To determine if EEG variables can risk stratify in-hospital mortality and discharge status of children with liver failure complicated by acute neurological injury in the pre- and post-transplant period.

Children with liver failure admitted to the Intensive Care Unit (ICU) in the pre- and post- transplant period are at risk for acute neurological complications. Previous studies have shown how EEG can detect actionable MRI findings in children with liver failure, but associations with mortality and discharge status are unknown.

This retrospective chart review included patients <25 years old admitted to a MedStar hospital between 2014 and 2022 with ICD code related to liver failure and MRI and EEG obtained during the same admission. Variables included clinical data (i.e. etiology of liver failure), demographic information (i.e. age), MRI findings (i.e. T2/T1 signal change, diffusion restriction, blood product), labs (i.e. AST, ALT), and qualitative and quantitative EEG features (i.e. organization, continuity, power, and symmetry). Primary outcome was death versus survival, and secondary outcome discharge location (home versus inpatient rehabilitation). Data were analyzed using XG-Boost with 5-fold cross validation.

746 unique patient encounters were screened. 70 patient encounters met inclusion criteria, and 61 survived to discharge. XG-Boost modeling of all dependent variables predicting survival yielded precision of 0.906 and recall of 0.951 (f1=0.928). Modeling to predict discharge location yielded precision of 0.697 and recall of 0.535 (f1=0.605). Overall, EEG and MRI features had greater importance in predicting discharge location, and clinical features and demographics had greater importance in predicting mortality.

EEG is an integral but not standalone diagnostic tool in predicting outcomes in children with liver failure complicated by acute neurological injury. Common lab values are in some cases more indicative of outcomes than EEG features.