A multi-stage SEIR(D) model of the COVID-19 epidemic in Korea

Background This paper uses a SEIR(D) model to analyse the time-varying transmission dynamics of the COVID-19 epidemic in Korea throughout its multiple stages of development. This multi-stage estimation of the model parameters offers a better model fit compared to the whole period analysis and shows how the COVID-19's infection patterns change over time, primarily depending on the effectiveness of the public health authority's non-pharmaceutical interventions (NPIs). Methods This paper uses the SEIR(D) compartment model to simulate and estimate the parameters for three distinctive stages of the COVID-19 epidemic in Korea, using a manually compiled COVID-19 epidemic dataset for the period between 18 February 2020 and 08 February 2021. The paper identifies three major stages of the COVID-19 epidemic, conducts multi-stage estimations of the SEIR(D) model parameters, and carefully infers context-dependent meaning of the estimation results to help better understand the unique patterns of the transmission of the novel coronavirus (SARS-CoV-2) in each stage. Results The original SIR compartment model may produce a poor and even misleading estimation result if it is used to cover the entire period of the epidemic. However, if we use the model carefully in distinctive stages of the COVID-19 epidemic, we can find useful insights into the nature of the transmission of the novel coronavirus and the relative effectiveness of the government's non-pharmaceutical interventions over time. Key messages Identifies three distinctive waves of the COVID-19 epidemic in Korea. Conducts multi-stage estimations of the COVID-19 transmission dynamics using SEIR(D) epidemic models. The transmission dynamics of the COVID-19 vary over time, primarily depending on the relative effectiveness of the government's non-pharmaceutical interventions (NPIs). The SEIR(D) epidemic model is useful and informative, but only when it is used carefully to account for the presence of multiple waves and context-dependent infection patterns in each wave.