Skip to main content
AAO-HNSF 2025 Annual Meeting & OTO EXPO Main banner
Icon Legend
This session is not in your schedule.
This session is in your schedule. Click again to remove it.
Rate ePoster
Video

Facial Plastic and Reconstructive Surgery

(A404) Artificial Intelligence Transformation of Online Resources on Gracilis Free Muscle Transfer for Facial Reanimation

Monday, October 13, 2025
12:00 PM - 1:00 PM EDT
Location: OTO EXPO, Posters

    Presenting Author(s)

  • Davis S. Chong, BS

    Medical Student
    UC Davis School of Medicine
    Sacramento, California, United States

    • Co-Author(s)

  • Davis S. Chong, BS

    Medical Student
    UC Davis School of Medicine
    Sacramento, California, United States

    • Presenting Author(s)

  • William Swift, MD

    Resident Physician
    Univ of California Davis Med Ctr
    Sacramento, California, United States

    • Co-Author(s)

  • William Swift, MD

    Resident Physician
    Univ of California Davis Med Ctr
    Sacramento, California, United States

  • Travis T. Tollefson, MD

    Professor
    Univ of California Davis Med Ctr
    Sacramento, California, United States

    • Senior Author(s)

  • AM

    Alexander P. Marston, MD

    Assistant Profressor
    UC Davis Department of Otolaryngology - Head and Neck Surgery
    Sacramento, California, United States

Disclosure(s):
Davis S. Chong, BS: No relevant relationships to disclose.
William Swift, MD: No relevant relationships to disclose.
Travis T. Tollefson, MD: No relevant relationships to disclose.
Alexander P. Marston, MD: No relevant relationships to disclose.


Introduction: Generative language models (GLMs) such as ChatGPT stand to revolutionize health literacy by improving accessibility to patient educational materials (PEMs). As a proof of concept, gracilis free muscle transfer (GFMT) exists today as a versatile solution for facial reanimation. However, as a complex and niche procedure, online information on GFMT may exceed the 6th grade reading level set by the American Medical Association (AMA) for PEMs. Analyzing the readability of patient-directed websites on GFMT and assessing the readability and integrity of GLM output will help assess the accessibility of education for this highly subspecialized surgery.

Methods: The first nine non-repeated patient-directed websites from a Google Search for “gracilis free tissue transfer” and other related searches were evaluated by six standardized readability formulas: Automated Readability Index (ARI), Coleman Liau Index (CLI), Gunning Fog Score, Simple Measure of Gobbledygook (SMOG), Flesch Kincaid Grade Level (FKGL), and Flesch Reading Ease (FRE) score. These websites were transformed by GLMs (ChatGPT 4, Claude 3.5, Llama 2) and reanalyzed for readability. Semantic similarity of GLM output to source was assessed by pairwise latent semantic analysis (LSA).

Results: The mean pooled FRE of the 9 websites was 44 (age 18+). The mean grade level by each test used was: FKGL 12.2 (age 17-18), GFS 14.6 (age 18+), SMOG 10.9 (age 14-15), CLI 13.5 (age 18+), and ARI 12.3 (age 17-18). After GLM transformation, the mean pooled FRE increased significantly from 44 to a mean of 64 (p < 0.05), indicating improved readability. Average FRE was similar between all three GLMs (67.2, 64.3. 59.7), but semantic similarity decreased with Claude 3.5 (cosine similarity [scale 0 to 1, 1 = identical] =0.37, vs. =0.52 [ChatGPT 4], =0.56 [Llama 2]).

Conclusions: GLMs can improve readability of current online educational resources of GFMT, but while readability improves, content integrity may suffer.