Abstract
Atrial fibrillation (AF) is the most common cardiac arrhythmia associated with increased risk of stroke and heart failure. AF is often asymptomatic and paroxysmal, making diagnosis challenging. Artificial intelligence (AI) applied to electrocardiogram (ECG) interpretation is a promising approach for improved diagnosis. While ECG-AI studies have shown promise, the common practice of evaluating based on data from single institutions may overestimate performance. External validation is essential to ensure AI models generalize well to diverse settings and populations.
This study aimed to externally validate an ECG-AI model for predicting 1-year AF risk.
In this retrospective study, ECG data from three clinical sites were aggregated and patients’ charts manually abstracted to define inclusion and exclusion criteria (age 65+ with no prior AF or history of pacer/defibrillator use), and endpoints (new diagnosis of AF within 1 year or 1 year of AF-free follow-up). The sensitivity and specificity of a risk score from an ECG-AI model were evaluated against pre-specified minimum values of 20% and 85%, respectively.
The analysis included 4,017 patients, with 240 (6.0%) developing AF within 1 year. The ECG-AI model returned an “increased risk” result for 391 patients (9.7%), including 74 who developed AF (sensitivity=31% ; 95% CI: 25–37%). A total of 3,626 patients had a “no increased risk” result, with 3,460 remaining free from AF (specificity=92% ; 95% CI: 91–92%).
The results validate the performance of the Tempus ECG-AF model and support its clinical use for AF risk stratification.