Introduction

Although the COVID-19 pandemic originated in China and quickly spread throughout Asia, it has had a far greater impact in the western world1. Within a few months of the outbreak, the epicentre of the pandemic shifted from Asia to Europe and then to the Americas. In addition to the grim impact on both physical and mental health in Europe and the Americas, lockdowns in the West caused wide-spread economic damage, the impact of which is likely to persist for years2,3.

During 2020 the COVID-19 mortality rates reported for North America and Europe were 114 times higher than in Western Pacific and South East Asian countries.11 The success of the East suggests the catastrophe of what ensued in the West could have been avoided. The ability of the East to prevent and manage SARS-CoV-2 transmission begs the question of how this was possible when advanced economies in the West, with established democracies and well-equipped health infrastructure, failed to contain the exponential growth.

This paper explores several potential answers to these questions by reporting on an ecological analysis that compares COVID-19 mortality with baseline country characteristics and national interventions and by focusing on the differences between the East and the West.

Methods

We conducted an ecological analysis to describe and investigate the uneven impact of the COVID-19 pandemic in countries with at least 1 million population.

Data and data sources

The main variable of interest was the cumulative number of reported COVID-19 deaths per million at December 31, 2020, and was retrieved for each observation from Our World in Data repository. No human data was analysed.

A large spectrum of variables (supplementary methods and Table S1) was used to explore the possible drivers of reported COVID-19 deaths grouped into three domains:

  1. 1.

    Demographic and health determinants (domain 1);

  2. 2.

    Factors linked with country systemic preparedness and response (domain 2);

  3. 3.

    Structural, economic, and political elements (domain 3).

Grouping

Countries were arbitrarily grouped into two subsets and labelled as the West (i.e., European Union, USA, and Canada) and the East (i.e., South-East Asia and the Western Pacific Region of the World Health Organization). The full list of countries is displayed in Table S2. While heterogeneous and evolving, the East and the West constructs have been used for various socioeconomic, geopolitical and cultural contrasts which framed the inclusion/exclusion criteria before our analysis. The comparison of these regions is relevant for pandemics given both overall levels of public health infrastructure and spending, and the origin (and initial spread) of COVID-19.

To explore the different timing of the adoption of correct behaviours to prevent SARS-CoV-2 infection, we used “face masks” using the relative search volume (RSV) in Google (supplementary materials).

Statistical plan

Quantitative variables were reported as median and range. COVID-19 deaths per million were logarithmically transformed to normalize the distribution of data. Likewise, gross domestic product (GDP) per capita, population size, and number of air travel passengers were non-normally distributed and consequently logarithmically transformed. All continuous exploratory variables were standardized before their inclusion in the ordinary least squares (OLS) regression models.

The association of the log of COVID-19 deaths per million with likely explanatory variables was further investigated using the Minimum Covariance Determinant (MCD) estimator and regression models with robust standard errors. For multiple OLS regressions, three exploratory sub-models were developed and tested based on specific domains described above. Statistically significant covariates, selected using a stepwise backward selection process, were merged in a unified final exploratory model. The inclusion of interaction terms in the regression models were considered, testing for non-additive effects of different combinations of predictor variables on the dependent variable. Issues of heterogeneity of variance, intragroup correlation and sensitivity to outliers were considered. A model-based clustering and classification estimation was used for corroborating the arbitrary West and East grouping. Additional information is available in the supplementary materials.

Statistical analysis was performed using STATA version 16.1 (StataCorp, College Station, Texas, USA) or R version 4.0.3 was used for the cluster analysis using the package mclust4.

Results

Demographics and health risk factors

As displayed in Table 1, the global median GDP per capita was 5,152 (range, 369 to 80,504) US dollars in 2019, the estimated proportion of the population aged 65 or older was 6.7% (range, 1.3–28.4%) in 2020, and the median prevalence of obesity was 20.2% (range, 2.1 to 37.9%).

