Forecasting the outcome and estimating the epidemic model parameters from the fatality time series in COVID-19 outbreaks (2004.08973v2)

Abstract: In the absence of other tools, monitoring the effects of protective measures, including social distancing and forecasting the outcome of outbreaks is of immense interest. Real-time data is noisy and very often hampered by systematic errors in reporting. Detailed epidemic models may contain a large number of empirical parameters, which cannot be determined with sufficient accuracy. In this paper, we show that the cumulative number of deaths can be regarded as a master variable, and the parameters of the epidemic such as the basic reproduction number, the size of the susceptible population, and the infection rate can be determined. In the SIR model, we derive an explicit single variable differential equation for the evolution of the cumulative number of fatalities. We show that the epidemic in Spain, Italy, and Hubei Province, China follows this master equation closely. We discuss the relationship with the logistic growth model, and we show that it is a good approximation when the basic reproduction number is less than $2.3$. This condition is valid for the outbreak in Hubei, but not for the outbreaks in Spain, Italy, and New York. The difference is in the shorter infectious period in China, probably due to the separation policy of the infected. For more complex models, with more internal variables, such as the SEIR model, the equations derived from the SIR model remain valid approximately, due to the separation of timescales.