*2.1. Data Collection*

We began by conducting a targeted literature search for potential risk factors for COVID-19 mortality through Pubmed, using variations on terms including "COVID-19 mortality risk factors" and "pandemic mortality risk factors". Our approach was similar to a study conducted by Morales et al., investigating H1N1 influenza risk factors that varied by country [12]. In our review of studies on the current pandemic as well as past SARS and influenza pandemics, we identified 24 risk factors (Table S1) for COVID-19 fatality that were worth investigating, including quarantine policies, air travel activity, age distribution, comorbidities, healthcare access, availability of diagnostic tests, cumulative and daily testing data, and environmental factors such as air pollution and climate [5–7,13–22]. Regarding quarantine policies, government responses varied by country. For purposes of this study, we defined this intervention as the first date per international news sources when recommendations were made or legislation was passed limiting gathering size, closing non-essential business, or encouraging social distancing (Table S2). School closures and international travel bans were not considered as they only applied to certain individuals within each country's population.

To calculate CFR, we used the total confirmed cases and deaths for a given country from Our World in Data [4]. As higher testing rates per capita could be associated with an increased record of mild cases, we included total cumulative tests and tests per 1000 to explore the relationship between CFR and testing capacity per capita. Additionally, we included socioeconomic factors such as GDP, level of education, and scientific production. To examine healthcare-related factors, we included hospital beds, physicians per 1000, and per capita healthcare expenditure. We also looked at the availability of CT scanners and radiologists per one million population, as CT chest imaging represents a possibly limited resource that can increase the diagnostic accuracy. This was particularly true during the first wave of the pandemic, during which reverse transcription polymerase chain reaction (RT-PCR) kits were limited and imaging was relied upon by multiple countries for diagnosis of the disease [23]. Finally, given recent data linking the comorbidities of obesity and chronic lung disease to increased disease severity and poorer outcomes in cases of COVID-19 [16,19,24], we included prevalence of obesity, chronic obstructive pulmonary disease (COPD), and tobacco use, along with particulate matter as a measure of air pollution.

We decided to include in our analysis all countries with over 5000 COVID-19 deaths at the time of writing [4]. This cutoff was chosen to generate a set of at least 20 countries located in all hemispheres with diverse quarantine measures, GDPs, and geographic locations. 39 countries were found to meet this criterion and were included in the study. For each country, data was generated for the 24 variables from the following sources: Our World in Data project [4], World Bank database [25], OECD database [26], United Nations World Population Prospects [27], Global Health Data Exchange [28], and various international news sources (Table S2).
