Delirium detection using relative delta power based on 1 minute single-channel EEG: a multicentre study

Dutch Delirium Detection Study Group

Research output: Contribution to journalArticleAcademicpeer-review

41 Citations (Scopus)

Abstract

Background: Delirium is frequently unrecognised. EEG shows slower frequencies (i.e. below 4 Hz) during delirium, which might be useful in improving delirium recognition. We studied the discriminative performance of a brief single-channel EEG recording for delirium detection in an independent cohort of patients. Methods: In this prospective, multicentre study, postoperative patients aged ≥60 yr were included (n=159). Before operation and during the first 3 postoperative days, patients underwent a 5-min EEG recording, followed by a video-recorded standardised cognitive assessment. Two or, in case of disagreement, three delirium experts classified each postoperative day based on the video and chart review. Relative delta power (1–4 Hz) was based on 1-min artifact-free EEG. The diagnostic value of the relative delta power was evaluated by the area under the receiver operating characteristic curve (AUROC), using the expert classification as the gold standard. Results: Experts classified 84 (23.3%) postoperative days as either delirium or possible delirium, and 276 (76.7%) non-delirium days. The AUROC of the relative EEG delta power was 0.75 [95% confidence interval (CI) 0.69–0.82]. Exploratory analysis showed that relative power from 1 to 6 Hz had significantly higher AUROC (0.78, 95% CI 0.72–0.84, P=0.014). Conclusions: Delirium/possible delirium can be detected in older postoperative patients based on a single-channel EEG recording that can be automatically analysed. This objective detection method with a continuous scale instead of a dichotomised outcome is a promising approach for routine detection of delirium. Clinical trial registration: NCT02404181.
Original languageEnglish
Pages (from-to)60-68
Number of pages9
JournalBritish Journal of Anaesthesia
Volume122
Issue number1
Early online date2018
DOIs
Publication statusPublished - 1 Jan 2019

Cite this