Datasets:

EPI-Eval
/

flusight-forecast-hub

pretty_name: FluSight Forecast Hub — flu hospital admission target
license: cc0-1.0
size_categories:
  - 1K<n<10K
tags:
  - cadence-weekly
  - geo-us
  - surveillance-respiratory
  - pathogen-influenza
  - tier-1
  - availability-open
schema_version: '0.1'
source_id: flusight-forecast-hub
source_url: https://github.com/cdcepi/FluSight-forecast-hub
manifest_section: §4.1
surveillance_category: respiratory
pathogens:
  - influenza
availability: open
availability_notes: null
access_type: github
tier: 1
cadence: weekly
geography_levels:
  - national
  - subnational-state
geography_countries:
  - US
gold_standard_for: []
vintaged_version_of: null
succeeds: null
derived_from:
  - nhsn-hrd
value_columns:
  - name: value
    unit: admissions/week
    value_type: incident
    description: >-
      Confirmed flu hospital admissions for the epiweek (FluSight scoring
      target)
    aggregation: sum
  - name: weekly_rate
    unit: admissions/week per 100k
    value_type: rate
    description: Weekly admission rate per 100k population
    aggregation: rate
notes:
  extra_columns:
    - column: location_name
      description: >-
        State name as published by the hub. `location_id` is FIPS for states /
        `US` for national.
  interpretation_caveats:
    - column: value
      caveat: >
        This is the *target truth* file the hub uses for scoring. It is derived
        from NHSN HRD

        (`derived_from`) and may lag NHSN by one cycle while the hub freezes the
        truth value.

        For the live admissions series use `nhsn-hrd.totalconfflunewadm`
        directly.
  general: >
    This source represents only the **truth** half of the hub. Model forecasts
    live in

    `model-output/` in the same repo and will become a separate
    `flusight-forecast-hub-models`

    dataset (forecasts are a different shape — one row per (team, model,
    reference_date,

    target, horizon, location, output_type, value) — and need a separate schema
    treatment).
extra:
  hub_format_note: Hubverse — see https://hubverse.io/
  forecast_models_separate: true
computed:
  last_ingested: '2026-04-26T12:48:54Z'
  row_count: 11660
  time_coverage:
    - start: '2022-02-05'
      end: '2026-04-18'
  geography_unit_count: 53
  observed_cadence_days: 7
  missing_gaps: []
  data_hash: 1364dce3a6023406

FluSight Forecast Hub — flu hospital admission target

This source represents only the truth half of the hub. Model forecasts live in model-output/ in the same repo and will become a separate flusight-forecast-hub-models dataset (forecasts are a different shape — one row per (team, model, reference_date, target, horizon, location, output_type, value) — and need a separate schema treatment).

Source: https://github.com/cdcepi/FluSight-forecast-hub

Coverage

Time: 2022-02-05 → 2026-04-18
Cadence: weekly (observed median spacing: 7 days)
Geography levels: national, subnational-state — 53 unique location IDs
Countries: US
Pathogens: influenza
Surveillance category: respiratory
Rows: 11,660

Columns

Column	Unit	value_type	Aggregation	Description
`value`	admissions/week	`incident`	`sum`	Confirmed flu hospital admissions for the epiweek (FluSight scoring target)
`weekly_rate`	admissions/week per 100k	`rate`	`rate`	Weekly admission rate per 100k population

Additional data columns

location_name — State name as published by the hub. location_id is FIPS for states / US for national.

Interpretation caveats

Things that may differ from how other sources define a similar measure. If you're combining this dataset with another, read these first.

value — This is the target truth file the hub uses for scoring. It is derived from NHSN HRD (derived_from) and may lag NHSN by one cycle while the hub freezes the truth value. For the live admissions series use nhsn-hrd.totalconfflunewadm directly.

Related datasets in EPI-Eval

Derived from: nhsn-hrd

Access

Availability: open
Access type: github
License: cc0-1.0
Tier: 1

Schema version 0.1 · Last ingested 2026-04-26T12:48:54Z · source_id: flusight-forecast-hub · Manifest section §4.1