[Sportschosun Reporter Jang Jong-ho] A study has found that artificial intelligence (AI) can predict the future risk of atrial fibrillation even in people without symptoms, and can also provide meaningful support for clinicians' decision-making.
A research team led by Professor Park Jun-beom of the Department of Cardiology at Ewha Womans University Mokdong Hospital, Professor Shin Tae-young of the Department of Urology, and Professor Kim Ye-ji of the Department of Cardiology at Ewha Womans University Seoul Hospital recently published a study titled "Impact of an AI algorithm for multi-day prediction of incident atrial fibrillation on clinical decisionmaking: PROVISION-AF study" in npj Digital Medicine, an international journal on integrative medicine with an impact factor of 18.0.
Atrial fibrillation is one of the most common arrhythmias and can increase the risk of stroke, heart failure, and death. However, because it often causes no symptoms or occurs only intermittently, it is frequently difficult to diagnose before serious complications develop. To address this, the team developed an AI model that can predict the risk of atrial fibrillation within three days to one month using only routine electrocardiogram results from people with no notable symptoms.
The model was validated from multiple angles using data from more than 110,000 patients, including data from Ewha Medical Center, large external data from Beth Israel Deaconess Medical Center (BIDMC) in the United States, and the prospective PROVISION-AF cohort involving seven university hospitals in Korea. In particular, the researchers applied a systematic three-stage AI training process, significantly improving both predictive accuracy and reliability. As a result, the model achieved an area under the receiver operating characteristic curve (AUROC) of 0.87 in in-house validation using Holter monitoring, and 0.75 in external validation involving real patients from other hospitals, demonstrating strong performance.
In addition, the team conducted a structured simulation survey of 70 physicians from Korea, the United States, and other countries to evaluate how AI-based risk predictions affect real-world clinical decision-making.
The results showed that when physicians used AI-generated risk predictions alongside their own assessments, their ability to distinguish atrial fibrillation risk improved significantly, with AUROC rising from 0.573 to 0.650. Negative predictive value also increased from 0.764 to 0.839. In particular, the ability to correctly identify normal patients, measured by the net reclassification improvement index (NRI 0.127), improved substantially. This suggests the model could help reduce unnecessary overmonitoring and support more efficient allocation of medical resources. The performance gains were observed regardless of specialty or years of experience, with the most notable improvement seen among general cardiologists.
Professor Park Jun-beom, the corresponding author, said, "This study is meaningful because it confirms that the AI-ECG model goes beyond pure prediction and has the potential to serve as a practical digital decision-support tool that complements clinicians' diagnostic accuracy and refines patient management strategies in real clinical settings." He added, "We will continue follow-up studies to determine whether it ultimately improves patient outcomes and increases medical efficiency in actual practice."
Meanwhile, the study was carried out with support from the Korea Health Industry Development Institute and the National Research Foundation of Korea (NRF).
Reporter Jang Jong-ho bellho@sportschosun.com
This article has been translated by GripLabs Mingo AI.
