Abstracts

Rationale: About 1/3 of patients undergoing EEG monitoring in the neuro-ICU have non-convulsive seizures (NCS) and almost ¾ of these cases are non-convulsive status epilepticus (NCSE) [1]. Obtaining an emergency EEG for the diagnosis of NCSE and NCS in the ICU raises logistic problems in most hospitals. Previous studies have looked into the hairline EEG for a broader population than the critically ill, revealing a sensitivity of 72% and a specificity of 93% for seizure detection and a sensitivity of 78% and a specificity of 87% for diffuse slowing [4]. We created a montage simple enough to be performed by caregivers with limited training in the performance of EEGs and rapidly achievable to meet the time constraints of a busy on call schedule. Methods: Seven electrodes (Fp1, Fp2, T3, T4, Cz, O1, O2) easily applied without the need for tape measure using anatomic landmarks (pupils, ears, vertex and inion) were used to configure three different montages (double diamond or “Espinosa montage”, circumferential and Cz referential). Reports of emergency EEGs performed in critically ill in the ICU between 1/1/06-5/28/08 were reviewed. 28 samples of diffuse slowing and 16 samples of epileptiform activity were randomly selected. All 20 min records were reviewed again by 2 neurophysiologists and 10 of them had undisputable seizures and were included in the study. A separate analysis was done for the 6 discordant samples. The 28 samples with slowing and the 10 samples with seizures were reviewed in the proposed montages by 2 neurophysiologists, who were asked to classify them as seizures versus diffuse slowing. Then, the sensitivities and specificities for seizure detection were estimated. Comparison was done with the previously used hairline montage. Results: The average sensitivity of the abbreviated montage for seizure detection by the attendings was 90%, while the specificity was 94%. For the one neurophysiologist, there was only 1 false positive interpretation that was due to epileptiform discharges not clearly representing a seizure. For the second neurophysiologist, there was 1 false positive interpretation due to an artifact and there were 2 false negative interpretations of seizures that were deemed PLEDs (1) and GPEDs (1) in the abbreviated montage. In comparison, the hairline EEG showed a sensitivity of 70% and a specificity of 100% for the detection of seizures. The 3 false negative interpretations were seizures read as PLEDs (1) and GPEDs (2) in the hairline montage. Conclusions: The proposed montage is a quick (9 minutes to perform) and reliable (90% sensitivity) screening tool in the ICU for the on call neurology resident, when routine EEG is restricted secondary to lack of support and time. Its specificity reaches 94%, with the false positives representing mostly other abnormal EEG patterns, such as PLEDs and GPEDs. In comparison to the previously studied hairline EEG, the proposed montage appears to have significantly higher sensitivity for seizure detection. Further analysis is ongoing to assess the sensitivity and specificity of the proposed montage by residents.