Background: Many patients with metastatic cancer receive high-cost, low-value care near the end of life. We examined interventions during terminal hospitalizations for patients with metastatic cancer to identify those with high likelihood of receiving futile care. Methods: A retrospective population-based cohort analysis of encounter-level data from the National Inpatient Sample was conducted, including records from 2010-2017 for patients ages ≥18 with metastatic cancer who died during hospitalization. We fit multivariable binomial logistic regression models to examine associations between exposures, including patient demographics, and the main outcome of aggressive, low-value, and high-cost medical care (Table). Results: Out of 321,898 hospitalizations among patients with metastatic cancer, 21,335 (6.6%) were terminal. Of these, 65.9% were white, 14.1% Black, 7.5% Hispanic, 58.2% were insured by Medicare or Medicaid, and 33.2% were privately insured. Overall, 63.2% were admitted from the Emergency Department (ED), 4.6% received systemic therapy, and 19.2% received invasive ventilation. Median total charges were $43,681. Black patients and publicly insured patients had higher likelihoods of admission from the ED and receiving ventilation, as well as higher total charges; similar trends emerged among patients of Asian race and Hispanic ethnicity. Patients hospitalized at urban teaching hospitals had higher likelihoods of receiving systemic therapy, ventilation, and incurring higher total charges (Table). Conclusions: Metastatic cancer patients of racial and ethnic minority groups and those with Medicare or Medicaid were more likely to receive low-value, aggressive interventions at the end of life. Further studies are needed to determine the underlying causes of these disparities in order to implement prospective interventions and advance appropriate end-of-life care.

^aWNH = White, non-Hispanic; BNH = Black, non-Hispanic; API = Asian or Pacific Islander; public = Medicare or Medicaid; UT= urban teaching hospital ^bOdds ratio (95% confidence interval); two-tailed p ^cDichotomized according to whether median total charges were greater than median for overall cohort.