Updated 2024/11/20 14:50
The US RSV Forecast Hub is as a collaborative forecasting effort to produce weekly short-term forecasts of weekly laboratory-confirmed RSV hospital admissions for the US and individual states. Each week, participants are asked to provide national- and jurisdiction-specific probabilistic forecasts of the weekly number of confirmed RSV hospitalizations for the following four weeks. The US RSV Forecast Hub is open to any team willing to provide projections at the right temporal and spatial scales. 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.
Those interested in participating, please see the README file in the Github repository.
If you are interested in longer-term scenario projections of RSV in the US, please take a look at the US RSV Scenario Modeling Hub.
RSV hospital admission forecast for the United States (national-level) for all age groups (0-130yr), weekly incident admissions:
Note: The dashed red line represents the peak of national RSV hospitalizations during the 2023-24 season.
State-level forecasts
National-level forecasts by Age
Evaluation will begin 4 weeks after the first model submissions.
Previous weeks’ forecasts are available in the archive and can be accessed in the links below. These forecasts are not updated with updated ground truth data, thus the ground truth data may differ between them as data are back-filled.
The Respiratory Syncytial Virus Hospitalization Surveillance Network (RSV-NET) is a network that conducts active, population-based surveillance for laboratory-confirmed RSV-associated hospitalizations in children younger than 18 years of age and adults. The network currently includes 58 counties in 12 states that participate in the Emerging Infections Program (California, Colorado, Connecticut, Georgia, Maryland, Minnesota, New Mexico, New York, Oregon, and Tennessee) or the Influenza Hospitalization Surveillance Program (Michigan and Utah). Age- and state-specific data on laboratory-confirmed RSV hospitalization rates are available for 12 states and the US from RSV-NET spanning 2017-18 to present (RSV-NET CDC Webpage). Age-specific weekly rates per 100,000 population are reported in this system.
The data has been standardized and posted on the rsv-forecast-hub GitHub target-data/ folder and is updated weekly. The target in this data is the weekly number of hospitalizations in each given state (inc_hosp variable), for all ages and for each age group. To obtain counts, we have converted RSV-NET weekly rates based on state population sizes. This method assumes that RSV-NET hospitals are representative of the whole state. To obtain national US counts, we have used the rates provided for the “overall RSV-NET network”. The data covers 2017-present. Reported age groups include: [0-6 months], [6-12 months], [1-2 yr], [2-4 yr], [5-17 yr], [18-49 yr], [50-64 yr], and 65+ years. The standardized dataset includes week-, state-, and age-specific RSV counts (the target), rates, and population sizes.
| Team | Model | Brief Description |
|---|---|---|
| CEPH Lab at Indiana University | Rtrend RSV | A renewal equation method based on Bayesian estimation of Rt from past hospitalization data. |
| Columbia University | RSV_SVIRS | Age-structured SVIRS model coupled with Ensemble Adjustment Kalman Filter. Inputs CDC RSV-NET, POLYMOD contact matrix, and seasonal absolute humidity |
| The Center for Systems Science and Engineering at Johns Hopkins University | CSSE Multi-pathogen Model for RSV | A data-driven multipathogen modeling approach to forecast RSV hospitalizations |
| Predictive Science Inc. | Package for Respiratory Disease Open-source Forecasting | We fit and extrapolate an age-stratified SIR compartmental model with four levels of natural immunity as well as infant/elder vaccination. |
| UGA_flucast | INFLAenza | A spatial time-series model that uses the R-INLA package for estimating forecast posterior distributions. |
| University of Michigan, Computer Science and Engineering | DeepOutbreak | Deep neural network model with conformal predictions. |