Assessing ChatGPT's Educational Potential in Lung Cancer Radiotherapy From Clinician and Patient Perspectives:...

Area of Science:

• Oncology
• Medical Informatics
• Artificial Intelligence

Background:

• Large language models (LLMs) are being explored for patient education in healthcare.
• In radiation oncology, LLMs may enhance patient understanding of complex treatments.
• The quality and acceptance of LLM-generated medical content need further investigation.

Purpose of the Study:

• To evaluate ChatGPT-4 as a patient education tool for lung cancer radiotherapy.
• To assess readability, content quality, and perceived usefulness of AI-generated responses.
• To gather perspectives from both clinicians and patients.

Main Methods:

• Developed 8 frequently asked questions on lung cancer radiotherapy.
• Submitted questions to ChatGPT-4o for responses.
• Evaluated responses using readability formulas (FRE, WSTF), expert clinician review (multicenter, 6 clinicians), and patient feedback (30 patients).

Main Results:

• Most responses were difficult to read (university/secondary school level).
• Clinicians rated relevance (4.5/5) and correctness (4.3/5) highly; completeness was lower (3.9/5).
• Patients found responses clear (4.4/5) and relevant (4.3/5), but less trustworthy (3.8/5) and usable (3.7/5). No harmful misinformation was found.

Conclusions:

• ChatGPT-4 has potential as a supplementary tool in radiation oncology patient education.
• Clarity and relevance were appreciated, but completeness and trustworthiness need enhancement.
• Clinician oversight and further optimization are crucial for safe and effective LLM integration in clinical practice.