Zum Hauptinhalt springen
14.12.2022

Global excess mortality due to COVID-19

     

  • updated estimation of WHO on excess mortality due to COVID-19
  •  

  • according to this 122.000 additional deaths in Germany
  •  

  • Methodology very rough for some countries, criticism on earlier results
  •  

An updated estimate of the World Health Organization (WHO) on global excess mortality caused by COVID-19 (see primary source) was published in the scientific journal "Nature" on 14th of December 2022. According to this estimate, about 14.8 million additional deaths occurred worldwide in 2020 and 2021. The estimate is nearly three times the officially reported 5.4 million COVID-19 deaths during that time. According to the researchers, the average global per capita mortality rate in 2020 was 0.06 percent, rising to 0.13 percent in 2021. This surpasses the influenza pandemics of 1957 (0.04 percent), 1968 (0.03) and 2009 (0.005). According to the WHO statistics, around 122.000 more people died in Germany than would have been expected. For almost half of all countries, it is not possible to state excess mortality with a high degree of certainty due to weak data.

The WHO statistics now published in "Nature" are an update of the results that were already published in May this year. At that time, the WHO's sometimes very complex methodology was criticized. The method was quite flexible, but also easily influenced by short-term random fluctuations [I]. This led to inaccuracies in Germany, for example: in 2018, there were an unusually large number of deaths in this country due to a strong wave of influenza, whereas in 2019 the figure was rather low. This circumstance was not sufficiently taken into account by the WHO method, critics said. This had led to implausible estimates of excess mortality for subsequent years. The WHO reacted quickly and published updated data in a preliminary preprint. This has now been published as a scientific peer-review in “Nature”.

In response to the criticism against the WHO publication in May, the SMC has asked statisticians to review the new estimate and to evaluate its validity globally and with regard to Germany.

Overview

     

  • Prof. Dr. Hanno Ulmer, Director of the Institute for Medical Statistics and Informatics, Medical University of Innsbruck, Austria
  •  

  • Jonas Schöley, Ph.D., Research Scientist in the Laboratory of Population Health, Max-Planck-Institute for demographic research, Rostock, Germany
  •  

  • Prof. Dr. Christoph Rothe, Professor of Statistics, University of Mannheim, Germany
  •  

Statements

Prof. Dr. Hanno Ulmer

Director of the Institute for Medical Statistics and Informatics, Medical University of Innsbruck, Austria

„The choice of statistical methods should be appropriate. The concrete calculations and results depend on the data available in the individual countries. A global estimation like this is problematic, as various limitations apply. The authors address some of these limitations themselves (page six, right column).“

„The WHO's desire to be able to classify the COVID-19 pandemic with a few figures on excess mortality is understandable. Ultimately, however, this is a very rough estimate that may not be accurate for individual countries. It seems necessary to shed light on the situation in individual countries separately so that excess mortality due to COVID-19 can be properly assessed.“

„Peru, for instance, has by far the highest excess mortality worldwide according to the study. However, there were apparently also strong dengue fever outbreaks during the pandemic years in Pure. High excess mortality can also happen if the reference years before the pandemic were particularly good years, that is, years without special outbreaks and with low mortality.“

„In the end, I wouldn't necessarily see this paper as an estimate of COVID-19 deaths, but as a paper on excess mortality in the pandemic years 2020 and 2021. A small but subtle difference that can vary greatly by country and region.“

Jonas Schöley, Ph.D.

Research Scientist in the Laboratory of Population Health, Max-Planck-Institute for demographic research, Rostock, Germany

„The WHO model is state of the art, and after the corrections it can reliably measure excess mortality in countries with well-developed demographic reporting systems - this includes Europe, North America, Australia, New Zealand and parts of Asia and South America. However, in large parts of Africa and parts of Asia and South America, death registration is very unreliable and therefore, excess mortality cannot be estimated from reported deaths. Instead, it is estimated indirectly from correlations between excess mortality and available variables such as gross domestic product. Here, the WHO model has the same limitations as alternative models from IHME (Institute for Health Metrics and Evaluation; editor's note) or Economist, namely that without knowing the number of deaths in a country, as in much of Africa, only a very unreliable estimate can be made.“

