Association Between Hospice Length of Stay, Health Care Utilization, and Medicare Costs at the End of Life Among Patients Who Received Maintenance Hemodialysis

Key Points

Question  What is the association between hospice length of stay and end-of-life health care utilization and costs among hemodialysis patients?

Findings  In this cross-sectional study of 770 191 Medicare beneficiaries, 20% of maintenance hemodialysis patients enrolled in hospice; of these, 41.5% received 3 days or fewer of hospice. While these patients with very short hospice stays were less likely than those without hospice stays to die in the hospital and receive intensive procedures, rates of hospitalization, intensive care unit admission, and Medicare costs were comparable to or higher than for patients who did not receive hospice.

Meaning  Late hospice referral may limit the effect of hospice on end-of-life health care utilization and costs among dialysis patients.

Importance  Patients with end-stage renal disease are less likely to use hospice services than other patients with advanced chronic illness. Little is known about the timing of hospice referral in this population and its association with health care utilization and costs.

Objective  To examine the association between hospice length of stay and health care utilization and costs at the end of life among Medicare beneficiaries who had received maintenance hemodialysis.

Design, Setting, and Participants  This cross-sectional observational study was conducted via the United States Renal Data System registry. Participants were all 770 191 hemodialysis patients in the registry who were enrolled in fee-for-service Medicare and died between January 1, 2000, and December 31, 2014. The dates of analysis were April 2016 to December 2017.

Main Outcomes and Measures  Hospital admission, intensive care unit (ICU) admission, and receipt of an intensive procedure during the last month of life; death in the hospital; and costs to the Medicare program in the last week of life.

Results  Among 770 191 patients, the mean (SD) age was 74.8 (11.0) years, and 53.7% were male. Twenty percent of cohort members were receiving hospice services when they died. Of these, 41.5% received hospice for 3 days or fewer. In adjusted analyses, compared with patients who did not receive hospice, those enrolled in hospice for 3 days or fewer were less likely to die in the hospital (13.5% vs 55.1%; P < .001) or to undergo an intensive procedure in the last month of life (17.7% vs 31.6%; P < .001) but had higher rates of hospitalization (83.6% vs 74.4%; P < .001) and ICU admission (54.0% vs 51.0%; P < .001) and similar Medicare costs in the last week of life ($10 756 vs $10 871; P = .08). Longer lengths of stay in hospice beyond 3 days were associated with progressively lower rates of utilization and costs, especially for those referred more than 15 days before death (35.1% hospitalized and 16.7% admitted to an ICU in the last month of life; the mean Medicare costs in the last week of life were $3221).

Conclusions and Relevance  Overall, 41.5% of hospice enrollees who had been treated with hemodialysis for their end-stage renal disease entered hospice within 3 days of death. Although less likely to die in the hospital and to receive an intensive procedure, these patients were more likely than those not enrolled in hospice to be hospitalized and admitted to the ICU, and they had similar Medicare costs. Without addressing barriers to more timely referral, greater use of hospice may not translate into meaningful changes in patterns of health care utilization, costs, and quality of care at the end of life in this population.

Use of hospice has increased markedly in the overall Medicare population in recent years,¹ with the percentage of Medicare beneficiaries who received hospice before death more than doubling from 23% in 2000 to 48% in 2014. While hospice offers clear benefits to patients approaching the end of life and to their families,^2-4 the Medicare hospice benefit, established in 1986, requires hospice enrollees to discontinue disease-modifying treatment for the condition listed as their hospice diagnosis. This may explain in part why, despite increases over time in the percentage of Medicare beneficiaries who receive hospice services at the end of life, hospice referral continues to occur very late in the course of illness for a substantial portion of those referred. From 2000 to 2009, the proportion of Medicare hospice beneficiaries not referred until the last 3 days of life increased from 22% to 28%.⁵ This is concerning because shorter hospice stays have been associated with lower-quality end-of-life care, including inadequate pain control and unmet emotional needs.^6,7

Nowhere is the tension between timely receipt of hospice care and continuation of treatments that may both extend life and palliate symptoms more pronounced than for those patients receiving maintenance hemodialysis. In instances where end-stage renal disease (ESRD) is viewed as the life-limiting condition, patients cannot receive concurrent Medicare coverage for hospice and hemodialysis. This Medicare payment policy, which has the effect of forcing patients to choose between receiving hospice services and continuing treatments considered to be “curative,”⁸ takes on a special significance for this population because most live for less than 1 week after stopping hemodialysis.⁹ Therefore, it is not surprising that patients who have been treated with maintenance hemodialysis are much less likely to be referred to hospice than other chronically ill populations.^10,11 However, less is known about the timing of hospice referral in this population.

