Abstracts

Rationale: Seizures in neonates are difficult to identify clinically, and video EEG (vEEG) monitoring is considered the gold standard to confirm the presence of seizures prior to initiating treatment. EEGs are often first screened by EEG technologists and later reviewed by clinical neurophysiologists. A previous study of pediatric critical care EEGs reported technologists had moderate aggregate agreement to seizures with sensitivity ranging from 44 to 93%. We hypothesized technologist agreement and sensitivity/specificity of neonatal EEGs at our single center would vary based on training and certification.




Methods: Fifty vEEG samples of 5 minutes were interpreted by 33 EEG technologists. Interpretations were dichotomized into the absence or presence of seizures, interictal epileptiform discharges (IEDs), brief rhythmic discharges (BRDs), and sharp transients. Background continuity, symmetry, and voltage were dichotomized into normal or abnormal. Two alternative interrater agreement statistics, Fleiss’ kappa and free-marginal multi-rater kappa, were presented with 95% confidence intervals (CI). Agreement for both forms of kappa was interpreted as slight (0-0.20), fair (0.21-0.40), moderate (0.41-0.60), substantial (0.61-0.80), or almost perfect (0.81-1.00). Seizure sensitivity and specificity were compared against the interpretations from board certified neurophysiologists. We assessed correlations with technologist certification and years of experience using the Kruskal-Wallis rank sum test.


Results: EEG technologist agreement was moderate for seizures (Fleiss’ kappa 0.414, free kappa 0.415) and slight for IEDs, BRDs, and sharp transients. Technologist agreement for background continuity, symmetry, and voltage ranged from slight to moderate.

EEG technologists had 80% sensitivity for seizures and 80% specificity for seizures. Technologists with CLTM certification had greatest specificity of 87%. Seizure specificity increased with years of experience from 57% (0-1 years) to 85% ( >2 years).


Conclusions: Technologist review of neonatal EEGs is most valuable to screen for seizures, as sensitivity and specificity were 80% and interrater reliability was moderate. However, interrater reliability was lower for interictal abnormalities and background interpretation. While screening for neonatal EEG for seizures by technologists is likely adequate, gaps in interpretation require continued oversight of neurologists.


Funding: This study was supported by the Nationwide Children’s Hospital Foundation.