Covid-19: Do DREES figures show a sharp drop in vaccine effectiveness? – Release

Question asked on 05/11/2021.

Several questions were addressed to CheckNews concerning the evolution of the proportion of vaccinated and unvaccinated among hospitalized patients in connection with Covid-19. You ask us in particular on graphs published on Twitter, in which the proportion of vaccinated hospitalized people is increasing, suggesting a decline in vaccine protection. In a tweet on November 5, Guillaume Rozier – computer engineer behind the CovidTracker site – wrote as follows: “It appears that the vaccine-induced reduction in the risk of admission to hospital is weaker and weaker in those over 80 years of age. The risk reduction was 10 times before the summer, it is only 4 times in this age group. “

Are these data correct and, if so, can they be interpreted as an illustration of a drastic drop in the efficacy of the vaccine?

Every week since June 2021, the DREES (statistical unit of the Ministry of Health) matches the information reported in the three databases dedicated to the Covid crisis : the Vac-Si (population vaccination), Si-Vic (hospital data), and Si-Dep (covid test results) databases. This pairing makes it possible in particular to determine the number of vaccinated and non-vaccinated who enter the hospital, in critical care, or tested positive.

On the basis of these data, CheckNews reconstructed the evolution, from the beginning of June to the end of October, of the number of hospital admissions of people positive for Covid-19, among a million vaccinated, then for a million unvaccinated. And this, for over 20 years and for four age groups. Before calculating the ratio between the number of hospitalizations among the unvaccinated and that of the vaccinated. In other words, the importance of the risk, for the unvaccinated, of experiencing a hospitalization in connection with the Covid-19, compared to the vaccinated (1).

For the population over 20 years of age, the unvaccinated thus had, at the end of October, “only” 7 times more risk of being hospitalized than the vaccinated, against 12 times more a month and a half earlier (curve yellow dotted).

A “pseudo” indicator of vaccine efficacy

Considering that this decrease directly reflects the evolution ofvaccine efficacy against hospitalization, this would be equivalent to moving, in a month and a half, from a protection of 91.7% (1 vaccinated hospitalized for 12 non-vaccinated hospitalized) to a protection of 85.7% (1 vaccinated hospitalized for 7 non-vaccinated hospitalized).

In fact, several studies have already confirmed the gradual decrease, over the months, of the protection against symptomatic forms conferred by the vaccine. As an example, according to monitoring of volunteers involved in the evaluation of the Pfizer vaccine, the effectiveness against the appearance of Covid-19 goes from 96% during the first two months post-vaccination to 84% six months after vaccination. Regarding protection against the risk of infection with the virus (which does not necessarily translate into symptoms), recent Israeli data shows that there are three to four times more cases in people vaccinated for six months compared to those vaccinated. since two months. According to the same data set, a decrease in protection is also observed for the risk of severe forms in people over 60 years of age (with uncertainty as to the extent of this decrease).

However, all the specialists interviewed agree in saying that the evolution of the proportion of vaccinated among hospitalized patients only very imperfectly reflects the evolution of vaccine efficacy. Contacted by CheckNews, the Drees therefore invites us to interpret these figures with caution: “We can deduce from this ratio (relative risk) a vaccine pseudo-efficacy which is similar to a descriptive statistic, but which is not as precise as the result of a case-control study where the [groupes de personnes] fully vaccinated and unvaccinated would have similar characteristics [en termes] age, department of residence, date of hospital events, sex ”. Therefore, believes the Drees, “It is not possible, at this stage, to calculate and observe a vaccine efficacy with the data made available in open data, because it would be necessary to neutralize the effects of age structures and other factors which play a important role in the impact of the vaccine ”.

Is the trend the same for all age groups?

Admittedly, the data published by DREES do not allow us to compare populations with the same characteristics, particularly in terms of comorbidities or places of residence. It is nevertheless possible to isolate the results by age groups. And observe the consequences of this stratification on the ratio of unvaccinated and vaccinated hospitalized.

Observation illustrated in the graphs below: this ratio seems to decrease all the more as the populations studied are older. A result that is not entirely surprising, both because of the vaccination schedule (the oldest people are also those who were the first to be eligible for vaccination) and the gradual drop in immune functions observed among the oldest.

In detail, among those over 80, the unvaccinated were 3.5 times more likely to be hospitalized than those vaccinated at the end of October, against 8 to 12 times more at the beginning of June. Or a division of the ratio by more than two or three in five months.

For those aged 60-79, the ratio, after having slowly increased, went from 18 at the end of September to 7.5 at the end of October.

For 40-59 year olds, the ratio – which has seen several significant oscillations in recent weeks – is now around 18, after a long increase since June, and a peak around 36 in mid-October.

Finally, for 20-39 year olds, the ratio has also experienced strong variations, and is now around 13, against 23 at the end of September.

To better understand the general evolution of this ratio (that is to say the relative risk of hospitalization between non-vaccinated and vaccinated), it is also possible to establish the “trend” of the data series. The recent drop in the ratio of hospitalizations between unvaccinated and vaccinated therefore appears to concern mainly those over 80 – who, once again, are among the first beneficiaries of the vaccination campaign – and, to a lesser extent, the last. period, the 60-79 years old.

Do not confuse this ratio with vaccine efficacy

But even by isolating the age parameter, it remains tendentious to estimate the extent of a drop in vaccine efficacy from these curves. Other biases can indeed widen the gap between the theoretical effect of the vaccine (which would be observed in populations which differ only in their vaccination status) and what is observed in real life.

“It may be that there is more risk taking among the vaccinated, sometimes asked to drop the mask and barrier gestures”, thus estimates the epidemiologist Antoine Flahault (University of Geneva). Conversely, “The health pass could have forced unvaccinated people to have fewer physical interactions and therefore to be less exposed to the virus”, adds his colleague Samuel Alizon (University of Montpellier).

“Another effect that we tend to forget, details Pascal Crépey (University of Rennes): the vaccinated [en limitant la circulation du virus] protect the unvaccinated. “ They therefore benefit from indirect protection, “Not sufficient to prevent all infections or an epidemic resumption – otherwise we would have achieved group immunity – but protection all the same, which increases with the number of vaccinated”.

So many behavioral parameters that make it difficult to interpret the raw data from the DREES. In addition, as researcher Samuel Alizon reminds us, “We could also add to the equation the replacement of the Alpha variant by the Delta variant in July, knowing that the latter seems to have a greater propensity to escape immunity, even if it remains very strong for vaccine immunity” .

For the statistics unit of the Ministry of Health, it is even impossible, “At this stage, to calculate and observe a vaccine efficacy with the data made available in open data”, as it would be necessary simultaneously “Neutralize the effects of age structures and other factors that play an important role in the impact of the vaccine”. La Drees nevertheless specifies that it is working “A model of the effectiveness of vaccines which would take into account all these effects”, for publication “in the next weeks”.

(1) Guillaume Rozier, on Twitter, proposed this same exercise for all Covid-19 cases, whether suspected or confirmed with a positive test.