Utility of Clinical Features in Identifying Electrographic Seizures in Hospitalized Patients Admitted for Non-Neurological Diagnoses.

Importance: Electrographic seizures (ESz) are seizures without prominent motor activity diagnosed with electroencephalogram and are a common complication in critically ill patients with alterations of consciousness. Previous studies suggested clinical signs, including ocular movement abnormalities, facial/periorbital twitching, or remote seizure risk factors, are sensitive for presence of ESz.

Objectives: To assess the utility of clinical features in identifying ESz in critically ill patients with alterations of consciousness.

Design, setting, and participants: This is a retrospective case-control study of 50 patients admitted to the University of North Carolina (UNC) Medical Center and UNC Rex Hospital. Inpatients older than 18 years old undergoing continuous video electroencephalogram (cEEG) were included. Patients admitted for neurologic diagnoses were excluded. A total of 25 patients with ESz (Sz-EEG) were matched with 25 controls by electroencephalogram duration ± 12 hours (No-Sz-EEG). Elements of patient's history and physical findings previously shown to be sensitive for presence of ESz were collected. Descriptive statistical analyses were used.

Results: Most patients were admitted to medical ICUs (72%; n = 36). There was no difference between groups in clinical findings previously shown to be sensitive for ESz. Positive and negative likelihood ratios for these findings generally fell between 1-2 and 0.5-1, respectively, indicating they are inaccurate predictors for ESz. Patients with ESz had significantly higher mortality (p = 0.012).

Conclusions and relevance: Our matched case-control study showed that in the critically ill patient population hospitalized in tertiary care centers and admitted for non-neurologic primary diagnoses, incidence of ocular movement abnormalities, facial/periorbital twitching, and presence of remote risk factors for seizures had low predictive accuracy for ESz. However, these findings are not generalizable to patients with neurologic diseases or to other practice settings with different levels of access to cEEG. We concluded that in this exploratory analysis of hospitalized critically ill patients with non-neurologic diagnoses, these clinical signs did not reliably stratify risk for ESz on cEEG. However, further prospective studies are needed to better evaluate these conclusions.