Description

Healthcare-associated infections (HAIs) have a negative impact on patient health, represent a significant healthcare and economic burden on healthcare systems and are considered the most preventable cause of serious adverse events in hospitalised patients.

Epidemiological surveillance is one of the eight core components of the World Health Organization (WHO) Infection Prevention and Control Programmes. These include surveillance programmes for surgical site infection (SSI), which have proven to be effective in all types of surgery and in a variety of settings.

For a programme to be effective, surveillance for HCAIs must be active, prospective and continuous, comprising a surveillance period up to 30-90 days post-intervention, to cover the high rate of SSIs detected after discharge.

At present, infection control teams perform a manual, prospective, time-consuming and almost artisanal work, which should make a transition to automated or semi-automated surveillance that leverages the possibilities offered by today\'s information technology.

The evolution of surveillance systems should benefit from this new possibilities offered by artificial intelligence, allowing automated detection of suspected SSI adverse events from clinical course text, microbiology reports or coding of diagnoses, procedures, complications and readmissions.

This study aims to evaluate the ability of ChatGPT to detect surgical site infections (SSI) at the three anatomical levels described by the CDC.

The study will retrospectively compare the results of the AI chatbot in diagnosing SSI, trained using the US CDC definition criteria, with a large cohort of elective colorectal surgery patients already evaluated through a nationwide nosocomial infection surveillance system, which will be the comparative gold standard.