Hospice Enrollment Saves Money For Medicare And Improves Care Quality Across A Number Of Different Lengths-Of-Stay

Abstract

Despite its demonstrated potential to both improve quality of care and lower costs, the Medicare hospice benefit has been seen as producing savings only for patients enrolled 53–105 days before death. Using data from the Health and Retirement Study, 2002–08, and individual Medicare claims, and overcoming limitations of previous work, we found $2,561 in savings to Medicare for each patient enrolled in hospice 53–105 days before death, compared to a matched, nonhospice control. Even higher savings were seen, however, with more common, shorter enrollment periods: $2,650, $5,040, and $6,430 per patient enrolled 1–7, 8–14, and 15–30 days prior to death, respectively. Within all periods examined, hospice patients also had significantly lower rates of hospital service use and in-hospital death than matched controls. Instead of attempting to limit Medicare hospice participation, the Centers for Medicare and Medicaid Services should focus on ensuring the timely enrollment of qualified patients who desire the benefit.

As of 2012, 5 percent of the most seriously ill Americans accounted for more than 50 percent of health care spending, with most costs incurred in the last year of life as a result of hospital-based treatment.^1–3 Despite those high and escalating health care costs, numerous studies demonstrate that seriously ill patients and their families receive suboptimal care, characterized by untreated pain and physical symptoms, spiritual and emotional distress, high family caregiving burdens, and unnecessary or unwanted treatments inconsistent with their previously stated wishes and goals for care.^4–11

Hospice has been shown to greatly improve the quality of care for patients and their families near the end of life. Under Medicare Part A, the hospice benefit covers palliative care services delivered by a team of professionals, including physicians, nurses, social workers, chaplains, home health aides, and volunteers, to dying patients—that is, patients with a life expectancy of six months or less—who are willing to forgo curative treatments.¹²

Studies have consistently demonstrated that hospice is associated with reductions in symptom distress, improved outcomes for caregivers, and high patient and family satisfaction.^8,13–15 Recent evidence also indicates that continuous hospice use reduces the use of hospital-based services—including emergency department visits and intensive care unit stays—and the likelihood of death in the hospital.¹⁶

The number of hospices has increased rapidly over the past twenty years, making hospice programs available to almost all eligible Americans.¹⁷ Medicare hospice spending has risen considerably with the growth and development of new hospice programs, particularly in the for-profit sector, and the resulting rise in the number of patients accessing the hospice benefit.^18,19

This increase in spending has led the Centers for Medicare and Medicaid Services to explore methods of containing Medicare hospice spending, such as through payment reform or investigation of hospices with long lengths-of-stay.²⁰ What is not known, however, is how the length of hospice enrollment relates to overall Medicare spending at the end of life—including what periods of enrollment might decrease net Medicare costs as compared to usual care and, if they do, by how much.

The length of hospice enrollment that might achieve the greatest cost savings to Medicare is the subject of considerable debate. Some scholars have argued that beneficiaries must be enrolled in hospice longer than current practice to achieve financial savings under Medicare.^21–23 Others have found that longer hospice length-of-stay is associated with higher Medicare spending—particularly for those with noncancer diagnoses.²⁴

In the largest and most rigorous study to date, Donald Taylor and colleagues observed that hospice enrollment 53–105 days before death maximized Medicare savings compared to usual nonhospice care.²³ However, this study has been criticized for its inability to control for factors not present in Medicare claims that are known to be associated with higher costs, such as patients’ functional status.²⁵

Another criticism cited notable differences between the hospice and control groups: Hospice users had greater costs in the period preceding hospice enrollment compared with their matched controls.²⁵ Such limitations cast doubt on the validity of the reported findings regarding both the timing of hospice enrollment to maximize savings and the magnitude of those savings.

Health care reform in the past decade has sharpened the focus on increasing the value of health care and on forging effective policy to guide that process. A clearer understanding of the value of existing Medicare programs thus is required. In this study we aimed to better understand the value of Medicare hospice by examining the relationship between length of hospice enrollment and overall Medicare costs.

Specifically, we compared Medicare costs for patients receiving hospice care to those of patients not receiving hospice care across four different periods of hospice enrollment: 1–7, 8–14, and 15–30 days before death, the most common enrollment periods, and 53–105 days before death. In addition, we investigated both the source of hospice-related savings, if any, such as decreased hospital admissions and fewer hospital and intensive care unit days, and the impact of hospice on selected measures of quality of care at the end of life, including thirty-day readmission rates and in-hospital death rates.

