Abstract
Context:
Hospice utilization is an end-of-life quality indicator. The Deep South has known disparities in palliative care that may affect hospice utilization.
Objectives:
To evaluate the association among Deep South patient and hospital characteristics and hospice utilization.
Methods:
This retrospective cohort study evaluated patient and hospital characteristics associated with hospice among Medicare cancer decedents aged ≥65 in 12 southeastern cancer centers between 2012-2015. We examined patient-level characteristics (age, race, gender, cancer type, and received patient navigation) and hospital-level characteristics (board-certified palliative physician, inpatient palliative care beds, and hospice ownership). Outcomes included hospice (within 90 vs 3 days of death). Relative risks (RR) and 95% confidence intervals evaluated the association between patient- and hospital-level characteristics and hospice outcomes using generalized log-linear models with Poisson distribution and robust variance estimates.
Results:
Of 12,725 cancer decedents, 4,142 (33%) did not utilize hospice. “No hospice” was associated with non-white (RR 1.24, 95% CI 1.17, 1.32) and non-navigated patients (RR 1.17, 95% CI 1.10, 1.25), and those at a hospital with inpatient palliative care beds (RR 1.15, 95% CI 1.10, 1.21). “Late hospice” (20%; n=1,458) was associated with being male (RR 1.31, 95% CI 1.19, 1.44) and seen at a hospital without inpatient palliative care beds (RR 0.82, 95% CI 0.75, 0.90).
Conclusion:
Hospice utilization differed by patient and hospital characteristics. Patients who were non-white, and non-navigated; and, hospitals with inpatient palliative care beds were associated with no hospice. Research should focus on ways to improve hospice utilization in Deep South older cancer patients.
Keywords: Hospice, Oncology, South, Disparities, Geriatric, Medicare
Introduction
Despite substantial growth of United States (U.S.) palliative care programs, opportunities remain for improving the quality of end-of-life care. Poor quality end-of-life care includes a high number of hospital and ICU days, low proportion of patients enrolled in hospice, and short hospice enrollment (within 3 days of death).(1-3) This poor quality was seen nationwide in 2011 when nearly one third of all hospice patients had short hospice stays, resulting in increased suffering for patients and their families.(2, 4)
Significant disparities in U.S. end-of-life care exist by geographic region and race.(5) Residents in Deep South states have particularly suboptimal hospice use patterns. In Alabama, Georgia, Louisiana, Mississippi, and South Carolina, the proportion of extended hospice enrollment and disenrollment were in the fourth quartile nationally,(2, 3) indicating that patients and families are not receiving the support they need at the appropriate time.(2) These observed disparities may be related, in part, to differences in demographics. The African American population is concentrated in the Deep South. Though hospice utilization is low for all racial groups, literature has shown that hospice utilization is much lower for African Americans, with 33% of African American decedents versus 44% of white decedents utilizing hospice prior to death.(6) Furthermore, when African Americans utilize hospice services, they have higher rates of hospice disenrollment and concerns about care than whites, such as care that is discordant with preferences, poor provider communication, and disparities in the assessment and treatment of pain.(7)
Institutional factors also contribute to disparities in hospice use.(2, 8, 9) Despite a tripling of palliative care use over the past two decades, only a fraction of appropriate patients received palliative care services.(8) Institutional measures of end-of-life care quality include the number of board-certified palliative care physicians, the number of inpatient palliative beds, and ownership of the inpatient hospice facility, i.e. whether the hospital or an independent organization owns the inpatient hospice facility, which may include a hospice unit or designated hospice beds.(9)
Where access to hospice care is limited, novel approaches to providing patient support are needed. Patient navigation is a patient-centric health care delivery system that establishes one-to-one relationships between the patient and navigator to minimize care barriers such as timely access to cancer diagnosis and treatment.(10-12) However, the benefits of navigation on end-of-life outcomes are still uncertain. Therefore, we sought to better understand the patient and institutional factors associated with hospice use by examining a geriatric cancer population of Medicare beneficiaries in the Deep South during implementation of the University of Alabama at Birmingham (UAB) Cancer Care Network (CCN). These findings may be useful in identifying potential intervention targets to address disparities and improve health outcomes.
