Key points are not available for this paper at this time.
Expanding recognition of sleep’s importance has created a vast commercial market for sleep monitoring devices. Device reliability is generally presumed, despite low performance: typically, high sleep-detecting sensitivity (≥95%) but poor wake-detecting specificity (≤40%). To describe the validity of a novel device and illustrate the impact of algorithm changes in a fast-adapting market, we compared versions 1 (V1) and 2 (V2) of the S+ by ResMed bedside monitor against PSG. Healthy adult sleepers underwent standard PSG, time-synchronized with the non-contact bedside device. Epoch-by-epoch V1 (N=27) and V2 (N=22) validity, and within-subject changes from V1 to V2 (N=22), were tested. Subjects were 41% female, 97% Caucasian, 15% married, aged 29. 1 (±12) years with a BMI of 27 (±6) ; they had 16 (±3) years education and median income 65, 000. Total sleep time TST per PSG was 338 (±57) minutes with normal sleep architecture. Full sleep staging agreement of V1 was 61 (±9) %, while V2 was 62 (±7) %. For sleep/wake comparison, sleep sensitivity of V1 and V2 were 93 (±6) % and 94 (±4) %, while wake specificity of V1 and V2 were 70 (±19) % and 73 (±20) %. Specificity of V1 and V2 for WBSO were 88 (±16) % and 90 (±15) %; for WASO they were 51 (±23) % and 53 (±22) %, respectively. Analysis of within-subject changes follows: overall sleep sensitivity of V2 was significantly lower (p=. 026) ; WBSO specificity did not differ; V2 WASO specificity was significantly higher (p=. 022). Stage-specific results and an actigraphy comparison will be presented at the meeting. Relative to other published evaluations of commercially available, wearable sleep-tracking devices, this bedside device better identifies WBSO - a major challenge in this industry. There is still room to improve WASO specificity of this and all commercial sleep trackers. 100% accuracy is an unrealistic goal; rather, devices should approach PSG human inter-scorer reliability of ~81% for wake specificity. Equipment, supplies, participant compensation, and student stipends were provided by ResMed. The faculty supervisor received no salary support or compensation from ResMed.
Schade et al. (Fri,) studied this question.