in the last week of life:
associations with end-of-life conversations. Archives of Internal Medicine,
169(5), 480-488.
Zheng, N. T., Mukamel, D. B., Caprio, T., Cai, S., & Temkin-Greener, H. (2011). Racial
disparities in in-hospital death and hospice use among nursing home residents at
the end of life. Medical Care, 49(11), 992-998.

165
Appendices
Appendix A
Appendix A. Weighted and Unweighted Example Description
White Black Hispanic Sig.
Gender (Unweighted) .613
Female 1,425 (42.65) 95 (43.38) 48 (47.52)
Male 1,916 (57.35) 124 (56.62) 53 (52.48)
Gender (Weighted) .024
Female 299,860 (42.16) 29,752 (57.92) 8,856 (33.78)
Male 411,424 (57.84) 21,616 (42.08) 17,364 (66.22)
Footnotes
Sample size: 3,661
Unweighted sample size: 788,872
Percentages are presented by column
Characteristics, No. (%)

166
Appendix B
Appendix B. Unweighted Full Sample Description (n = 3,661)
Total White Black Hispanic
(n = 3,661) (n = 3,341) (n = 219) (n = 101) Sig.
Age, Mean ± SD 82.30 ± 8.37 82.37 ± 8.39 81.52 ± 8.56 81.42 ± 7.13 0.197
Female, No. (%) 1,568 (42.83) 1,425 (42.65) 95 (43.38) 48 (47.52) 0.613
Marital Status, No. (%) 0.019
Married/Living with Partner 1,476 (40.32) 1,370 (41.01) 63 (28.77) 43 (42.57)
Widowed 1,706 (46.60) 1,543 (46.18) 118 (53.88) 45 (44.55)
Divorced/Separated 200 (5.46) 183 (5.48) 12 (5.48) 5 (4.95)
Never Married 151 (4.12) 130 (3.89) 17 (7.76) 4 (3.96)
Missing 128 (3.50) 115 (3.44) 9 (4.11) 4 (3.96)
Primary Diagnosis, No. (%) 0.012
Cancer 1,391 (38.00) 1,286 (38.49) 69 (31.51) 36 (35.64)
CHF/Heart Disease 527 (14.39) 479 (14.34) 33 (15.07) 15 (14.85)
Lung Disease 580 (15.84) 541 (16.19) 26 (11.87) 13 (12.87)
Neurological Diseases 645 (17.62) 563 (16.85) 62 (28.31) 20 (19.80)
Other 512 (13.99) 466 (13.95) 29 (13.24) 17 (16.83)
Missing 6 (0.16) 6 (0.18) 0 (0.00) 0 (0.00)
Comorbidity Count, Mean ± SD 3.55 ± 2.52 3.56 ± 2.52 3.48 ± 2.40 3.59 ± 2.49 0.883
Medicaid Enrollee, No. (%) 740 (20.21) 599 (17.93) 104 (47.49) 37 (36.63) <.001
Missing 30 (0.82) 30 (0.90) 0 (0.00) 0 (0.00)
Footnotes
SD: Standard Deviation
CHF: Congestive Heart Failure
Percentages are presented by column
Characteristics

