key: cord-0890908-6ed3two6
authors: Kuster, A. C.; Overgaard, H. J.
title: A novel comprehensive metric to assess COVID-19 testing outcomes: Effects of geography, government, and policy response
date: 2020-06-19
journal: nan
DOI: 10.1101/2020.06.17.20133389
sha: 7cfdc7148324e95da604b2868c8140fd6ff90fc6
doc_id: 890908
cord_uid: 6ed3two6

Testing and case identification are key strategies in controlling the COVID-19 pandemic. Contact tracing and isolation are only possible if cases have been identified. The effectiveness of testing must be tracked, but a single comprehensive metric is not available to assess testing effectiveness, and no timely estimates of case detection rate are available globally, making inter-country comparisons difficult. The purpose of this paper was to propose a single, comprehensive metric, called the COVID-19 Testing Index (CovTI) scaled from 0 to 100, that incorporated several testing metrics. The index was based on case-fatality rate, test positivity rate, active cases, and an estimate of the detection rate. It used parsimonious modeling to estimate the true total number of COVID-19 cases based on deaths, testing, health system capacity, and government transparency. Publicly reported data from 188 countries and territories were included in the index. Estimates of detection rates aligned with previous estimates in literature (R2=0.97). As of June 3, 2020, the states with the highest CovTI included Iceland, Australia, New Zealand, Hong Kong, and Thailand, and some island nations. Globally, CovTI increased from April 20 ([x]=43.2) to June 3 ([x]=52.2) but declined in ca. 10% of countries. Bivariate analyses showed the average in countries with open public testing policies (59.7, 95% CI 55.6-63.8) were significantly higher than countries with no testing policy (30.2, 95% CI 18.1-42.3) (p<0.0001). A multiple linear regression model assessed the association of independent grouping variables with CovTI. Open public testing and extensive contact tracing were shown to significantly increase CovTI, after adjusting for extrinsic factors, including geographic isolation and centralized forms of government. This tool may be useful for policymakers to assess testing effectiveness, inform decisions, and identify model countries. It may also serve as a tool for researchers in analyses by combining it with other databases.

Countries and territories were included in the index if at least one case was reported, and the 1 population was greater than or equal to 100,000. Data were accessed daily; however, this report 2 presents the results for data accessed as of 00:00 GMT on June 3, 2020 (n=188). 3 4 Definition of key indicators 5 Several key indicators were computed from the input data, representing important 6 epidemiological indicators used in the analysis. 7 8 Case Fatality Rate (CFR) 9 The CFR is the proportion of total deaths, D, among closed cases (sum of D and R). 10 However, some countries, including the Netherlands and United Kingdom, have not reported the 11 number of recoveries, and others have not tracked recoveries in real-time. Additionally, the closed-12 case definition of CFR can overestimate the CFR in the early stages of an epidemic because of the 13 relatively small number of closed cases [29] . Therefore, an alternative estimate of CFR was 14 calculated as the ratio of deaths, D, to cases, C. Logically, the ratio of D:C is lower than D:(D+R) 15 because C includes unresolved cases with unknown outcomes. Linear regression of these two ratios 16 using data from the Worldometer [26] showed the relationship to be: D:(D+R) = 1.99*(D:C) 17 (R 2 =0.58, n=178). Thus, the CFR used in our further analysis was the minimum of either the 18 reported closed-case CFR or 2 times the ratio of D:C (Eq. 1). 19 20 = min ( + , 2 ) (1) 21

Test Positivity Rate (TPR) 23 TPR was computed as the ratio of cases, C, to tests, T (Eq. 2). 24 25 . 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 June 19, 2020. . https://doi.org/10.1101/2020.06.17.20133389 doi: medRxiv preprint =

(2) 1 2 While the reported number of tests from specific countries or territories may represent multiple tests 3 conducted on a single individual or even number of specimens, no adjustment was attempted to 4 account for such heterogeneity in the various definitions of the TPR. In some cases, T was not 5 available and thus TPR was not calculated. 6 7 Tests per Capita (TPC) 8 TPC was computed as the ratio of tests, T, to population (in millions), P (Eq. 3). 9 10 =

