Developing clinically relevant predictive models requires the availability of electronic health record (EHR) data which is complicated by concerns around patient privacy. A novel framework, the ‘Model to Data’ framework (MTD), eliminates researchers’ direct interaction with patient data by using containerization software. We showcase the utility of the MTD framework via a community challenge, the EHR DREAM Challenge: Patient Mortality Prediction, showing that participants can still build accurate predictive models using protected EHR data.
Describe the new knowledge and additional skills the participant will gain after attending your presentation.: Attendees of this talk will see how our ‘Model to Data’ evaluation infrastructure enables healthcare organizations to crowdsource predictive analytics solutions using protected health data without exposing patient data and will be given the resources to implement this method within their own institution.
Timothy Bergquist (Presenter)
University of Washington
Yao Yan, University of Washington
Thomas Schaffter, Sage Bionetworks
Thomas Yu, Sage Bionetworks
Vikas Pejaver, University of Washington
Noah Hammarlund, University of Washington
Justin Prosser, University of Washington
Justin Guinney, Sage Bionetworks
Sean Mooney, University of Washington