Last updated: 2025-04-02 for Round 19 Scenarios.
https://covid19scenariomodelinghub.org/viz.html
Previous rounds (round 1 to round 18) are available in the COVID-19 Scenario Modeling Hub - Archive GitHub Repository
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.
From weather to infectious diseases, it has been shown that synergizing results from multiple models gives more reliable projections than any one model alone. In the COVID-19 pandemic this approach has been exemplified by the COVID-19 Forecast Hub, which combines the results of over 30 models (see a report on the first wave of the pandemic). Further, a comparison of the impact of interventions across 17 models has illustrated how any individual model can grossly underestimate uncertainty, while ensemble projections can offer robust projections of COVID-19 the course of the epidemic under different scenarios at a 6-month time scale.
The COVID-19 Forecasting Hub provides useful and accurate short-term forecasts, but there remains a lack of publicly available model projections at 3-6 month time scale. Some single models are available online (e.g., IHME, or Imperial College), but a decade of infectious disease forecasts has demonstrated that projections from a single model are particularly risky. Single model projections are particularly problematic for emerging infections where there is much uncertainty about basic epidemiological parameters (such as the waning of immunity), the transmission process, future policies, the impact of interventions, and how the population may react to the outbreak and associated interventions. There is a need for generating long-term COVID-19 projections combining insights from different models and making them available to decision-makers, public health experts, and the general public. We plan to fill this gap by building a public COVID-19 Scenario Hub to harmonize scenario projections in the United States.
We have specified a set of scenarios and target outcomes to allow alignment of model projections for collective insights. Scenarios have been designed in consultation with academic modeling teams and government agencies (e.g., CDC).
This repository follows the guidelines and standards outlined by the hubverse, which provides a set of data formats and open source tools for modeling hubs.
The COVID-19 Scenario Modeling Hub is be open to any team willing to provide projections at the right temporal and spatial scales, with minimal gatekeeping. We only require that participating teams share point estimates and uncertainty bounds, along with a short model description and answers to a list of key questions about design. A major output of the projection hub would be ensemble estimates of epidemic outcomes (e.g., cases, hospitalization and/or deaths), for different time points, intervention scenarios, and US jurisdictions.
Those interested to participate, please read the README file and email us at [email protected].
Model projections should be submitted via pull request to the data-processed folder of this GitHub repository. Technical instructions for submission and required file formats can be found here.
Round 19 focuses on the impact of different target populations for COVID-19 boosters combined with different timing of vaccination for the 2025-2026 season. The timeframe of projections will be Sun April 27, 2025 to Sat April 25, 2026 (52 weeks).
In all scenarios, boosters are expected to match the predominant variants circulating on June 30, 2025. Teams should use VE against COVID-19 hospitalization = 45% at the start of the vaccination campaign (either July 15 or Sep 1, 2025), in line with a recent US analysis of hospitalizations during September-December 2024. We note that in this study, VE against hospitalization was similar 0-2 vs 3-4 months after vaccine receipt, and among immuno-compromized vs healthy individuals. The study was only powered to estimate VE hospitalizations in individuals 65+ yo but it is generally accepted that COVID-19 VE does not depend on age. We recommend using the same VE across all population groups considered in round 19.
This VE is equivalent to a vaccine trial that would be performed at the start of the vaccination campaign in populations with varying levels of prior immunity at trial enrollment. Vaccinated individuals would have a 45% reduced risk of hospitalization compared to unvaccinated individuals on average in this trial, if VE was estimated a few days after the start of the campaign. Importantly, this stated hospitalization reduction includes the combined effects of protection against infection and protection against hospitalization given (breakthrough) infection. This is the same logic we used in round 18. Based on available evidence, we suggest that teams choose VE against infection in the range 35-57% at the start of the vaccination campaign (see VE Estimates 2023/2024 for detailed studies).
Two mechanisms will result in an effective decrease from the stated VE of 45% against hospitalization. The first mechanism is immune escape, with circulating strains moving gradually away from the vaccine. In this round, immune escape against infection and severe disease is at teams’ discretion (we return to this later).
The second mechanism is waning of immunity against infection, due to decline in immune responses over time. This applies to both vaccine-induced and natural immunity. Parameters for waning immunity against infection are left at teams discretion, although we suggest a 3-10mo waning time, with 40-60% reduction against baseline protection levels in waned state. Waning of vaccine-induced immunity against severe disease remains at teams discretion (but if present should wane at slower timescale than against infection).
