JoVE - Journal of Visualized Experiments
JoVE Visualize
Contact Us

Utility of an LLM-powered experts-in-the-loop chatbot for pre- and post-operative care of cataract surgery patients

  • 0Microsoft Research, Bangalore, India.
European Journal of Ophthalmology +

|

December 8, 2025

|

December 8, 2025

View abstract on PubMed

Summary

This summary is machine-generated.

CataractBot, an AI chatbot, effectively answers patient questions about cataract surgery. Expert verification improved its accuracy, demonstrating potential for enhanced patient-provider communication in ophthalmology.

Area Of Science

  • Ophthalmology
  • Artificial Intelligence
  • Medical Informatics

Background

  • Patient engagement and access to reliable information are crucial in managing ophthalmic conditions like cataracts.
  • Traditional communication channels can be time-consuming for both patients and healthcare providers.
  • Large Language Models (LLMs) offer potential for developing innovative patient support tools.

Purpose Of The Study

  • To evaluate the utility of CataractBot, an LLM-powered chatbot, in providing doctor-verified answers to patient inquiries about cataract surgery.
  • To assess the chatbot's performance and user experience among patients, attendants, and medical experts.
  • To determine the impact of expert verification on the accuracy and completeness of chatbot responses.

Main Methods

  • A 24-week mixed-methods study involving patients, attendants, ophthalmologists, and patient coordinators.
  • CataractBot utilized a curated knowledge base for instant responses, with asynchronous verification and edits by medical experts.
  • Analysis of interaction logs, including user questions, bot answers, and expert feedback, to assess utility and accuracy.

Main Results

  • 318 patients and attendants submitted 1,992 messages, with a higher volume of pre-surgery and medical questions.
  • Ophthalmologists rated 84.5% of CataractBot's medical answers as accurate and complete.
  • Expert edits improved the acceptance rate of bot answers by 19.0% over time, reducing the need for future expert intervention.

Conclusions

  • CataractBot effectively addresses patient medical questions regarding cataract surgery, incorporating expert feedback for continuous improvement.
  • The chatbot demonstrated potential in supporting patient-provider communication, improving information accessibility, and reducing expert workload.
  • LLM-powered chatbots represent a promising tool for enhancing patient education and support in ophthalmology.

Keywords: