LAUSR

Background/Aim: Demographic change and increasing complexity of therapy decisions lead to a growing burden on the healthcare system, necessitating efforts to simplify and enhance the efficiency of patient care. The present study evaluates ChatGPT’s ability to provide therapy recommendations for gynecological malignancies that are both in line with the local guidelines and individually tailored to the patient. Patients and Methods: Sixteen patients with endometrial, cervical, and ovarian cancer who were treated in the gynecological clinic of the University Hospital Erlangen from January 2022 to August 2023 were included in the analysis. Data collected within clinical routine care were communicated to the chat-based AI model ChatGPT (version 3.5). ChatGPT’s performance generating treatment plans were evaluated using an answer scoring system and descriptive analysis. Results: According to the answer scoring system [range: −1 point (minimum) to 2 points (maximum)], ChatGPT demonstrated a good potential in generating therapy recommendations with an average score of 0.75 points for patients with ovarian cancer, 0.7 points for cervical and 1.5 points for endometrial cancer patients. The most common deductions in points were about incomplete therapy recommendations, whereas contraindicated treatment modalities were rarely suggested. Individual patient characteristics were regularly considered by ChatGPT. ChatGPT reliably indicated aftercare and provided detailed information on preventive measures as well as supportive treatment. Conclusion: ChatGPT is a promising tool for the generation of therapy suggestions for gynecological carcinomas with high flexibility in response to individual patient differences. At the current state, however, ChatGPT is not suitable for replacing expert panels.