CLEAR-Arch: Consolidated LLM Extraction and Archival Reporting for Health Records

Introduction Digitized archives and health records offer growing opportunities for research, but approaches to extracting unstructured data remain heterogeneous. Large language models (LLMs) and generative artificial intelligence (GenAI) may facilitate extraction and synthesis, yet reporting guidance remains limited. This study develops CLEAR-Arch, a preliminary framework for transparent and reproducible reporting of LLM-assisted research. Method A scoping review–informed conceptual synthesis identified methodological and reporting considerations relevant to AI-assisted extraction, archival research, health informatics, and evidence synthesis. Candidate items were consolidated into reporting domains and used to develop performance indicators and a preliminary checklist. Results CLEAR-Arch proposes domains covering model specification, prompt design, document processing, extraction procedures, human validation, auditability, and data governance. Candidate indicators include precision, recall, F1-score, contextual accuracy, temporal fidelity, source-groundedness, inter-coder reliability, cross-source consistency, validation accuracy, and auditability. Conclusions CLEAR-Arch provides a preliminary, literature-informed framework for improving transparency, reproducibility, and methodological rigour in LLM-assisted research. Its domains, indicators, and reporting items are provisional, not established standards. Empirical validation, stakeholder consultation, content-validity assessment, and formal consensus development are needed before it can be considered standardized reporting guidance.