Methods
Study design and sample
To examine the hospice utilization patterns in the Deep South, we conducted a retrospective cohort study using data collected as part of a Center for Medicare and Medicaid Services (CMS) Innovation demonstration project from the UAB CCN. The CCN was established to improve patient outcomes through enhanced cancer care services using navigation and has been fully described elsewhere.(13-16) The UAB CCN includes 12 community hospitals and cancer centers located within five southeastern states, including Alabama, Georgia, Florida, Mississippi, and Tennessee located in rural and urban settings with 2 to 58 affiliated oncologists (Figure 1).(13) The patient sample included Medicare decedents age ≥65 diagnosed during or after 2008 with any type of cancer from who had received care within the UAB CCN facilities from 2012-2015. Data sources included local tumor registries linked to electronic medical records and claims abstracted from the CMS Chronic Condition Data Warehouse. Only patients with continuous Medicare Part A and B data and no health maintenance organization were included. This study was approved by the UAB Institutional Review Board.
Figure 1.
University of Alabama at Birmingham Cancer Community Network (13)
Outcomes
The primary study outcome was hospice utilization, determined by any Medicare claim of hospice within 90 days prior to death. The secondary outcome was late hospice utilization, defined as the first claim for hospice within three days prior to death.
Exposures
To explore patient and hospital characteristics associated with hospice utilization, multiple factors were considered. Patient-level characteristics included race (white, non-white), sex (male, female), age at death, cancer type (breast, lung, gastrointestinal, genitourinary, other), and navigation status (any receipt of navigation, no navigation). Race, sex, and age were abstracted from Medicare claims data. Cancer type, receipt of navigation services, and all hospital-level factors were obtained from local tumor registry-linked electronic medical record data. Hospital-level characteristics included the presence of a board-certified palliative care physician, dedicated inpatient palliative care beds, and hospital ownership of an in-patient hospice facility.
Statistical Analysis
Descriptive statistics were calculated using means (standard deviations) for continuous variables and frequencies (percentages) for categorical variables. To examine patient and hospital characteristics associated with hospice utilization, we estimated relative risks (RR) and 95% confidence intervals (CI) using generalized log-linear models with a Poisson distribution and robust variance estimates. In subset analyses of patients with any hospice utilization, RR and 95% CIs were used to determine patient and hospital characteristics associated with receipt of late hospice. We examined unadjusted models only due to the exploratory nature of this analysis.(17)
Results
The sample consisted of 12,725 UAB CCN Medicare decedents (average age at death = 77 [SD 7.8]); 32.6% did not utilize hospice prior to death. Table 1 shows patient- and hospital-level characteristics for the study sample. Patients who did not receive hospice care were more often non-white, non-navigated, and seen at a hospital with inpatient palliative care beds. Oxne in five patients who utilized hospice received late hospice care (n=1,458/8,583) and were more often male and seen at a hospital without inpatient palliative care beds (Table 2). The results of our unadjusted models estimating risk of no hospice compared to any hospice utilization are shown in Table 3. Non-white patients had a 24% increased risk of no hospice utilization compared to white patients (95% CI 1.17, 1.32), males had a 15% increased risk of no hospice utilization compared to females (95% CI 1.09, 1.21), and non-navigated patients had a 17% increased risk of no hospice utilization compared to navigated patients (95% CI 1.10, 1.25; Table 3). With respect to hospital-level characteristics, patients seen at hospitals with inpatient palliative care beds had a 15% increased risk for no hospice utilization compared to those without (95% CI 1.10, 1.21).
Table 1.
