Design of an IoT-Based Monitoring Sensor Network and Preliminary AI-Driven Data Analysis for Health Measurement in Residential Care Homes

Monitoring health and critical events in real time is crucial to ensuring well-being and enabling timely emergency response in Residential Care Homes. Despite the availability of numerous technological solutions, their application in real-life care environments remains limited. The Smart-RSA project addresses these challenges by developing an integrated, non-invasive Internet of Thing (IoT) monitoring system that combines environmental sensors, activity tracking technologies, and clinical-grade medical devices. Designed through a user-centered approach involving caregivers, healthcare professionals, and researchers, the system has the ambition to support four use cases: fall detection and prevention, nighttime monitoring, continuous vital signs and health monitoring, and, consequently, caregiver assistance. This paper presents the design methodology, IoT sensor network architecture, and a preliminary clinical data processing pipeline. A multilayer sensor infrastructure was deployed in a pilot facility, incorporating home automation devices, smart mattresses, RGB cameras, and health sensors for real-time monitoring. To assess the feasibility of health status classification, a stacked ensemble Machine Learning (ML) model (Random Forest, Decision Tree, Gradient Boosting) was trained and tested on the MIMIC-IV Emergency Department dataset, achieving an accuracy of 94.7%. These results validate the model's effectiveness in classifying clinical acuity levels based on vital signs. The Smart-RSA system has the potential to transform care delivery in long-term care facilities by enabling proactive, personalized, and data-driven care strategies.