167
Appendix C
Appendix C. Unweighted Decedent Sample Description (n = 3,006)
Total White Black Hispanic
(n = 3,006) (n = 2,787) (n = 139) (n = 80) Sig.
Age, Mean ± SD 82.07 ± 8.38 82.18 ± 8.41 80.45 ± 7.96 81.34 ± 7.39 0.043
Female, No. (%) 1,325 (44.08) 1,215 (43.60) 70 (50.36) 40 (50.00) 0.163
Marital Status, No. (%) 0.855
Married/Living with Partner 1,260 (41.92) 1,175 (42.16) 51 (36.69) 34 (42.50)
Widowed 1,357 (45.14) 1,257 (45.10) 65 (46.76) 35 (43.75)
Divorced/Separated 166 (5.52) 154 (5.53) 8 (5.76) 4 (5.00)
Never Married 120 (3.99) 109 (3.91) 8 (5.76) 3 (3.75)
Missing 103 (3.43) 92 (3.30) 7 (5.04) 4 (5.00)
Primary Diagnosis, No. (%) 0.370
Cancer 1,237 (41.15) 1,148 (41.19) 58 (41.73) 31 (38.75)
CHF/Heart Disease 398 (13.24) 379 (13.60) 10 (7.19) 9 (11.25)
Lung Disease 454 (15.10) 424 (15.21) 18 (12.95) 12 (15.00)
Neurological Diseases 492 (16.37) 443 (15.90) 32 (23.02) 17 (21.25)
Other 419 (13.94) 387 (13.89) 21 (15.11) 11 (13.75)
Missing 6 (0.20) 6 (0.22) 0 (0.00) 0 (0.00)
Comorbidity Count, Mean ± SD 3.55 ± 2.53 3.55 ± 2.53 3.58 ± 2.48 3.56 ± 2.47 0.991
Medicaid Enrollee, No. (%) 564 (18.76) 478 (17.15) 59 (42.45) 27 (33.75) <.001
Missing 25 (0.83) 25 (0.90) 0 (0.00) 0 (0.00)
Footnotes
SD: Standard Deviation
CHF: Congestive Heart Failure
Percentages are presented by column
Characteristics

168
Appendix D
Yes Missing
Age
65-75 144,556 (86.16) 7,980 (4.76)
76-82 165,684 (90.62) 688 (0.38)
83-87 171,992 (91.83) 1,456 (0.78)
88+ 235,804 (93.96) 228 (0.09)
Gender
Female 307,088 (90.73) 3,528 (1.04)
Male 410,948 (91.24) 6,824 (1.52)
Race/Ethnicity
White 652,144 (91.69) 9,976 (1.40)
Black 41,044 (79.90) 12 (0.02)
Hispanic 24,848 (94.77) 364 (1.39)
Marital Status
Married/Living Together 283,880 (90.46) 3,884 (1.24)
Widowed 340,156 (93.83) 1,820 (0.50)
Divorced/Separated 37,008 (87.81) 0 (0.00)
Never Married 27,152 (91.97) 12 (0.04)
Missing 29,840 (73.01) 4,636 (11.34)
Primary Diagnosis
Cancer 262,060 (91.21) 2,120 (0.74)
CHF/Heart Disease 88,016 (87.69) 1,832 (1.83)
Lung Disease 118,380 (92.46) 1,732 (1.35)
Neurological Diseases 155,616 (92.54) 1,848 (1.10)
Other 93,440 (89.73) 2,740 (2.63)
Missing 524 (60.65) 80 (9.26)
Comorbidity Count
0-1 190,464 (88.00) 8,396 (3.88)
2-3 224,692 (90.46) 628 (0.25)
4+ 302,880 (93.46) 1,328 (0.41)
Footnotes
CHF: Congestive Heart Failure
Percentages are presented by row among the full sample
Characteristics, No. (%)
Documented Advance Directive
Appendix D. Advance Directive Completion Sample Descriptives
(n = 788,872)

169
Appendix E
Yes Missing
Age
65-75 130,668 (77.88) 23,224 (13.84)
76-82 152,864 (83.61) 17,144 (9.38)
83-87 160,120 (85.49) 15,312 (8.18)
88+ 212,288 (84.59) 15,156 (6.04)
Gender
Female 282,876 (83.58) 31,380 (9.27)
Male 373,064 (82.83) 39,456 (8.76)
Race/Ethnicity
White 594,672 (83.61) 59,140 (8.32)
Black 37,296 (72.61) 10,324 (20.10)
Hispanic 23,972 (91.43) 1,372 (5.23)
Marital Status
Married/Living Together 260,656 (83.06) 29,940 (9.54)
Widowed 312,108 (86.10) 22,356 (6.17)
Divorced/Separated 32,652 (77.48) 5,136 (12.19)
Never Married 23,208 (78.61) 2,372 (8.03)
Missing 27,316 (66.83) 11,032 (26.99)
Primary Diagnosis
Cancer 234,296 (81.55) 25,248 (8.79)
CHF/Heart Disease 78,772 (78.48) 12,360 (12.31)
Lung Disease 104,912 (81.94) 9,648 (7.54)
Neurological Diseases 147,380 (87.64) 12,540 (7.46)
Other 90,076 (86.50) 10,700 (10.27)
Missing 504 (58.33) 340 (39.35)
Comorbidity Count
0-1 170,612 (78.83) 25,968 (12.00)
2-3 208,172 (83.81) 23,688 (9.54)
4+ 277,156 (85.53) 21,180 (6.54)
Footnotes
CHF: Congestive Heart Failure
Percentages are presented by row among the full sample
Characteristics, No. (%)
Documented DNR Order
Appendix E. Do Not Resuscitate Order Election Sample
Descriptives (n = 788,872)

