key: cord-1019633-qzb8a22l authors: Molenberghs, G.; Faes, C.; Aerts, J.; Theeten, H.; Devleesschauwer, B.; Bustos Sierra, N.; Braeye, T.; Renard, F.; Herzog, S.; Lusyne, P.; Van der Heyden, J.; Van Oyen, H.; Van Damme, P.; Hens, N. title: Belgian Covid-19 Mortality, Excess Deaths, Number of Deaths per Million, and Infection Fatality Rates (8 March - 9 May 2020) date: 2020-06-20 journal: nan DOI: 10.1101/2020.06.20.20136234 sha: d94b5de7bfa5b098051f7140cd951b9c5dac7d01 doc_id: 1019633 cord_uid: qzb8a22l Objective. Scrutiny of COVID-19 mortality in Belgium over the period 8 March-9 May 2020 (Weeks 11-19), using number of deaths per million, infection fatality rates, and the relation between COVID-19 mortality and excess death rates. Data. Publicly available COVID-19 mortality (2020); overall mortality (2009-2020) data in Belgium and demographic data on the Belgian population; data on the nursing home population; results of repeated sero-prevalence surveys in March-April 2020. Statistical methods. Reweighing, missing-data handling, rate estimation, visualization. Results. Belgium has virtually no discrepancy between COVID-19 reported mortality (confirmed and possible cases) and excess mortality. There is a sharp excess death peak over the study period; the total number of excess deaths makes April 2020 the deadliest month of April since WWII, with excess deaths far larger than in early 2017 or 2018, even though influenza-induced January 1951 and February 1960 number of excess deaths were similar in magnitude. Using various sero-prevalence estimates, infection fatality rates (IFRs; fraction of deaths among infected cases) are estimated at 0.38-0.73% for males and 0.20-0.39% for females in the non-nursing home population (non-NHP), and at 0.79-1.52% for males and 0.88-1.31% for females in the entire population. Estimates for the NHP range from 38 to 73% for males and over 22 to 37% for females. The IFRs rise from nearly 0% under 45 years, to 4.3% and 13.2% for males in the non-NHP and the general population, respectively, and to 1.5% and 11.1% for females in the non-NHP and general population, respectively. The IFR and number of deaths per million is strongly influenced by extensive reporting and the fact that 66.0% of the deaths concerned NH residents. At 764 (our re-estimation of the figure 735, presented by "Our World in Data"), the number of COVID-19 deaths per million led the international ranking on May 9, 2020, but drops to 262 in the non-NHP. The NHP is very specific: age-related increased risk; highly prevalent comorbidities that, while non-fatal in themselves, exacerbate COVID-19; larger collective households that share inadvertent vectors such as caregivers and favor clustered outbreaks; initial lack of protective equipment, etc. High-quality health care countries have a relatively older but also more frail population [1], which is likely to contribute to this result. At 764 (our estimate), the number of COVID-19 deaths per million leads the international ranking, but drops sharply to 262 in the non-nursing home population. global infection fatality rates [2] . These authors used asymptotic models to derive IFR as a limit of CFR. CFR is strongly influenced by testing strategy, and in several studies the delay between case confirmation and deaths is not accounted for. The handling of possible cases is ambiguous at best. We do not consider it here. Bias and precision in estimation of IFR is influenced by difficulties surrounding the estimation of sero-prevalence, such as sensitivity and specificity of the tests used [3] , time to IgM and in particular IgG seroconversion [4] , and potential selection bias occurring in data from residual sample surveys. A sensitivity analysis is undertaken by augmenting one primary with three auxiliary estimates of sero-prevalence. Because in Belgium there is a very close agreement between excess mortality on the one hand and confirmed and possible COVID-19 cases combined on the other, and because an international study [2] suggested that a fraction as high as 0.9 of possible cases could be attributable to COVID-19 [5] , it is a reasonable choice to use all Introduction Belgium's per million COVID-19 related mortality has been reported the highest worldwide over the period April -May 2020. For example, as reported on May 13, 2020 at https://ourworldindata.org/coronavirus-data, the figure is 756 for Belgium, versus 91 for Germany, 414 for France, 511 for Italy, 482 for the UK, 249 for the US, and 328 for Sweden. Because of its relative nature, this measure appears to be objective; nevertheless, it requires scrutiny. To this end, we examine COVID-19 reported mortality over the period 8 March -9 May 2020, and place it against the background of excess mortality in Belgium. The study period is chosen such that there is a sufficiently long data cleaning period, leading to accurate death counts. This allows one to gauge whether there is evidence for over-, under-or sufficiently accurate reporting of COVID-19 cases. Using data on the number of COVID-19 deaths and sero-prevalence estimates based on data from a repeated cross-sectional serological survey [3] , infection fatality rates (IFRs) and number of deaths per million (DPM) are estimated, overall, in relation to age and sex, for the total population, the nursing home population (NHP), and the non-NHP. Covid-19 mortality. The Belgium-based institute for health, Sciensano, reports daily COVID-19 mortality figures (https://epistat.wiv-isp.be/covid/). These daily data were extracted on 5 June 2020 and then binned to form age category by week mortality tables for each of the sexes and for the period 8 March -9 May 2020 (Week 11 - Week 19); age categories (in years) are 0-24, 25-44, 45-64, 65-74, 75-84, and 85+. These six categories are used throughout the analyses. Of the 8732 deaths reported, . 