Patient- and hospital-level characteristics by hospice utilization (N=12,725).
| No hospice utilization n=4,142 |
Any hospice utilization n=8,583 |
|
|---|---|---|
| Mean (SD) or N (%) | Mean (SD) or N (%) | |
| Age at death | 76.5 (7.6) | 77.1 (7.7) |
| Race | ||
| White | 3313 (80.0) | 7282 (84.8) |
| Non-white | 829 (20.0) | 1301 (15.2) |
| Sex | ||
| Male | 2356 (57.2) | 4452 (52.1) |
| Female | 1765 (42.8) | 4086 (47.9) |
| Cancer type | ||
| Lung | 973 (23.5) | 2212 (25.8) |
| Gastrointestinal | 852 (20.6) | 2131 (24.8) |
| Genitourinary | 578 (14.0) | 1033 (12.0) |
| Breast | 350 (8.5) | 654 (7.6) |
| Other | 1389 (33.5) | 2553 (29.7) |
| Navigated | ||
| Yes | 857 (20.7) | 2118 (24.7) |
| No | 3285 (79.3) | 6465 (75.3) |
| Hospital with board certified palliative care physician | ||
| Yes | 3122 (75.4) | 6561 (76.4) |
| No | 1020 (24.6) | 2022 (23.6) |
| Hospital with inpatient palliative care beds | ||
| Yes | 2251 (54.4) | 4217 (49.1) |
| No | 1891 (45.7) | 4366 (50.9) |
| Hospital with ownership of hospice facility | ||
| Yes | 920 (22.2) | 1982 (23.1) |
| No | 3222 (77.8) | 6601 (76.9) |
Table 2.
Patient- and hospital-level characteristics for patients who received hospice by timeframe of hospice utilization (N=8,583).
| Late hospice utilization n=1,458 |
Any other hospice utilization n=7,125 |
|
|---|---|---|
| Mean (SD) or N (%) | Mean (SD) or N (%) | |
| Age at death | 76.6 (7.7) | 77.2 (7.7) |
| Race | ||
| White | 1268 (87.0) | 6014 (84.4) |
| Non-white | 190 (13.0) | 1111 (15.6) |
| Sex | ||
| Male | 851 (58.6) | 3601 (50.8) |
| Female | 595 (41.2) | 3491 (49.2) |
| Cancer type | ||
| Lung | 402 (27.6) | 1810 (25.4) |
| Gastrointestinal | 312 (21.4) | 1819 (25.5) |
| Genitourinary | 200 (13.7) | 833 (11.7) |
| Breast | 103 (7.1) | 551 (7.7) |
| Other | 441 (30.3) | 2112 (29.6) |
| Navigated | ||
| Yes | 388 (26.6) | 1730 (24.3) |
| No | 1070 (73.4) | 5395 (75.7) |
| Hospital with board certified palliative care physician | ||
| Yes | 1107 (75.9) | 5454 (76.6) |
| No | 351 (24.1) | 1671 (23.5) |
| Hospital with inpatient palliative care beds | ||
| Yes | 644 (44.2) | 3573 (50.2) |
| No | 814 (55.8) | 3552 (49.9) |
| Hospital with ownership of hospice facility | ||
| Yes | 342 (23.5) | 1640 (23.0) |
| No | 1116 (76.5) | 5485 (77.0) |
Table 3.
Generalized log-linear models estimating risk for no vs. any hospice utilization (N=12,725).
| Unadjusted RR (95% CI) | |
|---|---|
| Age at death (unit, SD) | 0.99 (0.99, 1.00) |
| Race, non-white | 1.24 (1.17, 1.32) |
| Sex, male | 1.15 (1.09, 1.21) |
| Cancer type, breast vs. lung | 1.14 (1.03, 1.26) |
| Cancer type, gastrointestinal vs. lung | 0.93 (0.87, 1.01) |
| Cancer type, genitourinary vs. lung | 1.17 (1.08, 1.28) |
| Cancer type, other vs. lung | 1.15 (1.08, 1.23) |
| Non-navigated | 1.17 (1.10, 1.25) |
| Board certified palliative care physician | 0.96 (0.91, 1.02) |
| Inpatient palliative care beds | 1.15 (1.10, 1.21) |
| Ownership of hospice facility | 0.97 (0.91, 1.03) |
RR=relative risk, CI=confidence interval
In patients with any hospice utilization, we estimated risk of late hospice compared to any hospice utilization (Table 4). Non-white patients had a 16% decreased risk of receiving late hospice compared to white patients (95% CI 0.73, 0.97), males had a 31% increased risk of late hospice compared to females (95% CI 1.19, 1.44), and patients seen at hospitals with inpatient palliative care beds had a 18% decreased risk of late hospice utilization when compared to hospitals without inpatient palliative care beds (95% CI 0.75, 0.90).
