Survival by race after hospice revocation to pursue aggressive care.
Johnson KS, Kuchibhatla M, Tanis D, Tulsky JA. Racial Differences in Hospice Revocation to Pursue Aggressive Care. Arch Intern Med. 2008;168(2):218-224. doi:10.1001/archinternmed.2007.36
Copyright 2008 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2008
Hospice provides supportive care to terminally ill patients at the end of life. However, some enrollees leave hospice before death in search of therapies that may prolong survival. Because of a greater preference for life-sustaining therapies at the end of life, African American patients may be more likely than white patients to withdraw from hospice to seek life-prolonging therapies.
In a secondary data analysis of African American and white patients discharged from VITAS hospice programs between January 1, 1999, and December 31, 2003, we used logistic regression to examine the association between race and discharge disposition defined as hospice revocation to pursue aggressive care (eg, emergency medical care, chemotherapy, or invasive medical intervention) vs all other discharges. We used a Cox proportional hazards model to examine survival at 1 year after hospice revocation in a subgroup of enrollees from Florida hospice programs.
Of the 166 197 enrollees, 2.8% revoked hospice to pursue aggressive care, and African American patients were more likely than white patients to do so (4.5% vs 2.5%; P < .001). In multivariate analysis, African American patients had a 70% higher odds of leaving hospice to pursue life-prolonging therapies (odds ratio, 1.70; 95% confidence interval, 1.57-1.84). In the subgroup analysis, 48.4% of the enrollees who revoked hospice to pursue life-prolonging therapies were still alive at 1 year.
African American patients were more likely than white patients to revoke hospice to pursue life-prolonging therapies. Models of health care that couple curative and palliative therapies may be more attractive to African American patients and more effective at maximizing continuity throughout life-limiting illness.
Medicare pays for most hospice care in the United States.1 To qualify for the Medicare hospice benefit, patients must have a life-limiting illness with a prognosis of 6 months or less and must elect hospice care for management of the terminal illness in lieu of traditional Medicare Part A. Enrollees may revoke the hospice benefit at any time by simply signing a document, following which their traditional Medicare Part A benefits are restored. They may then seek potentially life-prolonging therapies for management of the illness for which they were previously receiving hospice care. The revocation process is similar for Medicaid and most private insurers.
Although hospice provides palliative and supportive care to terminally ill patients in the last months of life, approximately 15% of enrollees are discharged alive, including those who withdraw from hospice to seek life-prolonging therapies.1,2 African American patients, a group historically underrepresented in hospice, may be more likely to leave hospice before death. Compared with white patients, African American patients are more likely to request life-sustaining therapies at the end of life regardless of prognosis—treatment preferences that are inconsistent with the goals of hospice care.3- 5 If cultural differences in preferences for end-of-life care persist even among those who enroll in hospice, proportionately more African American patients than white patients may withdraw from hospice in search of therapies that could prolong survival.
In addition to understanding how cultural preferences for end-of-life care may influence decisions to withdraw from hospice, another important consideration is what happens to those who leave hospice. Most patients are referred to hospice shortly before death, when life-prolonging therapies, even if used, may not be effective.1,2 Because of greater preferences for life-sustaining therapies, African American patients may be even more likely to exhaust all treatment options regardless of the likelihood of success before considering hospice care. Therefore, despite their treatment goals, African American patients who withdraw from hospice to pursue life-prolonging therapies may still die in a short time and without access to high-quality end-of-life care.
Although several studies have examined barriers to hospice enrollment among African American populations, little research has focused on what happens to those who enroll in hospice or those who leave hospice before death.6- 10 Understanding the association between ethnicity and decisions to withdraw from hospice as well as survival after hospice revocation to seek life-prolonging therapies may inform the development of health care delivery models that provide culturally appropriate care to terminally ill patients of all ethnic groups. Therefore, we examined differences in hospice revocation to pursue life-prolonging therapies and subsequent 1-year survival in a large sample of African American and white hospice enrollees.
Data for these analyses were obtained from VITAS, the largest hospice care provider in the United States. After we obtained the required approval from the hospice care provider and the Duke University Health System Institutional Review Board, patient-level files were used to create the study database. The data were abstracted from the computerized central administrative and clinical database of the hospice care provider.
We included all African American and white patients discharged from VITAS hospice programs between January 1, 1999, and December 31, 2003. During the analysis period, there were 27 VITAS hospice programs in 8 states: California, Florida, Illinois, New Jersey, Ohio, Pennsylvania, Texas, and Wisconsin. The final sample included 166 197 enrollees.
