Preoperative characterization of baseline EEG recordings for risk stratification of post-anesthesia care unit delirium

Study objective: Delirium in the post-anesthesia care unit (PACU-D) presents a serious condition with a high medical and socioeconomic impact. In particular, PACU-D is among common postoperative complications of elderly patients. As PACU-D may be associated with postoperative delirium, early detection of at-risk patients and strategies to prevent PACU-D are important. We characterized EEG baseline signatures of patients who developed PACU-D following surgery and general anesthesia and patients who did not. Design and setting: We conducted a post-hoc analysis of preoperative EEG recordings between patients with and without PACU-D, as indicated by positive bCAM scores post general anesthesia and surgery. Patients and measurements: Preoperative baseline EEG recordings from 89 patients were recorded at controlled eyes-open (focused wakefulness) and eyes-closed (relaxed wakefulness) conditions. We computed power spectral densities, permutation entropy, spectral entropy and spectral edge frequency to see if these parameters can reflect potential baseline EEG differences between PACU-D (31.5%) and noPACU-D (68.5%) patients. Wilcoxon's Rank Sum Test as well as AUC values were used to determine statistical significance. Main results: Baseline EEG recordings showed significant differences between PACU-D and noPACU-D patients preoperatively. Compared to the noPACU-D group, PACU-D patients presented with lower power in higher frequencies during relaxed and focused wakefulness alike. These differences in power led to AUC values of 0.73 [0.59;0.85] (permutation entropy) and 0.72 [0.61;0.83] (spectral edge frequency) indicative of a “fair” performance to separate patients with and without PACU-D. Conclusions: The baseline EEG of relaxed wakefulness as well as focused wakefulness may be used to assess the risk of developing PACU-D following surgery under general anesthesia. Moreover, routinely used monitoring parameters capture these differences as well, potentially allowing an easy transfer to clinical settings.