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.
Video

Business of Medicine

(A011) Use of Artificial Intelligence to Improve Readability of Salivary Gland Cancer Patient Educational Materials

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

    Presenting Author(s)

  • Sree C. Chinta, BS

    Medical Student
    New Jersey Medical School
    Newark, New Jersey, United States

    • Co-Author(s)

  • Emma Thompson, BA

    4th Year Medical Student
    Rutgers New Jersey Medical School
    jersey city , New York, United States

  • Anthony M. Saad, BA

    Medical Student
    New Jersey Medical School
    East Brunswick, New Jersey, United States

  • Ariana Shaari, MD

    Resident
    Divison of Otolaryngology, Stony Brook Medicine
    Stony Brook, New York, United States

    • Senior Author(s)

  • DR

    Dylan Roden, MD, MPH

    Rutgers, United States

Disclosure(s):
Sree C. Chinta, BS: No relevant relationships to disclose.
Emma Thompson, BA: No relevant relationships to disclose.
Anthony M. Saad, BA: No relevant relationships to disclose.
Ariana Shaari, MD: No relevant relationships to disclose.


Introduction: Patient education materials (PEMs) should be written at a 6th-grade reading level, but it has been found within otolaryngology that they tend not to be. Artificial intelligence (AI) tools may be utilized to improve readability, such as for salivary gland cancers.

Methods: We queried Google for “salivary gland cancer” and 63 of the top 100 websites met inclusion criteria. We calculated six readability indices including the commonly used Flesch-Kincaid Grade Level [FKGL] for each of the webpages before and after prompting OpenAI’s ChatGPT-4 to “Convert the text to a sixth-grade reading level.” We evaluated the understandability and actionability of the original PEMs using the PEM Assessment Tool (PEMAT) for understandability (U) and actionability (A). T-tests were used to analyze PEMs' reading levels before and after ChatGPT modification and based on inclusion of text related to diagnosis, treatment, or prognosis of salivary gland cancer. Bivariate analysis was used to assess the correlation between PEMAT-U and -A.

Results: The plurality (44.4%) of queried webpages were from academic hospitals, with other main sources being general knowledge websites (9.5%), national organizations (11.1%), non-profit hospitals (14.3%), and private practices (12.7%). There was a significant correlation between PEMAT-U (mean 78.6%) and PEMAT-A (mean 10.9%) (p = 0.014) of the webpages. ChatGPT was able to improve the average FKGL from 11.7 to 9.3 (p < 0.001). ChatGPT decreased the average word count from 1,409 to 651 and words per sentence from 18.3 to 15.5 (p < 0.001). The average reading level was lower for ChatGPT-modified documents that did not contain a discussion of treatment for salivary gland cancers (FKGL 8.9% vs. 9.4%, p=0.042).

Conclusions: ChatGPT can improve the readability of PEMs related to salivary gland cancer, but not to the recommended sixth-grade reading level. ChatGPT may be more effective for PEMs that lack complex terminology, as may be the case with salivary gland treatment regimens. Future research will include comparing patient reported readability of original versus ChatGPT modified PEMs.