(3) 11 12 Similar to TPR, no adjustments were made to account for inter-country heterogeneity in 13 definitions of T. In cases where T was not available, TPC was also not available. 14 15 Estimating true number of infections and detection rate 16 It can be assumed that the reported number of cases, C, in a country represents a subset of 17 the true number of infections. Some infections will go undetected, but as detection of cases 18

increases, C will approach the true number of infections. Thus, the true number of infections (Inf) is  19 some factor, f, higher than the cases that have been identified (Eq. 4). 20 21 = (4) 22 23 We conceptualized this factor, f, to be a function of the level of testing, the approach to 24 testing (e.g., whether testing focuses on symptomatic, hospitalized, or general populations), and the 25 . 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 June 19, 2020. . https://doi.org/10.1101/2020.06.17.20133389 doi: medRxiv preprint quality and completeness of the data. We then used the CFR, TPR, TPC, Isys, and Idem to formulate 1 numerical values for f. Two separate formulations of f (f1 and f2) were defined, and the maximum of 2 each of the two factors was used in the following analysis, as described below. 3 4 Factor 1 (f1) 5 Health system capacity and government transparency affect the completeness of data. To 6 account for this, two multipliers were constructed: msys (Eq. 5) to adjust for the health system 7 capacity using I sys -an indicator of a health system's ability to detect and report cases; and m dem 8 (Eq. 6) to adjust for government transparency using Idem-an indicator of government transparency 9 in reporting data. The mathematical relationships between the index and the multiplier in each of these 15 equations follow declining relationships asymptotic to 1 and the principle that when health system 16 capacity (Isys) or government transparency (Idem) are reduced, the multipliers increase, representing 17 increased underreporting. 18

Additionally, another indicator of underreporting is a high ratio of deaths to cases, D:C. If 19 all infections have been identified as cases and all cases resolve, then this ratio will approach the 20 infection fatality rate (IFR), which has been estimated for COVID-19 to be around 1% or less [17] . 21

That is, for every recorded death, at least 100 infections occurred. Thus, if the ratio of 100D:C 22 exceeded 1, it was used together with msys and mdem to determine f1, otherwise 1 was used and f1 23 was determined by the product of msys and mdem only (Eq. 7). 24 25 . 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 June 19, 2020. . https://doi.org/10.1101/2020.06.17.20133389 doi: medRxiv preprint 1 = max ( 100 , 1) (7) 1 2 Factor 2 (f2) 3 Inadequate or low testing levels may also affect the completeness of the data. Therefore, a 4 second factor incorporated data on testing (Eq. 8). This factor was comprised of two multipliers: 5 mTPR (Eq. 9) based on TPR-an indicator of adequate testing relative to disease prevalence-and 6 mTPC (Eq. 10) based on TPC-an indicator of adequate testing relative to the population. 7

The World Health Organization (WHO) has suggested testing capacity is adequate when 11 TPR is <10% [30] . If TPR is greater than 10%, it was inferred that increasingly more cases were 12 undetected and that the multiplier mTPR would be equal to the ratio of the TPR to that 10% 13 benchmark (Eq. 9). 14

It was further assumed that if TPC was greater than 100,000 per million, or 10% of the entire 17 population, TPC was extensive enough not to contribute to underreporting (i.e., mTPC = 1). 18

Logically, as TPC decreases, the likelihood of undetected infections increases. The assumed 19 relationship between the multiplier mTPC and TPC followed a step-wise logarithmic relationship 20 (Eq. 10). 21

. 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 June 19, 2020. 

3 The true number of infections, Inf, was estimated as the product of C and the maximum of 4 the two factors, whichever was highest (Eq. 11). 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 June 19, 2020. computable with the methods described above. The relationship between each sub-index and its 1 indicator was built on two principles. First, each sub-index was scaled from 0 to 100 with 0 2 representing the worst indicator value and 100 the best indicator value. Second, a square root 3 mathematical relationship was used to scale the sub-index. A square root function was chosen as a 4 simplistic way to reflect the complex reality, in which each marginal change of an indicator in the 5 undesired direction (e.g., increase in TPR or decrease in DR) represents an increasingly higher risk 6 of an uncontrolled COVID-19 epidemic or a worse testing response. These sub-indices were 7 combined in a weighted average to compute CovTI. 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 June 19, 2020. . https://doi.org/10.1101/2020.06.17.20133389 doi: medRxiv preprint Case-Fatality sub-index (CFsi) 1 The WHO has suggested that the CFR should be around 1 to 2 percent if the response is 2 adequate. If CFR is higher, only severely infected patients are being diagnosed. Accordingly, if CFR 3 increases beyond the minimum benchmark of 2%, CFsi decreases (Eq. 13). If the CFR is less than 4 2%, the CFsi is 100. CFsi was given 20% weighting. If the CFR was 0%, a dummy value of 50 was 5 used. Active Case sub-index (ACsi) 10 Finally, a fourth sub-index accounts for how active the epidemic is in a country. If the 11 epidemic is relatively active in the country, it is less likely the testing is adequate, and the increase 12 reflects inadequate case identification. It provides a metric to incorporate progress as cases resolve. 13

