Predictive modeling has emerged as an essential tool in disease prevention, enabling healthcare providers to proactively identify patients at risk and implement timely interventions. However, conventional predictive models tend to be centralized, necessitating the aggregation of sensitive patient data from various sources, which raises significant privacy issues and restricts data availability. Federated learning presents a transformative solution by facilitating collaborative model training across decentralized data silos while ensuring that patient data remains secure and local. This paper investigates federated learning in disease prevention, emphasizing its potential to improve predictive accuracy, safeguard patient privacy, and enhance real-time healthcare applications. The study begins with a thorough overview of predictive modeling techniques, followed by a few case studies, with a particular focus on federated learning as a pivotal innovation in this field. We explore federated learning architecture and the challenges associated with federated learning. The paper concludes with insights into the future of federated learning in healthcare, highlighting the necessity for advanced privacy-preserving techniques, standardized model evaluation protocols, and interdisciplinary collaborations.