Delirium detection using EEG: what and how to measure.

Background: Despite its frequency and impact, delirium is poorly recognized in postoperative and critically ill patients. EEG is highly sensitive to delirium but, as currently used, it is not diagnostic. To develop an EEG-based tool for delirium detection with a limited number of electrodes, we determined the optimal electrode derivation and EEG characteristic to discriminate delirium from nondelirium.

Methods: Standard EEGs were recorded in 28 patients with delirium and 28 age- and sex-matched patients who had undergone cardiothoracic surgery and were not delirious, as classified by experts using Diagnostic and Statistical Manual of Mental Disorders, 4th edition, criteria. The first minute of artifact-free EEG data with eyes closed as well as with eyes open was selected. For each derivation, six EEG parameters were evaluated. Using Mann-Whitney U tests, all combinations of derivations and parameters were compared between patients with delirium and those without. Corresponding P values, corrected for multiple testing, were ranked.

Results: The largest difference between patients with and without delirium and highest area under the receiver operating curve (0.99; 95% CI, 0.97-1.00) was found during the eyes-closed periods of the EEG, using electrode derivation F8-Pz (frontal-parietal) and relative δ power (median [interquartile range (IQR)] for delirium, 0.59 [IQR, 0.47-0.71] and for nondelirium, 0.20 [IQR, 0.17-0.26]; P = .0000000000018). With a cutoff value of 0.37, it resulted in a sensitivity of 100% (95% CI, 100%-100%) and specificity of 96% (95% CI, 88%-100%).

Conclusions: In a homogenous population of nonsedated patients who had undergone cardiothoracic surgery, we observed that relative δ power from an eyes-closed EEG recording with only two electrodes in a frontal-parietal derivation can distinguish among patients who have delirium and those who do not.