Palliative Radiotherapy in Medicare-Certified Freestanding Hospices

Transcription

1 780 Journal of Pain and Symptom Management Vol. 37 No. 5 May 2009 Original Article Palliative Radiotherapy in Medicare-Certified Freestanding Hospices Stephanie L. Jarosek, RN, BSN, Beth A. Virnig, PhD, and Roger Feldman, PhD Department of Health Policy and Management, School of Public Health, University of Minnesota, Minneapolis, Minnesota, USA Abstract Hospice care is designed to provide a variety of services, including pain and symptom management, to terminally ill patients. Although palliative radiotherapy (PRT) has been shown to be effective in reducing pain and other symptoms related to tumor growth, only a few hospice patients receive this therapy. This analysis identifies Medicare-certified freestanding hospices that report use of radiotherapy using Medicare s Healthcare Cost Report Information System (HCRIS) dataset. Any reported cost for radiotherapy services was used to indicate provision of PRT because of the population served. The relationship of provider characteristics (ownership, profit status, percent of patients with a cancer diagnosis, geographic location, and size) with provision of PRT was analyzed. Overall, 23.8% of Medicare-certified freestanding hospices in the study population provided radiotherapy services in fiscal year Provision of radiotherapy services was associated with larger size (measured by total number of hospice days reported in the HCRIS), longer length of Medicare certification, not-for-profit status, and a higher proportion of patients surviving more than seven days after admission. The finding that size, length of Medicare certification, and profit status are associated with provision of radiotherapy services lends credence to suggestions that current reimbursement practices discourage the use of PRT in hospice care, particularly for low-volume hospices. J Pain Symptom Manage 2009;37:780e787. Ó 2009 U.S. Cancer Pain Relief Committee. Published by Elsevier Inc. All rights reserved. Key Words Medicare, hospice, palliative radiation Introduction There are two primary reasons for radiation therapy in oncologic settingsdtreatment and palliation. Palliative radiotherapy (PRT) is This work was supported by a grant from the American Cancer Society. Address correspondence to: Stephanie L. Jarosek, RN, BSN, Department of Health Policy and Management, School of Public Health, University of Minnesota, 420 Delaware Street, MMC 729, Minneapolis, MN 55455, USA. herb0079@umn.edu Accepted for publication: May 7, Ó 2009 U.S. Cancer Pain Relief Committee Published by Elsevier Inc. All rights reserved. given for relief of symptoms associated with cancer tumors and metastases without the intent to cure the disease. The most common indication for PRT is pain related to bone metastases. Other indications include complications relating to tumor mass (spinal cord compression, brain metastases, and obstruction of airway or gastrointestinal system) and control of bleeding. 1 Although palliative radiation is a costly service, studies demonstrate both effectiveness and cost-effectiveness of PRT vs. other forms of palliation, such as pain medication. 2 Hospice care is designed to provide a variety of services to terminally ill patients, primarily /09/$esee front matter doi: /j.jpainsymman

