Framework Whitepaper (PDF)
BC COVID-19 Modelling Group Presentation (PDF)
The speed and extent of the COVID-19 pandemic has challenged our abilities, as forecasters, like never before. Early data on the disease’s epidemiology is limited, records of cases and infections are incomplete, and the dynamics and scientific understanding of the disease are changing daily. Scientists from around the world have been quick to respond by developing a plethora of mathematical models to predict future COVID-19 infections and deaths.
Delivering this science to decision makers in an actionable form, however, remains a challenge. Without an inhouse team of modelers ready on stand-by, most policy makers are unable to direct modelling efforts towards their daily questions and circumstances. Instead they are often forced to rely on projections of questionable accuracy and relevance, made for different questions and/or jurisdictions, and often outdated by the time they are released.
Our solution to this challenge has been to develop a general software framework for providing real-time forecasts of COVID-19 infections and deaths that can be rapidly deployed for use anywhere in the world. Built upon our existing SyncroSim software platform, our framework allows end users to generate forecasts that are specific to their jurisdiction and questions. Through a collaboration with researchers from the University of British Columbia, Simon Fraser University, and the Pacific Institute for the Mathematical Sciences, the framework will provide access to the best of the world’s open-source forecasting models, along with real-time daily data, in a standardized and user-friendly format.
Using the latest source control techniques, scientists will be able to continually adapt and improve the framework’s underlying models as understanding of COVID-19 evolves over time. The framework also defines data formats and definitions for shared model inputs and outputs, ensuring that models can be consistently applied across jurisdictions regardless of the source of the input data; in this way the framework will allow decision makers to assess and compare alternative model projections for local accuracy and relevance, thus building confidence over time in their forecasts. Finally, the framework provides the flexibility for policy makers to introduce their own local “what-if” scenarios regarding the effects of possible future changes to public health measures. The result is a tool that generates locally responsive, meaningful, and ultimately actionable forecasts.
A prototype version of our framework, including a case study example providing real-time daily online COVID-19 forecasts for Canada and its provinces, is demonstrated through a short video; additional details are provided in an accompanying whitepaper.