Impact of vaccination against severe COVID-19 in the French population aged 50 years and above: a retrospective … – BMC Medicine

Study design

We conducted a retrospective ecological study based on French national surveillance data (see Data sources), from December 28, 2020 (week 532020 start of the vaccination campaign) to March 6, 2022 (week 92022 end of Omicron BA.1 wave). The study population was the French population aged 50years and above, stratified into four age groups: 5059, 6069, 7079 and 80years old and above. We estimated the impact of the first and second doses of primary vaccination and the first booster, in terms of the number of averted hospitalizations (admissions to all types of hospitalization services), ICU admissions and deaths.

To compute the number of averted events, we used a method initially developed for tuberculosis [13], which was later applied to influenza [14, 15] and more recently to COVID-19 [6, 7]. Details are given in Additional file 1: Text S1. Basically, the number of averted events (({N}_{mathrm{averted}})) can be estimated from the number of observed events (({N}_{mathrm{observed}})), vaccine coverage ((VC)) and vaccine effectiveness ((VE)), as follows:

$${N}_{mathrm{averted}}= {N}_{mathrm{observed}} times frac{VCtimes VE}{1 - (VC times VE)}$$

(1)

where the term (VCtimes VE) represents the proportion of the population that is protected by vaccination. This method only accounts for the direct effects of vaccination on severe outcomes (not its indirect effects such as impact on transmission) and therefore provides a lower bound estimate of the true number of averted events. Besides, in a scenario without vaccination, it is probable that additional control measures would have been implemented, which we do not account for here (i.e. we estimate the impact of vaccination under the assumption that the same NPIs would have been implemented over the study period). These two points will be further addressed in the discussion.

This formula can be extended to account for the week of observation ((w)) and the number of doses received ((k), ranging from one to three (two doses and one booster)) [6, 15]:

$${{N}_{mathrm{averted}}}_{w}= {{N}_{mathrm{observed}}}_{w} times frac{sum_{k=1}^{3}{VC}_{w,k} times {VE}_{k}}{1 - (sum_{k=1}^{3}{VC}_{w,k} times {VE}_{k})}$$

(2)

where ({VC}_{k}) represents the vaccine coverage of exactly (k) doses (not at least (k) doses).

This formula assumes that VE is the same for all vaccinated individuals, regardless of the time elapsed since vaccination. However, VE is not constant over time: it quickly increases in the first weeks following vaccination due to the build-up of immunity and declines over time due to waning immunity [16,17,18,19]. In order to account for the evolution of VE according to the time elapsed since vaccination (left(Delta right)) in weeks, we modified the formula as follows:

$${{N}_{mathrm{averted}}}_{w}= {{N}_{mathrm{observed}}}_{w} times frac{sum_{k=1}^{3}sum_{Delta =0}^{w-1}{VP}_{w,k,Delta } times {VE}_{k, Delta }}{1 - (sum_{k=1}^{3}sum_{Delta =0}^{w-1}{{VP}_{w,k, Delta } times VE}_{k, Delta } )}$$

(3)

where ({VE}_{k, Delta }) is the vaccine effectiveness of the (k)th dose (Delta) weeks after vaccination, and ({VP}_{w,k,Delta }) represents the proportion of people who received their last dose (k)exactly (Delta) weeks before the week of observation (w). Of note, the sum of ({VP}_{w,k,Delta }) over all (Delta) corresponds to the vaccine coverage of exactly (k) doses on week (w) ((sum_{Delta =0}^{w-1}{VP}_{w,k,Delta }={VC}_{w,k})).

This formula was applied separately for each age group and each variant, in order to account for different VE according to age groups and variants. The total number of events directly averted by vaccination (number of hospitalizations, ICU admissions and deaths) was then obtained by summing the number of averted events over all weeks, age groups and variants. All analyses were conducted in R software version 4.1.2 (R Foundation, Vienna, Austria).

For the number of observed events, we relied on hospitalization and death data from the SI-VIC database, maintained by the ANS (Agence du Numrique en Sant) and sent daily to Sant publique France, the French national public health agency. This database provides real-time data on patients hospitalized for COVID-19 in French public and private hospitals, including their age, date of hospitalization, type of hospitalization services and outcome (discharged/deceased). All COVID-19 cases are either biologically confirmed or present with a computed tomographic image highly suggestive of SARS-CoV-2 infection. For hospitalizations, we included patients hospitalized in all types of services (general wards, ICU, long-term care and rehabilitation, emergency care) and excluded patients hospitalized for reasons not linked to a COVID-19 infection. For deaths, we included all patients deceased in the hospitals with a COVID-19 infection and added individuals deceased in nursing homes (resident homes for elderly) with a COVID-19 infection from the SurvESMS database. The SurvESMS database, administered by Sant publique France, was designed for the monitoring of COVID-19 cases and deaths among the residents of nursing homes; it allows the distinction between residents who died in nursing homes from those who died in hospitals (already accounted for in the SI-VIC database). In the absence of data on age, individuals were considered 80+years old; this assumption was based on 2019 data which showed that the mean age of people deceased in nursing homes was 89years old [20]. Deaths at home could not be accounted for. The time series of hospitalizations, ICU admissions and deaths according to each variant were reconstructed using the SI-DEP database, the national surveillance system describing SARS-CoV-2 RT-PCR and antigen test results arising from all private and public French laboratories (see details in Additional file 1: Text S2). The three databases (SI-VIC, SurvESMS and SI-DEP) are intended to be exhaustive.

Regarding vaccine coverage, we used the VAC-SI database, the national information system developed by the French Health Insurance to monitor the implementation of COVID-19 vaccination campaigns, since the start of vaccine distribution in December 2020, across the country. Individual data include the number of doses, the date of vaccination, the type of vaccine, and socio-demographic information such as age. All types of vaccine were considered (Pfizer/BioNTech BNT162b2, Moderna mRNA-1273, Oxford/AstraZeneca ChAdOx1-S and Johnson & Johnson Ad26.COV2.S). For each week of observation (w) and each age group, the number of vaccinated individuals according to the number of doses received before the week of observation (w) and to the week of vaccination (in order to compute the time elapsed since vaccination, (Delta)) were extracted. For each individual, on a given week of observation (w), only the last dose received before that week was taken into account, to avoid counting the same person twice: e.g. when an individual received a second dose on week (w), they were no longer counted among the first-dose individuals from week (w) onwards (they were only counted among the first-dose individuals up to week (w-1)). We excluded individuals for which the date of the first dose was posterior to the date of the second dose or the booster dose, or the date of the second dose was posterior to the date of the booster. For the denominator (number of individuals per age group in the French population), we used 2022 demographic data from the National Institute of Statistics and Economic Studies (INSEE).

With regards to the effectiveness of COVID-19 vaccines against COVID-19 variants, we extracted data from a recent VE study conducted by Sant publique France [12]. The study investigated VE of the two mRNA vaccines Pfizer/BioNTech BNT162b2 and Moderna mRNA-1273 against Alpha (unpublished data), Delta and Omicron BA.1 severe outcomes (general ward hospitalizations and ICU/deaths) among immunocompetents50years old French individuals, between January 11, 2021, and February 10, 2022. Analyses were performed according to 5079 and 80years old and above age groups, and the number of vaccine doses (one dose, two doses and the first booster dose). VE according to time since vaccination were also estimated. These estimates were smoothed to remove random fluctuations over time.

Note that VE against Omicron severe outcomes were sourced from multiple studies available in the literature: the aforementioned French study for VE of the first dose against ICU/deaths [12], a test-negative casecontrol study in England for VE against hospitalizations among people aged 65years and older (which we applied to our 50+population) [21], and a test-negative casecontrol study in Canada for VE of the second dose and booster against deaths among people aged 18years and older (which we applied to our 50+population) [16].

In these three studies, the date associated to an individual is the date of the test or the date of symptom onset. In order to account for the time between the test or symptom onset and the event of interest (hospitalization, ICU admission or death), we applied a lag of 1week to VE against hospitalization and ICU admission and a lag of two weeks to VE against death. Finally, a linear decay of VE was assumed to account for the waning of immunity at longer time horizons. The decay values were extracted from the literature and set to0.5 points/week for the second dose after 21weeks [22, 23] and0.4 points/week for the booster dose after 21weeks [24]. In the absence of data from the literature for the first dose, we applied a coefficient twice higher than for the second dose (1 point/week). This coefficient was applied from the 13th week after vaccination [25]. In order to estimate uncertainty around the number of averted events, we used the 95% confidence intervals of VE estimates published in the three aforementioned studies [12, 16, 21].

Go here to see the original:

Impact of vaccination against severe COVID-19 in the French population aged 50 years and above: a retrospective ... - BMC Medicine