170
Appendix F
Yes Missing
Age
65-75 67,336 (40.13) 23,224 (13.84)
76-82 83,308 (45.57) 17,144 (9.38)
83-87 89,240 (47.64) 15,312 (8.18)
88+ 140,924 (56.15) 15,156 (6.04)
Gender
Female 152,464 (45.05) 31,380 (9.27)
Male 228,344 (50.70) 39,456 (8.76)
Race/Ethnicity
White 357,400 (50.25) 59,140 (8.32)
Black 15,800 (30.76) 10,324 (20.10)
Hispanic 7,608 (29.02) 1,372 (5.23)
Marital Status
Married/Living Together 128,576 (40.97) 29,940 (9.54)
Widowed 198,888 (54.86) 22,356 (6.17)
Divorced/Separated 20,632 (48.96) 5,136 (12.19)
Never Married 15,924 (53.94) 2,372 (8.03)
Missing 16,788 (41.07) 11,032 (26.99)
Primary Diagnosis
Cancer 12,6888 (44.16) 25,248 (8.79)
CHF/Heart Disease 49,680 (49.49) 12,360 (12.31)
Lung Disease 64,316 (50.24) 9,648 (7.54)
Neurological Diseases 86,272 (51.30) 12,540 (7.46)
Other 53,652 (51.52) 10,700 (10.27)
Missing 0 (0.00) 340 (39.35)
Comorbidity Count
0-1 84,852 (39.20) 25,968 (12.00)
2-3 125,416 (50.49) 23,688 (9.54)
4+ 170,540 (52.63) 21,180 (6.54)
Footnotes
CHF: Congestive Heart Failure
Percentages are presented by row among the full sample
Characteristics, No. (%)
Documented Healthcare Proxy
Appendix F. Healthcare Proxy Designation Sample Descriptives
(n = 788,872)

171
Appendix G
Yes Missing
Age
65-75 11,056 (6.59) 8,728 (5.20)
76-82 12,220 (6.68) 2,376 (1.30)
83-87 7,260 (3.88) 784 (0.42)
88+ 16,656 (6.64) 3,412 (1.36)
Gender
Female 19,536 (5.78) 3,512 (1.04)
Male 27,656 (6.14) 11,788 (2.62)
Race/Ethnicity
White 40,568 (5.70) 14,124 (1.99)
Black 4,864 (9.47) 1,176 (2.29)
Hispanic 1,760 (6.71) 0 (0.00)
Marital Status
Married/Living Together 17,292 (5.51) 4,388 (1.40)
Widowed 23,636 (6.52) 4,240 (1.17)
Divorced/Separated 3,320 (7.88) 8 (0.02)
Never Married 836 (2.83) 52 (0.18)
Missing 2,108 (5.16) 6,612 (16.18)
Primary Diagnosis
Cancer 15,996 (5.57) 6,172 (2.15)
CHF/Heart Disease 9,864 (9.83) 1,736 (1.73)
Lung Disease 8,876 (6.93) 1,284 (1.00)
Neurological Diseases 6,292 (3.74) 2,908 (1.73)
Other 6,164 (5.92) 2,460 (2.36)
Missing 0 (0.00) 740 (85.65)
Comorbidity Count
0-1 9,316 (4.30) 10,092 (4.66)
2-3 19,076 (7.68) 3,256 (1.31)
4+ 18,800 (5.80) 1,952 (0.60)
Enrolled in Medicaid
Yes 12,416 (9.92) 572 (0.46)
No 34,776 (5.31) 13,224 (2.02)
Missing 0 (0.00) 1,504 (17.20)
Advance Care Planning
No AD, No DNR order 8,116 (13.42) 536 (0.89)
Yes AD, No DNR order 4,440 (7.15) 900 (1.45)
Yes AD, Yes DNR order 34,624 (5.28) 7,072 (1.08)
Missing 12 (0.12) 6,792 (65.61)
Footnotes
CHF: Congestive Heart Failure
AD: Advance Directive
DNR: Do Not Resuscitate
Percentages are presented by row among the full sample
Characteristics, No. (%)
Emergent Care Utilization
Appendix G. Emergent Care Utilization Sample Descriptives
(n = 788,872)

