Abstracts

Rationale: Methods of assessment of fellowship training in clinical neurophysiology often poorly duplicate clinical practice and do not explicitly account for disagreements in expert interpretation. The purpose of this study was to use inter-reader reliability methods to assess expert and trainee agreement in interpretation of a sample of EEG recorded from term neonates. Methods: Two board-certified experts and two trainees (A=at end of fellowship, B=beginning of fellowship) independently ranked 1 hour long EEGs that were recorded from term infants with presumptive hypoxic-ischemic encephalopathy between 1-3 days of life according to presence of interictal epileptiform discharges, presence of epileptic seizures, and a 4 point encephalopathy severity scale. Inter-reader agreement and reliability, the latter measured via the kappa statistic, were measured for each category across the experts. The 95th confidence limit of kappa was selected for a threshold level by which the inter-reader reliability of each trainee was measured. Results: Agreement rates/reliability(reliability 95th confidence interval) between the two experts was IED=0.780/.560(0.283-0.837), seizures = 0.940/0.694(0.416-0.971), overall severity = 0.800/0.722(0.571-0.873). Reliability of Fellow A exceeded the lower confidence limit for all 3 criteria; reliability of Fellow B exceeded the threshold value in only seizure detection. The lower confidence limit of kappa varied linearly with the number of studies. Conclusions: Inter-reader agreement and reliability in interpretation of conventional multichannel EEG of term neonates is high (but less reliable in the designation of pathophysiological interictal epileptiform discharges). The use of fewer studies in this technique is feasible. Reliability measurements may provide a practical and clinically relevant means of interim and final fellowship assessment.