key: cord-1042926-2v4lrlcl authors: Pana, T. A.; Bhattacharya, S.; Gamble, D. T.; Pasdar, Z.; Szlachetka, W. A.; Perdomo-Lampignano, J. A.; McLernon, D.; Myint, P. K. title: Number of International Arrivals Predicts Severity of the first Global Wave of the COVID-19 Pandemic date: 2020-05-16 journal: nan DOI: 10.1101/2020.05.13.20100677 sha: 2f9a9b479d33af87d6abf9e26de55fccec15c11c doc_id: 1042926 cord_uid: 2v4lrlcl Background: Reported death rates from different countries during the COVID-19 pandemic vary. Lack of universal testing and death underreporting make between-country comparisons difficult. The country-level determinants of COVID-19 mortality are unknown. Objective: Derive a measure of COVID-related death rates that is comparable across countries and identify its country-level predictors. Methods: An ecological study design of publicly available data was employed. Countries reporting >25 COVID-related deaths until May 1, 2020 were included. The outcome was the mean mortality rate from COVID-19, an estimate of the country-level daily increase in reported deaths during the ascending phase of the epidemic curve. Potential predictors assessed were most recently published Demographic parameters (population and population density, percentage population living in urban areas, median age, average body mass index, smoking prevalence), Economic parameters (Gross Domestic Product per capita; environmental parameters: pollution levels, mean temperature (January-April)), co-morbidities (prevalence of diabetes, hypertension and cancer), health systems parameters (WHO Health Index and hospital beds per 10,000 population and international arrivals). Multivariable linear regression was used to analyse the data. Results: Thirty-one countries were included. Of all country-level predictors included in the multivariable model, only total number of international arrivals was significantly associated with the mean death rate: Beta 0.3798 (95% Confidence Interval 0.2414, 0.5182), P <0.001. Conclusion: International travel was directly associated with the mortality slope and thus potentially the spread of COVID-19. Stopping international travel, particularly from affected areas, may be the most effective strategy to control COVID outbreak and prevent related deaths. The atypical pneumonia caused by novel corona virus (SARS-CoV 2) detected in Wuhan, Hubei province, China at the end of 2019 has subsequently spread across five continents at a remarkable speed, with Europe and North America being the most affected regions of the world. The World Health Organisation (WHO) declared COVID-19 to be a pandemic of proportions similar to the Spanish Influenza of 1918. As of the 1 st May 2020, there have been over 224,172 deaths related to COVID-19 infection worldwide. 1 Data collated from across the world suggest that the overall case fatality rate is around 7%, with country-level estimates ranging between 0.5-14%. 2 These figures however are not useful for universal comparison as testing rates also vary by country and there is a lag phase in reported deaths that occur in the community. Consequently, there is wide variation in the reported country-specific death rates which may be attributed to variation in testing rates, underreporting or real differences in environmental, sociodemographic and health system parameters. The only previous ecological study to date assessing country-level predictors of the severity of the COVID-19 pandemic including data on 65 countries 3 has found that the cumulative number of infected patients in each country was directly associated with the case fatality rate, whilst testing intensity was inversely associated with case fatality rate. This study found no association between health expenditure and case fatality rate. However, other important country-level predictors were not evaluated and thus their relationship with pandemic severity remains unknown. Several risk factors for COVID-related mortality have been proposed, including older population, 4 higher population co-morbid burden, 5 smoking, 6 obesity, 7 pollution levels 8 and healthcare system performance. 9 Furthermore, countries outside China most severely hit by the pandemic were those with a high income, high GDP per capita and well-established healthcare systems, such as Italy, Spain, France, the United Kingdom and the United States. 10 In contrast, lower-and middle-income countries reported much lower COVID-19 incidence and mortality rates. 10 Whilst these differences may be attributable to case underreporting due to inadequate testing facilities in poorer countries, other factors may also be involved. In this study, we aimed to derive a comparable measure of COVID related death rates. In addition, we aimed to assess the determinants for this measure by examining the association between potential country level determinants driven by hypothesis based on currently available evidence and this measure using country level publicly available data and an ecological study design. An ecological study design was used. The chosen outcome was the steepness of the ascending curve of country specific daily reports of COVID related deaths from January to 1 st May 2020. The following predictors were used: demographic predictors (population and population density, percentage population living in urban areas, median age, average body mass index (BMI), smoking prevalence), economic