Are we there yet? Adaptive SIR model for continuous estimation of COVID-19 infection rate and reproduction number in the United States.

Abstract

BACKGROUND\nThe dynamics of the COVID-19 epidemic vary due to local population density and policy measures. When making decisions, policy makers consider an estimate of the effective reproduction number R_t which is the expected number of secondary infections by a single infected individual.\n\n\nOBJECTIVE\nWe propose a simple method for estimating the time-varying infection rate and reproduction number R_t .\n\n\nMETHODS\nWe use a sliding window approach applied to a Susceptible-Infectious-Removed model. The infection rate is estimated using the reported cases for a seven-day window to obtain continuous estimation of R_t. The proposed adaptive SIR (aSIR) model was applied to data at the state and county levels.\n\n\nRESULTS\nThe aSIR model showed an excellent fit for the number of reported COVID-19 positive cases, a one-day forecast MAPE was less than 2.6% across all states. However, a seven-day forecast MAPE reached 16.2% and strongly overestimated the number of cases when the reproduction number was high and changing fast. The maximal R_t showed a wide range of 2.0 to 4.5 across all states, with the highest values for New York (4.4) and Michigan (4.5). We demonstrate that the aSIR model can quickly adapt to an increase in the number of tests and associated increase in the reported cases of infections. Our results also suggest that intensive testing may be one of the effective methods of reducing R_t.\n\n\nCONCLUSIONS\nThe aSIR model provides a simple and accurate computational tool to obtain continuous estimation of the reproduction number and evaluate the efficacy of mitigation measures.\n\n\nCLINICALTRIAL