„The WHO deals with these uncertainties in the estimation in a transparent way and publishes uncertainty intervals around the estimated excess mortality, also providing a ranking of different countries in terms of probabilities. The uncertainty is higher for small countries (for example Denmark) and for countries without direct information on the number of deaths. Beyond this statistical uncertainty, however, there is always a remaining modelling uncertainty, also in the case of the WHO model. Similar to climate models, there are various plausible models which produce slightly different numbers. Here it is important to focus on the trends that are reliably reflected by all models (for example, the higher excess mortality in Eastern Europe or the generally historically high excess mortality), rather than interpreting small differences (for example, whether excess mortality in Denmark was higher or lower than in Sweden). There are also recent studies addressing the question of model robustness [1] [2].“

When asked how to assess the quality of estimates of excess mortality in Germany and why a different calculation method was chosen for Germany (and Sweden):
„Different models were chosen for Germany and Sweden, as the original estimates of expected mortality since 2020 were implausible for both countries. These errors were recognised by various parties and promptly publicly admitted, explained, and then corrected by Jon Wakefield (who is a statistician at the University of Washington and lead the WHO modelling of excess mortality due to COVID-19 [3]). The error has technical reasons and was well analysed: The WHO uses a model which, when extrapolating trends from annual deaths, is occasionally too strongly guided by what happened in 2019 instead of focusing on medium-term trends as intended. In combination with unnecessary corrections for deaths reported with time-delays, this led to implausible results for Germany and Sweden. It is not unusual for problems to occur occasionally in large models. Here, the same holds as for complex construction or software projects - when errors inevitably occur, they must be communicated transparently and then corrected. This is what happened here.“

When asked how the excess mortality calculated for Germany compares to that of the Federal Statistical Office:
„In the attached article, around 122,000 (101,000-143,000) premature deaths are estimated for Germany for the years 2020/21. As the Federal Statistical Office has not published comparable statistics to my knowledge, I cannot make a comparison here (Destatis publishes monthly excess mortality as a percentage).“

„The Federal Statistical Office uses the median of the monthly death rates of the four preceding years as a reference for calculating the monthly excess mortality. This method was presumably chosen because it can be explained in one sentence, and is thus transparent. However, there are more accurate, albeit complicated, methods, for example that of the WHO, to calculate expected deaths in the absence of the COVID-19 pandemic. An important difference between the WHO method and the method of the Federal Statistical Office is that the WHO gives uncertainty intervals around the estimated excess mortality. Since the number of expected deaths in the absence of the pandemic can only be estimated and is not known with certainty, this also results in uncertainty in excess mortality compared to expected mortality. Reporting this uncertainty is useful to avoid drawing erroneous conclusions based on random fluctuations.“

„The excess mortality caused by COVID-19 is unique in its magnitude and global scale in the last 70 years. We can also see this from the strong declines in period life expectancy for 2020/21 [4]. Such strong declines as seen since February 2020 have not been observed since the world wars, at least for Western Europe and the US.“

„Already in the first wave in 2020, it could be clearly seen that COVID-19-related excess mortality in severely affected countries is much higher than during a severe influenza season. Belgium is just one example among many: in week 15 2020, excess mortality in Belgium peaked at 99 percent (88 to 110 percent; 90 percent uncertainty interval) and then peaked again at 79 percent (70; 91 percent) in week 45. These levels are far beyond what happens at the peak of a severe flu season (here we would expect 20 to 30 percent), despite containment measures. Studies also show very clearly that the effect of COVID-19 on excess mortality in 2020 was far greater than that of the 2015 severe flu season in most European countries [5]. The exceptionally high excess mortality was repeated during the winter wave of 2020/21 and during the delta wave in winter 2021/22, leading to drastic levels of excess mortality, especially in Eastern Europe.“

