Old data, new insights
AI has found a sudden-cardiac-death risk signal hiding in routine ECGs, not by inventing a new test, but by re-reading a test already run everywhere.
Researchers trained AI on more than 440,000 ECGs from Sweden and validated it on U.S. and Taiwan datasets; the AI-flagged group had about 7% annual sudden-cardiac-death risk versus 4.6% for reduced-LVEF screening. More than 86% would not have been caught by that standard marker.
Enabled by a model that translated the black-box output into a visible clue: subtle slurring in the QRS complex on lead aVL, a pattern not previously described in the medical literature.
WHY IT MATTERS
Your organization may already own its richest behavioural sensors buried in call transcripts, ticket histories, process logs, learning records, calendar patterns, and your customer journeys.
What can the latent signals in the data exhaust people already create teach you?
WHAT TO WATCH FOR
The medical field is moving from 'cool idea' to prospective trials, workflow alerts, FDA-cleared ECG-AI tools, EHR integration, wearable-quality signals, and reimbursement pathways.
One AI-ECG alert trial reduced 90-day mortality from 4.3% to 3.6%, while EchoNext has now received FDA clearance for ECG-based structural heart disease screening.
LIMITATIONS
These results are still preliminary. AI-ECG evidence remains heterogeneous, and biased toward idealized settings, and are not yet a reliable benchmark for real-world clinical performance. Watch this space.
SOURCE
https://www.nature.com/articles/s41586-026-10674-6
https://www.nature.com/articles/s43856-024-00423-5
BESCI AI OPINION
Do you have rich insights hidden in data that you don't value. Is it time to mine the data that you have to gain greater insights?
One of our common AI use cases is to take anonymised secondary data, especially if there are rich data text fields and use it to assess organisational health.
