A Qualitative Requirements Analysis for the Representation of Clinical Guidelines

Clinical Guidelines (CPGs, CGs) are mostly static, narrative documents, which may cause usability issues at the point of care. Although computable CG standards are available, major publishing portals do not publish them alongside CGs. This study aims to provide a requirements analysis to formalize and visualize CGs, based on (1) a structured analysis of CGs sampled from four major CG portals, and (2) semi-structured qualitative interviews with clinicians from different medical specialties. The study seeks to identify a consolidated set of requirements to inform future efforts in CG formalization and design of CG-based CDSSs.

The structural document analysis included 20 full-version CGs obtained from AWMF (S3-level), NICE, SIGN and WHO. The extracted data covered structural features on the document and recommendation level and were synthesized descriptively. Interviews were conducted via video call (25-35 min) with 8 physicians from the Klinikum Rechts der Isar (TUM), guided by a 14-question interview guide. Participants were purposefully recruited from a range of clinical specialties and seniority levels. Whisper (OpenAI) wasused as a transcription tool, with subsequent human correction. Analysis was supported by Microsoft Copilot (Enterprise), via a structured prompt developed by the team. All outputs were reviewed by humans, with 3 quality checks applied.

All four portals publish CGs as static narrative documents with no computable representations, spread across different documents. Interview analysis produced 45 categories in five thematic clusters, with four of them reported consistently: outdated or conflicting versions, insufficient specificity, inadequate or missing patient-specific applicability, and evidence gaps for rare diseases. Clinician challenges correspond directly to the structural analysis' findings. 22 requirements (functional, structural and visual) were derived.

The 22 requirements based on two independent evidence sources provide a validated foundation for CG formalization and visualization.