„In general, the COVID 19 pandemic exacerbated existing inequalities in population health. The Scandinavian countries (Denmark, Norway, Finland, Sweden), which have excellent health care and high life expectancy, experienced relatively low excess mortality over the years 2020/21 in international comparison. Eastern Europe, on the other hand, which had a lower life expectancy compared to the rest of Europe even before the pandemic, experienced persistently high excess mortality over the years 2020/21. Among other things, differences in vaccination rates – which are high in Scandinavian countries and low in Eastern Europe - also play a role in these different outcomes.“

„The risk of dying from COVID-19 is not a variable that remains constant over time and is substantially reduced by vaccination. Thus, in autumn 2022, we saw that COVID-19 was no longer mainly responsible for excess mortality recorded in Germany. Instead, classical infectious diseases were probably again playing a larger role in seasonal excess mortality. However, the medium and long-term effects of the COVID-19 pandemic on mortality cannot yet be estimated.“

Prof. Dr. Christoph Rothe

Professor of Statistics, University of Mannheim, Germany

„The Nature study describes the method used by the WHO to calculate global excess mortality during the COVID-19 pandemic in 2020 and 2021, and its results. The figures are essentially the same as those published by the WHO back in May, and in that sense are not new."

"Calculating a global excess mortality is complicated, particularly because for many countries, there is no reliable data on deaths before and during the pandemic. The authors therefore estimate these values using statistical methods and the data of comparable countries with a better information base. The resulting global excess mortality of about 14.8 million deaths in 2020 and 2021 is therefore subject to some uncertainty, but should be of the right order of magnitude."

"The study also provides estimates for excess mortality in individual countries. These values are based on a statistical procedure which, unfortunately, can be quite strongly influenced by random fluctuations in the number of deaths before the pandemic. For instance, this resulted in an implausibly high value of 195,000 additional deaths for Germany. The authors of the study therefore use a separate ad hoc calculation for Germany, which, at 122,000 additional deaths in 2020 and 2021, is nevertheless still well above the more reliable estimates of the Federal Statistical Office of about 70,000 cases, and that from studies by De Nicola, Kauermann and Höhle of about 35,000 cases [6] [7].“

Declared interests

Jonas Schöley, Ph.D.: “Jon Wakefield and I were in contact last year regarding the WHO methodology. Mr Wakefield asked me for an assessment of part of the paper before the first figures were published. However, I am not a co-author of the study or otherwise involved in the publication of the figures. After the figures were published, I criticised the results for Germany and Mr Wakefield and I were in contact regarding the statistical reasons for the problem.”

Others: None received.

Primary reference

Msemburi W et al. (2022): The WHO estimates of excess mortality associated with the COVID-19 pandemic. Nature. DOI: 10.1038/s41586-022-05522-2.

References

[1] Schöley J (2021): Robustness and bias of European excess death estimates in 2020 under varying model specifications. Medrxiv. DOI: 10.1101/2021.06.04.21258353.
Editor's note: This is a pre-publication that has not yet been peer reviewed by independent experts.

[2] Nepomuceno MR et al. (2022): Sensitivity Analysis of Excess Mortality due to the COVID-19 Pandemic. DOI: 10.1111/padr.12475.

[3] Van Noorden (01.06.2022): COVID death tolls: scientists acknowledge errors in WHO estimates. Nature. News Feature.

[4] Schöley J et al. (2022): Life expectancy changes since COVID-19. Nature. DOI: 10.1038/s41562-022-01450-3.

[5] Islam N et al. (2021): Effects of covid-19 pandemic on life expectancy and premature mortality in 2020: time series analysis in 37 countries. BMJ. DOI: 10.1136/bmj-2021-066768.

[6] De Nicola G et al. (2022): On assessing excess mortality in Germany during the COVID-19 pandemic. AStA Wirtschafts- und Sozialistisches Archiv. DOI: 10.1007/s11943-021-00297-w.

[7] De Nicola G et al. (2022): An update on excess mortality in the second year of the COVID-19 pandemic in Germany. AStA Wirtschafts- und Sozialistisches Archiv. DOI: 10.1007/s11943-022-00303-9.

References, cited by SMC

[I] Rothe C (31.05.2022): WHO-Studie zur Corona-Übersterblichkeit nutzt störanfällige Methode. Leibniz-Institut für Wirtschaftsforschung. Press release.