This cohort study compares estimates of sepsis incidence, outcomes, and trends based on clinical data from US hospital electronic health record systems vs claims-based ICD-9 data.