Advancements in AI, Internet of Things (IoT) communication standards and digital twin development have led to a new field of engineering driven by the design and deployment of Cyber-Physical Systems (CPSs). In CPSs, devices monitor assets that generate readouts that are transmitted by means of IoT protocols to AI applications typically running on a cloud infrastructure to extract knowledge. This topology is based on the computational limitations of the smart devices that prevents them from running complex AI applications. The conversion of information into knowledge is common to all CPSs. In real-life scenarios, several initiatives ranging from Industry 4.0 to Home Automation have helped lower costs, improve safety and provide a better use of the available resources. This paper addresses one of these initiatives known as Connected Health. It introduces a CPS that enables the prediction of real-time conditions associated with emergency medicine, ranging from cardio-respiratory to neurological problems, in the context of Active Assisted Living (AAL). Specifically, this system integrates a large number of static and dynamic sensors, including wearables and streamers that are fed into a Machine Learning (ML) predictive model to estimate the potential for emergency medical conditions affecting senior citizens. In the context of network loss and latency, the system dynamically selects levels of Forward Error Correction (FEC) to guarantee that the most relevant information arrives at the predictor.