Table 1 Demographics, health risk factors, preparedness and response indicators, structural economical-political determinants.
Full size table

There was a positive and strong association between the log of GDP per capita and the log of COVID-19 mortality (MCD = 0.808, b = 0.512, p < 0.001). Likewise, age was associated with increased mortality both as median age (MCD = 0.803, b = 0.509, p < 0.001) and as proportion of population 65 + years of age (MCD = 0.744, b = 0.480, p < 0.001). While a higher proportion of the population living in urban settings was positively associated to COVID-19 mortality (MCD = 0.621, b = 0.484, p < 0.001), living in crowded households had an inverse association (MCD = -0.660, b = −0.491, p < 0.001).

Several health risks beyond age have been associated with poor COVID-19 outcomes. Prevalence of obesity, in particular, had a strong positive association with COVID-19 mortality (MCD = 0.674, b = 0.591, p < 0.001) followed by while the prevalence of raised blood pressure had a strong negative association with COVID-19 mortality (MCD = −0.486, b = −0.267, p < 0.001).

Health system preparedness and response indicators

Several preparedness indicators were significantly associated with COVID-19 mortality (Table 2). In particular, the service coverage index of universal health coverage (UHC) was positively associated with COVID-19 mortality (MCD = 0.718, b = 0.537, p = 0.002), as was the density of medical doctors (MCD = 0.584, b = 0.493, p < 0.001). Notably, scores used specifically to evaluate countries’ preparedness, like the average score from the country self-reporting preparedness within the 2005 International Health Regulations (IHR) (MCD = 0.484, b = 0.379, p < 0.001) and the global health security (GHS) index (MCD = 0.606, b = 0.422, p < 0.001), were also associated with increased COVID-19 mortality.

Table 2 Final multivariate model for global COVID-19 mortality adjusted for response measures.
Full size table

Structural socio-political determinants

The proportion of literate population was strongly associated with COVID-19 mortality (MCD = 0.462, b = 0.471, p < 0.001), as were the level of inefficient government bureaucracy (MCD = 0.481, b = 0.406, p < 0.001) and government effectiveness index (MCD = 0.403, b = 0.382, p < 0.001) (Table 2). Negative associations with COVID-19 mortality were notable for the proportion of GDP used on essential services (MCD = −0.227, b = −0.246, p = 0.002) and the Gini coefficient (r = −0.081, b = −0.172, p = 0.018).

Exploratory multivariable models

We developed three domain-specific OLS regression sub-models (Table S4). Only five variables were found to be associated with COVID-19 mortality in the final multivariate model (Table S5 and Fig. 1) and the significance of the outputs remained unchanged once adjusted for response measures (Table S6). The final adjusted OLS model, further refined by removing outliers as described in the supplementary results and in Figure S6, is reported in Table 2. In this, increased COVID-19 deaths per million are associated with higher median age (b = 0.370, p < 0.001), higher prevalence of obesity in adults (b = 0.358, p < 0.001), and higher Economist’s Intelligence Unit (EIU) democracy index (b = 0.353, p < 0.001). On the other hand, previous local SARS outbreaks (b = −0.311, p = 0.001), and the index for political stability (b = −0.297, p < 0.001) emerged as protective factors against COVID-19 mortality. The significance of the model was further tested for the geographical autocorrelation with only previous SARS cases losing its significance (Fig. 1).

Figure 1
figure 1

Forest plot of the final multivariate OLS regression model of COVID-19 mortality in 2020 adjusted for response measures, with spatial autocorrelation correction. In the figure, three countries (ie, Mongolia, Thailand, Papua New Guinea) were removed because outliers. Additional information can be found in the supplementary results.

Full size image

Comparing the East versus the West

On average, the log of COVID-19 mortality in the West compared to the East increase of 2 units using a univariate OLS model with robust standard errors (b = 2.012, robust 95%CI: 1.535–2.488, p < 0.001). The median COVID-19 mortality reported in the West was 68.4 per million, much higher than the median in the East 6.0. Among Western countries, Belgium had by far the highest mortality (1710.1 per million population) and Norway the lowest (81.5 per million). In the East the highest rates were registered in Indonesia (85.3 per million), the Philippines (82.4 per million), and Myanmar (49.1 per million population). Figure 2 displays the trends of COVID-19 mortality in the two regions throughout the 2020.

Figure 2
figure 2

Trends in COVID-19 mortality (per 1 million people) in the West and in the East. Displayed mortality rates are unstandardized as data disaggregated by age groups are not available in any public repository. All countries displayed in this figure are those listed in Table S2 of the supplementary methods, including those < 1 million population.

