Improving predictive accuracy of a survey measure of risk for narcolepsy

Abstract

Narcolepsy is a brain disorder that may go unrecognized and untreated for many years. The ability to use easily obtained survey information about symptoms of narcolepsy would facilitate identification of individuals potentially at risk for narcolepsy who could be referred for further testing. The purpose of the present study was to explore whether a survey instrument could successfully distinguish narcolepsy from other sleep disorders using data that could easily be obtained from a community or general patient sample. The hypothesized model added the Epworth Sleepiness Scale to a narcolepsy symptoms checklist to explore whether it improved accuracy of classification. Data related to symptoms were extracted from medical records of patients with a known diagnosis of narcolepsy, obstructive sleep apnea, or insomnia. The sample was then randomly split in half, allowing exploratory and confirmatory binary logistic regression. Adding the Epworth Sleepiness Scale score to the original list of symptoms more accurately classified those with or without narcolepsy. Although these findings require additional testing before they can be confirmed and generalized, they suggest that a self-report screening instrument for narcolepsy with acceptable accuracy is possible.