Will Changes to Medicare Payment Rates Alter Hospice's Cost-Saving Ability?

Abstract

Background:

On January 1, 2016, Medicare implemented a new “two-tiered” model for hospice services, with per diem rates increased for days 1 through 60, decreased for days 61 and greater, and service intensity add-on payments made retrospectively for the last seven days of life.

Objective:

To estimate whether the Medicare hospice benefit's potential for cost savings will change as a result of the January 2016 change in payment structure.

Design:

Analysis of decedents' claims records using propensity score matching, logistic regression, and sensitivity analysis.

Setting/Subjects:

All age-eligible Medicare decedents who received care and died in North Carolina in calendar years 2009 and 2010.

Measurements:

Costs to Medicare for hospice and other healthcare services.

Results:

Medicare costs were reduced from hospice election until death using both 2009–2010 and new 2016 payment structures and rates. Mean cost savings were $1,527 with actual payment rates, and would have been $2,105 with the new payment rates (p < 0.001). Cost savings were confirmed by reducing the number of days used for cost comparison by three days for those with hospice stays of at least four days ($4,318 using 2009–2010 rates, $3,138 for 2016 rates: p < 0.001). Cost savings were greater for males ($3,393) versus females ($1,051) and greatest in cancer ($6,706) followed by debility and failure to thrive ($5,636) and congestive heart failure ($1,309); dementia patients had higher costs (+$1,880) (p < 0.001). When adding 3 days to the comparison period, hospice increased costs to Medicare.

Conclusions:

Medicare savings could continue with the 2016 payment rate change. Cost savings were found for all primary diagnoses analyzed except dementia.

Introduction

The hospice benefit was added to Medicare in 1983, and has since grown into a multibillion-dollar industry as a growing proportion of Medicare beneficiaries elect hospice services. Of the 2.6 million U.S. deaths in 2014, 8 of 10 were Medicare recipients; this makes Medicare the largest insurer of healthcare provided during the last year of life.¹

Medicare pays for hospice services using a per diem approach, with the daily payment amount differing by degree and setting of care. The hospice benefit is designed to cover all supportive care related to a patient's terminal illness, and, in exchange, patients are required to forego curative treatments. We have previously demonstrated that the hospice benefit saves the Medicare program money, by ∼$2,300 per beneficiary after initiation of hospice, and more recent analyses confirm these findings.^2,3 However, recent methodological criticisms related to the identification of nonexperimental comparison groups used to assess the cost impact of hospice warrant attention.⁴

Beginning January 1, 2016, Medicare adopted a two-tiered per diem payment that represents the first change in Medicare hospice payment in over three decades.⁵ Under the new model, per diem payments are increased for days 1 through 60 of a hospice stay, payments are decreased for days 61 and greater, and service intensity add-on payments may be made retrospectively for the last seven days of life.

This change is designed to better align payment rates with the variance in service costs that hospice providers face and ensure that appropriate levels of care are delivered to patients to meet their more intensive needs at hospice admission and in the last days of life. The change also discourages unnecessarily long hospice stays that may represent wasteful, inappropriate, or even fraudulent care and billing practices. Table 1 details the payment changes and compares them to historical hospice per diem rates.⁵

Table 1.

Past and Current Medicare Hospice Payment Rates

Level of care	Historical hospice per diem model with FY 2015 rates	New U-shaped hospice per diem model with FY 2016 rates
RHC	$159.34	Days 1–60: $186.84
RHC	$159.34	Days 61+: $146.83
RHC last week of life	NA	Service intensity add-on: $38.75/h, up to four hours per day, for in-person nursing and social work visits
Inpatient respite care	$164.81	$164.81
General inpatient care	$708.77	$708.77
Continuous home care	$929.91	$929.91

NA, not applicable; RHC, routine home care.

A critical question is whether the hospice benefit's potential for cost savings to the Medicare program will be impacted by the new payment structure. Because the new payment rates went into effect at the beginning of 2016, not enough time has elapsed to allow for a robust analysis of Medicare claims to investigate this question. Instead, this article estimates the cost-saving potential of the new payment structure and rates by applying them to the claims records of Medicare decedents who died between 2009 and 2010.

Methods

To estimate the impact of the new hospice payment structure on costs to Medicare, we identified two groups of Medicare decedents who received care and died in North Carolina in calendar years 2009 and 2010: those who died while receiving hospice services and those who died while not receiving hospice services. We applied the new hospice payment model and rates to the claims history of the decedents in the sample and measured the difference in total costs to Medicare between the two groups. We narrowed our sample to decedents from North Carolina alone to hold constant the state-level policies and regulations, such as Certificate of Need laws, which impact the supply of hospice care accessible to Medicare beneficiaries.

