10-Year Risk Prediction of Higher-Grade AV Block in Patients with First-Degree AV Block

Area of Science:

• Cardiology
• Medical Informatics
• Artificial Intelligence in Medicine

Background:

• First-degree atrioventricular (AV) block, traditionally considered benign, is increasingly recognized as a potential risk factor for progression to higher-degree AV block.
• Early identification of individuals at risk for AV block progression is crucial for timely clinical intervention.

Purpose of the Study:

• To develop and externally validate a machine learning model for predicting the progression of first-degree AV block to higher degrees.
• To identify key electrocardiogram (ECG)-derived parameters and clinical factors predictive of AV block progression.

Main Methods:

• A retrospective cohort study utilizing 12-lead ECG data from two hospitals for model development and external validation.
• A Random Forest machine learning algorithm was employed, trained on six ECG parameters (RR interval, P duration, PR segment, PR interval, QRS duration, QT interval) and patient age and sex.
• SHAP (SHapley Additive exPlanations) analysis was used to interpret the model and identify important predictors.

Main Results:

• The machine learning model demonstrated strong predictive performance, achieving an area under the receiver operating characteristic curve (AUROC) of 0.823 in internal validation and 0.808 in external validation.
• Key predictors identified by SHAP analysis included PR segment duration, QRS duration, and patient age.
• The model's performance was further supported by area under the precision-recall curve (AUPRC) values of 0.719 (internal) and 0.894 (external).

Conclusions:

• A validated machine learning model can effectively stratify the risk of AV block progression using readily available ECG parameters and basic clinical data.
• This predictive tool can assist clinicians in making informed decisions for patients with first-degree AV block.
• The findings highlight the potential of AI-driven analysis of ECG data for early disease detection and risk assessment in cardiology.