172
Appendix H
Valid Cases Missing
Age
65-75 689 (96.77) 23 (3.23)
76-82 788 (97.16) 23 (2.84)
83-87 598 (96.14) 24 (3.86)
88+ 826 (95.93) 35 (4.07)
Gender
Female 1,277 (96.38) 48 (3.62)
Male 1,624 (96.61) 57 (3.39)
Race/Ethnicity
White 2,689 (96.48) 98 (3.52)
Black 136 (97,84) 3 (2.16)
Hispanic 76 (95.00) 4 (5.00)
Marital Status
Married/Living Together 1,213 (96.27) 47 (3.73)
Widowed 1,311 (96.61) 46 (3.39)
Divorced/Separated 160 (96.39) 6 (3.61)
Never Married 117 (97.50) 3 (2.50)
Missing 100 (97.09) 3 (2.91)
Primary Diagnosis
Cancer 1,209 (97.74) 28 (2.26)
CHF/Heart Disease 380 (95.48) 18 (4.52)
Lung Disease 433 (95.37) 21 (4.63)
Neurological Diseases 474 (96.34) 18 (3.66)
Other 399 (95.23) 20 (4.77)
Missing 6 (100.00) 0 (0.00)
Comorbidity Count
0-1 734 (94.71) 41 (5.29)
2-3 904 (97.10) 27 (2.90)
4+ 1,263 (97.15) 37 2.85)
Enrolled in Medicaid
Yes 2,328 (96.32) 89 (3.68)
No 548 (97.16) 16 (2.84)
Missing 25 (100.00) 0 (0.00)
Advance Care Planning
No AD, No DNR order 216 (95.58) 10 (4.42)
Yes AD, No DNR order 234 (96.69) 8 (3.31)
Yes AD, Yes DNR order 2,436 (96.59) 86 (3.41)
Missing 15 (93.75) 1 (6.25)
Footnotes
CHF: Congestive Heart Failure
AD: Advance Directive
DNR: Do Not Resuscitate
Percentages are presented by row among the decedent sample
Characteristics, No. (%)
Length of Stay: Full Episode
Appendix H. Decedent Hospice Length of Stay (Full Care Episode)
Sample Descriptives (n = 3,006)

173
Appendix I
n Median Sig.
Age 0.628
65-75 144,880 15.00
76-82 150,860 15.00
83-87 147,548 15.00
88+ 195,124 17.00
Gender 0.190
Female 361,604 15.00
Male 276,808 16.00
Race/Ethnicity 0.428
White 583,124 15.00
Black 36,304 19.00
Hispanic 18,984 17.00
Marital Status .316
Married/Living Together 262,216 15.00
Widowed 286,640 16.00
Divorced/Separated 31,956 14.50
Never Married 24,920 12.00
Missing 32,680 18.50
Primary Diagnosis <.001
Cancer 247,300 18.00
CHF/Heart Disease 80,536 21.00
Lung Disease 96,788 13.00
Neurological Diseases 133,176 14.00
Other 79,748 9.00
Missing 864 16.50
Comorbidity Count 0.695
0-1 174,004 15.00
2-3 194,320 16.00
4+ 270,088 15.00
Enrolled in Medicaid 0.849
Yes 95,952 15.00
No 534,904 16.00
Missing 7,556 10.00
Advance Care Planning 0.849
No AD, No DNR order 39,456 14.50
Yes AD, No DNR order 47,744 12.00
Yes AD, Yes DNR order 542,360 16.00
Missing 8,852 52.00
Footnotes
CHF: Congestive Heart Failure
AD: Advance Directive
DNR: Do Not Resuscitate
Characteristics
Decedent Length of Stay:
Full Episode
Appendix I. Decedent Hospice Length of Stay (Full Care
Episode) Median Comparisons (n = 638,412)

