Abstract / Summary
Sepsis-associated delirium (SAD) is thought to involve disrupted functional brain network connectivity. The aim of this study was to characterize the electroencephalographic (EEG) features of SAD and to develop a predictive model based on these features. Of 98 eligible patients with sepsis, 62 were enrolled and underwent continuous EEG monitoring, whereas 36 were not enrolled because monitoring was declined by the patients or their legally authorized representatives. All patients with SAD were diagnosed after the initiation of EEG monitoring. Continuous EEG monitoring was initiated within 24 h after ICU admission. Patients with an ICU stay of less than 24 h were not included because they could not complete the prespecified monitoring and observation protocol. Relative power spectral density across multiple frequency bands were extracted, and whole-brain differences in EEG features between the two groups were compared. A predictive model for SAD was constructed based on the EEG features and clinical data. Compared with patients without SAD, patients with SAD exhibited significantly lower alpha power and higher gamma power, primarily in the prefrontal and temporal regions. Relative gamma power was independently associated with SAD (OR = 2.085, 95% CI: 1.026–4.239, p = 0.042). Moreover, no significant association was observed between EEG features and 28-day mortality in patients with sepsis. Compared with patients without SAD, patients with SAD showed increased gamma oscillation power in the frontal lobe. Gamma-band activity in the prefrontal cortex may serve as a promising biomarker for SAD and improve clinical assessment of the condition.