Adapting to a global outbreak.

Helping clinicians on the frontline and ensuring efficient allocation of resources within healthcare institutions.

There is a pressing need to develop tools that can help physicians diagnose COVID-19 rapidly, determine if different disease presentations warrant different types of treatment, flag patients at high risk of deteriorating, and ensure healthcare resources are attributed efficiently and equitably.

Through an established partnership with 8 hospital sites in Québec and 1 site in Ontario, we have already built a large database of biological signal data from patients with COVID-19. We are developing a collaborative data analysis infrastructure to pool data from multiple sites while minimizing the exchange of patient-level information, distributed and federated learning techniques. Risk prediction models will be developed to identify patients at high risk of COVID prior to the availability of definitive testing, characterize the trajectories of patients presenting different manifestations of the disease, identify patients at high risk of clinical deterioration, and make forecasts to plan hospital resources and staffing. The accuracy of predictions will be continuously verified using new cases, which will be identified in real time at participating hospital sites.

These predictive models will be used to build tools that can help physicians better treat patients with COVID-19, and provide actionable recommendations to support Canada's response to COVID-19.