We used the rich survey data from the Health and Retirement Study, in combination with individual Medicare claims, and adjusted for previously unmeasured factors known to influence costs, such as functional status and social characteristics. These analyses revealed that net savings to Medicare are not limited to hospice enrollment 53–105 days prior to death but are also observed across the most common enrollment periods: 1–7, 8–14, and 15–30 days before death.

Study Data And Methods

We examined data from the Health and Retirement Study, a longitudinal survey administered to a nationally representative cohort of adults over age fifty. Serial interviews are conducted every two years and include information on participants’ demographic, economic, social, and functional characteristics. Each interview cycle, participants who died since the last interview are identified, and dates of death are drawn from the National Death Index. More than 80 percent of participants provided authorization to merge their survey data with Medicare claims,^26,27 a necessary step in the present analysis.

Sample

We sampled all survey participants who died during 2002–08. We included those age sixty-five or older who had continuous Medicare Parts A and B coverage for twelve months prior to death, while excluding those enrolled with Medicare managed care (for whom claims data were therefore incomplete). This methodology yielded a final sample of 3,069 people, both enrolled and not enrolled in Medicare hospice prior to death.

For the analyses of each enrollment period, we also excluded those who enrolled in hospice prior to the study outcome period (7, 14, 30, and 105 days, respectively) and those whose final predeath interview took place within the study period.

Measures

We categorized periods of enrollment in Medicare hospice before death based on the number of days prior to death that enrollment occurred, as follows: 53–105 days (the period expected to maximize reduction in Medicare spending),²³ 15–30 days, 8–14 days, and 1–7 days. For each period, the primary outcome was total Medicare spending measured from the beginning of the enrollment period to death.

We adjusted expenditures for inflation (2008 dollars) and for geographic differences in Medicare prices. We also examined six other measures of care utilization: hospital admissions, hospital and intensive care unit days, intensive care unit admission (any or none), thirty-day hospital readmission (any or none), and in-hospital death.

We selected independent variables based on our conceptual framework, “Determinants of Treatment Intensity for Patients with Serious Illness,” which postulates that treatment intensity is influenced by both regional and patient or family determinants.²⁸ We selected variables that could serve as empirical measures of each construct in the conceptual model: age; sex; race or ethnicity; education; net worth; marital status; insurance coverage; functional status; residential status; medical conditions; and regional supply of hospital beds, specialist physicians, and local hospital care intensity.

Variables were drawn from Health and Retirement Study data, individual Medicare claims, and the Dartmouth Atlas of Health Care.²⁹ Additional details are provided in the online Appendix.³⁰

Statistical Analyses

We employed doubly robust methods combining propensity score matching and regression adjustment.³¹ We first determined hospice enrollment in relation to date of death from individual Medicare hospice claims. For each enrollment period, we then developed propensity scores for hospice and nonhospice patients to estimate each subject’s likelihood of hospice enrollment during the specified period.

We used logistic regression to estimate the likelihood of hospice enrollment using all of the independent variables, described above, that may be associated with treatment intensity. Additionally, we included as a covariate the number of hospital days prior to the target hospice enrollment period up to six months before death, to account for prior utilization as a predictor of subsequent utilization.

We then matched hospice enrollees to one or many nonhospice controls within ±0.02 of the standard deviation of the propensity scores. Unmatched subjects were excluded. This procedure was completed for each enrollment period, resulting in the following sample sizes: 1,801 (1–7 days), 1,506 (8–14 days), 1,749 (15–30 days), and 1,492 (53–105 days).

We examined bivariate comparisons of unadjusted measures of spending and use, as well as patient characteristics, using the matched, weighted samples. We then conducted multivariable regressions for each of the outcome measures, once again adjusting for all independent variables.

Following the estimation of each fully adjusted regression, we examined the adjusted means, including 95 percent confidence intervals, and incremental effects in outcomes between groups of hospice enrollees and matched nonhospice controls. Additional details are provided in the online Appendix.³⁰ Analyses were conducted using the statistical analysis software Stata, version 11.

