When Precision Meets Penmanship: ChatGPT and Surgery Documentation.

Abstract

ChatGPT (Chatbot Generative Pre-Trained Transformer) is an artificial intelligence with several potential applications in the field of medicine. As a large language model, it is particularly good at generating text. This study investigates the use of ChatGPT in constructing operation notes for laparoscopic appendicectomy, one of the most common surgical procedures in the UK. We prompted ChatGPT-4, the latest generation of ChatGPT, to produce operation notes for laparoscopic appendicectomy, which were then evaluated against 'Getting It Right First Time' (GIRFT) recommendations. GIRFT is an organisation that has collaborated with the National Health Service (NHS) to improve surgical documentation guidelines. Excluding certain items documented elsewhere in patient records, the generated notes were assessed against 30 key points in GIRFT recommendations. This process was repeated three times to obtain an average score. Our results showed that ChatGPT generated operation notes in seconds, with an average coverage of 78.8% (23.66 out of 30 points) of the GIRFT guidelines, surpassing average compliance with similar guidelines from the Royal College of Surgeons (RCS). However, the quality of ChatGPT's output was found to be dependent on the quality of the prompt, highlighting the need for verification of the generated content. Additionally, secure integration with electronic health records is required before ChatGPT can be adopted into the NHS.