Abstract
Monitoring sleep status in psychiatric settings is crucial. However, psychiatric symptoms and cognitive impairments complicate traditional sleep assessments, such as polysomnography (PSG). To address this, we employed Nemuri SCAN (NSCAN, Paramount Bed Co. Ltd.), a contact-free patient sensor, and compared its performance with PSG in patients with psychiatric disorders. This cross-sectional study included 29 cases (median age: 61 years; 55.2% male) from August 2021 to January 2023. NSCAN showed lower specificity than PSG, often misclassifying still wakefulness as sleep. To improve this, we developed a logistic regression model named the Patient-Adjusted Cole Model (PAC Model), which incorporates 10 patient characteristics into the NSCAN decision algorithm based on the Cole–Kripke equation (Cole model). The agreement with PSG, sensitivity, and specificity were 77.8%, 97.3%, and 28.2% for the Cole model and 78.8%, 94.5%, and 38.9% for the PAC Model, respectively, where agreement represented the percentage of sleep/wake determinations by NSCAN that matched those by PSG. While sensitivity was slightly lower in the PAC Model, specificity improved notably, addressing a critical limitation of non-contact sensors. These findings highlight the importance of integrating patient characteristics into sleep monitoring algorithms to enhance the practicality and utility of NSCAN in psychiatric care.
