How often does routine pediatric EEG have an important unexpected result?

Background: Electroencephalogram recordings are requested for the assessment of many childhood disorders. To assess the utility of the EEG in children, we studied how often routine EEG results can be correctly predicted from the EEG requisition. Method: Five hundred consecutive initial EEG requests from the IWK Grace Health Centre from two time epochs were examined. All EEGs were 16 channel (10-20 electrode system). Based only on the requisition (patient demographics, referring physician, and reason for EEG), we coded our prediction of the result and then the actual result. When results were discordant from prediction, a judgment was made about the potential importance of the result. Results: Overall, EEG results were correctly predicted in 81%. Prediction for all nonepilepsy reasons was accurate in 91% (n=320) and 96% for paroxysmal nonepileptic events (n=158) but only 59% for epileptic disorders (n=141) (p<0.0001). Neurologists ordered 45% of EEGs, pediatricians 32%, and GP's 17%. Predictions were least accurate for neurologists' requests (p<0.006) however, neurologists were more likely to request EEG for epileptic disorders (p<0.0001), Age of the child and urban versus rural address did not affect the accuracy of prediction. Conclusion: Results of routine pediatric EEG for most nonepilepsy reasons appear highly predictable and therefore, possibly of little value to an experienced clinician. When requested for epilepsy, this "ancient" test remains full of surprises.