G307 A novel approach to recognizing community-acquired paediatric sepsis in an emergency department: a cross-sectional observational study

Abstract

Objective To determine the accuracy of a novel triage-based electronic screening algorithm and sepsis protocol (SP) to recognise community-acquired sepsis in a paediatric emergency department (ED). Design A retrospective, cross-sectional observational study. Setting Paediatric ED. Patients All children Primary outcome Culture-Positive Sepsis: 1) All children who received intravenous treatment for sepsis in ED and/or had a discharge diagnosis of sepsis following in-patient stay or 2) required PICU care for sepsis at point during their hospital attendance AND had positive blood cultures. Secondary outcomes Culture-negative/Clinical Sepsis: 1) All children who received intravenous treatment for sepsis in ED and/or had a discharge diagnosis of sepsis following in-patient stay or 2) required PICU care for sepsis at point during their hospital attendance BUT blood cultures. Re-attendance of any child that required sepsis bundle treatment, within 48 hours of ED attendance or in-patient discharge. Results 12301 children were included. The electronic screening tool (eST) had a 6.2% trigger alert rate. The eST had a 100% sensitivity and 89.8% specificity, with 4.97% positive predictive value (PPV) and 100% negative predictive value (NPV). The eST when combined with focussed clinician assessment (SP) had a 97.4% sensitivity and 99.3% specificity, with a 45.2% PPV and 99.9% NPV. 1.46% of screened children received IV treatment. The point prevalence of community-acquired sepsis in ED was 0.31%. Conclusions We reported the implementation of a novel electronic sepsis screening tool combining clinician assessment (SP) to predict community-acquired sepsis and sepsis at the point of care at ED triage. The SP shows a high degree of accuracy for recognising sepsis at triage within a paediatric ED. Paediatric sepsis screening tools are found to be most effective when utilised in conjunction with clinician assessment. The SP would be as a feasible model for trials in further multi-centre validation studies. Clinical decision support at ED triage using Artificial Intelligence technology was proposed as a future direction for recognising paediatric sepsis in ED.