The rapid spread of the current pandemic has left researchers and clinicians scrambling to fill in the gaps in our understanding of the clinical presentation and outcomes of COVID-19. With the global infection rate having surpassed 4 million, the study of patient health records is imperative to gaining a deeper understanding of the disease to, among other things, help establish patient-specific clinical risk scores and develop early warning tools.
Synapse, the open-source data analysis platform of Sage Bionetworks, has developed a platform to support the testing of hypothesis: analysis and ML hypothesis, on clinical data to help researchers rapidly discover and implement novel approaches for care.
The platform does not require data sharing works by initially giving researchers access to synthetic data in the cloud to determine if their models are functional. If they are, the University of Washington Medicine will apply the submitted models to data within UW Medicine, that does not contain patient identifiers, in a controlled computer environment. The algorithm accuracy will be shared with the community for review, having shared no clinical data.
The first EHR Dream Challenge posted is to use to ML and predictive algorithms, to understand the risk factors that lead to a positive test. At first, only the University of Washington Medicine clinical data will be used but the project hopes to expand the effort to include other focus topics and data from additional sites nationally.
Submissions are now open for the first challenge question:
“Of patients who have at least one clinical encounter/visit at UW Medicine and who were tested for COVID-19, can we predict who is positive?”
You can apply here to validate your COVID-19 hypothesis with EHRs once per day. The results of the highest performing models will be used in the Fred Hutch CovidWatch study to prioritise study recruitment.
