P-288 Chat Generative Pre-trained Transformer (ChatGPT) Proves to be an Effective Assistant for Clinical Embryologists in Laboratory Tasks: A Pilot Cross-sectional Study

Abstract

Abstract Study question What are the capabilities of ChatGPT in troubleshooting, fact-checking and generating report templates in the in vitro fertilization (IVF) laboratory? Summary answer Clinical embryologists perceived ChatGPT as accurate and comprehensive in troubleshooting, generating standard operating procedures (SOP), writing reports and fact-finding in the IVF laboratory. What is known already ChatGPT is an artificial intelligence (AI)-driven chat robot (chatbot) with 175 billion parameters in its natural language processing model. It is remarkable for its concise, human-like answers to user inquiries. With advanced AI technology, ChatGPT provides in-depth responses, handles complex problems, and addresses intricate questions. This chatbot has received significant recognition and is anticipated to encourage users to employ it for practical applications, including in the IVF laboratory. However, the abilities of ChatGPT in executing various tasks (such as troubleshooting, generating SOPs) in the in vitro fertilization (IVF) laboratory have not been investigated yet. Study design, size, duration The aim of this cross-sectional study is to assess the proficiency of ChatGPT in four tasks commonly performed by embryologists: troubleshooting, designing SOPs, composing reports, and fact-checking. An anonymous online survey of clinical embryologists (n = 40) was conducted to achieve this aim. It was performed between December 2022 and January 2023. Participants/materials, setting, methods Clinical embryologists participated in a five-point Likert scale (1-Very disagree to 5-Very agree) questionnaire. Participants were presented with eight vignettes generated by ChatGPT related to the investigated four tasks. Embryologists provide ratings about both their perceived accuracy and perceived completeness of the provided answers. We then asked about their intention to incorporate ChatGPT into their daily tasks. Main results and the role of chance The median years of experience of survey participants was 11.5 years (IQR 8-18). Among the participants, 37.5% held an undergraduate bachelor's degree, while the largest proportion, 62.5%, held graduate degrees (Masters and doctorates). Embryologists rated ChatGPT as having an accurate (mean Likert score of 3.45) and comprehensive (mean Likert score of 3.36) value for troubleshooting. They also considered it to be accurate (mean Likert score of 3.34) and comprehensive (mean Likert score of 3.26) for writing SOP templates. Furthermore, they found ChatGPT to have an accurate (mean Likert score of 3.71) and comprehensive (mean Likert score of 3.66) value for report writing. Additionally, they deemed ChatGPT to have both an accurate and comprehensive value (mean Likert score of 3.67) for verifying facts. Overall, experienced embryologists perceived an added value of ChatGPT (Likert scores ≥3). On average, they rated ChatGPT as accurate (average score of 3.54) and comprehensive (average score of 3.49) in all tasks tested. They also expressed an intention to use ChatGPT in their laboratory work (average Likert, 3.58). Limitations, reasons for caution The embryologists who participated in the survey were highly experienced. Wider implications of the findings The model has the potential to assist clinical embryologists in resolving issues and performing administrative duties, making it a valuable resource. Embryologists may benefit from integrating ChatGPT into their educational and certification processes, as well as their daily tasks. Trial registration number Not applicable