Machine learning method for predicting pacemaker implantation following transcatheter aortic valve replacement.

Area of Science:

• Cardiology
• Medical Informatics
• Biomedical Engineering

Background:

• Accurate prediction of permanent pacemaker implantation (PPI) risk post-transcatheter aortic valve replacement (TAVR) is vital for clinical decision-making.
• Investigating the utility of pre- and post-TAVR ECG data is essential for refining risk assessment.
• Comparing machine learning (ML) with traditional logistic regression (LR) can improve predictive accuracy.

Purpose of the Study:

• To evaluate the predictive value of pre- and post-TAVR ECG parameters for PPI risk.
• To compare the performance of ML algorithms against LR in predicting PPI after TAVR.
• To identify key ECG markers associated with increased PPI likelihood.

Main Methods:

• Analysis of 557 patients in sinus rhythm undergoing TAVR for severe aortic stenosis.
• Collection of baseline demographics, clinical data, pre-TAVR and post-TAVR ECGs, and echocardiographic data.
• Development and comparison of Random Forest (RF) and logistic regression models for PPI risk prediction.

Main Results:

• 17.1% of patients required PPI post-TAVR.
• Optimal delta PR cutoff of 20ms (sensitivity 0.82) and delta QRS cutoff of 13ms (sensitivity 0.68) were identified.
• RF model with post-TAVR ECG data achieved higher AUC (0.81) than without (0.72) and outperformed LR (AUC 0.69).

Conclusions:

• Random Forest methodology demonstrates superior performance over logistic regression for predicting PPI risk post-TAVR.
• Incorporating post-TAVR ECG data significantly enhances the accuracy of PPI risk prediction models.
• ML approaches offer a powerful tool for improving clinical decision-making in TAVR patients.