In the overall Medicare population, even very short hospice stays appear to be associated with lower rates of hospital admission, intensive care unit (ICU) admission, and in-hospital death,¹² which may reflect higher-quality end-of-life care because evidence suggests that most Americans would prefer not to die in the hospital.^13,14 Patients with ESRD may stand to gain much from hospice care: many have limited life expectancy^15-20 and a high symptom burden^21-29 and receive extremely intensive care at the end of life,^11,30-32 which may be inconsistent with their preferences.³³ Whether the association between the timing of hospice referral and health care utilization and costs for these patients is similar to that for the overall Medicare population—for whom even very short stays in hospice are associated with lower utilization and costs compared with those not referred to hospice—is not known. To address this question, we designed a study to examine hospice length of stay and its association with health care utilization and costs at the end of life among Medicare beneficiaries who had been treated with maintenance hemodialysis.

In this cross-sectional observational study, we identified all patients in the United States Renal Data System (USRDS) registry who died between January 1, 2000, and December 31, 2014, and were enrolled in Parts A and B of fee-for-service Medicare during their final year of life (n = 857 589). The dates of analysis were April 2016 to December 2017. Analyses were conducted among the subset of 770 191 of these patients 18 years or older for whom the most recent treatment modality before death was hemodialysis (vs kidney transplant or peritoneal dialysis) (Figure 1). The institutional review board at the University of Washington approved the study protocol.

We grouped patients according to whether they were enrolled in hospice at the time of death, and, if so, by the duration of enrollment. The groups were as follows: (1) not enrolled in hospice at death, (2) enrolled within 3 days of death, (3) enrolled 4 to 7 days before death, (4) enrolled 8 to 14 days before death, and (5) enrolled for 15 days or more before death.

We used USRDS standard analysis files to ascertain each patient’s age at the time of death (categorized as 18-44, 45-64, 65-74, 75-84, and ≥85 years), sex, race, and Hispanic ethnicity. We used the USRDS payer history file to identify patients with dual Medicare-Medicaid eligibility in the year before death. Comorbid conditions (diabetes, cirrhosis, dementia, cancer, coronary artery disease, stroke, peripheral arterial disease, congestive heart failure, and emphysema) and Quan comorbidity score³⁴ were ascertained using International Classification of Diseases, Ninth Revision (ICD-9) diagnosis codes in the Medicare institutional and physician supplier claims for both inpatient and outpatient care at a time point 1 year before death based on claims during the preceding year (ie, 12-24 months before death). Regional health care intensity was based on the average inpatient health care costs during the last 6 months of life in each patient’s hospital referral region of residence closest to the time of death, as reported in the Dartmouth Atlas of Health Care³⁵ for 2014, and categorized by quintile.

We determined whether patients had a primary hospice diagnosis of renal failure based on the ICD-9 code recorded in the Medicare hospice file (the eTable in the Supplement lists specific ICD-9 codes for different primary hospice diagnoses). Time from first hemodialysis to death was calculated by subtracting each patient’s date of first ESRD service from their date of death recorded in the USRDS patients file. Whether there was a decision to discontinue hemodialysis before death was ascertained from the Centers for Medicare & Medicaid Services (CMS) ESRD death notification form (CMS-2746).

