Patterns of Care and Costs for Older Patients With Colorectal Cancer at the End of Life: Descriptive Study of the United States and Canada

Abstract

PURPOSE:

End-of-life (EOL) cancer care is costly, with challenges regarding intensity and place of care. We described EOL care and costs for patients with colorectal cancer (CRC) in the United States and the province of Ontario, Canada, to inform better care delivery.

METHODS:

Patients diagnosed with CRC from 2007 to 2013, who died of any cancer from 2007 to 2013 at age ≥ 66 years, were selected from the US SEER cancer registries linked to Medicare claims (n = 16,565) and the Ontario Cancer Registry linked to administrative health data (n = 6,587). We estimated total and resource-specific costs (2015 US dollars) from public payer perspectives over the last 360 days of life by 30-day periods, by stage at diagnosis (0-II, III, IV).

RESULTS:

In all months, especially 30 days before death, higher percentages of SEER-Medicare than Ontario patients received chemotherapy (15.7% v 8.0%), and imaging tests (39.4% v 31.1%). A higher percentage of Ontario patients were hospitalized (62.5% v 51.0%), but 43.2% of hospitalized SEER-Medicare patients had intensive care unit (ICU) admissions versus 17.9% of hospitalized Ontario patients. Cost differences between cohorts were greater for patients with stage IV disease. In the last 30 days, mean total costs for patients with stage IV disease were $15,881 (SEER-Medicare) and $12,034 (Ontario) versus $19,354 and $17,312 for stage 0-II. Hospitalization costs were higher for SEER-Medicare patients ($11,180 v $9,434), with lower daily hospital costs in Ontario ($1,067 v $2,004).

CONCLUSION:

These findings suggest opportunities for reducing chemotherapy and ICU use in the United States and hospitalizations in Ontario.

INTRODUCTION

The United States and Canada are neighboring countries with many similarities, but there are differences in the funding, administration, and delivery of health care.^1-4 In Canada, publicly funded health care insurance plans, managed by provincial and territorial governments, cover the costs of medically necessary services for permanent residents, and most health care is free at the point of care.⁵ In Ontario, Canada’s most populous province (14.2 million in 2017),⁶ the Ontario Health Insurance Plan (OHIP) covers approximately 98.2% of the population.⁵ There is no universal health care program in the United States, although the Medicare program covers 97% of persons age ≥ 65 years.⁷ Unlike patients in Ontario, patients under Medicare have significant out-of-pocket costs for premiums, deductibles, copayments, and coinsurance.^8-10

Canada lacks a formal system of comprehensive end-of-life (EOL) care,^11-13 whereas Medicare offers a hospice program providing symptom control and psychosocial support for patients with a life expectancy of ≤ 6 months.¹⁴ Studies on EOL care found that hospitalization rates and mean per-patient hospital costs were higher in Canada than the United States.^15-17 Total health care costs, however, were higher in the United States, especially when patient-borne costs were considered.¹⁶

Because cancer stage at diagnosis may vary internationally, reflecting risk factor prevalence, prevention, screening, and detection, and may affect health care use and costs, even at the EOL,^18,19 accounting for stage is critical in cross-country studies of cancer care. Ultimately, such studies inform policies for efficient health care, especially at the EOL, a period of high costs fraught with issues regarding appropriate intensity and place of care.

This objective of this cross-country study is to describe use and costs of EOL care in older patients with colorectal cancer (CRC) by stage at diagnosis in the United States and Ontario. Such studies can be hypothesis generating and identify where reductions in use and costs might be possible. On the basis of previous studies,^15,16 we were particularly interested in hospitalization, imaging tests, and chemotherapy. We chose CRC because it is a common cancer that is diagnosed at all stages and is the second-leading cause of cancer deaths in the United States²⁰ and Ontario.²¹

METHODS

The Ontario arm of this study was approved by the Sunnybrook Health Sciences Centre Research Ethics Board, Toronto, Ontario, Canada. Analyses of the SEER-Medicare data were exempt from institutional review board review per the National Institutes of Health’s Office of Human Subjects Research.