As the proportion of total cases that are active decreases, it reflects the epidemic is declining and 14 passing, and the ACsi increases (Eq. 15). It was given a weight of 20%. 15 16 = 100 − 100 √ (15) 17

In the case of United Kingdom and the Netherlands that do not report A, the ACsi is not computable. 19

In such cases, a dummy value of 50 was used. 20 21 . 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 June 19, 2020. . https://doi.org/10.1101/2020.06.17.20133389 doi: medRxiv preprint COVID-19 Testing Index (CovTI) 1 The CovTI was calculated as the weighted average of the four sub-indices (Eq. 16) described 2 above with a heavier weighting given to the DRsi due to the importance of undetected cases in driving 3 uncontrolled epidemics and because it incorporates several factors into its derivation. 4

Statistical analyses with CovTI 8 Five independent grouping variables were assessed for their effect on COVID-19 testing 9 effectiveness by analyzing their association with CovTI (Table 1) . Testing and contact tracing 10 policy status were accessed from the Oxford COVID-19 Government Response Tracker [32] for 11

May 13, 2020, which is three weeks prior to June 3, 2020, approximately the average time from 12 symptom onset to death [19] . Islands were defined as any country that is an island, part of an island 13 (co-island), or has limited land connections (limited land) or archipelago (see details in S1 Table) . 14 Crude bivariate analyses using two-tailed two-sample t-tests and one-way analysis of variance 15 (ANOVA) were used to test whether the means between groups were different. A multiple linear 16 regression (MLR) model was developed by using forced entry of all factors with p>0. 20 

Operational Definition a n (%)

. 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 June 19, 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.

The copyright holder for this preprint this version posted June 19, 2020. the period prior to June 3, 2020, compared to the reported 6.47 million cases (mean multiplier 5 factor, f, =10.5, range = 1.5 to 165). In other words, for every reported case it is estimated that 9.5 6 infections have gone unreported. 7 8 Detection rate 9 Globally, the DR was estimated to be 9.5% (range = 0.6-66.6%). The countries estimated to 10 have the highest detection rates were Australia, Singapore, Iceland, and Hong Kong ( . 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 June 19, 2020. (Table 3) . 3

The results showed that this model's estimates were similar to previous estimates at comparable 4 time periods (R 2 =0.97). In many cases the estimates of true number of cases and DR closely 5 matched previous estimates, and in most cases the estimates were within the 95% confidence 6 interval of previous estimates. . 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 June 19, 2020. 

3 Countries in the top quartile of CovTI had lower TPR, lower CFR, lower proportion of 4 active cases, and higher DR (Fig 1) . The top 10 countries according to the index were dominated by 5 island nations and states that are effectively islands (e.g., Hong Kong) (Table 4a) . Among non-6 island nations, Thailand, Slovakia, and Israel had the highest CovTI (Table 4b) . Full results are 7 reported in Supporting Information (S1 and S2 Tables). DRsi, TPsi, CFsi, ACsi, as described in text with percentages indicating degree of weighting. Data per 6 00:00 GMT June 3, 2020. Complete data set in S1 . 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 June 19, 2020. Comparing the index from April 20 to June 3, the global average of CovTI increased nearly 3 10 points from 43.2 to 52.2 (Fig 2) . The index in most countries (89.9%) increased over this time 4 period. Some countries, such as Australia, increased substantially, while other countries, such as 5 Germany and USA, increased comparably to the global average increase. Other countries (10.1%) 6

had index values that decreased, including Russia, Canada, and Brazil. 

Bivariate analyses showed that testing policy and contact tracing policy were significantly 14 associated (p<0.0001) with CovTI (Table 5) . Additionally, the trend demonstrated that increasing 15 levels of testing and contact tracing was associated with improved CovTI. Island nations were 16 significantly more effective than non-islands, and significant differences in CovTI were found 17 . 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 June 19, 2020. . between forms of government. OECD members had better COVID-19 testing outcomes, but the 1 differences were not significant (p=0.11). 2 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 June 19, 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.

