IHQSE Highlights

---

Pediatric sepsis is a leading cause of death for children, and an emergency in which timely diagnosis and treatment are lifesaving. Our study Clinical Decision Support for Septic Shock in the Emergency Department: A Cluster Randomized Trial sought to address a gap in knowledge around approaches to diagnose pediatric sepsis in the earliest stages. In a cluster-randomized trial in four Children’s Hospital Colorado Emergency Departments, this study implemented decision support to predict septic shock and measured its effect on treatment and outcomes. The clinical decision support was developed at Children’s Hospital and the University of Colorado, based on models that use extant EHR data to predict septic shock. This was the first reported prospective, controlled trial of decision support intended to aid in early pediatric sepsis diagnosis and treatment. While papers have previously described implementation of consensus-based tools for sepsis diagnosis, few have been based on derived, validated models, and they have not been prospective trials.

We found that implementation of decision support based on machine learning models to predict septic shock was feasible and acceptable to clinicians. However, it did not change the outcomes for patients, in this setting of EDs with pre-existing high-quality sepsis care. 

Tell us about your approach to this project?
Many hospitals (including Children’s Hospital Colorado) have improved sepsis care and outcomes through quality improvement. This QI work has been very important and has usually included a tool to improve early diagnosis of sepsis. However, no tool had demonstrated optimal test characteristics, and every hospital was using a different tool. No one knew if the use of a diagnosis tool improved outcomes more than QI work alone. It was unclear if an early diagnosis tool was better than clinical recognition, and no studies in children had tested a sepsis recognition system in a controlled trial compared to clinical suspicion alone. We sought to answer this question with a controlled trial.

Why is this work important? 
This is the only prospective randomized controlled trial of any kind of sepsis diagnostic tool that has been published in pediatrics. In our trial, we did not identify a difference in the quality of patient care or outcomes using this tool, compared to clinical suspicion alone. However, clinicians really liked the tool we developed and wanted to keep it in clinical use after the trial ended. Notably, the tool was based on predictive models that we had developed with many investigators across the Anschutz campus (from ACCORDS, the Department of Pediatrics, and Children’s Hospital Colorado), which were integrated into the Electronic Health Record in a manner designed to minimize alert fatigue. Even though it didn’t change outcomes on a population level, clinicians felt like it was an unobtrusive safety net, that might help them avoid a potentially catastrophic miss on an individual level.

How do you think this will impact healthcare? 
As the only controlled trial of a sepsis screening tool in pediatrics, this has potential to affect sepsis guidance about the importance of a sepsis screening tool. It highlights that in a system with active quality improvement around all aspects of sepsis care, implementing a screening tool may not change outcomes. We conducted this study in 4 Children’s Colorado hospitals that had had an active sepsis QI program for 10 years, as well as high-quality Emergency Department triage systems. Globally, not all emergency systems have robust triage systems. We studied an EHR-embedded, predictive Clinical Decision Support system. While in a system like ours, a tool like ours didn’t change outcomes, and should be a lower priority aspect of sepsis care in similar systems. It is possible that screening tools may be more impactful in other systems, particularly those without ED triage, or the ability to engage in robust, sustained sepsis QI.