AI Models Revolutionize Early Sepsis Detection in Children

Sepsis, a severe and potentially life-threatening condition caused by infections, poses a significant risk to children globally. Researchers at Northwestern University and Ann & Robert H. Lurie Children’s Hospital in Chicago have developed advanced artificial intelligence (AI) models that can predict which children are at high risk for sepsis within 48 hours of their arrival in the emergency department. This groundbreaking work, recently published in JAMA Pediatrics, aims to enable early intervention and preemptive care for vulnerable patients.

The newly developed models utilize routine electronic health record (EHR) data collected during the initial four hours of a child’s emergency visit. These models represent a significant advancement in predicting sepsis, as previous efforts had not successfully improved early diagnosis. “The predictive models we developed are a huge step toward precision medicine for sepsis in children,” stated Dr. Elizabeth Alpern, a professor of pediatrics at Northwestern University and division head of emergency medicine at Lurie Children’s. She emphasized that the models effectively identify children who will develop sepsis without mistakenly labeling those who are not at risk.

Study Methodology and Findings

The study involved five health systems that are part of the Pediatric Emergency Care Applied Research Network (PECARN). This collaboration provided the researchers with access to a vast dataset that includes a diverse population of pediatric patients. The team analyzed data from emergency department visits between January 2016 and February 2020 to train their machine-learning models. They subsequently validated these models using data from 2021 to 2022.

Focusing on the first four hours of care, the researchers aimed to predict the likelihood of sepsis occurring within the subsequent 48 hours. They specifically excluded children who presented with sepsis at arrival or shortly thereafter, concentrating on those who were at risk but had not yet shown symptoms. Key predictive features included the emergency department triage score, heart rate, respiratory rate, and pre-existing medical conditions such as cancer.

“Our models showed robust balance in identifying children in the emergency department who will later develop sepsis,” Alpern noted. The ability to accurately predict sepsis is crucial for initiating timely therapies that can save lives.

Next Steps and Future Research

The researchers conducted thorough evaluations of their models to ensure they were free of biases that could affect results. Looking ahead, Alpern emphasized the need for future research to combine EHR-based AI predictions with clinical judgment, aiming to enhance the accuracy and effectiveness of sepsis detection further.

This innovative project received support from the National Institute of Child Health and Human Development (NICHD) through grant R01HD087363. The findings from this study underscore the potential of AI in transforming pediatric emergency care, ultimately aiming to reduce the incidence of sepsis-related complications and fatalities among children.