Shown are the mean unadjusted monthly Medicare costs before death (solid line) with 95% CIs (shaded) stratified by hospice status for oral cavity and pharyngeal cancers. Costs are in 2009 US dollars.
Shown are the mean monthly Medicare costs before death (solid line) with 95% CIs (shaded) stratified by hospice status in each stratum for patients with oral cavity cancer. Costs are in 2009 US dollars.
Shown are the mean monthly Medicare costs before death (solid line) with 95% CIs (shaded) stratified by hospice status in each stratum for patients with pharyngeal cancer. Costs are in 2009 US dollars.
Enomoto LM, Schaefer EW, Goldenberg D, Mackley H, Koch WM, Hollenbeak CS. The Cost of Hospice Services in Terminally Ill Patients With Head and Neck Cancer. JAMA Otolaryngol Head Neck Surg. 2015;141(12):1066-1074. doi:10.1001/jamaoto.2015.2162
Hospice care has been suggested as a way to preserve dignity and to lower costs at the end of life, which may be particularly important for patients with head and neck cancer because this disease is associated with considerable morbidity and a high mortality risk.
To use data from the Surveillance, Epidemiology, and End Results (SEER)–Medicare linked database to estimate monthly costs of all services used during the last months of life by patients with oral cavity and pharyngeal cancers and to determine whether those who used hospice services had lower costs.
Design, Setting, and Participants
Retrospective cohort analysis of SEER-Medicare data (January 1, 1995, to December 31, 2007). The setting was all participating SEER hospitals that treated patients with oral cavity or pharyngeal cancer. Participants were 65 years and older who were diagnosed as having oral cavity (n = 4205) or pharyngeal (n = 3178) cancer between January 1, 1995, and December 31, 2005, who subsequently died between January 1, 1995, and December 31, 2007.
Use of hospice services before death.
Main Outcomes and Measures
Hospice use was identified through Medicare claims. The primary outcome was all-cause Medicare expenditures, inflated to 2009 US dollars. We used a propensity score analysis to estimate the difference in the mean costs to Medicare in the last month of life between patients who used hospice services and patients who did not use hospice services.
Most patients (63.4% [1018 of 1605] with oral cavity cancer and 57.8% [644 of 1114] with pharyngeal cancer) who enrolled in hospice did so within 30 days of death. Patients who received hospice care had $7035 (95% CI, $6040-$8160) lower costs in the last month of life for oral cavity cancer and $7430 (95% CI, $6340-$9100) lower costs in the last month of life for pharyngeal cancer. These cost savings were greater in the last month of life when patients enrolled in hospice more than 30 days before death.
Conclusions and Relevance
Encouraging hospice admissions for patients with oral cavity and pharyngeal cancers provides not only compassionate, dignified care at the end of life but also an opportunity for substantial savings in health care costs.
In 1980, economist Eli Ginzberg commented on “the high cost of dying,”1 spurring decades of studies examining expenditures incurred at the end of life. Researchers have decried that costs at the end of life seem disproportionately high,2 with studies3- 5 demonstrating that 27% to 30% of Medicare payments each year are paid to the 5% to 6% of Medicare beneficiaries who die in that year. With an aging population and national health care costs in 2011 reaching $2.7 trillion,6 it is no surprise that there has been significant pressure from health care stakeholders to focus attention on elucidating and containing these expenditures.
Hospice care has been repeatedly suggested as a method for decreasing medical costs at the end of life. Multiple studies7- 10 have also demonstrated that 40% to 80% of terminally ill patients would prefer to end their lives at home with family rather than in a hospital bed. If tests and interventions are withheld at the end of life and patients are placed in hospice care, some researchers argue thatmillions of health care dollars could be saved,11- 14 with the added benefit of preserving patient autonomy and dignity at the end of life. However, other authors maintain that cost savings because of increased hospice use at the end of life are not likely to be substantial.2,15- 17 Therefore, the question of whether hospice care can significantly decrease end-of-life costs remains.
