Machine learning beats decades-old heart disease risk prediction

Researchers built a machine learning model that predicts which heart disease patients will have major cardiac events in the next 5 years — and it's roughly 50% more accurate than the standard clinical score that's been used for decades. This matters because doctors can now identify higher-risk patients more reliably, which changes who gets an implanted defibrillator and how aggressively they're monitored.

A clinical prediction tool that was dramatically more accurate than its predecessor finally exists and works on data doctors already collect — but this is still a preprint with no evidence it will change actual clinical practice or replace the entrenched standard score.