Abstracts

Rationale: Interpretation of neonatal EEGs is a challenge due to rapid maturational changes occurring over weeks. Review of these EEGs requires special training and experience. Since continuous EEG monitoring use is increasing in neonatal ICUs (NICU), identification of normal and abnormal patterns is vital. Variability in the interpretation of adult and pediatric EEGs has been discussed in literature, with moderate to low interrater reliability, depending on the parameter studied. However, there has been no such study in neonatal EEGs. The goal of this study was to determine the interrater agreement in the interpretation of EEGs of preterm babies.Methods: Continuous EEGs were obtained in preterm infants in the NICU of 2 academic centers as a part of an ongoing NIH-funded study of apnea/bradycardia/desaturation events. These babies underwent continuous EEG monitoring over 9 hours, using the microEEG® device (BioSignal Group) with an electrode cap (Waveguard™ ANT Neuro). The electrode cap consisted of 21 scalp electrodes in the 10-20 international system, with an additional EKG channel. The corrected gestational age at the first recording was between 28 to 32 weeks. Continuous EEG was repeated at 2-4 week intervals, with each neonate receiving 2-4 recordings. For the interrater study, 90-120 minutes of EEG were extracted from the 9 hour recordings. EEGs were reviewed independently by 4 board certified pediatric neurophysiologists. Each EEG was reviewed by 3 readers. The only information provided to the readers was the gestational age at birth and the day of life that the EEG was performed. The readers were instructed to review the EEGs as they would in a clinical setting. They also had to state the level of confidence in the final interpretation as a percentage. Gross percent agreement was determined. Interrater agreement was calculated using aggregated simple (Kc) and weighted (Kw) Kappa measures for nominal and ordinal variables respectively.Results: A total of 75 prolonged recordings were obtained from 24 babies. 17 EEGs were uninterpretable due to continuous artifact. 59 EEGs were reviewed, each EEG being read by 3 readers. Interrater agreement was low to moderate for various categories – Technically adequate/inadequate (76% agreement) – Kc 0.4; Normal/Abnormal/Uninterpretable (65% agreement) – Kc 0.31; Trace discontinu/alternant (82% agreement) – Kc 0.37; Gestational age of EEG (53% agreement) – Kw 0.64; EEG age appropriate/delayed for corrected gestational age (76% agreement) - Kc (–)0.04. Since most of the EEGs were normal, the low kappa values likely reflect the lack of variance and suggest that observed agreement could plausibly almost all be due to chance.Conclusions: The results of this study reveal low interrater agreement for the background and maturity of preterm EEGs, and reiterate the subjective nature of EEG interpretation. Limitations of this study include a large number of normal records, and lack of major abnormalities such as seizures that could highlight the variability of interpretation. Funded by NIH# R41HD072881