Ariel Karlinsky, who has been tucked away in his Tel Aviv apartment for the past 18 months, has searched the internet for data that could aid him in calculating the true death toll of Covid-19.
Although the 31-year-old student of economics at Hebrew University of Jerusalem was not familiar with health issues, he was disturbed by reports that Israel was not experiencing an increase in deaths.
He said that this was false. He said that excess mortality was evident and was clearly visible.
Other rumours arose. Another was that countries with minimal or no containment measures (e.g. Russia) were not suffering from significant excess mortality. It was false again, but it was more difficult to get the data.
Karlinsky realized that this was true for many countries. Even countries that regularly collected excess-mortality data did not often publish it until at most a year later. This meant that they didn't know of a sensitive indicator of pandemics scale or progress that could help them respond.
It was a challenge to collect the data in real-time for as many countries as possible.
He met Dmitry Kobak, a data scientist at the University of Tbingen, Germany, through Twitter. They agreed to work together. Karlinsky was looking for the numbers; Kobak did the analysis.
The World Mortality Dataset is the result. It forms the basis for estimates of Covid mortality published by the Economist and the Financial Times, and gives credibility to the official global death count of 4.8 million. For example, the Economist puts the actual number closer to 16 millions.
Karlinsky's and Kobaks efforts have been praised by those who assess the effects of public health catastrophes. This data revolution is similar to that in vaccine development and pathogen sequence, according to Lone Simonsen (Danish epidemiologist, Roskilde University) and Ccile Viboud (US National Institutes of Health).
There are many ways to measure the death toll from pandemics. Each method has its advantages and disadvantages. Although the official number is calculated from national Covid deaths reports, these are dependent on testing rates and almost always underestimated.
Moroccan Muslims worship at Casablanca's Hasan II mosque. It is one of the largest mosques in Africa. They observe Covid restrictions when praying. Photograph: Fadel Senna/AFP/Getty Images
According to Sondre Ulvund, a data journalist who heads the Economists pandemic tracking initiative, the official Covid death tolls for large groups of countries are not credible.
The excess mortality is defined as an increase in deaths due to all causes that exceeds the expected level based on historical trends. It does not depend upon testing rates. This is an old tool. It was used to estimate death tolls from historical pandemics, especially in cases where there was no test for the disease. However, it has been calculated retrospectively until now.
Karlinsky's and Kobaks innovative idea is to collect data and publish it during a pandemic. They used established statistical techniques to fill the gaps.
The downside to excess mortality is its composite nature. It includes Covid deaths as well as deaths indirectly related to the pandemic such cancer patients who couldn't get timely treatment or domestic abuse victims locked down during lockdowns. However, it doesn't give much information about each individual contributor.
Karlinsky & Kobak found that excess mortality in Covid is primarily due to deaths caused by the disease.
There are some unexpected results when calculating excess mortality. They reported in June in the journal eLife, that excess mortality was negative in Australia, South Korea, and Finland. This is because their pandemic control has been outstanding and they have almost eliminated flu by 2020. According to Simonsen & Viboud, these cases are an indicator of pandemics' tolls better than official Covid deaths.
The World Mortality Dataset provides information about more than 100 countries. Most African and Asian countries are missing, as well as many of the most populous in the world. They also include some of those most affected by Covid death. India is an example of this. While it does not release its national vital data routinely, some researchers believe that the Covid death toll could reach 4 million.
At a hospital in Abuja Nigeria, a health official receives the first dose of Covid vaccine. Photograph: Afolabi Sotunde/Reuters
Karlinsky, Kobak have retrieved subnational data from these data-poor nations or been provided them by journalists and academics. They then applied different techniques of extrapolation in order to produce national estimates.
They can also project from neighboring countries, where data is available.
Karlinsky and Kobak are not able to estimate the global death toll because of uncertainty in the data. However, they claim that national excess deaths are 1.4x higher than the reported Covid deaths. This would mean that a rough global count of 6.7 million could be derived.
Solstads modeling put the number between 9.9 and 18.5 millions, which Simonsen considered reasonable.
Viboud and she used excess-mortality figures from previous pandemics to put the numbers in historical context. They then adjusted these for 2020.
If they had occurred now, the death tolls from the four previous flu pandemics would have been 75 million (1918), 3 million (1957), 2 million (1968), and 0.4 million (2009).
They conclude that Covid is the most deadly pandemic in a century but it has not been anywhere near the death toll as 1918's pandemic.
This new data shows that countries that were criticized for suffering severe outbreaks (e.g. Spain, Italy) have actually not been the most affected.
Mexico, Bolivia and some eastern European countries are the worst affected. They have seen more than 50% increases in mortality. Peru is the worst affected country, with a 150% rise in mortality.
Because some data is trickling in with a delay, the dataset becomes more accurate over time. Some countries requested that their national statistical offices accelerate the collection of vital data. Others could not or refused to release it. Turkey was expected early in the summer to release its monthly vital data for 2020. It has not.
Istanbul market. Turkey has yet to release vital statistics for 2020. Photograph: Dilara Senkaya/Reuters
Karlinsky stated that Turkey is an example of a country where they have the numbers, but they don't release them because they don't want to explain the discrepancies.
He said that excess mortality could be a sidelight of government transparency. It is possible that the country did not have the necessary testing or vital registration capabilities to produce official Covid deaths that were lower than excess deaths.
Solstad believes excess mortality should continue to be monitored in the future. This would give better insight into all types of crises including wars and famines. He said that it is a fairly objective indicator of what is going wrong. Karlinsky is in agreement. Karlinsky agrees. In 2015, when Egypt was hit by a heatwave, the state media reported that 61 people died. His estimate was closer to 20,000.
Some countries might not want to do this. The World Health Organization made the first step in harnessing excess mortality to be a surveillance tool when it established an expert committee to evaluate Covid mortality.
If they are aware of a crisis coming, governments could respond more quickly and proportionally. A better ability to convince the public would be a plus. Simonsen stated that some people believe that the virus would not have been stopped if they hadn't done something. She said that the World Mortality Dataset revealed that a lot had happened in many countries.