2 Vol. 37 No. 5 May 2009 Palliative Radiotherapy in Freestanding Hospices 781 focusing on pain relief and symptom management. Medicare covered an estimated 80% of patients receiving hospice in the United States in 2005 through the hospice benefit. 3 The benefit is available to anyone covered by Medicare Hospital Insurance (Part A) who has been certified by a physician as having a prognosis of less than six months to live, who agrees to stop pursuing curative treatment, and who receives care from a hospice provider certified by the Medicare program. Medicare-certified hospices are charged with providing any services documented as reasonable and necessary for the palliation and management of the patient s terminal illness. 4 Medicare pays hospices on a capitated rather than cost basis (i.e., they receive a fixed per diem for each patient). In general, capitated payment systems provide an incentive for providers to reduce costs to a level below the payment leveldthe greater the difference between the capitated payment and the cost of treatment, the more money the firm keeps. Qualitative research has suggested that hospices will not accept patients who require high-cost services like palliative radiation. 5,6 Surveys of hospice providers point to a majority of hospices being willing to accept patients desiring PRT (78% 7 and 71% 8 ). At present, it is not clear how often the hospices actually provide the services. This analysis identifies the Medicare-certified freestanding hospices that provide radiotherapy services and the hospice characteristics associated with provision of services. Methods Data The publicly available Centers for Medicare and Medicaid Services (CMS) Healthcare Cost Reporting Information System (HCRIS) data for freestanding hospices (CMS, 1984e99) in combination with Medicare hospice claims and provider of service (POS) data present an opportunity to study the use of PRT in hospice. Since 1999, Medicare-certified hospices have been required to submit annual cost reports, including facility, operational and financial information, to a fiscal intermediary (FI). The data were downloaded from the CMS website in April Hospice claims were matched with the HCRIS and POS data by provider number. Information for all hospices was obtained from the POS file for the third quarter of 2003, Medicare HCRIS from fiscal year (FY) 2002 (hospice FYs beginning September 30, 2001 to September 29, 2002), and hospice claims data from 2000e2002. HCRIS data provided information on the location, ownership, size, profit status, number of years of certification, and provision of radiation services, whereas data on the proportion of patients who had cancer (calendar year (CY) 2002) and proportion of patients with short stays (seven or fewer days) (CY 2000e2002) were obtained from 100% hospice claims data. The POS data provided information to double check the validity of HCRIS data, in addition to providing information on the existence of radiation facilities within a hospice s service area. The HCRIS data present a challenge because they do not distinguish between fields that are not completed and those in which costs were truly zero. Given the range of services hospices are responsible for providing, we considered it unreasonable for a hospice to have zero drug costs. Thus, for this study, a cost report was considered complete if both drug costs and total costs were reported. Eighty-one providers (7.6%) were excluded from the analysis because of missing both drug and total costs (62 of the 81 cases) or missing drug cost data (19 of the 81 cases). Finally, inadequate information on the lengths of stay in some hospices and inability to match claims to providers in others resulted in the exclusion of 43 cases from the analysis. The final sample includes 953 freestanding hospices operating in 2002 (representing 89% of the freestanding hospices). Table 1 displays key characteristics, including all hospices (hospital, skilled nursing facility [SNF] and home health agency [HHA]-based hospices, in addition to freestanding hospices), all freestanding hospices, and only those hospices included in the final sample. Our sample includes a smaller proportion of small hospices than the entire sample of freestanding hospices, but is otherwise representative of freestanding hospices. Hospices in our sample are significantly different from all hospices in that freestanding hospices are more likely to be located in the Southern U.S. Census Region, and have lower proportions of cancer

3 782 Jarosek et al. Vol. 37 No. 5 May 2009 Table 1 Characteristics of All Hospices, Freestanding Hospices, and Freestanding Hospices in Final Sample (FY 2002) Hospice Characteristics Hospices Included in Final Sample All Freestanding Hospices All Hospices n % a n % a n % a Total Ownership Not-for-profit For-profit Region Northeast e Midwest e South e West Length of Medicare certification 3þ years c <3 years c Cancer case mix Missing d e 0%e32% e 33%e49% e 50þ% e Size b 0e10,900 patient days c 10,901e26,100 patient days ,101e502,281 patient days Urban influence Metropolitan Adjacent Rural % of stays #7 days Missing e e 0e e 24e e e a Percentages may not total to 100 due to rounding. b Size information not available for all hospices due to inadequate reporting in the HHA and SNF HCRIS. c P < d P < e P < 0.001, indicating a statistically significantly different proportion compared with hospices in the sample, using a z-test for different proportions. patients and lower proportions of patients with short lengths of stay. Medicare hospice HCRIS data have been used to determine service provision in hospices; 9 however, they are a relatively new data source. For this reason, a number of external validity checks were performed. Hospice costs were compared against the Medicare hospice claims data, and information on ownership was compared with the Medicare Hospice POS file, with adequate overlap. Variables 1. Provision of radiation (range of dollars [annual]: $0e$758,456): Hospices were considered to provide radiation services if they reported radiation costs greater than $0. 2. Size of hospice (range 14e502,281): Total number of patient days reported by the hospice, adjusted for the number of the months in the reported FY and classified into tertiles. 3. Cancer case mix of Medicare patients (range 3.8%e88%): Using the hospice claims, we divided the total number of patients with a cancer diagnosis by the total number of patients and classified the result into tertiles. 4. Urban influence: Using a zip code-county crosswalk file, we assigned an urban

