<![CDATA[Background: Radiology reports increasingly enter patient-facing digital health environments, yet their technical vocabulary, compressed syntax, and implicit diagnostic reasoning often exceed lay comprehension, particularly in multilingual Indonesian healthcare contexts. Objective: This study aims to compare how expert- and AI-generated simplifications transform Indonesian radiology-related information into patient narratives across readability, semantic fidelity, and narrative-pragmatic alignment. Method: A comparative qualitative-dominant corpus design was applied to 68 coded meaning units derived from 17 publicly verifiable radiology-related documents, with paired expert and AI simplifications analysed through health-literacy, clinical-risk, and patient-centred discourse matrices. Results: Findings show that AI-generated simplifications more frequently improved lexical accessibility, direct patient address, explanatory sequencing, orientation sentences, and actionable guidance. Expert-generated simplifications more consistently preserved uncertainty, diagnostic caution, warning, proportional reassurance, and source-bound clinical meaning. Implication: Results further indicate that readability gains and patient-centred voice do not automatically guarantee semantic accountability, because fluent AI explanations may introduce unsupported additions, softened risk, or overconfident reformulation. Novelty: This study contributes a triadic framework for evaluating medical simplification as readability transformation, semantic fidelity, and narrative-pragmatic alignment rather than as plain-language rewriting alone.]]>
Copyrights © 2026
