Evaluation of a Commercial Pediatric Sepsis Early Detection Alert in the Emergency Department

Importance Screening tools have been implemented with the goal of improving early detection of pediatric sepsis, with limited evidence of effectiveness. The Epic Systems Pediatric Sepsis Early Detection Model (PESM) and its associated clinical alert have been widely implemented in emergency departments (EDs) in the US but have not undergone independent evaluation. Objective To evaluate the diagnostic characteristics of the Epic PESM for (1) identifying sepsis and (2) identifying sepsis before patients met Phoenix sepsis criteria. Design, Setting, and Participants This retrospective multicenter cohort study of pediatric ED encounters was conducted across 8 EDs within a New York health system from January 2024 through June 2025. Children younger than 18 years were included; behavioral health encounters and ED deaths within 1 hour of arrival were excluded. Data were analyzed from August 2025 through July 2026. Exposure Triggering of the PESM alert in the first 24 hours after ED arrival. Main Outcomes and Measures Sepsis was defined as a Phoenix score of 2 or higher. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated for the outcome of sepsis within 24 hours of ED arrival. Secondarily, evaluation was restricted to only instances in which the alert triggered before sepsis onset. Sensitivity was evaluated in relevant subgroups, and a subgroup with a sensitivity 10% or more below the overall cohort sensitivity was considered clinically meaningful. Results Among 166 943 pediatric ED encounters, 189 cases of sepsis occurred in the first 24 hours. A total of 76 685 pediatric patients (45.9%) were female, and 79 032 patients (47.3%) were under the age of 5 years. Sensitivity of the PESM for identifying patients with sepsis was 36% (95% CI, 29%-43%); PPV was 3% (95% CI, 2%-4%), and NPV was greater than 99% (95% CI, 99%-99%). The alert triggered before sepsis with a sensitivity of 30% (95% CI, 24%-37%). Subgroup sensitivity was 10% or more below overall cohort sensitivity in Black children (22%; 95% CI, 9%-40%) and among children less than 5 years old (20%; 95% CI, 12%-30%), as well as in EDs outside of tertiary children’s hospitals (25%; 95% CI, 14%-40%). Conclusions and Relevance Per the results of this multicenter cohort study, the PESM alert had low sensitivity for identifying children with sepsis and does not support timely or equitable pediatric sepsis screening. Sensitivity was low in Black children, in children younger than 5 years, and outside of tertiary children’s hospitals.