Tele-electrocardiography and bigdata: The CODE (Clinical Outcomes in Digital Electrocardiography) study

Area of Science:

• Cardiology
• Biomedical Informatics
• Epidemiology

Background:

• Digital electrocardiographs are prevalent, generating vast amounts of data.
• Existing digital electrocardiogram (ECG) datasets are often fragmented or lack comprehensive clinical linkage.
• The CODE (Clinical Outcomes in Digital Electrocardiology) study addresses the need for a structured, large-scale digital ECG database.

Purpose of the Study:

• To develop and validate a large-scale digital ECG database (CODE study).
• To integrate ECG data with national mortality information for epidemiological research.
• To explore the clinical applications of the database, including machine learning for ECG diagnosis and mortality risk assessment.

Main Methods:

• Organized over 1.5 million digital ECGs from a Brazilian telehealth network (2010-2017).
• Employed a hierarchical free-text machine learning algorithm and the Glasgow ECG Analysis Program for automated diagnosis.
• Linked ECG data to the national mortality system using probabilistic methods and validated diagnoses through manual review.

Main Results:

• Achieved high performance (F1 >80%, specificity >99%) using a deep neural network for detecting 6 ECG abnormalities.
• Identified atrial fibrillation (AF) as a significant predictor of cardiovascular and all-cause mortality, with higher risk in women.
• Established a robust database of over 1.5 million patients with linked ECG and mortality data.

Conclusions:

• A comprehensive digital ECG database linked to mortality data is a valuable resource for cardiovascular epidemiology and clinical research.
• Machine learning models demonstrate high accuracy in automated ECG interpretation.
• The database facilitates the study of ECG abnormalities and their impact on mortality, particularly AF in women.