key: cord-0322404-dno22u52 authors: Levitt, M.; Zonta, F.; Ioannidis, J. P. A. title: Comparison of pandemic excess mortality in 2020-2021 across different empirical calculations date: 2022-05-10 journal: nan DOI: 10.1101/2022.05.09.22274863 sha: 8817ed70ad1baea07dcb5542583e6c75a6e7ccea doc_id: 322404 cord_uid: dno22u52 Different modeling approaches can be used to calculate excess deaths for the COVID-19 pandemic period. We compared 6 calculations of excess deaths (4 previously published and two new ones that we performed with and without age-adjustment) for 2020-2021. With each approach, we calculated excess deaths metrics and the ratio R of excess deaths over recorded COVID-19 deaths. The main analysis focused on 33 high-income countries with weekly deaths in the Human Mortality Database (HMD at mortality.org) and reliable death registration. Secondary analyses compared calculations for other countries, whenever available. Across the 33 high-income countries, excess deaths were 2.0-2.8 million without age-adjustment, and 1.6-2.1 million with age-adjustment with large differences across countries. In our analyses after age-adjustment, 8 of 33 countries had no overall excess deaths; there was a death deficit in children; and 0.478 million (29.7%) of the excess deaths were in people <65 years old. In countries like France, Germany, Italy, and Spain excess death estimates differed 2 to 4-fold between highest and lowest figures. The R values range exceeded 0.3 in all 33 countries. In 16 of 33 countries, the range of R exceeded 1. In 25 of 33 countries some calculations suggest R>1 (excess deaths exceeding COVID-19 deaths) while others suggest R<1 (excess deaths smaller than COVID-19 deaths). Inferred data from 4 evaluations for 42 countries and from 3 evaluations for another 98 countries are very tenuous Estimates of excess deaths are analysis-dependent and age-adjustment is important to consider. Excess deaths may be lower than previously calculated. Many studies estimate excess deaths in specific locations, countries, regions, or worldwide during the COVID-19 pandemic. [1] [2] [3] [4] Excess deaths reflect a composite of deaths from SARS-CoV-2 infection plus indirect effects of the pandemic (e.g. health system strain) and measures taken. 5, 6 It has been argued 7,8 that excess deaths are a more appropriate measure of impact than recorded COVID-19 deaths. Recorded COVID-19 deaths may be under-or overcounted in different time periods and locations 5 and do not capture indirect effects of the pandemic and the measures taken. However, excess deaths calculations require modeling of the expected deaths that entails many assumptions and analytical choices. To obtain excess deaths estimates one needs to define a control (reference) pre-pandemic period, use some model for extrapolating expected deaths in the pandemic period and compare them against observed deaths. There are many different possibilities on how to select the pre-pandemic reference period and on how to model data and extrapolations. 9,10 A major analytical decision is whether to account for changes in the age structure of the population over time. With an aging population (particularly in high-income countries), mortality rates may increase over time, countering the anticipated decrease in mortality from better healthcare and overall human progress. These changes may be better addressed by considering age structure 9,11-13 rather than simply relying on regression trends of overall population data regardless of age. Another pivotal dilemma in excess death calculations is what sources of data to use; and which countries are considered to have sufficiently reliable data. Death registration is substandard in most countries around the world: many deaths remain unrecorded. 14, 15 Information on causes of death has flaws even in the most developed countries 16, 17 while the pandemic generated new death coding challenges. 18 Changes in deaths over time may be confounded by changes in death registration and recorded COVID-19 deaths depend on coding. Even population counts have uncertainty and this applies also to age-stratified estimates from different sources using different imputations to estimate age-stratified population for recent years. Even in high-income countries, most of them have not had a formal census performed for many years. Finally, several highly visible studies that attempted to calculate excess mortality world-wide, 1-4 first estimated excess mortality in countries with trustworthy death data on all-cause mortality; then, they extrapolated across other countries worldwide, using the profile of various characteristics in these countries versus those with trustworthy data. Consequently, proper calculation of excess mortality in countries with most trustworthy mortality data has critical importance even for worldwide estimates. Here, we compare the results of different evaluations that have attempted to calculate global excess mortality during the COVID-19 pandemic in 2020-2021. [1] [2] [3] [4] We