Clinical and Electroencephalographic Predictors of Seizures and Status Epilepticus in 12,450 Critically Ill Adults: A Retrospective Cohort Study

This is the authors' abstract. We don't add key points for this paper.

Objectives Status epilepticus (SE) is associated with significantly higher morbidity and mortality than isolated seizures. Our objective was to identify clinical diagnoses and rhythmic and periodic electroencephalogram patterns (RPPs) associated with SE and seizures. Design Retrospective cohort study. Setting Tertiary-care hospitals. Subjects Twelve thousand four hundred fifty adult hospitalized patients undergoing continuous electroencephalogram (cEEG) monitoring in selected participating sites in the Critical Care EEG Monitoring Research Consortium database (February 2013 to June 2021). Interventions Not applicable. Measurements and main results We defined an ordinal outcome in the first 72 hours of cEEG: no seizures, isolated seizures without SE, or SE (with or without isolated seizures). Composite groups included isolated seizures or SE (AnySz) and no seizure or isolated seizures. In this cohort (mean age: 60 ± 17 yr), 1,226 patients (9.8%) had AnySz and 439 patients (3.5%) had SE. In a multivariate model, factors independently associated with SE were cardiac arrest (9.2% with SE; adjusted odds ratio, 8.8 [6.3-12.1]), clinical seizures before cEEG (5.7%; 3.3 [2.5-4.3]), brain neoplasms (3.2%; 1.6 [1.0-2.6]), lateralized periodic discharges (LPDs) (15.4%; 7.3 [5.7-9.4]), brief potentially ictal rhythmic discharges (BIRDs) (22.5%; 3.8 [2.6-5.5]), and generalized periodic discharges (GPDs) (7.2%; 2.4 [1.7-3.3]). All above variables and lateralized rhythmic delta activity (LRDA) were also associated with AnySz. Factors disproportionately increasing odds of SE over isolated seizures were cardiac arrest (7.3 [4.4-12.1]), clinical seizures (1.7 [1.3-2.4]), GPDs (2.3 [1.4-3.5]), and LPDs (1.4 [1.0-1.9]). LRDA had lower odds of SE compared with isolated seizures (0.5 [0.3-0.9]). RPP modifiers did not improve SE prediction beyond RPPs presence/absence ( p = 0.8). Conclusions Using the largest existing cEEG database, we identified specific predictors of SE (cardiac arrest, clinical seizures prior to cEEG, brain neoplasms, LPDs, GPDs, and BIRDs) and seizures (all previous and LRDA). These findings could be used to tailor cEEG monitoring for critically ill patients.