One of the many challenges of examining costs of hospice care is that, despite advances in modern medicine, predicting death with any certainty in most patients is almost impossible.2,16,18 In studies of medical expenditures and hospice care at the end of life, researchers may easily designate a patient as terminally ill retrospectively, but determining which patients are dying prospectively is a much more difficult matter.18 One of the rare exceptions is patients with cancer, for whom accurate predictions about survival are possible.18 Patients with head and neck cancer are of particular interest because their disease continues to remain a source of considerable morbidity and mortality, and a substantial proportion of patients will succumb to their disease.19- 22 The objective of this study was to use data from the Surveillance, Epidemiology, and End Results (SEER)–Medicare linked database between January 1, 1995, and December 31, 2007, to compare monthly costs of all services used during the last 12 months of life by patients with head and neck cancer who received hospice care vs those who did not use hospice services.
We obtained data for this study from the SEER-Medicare linked database, which combines tumor registry data from the National Cancer Institute’s SEER program with Medicare billing records for patients who are covered by Medicare. We limited our analysis to patients with a primary diagnosis of oral cavity or pharyngeal cancer. Oral cavity cancer included the following SEER cancer sites: lip, salivary gland, floor of mouth, oral tongue (excluding base of tongue and lingual tonsil), and gum and other mouth. Pharyngeal cancer included the following SEER cancer sites: nasopharynx, tonsil, oropharynx, hypopharynx, base of tongue (including base of tongue and lingual tonsil only), and other. We included patients 65 years and older who were diagnosed between January 1, 1995, and December 31, 2005; died between January 1, 1995, and December 31, 2007; had 1 primary tumor in their lifetime; and had a valid date of death in the Medicare records. The dates of diagnosis are January 1, 1995, to December 31, 2005. The dates of death are January 1, 1995, to December 31, 2007. We excluded patients who were enrolled in a health management organization for at least 1 month, patients who were diagnosed based on autopsy or death certificate alone, and patients who had a missing month of diagnosis in SEER data.
Our goal was to compare monthly Medicare costs of all services used during the last 12 months of life by patients with head and neck cancer who received hospice care vs those who never used hospice services. We determined hospice status based on hospice billing records in Medicare data. Patients who had a hospice claim at any time between the date of diagnosis and the date of death comprised the hospice group, and all other patients comprised the nonhospice comparison group. Medicare covers inpatient and in-home hospice, so both are included in the hospice group.
Monthly Medicare costs were calculated from Medicare claims for 12 months (360 days) before death. Costs represented all payments made by Medicare for all-cause treatments of inpatient stays, emergency department visits, outpatient procedures, physician bills, home health visits, durable medical equipment, and hospice care, but they excluded outpatient prescription drugs. All costs were adjusted for inflation to 2009 US dollars. Costs for inpatient stays and emergency department visits were inflated using the hospital and related services component of the consumer price index. Costs for outpatient visits, physician bills, home health visits, and hospice care were inflated using the medical care services component of the consumer price index. Costs for durable medical equipment were inflated using the medical care commodities component of the consumer price index.
We first summarized monthly costs for patients who enrolled in hospice and for patients who never enrolled in hospice by plotting monthly costs along with 95% CIs for 12 months before death. We used bootstrapping with 1000 samples and the percentile method to generate 95% CIs for the mean monthly costs because monthly cost distributions were highly skewed.
We obtained estimates for cost differences in the last month of life for each cancer site using a propensity analysis with stratification that accounted for the differences in patient characteristics between hospice groups. First, we created propensity models for hospice status separately for each cancer site using logistic regression. Hospice status was regressed on the following covariates: specific cancer site, age at diagnosis, year of diagnosis, race (white and nonwhite), marital status (married, single, divorced or separated, widowed, and unknown), urban or rural geographic code (big metropolitan city, metropolitan city, urban, less urban, and rural), and SEER historical stage (localized, regional, and distant). Age at diagnosis and year of diagnosis were modeled as linear effects on the log odds scale because nonlinear effects modeled using B-splines did not provide improved fits. Once the final models were fitted, we calculated the propensity score for each patient or the estimated probability of hospice enrollment.
