A machine-learning model has predicted ten-year cardiovascular risk with roughly 90% accuracy in a large validation study, outperforming conventional risk scores currently used in clinical practice.
The model draws on routine electronic health record data, including blood pressure trends, lipid panels and prescription history, and required no additional testing.
Clinicians noted that prospective validation is still required before the tool can guide treatment decisions, and cautioned that model performance may vary across populations not represented in the training data.
