Fax: (215) 573-8684
Race, treatment preferences, and hospice enrollment: Eligibility criteria may exclude patients with the greatest needs for care
Article first published online: 23 DEC 2008
Copyright © 2008 American Cancer Society
Volume 115, Issue 3, pages 689–697, 1 February 2009
How to Cite
Fishman, J., O'Dwyer, P., Lu, H. L., Henderson, H., Asch, D. A. and Casarett, D. J. (2009), Race, treatment preferences, and hospice enrollment: Eligibility criteria may exclude patients with the greatest needs for care. Cancer, 115: 689–697. doi: 10.1002/cncr.24046
Fax: (215) 573-8684
- Issue published online: 20 JAN 2009
- Article first published online: 23 DEC 2008
- Manuscript Accepted: 2 SEP 2008
- Manuscript Revised: 5 AUG 2008
- Manuscript Received: 16 JUN 2008
- American Cancer Society Mentored Scholar Research. Grant Number: MRSGT-08-013-01-CPPB
- University of Pennsylvania's National Institutes of Health-funded Center of Excellence for Cancer Communication Research
- Veterans Administration Advanced Research Career Development Award. Grant Number: R01CA109540
- Presidential Early Career Award for Scientists and Engineers
- Abramson Cancer Center
- Center for Clinical Epidemiology and Biostatistics
- Center for Health Equity Research and Promotion
- treatment preferences;
- conjoint analysis;
- Medicare Hospice Benefit;
- palliative care;
- African American;
The requirement that patients give up curative treatment makes hospice enrollment unappealing for some patients and may particularly limit use among African-American patients. The current study was conducted to determine whether African-American patients with cancer are more likely than white patients to have preferences for cancer treatment that exclude them from hospice and whether they are less likely to want specific hospice services.
Two hundred eighty-three patients who were receiving treatment for cancer at 6 oncology clinics within the University of Pennsylvania Cancer Network completed conjoint interviews measuring their perceived need for 5 hospice services and their preferences for continuing cancer treatment. Patients were followed for 6 months or until death.
African-American patients had stronger preferences for continuing their cancer treatments on a 7-point scale even after adjusting for age, sex, finances, education, Eastern Cooperative Oncology Group performance status, quality of life, and physical and psychologic symptom burden (adjusted mean score, 4.75 vs 3.96; β coefficient, 0.82; 95% confidence interval, 0.22-1.41 [P = .007]). African-American patients also had greater perceived needs for hospice services after adjusting for these characteristics (adjusted mean score, 2.31 vs 1.83; β coefficient, 0.51; 95% confidence interval, 0.11-0.92 [P = .01]). However, this effect disappeared after adjusting for household finances.
Hospice eligibility criteria may exclude African-American patients disproportionately despite greater perceived needs for hospice services in this population. The mechanisms driving this health disparity likely include both cultural differences and economic characteristics, and consideration should be given to redesigning hospice eligibility criteria. Cancer 2009. © 2008 American Cancer Society.
Patients with cancer have substantial needs for care throughout the course of their illness, from diagnosis through the end of life.1-8 Hospice offers a comprehensive program of services designed to meet these needs, including a visiting nurse, respite care, a chaplain, a home health aide, and a counselor. Approximately 500,000 patients with cancer enroll in hospice every year, and evidence indicates that hospice enrollment is associated with improved patient and family outcomes and higher satisfaction with care.9-14
African-American patients are less likely than white patients to use hospice.15-17 Although these disparities are well described, it is not known why they exist. It is possible that African Americans delay or avoid hospice enrollment because of the requirement that they forgo curative treatment.18 This criterion forces patients to make a ‘terrible choice’ between continued treatment and hospice services,19-21 a choice widely believed to obstruct timely hospice use.22-25 The criterion has been part of the Medicare Hospice Benefit since its creation 25 years ago; and, because nearly all hospice are Medicare-certified, they follow Medicare eligibility criteria for all patients regardless of age or insurance. This barrier may be particularly significant for African Americans, who are more likely than whites to want various life-prolonging interventions.26-29 However, it is not known whether African-American patients also have more aggressive cancer treatment preferences or whether such preferences help explain the disparity in hospice use.
