Stanford-CIDE Coronavirus
Simulation Model

Full technical details of the SC-COSMO modeling framework are posted in the Resources section; here, we provide a brief technical overview of its capabilities. Because the framework is intended to be used for modeling many different populations and geographies, it also includes the ability to easily load/parametrize population- and geography-specific inputs that are generally publicly available (e.g., the age-structure of a given county), methods to infer setting-specific inputs that may not be available (e.g., realistic contact matrices) based on demographic features and a database of existing empirically-derived contact matrices, and to calibrate to setting-specific COVID-19 data (e.g., detected cases time-series, testing time-series, etc.) accounting for incomplete detection and uncertainty. Benchmarking even using a single modern computer to date shows that the calibration and projection/forecasting features of the model under multiple scenarios are fast (e.g., on the order of 10s of seconds to 50+ states or counties), and the framework is designed with open source tools and in such a way that it can be scaled using symmetric parallel processing to further speed its simulations especially useful when propagating uncertainty to its outcome projections.