Online patient information on elective orthognathic surgery provided by UK National Health Services hospitals: assessment of readability and improvement using large language models.

Clear and understandable patient information is crucial for both informed consent and postoperative recovery following elective orthognathic surgery. Despite National Health Service (NHS) guidance recommending suitability for the reading level of an 11-year-old, current patient information leaflets (PILs) do not meet this standard. This study was undertaken to assess the readability of NHS orthognathic surgery patient information and whether large language models (LLMs) could improve readability to meet the recommended levels. A cross-sectional analysis was conducted of online PILs on elective orthognathic surgery from United Kingdom NHS hospital websites. Readability was assessed using five validated scoring systems: Flesch Reading Ease Score (FRES), Flesch-Kincaid Grade Level (FKGL), Gunning Fog Index (GFI), Coleman-Liau Index (CLI), and Simple Measure of Gobbledygook (SMOG). Three LLMs, ChatGPT, Claude, and Gemini, were prompted to simplify PILs to the reading age of an 11-year-old. Of 111 UK Hospital Trusts with Maxillofacial Surgery Units, 73 (65.8%) offered elective orthognathic surgery, of which 21 (28.8%) provided online PILs that yielded 24 eligible texts. Original PILs fell short of the recommended NHS readability standards on 4/5 indices, achieving acceptable levels only on FKGL (6.23 ± 1.26). Following revision, all three LLMs significantly improved readability across all indices (p < 0.001). Gemini achieved the strongest performance, meeting the target thresholds on 3/5 indices (FRES 80.35 ± 4.31, FKGL 4.73 ± 0.79, CLI 6.77 ± 0.59). Despite significant improvements, no model consistently met all NHS standards. NHS orthognathic surgery PILs are commonly written above the recommended readability levels. LLMs can improve comprehension; however, unrestricted LLM-generated materials still fail to meet the recommended NHS literacy standards, highlighting the need for targeted frameworks to improve readability.