The Epworth sleepiness scale: Reliably unreliable in a sleep clinic population

Abstract

The Epworth Sleepiness Scale is used frequently to measure excessive daytime sleepiness in research and clinical settings, although there is limited evidence on test–retest reliability, particularly among sleep clinic populations. The objective of this study was to evaluate the reliability of this instrument among adult patients recruited from a public hospital sleep clinic in Sydney, Australia. English‐speaking participants self‐completed the Epworth Sleepiness Scale on two occasions, at the specialist clinic visit and on the night of diagnostic polysomnography. Of the 108 participants included in the study, the majority were male (64%) and the mean age was 51 years. The median retest interval was 64 days. The primary outcome of test–retest reliability as measured using the intraclass correlation coefficient was 0.73 (95% confidence interval, 0.61–0.82). Despite moderate statistical reliability and a low mean difference of 1.1, Bland‐Altman analysis showed an unacceptably wide distribution of between‐score differences. The 95% limits of agreement were −8.5 to +10.6, and an absolute difference in scores of at least 3 was observed in 60 (56%) of the participants. Our results suggest that the Epworth Sleepiness Scale should not be used in clinical settings to make individual‐level comparisons, such as the effect of therapeutic interventions, or to prioritise access to services.

Low test–retest reliability of the Epworth Sleepiness Scale within a substantial short time frame

“…Correlations of the two score sets were then calculated and a high correlation was found. In contrast, other examinations failed to prove these ideal results over subsequent decades (Bourke, 2004; Knutson, 2006; Nguyen, 2006; Lee, 2020). Notably, the results varied from poor to fair.…”

“…Former studies using test–retest methods (Bourke, 2004; Johns, 1991; Knutson, 2006; Lee, 2020; Nguyen, 2002) applied different intervals between the two test sessions. The interval varied from 70 days to 6 months.…”

“…Compared to the statistical methods used to characterize the test–retest reliability of the ESS, we assumed the differences of the analyses may have contributed to the heterogeneous result of the reproducibility. The applied statistical method varied from the correlation analysis (Johns, 1991) to the intraclass correlation coefficient (Lee, 2020) and Lin's concordance coefficient (in our study). To compare our results with the first test–retest examination of the ESS by Johns (Johns, 1991), we calculated Pearson's correlation coefficient for our datasets, which showed good reliability.…”

“…Stability and similarity of the circumstances during the test and retest periods may increase the reliability of the applied questionnaire. All of the former examinations applied a relatively long inter‐investigation time (weeks to months) (Aloe, 1997; Bourke, 2004; Johns, 1991; Knutson, 2006; Lee, 2020; Nguyen, 2002). During these intervals, there are possibilities of rudimentary changes of the test population, which may negatively affect the test results.…”

“…During these intervals, there are possibilities of rudimentary changes of the test population, which may negatively affect the test results. Recently, Lee and coworkers’ (Lee, 2020) results suggested the Epworth Sleepiness Scale is not useful in clinical settings, in the evaluation of a therapeutic intervention or to use (prioritize) during the access to a service.…”