Limitations

Three study limitations are worth noting. First, the data are retrospective, following back from date of death—that is, we employed a mortality follow-back design. This retrospective approach artificially removed the prognostic uncertainty faced by patients and physicians when making treatment decisions. The mortality follow-back design and our inability to randomly assign patients to treatment groups may therefore have biased the results.

However, by using detailed survey data, propensity score matching procedures, and multivariable regression to adjust the results, we minimized the effect of this bias more than could have been achieved through the use of administrative claims data alone.

Second, we were unable to factor into the analysis direct measures of individual preferences and goals of care. We did, however, adjust for all available characteristics known to be potentially associated with treatment preferences, such as education, race, and debility.

Third, we were not able to fully assess quality of care, which, in combination with cost, determines value. We included among our secondary outcomes two markers of potentially low-quality care: thirty-day hospital readmission and in-hospital death. In addition, many prior studies have demonstrated high quality of and satisfaction with hospice and palliative care.^{8,13–15,32–36}

Study Results

Subject Characteristics

Among the 3,069 subjects, 1,064 (35 percent) were enrolled in hospice prior to death. The mean hospice length-of-stay was 49 days (median 16 days, range 1–362 days). Patient and regional characteristics of subjects are reported in Appendix Exhibit 1.³⁰ Subjects’ mean age at death was eighty-three years. Subjects were predominantly non-Hispanic white (80 percent), female (56 percent), covered by supplemental private insurance (50 percent), and educated through high school or beyond (58 percent). Fifty-eight percent reported needing no assistance with basic activities of daily living leading up to the study period, while 21 percent resided in a nursing home. Twenty-three percent were eligible for both Medicare and Medicaid.

Hospice Enrollment For 53–105 Days

Eighty-eight (70 percent) subjects enrolled in hospice for 53–105 days prior to death were matched to 1,404 decedents not enrolled in hospice for 53 days or more prior to death. There were no significant differences in patient or regional characteristics between the two groups (Appendix Exhibit 2).³⁰

In fully adjusted analyses of outcomes spanning the last 105 days of life, subjects enrolled in hospice for 53–105 days prior to death had significantly lower mean total Medicare expenditures than matched controls ($22,083 versus $24,644, p < 0.01) (Exhibit 1). Hospice enrollees during this period also had fewer hospital admissions, intensive care unit admissions, hospital days, thirty-day hospital readmissions, and in-hospital deaths (all p < 0.01) compared to nonhospice enrollees. Differences between the groups’ total intensive care unit days were not significant in the fully adjusted model (p ¼ 0.11). Additional details are provided in Appendix Exhibit 3.³⁰

Exhibit 1.

Health Care Use At The End Of Life For Subjects Enrolled In Hospice And Matched Nonhospice Controls

Measure of use	Hospice group, adjusted means	Propensity score matched controls, adjusted means
TOTAL MEDICARE EXPENDITURES, 2008 US DOLLARS
Last 105 daysa	22,083	24,644b
Last 30 daysc	10,383	16,814b
Last 14 daysd	5,698	10,738b
Last 7 dayse	4,806	7,457b
TOTAL HOSPITAL DAYS
Last 105 daysa	3.50	12.50b
Last 30 daysc	1.60	5.70b
Last 14 daysd	0.19	4.36b
Last 7 dayse	0.29	1.20b
TOTAL HOSPITAL ADMISSIONS
Last 105 daysa	0.58	1.22b
Last 30 daysc	0.34	0.74b
Last 14 daysd	0.08	0.48b
Last 7 dayse	0.12	0.35b
TOTAL ICU DAYS
Last 105 daysa	0.71	5.65
Last 30 daysc	0.31	2.91f
Last 14 daysd	0.03	1.61
Last 7 dayse	0.08	0.57b
PROPORTION WITH ICU ADMISSION
Last 105 daysa	0.15	0.37b
Last 30 daysc	0.10	0.31b
Last 14 daysd	0.02	0.23b
Last 7 dayse	0.05	0.15b
PROPORTION WITH 30-DAY HOSPITAL READMISSION
Last 105 daysa	0.11	0.26b
Last 30 daysc	0.02	0.12b
PROPORTION DYING IN THE HOSPITAL
Last 105 daysa	0.02	0.42b
Last 30 daysc	0.06	0.44b
Last 14 daysd	0.09	0.48b
Last 7 dayse	0.15	0.53b