174
Appendix J
Yes Missing
Age
65-75 58,400 (38.78) 5,704 (3.79)
76-82 56,316 (36.13) 5,004 (3.21)
83-87 66,264 (42.89) 6,940 (4.49)
88+ 66,368 (32.08) 11,760 (5.68)
Gender
Female 110,556 (37.99) 14,204 (4.88)
Male 136,792 (36.30) 15,204 (4.04)
Race/Ethnicity
White 229,872 (37.63) 27,696 (4.53)
Black 11,220 (30.26) 772 (2.08)
Hispanic 6,256 (31.40) 940 (4.72)
Marital Status
Married/Living Together 105,308 (38.49) 11,372 (4.16)
Widowed 105,612 (34.92) 15,836 (5.24)
Divorced/Separated 15,668 (48.06) 648 (1.99)
Never Married 11,128 (43.55) 632 (2.47)
Missing 9,632 (28.67) 920 (2.74)
Primary Diagnosis
Cancer 90,448 (35.39) 8,280 (3.24)
CHF/Heart Disease 34,124 (40.63) 3,460 (4.12)
Lung Disease 36,712 (34.84) 8,576 (8.14)
Neurological Diseases 51,284 (36.99) 5,456 (3.94)
Other 34,732 (41.65) 3,636 (4.36)
Missing 48 (5.56) 0 (0.00)
Comorbidity Count
0-1 68,128 (36.73) 11,472 (6.19)
2-3 79,668 (39.37) 8,044 (3.98)
4+ 99,552 (35.56) 9,892 (3.53)
Enrolled in Medicaid
Yes 30,044 (30.45) 2,728 (2.76)
No 214,484 (38.19) 26,680 (4.75)
Missing 2,820 (37.32) 0 (0.00)
Advance Care Planning
No AD, No DNR order 17,168 (41.21) 2,208 (5.30)
Yes AD, No DNR order 18,352 (37.55) 1,128 (2.31)
Yes AD, Yes DNR order 210,340 (37.01) 26,044 (4.58)
Missing 1,488 (16.76) 28 (0.32)
Footnotes
CHF: Congestive Heart Failure
AD: Advance Directive
DNR: Do Not Resuscitate
Percentages are presented by row among the decedent sample
Characteristics, No. (%)
Length of Stay: 0-7 Days
Appendix J. Decedent Hospice Length of Stay (0-7 Days) Sample
Descriptives (n = 667,820)