We examined the following measures of health care utilization during the last month of life: hospital admission, ICU admission, death in the hospital, and receipt of 1 or more inpatient intensive procedures. In addition, we assessed health care costs to the Medicare program under Medicare Parts A and B (hereafter referred to as “costs”) in the last week of life standardized to the 2014 consumer price index: these included costs for hemodialysis, hospice (paid to hospices at a per diem rate), acute inpatient care, and outpatient health care provider visits. Intensive care unit utilization was ascertained using intensive and coronary care unit revenue center codes in Medicare institutional claims (020x and 021x). Inpatient intensive procedures were identified using ICD-9 procedure codes in inpatient Medicare institutional claims and included intubation and mechanical ventilation (codes 96.04, 96.05, and 96.7x), tracheostomy (codes 31.1, 31.21, and 31.29), gastrostomy tube insertion (codes 43.2, 43.11, 43.19, 43.2, and 44.32), enteral or parenteral nutrition (codes 96.6 and 99.15), and cardiopulmonary resuscitation (codes 99.60 and 99.63).^11,36

We described patient characteristics by hospice length of stay using the mean (SD) or median (interquartile range [IQR]) values for continuous variables and percentages for categorical variables. We used generalized linear models to calculate marginal prevalence estimates and the mean Medicare costs by hospice length of stay after adjustment for potential confounders. Models were adjusted for age, sex, race, Hispanic ethnicity, dual Medicare-Medicaid eligibility, comorbid conditions (diabetes, cirrhosis, dementia, cancer, coronary artery disease, stroke, peripheral artery disease, congestive heart failure, and emphysema), Quan comorbidity score,³⁴ regional health care intensity, and whether there was a decision to discontinue hemodialysis before death. All analyses were conducted using Stata/SE (version 13.1; StataCorp LP), and 2-sided P values are presented.

Among 770 191 Medicare beneficiaries receiving maintenance hemodialysis who died between 2000 and 2014, a total of 154 186 (20.0%) were receiving hospice services at the time of death. Among these, the median length of stay in hospice was 5 days (IQR, 2-12 days). Overall, 41.5% of these patients enrolled in hospice within 3 days of death, 22.5% within 4 to 7 days, 14.3% within 8 to 14 days, and 21.7% within 15 days or more before death (Table). While the percentage of Medicare beneficiaries treated with hemodialysis who used hospice more than doubled from 11.0% in 2000 to 26.7% in 2014, the percentage of these patients spending 3 days or fewer in hospice was stable over this period (43.6% in 2000 to 41.7% in 2014).

As summarized in the Table, patients enrolled in hospice at the time of death were on average older than those not enrolled, but there were no large age differences among patients with differing hospice lengths of stay. Patients of black race were less likely than patients of white race to die in hospice but were more likely to spend longer time in hospice. Overall, comorbid disease burden was similar across groups, with the exception that a higher proportion of those who died in hospice had dementia or cancer compared with those who did not die in hospice. Dementia and cancer were also more prevalent among patients who spent a longer period in hospice. Only 13.9% of those who did not die in hospice discontinued hemodialysis before death compared with 66.3% of those who died in hospice. Among hospice users, 55.9% had a primary hospice diagnosis of renal failure (diagnoses for the remaining patients are listed in the eTable in the Supplement). Patients enrolled in hospice for longer periods were less likely to have discontinued hemodialysis and to have a primary hospice diagnosis of renal failure.

Figure 2 shows adjusted patterns of end-of-life health care utilization and Medicare costs by length of time in hospice (unadjusted findings are shown in the eFigure in the Supplement). The adjusted proportion of patients admitted to the hospital in the last month of life was lowest for those in hospice for 15 days or more (35.1%) and was highest for those in hospice for 3 days or fewer (83.6%) (ie, higher than for those who did not enroll in hospice [74.4%]). Intensive care unit admission in the last month of life followed a similar pattern, with 16.7% of those in hospice for 15 days or more admitted to the ICU compared with 54.0% of those in hospice for 3 days or fewer and 51.0% of those who did not enroll in hospice. The proportion of patients who received an inpatient intensive procedure was highest for those who did not enroll in hospice (31.6%) and was higher among those who spent a shorter time in hospice (ranging from 17.7% for those in hospice for ≤3 days to 3.0% for those in hospice for ≥15 days). While more than half of the patients not referred to hospice died in the hospital (55.1%), this was true for only 13.5% of those who spent 3 days or fewer in hospice and for less than 5% of patients who spent longer periods in hospice. Adjusted mean Medicare costs were similar for patients who spent 3 days or fewer in hospice and for those not referred to hospice ($10 756 vs $10 871; P = .08). Medicare costs and rates of all utilization measures decreased with increasing lengths of stay beyond 3 days.