An alternative explanation for the under-use of hospice among African Americans is that they may be less likely to want the services that hospice provides. To our knowledge, no studies have evaluated this possibility. Although some studies suggest that African Americans have less favorable attitudes toward hospice,30-34 their perceived needs for hospice services have not been described.
It is important to determine why these disparities in hospice enrollment exist.35, 36 If they are the result of aggressive treatment preferences among African Americans, then lower rates of hospice use, arguably, are unfair and represent a real disparity in care that could be eliminated by redesigning hospice eligibility criteria. Conversely, if African Americans are less likely to want hospice services, then changes to the benefit are not necessary, and, instead, modifications to the services offered may be warranted. Therefore, the objective of this study was to define and compare preferences for cancer treatment and perceived needs for hospice services among African-American patients and white patients.
MATERIALS AND METHODS
Setting and Sample
We conducted this study in a population of patients with cancer, because cancer is the most common hospice admitting diagnosis (approximately 50%).37 Patients with cancer were recruited over 18 months from 6 oncology clinics within the University of Pennsylvania Comprehensive Cancer Center Network. Patients were eligible if they had clinical or radiologic evidence of active cancer, were receiving chemotherapy or radiation therapy, and had a life expectancy of ≤6 months if they were to discontinue cancer-directed treatment, according to their oncologist. These patients met the prognostic eligibility criterion for hospice, which requires that patients have a prognosis of ≤6 months if their disease runs its usual course (eg, without treatment).18 Nursing staff at each clinic assisted a rotating team of 3 interviewers to identify and approach eligible patients. This study was approved by the institutional review boards of participating sites.
After providing informed consent, patients completed separate structured interviews using Tablet personal computers. First, they provided demographic data, including self-described race, age, marital status, and household finances (eg, money available at the end of the month).38 In addition, patients completed the Global Distress Index (GDI) of the Memorial Symptom Assessment Scale,39 the Functional Assessment of Cancer Therapy-General (FACT-G),40 the Medical Outcomes Survey Social Support Scale,41 a single-item global rating of health,42 the Eastern Cooperative Oncology Group performance status scale (ECOG-PS).43, 44 and assessments of Activities of Daily Living (ADLs) and Instrumental Activities of Daily Living (IADLs).45, 46
Patients' perceived needs for hospice services were assessed using a conjoint interview, which refers to a process by which respondents are asked to make choices among goods or services by considering several attributes jointly.47 One advantage of this approach over direct ratings is that the value of each service is determined while considering other services, revealing preferences in a more naturalized setting rather than eliciting them 1 by 1. Second, whereas direct rating tasks often suffer from ceiling effects, in conjoint tasks, respondents must choose those services that are most important.
For these interviews, we used Sawtooth Software's Adaptive Conjoint Analysis (ACA) package (ACA for Windows, version 4.0; Sawtooth Software, Sequim, Wash). ACA is a hybrid48-50 method of conjoint analysis that uses an interactive, self-administered computer program to decrease respondent burden while preserving the psychometric advantages of conjoint techniques. Similar to classic conjoint methods, hybrid conjoint methods such as ACA determine the importance that respondents place on various services by examining the choices that they make between pairs of services. In addition, however, hybrid methods also use respondents' direct ratings of the importance of those services, which substantially reduces the number of choices that each individual is required to make. Although they were developed for use in marketing,47 conjoint techniques and hybrid modifications have been used increasingly to study health-related preferences.51-58 Their validity, reliability, and predictive power are well established.59, 60
Patients reviewed a brochure describing 5 of the hospice services that are required by Medicare: a visiting nurse, a chaplain, a home health aide, a counselor, and respite care18 (Fig. 1). These services were selected based on previous studies indicating that they are among the most valuable to patients.61-63 The subsequent interview used abbreviated (4-6 word) descriptions of each service, and patients kept the brochure open for rapid reference. To assess patients' perceived needs for each service, they were asked to consider how much each service would help them if it were offered at that time (Fig. 2). First, they rated the importance of each service on a scale from 1 (‘not at all important’) to 7 (‘extremely important’). Second, they evaluated pairs of programs presented side-by-side, each containing 2 or 3 services, and rated their preference on a scale from 1 to 9 (1, ‘strongly prefer the program on the left’; 5, ‘indifferent’; 9, ‘strongly prefer the program on the right’).
