OBJECTIVES: Understanding sources of variation in acute care delivery may inform targeted strategies to promote evidence-uptake. We sought to characterize physician-level and ICU-level variation in corticosteroid prescribing for patients with septic shock. DESIGN: We performed a retrospective cohort study using the electronic health record of a multihospital health system. We identified ICU patients with septic shock admitted between 2018 and 2020. Using medication administration data, we determined which patients received corticosteroids within 2 days of vasopressor initiation. We linked each patient to their attending physician of record using digital signatures from clinical documentation. We then fit a hierarchical mixed-effects logistic regression model to identify factors associated with corticosteroid use and quantify variation in corticosteroid administration across physicians and ICUs. SETTING: Twenty-six ICUs across nine hospitals in the United States. PATIENTS: ICU patients with septic shock. MEASUREMENTS AND MAIN RESULTS: Of 5322 patients with vasopressor dependent septic shock, 1294 (24.3%) were treated with corticosteroids within 2 days of vasopressor initiation. We linked these patients to 174 unique attending physicians across 26 ICUs. At the ICU-level, median corticosteroid use was 21.8% (interquartile range [IQR], 18.5-25.7%). At the physician-level, median corticosteroid use was 22.0% (IQR, 11.9-32.7%). In the mixed-effects regression controlling for patient and physician characteristics, 16.5% of the variation in corticosteroid administration was attributable to the ICUs and 10.1% was attributable to the physicians. CONCLUSIONS: Both ICUs and physicians contribute to observed variation in the use of corticosteroids for vasopressor dependent septic shock. These findings underscore the need for multilevel interventions to standardize evidence-based practices in critical care.