SOURCE Authors’ analysis of Health and Retirement Study data linked to Medicare claims. NOTES Sample sizes vary across periods of enrollment. For enrollment 53–105 days before death: hospice patients, n = 88; matched controls, n = 1, 404. For enrollment 15–30 days before death: hospice patients, n = 133; matched controls, n = 1, 616. For enrollment 8–14 days before death: hospice patients, n = 90, matched controls, n = 1, 416. For enrollment 1–7 days before death: hospice patients, n = 308; matched controls, n = 1, 493. Multivariable regression models adjusted for age; sex; race/ethnicity; education; net worth; marital status; insurance coverage; functional status; residential status; medical conditions; and regional supply of hospital beds, specialist physicians, and local hospital care intensity. 95 percent confidence intervals for all estimates are available in the online Appendix (see Note 30 in text). ICU is intensive care unit.

^aHospice enrollment 53–105 days before death.

^bDifference between hospice and control groups statistically significant at p < 0.01.

^cHospice enrollment 15–30 days before death.

^dHospice enrollment 8–14 days before death.

^eHospice enrollment 1–7 days before death.

^fDifference between hospice and control groups statistically significant at p < 0.05.

Hospice Enrollment For 15–30 Days

One hundred thirty-three (80 percent) subjects enrolled in hospice for 15–30 days prior to death were matched to 1,616 decedents not enrolled in hospice for 15 days or more prior to death. There were no significant differences in patient or regional characteristics between the two groups (Appendix Exhibit 4).³⁰

In fully adjusted analysis of outcomes spanning the last thirty days of life, subjects enrolled in hospice for fifteen to thirty days prior to death had significantly lower average total Medicare expenditures than matched controls ($10,383 versus $16,814, p < 0.01) (Exhibit 1). Those enrolled in hospice during this period also had fewer hospital admissions, intensive care unit admissions, hospital days, intensive care unit days, thirty-day hospital readmissions, and in-hospital deaths (all p < 0.05). Additional details are provided in Appendix Exhibit 5.³⁰

Hospice Enrollment For 8–14 Days

Ninety (70 percent) subjects enrolled in hospice for 8–14 days prior to death were matched to 1,416 decedents not enrolled in hospice for 8 days or more days prior to death. Again, we found no significant differences in patient or regional characteristics between the two groups (Appendix Exhibit 6).³⁰

In fully adjusted analysis of outcomes spanning the last fourteen days of life, subjects enrolled in hospice for eight to fourteen days prior to death had significantly lower average total Medicare expenditures than matched controls ($5,698 versus $10,738, p < 0.01) (Exhibit 1). Once again, we found that those enrolled in hospice during this period also had fewer hospital admissions, intensive care unit admission, hospital days, and in-hospital deaths (all p < 0.01).

The hospice group had fewer intensive care unit days than the nonhospice group, but this difference did not reach statistical significance (p ¼ 0.11). Additional details are provided in Appendix Exhibit 7.³⁰

Hospice Enrollment For 1–7 Days

Three hundred eight (80 percent) subjects enrolled in hospice for 1–7 days prior to death were matched to 1,493 decedents not enrolled in hospice for 7 days or more prior to death. There were no significant differences in patient or regional characteristics between the two groups (Appendix Exhibit 8).³⁰

In fully adjusted analysis of outcomes spanning the last seven days of life, subjects enrolled in hospice for one to seven days prior to death had significantly lower average total Medicare expenditures than matched controls ($4,806 versus $7,457, p < 0.01) (Exhibit 1). Consistent with those patterns observed in other enrollment periods, those enrolled in hospice during this period also had fewer hospital admissions, intensive care unit admissions, hospital days, intensive care unit days, and in-hospital deaths (all p < 0.01).

Comparing Outcomes Across Hospice Enrollment Periods

Exhibits 2–4 compare the incremental effects in outcomes between subjects enrolled in hospice and nonhospice matched controls across the study periods. The adjusted savings in total Medicare spending ranged from $2,561 for those enrolled 53–105 days prior to death to $6,430 for those enrolled 15–30 days (Exhibit 2).

Exhibit 2.