focus primarily on high-income countries with the most reliable death registration systems and discuss the implications for extrapolations to a global level. We compared 4 widely publicized excess death calculations published in eLife, Economist, the Lancet, and by the World Health Organization (WHO) [1] [2] [3] [4] 19, 20 and included also our calculations. Our calculations explicitly explored what difference it would make to model deaths using separate death and population data for age strata. Besides raw excess mortality estimates, we focused on the ratio of excess mortality over recorded COVID-19 deaths. This ratio is critical in understanding whether excess mortality confers more information compared with just focusing on routinely recorded COVID-19 deaths. Ideally, one would like to see consistency in this ratio regardless of modeling and analytical . CC-BY 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 10, 2022. ; https://doi.org/10.1101/2022.05.09.22274863 doi: medRxiv preprint choices, while large inconsistency would put the added value of excess mortality calculations in question. Characteristics of the compared excess death calculations Table 1 summarizes the main features of the compared excess death calculations. As shown, the evaluations differed substantially in defining the reference period, the choice of analytical model, exclusions, use of adjustments, eligibility criteria, data sources, extrapolations and imputations. The Economist methods used complex machine learning not described in very broad terms; the WHO and Lancet methods were convoluted and not all can be easily reproduced. Comparison of excess deaths metrics in 33 countries with most-reliable data 33 countries were included in the main analysis, selected because they were high-income according to the World Bank, and they had detailed available weekly data on observed deaths in the three pre-pandemic years and also during 2020-2021 (Table 2) . These countries had a total population of almost 1 billion and almost 20 million deaths in 2020-2021, of which 9.5% were recorded as COVID-19 deaths. The total of calculated excess deaths in these 33 countries ranged from 2.0 million (eLife) to 2.8 million (Lancet), with WHO (2.1 million) and Economist (2.2 million) being closer to eLife. Our own calculations without age adjustment gave a similar total (2.3 million), but with age-adjustment the excess deaths were only 1.6 million. 8 of 33) had some calculations suggesting R>1 (i.e. excess deaths greater than COVID-19 deaths) and others suggesting R<1 (excess deaths smaller than COVID-19 deaths). Table 4 ) for a total of 10.7, 9.8, and 7.9 million, respectively (6-9 times higher than recorded COVID-19 deaths). The correlation of the . CC-BY 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) Across 33 high-income countries with total population of approximately 1 billion and highly thorough death registration, different empirical estimates of excess deaths in 2020-2021 ranged widely from 1.6 million, i.e. substantially fewer than the 1.9 million recorded COVID-19 deaths, to 2.8 million, almost a million more. Countries with highest estimates of excess deaths were more or less the same in all evaluations; and countries with the most favorable picture performed well across the different evaluations. However, large differences emerged in the magnitude of the estimates for each country. The largest divergence was produced by whether age adjustment was used in calculating excess deaths. Age-adjusted estimates were lower. They are more appropriate, since COVID-19 has impressive age-related risk gradient for death. 20,21 Modest changes in age structure, in particular with aging populations, may produce major differences in estimates. With age-adjustment, several high-income countries showed no or minimal excess deaths; and for many others, the estimates of excess deaths markedly decreased. However, even age-adjusted excess death estimates differed substantially depending on the choice of data source for age-stratified population in the recent years as well as modeling choices. Our age-adjusted analyses also revealed large differences across countries in the proportion of excess deaths accounted by non-elderly age strata. A few countries had more than a . CC-BY 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) insights on the total pandemic toll than COVID-19 recorded deaths. 