Real-time forecasts and risk assessment of novel coronavirus (COVID-19) cases: A data-driven analysis

Tanujit Chakraborty; Indrajit Ghosh

In the face of rapidly changing data for COVID-19, we calculated the case fatality ratio estimates for 50 countries from the day of starting the outbreak to 4 April 2020 from the following website 2 . A lot of preliminary analysis is done to determine a set of possible variables, some of which are expected to be critical causal variables for risk assessments of COVID-19 in these affected countries. Previous studies [22; 30; 18; 5] have suggested that the total number of cases, age distributions, and shutdown period have high impacts on the CFR values for some of the countries. Along with these three variables, we also considered seven more demographic structures and disease characteristics for these countries as input variables that are likely to have a potential impact on the CFR estimates. Therefore, the CFR modeling dataset consists of 50 observations having ten possible causal variables and one numerical output variable (viz. CFR), as reported in Table 3 . The possible causal variables considered in this study are the followings: the total number of COVID-19 cases (in thousands) in the country till 4 April, 2020, population density per km 2 for the country, total population (in millions) of the country (approx.), percentage of people in the age group of greater than 65 years, lockdown days count (from the starting day of lockdown till April 4, 2020), time-period (in days) of COVID-19 cases for the country (starting date to April 4, 2020), doctors per 1000 people in the country, hospital beds per 1000 people in the country, income standard (e.g., high or lower) of the country and climate zones (e.g., tropical, subtropical or moderate) of the country. The dataset contains a total of 8 numerical input variables and two categorical input variables.