P0502 - Decision-support algorithm for detection of severe sepsis in adults and children in a pediatric emergency department

Background: Pediatric emergency departments (PEDs) serve a growing adult patient population, who are often medically complex young adults. Embedded electronic health record sepsis screens have been shown to identify pediatric patients with sepsis leading to improved outcomes, but research has been limited in how these interventions perform for adults presenting to PEDs.

Objective: Evaluate the test characteristics of a decision-support algorithm to detect severe sepsis and compare performance in pediatric and adult patients in a PED.
Design/

Methods: Retrospective cohort of visits from March through November 2017 to a tertiary care PED. Visits were included based on an a priori defined high-risk cohort with any of the following: 1) PED orders for 2 saline boluses, intravenous antibiotics, and blood culture; 2) PED vasopressor use; 3) death or intensive care unit (ICU) admission within 24 hours; 4) sepsis treatment pathway initiation; or 5) ICD10 codes for sepsis/septic shock. Adults defined as ≥ 18 years old. The exposure is a two-step decision-support algorithm comprising (1) an electronic medical record vital sign based best-practice alert (BPA) and (2) prompted physician assessment and judgment (Figure 1). The outcome was severe sepsis as defined by systematic inflammatory response syndrome (SIRS) criteria with a) pediatric sequential organ failure assessment (pSOFA) criteria), b) clinically significant sepsis treatment of 60ml/kg of normal saline bolus and IV antibiotics administration in the PED, or c) initiation of the sepsis order set with admission to the ICU.

Results: 1660 visits included 1489 children (89.7%) and 171 adults (10.3%). The BPA was triggered for 59.2% of children and 87.1% of adults in the high-risk cohort. The BPA and physician judgment was positive for 36.4% of children and 36.8% of adults. The BPA alone had significantly higher sensitivity in adults, compared to children [87% (95% CI 77%-93%) vs 74% (95% CI 69-76%), respectively] and lower specificity in adults compared to children [13% (95% CI 7-21%) vs 48% (95% CI 45-52%); Table 1]. With the addition of provider judgment, there were no significant differences in sensitivity or specificity between adults and children.

Conclusion: The sensitivity of an electronic vital sign based BPA alone was high for adults in a PED, at a significant cost to specificity. However, an electronic screen coupled with provider judgment performed similarly in children and adults seen in a PED.