A study of inter-reviewer reliability of attacks recorded on ambulatory EEG

Eleven ambulatory EEG (A/EEG) cassette tapes were sent to 5 reviewers from different EEG laboratories in the U.K. Each tape included one attack recorded from a different patient. Attacks were chosen to include various possible EEG changes that can occur during attacks on A/EEG. Reviewers were asked to give details of EEG changes seen during attacks and to state whether they thought attacks were epileptic. Total agreement about the presence of abnormality was found on 2 tapes. Four out of 5 reviewers agreed on 4 tapes and on the remaining 5 tapes only 2 or 3 reviewers were in agreement. It was concluded that the amount of agreement amongst reviewers was poor but reasonable in view of the degree of difficulty of the tapes which were not a random sample. Compared to the final clinical opinion of referring consultants, these findings would have resulted in 16% false negatives, 7% false positives and 13% of cases in which no decision was made.