A CAD Risk-Prediction Tool Derived from Electronic Health Record Data

The EHR score performed better than both the pooled cohorts equation and a polygenic risk score.

Researchers wanted to know if the enormous amount of data available in electronic health records (EHRs) could be used to build a tool that predicts the presence of coronary artery disease (CAD) a year before it is clinically apparent — and if it could do so better than can current guidelines based on the pooled cohort equations (PCE) tool or a polygenic risk score for CAD.

To find out, the investigators applied a machine-learning framework to EHR data — comprising 107 laboratory variables and 9 clinical variables, measured as part of routine clinical care — from a multiethnic clinical cohort of 555 CAD cases and 6349 controls. For external validation, the same framework was applied to 3130 CAD cases and 378,344 controls from the UK Biobank, …

Reviewing Author

Karol E. Watson, MD, PhD, FACC

Disclosures

Consultant/Advisory BoardBoehringer Ingelheim; Amgen; Novartis; Lilly; Amarin

Speaker’s BureauBoehringer Ingelheim

Grant/Research SupportNIH

Editorial BoardsCirculation (Associate Editor)