Utilizing 2009–2010 claims records of all age-eligible Medicare decedents, we began with 1,784,829 North Carolina beneficiaries, of whom 253,625 had validated deaths. To better capture the trajectory of healthcare utilization of beneficiaries before the end-of-life period, we conditioned our sample on those persons for whom we observed at least 730 days of Medicare claims before death (the last two years of life). We also excluded those who received hospice, hospital, or skilled nursing facility outside of North Carolina; only had durable medical equipment (DME) claims; were not age-eligible Medicare beneficiaries; had Medicare Advantage coverage; had missing gender or race information; had Medicare payments totaling $0; and had no services within 30 days of death.

We also excluded N = 5234 decedents who had at least one claim signifying their receipt of hospital-based palliative care, as measured by a V code (International Classification of Diseases, Ninth Revision [ICD-9]: V66.7). These codes likely indicate supportive or comfort care similar to services offered by hospice, but not limited to patients that have elected hospice. Because of the uncertainty regarding the services rendered as indicated by V66.7 and the unreliability of its coding in the claims data (because use of the V code modifier does not yield an increased reimbursement), decedents with the palliative care V code were excluded to provide for the cleanest effect of hospice on Medicare costs.

The final analysis sample was N = 36,035 beneficiaries (sample methods detailed in Appendix 1), and was divided into two mutually exclusive groups of decedents who died while receiving hospice care (N = 18,647), and died while not receiving hospice care (N = 17,338).

The choice of hospice is nonrandom, so a simple comparison of costs between those who died while receiving hospice (hospice decedents) and those who died while not receiving hospice (controls) could produce a biased estimate of the impact of hospice costs. This methodological issue has been addressed in a variety of ways.^6–10 We employed a propensity score-matching approach that is similar to that used in past work by ourselves and others.^2,3 Specifically, we matched hospice decedents to controls who did not die while receiving hospice services, but who were otherwise similar by using a propensity score-matching model. The purpose of this matching is to produce a control group that is as similar as possible to hospice decedents in terms of key characteristics that impact care utilization and cost, such as diagnosis and age.

To do so, we first estimated a model predicting death in hospice, controlling for age, race, gender, and primary diagnosis using two years of Medicare claims history before death for each subject. Diagnosis was represented through disease-specific binary variables that took the value of 1 if an ICD-9-CM code for the following diseases was found in the primary diagnosis field of any claim [cancers of the following types: breast, colon, lung, prostate, other; chronic obstructive pulmonary disease; congestive heart failure (CHF); dementia of any type; and debility and failure to thrive, with the omitted category being “other”]. Using logistic regression, we predicted the underlying dependent variable (the probability of dying in hospice) that ranged from 0 to 1.0. We then took each member who died while receiving hospice services and matched that individual to an individual who died while not receiving hospice services, but who had the closest predicted probability of doing so. Our sample was large enough to match 16,238 decedents exactly, meaning decedent pairs had an identical probability of using hospice to the second decimal place (8119 pairs). Table 2 shows the characteristics of the observations included in the matched sample that was used to assess the cost-saving potential of hospice.

Table 2.

Sample Observation Characteristics

	Hospice decedents	Control
Total	8119	8119
Gender
Male	3654	3654
Female	4465	4465
Age groups
65–69	778	778
70–74	1067	1067
75–79	1301	1301
80–84	1683	1683
>84	3290	3290
Primary diagnosis
All cancer	1784	1784
CHF	1839	1839
COPD	1653	1653
Dementia	670	670
Debility/failure to thrive	624	624
Race
White	7086	7086
Black	1007	1007
N Am Native	12	12
Asian, other	14	14

CHF, congestive heart failure; COPD, chronic obstructive pulmonary disease.

Once the matched pairs were established, we measured the costs that Medicare paid for beneficiaries in the hospice decedents group compared with those paid for beneficiaries in the control group. Costs were measured as the amounts that Medicare paid for hospice services and facility utilization (inpatient hospital and skilled nursing facilities), DME, and intensive care unit days. We tested differences in costs from the point of initiation of hospice for hospice decedents and compared them to costs for controls. The cost comparison always included the same number of days before death for hospice decedents and controls. For example, if a hospice decedent received hospice services for the last 15 days of life (i.e., a 15-day length of stay in hospice), their costs were compared with the costs accrued in the last 15 days of life for their matched control.

