When Precision Meets Penmanship: ChatGPT and Surgery Documentation

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.


Introduction
ChatGPT (Chatbot Generative Pre-Trained Transformer) is an artificial intelligence developed by Open AI.It has several potential applications in the field of medicine, many of which have been published elsewhere.As a large-scale language model, it is particularly useful in generating text.In this study, we investigate the use of ChatGPT in constructing an operation note.
Writing an operation note is a crucial part of any surgical procedure.It allows future medical professionals to know what happened at the time of surgery.An accurate and detailed description of these events ensures patient safety and continuity of care.Furthermore, there are important medicolegal implications in providing clear documentation to prevent future litigation claims that often exploit gaps and ambiguities in patient notes.In the UK, the current cost of litigation against the National Health Service (NHS) is a staggering £128.6 billion [1] or approximately 1/3 of the yearly budget.Based on the previous litigation claims and feedback from NHS panel firm lawyers, the organisation 'Getting it right first time' (GIRFT) has produced guidelines on documentation of operation notes [2].The guidelines currently cover five common surgical procedures: appendicectomy, cholecystectomy, inguinal hernia repair, laparotomy, and bowel resection.
In a busy operating list, where efficiency is key, ChatGPT has the potential to help surgeons streamline the process of writing an operation note whilst making sure it is thorough and accurate.By reducing the time spent on documentation, it could alleviate some of the administrative burdens on surgeons and allow them to have more rest between cases.The significance of addressing surgeon fatigue and the importance of adequate breaks to prevent burnout is becoming increasingly recognised [3].A study published in the Lancet Digital Health has demonstrated that ChatGPT can generate adequate discharge summaries from a short prompt [4].This study investigates whether ChatGPT can do the same for operation notes, whilst maintaining GIRFT standards.The focus of this investigation is on laparoscopic appendicectomy, one of the most common surgical procedures performed in the UK.

Technical Report
We asked ChatGPT-4, the latest generation of GPT to produce an operation note for a laparoscopic appendicectomy based on the following prompt: "Complete an operation note for 'A 25 yo male, laparoscopic appendicectomy for migratory RIF pain and tenderness.GA. 40ml 0.5%.3 port.Appendix inflamed perforated, nil else healthy base.Suction washout.Endoloops and diathermy used.Closure. 5 days co-amox.No drain." The results were then evaluated against the GIRFT recommendations on best practice for laparoscopic appendicectomy documentation [5].Although the guidelines encompass 34 criteria, they acknowledge that certain elements might be documented elsewhere in the patient's record, such as accident and emergency (A&E) assessment, ward round entries, and a separate WHO (World Health Organisation) surgical checklist.Consequently, in the context of evaluating the comprehensiveness of an operative note, we excluded 'time of decision to operate,' 'consent process,' 'WHO checklist', and the surgeon's signature, considering the latter's incompatibility with digitisation.We assessed ChatGPT's response against the remaining 30 points, a score of 1 was given if the recommendation was documented, and a score of 0 was given if it was not.We then calculated a total score out of 30.This process was repeated three times with the same prompt to obtain an average score (Table 1).These generated outputs and results are as follows.1.The patient was brought to the operating room and placed in the supine position.General anaesthesia was induced, and the patient was intubated without complications.
2. The abdomen was prepped and draped in a sterile fashion.
3. A 40 ml solution of 0.5% bupivacaine was infiltrated at the planned incision sites for local anaesthesia.
4. A 3-port technique was utilized for the laparoscopic appendicectomy.A 10 mm umbilical port was inserted using the Hasson technique, and the abdomen was insufflated with CO2 to a pressure of 15 mmHg.Two additional 5 mm ports were placed in the left lower quadrant and suprapubic region under direct visualization.
5. Upon inspection, the appendix appeared inflamed and perforated.No other abnormalities were identified.
6.The mesoappendix was divided using a combination of endoloops and diathermy.
7. The base of the appendix was secured with two endoloops, and the appendix was divided between the loops using diathermy.
8. The specimen was retrieved using an endoscopic retrieval bag and removed through the umbilical port.

