ChatGPT in Urogynecology Research: Novel or Not?

Urogynecology (Phila). 2024 Mar 25. doi: 10.1097/SPV.0000000000001505. Online ahead of print.

Abstract

Importance: ChatGPT (Chat Generative Pre-trained Transformer) is an artificial intelligence (AI) chatbot that provides human-like responses to text prompts. Little is known regarding ChatGPT's ability to identify original research ideas in urogynecology.

Objective: Our objective was to evaluate the accuracy of ChatGPT in generating novel systematic review (SR) and research ideas in urogynecology.

Study design: In this cross-sectional study, we asked ChatGPT to generate 10 novel SR ideas that have not yet been published for each of the following 5 topics: (1) urogynecology, (2) tension-free vaginal tape slings, (3) pessaries for pelvic organ prolapse, (4) beta-3 agonist use for overactive bladder, and (5) sexual function with genitourinary syndrome of menopause. Research idea novelty was assessed by cross-referencing PubMed and Scopus to determine if any prior publications existed.

Results: ChatGPT proposed 50 total SR ideas, including 10 for each prompt. Overall, ChatGPT showed 54% overall accuracy in developing novel SR ideas. Nonnovel SR ideas had a median of 19 (interquartile range, 8-35) published SRs on the suggested topic. When stratified by prompt type, 50% of general and 40-70% of specific urogynecology proposed SR ideas were found to be novel. There were no publications of any type identified for 7 of the 50 suggested ideas.

Conclusions: ChatGPT may be helpful for identifying novel research ideas in urogynecology, but its accuracy is limited. It is essential for those using ChatGPT to review existing literature to ensure originality and credibility. As AI transforms health care, we encourage all urogynecologists to familiarize themselves with popular AI platforms.