Chemotherapy Use in the Months Before Death and Estimated Costs of Care in the Last Week of Life

Abstract

Context

Considerable attention has been paid to the disproportionately high costs of care for patients nearing death, yet little is known about the costs associated with chemotherapy use among end-stage cancer patients.

Objectives

To compare costs of care other than chemotherapy in the last week of life based on whether cancer patients were using chemotherapy in the months just prior to death.

Methods

A total of 311 patients with advanced cancer who died between 2002–2008 were studied. Data included medical records, patient baseline surveys (median four months before death), and postmortem interviews of caregivers and clinicians. Costs of care were estimated based on reports of death site and services other than chemotherapy received in the week before death (e.g., resuscitation). We tested whether end-of-life care preferences, do-not-resuscitate order completion, or end-of-life discussions accounted for relationships between chemotherapy use and estimated care costs.

Results

Half (50.5%) of patients were receiving chemotherapy at baseline. Estimated end-of-life care costs for patients with baseline chemotherapy use (median=$2681) were significantly higher than for patients without baseline chemotherapy use (median=$1092) (P=0.003). This relationship persisted after adjusting for sociodemographic and clinical characteristics in a generalized linear model (mean incremental cost=$2681, 95% confidence interval $611–$4751, P=0.01). None of the psychosocial variables accounted for the relationship between chemotherapy use and estimated care costs.

Conclusion

Chemotherapy for end-stage cancer patients is associated with higher estimated end-of-life care costs. Given evidence of limited benefit and potential harm of chemotherapy for end-stage cancer patients, the cost-effectiveness of such care is questioned and further study warranted.

Introduction

There is growing attention to the financial implications of treatment choices for patients with advanced cancer and other serious illnesses.^1–5 Of the 5% of Medicare beneficiaries responsible for 30% of Medicare costs, one-quarter are patients with cancer.⁶ In addition, costs of cancer care are expected to rise at least 26% between 2010 and 2020,^1,7 raising questions about the value obtained from high medical expenditures for the nation's health. Patients with advanced cancer are at high risk of medical debt and bankruptcy,^8–9 which are associated with worse quality of life.^10–11 Understanding the cost implications of decisions to begin treatments that may be of limited clinical benefit would inform United States health policy; communicating these cost implications is paramount to the ability of patients, families, and physicians to make informed treatment decisions.

One such treatment decision that many patients with advanced cancer and their family members and clinicians ultimately confront is whether to use additional lines of chemotherapy when disease progresses after an initial or multiple lines of chemotherapy. Chemotherapy for patients with incurable cancer is provided for symptom relief and/or life prolongation, although research suggests it may not enhance survival or improve quality of life.¹² Receipt of additional chemotherapy among metastatic cancer patients with chemo-refractory disease is associated with later referrals to hospice care and greater likelihood of aggressive life-prolonging care at the end of life (EOL).¹³ The American Society of Clinical Oncology (ASCO) recommends against the use of chemotherapy for patients with poor performance status who have not responded to earlier lines of treatment and who are not eligible for a clinical trial.⁴ Among patients with good baseline performance status who have not responded to earlier lines of treatment, however, chemotherapy is associated with worse quality of life in the week before death.¹²

Patients with advanced cancer may elect to receive additional lines of chemotherapy because of their own preferences,¹⁴ including hope for symptom relief or additional survival time. Others may misunderstand the goal of chemotherapy offered to them as curative when that is not its intent.¹⁵ In many cases, however, treatment decisions for seriously ill patients are primarily a reflection of physician or facility-level practice patterns.^16–18 Ensuring that patients with advanced cancer and their providers are well-informed about the risks (including financial ones) and benefits associated with treatments is essential to high quality care.¹⁹ Moreover, from a health systems perspective, reducing use of expensive care with questionable benefit leaves more resources for high-value care for patients with advanced cancer.¹

We used data from the prospective Coping with Cancer Study, which followed patients with advanced cancer a median of four months until their death, to examine the relationship between chemotherapy use at the time of study entry and estimated costs of care received in the last week of life. In addition, we sought to determine whether this relationship was affected by other factors associated with care received at the EOL and its estimated costs, including preferences for EOL care, do-not-resuscitate (DNR) order completion, and EOL discusssions.^20–22

