Abstracts

Rationale: Electrographic seizures (ES) are common in critically ill children with acute encephalopathy. ES identification by neurophysiologist review of conventional EEG is resource intense and often is not available for use in real-time. We aimed to evaluate the validity of neurologist interpretation of density spectral array (DSA) EEG for ES identification. Methods: We studied continuous EEG tracings from 21 consecutive critically ill children. Conventional EEG was scored for ES by a pediatric neurophysiologist, which served as the gold standard for seizure identification. EEG tracings were then converted into 8 channel DSA displays. Four DSA images, each containing two hours, were generated per patient. Eight neurophysiologists received brief and standardized DSA training and then circled DSA elements thought to represent seizures. Images were reviewed in random order with no clinical information (Group A) or with information regarding presence or absence of seizures in the initial 30 minutes and with patient images in order, as might occur in a clinical setting (Group B). Inter-rater reliability for accurately classifying an image as containing a seizure was assessed in both groups (kappa). Sensitivity and specificity for correctly identifying a seizure on each image and the area under the receiver operating characteristic (ROC) curve were calculated for each group. Using full array EEG, 73 ES were evaluated for duration, number of electrodes involved at maximal extent, morphology (spikes versus rhythmic), and evolution (slowly versus clearly evolving). Factors associated with DSA seizure identification were assessed (Wilcoxon rank-sums). Results: ES prevalence was 48% by neurophysiologist review. Inter-rater agreement was moderate in both groups (Group A: kappa 0.51 (95%CI 0.35-0.66), p<0.001; Group B: 0.46 (95%CI 0.33-0.59), p<0.001). Sensitivity and specificity were: Group A, 62% (95%CI 51.2-71.9%), 92.2% (88.1-95.2%) and Group B, 75% (95%CI 64.9-83.4%), 79.1% (95%CI 73.5-84%). ROC curves were not different between the groups (AUC 0.77 for both, p=0.99). Positive predictive values for Groups A and B were 75% (95%CI 63.7%-84.2%) and 57.5% (95%CI 48.1%-66.5%), respectively. Negative predictive values for Groups A and B were 86.5% (95%CI 81.8%-90.4%) and 89.4% (95%CI 84.5%-93.1%), respectively. Of the 73 seizures, 53% were identified by ≥6 raters. Predictors of seizure identification were seizure duration >2 minutes (p<0.001) and non-subtle evolution (p<0.001). 10% of images were falsely classified as containing a seizure and among those, the mean (±standard deviation) false positive rate per hour was 1.5±2. Conclusions: When interpreted by neurophysiologists, DSA may be a useful screening tool for ES identification, although it does not identify all seizures and has only moderate inter-rater reliability. False positives occur indicating that seizure confirmation by conventional EEG review is needed. Additional study is required to determine whether non-neurophysiologist raters can use DSA and whether bedside implementation of DSA leads to more rapid seizure identification.