In this round, we consider two different timings of vaccination. Scenarios B and D correspond to a “classic” timing of initiation of the seasonal vaccination campaign, with a start on Sep 1, 2025. This is the timing that is expected to operate in the 2025-26 season. We also consider hypothetical scenarios C and E, where vaccination is pushed 1.5 month earlier and starts on July 15, 2025 (meaning that the entire vaccine manufacturing and delivery process is moved earlier in time). These hypothetical scenarios may be useful to project the potential value of earlier vaccination campaigns in future COVID19 seasons, in light of substantial summer waves of COVID-19 observed in recent years. Vaccine coverage curves is provided for all scenarios and reflects the stipulated timing of vaccination in each scenario (available in the auxiliary-data/vaccination-coverage folder).
Vaccine coverage curves is provided for all scenarios and will be extrapolated from the 2024-25 data. The data are available in the auxiliary-data/vaccination-coverage folder.
No recommendation (scenario A): There is no future recommendation to get additional booster doses or receive additional vaccination. Teams should not model any future vaccination in the projection period, i.e., vaccine coverage is 0% in all population groups. Without recommendation, vaccines will not be covered by insurance or other sources.
Annual vaccination recommended for high risk groups (65+ and those with underlying risk factors)(scenarios B & C): Uptake of annual booster in high-risk groups (65+ and other individuals with underlying risk factors for severe COVID-19 outcomes) follows uptake observed for the booster dose during the 2024-25 season. Vaccination among non-recommended groups should be modeled as negligible (ie, 0%) as without recommendation, these groups will not be covered by insurance or other health care funding. Vaccine uptake data in 65+ and high-risk groups will be provided.
Annual vaccination recommended for currently eligible groups (ages 6 months and older) (scenarios D & E): Uptake of annual booster in all groups follows uptake observed for the booster dose during the 2024-25 season. Uptake data for all groups will be provided.
In this round, we assume that high-risk populations, of any age, are included in booster recommendations in scenarios B, C, D & E. We define high-risk groups as those with underlying conditions putting them at increased risk of severe outcomes from COVID-19. Data on the population size and vaccine coverage of high and low risk groups is provided by state and age in GitHub. Teams can choose to adjust VE for high-risk and low-risk groups based on available evidence, although the population-level average VE against hospitalization should equal 45% on September 1. Data on increased risk of COVID-19 hospitalization from high risk groups can be found here. If teams wish to adjust VE based on specific risk conditions, they can refer to data on VE in immunocompromised populations here.
SARS-CoV-2 immune escape away from existing immunity should proceed at a constant rate throughout the year, aligned with the diversity of strains that is now circulating in the population. The rate of immune escape is left at teams discretion but, in following recent prior rounds, we suggest that it should be bounded by 20-50% per year.
Immune escape will affect protection conferred by natural infection and vaccination. For instance, let’s assume that immune escape is 20%. Now, assume an individual is infected on June 15, 2025 and this infection confers X% protection against symptoms, compared to an individual who has not been recently infected. If this individual was instantaneously transported a year later, on June 15, 2026, with their antibodies from the 2025 infection intact, this individual's protection against variants circulating on June 15, 2026 would be X * 0.8 (20% immune escape). In this thought experiment, the decay of protection would solely be due to the effects of immune escape. In reality, moving away from the thought experiment, if this individual actually lived throughout an entire year without a new infection between June 2025 and June 2026, then their effective immunity on June 15, 2026 will be the combined ffects of antibody waning (at a rate and plateau left at teams’ discretion) and immune escape (at teams discretion as well).
Teams should note that the impact of immune escape is separate from the impact of waning immunity (especially because the impact of immune escape affects infection and vaccination differently), although these processes may be implemented similarly in models.
It is left to the teams’ discretion how to implement immune escape in their models. Teams may choose to sample over a range of immune escape values that is based on plausibility (for instance, within 20-50% per year), or informed by calibration to epidemiological or strain cycling data. Teams may choose to implement gradual escape of existing variants, or they can choose to introduce new discrete variants with levels of immune escape consistent with the epidemiology of SARS-CoV-2 in the past year, so long as these occur frequently.
Teams must incorporate waning of immunity against infection. The median waning time of protection against infection should range between 3-10 months (this should not be read to mean that waning is to complete loss of protection, see below). Teams can sample this range, or use any value within this range as a point estimate. Teams can consider differences in waning of natural and vaccine-induced immunity, or in waning after Omicron infection vs waning from other types of SARS-CoV-2 exposures; however the median waning time should remain within the 3-10 month range.
The rate and levels of waning are left to the best scientific discretion of the teams. We recommend that in the waned classes, teams consider a reduction from baseline levels of protection ranging between 40% and 60%, corresponding to x0.60 and x0.40 of the baseline levels reported immediately after exposure (vaccination or infection).
Teams may incorporate waning of immunity against severe disease, however the timescale of waning against severe disease must be slower than the timescale of waning against infection.