4 Vol. 37 No. 5 May 2009 Palliative Radiotherapy in Freestanding Hospices 783 influence code to hospices using the United States Department of Agriculture (USDA) s 2003 classification, which divides counties into three groups (metropolitan, adjacent, and rural) based on whether a county contains or is adjacent to a metropolitan area Ownership: Hospices were classified as for-profit (proprietary) and not-for-profit (including nonprofit and governmentowned) based on reporting of provider control in the cost report. 6. Medicare certification (recent/not recent): Hospices that had been certified by Medicare during each of the three years prior to the cost report were considered recently certified. 7. Region: Hospices were assigned to census region (northeast, midwest, south, and west) based on the state in which the facility is located. 8. Percent of Medicare patients with stays less than or equal to seven days (range 0%e55%): Using the hospice claims data, we divided the number of patients with stays less than or equal to seven days by the total number of patients in each hospice. 9. Therapeutic radiation provider in service area: POS data were used to compile a list of hospitals that provide therapeutic radiation and their zip codes. Service area maps from a previous analysis 11 listing zip codes served by each hospice were combined with POS data to create a dichotomous variable (Y/N) indicating the presence of a therapeutic radiation provider within the service area. Analysis Bivariate analysis of the association of provision of the radiation services with size of hospice, cancer case mix, urban influence, ownership, duration of Medicare certification, region and percent of stays less than or equal to seven days was conducted. Each of these characteristics was seen as important in this analysis. Urban influence (metro/adjacent/rural) could not be included in the multivariate analysis because of collinearity with the measure of size. Multivariate analysis consisted of estimating logistic regression in SAS for the provision of radiation services from key hospice characteristics mentioned earlier: Radiation (0,1) ¼ b 0 þ b 1 ownership þ b 2 region þ b 3 certification þ b 4 cancer þ b 5 size þ b 6 percent þ b 7 service in area þ error The range of reported costs for radiation services for those hospices reporting costs ($21e$758,456) includes low values, which caused us to question whether PRT could have been provided for the low reported cost. Costs of radiation therapy for bone metastases in the literature are estimated to range from $700 to $1700 for a single patient. 4 Although low costs may be possible if hospices have special financial relationships with providers (e.g., who donate of time or services to dying patients), because cost reports are primary financial reporting tools rather than analytic tools, it is necessary to assess the possibility of inaccurate values. The analysis was repeated labeling hospices as having provided radiation services only if reported radiation costs exceeded $500, $1000, and $2000 to examine the sensitivity of the results to this change. Results The freestanding hospices in the sample are slightly more likely to be nonprofit (52.7%), and most likely to be located in the south (50.7%) and in metropolitan areas (72%) than freestanding hospices not included in the sample. Fewer hospices included in our sample (13.8%) became certified in the last three years (Table 1). About one-fifth (227 of 953) of the hospices reported radiation costs (Table 2). Bivariate analyses reveal that the proportion of hospices reporting radiation costs significantly varies by hospice, cancer case mix, urban influence, ownership, duration of Medicare certification, and percent of stays less than or equal to seven days, but does not vary by census region. Differences in reporting radiation costs by size are the most dramatic, varying from 38.2% in the largest, 24.2% in mid-size hospices, and only 9.3% in the smallest. Multiple logistic regression results indicate that large size, not-for profit status, and the percent of patients with stays less than seven

5 784 Jarosek et al. Vol. 37 No. 5 May 2009 Table 2 Characteristics of Freestanding Hospices in the Final Sample Reporting Costs for Radiation Hospice Characteristics Hospices Reporting Radiation Costs n % Total Ownership a Not-for-profit For-profit Region Northeast Midwest South West Length of Medicare certification a 3þ years <3 years Cancer case mix a (%) 0e e þ Size a 0e10,900 bed days ,901e26,100 bed days ,101e502,281 bed days Urban influence a Metropolitan Adjacent Rural % of stays #7 days a 0e e e e Therapeutic radiation provider in service area a Yes No a Chi-square of characteristic by radiation provision (0,1) significant at P < level. days are significantly associated with providing radiation services. After controlling for all factors, large size appears to be the dominant characteristic associated with provision of PRT, with large hospices over seven times more likely to provide PRT as small hospices (odds ratio [OR]: 7.91, 95% confidence interval [CI]: 4.67e13.41) (Table 3). For-profit hospices are half as likely to provide PRT as not-for-profit. The percent of patients with stays less than or equal to seven days was also significant, with the hospices having a high percentage of patients with short stays less likely to provide PRT. Hospices with a provider of therapeutic radiation within the hospice s service area had significantly higher odds of providing PRT (OR: 2.101, 95% CI: 1.027e4.298). Results were similar as the threshold for the radiation costs counted as PRT increased, although effect sizes, especially for size, were larger. At a threshold of $1000, the first and second quartiles of short-stay patients reverse so that hospices in the first quartile are most likely to provide PRT. Discussion Many hospice professionals agree that consideration of radiotherapy is indicated in situations of painful bony metastases, spinal cord compression, painful tumor mass, and brain metastases. 7 Three of the four most common cancers that lead to mortalitydbreast, prostate, and lungdare likely to spread to the Table 3 Logistic Regression Analysis Predicting Reporting Radiation Costs in Medicare-Certified Freestanding Hospices in 2002 Hospice Characteristics Odds Ratio (95% CI) Ownership For-profit 1.00 Not-for-profit 2.50 a (1.631, 3.817) Region South 1.00 Midwest (0.972, 2.389) Northeast c (1.242, 3.882) West b (1.072, 2.773) Length of Medicare certification 3þ years 1.00 <3 years b (1.100, 5.845) Cancer case mix (%) 0e e (0.942, 2.550) 44e b (1.032, 3.048) Size 0e10,900 bed days ,901e26,100 bed days b (1.885, 5.146) 26,101e502,281 bed days a (4.673, ) Percent of patient stays #7 days 0e b (1.515, 4.764) 23e a (1.853, 4.952) 29e b (0.818, 2.139) 35e Therapeutic radiation provider in service area Yes b (1.027, 4.298) No 1.00 HosmereLemeshow: Chi-square ¼ , df ¼ 8, P ¼ CI ¼ confidence interval. a P < b P < c P < 0.01, based on Wald Chi-square statistic.