A recent criticism suggests that using the exact number of days before death for both hospice and control groups may be misplaced precision.⁴ Especially if hospice election is precipitated by exhaustion of aggressive (and costly) courses of treatment or triggered by an acute episode such as a hospital stay, then those high costs will not be measured in the hospice decedent group, whereas costs for comparable treatments could be captured in the matched control group.

To identify the sensitivity of this assumption, we conducted two additional analyses. In the first, we decreased the hospice length of stay by three days by subtracting three days from the admission dates for both the hospice decedents and control groups. Observations with lengths of stay of three days or less were dropped from this analysis, as reflected in the lower N values reported for this analysis in Table 4. We also conducted an analysis in which we simply altered our cost measurement method to include an additional three days of costs in the claims record directly preceding the hospice admission date, for both hospice decedent and control group. This analysis was designed to determine the costs of care immediately before the election of hospice. The results of these analyses are reported in Tables 3 and 4.

Table 3.

Comparing Cost Savings Using Old and New Hospice Payment Models and Per Diem Rates (2010 Dollars)

Group	Service category	Mean 2010 adjusted dollars, died outside hospice	Mean 2010 adjusted dollars, died In hospice
Old per diem	Total	$19,992	$18,465
Old per diem—3 days fewer^*	Total	$19,642	$15,324
Old per diem—3 days more^*	Total	$25,392	$30,507
New hospice payment—exact days	Total	$20,548	$18,443
New hospice payment—3 days fewer^**	Total	$16,037	$12,898
New hospice payment—3 days more	Total	$24,376	$30,911

Note: For the Old Per Diem method, the three days fewer and three days more estimates were based on per diem estimates, which had a mean = $1,095 and a range from $0 to $363,705. When reducing LOS three days, several records were deleted due to LOS less than three days; however when adding three days to LOS, no records were deleted—somewhat accounting for the variation in numbers.

Note: For the New Hospice Payment method, the three days fewer estimate for those who died in hospice was based on the “Old Per Diem – 3 Days Fewer” estimate minus the dollar adjusted rate for three days of Routine Home Care (one day @ the 0–60 rate; two days @ the 61+ day rate—proportional to the sample LOS). The three days fewer estimate for those who died in hospice was based on the “Old Per Diem – 3 Days Fewer” estimate. LOS, length of stay.

Table 4.

Differences in Costs Measured between Case and Control Using New Medicare Hospice Payment Model

	Exact hospice LOS comparison				−3 days hospice LOS				+3 days pre-hospice election
	N	Mean ^a	Difference ^a	p	N	Mean ^a	Difference ^a	p	N	Mean ^a	Difference ^a	p
Total	16,238	$19,496			13,990	$14,468			16,238	$27,644
Control	8119	$20,548			6995	$16,037			8119	$24,376
Hospice	8119	$18,443	−$2,105	< 0.001	6995	$12,898	−$3,138	< 0.001	8119	$30,911	+$6,535	< 0.001
Control
Male	3654	$21,312			3120	$16,556			3654	$25,804
Female	4465	$19,923			3875	$15,612			4465	$23,207
Hospice
Male	3654	$17,923	−$3,393		3120	$11,997	−$4,559		3654	$31,109	+$5,305
Female	4465	$18,872	−$1,051	< 0.001	3875	$13,637	−$1,976	< 0.001	4465	$30,750	+$7,542	< 0.001
Age groups
Control
65–69	778	$25,181			687	$19,565			778	$31,040
70–74	1067	$25,377			916	$20,001			1067	$31,428
75–79	1301	$23,900			1120	$18,674			1301	$28,945
80–84	1683	$19,912			1438	$15,488			1683	$23,525
>84	3290	$16,887			2834	$13,155			3290	$19,142
Hospice
65–69	778	$18,199	−$6,983		687	$10,653	−$8,912		778	$35,778	+$4,738
70–74	1067	$17,773	−$7,604		916	$11,112	−$8,889		1067	$32,976	+$1,548
75–79	1301	$18,566	−$5,334		1120	$12,597	−$6,077		1301	$32,306	+$3,361
80–84	1683	$19,189	−$724		1438	$13,656	−$1,832		1683	$31,904	+$8,379
>84	3290	$18,289	+$1,402	< 0.001	2834	$13,741	+$585	< 0.001	3290	$28,031	+$8,890	< 0.001
Primary diagnosis
All cancer
Control	1784	$21,404			1627	$17,301			1784	$24,479
Hospice	1784	$14,698	−$6,706	< 0.001	1627	$9,954	−$7,348	< 0.001	1784	$24,472	−$7	< 0.001
CHF
Control	1839	$20,501			1585	$16,643			1839	$23,710
Hospice	1839	$19,192	−$1,309	< 0.001	1585	$14,094	−$2,549	< 0.001	1839	$30,557	+$6,846	< 0.001
COPD
Control	1653	$24,202			1454	$20,179			1653	$28,064
Hospice	1653	$23,020	−$1,182	< 0.001	1454	$17,971	−$2,208	< 0.001	1653	$35,368	+$7,305	< 0.001
Dementia
Control	670	$18,149			592	$14,084			670	$19,023
Hospice	670	$20,029	+$1,880	< 0.001	592	$17,3443	+$3,259	< 0.001	670	$25,792	+$6,770	< 0.001
Debility/failure to thrive
Control	624	$23,940			549	$18,360			624	$24,383
Hospice	624	$18,304	−$5,636	< 0.001	549	$14,923	−$3,437	< 0.001	624	$25,374	+$992	< 0.001

