ICSD diagnostic criteria for narcolepsy: interobserver reliability.

STUDY OBJECTIVES: To estimate the reliability of the diagnosis of narcolepsy after clinical interview and polysomnographic evaluation among sleep medicine doctors, before and after training in application of the International Classification of Sleep Disorders (ICSD). SETTING: Videotaped semi-structured interviews of 10 patients complaining of daytime sleepiness of different etiologies. Questions referred to ICSD criteria for narcolepsy. A further series of 10 cases of narcolepsy without cataplexy were simulated, with at least a random one to three of the ICSD polysomnographic criteria at pathological levels. PARTICIPANTS AND DESIGN: Seventeen doctors were required to classify each videotaped case as "ascertained," "possible," or "excluded" narcolepsy, in two sessions: one before and one after discussion of ICSD criteria. The observers were invited to confirm or exclude the diagnosis of narcolepsy in the 10 simulated cases, according to the given polysomnographic findings, before and after an agreed proposal of the interpretation of ICSD polysomnographic criteria. Interobserver reliability was calculated using Kappa statistics. INTERVENTIONS: N/A. MEASUREMENTS AND RESULTS: Interobserver reliability of clinical judgement improved from "substantial" at baseline (Kappa 0.61) to "almost perfect" after training (Kappa 0.95). Interobserver reliability of polysomnographic findings was "fair" at baseline (Kappa 0.24), unanimous after the proposed interpretation of ICSD polysomnographic criteria. CONCLUSIONS: Baseline reliability of diagnostic judgement in suspected narcolepsy was found satisfactory among Italian sleep medicine doctors. Educational training, based on discussion of ICSD criteria, further improved agreement. Diagnosis based on polysomnographic findings, not reliable at baseline, needed a strict interpretation of ICSD criteria to attain standardization