In marketing materials and internal documents, national electronic health record vendor Epic Systems has touted the ability of its sepsis algorithm to crunch dozens of variables to detect the life-threatening condition in advance, enabling doctors to expedite the delivery of potentially lifesaving antibiotics.
But STAT has learned it is using a curious piece of data to make its prediction: whether a doctor has already ordered antibiotics.
The use of that information, which has not been publicly disclosed by the company, is contributing to a discrepancy between the accuracy of the algorithm in Epic’s internal testing and its performance in the outside world. Those problems came into view after multiple health systems attempted to validate the tool, but found it performed significantly worse than Epic advertised in their own hospitals.
About the Author Reprints
Create a display name to comment
This name will appear with your comment