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 20, 2020. Based upon considering various re-distribution methods (details not reported), the imprecision engendered by missingness is ignored, because it is dominated by uncertainty in sero-prevalence estimation, the latter of which is taken into account by precision estimation and statistical sensitivity analysis. Overall mortality. Weekly mortality per sex and age category, for the years 2009 -2019 (complete) and 2020 (until early May 2020) originate from the National Register. Statistics Belgium, the national statistical institute, processes these deaths and integrates them in Demobel, its demographic data warehouse. Open data by district (NUTS 3) can be found in [3] : https://statbel.fgov.be/en/open-data/numberdeaths-day-sex-district-age. Using the years 2009 -2019 combined, a weekly average . 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 20, 2020. . https://doi.org/10.1101/2020.06.20.20136234 doi: medRxiv preprint profile (termed reference year) is obtained, with pointwise corresponding 99% prediction bands. Population sizes. The Belgian population sizes (situation 1 January 2020), by age category and sex, are taken from Statistics Belgium (Demobel), based on National Register data: https://statbel.fgov.be/en/themes/population/structure-population. Sero-prevalences. Based on Herzog et al. [3] age-category specific sero-prevalences referring to April 2020, are used; details on sensitivity and specificity of the tests used can be found there. Data over the age range 0-101 years of age are available. A reanalysis of the data used in [3] provided estimates and confidence intervals for the age bins used in this paper. These estimates use the population structure for Belgium in 2020, as forecast by the Federal Plan Bureau (www.plan.be). Bias and precision in the determination of sero-prevalence depends on sensitivity and specificity of the tests used [1] , time to IgG seroconversion [4] , and potential selection bias occurring in data from residual samples surveys. A sensitivity analysis is undertaken by augmenting this primary with three auxiliary estimates of sero-prevalence for the general population, with in addition several forecast of the sero-prevalence in NH starting from extensive test results (https://covid-19.sciensano.be/sites/default/files/Covid19/COVID-19_Daily%20report_20200526%20-%20FR.pdf). Estimated number of COVID-19 cases. These are calculated by multiplying the age-sex-subgroup (NHP/non-NHP) population sizes with the corresponding seroprevalences, for various sero-prevalence estimates. Assuming an approximately constant sero-prevalence over the month of May, this is considered a sensible approach, even though no delay-adjustment is done. . 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 20, 2020. . https://doi.org/10.1101/2020.06.20.20136234 doi: medRxiv preprint Infection fatality rates. From the sero-prevalence and population size, per category, the actual number of infected cases is estimated. The IFR is calculated as the ratio of the number of deaths (confirmed and possible) over the number of infected cases. This is done per age and sex category, for the general population, the NHP, and the non-NHP. Given the uncertainty in the sero-prevalence estimates, delta-method-based confidence intervals supplement the primary IFR estimates. Case fatality rates (CFR), defined as the number of confirmed deaths (confirmed and possible) over the number of confirmed cases, ideally delay-distribution adjusted, will not be examined in this manuscript. Statistical software. The data analysis was performed using SAS Software, GAUSS, and R; visualizations were made using Vega. Python scripts to reproduce the analyses will be available at https://www.uhasselt.be/DSI. In line with international findings [7] , the number of deaths strongly increases with age. It is difficult to compare sexes in absolute terms, because the higher number of deaths in the female 85+ group, for example, is offset by the fact that the number of males in the 85+ category is less than half in size of the female category. Summary data are given in Table 1 and depicted in Figure 1 . Table 2 presents the figures for the NH population. Because of the large fraction of incomplete records, the redistribution over age categories in Table 2 (32.0% incomplete in terms of age and/or sex), is subject to uncertainty. There are more incomplete records in the NH sub-population than in the general population, because the dataset of NH residents who died in . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 20, 2020. . hospitals is currently not directly extractable from the hospital-deaths dataset, for which age and sex are virtually complete. 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 20, 2020. . https://doi.org/10.1101/2020.06.20.20136234 doi: medRxiv preprint . 