This study sought to characterize variations in severe sepsis mortality between hospitals in the United States. Hospital discharge data (2012) were used from the University HealthSystem Consortium (UHC), a cooperative of US not-for-profit academic medical centers and affiliated hospitals. Discharge diagnosis codes were used to define severe sepsis as the presence of a serious infection with at least 1 organ dysfunction on hospital presentation. Expected mortality was determined from UHC risk adjustment mortality models. Among the 188 hospitals in the analysis, there were 256 509 patients with severe sepsis on admission. The median number of severe sepsis cases per hospital was 1202 (interquartile range [IQR] = 718-1940). Severe sepsis observed mortality (median = 8.6%; IQR = 6.8%-10.3%; range = 0.9%-18.2%) and observed-to-expected (O:E) mortality ratios (median = 0.91; IQR = 0.77-1.05; range = 0.16-1.95) varied across the hospitals. Variations in institutional severe sepsis observed mortality rates and O:E mortality ratios were observed in this national consortium of major medical centers.