Adjusted for 2010 dollars.

Cost estimates based on Centers for Medicare & Medicaid Services reimbursement rates reported in Table 1 and adjusted for 2010 wage index adjustment (North Carolina mean for all counties = 0.9335).

Results

We confirmed that hospice reduced costs to Medicare from the point of hospice election until death using both the actual payment rates in place in 2009 and 2010, as well as if the new payment model implemented in 2016 had been in place then. Mean cost savings were $1,527 with the actual payment rates, and would have been $2,105 with the new payment model—in both cases estimated by comparing costs for hospice decedents and matched controls for the number of days in hospice before death (p < 0.001; Table 3).

We assessed the sensitivity of cost savings due to hospice by reducing the number of days used for cost comparison by three days for those with hospice stays of at least four days. Cost savings were confirmed ($4,318 using the actual payment model, $3,139 when applying the 2016 model; both comparisons, p < 0.001). When adding three days before the hospice election period to the cost comparison, we found that costs were much higher for persons who died while using hospice as compared with matched controls ($5,115 more when using the actual payment model, $6,535 more when applying the new one; p < 0.001 for both comparisons). This demonstrates that costs were much higher in the three days before hospice election as compared with controls.

We looked more closely at cost differences between hospice decedents and matched controls by key covariates (Table 4). Cost savings were greater for males ($3,393) than for females ($1,051) as well as for younger Medicare beneficiaries (max savings $7,604 for decedents 70–74 years of age; for the oldest old decedents [age 85+ years] hospice actually increased costs by $1,402). Disease-specific analyses show that hospice-related savings were the greatest among patients with a primary diagnosis of cancer ($6,706 in savings), followed by debility and failure to thrive ($5,636), and CHF ($1,309). Conversely, patients with dementia who died while receiving hospice services had higher costs than did those who did not (+$1,880). All results were statistically significant at the p < 0.001 level.

Our sensitivity analyses further confirmed cost savings to Medicare for the hospice decedent group relative to control. In the analysis that shortened length of stay (LOS) by three days, the mean cost savings was $3,138. However, in the final sensitivity analysis, in which costs in the three days preceding hospice admission were included for both groups, we found that mean costs for the hospice group were greater than the control group by $6,535. This analysis identified how expensive the prehospice costs were just before election (three days before hospice election observed Medicare-financed costs of around $12,500 for hospice users, compared with ∼$3,800 for controls).

Discussion

We confirmed the cost-saving potential of the Medicare hospice benefit found in past work using these data, and determined that this finding is somewhat robust to the precision of the matching period used to compare cases and controls.^2,3 Cost savings were also found when using the hospice payment rates at force when the care was used, as well as when applying the new (2016) hospice payment model and rates to past utilization. Past work has worried that the precision of the matching period used in this approach (the number of days of hospice use defining the number of days of cost comparison) was overly precise, potentially leading to spurious findings of cost savings. We find that using a shorter matching period (three days less) results in larger savings.

When we added three days to our cost comparison to further test the sensitivity of the matching period used for assessing the cost impact of hospice, we found that the three days before observed initiation of hospice care are much more expensive than are the same three days for matched controls. For controls, adding three days to the cost comparison increased costs by around $3,800, or a little over $1,000/day. For hospice decedents, however, the three days before the hospice admission that ended with the patient's death increased the cost comparison amount by ∼$12,500, or over $4,000/day (p < 0.001). This finding does not invalidate our findings of cost savings to Medicare for hospice relative to matched controls, which are used to estimate what they otherwise might have been, but it highlights two key points.

