key: cord-0776570-0r0zdpds authors: Leffler, C. T.; Ing, E. B.; Lykins, J. D.; Hogan, M. C.; McKeown, C. A.; Grzybowski, A. title: Association of country-wide coronavirus mortality with demographics, testing, lockdowns, and public wearing of masks. date: 2020-05-25 journal: nan DOI: 10.1101/2020.05.22.20109231 sha: 9d7e3fb4846e14ad352844fda91915404a43338e doc_id: 776570 cord_uid: 0r0zdpds Background. Wide variation between countries has been noted in per-capita mortality from the disease (COVID-19) caused by the SARS-CoV-2 virus. Determinants of this variation are not fully understood. Methods. Potential predictors of country-wide per-capita coronavirus-related mortality were studied, including age, sex ratio, temperature, urbanization, viral testing, smoking, duration of infection, lockdowns, and public mask-wearing norms and policies. Multivariable linear regression analysis was performed. Results. In univariate (but not multivariable) analyses, prevalence of smoking, per-capita gross domestic product, and colder average country temperature were positively associated with coronavirus-related mortality. In a multivariable analysis of 183 countries, urbanization, the duration of the infection in the country, and percent of the population at least 60 years of age were all positively associated with per-capita mortality, while duration of mask-wearing by the public was negatively associated with mortality (all p<0.001). In countries with cultural norms or government policies supporting public mask-wearing, per-capita coronavirus mortality increased on average by just 5.4% each week, as compared with 48% each week in remaining countries. In the multivariable analysis, lockdowns tended to be associated with less mortality (p=0.31), and per-capita testing with higher reported mortality (p=0.26), though neither association was statistically significant. Conclusions. Societal norms and government policies supporting the wearing of masks by the public are independently associated with less mortality from COVID-19. The COVID-19 global pandemic caused by infection with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has presented a major public health challenge. For reasons that are not completely understood, the per-capita mortality from COVID-19 varies by several orders of magnitude between countries. 1 Numerous sources of heterogeneity have been hypothesized. Higher mortality has been observed in older populations and in men. 2, 3 Patient-level behaviors, such as smoking, might also have an impact. 3 Other potentially relevant factors include economic activity, and environmental variation, such as temperature. 4 More urban settings and increased population density would be expected to enhance viral transmission. 5 In addition, public health responses to the COVID-19 pandemic may influence per-capita mortality. Various strategies have been implemented, ranging from robust testing programs to lockdown or stay-at-home orders, to mandates regarding social distancing and face mask usage. Practices with theoretical benefit, such as social distancing, stay-at-home orders, and implementation of mandates regarding use of masks in public spaces, must be assessed quickly, as implementation has the potential to reduce morbidity and mortality. Mask usage by the public is postulated to decrease infection by blocking the spread of respiratory droplets, 1 and was successfully implemented during other coronavirus outbreaks (i.e. SARS and MERS). 6 In the context of the ongoing pandemic, we assessed the impact of masks on per-capita COVID-19-related mortality, controlling for the aforementioned factors. We hypothesized that in countries where mask use was either an accepted cultural norm or favored by government policies on a national level, the per-capita mortality might be reduced, as compared with remaining countries. Methods. Data from 183 countries for which coronavirus mortality and testing data were available were retrieved from a publicly available source on May 9, 2020. 7 Archived testing data for April were also downloaded. 8 The date of the country's first reported infection and first death were obtained from the European Centre for Disease Prevention and Control. 9 Mean temperature in each country during the pandemic was estimated using the average monthly temperature in the country's largest city from public sources. 10, 11 Online news reports were searched to identify countries in which the public wore masks early in the outbreak based on tradition, as well as countries in which the . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 25, 2020. . https://doi.org/10.1101/2020.05.22.20109231 doi: medRxiv preprint national government mandated or recommended mask-wearing by the public before April 16, 2020. Population, 12 fraction of the population age 60 years and over, and age 14 and under, male: female ratio per country, 13 surface area, 12,13 gross domestic product per capita, 14 percent urbanization, 12,15 and adult smoking prevalence [16] [17] [18] [19] were obtained from publicly-available databases. Whether a nation was an isolated political entity on an island was also recorded. The prevalence of an infectious process subject to exponential growth (or decay) appears linear over time when graphed on a logarithmic scale. 1 Therefore, we postulated that the logarithm of the country-wide infection prevalence would be linearly related with the duration of the infection in each country. In addition, our analysis postulated that deaths from coronavirus would follow infections with some delay. On average, the time from infection with the coronavirus to onset of symptoms is 5.1 days, 20 and the time from symptom onset to death is on average 17.8 days. 21 Therefore, the time from infection to death is expected to be 23 days. 1, 22 These incubation and mortality times were prespecified. 1, 22 Therefore, the date of each country's initial infection was estimated as the earlier of: 5 days before the first reported infection, or 23 days before the first death. 8, 9, 23 Deaths by May 9, 2020 would typically reflect infections beginning 23 days previously (by April 16) . Therefore, we recorded the time from the first infection in a country until April 16. We also recorded the period of the outbreak: 1) from the mandating of activity restrictions until April 16, and 2) from when public mask-wearing was recommended until April 16. In univariate analysis, characteristics of countries with above-median per-capita mortality were compared with the remaining (lower mortality) countries by the two-group t-test. The odds ratio for being in the high-mortality group was calculated by logistic regression. Significant predictors of per-capita coronavirus mortality in the univariate analysis were analyzed by stepwise backwards multivariable linear regression analysis. The dependent variable was the logarithm (base 10) of per-capita coronavirus-related mortality. Because of the importance relative to public health, the weeks the country spent in lockdown and using masks, and per-capita testing levels, were all retained in the model. Statistical analysis was performed with xlstat 2020.1 (Addinsoft, New York). An alpha (p value) of 0.05 was deemed to be statistically significant. The study was approved by the Virginia Commonwealth University Office of Research Subjects Protection. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 25, 2020. . https://doi.org/10.1101/2020.05.22.20109231 doi: medRxiv preprint A total of 183 countries reported coronavirus mortality and testing data by May 9, 2020. The 91 lower-mortality countries had 1.1 deaths per million population, in contrast with an average of 101.0 deaths per million population in the 92 highermortality countries (p<0.001, Table 1 ). The median value was 4.7 deaths per million population. We assumed that island nations might find it less challenging to isolate and protect their populations. However, 21 of 91 low-mortality countries were isolated on islands, compared with 25 of 92 high-mortality countries. Country surface area and population were not associated with coronavirus mortality (Table 1) . . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted May 25, 2020. . https://doi.org/10.1101/2020.05.22.20109231 doi: medRxiv preprint Countries with older populations suffered higher coronavirus mortality. Countries with low mortality had on average 9.3% of their population over age 60, as compared with 18.9% in the high-mortality countries (Table 1) . Sex ratio was not associated with country-wide mortality (p=0.45, Table 1 ). Smoking prevalence was on average 14.0% in low mortality countries and 18.7% in high-mortality countries (p<0.001, Table 1 ). Colder countries were associated with higher coronavirus mortality in univariate analysis. The mean temperature was 21.7 C (SD 7.5 C) in the low-mortality countries, and 13.4 C (SD 9.1 C) in the high-mortality countries (p<0.001, Table 1 ). Urbanization was associated with coronavirus mortality. In low-mortality countries, on average 55% of the population was urban, as compared with 70% of the population in the high-mortality countries (p<0.001, Table 1 ). Richer countries suffered a higher coronavirus related mortality. The mean GDP per capita was $10,570 in the low-mortality countries, and was $27,600 in the high-mortality countries (Table 1 , p<0.001). In some Asian countries, masks were used extensively by the public, essentially from the beginning of the outbreak, 1,24 including: Thailand, 25 Japan, 24, 26 South Korea, 24, 27 Taiwan, 28 Hong Kong, 24, 29, 30 Vietnam, 31 Malaysia, 32 Cambodia, 33, 34 and the Philippines 35 (Table 2 ). In fact, in Mongolia 36 and Laos, 37 the public began wearing masks before any cases were confirmed in their countries. Despite the fact that the outbreak tended to appear in these 11 countries quite early, the countries had experienced a low per-capita coronavirus mortality by May 9 (mean 1.9 per million, SD 2.5, Table 2 ). The World Health Organization initially advised against widespread mask wearing by the public, as did the United States CDC. 1, 24 Nonetheless, some Western governments mandated or recommended wearing of masks by the general public during March 2020. Masks were required in public in Venezuela beginning March 13. 