7 Excess death calculations are dependent on how they are calculated. In some cases, like Eastern Europe, they can tell that COVID-19 deaths have been undercounted and/or the numbers of other deaths have escalated during the pandemic. However, they cannot differentiate the relative contribution of these two factors nor can they give a precise estimate of either or their combination. In a few countries with limited SARS-CoV-2 deaths they can be reassuring that indirect pandemic effects and measures did not escalate fatalities, at least during 2020-2021. However, for most countries, excess death calculations are so model-dependent that they should be seen with great caution. . CC-BY 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 10, 2022. Besides the high-income countries with meticulous weekly death registration data, excess death calculations for the rest of the world are very tenuous exercises. For 42 countries where eLife, Economist, Lancet and WHO generated estimates, on average excess deaths were ~2times the COVID-19 recorded deaths and among another 98 countries where Economist and Lancet provided estimates, overall R was 6-9. However, given the low reliability of the data, the immense uncertainty surrounding these estimates cannot be overstated. More importantly, these calculations offer no causal insights. Excess deaths may be due to the virus, the indirect pandemic effects, or/and disruptive measures taken, even more so in countries with very frail healthcare systems, widespread poverty, and/or even high rates of hunger. The 2020-2021 crisis may have indeed resulted in many deaths, but causes may be very complex. For some of these additional countries, calculations were run based on available mortality data. Even then, in most cases death registration is unreliable and the impact of changes in death registration during the pandemic compound any calculation. Perhaps age-adjustment would also lead to different estimates in these countries, as for the 33 most data-reliable countries. Even for these 33 countries, population estimates (including age-stratified population counts) are projected with different methods and typically no formal population census has been performed for several years. This adds further uncertainty to excess death calculations that are highly susceptible to minor differences especially in the population of elderly strata. For most countries, excess death calculations are indirectly imputed from the countries with mortality data. The methods employed by the Economist are not described in sufficient detail to allow probing validity and reproducibility. Lancet and WHO calculations provide more elaboration, mostly proving their complexity. The uncertainty in excess death estimates apparently far exceeds the width of published confidence intervals. . CC-BY 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 10, 2022. ; Hence, extrapolations from the 33 main analysis countries to other countries need to be extremely cautious. Among the total excess deaths, deaths due specifically to viral infection in the other countries may be proportionally far less. Other countries have far smaller percentage of elderly people and very few nursing home residents. Conversely, they have far more frail health systems, societies, and economies. Therefore, probably indirect effects of the pandemic and measures were perhaps more important contributors to total excess deaths than SARS-CoV-2 infection. Nepomuceno et al. have also assessed the impact of different analytical choices on excess death calculations for 2020. 8 They find modest differences among different approaches in countries with reliable death registration. Age-stratification has been also to be important in previous assessments for specific countries. 24, 25 Germany is a classic example. Our age-adjusted estimate is 55,000 excess deaths, while without age-adjustment we calculated 129,000 excess deaths and Lancet calculated 203,000 excess deaths -compared with 111,000 COVID-19 reported deaths. Baum 23 calculated only 22,000 excess deaths after age adjustment, while Koenig 24 without age adjustment found more excess deaths than recorded COVID-19 deaths. In Germany, the number of people aged >80 years increased from 4.6 million in 2016 to 5.8 million in 2020, so consideration of age is crucial. 11 Further diversity stems from whether modeling anticipates increasing life expectancy (decreasing mortality rate) over time. 11 However, this approach may calculate spuriously high excess deaths: calculations assume a desired mortality rate lower than even attained in reality in the past. Medical and overall progress cannot guarantee continuing to decrease overall mortality in high-income countries, especially in old, frail people. In fact, care of such people may have deteriorated over time (e.g. with privatization and deterioration of long-term care) and the pandemic brought this to light. 26 Moreover, data from . CC-BY 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 10, 2022. ; much longer-term periods of observation suggest multiple, overlapping, complex long-term trends in winter mortality. 27 Finally, some other models may diverge in their calculations, if they exclude certain periods. E.g., the popular Euromomo model using a Serfling model in its core but excludes weeks of high influenza activity from the modeling: thus it generates high excess death estimates. 