AI Models Show Promise in Predicting Pediatric โSepsis,โฃ Enabling Earlier โขIntervention
CHICAGO, โIL – Researchers have โขsuccessfully developed and validated artificial intelligence (AI)โ models โฃcapable of โฃaccurately identifying children at high risk ofโฃ developing sepsis – a life-threatening condition caused by the body’s overwhelming response to infection – within 48 hours of arrival at the Emergency Department (ED). the study, โคpublishedโ in JAMA Pediatrics, marks a significant step toward precision medicineโข in pediatric sepsis care, perhaps allowing for preemptive treatment โand improvedโฃ outcomes.
Led by Elizabeth Alpern, MD, MSCE, from Ann โค& Robert H. Lurie Children’s Hospitalโ of Chicago, โคtheโฃ multi-centerโฃ study utilized routineโ electronic health record (EHR) data collectedโข during the first fourโ hours of a child’s ED visit, before signsโ of organ dysfunction appeared. This is the first research to leverage AI โmodels based โคon the new Phoenix Sepsis Criteria for predicting sepsisโข in children.
The research teamโ analyzedโค data from five health systems within theโข Pediatric Emergency Care Applied Research Network (PECARN), providing access โคto โa large and diverse patient population. Children already exhibiting sepsis upon arrival or within the โinitial hours of ED careโ were excluded to focus specifically onโข predictive โขcapabilities and enable early therapeutic intervention.
“The predictive models we developed are a huge step toward precisionโค medicine for sepsis in children,” explained Dr. alpern, alsoโฃ Division Head of Emergency Medicine at Lurie Children’s and โฃprofessor โof โฃPediatrics atโข Northwestern University โFeinberg School of Medicine. “These models showedโฃ robust balance in identifying โขchildren in the ED who โwill later develop sepsis,โ without overidentifying those who are not at risk. Thisโ isโค very crucial because we want to avoid aggressive treatment for childrenโ who don’t need it.”
Researchers also prioritized addressing potential biases withinโ the models. Dr. Alpern notedโ that futureโ work will focus on โฃintegrating these AI-driven predictions with clinical judgment to further refine accuracy.
The study was supported by the Nationalโ Institute of Child Health and Human Advancement (NICHD) grant R01HD087363. Dr. Alpern holds โฃtheโข George M. Eisenberg Professorship inโฃ Pediatrics at Northwestern university Feinberg School ofโ Medicine.
