Challenges of Forecasting the Spread and Mortality of COVID-19   [open pdf - 2MB]

From the Key Takeaways: "[1] While statistical models can be useful tools for tracking COVID-19 [coronavirus disease 2019], they are only as accurate as the input assumptions, which depend on continually changing data. [2] Widespread, randomized testing is critical to generating credible forecasts and developing a full understanding of the disease. [3] To make well-informed decisions, policymakers must use a range of forecasts and fully understand how sensitive models are to the smallest changes in assumptions."