Neural Response to Familiar Names Predicts Outcome of Comatose ICU Patients: A Prospective Observational Cohort Study

Predicting the outcome of comatose patients in the intensive care unit (ICU) can inform decision making but remains challenging. Recent studies suggest that task-state electroencephalography (EEG) can detect covert cognition and facilitate patient prognosis. This study aimed to predict the outcome of comatose patients, by assessing covert processing of familiar names using a state-of-the-art EEG frequency tagging approach. Eighty-nine comatose patients following acute brain injury were recruited from five ICUs. Patients were presented with a rapid stream of familiar names and acoustically matched but unintelligible control sounds. EEG responses tracking the familiar names and control sounds were extracted in the frequency domain and utilised to predict the outcome of each patient, which was assessed at 1, 3, and 6 months post-injury using the Glasgow Outcome Scale-Extended (GOSE). Name-tracking EEG responses positively correlated with GOSE scores. A machine learning model integrating EEG responses and clinical characteristics achieved AUCs of 0.86, 0.88, and 0.86 in the test set, and 0.91, 0.90, and 0.85 in the external validation set, for predicting outcomes at 1, 3, and 6 months, respectively. These findings underscore that EEG assessment of residual processing of familiar names relates to patient outcomes and has the potential to predict outcome of comatose ICU patients.