28 Some other caveats should be discussed. First, finer adjustment for age (e.g. in more narrow age bins), and more comprehensive adjustment for other factors (e.g. gender, frailty, long-term care facility residence, comorbidities) may be able to offer even more accurate estimates of excess deaths. Second, even in countries with reliable data, some deaths are registered with delays, although with >4 months after the end of 2021 our main analyses for the 33 countries should not be affected much. Third, there can be debate on whether/how natural disasters and wars should be excluded. Fourth, long-term effects of both the pandemic and measures taken on healthcare, other aspects of health, education, society, and economy remain uncaptured in the 2020-2021 window. Comparisons of different approaches to excess death calculations should continue for longer periods of follow-up, also in the post-pandemic endemic phase. The boundaries between pandemic and endemic phase can be debated. 29 Regardless, the relative performance of different countries and their excess death ranking may change substantially over time. Sixth, analyzing specific causes of death may be informative, but suffers from major misclassification even in high-income countries. Acknowledging these caveats, our analyses map the magnitude and uncertainty of excess deaths during 2020-2021. In countries with reliable data, age-adjustment suggests that excess deaths is lower than what has been previously published in calculations without age-adjustment. Excess death calculations convey some broad picture, especially for countries that fared very . CC-BY 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 10, 2022. ; well or very poorly. Large differences in the impact of a pandemic across countries has been seen also in previous pandemics 30, 31 for reasons that often remain largely unexplained. In depth assessments with death certificate audits, medical record audits, and autopsies may yield more granular insights about deaths and their causes, but these approaches also have limitations and feasibility challenges. For most countries worldwide, the tremendous uncertainty in the sparse data and indirect inferences should be mostly a call for improving the completeness and accuracy of death registration and investment in more rigorous demography infrastructures in the future. We considered four previously published pandemic excess mortality evaluations [1] [2] [3] [4] 18, 19 that have calculated both country-specific and global estimates and which use diverse methods to extrapolate from pre-pandemic reference periods to the pandemic period. respectively. Both the eLife and Economist models allow updating of excess mortality estimates over time and we have used the Our World in Data resource 18,19 that includes such updates. We used the estimates of excess mortality for the 2020-2021 two-year period for all 4 evaluations to maximize comparability. The three evaluations used also different sources for capturing the recorded numbers of COVID-19 deaths, which resulted in mostly minor discrepancies. Again, to maximize comparability, we used the same set of recorded numbers of COVID-19 for comparing . CC-BY 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 10, 2022. ; against the excess mortality estimates of each evaluation: this set is identical to the set used by the Lancet evaluation, with the exception of Spain and UK where the Lancet numbers of recorded deaths were too high by 10% and 16%, respectively (perhaps due to clerical error) and where we used the Johns Hopkins data instead. 32 The reported cumulative counts of COVID-19 deaths are taken from the Johns Hopkins Evaluation considering age strata in the calculations . CC-BY 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 10, 2022. ; We performed also our own calculation of excess deaths focused on considering the impact of age-adjustment on the calculations. We considered three pre-pandemic years (2017 to 2019) as the reference period. We used the following age strata: 0-14, 15-64, 65-75, 75-85, and >85 years old. For each age stratum, we obtained the average mortality, the number of deaths per million for the population of the specific age stratum. Then we extrapolated to the two pandemic years, again correcting for the population size in the specific age stratum. Finally, expected deaths were summed across the population strata. An illustrative worked example of the ageadjusted calculations is shown in Figure 1 using the data for Germany. This analytical approach corrects for changes in total population of a country over time, as well as changes in the proportion of elderly people. If the total population and/or the proportion of older people is increasing in recent years over time, the expected deaths by the age-adjusted scheme will be more compared to analyses that do not consider population changes and age-stratification. The inverse will happen, if the total population and/or the proportion of older people is decreasing in recent years over time. We used the Human Mortality Database (https://www.mortality.org), 33 specifically the Short Term Mortality Fluctuations file (https://www.mortality.org/Public/STMF/Outputs/stmf.csv) that includes weekly