The EEG signals were recorded using Sedline brain function monitors (Masimo Corporation, Irvine, CA, USA), with 250 Hz sampling rate and 4 frontal electrodes. We re-referenced the signals to 2 bipolar channels: Fp1-F7 and Fp2-F8. The signals were first notch filtered at 60 Hz, then bandpass filtered between 0.5 Hz to 20 Hz, and finally downsampled to 62.5 Hz.
We took 1 h EEG segments 30 min before and 30 min after each RASS or CAM-ICU assessment. This is because the assessment times recorded by the ICU nurses may be imprecise, since they are recorded after performing assessments. We therefore included the longer EEG segment to ensure it included the actual assessment time.
EEG artifact were defined based on the presence of any of the following in any EEG channel: (1) maximum amplitude higher than 1000 µV; (2) standard deviation less than 0.2 µV; (3) overly fast changes of more than 900µV within 0.1 s; or (4) spuriously staircase-like spectrum, when the maximum value obtained by convolution with a predefined staircase-like kernel exceeds an empirical threshold of 10, indicating the presence of nonphysiologic single-frequency artifacts from ICU machines (e.g. cooling blankets or pumps).
We took 1 h EEG segments 30 min before and 30 min after each RASS or CAM-ICU assessment. This is because the assessment times recorded by the ICU nurses may be imprecise, since they are recorded after performing assessments. We therefore included the longer EEG segment to ensure it included the actual assessment time.
EEG artifact were defined based on the presence of any of the following in any EEG channel: (1) maximum amplitude higher than 1000 µV; (2) standard deviation less than 0.2 µV; (3) overly fast changes of more than 900µV within 0.1 s; or (4) spuriously staircase-like spectrum, when the maximum value obtained by convolution with a predefined staircase-like kernel exceeds an empirical threshold of 10, indicating the presence of nonphysiologic single-frequency artifacts from ICU machines (e.g. cooling blankets or pumps).