38, 39 In Slovakia, masks were mandated in shops and transit on March 15, 40 and more broadly in public on March 25. 41 Masks were required in public in Czechia on March 19, 42 and in Uzbekistan on March 25. 43 Masks were mandated in indoor public spaces in Slovenia on March 29. 44 In Austria, a mandate to wear masks in shops was announced on March 30, with the expectation that masks would be available by April 1. 45 In . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted May 25, 2020. 24 and was manifest by Jan. 31. 26 The government initially recommended them when in "confined, badly ventilated spaces". 24 South Korea Jan. 20 5.0 Use of masks is traditional. 24 South Korea initially had trouble obtaining enough masks, but at the end of February the government began to control the distribution of masks to the public. 27 Jan. 21 0.3 Use of masks is traditional. By January 27, the government had to limit mask exports and limit sales from pharmacies to those needed for personal use. 28 Jan. 23 29 0.5 Surgical masks were traditionally used, and also were recommended on public transport and in crowded places, on January 24, 2020. 24 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted May 25, 2020. By multivariable linear regression, significant predictors of the logarithm of each country's per-capita coronavirus mortality included: duration of infection in the country, duration of wearing masks, percentage of the population over age 60 , and urbanization (all p<0.001, Table 3 ). With the duration of infection in the country controlled for, there was a trend for time in lockdown to be negatively associated with mortality, and for percapita testing to be positively associated with mortality, though neither association was significant (Table 3 ). In a population not wearing masks, the per-capita mortality tended to increase each week by a factor of 10 0.171 = 1.48, or 48%. On the other hand, in a population wearing masks, the per-capita mortality tended to increase each week by a factor of 10 (0.171-0.148) = 1.054, or just 5.4%. Compared with the baseline condition (without masks), under lockdown, the per-capita mortality increased each week by 10 (0.171-0.039) = 1.36, or 36% (Table 3) . . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted May 25, 2020. . https://doi.org/10.1101/2020.05.22.20109231 doi: medRxiv preprint A country with 10% more of its population living in an urban environment than another country will tend to suffer a mortality 21% higher (10 0.0838 = 1.21). A country in which the percentage of the population age 60 or over is 10% higher than in another country will tend to suffer mortality 231% higher (10 0.52 = 3.31, Table 3 ). The trend towards a positive association with testing by May 9 probably reflects the greater recognition of coronavirus-related mortality with more testing, as well as the increased incentive countries have to test when they suffer a more intense outbreak. As noted above, testing was positively associated with mortality. By May 9, 2020, low-mortality countries had performed 1 test for every 510 members of the population, while high-mortality countries had performed 1 test for every 82.6 members of the population (p<0.001, Table 1 ). Presumably, testing performed early on would better help to contain the outbreak. By April 4 (the earliest testing data archived), lowmortality countries had tested 1 in 63,300 in the population, as compared with 1 in 1,710 in the high-mortality countries (p<0.001, Table 1 ). By April 16, 2020, low-mortality countries had tested 1 in 2,340 in the population, as compared with 1 in 251 in the highmortality countries (p<0.001, Table 1 ). If early testing lowers mortality, one might expect negative regression coefficients (more testing associated with less mortality). However, when (log) per-capita testing on April 4, 2020 was included in the model in Table 3 (which also includes later testing), early testing was not inversely related to mortality (coefficient = 0.01, p=0.83). Likewise, when (log) per-capita testing on April 16, 2020 was included in the model in Table 3 , early testing still tended not to be inversely related to coronavirus mortality (coefficient = 0.21, p=0.11). Only 5 countries had performed over 1 test for every 10 people in the country by May 9, 2020 (in order of most testing to least): the Faeroe Islands, Iceland, the Falkland Islands, the UAE, and Bahrain. The Faeroe and Falkland Islands reported no coronavirus-related deaths. The remaining 3 countries had per-capita mortality above the median value. The highest per-capita mortality among this group was 29.0 per million population (or 1 in 34,480 people), seen in Iceland. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted May 25, 2020. . These results confirm that over 4 months since the appearance of COVID-19 in late 2019, there remains a great deal of variation between countries in related mortality. Countries in the lower half of mortality have experienced an average COVID-19-related per-capita mortality of 1.1 deaths per million population, in contrast with an average of 101 deaths per million in the remaining countries. Independent predictors of mortality included urbanization, fraction of the population age 60 years or over, duration of the outbreak in the country, and the period of outbreak subject to cultural norms or government policies favoring mask-wearing by the public. These results support the universal wearing of masks by the public to suppress the spread of the coronavirus. 1 Given the low levels of coronavirus mortality seen in the Asian countries which adopted widespread public mask usage early in the outbreak, it seems highly unlikely that masks are harmful. One major limitation is that evidence concerning the actual levels of mask-wearing by the public are not available for most countries. Particularly in Western countries which only recommended (rather than mandated) mask-wearing by the public, such as the United States, the practice has been steadily increasing, but change has not occurred overnight. Our analysis based on norms and policies can be compared with "intention-to-treat" analysis in a clinical trial. Much of the randomized controlled data on the effect of mask-wearing on the spread of respiratory viruses relates to influenza. One recent meta-analysis of 10 trials in families, students, or religious pilgrims found that the relative risk for influenza with the use of face masks was 0.78, a 22% reduction, though the findings were not statistically significant. 48 Combining all the trials, there were 29 cases in groups assigned to wear masks, compared with 51 cases in control groups. 48 The direct applicability of these results to mask-wearing at the population level is uncertain. For instance, there was some heterogeneity in methods of the component trials, with one trial assigning mask wearing to the person with a respiratory illness, another to his close contacts, and the remainder to both the ill and their contacts. 48 Mask-wearing was inconsistent. The groups living together could not wear a mask when bathing, sleeping, eating, or brushing teeth. [49] [50] [51] In one of the studies reviewed, parents wore a mask during the day, but not at night when sleeping next to their sick child. 51 In a different trial, students were asked to wear a mask in their residence hall for at least 6 hours daily (rather than all the time). 49 The bottom line is that it is nearly impossible for people to constantly maintain mask wear around the people with whom they live. In contrast, wearing a mask when on public transit or shopping is quite feasible. In addition, as an infection propagates through multiple generations in the population, the benefits multiply exponentially. Even if one accepts that masks would only reduce transmissions by 22%, then after 10 cycles of the infection, mask-wearing would reduce the level of infection in the population by 91.7%, as compared with a non-mask wearing population (because 0.78 10 = 0.083). It is highly unlikely that anyone will ever randomize entire . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted May 25, 2020. . countries or populations to either wear, or not wear, masks. Public policies can only be formulated based on the best evidence available. A recent study demonstrated that in 4 patients infected with SARS-CoV-2 who were asked to cough without and with a cloth mask, the viral load on a nearby petri dish was reduced with the cloth mask in all patients, and in half was not detectable with the cloth mask. 52 Nationwide policies to ban large gatherings and to close schools or businesses, tended to be associated with lower mortality, though not in a statistically significant fashion. However, a full accounting would consider local policies, which often varied within each country. Moreover, businesses, schools, and individuals made decisions to limit contact, independent of any government policies. The adoption of numerous public health policies at the same time can make it difficult to tease out the relative importance of each. Colder average monthly temperature was associated with higher levels of COVID-19 mortality in univariate analysis, but not when accounting for other independent variables. Nonetheless, environmental factors which could influence either human behavior or the stability and spread of virus particles are worthy of further study. Per-capita testing both early (April 4, 16) and later (May 9) were positively associated with reported coronavirus-related mortality. It seems likely that countries which test at a low level are missing many cases. We previously identified just 3 countries (Iceland, the Faeroe Islands, and the UAE) which had performed over 75,000 tests per million population by April 16, and all 3 had mortality below 1 in 46,000 at that point. 