all-cause deaths and mortality values allowing the population size to be calculated as their ratio for each week. Some clarifications are required for these analyses. First, on the calculation of the Weekly Population from the weekly Death Count and Weekly Mortality given in the MDH file stmf.csv. As Population = Death/Mortality, when Death = 0, Mortality = 0, so that Population is not defined. In these cases, we get the Population value from adjacent weeks with non-zero Death in the same year. Another problem is that some years, like 2015 & 2020 have an extra week 53, a leap week. This is because 52 weeks is 364 days whereas an average year is 365.25 days (a leap . CC-BY 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 10, 2022. ; year every fourth year adds an extra day). Since mortality data are reported weekly, these years would get an apparent extra week of deaths. To correct this issue, we consider a standard year to be 365. 25 We averaged the mortality values for each age-stratum over the reference years. We did not exclude any deaths or periods due to heat waves or other natural or man-made events, since it is subjective to arbitrate which ones should be excluded; moreover, we wanted to compare the COVID-19 pandemic with three recent years that had not raised any concerns about undue levels of deaths. We also repeated the excess death calculations using the same exact process but without considering age-strata. We used population data, including also age-strata, from mortality.org. 34 We focused our primary comparison on countries that have excellent death registration, are high income, and include data with weekly deaths in the Human Mortality database. These countries have the most reliable evidence to allow proper excess death calculations and changes in deaths over time cannot be due to changes in death registration (e.g. either improvements in death registration during the pandemic as countries put more resources on capturing information or worsening of death registration during the pandemic due to the pandemonium and acute death . CC-BY 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. We calculated excess deaths for the full two years 2020-2021 for each country, E. We expressed them also as percentage above the expected deaths, E p and as excess deaths per million E m . E.g. if 60,000 deaths happened in 2020-2021 in one country of 4,000,000 people and 50,000 were expected, the percentage E P is (60,000-50,000)/50,000=12% and E m is (60,000-50,000)/4,000,000=2,500 per million. As the main metric of interest, we used the ratio of estimated excess mortality E divided by the number of officially recorded COVID-19 deaths during 2020-2021, hence called the excess ratio R. E.g. R=1.20 means that the estimated excess mortality is 20% higher than the officially recorded COVID-19 deaths and R=0.90 means that the estimated excess mortality is 10% lower than the officially recorded COVID-19 deaths. We calculated the total number of excess deaths across the eligible countries with each of the different calculation methods, and as compared with the total recorded COVID-19 deaths. We also calculated the Pearson correlation coefficients of excess death metrics across the eligible countries based on the different calculation methods (a full list of correlation coefficient estimates appears in Supplementary Table 5 ). . CC-BY 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 10, 2022. ; https://doi.org/10.1101/2022.05.09.22274863 doi: medRxiv preprint We noted in how many of the 33 eligible countries the different calculations of excess deaths had consistently R above 1 or R below 1 (i.e. they all agreed that excess deaths were more than the recorded COVID-19 deaths or they all agreed that excess deaths were less than the recorded COVID-19 deaths); and in how many countries had R values that were all within a range of 0.3 between the highest and lowest estimate (a higher range means that divergence in the excess death estimates exceeds 30% of the recorded COVID-19 deaths); and within a range of 1 (a higher range means that the divergence in the excess death estimates exceeds the number of recorded COVID-19 deaths itself). For the remaining, non-eligible countries, we evaluated excess death metrics and their comparison across different evaluations as exploratory analyses. All analyses were done independently by two analysts (ML and FZ) and then compared notes with arbitration (including discussion with the third author, JPAI) for any disagreements until both analysts obtained the same results. All data are in the manuscript, tables, and supplementary tables and in the publicly available databases listed in Supplementary Links to Data . CC-BY 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 10, 2022. ; 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 10, 2022. 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 10, 2022. 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 10, 2022. 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 10, 2022. . 