Incremental Savings In Medicare Expenditures, By Various Lengths Of Hospice Enrollment Before Death With Matched Nonhospice Controls

SOURCE Authors’ analysis of Health and Retirement Study data linked to Medicare claims. NOTE Total savings to Medicare denote the incremental difference in Medicare spending between hospice and nonhospice groups.

Exhibit 4.

Incremental Reductions In Hospital Deaths, Intensive Care Unit Admissions, And Thirty-Day Readmissions, By Various Lengths Of Hospice Enrollment Before Death With Matched Nonhospice Controls

SOURCE Authors’ analysis of Health and Retirement Study data linked to Medicare claims. NOTES Incremental reduction in various outcomes (in-hospital deaths, ICU admissions, and thirty-day hospital readmissions) is expressed as the incremental effect in proportion between hospice and nonhospice groups. ICU is intensive care unit.

The adjusted decrease in total hospital days ranged from 9.0 for those enrolled 53–105 days prior to death to 0.9 for those enrolled 1–7 days, and the decrease in intensive care unit days ranged from 4.9 for those enrolled 53–105 days to 0.5 days for those enrolled 1–7 days (Exhibit 3). The adjusted reduction in in-hospital deaths was similar across groups, and the adjusted reductions in intensive care unit admissions and thirty-day hospital readmissions were largest for those enrolled for 53–105 days (Exhibit 4).

Exhibit 3.

Incremental Reductions In Hospital Days And Intensive Care Unit Days, By Various Lengths Of Hospice Enrollment Before Death With Matched Nonhospice Controls

SOURCE Authors’ analysis of Health and Retirement Study data linked to Medicare claims. NOTE Hospital and intensive care unit (ICU) days avoided is expressed as the incremental effect in days between hospice and nonhospice groups.

Discussion

Medicare costs for patients enrolled in hospice were significantly lower than those of nonhospice enrollees across all periods studied: 1–7 days, 8–14 days, and 15–30 days, the most common enrollment periods prior to death, as well as 53–105 days, the period previously shown to maximize Medicare savings.²³

In addition, reductions in the use of hospital services at the end of life both contribute to these savings and potentially improve quality of care and patients’ quality of life. Specifically, hospice enrollment was associated with significant reductions in hospital and intensive care unit admissions, hospital days, and rates of thirty-day hospital readmission and in-hospital death.

Evidence Of Medicare Savings

Our results not only are consistent with prior studies for Medicare spending, but they also strengthen this evidence by replicating the results within a sample more thoroughly matched for individual health, functional, and social characteristics, as well as regional factors. Finding no difference between the hospice and control groups’ pre-enrollment health care use is evidence of this improved match, as compared to prior work.²³

Specifically, Taylor and colleagues reported a maximum reduction in Medicare spending among patients enrolled in hospice for 53–105 days prior to death.²³ We found Medicare savings among this group, too, but we also found a similar level of savings among those enrolled for 1–7 days and increased savings among those enrolled for 8–30 days prior to death. Furthermore, we demonstrated parallel reductions in hospital and intensive care unit use, hospital readmissions, and in-hospital death.

Increasing Value Through Medicare Hospice

These findings, albeit limited to enrollment up to 105 days, are of particular importance because they suggest that investment in the Medicare hospice benefit translates into savings overall for the Medicare system. For example, if 1,000 additional beneficiaries enrolled in hospice for 15–30 days prior to death, Medicare could save more than $6.4 million, while those beneficiaries would be spared 4,100 hospital days. Alternatively, if 1,000 additional beneficiaries enrolled in hospice for 53–105 days before death, the overall savings to Medicare would exceed $2.5 million.

Although our findings suggest that hospice enrollment results in savings to the Medicare program across a number of different lengths-of-stay, this work also highlights several areas for future research.

First, because of the limitations of our data set, we were unable to precisely determine the point at which hospice approaches usual care in terms of costs. Future studies will be needed to address this question.

Second, our data were also not able to identify the differential effects of hospice on specific diagnoses. This is of particular importance given the recent growth of for-profit hospices, which typically enroll more patients with noncancer diagnoses (and longer average lengths-of-stay) compared to not-for-profit programs.

We found that net Medicare savings for patients with longer lengths-of-stay are lower because of the per diem cost of hospice services. However, we note that if 1,000 additional beneficiaries enrolled in hospice for 53–105 days before death, these beneficiaries could avoid 9,000 hospital days at the end of life. Indeed, our findings suggest that substantial reduction in hospital days—a primary goal of health care reform—is achieved regardless of the length of hospice enrollment.

