USF Libraries
USF Digital Collections

Predictors and outcomes of hospice use among Medicare and Medicaid dual-eligible nursing home residents in Florida: a co...

MISSING IMAGE

Material Information

Title:
Predictors and outcomes of hospice use among Medicare and Medicaid dual-eligible nursing home residents in Florida: a comparison of non-Hispanic Blacks and non-Hispanic Whites
Physical Description:
Book
Language:
English
Creator:
Kwak, Jung
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla
Publication Date:

Subjects

Subjects / Keywords:
End-of-life
Place of death
Race
Ethnicity
Health services utilization
Dissertations, Academic -- Aging Studies -- Doctoral -- USF
Genre:
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Abstract:
ABSTRACT: This research investigated the racial/ethnic differences in hospice utilization and the effect of hospice in reducing the risk of hospital death at the end of life among non-Hispanic Black and non-Hispanic White dual-eligible (Medicare and Medicaid) older adults residing in a nursing home setting. The final study population included 30,765 non-Hispanic Black and non-Hispanic White nursing home residents who died between state fiscal years 2000-2002 in Florida.The behavioral model of health services use successfully predicted group membership in hospice use. In the full model, seven variables - female gender, non-Hispanic White race/ethnicity, being married, urban area of residence, and cancer and dementia/Alzheimer's disease as causes of death --^ predicted increased likelihood of hospice use. This study also found that race/ethnicity moderates the strength of the effect of the illness on the likelihood of using hospice. Among residents who died of cancer, no difference in hospice use was found between the two racial/ethnic groups while hospice utilization rate among non-cancer residents was lower for non-Hispanic Blacks than non-Hispanic Whites.The same predisposing, enabling and need factors tested in predicting hospice use were examined for association with the probability of survival time since hospice admission. The poor model fit and the small number of factors found to significantly affect the probability of survival time from the initial hospice enrollment suggest that the survival time might be influenced by external factors other than covariates examined in this study.This study found hospice to be a powerful predictor of place of death among nursing home residents. After controlling for other factors, hospice nurs ing home residents were 91 percent less likely to die in a hospital. At the same time, non-Hispanic Black residents were still 76 percent more likely to die in a hospital even after adjusting for the effect of hospice use and other variables.
Thesis:
Dissertation (Ph.D.)--University of South Florida, 2006.
Bibliography:
Includes bibliographical references.
System Details:
System requirements: World Wide Web browser and PDF reader.
System Details:
Mode of access: World Wide Web.
Statement of Responsibility:
by Jung Kwak.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 125 pages.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001909521
oclc - 173137304
usfldc doi - E14-SFE0001665
usfldc handle - e14.1665
System ID:
SFS0025983:00001


This item is only available as the following downloads:


Full Text
xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 001909521
003 fts
005 20070919094756.0
006 m||||e|||d||||||||
007 cr mnu|||uuuuu
008 070919s2006 flu sbm 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0001665
040
FHM
c FHM
035
(OCoLC)173137304
049
FHMM
090
HQ1061 (ONLINE)
1 100
Kwak, Jung.
0 245
Predictors and outcomes of hospice use among Medicare and Medicaid dual-eligible nursing home residents in Florida: a comparison of non-Hispanic Blacks and non-Hispanic Whites
h [electronic resource] /
by Jung Kwak.
260
[Tampa, Fla] :
b University of South Florida,
2006.
3 520
ABSTRACT: This research investigated the racial/ethnic differences in hospice utilization and the effect of hospice in reducing the risk of hospital death at the end of life among non-Hispanic Black and non-Hispanic White dual-eligible (Medicare and Medicaid) older adults residing in a nursing home setting. The final study population included 30,765 non-Hispanic Black and non-Hispanic White nursing home residents who died between state fiscal years 2000-2002 in Florida.The behavioral model of health services use successfully predicted group membership in hospice use. In the full model, seven variables female gender, non-Hispanic White race/ethnicity, being married, urban area of residence, and cancer and dementia/Alzheimer's disease as causes of death --^ predicted increased likelihood of hospice use. This study also found that race/ethnicity moderates the strength of the effect of the illness on the likelihood of using hospice. Among residents who died of cancer, no difference in hospice use was found between the two racial/ethnic groups while hospice utilization rate among non-cancer residents was lower for non-Hispanic Blacks than non-Hispanic Whites.The same predisposing, enabling and need factors tested in predicting hospice use were examined for association with the probability of survival time since hospice admission. The poor model fit and the small number of factors found to significantly affect the probability of survival time from the initial hospice enrollment suggest that the survival time might be influenced by external factors other than covariates examined in this study.This study found hospice to be a powerful predictor of place of death among nursing home residents. After controlling for other factors, hospice nurs ing home residents were 91 percent less likely to die in a hospital. At the same time, non-Hispanic Black residents were still 76 percent more likely to die in a hospital even after adjusting for the effect of hospice use and other variables.
502
Dissertation (Ph.D.)--University of South Florida, 2006.
504
Includes bibliographical references.
516
Text (Electronic dissertation) in PDF format.
538
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
500
Title from PDF of title page.
Document formatted into pages; contains 125 pages.
590
Adviser: William E. Haley, Ph.D.
653
End-of-life.
Place of death.
Race.
Ethnicity.
Health services utilization.
690
Dissertations, Academic
z USF
x Aging Studies
Doctoral.
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.1665



PAGE 1

Predictors and Outcomes of Hospice Use among Medicare and Medi caid Dual-Eligible Nursing Home Residents in Florida: A Co mparison of Non-Hispanic Blacks and NonHispanic Whites by Jung Kwak A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Aging Studies College of Arts and Sciences University of South Florida Co-Major Professor: David A. Chiriboga, Ph.D. Co-Major Professor: William E. Haley, Ph.D. Susan C. McMillan, Ph.D. Glenn E. Mitchell, Ph.D. William S. Rowe, Ph.D. Date of Approval: July 11, 2006 Keywords: end-of-life, place of death, race, ethnicity, health services utilization Copyright 2006 Jung Kwak

PAGE 2

Dedication I would like to dedicate this dissertati on research to my parents, Dr. Soo-Young and Mi-Soon K. Kwak. Dad and Mom, I love you more than I coul d ever express in words. How grateful and lucky I am to ha ve parents like you. Your unconditional love, encouragement, support and guidance helped me get through numerous moments of selfdoubts, and sometimes despair. Tha nk you for loving and believing in me.

PAGE 3

i Table of Contents List of Tables ……………………………………………………………………………….iii List of Figures……………………………………………………………………………….iv Abstract ……………………………………………………………………………….. ……v Preface ………………………………………………………………………………...……vii Chapter One: Background …………………………………………………………………..1 Overview……………………………………………………………………… …….1 Introduction ……………………………………………………………………....... .4 Health Services Utilization by Ol der Adults at the End of Life …………………….5 Overview of the Medicare and Medicaid Benefits……………………..... …8 Hospice as a Model of Care …………………………………………………9 Underutilization of Hospice……………………………………………… ..11 Health Services Use, Death and Dying, and Racial/Ethnic Diversity in End-ofLife Decision Making …………………………………………………...................13 Behavioral Model of H ealth Service Utilization …………………………..13 Theories on Death and Dying ……………………………………………...18 The Role of Race and Ethnicity in End-of-Life Decision Making……….. .21 Summary …………………………………………………………………...23 Factors Associated with Hospice Use ……………………………………………..24 Individual Determinants of Hospice Use ………………………………….24 Predisposing factors ……………………………………………….25 Enabling factors…………………………………………………….29 Need/Illness factors ……………………………….……………….32 Psychosocial factors ……………………………….……………….33 Hospice Service Provider Related Factors ………………..……………….34 Medicare Policy and Organizationa l Practice of Nursing Homes in Referring Nursing Home Residents to Hospice ……………………………35 Factors Associated with Hospice Length of Stay ………………….…………….38 Factors Associated with Place of Death …………………………..……………….40 Limitations of Previous Research ………………………………………………….41 Dual-eligible Medicare a nd Medicaid Beneficiaries ………………………41 Racial/Ethnic Differences between Blacks and Whites ……………………43 Hospice Utilization in Nursing Homes …………………….…………….46 The Effects of Hospice Use on Place of Death ……………………………47 Measurement Issues ………………………………………….…………….49 Summary ……………………………………………………..…………….51

PAGE 4

ii Research Questions …………………………………………………..…………….53 The Conceptual Model …………………………………………………….53 Research Questions and Hypotheses ………………………………………55 Research question 1 ……..…………………………………………55 Research question 2 ………………………………………………55 Research question 3 ……..…………………………………………58 Chapter Two: Research Methods …………………………………………………………..60 Sample……………………………………..……………………………………….60 Measures………………………………………………………………………… ...62 Dependent Variables ………………………………………………………62 Predictor Variables …………………………………………………………63 Predisposing factors ……………………………………………….63 Enabling factors ……………………………………………………63 Need factors ………………………………………………………..64 Statistical Approach ………………………………………………………………..65 Descriptive Analyses and Bivariate Correlations …………………………67 Multivariate Analyses ……………………………………………………68 Research question 1 ………………………………………………..68 Research question 2 ………………………………………………..69 Research question 3 ………………………………………………..71 Chapter Three: Findings………..…………………………………………………………..73 Population Characteristics………………………………………………………….73 Major Findings .………..…………………………………………………………..75 Research Question 1 ……………………………………………………..75 Research Question 2………………………………………………………..82 Research Question 3………………………………………………………..84 Chapter Four: Discussion and Conclusions ………………………………………………..89 The Behavioral Model of Health Services Use ……………………………………89 Racial/Ethnic Differences in U tilization of Hospice Services ……………………..93 Limited Contribution of Indi vidual Characteristics to Le ngth of Stay in Hospice ...97 The Role of Race and Ethnicity and Ho spice in Predicting In-Hospital Death ……98 Summary…… ……………………………………………………………………...99 Study Limitations…….. …………………………………………………………..100 Future Directions……………………………………………………………..…...102 List of References……………………………………………………………………. …..106 Appendices ..………………………………………………………………………….. ….122 Appendix A: Flow Chart of the Study Population Selection ……………………..123 Appendix B: Determining Nursing Home Status…………………………………124 About the Author……………………………………………………………………End Page

PAGE 5

iii List of Tables Table 1 Correlations be tween Variables for the popul ation of Medicaid Nursing Home Residents………………………………………..........................…...68 Table 2 Correlations between Variables fo r Medicaid Nursing Home Decedents Who Used Hospice…………………………………………………………69 Table 3 Nursing Home Residents Ag ed 65 and Older by Hospice Use, Race/Ethnicity, and Nursing Home Status ………………………………76 Table 4 Multiple Logistic Regression Models Predicting Hospice Use by Race/Ethnicity …………………………………………………………81 Table 5 Multiple Logistic Regression M odels Predicting Hospice Use For NonHispanic Whites ……………………………………………………… …...82 Table 6 Multiple Logistic Regression Models Predicting Hospice For nonHispanic Blacks ……………………………………………………………83 Table 7 Cox Regression Model to Examine th e Days of Hospice Length of Stay …85 Table 8 Multiple Logistic Regression Models Predicting Hospital Death by Race/Ethnicity and Hospice Use…………………………………… ……...88 Table 9 Multiple Logistic Regression Mode ls Predicting Hospital Death for NonHispanic Whites…………………………………………………………… 89 Table 10 Multiple Logistic Regression M odels Predicting Hospital Death For nonHispanic Blacks…………………………………………………………….90

PAGE 6

iv List of Figures Figure 1 Conceptual Model ………………………………………………………….56 Figure 2 Interaction Effect of Race/Ethn icity on Cause of Death – Cancer with Hospice Use ………………………………………………………………..80

PAGE 7

v Predictors and Outcomes of Hospice Use among Medicare and Medi caid Dual-Eligible Nursing Home Residents in Florida: A Co mparison of Non-Hispanic Blacks and NonHispanic Whites Jung Kwak ABSTRACT This research investigated the racial/et hnic differences in hospice utilization and the effect of hospice in reduc ing the risk of hospital deat h at the end of life among nonHispanic Black and non-Hispanic White dualeligible (Medicare and Medicaid) older adults residing in a nursing home setti ng. The final study population included 30,765 non-Hispanic Black and non-Hispanic White nursing home residents who died between state fiscal years 2000-2002 in Florida. The behavioral model of health servi ces use successfully predicted group membership in hospice use. In the full model, seven variables female gender, nonHispanic White race/ethnicity, being marrie d, urban area of residence, and cancer and dementia/Alzheimer’s disease as causes of death – predicted in creased likelihood of hospice use. This study also found that race/et hnicity moderates the st rength of the effect of the illness on the likelihood of using hospice. Among residents who died of cancer, no difference in hospice use was found between th e two racial/ethnic groups while hospice utilization rate among non-can cer residents was lower for non-Hispanic Blacks than nonHispanic Whites.

PAGE 8

vi The same predisposing, enabling and need f actors tested in predicting hospice use were examined for association with the probability of survival time since hospice admission. The poor model fit and the sma ll number of factors found to significantly affect the probability of survival time from th e initial hospice enrollment suggest that the survival time might be influenced by external factors other than covariates examined in this study. This study found hospice to be a powerf ul predictor of place of death among nursing home residents. After controlling fo r other factors, hospice nursing home residents were 91 percent less lik ely to die in a hospital. At the same time, non-Hispanic Black residents were still 76 pe rcent more likely to die in a hospital even after adjusting for the effect of hospice use and other variables.

PAGE 9

vii Preface This dissertation work would not have been completed without th e guidance and support of the following individuals and organizations: William E. Haley, Ph.D.: I would like to express my si ncere gratitude to you for your mentorship throughout my graduate school training. You taught me many invaluable skills and challenged me to thrive for excellence. David A. Chiriboga, Ph.D.: As a mentor and scholar, you helped me to look at questions and issues with new perspectives. I will always remember to face challenges as opportunities to solve problems Susan C. McMillan, Ph.D., William S. Rowe, Ph.D. & Glenn E. Mitchell, Ph.D.: Dr. McMillan, you have been a wonderful role mode l for female junior researchers like me. I am truly thankful for your constructive quest ions and feedbacks. Dr. Rowe, I deeply appreciate your time and guida nce for helping me think abou t and look at broader issues related to end-of-lif e research. Dr. Mitchell, I tha nk you for providing many challenges that helped me to learn and grow. Trevor Purcell, Ph.D.: Thank you so much for taking time to chair my dissertation defense. I am honored to have you chair the defense. Larry Polivka, Ph.D.: I am truly grateful to you for sh aring your knowledge, experience, and moral visions and ideals through your te achings and actions. Your wise advices throughout the dissertation proce ss were a vital source of su pport and encouragement. Thank you. Bill Kearns, Ph.D.: I would like to expr ess special thanks to you for providing RUCA codes for Florida. Thank you for your generosity and assistance. Faculty and Staff at the School of Aging Studies : I thank the School of Aging Studies for financial support and excellent educationa l and training opportuni ties. I am deeply

PAGE 10

viii thankful to Ms. Mary Kaplan for her frie ndship and personal mentorship. I thank Dr. Jennifer R. Salmon for her mentorship especi ally during the early phase of graduate school training. I am grateful to Scott Hint on for his help in learning SAS and Dr. Ross Andel and Dr. Brent Small for their inputs in in terpretation of statistical results. I would also like to thank staff, particularly Pam, Rosa, Nancy and Gail for helping me on countless occasions. Thank you so much for your superb skills, knowledge, and humanity. Agency for Health Care Administration & Florida Department of Health: I want to thank these two state agencies for providing funding and permission to use data used for the dissertation. State Data Center on Aging, Florida Policy Exchange Center on Aging : I have been so fortunate to work as a research assistan t to learn about and wo rk with administrative datasets stored and managed in the State Data Center on Aging (SDCA) I am grateful to the ADCA for providing financial and tec hnical assistance duri ng the dissertation process. AARP: I would like to thank AARP for providing me scholarship that provided partial funding for my dissertation research. AAR P has provided me invaluable opportunities throughout my graduate school training. AARP scholarship, in ternship and a research contract helped me gain better understanding of gerontological educ ation, and aging and health care policies. Thank you so much for your support. My Sister and Brother: Kyung, you have always been an important role model in my life. Through your life, you ha ve shown me the value of perseverance, dedication and commitment to work that matter. Thank you so much for your love and patience with me. I will be forever thankful for you putting up with my sporadic moments of insanity and hysteria. Isaac, I am so thankful for your prayer and encouragement. Thank you for believing in me. Allison : My best friend, confidant, and toughest critic. I don’t know how I could have kept my sanity without your loyalty and devo tion to our friendship. I am so honored and lucky to have a friend like you who po ssess brilliant and sh arp mind, insatiable intellectual curiosity, compassion, courage and great sens e of humor. Thank you for your love, support, and guidance thr oughout all these years. Most of all, thank you for being the best friend anyone can ever hope to have. Linda: Thank you so much for helping me to keep perspectives. I am thankful for your guidance, warmth, kindness, and generosity of hear t. I feel very lucky to be a recipient of your friendship.

PAGE 11

1 Chapter One: Background Overview A growing body of literature suggests that hospice is a critical service for individuals with terminal illness because it leads to such better outcomes as reduced hospitalization, better pain management and better sati sfaction with care (Baer & Hanson, 2000; Teno et al., 2004). Given better outcomes, the underutilization of hospice services and late entrance to hospice, esp ecially among older adults in nursing homes and members of racial/ethnic minority backgrounds has been a growing concern since the Medicare hospice benefit was created in 1982 (Gorden, 1996; Medicare Payment Advisory Commission [MedPAC], 2004; Rees e, Ahern, Nair, Schr ock, & Warren-Wheat, 1999). The study presented in this dissertati on research focuses on racial/ethnic differences in hospice utilization and the effect of hospice in reducing the risk of hospital death at the end of life among non-Hispanic Black and non-Hispanic White dual-eligible (Medicare and Medicaid ) older adults residing in a nur sing home setting. When studying the effects of race/ethnicity on hospice use dur ing the last year of life, Blacks are of special interest since they make up one of the largest ethnic minority groups, at 12.3 percent of the U.S. population (U.S. Census, 2000). In addition, disparities in access and in the use of medical services including end-of-life care be tween White and Black groups have been reported by a number of studi es (see Krakauer, Crenner & Fox, 2002 for

PAGE 12

2 review of this issue). The present stu dy specifically examines not only differences between the two racial/ethnic groups in quest ion, but also individua l differences within each group in hospice use. The reason for the within-group examination is that previous studies have reported that there is more vari ability within each racial/ethnic group than across groups (e.g., Dilworth-Anderson, Williams, & Gibson, 2002). In addition to examining the predictors of hospice use, fact ors associated with survival time from enrollment in hospice are also examined since late entrance to hospice (e.g. shorter survival after admission for the same diseas e state) has been a serious concern for practitioners and policy makers. The study also addresses an importa nt end-of-life care outcome, place of death, as a result of hospice use. For studying utilization of hospice serv ices, duration of hospice use, and outcomes of hospice use, the expanded version of the behavioral model of health services use (Andersen, 1995) originally developed by Andersen in 1968 was adopted. The behavioral model of health se rvice utilization has been used extensively in research, and in particular has been used to identify access disparity among racial/ethnic groups (Aday & Awe, 1997; Gabbe et al., 1995). According to the model, three categories of individual determinants covering predisposing, enabling, ne ed factors and psychosocial factors, as well as health care system determinants aff ect health services u tilization and outcomes (Aday & Andersen, 1974; Andersen, 1968; A ndersen & Newman, 1973; Bradley et al., 2002). In order to increase the model’s salience to end of life issues, selected theoretical models from the death and dying and literature on the role of race/et hnicity on end-of-life decision making were reviewed to facilitate us e of the behavioral model of health service utilization in this research.

PAGE 13

3 Chapter 1 begins with a brief overview of health service utiliza tion at the end of life, followed by a review of the literature on theories on health services utilization and death and dying, and a synopsis of predictors of hospice use, hospice length of stay and place of death. Specific research questions regarding predisposing, need, and enabling factors predicting hospice use, hospice length of stay, and place of death are presented. In both the reviews and research questions, the central focus is on the implications for race and ethnicity. The purpose of the liter ature review is to provide a critical and comprehensive background on theories and rese arch findings on end-of-life services use with a specific focus on the role of race/ethnic ity in hospice, length of stay in hospice and place of death to identify the gaps in the curr ent literature which led to development of research questions for this disse rtation research. The review pays particular attention to theoretical frameworks relevant to the rese arch effort, since theoretical frameworks facilitate both the selection and the examination of the effects of multiple variables (George, 2002). A review of the current medi cal and long-term care se rvices available in the U.S. is also necessary to better u nderstand the context in which people choose treatment options and services which often represent competing goals of treatment and care at the end of life (i.e., treatments with curative vs. palliative treatment purpose). In the review of the literature, findings from studies with varying methods of conceptualization and operati onalization of “end-of-life” issues are included, ranging from studies that examine the last few days before death to studies examining the last three years before death. Such a broad time frame for defining the end-of-life period is used in this review because there is no c onsensus on when the dying process actually is initiated – indeed much may depend on the i llness and the individual and there are many

PAGE 14

4 variations of definition of the end of life in the literature (Field & Cassel, 1997). As a report by the Institute of Medicine (IOM) ( 1997) recognizes, the dyi ng process can take from days to several months or even for y ears due to varying trajectories and progression of different illnesses. It thus appears to be appropriate to examine health service utilization patterns during a wide span of time before death. Introduction End-of-life service utilization among olde r adults is an in creasingly important area of research due to the complex needs fo r care at the end of life, and the high costs associated with such care. During the final months, older adults have special needs that are distinct from younger individuals. In part the differences arise because older adults are more likely to manifest multiple comorbidities, greater levels of functional impairment, and generally greater needs fo r care (World Health Organization [WHO], 2004). As a result, many older individuals die after a prolonged period of chronic illnesses and after spending time in a nursing home (Hogan, Lynn, Gabel, Lunney, O’Mara, & Wilkinson, 2000). A substantial po rtion of medical treatment and supportive services at the end of life is related to cari ng for older adults with declining health and functional status and increasing comorbid ity (Culler, Callahan, & Wolinsky, 1995; Stump, Johnson, & Wolinsky, 1995; Hogan et al., 2000) and cost of end-of-life care is highest during the final few months of life (Yang, Norton, & Stearns, 2002). Costs of caring for older dying patients on average are si x to seven times of those for survivors (Levinsky, Ash, Yu, & Moskowitz, 1999; Lubitz & Riley, 1993; Rile y, Lubitz, Prihoda, & Rabey, 1987).