First, it demonstrates the reality that hospice is often elected by a patient after a triggering event, such as an expensive hospitalization, during which a hospice-qualifying prognosis is made by a physician and delivered to the patient and family.

Second, it demonstrates that hospice cannot reduce costs to Medicare before a patient actually elects it. This is obvious, but much of the research that examines costs over the “last year of life” has found that hospice users have similar or even higher costs than nonhospice decedents, which indicates that hospice patients likely still get lots of high-cost care before hospice election. It may be literally true that, for the subjects in our study, hospice stays that began three days earlier would have reduced costs to Medicare relative to controls. However, for policymakers seeking to encourage earlier election of hospice to realize the associated cost savings, there is no obvious way to act on this insight, in large part because it is the actual experience of a high-cost hospitalization or acute event relatively near death that is the final piece of information that causes a patient and family to choose hospice.

This discussion demonstrates the difference between a retrospective perspective on end-of-life costs, in which you note after the fact that they are large (we have done this for four decades and it has failed to lead to effective changes), and a prospective orientation to the end of life. It will take prospective changes that result in patients electing hospice without first having very expensive care that is often viewed as nonproductive in retrospect, to truly impact end-of-life costs. This is a great policy challenge, and is a key impediment to truly reducing end-of-life costs.

Another area of concern for policymakers are long hospice stays that may indicate inappropriate care with hospice services functioning as a kind of long-term care benefit—a benefit that is not covered by Medicare—rather than end-of-life care services for patients with a prognosis of six months or less.⁵ The new payment structure, with its lower payment rates for days beyond 60, is designed to discourage inappropriately long stays. However, this fact highlights the biggest limitation of our study, which is that we have simply applied the new payment model and rates to historical claims data.

Before the revision in 2016, the old per diem payment structure had been in place since the inception of the benefit more than three decades ago. Had the new model and rates been in place in 2009 and 2010, hospice providers and patients may have behaved differently, resulting in different hospice utilization patterns, especially with regard to lengths of stay, but we have no evidence-based way of estimating such a potential impact. Future research that examines lengths of stay before and after the 2016 payment change may be able to estimate any such change.

Furthermore, two changes to Medicare hospice policy were implemented after our 2009–2010 study period that may also impact provider behavior and hospice utilization, and so limit the conclusions of our findings.

First, a provision of the Affordable Care Act that went into effect on January 1, 2011 requires that a physician or nurse practitioner must complete a face-to-face visit with hospice patients to determine continued eligibility before recertifying them for any hospice benefit periods beyond 180 days.¹¹ This change resulted from a MedPAC recommendation designed to improve accountability of the hospice benefit by ensuring that patients do indeed meet the hospice eligibility criteria and that those who do not are identified and discharged from enrollment.¹¹ This change, implemented in between the years studied and the payment change in 2016 may have had an impact on frequency of stays >180 days.

Second, starting October 1, 2014, “debility” and “failure to thrive” were eliminated as principal hospice diagnoses. For FY 2013, just before this change, debility was the most common principal hospice diagnosis, accounting for 9% of all hospice diagnoses, and failure to thrive was the sixth most common, accounting for ∼5% of all hospice diagnoses.⁵ Following the elimination of these diagnoses, the mix of diagnoses was necessarily changed and, most notably, Alzheimer's disease became the single most common diagnosis (9%) in 2014.⁵

Although we have no way to estimate whether patients that were or would have previously been diagnosed with debility or failure to thrive might have been recategorized with Alzheimer's or a related dementia diagnosis (and there was already an observed trend of increasing dementia diagnoses preceding the change), this change is especially notable given the results of our study, which found that dementia was the only diagnosis for which hospice was not cost saving relative to control (difference of +$1,880), whereas average cost savings for debility (difference of −$5,636) were second only to cancer.

The implications for policymakers are considerable since some have proposed a hospice payment model that is diagnosis specific. Our results for patients with dementia diagnoses are likely demonstrating that the lower likelihood of high-cost treatments or acute events for dementia patients means that they present less cost-saving potential than those with primary diagnoses such as cancer or heart failure. This finding illustrates the key point that delivery of hospice incurs cost to Medicare, but that hospice's cost-saving potential lies in its ability to avert high-cost care, such as emergency visits and hospital stays near the end of life that may outweigh the costs of hospice.

