Feasibility of Neurological Prognostication After Pediatric Cardiac Arrest with a Reduced Number of EEG Electrodes.

Abstract

Electroencephalography (EEG) for neurological prognostication after pediatric cardiac arrest (pCA) shows promise, but is resource-intensive, limiting implementation. Because postanoxic injury causes generalized EEG background pattern changes, fewer electrodes for monitoring may suffice, which could reduce workload. We assessed whether electrode reduction affects visual EEG assessment and the prognostic performance of machine learning models for the prediction of outcome after pCA. We retrospectively included children (<18 years old) after in- or out of hospital CA who were admitted to the pediatric intensive care unit between 2012-2023 and were monitored using EEG at 24 hours after return of circulation. We visually assessed the EEG based on three different bipolar montages (with all available-, 6- and 2-channels) based on the ACNS criteria. Additionally, three machine learning models for the prediction of a poor outcome (defined as a Pediatric Cerebral Performance Category score of 4-6 12 months post-pCA) were trained based on quantitative EEG (qEEG) analysis that was performed for all three montages separately. 82 children (median age 3.5 years) were included, with 60% having a poor outcome (PCPC 4-6). Inter-montage agreement for the visual classification of EEG background pattern was substantial for the 6- and 2-channel montages compared to the full montage (κ: 0.803 and κ: 0.682 respectively). There were no significant differences in model performance on patient level between the different machine learning models trained on qEEG. EEG continuity and amplitude features were most informative and robust across montages. 12-month outcome prediction after pCA using either visual- or qEEG is comparable between full and selected reduced electrode montages.