Utilization of Deterioration Index Model to Improve Sepsis Management in Medical Unit

Abstract

Problem and Purpose: One third of inpatient deaths are related to sepsis. In a medical unit, a sample of 245 sepsis patients in 2020, 28.4 % failed timely sepsis treatment and 35.4% lacked blood culture within 3 hours sepsis recognition. The implementation site was a 35-bed adult medical unit where the utilization of the existing sepsis best practice alert (BPA) in the EHR was substandard. From November 2020 to February 2021, unit nurses bypassed or ignored 84% of sepsis BPA. This DNP project sought to improve recognition of early sepsis complications and management in a medical unit by implementing a new EPIC BPA Deterioration Index with Sepsis Score (DISS) and sepsis bundle. Methods: Based on Press Ganey Sepsis Workflow, nursing staff in the unit were provided education and training on sepsis, sepsis bundle, and early identification and management of sepsis. Nurses were also provided training on DISS, which was incorporated into the hospital EPIC system. Measurements included DI BPA encounters, length of stay (LOS), and Blood culture draw compliance. The intervention site was a 35-bed adult medical unit. Results: Blood Cultures in <3 hours improved an average of 30% for the intervention period, and the average monthly length of stay trended down during the intervention. The number of alerts decreased significantly during the intervention period. Conclusion: Machine learning prediction models such as DISS with sepsis bundle can be utilized to identify early sepsis complications and improve timely blood culture compliance, and may lower LOS for patients with sepsis.