Finally, our findings cannot be extrapolated to novel models of health care delivery or reimbursement, such as the integration of hospice programs into accountable care organizations or graded per diem payment systems, higher reimbursement for earlier and later days of enrollment, and lower reimbursement for the middle days.^20,37 The ability of these models to achieve savings while maintaining or improving quality is unclear and must be evaluated.

Barriers To Timely Hospice Enrollment

Our results, when taken together with those of prior studies, suggest that hospice increases value by improving quality and reducing costs for Medicare beneficiaries at the end of life. Yet aggressive efforts to curtail Medicare hospice spending, including the Office of Inspector General’s investigation of hospices that enroll patients with late-stage diseases but unpredictable prognoses, are ongoing.

Our findings suggest that these efforts may be misguided. Indeed, this study reveals that savings are present for both cancer patients and noncancer patients and that reductions in the use of hospital services and numbers of hospital days, hospital admissions, and hospital deaths appear to grow as the period of hospice enrollment lengthens within the observed study period (up to 105 days). These outcomes not only are less costly but also have all been associated with higher quality of care and increased concordance with patients’ preferences.

Although sample-size limitations prevented us from examining enrollment beyond 105 days, the trend in our data and the projections by Taylor and colleagues support the idea that efforts to curtail hospice enrollment may actually increase use and spending overall. Instead of working to reduce Medicare hospice spending and creating a regulatory environment that discourages continued growth in hospice enrollment, the Centers for Medicare and Medicaid Services should focus on ensuring that patients’ preferences are elicited earlier in the course of their diseases and that those who want hospice care receive timely referral.

An additional barrier to timely hospice referral may be limited knowledge or misconceptions regarding hospice and palliative care.³⁸ In particular, the hospice requirement to forgo curative treatments—even if they might not be beneficial—may be difficult for patients and families to accept or prompt fears of health care rationing. Because some treatments may be used for both curative and palliative purposes, this regulation and the variability with which hospice providers interpret it may also cause clinicians to be uncertain about hospice eligibility.³⁹

Several recent state and federal policy initiatives are designed to promote patient-centered care, specifically by increasing palliative care education among all health professionals and requiring that clinicians apprise patients of palliative treatment options early in the course of a serious illness.^40–42 Such efforts to elucidate patients’ preferences and values early may increase timely referral to hospice.

Finally, highly specialized and fragmented care may also present a barrier to hospice access, particularly for patients with the most complex and highest-cost illnesses: those 5 percent of patients, many in their last year of life, who account for nearly half of the nation’s health care spending.^1–3 Not only is care for this group characterized by costly hospital-based treatment, but it is also often highly fragmented and of poor quality, particularly among those who are dually eligible for Medicare and Medicaid.⁴³ Although many demonstration projects seek to address this concern,⁴³ few target this population’s need for assistance in identifying individualized goals of care and developing comprehensive treatment plans to achieve those goals.

One such comprehensive treatment approach might be the enhancement of formal partnerships between hospital palliative care teams and hospice. Evidence from existing models that incorporate hospital palliative care services demonstrates improvement in quality indicators, heightened patient and family satisfaction, reduced hospital use, and increased rates of hospice referral.⁴⁴ These benefits may be even more substantial if formal relationships between established palliative care teams and community hospice programs were developed in order to offer a bridge to timely hospice enrollment.

Conclusion

Hospice enrollment during the longer period of 53–105 days prior to death and the most common period within 30 days prior to death lowers Medicare expenditures, rates of hospital and intensive care unit use, 30-day hospital readmissions, and in-hospital death. Building upon prior studies of hospice and palliative care that have demonstrated higher quality and improved patient and family satisfaction,^{8,13–15,32–36} this finding suggests that hospice and palliative care are critical components in achieving greater value through health care reform: namely, improved quality and reduced costs.

Medicare should thus seek to expand access to hospice services so that hospice can contribute to its full potential to the overall value of care. To do so, substantial barriers to timely hospice enrollment must be overcome. The Centers for Medicare and Medicaid Services should abandon efforts to reduce Medicare hospice spending and delay hospice enrollment and should instead focus on ensuring that people who want hospice care receive timely referral.

