AI screening flags probable undiagnosed heart disease
Electrocardiogram models flagged probable disease in 4.8% of 201,475 people, while heart ultrasound models flagged 6.4% of 6,692 people.
Edited by Boris Djordjevic · AgingResearch.News
An observational study evaluated 203,655 people aged 50 to 95 across seven US healthcare systems. Their average age was 70, and 105,969 (52.0%) were women. Participants had an electrocardiogram, which records the heart’s electrical activity, and an echocardiogram, a heart ultrasound, within one year of each other.
Artificial intelligence models screened for transthyretin amyloid cardiomyopathy, a condition involving protein deposits in the heart. The electrocardiogram model flagged 9,656 of 201,475 people (4.8%) for probable disease. The ultrasound model flagged 428 of 6,692 people (6.4%). Among 9,408 screen-positive people without confirmatory testing, positive screens were associated with higher death risk than in age- and sex-matched controls. Adjusted hazard ratios—a measure of relative death risk—were 1.74 for the electrocardiogram model and 1.28 for the ultrasound model.
Why it matters
The study addressed how to identify an underdiagnosed heart condition in adults aged 50 to 95. The authors suggested that broader screening could enable timely identification to maximize the benefit of new therapies.
Caveats
The mortality analysis was observational and involved 9,408 screen-positive people who had not undergone confirmatory testing. The findings came from a conference abstract and may not apply outside ages 50 to 95 or to people without both heart tests.