The copyright holder for this preprint this version posted June 19, 2020. [24]. However, this is the first published metric to comprehensively assess testing outcomes with a 8 focus on detection/underreporting. The results showed that testing and contact tracing 9 independently contribute to better detection and should be used in conjunction with one another. 10

The results showed that the parsimonious model yielded estimates of true period prevalence 11 consistent with previous estimates as shown below. 12 13 National testing policy impact on testing outcomes 14 Testing effectiveness, as indicated by CovTI, was strongly associated with contact tracing 15 and testing policies. This validates the model's ability to track COVID-19 testing outcomes and 16 . 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 June 19, 2020. . further provides confirmation that countries that prioritize policies and dedicate resources 1 specifically to testing and contact tracing have effectively reduced underreporting and improved 2 testing-related health outcome metrics, such as the CFR. 3

The results also show that increasingly more inclusive policies for testing (for example,  4 open public testing vs. symptomatic testing) and contact tracing (e.g., extensive vs. limited) yielded 5 progressively better detection rates and testing outcomes. Therefore, countries should aim to 6 increasingly expand these efforts. Additionally, contact tracing and testing independently 7 contributed to better outcomes, even after adjusting for other factors. Hence, policies should ensure 8 increased testing is paired with increased contact tracing capacity. for autocracies, i.e. democracies have higher CFR than autocracies [21] . A higher CFR indicates 24 poorer testing effectiveness, i.e. lower CovTI (Fig 1) . Here we show that centralized governments 25 have higher CovTI and lower CFR, which is consistent with Sorci et al. [21] . Thus, extrinsic factors 26 . 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 June 19, 2020. . affect a country's ability to implement testing strategies. Combining this metric with other 1 databases that can account for other possible factors, such as trust in government institutions, 2 demographics, or urban/rural distribution, could further elucidate other extrinsic factors related to 3 COVID-19 testing. showing that true number of infections are many times higher than reported cases [9] [10] [11] . In (Table 3) . 17

Seroprevalence of antibodies is often touted as a reliable means to estimate period 18 prevalence and past exposure to SARS-CoV-2. Efforts are underway in June 2020 across England, 19

Germany, and the United States, among others, to randomly sample the population and determine 20 country-level seroprevalence. However, due to the specificity of the serological tests, false positives 21 can substantially affect the accuracy of the results[42] Therefore, models and parsimonious 22 estimates may continue to play an important role in estimating the true number of infections. 23 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 June 19, 2020. . https://doi.org/10.1101/2020.06.17.20133389 doi: medRxiv preprint Herd immunity 1 The estimates of this model further agree with models in the hardest hit countries, such as 2 Italy or Spain, which estimate that more than five percent of the population in those countries have 3 been infected [41] . However, such a low proportion of the population presumably with antibodies is 4 far from conferring herd immunity that may inhibit future disease transmission. It is important to 5 note that, while these proportions are much higher than the officially reported cases, they do not 6 represent herd immunity-a concept considered important to fully reopening society. Although 7 herd immunity depends on the basic reproductive number (R0) [43] , which varies with effectiveness 8 of interventions, some estimates specify a threshold of 50 to 60 percent seroprevalence to achieve 9 herd immunity [44], while others, accounting for differential susceptibility, estimate the threshold 10 may be as low as 20 percent [45] . Nevertheless, these estimates suggest that herd immunity is not 11 yet occurring at the national level of the countries analyzed. 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 June 19, 2020. . policy decisions that systematially exclude some deaths. While proxy factors, including the 1 adjustments for government transparency and health system capacity, were included to mitigate this 2 factor, it is likely the model estimates will not sufficiently adjust in cases of excessive deaths across 3 such wide ranging scenarios. Including data on excessive deaths could improve the validity of the 4 model; however, such data is usually reported with a lag of several weeks or not reported at all. 5 6 Sources of uncertainty 7 Several assumptions were made to generate this model. Most notably, the model assumes 8 specific and universal relationships between deaths and total number of infections, implying an 9 inherent IFR. Advances in therapeutics and differences in health system capacity will influence this 10 rate, though [21] . In addition, several factors including age, sex, hypertension, diabetes, and blood 11 groups are known to affect the mortality and hospitalization rates [50-56]. Future analyses may 12

combine CovTI with databases that include these factors. The model also assumes specific 13 relationships between proxy indicators, such as the Global Health Security Index and Democracy 14 Index, and data outcomes. While the direction of the relationship is arguably evident, the magnitude 15 and shape of the relationship is unknown. Furthermore, this model aims to be parsimonious (i.e., 16 not introducing excessive parameters or uncertainty) and is, by nature, deterministic. The decision 17 to include a minimum number of variables and data was strategic, but a stochastic approach could 18 better illustrate the uncertainty and sensitivity to the above assumptions. The model also reported 19 estimates of total number of cases and detection rates. These values should be used cautiously as a 20 comparative tool, rather than exact values. 21