Full size image

The West had a median GDP per capita 8-times higher than the East, and their populations were older (median age 43.2 vs. 31.4 years, respectively)(Table 3). Western countries were more urbanized, while the East had a higher proportion of households with 4 or more members (median 58.6 vs. 22.2%). The prevalence of obesity was much higher in the West (median 23.1%) than in the East (6.2%), and only slightly more so for smoking.

Table 3 Demographics, health risk factors, preparedness and response indicators, structural economical-political determinants within the two regions of interest.
Full size table

The West had significantly better indicators of pandemic preparedness (ie, IHR and GHS indices), health system performance (ie, density of medical doctors, hospital beds), and health spending. Two thirds of the countries in the East experienced SARS in 2002–2003 compared to 37% in the West. Western countries had better indicators of governance, literacy and income inequality, and were more likely to enjoy political stability free of violence compared to the East. Lastly, of the countries included in this analysis more countries in the East were islands than in the West (33.3% vs. 7.4%).

When analysing how the subset of countries clustered based on the Table 2 features, we corroborated the empirical East–West classification. However, countries of the Eastern Europe were observed to form a distinct cluster compared to countries of the Western Europe and North America. (details and plots in the supplementary results).

Timeliness and the maintenance of response measures to COVID-19

As shown in Fig. 2, the cumulative mortality from COVID-19 in the West skyrocketed nearly 8000-fold from March 1st to May 31st, but only 4.5 times in the East; from May 31st to December 31st the relative increases were 2.9 and 6.0, respectively.

The East and the West had a similar stringency index for their response to COVID-19, but the West was delayed and more reactive and it failed to halt an early and catastrophic exponential rise in cases and deaths (Table 4). The effective reproduction rate for COVID-19 was higher in the first quarter of 2020, and more so in the West vs. the East; the opposite was true for the stringency of the response. The positive rate of tests performed was much higher in the West.

Table 4 COVID-19 Epidemiology and response during 2020.
Full size table

In the East internet users increased searches for masks at the end of January, while the West such web-searches were delayed until March–April (Fig. 3). The temporal delay persists after adjusting for the date of each country’s first confirmed case.

Figure 3
figure 3

Weekly trend in online searches of “face masks” in Google, expressed in RSV (%), in the West (in red) and in the East (in purple). As extensively explained in the supplementary materials, data from 1st December 2019 to 30th June 2020, accounting for 1-month period of baseline to account for possible confounding for some countries in the East with higher baseline values (ie, Japan and Viet Nam). Additional information on Google Trend RSV can be found in the supplementary methods.

Full size image

Discussion

In 2020, there were 2.24 million confirmed COVID-19 deaths reported worldwide and this figure is estimated to surpass 10 million in reality5. Significant elements positively associated with increased COVID-19 mortality were median age, obesity prevalence, and EIU democracy index; political stability and previous SARS cases were associated with lower mortality. Despite having robust economies and health systems, Western countries experienced a COVID-19 mortality 114-times higher than countries in the East. Aging5 and obesity6, well established risk factors for COVID-19 severity, were more prevalent in the West and may have contributed to the higher mortality there. Having experienced SARS, on the other hand, seemed protective and may have helped the East mount a swifter response.

Disappointingly, the GHS index, the average score of IHR self-assessment, UHC, and the density of both hospital beds and medical doctors did not seem to protect against COVID-19 mortality. Structural factors played a less significant role at the global level, including the Gini coefficient which has been established as important at the subnational level7. The index used to measure political stability was associated with lower COVID-19 mortality, while, the EIU democracy index was associated with worse COVID-19 mortality; this suggests that authoritarian regimes with command-and-control leadership and more pliable societies may have been more effective in containing the epidemic8.

Delays in responding to the pandemic may have played a large role in the health and social crisis. The reproduction number of COVID-19 was much higher in the West in the first quarter of 2020 and cumulative mortality skyrocketed nearly 8,000-fold from March 1st to May 31st, but rose less than tenfold in the East and from May 31st to December 31st. During the first quarter of 2020, Asian countries tested 2.7 times more patients per reported COVID-19 case than in Western countries9. Testing is paramount not only for surveillance but for early contact-tracing and prompt isolation10—time-honoured public health measures that South Korea and other neighbouring countries implemented early and with rigor. Similarly, the use of face masks was more prevalent in Asia11 and interest in them surged several weeks before they did in the West. In addition to masks becoming politicized in the West12, their slow uptake in this region may have been due to the WHO’s slow recognition of the effectiveness of masks and therefore lack of leadership in this area13.