PAGE 15

5 The high cost is associated with the many types of medical and social services are used by older adults during th e last year of life, incl uding hospital care, physician services, skilled nursing care, and hospice care. Alone among these services, hospice is a program specifically designed to meet various medical, psychosocial, and spiritual needs of patients at the end of life. Hospice services are reimbursed by Medicare, the primary source of insurance for almost all older Americans aged 65 and olde r, and Medicaid, the primary payment source for long-term care for dual-elig ible older adults, [National Hospice and Palliative Care Organization (NHPCO), 2002]. Despite the support of medical insurers, and the increasing empirical evidence supporting th e benefits of hospice use (Christakis & Iwashyna, 2003; Miller, Gozalo, & Mor, 2001; Miller, Mor, Wu, & Lapane, 2002; Teno et al., 2004), underutilization of hospice has been reported among Medicare beneficiaries. This is especially the case for racially and ethnically diverse ol der adults and nursing home residents (Hoffmannn & Tarzian, 2005; Iwashyna, Chang, Zhang, & Christakis, 2002; MedPAC, 2004; Virnig, Kind, McBean, & Fisher, 2000). Although timely access to hospice would provide better opportunities for patients, families and hospice providers to develop and coordinate an a ppropriate plan of care, persis tent pattern of short duration of hospice use remains as a serious concern. Health Services Utilization by Ol der Adults at the End of Life Health services use by older Medicare benefi ciaries in the last years of life have been studied extensively. With declining functional abil ity, deteriorating health and increasing severity of illnesses, older adults utilize various types of medical and longterm care services as they approach the e nd of life. These services are accessed at

PAGE 16

6 various health care and long-term care settings Many older adults transition in and out of different health care institutions including hospital and nursing homes as their illnesses progresses near the end of life. Among health care inst itutions, hospitals are the most frequent site of death: from 50 to 60 percent of all deaths occur in hospitals (Hogan et al., 2000; Weitzen, Teno, Fennell, & Mor, 2003). Hosp itals are also used more frequently by older adults after they receive a diagnosis of serious progressive illnesses (Field & Cassel, 1997). A recent study of decedents enro lled in the Program of All-Inclusive Care for the Elderly (PACE) (Mukamel, Bajors ka, & Temkin-Greener, 2002) shows that hospitalization use increases as early as 7 mont hs before death and continues to increase until the last month of life; there was an incr ease of almost 250 percent in the last month among this group of older adults. Many patients, who spend time at hospitals during the last year of life, or during the last few days before death, are transfe rred from nursing homes. At the same time, nursing homes are another common site of deat h; this is where 20 percent of all U.S. deaths occur (Mezey, Dubler, Mitty, & Brody, 2002). A significant minority of older adults spend varying periods of time at a nursi ng home in their last year of life. An analysis of the 1993 National Mortality Foll owback Survey by Hogan and colleagues (2000) showed that in 1993 about 38 percent of older adults resided in a nursing home for either a full (22 percent) or pa rtial year (16 percent) and 59 pe rcent of older adults resided at home during the last year of life. The same analysis (Hoga n et al., 2000) indicated that the majority of nursing home decedents who spen t a full year in a nursing home died in a nursing home (67 percent) while about 28 percent died in hosp itals and 1 percent died at home. Of those who stayed part of the year 1993 in a nursing home, 48 percent, 35

PAGE 17

7 percent and 10 percent died in a nursing home, hospital and home respectively. These results suggest that older adul ts who enter a nursing home for a short-term stay are more likely to be transferred to hospitals and die at hospitals than lo ng-term nursing home residents. The literature on hospitalization of nursi ng home residents reveals the complex process of decision making on hospitaliza tion of nursing home residents and both beneficial and harmful effects of hospitaliza tion of nursing home resi dents at the end of life. Hospitalization is often inevitable and also potentially appropriate for older adults who die of organ failure (Mezey et al., 2002) On the other hand, a number of studies report that hospitalization is not necessari ly appropriate for all dying nursing home residents (Engle, 1998; Mezey et al., 2002). Th e literature documents that transfer from skilled nursing facilities to hos pitals can lead to adverse outcomes such as sudden death, disruption of care plan, disori entation, and unnecessary financ ial costs associated with transfer (Creditor, 1993 ; Fried, & Mor, 1997; Sanders & Morley, 1993). Hospice has been found to not only provide better pain and symptom management (Baer & Hanson, 2000; Teno et al ., 2004) but also to reduce ove rall hospitaliz ation rates at the end of life and to facil itate dying in place (Miller et al., 2001; Teno et al., 2004). In the U.S., hospice is the main source of pallia tive care which has been identified as one of the most important services that can be provided to terminally ill patients and their families [Gage et al., 2000; NHPCO, 2002; WHO, 2004]. For hospice care, Medicare provides the bulk of financing. It was the pr imary payer for 80 percent of all patients served by hospice organizations in 2002 (NHPCO, 2002). Medicaid plays a somewhat smaller role: it covers hospice care for Medi caid beneficiaries wit hout other sources of

PAGE 18

8 insurance and in 45 states and the District of Columbia it reimburses for nursing home care for dual-eligible nursing home residents who elect hospice care. (Tilly & Wiener, 2001). Therefore, it is important to review some of the ma jor features and requirements of the Medicare hospice benefit (MHB ) and Medicaid hospice benefit. Overview of the Medicare and Medicaid Hospice Benefits The Medicare hospice benefit (MHB) was created under the Tax Equity and Fiscal Responsibility Act of 1982 (TEFRA) (Gag e et al., 2000). A new type of formal service that specializes in the care of the dying, the Medicare hospice benefit covers a broad set of palliative services for beneficiaries with a diagnosed terminal illness who are expected to die within six mont hs if their illness runs a normal course (Gage et al., 2000). Since 1974 when hospice was first introduced in the United States, the number of health care organizations providing hospice servic es and public recognition of patient preferences to receive palli ative care at home has grow n rapidly (Greer & Mor, Sherwood, & Birnbaum, 1983). By adding the ho spice benefit to the Medicare, Congress allowed beneficiaries with a terminal illne ss to choose between comprehensive palliative care and conventional care (MedPAC, 2004; Moon & Boccuti, 2002). Under the 1982 Medicare hos pice legislation, all Medica re Part A beneficiaries became eligible to receive hospice benefits if the Medicare beneficiary met both of the following criteria: (a) cer tification by both an attending physician and the hospice medical director (or other physician affiliated with the hospice) that the individual has a terminal illness with a likely prognosis of six months’ or less anticipated survival if the disease runs its normal course; and (b) benefici ary’s agreement in writing to a care plan focused on palliative care of their terminal illness while forgoing curative treatment

PAGE 19

9 (Gage et al., 2000). Thus, to receive the Medicare hospice benefit, beneficiaries must waive all Medicare coverage of non-palliative care related to their terminal illness. Every state except for Connecticut, Nebr aska, New Hampshire, Oklahoma, and South Dakota provides a hospice benefit unde r Medicaid (Tilly & Wiener, 2001). Although there are some variations in the overall structure and design of Medicaid hospice programs across states, the eligibil ity, benefit coverage, and payment rate structure of Medicaid hospice programs are designed similarly to the MHB (Tilly & Wiener, 2001). A few notable differences be tween the Medicaid and Medicare hospice programs need to be mentioned however. Medicaid programs cannot impose copayments on hospice patients while the MH B does; instead, the programs must pay Medicare hospice co-payments for dually eligible benefici aries (Tilly & Wiener, 2001). In addition, when a dual-eligible nursing facil ity resident chooses hospice in the nursing home, Medicaid allocates funds to the hospice, which is responsible for paying the nursing home 95 percent of the Medicaid nur sing facility room and board rate. Hospice as a Model of Care Hospice care is considered as a model of care for the dying in terms of comprehensiveness of services provided a nd its focus on patientand family-centered care and continuity of care by practitioners, consumers, and researchers (Lynn, 2001). Hospice care represents a patientand fam ily-centered approach th at is provided by an interdisciplinary team of health care provi ders including physician, registered nurse, social worker, pastoral or other counsel or, and volunteers (NHPCO, 2002). The MHB and Medicaid hospice benefit programs provide a comprehensive coverage for medical care (focusing on symptom and pain manage ment), prescription drugs, bereavement

PAGE 20

10 counseling, respite care, and other services not ordinarily covered by Medicare. Continuity and coordination of care are consid ered key strengths of hospice care (Lynn, 2001). This is because hospice staff take re sponsibility for all as pects of a Medicare beneficiary’s care related to the terminal illne ss. Family members also receive education and support from hospice providers during caregiving and bereavement and hospice appears to decrease the bur den of bereavement among surviving spouses (Christakis & Iwashyna, 2002). While the exact numbers of patients receivi ng hospice service in different settings are not clear, accordi ng to the 2002 NHPCO National Data Set Summary Report, the majority (58 percent) of terminally ill patients were at home (i.e., private residence) at the time of enrollment while 22 percent of hospi ce patients were in a nursing home at the time of enro llment in 2002 (NHPCO, 2003). Recent empirical studies also provide ev idence that hospice care leads to better end-of-life outcomes. These outcomes include lower hospitalization (M iller et al., 2001) among nursing home residents and better pain management among both community dwelling patients and nursing home residents (G reer et al., 1986; Miller et al., 2002; Teno et al., 2004), better satisfaction among patie nts and families (Kane et al., 1984; Teno et al., 2004) and a greater likeli hood of dying at home (Hogan et al., 2000; Iwashyna et al., 2002; Moinpour & Polissar, 1989). Although strengths of hospice care ar e supported by empirical evidence and recognized by patients and familie s and practitioners, recent studi es demonstrate that cost savings due to hospice service are found onl y among a selected group of dying patients (Campbell, Lynn, Louis, & Shugarman, 2004) and cost saving is sensitive to time frame of assessment (Miller, Intrator, Gozalo, R oy, Barber, & Mor, 2004). One recent study of

PAGE 21

11 Medicare decedents (Campbell, Lynn, Loui s, & Shugarman, 2004) found cost saving among patients who were younger or diagnosed with cancer. Cam pbell and colleagues (2004) also found that costs in creased for older patients, especially those who were diagnosed with non-cancer illnesses. A nother study among dual -eligible nursing home residents by Miller and colleag ues (2004) found that regardle ss of length of stay among all nursing home residents, mean total public expenditures in the last month of life were significantly less for hospi ce than non-hospice reside nts. In the last six months of life, however, total mean public expenditures were significantly greater for hospice than nonhospice residents. Underutilization of Hospice Since the MHB benefit was created in 1982, the number of MHB beneficiaries has grown rapidly. In 2001, 580,000 Medicare beneficiaries en rolled in hospice compared to only slightly more than 150,000 in 1992 (MedPAC, 2004). Many beneficiaries who are eligible for and benef it from hospice, however, do not utilize it. Using 1996 data, Virnig and colleagues (2000) found that only a bout 15.5 percent of Medicare decedents used hospice. One im portant caveat lies in their approach to calculating the utilization rate of hospice: using all Medicare decedents as a denominator may underreport utilization rate among eligible beneficiaries since some beneficiaries experience sudden, unexpected death. Still, Lunney and colleagues (2002) estimated that hospice use rate was less than 50 percent ev en among decedents who died of identifiable clinical causes other than s udden death or unidentifiable cause of death. Moreover, among cancer patients who are characterized w ith relatively predictable trajectory of illness progression, and therefore prime candida tes for hospice use (Lunney et al., 2003),

PAGE 22

12 Lunney and associates (2002) found that less than half used hospice. Of important note is that the utilization rate of hospice differs across states. In the state of Florida, for example, approximately 35 percent of eligib le adults aged 65 a nd older are served by hospice organizations (NHPCO, 2003). In a ddition, 60 percent of hospice enrollees in Florida are diagnosed with illnesses ot her than cancer (NHPCO, 2003). Overall, the results suggest that hospi ce services are underutilized even among those beneficiaries who have relatively be tter access to hospice due to their disease characteristics. Moreover, short length of st ay in hospice is anothe r serious issue related to hospice utilization. In 2003, over 30 percent of hospice enrollees died within 7 days of admission and the median length of stay in hospice was 22 days (NHPCO, 2005). Although underutilization of hos pice benefit and short durat ions of hospice stay among terminally ill patients have become a fo cus of increasing research, most of the research examining predictors of and potenti al barriers to hospice use employ exploratory study designs. The absence of a theoretical basis for most of the existing empirical studies on end-of-life care and end-of-life decision making has been identified as another major limitation of current end-of-life rese arch (George, 2002; Tilden, Tolle, Drach, & Hickman, 2002). For a better understanding of the co mplex processes by which a number of demographic, clinical, organizational and health care market area factors influence hospice use, more research w ith a theoretical framework is needed (George, 2002). Without a theoretical or conceptual model, it is difficult to identify appropriate constructs and their specific relationship s to one another (Dilworth-Ande rson et al., 2002). The next section of the literature revi ew, therefore, discusses the behavioral model of health

PAGE 23

13 service use and several theories on death and dying, in the context of end-of-life. The behavioral model of health se rvice utilization was chosen as the centerpiece of the review because it has been used extensively in serv ice utilization research and also to identify access disparity among racial/ethnic groups. The behavioral model of health service utilization’s strengths lie in its adaptabili ty in conceptualizing and categorizing the myriad of correlates of use of health and social services (Grabbe et al., 1995) and identifying utilization of subgroups, especia lly racial/ethnic minor ity groups (Aday & Awe, 1997). Thus, this model can be applied to studying diversity a nd disparity in access to end of life care serv ices such as hospice. In a ddition, theories on death and dying provide contextual information that is specifi c to the experience at the end of life. The literature suggests that the complex process of making the decision to use hospice is influenced by many factors including knowledge and perception of th e severity of illness and the possibility of recovery by the pa tient, physician and family during which race/ethnicity plays an important role. Literature on Health Services Use, Deat h and Dying, and Racial/Ethnic Diversity in Endof-Life Decision Making Behavioral Model of Health Service Utilization The behavioral model of health servi ce utilization was in itially developed by Andersen in 1968 to explain use of health se rvices by families and to define and measure access to health care (Andersen, 1968). Sin ce the initial model was introduced, it has been modified and expanded by Andersen and colleagues (Aday & Andersen, 1974; Andersen & Newman, 1973; Andersen, Davidson, & Ganz, 1994; Andersen, 1995; Bradley et al., 2002; Evans & Stodart, 1990) The four major components of the

PAGE 24

14 behavioral model of health service utilization (Aday & Andersen, 1974; Andersen, 1968; Andersen & Newman, 1973; Andersen, 1995) in clude the individual determinants, health care system determinants, health behavior and outcomes. Each includes several categories. The individual determinant component cons ists of three categories of factors predisposing, enabling and need factors (Aday & Andersen, 1974; Andersen, 1968; Andersen & Newman, 1973). Predisposing ch aracteristics are char acteristics of an individual that would either predispose or deter the pe rson from using the service and include demographic characteristics (i.e. age, gender), social struct ure (i.e. education, occupation, race/ethnicity), and health belie fs (i.e. attitudes, beliefs, knowledge of medical care, provider and disease). Health be liefs can be shaped by culture and social group. The enabling resources component incl udes the means by which an individual could use services such as personal (i.e. income, insurance coverage, having a regular source of care, the nature of the regular source of care, and the accessibility of the source), family (i.e. social network and social support) and community (i.e. availability of health personnel and facilities, geographic location – rural or urban nature – of the community in which the individual lives) res ources. The need component, in turn, refers to illness level or need for care and is ofte n the most immediate cause of health service use (Andersen, 1995). The need component has two parts: illness level perceived by the individual and illness evaluated by the deliver y system. Two versions of the model (Aday & Andersen, 1974; Andersen & Newman, 1973) however, do not make a conceptual distinction between need for services and illness level or measurement approaches to

PAGE 25

15 these concepts. Instead, illness level perceived by individuals (e.g., days of disability) or evaluated illness (e.g., diagnoses) are used. Health care system determinants refer to the arrangements made to provide care to potential consumers and include volume and distribution of resources, access to resources, structure of resource organizat ion, and health policy (Andersen, 1995). Volume and distribution of re sources refer to personnel/popu lation ratios for types of personnel or facilities and the way in which resources are distributed within a geographic area. The access component of the health car e system refers to the means by which the patient gains entry to the medical care system and continues the tr eatment process (e.g., diagnostic criteria for receivi ng certain treatments or out-ofpocket cost). The structure component is defined as characteristics of th e organization (e.g., medi cal facility) that the patient deals with following entry (e.g., mean s of admission to the hospital or processes of referral to other sources of care). However, Andersen and Newman (1973) acknowledged that the structural component is difficult to define and measure due to the many facets of structure and its high correlation with other co mponents in the model such as policy variables. Health policy factors (Aday & Anders en, 1974) include fi nancing, education, manpower, and organization. State-to-state va riability of public he alth policy variables such as financial and need eligibility criter ia for national programs such as Medicaid services is an example of the policy variable. Health service utiliz ation is the health behavior co mponent and refers to actual utilization of services measured by exam ining type (e.g. hospital, physician, nursing home, etc), purpose (i.e. primar y care, secondary care, tertia ry care, custodial care) and

PAGE 26

16 unit of service (i.e. contact, volume, episodi c care) (Aday & Anders en, 1974; Andersen & Newman, 1973). Several studies published during the 1990s suggested inclusion of personal health behavior as one of subcompone nts, along with use of health services, in the health behavior component and perceived health status, evaluate d health status and consumer satisfaction as outcome measures (Andersen et al., 1994; Evans & Stodart, 1990). According to the behavioral model of services utilizati on (Aday & Andersen, 1974; Andersen & Newman, 1973; Andersen, 1 995), health service use behavior is determined directly by indivi dual determinants and indirect ly by health care system factors. Health care system factors directly influence Individual determinants and outcomes and indirectly influence health se rvices use through individual determinants (Andersen, 1995). Within the individual dete rminant components, predisposing factors precede the enabling factors which are follo wed by need factors (Aday & Andersen, 1974; Andersen & Newman, 1973). Howeve r, the expanded model by Bradley and colleagues (2002) conceptualizes psychosocia l determinants such as health beliefs, knowledge and attitudes as related but separa te construct from pr edisposing factors and posits that psychosocial factors may be mech anisms by which race and ethnicity (one of the predisposing variables) may influence use. This version of the behavioral model of services utilization (Bradley et al., 2002) suggests that psyc hosocial factors follow rather than precede enabling and need factors and se rve as mediating factors. Lastly, health services use directly influence outcomes. The behavioral model of services utilizat ion has been utilized by and applied to a wide range of health services use among olde r adults including gene ral health services

PAGE 27

17 utilization (e.g., hospital and physician services), long-term care (e.g., nursing home and home care use), social services, adult day care and dental services fo r older adults (see Aday and Awe, 1997, for a review of this issu e). The behavioral model of services utilization was also recently used as a framework for studying hospice visit volume across nursing home and non-nursing home se ttings (Miller, 2004). Moreover, a number of studies examining racial/e thnic disparities in service utilization among older adults have been based on this model (e.g., Br adley et al., 2002; Dunlop, Manheim, Song, & Chang, 2002; Kart, 1991; Mui & Burnette 1994; Wallace, Levy-Storms, Kington, & Andersen, 1998; White-Means & Rubin, 2004). Th e behavioral model of health service utilization’s concept of equitable access provides a framew ork for examining the role of potential access measures (enabling resources) in predicting different groups’ actual rate of utilization or health care outcome (Ada y & Awe, 1997). By identifying mutable (i.e., being amenable to changes such as public policy, health care be liefs and attitudes, knowledge and sources of health care informa tion, insurance coverage, regular source of care, ease of getting care) and immutable popul ation characteristics such as predisposing and need factors, the model offers a systema tic approach to examining the relative effect of each component of the model and the opportunity to determine whether subgroups with distinct immutable attr ibutes experience equitable access to health services (Andersen, 1995). Keeping in mind the strengths of and cont ributions made by the behavioral model of services utilization in health services re search, it is nonetheless important to recognize potential weaknesses of the model. Several i nvestigators have raised concerns about the conceptualization and operati onalization of the major model constructs, the specification

PAGE 28

18 and testing of relationships among constructs, and the robustness and generalizability of findings based on the various versions of the model (e.g., Aday & Awe, 1997; Porter, 2000). One criticism relates to the concern that predictors ma y not have been adequately captured due to inconsistent conceptualization and measurement of concepts such as need and illness. A second general criticism of the Andersen mode l is that the original and modified versions of the model do not spec ify the relationships between psychosocial variables (e.g., beliefs and know ledge regarding illness), en abling (e.g., social network), need and service use (Bradley et al., 2002; Kart, 1991). A third general criticism concerns the Andersen mode l’s overall predictiv e and explanatory capacity and the prioritizing of need as the dominant predic tor (Aday & Awe, 1997). The model tends to explain a small amount of variability in predicting service utilization, with need consistently being the dominant predictor (Aday & Awe, 1997). Notwithstanding the several limitations of th e behavioral model of health service use, the flexibility of the model suggest s that improvements can be made. Many researchers have expanded the Andersen m odel by adding explanator y variables that are relevant and specific to the type of utilizati on behavior explored in their studies. For example, some studies have added variables m easuring characteristics of kin or caregiver of the individual that may affect the servic e utilization (Bass & No elker, 1987; Coulton & Frost, 1982; Counte & Glandon, 1991; Freed man, 1993; Mutran & Ferraro, 1988). Theories on Death and Dying Major theories on death and dying develope d in the last three decades include the stage of dying theory (Kubler-Ross, 1969), th e context of awarene ss theory (Glaser & Strauss, 1965), dying trajectories (Gla ser & Strauss, 1968), the livingdying

PAGE 29

19 interval/phase theory of dyi ng (Pattison, 1977), the task-based approach to coping with dying (Corr, 1992), and the readiness-to-d ie theory (Copp, 1996) (see the review by Copp, 1998). Of these theories, the dying tr ajectories idea introduced by Glaser and Strauss (1968) and the readin ess-to-die theory (Copp, 1996) ar e particularly useful in understanding the clinical and psychological context in which people make decisions on treatments and services to use at the end of life. Dying trajectories (Glaser & Stra uss, 1968) provides a framework for understanding two important elements in an individual’s dying traj ectory: duration and shape. Glaser and Strauss (1968) proposed th at depending on the type of major illness, individuals may experience different durati ons (e.g., rapid or slow progression toward death) of clinical and functional changes of varying trajectories (e.g., short term improvements and relapses, crisis, plateaus, etc). According to the dying trajectories theory (Glaser and Strauss, 1968; Lunne y et al., 2002; Lunney, Lynn, Foley, Lipson, & Guralnik, 2003), there are four modal trajectories: sudden d eath, terminal illness, organ failure, and frailty. Some people die s uddenly without experiencing any significant change in functional ability (i.e., sudden d eath). Individuals in the terminal illness category are characterized with a reasonably hi gh functional status th at rapidly declines within 6 weeks to three months before deat h, which is typically shown among some types of cancer patients. The organ failure group ha s serious organ system failure related to illnesses such as congestive health failure or chronic obstructive disease and they experience gradually decreasing functional abil ity that is often accompanied by periodic exacerbations of conditions. The last categor y is the frailty group in which individuals experience slow decline of functional ability with steadily progressive disability due to

PAGE 30

20 illnesses such as stroke or dementia. Lunney and colleagues (Lunney et al, 2003) have found support for the idea of different functional trajectories of dying in older adults with different main illnesses. Understanding diffe rent trajectories of illnesses is important in the context of hospice use at the end of life. For older adults to receive the hospice benefit through Medicare in the U.S., a physician must certify th at they have less less than 6 months to live (Gage et al., 2000). Ho wever, the majority of older adults die of serious illnesses without clear f unctional trajectories that indi cate whether the patient is in terminal phase. According to the analysis of Medicare beneficiar ies who died between 1993 and 1998 (Lunney et al., 2002), the majority of decedents experienced organ system failure or frailty trajectories and only about 20 percent of decedents died of cancer which follows more predictable terminal trajectory mode of dying. Physicians have been found in numerous studies to greatly overestimate likely survival time (Lamont & Christakis, 2001), and non-cancer diagnoses present particul ar challenges. Indeed, the clinical characteristic of the patient appears to be a strong predictor of hospice use as studies consistently have shown cancer diagnosis as a significant predic tor of hospice use (Iwashyna et al., 2002; MedPAC, 2004; Moon & Boccuti, 2002; Virnig et al., 2000). Even if serious physical and functional decline may indicate impending death, the person may not be psychologically ready to accept the poor prognosis. The readiness-todie theory (Copp, 1996) argues that while the bo dy (i.e. biological st ate) is approaching death, the person (i.e. psychologica l state) may not be psychol ogically ready to die. The state of readiness of the pers on can be influenced by a number of factors, especially by previous treatment history of the illness. If the initial diagno sis of the illness is made at an advanced stage of the illness, the person may not be able to move to the ready state as

PAGE 31

21 quickly as the person who has received a nd pursued curative treat ments during earlier stages of the illness and/or has simply had time to think about the seriousness of the problem. The person may also appraise the illness as curable despite poor prognosis due to such reasons as religious belief in miracle, or may perceive the hospice option as a form of denial of care rather than an appropria te care approach. This latter perception may results from previous experience of discriminati on in the health care system. In this case, the coping response of the person may focus on tasks of survival a nd cure rather than tasks of dying. Thus, coping responses can be either pursuing curative treatment or palliative treatment depending on th e appraisal of the illness. The Role of Race and Ethnicity in End-of-Life Decision Making The literature details the importance of the cultural norms and behaviors of a racial/ethnic group in influenc ing the individual response to the terminal illness and decision to use hospice (Ersek et al., 1998; Kalish & Reynolds 1976; Reese et al., 1999). A classic study on death and dying among et hnic groups by Kalish and Reynolds (1976) suggests that ethnic variation is an important factor influe ncing attitudes, beliefs and expectations regarding deat h and dying although individual differences within any racial/ethnic groups also exist. Persons of ethnic minority backgrounds may be more likely to turn to traditional norms and practi ces at the end of life because religious and cultural beliefs and norms can provide them with meaning for thei r illnesses and guide them in making decisions rega rding treatment and care options (Ersek et al., 1998; Reese et al., 1999). Thus, a careful examination of the role of culture associated with

PAGE 32

22 racial/ethnic groups in health care decision making is needed to better understand how individuals make decisions to use hea lth services such as hospice. A recent review of the role of race and ethnicity in end-of-life decision making by Kwak and Haley (2005) documents that race/ethn icity plays an important role in shaping preferences for and actual deci sions in end-of-life care pla nning and use of treatments. Among 33 empirical studies, including both qualita tive and quantitative studies that were published between 1992 and 2003, Kwak and Haley (2005) found multiple potential explanatory factors for racia l/ethnic diversity in end-oflife decisions, including cultural values and norms regarding death and dying, individual level of acculturation and assimilation as well as sociodemographic char acteristics, and shared history within the U.S health care system among older adults of racial/ethnic minority backgrounds. Kwak and Haley (2005) also report that Blacks were the most frequently studied racial/ethnic group (28 studies out of 33 studies included Blac ks in their study samples) and were consistently found to prefer more a ggressive treatments th an other racial/ethnic groups as indicated by: (a) advance directiv e documents (Degenholtz et al., 2002; Eleazer et al., 1996; Hopp, 2000; Hopp & Duffy, 2000; Kiely, Mitchell, Marlow, Murphy, & Morris, 2001; McKinley et al., 1996; Mebane et al., 1999; Murphy et al., 1996; Phipps, True, Harris, Chong, Tester, Chavin et al., 2003); (b) inte rviews (Allen-Burge & Haley, 1997; Blackhall et al., 1999; Ca ralis et al., 1993; Cicirelli, 1997; Garrett et al., 1993; Klessig, 1992; McKinley et al., 1996; Mebane et al., 1999; O’Brien et al., 1995; Phillips et al., 2000; Phipps et al., 2003; Waters, 2001); and (c) actual use (Gessert, et al., 2001; Hopp & Duffy, 2000).

