A Quadratic Trend-based Time Series Method to Analyse the Early Incidence Pattern of COVID-19

Journal: Biostatistics & Epidemiology

Abstract: COVID-19 has already affected more than 300,000 people. In this study, the authors propose an appropriate auto-regressive integrated moving average model with time-varying parameters to analyse the trend pattern of the early incidence of COVID-19 outbreak, and subsequently estimate the basic reproduction number R0 for different countries. They also incorporate information on total or partial lockdown into the model. The model is concise and flexible in structure. For R0, they use maximum likelihood method and estimate it for different serial interval distributions. Proper diagnostic measures establish that a time-varying quadratic trend successfully captures the incidence pattern of the disease. They find that the number of affected cases starts increasing more rapidly 3 to 4 weeks after the first case is identified. Countrywide lockdown has been effective in reducing the growth rate of the disease in Italy. Estimated R0 of COVID-19 ranges between 1.4 to 3.2 in different countries, except for the United States where it is higher. A much needed outcome is that the method gives insight on what epidemiological stage a region is in. This has the potential to help in prompting policies to address COVID-19 in different countries.