175
Appendix K
Yes Missing
Age
65-75 99,052 (65.78) 5,704 (3.79)
76-82 100,312 (64.36) 5,004 (3.21)
83-87 106,264 (68.78) 6,940 (4.49)
88+ 116,704 (56.41) 11,760 (5.68)
Gender
Female 190,856 (65.68) 14,204 (4.88)
Male 231,476 (61.43) 15,204 (4.04)
Race/Ethnicity
White 382,168 (62.57) 27,696 (4.53)
Black 27,764 (74.88) 772 (2.08)
Hispanic 12,400 (62.24) 940 (4.72)
Marital Status
Married/Living Together 180,868 (66.11) 11,372 (4.16)
Widowed 182,292 (60.27) 15,836 (5.24)
Divorced/Separated 25,156 (77.16) 648 (1.99)
Never Married 18,184 (71.16) 632 (2.47)
Missing 15,832 (47.12) 920 (2.74)
Primary Diagnosis
Cancer 171,096 (66.94) 8,280 (3.24)
CHF/Heart Disease 50,424 (60.03) 3,460 (4.12)
Lung Disease 58,488 (55.51) 8,576 (8.14)
Neurological Diseases 81,428 (58.74) 5,456 (3.94)
Other 60,156 (72.14) 3,636 (4.36)
Missing 740 (85.65) 0 (0.00)
Comorbidity Count
0-1 120,068 (64.74) 11,472 (6.19)
2-3 135,508 (66.96) 8,044 (3.98)
4+ 166,756 (59.56) 9,892 (3.53)
Enrolled in Medicaid
Yes 57,108 (57.87) 2,728 (2.76)
No 360,576 (64.21) 26,680 (4.75)
Missing 4,648 (61.51) 0 (0.00)
Advance Care Planning
No AD, No DNR order 27,320 (65.57) 2,208 (5.30)
Yes AD, No DNR order 33,032 (67.59) 1,128 (2.31)
Yes AD, Yes DNR order 359,764 (63.29) 26,044 (4.58)
Missing 2,216 (24.95) 28 (0.32)
Footnotes
CHF: Congestive Heart Failure
AD: Advance Directive
DNR: Do Not Resuscitate
Percentages are presented by row among the decedent sample
Characteristics, No. (%)
Length of Stay: 0-30 Days
Appendix K. Decedent Hospice Length of Stay (0-30 Days) Sample
Descriptives (n = 667,820)

176
Appendix L
Yes Missing
Age
65-75 80,480 (53.45) 34,456 (22.88)
76-82 80,996 (51.97) 23,808 (15.27)
83-87 77,624 (50.25) 20,928 (13.55)
88+ 87,396 (42.24) 28,036 (13.55)
Gender
Female 159,756 (54.90) 48,720 (16.75)
Male 166,740 (44.25) 58,508 (15.53)
Race/Ethnicity
White 298,268 (48.83) 97,288 (15.93)
Black 18,980 (51.19) 6,500 (17.53)
Hispanic 9,248 (46.42) 3,440 (17.27)
Marital Status
Married/Living Together 156,780 (57.31) 47,812 (17.48)
Widowed 133,836 (44.25) 38,512 (12.73)
Divorced/Separated 16,928 (51.92) 7,168 (21.99)
Never Married 7,460 (29.20) 6,292 (24.62)
Missing 11,492 (34.20) 7,444 (22.15)
Primary Diagnosis
Cancer 158,424 (61.99) 46,076 (18.03)
CHF/Heart Disease 39,972 (47.59) 11,836 (14.09)
Lung Disease 47,916 (45.48) 18,560 (17.62)
Neurological Diseases 50,624 (36.52) 15,020 (10.83)
Other 29,176 (34.99) 15,276 (18.32)
Missing 384 (44.44) 460 (53.24)
Comorbidity Count
0-1 87,304 (47.07) 31,796 (17.14)
2-3 106,192 (52.48) 32,056 (15.84)
4+ 133,000 (47.50) 43,376 (15.49)
Enrolled in Medicaid
Yes 23,852 (53.55) 91,532 (16.30)
No 300,712 ()24.17 10,820 (10.96)
Missing 1,932 (25.57) 4,876 (64.53)
Advance Care Planning
No AD, No DNR order 20,392 (48.94) 5,312 (12.75)
Yes AD, No DNR order 32,320 (66.13) 5,252 (10.75)
Yes AD, Yes DNR order 272,848 (48.00) 88,748 (15.61)
Missing 936 (10.54) 7,916 (89.14)
Footnotes
CHF: Congestive Heart Failure
AD: Advance Directive
DNR: Do Not Resuscitate
Percentages are presented by row among the decedent sample
Characteristics, No. (%)
Site of Death: Home
Appendix L. Site of Death (Home-Like Setting) Sample
Descriptives (n = 667,820)