PAGE 33

23 The review by Kwak and Haley (2005) reports distrust of the health care system as one of the most frequently mentioned e xplanation for racial/ethnic differences in advanced care planning and preferences for e nd-of-life care treatments. It is suggested that low level of trust in Blacks in the U. S. health care system is due to personal experience with poor access to me dical care and awareness of abuses in medical research in the past, for example, the Tuskegee s yphilis study (Berger, 1998; Krakauer et al., 2002; Murphy et al., 1996; Waters, 2001). One of the key issues that needs to be a ddressed in future re search is individual variation regarding end-of-lif e related issues within a single racial/ethnic group (Kwak & Haley, 2005). As within-group differences among diverse groups are influenced by sociodemographic factors such as age, gende r, education, accultura tion and location (e.g. rural vs. urban) (Barker, 1992), it is suggested that these f actors and a greater awareness of their potential theoretical significance should be more fr equently incorporated into research designs for the study of endof-life care (Kwak & Haley, 2005). Summary The current literature on health services use suggests that the Andersen model’s strengths lie in its usefulness in conceptualiz ing and categorizing the myriad of correlates of use of health and social services (Grabbe et al., 1995) and identifying patterns of service use by subgroups, and its flexibility and openness to modification (Aday & Awe, 1997). Moreover, explanatory power of th e model can be improved by inclusion of additional variables and further specifica tion of relationships between individual determinants which can be informed by dying trajectories, readiness-to-die theory, and the literature on endof-life decision making. Dying traject ories (Glaser & Strauss, 1968),

PAGE 34

24 and readiness-to-die theory (Copp, 1996) provide the soci al and clinical context for understanding end-of-life decision making proc ess while the litera ture on end-of-life decision making across diverse racial/et hnic groups suggests the importance of recognizing the diversity across ra cial/ethnic groups. Illness a nd need factors, such as the principal illness and associated trajectory, a nd psychosocial factors, are likely to have strong effects on the decision making process and service utilization. Factors Associated with Hospice Use Underutilization of hospice among termin ally ill patients who can potentially benefit from its comprehensive benefit packag e of services has led to an increase in research efforts in identifying predictors of hospice use. A literatur e review of empirical research, expert opinions, and commentaries on this topic identified a number of potential determinants of hospice enrollment. Thes e consist of (a) individual determinants including predisposing, enabling, need and psychosocial factors; (b) hospice service provider related factors; a nd (c) Medicare policy and orga nizational practice of nursing home factors. The literature on factors associ ated with length of stay in hospice is less extensive, but many factors that have been st udied are also predicto rs of hospice use. Although most empirical studies that examined predictors of hospi ce use reviewed here are atheoretical, factors identified by thes e studies reflect many elements of the behavioral model of health serv ice use. Therefore, the revi ew of the literature on hospice use is organized within the framework of the behavioral model of health service use and factors associated with hospice length of stay are also summarized. Individual Determinants of Hospice Use

PAGE 35

25 Predisposing factors. Among predisposing factors, ag e, gender, education and race have been identified to be associated with hospice use. Two studies examining Medicare decedents (Iwashyna, Chang, Zhang, & Christakis, 2002; Virnig et al., 2000) and a study of Medicaid and Medicare dual-elig ible decedents in California (Enguidanos, Yip, & Wilber, 2005) found that being older was significantly asso ciated with higher likelihood of using hospice after controlli ng for gender, race, and main principal diagnosis. Some studies suggest that older patients are treated differently in health care settings. Even after adjusting for disease seve rity and patient prefer ences, older adults are less likely to receive intensive treatm ents than younger patients (Lubitz & Riley, 1993; Riley, Lubitz, Prihoda, & Rabey, 1987; Sp ector & Mor, 1984). Thus, it is possible that health care providers su ch as physicians may see it as more appropriate for older patients with severe illnesse s to receive hospice care than younger patients, or older adults are more willing to forego life sustaini ng treatments to enhance quality of life. The same studies (Enguidanos et al., 2005; Iwashyna et al., 2002; Virnig et al., 2000) also found that women were more likely to use hos pice. Other studies reported that among both community dwelling and nursing home re sidents, older women were less likely to receive intensive treatments or care such as hospitalization (Barker, Zimmer, Hall, Ruff, Freundlick, Eggert, & Rates, 1994; Fried & Mo r, 1997), dialysis, transplantation, timely diagnosis of lung cancer, and specific interven tions for heart disease adjusting for disease type and severity (American Medical Associ ation, 1991). However, a recent study by Bird and colleagues (2002) on health car e utilization and sp ending of Medicare beneficiaries in the last year of life demonstrates that ge nder differences in hospice care may be in part due to women’s greater longe vity and the tendency to treat younger adults

PAGE 36

26 more aggressively by physicians. It is al so possible that differential use of hospice between women and men may reflect treatment preference difference between genders. However, treatment preferences among men and wo men in this study are unknown. Two studies found education level to be associated with hospice use (Chen, Haley, Robinson, & Schonwetter, 2003; Grei ner, Perera, & Ahlu walia, 2003), but the studies reported the associati ons to be in opposite directions Chen and colleagues (2003) found lower education to be associated w ith hospice use after controlling for other demographic characteristics whereas Greine r and colleagues (2003) found that increased level of education increased the probability of receiving hospice at the end of life. Inconsistent findings from thes e two studies may be due to di fference in characteristics of study samples. The study by Chen and co lleagues (2003) was based on a convenience sample of older patients (i.e., aged 65 and ol der) with selected type s of cancer diagnoses while Greiner and colleagues (2003) examined a random, probability national sample of decedents aged 15 and older in 1993 us ing the 1993 National Mortality Followback Survey (NMFS). Although it may be hypothesize d that higher level of education allow the person greater access to information a bout hospice service, th e small number of studies examining this issue precludes a ny conclusion on the role of education. Older Whites have been consistently found to use hospice more than older adults from other race/ethnic groups, especially Blacks. This is true for older adults with differing diagnostic characteristic s at the end of life (Enguida nos et al., 2005; Greiner et al., 2003; Iwasyna et al., 2002; Lackan, Ostir, Freeman, Kuo, Zhang, & Goodwin, 2004; Lackan, Ostir, Freeman, Mahnken, & Goodwin, 2004; Virnig, McBean, Kind, & Dholakia, 2002; Virnig, Morgan, Persily, & DeVito, 1999; Welch, Teno, & Mor, 2005).

PAGE 37

27 Five studies using multivariate analyses (E nguidanos et al., 2005; Iwasyna et al., 2002; Lackan, Ostir, Freeman, Mahnken et al., 2004; Virnig et al., 1999; Virnig et al., 2002) report that Blacks are less likely to use hos pice. One multivariate study by Greiner and colleagues (2003) found that the role of race/ethnicity is attenuated by presence of a living will and another recent study using uni variate analysis by Welch and colleagues (2005) found no difference in hospice us e between Blacks and Whites. The complexity of the relationship betw een race/ethnicity and hospice use is illustrated by the work of Virnig and co lleagues (2002). Examining Medicare cancer patients who died in 1996, Virnig and collea gues (2002) report that Blacks are 25 percent less likely to use hospice compared with people from other raci al/ethnic groups after adjusting for age and gender. Similar findings were reported by studies adjusting for various sociodemographic and clinical charact eristics. Adjusting for age, gender, and income, Virnig and colleagues (1999) found that Black Medicare decedents in South Florida were less likely to use hospice. Virn ig and associates (1999) also found that the racial difference in hospice use was reduced among Medicare decedents in managed care programs compared to those in fee-for-ser vice programs. Iwashyna and colleagues (2002) found results similar to Virnig and co lleagues (1999) after ad justing for the age, gender, income and Medicaid insurance status of Medicare decedents with diagnosis of lung cancer, colorectal cancer, stroke, and heart attack. Lackan and colleagues (2004) found that Black Medicare decedents with breas t, colorectal, lung, or prostate cancer who died between 1991 and 1999 were less likely to use hospice in all years between 1991 and 1999 after adjusting for age, gender, and education. Enguidanos and colleagues

PAGE 38

28 (2005) reported that Blacks were less likely to use hospice after controlling for age, gender, marital status, nursing hom e stay, and cause of death. The findings by the study by Greiner a nd colleagues (2003) using the 1993 NMFS data provide additional insights into the ro le of race/ethnicity and other factors in predicting hospice use. Greiner and collea gues (2003) initially f ound that the rate of hospice use was significantly different betw een Black and White groups after controlling for a number of sociodemogra phic and clinical characterist ics. However, Greiner and associates also found that the negative re lationship between Black race/ethnicity and hospice use was reduced when the presence of a living will was controlled for. Examining only decedents who were aged 55 an d older, the investigators found that the role of Black race/ethnicity was no l onger significant when income, education, homeownership, and presence of a living will were included in the final model. Moreover, they found a significant interacti on between race/ethnicity and access to care difficulties on hospice use. Blacks were less likely to use hospice when they experienced difficulty accessing health car e while Whites and Hispanics showed increased hospice use when experiencing difficulty in accessing he alth care. It may be the case that Blacks are more likely to experience difficulty in ac cessing all health care in general including hospice, than for other racial/e thnic groups. It is also possi ble that the difficulties in accessing health care in general may prompt Bl acks to pursue aggressive care. Findings from this study suggests that while the eff ect of race/ethnicity is not independent of income or access to care variables, the significan t effect of the presen ce of a living will in explaining the racial/ethnic difference in hospice use may indicate cultural differences toward end-of-life decision making. The inve stigators hypothesized that lower hospice

PAGE 39

29 use by Blacks results from the same forces th at lead to the frequent Blacks’ aversion toward completing advance directives or treatments limiting choices. Enabling factors. Several studies have examined the role of individual, family, and community resource factors on hospice us e. Enabling factors examined in these studies include income, marital status, careg iver availability, so cial support, living arrangement, rural/urban location of the primary residence and health care provider availability. On individual and family res ource factors, however, much a smaller number of studies are available (Chen et al., 2003; Enguidanos et al., 2005; Greiner et al., 2003; Iwasyna et al., 2002; Virnig et al ., 1999). Of four studies examining the role of income in hospice use, one study (Virnig et al., 1999) us ed the average or median household income per zip code from the 1990 U.S. Census data as a proxy for income, another study (Iwasyna et al., 2002) used the same method as Virnig and colleagues (1999) in addition to Medicaid status as a sec ond proxy for the income status and the other two studies (Chen et al., 2003; Greiner et al., 2003) measured househol d income by interviewing the patient or the proxy for the decedent. In the multivariate analysis controlling fo r age, gender, race, primary diagnosis and comorbidity, Iwashyna and colleagues (2002) found little eff ect of income on hospice use when examining the median income alone. However, Iwashyna and associates found that Medicaid recipients were 0.69 time less likely to use hospice than those without Medicaid insurance, sugges ting that poor Medicare recipients (i.e., Medicare and Medicaid dual-e ligible beneficiaries) were less likely to use hospice. Virnig and colleagues (1999) found that the hi gher the income was, the more likely Medicare beneficiaries were to use hospice af ter controlling for age, gender and race.

PAGE 40

30 Greiner and colleagues (2003) also found havi ng a household income of more than $25,000 significantly increased the likelihood of using hospice in their multivariate analysis of all decedents in the 1993 NMFS sample. In contrast, Chen and colleagues (2003) did not find any significant effect of income on hospice use in their multivariate analysis. Given the small number of studies ex amining the role of income and variability in the method of measuring income across th ese studies, the role of income on hospice use is unclear. It may be that income is not a significant enabling factor for access to hospice care among older adults since all term inally ill Medicare and most Medicaid beneficiaries are eligible to receive it. Findings on the role of soci al support are also inconclu sive due to few studies available on this issue. Chen and collea gues (2003) found a positive association between the number of individuals living in the same household and hospice use but did not find marital status or caregiver relationship to the patient to be significant in predicting hospice use. Chen and colleagues (2003) also found that the final decision to use hospice was made largely by family members (42 percent of hospice patients) followed by patients themselves (28 percent) and physicia ns (27 percent). E nguidanos and colleagues (2005) did not find a significant role of marital status on predicting hospice use among Medicare and Medicaid dual-eligible older adults either. At the same time, Greiner and colleagues (2003) found that being never married significantly increase d the likelihood of receiving hospice while being widowed signi ficantly reduced the chance of receiving hospice. It is possible that such difference found by Greiner and associ ates (2003) may be due to age differences between these two groups in this study. It is also possible that a

PAGE 41

31 significant proportion of never married decedents were younger adults while older adults disproportionately represented widowed individuals who were also likely to have fewer social support resources and more likely to be in nursing homes and thus, they may have less access to hospice. In addition, Greine r and associates (2003) found the increased social support measured by frequency of phone calls was positively associated with likelihood of hospice use. An implication of th ese findings is that marital status may not be as significant as the av ailability of caregiver residing in the same household in predicting hospice use since most of hospi ce care is provided at home and informal caregivers are needed to provide substantial levels of care for hospice patients (Lorenz, Asch, Rosenfeld, Liu, & Ettner, 2004). The role of rural/urban location of resi dence has been studied and found to be significant in predicting hospice use. Although it is estimate d that 20 to 25 percent of Americans live in rural areas and over 18 perc ent of rural residents are over the age of 65, compared to 15 percent in metropolitan area s (MedPAC, 2001), the availability of hospice in rural areas is more limited than in urban areas. Virnig and colleagues found that a higher proportion of counties in urba n areas (69 percent) had a hospice agency physically located within them than did ru ral counties (34 percent) (Virnig, Moscovice, Kind, & Casey, 2002). Virnig and associates al so found that the rate of hospice use was lower among rural decedents. Among the Medicare beneficiaries who died in 1999, Virnig and colleagues (2004) reported that afte r controlling for age, sex, and race that the hospice use rate was 22.2 percent for those living in urban areas and 15.2 percent for those living in rural non-adjacen t areas. Although the percen tage of rural beneficiaries enrolled in hospice increased from 1992 to 2000, rural beneficiaries still used hospice at

PAGE 42

32 75 percent of the urban rate (MedPAC, 2002). T hus, it is expected that residents in rural areas have less access to hospice and consequently, are less likely to use hospice. Three studies examined the ro le of availability of certain types of physicians in predicting hospice use. Virnig and colleague s (2000) report that the number of hospitals per Health Service Area (HSA) is negativel y associated with hospice use while the number of generalists and number of doctors per capita increased hos pice use. The study by Iwashyna and colleagues (2002) however, did not find the effects of the number of hospitals and generalists in predicting hospice use at county level. The findings of this study (Iwashyna et al., 2002) s uggested that when indivi dual-level differences in diagnosis and wealth are taken into a consid eration, market variable s at the county level did not explain intercounty vari ability in hospice use rate wh ereas the study of Virnig and colleagues (2002) which only adjusted for age, sex, and race and examined market variables at the HSA level did find signif icant effects of market variables. A study examining differences in attit udes toward hospice between physicians with different specialties (Bradley, Cramer, Bogardus Kasl, Johnson-Hurzeler, & Horwitz, 2002) provides an insight into the role of physician speci alty. Bradley and colleagues (2002) found that physicians’ posi tive attitudes toward hospice care and increased communication about hos pice care increased referral for hospice. Bradley and colleagues (2002) also found differences in self knowledge of and attitude toward hospice by physicians’ specialty: oncologi sts with the least favorable attitude toward hospice and cardiologists with the leas t knowledge of hospice. Need/illness factors Studies have consistently shown that having a cancer diagnosis is one of the strongest pred ictors of hospice use (Iwashyna, Zhang, &

PAGE 43

33 Christakis, 2002). Recent studies examining disease trajectories of varying illnesses document the limited current knowledge base to understand trajecto ries of different illnesses and to identify the point where the prognosis slips below 6 months (Field & Cassel, 1997; Lunney et al., 2002). Studies report that physicians experience the difficulties inherent in accurate prognostication of life expectancy and are reluctant to identify patients as dying (Christaki s & Lamont, 2000; Mahoney, 2002 as cited in MedPAC, 2002). Among all major diagnosti c groups, however, cancer has a relatively predictable disease trajectory that is charac terized by a precipitous decline of function as the patient approaches death (Teno, Weitzen, Fennel, & Mor, 2001). Therefore, the sixmonth prognosis rule of the MHB is usua lly appropriate for patients with cancer diagnosis but it is much more difficult to furnish a e six month prognosis for patients dying of other illnesses. Psychosocial factors. Knowledge and accurate understanding of what hospice is and what type of services offered by hospice pr oviders is clearly an important factor in order for the utilization to take place. Howe ver, the lack of knowledge or misconceptions about hospice has been frequently cited as one of major barriers to hospice enrollment by commentators and reviewers on this issue, alt hough very few studies have examined this issue empirically. Of the few empirical studi es that explore this issue, a recent study about hospice decision making among patients and families (Casarett, Crowley, Stevenson, Xie, & Teno, 2005) found that only 40 percent of 237 patients and family members who were interviewed reported they knew anything about hospice and only 27 participants knew that the focus of ho spice was on comfort care and symptom management. Although this study was based on a convenience sample of patients and

PAGE 44

34 families who were recruited from hospice info rmation visits, the result suggests that the misconception and lack of knowledge about hospice may be common among terminally ill patients. The literature also suggests that forced choice between palliative and curative treatment required by the MHB makes it difficult for individual patients and families to choose hospice care. Case studies and review s on hospice utilizati on report that many patients and families as well as physicians find it difficult to accept poor prognosis and are willing to seek curative treatmen t despite poor prognosis (Schulman-Green, McCorkle, Cherlin, & Bradley, 2005). A ccording to a Gallup poll in 1996, most Americans are willing to seek curative care if faced with a terminal illness although a majority expresses an interest in hospice ca re (GAO, 2000). Preference to seek curative treatment even in terminal phases of an illness may be even stronger among minority groups. Due to differences in cultural be liefs and practice rega rding death and dying, members of minority groups are less likely to choose hospice care; differences in religion, socialization and educat ion and disparities in access to healthcare in general also play a role (Crawley et al., 2000). Consistent findings that Blacks pr efer more aggressive care and treatment at the end of life compar ed with Whites provides support for this hypothesis (Kwak & Haley, 2005). Hospice Service Provider Related Factors A recent study examined the role of hospice organization admission practices (Lorenz et al., 2004). Based on a statewide surv ey of hospice organizations in California, Lorenz and colleagues (2004) found that a si gnificant minority of hospices restricted admission on criteria such as lack of caregiver (26 percent), and unwillingeess to forgo

PAGE 45

35 hospitalization (29 percent); in addition many hospices were unwilling to accept patients with complex medical care needs such as to tal parenteral nutriti on (TPN) (38 percent), tube feeding (3 percent), transfusions ( 25 percent), radiotherapy (36 percent), and chemotherapy (48 percent). Lorenz and associ ates also found that larger hospices were less likely to restrict admissi on based on these criteria, excep t for TPN or tube feeding, suggesting that eligible benefi ciaries living in areas served by smaller hospices may be likely to experience provider-related barriers to hospice. As one of the goals of hospice care is to promote home-based care, and hos pice organizations stress the need for a family member or someone else to serve as a primary caregiver for the patient who wants to receive home-based hospice care, some hospices do not accept patients without a designated caregiver (NHO, 1996b). However, it a ppears that the use of such restrictions imposed by hospice organizations may vary ac ross states. The state of Florida for instance, has a higher proportion of hospi ce organizations with freestanding hospice inpatient or residential beds (12 percent) than the national average (8 percent) (Miller & Lima, 2004). The implication of the higher level of availability of inpatient or residential hospice beds in states such as Florida is th at patients who are unable to stay in their homes in the community due to lack of careg iver may experience fewer barriers to access to hospice because they may be able to rece ive hospice care in these hospice residential beds. For nursing home resident, however, l ack of available careg iver may not function as a barrier. In a nursing home facility, hospice and other long-term care are coordinated between hospice and nursing home staff if a nursi ng home resident elec ts a hospice care. Medicare policy and organizational practice of nursing homes in referring nursing home patients to hospice

PAGE 46

36 Nursing home residents are one of the mo st rapidly growing segments of hospice beneficiaries, with representation going from 11 percent of hospice enrollees in 1992 to 36 percent in 2000 (Hogan, 2002; MedPAC, 2002) Nevertheless, the rate of hospice enrollment among nursing home resi dents is reported to be lowe r than the enrollment rate among community-dwelling residents (Hoffm ann & Tarzian, 2005). Some studies and commentaries suggest several major barriers to hospice in nursing home exist (Hoffmann & Tarzian, 2005; Mezey et al., 2002; Miller et al., 2003; Zerzan, Stearn, & Hanson, 2000). Jones and colleagues (1997) report th at differential use of hospice in nursing homes is influenced by the nur sing home administrator’s at titude toward and knowledge of hospice care. Another poten tial barrier is that the majo r goal of care in nursing homes is to promote functional maximization and re habilitation while hospice care focuses on palliative care, not curative care (Hoffmann & Tarzian, 2005; Mezey et al., 2002; Zerzan, Stearn, & Hanson, 2000). The potential conf licts of care philosophy between nursing home and hospice care may prevent enrollment of nursing home residents in hospice. The six-month prognosis rule of MHB and Medicare nursing home reimbursement rates are reported to be othe r potential barriers to hospice enrollment among nursing home residents. The majo rity of nursing home residents can be characterized as having chronic illnesses related to organ syst em failure or frailty (Mezey et al., 2001), types of illnesses in which the dying traject ory is difficult to be determined (Lunney et al., 2003). Thus, it is more diffi cult to identify nursing home residents as dying by clinical measures at the end of life. In addition, there is often a financial incentive to choose skilled nursing care over hospice for both nursing facilities and Medicare beneficiaries. This is especially the case for Medi care beneficiaries with a need

PAGE 47

37 for nursing home placement following hospita lization, since Medicare does not cover skilled nursing care and hospice care simultaneously. For Medicare beneficiaries without Medicaid or othe r supplemental insurance coverage for nursing home care, there is a financial disincentive for them to choose hospice and to forgo the skilled nursing care be nefit. A potential fi nancial disincentive also exists for the facility with respect to Medicare and Medicaid dual-eligible nursing home residents. When a dual-eligible nursing facility resident chooses hospice, Medicaid allocates funds to the hospice, which is responsible for paying the nursing home 95 percent of the Medicaid nursing facility room and board rate. It has been suggested that this “pass thru” practice not only creates conf usion for the nursing facility regarding the source of payment for the room and board, but also creates financial disadvantages (Miller & Mor, 2002). Also, the Medicare nursing home per diem payment rate for skilled nursing care is higher w ith an average per diem rate of $265 than for Medicaid nursing home care ($118) (Hoffm ann & Tarzian, 2005). Thus, even for Medicare and Medicaid dual-eligible benefici aries, nursing home facilities have financial incentives to enroll their residents under the skilled nur sing benefit rather th an hospice care since reimbursement rate under Medicare skill nursing benefit is higher than hospice room and board rate under Medicaid. The presence of financial disincentives doe s seem to have an impact. Two recent studies suggest that there is at least a delay in access to hospice care if not implicit denial of access to hospice care due to the nursing home care reimbursement rate policy. One of these studies examined nursing home resident s and hospice use in five states. Miller, Gozalo and Mor (2000) found that about 26 pe rcent of nursing home residents who

PAGE 48

38 received hospice were admitted to hospice within 1 day after being discharged from a skilled nursing facility using the Medicare skille d nursing benefit. It is possible that due to the fact that skilled nu rsing benefit is covered under Medicare, albeit for a limited period of time, some older adults (without Medicaid) with nursi ng home care need may opt to choose nursing home benefit over hos pice care and other older adults who are dually eligible for Medicare and Medicaid, nursing home facility may find a financial incentive to enroll residents under Medica re skilled nursing benefit instead of under Medicaid hospice room and board benefit. The second study (Moon & Boccuti, 2002) found that hospice use was higher among Medicare decedents who entered a nursing facility during th e calendar year of death (19 percent) than decedents who reside d in a nursing home from the start of the calendar year of death (11 per cent). Combined, these findings indicate a need for further assessment of the process of hospice enro llment among both short-term and long-term nursing home residents. That is because a significant minority of older adults are admitted into a nursing home for a short duration of time using the Medicare skilled nursing benefit during the last fe w months of life (Miller, In trator, Gozalo et al., 2004; Miller, Intrator, Laiberte, & Cang, 2004). Su ch transition from community to nursing facilities for example may occur when an acu tely-ill older adult is discharged to a nurscing facility following hospitalization (Mukamel, Bajorska, & Temkin-Greener, 2002). These short-term nursing home resi dents may experience access barriers to hospice during the last few months. Factors Associated with Hospice Length of stay

PAGE 49

39 Timely access to hospice would provide better opportunities for patients, families and hospice providers to devel op and coordinate an appropria te plan of care. Short length of stay among hospice enrollees has b ecome a major concern for hospice providers and policy makers. The analysis by the G overnment Accountability Office (2000) on the length of enrollment under hospice care shows that the average lengt h of hospice use per hospice enrollee declin ed from 74 days in 1992 to 59 days in 1998. Although patients become eligible during 6 months before expe cted time of death, half of hospice users now receive care for 19 or fewer days, a nd care for one week or less is common (NHPCO, 2004). The literature on the factors associated with shorter length of stay include diagnoses such as renal failure, leukemia or lymphoma, and liver or biliary cancer (Christakis & Escarce, 1996), male gender (Miller et al., 2000; Somova, Somov, Lawrence, & Frantz, 2000; Christakis & Iwashyna, 2000), White race (Christakis & Escarce, 1996; Christakis & Iwashyna, 2000), pr ivate insurance status (Somova et al., 2000), enrollment in Medicare’s fee-for-service program (Virnig et al., 2000), living with a caregiver (Somova et al., 2000), having a hospital referral sources (Somova et al., 2000), and states and health care market ch aracteristics (Christa kis & Iwashyna, 2000; Virnig et al., 2000). Of pa rticular interest among these factors is race/ethnicity. Although two previous studies (Christakis & Escarce, 1996; Christakis & Iwashyna, 2000) found Whites to be more likely to use hospice, among hospice users non-Whites are likely to have a longer st ay than Whites. However, these studies do provide clear reasons for differences in the rate of hospi ce use and length of stay between these White and non-White racial/ethnic groups.

