Improved real-time surveillance of COVID-19 patients’ electronic health records using transfer learning and ordinal regression

“MIDAS enthusiastically supports Drs. Admon and Gillies research using data science to improve outcomes for COVID-19 patients by predicting, for each patient, which respiratory treatment is needed and at which time point.”

-H.V. Jagadish, Director, MIDAS

Led by Drs. Andrew Admon (Internal Medicine) and Christopher Gillies (Emergency Medicine), this team is using Machine Learning, a powerful data science tool, to build a real-time patient surveillance system. During times of unprecedented strain on healthcare personnel and clinical resources, this will help clinicians identify COVID-19 patients who need more intensive monitoring, closer nursing care, or urgent physician intervention.

The MIDAS COVID-19 Propelling Original Data Science (PODS) grants were awarded on May 13th, 2020, each of the 7 teams received funding of up to $30,000 with projects starting immediately and expected to finish by the end of 2020. These projects demonstrate the resolve, expertise, and creativity of U-M data scientists facing a public health crisis. To learn more about the other projects please visit our COVID-19 PODS awards page.

Andrew Admon, Clinical Lecturer, Pulmonary Diseases & Internal Medicine, University of Michigan