The development of Healthcare 5.0 is expected to offer intelligent, interconnected, and personalized medical services based on new digital technologies. But to deliver precision medicine at scale, we need to have collective learning between institutions, while still protecting patient privacy and data provenance. In this work, we introduce a blockchain-empowered FL framework to support privacy-preserving predictive analytics in Healthcare 5.0 networks. The architecture enables multiple providers to collaboratively train machine learning models on decentralized, sensitive patient data without explicitly sharing the data. In other words, thanks to blockchain, immutable audit trails, decentralized trust management, secure updating of models are ensured among the participants. We test our framework on large-scale multi-institutional synthetic datasets created to emulate predictive tasks such as early detection of disease and personalized treatment recommendations. Experimental results show that our blockchain-empowered FL model achieves high predictive accuracy, effective cooperation, and robust security with the capability of handling adversarial attacks while satisfying the regulatory and ethical demands of data privacy. The work and contributions are made towards scalable, trustworthy, and intelligent health care analytics for next-generation precision medicine.