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 20, 2020. . https://doi.org/10.1101/2020.06.20.20136234 doi: medRxiv preprint . 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 20, 2020. . https://doi.org/10.1101/2020.06.20.20136234 doi: medRxiv preprint excess mortality) (https://epistat.wiv-isp.be/docs/momo/Be-MOMO%20winter%202017-18%20report_FR.pdf; https://epistat.wiv-isp.be/momo/) the black curve is the average over 2009 -2019, with dashes 99% pointwise prediction bands; the red curve refers to 2020. That the peak is strongly driven by the older age category is clear from Figure 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 20, 2020. . https://doi.org/10.1101/2020.06.20.20136234 doi: medRxiv preprint In the above, reference was made to Belgium's extensive COVID-19 death reporting, as also internationally noted. As we saw in Figure 2 , mortality in January and February 2020 was below the average over 2009-2019, although coherent with the prediction interval. Then, the peak emerges and, most important for this study is the near coincidence of excess and COVID-19 mortality. . 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 20, 2020. . https://doi.org/10.1101/2020.06.20.20136234 doi: medRxiv preprint Now that the close agreement between COVID-19 and excess mortality has been established, it is useful to consider the number of deaths per million inhabitants (DPM). Belgium's DPM has been reported as the highest worldwide over the period April -May 2020. For example, according to https://ourworldindata.org/coronavirusdata (accessed on June 16, 2020), the figures are: 735 for Belgium on 9 May 2020, versus 500 for Italy, 88 for Germany, 402 for France, 460 for the UK, 233 for the US, and 314 for Sweden. Table 3 displays age-and sex-specific DPM, for the general Belgian population, the population without nursing homes, as well as the NHP. Because the Belgian mortality data have been cleaned and verified between the end of the study period (May 9, 2020) and final analysis (June 16, 2020), the overall figure of 764 in Table 3 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 20, 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 20, 2020. . https://doi.org/10.1101/2020.06.20.20136234 doi: medRxiv preprint From Table 3 we deduce a strong age and important sex effect. But what is most striking is the extent of the impact in the NHP. For the non-NHP, the overall figure goes down to 262, which internationally might not stand out, but then reliable figures are needed of other countries' NHP/non-NHP as well. For a nursing home population of around 1% of the total population, this effect is striking. It is clear from Table 3 that the overall number is not very informative, but rather an age, sex, and populationspecific breakout is necessary. For a coherent interpretation, the IFR need to be considered as well. IFR are displayed in Table 6 , along with supporting quantities (population sizes, estimated number of cases and sero-prevalences) in Tables 4 and 5 . We use in the numerators not only (lab) confirmed cases, but also possible cases. This would create difficulty for CFR estimation (as the denominator would be lab-confirmed cases only), but is a sensible choice for IFR estimation, especially because confirmed plus possible COVID-19 deaths nearly coincide with the excess death rates in Belgium ( Figure 4 ). This coincidence is not a proof for the fact that all excess deaths are COVID-19 related, although it has been reported internationally that around 90% of possible cases are proper COVID-19 [2] . While it may be possible, for example, that some excess deaths are related to other factors, such as lockdown-induced stress, the plausible assumption is made that this effect on mortality is minor. Further examination is warranted as soon as the cause-specific mortality database becomes available, typically after a three-year interval. We make no claims regarding nonmortality related effects. . 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 20, 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 20, 2020. . https://doi.org/10.1101/2020.06.20.20136234 doi: medRxiv preprint . 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 20, 2020. . https://doi.org/10.1101/2020.06.20.20136234 doi: medRxiv preprint . 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 20, 2020. . https://doi.org/10.1101/2020.06.20.20136234 doi: medRxiv preprint 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 20, 2020. . https://doi.org/10.1101/2020.06.20.20136234 doi: medRxiv preprint 19_Daily%20report_20200526%20-%20FR.pdf), 0.70 reflects test sensitivity and 3 is an ad-hoc scale factor. Of note, bias and precision in IFR estimates are influenced by issues surrounding the estimation of sero-prevalence, such as sensitivity and specificity of the tests used [3] , time to IgG seroconversion [4, 11, 12] detectability [11] and clearance [12] , and potential selection bias occurring in data from residual sample surveys. The overall values are re-estimated and reported in Table 7 Table 6 , for the entire population, as well as for the non-NHP. . 