Prognostic Significance of AI-Enhanced ECG for Emergency Department Patients

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Informatics

Background:

• Objective risk stratification using artificial intelligence (AI)-enabled electrocardiogram (ECG) analysis can aid clinical decision-making.
• The prognostic value of serial ECGs, especially pre-discharge follow-up ECGs, for predicting patient outcomes remains under-explored.
• Dynamic changes in AI-predicted ECG risk scores may enhance the prediction of post-discharge adverse events.

Purpose of the Study:

• To evaluate the prognostic utility of AI-enabled ECG risk scores derived from serial ECGs.
• To determine if dynamic changes in AI-ECG risk scores improve post-discharge outcome prediction.
• To assess the value of pre-discharge follow-up ECGs in risk stratification for emergency department (ED) patients.

Main Methods:

• Retrospective cohort study of 11,508 ED visits with direct discharge and two ECGs.
• AI-enabled ECG models were used to stratify mortality risk (low, medium, high) based on initial and follow-up ECGs.
• Area Under the Curve (AUC) calculated for predictive performance; Kaplan-Meier curves and Cox proportional hazards models analyzed 90-day mortality.

Main Results:

• AI-enabled ECG risk prediction models showed dynamic risk transitions between initial and follow-up ECGs.
• The AUC for mortality risk prediction was 78.6% for the initial ECG and 83.3% for the follow-up ECG.
• A significant increase in 90-day mortality was observed in patients transitioning from low to medium/high risk at discharge (HR: 6.01).

Conclusions:

• AI-enabled pre-discharge ECGs provide superior mortality risk stratification compared to initial ECGs for ED patients.
• Patients identified as medium or high risk at discharge warrant closer clinical evaluation.
• AI-ECG serves as a valuable tool to support clinical judgment and enhance safe discharge decisions in dynamic ED settings.