Quantitative EEG signatures of delirium and coma in mechanically ventilated ICU patients

•We identified four spectral metrics of EEG that contribute independently to delirium or coma detection in ventilated ICU patients.•A linear combination of these metrics showed good internal validity as an EEG-based indicator of delirium or coma.•Evaluated on 24 hrs continuous EEG, our delirium/coma...

Full description

Saved in:
Bibliographic Details
Published in:Clinical neurophysiology 2023-02, Vol.146, p.40-48
Main Authors: Williams Roberson, Shawniqua, Azeez, Naureen A., Fulton, Jenna N., Zhang, Kevin C., Lee, Aaron X.T., Ye, Fei, Pandharipande, Pratik, Brummel, Nathan E., Patel, Mayur B., Ely, E. Wesley
Format: Article
Language:English
Subjects:
Coma - diagnosis
Critical care
Delirium
Delirium - diagnosis
Electroencephalography
Encephalopathy
Humans
Intensive Care Units
qEEG
Respiration, Artificial
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:•We identified four spectral metrics of EEG that contribute independently to delirium or coma detection in ventilated ICU patients.•A linear combination of these metrics showed good internal validity as an EEG-based indicator of delirium or coma.•Evaluated on 24 hrs continuous EEG, our delirium/coma indicator showed good separation between groups and stability across hours. To identify quantitative electroencephalography (EEG)-based indicators of delirium or coma in mechanically ventilated patients. We prospectively enrolled 28 mechanically ventilated intensive care unit (ICU) patients to undergo 24-hour continuous EEG, 25 of whom completed the study. We assessed patients twice daily using the Richmond Agitation-Sedation Scale (RASS) and Confusion Assessment Method for the ICU (CAM-ICU). We evaluated the spectral profile, regional connectivity and complexity of 5-minute EEG segments after each assessment. We used penalized regression to select EEG metrics associated with delirium or coma, and compared mixed-effects models predicting delirium with and without the selected EEG metrics. Delta variability, high-beta variability, relative theta power, and relative alpha power contributed independently to EEG-based identification of delirium or coma. A model with these metrics achieved better prediction of delirium or coma than a model with clinical variables alone (Akaike Information Criterion: 36 vs 43, p = 0.006 by likelihood ratio test). The area under the receiver operating characteristic curve for an ad hoc hypothetical delirium score using these metrics was 0.94 (95%CI 0.83–0.99). We identified four EEG metrics that, in combination, provided excellent discrimination between delirious/comatose and non-delirious mechanically ventilated ICU patients. Our findings give insight to neurophysiologic changes underlying delirium and provide a basis for pragmatic, EEG-based delirium monitoring technology.
ISSN:1388-2457
1872-8952
DOI:10.1016/j.clinph.2022.11.012