Large Language Model–Enabled Health Information Systems: A Systematic Review of Fhir-Based Interoperability, Clinical Decision Support, Privacy, and Trustworthy Ai Governance
Large Language Model (LLM)–enabled Health Information Systems are increasingly transforming how clinical data are captured, exchanged, interpreted, and governed. This systematic review examined recent evidence on the integration of LLMs into Health Information Systems, with emphasis on FHIR-based interoperability, clinical decision support, privacy protection, and trustworthy AI governance. The review adopted a systematic methodology and synthesised studies published from 2020 onward across health informatics, clinical artificial intelligence, digital health, interoperability, privacy-preserving computing, and governance literature. The findings show that LLMs can support health systems by converting unstructured clinical narratives into structured data, improving documentation, assisting medical evidence summarisation, enhancing patient communication, and strengthening decision-support workflows. FHIR and SMART on FHIR emerged as important foundations for standardised and reusable clinical data exchange. However, the review also found that LLM-enabled systems remain limited by hallucination, weak clinical validation, privacy leakage, algorithmic bias, limited transparency, and uncertain accountability. Privacy-preserving approaches such as federated learning, secure computation, encryption, and controlled retrieval offer useful safeguards, but they require strong institutional governance and continuous monitoring. The review concludes that LLM-enabled Health Information Systems should be implemented as human-supervised, standards-based, and governance-driven clinical infrastructures rather than autonomous decision-making tools. Responsible adoption requires FHIR-conformant architecture, clinician oversight, privacy-by-design, fairness evaluation, transparent reporting, regulatory alignment, and lifecycle governance.
Keywords: Large Language Models, Health Information Systems, FHIR Interoperability, Clinical Decision Support, Trustworthy AI Governance
