Epic Deterioration Index accuracy, bias
Occurred: May 2021
Physicians Vishal Khetpal, MD, and Nishant Shah, MD have published an op-ed on the Epic Deterioration Index (EDI), a machine learning algorithm developed by private electronic health record company Epic Systems that helps identify when to move a patient in or out of intensive care.
Rolled out at speed during the COVID-19 pandemic, the proprietary prediction tool has been taken up by hundreds of hospitals across the US.
However, despite general concerns about medical system accuracy and bias, Epic is limiting expert access to raw data and equations. Nor was the system independently validated or peer-reviewed before launch.
Epic holds over 250 million patients' electronic records in the US.
Operator: Parkview Health; University of Michigan; Multiple
Developer: Epic Systems Corporation
Purpose: Predict patient outcomes
Technology: Machine learning
Issue: Accuracy/reliability; Bias/discrimination - race, gender
Transparency: Governance; Black box
Published: December 2021