A Precision Health Initiative for Chronic Conditions: Design and Cohort Study Utilizing Wearable Technology, Machine Learning, and Deep Learning

Abstract

Abstract:  This study introduces a comprehensive precision health service designed for health promotion and prevention of chronic diseases. It involves continuous real-time monitoring of lifestyle and environmental factors through the integration of wearable devices, open environmental data, indoor air quality sensors, a location-based smartphone application, and an AI-assisted telecare platform. The AI-assisted telecare platform offers detailed insights into patients' clinical, lifestyle, and environmental data, facilitating reliable predictions of future acute exacerbation events. Over a 24-month follow-up period, data from 1,667 patients were prospectively collected, leading to the identification of 386 abnormal episodes. Machine learning and deep learning algorithms were employed to develop modular chronic disease prediction models, which underwent external validation. These models successfully predicted obesity, panic disorder, and chronic obstructive pulmonary disease with an average accuracy of 88.46%, sensitivity of 75.6%, specificity of 93.0%, and an F1 score of 79.8%. By incorporating objective comprehensive data and feature selection, this study improved prediction model performance compared to previous approaches. It highlights the strong correlation between lifestyle, environmental factors, and patient health, suggesting their potential for more accurate prediction of future abnormal events compared to using only questionnaire data. Additionally, the study demonstrates the feasibility of constructing cost-effective prediction models with minimal features, enhancing their applicability in real-world settings.