Patients

Patients in Ontario were identified from the Ontario Cancer Registry linked to administrative databases using unique encoded identifiers. Patients in the United States were identified from the National Cancer Institute’s SEER cancer registries (SEER 18 excluding Alaska) linked to Medicare claims.²² Both registries collect cancer site, diagnosis date, and stage and age at diagnosis.^7,23

Inclusion criteria required that between January 1, 2007 and December 31, 2013 patients were diagnosed with pathologically confirmed CRC, with no previous or concurrent cancer, and died of any cancer at age ≥ 66 years. We specified any cancer because cause of death might be recorded as the site of metastatic CRC. We excluded patients with unknown stage at diagnosis, missing sex, or survival < 30 days after diagnosis. Patients had to be covered by OHIP or have continuous Medicare fee-for-service coverage and Part A/B enrollment for 12 months before death or from diagnosis to death for patients who survived < 12 months after diagnosis. Data used for cohort selection are described in Appendix Table A1 (online only).

Health Care Data

Ontario.

Administrative data record health care resource use, dates, and diagnostic and procedure codes for all OHIP recipients (Appendix Table A2, online only). Canada Census data include information about patients’ areas of residence, including urbanicity and median household income.²⁴ The Ontario data were analyzed at the Institute for Clinical Evaluative Sciences in Toronto, Ontario.

United States.

Health care use data for the SEER-Medicare cohort, including dates, diagnostic and procedure codes, and payments, were obtained from Medicare claims (Appendix Tables A2 and A3, online only).¹⁹ Urbanicity and median household income for patients’ areas of residence were obtained from US Census data. US data analyses were led by investigators at the National Cancer Institute in Bethesda, Maryland.

Health Care Resource Use and Costs

Our primary cost analysis was conducted from the public payer perspective. We estimated use and costs of health care resources that were defined similarly and publicly covered in both systems (Appendix Table A2): inpatient hospitalizations, emergency room (ER) visits, ambulatory procedures, specific imaging tests (computed tomography [CT], magnetic resonance imaging [MRI], positron emission tomography [PET]), outpatient chemotherapy, physician services, and home care services. We examined intensive care unit (ICU) use among hospitalized patients and death in hospital. The Medicare data combine the costs of ER visits that resulted in hospitalization with the costs of the subsequent hospitalization. Therefore, we did the same for the Ontario data, and ER visits refer to those not resulting in a hospitalization. Because of differences in reporting and costing methodology between health care systems, we could not include use or costs for radiation therapy, outpatient prescription drugs, oral antineoplastic therapies, and laboratory tests.

We examined services specific to the EOL that were not available in both cohorts. In Ontario, we examined use of complex continuing care facilities, which provide medically complex and specialized services to patients with long-term illnesses or disabilities, and palliative care services, which included a selection of the resources mentioned above: hospice care in complex continuing care facilities, physician house calls and palliative care consults, stays in inpatient palliative care units, supportive or EOL home care, and palliative care in the ER (Appendix Table A2). These were identified by codes in the Ontario data,¹³ but codes for palliative care services were not consistently available from Medicare claims during the study period. In the SEER-Medicare cohort, we examined use of hospice and skilled nursing facilities.

All resources defined in this study were funded by the Ontario Ministry of Health. We therefore assumed that our estimated health care costs for the Ontario cohort were fully paid by the public payer. Because Ontario databases mostly record resource use rather than costs, we estimated costs using standard methods for administrative data (Appendix Table A2).^16,25 We used Medicare payments to reflect public costs for US patients.^16,19,26 However, Medicare payments account for only 60%-70% of total medical costs for cancer survivors. The remainder is borne by patients and other payers through insurance premiums, deductibles, coinsurance, and copayments.^8-10 In supplemental analyses, we estimated these patient responsibility costs, constructed from a series of variables in the Medicare claims data files (Appendix Table A2).

We reported all costs in 2015 US dollars. We adjusted costs in Ontario to 2015 Canadian dollars using the Statistics Canada Consumer Price Index for health care for Ontario^27,28 and then expressed them in US dollars using the Purchasing Power Parity for Health in 2015 (1 Canadian dollar = 0.927 US dollar).²⁹ US costs from all years were adjusted using the Consumer Price Index for Medical Care from the Bureau of Labor Statistics.³⁰

Analysis

Data for the Ontario and SEER-Medicare cohorts were analyzed independently. We reported use and costs for all patients in each cohort and by stage at diagnosis (0/I/II, III, and IV) for each of the 12 30-day periods before death, starting from date of death. The last year of life is one of the most common periods in EOL studies,³¹ and examining 30-day intervals allowed us to make adjustments when estimating costs for short-term survivors. All patients survived at least 30 days after diagnosis, so we examined the last 30 days of life for the entire cohort. Patients who survived at least 360 days after diagnosis had 12 30-day periods before death. However, those who survived between 31 and 359 days after diagnosis could have an incomplete 30-day period, or short month, immediately after diagnosis. When reporting use, we made no adjustment for this short month. However, when reporting longitudinal costs, we added the costs and number of observation days in the short month to the numerator and denominator, respectively, of the adjacent full 30-day period. We calculated a daily cost for this entire extended month and multiplied it by 30 to obtain the 30-day cost. Because distributions of characteristics were similar between cohorts and within stage, we did not adjust costs for characteristics or conduct multivariable analyses but described proportions, means, and 95% CIs by cohort and stage.

