Machine-learning models predict death risk after heart attack better than standard score

Machine-learning models developed at Amsterdam UMC predicted death in hospital and within six months after treatment for a severe heart attack better than the widely used GRACE 2.0 score, according to research presented at the European Society of Cardiology's Digital & AI Summit, the ESC said.

The PERISCOPE models were trained on 38,489 US patients, tested on 119,035 more and validated on 6,776 Amsterdam UMC records, with consistent results across all three. Presenter Andrea Rafanelli said they could help find high-risk patients after stenting, but their effect on outcomes still needs prospective studies.