predictors (gross Domestic Product (GDP) per capita), environmental predictors (pollution levels, mean temperature (January-April) [2010] [2011] [2012] [2013] [2014] [2015] [2016] ), prevalent co-morbidities (diabetes, hypertension and cancer), health systems predictors (WHO Health Index and hospital beds per 10, 000 population) and international arrivals, as a proxy measure of the globalisation status of each country. Given the study design and the use of publicly available data, no ethical approval was necessary. Countries reporting at least 25 daily deaths up to the 1 st of May 2020 with available data for all chose predictors were included. A total of 31 countries were included in the analysis: Algeria, Austria, Belgium, Brazil, Canada, the Dominican Republic, Ecuador, Egypt, Finland, France, Germany, Hungary, India, Indonesia, Ireland, Italy, Japan, Mexico, the Netherlands, Peru, the Philippines, Poland, Portugal, Romania, the Russian Federation, Spain, Sweden, Switzerland, Turkey, the United Kingdom and the United States. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The data regarding the median population age and population density were 18 The world health organisation health index was extracted from the WHO Global Partnership for Education (GPE) paper series published in 2000. 19 Country-level total hospital beds per 10,000 population data were extracted from the World Bank Dataset "World Bank Indicators of Interest to the COVID-19 Outbreak". 20 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) 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 16, 2020. . https://doi.org/10.1101/2020.05.13.20100677 doi: medRxiv preprint Whilst previous ecological studies of other epidemics have utilised case or death counts as outcome, 21 these variables may be prone to bias due to variations in country level control measures including different testing strategies, 22 variations in population movement controls and differences in secondary attack rates within community cohorts 23 . The mean mortality rate was thus chosen as outcome instead, since it is independent of these highly variable parameters and may thus represent a more reliable indicator of the country-level severity of the COVID-19 pandemic Mean mortality rate was defined as the slope of the mean mortality curve (Figure 1 ), measured from the first day when more than 2 COVID-19 deaths were reported until either the mortality curve reached a peak value or the 1 st of May 2020, whichever occurred first. Before slope calculation, the mortality curve in each country was smoothed using a locally weighted (Lowess) regression using a bandwidth of 0.4. In order to ensure a good fit of the Lowess regression line, only countries having reported at least 25 daily deaths until the 1 st of May 2020 were included. The mean mortality rate thus represents an estimate of the country-level daily increase in reported deaths during the ascending phase of the epidemic curve. Data on population density were extracted as the country-level population per square kilometre in 2019. 24 Data on ambient air pollution were extracted as the countrylevel mean concentration of fine particulate matter (PM2.5) measured in 2016. 25 Temperature data were extracted as the mean temperature recorded in each country between January and April between 2010 and 2016. 15 Data on International Arrivals were extracted as the total number of country-level international arrivals in 2018. 26 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) 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 16, 2020. . https://doi.org/10.1101/2020.05.13.20100677 doi: medRxiv preprint Data on prevalent diabetes were extracted as the percentage of the population aged 20 to 79 years in 2019. 16 Data on prevalent cancers were extracted as the age-standardized cancer prevalence among both sexes in 2017, expressed as percentages. 27 Data on prevalent hypertension were extracted as the age-standardised percentage of the population over 18 years of age with systolic blood pressure ≥140 mmHg or diastolic blood pressure ≥90 mmHg in 2015. 28 Data on BMI were extracted as the age-standardised mean body mass index trend estimates for both sexes amongst adults (≥18 years) in 2016. 29 Data on daily cigarette smoking were extracted as the age-standardised rate on both sexes amongst adults (≥18 years) in 2013. 30 Whilst the definition of "daily cigarette smoking" varies across surveys, it habitually refers to current smoking of cigarettes at least once a day. 30 Data on GDP were extracted as GDP per capita by Purchasing Power Parity (PPP) in current international dollars in 2018. 31 The percentage of population living in urban areas was defined as the percentage of de facto population living in areas classified as urban according to the criteria used by each area or country. 