RESULTS

Patients

The cohort selection is shown in Appendix Fig A1 (online only). We excluded 8,116 SEER-Medicare patients who did not have continuous fee-for-service coverage and 1,597 who did not have Medicare Parts A and B. In the final selection step, 2,142 of 18,707 SEER-Medicare patients (11.4%) and 706 of 7,293 Ontario patients (9.7%) were excluded because they had unknown stage at diagnosis.

The SEER-Medicare (n = 16,565) and Ontario (n = 6,587) patients were similar with respect to most demographic and clinical characteristics. In both cohorts, 47% of patients were diagnosed with stage IV cancer, and > 70% had colon cancer. Patient characteristics varied little by stage at diagnosis (Appendix Table A4, online only).

Health Care Use and Costs in the Last 30 Days of Life

Chemotherapy use (%, 95% CI) in the last 30 days of life was higher among SEER-Medicare patients than among Ontario patients (15.7%; 95% CI, 15.2% to 16.3%; and 8.0%; 95% CI, 7.3% to 8.6%, respectively). In both cohorts, patients diagnosed with stage IV cancer had the highest chemotherapy use in the last 30 days (18.6%; 95% CI, 17.7% to 19.4%; and 11.0%; 95% CI, 9.9% to 12.1%, respectively; Table 1).

TABLE 1.

Resource Use and Costs in the Last 30 Days of Life for Patients Diagnosed With Colorectal Cancer and Dying of Any Cancer, 2007-2013, Ontario and United States SEER-Medicare Cohorts

Use of CT, MRI, and PET was higher in the SEER-Medicare cohort, particularly among patients diagnosed at advanced stages. Approximately 40% of SEER-Medicare stage III and IV patients received a CT, MRI, or PET test in the last 30 days of life, as did 33% (stage III) and 28% (stage IV) of Ontario patients (Table 1).

Hospice enrollment among SEER-Medicare patients was higher in patients with stage IV disease (71.1%) than in those with stage III (66.3%) or stage 0-II (63.6%). In Ontario, 93% of stage IV, 86.3% of stage III, and 77.5% of stage 0-II patients used at least one palliative care service (Table 1).

Higher percentages of Ontario patients than SEER-Medicare patients had ER visits (14.7%; 95% CI, 13.8% to 15.5%; v 6.7%; 95% CI, 6.4% to 7.1%) and inpatient hospitalizations (62.5%; 95% CI, 61.3% to 63.7%; v 51.0%; 95% CI, 50.2% to 51.7%) during the last 30 days of life. Ontario patients had more days in the hospital (mean, 8.8 days; 95% CI, 8.6 to 9.1 days; v 5.6 days; 95% CI, 5.5 to 5.7 days) and hospital deaths than SEER-Medicare patients (42.0%; 95% CI, 40.8% to 43.2%; v 24.3%; 95% CI, 23.6% to 24.9%). However, 43.2% of SEER-Medicare patients who were hospitalized were admitted to the ICU, but only 17.9% of Ontario patients. In both cohorts, use of inpatient hospital resources was lowest in patients with stage IV disease and highest in those with stages 0-II (Table 1).

Hospitalization was the largest cost driver in both cohorts, accounting for > 60% of total costs in the last 30 days. Despite high hospitalization use in Ontario, the mean per-patient cost for inpatient hospitalization was lower in Ontario ($9,434; 95% CI, $9,073 to $9,796) than in SEER-Medicare ($11,180; 95% CI, $10,897 to $11,463). The mean cost per hospital day was almost twice as high in SEER-Medicare ($2,004; 95% CI, $1,995 to $2,012) as in Ontario ($1,067; 95% CI, $1,039 to $1,095). In both cohorts, costs of hospitalization were lower with higher stage, consistent with use patterns (Table 1).

