key: cord-0757057-s23o6mu0 authors: Hoverman, J. Russell; Mann, B. Brooke; Phu, Sara; Nelson, Philip; Hayes, Jad E.; Taniguchi, Cynthia B.; Neubauer, Marcus A. title: Hospice or Hospital: The Costs of Dying of Cancer in the Oncology Care Model date: 2020-06-25 journal: Palliat Med Rep DOI: 10.1089/pmr.2020.0023 sha: dbec178d7a5417931d5e08cf8bef3be6489c935e doc_id: 757057 cord_uid: s23o6mu0 Background: End-of-life management is a difficult aspect of cancer care. With the oncology care model (OCM), we have data to assess both clinical outcomes and total cost of care (TCOC). Objective: To measure and characterize the TCOC for those who received less than three days of hospice care (HC) at the end of life compared with those who received three days or more. Design: Assess data on costs and site and date of death from Medicare claims on patients identified in the OCM who received chemotherapy in the six months before death. Standard statistical methods were used to characterize both populations. Setting/Subjects: Subjects were Medicare patients with cancer who died while managed by U.S. oncology practices in the OCM. Measurements were TCOC in 30-day intervals for the last months of life, cost by site of care at the end of life, and demographic characteristics of the population and association with HC. Results: There were 7329 deaths. Dying in the hospital was twice the cost of dying at home under HC ($20,113 vs. $10,803). Of demographic groups measured, only black race and a lymphoma diagnosis had <50% hospice enrollment for three days or more before death. Conclusions: This study reinforces previous studies regarding costs in the last 30 days of life. The graphic representation highlights the dollar cost and the costs of lost opportunity. Using these data to improve communication, addressing socioeconomic support, and formal palliative care integration are potential strategies to improve care. The cancer community, including patients, families, and providers, are of two minds when it comes to making decisions at the end of life. This dichotomy can be represented by the following two competing narratives. (The patients in these narratives are abstractions from many patients and do not represent any single patient's story.) Narrative #1: S.G. was a 75-year-old gentleman with lung cancer. After a brief response with combined therapy including immunotherapy, he was started on sec-ond line therapy. After two cycles he had lost weight and was short of breath. His physician suggested his disease may be progressing. On discussing options, the patient stated he did not want to ''give up.'' They chose a third line of chemotherapy. Within days he was admitted to the hospital with fever and profound weakness. After discharge to a rehabilitation facility, he developed fever and respiratory failure, was readmitted, and died in the hospital. Narrative #2: W.H. was an 85-year-old woman with metastatic colon cancer. She tried standard first line 1 U.S. Oncology Network, Texas Oncology, Dallas, Texas, USA. 2 McKesson Corp., The Woodlands, Texas, USA. *Address correspondence to: J. Russell Hoverman, MD, PhD, U.S. Oncology Network, Texas Oncology, No. 500, 12221 Merit Drive, Dallas, TX 75251, USA, E-mail: jrussell.hoverman@usoncology.com therapy but stopped due to side effects. She tried a clinical trial (''maybe it will help someone'') but again stopped due to toxicity. She was adamant that she did not want to spend resources on heroic measures and on only what was needed to keep her comfortable. She enrolled in hospice care (HC) and died at home six weeks later. These narratives are both examples of patients doing what they think is ''everything,'' or at least, not doing ''nothing.'' The dichotomy between hospital and hospice was demonstrated by a recent report that showed that, for the first time in over half a century, more patients died at home than in the hospital. 1 Even with this apparent cultural change, an end-oflife discussion remains one of the most difficult aspects of cancer care. In 2016, funded by provisions in the Affordable Care Act, the Centers for Medicare and Medicaid Innovation began a program with Medicare called the oncology care model (OCM). The intention was to provide support for practices to examine and improve their processes to meet the goals of improving quality and reducing costs. A chief metric was to increase the use of HC at the end of life. 2, 3 With the OCM, oncology providers have access to all claims data associated with their Medicare cancer patients. The U.S. Oncology Network (USON) had 16 practices in 13 states participating in the OCM. The data from the first two years of the program (2016-2018) on 7329 patients are presented here to display the costs of care in the last six months of life, and especially in the last 30 days, to better inform our providers and patients as these decisions are made and give insight into how to improve care. Centers for Medicare and Medicaid Services provided claims data and demographic details, including date of death, for OCM-enrolled patients attributed to USON practices. We measured 30-day interval OCM episode expenditures for n = 7329 deceased patients, as well as patient demographics, clinical outcomes (hospice, hospital/nonintensive care unit [ICU], ICU/hospital, and emergency room visits), and treatment categories in the last 30 days. The population was divided into two cohorts: those with three days or more of HC and those with fewer than three days or no hospice care (NOHC). The three-day differentiation is based on data that hospice for less than three days is no better than NOHC. This metric is validated by the National Quality Forum (NQF). 