Predicting Sleep Quality in Family Caregivers of Dementia Patients From Diverse Populations Using Wearable Sensor Data.

Abstract

This study aimed to use wearable technology to predict the sleep quality of family caregivers of people with dementia among underrepresented groups. Caregivers of people with dementia often experience high levels of stress and poor sleep, and those from underrepresented communities face additional burdens, such as language barriers and cultural adaptation challenges. Participants, consisting of 29 dementia caregivers from underrepresented populations, wore smartwatches that tracked various physiological and behavioral markers, including stress level, heart rate, steps taken, sleep duration and stages, and overall daily wellness. The study spanned 529 days and analyzed data using 70 features. Three machine learning algorithms-random forest, k nearest neighbor, and XGBoost classifiers-were developed for this purpose. The random forest classifier was shown to be the most effective, boasting an area under the curve of 0.86, an F1 score of 0.87, and a precision of 0.84. Key findings revealed that factors such as wake-up stress, wake-up heart rate, sedentary seconds, total distance traveled, and sleep duration significantly correlated with the caregivers' sleep quality. This research highlights the potential of wearable technology in assessing and predicting sleep quality, offering a pathway to creating targeted support measures for dementia caregivers from underserved groups. The study suggests that such technology can be instrumental in enhancing the well-being of these caregivers across diverse populations.