Methods

Sample

We studied 311 patients with advanced cancer from the Coping with Cancer cohort who died during the study period and had information on both baseline chemotherapy use and health care used in the week before death. The Coping with Cancer study followed patients with advanced cancer and their caregivers from a baseline interview until death (a median of four months) and explored relationships among EOL care preferences, beliefs, treatment decisions, patient quality of life, and caregiver bereavement adjustment. Patients were recruited and enrolled from seven outpatient clinics between January 1, 2002 and February 28, 2008; patients from one site with missing assessments were excluded.¹²

Patients with distant metastases and disease refractory to at least one line of chemotherapy, who spoke English or Spanish, and who were physically well enough to participate in an interview were eligible for this study. Patients with dementia (Short Portable Mental Status Questionnaire score ≥ 6) or delirium were excluded from the study.²³ Patients identified their primary unpaid caregivers (family member or friend); caregivers also were interviewed in this study. Of the 661 patients in the six clinics who participated in the study, 384 died during the study period (2002–2008). Baseline chemotherapy use was not significantly associated with death during the study period.¹² Consistent with an earlier analysis of quality of life and chemotherapy, we excluded patients on a clinical trial at the baseline interview (n= 33), and others missing data on functional status, quality of life in the week before death, and/or baseline chemotherapy use (n= 39).¹² One additional patient with incomplete information on care provided in the week before death was excluded from the sample, for a total sample size of 311. Written informed consent was obtained from both patients and caregivers, and the study was approved by the institutional review board of each site.

Data Sources

Information on chemotherapy use came from the patient's medical record. Patient sociodemographic characteristics, preferences for care, DNR order completion, and EOL discussions were reported by the patients in baseline interviews (median of four months before death). Information on health services used in the week before death was obtained from patients' medical records and through post-mortem interviews with the caregiver (n=134 [43%]) or clinician (n = 177 [57%]) most familiar with the circumstances surrounding the patient's death. Estimated costs were derived in part from the 2008 Health Care Utilization Project National Inpatient Sample (HCUP SID) database.²⁴

Variables

Baseline Chemotherapy Use

This was a dichotomous variable coded as yes (1) if the patient's medical record indicated use of chemotherapy at the time of study enrollment (0=no).

Estimated Costs of Care in the Last Week of Life

Following methods we have used in previous studies,^21–22,25 we estimated costs of care in the last week of life based on the location of the patient's death and services provided to the patient in the last week of life. From the post-mortem interviews, we identified the patient's location of death (home, hospital outside of intensive care unit [ICU], ICU, nursing home, or inpatient hospice). We also identified services provided to the patient in the week before death, including ventilation, cardiopulmonary resuscitation (CPR), and hospice. Using mean hospital costs for decedents in the HCUP NIS with a cancer diagnosis anywhere in the record (excluding maternal and neonatal discharges), Medicare and Medicaid reimbursement rates for nursing home and hospice care,^26–27 and other published estimates of costs of care near the EOL,^28–31 we estimated the costs associated with health care other than chemotherapy provided to patients in the last week of life (see the methodological Appendix for further details; available at jpsmjournal.com). All costs were adjusted to 2014 United States dollars. ³²

Preferences for Care

At the baseline interview, patients answered questions about preferences for life-prolonging care (1 = prefers to “extend life as much as possible”; 0=prefers to “relieve pain or discomfort as much as possible”), heroic measures: “…everything possible to keep you alive even if you were going to die in a few days anyway?” (1=yes, 0=no), and death outside of an ICU (1=yes, 0=no).

DNR Order Completion

Patients indicated whether they had completed a DNR order at baseline (1= yes, 0 = no).

EOL Discussion

Patients were asked “Have you and your doctor discussed any particular wishes you have about the care you would want to receive if you were dying?” (1=yes, 0=no).

Sociodemographic and Clinical Characteristics

Our final models included patient age, gender, possession of health insurance (1 = yes, 0 = no), self-reported race/ethnicity (White, Black, non-White Hispanic, or other), Charlson Comorbidity score,³³ cancer site (lung, colon, pancreatic, other gastrointestinal, breast, or other), Eastern Cooperative Oncology Group (ECOG) performance status,³⁴ and study recruitment site.

