2168 - Designing a Question-Based Patient Education System Powered by Generative AI to Narrow the Information Gap for Underserved Populations

Y. Wang, and J. A. Efstathiou; Department of Radiation Oncology, Massachusetts General Hospital, Harvard Medical School, Boston, MA

Purpose/Objective(s): To narrow the information gap in radiation therapy (RT) for underserved populations through generative AI. The objective is to design an interactive, question-based system for personalized patient education, addressing disparities in comprehension and facilitating informed decision-making. Materials/

Methods: Designed by experts with substantial experiences in clinical oncology, radiation physics, and AI, the system’s foundation is a knowledge framework crucial for patient comprehension of RT options, procedures, and expectations. In this innovative question-based approach, the system interacts with a chatbot powered by generative AI. The chatbot’s content creation is tailored to underserved patients’ needs, utilizing a comprehensive question bank curated by the experts, and algorithms to suggest follow-up questions and match a free-text question to best entry in the bank. After providing basic disease information and underserved factors, patients begin learning by selecting one of the three common questions suggested by the system. Serving as an interface to prompt the chatbot, the system allows patients to choose from three suggested follow-up questions, input a free-text question, or revisit a previous level. The system offers feedback on progress, summarizes key perspectives, and previews upcoming information. Impact assessment is conducted through pre- and post-intervention surveys (Table 1), evaluating knowledge acquisition, confidence, anxiety reduction, information accessibility, and AI tool perceptions.

Results: Comprehensive modules are created for each common cancer, covering treatment options, procedure expectations, potential outcomes and side effects, and post-treatment care. The system guides the chatbot to provide highly personalized responses, considering five pivotal factors influencing decision-making for underserved patients: socioeconomic status, education level, healthcare access, cultural and linguistic diversity, and mental well-being.

Conclusion: This study lays the groundwork for a novel approach to patient education, potentially mitigating information disparities and empowering underserved patients in decision-making processes. Table 1: Pre- and post-intervention surveys with Likert-scale questions

	Pre	Post
1	How knowledgeable are you about your RT options and procedures?	How knowledgeable are you about your RT, compared to before using the AI tool?
2	How confident are you in your ability to make informed decisions about your RT?	How confident are you in your decision-making ability, compared to before?
3	How anxious do you feel about starting your RT treatment?	Did using the AI tool help reduce your anxiety about your RT?
4	Do you have access to reliable sources of information about your RT?	Did the AI tool provide helpful information that empowered your decision-making?
5	Are you interested in using an AI tool to learn more about your RT?	Would you recommend the AI tool to other patients facing similar situations?