PAGE 50

40 Factors Associated with Place of Death The place of death has been considered as one of important end-of-life care outcomes as increasing number of studies have documented the gaps between the preferred places of death and act ual deaths of place. Most of studies that examined the issue of site of death vary greatly in samp le characteristics – cancer patients (Bruera et al., 2003), individuals en rolled in specialized programs such as hospice or the PACE programs (Fried, Drickamer, & Tinetti, 1999; Tempkin-Greener & Mulamel, 2002), seriously ill patients who were hospitalized (Pritchard et al., 1998), skilled nursing facility residents (Levy, Fish, & Kramer, 2004), Medicare beneficiaries of selected states (Emanuel et al., 2002), dual-eligible older adu lts from a western st ate (Enguidanos et al., 2005) and U.S. decedent population (Iwashyna & Chang, 2002; Mitchell, Teno, Miller, & Mor, 2005; Weitzen et al., 2003). Despite tremendous variations in sampli ng and methods of these studies, two factors are consistently found to influence lo cation of death: race/e thnicity and hospice use. Studies report Black race /ethnicity to be associated with higher odds of in-hospital death (Iwashyna & Chang, 2002; Pritchard et al ., 1998; Weitzen et al., 2003). In addition two studies, although based on convenience samp les of community dwelling residents, found that Blacks were more likely to prefer to die in a hospital than their White counterparts (Kalish & Reynolds, 1976; Neubauer & Hamilton, 1990). Another factor in place of death is hospi ce use. Hospice use has been associated with greater likelihood of ho me deaths and lower likelihood of in-hospital deaths (Emanuel et al., 2002; Enguidanos et al., 2005; Moinpour & Polissar, 1989; Pritchard et al., 1998). Some studies have suggested that the higher rate of in-hospital deaths among

PAGE 51

41 racial/ethnic groups may be due to the lack of resources that may help them receive endof-life care at home since minority groups are le ss likely to use hospice. However, it is not clear how hospice use and race/ethnicity may affect place of death among nursing home residents, particularly among dualeligible nursing home residents. Limitations of Previous Research The present review of the li terature on health services use by ethnically/racially diverse older adults at the e nd of life, specifically hospice us e, suggests areas of further research and areas of methodological improvement that this dissertation research seeks to address. Areas of further research include : (a) hospice utilizati on among Medicare and Medicaid dual-eligible older adults; (b) with inand between-group differences in hospice utilization among racial/ethnic groups; (c ) hospice utilization among nursing home residents; and (d) the effects of hospice use on place of death between and within racial/ethnic groups of nursing home residents. Dual-eligible Medicare and Medicaid Beneficiaries Little is known about hospice use within special populations such as older adults eligible for both Medicare and Medicaid. A lthough it has been shown that dual-eligible older adults are less likely to use hospice th an Medicare-only bene ficiaries (Iwashyna et al., 2002; Moon & Boccuti, 2002), reasons for th e lower utilization by dual-eligibles are unclear. Studies that found differential rate s of hospice use between these two groups (i.e., dual-eligible and non-dual-eligible Medicare beneficiaries) suggest that the difference may be due to income differe nces. Nevertheless, given the unique characteristics of dual-eligibles, the role of predictors of hospice us e previously identified

PAGE 52

42 among general Medicare population may be diff erent within the dual-eligible population from those of the non-dual-eligib le older adult population. Dual-eligibles are considered a vulnerab le subgroup of Medicare beneficiaries as they tend to be poor, are more likely to be members of non-White minority groups, and report lower health status than other bene ficiaries (MedPAC, 2004). A report on dualeligibles provided by Medicare Payment A dvisory Commission (2004) shows a detailed profile of dual-eligibles on multiple characteri stics. According to the analysis of dualeligibles based on the 2001 Medicare Current Beneficiary Survey by MedPAC (2004), 60 percent of them live below the poverty leve l and 94 percent live below 200 percent of poverty. Blacks and individuals with less than a high school level of education make up more than 20 percent and 60 percent of the dual-eligible population respectively while of Medicare beneficiaries wit hout Medicaid only seven perc ent are Blacks and 28 percent have less than high school education. Sixty two percent of the dua l-eligible group are female while only 55 percent of the non-dual eligible Medicare group are female. Dual-eligibles are also mo re likely to have poor health compared to non-dual eligibles and to suffer from cognitive impairme nt, mental disorders, diabetes, pulmonary disease, stroke, and Alzheimer’s disease than are non-dual eligibles (MedPAC, 2004). Moreover, between 1993 and 1998, dual-eligible s comprised 21 percent of all Medicare decedents while they made up for 13 percent of survivors during these years (Hogan et al., 2000) indicating higher morta lity rate among dualeligible beneficiaries. Furthermore, dual-eligibles are also more like ly to receive care in long-te rm care facilities than other Medicare beneficiaries with almo st one-quarter of dual-eligible s residing in an institution, compared with three percent of non-dual eligibles (MedPAC, 2004).

PAGE 53

43 Racial/Ethnic Differences between Blacks and Whites Although a few recent studies examine the ro le of race/ethnicity in hospice use, all empirical studies examining hospice use reviewed examined between-group differences only. Blacks have b een consistently found to use hospice less frequently than Whites as discussed previously. The end-of -life literature suggests that cultural and institutional barriers may be reasons for underu tilization of hospice by Blacks. Suggested institutional barriers include a lack of knowledge, trust of hospice and the health care system, and the lack of a full time primary caregiver. Cultural barriers include Black spiritual and religious values, and cultural va lues that are often in conflict with hospice philosophy and medical care. There are at least five areas in which in stitutional and cultural barriers have been explored. The first is that Blacks are suggested to be le ss knowledgeable about hospice services than Whites. While a lack of know ledge concerning hospice services has been suggested as a barrier to hospice access and use among the general population (Reese et al., 1999), it is also suggested that Blacks may be even less likely than the general population to know of such services. Rees e and colleagues found in focus groups that most community leaders such as pastor were not familiar with hospice and most Blacks who are religious often seek a dvices regarding health care fr om their ministers (Reese et al., 1999). Moreover, Blacks are likely to have a lower socioeconomic status: Blacks in general have lower levels of education and income compared to Whites (MedPAC, 2004). Socioeconomic status (SES), typically measured in income and education, affects access to health services because knowledge an d financial resources affect the ability to

PAGE 54

44 obtain health information about innovative services such as ho spice. Thus, it is possible that the lower SES among Blacks contributes to some of the differences in hospice use between Blacks and Whites. In fact, a multivariate analysis of the national sample of adult population who were aged 55 and older and died in 1993 by Greiner and colleagues (2003) supports this. Greine r and associates (2003) found that the negative relationship between Black race/ethnicity and hospice us e was attenuated and became statistically insignificant when income was controlled for. Second, mistrust among Blacks toward health care system is another potential barrier to hospice. It can be hypothesized th at Blacks may perceive hospice service as a care option that will prevent them from receivi ng adequate care. Some speculate that the historical experience of receiving inadequate and disrespectful treatment by the medical and health care community has lead Black gr oup to distrust the overall medical system and serve as a barrier to hospice (Born, Greiner, Sylvia Butler, & Ahluwalia, 2004; Burrs, 1995; Gorden, 1995; Neubauer & Hamilton, 1990). The infamous Tuskegee Syphilis Study is often cited as a major influence on Black opinions concerning medical care. Distrust toward medical system has been found to be associated with Blacks’ attitude toward other end-oflife care options such as life support and advance directives (see the review by Kwak & Hale y, 2005 on this issue). Third, Blacks may seek or obtain access to ca re at later stages than do Whites, and therefore may present with illnesses at adva nced stages during encounters with health care providers. Among patients wi th cancer or stroke, Blacks are more likely to be diagnosed at later stages of the disease than Whites (American Cancer Society, 2003; Manderlblatt, Andrews, Kao, Wallace, & Kern er, 1996). It also may be that in the

PAGE 55

45 context of having serious illnesses such as a cancer diagnosis, the later entry point into the health care system for Blacks results in the patient and family having less time to contemplate the significance of the diagnosis Difficulties in accessing care can also have a ripple effect: Greiner and colleagues (2003) report that the more health care access problems Blacks experienced, the less they we re likely to use hospice. As a result, Blacks may be likely to be more focused on pur suing all possible curative treatment with a goal of recovery rather than pursuing palliative care. Lastly, traditional Black spiritual and re ligious and cultural va lues are often in disagreement with hospice philosophy of giving up curative treatment (Burrs, 1995; Gorden, 1995; Reese et al., 1999). The trad itional hospice philosophy is often in disagreement with their prefer ences for life-sustaining treatm ent and against advance care planning such as completing advance directiv es (Blackhall et al ., 1995, 1999; Caralis, 1993; Klessig, 1992; Waters, 2001). Such pers pectives on end-of-lif e care options are often related to spiritual/relig ious values (Burrs, 1995; Reese et al., 1999). Traditionally, Blacks believe in God’s omnipotence and mir acles especially regarding medical care decisions and Blacks rely on thei r religious faith for recovery or cure at the end-of-life rather than accept terminality (Reese et al., 1999). Through their quantitative study, Reese and colleague (1999) found that Blacks were less likely to agree with hospice philosophy controlling for gender, age, education and income. Overall, the literature suggests that Bl acks are less likely to use hospice due to different sociodemographic and their unique hi storical and cultural experiences in the medical system. Such results should be cons idered in the context of the diversity of Blacks and there is a need for further examin ation of individual variation within each

PAGE 56

46 racial/ethnic group in addition to between-di fferences to better understand the mechanism by which race/ethnicity affects hospice use. Another related issue regardi ng the role of race /ethnicity in hospice use is length of stay. Two previous studies (Christa kis & Escarce, 1996; Christakis & Iwashyna, 2000) found that non-Whites are likely to enter hospice sooner, measured by survival time from the time of initial hospice enrollment. The reason for such finding is not clear. Comparing the predictors of use of and lengt h of stay under hospice across racial/ethnic group may help identify factors that may facilitate hospice use as well as timely access to hospice service. Hospice Utilization in Nursing Homes Better understanding of endof-life experiences particularly hospice utilization and its outcomes, among nursing residents is ne eded. The current literature on hospice utilization among nursing home residents is limited by a small number of studies that have comprehensively examined two differe nt types of nursing home residents: shortterm and long-term nursing home residents. Although a study of dual-eligibles in California by Enguidanos and colleagues (2005) included a variable indicating long-term nursing home admission in their analysis, thei r conceptualization of long-term nursing home admission is limited in capturing a subgro up of nursing home residents, short-term nursing home residents. Engui danos and colleagues (2005) defined individuals in a custodial nursing facili ty as having a minimum of 90 consecu tive days in their last year of life. However, a significant minority of ol der adults enter into a nursing home for a short duration of time using the Medicare skilled nursing benef it (Miller, Intrator, Laliberte et al., 2004). Miller and colleagues (Miller, Intrat or, Laliberte et al., 2004)

PAGE 57

47 found that 31 percent of all nursing home resi dents who died in Flor ida were enrolled in the Medicare skilled nursing benefit duri ng the last 60 days before death. The literature also suggests that ther e may be unique differences among shortterm and long-term nursing home residences at the end of life. Among nursing home residents, short-term nursing home resident s are more likely to be younger and have a cancer diagnosis (Miller, Intrator, Gozalo et al ., 2004). Similar to hospice use, Blacks are in general less likely to use nursing homes afte r adjusting for other confounders (Salive, Collins, Foley, & George, 1993; Stevens, Owen, Roth, Clay, Bartolucci, & Haley, 2004). Therefore, it is important to account for the unique differences between subpopulations of nursing home residents when examining hospice use at the end of life. Moreover, few studies comprehensively examine how hospice use affects outcomes within and between racial/ethnic groups residi ng in nursing homes. The Effects of Hospice Use on Place of Death One of the key issues in end-of-life ca re is where people die because place of death shows where the person was receiving car e at the very end of life and suggests areas of improvement for that particular setting of care (Flory, Young-Xu, Gurol, Levinsky, Ash, & Emanual, 2004). In addition, there is a gap between preferred place of death and actual place of death. A national surv ey shows that the majority (65 percent) of older adults aged 65 and older prefer to die at home or in a hospice (Gallup, 2000) while over 50 percent of older adults actually died in acute care hospitals (Hogan et al., 2000). Although hospitalization at the end of life can be both nece ssary and unnecessary depending on the clinical needs of the patien t, the experts on end-of-life care generally agree that hospitalization at the end of life should be avoi ded if possible (Engle, 1998;

PAGE 58

48 Mezey et al., 2002; Mor, Papandonatos & Mi ller, 2005). Early studies on home and inpatient-based models of hospice care have found that hospice reduces hospitalization among non-nursing home residents and fac ilitates home death among communitydwelling terminally ill pateints (Kidder, 1992; Mor & Kidder, 1985). Among nursing home residents, a recent study shows that hos pice reduces hospitaliz ation during the last month of life (Miller et al., 2001). However, more research is needed to replicate these study findings among dual-eligible older adults in a nursing home setting. Of particular interest is to explore patt erns of hospital death between Blacks and non-Hispanic Whites and how they are affected by hospice use. A recent study of nursing home reside nts in two states, New York and Mississippi, by Mor, Papandonatos and Miller (2005) holds importa tion implications for a study of site of death among nursing home re sidents. Mor and co lleagues (2005) found that Blacks were more likely to be hospitaliz ed during the last 90 days of life than Whites after adjusting for demographi c and clinical characteristics, patient preference (using donot-resuscitate order) and facility resour ces. Mor and colleagues (2005) also found interactions between race and age and race and functional impairment on hospitalization rates. They found that the ol dest and most impaired Blac ks experienced significantly higher rates of hospitalization than nonHispanic White counterparts. Mor and colleagues also speculated that these findings mi ght be attributed to racial segregation in nursing homes and/or cultural e ffects. Their study found that Black residents resided in nursing facilities with fewer resources and th erefore, these nursing facilities might be predisposed to discharge to hospital those patien ts with high levels of clinical needs. Mor

PAGE 59

49 and associates (2005) also recognized that higher rate of hospitalization by Black residents might be due to their preference for ag gressive treatments at the end of life. For nursing home residents, particular ly long-term nursing home residents, nursing homes can be considered as an appr opriate place of deat h. With increased functional impairment and physical deterior ation along with diminished capacity of informal caregivers to provide adequate care in a community setting, many frail older adults enter nursing homes to receive long-term care that ot herwise may not be available in a community setting. For these nursing home residents, some of the important goals of end-of-life care are to minimize unnecessary transfers to hosp itals while providing adequate medical care that is appropriate for terminally ill patients and to reduce hospital death. In fact, a recent randomized controlled study of a hospice educational and referral interventions for nursing home residents and their families (Casarett, Kwalawish, Morales, Crowley, Mirsch, & Asch, 2005) s hows that a simple communication studies significantly improved referral of eligible nursing home reside nts to hospice with better outcomes such as fewer acute care admissions and higher ratings on quality of care. Thus, providing hospice care for nursing home residents appear to make significant improvement in quality of care and facilitate nursing home residents experience dying in place at the end of life. Measurement issues The operationalization of key concepts such as race and ethnic ity is critical in understanding disparities and vari ations in access to and qual ity of health care services (Arispe, Holmes, & Moy, 2005). Race classifi cation and selection of variables influence almost all aspects of the re search including methodological approach, analysis, and the

PAGE 60

50 presentation and framing of results (Arispe et al., 2005). Yet, it is generally recognized that there is tremendous variations in the ava ilability, reliability and va lidity of racial and ethnic-related variables across data systems (Arispe et al., 2005). Chen, Chen and Mehra (2005) for example found that while the id entification of Whites and Blacks in a statewide Medicaid data set was over 90 pe rcent accurate, accuracy was substantially lower for other groups. Limitations in using administrative record or vital statistics records should be noted because many studies examining the endof-life health servic es utilization (e.g., Enguidanos et al., 2005) use such records as their data sources. Generally, administrative data contain enrollment files wi th information on eligibility of the person for a health plan and demographic characterist ics and encounter or cl aims data including information on individual services or sets of services (Iezzoni, 2002). Since administrative data are usually generated from billing records or discharge data provided by health service providers who are not re quired to collect information on race and ethnicity, rates of missing data are relative ly higher than what is found in populationbased surveys. Quality of data is also of ten in question because designation of race and ethnicity may reflect patient, proxy or provider report. Moreover, large data sources such as Medicare or Medicaid do not provide readily available in formation regarding enrollees other than age, race, and gender and certain proxies for income and ethnicity (Fisher et al., 1992; Iwashyna, Brennan, Zhang, & Christakis 2002). While the literature suggests the importance of the characteristics of family members in the household of the patient in medical decision making, Medicare claims da ta, for example, do not provide reliable information concerning such variables as mar ital status although there have been some

PAGE 61

51 efforts to improve the accuracy of sociodemographic informa tion in the claims data set and to identify social information such as marital status (Iw ashyna et al., 2002). Furthermore, there are limitations of using diagnostic and procedure codes recorded in the administrative data. Diffe ring systems of diagnos tic categorization (e.g., International Classification of Diseases ICD-9-CM vs. ICD-10 or DSM-IV-R) and procedure codes (Current Procedure Termi nology, CPT-4) and inconsistent accuracy and completeness of coding of administrativ e data are well documented (Iezzoni, 2002; Iezzoni, Daley, Heeren, Foley, Fisher, D uncan et al., 1994; McCarthy et al., 2000). Different diagnosis and procedure codes may describe the same event, and diagnostic codes do not allow researchers to determine the severity of illness of the patient (Iezzoni, 2002; McGlynn, Damberg, Kerr, & Brook, 1998). Mo reover, substantia l variability is found in the accuracy of dia gnostic coding across hospitals and physicians and up-coding practices (i.e., using codes to ensure the highe st payment and not the condition that is the most clinically salient) by some hospitals has been found to occur in an effort to maximize reimbursement through Medicare’s Prospective Payment System (PPS) (Iezzoni, 1997; Iezzoni, 2002; McGl ynn et al., 1998). Yet another problem is the validity of coding of cause of death on death certifi cates, which has also been questioned (Hogan et al., 2000). There is some evidence, however that the accuracy of cause of death codes has improved in recent years (Hoyert and Rosenberg, 1999). Summary This dissertation research seeks to fill th e gap in the end-of-life literature in the areas addressed above. Specifical ly, this dissertation research investigates the role of multiple determinants of hospice use among dua l-eligible older adults in nursing homes

PAGE 62

52 to provide implications for practice and po licy to improve access to hospice among this unique population of older adults by identifyi ng potential access barriers to hospice that may be unique to dual-eligible nursing home re sidents. Another major aim of this study is to explore individual variation within each racial/ethnic group in addition to betweengroup differences to better understand the mechanism by which race/ethnicity affects hospice use and length of stay in hospice. Fi nally, the place of deat h as an outcome of hospice use is examined between and with in each racial/eth nic group, non-Hispanic Whites and non-Hispanic Blacks. In an effort to contend with measurement issues, this study uses demographic and clinical data of d ecedents from multiple data sources such as Florida Medicaid eligibility and claims data se ts and Florida Death Ce rtificate record set to improve accuracy of race/ethnicity and cause of death variables. The decisions to focus on dual-eligible in dividuals and resident s of nursing homes were made for several important reasons. Fi rst, dual-eligibles are a group that is of special importance to policy makers, due to th eir very high utiliza tion of public funds (both Medicaid and Medicare). Second, using dual-eligibles is adva ntageous in studying racial/ethnic differences, because this group incl udes a restricted range of financial assets, minimizing some of the problems typically en countered in such comparisons because of the large disparities in income in the gene ral population of older NHW and NHB. Since nursing homes are an increasing site of deat h, and use of hospice use in nursing homes has not been extensively studie d, research focused in this se tting can also provide useful policy information. Finally, because of the complex nature of Medicare and Medicaid funding for hospice care for dual eligibles, and limitations of the available administrative data sets, we have a much better accuracy of ascertaining whether hospice care was

PAGE 63

53 utilized among a nursing home population than in a community-dwelling group. For all of these reasons, the project focuses on dualeligible individuals who were nursing home residents. Research Questions The Conceptual Model Within the framework of the behavioral model of health service utilization (Andersen, 1995), the relationships between race/ethnicity, nursi ng home use, hospice, and hospital death are examined systematically in this study. The behavioral model of health service utiliza tion is modified to include pred isposing, enabling and need factors predicting hospice service use which is expect ed to reduce the risk of hospital death as identified in the literature (Figure 1). Thus the final outcome in this model is hospital death. For hospice, three sets of individual determinants – predisposing, enabling and need factors will predict the use of this service. Among predisposing factors, older age, female gender, and Black race/ethnicity have been consistently f ound to predict hospice use while the role of educa tion is not clear. As enab ling factors, marital status, rural/urban location of the primary residen ce, and short-term nursing home status have been found to predict higher likelihood of hospice use while a cancer diagnosis is found to be the strongest predictor of hospice use among need factors. The literature on place of death suggests that Black race, presence of the principal diagnosis other than cancer, a nd hospice use are associated w ith hospital deat h (Miller et al., 2001; Mor et al, 2005). In this study, the us e of hospice, in turn, is expected to reduce

PAGE 64

54 Figure 1. Conceptual Model Independent Variables Intermediate Dependent Variables Dependent Variable Predisposing Factors Age Gender Race/Ethnicity Education Enabling Factors Marital Status Urban/Rural Location of the Primary Residence Nursing Home Status Need Factors Cause of Death: Cancer Use of Hospice Place of Death: Hospital A B C D F E G 54

PAGE 65

55 the risk of hospital death. In predicting hospital death, some of predisposing, enabling and need factors predicting hospice are expected to directly influe nce hospital death and hospice use is expected to reduce the risk of hospital death after controlling for these covariates. In this study, not only betweenbut also within-group differences across racial/ethnic groups are examined. In addition, factors associated with the length of stay in hospice will be studied among individuals who use hospice. This is an exploratory component of the dissertation study; the expl oratory component has been added because there is no substantial theoretical base in th e literature to explai n the timing of hospice use, and thus, the same predictors identified for hospice use will be assessed in predicting the length of stay in hospice. Research Questions and Hypotheses Research question 1. Do predisposing, need, and enab ling factors have significant associations with hospice use in the last y ear of life among dual-eligible nursing home residents? Hypothesis 1: Four predisposing factors – age, race, education, and gender are significantly associated with hospice use (Figure 1, A). Older age, non-Hispanic White race, higher educational level, and female gender are significantly associated with increased likelihood of using hospice. Race/ethnicity is significantly associat ed with hospice use after controlling for other variables: non-Hispanic Blacks are less likely to use hospice compared with non-Hispanic Whites.

PAGE 66

56 Hypothesis 2. Enabling factors –martial status geographic location of the primary residence (i.e., rural or urban area code), and short-term nursing home status significantly predict hospi ce use (Figure 1, B). Being married, urban residence, and long-term nursing home status are significantly associated with increased hospice use. Hypothesis 3: An illness-need factor, diagnosis of cancer directly influences hospice use (Figure 1, C). Having a diagnosis of cancer of any type is significantly and positively associated with hospice use. Hypothesis 4: Race/ethnicity moderates the im pacts of education, marital status, and cancer diagnoses on hospi ce use. This is an exploratory component of the analysis of predic tors of hospice use. The moderating effect of race/ethnicity on other enab ling factors with hospice use has not been explored extensively by previous research on hospice utilization. However, the literature on racial/eth nic differences in end-of-life decision making and service utilization do cuments a strong influence of race/ethnicity. Thus, it is possible th at race/ethnicity and its associated cultural norms and expectations may infl uence the strength or direction of effects of other enabling and need factors known to facilitate use of hospice. In this sub-analysis, pot ential interaction effects between race/ethnicity and education, marital st atus and cancer as a cause of death with hospice use are explored.

