Automated EEG Background Analysis and 2-Year Outcomes in Neonatal Hypoxic-Ischemic Encephalopathy

Hypoxic-ischemic encephalopathy (HIE) remains an important contributor to neonatal mortality and morbidity despite therapeutic hypothermia. Accurate early prognostication of outcomes is essential for clinical management and risk stratification in future trials.

To evaluate the feasibility and estimative ability of automated electroencephalographic (EEG) background analysis in projecting neurodevelopmental outcomes in neonates with HIE using the Brain State of the Newborn (BSN) score.

This cohort study was a secondary analysis of the High-Dose Erythropoietin for Asphyxia and Encephalopathy (HEAL) trial, which enrolled infants born between January 25, 2017, and October 9, 2019, and were followed up at age 2 years (between age 22 and 36 months). Nine academic centers across the US provided raw EEG data starting within the first 24 hours of life. The data were analyzed between August 15, 2024, and August 25, 2025.

The primary outcome was death or severe neurodevelopmental impairment (NDI) at age 2 years. The BSN scores were computed using a cloud service and correlated with expert reader interpretations. Generalized linear mixed models were used to assess the estimative capabilities (quantified via cross-validated area under the receiver operating characteristic curve [AUROC]) of clinical variables and BSN score for severe NDI or death.

Among 500 infants enrolled in the HEAL trial, 203 were included in the current analysis (median [IQR] gestational age, 39.3 [38.0-40.3] weeks; 121 male [59.6%]). Of these infants, 21 (10.3%) experienced severe NDI, and 28 (13.8%) died. The BSN scores correlated with expert reader EEG background classifications (Pearson correlation coefficient for median BSN, 0.69; 95% CI, 0.64-0.73). Adding median overall BSN score (AUROC, 0.90; 95% CI, 0.84-0.97) or median BSN score at all time points (AUROC, 0.93; 95% CI, 0.88-0.98) significantly improved the prognosis of severe NDI or death compared with clinical variables alone (AUROC, 0.79; 95% CI, 0.70-0.87). Prognostic accuracy of median BSN score overall was similar to expert assessments of EEG background (AUROC, 0.90; 95% CI, 0.81-0.98).

These findings suggest that automated EEG background analysis may provide an objective method for early prognostication in neonates with HIE. The BSN scores correlated with expert classifications and may aid in risk stratification. The BSN's role in the treatment of neonates with HIE in environments lacking expert EEG interpretation requires further study.