Patients' preferences for cancer treatment were assessed by asking whether they would want to continue receiving their current cancer treatment to achieve various probabilities of surviving for 6 months (nearly 100%, 90%-99%, 50%-89%, 10%-49%, 1%-9%, or nearly 0%). This question was adapted from previous work in seriously ill populations.64, 65 These probabilities were varied systematically (from lowest to highest and then back to the second lowest) to balance order effects. The lowest probability of 6-month survival was recorded for which a patient would be willing to continue receiving treatment. Responses were used to create a 7-level ordinal variable that described the strength of a patient's treatment preferences (1 level for each of the 6 responses above plus 1 for those patients who would not want to continue treatment even for a nearly 100% chance of surviving for 6 months). This last category was used to create a dichotomous variable that identified patients who would not want treatment even for a nearly 100% probability of surviving for 6 months and, thus, who would have been eligible for hospice.
By using the scale described above, preferences for aggressive cancer treatment were compared using a Wilcoxon rank-sum test in bivariate analysis. Adjusted analysis of preferences was performed by using an ordinal logistic regression model. On the basis of respondents' direct ratings and their choices among service options, ACA uses least-squares regression to calculate the utility, or usefulness, of each service for each patient.66 These utilities reflect the perceived need that a patient has for each service, and higher utilities correspond to a greater perceived need. We analyzed patients' perceived needs for each of the 5 services individually as well as patients' perceived needs for all 5 services combined (their total utilities). Because utilities were not distributed normally, we used the Wilcoxon rank-sum test to compare perceived needs among patient subgroups and linear regression models with a log transformation and exponentiated coefficients to identify variables that were associated independently with service needs. Variables with a significance value <.25 were considered for inclusion in a multivariable linear regression model,67 also with a log transformation and exponentiated coefficients.
A sample of 283 patients would be adequate to detect a small difference in the summed utilities of all 5 hospice services between African-American and white patients, assuming at least 20% of the sample would be African American (ability to detect 0.33 standardized difference, α = .05; 2-sided). Stata software (Stata for Windows, version 8.0; StataCorp, College Station, Tex) was used for all statistical analysis.
Patients who agreed to participate (300 of 352; 85%) were similar to those who refused with respect to age, ECOG-PS, race, and education (data not shown). However, women were significantly more likely to participate compared with men (152 of 169 women [92%] vs 148 of 183 men [82%]; Fisher exact test; P = .02). Of the 300 patients who consented, this sample was restricted to those patients (n = 283; 94%) who self-identified as either African-American (n = 81;2 9%) or white (n = 202; 71%). The characteristics of these patients are described in Table 1.