PAGE 67

57 Hypothesis 5: The same predictors identified by Hypotheses 1-3 with the exception of the direct effect of race/ethnicity will affect hospice use within each of the racial/ethnic groups. The same predictors include predisposing (older age, higher educational level, female gender), enabling (being married, urban residence, long-term nursing home status), and need (cancer as a cause of death) characteristics will be associated with higher likelihood of hospice use within each racial/eth nic group. Little is known about within racial/ethni c group difference on predictors of hospice use, but there is no evidence that the elements of the Ande rsen Model should function differently across the two groups being studied. Research question 2. Do major predisposing, need, and enabling factors have significant associations with length of st ay in hospice in the last ye ar of life among dual-eligible nursing home residents? Hypothesis1. The same predictors identified from the Research Question 1 are tested in their roles in predicting longer leng th of stay. Specifical ly, predisposing (older age, higher educational level, female ge nder, non-Hispanic White), enabling (being married, urban residence, long-term nursing home status), and need ( cancer as a cause of death) characteristics will be associated with longer survival time from the time of initial hospice enrollment for the overa ll nursing home residents enrolle d in a hospice program. Due to limited theoretical understanding on th e effects of multiple factors on predicting the short length of stay in hospice, the same predisposing, enabling and need factors predicting hospice use will be assessed in this analysis. Among predisposing, enabling and need factors, characteri stics associated w ith longer survival time identified in previous research are hypothesized to be a ssociated with longer survival time.

PAGE 68

58 Hypothesis 2. The same predisposing, enabling and n eed predictors for longer survival time from the Hypothesis 2 of the Research Question 2 are tested within each of the racial/ethnic groups. There is no clear evidence in the literature that indicates the factors influencing length of stay will be different w ithin each racial/ethnic group. Thus, relative importance of these factors will be explored within each racial/ethnic group, with the guiding hypothesis being that factors influencin g length of stay will be the same across both groups. Research question 3. Does hospice use reduce hospital death after controlling for covariates? Hypothesis 1. Among predisposing, enabling and n eed factors, female gender, non-Hispanic White race/ethnicity, presence of cancer, and hospice us e are significantly associated with lower risk of hospital death (Figure 1, D G). Hospice use is significantly associated with lower risk of hospital death after controlling for covariates. In addition, the potential moderating effect of race/ethnicity on age with hospital death is tested as a previous study of hos pitalization rates among nursing home residents (Mor et al., 2005) found a moderating effect of race/ethnicity on age with hospitalization rates and hospitalization pr ecedes in-hospital death. Hypothesis 2. Among predisposing, enabling and need factors, female gender, presence of cancer, and hospice use are signif icantly associated with non-hospital death within each of the racial/ethnic groups. The effect of hospice use on lower risk of hospital death after controlling for other covariates is not significantly different for nonHispanic Blacks compared with non-Hispanic Whites. There is no evidence found in the

PAGE 69

59 literature to suggest that th ere will be differential effect s of hospice use across race on hospital death.

PAGE 70

60 Chapter Two: Research Methods Sample This doctoral research consists of a series of analyses of secondary data utilizing a Florida Medicaid Long-Term Care analysis file which was built from Florida Medicaid eligibility and claims records and Florida Depa rtment of Health Deat h Certificates stored and managed at the State Data Center on Agi ng at the Florida Policy Exchange Center on Aging, School of Aging Studies, University of South Florida. In addition, an extended hospice claims file from Medicaid Program Analysis data file was provided by the Agency for Health Care Administration. T hose desiring further information on hospice claims should contact the Agency for H ealth Care Administration for additional information on the nature and type of hospi ce services that were utilized. The sampling frame consists of dual-elig ible nursing home residents aged 65 and older who died between three state fiscal years (SFY) 2000-2001, 2001-2002, and 20022003 in Florida. To be included in the final sample of this study of decedents, individuals needed to be dually eligible for nine consecuti ve months during the la st 12 months of life and be either non-Hispanic Blacks or non-Hispan ic Whites. Individuals with traumatic or sudden death were excluded from the analys is since these individuals would not have been eligible for hospice service. The data sources were linked by a prob abilistic record matching method. The probabilistic matching method is a statistically validated technique that uses multiple

PAGE 71

61 passes combined with clerical review of candidate matches (G. E. Mitchell, personal communication, October 12, 2004). Social s ecurity number was a common identifier variable for all three data s ources (i.e., Florida Medicaid L ong-Term Care analysis file, extended hospice claims file, death certificate record file) while Medicaid data files also included a unique Medicaid recipient identi fication number for each recipient. Thus, first, Medicaid recipient identification number and social security number were used to match and merge a Medicaid Long-Term Care analysis file and an extended hospice claims file. The merged Medicaid file wa s matched and merged with death certificate record file using social security number. The initial study population included a to tal of 118,703 Medicaid beneficiaries aged 18 and older, who died between SFY 2000 and 2002 (see Appendix A). From the 118,703 Medicaid decedents during those three state fiscal years, there were 117,667 beneficiaries who had a death ce rtificate record a nd manner of death recorded as natural or unknown. A further 1,036 Medicaid recipients were excluded from analyses, since individuals with traumatic or sudden death would not have been eligible for hospice service. Of 117,667 Medicaid decedents, 67,562 (57 pe rcent) were dua lly eligible for Medicare and Medicaid being eligible for both Medicare a nd Medicaid for at least 9 months. Of 67,562 dual-eligible beneficiaries, 61,137 (90 pe rcent) were aged 65 and older and 34,081 (50 percent) were identified as either long-term or short-term nursing home residents. Among 34,081 dual-eligible nursing home residents aged 65 and older, 30,765 residents were identified as either nonHispanic Black or non-Hispanic White and thus, there were 30,765 nursing home resi dents in the final study population.

PAGE 72

62 Measures Dependent Variables The dependent variable for the Research Question 1 was hospice use. Any use of hospice room and board care during the last year of life recorded in the Medicaid claims record was considered as use of hospice (hos pice vs. no hospice use). Hospice utilization information among the study population (i.e., dual-eligible nursing home residents in Florida) is based on Florida Medicaid claims record file. However, nearly all dualeligible nursing home residents receive hospice room and board care reimbursed by Medicaid. Although dual-eligible nursing home residents may receive hospice or nursing home care reimbursed by private sources other than Medicare or Medicaid, the author is not aware of actual or estimated number or perc entage of such cases in Florida or in the U.S. Thus, hospice utilization rate reporte d in this study is likely to reflect overall hospice utilization experience of dual-eligible nursing home residents in Florida. For Research Question 2 the dependent variable was survival from hospice enrollment until death in days. Since identify ing factors that affect timing of initiation of hospice service is a major interest in this study, number of days from the beginning date of hospice service recorded in the Medicaid Pr ogram Analysis file and the date of death from the death certificate record was calculate d and used as a dependent variable for the Research Question 2 The dependent variable for Research Question 3 was place of death recorded in the death certificate r ecord and includes: hospital inpatient, emergency room, outpatient medical facilitie s, dead on arrival; nursing home; private residence and others. This variable was reco ded into three discrete variables: death at hospital (1=yes, 0=no), death at nursing hom e and death at other (1=yes, 0=no).

PAGE 73

63 Predictor Variables For all three research questions, indepe ndent variables incl uded predisposing factors (age, gender, race, education), enab ling factors (marital status, urban/rural location of the primary residence, nursing home status), and need factors (principal diagnosis cancer, non-cancer COPD, CHF, stroke, dementia, or other non-cancer illness). For the Research Question 3 an additional independe nt variable, hospice use, was included. All these variables except educat ion and Hispanic ethnicity were available from the Medicaid Long-Term Care Analysis fi le. While Hispanics were not included in the study, the Hispanic ethnicity variable was used to id entify Black and White nursing home residents who were not of Hispanic ethnicity. The variable for education and Hispanic ethnicity were available fr om the death certificate record. Predisposing factors. Age and education were conti nuous variables. Categorical variables included gender (female=1 and ma le=0) and race/ethnicity (non-Hispanic Black=1, non-Hispanic White=0). Enabling factors. Marital status was recoded an d categorized into two groups: married as 1 and non-married as 0. The zip code was reviewed based on the Rural-Urban Commuting Areas (RUCA) code developed for th e state of Florida by Bill Kearns at the Department of Aging and Mental Health of th e Louis de la Parte Florida Mental Health Institute at the University of South Florida and recoded into two cat egories: urban as 1 and rural as 0. To be considered as a long-term nursing home resident, the decedent had to have a documented record of nursing home stay iden tified in the nursing home variable in the Medicaid Long-Term Care file during the last three consecutive months of life. If a

PAGE 74

64 recipient had a record of nursing home stay id entified in that file during any of the last three consecutive months of life without meeting the above duration criterion, the person was considered as a short-term nursing home resident. The nursing home variable in the Medi caid long-term care file is based on a complex algorithm that considers multiple combinations of different types of nursing home stays and determines the nursing home status based on the to tal number of nursing home stays during three cons ecutive months. However, the nursing home status algorithm identifies Medicaid recipients who el ect hospice in a nursing home setting as a hospice recipient and not as a nursing home resident during a given month because the priority is given to hospice status over nursi ng home stay in the administrative data. Thus, if the person elects hospice, his/her nu rsing home status is not captured by the nursing home variable in the data. Data re garding the nature a nd location of hospice services – nursing home room and board and other hospice services was obtained from the Medicaid Program Analysis file for all Medicaid decedents in this study. The flowchart of the decision ma king process for assigning nursing home status is described in the Appendix B. Need factors. The principal diagnosis of cancer or non-cancer condition as a cause of death was determined by examining the reported cause of death in the death certificates. The principal cause of death vari able was categorized a nd recoded into eight discrete variables: cancer, heart, congestive heart failure (CHF), chronic obstructive pulmonary disease (COPD), stroke, demen tia, renal failure, dementia/Alzheimer’s disease, and other. Among principal causes of death, CHF was distinguished from other health disease group. CHF has been identified as a condition characterized with a

PAGE 75

65 trajectory that consists of abrupt changes in functional status whic h is similar to COPD (Lunney et al., 2002). Statistical Approach Descriptive Analyses and Bivariate Co rrelations for Research Questions 1-3 Descriptive analyses were conducted of all predisposing, enabling and need factor variables and dependent variab les hospice use, hospice le ngth of stay, and hospital death – by hospice use, race/ethnicity, and nur sing home status. Descriptive analysis results include means with standard deviat ions and percentages and are described in Chapter 3. Pearson’s correlations were pe rformed on all independent and dependent variables for the overall study population (Tab le 1) and the sub-group of hospice users (Table 2). Although 10 out of 14 variable s are dichotomous variables, Pearson’s correlations were chosen to examine the strength of relationship between variables because studies have shown that variant forms of correlation for dichotomous variables such as point-biserial correlation rarely aff ect research conclusions (Tabachnik & Fidell, 2001). Significance levels were not included since these an alyses used a population, and not a sample. Table 1 is a display of the results of th e full correlation matrix for the overall study population including two dependent vari ables, hospice use and place of death. Pearson’s correlations analyses for bo th hospice and non-hospice groups found no evidence for multicollinearity among variables except for a very high negative correlation, r= -0.91, between hospital death and nursing home death. However, the high negative correlation between nursing home death and hospi tal death is due to other places of death accounting for only four percent of all deaths. Hospice use was negatively correlated

PAGE 76

66 Table 1. Correlations among Variables for the po pulation of Medicaid Nursing Home Residents 2 3 4 5 6 7 8 9 10 11 12 13 1 Age .24 .-11 -.07 -.23 -.12 02 -.09 .07 -.15 .15 .00 .02 2 Female -.07 .00 -.37 -.05 00 -.05 .05 -.06 .05 .02 .02 3 Non-Hispanic Black -.25 -.02 .10 .02 .04 -.08 .14 -.12 .00 -.06 4 Education .09 -.04 09 .00 .04 -.04 .03 .00 .04 5 Married .00 -.01 -.01 .03 .01 -.01 -.01 .02 6 Short-term Nursing Home .00 .00 -.07 .18 .19 .02 -.25 7 Urban area of residence -.01 .00 -.02 .00 .01 .04 8 Cause of Death: Cancer -.10 -.05 .05 .00 .11 9 Cause of Death: Dementia -.15 .13 .00 .09 10 Hospital death -.91 -.08 -.33 11 Nursing home death -.19 .32 12 Home death .01 13 Any hospice use 66

PAGE 77

67 Table 2. Correlations among Variab les for Medicaid Nursing Home Decedents Who Used Hospice 2 3 4 5 6 7 8 9 10 11 12 13 1 Age .23 -.10 -.10 -.25 -.07 04 -.15 .06 -.06 .06 .00 -.03 2 Female -.06 -.03 -.41 -.04 .03 -.07 .06 -.03 .00 .04 .00 3 Non-Hispanic Black -.25 -.02 .05 -.02 .09 -.07 .09 -.03 -.03 .04 4 Education .11 -.03 09 -.02 .03 -.03 .00 .00 -.04 5 Married -.01 -.02 -.03 .05 .02 -.01 -.04 -.02 6 Short-term Nursing Home .00 .07 -.04 .02 -.02 .01 .03 7 Urban area of residence -.04 .01 -.03 -.01 .02 -.03 8 Cause of Death: Cancer -.16 .00 .00 -.01 .00 9 Cause of Death: Dementia -.05 .03 .00 -.04 10 Hospital death -.07 -.03 .11 11 Nursing home death -.05 -.09 12 Home death -.09 13 Length of Stay in Hospice 67

PAGE 78

68 with short-term nursing home status (r=.25) and hospital death (r=-.33) and positively correlated with cancer as a cause of death (r=.11). Hospital death was positively correlated with non-Hispanic Black race/et hnicity (r=.14) and short-term nursing home status (r=.18) and negatively associated with age (r=-.15) and dementia as a cause of death (r= -.15). No variable except for death in a hospital (r=.11) showed a correlation of .10 or greater (e.g. accounting fo r 1 percent of more of variance) with hospice length of stay. Multivariate Analyses Research Question 1. The choice of multivariate analysis method for examining the role of multiple factors in predicting hos pice use for this analysis was a sequential binomial (or binary) logistic regression. Bi nominal logistic regression was used since the dependent variable is dichonomous (1=hos pice use, 0=no hospice use) (Tabachnick & Fidell, 2001). Unlike ordinary least square (OLS) regression, the logistic regression does not assume linearity of the relationship betw een the independent variables and dependent variable, normal distribution of variables, and homoscedastisity. A sequential binomial model is used over a direct model when th e underlying theoretical model being tested suggests a sequence of events (Tabachnick & Fi dell, 2001). In the behavioral model of service utilization, predisposing factors pr ecede enabling factors both of which are underlying factors that are trigge red by need factors. The model described took the form of: {log (P/1-P) = + 1X1 + 2X2 + 3X3 + 4XZ4 + e}, where P represents the probability of using hospice during the last year of life, X1 represents a set of predisposing fact ors (age, gender, race, education), X2 represents a set of enabling factors (urban/rural residen ce, short-term nursing home

PAGE 79

69 status), X3 represents a set of need factors (cancer, dementia), XZ4 represents interaction terms (non-Hispanic Black race/ethnici ty*being married; non-Hispanic Black race/ethnicity*cancer) while e represents an error term. The model was be tested by entering each set of individual determinant in sequence. Each variable within predisposing, enabling and need factors was a ssessed for main effect on hospice use first, and then, for potential interaction effects of marital status, and cancer diagnosis on hospice use. The multivariate analysis result revealed no meaningful (odds ratio of 1.01) effect of education on th e likelihood of hospice use while non-Hispanic Black race/ethnicity (odds ratio of 0.77), being ma rried (odds ratio of 1.16) and cancer (odds ratio of 3.2) had meaningful direct effect s on hospice use after c ontrolling for other factors. Thus, only two interaction effect models were tested in the final model: interaction between non-Hispanic Black race/e thnicity and being ma rried and interaction between non-Hispanic Black ra ce/ethnicity and cancer as a cause of death. For multivariate analysis results, upper and lower confidence intervals are provided, and results are not interpreted if confidence intervals cross or include 1.0. However, confidence intervals should not be interpreted for the purpose of significance testing since this study is based on a population of dual-eligible nursing home residents in Florida. Model fit statistics are presente d in the –2log likelihood, the adjusted R2, the receiver operating curve (roc) st atistic and the Hosmer-Lemeshow test for model fit. The same analysis was repeated to identify pr edictors of hospice use for each racial/ethnic group excluding race/ethnicity variable. Research Question 2

PAGE 80

70 The main goal of this analysis was to examine the relative risk of predictor variables in predicti ng the timing of the hospice enrollm ent among all hospice users with the focus on the role of race/ethnicity. As a choice of multivariate analysis method, a Cox proportional-hazards model was perf ormed to assess the multiple factors determining survival from hospice enrollmen t until death in days. In a sequential Cox regression, covariates enter the equation in an order sp ecified by the researcher (Tabachnik & Fidell, 2001). In this analysis, the same assu mptions were made regarding the order of entering predisposing, enabling and need factors since th e literature does not suggest otherwise. The assumptions regardi ng normality distribution, proportionality of hazards, and multicollinearity were assessed by examining the interactions between each variable and the natural logarithm of the time variable. No violation of these assumptions was found. Also, no censoring was involved si nce all individuals by definition of the population experienced d eath within 12 months of time period. The association between a variable and le ngth of stay in hospice was assessed by examining the odds ratios (or hazard ratio or ri sk ratio) indicated as in Exp (B). An odds ratio greater than unity is associated with a higher risk of death and therefore represents a shorter survival time after hospice enrollment, which is interpreted as shorter length of stay. Upper and lower confidence interval s are provided. The significant change as a result of entering additional sets of f actors (predisposing, enabling, and need) was assessed by examining the Wald test and the likelihood ratio test. The inverse was tried to reduce severe positive skewne ss of the survival time. Ho wever, the transformation of the dependent variable did not change the outcome of the Cox regression survival analysis, and thus, the results are based on the model with the untransformed dependent

PAGE 81

71 variable. The model fit did not improve by sequentially entering predisposing, enabling and need factors and the overall model fit was poor: -2 log like lihood of the model fit without covariate was 134313 while -2 log like lihood of the model fit with covariate was 134285. Thus, only the final model is presen ted for the overall nursing home hospice population and for each raci al/ethnic group. Research Question 3 The main goal of a series of analyses for Research Question 3 was to determine the unique effect of hospice use on hospital death after controlling for other potential covariates. The choice of multivariate anal ysis method for examining the effects of multiple factors on hospital death was a sequential binomial (or binary) logistic regression since the dependent variable is dichonomous (1=hospital death, 0=no hospital death) (Tabachnick & Fidell, 2001). The mode l described took the form of: {log (P/1-P) = + 1X1 + 2X2 + 3X3 + 4X4 + 5XZ5 e}, where P represents the probability of hospital death, X1 represents a set of predisposing fact ors (age, gender, race, education), X2 represents a set of enabling factors (urb an/rural residence, short-term nursing home status), X3 represents a set of need factors (cancer dementia), X4 represents hospice use, and XZ5 represents the interaction term for non-Hi spanic Black race/ethnicity and age on hospital death while e represents an error te rm. The model was tested by entering each set of individual determinant in sequence. Each variable within predisposing, enabling and need factors were assessed for main e ffect on hospital death and hospice use was entered in the final model to examine the effect of hospice on reducing hospital death after controlling for other covariates. As the direct effect of age on in-hospital death was

PAGE 82

72 not found, interaction effect of race/ethnicity and age on hosp ital death was not entered in the final model. For multivariate analysis results, upper and lower confidence intervals are provided. Model fit statisti cs are presented in the –2l og likelihood, the adjusted R2, the receiver operating curve (roc) st atistic and the Hosmer-Lemeshow test for model fit. The same analysis was repeated to identify pr edictors of hospice use for each racial/ethnic group excluding race/ethnicity variable.

PAGE 83

73 Chapter Three: Findings Population Characteristics The dual-eligible nursing home reside nts in this study (n=30,765) were on average 86 years old, female (73 percent), non -Hispanic White (85 percent), not married (82 percent), residing in a urban area, a l ong-term nursing home residents (74 percent) and had 11 years of formal education (Table 3). About five percent of the nursing home residents died of cancer while 37 percent a nd 15 percent died of heart disease and or some form of dementia respectively. The most common place of death was nursing home (69 percent) followed by hospital (27 pe rcent) and home (1.5 percent). Twenty eight percent of nursing home residents used any hospice service and of those who used hospice, the average days between the date of enrollment and the date of death was 53 days. For those who used hospice, the aver age length of stay was 48 days while the median length of stay was 21 days (data not shown). Hospice nursing home residents, compared with non-hospice residents, were more likely to be female (74 percent vs. 72 percen t), married (19 percent vs. 18 percent), living in an urban area (90 percent vs. 86 percent), to have died of cancer (10 percent vs. 4 percent) and dementia or Alzheimer’s disease (20 percent vs. 13 percen t), and to die in a nursing home (93 percent vs. 60 percent) compared with non-hospice nursing home residents. Non-hospice nursing home residents, compared with hospice residents, were more likely to be non-Hispanic Black (16 pe rcent vs. 11 percent) and short-term nursing

PAGE 84

74 Table 3. Nursing Home Residents Aged 65 and Older by Hospi ce Use, Race/Ethnicity, and Nursing Home Status (N=30,765) All Hospice Non-Hospice Non-Hispanic White Non-Hispanic Black LTNH1 STNH2 Number (%) or Mean SD PREDISPOSING Age 85.98 868 85.88 86.37.9 83.78.9 86.58 848 Gender: Female 73 74 72 72.6 73.9 74 69 Education 10.82.9 112.8 10.72.9 11.12.7 93.3 113 113 Race: Non-Hispanic Black 14 .6 11.3 15.9 0 100 13 21 ENABLING Marital status: Married 18.1 19.4 17.7 18.5 16.2 18 18 Area of Residence Urban 87.3 89.5 86 87 88 88 88 STNH 2 25.9 8 32.8 24.1 36.6 0 100 NEED Cause of Death Cancer 5.4 9.6 3.9 5.1 7.7 5 6 Heart 36.7 32.9 38.2 37 37 37 37 Dementia 14.9 20.3 12.9 16 8.4 16 11 DEPENDENT VARIABLES Any Hospice use 27.7 100 0 28.8 21.4 Time from hospice enrollment until death 5368 5368 0 5368 5167 5166 78101 Place of death Hospital 26.9 3.6 35.8 24.4 41.7 22 41 Nursing home 69 93 60 71.4 55.5 74 54 Home 1.5 1.8 1.4 1.6 1.3 1.4 2 Note: 1 LTNH: Long-term nursing home. 2 STNH: Short-term nursing home. 74

PAGE 85

75 home resident (33 percent vs. 8 percent), to die of heart disease (38 percent vs. 33 percent), and to die in a hosp ital (36 percent vs. 4 percent). Compared with non-Hispanic White nursing home residents, non-Hispanic Bl ack nursing home residents were likely to be younger (84 years old vs. 86 years old) and ha ve lower level of education (9 years vs. 11 years). Non-Hispanic Black nursing home residents were also less likely to be married (16 percent vs. 19 percent), to die of dementia or Alzheimer’s disease (8 percent vs. 16 percent), and use hospice. They were more likely to be short-term nursing home resident, to die of cancer (8 pe rcent vs. 5 percent) and to di e in a hospital (42 percent vs. 24 percent). Mean length of stay in hospice was somewhat shorter for non-Hispanic Black nursing home residents compared with non-Hispanic White (45 days vs. 47 days) nursing home residents, although the variance was quite large compared to the magnitude of difference. Length of stay in facility yielded gene rally expectable results. Long-term nursing home residents were likely to be older (87 years vs. 84 years), female (74 percent vs. 69 percent), to die of dementia/Alzheimer’s dis ease (16 percent vs. 11 pe rcent) and to die in a nursing home (74 percent vs. 54 percent). Major Findings The major findings for each of three re search questions are presented next. Research question 1. Do major predisposi ng, need, and enabling factors have significant associations with hospice use in the la st year of life among dual-eligible nursing home residents? The first step of the sequential binomial logistic regression model was to enter four predisposing characteristics (i.e., age, gender, race/ethnicity, education). The second

PAGE 86

76 step included three characteristics (i.e., marita l status, urban/rural status of residence, nursing home status) which were followed by th ird step of entering two major causes of death: cancer and dementia/Alzheimer’s dis ease. The final step was to enter the interaction terms between race/ethnicity and marital status and race/ethnicity and cancer as a cause of death. All steps for the ove rall nursing home populat ion are displayed in Table 4. All steps of analysis were repeated for each racial/ethnic group and displayed in Table 5 for the non-Hispanic White group and Table 6 for the non-Hispanic Black group. The race/ethnicity variable, however, was ex cluded from entering into the model for within-group analysis. For the nursing home population (Table 4), there was a good model fit based on improved chi-square statistic for the -2log li kelihood parameters which were significant (p .001) at each step. In addition, the Hosm er and Lemeshow goodness-of-fit test was not significant at each step, so the null hypothe sis (that there is a better model than this one) was rejected. Finally, adjusted R2 of the final model was 0.14. The odds ratio provides the relative likelihood that a characteristic cont ributes to a nursing home resident receiving hospice care. Odds ratios that are greater than one suggest that the characteristics increases the likelihood of r eceiving hospice care while odds ratios that are less than one suggests that the characteristic decreases the likelihood of receiving hospice care. Where confidence interval s cross or include 1.0, results ar e considered to be of low magnitude and are not interpreted. When just predisposing characteristics were entered (Model 1 in the Table 4), being non-Hispanic Black was negatively a ssociated with hospice use. Non-Hispanic Blacks were 29 percent less likel y to use hospice after contro lling for other predisposing

