Hospital-Based End-of-Life Care and Costs for Older Patients With Malignant Brain Tumors

Abstract

This study analyzes Medicare claims data associated with end-of-life hospitalization care and costs for elderly patients with a malignant brain tumor.

Approximately half of patients with a diagnosis of primary malignant brain tumor (PMBT) or secondary malignant brain tumor (SMBT) are older than 65 years and experience disproportionate mortality and symptom burden. End-of-life care for patients with terminal cancer is often aggressive, costly, and discordant with patient preferences. However, a lack of knowledge remains about patterns of end-of-life care for the growing population of elderly people with a malignant brain tumor. This study compares hospital-based care and costs in the last 30 days of life for older patients with PMBT and SMBT, identifies potential risk factors for aggressive care, and evaluates the association between aggressive care and cost.

Methods

Medicare claims data, derived from a lay navigation program in the southeastern United States, were used to identify decedents from January 1, 2012, to December 31, 2015, who were 65 years or older with either PMBT (International Classification of Diseases, Ninth Revision [ICD-9] code 191.X) or SMBT (ICD-9 code 198.3). Those with claims for both codes were excluded. Total costs to Medicare and hospital-based care (emergency department visits, intensive care unit admissions, and hospital admissions) in the 30 days prior to death, as described in Medicare core quality measures, were determined for each patient and compared using the Mann-Whitney test for continuous variables (cost) and χ² test for categorical variables (all others). A 2-sided P = .05 indicated statistical significance.

Regression analyses of risk for hospital-based care and costs were performed, with adjustment for sociodemographic factors (including self-reported race/ethnicity), Charlson comorbidity index score (excluding cancer), and receipt of chemotherapy, radiotherapy, and hospice care. Risks for hospital-based care were calculated by generalized log-linear models using Poisson distribution with robust variance estimates. Linear mixed-effect models accounting for random effects were constructed to assess costs. Analyses were performed from August 31, 2016, to February 16, 2017, using SAS software, version 9.4 (SAS Institute Inc). The University of Alabama at Birmingham institutional review board approved the study and the waiver of patient informed consent.

Results

Of the 12 725 decedents, 1323 (10.4%) had either PMBT (n = 383) or SMBT (n = 940). Sociodemographic characteristics were similar between groups. In the last 30 days of life, patients with SMBT were more likely than those with PMBT to have an emergency department visit (470 [50%] vs 152 [40%]; P < .001) or hospitalization (472 [50%] vs 162 [42%]; P = .009), but there was no difference in intensive care unit admissions (Table 1). Total costs to Medicare were similar for PMBT and SMBT ($8592 vs $9964).

Table 1. Hospital-Based Care Utilization and Total Costs in the Last 30 Days of Life for Decedents With Primary or Secondary Malignant Brain Tumor.

Variable	PMBT (n = 383)	SMBT (n = 940)	P Valuea
Hospital-based care, No. (%)
Any	186 (48.6)	539 (57.3)	.004
Emergency department visit	152 (39.7)	470 (50.0)	<.001
Hospital admission	162 (42.3)	472 (50.2)	.009
Intensive care unit admission	66 (17.2)	143 (15.2)	.36
Total Medicare costs, median (IQR), US$	8592 (5406-19 214)	9964 (5593-17 714)	.75

Abbreviations: IQR, interquartile range; PMBT, primary malignant brain tumor; SMBT, secondary malignant brain tumor.

^aDetermined with the χ² test of independence or Mann-Whitney test.

Among those with PMBT, men (relative risk [RR] = 1.28; 95% CI, 1.03-1.60) and those with a Charlson score of 1 or higher (RR = 1.52; 95% CI, 1.12-2.06) had increased risk for hospital-based care (Table 2). Among those with SMBT, increasing age (RR = 0.92; 95% CI, 0.86-0.99) and Charlson score of 1 or higher (RR = 1.81; 95% CI, 1.47-2.22) were associated with hospital-based care. Hospital-based care was associated with increased costs for both PMBT ($16 303 increase; P < .001) and SMBT ($13 132 increase; P < .001) (Table 2).

Table 2. Risk Factors for Hospital-Based Care Utilization and Costs in the Last 30 Days of Life for Decedents With Primary or Secondary Malignant Brain Tumor.

Variablea	PMBT (n = 355)b		SMBT (n = 859)b
	RR (95% CI)	P Valuec	RR (95% CI)	P Valuec
Hospital-based care
Age	0.97 (0.87 to 1.09)	.65	0.92 (0.86 to 0.99)	.02
Male	1.28 (1.03 to 1.60)	.03	1.04 (0.93 to 1.16)	.53
Nonwhite	1.03 (0.75 to 1.42)	.86	1.02 (0.88 to 1.19)	.75
Charlson comorbidity index score ≥1d	1.52 (1.12 to 2.06)	.007	1.81 (1.47 to 2.22)	<.001
Variablea	Estimate (95% CI), US$	P Valuec	Estimate (95% CI), US$	P Valuec
Total Medicare costs
Age	−187 (−501 to 126)	.24	17 (−115 to 149)	.80
Male	−2788 (−7077 to 1502)	.20	−1046 (−2819 to 526)	.18
Nonwhite	1696 (−4760 to 8151)	.61	1083 (−1264 to 3430)	.37
Charlson comorbidity index score ≥1d	2233 (−2714 to 7179)	.38	2281 (191 to 4370)	.03
Any hospital-based care	16 303 (11 688 to 20 918)	<.001	13 132 (11 323 to 14 940)	<.001

Abbreviations: PMBT, primary malignant brain tumor; RR, relative risk; SMBT, secondary malignant brain tumor.

^aCollected from Medicare claims data and lay navigation program data.

^bOf this total, 28 patients (7.8%) with PMBT and 81 patients (9.4%) with SMBT had at least 1 missing data point.

^cDetermined with generalized log-linear models predicting risk for hospital-based care and linear mixed-effect models predicting total Medicare costs.

^dThis score excludes cancer diagnosis.

Discussion

Hospital-based care was used by 725 of 1323 older patients (54.8%) with malignant brain tumors in the final days of life, which was associated with a $13 000 to $16 000 increase in cost. Patients with SMBT were more likely than those with PMBT to receive hospital-based care, as were men and younger patients. It is possible that less certain disease trajectories in patients with SMBT led to the choice of aggressive care at a juncture not perceived to be the end of life. Furthermore, certain sociodemographic characteristics, such as being male, are consistent risk factors for aggressive end-of-life care, which may represent different preferences as opposed to inequities in care. Comparable rates of intensive care unit admission for patients with PMBT and SMBT are likely attributable to lower relative incidence than in other hospital-based care. High overall rates of hospital-based care and costs may also be attributable to caretakers being unprepared or unequipped to manage common end-of-life symptoms, particularly neurological decline. These findings underscore the need for interventions that facilitate earlier communication about common illness trajectories and care preferences in the brain tumor population—specifically, discussions about how and where the patient would like to manage a foreseeable neurologic decline.