This report described a novel comprehensive metric (COVID-19 Testing Index, CovTI) that 23 evaluates the overall effectiveness of COVID-19 testing in the current pandemic using real-time 24 publicly reported data among 188 countries and territories. The metric incorporated case-fatality 25 . 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 June 19, 2020. . https://doi.org/10.1101/2020.06.17.20133389 doi: medRxiv preprint rate, test positivity rate, proportion of active cases, and an estimate of detection rate based upon 1 reported death data by adjusting for heterogeneity in testing levels, health system capacity, and 2 government transparency. The estimated detection rate of COVID-19 aligned satisfactorily with 3 previous empirical and epidemiological models. National policies that allow open public testing and 4 extensive contact tracing were significantly associated with higher values of CovTI, which reflects 5 improvements in the estimated detection rate. Extrinsic factors, including geographic isolation and 6 centralized forms of government, were also shown to be associated with higher COVID-19 testing 7 outcomes. Countries should commit to expanding policies on testing and contact tracing in order to 8 reduce levels of undetected infections and reduce disease transmission. Applications of this metric 9 include combining it with different databases to identify other factors that affect testing outcomes or 10 using it to temporally track a holistic measure of testing outcomes at the national level. . 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 June 19, 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 June 19, 2020. . https://doi.org/10.1101/2020.06.17.20133389 doi: medRxiv preprint CovTI on June 3, 2020 among eligible countries and territories (n=188). Further details of each variable are 23 described in the text. C=cases, D=total deaths, R=total recovered, A=active cases, P= population (in 24 millions), T=total tests, CFR=case fatality rate, TPR=test positivity rate, TPC=tests per capita, msys=health 25 . 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 June 19, 2020. 13, 2020; and COVID-19 contact tracing policy as of May 13, 2020 among eligible countries and 10 territories with complete data (n=163). Further details of each variable are described in the text. 11

. 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 June 19, 2020. . https://doi.org/10.1101/2020.06.17.20133389 doi: medRxiv preprint

Evidence based management 13 guideline for the COVID-19 pandemic -Review article

17 3. World Health Organization. World Health Organization. Critical preparedness, readiness and 18 response actions for COVID-19 (Interim Guidance)

Physical distancing, 5. US CDC. CDC Activities and Initiatives Supporting the COVID-19 Response and the

19-Response.pdf 4 6. UK Government. Our plan to rebuild: The UK Government's COVID-19 recovery strategy

Estimating the asymptomatic proportion of 8 coronavirus disease 2019 (COVID-19) cases on board the Diamond Princess cruise ship

Euro Surveill Bull Eur sur les Mal Transm = Eur Commun Dis 10

Asymptomatic Transmission, the Achilles' Heel of 12

Current Strategies to Control Covid-19

Estimating COVID-19 Prevalence in the United States: A 15

Estimating the Prevalence of COVID-19 in the 18 United States: Three Complementary Approaches

Level of underreporting including underdiagnosis before the first 21 peak of COVID-19 in various countries: Preliminary retrospective results based on wavelets 22

Estimation of COVID-19 outbreak size in Italy

Modelling 4 the COVID-19 epidemic and implementation of population-wide interventions in Italy

Why daily death tolls have become unusually important in understanding the 7 coronavirus pandemic

Pandemic potential of a strain of influenza A (H1N1): early findings

A systematic review and meta-analysis of published research 12 data on COVID-19 infection-fatality rates

COVID-19: in the footsteps of Ernest Shackleton

Estimates of the 17 severity of coronavirus disease 2019: a model-based analysis

Estimating the infection and case fatality ratio for coronavirus disease (COVID-19) using 21 age-adjusted data from the outbreak on the Diamond Princess cruise ship

Relation to COVID-19 in Italy

On Target 2 [Internet]. 2020

COVID-19: how to make between-country comparisons

Whose coronavirus strategy worked best? Scientists hunt most effective policies

Global Health Security Index

Estimating case fatality rates of COVID-15 19. The Lancet. Infectious diseases

Coronavirus Testing Still Falls Short. How's Your State 17 Doing?

States Meet WHO Recommended Testing Criteria?

COVID-19 and ABO blood group: another viewpoint

Association between ABO blood groups and 14 risk of SARS-CoV-2 pneumonia

Risk factors of critical & mortal 16 COVID-19 cases: A systematic literature review and meta-analysis

Raw data inputs, key epidemiological indicators, multipliers, factors, and sub-indices used to compute