PURPOSE: End-of-life care for women with ovarian cancer is persistently aggressive, but factors associated with overuse are not well understood. We evaluated physician-level variation in receipt of aggressive end-of-life care and examined physician-level factors contributing to this variation in the SEER-Medicare data set. METHODS: Medicare beneficiaries with ovarian cancer who died between 2000 and 2016 were included if they were diagnosed after age 66 years, had complete Medicare coverage between diagnosis and death, and had outpatient physician evaluation and management for their ovarian cancer. Using multilevel logistic regression, we examined physician variation in no hospice enrollment, late hospice enrollment (≤ 3 days), > 1 emergency department visit, an intensive care unit stay, terminal hospitalization, > 1 hospitalization, receiving a life-extending or invasive procedure, and chemotherapy (in the last 2 weeks). RESULTS: In this sample of 6,288 women, 51% of women received at least one form of aggressive end-of-life care. Most common were no hospice enrollment (28.9%), an intensive care unit stay (18.6%), and receipt of an invasive procedure (20.7%). For not enrolling in hospice, 9.9% of variation was accounted for by physician clustering ( P < .01). Chemotherapy had the highest physician variation (12.4%), with no meaningful portion of the variation explained by physician specialty, volume, region, or patient characteristics. CONCLUSION: In this study, a meaningful amount of variation in aggressive end-of-life care among women dying of ovarian cancer was at the physician level, suggesting that efforts to improve the quality of this care should include interventions aimed at physician practices and decision making in end-of-life care.