COVID-19 Hospital Admissions Forecast

National and state forecasts

Interpretation of forecasts

• Reported and forecasted new COVID-19 hospital admissions as of March 26, 2025. This week, 12 modeling groups contributed 13 forecasts that were eligible for inclusion in the ensemble forecasts for at least one jurisdiction. Contributing teams are listed below.

• Ensemble forecasts combine forecasts from all submitted models into one forecast. These forecasts have been among the most accurate for previous COVID-19 forecasting efforts but may not reliably predict rapid changes in disease trends.

• The figure shows the number of new laboratory-confirmed COVID-19 hospital admissions reported in the United States each week from August 8, 2020 through March 22, 2025 and forecasted new COVID-19 hospital admissions per week for this week and the next 2 weeks through April 12, 2025. The colored areas on the graph show prediction intervals, which indicate the bounds of uncertainty around the forecast estimates.  

• The forecasted hospital admission rates (for both median and prediction intervals) use Census Vintage 2023 population estimates as a population denominator for the state. We note that the denominator for rates may not capture the exact catchment populations of the hospitals in each state, but rates provide a rough adjustment for population size, allowing comparison across states with different population sizes.

• Hospitals are required to report daily laboratory-confirmed COVID-19 hospitalizations to the National Healthcare Safety Network (NHSN). Review Hospital Respiratory Data for additional details on this guidance. Historical data from NHSN are not shown from April 27-October 31, 2024, due to the pause in mandatory reporting.

• The following jurisdictions had <80% of hospitals reporting for the most recent week: Alaska, Arkansas, Arizona, District of Columbia, Georgia, Hawaii, Idaho, Iowa, Kansas, Maine, Massachusetts, Mississippi, Missouri, Montana, Nebraska, Nevada, New Jersey, New Mexico, Ohio, Oklahoma, Oregon, Pennsylvania, Puerto Rico, Texas, Vermont, and Wyoming. Lower reporting rates could impact forecast validity. Percent of hospitals reporting is calculated based on the number of active hospitals reporting complete data to NHSN for a given reporting week.

• Forecast submissions are collected and stored in the COVID-19 Forecast Hub GitHub repository, which also contains more detailed information about the forecast data and submission process.

Contributing teams and models

Individual model websites are linked where available.

• CEPH Lab at Indiana University (Model: Rtrend_covid)

• Carnegie Mellon Delphi Group (Model: TimeSeries)

• Center for Forecasting and Outbreak Analytics (CFA/CDC) Pyrenew team (Model: Pyrenew HE_COVID)

• Center for Systems Science and Engineering at Johns Hopkins University (Model: CSSE_Ensemble)

• Google Science AI (Model: Ensemble)

• Metaculus (Model: cp)

• MOBS Lab at Northeastern University (Model: GLEAM_COVID)

• MOBS Lab at Northeastern University and ISI Foundation (Model: AdaptiveEnsemble)

• One Health Trust and Johns Hopkins University (Model: nbxd)

• UGA_flucast (Model: INFLAenza)

• University of Michigan, Computer Science and Engineering (Model: DeepOutbreak)

• UMass-Amherst (Model: ar6_pooled)

• UMass-Amherst (Model: gbqr)

Models not included in the CovidHub ensemble

• Carnegie Mellon Delphi Group (Model: climate baseline)

• CovidHub-hub-models (Model: baseline)

• Center for Forecasting and Outbreak Analytics (CFA/CDC) Pyrenew team (Model: Pyrenew H_COVID)