Loading...
Algorithms are increasingly embedded in the electronic health record (EHR) with the goal of improving practice. However, there is a concomitant need to evaluate these tools and scrutinize their results. In this systematic review, researchers sought to evaluate models predicting incident atrial fibrillation from community-based EHRs.
They identified 11 suitable studies based on 9 cohorts from 8 EHR databases. The data used for the models were largely simple, such as the presence or absence of chronic heart failure. Only one model had variables capturing change over time, such as change in body-mass index. The major issue identified was that 96% of the model results were considered to have a high risk of bias. This assessment was due to exclus…