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

Tanujit Chakraborty; Indrajit Ghosh

From the variable importance plot based on the complexity parameter of the RT model (also see Figure 7 ), seven causal variables are obtained out of 10 potential input variables having higher importance. These seven causal variables that significantly affect the CFR for 50 most affected countries are the followings: total number of COVID-19 cases in the country (in thousands), percentage of people in the age group of greater than 65 years, total population (in millions) of the country, doctors per 1000 people in the country, lockdown period (in days) for the country, time-period (in days) of COVID-19 cases for the country, and hospital beds per 1000 people in the country. Our results are consistent with previous results obtained by [30; 18; 5] , where the authors suggested that the total number of cases, age distributions, and shutdown period have high impacts on the CFR estimates. But interestingly, we obtained four more essential causal variables that will provide some new insights into the study of risk assessments for COVID-19 affected countries. Out of these 7 numerical input variables, there are four control variables (number of cases, people of age group > 65 years, lockdown period, and hospital beds per 1000 people) present that can be managed to fight against this deadly disease. Once these variables are taken care of, the respective country may reduce their case fatality rate at a significant rate. x.x3