Interrater reliability of the multiple sleep latency test

The purpose of this study was to evaluate interrater reliability in the interpretation of the multiple sleep latency test (MSLT). We prospectively analyzed MSLTs performed on 21 patients with excessive daytime sleepiness. MSLTs were recorded on Grass Model 78 polygraphs with EEG, electro-oculogram, and chin EMG. Each test was performed simultaneously at paper speeds of 10 and 30 mm/sec and was scored blindly by 3 readers using standard criteria. For the quantitative variable (sleep latency), a LISREL model was used. For the binary variable (REM present or not), a kappa coefficient was used. Interrater reliability of sleep latency was 0.850 at speed 10, and 0.884 at speed 30. There was no significant difference between speed 10 and 30. Interrater reliability for the presence or absence of REM was 0.515-0.563 at speed 10, and 0.447-0.525 at speed 30. On the MSLT, the estimation of sleep latency showed excellent consistency between different readers. The determination of the presence or absence of REM only showed fair to good agreement among observers. There was no significant difference between a paper speed of 10 vs. 30 mm/sec.