6 Vol. 37 No. 5 May 2009 Palliative Radiotherapy in Freestanding Hospices 785 bone, 12 resulting in a large number and proportion of hospice patients who could benefit from the treatment. The low proportion of hospices reporting radiation costs is surprising, given the general agreement regarding the efficacy of palliative radiation and the frequency of the conditions that can benefit from it in the hospice setting. Current reimbursement practices may discourage the use of PRT, because the high costs of treatment outpace the hospice capitation. The association between large size and provision of services suggests that large hospices are better able to absorb the high costs of offering radiotherapy, a finding similar to previous survey results. 8 While survey results by Lorenz et al. 8 found that for-profit providers were as likely to offer radiation services as notfor-profit providers, our findings are similar to McCue and Thompson 9 that not-for-profit hospices are much more likely to provide the service. This finding is consistent with previous work, which suggests that pure profit maximization results in different strategies than maximization that includes a nonprofit motive. 13 The finding of no significant regional variation in report of radiation costs may be related to the lack of regional variation in hospice payment policy. There are two possible explanations for the statistically significant finding that hospices with a high percentage of stays less than or equal to seven days are less likely to offer palliative radiation among the percent of patients with stays less than or equal to seven days. First, hospices with many patients with shorter lengths of stay may have less ability to spread the costs of PRT, because the per diem payment does not take into account the higher costs during admission and just before death. With very short lengths of stay, there are fewer lower cost days in which to recoup those costs, making patients with short lengths of stay expensive to the hospice. It is also possible that practice style in areas with hospices with a higher proportion of patients with shorter lengths of stay favors providing palliative radiation before referral to hospice, thus increasing the number of patients with very short stays. Another explanation may be that patients with very short lengths of stay do not have the opportunity to benefit from PRT, because there is not enough time to initiate the therapy or because when death is imminent, the patient is simply too ill to face transportation to radiation centers. If this were the case, however, the only reason for zero radiation costs at the hospice level would be that all patients had lengths of stay that were too short to initiate PRT. It is likely that both factors play a role. Individual-level studies of decision making regarding PRT will help to inform this finding. Medicare s HCRIS is intended for reimbursement purposes, and thus presents many challenges for use in research. First, palliative radiation costs cannot be separated from other radiation costs as both are lumped into the radiation cost category. Sensitivity analysis demonstrates that our findings persist despite the range of reported costs for radiation, including low values. To the extent that non-palliative radiation services are counted or hospices with truly zero radiation costs were counted as missing, these findings may be biased. However, given the nature of the patient population and the focus on quality of life rather than cure, it is unlikely that a large amount of diagnostic imaging or procedures requiring radiation occur. Furthermore, reporting requirements differ across provider types. Although cost information for other providers of hospice services (hospital, HHA, or SNF) is also reported on hospital, HHA and SNF reports, over half of the hospital and SNF records had missing information on total costs or drug costs for hospice services. Besides, the hospice-level nature of the HCRIS data forced us to aggregate claims data to the provider, and did not permit conjecture as to the number of patients in Medicare-certified hospices who receive palliative radiation. We address these limitations of these data by focusing on freestanding hospices, which have the least missing data for drug costs and total costs. However, these findings cannot be generalized to non-freestanding providers of hospice services or to hospices that are not Medicare certified. Finally, small hospices were more likely to have missing drug costs or total radiation costs and hence to be excluded from this analysis. To the extent that these hospices had radiation costs that were truly missing, these results may be biased and must be interpreted with care. However, the trend in the proportion of providers reporting radiation costs from small,