14 The World Health Organisation (WHO) heath index is a composite index that aims to evaluate a given countries healthcare system performance relative to the maximum it could achieve given its level of resources and non-healthcare system determinants. It was calculated in the year 2000. The index uses five weighted parameters: overall or average disability-adjusted life expectancy (25%), distribution or equality of disability-adjusted life expectancy (25%), overall or average healthcare system responsiveness (including speed of provision and quality of amenities; 12.5%), distribution or equality of healthcare system responsiveness (12.5%) and healthcare expenditure (25%). Data on hospital beds per 10,000 population were defined by the World Bank as including "inpatient beds available in public, private, general, and specialized All rights reserved. No reuse allowed without permission. (which was not certified by peer review) 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 16, 2020. . https://doi.org/10.1101/2020.05.13.20100677 doi: medRxiv preprint hospitals and rehabilitation centers. The published data for countries included was from 2000 to 2017. In most cases beds for both acute and chronic care are included. 20 All analyses were performed in Stata 15.1SE, Stata Statistical Software. A 5% threshold of statistical significance was utilised for all analyses (P <0.05). Linear regression was performed to assess the univariable relationship between each country-level predictor and the calculated mean mortality rate for each country. The following predictors were included in the univariable analyses: population in 2018, median age, pollution levels, mean temperature (January-April), international arrivals, population density, prevalent diabetes, prevalent neoplasms, median BMI, prevalent hypertension, smoking prevalence, hospital beds (per 10,000 population), WHO health index, percentage population living in urban areas and GDP per capita (PPP). Predictors reaching a P-value <0.3 at univariable level were then included in a multivariable logistic regression model to predict the mean mortality rate outcome: median age, pollution levels, international arrivals, prevalent neoplasms, median BMI, prevalent hypertension, WHO health index, percentage of population living in urban areas and GDP per capita. (which was not certified by peer review) 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 16, 2020. . Table 2 In this ecological study including data from 31 countries which were most severely affected by COVID-19 in the first wave of current Global pandemic, we assessed 14 countrylevel socioeconomic, environmental, health and healthcare system, and globalisation parameters as potential predictors of variation in death rates from COVID 19 infection. In the multivariable linear regression model, the only predictor that reached statistical significance was international arrivals, a proxy of global connection. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) 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 16, 2020. . https://doi.org/10.1101/2020.05.13.20100677 doi: medRxiv preprint A recently published ecological study analysed the country-level predictors of the case fatality rate of the COVID-19 pandemic using data from 65 countries. 3 This study found that upon adjustment for epidemic age, health expenditure and world region, the case fatality rate was significantly associated with increasing cumulative number of COVID-19 cases and decreasing testing intensity. 3 Nevertheless, no other country-level predictors were included in this study. Comorbidities may account for differences in mortality rates across countries. A study among laboratory-confirmed cases of COVID-19 in China showed that patients with any comorbidity, including diabetes, malignancy and hypertension, had poorer clinical outcomes than those without. 5 We thus accounted for country-level data on a selection of key comorbidities in our analysis which included prevalent diabetes mellitus, neoplasms, and hypertension. Diabetes mellitus is significantly associated with all-cause and cardiovascular disease mortality globally. BMI ≥40kg/m2 has been identified as an independent risk factor for severe COVID-19 illness. 7 Finally, a recent systematic review on 5 studies from China showed that smoking is likely associated with negative outcomes and progression of COVID-19. 6 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) 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 16, 2020. Interestingly, during the COVID-19 pandemic, some countries (such as Thailand) have adopted aggressive international travel screening and isolation policies, which may have led to lower infection rates. 38 Our study suggests that travel restrictions have the potential to influence the impact of the COVID-19 pandemic and should be part of a structured and rapidly instigated pandemic preparedness plan. Any policy on the restriction of international All rights reserved. No reuse allowed without permission. (which was not certified by peer review) 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 16, 2020. . https://doi.org/10.1101/2020.05.13.20100677 doi: medRxiv preprint travel should be developed taking into account the economic and social impacts of such restrictions. The main strength of this study lies in its use of comparable and relevant outcome data derived from contemporary death reporting from countries affected by COVID-19. As testing rates for the virus vary across countries, the incidence or prevalence of the disease cannot be compared between countries. While death from the disease is a hard outcome, the denominator information to calculate death rates make between-country comparisons difficult. In addition, the deaths in the community, particularly in the elderly living in care homes, often go untested and thus firm diagnosis remains impossible. Therefore, in this study we have adopted an outcome that is