We separated patients into 5 equally sized strata defined by the quintiles of the propensity score. In each propensity stratum, we calculated the mean difference in costs in the last month of life between patients who used and did not use hospice services and then averaged these cost differences across strata to obtain the final estimates. We used bootstrapping and the percentile method to obtain 95% CIs. In each of the 1000 bootstrap samples, we fitted the propensity model, created strata based on the estimated propensity scores, and obtained the final estimates after stratification. Therefore, bootstrapping accounted for the uncertainty in fitting the propensity model and in estimating the final cost differences after stratification.
We also conducted a subgroup analysis whereby patients receiving hospice services were grouped according to the time that hospice coverage began (within 30 days of death, between 31 and 90 days, and >90 days before death). For each subgroup, we calculated the difference in the mean cost in the last month of life compared with the nonhospice group using the same methods after stratifying on the propensity score.
The sample comprised 7383 patients 65 years and older at diagnosis who died between January 1, 1995, and December 31, 2007, including 4205 patients (57.0%) with oral cavity cancer and 3178 patients (43.0%) with pharyngeal cancer. At the time of death, 1605 patients (38.2%) with oral cavity cancer and 1114 patients (35.1%) with pharyngeal cancer received hospice services. Table 1 summarizes patient characteristics stratified by hospice status for each cancer site. Patients who used hospice services were more likely to be female and of white race. For pharyngeal cancer, patients who used hospice services were slightly older at the time of death. For oral cavity cancer, patients who used hospice services were more likely to have regional, distant, or unstaged disease. Most patients had a first claim for hospice services within 30 days of death (63.4% [1018 of 1605] for oral cavity cancer and 57.8% [644 of 1114] for pharyngeal cancer).
Table 2 summarizes the number of patients who were alive with cancer for each month (30 days) before death. For oral cavity cancer, 951 of 1605 patients (59.3%) who received hospice services at the time of death had lived with cancer for at least 12 months (360 days) compared with 1460 of 2600 patients (56.2%) who did not use hospice services. The percentages were lower for pharyngeal cancer, including 614 of 1114 patients (55.1%) who received hospice services and 942 of 2064 patients (45.6%) who did not receive hospice services.
The mean unadjusted monthly Medicare costs before death are shown for each cancer site stratified by hospice status in Figure 1. The same pattern was observed for each cancer site, and patients who did and did not use hospice services had similar costs until approximately 60 days before death, at which time costs for patients who did not use hospice services increased sharply while costs for patients who received hospice services increased gradually. The similarity in monthly costs for earlier time points was expected because most patients were not yet enrolled in hospice. In the last month (30 days) of life for patients with oral cavity cancer, the mean cost per patient who used hospice services was $7880 compared with $14 990 per patient who did not use hospice services, a difference of $7110. Similarly, for pharyngeal cancer, the difference in the mean cost between patients who used hospice services ($8790) and patients who did not ($16 390) was $7600.
We conducted a propensity analysis with stratification separately for each cancer site to obtain adjusted estimates of costs after accounting for baseline differences between hospice and nonhospice enrollees. The final propensity models for each cancer site indicated that female sex, white race (relative to other races), married status (relative to single), and increasing year of diagnosis were significantly associated with higher odds of hospice use. For oral cavity cancer, regional disease (relative to localized disease) was also associated with higher odds of hospice use. Propensity scores ranged from 0.11 to 0.64 for oral cavity cancer and from 0.17 to 0.60 for pharyngeal cancer and were stratified into 5 strata for each cancer site. Box plots showed that propensity score distributions were similar between hospice groups within each stratum for each cancer site.