177
Appendix M
Yes Missing
Age
65-75 15,944 (10.59) 34,456 (22.88)
76-82 15,820 (10.15) 23,808 (15.27)
83-87 12,052 (7.80) 20,928 (13.55)
88+ 13,420 (6.49) 28,036 (13.55)
Gender
Female 23,364 (8.03) 48,720 (16.75)
Male 33,872 (8.99) 58,508 (15.53)
Race/Ethnicity
White 52,744 (8.63) 97,288 (15.93)
Black 2,280 (6.15) 6,500 (17.53)
Hispanic 2,212 (11.10) 3,440 (17.27)
Marital Status
Married/Living Together 25,316 (9.25) 47,812 (17.48)
Widowed 23,752 (7.85) 38,512 (12.73)
Divorced/Separated 3,804 (11.67) 7,168 (21.99)
Never Married 972 (3.80) 6,292 (24.62)
Missing 3,392 (10.10) 7,444 (22.15)
Primary Diagnosis
Cancer 18,148 (7.10) 46,076 (18.03)
CHF/Heart Disease 7,148 (8.51) 11,836 (14.09)
Lung Disease 12,340 (11.71) 18,560 (17.62)
Neurological Diseases 9,508 (6.86) 15,020 (10.83)
Other 10,072 (12.08) 15,276 (18.32)
Missing 20 (2.32) 460 (53.24)
Comorbidity Count
0-1 23,692 (12.77) 31,796 (17.14)
2-3 15,056 (7.44) 32,056 (15.84)
4+ 18,488 (6.60) 43,376 (15.49)
Enrolled in Medicaid
Yes 5,448 (9.21) 91,532 (16.30)
No 51,720 (5.52) 10,820 (10.96)
Missing 68 (0.90) 4,876 (64.53)
Advance Care Planning
No AD, No DNR order 5,860 (14.06) 5,312 (12.75)
Yes AD, No DNR order 3,304 (6.67) 5,252 (10.75)
Yes AD, Yes DNR order 48,072 (8.46) 88,748 (15.61)
Missing 0 (0.00) 7,916 (89.14)
Footnotes
CHF: Congestive Heart Failure
AD: Advance Directive
DNR: Do Not Resuscitate
Percentages are presented by row among the decedent sample
Characteristics, No. (%)
Site of Death: Hospital
Appendix M. Site of Death (Hospital) Sample Descriptives
(n = 667,820)

178
Appendix N
Appendix N. Original Site of Death: Full Responses (n = 667,820)
Total White Black Hispanic
Inpatient Hospice Agency
β
100,012 (14.98) 90,072 (14.75) 6,500 (17.53) 3,440 (17.27)
Private Home/Apartment 289,900 (43.41) 262,260 (42.94) 18,392 (49.61) 9,248 (46.42)
Residental Care Place 36,596 (5.48) 36,008 (5.90) 588 (1.59) 0 (0.00)
Nursing Home/SNF 176,572 (26.44) 162,232 (26.56) 9,316 (25.13) 5,024 (25.22)
Hospital 57,236 (8.57) 52,744 (8.63) 2,280 (6.15) 2,212 (11.10)
Other 492 (0.07) 492 (0.08) 0 (0.00) 0 (0.00)
Missing 7,012 (1.05) 7,012 (1.15) 0 (0.00) 0 (0.00)
Footnotes
β
Inpatient Hospice Agency excluded from study analyses due to operationalization problems
SNF: Skilled Nursing Facility
Percentages are presented by column

Abstract (if available)

Abstract Despite the rapid growth of hospice care in the United States over the past several decades, racial/ethnic minorities continue to utilize higher levels of aggressive life‐prolonging interventions at end‐of‐life, often resulting in poorer care experiences. While previous research has expanded understanding of racial/ethnic end-of-life disparities outside of hospice, an in‐depth analysis of the relationship between patient care preferences and key end‐of‐life outcomes within a racially/ethnically diverse hospice population remains to be conducted. ❧ Using the 2007 wave of the National Home Health and Hospice Care Survey (NHHCS), a retrospective analysis of clinical and service use outcomes was conducted to test for racial/ethnic variation following hospice enrollment. Key outcomes of interest included advance care planning, emergent care utilization, hospice length of stay, and site of death. In total, 3,661 White, Black, and Hispanic Medicare hospice patients were analyzed, representing approximately 788,872 older Americans. Results indicated that advance care planning varied by race/ethnicity, with Blacks less likely to complete an advance directive, Hispanics more likely to elect a do not resuscitate order, and both Blacks and Hispanics less likely to designate a healthcare proxy. Findings also indicated that Blacks were less likely to utilize emergent care following adjustment for advance care planning. While Hispanics were more likely to die in the first week of hospice care, Blacks were more likely to die in the first month of hospice care. Concerning site of death, Blacks and Hispanics were more likely to die in a home‐like setting, and Hispanics were also more likely to die in a hospital. Results also indicated that advance care planning reduced the likelihood of emergent care utilization, death in the first week of hospice care, death in the first month of hospice care, and death in a hospital. Lastly, patients engaging in advance care planning were also more likely to die in a home‐like setting. ❧ Findings support racial/ethnic variation following hospice enrollment, but suggest that differences within hospice contrast with those in the larger healthcare system. Furthermore, results support the protective effect of advance care planning among hospice enrollees. The data presented have substantial clinical and policy implications for improving the care of all patients at end of life. Additional research is needed to better understand and address reported racial/ethnic differences following hospice enrollment.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Exploring end-of-life care experiences and disparities among Hispanic populations in the United States