4 Expenditures were compared between HC and NOHC at the univariate level using a Mann-Whitney U test; categorical variables were compared using chi-squared tests. Multivariate regression was used to determine the association of HC with expenditures in the 30 days before death (EXP30) adjusted for demographic and disease factors. To identify each patient's place of death, we first evaluated discharge status codes on hospice claims where the discharge date was the same as the patient's date of death. If hospice claims were not available (e.g., in cases wherein the patient did not die while in HC), then we evaluated revenue center codes and patient discharge status codes from any available inpatient claims wherein the discharge date was the same as the patient's date of death. If neither hospice claims nor inpatient claims yielded a place of death, the patient's place of death was categorized as ''unknown.'' Of 50,569 individual patients, there were 7329 deaths (14.5%). HC had mean last EXP30 reduction of À$5,641 versus NOHC (95% confidence interval À$6,009 to À$5,252) adjusted for demographics and disease. HC had lower rates of death in hospital (0% vs. 42.1%, p < 0.0001). Mean expense by days before death was as follows: NOHC values 0-30 days = $20,113; 31-60 days = $12,438; and 61-90 days = $10,258; and HC values 0-30 days = $10,803; 31-60 days = $10,960; and 61-90 days = $9,311. These data are presented in graphic format and Figure 1 shows the average for both groups of patients over time, illustrating a difference of *$10,000 in the last 30 days. Table 1 summarizes hospice versus nonhospice use by demographic group. Table 2 gives expenditures by hospice status and place of death. The total cost of care for episodes ending in death was $318.6 million with the cost of the last 30 days accounting for 34.2% ($109 million) of total cost. All categories of care (hospital, drugs, professional visits, laboratories imaging, etc.), with the exception of hospice, were higher in the NOHC cohort. The population of this study is unique in that these are only cancer patients who have received treatment within the last six months, the population is representative of the population of the country, there are data for different racial and ethnic groups, there are diseasespecific data, and the treatment options include current drug choices. The costs associated with the metric of hospice for three days or more (HC) are also presented in a cancer population for the first time. A critical value of this study is the graphic representation of the trajectory of end-of-life costs and potential lost opportunities depending on treatment choices, particularly HC. The current experience with ICU deaths in the coronavirus (COVID-19) epidemic and the pervasive unwanted specter of dying alone make the choices depicted by these data even more compelling. The overall rate of hospice admission was 56.2%. The demographic data confirm previous studies in that black patients generally have lower hospice use rates (45.7%) than other racial groups, and those with hematologic malignancies also have lower rates (42.3%). Each group, black patients and those with lymphoma, has a significant proportion of patients who chose HC. In the case of race, there may be opportunities to address socioeconomic issues both late in disease 5 but before hospice admission, and to reduce hospice disenrollment shortly before death. 6 There is no racial group where the majority would choose to die in the hospital. 7 In the case of hematologic malignancies, this may be an opportunity to target palliative care involvement. [8] [9] [10] [11] The demographic differences do not alone determine decisions. Obermeyer et al., 12 using an extensive retrospective Medicare database, showed increased costs in the last month of life for patients with poor prognosis cancer. Gidwani-Marszowshi et al., 13 in a 98% male Veterans Administration database, demonstrated that costs with HC were substantially lower and associated with much less aggressive care than NOHC. Chastek et al., 14 in a commercially insured population, reported similar cost findings. These studies, and others, consistently find that aggressive end-of-life care is more costly without an improvement in survival and is associated with a demonstrated detrimental impact on the satisfaction and well-being of caregivers and family. 15 This study also shows high cost for those dying in a hospital, particularly in the ICU. There is likely a disproportionate number of black patients and those with heme malignancies and hospice duration of less than three days who die in hospital-based hospices. 