53 By May 9, we could add to this "high-testing" group, the Falkland Islands and Bahrain, as all 5 countries had tested over one tenth of their population. All 5 countries had a mortality of 29 per million (1 in 34,480 people) or less. The degree to which these results would apply to larger, less isolated, or less wealthy countries is unknown. Statistical support for benefit of high levels of testing might be demonstrated if additional and more diverse countries are able to test at this level. One limitation of our study is that the ultimate source of mortality data is often from governments which may not have the resources to provide a full accounting of their public health crises, or an interest in doing so. In summary, older age of the population, urbanization, and longer duration of the outbreak in a country were independently associated with higher country-wide percapita coronavirus mortality. Lockdowns and the amount of viral testing were not statistically significant predictors of country-wide coronavirus mortality, after controlling for other variables. The use of masks in public is an important and readily modifiable public health measure. Societal norms and government policies supporting maskwearing by the public were independently associated with lower per-capita mortality from COVID-19. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 25, 2020. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 25, 2020. . https://doi.org/10.1101/2020.05.22.20109231 doi: medRxiv preprint Final Country-wide Mortality from the Novel Coronavirus (COVID-19) Pandemic and Notes Regarding Mask Usage by the Public Age-dependence of mortality from novel coronavirus disease (COVID-19) in highly exposed populations: New York transit workers and residents and Diamond Princess passengers Covid-19: risk factors for severe disease and death Association between ambient temperature and COVID-19 infection in 122 cities from China. Science of The Total Environment Evaluating the determinants of COVID-19 mortality: A cross-country study. medRxiv Journey through an epidemic: some observations of contrasting public health responses to SARS COVID-19 Coronavirus Pandemic COVID-19 Coronavirus Pandemic European Centre for Disease Prevention and Control. Download today's data on the geographic distribution of COVID-19 cases worldwide List of Cities by Average Temperature World Population Prospects: The 2019 Revision. Population, surface area, and density Report for Selected Countries and Subjects Urban population (% of total population) The American Cancer Society. The Tobacco Atlas. 2020 Available from Surveys of adult tobacco use in WHO Member States Smoking prevalence, males (% of adults) Smoking prevalence, females (% of adults) The incubation period of coronavirus disease 2019 (COVID-19) from publicly reported confirmed cases: estimation and application Estimates of the severity of COVID-19 disease Prevention of the spread of coronavirus using masks European Centre for Disease Prevention and Control. COVID-19 Coronavirus data Rational use of face masks in the COVID-19 pandemic Ministry of Public Health Enhances Disease Prevention and Control. The Bangkok Insight Amid virus outbreak, Japan stores scramble to meet demand for face masks How South Korea Solved Its Face Mask Shortage: Neighborhood pharmacists and government intervention were the secret weapons Taiwan ups Chinese visitor curbs, to stop mask exports. Reuters China coronavirus: death toll almost doubles in one day as Hong Kong reports its first two cases. South China Morning Post I've been traveling in Asia for 3 weeks amid the deadly coronavirus outbreak, and actually catching the virus is far from my biggest fear. Business Insider Demand for face masks, hand sanitisers soars. New Straits Times Cambodia Confirms First Coronavirus Case. VOA Khmer Cambodian Businesses Embrace Protective Measures Against Coronavirus. VOA Khmer Why Face Masks Are Going Viral. Sapiens Laos reports no case of COVID-19, to import face masks from Vietnam. Vietnam Times Coronavirus: Maduro tells Venezuelans to make their own masks Venezuela confirms coronavirus cases amid public health concerns. Reuters Slovensko pritvrdilo v boji s vírusom. Nosenie rúšok je povinné". Pravda Opatrenie Úradu verejného zdravotníctva Slovenskej republiky pri ohrození verejného zdravia Could Czech's Measure to Fight Coronavirus Save Thousands of Lives? Prague Morning COVID-19 Information. US Embassy in Uzbekistan STA. COVID-19 & Slovenia, Night 29 March: Movement Restrictions, Mandatory Masks, More Aid for Individuals. Total Slovenia News Austria widening face-mask requirement while loosening lockdown Preliminary findings of a randomized trial of non-pharmaceutical interventions to prevent influenza transmission in households Face mask use and control of respiratory virus transmission in households Effectiveness of surgical and cotton masks in blocking SARS-CoV-2: a controlled comparison in 4 patients Suppression of the Covid-19 Outbreak by Mass Testing and Tracing, and Other Measures: Real-World Data