Checchi, F., & Roberts, L. (2005). Interpreting and using mortality data in humanitarian emergencies. Humanitarian Practice Network, 52.
For example, because no COVID-19 test was conducted or a country's death reporting system failed to register the death as from COVID.
Conditions such as health systems being overwhelmed, resources being diverted away from other health problems, or fewer people seeking treatment for other health risks.
A working paper with data from England estimates that for every 30 COVID deaths there is at least one avoidable non-COVID excess death in hospitals. See Fetzer and Rauh (2022) Pandemic Pressures and Public Health Care: Evidence from England.
We use this estimate as of 20 September 2021.
Or as many years from this period as are available.
Before 18 January 2022, WMD published a projection only for 2020 to avoid further extrapolation from the historical data. As of 18 January, they do publish-and we use-separate projections for 2020, 2021, and 2022.
Karlinsky, A. and Kobak, D. (2021). Tracking excess mortality across countries during the COVID-19 pandemic with the World Mortality Dataset. eLife, DOI: 10.7554/eLife.69336.
Except for a few countries for which we only have data from the years 2016 or 2017 to 2019; for details see this spreadsheet and the Human Mortality Database metadata.
For instance, for countries that have an increasing trend in mortality like the US and South Korea, the five-year average will overestimate excess deaths; while for countries that have a decreasing trend in mortality like Russia, it will underestimate excess deaths.
We calculate P-scores using the reported deaths data for 2020-2022 from HMD and WMD - see here for country by country source information - and the projected deaths for 2020-2022 from WMD (which we use for all countries and regions, including for deaths broken down by age group).
See this spreadsheet for the UN-estimated death registration coverage of the countries in our dataset. Despite the estimates, the actual coverage might be lower due to the burden of the pandemic. For analysis of this under-reporting see the recent paper by Whittaker et al (2021) Under-reporting of deaths limits our understanding of true burden of covid-19.
For instance, a 2016 CDC study of the delay in the US found that after four weeks, only 54% of deaths had been registered; by eight weeks the figure was 75%, and it didn't reach 100% until almost a year after the date of death. Though the CDC does note that "Data timeliness has improved in recent years, and current timeliness is likely higher than published rates." In fact the CDC currently estimates that "63% of all U.S. deaths are reported within 10 days of the date of death, but there is significant variation between states."
Clearly incomplete data is marked by a large, abrupt drop in the death count - often well below the five-year average - and a pattern of substantial upward revision to the count from recent periods. For a detailed list of the data we exclude for each country see this spreadsheet.
In 2020, for instance, there were an estimated ~2500 such deaths; see here for details.
Before 27 September 2021, we did not include these deaths in Sweden's data series because our source at the time, the Human Mortality Database, did not include them.
As of 27 September 2021, however, we do include these deaths with unknown date - we switched sources for Sweden to the World Mortality Dataset (WMD), which does include these deaths. See here for how WMD does this: https://github.com/akarlinsky/world_mortality#sweden-weekly.
For instance, England & Wales define the week as from Saturday to Friday.
This enables easy comparisons of weekly deaths across the years in the chart, but it means we show a date that is slightly incorrect (plus or minus a few days) for the other years. This is because the same numbered week falls on slightly different dates in different years; for example, Week 1 2020 ended on 5 January 2020, while Week 1 2021 ended on 10 January 2021. For more details see this resource with ISO 8601 week dates across the years.
The reason for this is that the monthly data smooth the weekly fluctuations, resulting in lower estimates. Source: D. Jdanov, Human Mortality Database, personal communication, 11 February 2021.
For example, compared to the death rate from COVID-19 for ages 5-17, the death rate for ages 65-74 is 1100 times higher, for ages 75-84 it is 2800 times higher, and for ages 85+ it is 7900 times higher. These estimates are based on US data from the CDC: COVID-19 Hospitalization and Death by Age.
These baselines are produced according to the same method described in Karlinsky and Kobak 2021.
To read The Economist's article presenting the model estimates, see: https://www.economist.com/graphic-detail/coronavirus-excess-deaths-estimates
To read about the model methodology, see: https://www.economist.com/graphic-detail/2021/05/13/how-we-estimated-the-true-death-toll-of-the-pandemic
Similar results were reported in March 2022 by researchers from the Institute for Health Metrics and Evaluation, publishing in the journal The Lancet.
They found that "Although reported COVID-19 deaths between Jan 1, 2020, and Dec 31, 2021, totalled 5.94 million worldwide, we estimate that 18.2 million (95% uncertainty interval 17.1-19.6) people died worldwide because of the COVID-19 pandemic (as measured by excess mortality) over that period."
The full citation is: Wang, H., et al. (2022). Estimating excess mortality due to the COVID-19 pandemic: a systematic analysis of COVID-19-related mortality, 2020-21. The Lancet.
HMD only includes countries with the highest quality and most comprehensive mortality statistics, with breakdowns by age and sex and many years of historical data. Because of this, only select countries with very robust and capable statistical agencies are included.
Németh L., Jdanov D.A., Shkolnikov V.M. (2021) An open-sourced, web-based application to analyze weekly excess mortality based on the Short-term Mortality Fluctuations data series. PLoS ONE 16(2): e0246663. https://doi.org/10.1371/journal.pone.0246663
This criterion was changed from four years to three years of historical data on 30 May 2021.
The full list of excluded countries and reasons for exclusion can be found in this spreadsheet. It is possible we will amend this list on the basis of new information.
Though WMD does provide the projected baselines used for calculating P-scores in the by-age chart.
Before 18 January 2022, WMD published a projection for 2020 only to avoid further extrapolation from the historical data. As of 18 January, they do publish-and we use-separate projections for 2020, 2021, and 2022.
Karlinsky, A. and Kobak, D. (2021). Tracking excess mortality across countries during the COVID-19 pandemic with the World Mortality Dataset. eLife, DOI: 10.7554/eLife.69336.
2 May 2022: Switched source back to Human Mortality Database as they now include total deaths, not just doctor-certified deaths.
Had previously (on 18 January 2022) switched source from Human Mortality Database to World Mortality Dataset to get an estimate of total deaths and not just the doctor-certified deaths that HMD publishes. For more details see: https://github.com/akarlinsky/world_mortality#australia-weekly.
27 September 2021: Switched source from Human Mortality Database to World Mortality Dataset to account for deaths with unknown date.
A working paper with data from England estimates that for every 30 COVID deaths there is at least one avoidable non-COVID excess death in hospitals. See Fetzer and Rauh (2022) Pandemic Pressures and Public Health Care: Evidence from England.
All visualizations, data, and code produced by Our World in Data are completely open access under the Creative Commons BY license. You have the permission to use, distribute, and reproduce these in any medium, provided the source and authors are credited.
The data produced by third parties and made available by Our World in Data is subject to the license terms from the original third-party authors. We will always indicate the original source of the data in our documentation, so you should always check the license of any such third-party data before use and redistribution.
Our articles and data visualizations rely on work from many different people and organizations. When citing this entry, please also cite the underlying data sources. This entry can be cited as: