Polysomnographic validation of an under-mattress monitoring device in estimating sleep architecture and obstructive sleep apnea in adults

• Published in: Sleep medicine Vol. 96; pp. 20 - 27
• Main Authors: Ding, Feihong, Cotton-Clay, Andrew, Fava, Laura, Easwar, Venkat, Kinsolving, Arthur, Kahn, Philippe, Rama, Anil, Kushida, Clete
• Format: Journal Article
• Language: English
• Published: Netherlands Elsevier B.V 01.08.2022
• Subjects: Obstructive sleep apnea; Polysomnography; Sleep health; Sleeptracker-AI; Validation; Validation; Sleep health; Sleeptracker-AI; Polysomnography; Obstructive sleep apnea
• ISSN: 1389-9457; 1878-5506
• DOI: 10.1016/j.sleep.2022.04.010

The objective of this study is to evaluate the validity of an under-mattress monitoring device (Fullpower Technologies) in estimating sleep continuity and architecture, as well as estimating obstructive sleep apnea in an adult population. Adult volunteers (n=102, 55% male and 45% female, aged 40.6 ± 13.7 years with a mean body mass index of 26.8 ± 5.8 kg/m2) each participated in a one-night unattended in-lab study conducted by Fullpower Technologies. Each participant slept on a queen-sized bed with Sleeptracker-AI Monitor sensors placed underneath the mattress. Standard polysomnography (PSG) was simultaneously recorded on the same night. Researchers (FD and CK) were provided de-identified sleep studies and datasets by Fullpower Technologies for analysis. Sleep continuity measures, 30-s epoch-by-epoch sleep stages, and apnea and hypopnea events estimated by an automated algorithm from the Sleeptracker-AI Monitor were compared with the PSG recordings, with the PSG recordings serving as the reference. Overall, the Sleeptracker-AI Monitor estimated similar sleep continuity measures compared with PSG. The Sleeptracker-AI Monitor overestimated total sleep time (TST) by an average of 6.3 min and underestimated wake after sleep onset (WASO) by 10.2 min. Sleep efficiency (SE) was similar between the Sleeptracker-AI Monitor and PSG (87.6% and 86.3%, respectively). The epoch-by-epoch accuracy of Sleeptracker-AI Monitor to distinguish 4-stage sleep (wake, light, deep, and REM sleep) was 79.0% (95% CI: 77.8%, 80.2%) with a Cohen's kappa of 0.676 (95% CI: 0.656, 0.697). Thirty-five participants (34.3%) were diagnosed with obstructive sleep apnea (OSA) with an apnea-hypopnea index (AHI) ≥ 5 based on PSG. Accuracy, sensitivity, and specificity for the Sleeptracker-AI Monitor to estimate OSA (an AHI ≥5) were 87.3% (95% CI: 80.8%, 93.7%), 85.7% (95% CI: 74.1%, 97.3%), and 88.1% (95% CI: 80.3%, 95.8%) respectively. The positive likelihood ratio (LR+) for AHI ≥5 was 7.18 (95% CI: 3.69, 14.0), and the negative likelihood ratio (LR-) for AHI ≥5 was 0.16 (95% CI: 0.072, 0.368). The Sleeptracker-AI Monitor had high accuracy, sensitivity, and specificity in estimating sleep continuity measures and sleep architecture, as well as in estimating apnea and hypopnea events. These findings indicate that Sleeptracker-AI Monitor is a valid device to monitor sleep quantity and quality among adults. Sleeptracker-AI Monitor may also be a reliable complementary tool to PSG for OSA screening in clinical practice. •An under-mattress sleep monitoring device (Sleeptracker-AI Monitor) was compared with polysomnography (PSG).•Sleeptracker-AI Monitor estimated comparable sleep measures and had high agreement in sleep stage classification with PSG.•Sleeptracker-AI Monitor had high accuracy in estimating obstructive sleep apnea.