Abstracts

Rationale: Epilepsy is the most common serious non-infectious neurological condition worldwide. Physical examination rarely contributes to the diagnosis and the interictal EEG is commonly not diagnostic. Patient interview, on the other hand, is usually essential in making an accurate clinical diagnosis. A detailed clinical history should be amenable to a standardized questionnaire approach. A structured 2-part parent questionnaire was developed to simulate the logic process that an expert child epileptologist might follow in making a seizure diagnosis. It details the various clinical characteristics associated with seizure-like events. We sought to validate this parent questionnaire in a population of children enriched for neurological morbidities.Methods: English-speaking parents of participants aged 6 months to12 years were consented in the Division of Pediatric Neurology at Boston Medical Center or one of its affiliated outpatient clinics. Parents were interviewed using the structured computerized questionnaire we developed. A computer-driven algorithm rendered a 4-level ranking scale of seizure probability for each patient and each event experienced by the patient: 1) not likely seizure, 2) indeterminate for seizure, 3) probable seizure, 4) almost certain seizure. Blinded to the questionnaire results, the treating pediatric neurologist independently ranked each patient on the same scale based on their history and examination. The performance of the seizure questionnaire relative to the neurologist's diagnosis was described through sensitivity and specificity, with 95% confidence intervals (CI). Sensitivity and specificity was compared across patient subgroups through the chi-square test.Results: 150 of 177 (84.7%) enrolled parents completed the questionnaire. The prevalence of seizures in this enriched sample was 39.3% based on the clinician s diagnosis. The seizure questionnaire yielded a fitted Receiver Operating Characteristic (ROC) area of 0.93 (95%CI 0.89-0.97). Optimal sensitivity and false positive fraction is at the cutoff rating of 3 for true positives; thus, we dichotomized the questionnaire results as consistent with seizure (levels 3 and 4) or without seizure (levels 1 and 2). Overall, findings included a 91.5% sensitivity (95%CI 84.4%-98.6%), and an 83.5% specificity (95%CI 75.9%-91.1%). The Positive Predictive Value was 78.3% (95%CI 68.5%-88.0%) and Negative Predictive Value 93.8% (95%CI 88.6%-99.1%). The questionnaire had higher specificity when administered in-person (n=44, sensitivity 100%, specificity 93.1%) compared to its administration over the phone (n=42, sensitivity 88.9%, specificity 70.8%, p=0.03).Conclusions: The 2-part seizure questionnaire developed to identify childhood seizures achieved high levels of sensitivity (91.5%) and specificity (83.5%) when compared to expert Child Neurology clinical diagnosis. In-person interview responses were more specific than those conducted by telephone interview. We aimed to enroll 250 participants in this validation study; therefore, this report represents an interim set of results.