AI flags kidney injury risk before operation

Researchers have developed an AI model that can identify the risk of acute kidney injury after surgery using only eight pieces of information gathered before the operation. 
The model relies on existing test results and doesn’t require additional special examinations, and it came close to matching the accuracy of more complex models that also use data collected during and after surgery.
The AI model predicts the risk of acute kidney injury based on data from patients who underwent surgery under general anaesthesia. 
Acute kidney injury after surgery can raise the risk of death, lengthen hospital stays and lead to long-term deterioration in kidney function. 
Several prediction models have been developed, but many required information obtained during or after surgery or covered only patients undergoing specific procedures, such as cardiac surgery, limiting their usefulness in pre-operative care.
The team analysed data on 184,560 adult patients who underwent surgery lasting at least one hour under general anaesthesia at Severance Hospital and Gangnam Severance Hospital between 2006 and 2022. 
After comparing an AI model that used pre-, intra-, and post-operative information with one that used only pre-operative data, the researchers narrowed the inputs to the eight most important variables to create a simplified model.
These inputs include information routinely checked before surgery, such as kidney function, overall physical condition, and nutritional and anaemia status, so no separate testing is needed.
The research team was led by Professor Lee Joo-han of the Department of Transplantation Surgery and Professor Kim Hyoung-woo of the Division of Nephrology at Yonsei University Severance Hospital, along with Professors Yoo Seung-chan and Kim Sun-ji of the Department of Biomedical Systems Informatics at the medical school.
Lee said: ‘What matters is having a tool that is easy to use in actual clinical settings while still offering sufficient predictive power. We expect this model to be used in identifying and managing high-risk patients. We plan to link the system to electronic medical records so that a patient's risk level is presented automatically before surgery, and to verify whether it actually leads to fewer complications.’
The findings were published in the International Journal of Surgery – Lee JH, Kim HW, Yoo SC, Kim SJ. Development of a preoperative artificial intelligence model to predict postoperative acute kidney injury using eight clinical data points. International Journal of Surgery. 2026 Oct.