PAGE 87

77 factors. Age, gender, and education had limited effects on the lik elihood of hospice use, odds ratio of 1.01, 1.07, and 1.02 respectively. When predisposing and enabling factors we re entered together (Model 2 in the Table 4), short-term nursing home residents we re 82 percent less likely to use hospice, while being married and resident of urban areas increased the li kelihood of receiving hospice by 15 percent and 34 percent respec tively. After controlling for these predisposing and enabling factors, non-Hispan ic Black nursing home residents were still 20 percent less likely to use hospice. When cancer and dementia/Alzheimer’s di sease were entered in the third step (Model 3 in the Table 4) along with predispos ing and enabling characteristics, having died from cancer increased the likelihood of using hospice more than threefold and dementia/Alzheimer’s disease as a cause of death increased the likelihood by 69 percent. The relative likelihood of using hospice among non-Hispanic Black remained the same (odds ratio of .80) in the third model. The final model (Model 4 in the Table 4) revealed a relatively strong interaction effect of race/ethnicity and cancer as a cause of death on hospice use (odds ratio of 1.35). However, the interaction effect between race/e thnicity and marital status was found to be relatively small (odds ratio of .97). The regression line for non-Hispanic Black and nonHispanic White nursing home residents with cancer and non-cancer causes of death was calculated and is displayed in Figure 2. Non-Hispanic Black nursing home residents were less likely than White residents to us e hospice if they had a non-cancer cause of death but non-Hispanic Black nur sing home resident with cancer as a cause of death were

PAGE 88

78 Table 4. Multiple Logistic Regression Models Pr edicting Hospice Use by Race/Ethnicity (N=30,765) Model 1 Model 2 Model 3 Model 4 Odds Ratio (95% Confidence Interval) PREDISPOSING Age 1.01 (1.00-1.01) 0.99 (0.99-1.00) 0.99 (1.00-1.00) 0.99 (1.00-1.00) Gender: Female 1.07 (1.00-1.13) 1.08 (1.02-1.16) 1.09 (1.02-1.16) 1.09 (1.02-1.16) Race: Non-Hispanic Black 0.71 (0.66-0.77 ) 0.80 (0.74-0. 87) 0.80 (0.73-0.86 ) 0.80 (0 .65-0.98) Education 1.02 (1.01-1.03) 1.02 (1.01-1.02) 1.01 (1.00-1.02) 1.01 (1.00-1.02) ENABLING Marital status: Married 1.15 (1.06-1.23) 1.16 (1.07-1.25) 1.16 (1.07-1.25) Area of Residence Urban 1.34 (1.23-1.46) 1.36 (1.25-1.48) 1.36 (1.25-1.48) Short-Term Nursing Home Resident 0.18 (0.17-0.20) 0.18 (0.17-0.20) 0.18 (0.16-0.20) NEED Cause of Death Cancer 3.4 (3.07-3.8) 3.01 (2.65-3.42) Dementia 1.69 (1.56-1.81) 1.69 (1.56-1.81) INTERACTION NHBlack Marital Status Married 0.97 (0.78-1.21) NHBlack*Cancer 1.35 (1.03-1.77) -2 log likelihood 36183 33891 33246 33241 Hosmer and Lemeshow 0.24 0.13 0.37 0.55 Max-rescaled R-square 0.01 0.11 0.14 0.14 c 0.54 0.65 0.68 0.68 78

PAGE 89

79 as likely as non-Hispanic White nursing home re sidents to use hospice. The race effect remained the same at this step. 0 0.1 0.2 0.3 0.4 0.5 0.6 Cancer Non-Cancer Cause of DeathProbability of Hospice Use Non-Hispanic Black Non-Hispanic White Figure 2. Interaction Effect of Race/Ethnicity on Cause of Death – Cancer with Hospice Use The results of sequential binomial logistic regression model for the non-Hispanic White and non-Hispanic Black groups are found in Table 5 and Table 6. For non-Hispanic White nursing home residents, female gender, being married, and two causes of death – cancer and dementia increased the likelihood of receiving hospice care while short-term nursing home status decreased the likelihood of hospice use. However, only short-term nursing home status and two causes of d eath, cancer and dementia, predicted the likelihood of hospice use among non-Hispanic Black nursing home residents (all three results with the same direction). Specifical ly, urban location of re sidence led to greater likelihood of using hospice for non-Hispanic Whites (odds ratios of 1.39 vs. 1.01)

PAGE 90

80 Table 5. Multiple Logistic Regression Models Pred icting Hospice Use For Non-Hispanic Whites (n=26,271) Model 1 Model 2 Model 3 Odds Ratio (95% Confidence Interval) PREDISPOSING Age 1 (1.00-1.01) 1 (0.99-1.00) 1 (0.99-1.00) Gender: Female 1.07 (1.00-1.14) 1.09 (1.01-1.16) 1.08 (1.01-1.16) Education 1.02 (1.01-1.03) 1.01 (1.00-1.02) 1.01 (1.00-1.02) ENABLING Marital status: Married 1.16 (1.08-1.26) 1.15 (1.07-1.25) Area of Residence Urban 1.39 (1.28-1.53) 1.4 (1.29-1.55) Short-Term Nursing Home Resident 0.18 (0.17-0.20) 0.18 (0.17-0.20) NEED Cause of Death Cancer 3.21 (2.85-3.62) Dementia 1.67 (1.56-1.80) -2 log likelihood 31516 29578 29076 Hosmer and Lemeshow 0.05 0.85 0.57 Max-rescaled R-square 0.00 0.10 0.13 c 0.52 0.64 0.67 80

PAGE 91

81 Table 6. Multiple Logistic Regression Models Predicting Hospice For non-Hispanic Blacks (n=4,494) Model 1 Model 2 Model 3 Odds Ratio (95% Confidence Interval) PREDISPOSING Age 1 (0.99-1.01) 1 (0.99-1.01) 1 (0.99-1.01) Gender: Female 1.08 (0.92-1.26) 1.08 (0.92-1.26) 1.17 (0.98-1.40) Education 0.99 (0.98-1.02) 0.99 (0.98-1.02) 0.99 (0.97-1.02) ENABLING Marital status: Married 1.12 (0.91-1.39) 1.17 (0.94-1.46) Area of Residence Urban 1.01 (0.80-1.27) 1.06 (0.83-1.33) Short-Term Nursing Home Resident 0.18 (0.15-0.22) 0.18 (0.14-0.22) NEED Cause of Death Cancer 4.41 (3.44-5.64) Dementia 1.82 (1.43-2.33) -2 log likelihood 4662 4304 4157 Hosmer and Lemeshow 0.09 0.61 0.99 Max-rescaled R-square 0.00 0.12 0.17 c 0.52 0.67 0.71 81

PAGE 92

82 compared with non-Hispanic Blacks. The two causes of death studied, cancer and dementia/ Alzheimer’s disease, had greater as sociation with the lik elihood of hospice use (odds ratio of 4.41 and 1.82 vs. 3.21 and 1.7) w ithin the non-Hispani c Black group than non-Hispanic White group. Research question 2. Do major predisposing, need, and enabling factors have significant associations with length of stay in hospice in the last year of life among dual-eligible nursing home residents? The multivariate analysis result of the full model for the overall hospice user population, the non-Hispanic White, and non-Hi spanic Black groups (Table 7) suggests that the covariates (i.e., predisposing, enabli ng and need factors) in cluded in the model do not explain the model very well. The factor s that significantly incr eased the probability of survival time were female gender (haza rd ratio, .93) and urban area of residence (hazard ratio, .92) while stroke (hazard ratio, 1.09) reduced the probability of survival time. The poor model fit and the small num ber of factors significant affect the probability of survival time sugge st that the survival time might be influenced by external factors other than covariates examined in this study. Some differential effects of covariates within each group included increase d probability of survival time associated with female gender for non-Hispanic Blacks a nd stroke as a cause of death associated with decreased probability of survival time fo r non-Hispanic Whites. Nonetheless, poor model fit and low explanatory pow er of all three full models s uggest that exte rnal factors other than covariates entered in this gr oup of nursing home population may explain the timing of hospice enrollment.

PAGE 93

83 Table 7. Cox Regression Model to Examine the Days of Hospice Length of Stay All Non-Hispanic White Non-Hispanic Black PREDISPOSING Hazard Ratio P -value Hazard Ratio P -value Hazard Ratio P -value Age 1.00 0.27 1.00 0.75 1.01 0.07 Gender: Female 0.93 0.007 0.95 0.05 0.84 0.02 Race: Non-Hispanic Black 1.00 0.94 N/A N/A N/A N/A Education 1.00 0.21 0.99 0.14 1.00 0.64 ENABLING Marital status: Married 0.99 0.80 0.99 0.88 1.04 0.87 Area of Residence Urban 0.92 0.03 0.94 0.08 0.88 0.14 Short-Term Nursing Home Resident 1.03 0.42 1.03 0.59 1.10 0.35 NEED Cause of Death Cancer 0.97 0.43 0.93 0.25 0.94 0.96 Heart 1.02 0.51 1.02 0.24 0.81 0.14 CHF or COPD 0.94 0.83 0.96 0.88 0.68 0.09 Stroke 1.09 0.01 1.10 0.02 0.96 0.64 Dementia/Alzheimer's Disease/Senility 0.96 0.96 0.97 0.99 0.82 0.84 -2 log likelihood 83

PAGE 94

84 Research question 3. Does hospice use re duce hospital death a fter controlling for covariates? For the overall nursing home resident population (Table 8), when predisposing factors were entered, a one unit increase of age and education led nursing home residents to be 4 percent and 2 percent less likely to die in a hospita l. Similarly, a female nursing home resident was 9 percent less likely to die in a hospital. At the same time, a nonHispanic Black nursing home resident was 96 percent more likely to experience inhospital death. When enabling factors were entered along with predisposing factors, being married and residing in an urban area was ne gatively associated with the likelihood of inhospital death. The odds ratios were 0.89 a nd 0.88, respectively. On the other hand, shortterm nursing home residents were twice as li kely as to experience in-hospital death. As shown in the Model 3 of the Table 8, those w ho died of cancer and dementia/Alzheimer’s disease were 62 percent and 66 percent less likely to die in a hospital. Hospice use, which was entered in the final, substantially reduced the likelihood of in-hospital death (after controlling for all other factors) with an odds ratio of 0.08: nursing home residents who used hospice were 92 percent less likely to die in a hospital. At the same time, the relative influence of urban lo cation of residence, shortterm nursing home status, and cancer and dementia/Alzheimer’s disease as causes of death were reduced. After controlling for all other predis posing, enabling and need f actors and hospice use, nonHispanic Black race/ethnicity was still strong ly and positively associated with in-hospital death (odds ratio of 1.76) which suggests th at despite the powerful effect of hospice

PAGE 95

85 preventing hospitalization and subsequent in -hospital death, non-Hispanic Blacks were more likely to die in a hospital. Within-group analyses of non-Hispanic White and non-Hispanic Black groups (Table 9, Table 10) found that variables such as gender, marital status, urban location of residence, dementia/Alzheimer’s disease as a cause of death and use of hospice have differential effects on the like lihood of in-hospital death wi thin the two racial/ethnic groups. Comparing the results of the final model from each racial/ethnic group suggests that female non-Hispanic Whites are 11 percen t less likely to experi ence in-hospital death while the effect of gender appears to be minimal in Blacks. Dying from dementia/Alzheimer’s disease reduces the like lihood of in-hospital death by 63 percent in Whites while it reduces the lik elihood by 57 percent among nonHispanic Blacks. Lastly, hospice appears to have a stronger influen ce among non-Hispanic Whites (odds ratio of 0.07) than non-Hispanic Blacks (odds ratio of 0.10). Of note is that hospice use has similar effects in reducing the likelihood of hospital death ac ross race/ethnicity.

PAGE 96

86 Table 8. Multiple Logistic Regression Models Predic ting Hospital Death by Race/Ethnicity and Hospice Use Model 1 Model 2 Model 3 Model 4 Odds Ratio (95% Confidence Interval) PREDISPOSING Age 0.96 (0.96-0.97) 0.96 (0.96-0.97) 0.96 (0.96-0.97) 0.96 (0.96-0.97) Gender: Female 0.91 (0.86-0.96) 0.9 (0.84-0.95) 0.9 (0.84-0.95) 0.9 (0.84-0.95) Race: Non-Hispanic Black 1.96 (1.82-2.10 ) 1.84 (1.711.97) 1.81 (1 .68-1.94) 1.80 (1.67-1.94) Education 0.98 (0.92-0.99) 0.99 (0.98-0.99) 0.99 (0.98-1.00) 0.99 (0.98-1.00) ENABLING Marital status: Married 0.89 (0.82-0.95) 0.9 (0.84-0.97) 0.94 (0.87-1.01) Area of Residence Urban 0.88 (0.82-0.95) 0.88 (0.81-0.95) 0.94 (0.87-1.2) Short-Term Nursing Home Resident 2.13 (2.02-2.26) 2.08 (1.97-2.20) 1.44 (1.36-1.53) NEED Cause of Death Cancer 0.38 (0.34-0.44) 0.57 (0.49-0.66) Dementia 0.34 (0.31-0.38) 0.38 (0.34-0.42) HOSPICE 0.08 (0.07-0.08) -2 log likelihood 34699 33982 33228 30000 Hosmer and Lemeshow 0.16 0.02 0.37 0.03 Max-rescaled R-square 0.05 0.08 0.12 0.25 c 0.62 0.65 0.682 0.76 86

PAGE 97

87 Table 9. Multiple Logistic Regression Models Predicting Hospital Death for Non-Hispanic Whites Model 1 Model 2 Model 3 Model 4 Odds Ratio (95% Confidence Interval) PREDISPOSING Age 0.96 (0.96-0.97) 0.96 (0.96-0.97) 0.97 (0.96-0.97) 0.96 (0.96-0.97) Gender: Female 0.87 (0.84-0.95) 0.89 (0.81-0.93) 0.88 (0.82-0.94) 0.89 (0.83-0.95) Education 0.98 (0.97-0.99) 0.98 (0.97-0.99) 0.99 (0.98-1.00) 0.99 (0.98-1.00) ENABLING Marital status: Married 0.85 (0.79-0.92) 0.87 (0.80-0.95) 0.9 (0.83-0.99) Area of Residence Urban 0.84 (0.77-0.91) 0.83 (0.76-0.90) 0.9 (0.82-0.98) Short-Term Nursing Home Resident 2.16 (2.03-2.29) 2.09 (1.97-2.23) 1.45 (1.35-1.54) NEED Cause of Death Cancer 0.39 (0.33-0.46) 0.57 (0.48-0.67) Dementia 0.34 (0.30-0.37) 0.37 (0.33-0.41) HOSPICE 0.07 (0.06-0.08) -2 log likelihood 28653 28050 27409 24686 Hosmer and Lemeshow 0.2832 0.643 0.668 0.39 Max-rescaled R-square 0.03 0.06 0.10 0.23 c 0.592 0.64 0.09 0.754 87

PAGE 98

Table 10. Multiple Logistic Regression Models Pr edicting Hospital Death For non-Hispanic Blacks Model 1 Model 2 Model 3 Model 4 Odds Ratio (95% Confidence Interval) PREDISPOSING Age 0.97 (0.97-0.98) 0.97 (0.97-0.98) 0.97 (0.97-0.98) 0.97 (0.97-0.98) Gender: Female 0.96 (0.84-1.09) 0.96 (0.86-1.12) 0.96 (0.83-1.10) 0.99 (0.86-1.15) Education 1.00 (0.98-1.02) 1.00 (0.98-1.02) 1.00 (0.98-1.02) 1.00 (0.98-1.02) ENABLING Marital status: Married 1.05 (0.88-1.25) 1.04 (0.87-1.24) 1.09 (0.91-1.32) Area of Residence Urban 1.19 (0.98-1.44) 1.18 (0.97-1.44) 1.22 (0.99-1.50) Short-Term Nursing Home Resident 2.03 (1.79-2.30) 2 (1.76-2.27) 1.42 (1.24-1.62) NEED Cause of Death Cancer 0.38 (0.29-0.49) 0.57 (0.43-0.75) Dementia 0.4 (0.31-0.51) 0.43 (0.33-0.56) HOSPICE 0.10 (0.08-0.13) -2 log likelihood 6033 5905 5794 5288 Hosmer and Lemeshow 0.53 0.06 0.34 0.84 Max-rescaled R-square 0.02 0.06 0.09 0.22 c 0.57 0.62 0.65 0.72 88

PAGE 99

89 Chapter Four: Discussion and Conclusions Identifying and reducing barriers to quali ty end-of-life care among diverse older adults has become an increasingly important area of research. The findings from this study provide implications for theory, rese arch, practice and policy in regard to improving access to and outcomes of hospice serv ices for frail nursing home residents at the end of life. These findings and implications are addressed in terms of: (a) the utility of the behavioral model of health servi ces use (Andersen, 1995) ; (b) racial/ethnic differences in utilization of hospice servic es; (c) the limited cont ribution of individual characteristics to length of stay in hospice; and (d) the role of race /ethnicity and hospice in in-hospital death. The Behavioral Model of Health Services Use The first question raised by this research was aimed at identifying characteristics that facilitate hospice use am ong dual-eligible nursing home residents. Analyses were guided by the behavioral model of health se rvices use (Andersen, 1995). The behavioral model of health services use successfully predicted group membership in hospice use. With each sequential step of the model, the model fit improved compar ed to the Intercept only and all improvements were significant. In the full model, seven variables female gender, non-Hispanic White race/ethnicity, be ing married, urban area of residence, and cancer and dementia/Alzheimer’s disease as causes of death – predicted increased hospice use.

PAGE 100

90 Although it was hypothesized that older ag e and higher level of education would be associated with the incr eased likelihood of hospice use, this study did not find any meaningful effects of these two predisposing characteristics. At the same time, as predicted, race/ethnicity was th e prevailing predisposing charac teristic contributing to the model. Non-Hispanic Black nursing home re sidents were 20 percen t less likely to use hospice after controlling for all other factors. This finding is discussed further with findings on within-group differences later in this chapter. All three enabling factors te sted in this study affected the likelihood of hospice use. Married and urban nursing home resident s were more likely to use hospice by 16 percent and 36 percent, respectively, whereas short-term nursing home residents were 82 percent less likely to use hospice. The role of marital status is of particular note since, while it is consistent with the behavioral model followed in this research, the relationship has not been found in other research. Two studies (Chen et al., 2003; Enguidanos et al., 2005) found no significant effect of marita l status while a study by Greiner and colleagues (2003) found that being never married significantly increase d the likelihood of receiving hospice while being widowed signi ficantly reduced the chance of receiving hospice. One of the reasons for the discrepancy in findings for marital status may lie in the unique nature of the population under study. Dual -eligible individuals who are in fact eligible for both Medicare and Medicaid be nefits, have generally not received much attention in past research. The study population of the study by Enguidanos and colleagues (2005) resemble the characteristic s of the population of the present study the most, as Enguidanos and colleagues examined hospice use among dual-eligible decedents

PAGE 101

91 in California. However, this dissertati on examined the experiences among nursing home residents while the study by Enguidanos and colleagues included both communitydwelling and nursing home residents. It is possible that the deci sion to use hospice is influenced by preferences and influence of spouses of nursing home residents in the current study. However, information on s pouses of nursing home residents is not available from the data sources used for this study. The finding that urban residents were more likely to use hospice is consistent with previous research findings that the availabil ity of hospice in rural areas is more limited than in urban areas and the rate of hos pice use is correspondingly lower among rural decedents (Virnig, Moscovice, Kind, & Casey, 2002). Although Florida serves a higher percentage of terminally ill patients than mo st of other states (Miller and Lima, 2004), it may be that hospice services remain less availa ble in rural areas than in urban areas and nursing home residents in rural area still e xperience access barriers to hospice care. As predicted, short-term nursing home resi dents in this study were less likely to use hospice. Short-term nursing home resident s in this study were different from longterm nursing home residents. Short-term nursing home residents were, on average, younger, more likely to be males, and more lik ely to be non-Hispanic Blacks. Short-term nursing home residents may be a unique subgroup of nursing home residents who enter nursing home with different demographic ch aracteristics under different circumstance from long-term nursing home residents. It is also possible that short-term nursing home residents were more likely to be receiving Medicare skilled nursing benefit which does not allow simultaneous enrollment in hospi ce. Further examina tion of end-of-life experiences of short-term nur sing home residents, particul arly on their experience in

PAGE 102

92 transitions in and out of health and long-term care institutions will help us to better understand end-of-life experi ences of this unique subpopulation of nursing home residents. Better understanding of the factors that facilitate and prevent short-term nursing home residents from hospice use will help identify barriers to hospice care and develop policy initiatives such as reimburseme nt policy to reduce such barriers that may exist at individual and health care facility levels. As previous research testing the behavi oral model of health services use has shown, the primary need factor used in th is study, cause of death, was the strongest predictor of hospice use. Nursing home residents dying from cancer were three times more likely to use hospice than those who di ed of illnesses other than cancer. Another unexpected need factor found in this study wa s dementia/Alzheimer’s disease as a cause of death leading to increase in the likeli hood of using hospice by 69 percent. It is not clear as to why dementia/Alzheimer’s diseas e as a cause of death would predict higher likelihood of hospice use compared to non-demen tia causes of death, since historically dementia patients have vastly underutilized hospice care (Christakis & Escarce, 2001). However, Florida is a state with the hi ghest proportion of dying older adults receiving hospice care, 35 percent. It al so has a high proportion of non-cancer hospice patients (60 percent), second only to Ariz ona (Miller & Lima, 2004). In addition, a higher proportion of nursing home residents is served by hospice providers in Florida than in 44 of the 48 states studied by M iller and colleagues (Miller & Lima, 2004). Given a high level of penetration of hospice se rvice providers in the state of Florida and that a significant proportion of nursing home residents have dementia or other cognitive impairment, it is possible that nursing home residents with dementia /Alzheimer’s disease

PAGE 103

93 are more likely to be identified as appropriate candidates for hospice ca re than those with illnesses other than cancer or dementia/Alzheim er’s disease. In addition, the guidelines for determining 6-month prognosis, origin ally developed by National Hospice and Palliative Care Organization and subsequent ly adopted by Medicare, includes guidelines for not only cancer diagnoses but also dementia diagnoses. Such guidelines for dementia patients may have made it easier for nursing home care providers, physicians and hospice providers to identify and refer and enroll nur sing home residents with dementia into a hospice program. This finding is encouraging given that hos pice services have historically been utilized primarily by terminally ill patients with a cancer diagnosis. Although recent data from NHPCO reports that an increasing number of patient s with non-cancer diagnoses receive hospice care, cancer pa tients still made up 46 per cent of all hospice patient population in 2004 (NHPCO, 2005), while cancer cause s 23 percent of deaths in the U.S. (Anderson & Smith, 2005). Further examin ation of how the decision of hospice enrollment is made among nursing home reside nts with different clinical characteristics at individual, family, nursing home provider, physician, and hospice provider levels will improve our understanding of the ef fects of principal illness. Racial/Ethnic Differences in U tilization of Hospice Services This study found a significantly lower le vel of hospice utiliz ation rate in nonHispanic Black as compared to non-Hispanic White nursing home residents, even after controlling for other factors. This finding is consistent with previous research. The explanation for this finding may be twofold.

PAGE 104

94 First, the racial/ethnic difference in hospice utilization may reflect cultural differences in preferences for end-of-life car e between the two groups as race/ethnicity was viewed as a proxy for culture in this st udy. The literature sugge sts that non-Hispanic Blacks are less likely to use hospice due to different sociodemogra phic characteristics and their unique historical and cultural experiences in the medical system. However, the fact that non-Hispanic Black nursing home resi dents were less likely to use hospice even among dual-eligible nursing hom e residents (with low income and poorer health) after controlling for the level of education suggest s that the racial/et hnic difference found in this study is not likely to be a result of differential socioeconomic status of the two groups. Furthermore, this study excluded fr om the study White and Black nursing home residents who were identified as Hispanic Thus, the unique historical and cultural experiences of non-Hispanic Blacks may attribut e, at least partiall y, to the differential rates of hospice use. A second reason for racial/ethnic diffe rence may be the potential effect of residential segregation reported in the nursing home literature. A recent national study of nursing homes (Mor, Zinn, Angelelli, Te no, & Miller, 2004) found that Black nursing home residents were significantly overrepresented in poor-quality nursing home facilities measured by quality of care indicators incl uding staffing and pain control. It is conceivable that poor-quality nursi ng facilities are less likely to facilitate residents’ use of hospice care. Poor-quality nursing facilities are more likely to experience the shortage of highly trained health care staff who can monitor the progress of the illness and functional decline of the residents and identify and refer residents for whom hospice care may be appropriate compared with high-quality nursing facilities.

