Creating an algorithm to identify indices of sleep quantity and quality from a wearable armband in adults

Objective: To develop an algorithm to quantify indices of sleep quantity and quality using the SenseWear armband (SWA) and to compare indices of sleep from this novel algorithm to standard wrist actigraphy (Actiwatch 2; AW2) under free-living conditions.

Material and methods: Thirty participants (47±10 years; 33.0±4.8kg/m²) wore the SWA and AW2 for seven consecutive days. Participants self-reported bedtime and waketime across these 7 days. Bedtime, sleep onset, sleep offset, waketime, total sleep time (TST), time in bed (TIB), sleep effciency (SE), sleep onset latency (SOL), wake after sleep onset (WASO), sleep fragmentations (SF), sleep regularity (calculated as SD of waketime), and mid-point of sleep were calculated using each device.

Results: There was significant evidence for equivalence of means (or mean ranks) for bedtime, sleep onset, sleep offset, waketime, TST, TIB, SOL, WASO, and midpoint of sleep measured by the SWA and AW2 (p<0.05). There was insuffcient evidence for equivalence of means in SF (SW: 25±6 vs. AW2: 10±3 events; p=1.0), mean ranks in sleep regularity (SW: 58±33 vs. AW2: 68±40 min; p=0.11), and mean ranks in SE (SW: 84.7±5.1% vs. AW2: 86.3±5.5%; p=0.05). When comparing minute-by-minute sleep/wake status, the sensitivity and specificity of the SWA were 0.94 (95%CI: 0.93, 0.95) and 0.88 (95%CI: 0.85, 0.90), respectively, using AW2 as the criterion measure.

Conclusion: The algorithm developed for the SWA produced relatively accurate and consistent measurements of sleep quantity, timing, and quality compared to the AW2 under free-living conditions. Thus, the SWA is a viable alternative to standard wrist actigraphy.