The outcome measure for the primary analysis was disposition at discharge. VITAS assigns to all discharges 1 of 18 unique qualifiers that indicate the reason for discharge. Examples include alive, moved; alive, prognosis extended; died, home; died, inpatient; and alive, revoked for aggressive therapy. We categorized discharge disposition as alive, revoked for aggressive therapy vs all other discharges. Although the specific reason for revocation is not listed in the database, revocations to pursue aggressive therapy include voluntary withdrawal from hospice to seek emergency medical services, invasive medical interventions, chemotherapy, or other potentially life-prolonging therapies.
Variables for the final model were chosen based on their relevance as potential confounders of the relationship between race and disposition at discharge. The final model included demographic and hospice use variables. In addition to race, demographic variables included age (<65 or ≥65 years), sex, marital status (married or other), relationship of caregiver to enrollee (spouse, child, other relative, nonrelative, or no caregiver identified), and median household income, which was generated by matching the patients' zip codes to 2000 US Census tract data. The income measure was divided into 3 categories, each containing one-third of the income distribution of the sample (low, ≤$37 280; middle, $37 281 to $48 541; and high, ≥$48 542). Hospice use variables included admission level of care (routine care in a private residence, routine care in a nursing home, continuous care, or inpatient care), payment source (Medicare, Medicaid, charity, or other), health maintenance organization (yes or no), diagnosis (cancer or other), length of stay (0-7 days, 8-30 days, 31-180 days, or >180 days), and hospice location by state.
We used χ2analyses to compare African American and white enrollees across categorical demographic and hospice use variables. Differences were considered statistically significant at P < .05.
We used a logistic regression model in the multivariate analysis. As noted, all variables hypothesized a priori to be important predictors of disposition at discharge were included in the final model. We also tested for interactions between race and all other covariates. Because these interactions were not statistically significant (P > .05), we did not include them in the final model. The final model included only main effects. We used the c-index to assess the discrimination of the model.
We examined survival at 1 year among a subgroup of enrollees who revoked hospice to pursue aggressive care. This analysis included African American and white patients discharged from 3 Florida VITAS hospice programs between January 1, 1999, and December 31, 2001. To determine the date of death, we searched the Florida death registry for the years 1999 through 2002 using the patients' last name and date of birth.
We used a Cox proportional hazards model to examine predictors of survival 1 year after hospice revocation to pursue aggressive care. In addition to race, the final model included the following variables: age, sex, marital status, payment source, health maintenance organization enrollment, diagnosis, admission level of care, relationship of caregiver to enrollee, and length of stay. We used the likelihood ratio test to examine the fit of the model. All analyses were performed using SAS statistical software, version 9.1 (SAS Institute Inc, Cary, North Carolina).
Between January 1, 1999, and December 31, 2003, 166 197 African American and white patients were discharged from VITAS hospice programs. African American patients represented 14.6% of the sample. Most enrollees of both races died before discharge (93.0% of African American patients and 93.6% of white patients). A total of 4648 enrollees (2.8%) revoked hospice to pursue aggressive care, and African American patients were more likely than white patients to do so (4.5% vs 2.5%; P < .001). The median time from admission to revocation was 22 days for both groups. Sample characteristics are listed in Table 1.
Compared with white patients, a greater proportion of African American patients were younger than 65 years, unmarried, in the lower third of the distribution of median household income, admitted with a diagnosis of cancer, and used Medicaid or charity as a payment source. After controlling for demographic and hospice use variables, African American patients remained significantly more likely than white patients to revoke hospice to pursue aggressive therapy (odds ratio, 1.70; 95% confidence interval, 1.57-1.84). Results of the multivariate analysis are given in Table 2. The c-index of the model was 0.66; the model explained 66% of the variation in discharge disposition.
Of the 32 216 African American and white patients who were discharged from the 3 Florida hospice programs included in this analysis between January 1, 1999, and December 31, 2001, 2.6% revoked hospice to pursue aggressive care. Similar to the findings described herein, African American patients were more likely than white patients to do so (4.5% vs 2.2%, P < .001). The median time from enrollment to revocation was 29 days for African American patients and 22 days for white patients. Sample characteristics for this subgroup are given in Table 3.