Hospice utilization at the end of life for Medicare beneficiaries with ESRD who had been treated with maintenance hemodialysis increased markedly between 2000 and 2014. However, there was little change during this time in hospice length of stay for members of this population, with 41.5% of those dying in hospice enrolled for 3 days or less. For these patients enrolled in hospice very close to the time of death, rates of hospital admission and ICU admission were higher than, and Medicare costs similar to, those for patients not enrolled in hospice. Longer lengths of stay in hospice beyond 3 days were associated with progressively lower rates of health care utilization and costs, especially for those referred more than 15 days before death.

To our knowledge, this study (which uses data from 2000 to 2014) is the first to look at rates of hospice utilization among Medicare beneficiaries with ESRD since Murray et al³⁷ reported a rate of 14% using data from 2000 to 2003. Our study demonstrates that from 2000 to 2014 hospice utilization among Medicare beneficiaries who had been treated with maintenance hemodialysis more than doubled from 11.0% to 26.7% but still lags substantially behind that for the broader Medicare population. Over a similar period (2002-2012), rates of hospice utilization among the overall population of older Medicare beneficiaries increased from 26% to 47%.³⁸ Most strikingly, we observed comparatively late hospice referrals among Medicare beneficiaries who had been treated with maintenance hemodialysis that persisted throughout the period under study: 41.5% of hospice users in our study did not enroll until the last 3 days of life compared with the 25% reported for older Medicare hospice beneficiaries.⁵ Almost two-thirds (64.0%) of hospice users in our study received 1 week or less of hospice care compared with 39%, 36%, and 34% reported for Medicare hospice beneficiaries with heart failure, colorectal cancer, and dementia, respectively.³⁹ This is concerning because short hospice stays have been associated with inadequate pain control and unmet emotional needs.^6,7

To our knowledge, this study is the first to examine the association between hospice length of stay and patterns of health care utilization and costs near the end of life in the ESRD population. While prior studies among patients with ESRD have described strong associations between hospice referral and health care utilization^11,37 and Medicare costs,³⁷ we found that (consistent with prior literature in other patient populations^12,39-43) these associations are heavily contingent on the time spent in hospice and the specific measures examined. We found that longer lengths of stay in hospice were generally associated with lower health care utilization and costs. Being in hospice for more than 2 weeks was associated with markedly lower hospital and ICU admission rates, a total of just 4 to 7 days of hospice was associated with much lower Medicare costs and rates of intensive procedures, and even very short stays in hospice were associated with markedly lower rates of death in the hospital. However, for a number of measures, such as hospital and ICU admission in the last month of life and Medicare costs in the last week of life, those who enrolled in hospice within 3 days of death had similar costs and more intensive patterns of health care utilization than those not referred to hospice. While these patterns of utilization and costs may seem counterintuitive, they likely reflect a crisis-driven approach to hospice referral in which hospice serves as a last-minute “add on” to the intensive and costly patterns of end-of-life care previously described for members of this population.^44,45 Our findings also suggest that as long as close to half of hospice referrals occur within the last 3 days of life, efforts to promote hospice utilization in patients receiving maintenance hemodialysis are unlikely to have a significant effect on end-of-life costs and health care utilization.

While having the option to discontinue hemodialysis and exercise some control over the timing of death has been described by some as “a blessing,”⁹ others may be forced to discontinue hemodialysis before they are ready so as to receive hospice services. To better understand the implications of late referral to hospice for patients receiving maintenance hemodialysis, more work is needed to gain a deeper understanding of the end-of-life experiences of patients treated with maintenance hemodialysis and their family members and friends. For those in whom more timely receipt of hospice care would be beneficial, efforts are needed to understand barriers to and facilitators of hospice referral and identify opportunities for more timely transitions to hospice.

