The rapid rise of COVID-19 has challenged healthcare globally. The underlying risks and outcomes of infection are still incompletely characterized even as the world surpasses 1 million infections. Due to the importance and emergent need for better understanding of the condition and the development of patient-specific clinical risk scores and early warning tools, CD2H has developed a platform to support analytic and machine learning hypotheses on clinical data, without data sharing, to rapidly discover and implement approaches for care. CD2H and the University of Washington (UW) School of Medicine previously applied this approach in the successful EHR DREAM Challenge that focused on Patient Mortality Prediction.
COVID-19 DREAM Challenge questions:
Question 1: Of patients who had at least one encounter/visit at University of Washington Medicine and who had a COVID-19 test, can we predict who tested positive?
Question 2: For patients with a positive RT-PCR for COVID-19 and who were tested at an outpatient visit, which patients were admitted to the hospital within 21 days of their RT-PCR test?
Submitted models can be trained and evaluated on University of Washington EHR data. The dataset is updated regularly. (Updates are nonbreaking; models previously submitted will still work on the updated infrastructure.)
Useful resources are available and include:
The EHR COVID-19 DREAM challenge is made possible by funding from the National Center for Advancing Translational Sciences (NCATS) and partnerships with the University of Washington School of Medicine Departments (Anesthesiology and Pain Medicine, Radiology and Biomedical Informatics, and Medical Education), UW Medicine Information Technology Services, the Institute of Translational Health Sciences (ITHS), the National Center for Data to Health (CD2H), Sage Bionetworks, and the CLEAR Center.