Abstracts

Rationale: Neonates with acute brain injury often undergo electroencephalogram (EEG) monitoring. While EEG is often reduced to a ‘seizure detector’, the EEG background data contain a vast amount of information about brain function. EEG background features may be useful biomarkers for neurobehavioral outcome, and therefore it is important to ensure that interpretation is consistent across electroencephalographers. The objective of this study is to assess the inter-rater agreement among electroencephalographers interpreting neonatal EEG background patterns using standardized terminology. Methods: Five pediatric electroencephalographers at The Children’s Hospital of Philadelphia reviewed 5-minute epochs of EEG from 27 full term neonates with hypoxic-ischemic encephalopathy who underwent continuous conventional video EEG recording. The EEG assessment tool was based on the American Clinical Neurophysiology Society’s guideline for neonatal EEG terminology. Inter-rater agreement was assessed using percent agreement and Kappa coefficients [poor ( < 0), slight (0.01-0.2), fair (0.21-0.4), moderate (0.41-0.6), substantial (0.61-0.8), almost perfect (0.81-1), perfect (1)]. Results: Some EEG features could not be assessed meaningfully using Kappa scores due to insufficient samples, including synchrony, symmetry, discontinuity type, interburst abnormalities, presence of dysmaturity features, and behavioral state. Table 1 summarizes the data for the EEG features with adequate samples. Based on kappa’s coefficient, inter-rater agreement was substantial for voltage (0.62); moderate for binary overall impression (0.51), burst abnormality type (0.49), EEG transient type (0.48), categorical overall impression (0.47), burst voltage (0.45), continuity (0.44); fair for presence of graphoelements (0.38), presence of seizures (0.38); and slight for presence of EEG transients (0.19), interburst voltage (0.09), interburst type (0.07), discontinuity type (0.02). Conclusions: There is variability across electroencephalographers in neonatal EEG background interpretation indicating a need for improvement using educational and technological strategies. If electroencephalographers’ agreement cannot be improved with these types of interventions, these results demonstrate an increased need for quantitative analysis of EEG to remove the subjectivity of EEG interpretation. Future studies of conventional neonatal EEG features as predictors of outcome should utilize features with higher inter-rater agreement to ensure reliability and generalizability of results. Funding: NIH Neuroepidemiologic T32 – University of Pennsylvania