Study links older AI-estimated heart age to higher long-term mortality

Heart ultrasound clips recorded more than 20 years ago have revealed a possible risk marker. Over 20 years, adults whose hearts looked older to an AI model had higher rates of death from cardiovascular or dementia-related causes. Presented November 13 at the ESC Digital & AI Summit 2026, the retrospective study does not show that using the estimate improves care.

Researchers asked whether a link seen in people with cardiovascular disease also appeared in the wider community. Earlier research had tied a gap between AI-estimated heart age and actual age to worse outcomes in cardiovascular patients, said study presenter Dr. Jeffrey G. Malins of Mayo Clinic in Rochester. The relationship in the general population had been less clear.

To investigate, researchers used records from 568 adults aged 50 or older who underwent echocardiography—ultrasound imaging of the heart—between 2002 and 2004.

A previously developed AI model estimated age from the heart-ultrasound clips. Researchers averaged its predictions for each participant and subtracted the person’s actual age to calculate the cardiac age gap. A positive gap meant the model estimated an older heart; the study tested whether larger gaps were linked to later deaths.

The first analysis accounted for participants’ age and sex. The second also accounted for body mass index and the clinical measurement called left ventricular ejection fraction. The association weakened with those additional factors included, but remained statistically significant. Its 95% confidence interval ran from 1.13 to 1.90, with a reported p-value of 0.004.

These figures show an association across the study group. They are not an individual’s chance of dying, and they do not prove that an older AI-estimated heart age causes death. The study linked measurements from earlier scans to later outcomes; it did not test whether sharing the estimates with doctors changes treatment or survival.

Malins said the findings suggest AI-derived cardiac aging patterns may help identify people with long-term vulnerability. He also proposed studying whether lifestyle changes or medical interventions can reduce the cardiac age gap.

Another next step is to separate the outcomes. Malins said cardiovascular and dementia-related mortality were combined for practical reasons. Analyzing them separately could help clarify the relationship between cardiovascular and neurodegenerative diseases, rather than treating the combined result as evidence that the AI identifies each risk equally well.

For now, the proposed next move is larger, independent prospective studies—research that follows participants forward. Malins called for those evaluations to test AI-derived cardiac aging signatures beyond this 568-person cohort.