Interestingly, countries that reported local SARS outbreaks in 2002–2003 had a lower COVID-19 mortality in 2020. While the viruses causing SARS and COVID-19 are closely related, protective cross-immunity in 2020 is improbable because of the very limited number of SARS cases and relatively short duration of neutralizing antibodies14. More plausible, the “social memory” derived from experiencing a highly lethal outbreak and associated economic shock, especially in Asia, may have contributed to a faster and consistent response at individual and community levels. Conversely the West, with more robust economies and stronger health systems, may have been overconfident causing delays in response and losing control of the epidemic early on which ultimately required more drastic countermeasures15. The costs of lock-downs will persist over several years and will impact both the political economy as well as the physical and mental health of the population.

In the summer of 2020, there was concern that the lack of population-level immunity and exhaustion with strict behavioural measures would lead to a spike of COVID-19 deaths in the East by the end of the year like the one experienced by the West in the spring. Instead, the West missed a second chance at controlling the pandemic16 and experienced an even greater wave (in absolute numbers) by the end of 2020. This included Sweden, whose more liberal (and sustainable) policies permitted the un-interrupted transmission of the virus among young and healthy people17. A sustained response to COVID-19 was clearly more important than slowly growing levels of natural immunity18.

With much lower levels of natural immunity in 202119, Asian populations remain very vulnerable to COVID-19 as documented by the rapidly increasing burden in some countries in the first half of 2021. This is a clear signal that these countries should not lower their guard until enough people are vaccinated, starting with the elderly, health care workers and other vulnerable populations. The wave of cases in India in early 2021 is a case in point, as is the recent increase in mortality rates in Taiwan, Japan and Thailand.11 The prognosis in Africa, also reporting low COVID-19 mortality in 2020, is less clear given its younger population20.

This study has some limitations as data from different sources of data are far from perfect nor complete and do not disaggregate by key elements such as age or socioeconomic status. Further, numerous COVID-19 cases, as well as deaths, likely went undetected or unreported which may have been more prevalent in the East.20 Noteworthy, different types of definitions were used by countries to define a COVID-19 death, either using the current WHO definition (i.e., clinical diagnosis-based for confirmed and probable cases)21 or a test-based diagnosis22. To attenuate this challenge our analysis focused on mortality rather than incidence. We did not analyze the importance of culture, climate change or biological indicators (genetics, mutations), although we note that in the United States persons of Asian descent had the same COVID-19 mortality as non-Hispanic Whites23. Several paradoxical findings in univariate analyses, like higher mortality in countries with better health preparedness indicators, may have been confounded by higher case detection or, more likely, by delays activating their public health capability before contagion spun out of control. Furthermore, as described with the availability and reliability of data on COVID-19 epidemiology, other predictors used in the models might be affected by the same under-estimation (e.g. obesity prevalence and median age). These, however, are estimated from international organizations, while COVID-19 cases and deaths are those directly reported by countries and not based on estimates accounting for under-reporting.

Conclusion

The East was successful in stopping the exponential first phase of COVID-19 in 2020, while the West has suffered mortality rates 114 times higher. Older age and obesity emerged as significant ecological risk factors for COVID-19 and might have contributed to the vast differences between these two regions. Rapid response from governments, including early lock-downs in China, higher testing coverage per case in South Korea, and early mask wearing in most of Asia may have contributed to the East’s ability to contain the epidemic. Similarly, countries that experienced SARS in 2002–2003 had lower deaths from COVID-19 in 202024. Whereas delays in response and ineffective leadership and communication hindered western countries’ ability to control the virus, despite their stronger economies and more robust health systems 25,26, While the early success of the East was sustained throughout 2020, low natural immunity in the region requires continued public health vigilance until threshold levels of vaccination-acquired immunity are reached or the virus otherwise disappears from the face of the Earth.