Objectives: Variables used in prediction rules and clinical guidelines should show acceptable agreement when assessed by different observers. Our objective was to determine the interobserver agreement of patient history and physical examination variables used to assess children undergoing emergency department (ED) evaluation for a first seizure not provoked by a known precipitant such as fever or trauma (ie, an unprovoked seizure).
Methods: We conducted a prospective cohort study of children aged 28 days to 18 years evaluated for unprovoked seizures at 6 tertiary care EDs. We excluded patients if previously evaluated for a similar event. Two clinicians independently completed a clinical assessment before neuroimaging. We determined agreement for each clinical variable by using the unweighted κ statistic.
Results: A total of 217 paired observations were analyzed; median patient age was 53.5 months, and 38% were younger than 2 years. Agreement beyond chance was at least moderate (κ ≥ 0.41) for 21 of 31 (68%) variables for which κ could be calculated. κ was ≥0.41 for 7 of 11 (64%) general history variables, all 8 seizure-specific history variables (including seizure focality), and 6 of 12 (50%) physical examination variables. Agreement beyond chance was substantial or better (κ ≥ 0.61) for 2 of 11 (18%) general history variables, for 5 of 8 (63%) seizure-specific history variables, and for 2 of 12 (17%) physical examination variables.
Conclusions: For children with first unprovoked seizures evaluated in the ED, clinicians frequently assess findings from seizure-specific history with substantial agreement beyond chance. Those clinical variables that have been associated with the presence of intracranial abnormalities and show reliability between assessors, such as seizure focality and the presence of any focal neurological finding, may be more useful in the ED assessment of children with first unprovoked seizures.