PAGE 105

95 Poor-quality facilities also may be less able to coor dinate comprehensive end-oflife care with hospice providers for their dy ing nursing home residents. If a higher proportion of non-Hispanic Black residents in the current study were living in poorquality facilities than non-Hispanic White residents, it is possible that non-Hispanic Black nursing home residents experienced an access barrier to hospice care due to the facility effect. However, the current study is limited to examining only individual level characteristics and hospice use due to the da ta limitations. Ideally, residents in poor quality facilities should be more likely to receive hospice care because poor-quality facilities are more likely to l ack the capacity to provide quality pain control and symptom management on their own. The results of the interac tion effect of race/ethnicity and cancer as a cause of death with hospice use, and w ithin-group analyses, also sugge st unique differences in the role of predisposing, enabling and need fact ors in predicting hospice use. This study found that race/ethnicity mode rates the strength of the effect of the illness on the likelihood of using hospice. Among reside nts who died of cancer, no difference in hospice use was found between the two racial/ethnic groups while hospice utilization rate was lower for non-Hispanic Blacks than non-Hispanic Whites among non-cancer residents. It may be that cultural differences in the perception of illness may exist between the two groups. As cancer has a re latively predictable illness trajectory compared to other illnesses, and is generally recognized to be a lethal illness, it is possible that there may not be a significan t difference in accepting a prognosis of terminal cancer between the tw o groups and thus, cancer may not lead to differential use of hospice between the two groups. However, for other illnesses w ith less predictable

PAGE 106

96 illness trajectories, non-Hispanic Blacks may be more likely to pursue curative treatments than hospice care. Further analysis of predictors of hospice use within each group, an exploratory component of the study, found that marital st atus and urban/rural location of primary residence influenced the likelihood of using hospice. This was the case only for the nonHispanic White group. In addition, the tw o causes of death studied, cancer and dementia/Alzheimer’s disease, exerted a st ronger influence on th e likelihood of using hospice among non-Hispanic Blacks compared with non-Hispanic Whites. These findings may imply that the current behavioral model of health serv ice use does a better job of predicting the likel ihood of use of hospice for the non-Hispanic White group compared with non-Hispanic Black group. In corporating psychosocial factors in the model may enhance the predictive function of the behavioral m odel of health services use model for non-Hispanic Black nursing home residents. In sum, this study found that a signifi cant racial/ethnic difference in hospice utilization exists among dual-e ligible nursing home residents. Such differences may be due to cultural preferences for types of end-of-life care and treatment of each racial/ethnic group or other extern al factors such as facility effects. At the individual level, better understanding of the mechan isms by which race/ethnicity affects the decision-making process and access to hospice is needed. Race/ethnicity is only a proxy for culture. Further research is needed to better understa nd racial/ethnic differences by incorporating psychosocial measures in the theoretical model that examines the expectations and views regarding specific illnesses, coping methods, death and dying, and the role of health and long-term care pr oviders. Moreover, exam ining the effects of

PAGE 107

97 nursing home facility and hos pice service provider will help better understand the broader, more complex process of access to and utilization of hospice among nursing home residents. Furthermore, nursing ho me and hospice staff may need to acknowledge racial/ethnic disparity in u tilization of hospice and gain a better understanding of the historical and cultural differe nces among members of variou s racial/ethnic groups which may help them to develop effective strate gies for nursing home community outreach and education. Limited Contribution of Indi vidual Characteristics to Le ngth of Stay in Hospice Another exploratory part of the overall study was the examination of the factors associated with length of stay in hospi ce among a sub-group of nursing home residents who used hospice. However, as the multivaria te results shows, the behavioral model of health service did not fit well and explained little variance. There may be two potential explanations for this result. First, beca use of the limited accuracy of the current prognostic criteria and tools and the six mont h prognosis rule, physicians may prefer to err on the side of overestimati ng survival time. As a resu lt, many, if not most, nursing home residents may not be referred to hospi ce until death is imminent. This is illustrated by the fact that the mean and median le ngths of hospice stay among nursing home residents in the current study were 47 days and 21 days, respectively. Second, it is also possible that factors ot her than individual characteristics may affect the length of stay in hospice. As a study of hospice survival time of Medicare beneficiaries (Christakis et al., 2000) reports, both nursing home and hospice provider level factors as well as health care market factors may affect the length of stay among nursing home residents. Further study of nursing home resident s that include facility and

PAGE 108

98 health care market level characteristics may improve the model fit and better explain the factors associated with hos pice length of stay. The Role of Race/Ethnicity and Hospi ce in Predicting In-Hospital Death The third question explored in this study was to examine the effect of race/ethnicity and hospice use in determining place of death. This dissertation research found that hospice is a powerful predic tor of place of death among nursing home residents. Ninety-three percent of nursing home residents who used hospice died in a nursing home compared with 60 percent of non-hospice nursing home residents dying in a nursing home. After control ling for other factors, hospice nursing home residents were 92 percent less likely to die in a hospital. At the same time non-Hispanic Black residents were still 80 percent more likely to die in a hospital even after adjusting for the effect of hospice use and other variables. The findings from the current study are par tially consistent with the finding of a recent study of dual-eligible decedents fr om California (Enguidanos et al., 2005). Enguidanos and colleagues (2005) found that am ong dual-eligible decedents in California (including both community-dwelling and nursi ng home residents) hospice users were three times more likely to die at home. At the same time, Enguidanos and colleagues found that Blacks were significantly more likel y to die at home although they were less likely to receive hospice care. However, different results between this current dissertation study and the study by Enguidanos et al. may be due to the fact that the study population of the current dissertation study only examines the experiences of nursing home residents in Florida. Moreover, the main outcome of the current study is inhospital death while the main outcome of the study by Enguidanos and colleagues was

PAGE 109

99 home death and thus, the findings cannot be directly compared with each other. Despite these differences, a similar conclusion to th e study of Enguidanos and colleagues (2005) can be drawn: hospice significantly reduces in-hospital death. Nursing homes often represent the last “home” fo r older adults with substant ial long-term care needs and hospice is effective in facilitating nursing home residents to die in place in this study population. Summary This dissertation research found racial/et hnic differences in utilization of hospice services among dual-eligible nursing home resi dents in Florida. The significantly lower hospice use among non-Hispanic Black nursing home residents in Florida, a state with high level of penetration and availability of hospice service providers, suggests that further outreach and education efforts ar e needed to increase communication and knowledge among nursing home residents, physic ians and hospice providers about the hospice program. Decision making, access and utiliz ation of health serv ices at the end of life are complex processes which are influen ced not only by individua l characteristics, but also service provider and health care mark et characteristics. Better understanding of this process will be achieved by further examining not only predisposing, enabling and need factors but also psychosoc ial factors that influence th e individual’s perceptions of illness and prognosis attitudes and beliefs about health care providers and system. Given the powerful effect of hospice reducing in in-hospital death and the competency and skills of hospice care professiona ls in providing quality end-oflife care, further efforts in reducing the gap in hospice utilization among minority nursing home residents through culturally sensitive strategies are needed.

PAGE 110

100 Study Limitations This dissertation study has several limitations. First, although indi viduals with any record of Medicaid claims in this study can be expected to represent the population of dual-eligible older adults who died in Florida during a certai n time period, SFY 20002001, SFY 2001-2002, and SFY 2002-2003, it is also recognized that the study population may not accurately represent all dua l-eligible older adults who received nursing home care. Not all dual-eligible older adults receive nursing home care reimbursed through the Medicaid program. It is possible that some minority of older adults in this group may have had their nur sing home care paid by a private long-term insurance benefit or other pr ivate resources, although the dua l eligible group is highly unlikely to have such resources available. These nursing home re sidents would not be captured in the data sources for this study. Although the majority of dual-eligible nursing home residents receive hospice care reimbur sed by Medicare first and then by Medicaid for the room and board part of hospice car e and this study specif ically focuses on the dual-eligible nursing home population in Flor ida, there may be unknown number of dualeligible nursing home resi dents who receive hospice care or nursing home care reimbursed by private sources such as longterm care insurance or private pay. The hospice utilization of these nursing home resi dents are not captured in this study. Second, although one of the major objectiv es was to examine the role of race/ethnicity, this study does not have info rmation on personal pref erences for hospice service or individual or ientation toward the self-identified race/ethnicity. Thus, the potential effects of cultural norms and indi vidual psychosocial fact ors were not assessed in this study. Moreover, it should be recognized that th ere may be sub-ethnic groups

PAGE 111

101 within each of the racial/et hnic groups studied. For example, there are various subgroups within non-Hispanic Whites such as Italian or Irish Ameri cans who may have different cultural norms. In addition, the data set does not include any se lf-identification of race or ethnicity by the patient or their family, and relied on the so-called “race” variable in Medicaid administrative files. As shown by Chen, Chen and Mehra (2005), there is a possibility that up to 7 or 8 percent of the Medicaid enrollees could have been misclassified. Third, certain types of nursing home resident s may not have been captured in this study due to the methodology chosen for this st udy. To be considered as a long-term nursing home resident in this study, the deced ent had to have a documented record of nursing home stay identified in the nursing home variable in the Medicaid Long-Term Care file during the last three consecutive months of life. If a recipient had a record of nursing home stay identified in that file duri ng any of the last thr ee consecutive months of life without meeting the above duration cr iterion, the person was c onsidered as a shortterm nursing home resident. However, th ere may be other long-term nursing home residents who were not categori zed as long-term nursing home re sidents. For instance, as the hospitalization increases toward the e nd of life, some long-term nursing home residents (who may have spent most of their la st year of life as nur sing home residents) may experience hospitalization and may spend the last month of life a nd die in a hospital. Although these residents should be considered as long-term nursing home residents, they were captured as short-term nursing home reside nts in this study. Thus, the effects of the short-term nursing home status on the outcome s need to be interpreted with caution.

PAGE 112

102 Fourth, there may be other potential fact ors affecting hospice use, timing of hospice initiation and place of death which are not included in this study. They include health care market factors, physician fact ors and organizational factors of hospice providers and nursing homes. The effects of th ese factors were not as sessed in this study due to data limitations. Lastly, the study population in this resear ch is dual-eligible older adults in Florida. Findings from this study regarding f actors associated with hospice use, timing of initial hospice enrollment, a nd the effects of hospice use on place of death may not be generalizable to the national population of dua l-eligible older adults. Due to several unique characteristics of Florida such as the st ate’s high level of hospi ce availability and high proportion of residential or inpatient be ds of hospice organizations, the rate of hospice may be higher and barriers to acce ss to hospice may be lower among dualeligible older adults studied in this research compared to dual-eligible older adults in other states. Future Directions This study found racial/ethnic differences in both hospice use and in –hospital death among dually eligible nurs ing home residents in Flor ida. In order to better understand the effect of race/ethnicity on end-of-life experiences of older adults, additional study on this topic should be c onducted utilizing Medi care and Medicaid claims and eligibility data and other facility level data such as On-Line Survey Certification And Reporting (OSCAR). Th ese types of data sources will provide additional important information on factors that may influence end-of-life services

PAGE 113

103 utilization experience of dual-e ligible older adults. In addi tion, comparison of end-of-life experiences across different long-term care settings can be examined. Addressing ethnic/racial divers ity and disparity in end-of-life service utilization is an important and challenging task. Although review of the literature and findings from this dissertation study suggest clearly distin ct patterns of services use by different racial/ethnic groups, the pathway by which the race/ethnicity influenc es services use is unclear. Incorporating psychosocial factors su ch as cultural norms and values regarding illness and dying, attitudes toward the health care system, and knowledge of the purpose and availability of hospice in addition to provider and organization factors affecting actual service use in the A ndersen model will help identify mechanisms by which race/ethnicity influence the decision to use and actual use of hospice. An increasing number of studies have found strong evidence for the effectiveness of hospice care in improving end-of-life car e outcomes and the efficiency of simple, feasible interventions that can improve hospice referral among community-dwelling and nursing home residents. The doc toral dissertation research re ported here finds a strong evidence for the efficacy of hospice in reducing in-hospital death among dually eligible nursing home residents in Florida. At the same time, the current study also found that racial/ethnic disparities in hospice use exist among nursing home residents. Given that these findings are based on a large sub-popula tion of nursing home residents, further efforts should be made in developing and providing hospi ce education and communication interventions for nursing home resi dents of diverse cultu ral backgrounds. A recent randomized controlled study of nur sing home residents (Casarett et al, 2005) shows that a simple educational and co mmunication intervention can significantly

PAGE 114

104 improve appropriate referral of nursing ho me residents to hospice care even among racially/ethnically diverse nursing home residents. Replicating such a simple intervention as an intervention used in th e study by Casarett and colleagues (2005) with additional efforts in improving cultural sens itivity and awareness of nursing home and other medical staff may help end-of-life decision-making processes among nursing home residents and their families and health care staff. Casarett and colleagues (2005) found that at both the beginning and 6-mont h follow-up, hospice referral rates were significantly higher in the intervention group compared with the control group. In addition, this study did not find significantly di fferent hospice enrollment rates between the two racial/ethnic groups. This fi nding suggests that when an effective communication and referral intervention is em ployed, racial/ethnic differences in hospice use among nursing home residents may be signif icantly reduced if not all differences may be reduced. Ensuring that nursing home residents and their family members receive appropriate and timely information on all f easible and appropriate end-of-life care options will be a critical st ep toward improvement in theoretical understanding of racial/ethnic differences in end-of-life decision making and service utilizations. Identifying institutional barriers such as lack of information or knowledge of end-of-life care options such as hospice care will be an important task that is needed for better understanding the role of cultu ral norms and expectations in perceptions and decision making processes by older adults and their fa mily members of diverse backgrounds at the end of life. With a better unde rstanding of instit utional and cultural factors influencing end-of-life care preferences and decision making processes, culturally sensitive and

PAGE 115

105 effective interventions can be developed to re duce racial/ethnic disp arities in end-of-life care access, utilization and out comes among diverse older adults and their families.

PAGE 116

106 List of References Aday, L. A., & Andersen, R. (1974). A framework for the study of access to medical care. Health Services Research, 9 208-220. Aday, L. A., & Awe, W. C. (1997). Health serv ice utilization models. In D. S. Gochman (Ed.), Handbook of health behavior research I: Personal and social determinants (pp. 153-172). New York: Plenum Press Ajzen, I. (1991). The theory of planned behavior, Organizational Behavior and Human Decision Processes, 50, 179-211. Allen-Burge, R. & Haley, W. E. (1997). I ndividual differences and surrogate medical decisions: Differing preferences for life-sustaining treatments. Aging and Mental Health, 1 121-131. American Cancer Society. (2003). Cancer fact s and figures for African Americans: 20032004. Retrieved October 1, 2004, from h ttp://www.cancer.org/downloads/STT/ 861403.pdf American Medical Association (1991). Gende r disparities in clin ical decision making. Journal of the American Medical Association, 266 559-562. Andersen, R. (1968). A behavioral model of families’ us e of health services (Research Series 25) Chicago: The University of Chica go Center for Health Administration Studies. Andersen, R. (1995). Revisiting the behavioral model and access to medical care: Does it matter? Journal of Health and Social Behavior, 36 1-10. Andersen, R. M., Davidson, P., & Ganz, P. ( 1994). Symbiotic relationships of qualify of life, health services researc h, and other health research. Qualify of Life Research, 3, 365-371. Andersen, R., & Newman, J. F. (1973). Societal and individual determinants of medical care utilization in the United States. Milbank Memorial Fund Quarterly, 51 95124.

PAGE 117

107 Anderson, R. N., & Smith, B. L. (2005). Deaths: Leading causes for 2002 National Vital Statistics Reports. 53, 17. Hyattsville, Maryland: Nati onal Center for Health Statistics. Arispe, I. E., Homes, J. S., & Moy, E. (2005) Measurement challenges in developing the National Healthcare Quality Report and the National Healthcare Disparities Report. Medical Care, 43, I17-23. Baer, W. M., & Hanson, L. C. (2000). Families’ perception of the added value of hospice in the nursing home. Journal of the American Geriatrics Society, 48 879-882. Barker, W.H., Zimmer, J. G., Hall, W. J., Ruff, B. C., Freundlich, C. B., & Eggert, G. M. (1994). Rates, patterns, causes, and co sts of hospitalization of nursing home residents: A population-based study. American Journal of Public Health, 84, 1615-1620. Bass, D., & Noelker, L. (1987). The influene of family caregivers on elder’s use of inhome services: An expanded conceptual framework. Journal of Health and Social Behavior, 28, 184-196. Berger, J. T. (1998). Culture and ra cial/ethnicity in clinical care. Archives of Internal Medicine, 158, 2085-2090. Bharucha, A.J., Pandav, R., Shen, C., Dodge, H. H., & Ganguli, M. (2004). Predictors of nursing facility admission: A 12-year epid emiological study in the United States. Journal of the American Geriatrics Society, 52 434-439. Bird, C. E., Shugarman, L. R., & Lynn, J. ( 2002). Age and gender differences in health care utilization and spending fo r Medicare Beneficiaries in their last years of life. Journal of Palliative Medicine, 5, 705-712. Blackhall, L. J., Frank, G., Murphy, S. T., Mich el, V., Palmer, J. M. & Azen, S. P (1999). Ethnicity and attitudes towa rds life sustaining technology. Social Science and Medicine, 48, 1779-1789. Blackhall, L. J., Murphy, S. T., Frank, G., Mi chel, V., Palmer, J. M., & Azen, S. (1995). Ethnicity and attitudes toward patient autonomy. Journal of the American Medical Association, 274, 820-825. Born, W., Greiner, A., Sylv ia, E., Butler, J., & Ahluwa lia, J. S. (2004). Knowledge, attitudes, and beliefs abou t end-of-life care among inne r-city African Americans and Latinos. Journal of Palliative Medicine, 7, 247-256.

PAGE 118

108 Bradley, E. H., Cramer, L. D., Bogardus, S. T., Kasl, S. V., Johnson-Hurzeler, R., & Horwitz, S. M. (2002). Physician’s ratings of their knowledge, attitudes, and endof-life-care practices. Academic Medicine, 77 305-311. Bradley, E. H., McGraw, S. A., Curry, L., Buck ser, A., King, K. L., Kasl, S. V., et al. (2002). Expanding the Andersen model: The role of psychosocial factors in longterm care use. Health Services Research, 37 1221-1242. Burrs, F. A. (1995). The African American expe rience: Breaking the barriers to Hospices. The Hospice Journal, 10 15-18. Campbell, D. E., Lynn, J., Louis, T. A., & Shugarman, L. R. (2004). Medicare program expenditures associated with hospice use. Annals of Internal Medicine, 140, 269277. Caralis, P.V., Davis, B., Wright, K., & Marcia l, E. (1993). The influe nce of ethnicity and race on attitudes toward advance dire ctives, life-prolonging treatments, and euthanasia. The Journal of Clinical Ethics, 4, 155-165. Casarett, D. J., Crowley, R. L., & Hirsch man, K. B. (2004). How should clinicians describe hospice to patients and families? Journal of the American Geriatrics Society, 52, 1923-1928. Casarett, D. J., Crowley, R., Stevenson, C., Xie, S., & Teno, J. (2005). Making difficult decisions about hospice enrollment: What do patient and families want to know? Journal of the American Geriatrics Society, 53 249-254. Casarett, D. J., Karlawish, J., Marales, K ., Crowley, K., Mirsch, T., & Asch, D. A. (2005). Importance of the use of hos pice services in nursing homes: A randomized controlled trial. Jounal of the American Me dical Association, 294, 211-217. Centers for Medicare and Me dicaid Services. (2003). National health care by type of expenditure: Calendar year 2001 Washington, DC: Office of the Actuary, U.S. Bureau of the Census. Chen, H., Haley, W. E., Robinson, B. E., & Schonwetter, R. S. (2003). Decisions for hospice care in patients with advanced cancer. Journal of the American Geriatrics Society, 51 789-797. Chen, H.J., Chen, R., & Mehra, S. (2005). Raci al and Ethnic Disparities in Health Service Use and Unmet Health Needs Among Me dicaid Recipients. Presentation to Florida Agency for Health Care Ad ministration, June 3, 2005, Tallahassee, Florida.

PAGE 119

109 Christakis, N. A., & Escarce, J. J. (1996) Survival of Medicare patients following enrollment in hospice programs. New England Journal of Medicine, 335 172-178 Christakis, N. A., & Lamont, E. B. (2000). Ex tent and determinants of error in doctor’s prognoses in terminally ill pa tients: Prospective cohort study. British Medical Journal, 320, 469-472. Christakis, N. A., & Iwashyna, T. J. (2000). Th e impact of individual and market factors on the timing of initiation of hospice terminal care. Medical Care, 38, 528-541. Christakis, N. A., & Iwashyna, T. J. (2003). Th e health impact of h ealth care on families: A matched cohort study of hospice use by decedents and mortality outcomes in surviving, widowed spouses. Social Science & Medicine, 57, 465-75. Cicirelli, V. G. (1997). Relationship of ps ychosocial and background variables to older adults’ end-of-life decisions. Psychology and Aging, 12, 72-83. Copp, G. (1996). Facing impending death: The experie nces of patients and their nurses in a hospice setting. Unpublished doctoral dissertation, Oxford Brooks University, Oxford. Copp, G. (1998). A review of current theories of death and dying. Journal of Advanced Nursing, 28 382-390. Corr, C. A. (1992). A task-based approach to coping with dying, Omega, 24 81-94. Coulton, C., & Frost, A. K. (1982). Use of social and health se rvices by the elderly. Journal of Health and Social Behaviors, 23 330-339. Counte, M., & Glandon, G. (1991). A panel study of life stress, social support, and the health services utiliz ation of older persons. Medical Care, 29 348-361. Crawley, L., Payne, R., Bolden, J., Payne, T., Washington, P., & Williams, S. (2000). Palliative and end-of-life care in the African American community. Journal of the American Medical Association, 284 2518-2521. Creditor, M. (1993). Hazards of hospitalization of the elderly. Annals of Internal Medicine, 118, 219-223. Culler, S.D., Callahan, C. M., & Wolinsky, F. D. (1995). Predicting hospital costs among older decedents over time. Medical Care, 33, 1089-1105. Degenholtz, H. B., Arnold, R. A., Meisel, A ., & Lave, J. R. (2002). Persistence of racial/ethnic disparities in advance ca re plan documents among nursing home residents. Journal of the American Geriatrics Society, 50, 378-381.

PAGE 120

110 Deyo, R. A., Cherkin, D. C., & Ciol, M. A. (1992). Adapting a clinic al comorbidity index for use with ICD-9-CM Ad ministrative databases. Journal of Clinical Epidemiology, 45 613-619. Dilworth-Anderson, P., Williams, I. C., & Gibson, B. E. (2002). Issues of race, ethnicity, and culture in caregiving resear ch: A 20-year review (1980-2000). The Gerontologist, 42, 237-272. Dunlop, D. D., Manheim, L. M., Song, J ., & Chang, R. W. (2002). Gender and ethnic/racial disparities in health care utilization among older adults. Journal of Gerontology: Social Sciences, 57B S221-S233. Eleazer, G. P., Hornung, C. A., Egbert, C. B., E gbert, J. R., Eng, C., Hedgepeth, J., et al. (1996). The relationship between ethnicity a nd advance directives in a frail older population. Journal of the American Geriatrics Society, 44, 938-943. Emanuel, E. J. (1996). Cost savings at th e end of life: What do the data show? Journal of the American Medical Association, 275 1907-1914. Engle, V. F. (1998). Care of the living, care of the dying: Rec onceptualizing nursing home care. Journal of the American Geriatrics Society, 46, 1172-1174. Enguidanos, S., Yip, J., & Wilber, K. (2005). Ethn ic variation in site of death of older adults dually eligible fo r Medicaid and Medicare. Journal of the American Geriatrics Society, 53 1411-1416. Ersek, M., Kawaga-Singer, M., Barnes, D ., Blackhall, L., & Koenig, B. A. (1998). Multicultural considerations in the use of advance directives. Oncology Nursing Forum, 25, 1683-1690. Evans, R. G., & Stoddart, G. L. (1990). Produc ing health, consuming health care. Social Science and Medicine, 31 1347-1363. Evashwick, C., Rowe, G., Diehr, P., & Branc h, L. (1984). Factors explaining the use of health care services by the elderly. Health Services Research, 19, 357-382. Field, M.J., & Cassel, C.K. (Eds.). (1997). Approaching death: Improving care at the end-of-life Washington, D.C.: National Academy Press. Fisher, E.S., Whaley, F.S., Krushat, M., Male nka, D.J., Fleming, C., Baron, J.A. et al., (1992). The accuracy of Medicare’s hospita l claims data: Progress has been made, but problems remain, American Journal of Public Health, 82, 243–248. Foster, E. M. (2003). Propensity score ma tching: An illustrative analysis of dose response. Medical Care, 41 1183-1192.

