Electronic Health Record-Based Machine-Learning for Cardiac Arrhythmia Identification in Adults with High-Risk Chronic Disease: Model Development and Validation Study

XGBoost trained on 2018–2025 EHR data identifies documented cardiac arrhythmia diagnoses in adults with high-risk chronic disease, achieving AUROC 0.894 in a 3,051-patient holdout set.

• Why it matters: Arrhythmias in adults with chronic conditions may go unrecognized, and existing risk tools focus mainly on atrial fibrillation, leaving other arrhythmias harder to detect.
• What they did: Using retrospective EHR records from one healthcare organization, logistic regression, random forest, and XGBoost were compared; XGBoost was reduced to 30 features via Shapley additive explanations and tested on a holdout set ending one day before the index date.
• The result: Among 30,356 patients, 6,835 (22.5%) had qualifying arrhythmia diagnoses; at a 0.5 threshold sensitivity was 0.564, specificity 0.971, and precision 0.857, and atrial fibrillation/flutter AUROC was 0.883 versus 0.712 (CHARGE-AF) and 0.554 (HARMS2-AF).

AI summary of the abstract · may contain errors · not medical advice · read the paper.