Epidemic Trends and Rt: COVID-19, Influenza, and RSV at National, State, Local Level

An archive of time-varying reproductive number (Rt) estimates and epidemic trend categories.

CDC estimates epidemic trends, based on the time-varying reproductive number (Rt), for COVID-19, influenza, and RSV at the national, state, and health service area (HSA) levels. Estimates are released weekly at https://www.cdc.gov/cfa-modeling-and-forecasting/rt-estimates.

------- About Rt estimates ------- These estimates indicate whether infections are increasing (Rt > 1) or decreasing (Rt < 1). Rt is estimated from daily incident emergency department visits reported through the National Syndromic Surveillance Program.

------- Methods and resources ------- Read more about the hierarchical generalized additive models (HGAMs) used to generate these estimates in:

[1] Behind the Model: CDC’s Tools to Assess Epidemic Trends (https://www.cdc.gov/cfa-behind-the-model/php/data-research/rt-estimates/index.html) [2] Modeling Handbook chapter on Rt (https://www.cdc.gov/cfa-modeling-and-forecasting/modeling-handbook/mh-rt.html)

------- Geographic resolution ------- We release estimates at the national, state, and local levels. Local estimates are for health service areas (HSAs). An HSA is a geographic area that combines one or more counties that are next to each other to form a region that is relatively self-contained with respect to hospital care. CFA uses NCI modified HSAs, where all counties within an HSA are in the same state. We provide one row for each county, to support searchability and merging with other data sets. However, Rt estimates are at the HSA level, so for each county in an HSA, the estimated values are identical (for a given disease, origin_date, and target_date).

------- Dates, release cadence, and version history ------- Weekly releases include an Rt estimate for each location, and disease, across the most recent 8 weeks. The target_date column indicates the date, t, for which the Rt estimate was made.

The origin_date column indicates the vintage of the data used to generate all the estimates in the release (think of this as a version ID, release date, or as-of date). The origin_date is typically the final date available in the surveillance data at the time the estimates were generated. For example, if models were run using data available through Tuesday, 9/15, then the origin_date for every estimate in the run (across 8 weeks of target_dates) is 9/15.

Finally, the horizon is defined as the difference between the target_date and origin_date. A negative horizon implies the target_date falls in the observed period; the target_date had already passed at the time the estimate was generated.

Our date definitions and terminology follow standards used by the forecast hubverse (https://docs.hubverse.io/en/latest/user-guide/tasks.html#standard-task-id-variables).

------- How to work with this large dataset ------- This is a very large dataset.

Before downloading data, run a query to filter for the specific locations, diseases, and origin_dates (release versions) you want to access. To filter the data in your browser use the “Actions” dropdown menu to open the "Query” tool. If you’re a developer you can also run SoQL queries using the available SDKs, including for R and Python.

We do not recommend using the “Export” button, which will download every estimate we have ever published (including weekly releases for >800 HSAs, 50 states, DC, and the US overall, across three diseases).