Validation of self-reported epilepsy for purposes of community surveillance

Abstract

We evaluated the validity of questions designed to identify lifetime and active epilepsy, medication use, and seizure occurrence on population-based surveys. Subjects were interviewed by telephone, and responses were compared with information in their medical records. Prevalence, sensitivity, specificity, and positive predictive value (PPV) were calculated. The prevalence of ever having been diagnosed with epilepsy was 3.1% by self-report and 2.7% by medical record review. Sensitivity was 84.2%, specificity was 99.2%, and PPV was 73.5% for self-reported lifetime epilepsy, and values were similar for active epilepsy. By comparison, sensitivity was higher and specificity was lower for epilepsy medication use and seizure occurrence. The PPV for seizure occurrence was substantially higher for a recall period of 12 months than for 3 months. These results compare favorably with results for other chronic conditions, such as diabetes and arthritis, and indicate that questionnaires can be used to identify epilepsy at a population level.