"This rate does not come from the sky": how do we know that 17% of French adults have contracted the coronavirus?

One year after the start of the pandemic, less than one in five French people over the age of 20 and living in metropolitan France was infected with SARS-CoV-2.

17% precisely, but with immense disparities from one territory to another, and from one age group to another.

It is the Institut Pasteur, in collaboration with Inserm and Public Health France, which unveiled this estimate Wednesday as part of the update of its mathematical modeling work.

The health disaster has highlighted this kind of ultra-delicate exercise, especially when it comes to tracking down a phenomenon in real time.

But how do we actually achieve this type of quantified results?

What does he say about the evolution of the epidemic?

And what does it not allow to say?

Let's start with the calculation method.

It is obviously impossible to know the precise number of people infected cumulatively.

There have been 3.6 million positive tests in the past year, but for how many undetected cases?

The poor screening capacities during the first wave as well as the massive proportion of asymptomatics make testing an imperfect indicator for obtaining such a response.

Hospital admissions, a more reliable indicator

Researchers from the Institut Pasteur, led by Simon Cauchemez, therefore turned to hospital admissions and to a large seroprevalence study (SAPRIS) conducted last spring in Île-de-France and the Grand Est.

As a reminder, this type of survey consists in detecting a posteriori the presence or absence of antibodies in a large sample.

By analyzing admissions against these serological data, they were able to establish the risk for an infected individual to end up in hospital, depending on their age group.

By applying this ratio to all the hospitalizations recorded in the territory since the start of the pandemic, they mathematically deduced the total number of infections in France.

According to their calculations, we would have detected approximately “between 46.9 and 60.8%” of the total contaminations of adults which occurred between June and November 2020 in France.

We are also rediscovering to what extent the virus does not affect everyone and everywhere in the same way: 40% of Ile-de-France residents aged 30 to 39 have thus been infected against 4% of those over 70 in Corsica or New Aquitaine. .

A necessity in the face of "the absence of regular, large-scale and representative serological surveys"

These data must of course be shown the methodological limits.

“For example, we hypothesize that the probability of being hospitalized when one is infected has remained constant in each age group.

It is nevertheless possible that these probabilities have varied during the pandemic or from one region to another, ”indicates the Institut Pasteur on its website.

Advances in treatment can have an impact.

It was also necessary to take into account the imperfect sensitivity of the serological tests.

The youngest, who develop few serious forms, may also be under-represented in the event of an epidemic peak.

Because only the most at-risk cases then get a bed.

It is precisely to avoid this pitfall as much as possible that the models of the Institut Pasteur do not take into account those under 20.

In short, this method is therefore not perfect.

But it compensates at best "the absence of regular serological surveys, on a large scale and representative", argues the Pasteur Institute in a study being re-read by "peers".

This 17% rate does not surprise Samuel Alizon.

Research director at the CNRS in Montpellier, this specialist in the modeling of infectious diseases relies on other indicators (mortality and intensive care) but falls more or less on the same proportion.

"The number of people infected on February 20 was between 6.4 and 7.1 million people (excluding nursing homes)," he told Le Parisien, emphasizing the difficulty of this exercise in the face of the unknown represent the asymptomatic.

“This 17% rate does not come from the sky.

It does not come from a theoretical model disconnected from the data ”, abounds Pascal Crepey, teacher-researcher in epidemiology and biostatistics.

If it cannot be exact, it still gives an "interesting" idea.

"15, 17, 18 or 19%, that would not change much".

Do not confuse infection rate and immunity rate

This professor at the School of Advanced Public Health Studies (EHESP) in Rennes prefers to warn about a potential annoying confusion.

That which takes place in the spirits between the rate of infection of a population and its level of immunity.

“We still have little information on the lifespan of these antibodies.

It is possible that the active immunity obtained after an infection declines over time, it is even quite probable ”.

Same warning from the Institut Pasteur: "If the infection undoubtedly confers short-term immunity, it is possible that the latter fades over time".

Morning essentials newsletter

A tour of the news to start the day

Subscribe to the newsletterAll newsletters

This means that the immunity rate in mid-February is probably below the 17% mark.

And the discovery of new strains is unlikely to help matters.

“The Brazilian and South African variants seem to be able to escape this immunity.

These 17% of infected are therefore not immune to reinfection, ”adds Pascal Crepey.

READ ALSO>

Covid-19 reinfections: do the variants call into question immunity?

This Wednesday, Simon Cauchemez also alerted to the impact of these changes.

"We see that vaccination" has a real "impact on the health system", he explained.

But even if the campaign "succeeds very strongly in mitigating the impact of variants", the situation will remain "complicated without further reduction in transmission rates".