What is the clinical evidence from this study?

Study design: Observational. Population: Atrial Fibrillation (n=4017). Intervention: Tempus ECG-AF model (ECG-AI). Primary outcome: New diagnosis of AF within 1 year (Sensitivity 31%, Specificity 92%, 95% CI 25-37 (sensitivity), 91-92 (specificity)).

What question did this study set out to answer?

The study aimed to externally validate an artificial intelligence-enabled ECG model for predicting the 1-year risk of atrial fibrillation.

April 10, 2026Open Access

Multi-Center Validation of an Artificial Intelligence-Enabled ECG Model to Predict 1-Year Risk of Atrial Fibrillation or Flutter

Key Result

The Tempus ECG-AF model predicted 1-year AF risk with a sensitivity of 31% (95% CI 25-37%) and a specificity of 92% (95% CI 91-92%).

Key Points

The study aimed to externally validate an artificial intelligence-enabled ECG model for predicting the 1-year risk of atrial fibrillation.
Retrospective study aggregating ECG data from three clinical sites.
Patients aged 65+ with no prior atrial fibrillation or pacer/defibrillator history were included.
Sensitivity and specificity of the ECG-AI model were evaluated against minimum thresholds of 20% and 85%, respectively.
4,017 patients analyzed, with 240 (6.0%) developing atrial fibrillation within 1 year.
ECG-AI model identified 391 patients as 'increased risk', with 74 actually developing AF (sensitivity 31%, 95% CI: 25-37%).
Specificity was 92% (with 3,460 out of 3,626 remaining AF-free, 95% CI: 91-92%).

Study Design

Type

Observational (n=4,017)

Multicenter

Yes

Structured PICO

Does an artificial intelligence-enabled ECG model accurately predict the 1-year risk of atrial fibrillation in patients aged 65 and older without prior AF?

Population

4,017 patients aged 65 and older with no prior atrial fibrillation or history of pacemaker/defibrillator use.

Intervention

Tempus ECG-AF artificial intelligence model applied to electrocardiograms to predict 1-year AF risk

Outcome

New diagnosis of atrial fibrillation within 1 year (evaluated via sensitivity and specificity against pre-specified minimums)

An AI-enabled ECG model successfully predicted 1-year incident atrial fibrillation with 92% specificity and 31% sensitivity in a multicenter external validation cohort of older adults.

Main Result

Effect estimate: Sensitivity 31%, Specificity 92% (95% CI 25-37 (sensitivity), 91-92 (specificity))

Abstract

BackgroundAtrial 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. ObjectiveThis study aimed to externally validate an ECG-AI model for predicting 1-year AF risk. MethodsIn 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. ResultsThe 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%). ConclusionThe results validate the performance of the Tempus ECG-AF model and support its clinical use for AF risk stratification.

Bookmark

View Full Paper