Predicting Patient Risk and Outcomes Using Artificial Intelligence-Predicting Patient Risk and Outcomes Using Artificial Intelligence

Video Transcription

Video Summary

At the HRS 2025 conference, a presenter discussed a new artificial intelligence (AI) model designed to predict all-cause mortality using electrocardiograms (ECGs). The model leverages AI to discern subtle changes and patterns in ECG wave intervals, which are typically undetectable to the human eye. The importance of mortality prediction was emphasized, especially in seemingly stable patients who can deteriorate rapidly. Utilizing a large database of over 5 million ECGs, each paired with mortality data, the model was trained and validated to predict mortality over various timeframes.<br /><br />The research involved cleaning ECG data with frequency filtering and employing a deep neural network model to predict outcomes. The model demonstrated high accuracy, particularly for short and medium-term mortality predictions, with an AUC greater than 0.9 in internal validation, but performed less effectively in long-term predictions on external datasets. The study highlighted AI's potential in predicting mortality, particularly in short and medium-term scenarios, while acknowledging challenges in long-term prediction and emphasized future plans for clinical integration and trial evaluations to enhance patient care.

Keywords

AI model

all-cause mortality

electrocardiograms

ECG patterns

deep neural network

mortality prediction

short-term prediction

medium-term prediction

clinical integration