It is important that scenarios are directly comparable in the amount of immune escape and waning, both in terms of the proportion of population in different immune classes at the start of projections, and throughout the projection period (because immune escape and waning are not part of the scenario axes). These shared assumptions should include rate of immune escape (or number of variants modeled), timescale of immunity decline and plateau reached after immunity has waned, if any. Ideally, simulations should be paired across scenarios, with each simulation of the “counterfactual” (scenario A) having a comparable simulation in each of the other intervention scenarios in terms of immune escape and waning assumptions during both calibration and projection periods. Alternatively, if pairing is not feasible, each scenario should similarly draw from shared distributions to ensure comparability, with no variation in assumptions/specifications across the scenarios during the calibration period. Only by having paired simulations or completely comparable starting conditions at the start of the projection period for all scenarios can we evaluate the impact of different vaccine coverage and timing assumptions in the projection period. If past immune escape and/or waning immunity parameters are unobservable from the recent data, estimates can be drawn from the literature to help with calibration. It is also acceptable to use the midpoint of the recommended immune escape bounds (35%, midpoint between 20% and 50%) for calibration of immune escape in the recent past.
Recent years have seen marked levels of COVID-19 activity in the summer and fall, along with a winter wave of varying magnitude. Seasonality has also varied geographically, with Southern states experiencing more pronounced summer waves. It is important to try to reproduce this feature of COVID-19 epidemiology as COVID-19 timing interacts with the timing of the vaccination campaign, which is one of the scenario axes.
Teams should include their best estimate of COVID-19 seasonality in their model. We do not prescribe a specific level or shape of seasonal forcing but we ask that teams check that their models are able to reproduce the observed timing of COVID-19 activity in the past year in their calibration step. Note that reporting to the NHSN hospitalization dataset was paused during May-November 2024, with reporting from fewer than 75% hospitals over the period. Hospitalization levels reported in this dataset should not be taken at face value. Death data is complete however. Teams will also be provided with auxiliary datasets from other surveillance systems to help with calibration (eg, wastewater surveillance, ED visits, COVID-net).
Teams should assume that the projection period will not see the emergence of any unusual variants, other than those implied by the level of immune escape chosen for a specific simulation. Treatment of variants existing at the start of the projection period is left to the discretion of the teams. Intrinsic transmissibility and severity of disease in a naive individual is assumed to be constant across all currently-circulating and future variants.
Teams should NOT include reactive changes in NPIs imposed by health authorities to curb transmission, e.g., reinstatement of mask mandates, or closure of schools and businesses. However, teams can incorporate inherent changes in population behavior in response to increasing or decreasing incidences (eg, changes in contacts or masking), if these changes were inferred from earlier phases of the pandemic and are already part of the model.
Database tracking of NPIs: teams may use their own data if desired, otherwise we recommend the following sources as a common starting point:
- Coronavirus Government Response Tracker | Blavatnik School of Government (ox.ac.uk)
- Coronavirus State Actions - National Governors Association (nga.org)
The mix of circulating strains at the start of the projection period is at the discretion of the teams based on their interpretation/analysis of the available data. Variation in initial prevalence between states is left at teams’ discretion.
Targets will be similar to Round 18 and consist of weekly state- and national-level COVID-19 hospitalizations and deaths (no case projections). Ascertainment of hospitalizations and deaths will proceed at the same level as they were at the start of the projection period. NHSN hospitalization will be used as the source of hospitalization data and NCHS will be used as the source of gold-standard death data. Note that NCHS data source counts deaths on the dates they occurred, not on the date they were reported. In accordance with the data, the death target should give deaths on the date they occur.
Both weekly state- and national-level COVID-19 hospitalizations and deaths should be provided for the following age groups: 0-64, 65+ and overall population
Whether or not to include demographic dynamics (aging, birth) is at the discretion of the teams.
All of the teams' specific assumptions should be documented in metadata and abstract.
Projection Time Horizon: We consider a one-year projection period.
| Scenario | Scenario name | Scenario ID for submission file ('scenario_id') |
|---|---|---|
| Scenario A. No booster, no immunization timing (counterfactual) | noBoo_noTiming | A-2025-04-01 |
| Scenario B. 65+ and high-risk booster, classic immunization timing | HighRiskVax_classicTiming | B-2025-04-01 |
| Scenario C. 65+ and high-risk booster, early immunization | HighRiskVax_earlyTiming | C-2025-04-01 |
| Scenario D. All booster, classic immunization timing | allVax_classicTiming | D-2025-04-01 |
| Scenario E. All booster, early immunization | allVax_earlyTiming | E-2025-04-01 |
-
Due date: Tue May 6, 2025
-
End date for fitting data: April 26, 2025 (no later than April 26, no earlier than April 20)
-
Start date for scenarios: April 27, 2025 (first date of simulated transmission/outcomes)
-
Simulation end date: April 25, 2026 (52-week horizon)
Submission Target
- Weekly Incident Deaths
- Weekly Incident Hospitalization
Other submission requirements
-
Simulation trajectories: We ask that teams submit a sample of 100 to 300 simulation replicates.
