Background: Although portable polygraphy or portable monitoring of respiratory parameters (PM) is commonly used to confirm obstructive sleep apnoea syndrome, agreement on apnoea hypopnoea index (AHI), the main measure of disease severity, has not been evaluated. The aim of this study was to assess the agreement on AHI among multiple observers as well as between individual observers and automated analysis.
Methods: A total of 88 ambulatory sleep recordings ("Embletta") were independently scored by 8 physicians (observers). Agreement on AHI, using intraclass correlation coefficient (ICC), was measured among observers. Bland Altman plots were built to compare individual observers with PM.
Results: Among observers, ICCs were .73 for agreement on AHI, .71 for hypopnoea index and .98 for desaturation index. Compared to visual analysis, automated analysis underestimated AHI by 5.1 events on average. When comparing individual observers with automated analysis, systematic bias varied from -1. to +1 .5 events/h on AHI.
Conclusions: Among observers who used PM in a clinical setting, agreement on AHI was limited. When automated and individual visual analyses were compared, the systematic bias varied from almost zero to values sufficient to affect clinical diagnosis. Much of the discordance was due to different counts of hypopnoea, whereas agreement on apnoea and desaturation index was better. Efforts should be directed towards standardisation of visual analysis, improvement and quality control of ambulatory sleep studies.