|Characteristic||No. of Patients (%)||P|
|Overall, N=300||African American, N=81||White, N=202|
|Men||148 (49)||47 (58)||95 (47)|
|Mean [range]||58 [20-89]||63 [31-89]||57 [20-89]|
|>65||84 (28)||33 (41)||49 (24)|
|Single||45 (15)||19 (23)||23 (11)|
|Married||174 (58)||28 (35)||136 (67)|
|Divorced/separated||45 (15)||19 (23)||23 (11)|
|Widowed||22 (7)||9 (11)||13 (6)|
|Living w/ partner||14 (5)||6 (7)||7 (3)|
|Household finances (funds available at the end of the mo)||<.001|
|Has money left over||142 (47)||14 (17)||122 (60)|
|Has just enough to make ends meet||80 (27)||29 (36)||47 (23)|
|Does not have enough money to make ends meet||52 (17)||27 (33)||20 (10)|
|Refused||26 (9)||11 (14)||13 (6)|
|Did not graduate from high school||24 (8)||14 (17)||8 (4)|
|High school||85 (28)||32 (40)||50 (25)|
|Some college or technical school||76 (25)||19 (23)||53 (26)|
|College||55 (18)||8 (10)||43 (21)|
|Graduate school||60 (20)||8 (10)||48 (24)|
|ECOG performance score||.13|
|0||90 (30)||23 (28)||64 (32)|
|1||32 (11)||6 (7)||23 (11)|
|2||124 (41)||32 (40)||85 (42)|
|3||46 (15)||17 (21)||26 (13)|
|4||8 (3)||3 (4)||4 (2)|
|Breast||84 (28)||20 (25)||58 (29)|
|Gastrointestinal||53 (18)||15 (19)||35 (17)|
|Hematologic||37 (12)||4 (5)||31 (15)|
|Lung||36 (12)||9 (11)||26 (13)|
|Prostate||35 (12)||20 (25)||14 (7)|
|Oropharynx||29 (10)||8 (10)||19 (9)|
|Genitourinary||18 (6)||1 (1)||16 (8)|
|Sarcoma||5 (2)||4 (5)||1 (0.5)|
|Melanoma||2 (1)||0 (0)||2 (1)|
|Glioblastoma||1 (0)||0 (0)||0 (0)|
|Global distress index symptoms|
|Lack of appetite||111 (37)||34 (42)||69 (34)||.22|
|Lack of energy||204 (68)||51 (63)||150 (69)||.30|
|Pain||164 (55)||50 (62)||101 (50)||.07|
|Drowsiness||171 (57)||47 (58)||113 (56)||.75|
|Constipation||94 (31)||30 (37)||55 (27)||.10|
|Dry mouth||137 (46)||43 (53)||83 (42)||.07|
|Sadness||116 (39)||29 (36)||80 (40)||.55|
|Worrying||170 (57)||47 (58)||115 (57)||.87|
|Irritability||121 (40)||32 (40)||80 (40)||.99|
|Nervousness||116 (39)||30 (37)||78 (39)||.81|
|Continue cancer treatment with probability of survival, %||.002|
|Nearly 100||140 (47)||48 (59)||86 (43)|
|90-99||22 (7)||9 (11)||11 (5)|
|50-89||34 (11)||3 (4)||28 (14)|
|10-49||50 (17)||16 (20)||31 (15)|
|1-9||9 (3)||0 (0)||9 (4)|
|Almost 0||7 (2)||0 (0)||5 (2)|
|No treatments||38 (13)||5 (6)||32 (16)|
Patient Race and Cancer Treatment Preferences
African-American patients had stronger preferences for aggressive cancer treatment (using the scale of choices from 1 to 7) compared with white patients (median score, 6 vs 4; Wilcoxon rank-sum test; P = .006). That is, they were willing to continue their treatment in return for a smaller likelihood of 6-month survival. In an ordinal logistic regression model, African-American patients had stronger preferences for aggressive cancer treatment (adjusted means, 4.75 vs 3.96; β coefficient, 0.82; 95% confidence interval [95% CI], 0.22-1.41 [P = .007]) after adjusting for age, sex, household finances, education, ECOG -PS, quality of life (FACT-G), and physical and psychologic symptom burden (physical and psychologic subscales of the GDI, respectively). Compared with white patients, African-American patients also were less likely to say that they would refuse all cancer treatment even if it were associated with an ‘almost 100%’ probability of surviving for 6 months (5 of 81 patients [6%] vs 32 of 202 patients [16%]; odds ratio, 2.86; 95% CI, 1.04-9.74; P = .03).
Patient Race and Perceived Needs for Hospice Services
Compared with white patients, African-American patients reported greater needs for all hospice services combined (median utilities: 2.26 vs 1.79; rank-sum test; P = .005). African-American patients assigned higher utilities to all hospice services except a home health aide (Fig. 3).
In a multivariate regression model, African-American race remained associated with greater perceived needs for services after adjusting for age, education, ECOG-PS, dependencies in ADLs, and physical and psychologic symptom burden (GDI physical and psychologic subscales, respectively; adjusted mean utilities: 2.31 vs 1.83; β coefficient, 0.51; 95% CI, 0.11-0.92 [P = .01]). This model did not include economic factors, which are considered separately below.