16 The course of a cancer treatment is usually described as a representation of the bell-shaped curve skewed to the right with a survival tail-''a median survival of eight months with a two-year survival of 20% and a five-year survival of 5%,'' for example. Physicians and the cancer community are comfortable with pursuing improbable events-moderate to high risk with high reward. However, the community is not well equipped to identify highly improbable events based on age, fitness, comorbidity, tumor responses, line of therapy, and previous toxicities, for example. 17, 18 The profession and the public are enamored with the tail of the curve (the five-year survival of 5%). This image was brought to the public domain by Stephen J. Gould's article 19 -The Median is not the Message-where his own experience as a young man with peritoneal sarcoma argued for pursuing improbable outcomes. However, these bell-shaped curves do not represent the potential nonmonetary costs of treatment. The ''hockeystick'' shape of the graph presented in Figure 1 depicts not only dollar costs but also the costs of lost opportunity. There is loss of the opportunity to maximize comfort, to make choices to be at home, to affirm relationships, to reduce the burden on caregivers and family, and to say goodbyes. 20 The cautionary tale from this image is to be aware of the costs of the one surgery or one dose of chemotherapy too many that leads to dying in the hospital or ICU. Dr. Gould was very aware of the outcomes of pursuing highly improbable choices. Twenty-eight years after his initial episode with sarcoma, he was diagnosed with lung cancer. He was older and the disease had spread widely. He died in his library at home surrounded by his family and his books. 21 These data can provide benchmarks using the OCM (NQF) hospice metric for assessing the success of endof-life programs. They also suggest support where socioeconomic conditions place barriers to HC and suggest areas where formal palliative care evaluation could inform decision making. There are limitations to this study. The cause of death is unspecified and we cannot be sure of the direct relation to cancer. There may be regional or practice cultural differences that impact the use of HC. However, the consistency of the data suggests the trajectory of the curve is likely to be little changed. Also, the place of death for 13% of the population is unknown. Given the approximation of cost for this group to the HC group, it may be that many of these died at home, either unexpectedly or with the care of family. Within the USON, the change in payment structure and the continual review of quality metrics have focused our attention on end-of-life care. We have developed communication tools to begin an assessment of values early in the course of disease. 22 The network has emphasized improved communication with patients with more follow-up and symptom management calls, and internally with team building and systematic review of high-risk patients. 23 A promising aspect of the OCM is the interrogation of data at multiple sites both in and out of USON to uncover and share best practices. 24 Hospice utilization has increased over the course of the OCM but much remains to be done. Changes in the place of death in the United States Using an episode-based payment to improve oncology care The Centers for Medicare and Medicaid Services: Oncology care model halfway through-perspectives from diverse participants National Quality Forum: Palliative and End-of-Life Care Project: 2015-2016. Palliative and End-of-Life_Care_Project:2015-2016/Draft_Report_for_Comment.aspx (Last accessed Racial and ethnic differences in end-oflife care in the United States: Evidence from the Health and Retirement Study (HRS) Live Discharge from hospice due to acute hospitalization: The role of neighborhood socioeconomic characteristics and race/ethnicity Racial and ethnic differences in preferences for end-of-life treatment Patterns of hospice use in patients dying from hematologic malignancies Early palliative care in hematologic malignancies: A review of current knowledge The impact of a of a home-based palliative care program in an Accountable Care Organization Bridging the gap: A palliative care consultation service in a hematological malignancy-bone marrow transplant unit Association between the Medicare hospice benefit and health care utilization and costs for patients with poor-prognosis cancer Health system and beneficiary costs associated with intensive end-of-life medical services Health care costs for patients with cancer at the end of life Place of death: Correlations with quality of life of patients with cancer and predictors of bereaved caregivers' mental health Hospice admissions for cancer in the final days of life: Independent predictors and implications for quality measures Managing cancer patients' expectations amid hope and hype Keeping expectations in check with immune checkpoint inhibitors The median isn't the message Factors considered important at the end of life by patients, family, physicians and other care providers Gould 60 is dead: Enlivened evolutionary history If we don't ask, our patients might never tell: The impact of the routine use of a patient values assessment From angst to analytics: Lessons learned from an oncology care model internal pilot Evaluation of the Oncology Care Model: Performance Period One No funding was received.