PAGE 121

111 Freedman, V. (1993). Kin and nur sing home lengths of stay: A backward recurrence time approach. Journal of Health and Social Behavior, 34 138-152. Fried, T., & Mor, V. (1997). Frailty and hos pitalization of long-te rm stay nursing home residents. Journal of the American Geriatrics Society, 45, 265-269. Friedman, S.M., Steinwachs, D.M., Rathouz P.J., Burton, L.C., & Mukamel, D.B. (2005). Characteristics predicting nursi ng home admission in the Program of AllInclusive Care for Elderly People. The Gerontologist, 45 (2), 157-166. Gage, B., Miller, S. C., Coppola, K., Harve ll, J., Laliberte, L., Mor, V., & Teno, J. (March, 2000). Important questions fo r hospice in the next century. The MEDSTAT Group. Prepared for the U.S. Department of Health and Human Services. Contract number: #100-97-0010. Accessed on August 15, 2005 at http://aspe.hhs.gov/daltcp/reports/impques.htm Gallo, W. T., Baker, M. J., & Bradley, E. H. (2001). Factors associated with home versus institutional death among cancer patients in Connecticut. Journal of the American Geriatrics Society, 49, 771-777. Gallup, G. J. (2000). Spiritual beliefs and the dying process: A report on a national survey. The Nathan Cummings Foundation and Fetzer Institute. Garrett, J. M., Harris, R. P., Norburn, J. K ., Patrick, D. L. & Danis, M. (1993). Lifesustaining treatments during termin al illness: Who wants what? Journal of General Internal Medicine, 8 361-368. Gessert, C. E., Curry, N. M., & Robinson, A. (2001). Ethnicity and end-of-life care: The use of feeding tubes. Ethnicity and Disease, 11 97-106. Gaugler, J.E., Kane, R.L., Kane, R.A., Clay, T., & Newcomer, R. (2003). Caregiving and institutionalization of cognitively impa ired older people: Utilizing dynamic predictors of change. The Gerontologist, 43 (2), 219-229. General Accounting Office. (2000). Report to congressional requesters: Medicare: More beneficiaries use hospice, but for fewer days of care. GAO/HEHS-00-182. September 18, 2000. Retrieved on May 1, 2005 at www.gao.gov/archive/2000/he00182.pdf. George, L. K. (2002). Research design in end-of-life research: State of science. The Gerontologist, 42 (Special Issue, 3), 86-98. Glaser, B. G., & Strauss, A. L. (1965). Awareness of dying New York: Aldine Publishing Glaser, B. G., & Strauss, A. L. (1968). Time for dying Chicago: Aldine Publishing.

PAGE 122

112 Gorden, A. K. (1996). Hospice and minorities: A national study of organizational access and practice. The Hospice Journal, 11, 49-70. Gordon, A. K. (1995). Deterrents to access a nd service for blacks and Hispanics: The Medicare hospice benefit, healthcar e utilization, and cultural barriers. Hospice Journal, 10 65-83. Grabbe, L., Demi, A. S., Whittington, F., Jone s, J. M., Branch, L. G., & Lambert, R. (1995). Functional status and the use of formal home care in the year before death. Journal of Aging and Health, 7, 339-364. Greer, D. S., Mor, V., Morris, J. N., Sher wood, S., Kidder, D., & Birnbaum, D. (1986). An alternative in terminal care: Results of the National Hospice Study. Journal of Chronic Diseases, 36 737-780. Greiner, K. A., Perera, S., & Ahluwalia, J. S. (2003). Hospice usage by minorities in the last year of life: Results from th e National Mortality Followback Survey, Journal of the American Geriatrics Society, 51, 970-978. Hareven, T.K. (2001). Historical perspectives on aging and family relations. In R.H. Binstock & L.K. George (Eds.), Handbook of aging and the social sciences 5th ed (pp. 141-159). New York: Academic Press. Hing, E., Sekscenski, E., & Strahan, G. ( 1989). The National Nursing Home Survey: 1985 Summary for the United States. Vital and Health Statistics, Series 13 number 97, DHHS Publication num ber 89-1758. Washington, DC: US Government Printing Office. Hoffmann, D. E.,& Tarzian, A. J. (2005). Dyi ng in America: An examination of policies that deter adequate end-of -life care in nursing homes. The Journal of Law, Medicine, & Ethics, 33 294-309. Hogan, C., Lunney, J., Gabel, J., & Lynn, J. (2001) Medicare beneficiaries’ cost of care in the last year of life. Health Affairs, 20, 188-195. Hogan, C., Lynn, J., Gabel, J., Lunney, J., O’Mara, A., & Wilkinson, A. (May 1, 2000). Medicare beneficiaries’ costs and use of care in the last year of life: Submitted to the Medicare Payment Advisory Commission Retrieved October 14, 2004, from the MediCaring Web site: http://www.medicaring.org/e ducate/download/medpac.pdf. Hopp, F. P. & Duffy, S. A. (2000). Raci al variations in end-of-life care. Journal of the American Geriatrics Society, 48 658-663.

PAGE 123

113 Hopp, F. P. (2000). Preferences for surrogate decision makers, informal communication, and advance directives among community dwelling older adults : Results from a national study. The Gerontologist, 40, 449-457. Hoyert, D. L., & Rosenberg, H. M. (1999). Mortality from Alzheimer’s disease: An update. National Vital Statistics Reports, 47 1-8. Hoyert, D. L., Kochanek, K. D., & Murphy, S. L. (1999). Final data for 1997. National Vital Statistics Reports, 47, 6-30. Hyatts ville, Maryland: National Center for Health Statistics. Iezzoni, L. I. (Ed.). (1997). Risk adjustment for measur ing health care outcomes (2nd ed.). Chicago: Health Administration Press. Iezzoni, L. I. (2002). Using administrative data to study persons with disabilities. The Milbank Quarterly, 80, 347-379. Iezzoni, L. I., Daley, J., Heeren, T., Foley, S. M., Fisher, E. S., Duncan, C. et al. (1994). Identifying complications of care using administrative data. Medical Care, 32, 700-715. Iwashyna, T. J., Brennan, G., Zhang, J. X., & Christakis, N. A. (2002). Finding married couples in Medicare claims data. Health Services & Outcomes Research Methodology, 3, 75–86. Iwashyna, T. J., Chang, V. W., Zhang, J. X., & Christakis, N. A. (2002). The lack of effect of market structure on hospice use. Health Services Research, 37, 15311551. Iwashyna, T. J., Zhang, J. X., & Christakis, N. A. (2002). Disease-specific patterns of hospice and related health care use in an incidence cohort of seriously ill elderly patients. Journal of Palliative Medicine, 5, 531-538. Jones, B., Nackerud, L., & Boyle, D. (1997). Di fferential utilization of hospice services in nursing homes. Hospice Journal, 12 41-57. Kalish, R. A., & Reynolds, D. K. (1976). Death and ethnicity: A psychocultural study New York: Baywood. Kane, R. L., Wales, J., Berstein, L., Lwibow itz, A. et al. (1984). A randomized controlled trial of hospice care. Lancet, 8282 890-894. Kart, C. S. (1991). Variation in long-te rm care services use by aged blacks. Journal of Aging and Health, 3, 511-526.

PAGE 124

114 Kidder, D. (1992). The effects of hospi ce coverage on Medicare expenditures. Health Services Research, 27, 195-217. Kiely, D. K., Mitchell, S. L., Marlow, A., Mu rphy, K. M., & Morris, J. N. (2001). Racial and state differences in the designation of advance directives in nursing home residents. Journal of the American Geriatrics Society, 49, 1346-1352. Klessig (1992). Cross-cultural medicine a de cade later (special issue): The effect of values and culture on life-support decisions. The Western Journal of Medicine, 157, 16-322. Krakauer, E. L., Crenner, C., & Fox, K. (2002) Barriers to optimum end-of-life care for minority patients. Journal of the American Geriatrics Society, 50, 182-190. Kubler-Ross, E. (1969). On death and dying New York: McMillan. Kwak, J., & Haley, W. E. (2005). Current re search findings on end-of-life decision making among racially/et hnically diverse groups. The Gerontologist, 45, 634641. Lackan, N. A., Ostir, G. V., Freeman, J. L., Kuo, Y., Zhang, D. D., & Goodwin, J. S. (2004). Hospice use by Hispanic and nonHispanic White Cancer decedents. Health Services Research, 39, 969-983. Lackan, N. A., Ostir, G. V., Freeman, J. L., Mahnken, J. D., & Goodwin, J. S. (2004). Decreasing variation in th e use of hospice among olde r adults w ith breast, colorectal, lung, and prostate cancer. Medical Care, 42 116-122. Lamont, E. B., & Christakis, N.A. (2001). Prog nostic disclosure to patients with cancer near the end of life. Annals of Internal Medicine, 134, 1096-1105. Levinsky, N. G., Ash, A. S., Yu, W., & Moskow itz, M. A. (1999). Patterns of use of common major procedures in me dical care of older adults. Journal of the American Geriatrics Society, 47, 553-558. Lewis-Beck, M. S. (1980). Applied regression : An introduction. Newbury Park: Sage Publications. Lorenz, K. A., Asch, S. M., Rosenfeld, K. E., Liu, H., & Ettner, S. L. (2004). Hospice admission practices: Where does hospice fit in the continuum of care? Journal of the American Geriatrics Society, 52, 725-730. Lubitz, J. D., & Riley, G. F. (1993). Trends in Medicare payments in the last year of life. New England Journal of Medicine, 328 1092-1096.

PAGE 125

115 Lunney, J. R., Lynn, J., & Hogan, C. (2002). Profiles of older Medicare decedents. Journal of the American Geriatrics Society, 50 1108-1112. Lunney, J. R., Lynn, J., Foley, D. J., Lipson, S., & Guralnik, J. M. (2003). Patterns of functional decline at the end of life. Journal of the American Medical Association, 289, 2387-2392. Lynn, J. (2001). Serving patients who may die soon and their families: The role of hospice and other services. Journal of the American Medical Association, 285, 925-932. Mandelblatt, J., Andrews, H., Kao, R., Wallace, R., & Kerner, J. (1996). The late-stage diagnosis of colorectal cancer: Dem ographic and socioeconomic factors. American Journal of Public Health, 86 1794-1797. McCarthy, E. P., Iezzoni, L. I., Davis, R. B., Palmer, R. H., Cahalane, M., Hamel, M. B. et al. (2000). Does clinical evidence support ICD-9-CM diagnosis coding of complications? Medical Care, 38, 868-876. McGlynn, E. A., Damberg, C., Kerr, E. A., & Brook, R. H. (Eds.). (1998). Health information systems: Design i ssues and analytic applications. Santa Monica: Rand. McKinley, E.D., Garrett, J. M., Evans, A. T. & Danis, M. (1996). Differences in end-oflife decision making among black and white ambulatory cancer patients. Journal of General Internal Medicine, 11 651-656. Medicare Payment Advisory Commission. (2002). Report to the Congress: Medicare beneficiaries’ access to hospice Retrieved November 1, 2004, from http://www.medpac.gov/publications/congr essional_reports/may2002_HospiceAc cess.pdf. Medicare Payment Advisory Commission. (2004). Report to the Congress: New approaches in Medicare. Re trieved November 1, 2004, from http://www.medpac.gov/publications/congressional_ reports/June04_Entire_Report.pdf. Medicare Payment Advisory Commission. (2001). Report to Congress: Medicare in rural America. Retrieved November 1, 2004, from http://www.medpac.gov/publications/congressional_ reports/June01_Entire_Report.pdf. Mezey, M., Dubler, 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? The Gerontologist, 42 54-67.

PAGE 126

116 Miller, S. C. (2004). Hospice care in nursing hom es: Is site of care associated with visit volume? Journal of the American Geriatrics Society, 52, 1331-1336. Miller, S. C., Gozalo, P., & Mor, V. (2000). Use of Medicare’s hospice benefit by nursing facility residents. Report 5. In: Synthesis and an alysis of Medicare’s hospice benefit. Washington, DC: Office of Disabi lity, Aging and Long Term Care Policy in the Office of Assistant Secretary for Planning and Evaluation, U.S. Department of Health and Human Services. Miller, S. C., Gozalo, P., & Mor, V. (2001) Hospice enrollment and hospitalization of dying nursing home residents. American Journal of Medicine, 111 38-44. Miller, S. C., Intrator, O., Gozalo, P., Roy, J ., Barber, J., & Mor, V. (2004). Government expenditures at end-of-life short a nd long-stay nursing home residents: Differences by hospice enrollment. Journal of the American Geriatrics Society, 52 1284-1292. Miller, S. C., Intrator, O., Laliberte, L., & Cang, Y. (2004). End-of-life care for Medicare skilled nursing facility residents: Pa tterns of care and expenditures by hospice enrollment status. Journal of the American Geriatrics Society, S193 P513. Miller, S. C. & Lima, J. (2004). The Florida Model of Hospice Care: A report for Florida Hospices and Palliative Care, Inc Miller, S. C. & Mor, V. (2002). The role of hospice care in the nursing home setting. Journal of Palliative Medicine, 5 271-277. Miller, S.C., Mor, V., Wu, N., Gozalo, P., & Lapane, K. (2002). Does Receipt of Hospice Care in Nursing Homes Improve the Mana gement of Pain at the End-of-Life? Journal of the American Geriatrics Society, 50, 507-515. Mitchell, S. L., Morris, J. N., Park, P. S., & Fries, B. E. (2004). Terminal care for persons with advanced dementia in the nu rsing home and home care settings. Journal of Palliative Medicine, 7, 808-816. Moinpour, C. M., & Polissar, L. (1989). Factor s affecting place of death of hospice and non-hospice cancer patients. American Journal of Public Health, 79, 1549-1551. Moon, M., & Boccuti, C. (Summer, 2002). Medicare and End-of-Life Care Washington DC: Last Acts. Mor, V., & Kidder, D. (1985). Cost savings in hospice: Final results of the National Hospice Study. Health Services Research, 20, 407-422.

PAGE 127

117 Mor, V., Papandonatos, G., & Miller, S.C. ( 2005). End-of-life hospitalization for African American and non-Latino white nursing hom e residents: Variation by race and a facility’s racial composition. Journal of Palliative Medicine, 8 58-68. Mor, V., Zinn, J., Angelelli, J., Teno, J. M., & Miller, S. C. (2004). Driven to tiers: Socioeconomic and racial disparities in the quality of nursing home care. The Milbank Quarterly, 82, 227-256. Mui, A. C., & Burnette, D. (1994). Long-term care service use by frail elders: Is ethnicity a factor? The Gerontologist, 34 190-198. Mukamel, D. B., Bajorska, A., & Temkin-G reener, H. (2002). H ealth care services utilization at the end of life in a ma naged care program integrating acute and long-term care. Medical Care, 40 1136-1148. Murphy, S. T., Palmer, J. M., Azen, S., Frank, G., Michel, V. & Blackhall, L. J. (1996). Ethnicity and advance directives. Journal of Law and Medical Ethics, 24 108117. Mutran, E., & Ferraro, K. (1988). Medical need and use of services among older men and women. Journal of Gerontology: Social Sciences, 43 S162-S171. National Hospice and Pallia tive Care Organization (2002) 2002 NHPCO National Data Set Summary Report. Available at http://www.nhpco.org/files/members/2002Na tionalDataSet.pdf Accessed on October 4, 2005. National Hospice and Palliativ e Care Organization (2003). Hospice fact Sheet Arlington, VA: NHPCO. National Hospice and Palliativ e Care Organization (2005). Hospice fact Sheet Arlington, VA: NHPCO. National Hospice Organization (1996a). Press release: New findings address escalating end-of-life debate. Arlington, VA: NHO. National Hospice Organization (1996b). Hospice Fact Sheet Arlington, VA: NHO. Neubauer, B. J., & Hamilton, C. L. (1990). Raci al differences in attitudes toward hospice care. Hospice Journal, 6 37-48. O’Brien, L. A., Grisso, J. A., Maislin, G., La Pann, K., Krotki, K. P., Greco, P.J., et al. (1995). Nursing home residents’ preferen ces for life-sustaining treatments. Journal of the American Medical Association, 274, 1775-1779.

PAGE 128

118 Parker-Oliver, D., Porock, D., Zweig, S., Ra ntz, M., & Petroski, G. F. (2003). Hospice and nonhospice nursing home residents. Journal of Palliative Medicine, 6 69-75. Pattison, E. M. (1977 ). The experience of dying Englewood, NJ: Prentice-Hall. Phillips, R. S., Hamel, M. B., Teno, J. M., Soukup, J., Lynn, J., Califf, R. et al. For the SUPPORT Investigators. (2000). Patient race and decisions to withhold or withdraw life-sustaining treatments for seriously ill hospitalized adults. American Journal of Medicine, 108, 14-19. Phipps, E., True, G., Harris, D., Chong, U., Tester, W., Chavin, S. I. et al. (2003). Approaching the end of life: Attitudes, preferences, and behaviors of AfricanAmerican and White patients a nd their family caregivers. Journal of Clinical Oncology, 21, 549-554. Porter, E. J. (2000). Research on home care ut ilization: A critical analysis of the preeminent approach. Journal of Aging Studies, 14 25-38. Pritchard, R. S., Fisher, E. S., Teno, J. M., Sh arp, S. M., Reding, D. J., Knaus, W. A. et al. (1998). Influence of patient preferences and local health system characteristics on the place of death. Journal of the American Geriatrics Society, 46, 1242-1250. Reese, D. J., Ahern, R. E., Nair, S., Sc hrock, J. D. O., & Warren-Wheat, C. (1999). Hospice access and use by African Am ericans: Addressing cultural and institutional barriers through pa rticipatory action research. Social Work, 44, 549559. Riley, G., Lubitz, J., Prihoda, R., & Rabey, E. (1987). The use and costs of Medicare services by cause of death. Inquiry, 24, 233-244. Rosenbaum, P., & Rubin, D. B. (1983). The cen tral role of the propensity score in observational studies for causal effects. Biometrika, 70, 41-55. Rubin, D. B. (1997). Estimating causal effect s from large data sets using propensity scores. Annals of Internal Medicine, 127 757-763. Salive, M E., Collinas, K. S., Foley, D. J., & George, L. K. (1993). Predictors of nursing home admission in a biracial population. American Journal of Public Health, 83 1765-1767, Sanders, A., & Morley, J. (1993). The olde r person and the emergency department. Journal of the American Geriatrics Society, 41, 880-882.

PAGE 129

119 Schulman-Green, D., McCorkle, R., Cherlin, E ., Johnson-Hurzeler, R., & Bradley, E. H. (2005). Nurses’ communication of pr ognosis and implications for hospice referral: A study of nurses caring for te rminally ill hospitalized patients. American Journal of Critical Care, 14 64-70. Seidlitz, L., Duberstein, P. R ., Cox, C., & Conwell, Y. (1995). Attitudes of older people toward suicide and assisted suicide: An analysis of Gallup Poll Findings. Journal of the American Geriatrics Society, 43 993-998. Somova, M. J., Somov, P. G., Lawrence, J. C ., & Frantz, T. T. (2000). Factors associate with length of stay in a mid-sized, urban hospice. American Journal of Hospice and Palliative Care, 17, 99-105. Spector, W. D., & Mor, V. (1984). Utilziation and charges for terminal cancer patients in Rhode Island. Inquiry, 21 328-337. Stevens, A., Owen, J., Roth, D., Clay, O., Bart olucci, A., & Haley, W. (2004). Predictors of time to nursing home placement in Wh ite and African American individuals with dementia. Journal of Aging and Health, 16 (3), 375-397. Stump, T.E., Johnson, R.J., & Wolinsky, F.D. (1995). Changes in physician utilization over time among older adults. Journal of Gerontology: Social Sciences, 50, S45S58. Tabachnick, B. G., & Fidell, L. S. (2001). Using multivariate statistics (4th ed.). Needham Heights, MA: Allyn and Bacon. Teno, J. M., Clarridge, B. R., Casey, V., Welch, L. C., Wetle, T., Shield, R. et al. (2004). Family perspectives on end-of-life care at the last place of care. Journal of the American Medical Association, 291, 88-93. Teno, J. M., Weitzen, S. Fennerl, M. L., & Mor, V. (2001). Dying traj ectory in the last year of life: Does cancer tr ajectory fit other diseases? Journal of Palliative Medicine, 4, 457-464. Tilden, V. P., Tolle, S., Drach, L., & Hickma n, S. (2002). Measurement of quality of care and quality of life at the end of life. The Gerontologist, 42 (Special Issue, 3), 7180. Tilden, V. P., Tolle, S., Drach, L., & Hickma n, S. (2002). Measurement of quality of care and quality of life at the end of life. The Gerontologist, 42 (Special Issue, 3), 7180. U.S. Census Bureau. (2000 ). Projections of the total re sident population by 5-year age groups, U S. Census 2002 race, and Hispanic origin with special age categories:

PAGE 130

120 Middle 2050-2070. Retrieved January 30, 2003, from http://www.census.gov/population/projec tions/nation/summary/np-t4-a.txt. Virnig, B. A., Kind, S., McBean, M., & Fish er, E. S. (2000). Geogr aphic variation in hospice use prior to death. Journal of the American Geriatrics Society, 48, 11171125. Virnig, B. A., Moscovice, I. S., Durham, S. B., & Casey, M. M. (2004). Do rural elders have limited access to Medicare hospice services? Journal of the American Geriatrics Society, 52 731-735. Virnig, B., Persily, N., Morgan, R., & De Vito, C. (1999). Do Medicare HMOs and Medicare FFS differ in their use of the Medicare hospice benefit? The Hospice Journal, 14, 1-12. Virnig, B.A., McBean, A. M., Kind, S., & Dholak ia, R. (2002). Hospice use before death: variability across cancer diagnosis. Medical Care, 20, 73-78. Wallace, S. P., Levy-Storms, L., Kington, R. S., & Andersen, R. M. (1998). The persistence of race a nd ethnicity in the use of long-term care. Journal of Gerontology: Social Sciences, 53B S104-S112. Waters, C. M. (2001). Understanding and suppor ting African Americans’ perspectives of end-of-life care planni ng and decision making. Qualitative Health Research, 11 385-398. 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, 323-335. Welch, L. C., Teno, J. M., & Mor, V. (2005). End-of-life care in black and white: Race matters for medical care of dying patients and their families. Journal of the American Geriatrics Society, 53 1145-1153. White-Means, S. I., & Rubin, R. M. (2004). Is there equity in the home health care market? Understanding racial patterns in the use of formal home health care. Journal of Gerontology: Social Sciences, 59B S220-S229 World Health Orgnization. (2004). Better palliative care for older people Milan: World Health Organization. Retrie ved October 10, 2004, from http://www.euro.who.int/document/E82933.pdf. Yang, Z., Norton, E. C ., & Stearns, S. C. (2002). Longevity and health care expenditures: The real reas ons older people spend more. Journal of Gerontology: Social Sciences, 55B S2-10.

PAGE 131

121 Zerzan, J., Stearn, S. & Hanson, L. (2000). Access to palliative care and hospice in nursing homes. Journal of the American Medical Association, 284 2489-2494.

PAGE 132

122 Appendices

PAGE 133

123 Appendix A: Flow Chart of the Study Population Selection All Medicaid Decedents N=118,703 Medicaid Decedents with Death Certificate Records and Manner of Death Being ‘Natural’ or ‘Not Stated’ N=117,667 Dual-Eligible N=67,562 Aged 65 and Older N=61,137 Main Study Population: LTNH and STNH Residents who were either Non-Hispanic Black or Non-Hispanic White N=30,765

PAGE 134

124 Appendix B: Determining Nursing Home Status 1 HSPC R&B: Hospice Room & Board 2 IMC : Intermediate Care Units Month variable 3 NH Month: Nursing Home Month variable 4 LTNH: Long-Term Nursing Home 5 STNH : Short-Term Nursing Home Determining NH Status (n=117,667) Any HSPC R&B 1 Claims (n=19.474) NH 3 Month 10, 11, 12 = 1 (LTNH 4) IMC 2 Units For Month 10, 11, or 12 >=1 (STNH 5) Yes, n=19,996 No, n=66,780 Yes, n=24,012 No, n=93,655 With Any Preceding NH Mo.: n=1,324 Without Any Preceding NH Mo.: n=5,801 With 2 Preceding NH Mo. >=1 n=6,077 With 2 Preceding NH Mo.=1 or IMC Mo>=28 n=47 With 1 Preceding NH Mo.=1 and 1 IMC Mo >=1 n=364 With 2 Preceding IMC Months Units >=1 n=391 Individuals with any HSPC R&B Claims identified as a nursing home residents: Group 1, 2, 3, & 4 (n=13,673) Any HSPC R&B Claims n=3,250 Any HSPC R&B Claims n=16,224 Any HSPC R&B Claims n=2,220 Any HSPC R&B Claims n=7,125 Group 1 Group 2 Group 5 Group 3 Group 4

PAGE 135

125 About the Author Jung Kwak received her Bachelor’s of Sc ience Degree in Business Administration from University of South Carolina in August of 1998 and Master of Social Work Degree from University of South Carolina in Ma y of 2001. She entered the Ph.D. in Aging Studies program at the University of S outh Florida in the Fall of 2001. Ms. Kwak’s scholarly work include research article s published in journals including The Gerontologist and the Journal of Pain a nd Symptom Management, and a number of technical reports prepared fo r Florida Agency for Health Care Administration. Ms Kwak is also a recipient of AARP Scholarship, Gr antmakers in Aging Fellowship, Institute on Aging Fellowship and Center for Hospice, Pa lliative Care, and Endof-Life Studies Pilot Grant.