Round 5
“Assessing the impact of COVID-19 booster uptake on the 2023/24 winter burden”
Report created 2024-02-07
Scenarios
|
No vaccination
|
Pessimistic vaccination
|
Optimistic vaccination
|
Theoretical upper boundary vaccination
|
|
Optimistic waning
|
Scenario A | Scenario B | Scenario C | Scenario D |
Pessimistic waning
|
Scenario E | Scenario F | Scenario G | Scenario H |
See also the full scenario details for more detail on the common set of assumptions teams used to create their models.
In Round 5, we asked modellers to start their projections from the 2023-08-01. Data after this date were not included, and as a result, model projections are unlikely to fully account for later information on, for example, the changing variant landscape or behavioural patterns.
Assumptions left to modeller judgement
Modellers should use their own judgement and relevant literature if making assumptions about a number of modelling aspects. Note that the resulting between-team differences due to these assumptions are desired, and reflect uncertainty regarding these aspects. They are a main contributor to the uncertainty of the analysis results.
- Vaccination (and uptake levels) of at-risk individuals of all ages as well as vaccination of healthcare workers.
- Functional shape of waning immunity against infection and against severe outcomes (following the above specified decrease of immunity against infection within the specified time frame).
- The absolute value of vaccine effectiveness and protection from past infections for all outcomes.
- The under-ascertainment factor and the impact of downscaled testing.
- Contact rates in response to infection rates.
- Impact of seasonality, including seasonal behavioural changes.
Round 5 submissions
Round 5 has received submissions from six international modelling teams from across the EU/EEA and beyond. Submissions are from a diversity of modelling approaches, adding to the between-team variability of projections. One team submitted projections for 30 EU/EEA countries, one team submitted for 6 countries, and 4 teams for one EU/EEA country.
Table 1. Participating teams by number of modelled countries, scenarios, age-disaggregation
Results and limitations
We first ask, what is the expected 2023/24 winter COVID-19 burden without any further vaccination? And which age groups would be most impacted? We find that without further vaccination, COVID-19 death burden is projected to be around X-X% (80% credible interval) of last year’s death. We focus here on deaths because other outcomes are less comparable across the years. Without further vaccination, elderly (60+) are projected to contribute little to infections (12-17%), but elderly contribute the majority of hospitalisations (92-98%) and deaths (96-99%). This is because the per-capita rate of hospitalisation for elderly is approximately 5 times higher than that of younger individuals, and for death it is roughly 30 times higher.
We then ask, what is the impact of a high autumn COVID-19 booster uptake on expected burden compared to a low uptake? The results show that a high-uptake booster campaign can reduce deaths substantially. Specifically, while a low-uptake autumn booster campaign reduces projected deaths by up to X% (50% upper credible interval bound, the lower being X), this impact increases to up to X% for a high-uptake booster campaign (lower bound being X). The results are similar for prevention of hospitalisations (X% - X% for a high-uptake campaign). The impact of a booster campaign increases when there is fast immune evasion by variants and/or fast immunity waning. The overall impact of a booster campaign on infections is minimal (<4%), so that for the reduction of COVID-19 spread other measures would be required.
There is a substantial variation in the above estimates, which is partly due to the aggregation across different countries, as well as variation between models, and model assumptions. This variations reflects large unknowns with respect to the level and waning of protection from prior infection, vaccination, and hybrid immunity, characteristics of future variants, seasonality and social behaviour of contact mixing and mobility, the under-ascertainment factor. Note that immunocompromised subgroups, other high-risk groups, or specific settings (e.g., hospitals, LTCF) are not addressed with this modelling.
Methods
COVID-19 burden 2023/24 relative to last year
- For each country, scenario, model, and projection sample, we divide the cumulative deaths projected for the time window 1 August 2023 to 1 June 2024 by the reported deaths in the time window from the previous year.
- We include only the scenarios without vaccination (Scenario A and E). And we included only 5 models, because 1 model did not include projections for death burden.
- In a Bayesian framework, we model the projected burden (relative to last year and log-transformed) as a function of individual countries, and the projecting model.
- We did not include an explicit effect of waning, because projections for both waning scenarios are balanced.
- Variability between projection samples are modelled through model-specific uncertainties.
COVID-19 burden 2023/24 by age group
- For each projection sample we divide projected burden (COVID-19 infections, hospitalisations, and death) in elderly (60+) by the projected burden in the remaining age groups (<60).
- We include only the scenarios without vaccination (Scenario A and E). Only 5 models provided projections split by age group and could be used for this analysis.
- In a Bayesian framework, we model projected age-specific burden as a function of individual countries, and the projecting model.
- Variability between projection samples are modelled through model-specific uncertainties.
Impact of booster uptake on COVID-19 burden 2023/24
Accounting for imbalanced submissions
Figure 1: Reduction in cumulative COVID-19 death burden through a 2023 autumn booster campaign relative to the no-booster scenario. Vaccine impact estimates are shown for the high-uptake (top) and the low-uptake (bottom) conditions. Those estimates are then further split by fast and slow immunity waning. The light and dark bars show the 80% and 50% quantiles, respectively..
Additional notes
…