Attitudes toward large language model-based Artificial Intelligence systems as an information source for shared...

Area of Science:

• Oncology
• Medical Informatics
• Artificial Intelligence

Background:

• Structured shared decision-making (SDM) in radiation oncology necessitates reliable patient information.
• Patients often lack knowledge and hold misconceptions about cancer therapy and side effects, hindering informed decision-making.
• Large Language Model (LLM)-based AI offers potential for accessible, evidence-based patient information but faces adoption barriers.

Purpose of the Study:

• To investigate patient and healthcare professional (HCP) perspectives on LLM adoption in radiation oncology.
• To assess current and anticipated use of LLMs for patient information and shared decision-making (SDM).
• To identify factors influencing LLM acceptance in cancer care.

Main Methods:

• A survey was administered to radiation therapy patients (n=400) and HCPs (n=200) between March 2024 and February 2025.
• Electronic questionnaires assessed sociodemographics, SDM status, information sources, and LLM perceptions.
• Data analysis included descriptive statistics and logistic regression.

Main Results:

• The internet is the primary information source for patients; current LLM use is low (18.2% patients vs. 69.5% HCPs).
• HCPs are optimistic about future patient reliance on LLMs (77%), but most patients remain skeptical (29.1% agreement).
• Patients prioritize physician trust (65.8%) over LLMs for SDM data; technological familiarity predicts patient LLM use.

Conclusions:

• A significant gap exists between HCP optimism and patient skepticism regarding LLM utility in radiation oncology.
• Patient trust in physicians and technological familiarity are key factors influencing LLM adoption for SDM.
• Further research is needed to bridge the gap and facilitate effective LLM integration into cancer care.