7 786 Jarosek et al. Vol. 37 No. 5 May 2009 medium, and large hospices increases confidence in the findings. HHA, SNF, and hospital-based hospices had large proportions of missing data on the number of bed days; hence, we are unable to determine the extent to which the study population differs from the total population. Other barriers to providing radiation therapy for individuals in hospice settings are the length of survival after entry into hospice, burden of travel and time for radiation treatments, unwanted side effects, and, for some, a desire for minimal medical intervention at the end of life. Individual-level factors are outside the scope of this analysis. Only two of the factors, burden of travel for radiation treatments and a hospice s commitment to removing traditional medical treatments from the end of life, could affect all patients of any hospice. Individual patient preferences would not impact the final result, which does not predict the amount of radiotherapy but rather whether or not any patient in care of the hospice received it. The existence of a facility that provides radiation in the hospice s service area is controlled for and would, in theory, mean reduced travel time to radiation therapy for patients. It is noteworthy that the lack of such a facility within a hospice s service area does not imply that no patient will receive radiation. Conclusion While PRT is well within the scope of services covered under the hospice benefitd those documented as reasonable and necessary for the palliation and management of the patient s terminal illnessdthe high costs associated with radiation therapy make it a financial burden for hospices, in part because Medicare payments are not associated with the cost of individual treatments. The finding that size and profit status are associated with provision of radiotherapy services lends credence to previous findings that current reimbursement practices discourage the use of PRT in hospice care, particularly for low-volume hospices. In effect, this payment strategy means that a hospice is an HMO for dying patients (as termed by the CEO of a large hospice in a previous qualitative research report), 14 meaning that hospices rely on volume to balance the cost of patients with less expensive stays with those requiring more expensive treatments. Large hospices also may use their volume of patients to negotiate lower prices. 5,15 These findings have an important implication for patients who could be served by a low-volume hospice; physicians may delay referral to hospice because of concern about the effect of high-cost treatments on the hospice s bottom line, or because they know that the hospice will not provide the service. 5 A greater understanding of whether the need for PRT impacts the choice of hospice and the timing of hospice enrollment is needed. Additional analysis of the costs and benefits of amending hospice payment to more appropriately reimburse for palliative radiation expenses is also warranted if the goal is optimum symptom management for terminally ill cancer patients. References 1. Fine PG. Palliative radiation therapy in endof-life care: evidence-based utilization. Am J Hosp Palliat Care 2002;19(3):166e Konski A. Radiotherapy is a cost-effective palliative treatment for Patients with bone metastasis from prostate cancer. Int J Radiat Oncol Biol Phys 2004;60(5):1373e NHPCO s facts and figures-2005 findings, National Hospice and Palliative Care Organization. Alexandria, VA: National Hospice and Palliative Care Organization, Hospice Care-Covered Services. 2005:42 CFR Huskamp HA, Buntin MB, Wang V, Newhouse JP. Providing care at the end of life: do Medicare rules impede good care? Health Aff 2001;20(3):204e Medicare Payment Advisory Commission. Medicare beneficiaries access to hospice. Report to the Congress. May Available from medpac.gov/publications/congressional_reports/ may2002_hospiceaccess.pdf. Accessed June Lutz S, Spence C, Chow E, Janjan N, Connor S. Survey on use of PRT in hospice care. J Clin Oncol 2004;22(17):3581e Lorenz KA, Ettner SL, Rosenfeld KE, et al. Cash and compassion: profit status and the delivery of hospice services. J Palliat Med 2002;5(4):507e McCue MJ, Thompson JM. Operational and financial peformance of publicly traded hospice companies. J Palliat Med 2005;8(6):1196e United States Department of Agriculture, Economic Research Service. Measuring rurality:

8 Vol. 37 No. 5 May 2009 Palliative Radiotherapy in Freestanding Hospices 787 urban influence codes. Available from ers.usda.gov/briefing/rurality/urbaninf/. Accessed June Virnig BA, Ma H, Hartman LK, Moscovice I, Carlin B. Access to home-based hospice care for rural populations: identification of areas lacking service. J Palliat Med 2006;9(6):1292e Mundy GR. Metastases to bone: causes, consequences and therapeutic opportunities. Nat Rev Cancer 2002;2:584e Sloan FA. Not-for-profit ownership and hospital behavior. In: Culyer AP, Newhouse JP, eds. Handbook of health economics. Amsterdam: Elsevier, Casey MM, Moscovice IS, Virnig, Kind, SB. Models for providing hospice care in rural areas: successes and challenges. Rural Health Research Center. Working Paper Series #46. University of Minnesota, Minneapolis, MN. Available from pdf. Accessed June Casey MM, Moscovice IS, Virnig, Durham SB. Providing hospice care in rural areas: challenges and strategies. Am J Hosp Palliat Care 2005;22(5): 363e368.