Two measures of reliability assessed inter-rater agreement regarding n
eonatal seizure scoring: coefficient of variation and intraclass corre
lation coefficient. Ten neonates with electrographic seizures recorded
during each of 2 h of an EEG-sleep study were submitted for visual an
alysis. The onset, spread, and termination of seizures were coded. The
number of anatomical sites with electrographic seizures as well as th
e durations for each of three phases of the seizure were calculated. S
even EEG scores for these 10 patients were coded independently by thre
e different raters. The number of sites at the onset of each seizure a
nd the duration of seizures, both at the onset and for the entire dura
tion of the seizure, had moderate-to-excellent reliability. Lower reli
ability, however, was noted for the number of seizures during the phas
es of spread and termination. Before the development of computerized p
rograms to detect and document the evolution of neonatal seizures, par
ticularly in the context of treatment with different medications, prac
tical strategies based on visual analysis need to be developed. Althou
gh the onset and total duration of seizures can be reliably scored amo
ng raters, the more variable description of spread and termination of
the seizure is more problematic and requires greater attention and exp
erience. This technique may be useful in evaluating the medication eff
icacy.