Inter-observer variability of the EEG diagnosis of seizures in comatose patients

OBJECTIVE To assess the inter-observer agreement of the electroencephalogram (EEG) diagnosis of (non-convulsive) seizures in comatose patients. DESIGN/SETTING/PATIENTS Nine clinicians with different levels of experience in clinical neurophysiology were asked to evaluate in a strictly controlled way 90 epochs (10s each) of 30 EEG's of 23 comatose patients admitted to the intensive care unit (ICU). For each EEG clinicians had to decide whether there was an electrographic seizure or not. Furthermore, Young's EEG criteria for (non-convulsive) seizures were scored in detail for all EEG's. Agreement was determined by calculating kappa values. RESULTS The inter-observer agreement of an EEG diagnosis of seizure was limited. The overall kappa score for the five experienced raters was 0.5, and the kappa score for less experienced raters was 0.29. Kappa values for the individual Young's criteria were highly variable, indicating discrepancies in the interpretation of specific phenomena. Especially, some types of periodic discharges gave rise to different interpretations. CONCLUSIONS The EEG diagnosis of (non-convulsive) seizures in ICU patients is not very reliable, even when strict criteria such as proposed by Young are applied. There is a need for less ambiguous EEG criteria for (non-convulsive) seizures and status epilepticus.