Artificial Intelligence-Enabled Electrocardiography Predicts Future Pacemaker Implantation and Adverse Cardiovascular...

Area of Science:

• Cardiology
• Artificial Intelligence in Medicine
• Medical Diagnostics

Background:

• Increasing life expectancy leads to more permanent pacemaker implants.
• Predicting pacemaker implantation (PMI) is challenging due to non-specific symptoms in conditions like sick sinus syndrome.
• Electrocardiogram (ECG) data holds potential for predicting future cardiovascular events and the need for PMI.

Purpose of the Study:

• To develop a deep learning model (DLM) for predicting future PMI from ECG data.
• To evaluate the DLM's ability to predict cardiovascular mortality and incident cardiovascular disease.
• To identify key ECG features predictive of PMI and adverse cardiovascular outcomes.

Main Methods:

• Trained a DLM on a large dataset of 158,471 ECGs from academic medical center patients.
• Validated the DLM on independent datasets from medical centers (25,640 patients) and community hospitals (26,538 patients).
• Analyzed prediction accuracy for PMI within 90 days and assessed risks for mortality and cardiovascular events using hazard ratios.

Main Results:

• The DLM achieved high accuracy in predicting PMI within 30, 60, and 90 days (AUCs 0.870-0.883) with excellent sensitivity and specificity.
• Identified significant ECG predictors including PR interval, corrected QT interval, heart rate, and QRS duration.
• Patients identified by the AI-DLM showed significantly higher risks of PMI, all-cause mortality, CVD mortality, and new cardiovascular events.

Conclusions:

• A deep learning model utilizing ECG data can accurately predict future pacemaker implantation.
• The AI-DLM effectively identifies patients at elevated risk for adverse cardiovascular outcomes and mortality.
• This AI tool can assist clinicians in timely intervention for patients at risk of PMI and related complications.