Data sources
We compiled geographic data on the number of reported COVID-19 cases per day from December 2019 to June 30, 2020. We collected the numbers of COVID-19 cases for 1,020 countries/regions from various sources (see S1 Appendix for a list of data sources for the COVID-19 cases). We then calculated the length of time (in days) since the onset of COVID-19 spread as defined by the date of the first confirmed case in each country or region. We also examined the number of SARS-CoV-2 PCR tests conducted based on data published by the World Health Organization (WHO) (https://ourworldindata.org/covid-testing) to assess the influence of sampling effort on the number of confirmed cases of COVID-19.
For each country or region, we compiled several environmental variables. For mapping cases of COVID-19, the longitude and latitude of the largest city and area for each country or region were extracted from GADM maps and data (https://gadm.org/index.html). Based on the geocoordinates of the cities, we collected the climatic data of mean precipitation (mm month–1) and temperature (°C) from January to June (WorldClim) using WorldClim version 2.1 climate data (https://www.worldclim.org/data/worldclim21.html) at a resolution of 2.5 arc-minutes grid cells that contained a country or region.
Regarding international travel linked to the disease transmission, we compiled the average annual number of foreign visitors (per year) for individual countries/regions from data published by the World Tourism Organization (https://www.e-unwto.org/toc/unwtotfb/current). We then calculated the relative amount of foreign visitors per population of each country or region to use in the analysis.
Regarding region-specific host susceptibility to COVID-19, we collected data on the following three epidemiologic properties: the proportion of the population aged over 65 years, the malaria incidence (per year), and information regarding bacillus Calmette–Guérin (BCG) vaccination. We included these attributes in our analyses based on the assumptions that BCG vaccination and/or recurrent treatment with anti-malarial medications could be associated with providing some protection against COVID-19 [13, 14]. We compiled BCG data from the WHO (https://www.who.int/malaria/data/en/) and (https://apps.who.int/gho/data/view.main.80500?lang=en) and the BCG Atlas Team (http://www.bcgatlas.org/) on the following five attributes: i) the number of years since BCG vaccination was started (BCG_year); ii) the present situation regarding BCG vaccination (BCG_type), split into all vaccinated, partly vaccinated, vaccinated once in the past, or never vaccinated; iii) the relative frequency of post-1980 (i.e., the past 40 years) BCG vaccination for people aged less than 1 year old (BCG_rate); iv) the number of BCG vaccinations (MultipleBCG), describing countries as never having vaccinated their citizens with BCG, vaccinated their citizens with BCG only once, vaccinated their citizens with BCG multiple times in the past, or currently vaccinate their citizens with BCG multiple times; and v) tuberculosis cases per 1 million people (TB). These BCG-related variables are strongly intercorrelated. Therefore, we reduced the dimensions of these variables (BCG_year, BCG_type, BCG_rate, MultipleBCG, and TB) by extracting the first axis of the PCA analysis: the score of the PCA 1 axis was negatively correlated with the five variables, so the PCA 1 score multiplied by –1 was defined as the BCG vaccination effect.
We also compiled socioeconomic data for each country or region. The population size, population density (per km2) (Gridded Population of the World GPW, v4.; https://sedac.ciesin.columbia.edu/data/collection/gpw-v4), gross domestic product (GDP in US dollars), and GDP per person were obtained from national census data (World Development Indicators; https://datacatalog.worldbank.org/dataset/world-development-indicators).