Abstract Body (Do not enter title and authors here): Background: Cardiovascular mortality risk prediction traditionally relies on standard modifiable risk factors (smurfs) including hypertension, diabetes, smoking and cholesterol. However, patients without smurfs, so-called “smurfless” individuals, also face substantial cardiovascular risk that is poorly identified by current models.

Objectives: We evaluated the ten year prognostic utility of the embedded survival Contrastive Augmentation by Patient Electrocardiogram (esCAPE) model for predicting cardiovascular mortality, particularly in traditionally low risk smurfless populations.

Methods: CAPE, a pre-trained AI-enhanced ECG model from over six million multicontinental ECGs, was embedded with a survival architecture (esCAPE) using a time-to-event loss. Internal validation included 10-year predictions for cardiovascular death and secondary lethal events (heart failure, stroke, intracranial haemorrhage, sudden cardiac death, acute myocardial infarction) in the BIDMC dataset (n=62,613). Cox models with esCAPE scores were evaluated using C-index and NRI compared to SCORE2 & PREVENT. Echo-wide association studies identified structural correlates of high esCAPE scores.

Results: esCAPE achieved a C-index of 0.834 for cardiovascular mortality, with equivalent or better performance in smurfless individuals (C-index: 0.846). Predictive performance declined stepwise with more smurfs (down to 0.734 for four smurfs). esCAPE significantly improved discrimination when added to SCORE2 and PREVENT (NRI 0.48 & 0.57 in smurfless cohort). Echo-wide analysis showed high-risk esCAPE scores correlated with larger ventricular size, lower ejection fraction, and higher right-sided pressures in both smurfless and smurf groups. Secondary lethal endpoints had C-indices ranging 0.769-0.876, with 0.826 for sudden cardiac death.

Conclusions: esCAPE offers a precise, low-cost, non-invasive cardiovascular risk prediction tool with particularly strong, additive performance in smurfless individuals. Following external validation, it may support prospective clinical integration for preventive strategies in both high- and low-risk patients.