However, if patients that would have been diagnosed with debility and were found to be cost saving on average are now receiving dementia-related principal hospice diagnoses following the 2014 policy change, then results from a similar study design could differ significantly compared with our findings based on the 2009–2010 data. This suggests that any proposal for diagnosis-specific payments must wait to be informed by robust analysis of the change of diagnosis patterns following the 2014 policy change and any subsequent changes in costs to providers and cost savings to Medicare by diagnosis.

The “two-tiered” payment model is designed to better align Medicare's payment for hospice services with the variation in service levels that patients require over the course of a hospice stay. Our results suggest that if hospice utilization patterns remain similar to those observed in 2009 and 2010, then paying for hospice care will still be cost saving to Medicare. However, the new payment model could have an impact on hospice utilization, and our results illustrate the ways that changes in utilization, especially in hospice lengths of stay, can have a significant impact on costs and the potential for cost savings to the Medicare program.

Footnotes

Acknowledgment

The authors acknowledge the work of Matthew Harker on this project.

Author Disclosure Statement

Donald Taylor has provided paid expert testimony in CON hospice cases in one state in the past two years.

Appendix 1.

Sample Construction Methodology

Method	Result
All 2007–2010 beneficiaries in files	1,784,829 Benes
All 2007–2010 beneficiaries in files	253,625 deaths
Excluded those who received hospice, hospital, or SNF outside of North Carolina	242,800 deaths
Limit to North Carolina Benes only, who died, with at least 730 days of claims before death (2007–2010)	118,875 deaths
Excluded those who were not 65+, missing sex or race information	107,037 deaths
Excluded those with no Medicare payments	97,498 deaths
Excluded those who were DME only	56,842 deaths
Excluded if Medicare Advantage	46,519 deaths
Excluded those with no services within 30 days of death	41,269 deaths
Excluded those with a V66.7 code in hospital	36,035 deaths
5234 deaths	36,035 deaths
2978 died in hospice	36,035 deaths
2256 died outside hospice	18,647 died in hospice
Excluded those if death was not validated	17,388 died outside hospice
Result	16,238 deaths
Final Result (after applying matching criteria)	8119 matched pairs

DME, durable medical equipment; SNF, skilled nursing facility.

References

Cubanski

, Neuman

, Griffin

, Damico

: Medicare spending at the end of life: A snapshot of beneficiaries who died in 2014 and the cost of their care. The Henry J. Kaiser Family Foundation website. http://kff.org/medicare/issue-brief/medicare-spending-at-the-end-of-life. 2016. (Last accessed August 7, 2017).

Taylor

Jr. , Ostermann

, Van Houtven

, et al.: What length of hospice use maximizes reduction in medical expenditures near death in the US Medicare program?. Soc Sci Med, 2007; 65:1466–1478.

Kelley

, Deb

, Du

, et al.: Hospice enrollment saves Money for Medicare and improves care quality across a number of different lengths-of-stay. Health Aff (Millwood), 2013; 32:552–561.

Hogan

: Spending in the Last Year of Life and the Impact of Hospice on Medicare Outlays. Washington, DC: MedPAC, 2015.

Centers for Medicare & Medicaid Services (CMS), HHS: Medicare program; FY 2016 hospice wage index and payment rate update and hospice quality reporting requirements; Final rule. Fed Regist, 2015; 80:47141–47207.

Bretscher

, Rummans

, Sloan

, et al.: Quality of life in hospice patients. A pilot study. Psychosomatics, 1999; 40:309–313.

Campbell

, Lynn

, Louis

, Shugarman

: Medicare program expenditures associated with hospice use. Ann Intern Med, 2004; 140:269–277.

Emanuel

: Cost savings at the end of life. What do the data show?. JAMA, 1996; 275:1907–1914.

Mularski

, Dy

, Shugarman

, et al.: A systematic review of measures of end-of-life care and its outcomes. Health Serv Res, 2007; 42:1848–1870.

10.

Pyenson

, Connor

, Fitch

, Kinzbrunner

: Medicare cost in matched hospice and non-hospice cohorts. J Pain Symptom Manage, 2004; 28:200–210.

11.

National Hospice and Palliative Care Organization: Medicare Program: Home Health Prospective Payment System Rate Update for Calendar Year. 2011; Changes in Certification Requirements for Home Health Agencies and Hospice. www.nhpco.org/sites/default/files/public/regulatory/FacetoFace.pdf. 2010. (Last accessed November 8, 2017).