The mean monthly Medicare costs are shown for each stratum in Figure 2 (oral cavity cancer) and Figure 3 (pharyngeal cancer). In all strata for both cancers, the patterns were similar, with patients receiving hospice services having lower costs for 2 to 3 months before death. After stratifying on the propensity score, patients who used hospice services had $7035 (95% CI, $6040-$8160) lower mean Medicare costs for oral cavity cancer and $7730 (95% CI, $6340-$9100) lower mean Medicare costs for pharyngeal cancer than patients who did not use hospice services in the last month of life. These propensity-adjusted estimates were similar to the unadjusted estimates reported above.
In subgroup analysis after stratifying on the propensity score, patients with oral cavity cancer and pharyngeal cancer who enrolled in hospice within 30 days of death had $4660 (95% CI, $3580-$5820) and $4590 (95% CI, $3010-$6000) lower costs in the last month before death, respectively, than patients who did not enroll in hospice care. These estimates were 33% and 41% lower than the respective overall estimates. However, patients who enrolled in hospice between 31 and 90 days before death had larger cost reductions in the last month before death, namely, $11 200 (95% CI, $10 090-$12 330) for oral cavity cancer and $12 410 (95% CI, $11 270-$13 650) for pharyngeal cancer. These reductions were similar to the reduction in costs for patients who enrolled in hospice more than 90 days before death, namely, $10 690 (95% CI, $9490-$11 810) for oral cavity cancer and $11 410 (95% CI, $9630-$13 050) for pharyngeal cancer. Therefore, earlier enrollment (>30 days before death) in hospice was associated with lower costs in the last month before death.
The high cost of medical care at the end of life is not a new finding, nor is it a phenomenon that has developed in recent years. Even before the advent of Medicare, data from 1961 suggested that hospital costs of patients who died were 3 times higher than those of other patients.18 In 1993, Lubitz and Riley3 reported that more than one-fourth of the US Medicare budget was spent on health care in the last year of life. Consequently, attention has been focused on measures to control rising costs at the end of life. Hospice care has been proposed not only as a method to ensure patient autonomy and provide quality care at the end of life but also as a way to reduce medical expenses by preventing unnecessary, costly interventions.
Multiple studies have shown the cost benefit of hospice care. When comparing hospice and conventional hospital care in the last month of life, major investigations found between 31% and 64% savings in medical care costs with home hospice services.23- 28 According to findings by the National Hospice Study,26 most of this savings was largely attributable to substituting less costly hospice care for more costly conventional inpatient care. Accordingly, a meta-analysis by Hughes et al29 demonstrated that home care services, including hospice, reduced the number of hospital days by 2.5 days to 6 days. Brumley et al30 showed that patients treated in hospice had reduced trips to the emergency department and thus fewer hospitalizations, resulting in 45% lower costs compared with patients not treated in hospice.
Nevertheless, the studies on cost savings from hospice are not definitive. Some researchers speculate that the variability of savings is due to selection bias.2,15,26,27 Patients who choose hospice care are inherently different from patients who do not. Hospice patients have chosen to avoid aggressive care at the end of life and refuse interventions near death; therefore, the savings attributed to hospice care may actually be because of a characteristic of the patient rather than hospice itself.15 In a randomized trial, which should eliminate this selection bias, Kane et al31 found that hospice care was not associated with reduced use of hospital inpatient days or therapeutic procedures and was at least as expensive as conventional care. In addition, Campbell et al17 found that, while hospice use correlated with reduced Medicare expenditures among younger patients, it increased expenditures among patients without cancer and those older than 84 years.
Hospice studies of patients with cancer are of particular interest because cancer is one of a few diseases in which fairly accurate predictions of death are possible18; therefore, hospice stays could be used earlier in treatment courses compared with other terminal diseases. Cancers account for more than 35% of all hospice admissions, or more than double than for any other terminal diagnosis.32 In addition, compared with other decedents, patients with cancer have higher costs at the end of life.33 Because of higher spending and more accurate prognoses, patients with cancer may theoretically have more expenditures to be saved15 and may represent a potential target for substantial cost savings at the end of life.