Table 4.
Generalized log-linear models estimating risk for late vs. any hospice utilization (N=8,583).
| Unadjusted RR (95% CI) | |
|---|---|
| Age at death (unit, SD) | 0.99 (0.98, 1.00) |
| Race, non-white | 0.84 (0.73, 0.97) |
| Sex, male | 1.31 (1.19, 1.44) |
| Cancer type, breast vs. lung | 0.87 (0.71, 1.06) |
| Cancer type, gastrointestinal vs. lung | 0.81 (0.70, 0.92) |
| Cancer type, genitourinary vs. lung | 1.07 (0.91, 1.24) |
| Cancer type, other vs. lung | 0.95 (0.84, 1.07) |
| Non-navigated | 0.90 (0.81, 1.00) |
| Board certified palliative care physician | 0.97 (0.87, 1.08) |
| Inpatient palliative care beds | 0.82 (0.75, 0.90) |
| Ownership of hospice facility | 1.02 (0.91, 1.14) |
RR=relative risk, CI=confidence interval
Discussion
This study found differences in hospice utilization for older patients with cancer living in the Deep South along the lines of race, gender, and navigation status that are consistent with other research.(18-20) Additionally, the institutional variable of a hospital having inpatient palliative care beds was associated with hospice utilization. These findings add to the sparse literature demonstrating disparities in hospice usage patterns based on race and geographic location.(21, 22)
Both our study and previous studies suggested that patients who were ≥ 65 years and non-white did not have optimal hospice use, potentially indicating poor quality of care at the end-of-life.(18, 19) In the Deep South, the non-white population is predominantly African American. Noted barriers to hospice enrollment by patients identifying as African American have included poor knowledge of hospice, low health literacy, misconceptions about the purpose of hospice, and cultural and spiritual beliefs.(6) Research found that the more a person uses religious coping, the more often they opt for life-sustaining aggressive treatment.(19, 23) African American patients are more likely to use religious coping and to request life-prolonging treatment.(19, 23, 24) Gender was also a significant finding with men being at greater risk of no hospice. In addition to not receiving hospice, other research found that being male was a risk factor for more aggressive end-of-life treatment.(25) Males, once in hospice, were also more likely to have short hospice stays due to late enrollment.(25) These findings suggest there may be cultural or gender differences to who uses hospice and strategies should consider such cultural competency.
Our analysis of late hospice utilization showed some trends that differed from our findings of any hospice use. Non-white patients had a decreased risk of receiving late hospice, as did non-navigated patients. We speculate that the greatest barrier to hospice for non-whites and non-navigated patients exist in entering hospice at all. In contrast, like our analysis of no versus any hospice, males were at increased risk for late hospice suggesting that males may experience barriers for timely hospice.
On the institutional level, we found that hospitals with inpatient palliative services were associated with a decreased likelihood of hospice utilization. Another study found that inpatient hospice units and patients admitted to inpatient hospice were more likely to have three or less days on hospice if they had a hematologic malignancy, were male, married, and younger than 65 years.(18) Prolonged intensive care unit stays, rather than discharging patients to hospice, may account for our study findings.(2, 26) Hospitals with palliative care beds may be treating patients at true end-of-life, instead of utilizing hospice. This is an important consideration given that hospice use is a quality metric of end-of-life.(23)
We observed that navigation may be a promising strategy for improving access to hospice. This study found that patients who received lay navigation support were more likely to use hospice. Others have found that navigation was useful in facilitating palliative care for older rural dwelling adults with advanced illness.(20) A navigation program using nurses for 25 rural older adults with mixed diagnoses including cancer found that navigators were able to provide a variety of services and there was high patient satisfaction. Patients were followed over two years and all patients died in their preferred place of death. Lay navigation may also be a way to improve patients’ understanding of hospice and to facilitate access to end-of-life services.
