AB223. SOH25_AB_396. ConsentGPT?—Artificial intelligence and its accuracy in providing peri-operative information to patients undergoing elective arthroplasty

Background: ChatGPT’ and other artificial intelligence applications are becoming increasingly more popular for both personal and professional use. Patients may seek out information from these services to assist them with decision making surrounding surgery and to explain to them the risks with these options. Monitoring and surveillance of these services is vital to ensure that they complement the human aspect of consent and keep patient safety of paramount importance.

Methods: Three commonly used artificial intelligence based chatbots (‘ChatGPT’, ‘AskAI’, and ‘Gemini’) were examined based off their generated responses in relation to queries surrounding common orthopaedic elective procedures. These were compared against each other, as well as compared to evidence-based answers for the same questions using systematic reviews and consensus literature for the same orthopaedic procedures in question.

Results: Discrepancies and inaccuracies were identified between the three chatbots that were tested. ChatGPT was found to provide the most extensive answers with relatively accurate risks of complications. Certain risks associated with the procedures were omitted by all three chatbots. All three chatbots did notably highlight confounding effects of individual patient risk factors surrounding the peri-operative period.

Conclusions: There is little doubt that artificial intelligence will continue to have an increasingly powerful role in the delivery of healthcare and the patient-practitioner relationship. It is important to be mindful of the sources of information and the potential for inaccuracies in their quality. Further review of online chatbots and contemporary literature is advised to ensure consistency of information and delivery of high-quality patient safety and care.

Keywords: Arthroplasty; ChatGPT; consent; intelligence; perioperative