Purpose: Nationally, children’s hospitals have mobilized to implement clinical best practices and leverage health information technology to combat what the World Health Organization recognizes as a global threat to children – pediatric sepsis. Researchers estimate 1.2 million cases of children develop sepsis annually around the world. In the United States, sepsis accounts for approximately 75,000 pediatric inpatient admissions a year, with a mortality rate ranging from 5% to 20%. If not promptly identified and treated, pediatric sepsis remains the leading cause of mortality in children, commanding the need for high-reliability, quality improvement work processes.
The project aim was to leverage the electronic health record (EHR) clinical decision support (CDS) tools and best practice models to support early identification of pediatric sepsis, ensure standardization and compliance with treatment guidelines, support practice, improve care delivery, and optimize patient outcomes. This project evaluates the scalable, phased implementation approach to optimizing an electronic infrastructure utilizing data analytics, retrospective case reviews, and interdisciplinary collaboration. The institution examines alert statistical performance analysis to assess changes and improve design. Post-implementation time-to-intervention metrics and patient outcomes are reviewed.
Description: Phase 1 of the project involved modifying the alert/screening criteria (vital sign parameters aligned with evidence-based practice (EBP) and state recommendations), and alert behavior (action-oriented display information, impacted users, and alert lockout behavior) based on data analysis. In phase 2, the project team applied CDS tools to facilitate workflow and practice compliance. Retrospective case reviews and EBP drove changes, resulting in improved case sensitivity. Pre-implementation alert analysis using statistical performance provided predictive screening behavior. Implementation of the sepsis bedside huddle supported best practices to communicate key predictive risk factors among the team.
Evaluation/outcome: The alert data analysis showed an increased alert generation volume and an increased case identification sensitivity to an average of 98% without significantly reducing specificity. In all pediatric units observed, the alert showed an increase in the negative predictive value. In a statistically analysis of 420 cases in the ED, the alert sensitivity increased from 33% to 100%, and a specificity of 92% to 99% after modifications. Alert revisions met the aim of improving sensitivity to support early identification of non-severe sepsis. For non-severe sepsis cases, 100% of cases met the one hour aim of time-to-antibiotics from go-live to May 2019. Greater than 90% of cases met the one hour aim of time-to-fluid-bolus. Both metrics demonstrated a decrease in total turnaround time. Initiatives toward reducing alert fatigue include modification to the inclusion criteria and ongoing analysis of the alert’s statistical performance.
Best practice application, workflow standardization, data analytics, and interdisciplinary collaboration contributed to improved outcome measures. Ongoing surveillance of overall length of stay, mortality, and resource utilization are continued initiatives. The project next steps include analyzing ways to leverage clinical data to generate a predictive score for non-severe pediatrics sepsis and application of the predictive score for risk stratification and early care intervention. Predictive analytics work is needed to optimize recognition, and provide meaningful predictions for at-risk pediatric patients.