Evaluation of a Retrieval-Augmented Generation-Powered Chatbot for Pre-CT Informed Consent: a Prospective Comparative...

Area of Science:

• Radiology and Medical Imaging
• Artificial Intelligence in Healthcare
• Patient Education

Background:

• Standard pre-CT scan information delivery relies on physician consultations, which can be time-consuming.
• Ensuring patient understanding and addressing concerns is crucial for informed consent and satisfaction.
• AI-powered tools offer potential for efficient and accessible patient information delivery.

Purpose of the Study:

• To evaluate the feasibility, usability, and effectiveness of a Retrieval-Augmented Generation (RAG)-powered Patient Information Assistant (PIA) chatbot for pre-CT counseling.
• To compare the PIA chatbot's performance against standard physician consultations.
• To assess patient satisfaction, information comprehension, and consultation duration.

Main Methods:

• A prospective comparative study involving 86 patients scheduled for CT imaging.
• Random assignment to either a PIA chatbot group (n=43) or a standard physician consultation group (n=43).
• Assessment of patient satisfaction, information clarity, comprehension, and concerns using Likert-scale questions; measurement of consultation duration; radiologist ratings of PIA chats.

Main Results:

• Both groups reported similar high ratings for information clarity and overall comprehension.
• The physician consultation group was more effective in alleviating patient concerns (p=0.003).
• The PIA group had significantly shorter subsequent consultation times (p=0.04), with radiologists rating PIA quality highly.

Conclusions:

• A RAG-powered PIA chatbot is a feasible tool for delivering pre-CT information, comparable to physicians in clarity and comprehension.
• The PIA chatbot significantly reduces physician consultation time.
• Physician consultations remain superior in addressing patient worries and concerns, despite comparable satisfaction and comprehension levels with the PIA.