Demographers have been thrust to the frontlines of the world’s efforts to evaluate the coronavirus pandemic, but so far without any weapons. Lacking data, they cannot reliably assess the situation. And this is despite the fact that the internet, it would seem, is flush with statistics. A webinar hosted by the HSE International Laboratory for Population and Health discussed the paradoxes of quantitative approaches to COVID-19. IQ.HSE spoke with webinar participants Vladimir Shkolnikov, Inna Danilova, and Dmitry Jdanov.
According to graphs and data available online, it is clear that countries are very different both in terms of the number of confirmed cases and the number of deaths. The Case Fatality Ratio (CFR), that is, the ratio of deaths to confirmed cases of the disease, is different even in countries with commensurate levels of economic development in which the epidemic began at about the same time. For example, in Germany the CFR is about 2%, while in France it is more than 10%.
In addition, CFR calculation methods have been changing, and this affects the recorded mortality numbers.When the CFR sharply increased in France in early April, it was simply because scientists began including deaths that had occurred in nursing homes. Prior to that change, the statistics had only included mortality data collected from hospitals.
‘Despite the availability of data, it is not yet clear what it includes, and how comparable statistics are between countries,’ says Inna Danilova.
In these conditions, it is important for demographers to consider three parameters:
the extent to which the disease has spread: how much of the population is sick and how that will change;
mortality rate: what proportion of patients die, and how the rate differs by sex and age;
the disease’s effect on overall mortality rate: the increase of additional deaths during the epidemic and their impact on overall mortality trends and life expectancy.
These questions cannot be answered without understanding how COVID-19 is diagnosed and recorded as a cause of death.
The numbers of confirmed cases of infection in most countries are in accordance with the definition put forth by the World Health Organization (WHO). A confirmed case is a person with laboratory-recorded case of COVID-19, regardless of clinical manifestations or symptoms.
Countries have varying levels of test accessibility, however, so this causes discrepancies in published statistics.
The virus was first identified a little over three months ago. During this time, tests had to be developed, approved, and produced, and laboratories had to be prepared. Some countries, such as South Korea, Singapore, and Australia, managed to do this in the early stages of the virus’s spread—already in January, wide-scale testing was being conducted in these countries. In other countries, the USA in particular, there were problems with making tests available.
A huge gap in the scale and quality of testing for COVID-19 exists between countries to this day.
In March, the WHO reported a global shortage of reagents due to the large number of tests being conducted. In this case, the need to introduce diagnostic criteria is understandable. And if some places do have diagnostic criteria, while other places continue to test all suspected cases, the ratios of detected to undetected cases will be different.
Currently, testing is mainly being conducted based on criteria and decisions determined by doctors. In some countries, such as Russia or Iceland, anyone who wishes can be tested for the virus. Thanks to this practice, Iceland leads in the number of tests performed per capita: as of April 7, there were about 30,000 tests performed for a population of 364,000 people.
Published statistics regarding infection rates are also affected by the scale of testing, which is different in different countries and changes over time.
In Germany, on the website of the Robert Koch Federal Institute, recommendations for which cases to send for analysis were updated several times between January and March. For comparison:
In the January 21 instructions, diagnostic criteria for COVID-19 included confirmed pneumonia, travel to the epidemic zone (specifically, the city Wuhan and the Hubei province in China), acute respiratory symptoms, and contact with the coronavirus carrier;
in the instructions on March 24, diagnostic criteria were narrowed to include only those who are at risk; only medical and nursing home personnel who show symptoms can be tested. Others may be tested only if there is sufficient testing available.
In accordance with the criteria, it is mainly patients with accompanying illnesses and chronic conditions who are tested, and these are generally elderly people. Their infection is more severe, and they more have to see a doctor.
Perhaps this leads to the fact that the ratio of detected to undetected cases differs in age.
For example, in Italy, which has the highest infection rate and mortality rate in Europe, it is those with serious symptoms who are more often tested. There are more cases of infection recorded amongst the older generations there. In countries where testing is widespread and the criteria are more flexible (such as in Iceland, South Korea, and Australia), the demographics of infection correspond approximately to those of the entire population.
Mortality rates also depend on diagnostic coverage. But how are coronavirus deaths classified?
The WHO has introduced two new causes of death into the international classification of diseases:
UO7.1 COVID-19. The virus has been identified: lab testing has confirmed the presence of COVID-19, regardless of the severity of the clinical signs or symptoms;
UO7.2 COVID-19. The virus has not been identified: COVID-19 was diagnosed clinically or epidemiologically, but laboratory tests were inconclusive or unavailable.
However, the approach to determining which deaths to include in coronavirus statistics varies among countries: those who died directly from infection or those who, in addition to the infection, had concomitant diseases or conditions.
In Italy, where this question was probably raised earlier than in other countries, all deaths with a diagnosis of COVID-19 are recognized as coronavirus fatalities. However, according to estimates by the Italian National Institute of Health, the virus is the direct culprit of only 12% of deaths classified as coronavirus deaths. In the remaining 88% of deaths, patients had an underlying condition, or, oftentimes, several underlying conditions.
‘Looking at the infection and mortality rates of COVID-19, we are only able to see the trajectory and speed of changes. We are unable to gauge the extent of the disease on a consistent scale,’ says Vladimir Shkolnikov.
Shkolnikov suggests the scale of losses from the epidemic can be estimated indirectly: not based on coronavirus data, but by analyzing mortality from all causes of death and looking at its fluctuations by months and weeks.
This method would consist of measuring excess mortality in every country — that is, the difference between the usual (expected) mortality rate at a certain point in time and the mortality rate during the epidemic.
The EuroMOMO project is using this method to study 24 EU countries and certain regions in within them. So far, the overall mortality rate in these countries is comparable to its seasonal peak in 2016-2017, even in Italy and Spain. But the excess mortality associated with COVID-19 is increasing, and no one knows the consequences.
The same goes for the current situation. According to Dmitry Jdanov, statistics as a whole was not ready for the challenges presented by the pandemic, and society remains unable to truly assess the danger.