Oral Presentation ESA-SRB-ANZBMS 2024 in conjunction with ENSA

Background and Aim:  This project aimed to leverage chatGPT to develop a user-friendly tool featuring a conversational interface for personalised fracture risk assessment and guidance on osteoporosis treatment and prevention.

Methods: We have previously developed BONEcheck^TM, a comprehensive tool for personalised fracture risk prediction. In this study, we extend the tool to include risk interpretation and advice on osteoporosis treatment and prevention. We began by extracting predicted risks of fracture, refracture, and mortality from BONEcheck and integrating them into ChatGPT. We then updated ChatGPT using nine clinical osteoporosis textbooks, fifteen osteoporosis management guidelines from recognized professional societies, pivotal randomized controlled clinical trials, and 220 questions related to the biology, diagnosis, prognosis, treatment, and prevention of osteoporosis. The new tool is called BONEcheckGPT, and we tested the tool with new 25 questions to evaluate its accuracy and reproducibility.

Results: In multiple tests utilizing either voice or text, BONEcheckGPT produced accurate and 100% reproducible risk of fracture, refracture, and mortality comparable to those generated by BONEcheck. In all tests involving clinical questions related to the diagnosis, treatment, and prevention of osteoporosis, BONEcheckGPT provided answers aligned with treatment guidelines and textbooks. Remarkably, BONEcheckGPT delivered this information in over 80 languages. A demo version of BONEcheckGPT is available at the following site: https://youtu.be/-pw4xYJSMIc.

Conclusions: BONEcheckGPT can serve as a valuable resource for promoting osteoporosis prevention by offering user-friendly and detailed answers to enquiries in multiple languages, making it particularly useful for underdeveloped regions in the world.