Developing a predictive model for early identification of hypoxic-ischemic encephalopathy in high-risk neonates using clinical data

Background Neonatal encephalopathy is a leading cause of death and disability in neonates. Most cases are due to hypoxic-ischemic encephalopathy (HIE) which is treated with therapeutic hypothermia (TH). Treatment is only effective if initiated within 6 h of birth, leaving a narrow window for diagnosis. Early recognition is challenging because clinicians must rely on nonspecific clinical signs and basic laboratory indicators. Although multiple biomarkers, including metabolites and electroencephalography, show potential, none have demonstrated adequate diagnostic accuracy or practicality for universal bedside use. As a result, many infants at risk remain unidentified in time for treatment. Therefore, we developed a predictive model using early clinical data to identify neonates at risk for HIE requiring TH. Methods A secondary analysis was conducted on 362 neonates from the Midwest Neonatal Encephalopathy Registry. Eligible neonates had HIE receiving TH or controls requiring positive pressure ventilation in the delivery room but no HIE/TH. Logistic regression was used to predict HIE. Results HIE was associated with higher ALT, AST, creatinine, lactate, nucleated red blood cells, and lower pH compared to those receiving resuscitation in the delivery room but without a diagnosis of HIE receiving TH. The regression model incorporating base excess, signs of possible fetal hypoxia, and 5-min Apgar scores provided the best discrimination with an AUC of 0.831 in 259 infants. Conclusion This simple model demonstrates robust accuracy as a screening tool for HIE using data available within the first hours of life but requires validation in larger cohorts.