Models based on routine electronic health record data
By Lori Solomon HealthDay Reporter
THURSDAY, Oct. 16, 2025 (HealthDay News) — Artificial intelligence can help predict which children in the emergency department will develop sepsis within 48 hours, according to a study published online Oct. 13 in JAMA Pediatrics.
Elizabeth R. Alpern, M.D., from the Ann & Robert H. Lurie Children’s Hospital of Chicago, and colleagues developed machine learning models to estimate the probability of developing sepsis in the subsequent 48 hours in children aged ≥2 months to <18 years. Electronic health record data for the training cohort included more than 1.6 million emergency department visits and 719,298 visits in the test cohort.
The researchers found that performance characteristics for the Phoenix Sepsis Criteria (PSC) sepsis prediction models were an area under the receiver operating characteristic (AUROC) of 0.92 for logistic regression and 0.94 for gradient tree boosting. AUROCs for PSC shock models were ≥0.92. The gradient tree boosting models showed positive likelihood ratios of 4.67 to 6.18 for sepsis and 4.16 to 5.83 for septic shock. Emergency severity index, age-adjusted vital signs, and medical complexity were predictive features. For all demographic characteristics except payer, assessment of model performance fairness was similar. Compared with patients with commercial payers, the AUROC for patients with Medicaid insurance was better.
“The predictive models we developed are a huge step toward precision medicine for sepsis in children,” Alpern said in a statement. “These models showed robust balance in identifying children in the emergency department who will later develop sepsis, without overidentifying those who are not at risk. This is very important because we want to avoid aggressive treatment for children who don’t need it.”
Abstract/Full Text (subscription or payment may be required)
Copyright © 2025 HealthDay. All rights reserved.