The Figure displays survival curves by race in the year after revocation. No significant difference was found in survival rates between the 2 groups. One year after revoking hospice to pursue life-prolonging therapies, 48.1% of African American patients and 48.5% of white patients were still alive (P = .93). Among those who died, the median time to death was 42 days for African American patients and 48 days for white patients (P = .33). In a Cox proportional hazards model, only a diagnosis of cancer and admission level of care of continuous care were associated with an increased risk of death at 1 year. As hospice length of stay increased by category, the risk of death at 1 year decreased (Table 4). The likelihood ratio test for the model fit was statistically significant (P < .001), indicating that the full model was associated with significantly better predictions of the outcome than a model with no independent variables.
Consistent with national data, most hospice enrollees in this analysis died while receiving hospice care.1 African American patients were significantly more likely than white patients to leave hospice care before death to pursue life-prolonging therapies. In a subgroup of patients who revoked hospice to pursue life-prolonging therapies, almost half were still alive at 1 year. These findings highlight cultural differences in preferences for end-of-life care and have implications for the design of health care delivery models that support the needs and preferences of terminally ill patients throughout the course of life-limiting illness.
African American patients are more likely than white patients to request life-sustaining therapies, including cardiopulmonary resuscitation and mechanical ventilation, at the end of life regardless of prognosis.3- 5 These beliefs and values that emphasize longevity and deny death are in direct opposition to the hospice philosophy of care. Because hospice care recognizes death as the final stage of life and emphasizes quality rather than quantity of life, it seems likely that terminally ill patients whose goals of care involve unconditionally prolonging life, regardless of their ethnicity, would decline hospice enrollment. However, many patients and families know little about hospice care before enrollment and are more often motivated to enroll because of the services that hospice provides instead of its emphasis on comfort rather than cure, which patients and families may learn about only after enrollment.11 This lack of information may be even more common among African American patients, who may have less general knowledge of hospice care.6,8,12 African American patients who learn about the hospice philosophy of care after enrollment may subsequently revoke hospice if their treatment preferences involve the use of life-prolonging therapies.
Another possible explanation for our findings is that African Americans who choose to forgo life-prolonging therapies may be more likely than white patients to change their treatment preferences. Possession of an advance directive has been associated with more stable preferences to forgo life-prolonging therapies.13 Our data did not include information about advance directives. However, because African American patients are less likely to participate in advance care planning, they may spend less time thinking about or discussing their preferences for end-of-life care before enrolling in hospice.3,5,12,14 Because hospice care facilitates advance care planning, hospice enrollment may provide African American patients with the opportunity to solidify and discuss their treatment preferences. When these preferences conflict with the hospice philosophy of care, they may withdraw from hospice.
Although those who enroll in hospice are required to have a prognosis of 6 months or less, almost half of the patients in our subgroup analysis were still alive 1 year after revoking hospice to pursue life-prolonging therapies. No racial differences were found in rates of survival. We do not know what therapies enrollees received after hospice revocation or if these therapies prolonged survival. A diagnosis of cancer and an admission level of care of continuous care were the only variables associated with an increased risk of death at 1 year. This finding suggests that patients who required intensive symptom management at admission (continuous care) had a poorer prognosis. In addition, patients with cancer have a more predictable trajectory of decline at the end of life than those with noncancer diagnoses.15 Physicians may be more accurate at identifying patients with cancer who have limited life expectancies than patients with other diagnoses such as congestive heart failure or dementia in whom the dying trajectory is less predictable and prognostication more difficult. Therefore, survival at 1 year may reflect prognosis at hospice admission rather than response to a specific treatment received after hospice revocation. Also, of the 15% of hospice enrollees who are discharged alive each year, some no longer meet the eligibility criteria. Thus, it is likely that some of those who were alive 1 year after revocation may have survived even if they had continued to receive hospice care. Others may have survived because of therapies that they received after revoking hospice. With continued advances in medical care, some terminally ill patients with even advanced disease may benefit from life-prolonging therapies.
Those who revoke hospice may later resume hospice care if they meet the eligibility criteria; however, none of the patients in the subgroup analysis who died in the year after revocation returned to a VITAS hospice program before death. This is important because without hospice care, terminally ill patients, regardless of ethnicity, are less likely to have access to the services that hospice provides, including pain and symptom management, spiritual and psychosocial support, nursing care, and bereavement care for family.16,17 In a previous study of enrollees who left hospice before death, African American patients were significantly less likely than white patients to return to hospice.18 Because terminally ill African American patients are more likely than white patients to experience inadequate pain management, to experience problematic communication with their health care professional, and to report less overall satisfaction with care, African American patients who do not return to hospice before death may be at risk for dying without access to high-quality end-of-life care.19- 21
Of the enrollees who died in the year after revocation, the median time to death was only 45 days; the median length of stay among hospice enrollees nationally is 26 days.1 If effective, the therapies that these patients received after revocation may have prolonged survival by only days to weeks at most. As a result, many of them may have spent their last days in the hospital receiving life-prolonging therapies at substantial health care costs without realizing their treatment goals.
