What variables can better predict the number of infections and deaths worldwide by SARS-CoV-2? Variation through time

Using data from 50 very different countries (which represent nearly 70% of worlds population) and by means of a regression analysis, we studied the predictive power of different variables (mobility, air pollution, health & research, economic and social & geographic indicators) over the number of infected and dead by SARS-CoV-2. We also studied if the predictive power of these variables changed during a 4 months period (March, April, May and June). We approached data in two different ways, cumulative data and non-cumulative data. The number of deaths by Covid-19 can always be predicted with great accuracy from the number of infected, regardless of the characteristics of the country. Inbound tourism emerged as the variable that best predicts the number of infected (and, consequently, the number of deaths) happening in the different countries. Electricity consumption and air pollution of a country (CO2 emissions, nitrous oxide and methane) are also capable of predicting, with great precision, the number of infections and deaths from Covid-19. Characteristics such as the area and population of a country can also predict, although to a lesser extent, the number of infected and dead. All predictive variables remained significant through time. In contrast, a series of variables, which in principle would seem to have a greater influence on the evolution of Covid-19 (hospital bed density, Physicians per 1000 people, Researches in R & D, urban population...), turned out to have very little - or none- predictive power. Our results explain why countries that opted for travel restrictions and social withdrawal policies at a very early stage of the pandemic outbreak, obtained better results. Preventive policies proved to be the key, rather than having large hospital and medical resources.