COVID-19

Scenario Modeling Hub

A Note on Scenario Modeling Hub Round 17 (June 26th, 2023)

In a new round of projections, the Scenario Modeling Hub evaluated the trajectory of COVID-19 during April 16, 2023 to April 19, 2025 (104-week horizon), under 6 scenarios about the annual uptake of reformulated boosters (minimal uptake, uptake in 65+ corresponding to 2021 booster levels, or uptake in all ages corresponding to 2021 levels) and extent of immune escape of circulating variants (50% vs 20% annually). Eight teams contributed both national and state-specific projections, while one team submitted results for a subset of states.

Our main findings include:

  • Based on the national ensemble, the main period of COVID19 activity is expected to occur in late fall and early winter over the next 2 years, with median peak incidence between November and mid January. Lowest incidences are projected to occur in August of each year.
  • For the range of scenarios considered, weekly hospitalizations and deaths are likely to stay within last year’s range, and unlikely to hit Delta or Omicron peaks. Further, weekly hospitalizations are likely to remain at low or medium community transmission levels and unlikely to reach high transmission levels (>20 weekly hospitalizations per 100,000), as defined by the CDC.
  • In the most pessimistic scenario (no booster, high immune escape) we project 2.1 million hospitalizations (1.4 million-4.5 million) and 209,000 deaths (138,000-479,000) over the 2-year projection period, with 839,000 hospitalizations and 87,000 deaths in the first cold month season (Sep 2023-Apr 2024). In the most optimistic scenario (boosters for all ages, low immune escape) this reduces to 1.4 million hospitalizations (907,000-2 million) and 122,000 deaths (55,000-201,000) over 2 years, with 484,000 hospitalizations and 45,000 deaths occurring in the first cold month season. The Sep 2024-Apr 2025 season is projected to be slightly (5-22%) more severe than the coming season.
  • Vaccination of 65+ and of all ages would significantly reduce disease burden compared to no vaccination scenarios, irrespective of immune escape assumptions. Under low and high immune escape scenarios vaccination of 65+ reduces hospitalizations by 11% and 9%, and reduces deaths by 16% and 13%; targeting all ages reduces hospitalizations by 23% and 17%, and deaths by 26% and 20%, compared to no vaccination. In absolute numbers vaccinating 65+ would result in 202,000 (114,000-290,000) fewer hospitalizations and 28,000 (13,000-44,000) fewer deaths nationally over the two year projection period in low immune escape scenarios, compared to no vaccination. Expanding vaccination to all ages increases these reductions to 430,000 (264,000-598,000) hospitalizations and 49,000 (29,000-69,000) deaths under low immune escape assumptions. Reductions in numbers of deaths and hospitalizations are similar, but slightly higher, in high immune escape scenarios.
  • A few caveats are worth noting:
    • We assumed the VE of reformulated boosters would be 65% against symptomatic disease at the time of reformulation in June of each year. The effectiveness of reformulated boosters against existing and new variants remains unclear, as does the pace of waning after multiple booster shots and repeat infections.
    • We assumed continuous immune escape rather than discrete variants, mirroring observations of evolutionary changes in the last year. We did not consider the impact of a significant new variant that would have accumulated a large amount of antigenic changes over a very short period, nor increase in transmissibility (akin to Delta or Omicron variants). We also assumed that the intrinsic severity (i.e. severity in naive populations) of future circulating strains would remain similar to that of the Omicron lineages.
    • There is considerable heterogeneity between states and between individual models. In particular, in the high immune escape scenarios some models project a second smaller peak in late Spring.
    • We switched calibration of death data to a different dataset (NCHS) following the end of the CSSE surveillance system, which may introduce small differences when comparing Round 17 death projections with those of past rounds. Further, we no longer project case trajectories. Hospitalizations from HHS protect continue to be a stable outcome, although reporting to this system remains unclear over the full 2 year projection period.
    • These are results based on projections from 9 teams, of which 8 provided national estimates for hospitalizations and 7 provided national estimates for deaths. Results will be updated should projections from additional models become available.

Figures. Estimates of burden averted by increasing target age groups for vaccination (pooled national estimates across models for the full projection period

Table 1. COVID-19 Scenario Modeling Hub round 17 scenarios. More detailed scenario definitions and model characteristics can be found at https://github.com/midas-network/covid19-scenario-modeling-hub.


Rationale

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. The need for long-term epidemic projections is particularly acute in a severe pandemic, such as COVID-19, that has a large impact on the economy; for instance, economic and budget projections require estimates of outbreak trajectories in the 3-6 month time scale.