Predicting Mortality Risk with a Single Preoperative ECG: "AI Also Identifies Candidates for Detailed Cardiac Testing"

Bundang Seoul National University Hospital research team analyzes 46,000 non-cardiac surgeries
92% of all patients classified as "low-risk"
Potential to reduce unnecessary tests

A recent study has found that artificial intelligence (AI) can predict the risk of mortality after surgery and identify patients who require additional cardiac examinations, using just a single preoperative electrocardiogram (ECG).

From the left, Professor Hongmi Choi of the Department of Cardiology at Seoul National University Bundang Hospital, resident Yerin Kim, Professor Youngjin Cho, and Professor Inae Song of the Department of Anesthesiology and Pain Medicine. Seoul National University Bundang Hospital

The research team comprising Hongmi Choi, Professor of Cardiology at Seoul National University Bundang Hospital, Resident Yerin Kim, Professor Youngjin Cho, and Professor Inae Song from the Department of Anesthesiology and Pain Medicine, analyzed 46,000 non-cardiac surgeries performed between 2020 and 2021. On July 20, they announced that the AI-ECG solution was able to predict the 30-day postoperative mortality risk with high accuracy and effectively identify patients who should undergo further cardiac examinations.

ECG is a test that records the electrical signals produced when the heart beats as waveforms. It is very difficult for the human eye to interpret all the minute differences and complex patterns present. Previously, Professor Junghi Kim from the Department of Emergency Medicine and Professor Youngjin Cho from the Department of Cardiology at Seoul National University Bundang Hospital recognized that AI could quantify risks that conventional diagnosis might miss, and developed “ECG Buddy,” an AI solution that is now the most widely used for ECG analysis in emergency rooms nationwide.

In this study, the researchers utilized the “AI Severity Score (QCG-Critical score)” generated by AI to predict the risk of death within 30 days after surgery. The results showed a clear correlation between the AI risk assessment and actual mortality: patients with an AI risk score below 10 had a 30-day mortality rate of 0.1%, while among the high-risk group with scores exceeding 40 points, the rate soared to 11.7%.

The predictive performance also exceeded that of existing evaluation methods. The area under the receiver operating characteristic curve (AUROC), representing 30-day postoperative mortality prediction accuracy, was 0.909—higher than the European Society of Cardiology (ESC) risk assessment (0.728), the Revised Cardiac Risk Index (RCRI, 0.725), and the American Society of Anesthesiologists (ASA, 0.886) classification.

The team also evaluated whether AI-ECG could effectively identify patients requiring preoperative cardiac imaging such as echocardiography or coronary CT. By combining eight AI biomarkers that reflect cardiac function and disease, patients were classified as “low-risk” if all values were in the normal range, or as “high-risk” if at least one abnormality was present.

As a result, 92.3% of all surgical patients were classified by the AI as low-risk. Among these, the rate of death or emergency coronary intervention within 30 days after surgery was only 0.2%. Notably, in 91% of preoperative cardiac imaging procedures performed on the low-risk group, no actual abnormalities were detected. The associated examination costs accounted for 62.8% of the total preoperative cardiovascular imaging expenses, indicating that using AI could reduce unnecessary tests and the overall medical cost burden.

The researchers highlighted the biggest strength of this study: the ability to assess risk using just a single preoperative ECG, without the need for complex clinical information or blood test results. ECGs are inexpensive and quick to perform, making them highly practical for use in clinical settings.

However, the team emphasized that AI-ECG should be used as an auxiliary tool to improve the efficiency of selecting high-risk patients for further testing, rather than as a replacement for conventional preoperative cardiac examinations.

Professor Youngjin Cho said, “AI-based ECG analysis is already being used in emergency care settings as a triage tool before more detailed cardiac testing. Because it is fast, simple, and cost-effective, this approach could also help efficiently identify high-risk patients and reduce the testing burden during routine preoperative assessments.”

The results of this study were published as two separate papers, one on AI-ECG’s mortality risk prediction and the other on its effectiveness for screening candidates for detailed cardiac exams, in the international journals 'European Heart Journal-Digital Health' and 'Journal of Medical Internet Research'.

This content was produced with the assistance of AI translation services.