Officially, COVID-19 deaths are now more than 5.6 million. However, the actual number of deaths occurring globally is estimated to be two-to-four times the official figures. In specific country cases, this estimate may even be higher.
Many country-level studies and investigations by epidemiologists, other researchers and journalists have shown that the official figures underestimate the death numbers. Reviewing some of these prominent investigations, journalist and author David Adam, in his piece “The Effort to Count the Pandemic’s Global Death Toll” (Nature, January 18, 2022) discusses some these efforts across the globe.
Various methodologies -- satellite images of cemeteries, door-to-door surveys and machine-learning computer models -- have been used to extrapolate global estimates from available data. Although each of these methodologies has limitations of their own, the one unambiguous conclusion that most of the studies arrive at is that the official deaths are only a fraction of the deaths that are estimated.
The mistrust over official data and an attempt to arrive at better estimates is prompted by various factors. These include not just plain observation of divergence between the real time deaths happening all around and the official numbers that do not seem to capture these. Other factors, as Adam points out, include varying approaches to recording COVID deaths. For instance, while the Netherlands counted only those individuals who died in hospital after testing positive for the SARS-CoV-2 virus, Belgium included deaths in the community and everyone who died after showing symptoms of the disease, even if they were not diagnosed.
Calculating Excess Mortality
Counting excess mortality is one of the commonest and relatively easier techniques that the researchers and journalists have used to arrive at a closest-to-correct number of COVID mortality. Excess mortality is a metric that involves comparing all deaths recorded with those expected to occur. It is calculated by comparing deaths during the pandemic with the average recorded over the previous five years or so. The excess mortality estimates that have come up over the past year show many more people than the officially recorded have died in the pandemic.
While measuring excess deaths this way, demographer Tom Wilson, at the University of Melbourne, Australia and Felix zur Nieden, at Germany’s statistical office, says that raw numbers should be adjusted to take age structure and other subtleties into account.
World Mortality Dataset (WMD)
Accounting for such biases, Ariel Karlinsky, an economist at the Hebrew University of Jerusalem, Israel, and Dmitry Kobak, a data scientist at the University of Tübingen, Germany have worked out more sophisticated and comprehensive estimates of excess mortality. Covering about 116 countries, they have developed the World Mortality Dataset (WMD) using official death statistics collected and published by national offices and governments and estimated excess mortality taking into account factors such as death tolls associated with armed conflict, natural disasters and heatwaves.
According to their estimates, in 2021, cumulative COVID-19 deaths were around 6.5 million, or 1.6 times more than the then official figure of 4.1 million deaths; in some countries, the relative impact of the virus is even higher. The WMD, however, does not report excess-death estimates for more than 100 countries because those countries either do not collect death statistics or do not publish them timely.
Modelling Global Deaths
In an attempt to overcome the biases and shortcomings of these exercises, some researchers have relied on modelling techniques to arrive at more precise estimates.
The Economist magazine has used a machine-learning approach to produce an estimate of 12 million to 22 million excess deaths — or between two and four times the pandemic’s official toll so far.
This approach identifies more than 100 national indicators -- official deaths, the scale of COVID-19 testing, results of antibody surveys, geographical latitude, degree of internet censorship, number of years a country has been a democracy, etc. – that seem to correlate with excess deaths in more than 80 countries where data are available.
Plugging numbers for these indicators for a country that does not produce mortality data into the model, algorithms estimate that country’s excess deaths, though some researchers argue that it is erroneous to apply a model based on rich countries with high life expectancy to poor countries with low life expectancy.
Overall, the model suggests that lower middle-income countries have suffered at least as severely in per-capita deaths as rich countries — in contrast to the picture given by official figures, despite the fact that these poorer countries have younger populations.
The Institute for Health Metrics and Evaluation (IHME) in Seattle, Washington, offers daily updates of modelled results and projections of how quickly the global toll might rise. The IHME’s modelling puts the death-toll between 9 million and 18 million.
In all these different approaches/models although the global estimates are in congruence, some of their country estimates differ by a wide margin.
The World Health Organisation (WHO) is working on arriving at estimates in countries that do not have reliable national mortality data. According to researchers, this can be done by extrapolating from smaller regions in a country, where limited data might be available.
Another method, being used in a few countries, is to survey a representative sample of households to ask them about deaths. Some of these surveys have shown that excess mortality is several times higher than official COVID deaths. A study in the University of Toronto, Canada, used the results of a telephone survey in India to find more than 3 million COVID-19 deaths up to July 2021 – much higher than the official estimates.
A more direct method used to estimate deaths is counting graves. A study at the London School of Hygiene and Tropical Medicine (LSHTM), used satellite images of 11 cemeteries in Aden province in Yemen which suggested that the estimated excess deaths for the region were 2,120 during the same period. Similarly, in Mogadishu, Somalia, estimated excess deaths were 3,200 to 11,800 between January and September 2020.
These numerous efforts all point out the fact that the official figures are not even close to the actual numbers of deaths. It is a sad state that despite technological progress such simple data as real time mortality figures are not readily available which may adversely affect public health policy measures.