14 Writing To Patients With Vestibular Schwannoma Transitioning from complexity to clarity leveraging Learning language model

Abstract Introduction Effective written communication is central to patient-centred and safe care. NHS guidance recommends that outpatient clinic letters be written directly to patients at a readability level equivalent to Flesch Reading Ease (FRE) ≥60. However, neurosurgical correspondence often remains highly technical. Vestibular schwannoma (VS) management, involving complex decisions regarding surgery, radiotherapy, or surveillance, provides a suitable model to examine readability and evaluate whether large language models (LLMs) can improve clarity. Method Eighty-three anonymised VS clinic letters written between 2014 and 2025 from a tertiary skull-base service were analysed. Readability was assessed using Flesch Reading Ease, Flesch–Kincaid Grade Level, and Gunning Fog Index. Letter length, estimated dictation time, estimated reading time, temporal trends, and recipient type (GP vs patient) were examined. A structured LLM prompt was developed to rewrite letters targeting improved accessibility (FRE ≥60), while preserving clinical accuracy. Results Letters were consistently difficult to read (mean FRE ∼42; mean Flesch–Kincaid Grade 13; mean Gunning Fog 16). No letters met NHS-recommended thresholds. From 2019 onwards, letter length and estimated dictation and reading times increased, peaking in 2022–2023 (>1,000 words; >7 minutes dictation). Letters addressed directly to patients did not demonstrate meaningful improvement in readability. AI-assisted rewriting improved readability metrics (FRE increased to high-40s/low-50s; Grade reduced to ∼10–11) while maintaining clinical tone. Conclusions VS clinic letters remain written above recommended health-literacy standards. LLM-assisted rewriting offers a scalable approach to improving clarity while preserving professional accuracy, with potential to enhance patient understanding and shared decision-making.