Discussion of “An epidemiological forecast model and software assessing interventions on the COVID-19 epidemic in China”

Abstract

Predictions during the early stage of an epidemic are essential to inform quarantine protocols, plan medical resources, and implement economic strategies. One of the major obstacles for making accurate predictions is imperfect available data on the current state of the disease dynamics typically summarized via the number of infected and recovered cases, and disease-related deaths. In the case of rapidly evolving COVID-19 pandemic, major challenges include underreporting of infected and recovered cases due to the shortage of available tests, non-uniform performance and not sufficiently high sensitivity/specificity of currently available tests, a long incubation period, and a significant number of asymptomatic/unconfirmed cases, inconsistent accounting practices in death classification. All these make it quite challenging to accurately predict the future trajectory of this pandemic.Putting aside the imperfections in the currently available data, choosing an appropriate model is another essential component of the modelling endeavor. Modifications of the classical SIR (Susceptible-Infectious-Removed) and SEIR (Susceptible-Exposed-Infectious-Removed) models are among the most popular modelling frameworks used during the early stage of this pandemic. Ironically, even the notorious Institute for Health Metrics and Evaluation at the University of Washington that originally proposed an misleadingly inflexible curve-fitting model (The Institute for Health Metrics and Evaluation, 2020) that received vigorous criticism within statistical and epidemiological communities (see (Brookmeyer, 2020;Jewell et al., 2020;Jin