O036 Performance characteristics of a contactless under-mattress sleep analyser in a clinical sleep laboratory: Are automated sleep analysers ready for prime time?

Abstract Introduction Contactless under-mattress sleep analysers which estimate sleep architecture and breathing disturbances have the potential to alleviate wait times for diagnostic sleep studies and enable multi-night sleep assessments. Objective To evaluate the performance characteristics of a commercially available pneumatic under-mattress sleep analyser (UMSA) during overnight laboratory-based polysomnography (PSG) in an unselected sleep clinic population. Methods To date, simultaneous UMSA and PSG recordings have been obtained in 74 individuals. After removing studies with insufficient sleep duration (i.e., Total Sleep Time (TST) <5-h, n=7) or recording discrepancies (n=12), TST, rapid-eye movement (REM) duration and apnoea-hypopnea index (AHI) were compared between UMSA and PSG. Results Average TST (416±56 vs 355±7min, p<0.001) and REM duration (83±48 vs 60±34min, p=0.006) were significantly longer for UMSA than PSG, while mean AHI (19.4±24.2 vs 16.3±21.6 events/h, p=0.495) was similar. Mean AHI difference between UMSA and PSG was 3.1 events/h with 95% limits of agreement between -23.6 to 29.8 events/h. The sensitivity, specificity, and area under the receiver operating curve (AUC) at a threshold of AHI≥5 was 78.6%, 56.5% and 0.744AUC. At a threshold of AHI≥15 the sensitivity, specificity, and AUC were 80.0%, 69.4% and 0.902AUC. Conclusions Estimates of TST and REM duration from an under-mattress sleep analyser were not comparable to laboratory-based polysomnography. Mean AHI was similar albeit with large limits of agreement and the under-mattress sleep analyser had moderate diagnostic accuracy to detect an AHI>5 when compared to polysomnography. Caution is required in the application of under-mattress sleep analysers for research and clinical purposes.