Health Care and Costs in the Last 360 Days of Life

Chemotherapy use was higher among SEER-Medicare patients than Ontario patients for all 30-day periods before death (Fig 1A). The percentage of patients who received CT, MRI, or PET tests increased toward death and was similar in both cohorts until the last month of life, when it decreased in Ontario (Fig 1B). Higher percentages of hospitalized SEER-Medicare patients than Ontario patients were admitted to the ICU in each month; in the 2 months before death, use increased among SEER-Medicare patients and decreased among Ontario patients (Fig 1C).

Fig 1.

Percentages (± 95% CI) of patients who used resources in the last 360 days of life, by cohort. Solid lines with square symbols represent the Ontario cohort, and dashed lines with circle symbols represent the SEER-Medicare cohort. (A) Received chemotherapy. (B) Received CT, MRI, or PET scan. (C) Were in the intensive care unit (% of hospitalized patients).

Mean monthly total costs were consistently higher for SEER-Medicare patients than for Ontario patients, with larger differences for those diagnosed with stage III or stage IV cancer (Fig 2). Patients diagnosed with stage IV cancer (Fig 2C) had higher mean costs than stage 0 to III patients (Figs 2A and 2B) until just before death. Costs increased sharply in the last 3 months before death in both cohorts, for all stages. For the SEER-Medicare cohort, patient responsibility costs ranged from approximately $400 to $1,200 per month and were much higher in patients with stage IV disease until the last month of life, when they were approximately $1,000 for all patients (Fig 2).

Fig 2.

Mean (± 95% CI) 30-day costs in the last 360 days of life for patients by stage at diagnosis, by cohort. Solid lines with square symbols represent the Ontario cohort, dotted lines with circle symbols represent the Medicare-paid costs for the SEER-Medicare cohort, and dashed lines with diamond symbols represent the Medicare-paid costs plus patient responsibility costs for the SEER-Medicare cohort. (A) Patients diagnosed with stage 0-II colorectal cancer (CRC). (B) Patients diagnosed with stage III CRC, by cohort. (C) Patients diagnosed with stage IV CRC. US$, US dollars.

DISCUSSION

This study describes health care use and costs at the EOL in two large similarly defined population-based cohorts of older patients with CRC by stage at diagnosis in the United States (SEER-Medicare) and Ontario, Canada. SEER-Medicare patients received more chemotherapy, ICU care, and CT, MRI, and PET tests and incurred costs that were $1,000 to $3,000 higher in each month before death.

Our findings add to the growing body of research describing health care use and costs among patients in different systems^15-17,32-35 to inform efforts to improve organization and delivery of care. Our study is unique in using cancer registries to select similar cohorts of patients and administrative data to define similar services in two settings. We stratified analyses by stage at diagnosis and found that, for patients diagnosed with stage III and IV disease, mean costs in the last 30 days of life were $3,800 higher in SEER-Medicare patients than in Ontario patients but only $2,000 higher in patients diagnosed with stages 0-II disease. These differences were even greater when SEER-Medicare patient responsibility costs were considered.

The United States spends more money on health care than most developed countries, including Canada.^35-37 Possible reasons include key structural differences between health care systems in the United States and in Canada and many European countries.^1-3,38,39 Higher unit prices for health care services, including clinician salaries,^{3,35,37,40,41} pharmaceuticals, and devices, have been reported in the United States compared with other countries.^3,37,39 The United States performs more surgical procedures and advanced diagnostic tests and has more MRI, CT, and mammography units per capita than Canada and other developed countries.³⁵

Hospitalization was the largest cost driver in both cohorts, accounting for > 60% of total costs in the last 30 days of life. The mean number of hospital days was lower in the SEER-Medicare patients than Ontario, but the mean cost per hospital day was almost twice as high in SEER-Medicare patients ($2,004 v $1,067). If its daily hospital costs could be reduced to a level similar to those in Ontario, the Medicare program could potentially save approximately $265 million in costs of EOL care in the last 30 days of life alone for the approximately 51,020 patients with CRC projected to die of cancer in 2019.⁴² Conversely, if the 3,359 patients with CRC expected to die in Ontario in 2018⁴³ had the same number of hospital days in their last 30 days of life as the SEER-Medicare patients in our study, almost $12 million (United States) in savings could be realized. Identifying opportunities to reduce ICU use in SEER-Medicare patients and hospitalizations in Ontario patients has the potential to improve care delivery and reduce costs.