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 20, 2020. . https://doi.org/10.1101/2020.06.20.20136234 doi: medRxiv preprint Based on the various sero-prevalence estimates, IFRs across all ages are estimated at 0.38 -0.73% for males and 0.20 -0.39% for females in the non-nursing home population (non-NHP), and at 0.79 -1.52% for males and 0.88 -1.31% for females in the entire population. Estimates for the NHP range from 41 to 73% for males and over 24 to 37% for females. The IFRs rise from nearly 0% under 45 years, to 4.3% and 13.2% at 85+ for males in the non-NHP and the general population, respectively, and to 1.5% and 11.1% for females in the non-NHP and general population, respectively. In the NHP, the IFR may well be above 50% in males and about 30% in females of 85+. . 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 20, 2020. . https://doi.org/10.1101/2020.06.20.20136234 doi: medRxiv preprint The IFR is strongly influenced by extensive death cases reporting and the fact that 66.0% of the deaths concerned NH residents. Apart from a strong age-related gradient, also for each age category, IFRs are substantially higher in males than in females. The average, obtained from pooled figures, shows the reverse ordering in the overall population (1.2% in males and 1.3% in females), although the difference is nonsignificant. In the non-NHP, the effect is reversed (0.6% in males, 0.3% in females). This phenomenon, known as Simpson's paradox, stems from the fact that age-specific IFRs are lower in females, but the age-specific population size in older age groups is larger in females (e.g., 200,000 females versus 100,000 males in 85+). Like with DPM, also the IFR strongly differs between the NHP and the non-NHP. A cautionary remark is in place. The NHP is relatively small and it is exactly in this population that age and sex suffers from incompletes. The more reliable IFR estimates, therefore, are for the overall and non-NHP populations, because of stable denominators, but for the NHP, the age and/or sex specific values should be interpreted qualitatively, in terms of ranges, only. In the NHP the overall IFR climbs to 28 -45%. The difference between male and female IFR for 85+ (to a lesser extent also in 75-84) appears to be more pronounced in the non-NHP than in the general population. This is another instance of Simpson's paradox: At 85+, the female population is twice in size the male population, but in nursing homes, there are four times as many females than males. Thus, the general population at 85+ is mixed quite differently over NH and non-NH settings, with the NHP the more vulnerable fraction. . 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 20, 2020. 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 20, 2020. . https://doi.org/10.1101/2020.06.20.20136234 doi: medRxiv preprint not show in the overall sero-prevalence, but may account for large differences in DPM, is the age distribution of a country, because of the steep age gradient in IFR, but also in the DPM itself. A relatively older population will lead to considerably more deaths, for constant IFR structure. As reported by Wyper et al. [1] , in Belgium 13.1% of the population is 70+ (the European continent extremes being Israel for 7.7% and Italy for 16.4%), among whom further 42.2% are 80+ (with extremes 31.5% in Czechia and 44.5% in France). Larger elderly fractions may have various demographic reasons. To some extent, one of them is that countries with high-quality health care facilities are accommodating to a somewhat more frail elderly population, with underlying comorbidities such as high blood pressure and diabetes that are known to be risk factors for COVID-19 mortality. This point is partially underscores by the higher DPM and IFR in the NHP, for given age and sex. (7) Directly related to this, but worth separate mention, is that the epidemic has been very severe in the NHP, which shows in both the DPM and IFR, suggesting a nuanced explanation. While figures should be interpreted with caution, the IFR in the 85+ NHP appears to be roughly 15 times that in the non-NHP, pointing to increased frailty and higher prevalence of underlying comorbidities. Also, the effect of vectors, such as caregivers, should not be underestimated and protection and preventive measures taken in view of possible future outbreaks. In summary, the very large DPM in the NHP versus the non-NHP, when compared within a given age and sex group, arguably results from a larger sero-prevalence, in combination with an increased IFR. All in all, the outbreak in Belgian nursing homes was extremely serious, in line with international findings [13] . A more detailed study and further international comparison is urgent, as well as the implementation of targeted non-pharmaceutical interventions, while awaiting promising pharmaceutical development. For the general . 