39 The role of artificial intelligence in surgical infection management: Eealuation of a bounded chatbot for clinical practice and education

Abstract Introduction Artificial intelligence (AI) is a tool whose usefulness is being explored in the medical, surgical, and academic fields. Its use became widespread a couple of years ago, with different IA powered engines being launched and updated at a high pace. There’s still a lack of knowledge about how it can assist in daily clinical practice, how to interpret the results obtained from it, and if these are trustworthy. Regarding decision-making in surgical infections, this tool has an enormous potential. In daily practice, decisions about surgical infections can be complex and infrequent cases require multidisciplinary decision-making. Unfortunately, in many cases international guidelines are difficult to consult and apply clinically. The main objective of this study was to develop and evaluate a chatbot using AI, limited to selected information provided, as a tool for clinical practice and surgical education on surgical infections. Methods To achieve this, we designed a cross-sectional observational study utilising NotebookLM (a Google® Company product) as the AI tool. NotebookLM processes information provided in formats such as PDFs and other documents, producing deep summaries, structured data, comparisons, and accurate responses to queries related to the uploaded content. Surgical infection experts selected five key documents to train the tool. Subsequently, three practical clinical cases of surgical infection of increasing complexity were analysed, requesting differential diagnoses, the most likely diagnosis, and, where relevant, the selection of antibiotic therapy, including the drug, dosage, and timing, with justifications provided. An anonymous survey evaluated the level of interest in the AI responses using a Likert scale, compared AI responses to participants’ own, and assessed whether participants would rely on the AI response for real patient decisions. Two yes/no questions were also included to evaluate the usefulness of the tool in daily practice and training. Results Among the 11 respondents, 6 were doctors, 4 were surgical infection specialists, and 1 was a resident. Of these, 90% found the AI tool very interesting. However, 50% believed their own responses were better and more accurate. Only one respondent indicated they would rely solely on the AI for decisions regarding a real patient. Nonetheless, 80% considered it a highly useful tool for routine clinical practice, citing its ability to shorten decision times. All 11 respondents agreed on its educational value. Drawbacks identified included its reliance solely on scientific evidence and the need for adaptation based on clinical judgement. Conclusions According to the data obtained, this expert-trained chatbot is a promising tool for decision-making in surgical infection cases. Bounded chatbots such as NotebookLM, which rely on curated sources rather than uncontrolled information, are particularly useful for applying clinical guideline algorithms in decision-making. AI is designed to recognise patterns and develop responses accordingly, making it a valuable asset in surgical infection cases, where decisions are supported by published algorithms and guidelines. Additionally, the responses generated are formative, explanatory, and conduce to structured and reasoned knowledge, aiding the study and training of residents who encounter surgical infections, as this is a complex subject with vast amounts of information that can be challenging to interpret.