To describe the relation between perceived needs for hospice services and economic factors, which often are associated closely with race,68 we used household finances as an indicator of economic status (not enough, enough, or more than enough money left at the end of the month). In unadjusted analysis, patients with fewer financial resources had greater perceived needs for hospice services (more than enough money left over, 1.51; enough money left over, 2.34; not enough money left over, 2.53). When the perceived needs of African-American and white patients were compared after adjusting for household finances, there was no significant difference (adjusted mean utilities: 2.38 vs 1.75; β coefficient, 0.24; 95% CI, 0.22-0.70 [P = 31]). Similarly, in a combined model that included household finances as well as all of the patient characteristics used in earlier models, African Americans did not have greater perceived needs for hospice services.
In subanalyses, we compared the perceived needs for hospice services of African-American patients (n = 76) and white patients (n = 170) whose preferences excluded them from hospice. In these groups, African-American patients had significantly greater perceived needs than white patients (total utilities: 2.34 vs 1.81; rank-sum test; P = .006). We also compared the perceived needs of African-American patients whose treatment preferences made them ineligible for hospice (n = 76) and white patients whose treatment preferences were consistent with hospice (n = 32). After controlling for health status by adjusting for age, ECOG-PS, dependencies in ADLs, and GDI physical and psychologic subscale scores, ineligible African-American patients had greater perceived needs for hospice services than eligible whites (adjusted means: 2.31 vs 1.73; β coefficient, 0.70; 95% CI, 0.05-1.35 [P = .03]). However, after adjusting for household finances, these 2 groups had similar utilities for hospice services (1.91 vs 1.83; β coefficient, −0.34; 95% CI, −0.56-0.11 [P = .27]). Because of the small number of African-American patients whose preferences made them eligible for hospice (n = 5), a comparison of these patients' perceived needs with those of ineligible African Americans was not possible.
The Medicare Hospice Benefit and similar benefits of other insurers were designed to ensure that patients in the last 6 months of life have access to high-quality palliative care. However, African-American patients are less likely to use hospice than white patients. The results from the current study suggest that disparities in hospice use among patients with cancer may be the result of cancer treatment preferences among African-American patients that make them ineligible despite greater perceived needs for hospice care. In particular, this study offers 3 main insights into these disparities in hospice access.
First, we observed that African-American patients with cancer are more likely than white patients to want aggressive cancer treatment. This difference in preferences persists after adjusting for clinical and demographic characteristics. This result is important because, although other studies have observed racial differences in preferences for life-sustaining technologies,26-29 they have not examined preferences for cancer treatment. Preferences for aggressive cancer care, as well as other aggressive technologies, could be linked to medical distrust, quality of care, or communication issues between patients and providers.
Second, we observed that African-American patients with cancer had greater perceived needs for specific hospice services compared with white patients. This difference appears to be independent of clinical characteristics (eg, symptom burden, performance status) that are associated with greater needs. Furthermore, African-American patients whose treatment preferences would have excluded them from hospice had greater perceived needs than white patients who would have been eligible.
Third, we observed that that the greater perceived need for hospice services among African Americans was attributable largely to differences in self-reported finances. That is, the observed differences in perceived needs between African-American and white patients may reflect financial resources, because those who reported the fewest financial resources also reported the greatest need for services. This finding is consistent with other studies suggesting that the mechanisms driving health disparities include not only cultural differences associated with race but also economic characteristics that fundamentally may be unrelated to race.36, 68, 69 Thus, these economic factors should be explored further to determine why poorer patients want more services. For instance, it is possible that wealthier patients have more resources for care at the end of life and use these resources to obtain services outside of hospice. If so, then wealthier patients may be in a better position to avoid the ‘terrible choice’ requiring that they forgo treatment to receive hospice benefits. However, this is not known.