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 10, 2022. 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 10, 2022. 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 10, 2022. 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 10, 2022. . 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 10, 2022. 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 10, 2022. 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 10, 2022. 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 10, 2022. 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 10, 2022. 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 10, 2022. 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 10, 2022. Supplementary 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 10, 2022. 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 10, 2022. Table 3, supplementary Table 1, supplementary Table 2, supplementary Table 3, and supplementary Table 4 ) between the different evaluations R ( 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 10, 2022. Ep ( 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 10, 2022. Em (Table S2) 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 10, 2022. Em (Table S3) 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 10, 2022. . 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 10, 2022. Figure 1 : Schematic Representation of the Calculation of Age Adjusted Excess Death. Populations and total amounts of death on a single year are used to obtain the mortality values for the reference years (2017-2019) for each of the age strata. The average of this value is taken as the reference mortality for that strata. Expected deaths for a non COVID scenario for 2020 and 2021 are obtained from the population and reference mortality data for each strata. Excess deaths are calculated as the difference between the actual deaths and the expected deaths. Expected and excess deaths for the non age adjusted case are also reported for comparison. The table reports the real values for Germany as an example. This table is provided as Excel in the Supplement so that method can be easily used on other data. . 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 10, 2022. Tracking excess mortality across countries during the COVID-19 pandemic with the World Mortality Dataset. Elife Tracking COVID-19 excess deaths COVID-19 Excess Mortality Collaborators. Estimating excess mortality due to the COVID-19 pandemic: a systematic analysis of COVID-19-related mortality World Health Organization, Global excess deaths associated with COVID-19 Over-and under-estimation of COVID-19 deaths Every Body Counts: Measuring Mortality From the COVID-19 Pandemic Excess deaths" is the best metric for tracking the pandemic Covid-19: excess deaths should be the outcome measure Sensitivity analysis of excess mortality due to the COVID-19 pandemic Excess deaths associated with covid-19 pandemic in 2020: age and sex disaggregated time series analysis in 29 high income countries Excess mortality due to Covid-19? A comparison of total mortality in 2020 with total mortality in Gender specific excess mortality in Italy during the COVID-19 pandemic accounting for age Excess mortality due to COVID-19 in Germany Estimating the completeness of death registration: An empirical method A global assessment of civil registration and vital statistics systems: monitoring data quality and progress Ill-defined and multiple causes on death certificates--a study of misclassification in mortality statistics Swiss National Cohort Study Group. Cause of death coding in Switzerland: evaluation based on a nationwide individual linkage of mortality and hospital in-patient records Different approaches to the analysis of causes of death during the COVID-19 epidemic Estimated excess death count based on Karlinsky and Kobak methods (as of Estimated excess death count from the Economist (as of Age-specific mortality and immunity patterns of SARS-CoV-2 Infection fatality rate of COVID-19 in community-dwelling elderly populations Time-Varying Death Risk After SARS-CoV-2-Infection in Swedish Long-Term Care Facilities Bollmann excess mortality exceeding COVID-19-attributed fatalities COVID-19 excess mortality among long-term care residents in Ontario Trends in Excess Winter Mortality (EWM) from 1900/01 to 2019/20-Evidence for a Complex System of Multiple Long-Term Trends Robustness and bias of European excess death estimates in 2020 under varying model specifications The end of the COVID-19 pandemic Global Mortality Impact of the 1957-1959 Influenza Pandemic Hong Kong influenza pandemic: evidence for a smoldering pandemic Methods protocol for the human mortality database Human mortality database. Encyclopedia of gerontology and population aging World population prospects 2019: department of economic and social Affairs Economist 3-Apr-2022 https://ourworldindata.org/excess-mortality-covid#estimated-excess-mortality-from-theeconomist excess-deaths-cumulative-economist-single-entity.csv