Randall Grout, M.D., M.S., explains how the UNAFIED model predicts AFib using existing EHR data, eliminating the need for additional testing.
Transcript:
One of the strengths of the UNAFIED predictive model is that it relied on data that was generally available inside the electronic health record. It didn’t require extra purposeful testing just for detecting atrial fibrillation. Rather, we could capture things that were already in the EHR like age, height and weight, prior diagnoses of heart disease or kidney disease or COPD, as well as common lab values that were likely to have already occurred in a patient’s history. In the case that those lab values or those variables were not available, we also accounted for that and allowed for missing values in the algorithm.
Dr. Grout shares how testing a predictive model algorithm can help physicians without burdening clinics.
Transcript:
Before a model gets implemented into routine use, it is important to do these testing steps. First, the testing in the development side before it touches any patients, either with simulated data or real-world data, and then to test it in a controlled environment and then increasingly in more routine environments. Overall, the physicians who used the algorithm on most clinic days were more likely to indicate that it was easy to use, it wasn’t time-consuming and it helped improve patient care more than those physicians who only used it occasionally. This was reassuring to us to know that the way we put it into practice could be useful and not an extra burden on a busy clinic.
