0291 Evaluating Sleep Quality Metrics Using Zero-Effort Technology: Implications for Public Health Dynamics

Jasleen Kaur,Arlene Oetomo,Vivek Chauhan,Plinio Morita

DOI: https://doi.org/10.1093/sleep/zsae067.0291

IF: 6.313

2024-04-20

SLEEP

Abstract:Abstract Introduction Sleep quality is critical to human health and well-being, with implications for manifold physiological and psychological processes. The quality and reliability of the data due to recall bias and subjective interpretation often limit traditional methods of sleep data collection. This research presents a novel framework that can objectively measure and evaluate sleep quality using smart thermostats equipped with motion sensors, providing non-invasive and effortless sleep monitoring. Methods We leveraged the ecobee 'Donate Your Data' initiative, which collects data from smart home sensors, to analyze 8 Terabytes of data from 178,706 households. In our time-series model, sensor activation values were transformed into signals to model sleep features. We developed a data pipeline integrating data preprocessing, feature engineering, and various machine learning models. These models, such as RNN, VAE, K-means clustering, PCA, and Random Forest classifiers, were used to discern sleep quality indicators. These indicators included Absolute Sleep Duration, Normalized Disturbance Time, Wakeup Onset and Sleep Onset Time, Time in Room, and Sleep Efficiency (with and without onset time) derived from motion sensor data. Results Our findings show three distinct sleep quality clusters, with clear variations in sleep duration, disturbances, and efficiency. Cluster 0 profiled a pattern of fewer disturbances and higher sleep efficiency, whereas Cluster 1 indicated a moderate disturbance with prolonged sleep onset and wakeup duration. The lowest average sleep duration and varied disturbances characterized Cluster 2. Comparative analysis underscores the heterogeneity in sleep quality, highlighting the potential of Internet of Things (IoT) devices in identifying sleep patterns and contributing to sleep research without invasive monitoring. These clusters demonstrate the heterogeneity of sleep quality and showcase the potential of Internet of Things (IoT) devices in sleep pattern identification. Conclusion This novel approach shows smart thermostats as a viable data source for evaluating sleep quality, providing a new perspective on technology in health surveillance and paving the way for personalized sleep improvement strategies for sleep technologists and healthcare policymakers. Furthermore, integrating temperature data from ecobee datasets could deepen insights into the connection between indoor temperature and sleep, impacting health monitoring and policy in the context of increasing global temperatures from climate change. Support (if any)

neurosciences,clinical neurology