Leveraging artificial intelligence for surgical site infection surveillance: A comparison of 5 large language models

We conducted a retrospective study to evaluate the performance of 5 large language models in detecting surgical site infections (SSIs), compared with manual surveillance by an infection preventionist nurse. Forty abdominal surgery patients were included. Manual review achieved 100% diagnostic accuracy. All large language models demonstrated high accuracy (90%-95%) and strong agreement with manual review (κ = 0.80-0.90), with no statistically significant differences in performance (P > .05). AI-based tools may enhance the efficiency of surgical site infection surveillance.