This study is limited by the data being drawn from hospitals with varying levels of palliative care services. Data drawn from Medicare claims and from institutions in the Deep South may not apply to populations who have non-Medicare insurance or who live outside of the Deep South. Additionally, when considering findings related to navigation, it is important to note that patients were not randomly assigned to navigation and there may have been selection bias. Patients were more likely to be referred to navigation if they were considered high-risk, for example those with metastatic cancer, high-morbidity cancers (e.g. pancreatic, ovarian, and lung), high-risk comorbidities (e.g. diabetes, heart failure, etc.) or history of hospitalization within the preceding month. We were also unable to examine whether the length of time a patient spent in navigation could affect late entry into hospice. Finally, this study was unable to determine if differences in hospice utilization were based in patient preferences and/or other healthcare delivery system factors.
In conclusion, this study indicates that patient and hospital characteristics have a potential role in who accesses timely hospice, and continued opportunity exists to improve hospice use for Medicare beneficiaries in the Deep South. Clinicians should be aware of specific patient characteristics, such as non-white race, and hospital characteristics, such as presence of inpatient palliative care services, which place patients at greater risk for no or late hospice utilization. Future research should focus on understanding the causes of disparities that place people of certain demographic traits at greater risk, finding ways to alleviate disparities, and exploring the potential role of lay navigation to enhance use of hospice use. Additional research should examine whether institutional factors contribute to disparities and support policies that improve hospice utilization.
Acknowledgments
Patients
Lay Navigators
UAB CMMI Team
UAB Comprehensive Cancer Center
UAB Cancer Community Network
Funding
This work was supported by the Department of Health and Human Services, Centers for Medicare and Medicaid Services (grant number 1C1CMS331023). The contents of this publication are solely the responsibility of the authors and do not necessarily represent the official views of the U.S. Department of Health and Human Services or any of its agencies. This funding source had no role in the design and conduct of the study; collection, management, analysis, or interpretation of the data; and preparation, review, or approval of the manuscript for publication. Dr. Turkman is supported by the National Palliative Care Research Center through a Junior Faculty Career Development award. Dr. Dionne-Odom is supported by the National Institute of Nursing Research (grant number R00NR015903). Dr Bakitas receives support from NR0136650-1A1, NR011871-01, PCORI PLC-1609-36381 and PLC-1609-36714.
Footnotes
Disclosures
The authors have no relevant conflict of interests to disclose.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
References
- 1.Fairfield KM, Murray KM, Wierman HR, Han PK, Hallen S, Miesfeldt S, et al. Disparities in hospice care among older women dying with ovarian cancer. Gynecol Oncol. 2012;125(1):14–8. [DOI] [PubMed] [Google Scholar]
- 2.Wang SY, Aldridge MD, Gross CP, Canavan M, Cherlin E, Johnson-Hurzeler R, et al. Geographic Variation of Hospice Use Patterns at the End of Life. J Palliat Med. 2015;18(9):771–80. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Teno JM, Plotzke M, Gozalo P, Mor V. A national study of live discharges from hospice. J Palliat Med. 2014;17(10):1121–7. [DOI] [PubMed] [Google Scholar]
- 4.Wang X, Knight LS, Evans A, Wang J, Smith TJ. Variations Among Physicians in Hospice Referrals of Patients With Advanced Cancer. J Oncol Pract. 2017;13(5):e496–e504. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Care CtAP. America’s Care of Serious Illness: 2015 State-By-State Report Card on Access to Palliative Care in Our Nation’s Hospitals. New York, NY: Center to Advance Palliative Care; 2015. [Google Scholar]