Together, these findings suggest that the hospice eligibility criteria of Medicare and other insurers requiring patients to give up cancer treatment contribute to racial disparities in hospice use. Moreover, these criteria do not select those patients with the greatest needs for hospice services. To the degree that this eligibility requirement prevents hospice use by the patients with the greatest needs, it fails to fulfill its purpose and should be reconsidered. It also is unfair: Other Medicare-supported services do not require that patients forgo 1 treatment to get another.70 Instead, our study suggests that hospice access could be made fairer by using eligibility criteria that are more directly need-based. For instance, eligibility might be determined better by assessing needs for specific hospice services, such as pain or symptom management. This would make eligibility for hospice similar to eligibility for the Medicare Home Care Benefit, which perhaps is the closest Medicare program.70
This study has several limitations. First, the results reported here are based on patients' self-reported needs for hospice services, but it is possible that these patients could not appreciate adequately how the services described could benefit them.63 However, patients arguably are the best judges of their own needs, even if those assessments are imperfect.
Second, this study assessed only 5 hospice services. It is possible that the finding of greater perceived needs reported here by African-American patients would not be true for other services that hospice provides (eg, delivery of medications or durable medical equipment). Although this is possible, there is no reason to believe that the pattern observed for the 5 services described here are unique.
Third, in this study, we did not assess actual choices regarding hospice enrollment, which may be influenced by factors such as the availability of family members who can assist with home care. A patient without adequate informal caregiving support and supervision at home may be less likely to enroll. Therefore, further research is needed to define which patient characteristics, such as the availability of family support, influence actual choices regarding hospice enrollment.
For 25 years, hospice services have been restricted to patients in the last 6 months of life who are willing to give up curative treatment.18 Although this restriction was intended to control costs, it has substantially reduced access to hospice for many patients who have needs for hospice services, and a disproportionate number of these patients are African American. This eligibility criterion should be reconsidered, and needs-based criteria should be considered to make hospice eligibility criteria both fairer and more consistent with eligibility criteria for other health insurance benefits.70
Conflict of Interest Disclosures
Supported by American Cancer Society Mentored Scholar Research grant MRSGT-08-013-01-CPPB (to J.F.), the University of Pennsylvania's National Institutes of Health-funded Center of Excellence for Cancer Communication Research grant (to J.F.), grant R01CA109540 (to D.J.C.), a Veterans Administration Advanced Research Career Development Award, a Presidential Early Career Award for Scientists and Engineers (to D.J.C.), and by grants from the Abramson Cancer Center (to P.O.), the Center for Clinical Epidemiology and Biostatistics (to J.F.), and the Center for Health Equity Research and Promotion (to D.J.C. and J.F.).
- 18United States Congress. Medicare Hospice Regulations: 42 Code of Federal Regulations, Part 418. Washington, DC: Federal Register; 1996.
- 20United States Congress. Testimony Presented Before the Congress of the United States House of Representatives Committee on Government Reform. October 19, 1999. Available at:http://www.dyingwell.org/uschrtest.htm. Accessed May 25, 2007.
- 21Improving Palliative Care for Cancer: Summary and Recommendations. Washington, DC: National Academies Press; 2001., .
- 32Attitudes, values, and questions of African Americans regarding participation in hospice programs. J Hospice Palliat Nurs. 2006; 8: 77-85..
- 35Improving Palliative Care for Cancer: Summary and Recommendations. Institute of Medicine and National Research Council. Washington, DC: National Academies Press; 2001., .
- 37National Hospice and Palliative Care Organization. National Trend Summary 2005. Washington, DC: National Hospice and Palliative Care Organization; 2006.
- 38Establishing populations for epidemiologic studies of the elderly: study design and methodology. Aging. 1993; 5: 27-37., , , et al.
- 46Functional evaluation: the Barthel Index. Md State Med J. 1965; 14: 62-65., .
- 66Sawtooth Software, Inc. ACA/Hierarchical Bayes Technical Paper.Sawtooth Software Technical Paper Series. Sequim, Wash: Sawtooth Software, Inc.; 2001.
- 70Centers for Medicare & Medicaid Services. Your Medicare Benefits. Available at: http://www.medicare.gov/Publications/Pubs/pdf/10116.pdf Accessed June 4, 2007.