Background: Sepsis is a leading cause of mortality among hospitalized patients. Early detection and intervention reduces sepsis-related mortality. We implemented a novel early warning system (EWS 2.0) based on a machine-learning algorithm to prospectively identify patients with increased risk of severe sepsis or septic shock. Validation suggested excellent predictive characteristics, including a positive likelihood ratio of 13. Stakeholder acceptance of such a system is crucial for process and outcome improvement.
Methods: We developed surveys to assess clinicians’ perceptions of the EWS 2.0 alert’s helpfulness and impact on care. During a 6-week study period conducted 5 months after EWS 2.0 alert implementation, clinicians, including nurses (RNs) and providers (physicians, advanced practice providers), were surveyed twice for each alert. The first 16-item survey was completed within 6 hours of the alert, and the second 11-item survey was completed 48 hours post-alert.
Results: For the 247 alerts that triggered, 132 nurses (61%) and 83 providers (39%) completed the first survey. Of these, 26 nurses and 30 providers completed a second survey at 48 hours. Few (20% RNs, 10% providers) identified a new clinical finding after responding to the alert. Perceptions of sepsis at the time of alert were discrepant between nurses (38%) and providers (67%). The majority of both groups felt the alert did not change their perception of the patient’s risk for sepsis (58% RNs, 64% providers). Few RNs (8%) and less than half of providers (40%) felt the patient had sepsis at 48 hours. Few providers (6%) but a quarter of RNs (28%) reported the alert changed management. Almost half of nurses (40%) but less than a fifth of providers (13%) found the alert helpful at 6 hours, though the proportion of providers finding the alert helpful more than doubled by 48 hours (33%). Reported helpful features included improved team communication and more frequent monitoring. Those who found the alert unhelpful cited triggering for irrelevant clinical findings and known abnormalities. Nurses were more likely to describe the alert as improving care at both 6 hours (38% vs. 10% providers) and 48 hours (38% vs. 10% providers).
Conclusions: In general, clinical perceptions of the EWS were poor. Improved provider perception at 48 hours may suggest increasing provider agreement with the tool’s initial assessment of sepsis risk as the patient’s clinical course evolves. Nurses and providers differed in their perceptions of sepsis and EWS alert benefits. Although both groups felt the alert frequently did not change management, the value of the alert may confer differential benefits to each group and foster interdisciplinary communication. Consideration of both clinician perceptions and patient outcomes will be critical to improving the system to best meet clinicians’ needs.
To cite this abstract:Ginestra, JC; Schweickert, WD; Meadows, L; Lynch, M; Pavan, K; Chivers, C; Draugelis, M; Donnelly, P; Fuchs, BD; Umscheid, CA . CLINICIAN PERCEPTION OF THE EFFECTIVENESS OF AN AUTOMATED EARLY WARNING SYSTEM TO PREDICT SEPSIS. Abstract published at Hospital Medicine 2017, May 1-4, 2017; Las Vegas, Nev. Abstract 232. Journal of Hospital Medicine. 2017; 12 (suppl 2). https://www.shmabstracts.com/abstract/clinician-perception-of-the-effectiveness-of-an-automated-early-warning-system-to-predict-sepsis/. Accessed January 24, 2020.