Unlike Ontario’s publicly funded health system, the US Medicare program does not cover all medically necessary health care costs, and almost one-third of Medicare beneficiaries with cancer spent at least 20% of their income on health care.⁴⁴ We did not have information on indirect costs, including productivity losses, or out-of-pocket costs, such as travel, which can be substantial for patients in both Ontario and the United States.^10,45,46

Forty-two percent of Ontario patients died in the hospital (v 24.3% of SEER-Medicare patients). An earlier study found that approximately 46% of patients with advanced lung cancer who died between 1999 and 2003 in Ontario died in the hospital.¹⁷ In 2005, the Ontario Ministry of Health and Long-Term Care began establishing networks to improve the coordination of EOL care and shift palliative care from the hospital to home,⁴⁷ but evidence for significant reductions of in-hospital deaths is not clear.⁴⁷ Differences in in-hospital deaths between Ontario and SEER-Medicare patients suggest that better coordination is needed in care delivered by physicians, home care workers, and nurses.¹²

Concerns have been raised about the approach of identifying patients who died and retrospectively evaluating health care, because not all patients who died appeared to be dying and hence were not offered care for the dying.⁴⁸ We selected patients on the basis of years of diagnosis and death (2007-2013), which created a bias toward short-term survivors and advanced stage at diagnosis. We studied only Medicare fee-for-service beneficiaries residing in SEER areas and patients residing in one province in Canada. Individuals age ≥ 65 years residing in SEER areas were more likely to be urban residents than those in the entire United States, but otherwise they were demographically similar.⁷ In Canada, all provinces have universal health care coverage with similar insurance plans,⁴⁹ and the per-person availability of health care resources and use of health care in Ontario, and health care costs, are similar to, and certainly no lower than in other provinces.⁵⁰

Administrative data lack information about patient and family experiences, quality of life, symptom control, and preferences. Some services were not available in both cohorts or comparable between cohorts, including complex continuing care facilities, palliative care, hospice, and skilled nursing facilities. Because of differences in reporting and costing methodology in the two health care systems, we could not include radiation therapy, laboratory tests, outpatient prescription drugs, or oral antineoplastic therapies. However, our previous study showed that the service definitions used in this study account for the vast majority of total costs in patients with cancer at the EOL in SEER-Medicare.¹⁶

In conclusion, our descriptive study of health care use and costs at the EOL in similar groups of older patients with CRC, although not supporting a direct comparison of two health systems, generated hypotheses concerning areas for improvement in service delivery and lower costs in both settings. In Ontario, improving coordination of EOL care and reducing hospitalizations and in-hospital deaths could provide savings. Reducing daily hospital costs and intensity of health care services for SEER-Medicare patients, especially those with stage IV disease at diagnosis, could reduce costs to the Medicare program and decrease the financial burden on patients and families.

APPENDIX

Fig A1.

Selection of cohorts. CRC, colorectal cancer; OHIP, Ontario Health Insurance Plan.

TABLE A1.

Data and Variables Used for Cohort Selection

TABLE A2.

Data and Variables Used for Service Use and Costs

TABLE A3.

ICD-9, HCPCS/CPT, and Revenue Center Codes Used to Identify Specified Health Care Services in Medicare Data

TABLE A4.

Characteristics of Patients Diagnosed With Colorectal Cancer Who Died of Any Cancer at Age ≥ 66 Years, 2007-2013, in Ontario and the United States SEER-Medicare Data

AUTHOR CONTRIBUTIONS

Conception and design: Karen E. Bremner, K. Robin Yabroff, Diarmuid Coughlan, Joan L. Warren, Claire de Oliveira, Angela B. Mariotto, Kelvin K.-W. Chan, Murray D. Krahn

Financial support: Murray D. Krahn

Administrative support: Murray D. Krahn

Collection and assembly of data: Karen E. Bremner, Diarmuid Coughlan, Ning Liu, Christopher Zeruto, Joan L. Warren, Claire de Oliveira, Angela B. Mariotto, Clara Lam, Michael J. Barrett, Kelvin K.-W. Chan, Murray D. Krahn

Data analysis and interpretation: Karen E. Bremner, K. Robin Yabroff, Diarmuid Coughlan, Ning Liu, Christopher Zeruto, Joan L. Warren, Claire de Oliveira, Angela B. Mariotto, Clara Lam, Michael J. Barrett, Kelvin K.-W. Chan, Jeffrey S. Hoch, Murray D. Krahn

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

Patterns of Care and Costs for Older Patients With Colorectal Cancer at the End of Life: Descriptive Study of the United States and Canada

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/op/site/ifc/journal-policies.html.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).

No potential conflicts of interest were reported.