AI can rapidly assess a patient’s risk of postoperative infection immediately after surgery by analysing vital signs recorded during the procedure.
Researchers at the University of Bern and Inselspital, Bern’s University Hospital, have developed CARESCORE, an AI model that calculates a patient’s individual risk of postoperative infection within seconds of an operation ending.
The period between a successful operation and full recovery is obviously critical, during which complications such as wound infections, pneumonia and sepsis must be detected early.
Ordinarily, risk assessment relies primarily on factors known before surgery, such as age, pre-existing conditions and, in some cases, genetic predispositions.
The Bern team noted: ‘Infections after surgery remain a leading cause of morbidity and mortality, yet reliable risk stratification at the end of surgery is limited. Intraoperative vital signs are continuously recorded in modern operating rooms but remain an underexploited source of real-time prognostic information.’
The team, which brings together clinical researchers, data scientists and infectious disease specialists, set out to determine whether the body’s responses during the operation itself carry useful information.
For the study, the team analysed data from more than 10,000 surgeries at Inselspital, combining routinely collected clinical information such as patient age, pre-existing conditions and the type and duration of surgery with vital signs monitored continuously during the procedure, including blood pressure, heart rate and oxygen saturation.
According to co-study leader Guido Beldi, chief of visceral surgery at Inselspital, specific patterns of blood oxygen saturation and changes in heart rate were associated with a higher rate of complications.
The resulting model predicts infection risk more accurately than approaches based on preoperative data alone, allowing clinicians to monitor and treat at-risk patients earlier.
The researchers describe it as a support for medical care rather than a substitute for clinical judgement.
CARESCORE has so far been developed and tested exclusively on data from Inselspital, and must now be validated in other hospitals and under different surgical conditions before it can enter routine clinical practice.
The research is published in the journal npj Digital Medicine.