comparable in terms of the increase in the rate of death, rather than death rates per se. Therefore, this may better represent the spread and seriousness of pandemic in individual countries when comparing countries at different stages of the pandemic. The country-level parameters assessed as potential predictors have all been implicated at some point to be associated with severity and consequently mortality. We however found that the only significant predictor to be total number of international arrivals in the country (2018 figures), signifying transmission of the infection through travel. Although the data was from 2018, there is no reason to believe that international travel figures between countries would be different in early 2020. Our model had a reasonably good fit to the data, explaining around 77% of the between country variation in mean death rates. The main limitation of the study stems from the ecological study design. Despite the fact that we did not find any association between comorbidities such as diabetes, cancer All rights reserved. No reuse allowed without permission. (which was not certified by peer review) 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 16, 2020. Out of all the country-level parameters assessed, international travel was the only significant predictor of the severity of the first global wave of the COVID-19 pandemic. Given that many of world middle and lower-income countries are showing signs of continued rise in infection rates, international travel restrictions applied early in the pandemic course may be an effective measure to avoid rapidly increasing infection and death rates globally. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) 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 16, 2020. Table 1 . Observed mean mortality rate and number of international arrivals in 2018 (millions) for each country included in the analyses. Countries were categorised in 3 groups: high mean mortality rate group (>20 additional daily deaths), medium mean mortality rate group (2-20 additional daily deaths) and low mean mortality rate group (<2 additional daily deaths). (which was not certified by peer review) 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 16, 2020. . Japan 0.01 31.19 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) 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 16, 2020. . BMI -body mass index; WHO -world health organisation; GDP -gross domestic product; PPP -purchasing power parity; Figure 1 . Graphical representation of the smoothed* number of daily deaths of each country (before reaching mortality peak, if applicable) as a function of the number of days passed since the first day when an excess of 3 deaths were reported. Countries with higher mortality rates are depicted in blue, while those with lower mortality rates are depicted in red. *smoothed using a local regression (lowess) function with a bandwidth of 0.4 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) 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 16, 2020. . https://doi.org/10.1101/2020.05.13.20100677 doi: medRxiv preprint Figure 2 . Predicted (based on the results of the multivariable linear regression) and observed country-level mortality rate (mean daily increase in deaths until the peak in mortality) as a function of the recorded country-level number of international arrivals in 2018 (millions). World Health Organization. Coronavirus disease (COVID-19) Situation Report-102 Mortality Risk of COVID-19 Flattening-the-curve associated with reduced COVID-19 case fatality ratesan ecological analysis of 65 countries Demographic science aids in understanding the spread and fatality rates of COVID-19 Comorbidity and its impact on 1590 patients with Covid-19 in China: A Nationwide Analysis COVID-19 and smoking: A systematic review of the evidence COVID-19 and obesity-lack of clarity, guidance, and implications for care. The Lancet Assessing nitrogen dioxide (NO2) levels as a contributing factor to coronavirus (COVID-19) fatality Potential association between COVID-19 mortality and health-care resource availability World Health Organization. Coronavirus disease (COVID-19) Situation Report-108 European Centre for Disease Prevention and Control. European Centre for Disease Prevention and Control United Nations. Department of Economic and Social Affairs Population Dynamics United Nations Statistics Division. Population density and urbanization The World Bank Group. Data Bank, World Development Indicators Measuring progress towards the Sustainable Development Goals World Health Organization. Global Health Observatory indicator views Measuring overall health system performance for 191 countries World Bank Indicators of Interest to the COVID-19 An ecological study of the determinants of differences in 2009 pandemic influenza mortality rates between countries in Europe United Nations Department of Economics and Social Affairs Annual mean concentration of particulate matter of less than 2.5 microns of diameter The World Bank. International tourism, number of arrivals World Health Organization. Prevalence of raised blood pressure (SBP≥140 OR DBP≥90 World Health Organization. Mean BMI (kg/m²) (age-standardized estimate World Health Organization. Daily smoking of cigarettes (age-standardized rate) We would like to thank Dr Kathryn Martin who provided valuable advice in study design. None. None.