(1462) EEG Profiles of Clinical Delirium Phenotypes in Critical Illness

Introduction: Delirium is a common syndrome in the critically ill that involves acute disturbances in attention and awareness. It can be categorized into distinct clinical phenotypes based on likely precipitating etiologies. We previously identified an electroencephalography (EEG)-based signature of delirium and a composite delirium score using spectral and complexity metrics. Here, we hypothesize that clinical delirium phenotypes may be distinguished by these metrics.

Methods: In this retrospective exploratory study, we identified patients aged 18 and older who had EEGs recorded while admitted to Vanderbilt critical care units between January 1, 2021 and December 31, 2022. Presence of delirium was determined based on the Confusion Assessment Method for the ICU (CAM-ICU). Among CAM-ICU positive participants, we assigned clinical delirium phenotypes using criteria established by Girard et al (2018). For each participant, we extracted 5-minute EEG segments and computed 28 spectral and complexity metrics. Grouping the EEGs by phenotype, we identified each phenotype’s EEG signature and compared metrics using two-tailed Student’s t-tests.

Results: We identified 266 ICU patients with EEGs during the study timeframe. Of 133 patients analyzed to date, 66 were CAM-ICU positive. Eight patients were excluded due to cranial skull breach and one due to absence of raw EEG data. Of the remaining 57, 44 met criteria for the septic phenotype, 44 for sedative-associated, 19 for metabolic, 18 for hypoxic, and 3 were unclassified. Several EEG metrics showed significant changes based on phenotype. Relative alpha power and alpha-delta ratio were lower with metabolic delirium and approximate entropy (apEn) was lower with hypoxic delirium. Sedative-associated delirium showed multiple changes: lower relative theta power, high-beta variability, apEn, low-frequency and peak interhemispheric coherence, and a lower composite delirium score (p < 0.05 for all).

Conclusions: In this preliminary analysis, we identified several EEG metrics that appear to characterize sedative-associated delirium and may serve as indicators for this phenotype. Since sedative-associated delirium is strongly associated with long-term cognitive impairment, its early identification using EEG metrics can be tested as a tool to reduce this complication of ICU care.