Statistical Analyses

We used generalized linear models (GLMs) with a gamma distribution and a log link function to model relationships among baseline chemotherapy use and estimated costs of care in the last week of life. A Pregibon link test³⁵ verified our choice of link function. In order to understand whether preferences, DNR order completion, and EOL discussions account for relationships among chemotherapy use and estimated costs of care, we first ran unadjusted logistic regression models of each variable on baseline chemotherapy use, and GLMs with estimated costs of care regressed on each variable individually. Variables that were associated with both baseline chemotherapy use and estimated costs of care were included in a GLM modeling costs and chemotherapy use. We calculated mean incremental effects of chemotherapy on estimated costs (mean change in estimated costs of care in the last week of life when switching the chemotherapy indicator from 0 to 1 for each patient, holding all other variables constant at their original values). In addition, we controlled for sociodemographic and clinical characteristics as well as fixed effects for recruitment site.

Results

Our sample was diverse in age (mean = 59 years, range 22–93) and race (61% White, 21% Black, and 17% non-White Hispanic) (Table 1). Of the 311 patients in our sample, 50.5% (n=157) were receiving chemotherapy at baseline. In previous analyses of this sample, patients receiving chemotherapy at baseline had more years of education, fewer comorbidities, better performance status, were significantly younger, and were more likely to have breast cancer and less likely to have gastrointestinal cancers other than pancreatic cancer than patients not receiving chemotherapy.¹² Baseline chemotherapy was not significantly associated with time until death in previous analyses.¹³ Estimated costs of care in the last week of life among patients receiving baseline chemotherapy were significantly higher than those of patients not receiving baseline chemotherapy (with baseline chemotherapy: median = $2681, interquartile range [IQR] = $1092–$13,584; without baseline chemotherapy: median =$1092, IQR = $1092–$4858; Wilcoxon rank-sum test z=−2.965, P=0.003) (Fig. 1).

Table 1.

Sociodemographic and Clinical Characteristics of Sample (n=311)

Variable	N (%) or Mean(SD)
Age, yrs	59 (13)
Race/Ethnicity
White	191 (61%)
Black	64 (21%)
Non-White Hispanic	52 (17%)
Other	4 (1%)
Male	170 (55%)
Insureda	164 (54%)
Charlson Comorbidity Scoreb	8.6 (2.6)
Cancer site
Lung	72 (23%)
Colon	40 (13%)
Pancreatic	23 (7%)
Other gastrointestinal	38 (12%)
Breast	42 (13%)
Other	97 (31%)
ECOG performance status
0 = Asymptomatic	9 (3%)
1 = Symptomatic, ambulatory	122 (39%)
2 = Symptomatic, bedridden <50% of time	115 (37%)
3 = Symptomatic, bedridden > 50% of time	58 (19%)
4= 100% bedridden	7 (2%)

ECOG = Eastern Cooperative Oncology Group; SD = standard deviation

^aMissing insurance data for 9 patients.

^bMissing Charlson data for one patient.

Fig. 1.

Boxplot of estimated costs in 2014 U.S. dollars in last week of life by baseline chemotherapy status. Boxes represent interquartile ranges (IQR) of estimated costs; whiskers extend to 1.5 IQR beyond the upper and lower quartiles; dots represent patients with outlier values for estimated costs.

In unadjusted regression models of preferences, DNR order completion, and EOL discussions on baseline chemotherapy use, baseline chemotherapy use was significantly negatively associated with preferences for death outside an ICU and an EOL discussion with the physician. In unadjusted GLMs of estimated costs of care on these same variables, both preference for comfort care over life-prolonging aggressive care and EOL discussion with the physician reported at the baseline interview were significantly associated with lower estimated costs of care in the last week of life.

Because EOL discussion was the only variable significantly associated with both baseline chemotherapy use and estimated costs of care, we included it in a GLM of estimated costs of care as a function of baseline chemotherapy use. Baseline chemotherapy use was significantly associated with higher estimated costs of care in models with and without EOL discussions included as a covariate. In an unadjusted model, the mean incremental effect of baseline chemotherapy use on estimated costs of care was $2740 (95% confidence interval [CI] $1229 – $4251, P<0.001; Table 2). When EOL discussions were added to the model, the mean incremental effect of baseline chemotherapy remained largely unchanged ($2496, 95% CI = $967–$4026, P=0.001). The mean incremental effect remained significant and of similar magnitude after adjusting for patient sociodemographic and clinical characteristics and recruitment site (Table 2).

Table 2.

