As the aging population grows and medical technology advances, the number of patients aged 65 and older undergoing surgeries such as cancer and joint procedures is rapidly increasing.

In response, Korean researchers have developed an artificial intelligence model capable of predicting cardiovascular disease -- a major postoperative complication in older patients -- before surgery.

Seoul National University Bundang Hospital (SNUBH) announced Tuesday that its research team, led by Professor Suh Jung-won of the Department of Cardiology, developed a machine learning-based algorithm to assess the risk of cardiovascular disease in older adults undergoing non-cardiac surgery. The algorithm analyzes electronic medical records (EMR) to make predictions in advance.

Cardiovascular complications such as myocardial infarction and stroke are among the most serious risks following surgery in elderly patients. Because many older adults also suffer from chronic conditions like hypertension, diabetes, and heart disease, undergoing general anesthesia, intraoperative bleeding, and inflammatory responses can place significant stress on the cardiovascular system.

Until now, the Revised Cardiac Risk Index (RCRI) has been the standard tool for assessing cardiovascular risk before surgery. However, RCRI relies on limited factors such as age, history of heart disease, and type of surgery, which has been criticized for reducing predictive accuracy.

Key information such as blood test results, medications, and prior diagnoses is excluded from RCRI assessments, making it harder for physicians to accurately determine a patient’s risk.

To address this gap, the SNUBH research team developed an AI model that integrates comprehensive patient data—blood test results, underlying diseases, medications, and surgery type—stored in EMRs. The model accurately predicts cardiovascular complications within 30 days following general (non-cardiac) surgery. The study was based on data from 46,000 patients at SNUBH and was externally validated using a cohort from Asan Medical Center.

The model achieved a predictive accuracy, measured by area under the receiver operating characteristic curve (AUROC), of up to 0.897—significantly outperforming the RCRI’s 0.704. The research team emphasized the significance of being able to predict postoperative cardiovascular risk quickly and easily in clinical settings without the need for additional diagnostic tests. Because the model was developed through a standardized process, it is expected to be applicable across various hospitals.

“Elderly patients show wide variation in health status even at the same age, so accurate prediction of postoperative cardiovascular complications is crucial for patient safety,” said Professor Suh. “We plan to continue developing the model so that it can be easily and quickly used by medical staff in conjunction with hospital systems.”