While a change in Medicare payment policy to allow for concurrent receipt of hemodialysis and hospice services for these patients would be needed to allow for greater flexibility in the timing of hospice referral in the ESRD population, this alone may not be sufficient to improve access to timely hospice care for members of this population. Interventions targeted at other potential barriers to timely hospice referral may also be needed. Compared with most other broadly defined populations with advanced chronic illness, a distinguishing feature of patients treated with maintenance hemodialysis is that they are already receiving 1 form of chronic life support. In so doing, they have already been drawn into a highly medicalized “life at any cost”⁴⁶ disease-oriented model of care in which patients may have unrealistic prognostic expectations.⁴⁷ Furthermore, renal failure may be seen by both patients and clinicians as a problem that can be “fixed” with hemodialysis, with hospice and palliative care viewed as treatments of last resort. Illness trajectories may also be less predictable for patients with advanced organ failure than for those with cancer, perhaps contributing to shorter hospice stays.^48-50 Nevertheless, hospice referral among members of our cohort who had been receiving hemodialysis appears to occur much later in the course of illness than for those with other forms of organ failure (eg, heart failure and chronic lung disease) characterized by a similar degree of prognostic uncertainty.³⁹

Earlier and more frequent integration of palliative care services into the care of patients receiving hemodialysis is an intervention that could potentially target a number of these barriers. While Medicare payment policy serves as a barrier to concurrent use of hemodialysis and hospice services for those with a life-limiting diagnosis of renal failure, rates of palliative care in this population are also low,³³ despite the absence of such policy barriers to concurrent receipt of palliative care services and hemodialysis. Earlier integration of palliative care could likely address the substantial and often unrecognized symptom, functional, and caregiving burden faced by many hemodialysis patients and their family members and friends, especially for those who are older or frail. Concurrent receipt of hemodialysis and palliative care services earlier in the illness trajectory could perhaps also allow for a smoother, less crisis-driven transition to hospice closer to the end of life.

Our study has several limitations. Although the USRDS registry includes all US patients receiving maintenance hemodialysis, the registry does not include patients with ESRD who are not treated with hemodialysis. Therefore, we cannot report on hospice utilization among this important subset of the population with ESRD. Moreover, the generalizability of our findings to patients enrolled in Medicare Advantage, patients with private insurance, and those covered by federal programs, such as the Department of Veterans Affairs, requires further study. In addition, the USRDS does not collect information about patient and family treatment preferences; therefore, we could not assess whether care was concordant with such preferences. Another limitation is that the CMS ESRD death notification form (CMS-2746) provides limited information on the clinical context in which hemodialysis discontinuation occurred. Last, and perhaps most important, although our study showed strong associations between hospice length of stay and health care utilization and costs, it is an observational study, and conclusions cannot be drawn about whether lengthening hospice stays would lead to decreased utilization and costs.

Hospice referral among patients treated with maintenance hemodialysis occurs very late in the course of illness as an apparent add-on to intensive patterns of care. Without measures to make hospice accessible earlier in the illness trajectory or efforts to better integrate palliative care services into the care of patients receiving maintenance hemodialysis, greater use of hospice is unlikely to translate into meaningful changes in patterns of health care utilization, costs, and quality of end-of-life care in the ESRD population.

Accepted for Publication: January 13, 2018.

Corresponding Author: Melissa W. Wachterman, MD, MSc, MPH, Section of General Internal Medicine, Veterans Affairs Boston Healthcare System, 150 S Huntington Ave, Building 9, Boston, MA 02130 (mwachterman@partners.org).

Published Online: April 30, 2018. doi:10.1001/jamainternmed.2018.0256

Author Contributions: Dr O’Hare 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: Wachterman, Keating, O’Hare.

Acquisition, analysis, or interpretation of data: Wachterman, Hailpern, Kurella Tamura, O’Hare.

Drafting of the manuscript: Wachterman, Hailpern.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Wachterman, Hailpern, O’Hare.

Obtained funding: Wachterman, Kurella Tamura, O’Hare.

Administrative, technical, or material support: Kurella Tamura, O’Hare.

Study supervision: O’Hare.

Conflict of Interest Disclosures: None reported.

Funding/Support: This work was supported by grant U01DK102150 from the National Institute of Diabetes and Digestive and Kidney Diseases. Dr Wachterman also received support from the National Palliative Care Research Center Junior Faculty Career Development Award and by grant 1K23AG049088 from the National Institute on Aging.

Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: The interpretation of these data is the sole responsibility of the authors and does not reflect the opinion of the Department of Veterans Affairs or the United States Renal Data System.

Additional Contributions: Gabrielle Kate Alicante, BS, Harvard Medical School, assisted with the preparation of the manuscript, including formatting tables and figures, as part of her duties as a paid research assistant.