The Quality of End-of-Life Care in Medicare versus the VA

Tuesday, June 12, 2018
Lullwater Ballroom - Garden Level (Emory Conference Center Hotel)

Presenter: Todd Wagner

Co-Authors: Risha Gidwani-Marszowski; Jack Needleman; Vincent Mor; Katherine Faricy-Anderson; Derek Boothroyd; Gary Hsin; Karl Lorenz; Manali Patel; Vilija Joyce; Samantha Murrell; Kavitha Ramchandran; Steven Asch


Concerns over veteran access to care have prompted Congressional recommendations that the Veterans Health Administration (VA) shift to an increasingly Medicare-like role in purchasing care in the community for veterans. The question then arises whether a purchasing system with fee-for-service incentives (Medicare) can provide similar or better quality than one in which care is directly provided by salaried clinicians (VA).

We studied quality differences between Medicare and VA by evaluating end-of-life (EOL) care for cancer patients. EOL care in the U.S. represents a triple challenge: it is low quality; has high per-person costs; and affects much of the population. Cancer specialty societies recommend a reduction in intensive care at EOL. We evaluated the quality of EOL cancer care provided by Fee-for-Service (FFS) Medicare and VA, using well-accepted quality-of-care metrics. More than 90% of older Veterans are enrolled in Medicare; this dually-eligible population thus allows for investigation of quality differences that are healthcare system-based.


We conducted a retrospective analysis of FFS Medicare and VA administrative data (FY 2010-2014). Our primary goal was to evaluate quality of EOL care in the last 30 days of life for veterans dying of cancer.

We evaluated quality of care as the proportion of patients who experienced: 2+ Emergency Department (ED) visits; chemotherapy; ICU stay; hospital admission; death in hospital; and number of days spent in hospital. Poor-quality care is indicated by higher proportions of patients receiving these services. Patients were allocated to Medicare or VA based on the system in which they received the most overall healthcare.

Outcomes were evaluated with hierarchical regression models using geographic area as a fixed effect. We adjusted for variables previously shown to influence veteran reliance on Medicare versus VA. We also adjusted for comorbidity burden and included year fixed effects. Missing data were multiply imputed and analyses adjusted for multiple hypothesis testing.

Our secondary goal was to investigate whether EOL care quality was influenced by palliative care. We regressed patient-level outcomes of EOL care quality on facility-level palliative care penetration quintile, using VA data only.


Our cohort consisted of 87,251 cancer decedents. Medicare-reliant patients received more-intensive (lower-quality) EOL care relative to VA-reliant patients with respect to: chemotherapy (aOR: 1.73; p <0.001); hospital admission (aOR: 1.12; p <0.001); ICU admission (aOR: 1.43; p <0.001), number of days spent in hospital (aIRR): 1.11; p <0.001); and death in hospital (aOR: 1.19; p <0.001). Conversely, Medicare-reliant patients were less likely to have 2+ ED visits in the last 30 days of life (aOR: 0.77; p <0.001). Results from models evaluating the effect of facility-level palliative care penetration on EOL care found no association.


After adjusting for variables shown to influence selection into VA versus Medicare, we found veterans treated under FFS Medicare were more likely to get unduly intensive healthcare at EOL compared to those treated by VA. Although purchasing care in the community may increase access to care for some veterans, our work indicates it may do so at the cost of more intensive, potentially burdensome care.