An Early Warning System (EWS) was developed in our electronic health record (EHR) to monitor in real time laboratory and vital sign (VS) data and notify providers about non-ICU inpatients with sepsis at risk for developing clinical deterioration. If a patient had at least 4 predefined VS and lab abnormalities an alert was sent to the provider, nurse and rapid response coordinator. Compared to a prospective control period the intervention resulted in a significant increase in antibiotics and IV fluids and a non-significant decrease in mortality. Introduction and background: Early recognition and timely intervention significantly reduce sepsis-related mortality. The recent introduction of VS and provider data in our EHR created the opportunity to detect sepsis early and notify providers with the goal of reducing our higher than expected sepsis mortality. Methods: The EWS was designed to monitor real time laboratory and VS data and trigger if patients fulfilled at least 4 criteria at any one time (Temp. 38°C, HR > 90, RR >20 or PCO2 12000 or >10% bands, Lactate >2.2, SBP <100). Notifications were sent to the covering physician, nurse and rapid response coordinator and directed them to meet at the bedside within 30 minutes to assess the patient. The EWS was initially activated for a “silent period” (6/6 to 9/4/2012) to validate the predictive model. 595 patients out of 15,570 admissions triggered the alert, but no notifications were sent during this period. In the intervention period (9/12 to 12/11/2012) 731 patients out of 16,103 admissions triggered the alert resulting in notifications to the care team. Process measures included new antibiotic and fluid bolus orders within 3 hours of the trigger. Outcome measures included ICU transfer rates within 6 hours and in-hospital mortality. Chi square tests compared proportions. Logistic regression models adjusted for age, gender, Charlson index on admission and admitting service. Results: In unadjusted analysis antibiotic and IV fluid boluses increased significantly. The ICU transfer rate increased but not significantly. Mortality decreased and reached statistical significance at one of our 3 hospitals. Results remained significant after adjustment except for mortality reduction at PMC (OR 0.40, 95% CI 0.13-1.23). Discussion: By leveraging readily available electronic data, a predictive model identified at risk patients and auto generated notifications to care teams resulting in more timely sepsis care and reduced mortality. This alert is scalable to other health systems.