9.
Copious suction and washout of the peritoneal cavity were performed.
10. Haemostasis was confirmed, and the appendiceal stump was inspected for any signs of bleeding or contamination.
11.The ports were removed under direct visualization, and the fascia at the umbilical port site was closed using absorbable sutures.
12. The skin incisions were closed with subcuticular absorbable sutures and covered with sterile dressings.
13.The patient was extubated and transferred to the recovery room in stable condition.

Estimated Blood Loss: Minimal
Complications: None Disposition: The patient was admitted for postoperative care and monitoring.
Postoperative Instructions: The patient will receive a 5-day course of co-amoxiclav as per hospital protocol.
No drain was placed.
Follow-up: The patient is to follow up with the surgeon in 2 weeks for a postoperative check-up and suture removal, or earlier if any complications arise.

ChatGPT generated operation note 2
Patient: 25-year-old male Procedure: Laparoscopic Appendectomy 1.The patient was brought to the operating room and placed in a supine position on the operating table.
General anaesthesia was induced by the anaesthesia team, and the patient was intubated without any complications.
2. The abdomen was prepped and draped in a sterile fashion.A time-out was performed to confirm the correct patient, procedure, and operative site.
3. A 1 cm infraumbilical incision was made, and a Veress needle was inserted into the peritoneal cavity.The abdomen was insufflated with CO2 to a pressure of 15 mmHg.A 10 mm trocar was introduced through the infraumbilical incision.
4. Under direct vision, two additional ports were inserted: a 5 mm port in the left lower quadrant and another 5 mm port in the suprapubic area.

5.
A laparoscopic exploration of the abdominal cavity was performed.The appendix was found to be inflamed and perforated.No other abnormalities were detected.
6.The mesoappendix was divided using diathermy.Two endoloops were applied to the base of the appendix, and the appendix was transected between the endoloops.
7. The specimen was placed in an endoscopic retrieval bag and removed through the 10 mm infraumbilical port.
8. The abdominal cavity was thoroughly irrigated with warm saline, and suction was used to remove any residual fluid.
9. The three trocar sites were inspected for bleeding or injury.Haemostasis was confirmed.
10.The pneumoperitoneum was deflated, and the trocars were removed under direct vision.
11.The fascia of the 10 mm infraumbilical port site was closed using an absorbable suture.All incisions were closed with absorbable subcuticular sutures and covered with sterile adhesive dressings.
12. The patient was extubated and transferred to the recovery room in stable condition.over-reliant on the automated process without adequate checking of the produced document.
Another issue is that confidential patient information should not be stored on insecure servers abroad.If intelligent large language models can be integrated into secure electronic health records, patient details such as past medical history, operation dates, surgeon, and anaesthetist information could be auto-filled making the process even more streamlined.NHS trusts are currently trialling the use of AI-powered voice recognition with various programmes such as dictate IT [8].Open AI offers the Application Programming Interface (API) of ChatGPT to third-party developers to be integrated into their own products which may allow more widespread licencing and use within the NHS.These technologies could be integrated with a large language model (LLM) such as ChatGPT to allow secure voice dictation of comprehensive operation notes and could allow the outputs to be seamlessly processed into operative notes instead of requiring the user to copy and paste into a Word document to print as in the current iteration.Furthermore, these operation notes could be integrated with medical inpatient notes to streamline the creation of discharge summaries.

Conclusions
ChatGPT can serve as an assistant for writing detailed operation notes, helping to save time and alleviating pressure on medical professionals, a factor shown to increase errors.However, care must be taken to ensure that the generated note contains the correct procedure technique and case-specific details.After repeated uses and requests, ChatGPT does become an intelligent and accurate surgical note-making assistant.