We aim to identify electroencephalographic markers as an indicator of mortality in non-neurological patients admitted to an intensive care unit (ICU) presenting with altered mental status (AMS) and seizures (subtle and non-convulsive).

Seizures can be seen in up to 30% of patients with AMS and can be associated with worse outcomes. However, most studies include patients with acute neurological illnesses. EEG features can provide physiological biomarkers of disease severity that can be utilized in predicting outcomes of patients with AMS and seizures.

Matched case-control study from a retrospective review of 432 patients admitted to our institution and underwent continuous electroencephalogram (cEEG) due to AMS. We included patients >18 years and excluded if admitted for neurological diagnosis and non-first-time EEG recording. A total of 26 patients with seizures were found (Sz-EEG). We matched 26 controls to Sz-EEG patients by demographic and cEEG duration ± 12 hours (NoSz-EEG). APACHE-II score, Charlson comorbidity index (CCI), and EEG characteristics on seizure day were collected.

NoSz-EEG group had higher prevalence of coronary artery disease (p=0.04). After adjusting for confounders, including critical illness severity scores, Sz-EEG group had significantly higher mortality than NoSz-EEG group (p=0.036). The presence of spontaneous burst-suppression (BS, p=0.0029), sporadic epileptiform activity (EA, p<0.0001) and lateralized periodic discharges (LPDs, p=<0.0001) was significantly higher in the Sz-EEG group. BS pattern, EA and LPDs were present in 30% (8/26), 23% (6/26) and 23% (6/26), respectively, of Sz-EEG patients that died during hospitalization.

Our matched case-control study showed that non-neurological patients with AMS and seizures had significantly higher mortality. BS, EA and LPDs patterns were the EEG features with higher prevalence in the Sz-EEG patients. These results suggest that EEG could be a marker of severity of encephalopathy in critically-ill patients. Larger studies are required to assess the strength of predictability of EEG biomarkers and outcomes.