Mean Incremental Effects of Baseline Chemotherapy on Estimated Costs of Care in the Last Week of Life

	Model 1 (Baseline chemotherapy only; unadjusteda)		Model 2 (Baseline chemotherapy and EOL discussion; unadjusteda)		Model 3 (Baseline chemotherapy, EOL discussions; adjusted for sociodemographic and clinical characteristicsb)
Variable	MIE (95% CI)	P-value	MIE (95% CI)	P-value	MIE (95% CI)	P-value
Baseline chemotherapy use	$2740 ($1229–$4251)	<0.001	$2496 ($967–$4026)	0.001	$2681 ($611–$4751)	0.01

CI = confidence interval; MIE = mean incremental effect

^aModels run on 305 patients with data on EOL discussions.

^bCharacteristics included age, race, gender, health insurance possession, Charlson Comorbidity Index, cancer site, ECOG performance status, and study recruitment site. N=297 for Model 3 because of missing data on insurance possession and Charlson Comorbidity Index; results for Models 1 and 2 are substantively similar when run on the smaller subset used for Model 3.

Discussion

Chemotherapy use was significantly associated with higher estimated costs of care in the last week of life in our sample of patients with advanced cancer. Previous studies have found that chemotherapy for patients with late-stage cancer has low response rates and is associated with more aggressive life-prolonging care and less hospice care with no evidence of improved quality of life in the week before death.^12–13,36 This study adds to these findings by showing that not only is chemotherapy for patients with advanced metastatic cancer associated with worse end-of-life outcomes, it may be associated with increased costs of EOL care.

We did not find evidence that the relationship between chemotherapy use and higher estimated costs of care at the EOL was accounted for by patients' EOL preferences, DNR order completion, or discussions of preferences. Surprisingly, use of baseline chemotherapy was not significantly associated with preferences for aggressive life-prolonging care, even though baseline chemotherapy is predictive of life-prolonging care use near the end of life.¹³ Decisions about treatments should optimally be made with careful consideration of a patient's informed preferences and values.³⁷ However, we found that baseline chemotherapy use was strongly associated with higher estimated costs of care at the end of life even after accounting for reported discussions of EOL preferences with physicians. When it is unclear that the benefits of a treatment outweigh its disadvantages, many treatment decisions among seriously ill patients are made by health care providers who may pursue a treatment rather than risk “abandoning” their patients.^14,38 Future analyses should explore whether and the extent to which the heightened estimated costs we observed may be attributed to care to treat the adverse effects of chemotherapy³⁹ and/or reflect physician practice patterns and communication styles.

Patients often want an honest assessment of the benefits and risks of a treatment, as well as information on the cost implications of a treatment.⁴⁰ ASCO guidelines ask practitioners to reduce inappropriate chemotherapy among patients with end-stage cancer and poor performance status, and the benefits of chemotherapy for patients with better performance status are also questionable.^4,12 Our findings add to the list of implications patients may wish to consider before making a treatment decision. They are consistent with the motivation for new ASCO models that strive to determine value in end-stage cancer care and clinicians' calls for transparent information about cost implications of treatment decisions.^1,10

Limitations

Our study is limited in that we use estimated costs based on average hospital expenditures, Medicare payment rates, and published estimates in the literature rather than actual costs. Because actual health care costs for patients vary widely¹⁷ and because some of our cost estimates are derived from averages over a hospital stay or longer time period, these results need to be replicated in a dataset that includes more recent data on actual costs incurred by both the health care system and by patients. That said, we used a more conservative basis for our cost estimates (basing them on median hospital length of stay rather than mean length of stay) than we have used in previous estimates.^21,25 In addition, our cost estimate does not account for chemotherapy or other anti-cancer drug costs and thus may be conservative because we were interested in additional costs associated with chemotherapy use and not chemotherapy costs, per se.⁴¹ By using average cost estimates, differences between patients were minimized, which likely leads to conservative estimates. Our study is also limited in that recruitment for this study ended in 2008, and great strides in palliative care for patients with advanced cancer have been made since then. However, rates of aggressive life-prolonging care measures, including mechanical ventilation and ICU use, continue to increase among patients with cancer.⁴² Finally, future studies should take into account physician practice patterns, changes in patient EOL preferences and advance care plans over time, and dose and type of chemotherapy used, variables that are not available in this dataset.

Conclusions

Chemotherapy use in end-stage cancer patients is associated with higher estimated costs of care in the last week of life. Given evidence of limited benefit and potential harm of chemotherapy for end-stage cancer patients, the cost-effectiveness of such care is questioned and further study warranted. In addition, practice guidelines for chemotherapy use for patients with advanced cancer should consider health care costs as well as patient quality of life at the EOL.