AI Identifies Children in ER Likely to Develop Sepsis Within 48 Hours

Sepsis, a severe infection that causes life-threatening organ dysfunction, remains one of the leading causes of death among children worldwide. Early identification is critical, yet the condition can develop rapidly and unpredictably, often after a child arrives at the Emergency Department (ED) without obvious symptoms. To improve early diagnosis and intervention, researchers have now developed artificial intelligence (AI)-based predictive models that identify children at risk of developing sepsis within 48 hours, even before organ dysfunction becomes apparent.

A study conducted by researchers at Northwestern University (Evanston, IL, USA) and Ann & Robert H. Lurie Children’s Hospital of Chicago (Chicago, IL, USA) represents the first use of AI models to predict pediatric sepsis based on the new Phoenix Sepsis Criteria. The models were developed using routine electronic health record (EHR) data collected during the first four hours of a child’s stay in the ED. This approach enabled early risk detection while excluding cases where sepsis was already present upon arrival.

The research involved data from five health systems that are part of the Pediatric Emergency Care Applied Research Network (PECARN), ensuring a large and diverse patient population. The AI models were trained and validated to identify early signs of sepsis while minimizing false positives, aiming to support the timely initiation of lifesaving treatments.

Findings from the study, published in JAMA Pediatrics, demonstrated that the AI models achieved robust accuracy in distinguishing children who were likely to develop sepsis from those who were not at risk. The system showed a strong balance between sensitivity and specificity, allowing precise predictions without overidentifying low-risk patients. These results indicate that AI-driven analysis of EHR data can serve as a reliable and efficient tool for early sepsis prediction in emergency care settings.

The study’s success highlights the growing role of AI in precision medicine for pediatric care. By enabling preemptive treatment before organ dysfunction develops, the models have the potential to reduce mortality rates and improve clinical outcomes. Future research aims to integrate AI predictions with clinician judgment to refine accuracy further and ensure unbiased, patient-centered application. This combined approach could set a new standard for proactive pediatric sepsis management worldwide.

“The predictive models we developed are a huge step toward precision medicine for sepsis in children,” said corresponding author Elizabeth Alpern, MD, MSCE. “Future research will need to combine EHR-based AI models with clinician judgment to make even better predictions.”