Artificial Intelligence-Based ChatGPT Responses for Patient Questions on Optic Disc Drusen

Area of Science:

• Ophthalmology
• Artificial Intelligence in Healthcare

Background:

• Optic disc drusen (ODD) are asymptomatic optic nerve head deposits, but can cause visual field defects and mimic papilledema, leading to unnecessary investigations.
• Accurate patient information is vital for managing ODD and preventing misdiagnosis.
• Distinguishing ODD from papilledema is challenging, increasing the risk of extensive clinical workups.

Purpose of the Study:

• To evaluate the accuracy of ChatGPT responses to common patient inquiries regarding optic disc drusen (ODD).

Main Methods:

• Twenty typical patient questions about ODD were developed by two content experts.
• Five distinct ChatGPT responses were generated for each question, totaling 100 responses.
• Two independent experts rated the accuracy of each response on a 5-point Likert scale.

Main Results:

• Expert agreement on response accuracy was substantial (Cronbach's alpha: 0.64).
• 17% of responses were accurate, 78% had non-harmful inaccuracies, and 5% contained potentially harmful inaccuracies.
• Lowest accuracy was observed for questions concerning ODD treatment and prognosis.

Conclusions:

• ChatGPT frequently offers relevant information for optic disc drusen (ODD) patient queries.
• However, inaccuracies in responses related to treatment and prognosis pose potential harm.
• Patients should exercise caution and verify AI-generated medical advice, especially regarding therapeutic strategies and long-term outlooks.