Skip to main content

University of Michigan School of Information


Singh: independent investigation into healthcare algorithm reveals inaccuracy

Headshot of Karandeep Singh placed next to graphic of a speech bubble

Wednesday, 06/23/2021

The Verge covered a new study led by Karandeep Singh, M.D., assistant professor at the University of Michigan School of Information and School of Medicine. Singh investigated the accuracy of a predictive modeling tool created to identify early signs of sepsis in patients.

Epic Systems, the largest electronic medical records company in the United States, claims that they developed a machine learning tool which scans patient testing data to identify signs of sepsis. Hundreds of hospitals have adopted Epic’s prediction tool including the University of Michigan health system. By investigating data reported from Michigan Medicine, Singh determined that Epic’s algorithm fails to dramatically increase identification and produces significant false positive alerting.

Read the full article from The Verge and Singh’s Twitter thread about the study.

Read the study: External Validation of a Widely Implemented Proprietary Sepsis Prediction Model in Hospitalized Patients