“Eleven times more vaccinated in the hospital”: how to make lie real statistics on the Covid-19 – archyde

In France there are, in proportion, seven times more unvaccinated people than people vaccinated in hospital, according to the latest data from the Directorate of Research, Studies, Evaluation and Statistics (DREES) for the end of October. Would the Pays de la Loire be an exception to this pattern? This is what Alexis Poulin, who presents himself as a French-speaking columnist for the Russian channel RT, said in a message posted on Twitter on November 15, and very shared by opponents of vaccination.

According to him, in this region, “Eleven times more vaccinated than unvaccinated are in the hospital”. But in arriving at this conclusion, he made a series of gross statistical errors.

1. Look for a maximum deviation on a particular day

The curve shared in this message indeed comes from the Drees site and corresponds to new hospital admissions for Covid-19 in the Pays de la Loire region, depending on vaccination status, for the week of October 25 to 31.

Alexis Poulin looked in the curve for the day when the difference between the two curves was the highest: on Sunday, October 31, 0.7 unvaccinated person was admitted to conventional hospitalization for Covid-19 against 8.5 people vaccinated in region (these non-integer figures are due to DREES data recovery methods when it crosses the hospitalization and vaccination databases). This in fact results in a ratio of 1 to 11.8. He would have chosen the day before or the day before, he would have found very different ratios.

To mitigate this effect, it is customary, as DREES does, to average the data over a longer period – a week or a month – in order to avoid magnifying effects. For the week of October 25 to 31, there are, in the Pays de la Loire, 20 unvaccinated people in conventional hospitalization for Covid-19, against 75 vaccinated people: the ratio is only 1 to 3.7.

2. Compare populations of very different sizes

The main pitfall of the present reasoning is that it compares phenomena which occur in groups of very different sizes: the number of vaccinated has become largely the majority, rendering ineffective the comparisons in absolute values, as we had already underlined for the examples English Where Israeli, from summer.

In the theoretical example above, we have 28 vaccinated and 5 unvaccinated. An unvaccinated person represents 20% of the vaccinated population, while a vaccinated person weighs only 3.5% of all vaccinated. Therefore, in order for the data to be meaningful, it is necessary to relate hospitalizations to the group to which the hospitalized person belongs.

For the example of Pays de la Loire, at the period concerned, there were 828,852 people unvaccinated against Covid-19 – a figure that can be reduced to 294,974 if we subtract all children under 12 years old. , not eligible for vaccination. The number of people vaccinated in the region amounted to 2,946,599 people.

It is easily understood that 100 people in a group of 294,974 have a much greater weight when diluted in a group of 2.9 million, ten times larger. It is more correct to reformulate the calculation of conventional hospital admissions in Pays de la Loire:

• the 20 admissions of unvaccinated people represent 68.2 per million unvaccinated people from Loire Valley (among those over 12 years old),
• the 75 vaccinated correspond to 25.7 per million vaccinated.

In addition, since October 29, the Drees took the under-20s out of its calculations because, “Due to the very low numbers of hospitalized under 20s, the inclusion of this category of the youngest disrupts the analysis of the statistics produced” -, but the organization continued to make its raw figures available online.

3. Select a territory with a reduced population

Another methodological pitfall: choosing a reduced sample increases the risk of having figures that deviate from the average and problems of representativeness. The Pays de la Loire has only 3.8 million inhabitants (Insee, 2021), and only 3.3 million people over 12 eligible for vaccination, much less than the 58 million French eligible.

Mr. Poulin could have chosen Corsica, even if it means having a really small sample. On the island, four unvaccinated and four vaccinated were hospitalized the week of October 25 to 31. And yet, there too, once reported to their group, it is the unvaccinated (45.16 per million without those under 12) who were the most numerous (against 20.14 per million among the vaccinated) .

Finally, let us recall that two studies ofPPE-Lighthouse – a scientific interest group placed under the dual supervision of the National Medicines Safety Agency (ANSM) and the National Health Insurance Fund (CNAM) – carried out on 22.6 million people were published on October 11 and conclude that “Vaccinated people aged 50 and over have nine times less risk of being hospitalized or dying from Covid-19 than unvaccinated people”.