- 6.Hospice Arenella C. and Palliative Care for African Americans: Overcoming Disparities. J Palliat Med. 2016;19(2):126. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Johnson KS. Racial and ethnic disparities in palliative care. J Palliat Med. 2013;16(11):1329–34. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Care CtAP. Growth of Palliative Care in U.S. Hospitals: 2018 Snapshot (2000–2016). 2018. [Google Scholar]
- 9.Rogers MA, Dumanovsky T. How We Work: Trends and Insights in Hospital Palliative Care. The Center to Advance Palliative Care; 2017 February 2017. [Google Scholar]
- 10.Freeman HP, Rodriguez RL. History and principles of patient navigation. Cancer. 2011;117(15 Suppl):3539–42. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Meade CD, Wells KJ, Arevalo M, Calcano ER, Rivera M, Sarmiento Y, et al. Lay navigator model for impacting cancer health disparities. J Cancer Educ. 2014;29(3):449–57. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Shusted CS, Barta JA, Lake M, Brawer R, Ruane B, Giamboy TE, et al. The Case for Patient Navigation in Lung Cancer Screening in Vulnerable Populations: A Systematic Review. Popul Health Manag. 2018. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Rocque GB, Partridge EE, Pisu M, Martin MY, Demark-Wahnefried W, Acemgil A, et al. The Patient Care Connect Program: Transforming Health Care Through Lay Navigation. J Oncol Pract. 2016;12(6):e633–42. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Rocque GB, Dionne-Odom JN, Sylvia Huang CH, Niranjan SJ, Williams CP, Jackson BE, et al. Implementation and Impact of Patient Lay Navigator-Led Advance Care Planning Conversations. J Pain Symptom Manage. 2017;53(4):682–92. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Rocque GB, Pisu M, Jackson BE, Kvale EA, Demark-Wahnefried W, Martin MY, et al. Resource Use and Medicare Costs During Lay Navigation for Geriatric Patients With Cancer. JAMA Oncol. 2017. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Rocque GB, Taylor RA, Acemgil A, Li X, Pisu M, Kenzik K, et al. Guiding Lay Navigation in Geriatric Patients With Cancer Using a Distress Assessment Tool. J Natl Compr Canc Netw. 2016;14(4):407–14. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Westreich D, Greenland S. The table 2 fallacy: presenting and interpreting confounder and modifier coefficients. Am J Epidemiol. 2013;177(4):292–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.O'Connor NR, Hu R, Harris PS, Ache K, Casarett DJ. Hospice admissions for cancer in the final days of life: independent predictors and implications for quality measures. J Clin Oncol. 2014;32(28):3184–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Johnson KS, Payne R, Kuchibhatla MN, Tulsky JA. Are Hospice Admission Practices Associated With Hospice Enrollment for Older African Americans and Whites? J Pain Symptom Manage. 2016;51(4):697–705. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Pesut B, Hooper B, Jacobsen M, Nielsen B, Falk M, BP OC. Nurse-led navigation to provide early palliative care in rural areas: a pilot study. BMC Palliat Care. 2017;16(1):37. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Kao YH, Chiang JK. Effect of hospice care on quality indicators of end-of-life care among patients with liver cancer: a national longitudinal population-based study in Taiwan 2000-2011. BMC Palliat Care. 2015;14:39. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Foley KM, Gelband H. Improving palliative care for cancer. Washington, D.C.: National Academy Press; 2001. xvii, 325 p. p. [PubMed] [Google Scholar]
- 23.LoPresti MA, Dement F, Gold HT. End-of-Life Care for People With Cancer From Ethnic Minority Groups: A Systematic Review. The American journal of hospice & palliative care. 2016;33(3):291–305. [DOI] [PubMed] [Google Scholar]
- 24.Aldridge MD, Canavan M, Cherlin E, Bradley EH. Has Hospice Use Changed? 2000-2010 Utilization Patterns. Med Care. 2015;53(1):95–101. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Saito AM, Landrum MB, Neville BA, Ayanian JZ, Weeks JC, Earle CC. Hospice care and survival among elderly patients with lung cancer. J Palliat Med. 2011;14(8):929–39. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Anderson NB, Bulatao RA, Cohen B, National Research Council (U.S.). Panel on Race Ethnicity and Health in Later Life Critical perspectives on racial and ethnic differences in health in late life. Washington, D.C.: National Academies Press; 2004. xv, 735 p. p. [PubMed] [Google Scholar]