-
Trajectories will need to be paired across horizons, age groups, targets, and scenarios.
-
Projection quantiles for incident outcomes are optional but encouraged. Similarly projections of cumulative outcomes (either as quantiles or cumulative trajectories) are optional.
- Weekly incident deaths
- Weekly incident hospitalizations
- Weekly cumulative deaths since simulation start
- Weekly cumulative hospitalizations since simulation start
- For teams who wish to submit quantiles, the format is in accordance with prior rounds. We ask for the following quantiles: 0.01, 0.025, 0.05, every 5% to 0.95, 0.975, and 0.99. Mean is optional.
-
Weeks will follow epi-weeks (Sun-Sat) dated by the last day of the week
-
Submission file type: gz.parquet (from Apache Arrow) is now required. The submission file can be partitioned by "origin_date" and "target". For more information, please consult the associated README
-
Abstract: We require a brief abstract describing model assumptions and results, from all teams.
-
Metadata: We require a brief metadata form, from all teams.
Groups interested in participating can submit model projections for each scenario in a PARQUET file formatted according to our specifications, and a metadata file with a description of model information. See here for technical submission requirements.
The target-data/ folder contains the target data in a hubverse compliant time-series format.
The data are automatically updated on Monday morning. The code to generate the
data is available in the src folder.
The past version of the time-series files are stored in the
auxiliary-data/target-data_archive
folder, with the date the data was archived append to the filename.
National Center for Health Statistics (NCHS) Mortality Surveillance Data data for weekly incidence COVID-19 deaths extracted from the FluView Interactive - Mortality CDC dashboard will be used for incidence death target . These data are weekly and pertain to date of death, not report date.
Due to the delay and backfilling of these data, the cumulative death target will also start from the date of projection (instead of cumulative since pandemic start).
Weekly Hospital Respiratory Data (HRD) Metrics by Jurisdiction from the National Healthcare Safety Network (NHSN) will be used for incidence hospitalization target. The data are weekly.
The repository stores and updates additional data relevant to the COVID-19 modeling efforts in the auxiliary-data/ folder:
-
Vaccination Coverage: data on vaccination coverage that can be used for a specific round.
-
Reports: Reports from COVID-19 Scenario Modeling Hub rounds results. Each report contains an executive summary with key messages and results, and analyses of ensemble and individual projections.
-
Hospitalization: Data from local state authorities to complete NHSN data. Does not include all states.
-
Wastewater & Other Topic of Interest: List of source and link that might be of interest for COVID-19 modeling. The list contains link to cases, deaths, tests, vaccination, emergency department visit, variants, wastewater, demographics data.
-
Population and census data: National and State level name and fips code as used in the Hub and associated population size.
-
Rounds: Information on ongoing round and previous round available in the repository
For more information, please consult the associated README file.
We aim to combine model projections into an ensemble.
We are grateful to the teams who have generated these scenarios. The groups have made their public data available under different terms and licenses. You will find the licenses (when provided) within the model-specific metadata files in the model-metadata directory. Please consult these licenses before using these data to ensure that you follow the terms under which these data were released.
All source code that is specific to the overall project is available under an open-source MIT license. We note that this license does NOT cover model code from the various teams or model scenario data (available under specified licenses as described above).
Those teams interested in accessing additional computational power should contact Katriona Shea at [email protected].
Additional resources might be available from the MIDAS Coordination Center, please contact [email protected] for information.
Scenario modeling groups are supported through grants to the contributing investigators.
The Scenario Modeling Hub site is supported by the MIDAS Coordination Center, NIGMS Grant U24GM132013 (2019-2024) and R24GM153920 (2024-2029) to the University of Pittsburgh.
- Justin Lessler, University of North Carolina
- Katriona Shea, Penn State University
- Cécile Viboud, NIH Fogarty
- Shaun Truelove, Johns Hopkins University
- Claire Smith, Johns Hopkins University
- Emily Howerton, Penn State University
- Nick Reich, University of Massachussetts at Amherst
- Harry Hochheiser, University of Pittsburgh
- Michael Runge, USGS
- Lucie Contamin, University of Pittsburgh
- John Levander, University of Pittsburgh
- Jessi Espino, University of Pittsburgh
- Sara Loo, Johns Hopkins University
- Erica Carcelen, John Hopkins University
- Sung-mok Jung, University of North Carolina
- Samantha Bents, NIH Fogarty
- Katie Yan, Penn State University
- Wilbert Van Panhuis, University of Pittsburgh
- Jessica Kerr, University of Pittsburgh
- Luke Mullany, Johns Hopkins University
- Kaitlin Lovett, John Hopkins University
- Michelle Qin, Harvard University
- Tiffany Bogich, Penn State University
- Rebecca Borchering, Penn State University