Models of health care delivery that couple curative and palliative therapies may address the discontinuity of care that results from revoking hospice to pursue life-prolonging therapies and may also ensure access to the comprehensive care that hospice provides. Several large hospices and private health insurers are piloting programs that allow patients to receive hospice care while continuing to receive therapies that may slow or stop the progression of their disease.22,23 This type of open access supports a spectrum of decisions and changing preferences among patients with life-limiting illness while ensuring access to palliative and supportive care. Such models may be especially attractive to African American patients by allowing them to receive life-prolonging therapies without forgoing services that may enhance their quality of life. In addition, large insurers who are currently piloting open access hospice programs report that the increased cost is negligible compared with the cost of other programs and report cost savings associated with reduced hospitalizations.22 However, because of Medicare reimbursements, most small hospices cannot afford to offer complex medical treatments such as chemotherapy or radiation therapy and may restrict admission to those who are willing to forgo these therapies, resulting in patients having to choose between potentially life-prolonging therapies and hospice care.24
Hospice bridge programs are another alternative that allows terminally ill patients to choose from a range of treatment options while receiving some of the services offered by hospice.25 These programs are often operated by a home health care agency and are affiliated with a hospice but do not restrict admission to patients with a prognosis of 6 months or less or those who are willing to forgo life-prolonging therapies. An interdisciplinary team manages the physical and psychosocial needs associated with life-limiting illness and provides a smooth transition to hospice when the patient and family are ready. Currently, there are few hospice bridge or open access programs in operation.
This analysis has some limitations. It is not clear whether our findings reflect cultural differences in preferences for end-of-life care or other factors not included in our model. We did not have data on functional status, symptom severity, or do-not-resuscitate orders, which have all been associated with withdrawal from hospice.26,27 Another limitation is that the survival analysis relied entirely on data from the Florida death registry. We were not able to include enrollees who may have left the state or were not included in the registry because of error. This likely represented a small proportion of enrollees. Our findings are similar to other analyses that indicate that more than half of those who withdraw from hospice are still alive at 6 months.28 The data were also drawn entirely from a for-profit hospice and may not be generalizable to the population of hospice enrollees throughout the United States. Finally, the accuracy of our analysis depends on the accuracy of disposition coding by hospice administrative staff. Ongoing systemwide data quality checks monitor the accuracy and timeliness of data entry at each program site.
In conclusion, as the profile of hospice enrollees becomes more diverse, physicians and policymakers face the challenge of accommodating a range of individual and cultural differences in values, beliefs, and preferences. Health care delivery models that accommodate either life-prolonging therapies or palliative and supportive care fail to account for the roles of each throughout the continuum of life-limiting illness, roles that are influenced by cultural beliefs and preferences. Future research should design and test innovative models of care that support the varied needs and preferences of terminally ill patients of all racial and ethnic groups and facilitate continuity throughout the course of life-limiting illness.
Correspondence: Kimberly S. Johnson, MD, MHS, Duke University Medical Center, Division of Geriatrics, DUMC Box 3003, Durham, NC 27710 (firstname.lastname@example.org).
Accepted for Publication: September 10, 2007.
Author Contributions: Dr Johnson had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Johnson, Kuchibhatla, and Tulsky. Acquisition of data: Johnson and Tanis. Analysis and interpretation of data: Johnson, Kuchibhatla, and Tulsky. Drafting of the manuscript: Johnson and Kuchibhatla. Critical revision of the manuscript for important intellectual content: Johnson, Tanis, and Tulsky. Statistical analysis: Kuchibhatla. Obtained funding: Johnson. Study supervision: Tulsky.
Financial Disclosure: This analysis uses data from the VITAS Healthcare Corporation. David Tanis, PhD, is the Director of Clinical Research for VITAS Healthcare.
Funding/Support: This study was supported by the Duke Mentored Clinical Research Scholars Program (NIH/NCRR 1K12RR17630).
Disclaimer: This analysis was not funded by VITAS and does not reflect the views of the VITAS Healthcare Corporation.
Previous Presentations: This study was presented in part at the American Geriatrics Society Annual Meeting; May 13, 2005; Orlando, Florida; and the American Geriatrics Society Annual Meeting; May 5, 2006; Chicago, Illinois.