<p>With the acceleration of globalization, public health issues have become global challenges, including cross-border transmission of infectious diseases and global spread of non-communicable diseases. The development of federated learning and large language model technologies offers new possibilities for addressing these challenges. This article, which introduces global public health challenges and the rise of federated learning and large language model technologies, and explores the potential of these technologies in disease control and proactive health management. This article reviews the key technical foundational theories of federated learning, including sensor networks and data collection, wireless communication technologies, edge computing and cloud computing collaboration, and large language model training. In the case analysis section, this article deeply analyzes the typical applications of federated learning, large language models, and Internet of Things technologies in proactive disease early warning in health management, and discusses their potential advantages and challenges. Finally, this article proposes a multi-dimensional perception and multi-point triggered early warning system framework based on federated learning and multi-modal continuous learning models. This system conducts multi-level information processing through Internet of Things devices, federated learning servers, and large language models to achieve precise disease early warning and dynamic resource scheduling. The research in this article provides theoretical support and technical paths for intelligent public health systems based on federated learning, helps improve the efficiency of disease prevention and health management, and provides technical support for the optimization allocation of global medical resources, precision medicine, and public health security.</p>