Abstracts

Rationale: We designed and validated a 2-stage parent questionnaire to identify seizures in children at risk for epilepsy (11 question screen followed by 29 confirmatory questions). We sought to understand the contribution of individual questions or sets of questions to the performance of the questionnaire in its ability to distinguish a paroxysmal event as seizure from non-seizure.Methods: For all participants with paroxysmal events, we compared responses in the confirmatory portion of the questionnaire to the pediatric neurologist s clinical diagnosis. Factor analysis was carried out on the 29 confirmatory algorithm variables. Sensitivity and specificity of the extracted factors and individual variables comprising the factors were determined. Additionally, we evaluated the impact of participant gender, age of onset, event frequency, and time to last episode on the sensitivity and specificity using Chi square.Results: Of 152 participants, 51 recorded no events. The remaining 101 children had 144 events (range: 1 4/ participant). Seven factors (sets of items) explained 71% of the variance in the data. Sensitivity for the individual factors ranged from 59% (focal jerking, drooling and gurgling during events) to 99% (behavioral arrest and postictal lethargy), and specificity ranged from 26% (events occurring only when upset, mad, or in pain) to 80% (focal jerking, drooling and gurgling; figure). Combining certain factors improved sensitivity while reducing specificity. Sensitivity and specificity for individual questions varied widely. Neither sensitivity nor specificity was associated with participant gender, age of onset, event frequency, or time since the last episode.Conclusions: Seven latent variables (factors) were extracted from the confirmatory items of the questionnaire. Although certain factors and individual questions had very high sensitivity or specificity, none attained the combined sensitivity and specificity values of the overall algorithm. Use of multiple differing questions to address a concept (e.g., postictal symptoms or motor characteristics) improved sensitivity. Questions about alteration of awareness and postictal lethargy had a higher sensitivity for seizures than questions about motor characteristics. Based on this factor and item analysis, refinement of the questionnaire to improve efficiency and strike a balance between sensitivity and specificity is possible.