Our study of terminally ill patients with head and neck cancer demonstrated that in the last month before death, the mean cost per patient who was enrolled in hospice was more than $7000 less than the mean cost per patient who did not use hospice services. Moreover, the mean monthly costs per patient were significantly lower if hospice care was used earlier rather than later in the treatment course, with more than a $10 000 difference per month for patients with hospice care that started more than 90 days before death compared with patients with no hospice care. These findings add to the growing body of evidence that demonstrates that patients with cancer who use hospice services have significantly lower expenditures.
Investigations on cost savings at the end of life are plagued by claims that selection bias may explain cost differences found in nonrandomized studies.2,15,25,26 Patients who enroll in hospice care are fundamentally different from patients who do not. They tend to be younger and of white race and have had cancer longer than patients with cancer who do not use hospice services.2,15,34 In addition, they tend to be from higher socioeconomic groups and have informal support structures that enable them to obtain additional services that may not be measured on traditional cost estimates.34 To address this bias, we performed a propensity score analysis adjusting for specific cancer site, age at diagnosis, year of diagnosis, race, marital status, urban or rural geographic code, and SEER historical stage. The propensity score–matched results were similar to the unadjusted results, demonstrating that among patients with head and neck cancer those who received hospice services had more than $7000 lower costs in the last month of life compared with those who did not receive hospice services. Therefore, despite fundamental differences in patient populations, there still remained a greater cost savings between those who used hospice services and those who did not. Although there may be inherent differences between patients who choose hospice admission and those who do not use hospice services, hospice admission still has a significant effect on costs.
There are several limitations to our study. Patients with cancer have a different treatment course than patients with other terminal diseases; therefore, our findings lack generalizability to other diseases. Also, while propensity score matching addresses selection bias due to measured confounders, it cannot account for differences in the cohorts not captured by the database. In addition, because an administrative database was used, other health outcomes, such as quality of life, patient satisfaction, or clinical details, could not be included. Unpaid caregiving by family members also could not be measured and may have contributed to cost differences.
Our findings suggest that patients with head and neck cancer who enroll in hospice care have significantly lower costs in the final month of life relative to patients with head and neck cancer who do not receive hospice care. Health care stakeholders may want to increase hospice funding and significant expansion of these services. Encouraging hospice admissions for patient care provides not only compassionate, dignified treatment at the end of life but also an opportunity for substantial savings in health care costs.
Submitted for Publication: April 2, 2015; final revision received August 11, 2015; accepted August 25, 2015.
Corresponding Author: Christopher S. Hollenbeak, PhD, Department of Surgery, College of Medicine, The Pennsylvania State University, 500 University Dr, Mailbox Code H151, Hershey, PA 17033-0850 (email@example.com).
Published Online: October 8, 2015. doi:10.1001/jamaoto.2015.2162.
Author Contributions: Mr Schaefer and Dr Hollenbeak had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Schaefer, Hollenbeak.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Enomoto, Schaefer, Hollenbeak.
Critical revision of the manuscript for important intellectual content: All authors.
Study supervision: Goldenberg, Koch, Hollenbeak.
Conflict of Interest Disclosures: None reported.
Funding/Support: This study was supported by grant 5 R03 DEOOO019511 from the National Institute of Dental and Craniofacial Research, National Institutes of Health.
Role of Funder/Sponsor: The funding source 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: This study used the Surveillance, Epidemiology, and End Results (SEER)–Medicare linked database. The interpretation and reporting of these data are the sole responsibility of the authors.
Previous Presentation: This study was presented at the Annual Meeting of the American Head and Neck Society; April 22, 2015; Boston, Massachusetts.
Additional Contributions: We acknowledge the efforts of the Applied Research Program, National Cancer Institute; the Office of Research, Development, and Information, Centers for Medicare & Medicaid Services; Information Management Services, Inc; and the SEER program tumor registries in the creation of the SEER-Medicare linked database.