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 20, 2020. . https://doi.org/10.1101/2020.06.20.20136234 doi: medRxiv preprint population, the steep age-related gradient in mortality, expressed in IFR or number of deaths per million, contributes useful information to policymakers for differential non-pharmaceutical interventions. It is difficult to compare COVID-19 figures to these of countries that have a less extensive reporting strategy, in particular when the gap between excess deaths and COVID-19 mortality is large, such as in the Netherlands, Italy, or Austria [14] . Excess mortality across countries is a better base for comparison. This is definitely true in view of the very strong difference in DPM for the nursing home population as opposed to the general population, even though the figures are generally very high for the older ages groups. Countries that underreport deaths in nursing homes are therefore not a basis for comparison. Arguably, excess mortality is a better basis for comparison. Currently, EuroMOMO allows for this by means of the Z-score, a useful metric that indicates how unusual mortality is over a given period, relative to average mortality in that same period. It does allow for within-country comparisons only. For example, countries with less extreme variations in the reference period will have a smaller variance and for the same deviation in the epidemic period a larger Z, compared to a country with more variation in the reference period. It may thus be useful to supplement it with other metrics, such as the excess mortality rate, relative to reference mortality. For instance, in the second week of April 2020, mortality is about double the average over 2009 - Because of its inherent limitations, especially dependence on testing strategy, CFR is a flawed metric for international comparisons [7, 15] , and should be used with extreme caution. When compared, the delay distribution between confirmation and death should be taken into account [16] , also to accommodate under-reporting of cases. . 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 20, 2020. . https://doi.org/10.1101/2020.06.20.20136234 doi: medRxiv preprint The IFR is preferred, even though care should be taken when estimating seroprevalence, and its sources for bias and uncertainty quantified. This suggests the use of statistical sensitivity analysis, along with the reporting of interval estimates. COVID-19-related mortality figures suggest the seriousness of the epidemic. Belgium's April 2020 mortality was the highest among all months of April since WWII. In the week of April 5, 2020, COVID-19 mortality was twice as high as longterm average-mortality for that week. Summary of limitations. The large fraction of missing age and sex data adds uncertainty to age and sex specific estimates, especially in those pertaining to the NHP. Should more complete data become available, future adjustments will be possible. The determination of sero-prevalence is naturally surrounded with uncertainty, especially in the NHP, for reasons related to survey sampling and at this point unfolding virological knowledge. We are grateful for the ability to use open data on COVID-19 mortality and causespecific mortality (Sciensano, Belgium), general mortality and population figures (Statistics Belgium, Demobel; National Register). The data providers hold no responsibility for the analyses reported in this manuscript. We thank Sciensano colleagues Sophie Quoilin, Katrien Tersago, Dominique Van Beckhoven , Nina Van Goethem, and others, for suggesting relevant data sources from among their publicly available data, for useful comments and critical reflections on the analysis strategy, and for comments on earlier drafts of the manuscript. Particular thanks go out to Sciensano colleagues Sara Dequeker and Eline Vandael from the Nursing Homes . 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 20, 2020. The sero-prevalence study of which the results are used in this manuscript has been sponsored by the University of Antwerp's Research Fund. Other declarations: The investigators were independent from the funders; all authors had full access to the data and can take responsibility for the integrity of the data and the accuracy of the data analysis; the lead author affirms that the manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted. Data sharing: data and software code used for the tabular and graphical displays in this study are publicly available from https://www.uhasselt.be/DSI]. This works reflects the views of the authors and not necessarily the official position of the institutions they belong to. . 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 20, 2020. . https://doi.org/10.1101/2020.06.20.20136234 doi: medRxiv preprint . 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 20, 2020. . https://doi.org/10.1101/2020.06.20.20136234 doi: medRxiv preprint Population vulnerability to COVID-19 in Europe: a burden of disease analysis Estimating the global infection fatality rate of COVID-19 Seroprevalence of IgG antibodies against SARS coronavirus 2 in Belgium: A prospective cross-sectional study of residual samples Quantifying antibody kinetics and RNA shedding during early-phase SARS-CoV-2 infection All-cause mortality supports the COVID-19 mortality figures in Belgium Missing Data in Clinical Studies Cross-country comparison of case fatality rates of COVID-19/SARS-CoV-2. Osong Public Health and Research Perspective Mortality in Belgium from nineteenth century to today Estimating the infection and case fatality ratio for coronavirus disease (COVID-19) using age-adjusted data from the outbreak on the Diamond Princess cruise ship Using a delay-adjusted case fatality ratio to estimate under-reporting SARS-CoV-2 serological analysis of COVID-19 hospitalized patients, pauci-symptomatic individuals and blood donors A systematic review of antibody mediated immunity to coronaviruses: antibody kinetics, correlates of protection, and association of antibody responses with severity of disease High impact of COVID-19 in long-term care facilities, suggestion for monitoring in the EU/EEA