PDF

Racial disparities in advance care planning and directives completion and end-of-life outcomes in the United States

PDF

Investigating racial and ethnic disparities in patient experiences with care and health services use following colorectal cancer diagnosis among older adults with comorbid chronic conditions

PDF

Palliative care: breaking the knowledge barrier among community-based older adults

PDF

Eliciting perspectives on palliative care: outpatient visits, advance care planning, and the impact of COVID-19

PDF

Getting to end-of-life discussions in advanced cancer care: barriers and attitudes that limit end-of-life communication for disadvantaged Latinos

PDF

Racial/ethnic differences in colorectal cancer patient experiences, health care utilization and their association with mortality: findings from the SEER-CAHPS data

PDF

Racial/ethnic differences in patient experiences with health care in association with the Gleason’s score level at prostate cancer diagnosis: findings from the SEER-CAHPS data

PDF

Three essays in health economics

PDF

Crosscultural differences in dementia diagnosis and care-seeking in Hispanic and non-Hispanic white outpatients

PDF

Exploring mental health care utilization and academic persistence among college students: evaluating racial and stigma-related disparities

PDF

A biodemographic approach to understanding sociodemographic disparities in kidney functioning on three dimensions: individual, population, and cross-national

PDF

Cognitive health and self-rated memory in later life: linkages to race/ethnicity, multimorbidity, and survival

PDF

Disparities in gallbladder, intra-hepatic bile duct, and other biliary cancers among multi-ethnic populations: a California Cancer Registry study

PDF

Prostate cancer disparities among Californian Latinos by country of origin: clinical characteristics, incidence, treatment received and survival

PDF

Clinical outcomes of allogeneic hematopoietic stem cell transplant in acute lymphoblastic leukemia patients: a quality improvement project and systematic review meta-analysis

PDF

Chronic eye disease epidemiology in the multiethnic ophthalmology cohorts of California study

PDF

Grassroots market-making: the production, policing and politics of value in Los Angeles’ transforming cannabis industry

PDF

Ancestral/Ethnic variation in the epidemiology and genetic predisposition of early-onset hematologic cancers

Asset Metadata

Creator Laguna, Jeffrey Ryan (author)

Core Title Racial/ethnic variation in care preferences and care outcomes among United States hospice enrollees

School Leonard Davis School of Gerontology

Degree Doctor of Philosophy

Degree Program Gerontology

Publication Date 08/12/2014

Defense Date 05/13/2014

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag end‐of‐life,health services research,hospice,OAI-PMH Harvest,palliative care,racial/ethnic disparities

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Enguídanos, Susan M. (committee chair), Aranda, Maria P. (committee member), Silverstein, Merril (committee member)

Creator Email drjefflaguna@gmail.com,laguna@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-460539

Unique identifier UC11287167

Identifier etd-LagunaJeff-2814.pdf (filename),usctheses-c3-460539 (legacy record id)

Legacy Identifier etd-LagunaJeff-2814.pdf

Dmrecord 460539

Document Type Dissertation

Format application/pdf (imt)

Rights Laguna, Jeffrey Ryan

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA