Scenario Modeling Hub

In the first influenza round of 2023-24, we generated pre-season projections for the period Sep 3, 2023 to June 1, 2024. We considered 6 scenarios representing the impact of 3 different levels of vaccine coverage (similar to the 2021-22 season, and 20% higher or lower), combined with the dominance of the influenza A/H3N2 or A/H1N1 subtype. Ensemble projections are based on contributions from 10 teams (including 9 contributing national projections).

Our main findings include:

• The hospitalization and death burden of the next influenza season will be more heavily driven by the dominant subtype than by vaccination coverage, with an A/H3N2-dominant season projected to have moderate to high impact, and A/H1N1 low to moderate impact, compared to influenza epidemics in the past decade.
• In our most optimistic scenario (scenario B, high vaccine coverage, A/H1N1 dominance), median weekly hospitalizations would peak at 10,400 (95%PI 900-22,600). In our most pessimistic scenario (scenario E, low vaccine coverage, A/H3N2 dominance), weekly hospitalizations would peak at 24,100 (95%PI 1,300-41,700).
• In all scenarios, ensemble projections suggest a prolonged period of high influenza activity before and after the New Year, in part due to differences in projected peak timing across models. Periods of high influenza activity tend to occur earlier in A/H3N2 projections while A/H1N1 projections have more protracted activity lasting into the Spring.
• A 20% relative increase in vaccine coverage, compared to usual, would avert 9% (95% CI 3%, 16%) of influenza-related hospitalizations in the A/H3N2 scenario, and 9% (0.5-18%) in the A/H1N1 scenario. A 20% drop in vaccine coverage compared to usual, potentially fueled by a rise in vaccine hesitancy, would increase influenza-related hospitalizations by 10% (95% CI 4-15%) in the A/H3N2 scenario, and 12% (95%CI 5-19%) in the A/H1N1 scenario. Projected percent changes in deaths are less pronounced than in hospitalizations. In absolute terms, this represents differences in the order of 22,000 to 34,000 hospitalizations and 800-1,400 deaths (range across medians).
• Based on median ensemble projections, the combined impact of influenza and COVID-19 on hospitalizations is projected to be lower than that last season (2022-23). This is based on the assumption of high immune escape for COVID-19, and moderate COVID-19 booster uptake in all age groups (Round 17 scenario A, https://covid19scenariomodelinghub.org/). This combined impact is based on the median of both flu and COVID19 scenarios and there is considerable variability within these projections that is not captured by the median.
• A few caveats are worth noting:
  • We assumed a fixed VE of 40% against medically attended illnesses in all flu scenarios, which anticipates a good match between circulating viruses and vaccine strains.
  • These are pre-season projections, and hence there is no calibration data on the dynamics of the upcoming epidemic. These projections are primarily based on the historical dynamics of influenza A/H3N2 and A/H1N1 epidemics. As a result, there is high variability in the projected dynamics even within a single model and scenario, which in turn increases uncertainty in projected vaccine benefits. There is also considerable variability between models as regards projected timing and severity of epidemics, in part due to differences in underlying assumptions regarding seasonality and seeding.
  • Only 7 participating models contributed national death projections. Together with more limited calibration data available for deaths, our projections for deaths may be somewhat less reliable than for hospitalizations.
  • Testing practices continue to evolve in the wake of the COVID-19 pandemic, including increased use of multi-pathogen testing in clinical settings, which may affect reported hospitalizations. This in turn will affect comparison with our projections and with prior year hospitalization data.

Table 1. Flu Scenario Modeling Hub round 1 2023-2024 scenarios. More detailed scenario definitions and model characteristics can be found at https://github.com/midas-network/flu-scenario-modeling-hub.

Even the best models of emerging infections struggle to give accurate forecasts at time scales greater than 3-4 weeks due to unpredictable drivers such as a changing policy environment, behavior change, the development of new control measures, and stochastic events. However, policy decisions around the course of emerging infections often require projections in the time frame of months. The goal of long-term projections is to compare outbreak trajectories under different scenarios, as opposed to offering a specific, unconditional estimate of what “will” happen.

As such, long-term projections can guide longer-term decision-making while short-term forecasts are more useful for situational awareness and guiding immediate response.