Within the current Medicare hospice benefit, several approaches may expand access and increase appropriate and timely referral to hospice. These approaches include formalized partnerships between hospital palliative care programs and community hospice programs and the promotion of patient-centered care by educating patients, families, and physicians about the availability and benefits of hospice and palliative care services.

Finally, ongoing demonstration projects and novel models of health care delivery and reimbursement should place a high priority on the rigorous evaluation of hospice service use and its impact on the value of care.

Biographies

Amy S. Kelley is an assistant professor in the Brookdale Department of Geriatrics and Palliative Medicine, Icahn School of Medicine at Mount Sinai.

In this month’s Health Affairs, Amy Kelley and coauthors report on their study examining Medicare costs for hospice patients enrolled for different lengths-of-stay, ranging from 1 day to 105 days. Using data from the Health and Retirement Study and individual Medicare claims, they found savings for Medicare across all lengths-of-stay examined. Hospice patients also had less hospital use than matched controls, and thus a higher quality of life. The authors argue that instead of attempting to limit Medicare hospice participation for fear of not seeing savings, the Centers for Medicare and Medicaid Services should focus on ensuring the timely enrollment of qualified patients who desire the benefit.

Kelley is an assistant professor in the Brookdale Department of Geriatrics and Palliative Medicine, Icahn School of Medicine at Mount Sinai, and is a board-certified physician in internal medicine, geriatric medicine, and palliative medicine. Her research focuses on improving the quality of care for older adults with serious medical illness. She is particularly interested in regional practice variations and the relationship between patient characteristics and treatment intensity.

In 2012 Kelley was selected for the Paul B. Beeson Career Development Award in Aging Research from the National Institute on Aging and won the American Geriatrics Society’s best paper award in geriatrics research. Kelley earned a master’s degree in health services from the University of California, Los Angeles, and a medical degree from Cornell University.

Partha Deb is a professor and director of graduate studies in the Department of Economics at Hunter College.

Partha Deb is a professor and director of graduate studies in the Department of Economics at Hunter College and a professor at the Graduate Center, City University of New York. He is also an adjunct professor at the School of Public Health, Hunter College; a senior adviser at the Center for Medicare and Medicaid Innovation, Department of Health and Human Services; a research associate at the National Bureau of Economic Research; and a faculty fellow at the Brookdale Center for Healthy Aging and Longevity, Hunter College. Deb also serves on the editorial board of Health Services Research. He earned a master’s degree and a doctorate in economics from Rutgers University.

Qingling Du is a statistician in the Brookdale Department of Geriatrics and Palliative Medicine, Icahn School of Medicine at Mount Sinai.

Qingling Du is a statistician in the Brookdale Department of Geriatrics and Palliative Medicine, Icahn School of Medicine at Mount Sinai. Her work focuses on developing statistical models to study health care delivery systems. Du earned a master’s degree in statistics from the University of Chicago.

Melissa D. Aldridge Carlson is an assistant professor in the Brookdale Department of Geriatrics and Palliative Medicine, Icahn School of Medicine at Mount Sinai.

Melissa Aldridge Carlson is an assistant professor in the Brookdale Department of Geriatrics and Palliative Medicine, Icahn School of Medicine at Mount Sinai, and the director of research methods training for the Mount Sinai Medical Student Training in Aging Research Program. She is a member of the National Palliative Care Research Center’s Scientific Review Committee and serves on the editorial board of the Journal of Palliative Medicine. She earned an MBA from New York University, a master’s degree in public health from Columbia University, and a doctorate in health policy and administration from Yale University.

R. Sean Morrison is a tenured professor in the Brookdale Department of Geriatrics and Palliative Medicine, Icahn School of Medicine at Mount Sinai.

Sean Morrison is a tenured professor in the Brookdale Department of Geriatrics and Palliative Medicine, Icahn School of Medicine at Mount Sinai; director of the school’s Hertzberg Palliative Care Institute; and the Herman Merkin Professor of Palliative Care. He is the director of the National Palliative Care Research Center and was the president of the American Academy of Hospice and Palliative Medicine. Morrison serves on the editorial board of Palliative Medicine and is the senior associate editor of the Journal of Palliative Medicine. He earned a medical degree from the University of Chicago.