NSSP Emergency Department Visits

Attribute Details
Source Name nssp
Data Source National Syndromic Surveillance Program (NSSP) via CDC Socrata
Geographic Levels county, hsa_nci, hrr, msa, state, hhs, census_division, census_region, nation
Temporal Granularity Week, ending Saturday
Reporting Cadence Weekly
Temporal Scope Start 2022-10-01
Temporal Scope End Ongoing
Extra Key Columns None
License Public Domain US Government

This source reproduces the legacy V4 nssp source with the same signal definitions plus an additional acute respiratory illness signal. V5 adds census division and census region levels, exposes the aggregation choice through fill_method, and serves revision history through the /archive/ and /snapshot/ endpoints. See Relationship to V4.

Table of contents

  1. Overview
  2. Indicators (Signals)
  3. Estimation
    1. Metric Definition
    2. Smoothing
    3. Temporal Handling
    4. Geographic Handling
  4. Relationship to V4
  5. Schema
    1. Columns
  6. Missingness & Privacy
  7. Limitations
  8. Lag & Backfill
  9. Source and Licensing

Overview

The National Syndromic Surveillance Program (NSSP) tracks the share of emergency department (ED) visits associated with respiratory illness across participating facilities in the United States. The CDC publishes weekly percentages for COVID-19, influenza, Respiratory syncytial virus (RSV), a combined category of the previous three diseases listed, and broader acute respiratory illness (ARI). Delphi ingests the CDC Socrata release, with the CDC forecast-hub GitHub mirror as a fallback when Socrata is unavailable.

Indicators (Signals)

Indicator Name Pathogen or Disease Metric Type Description
pct_ed_visits_covid COVID-19 Percentage Share of ED visits with a discharge diagnosis of COVID-19.
pct_ed_visits_influenza Influenza Percentage Share of ED visits with a discharge diagnosis of influenza.
pct_ed_visits_rsv RSV Percentage Share of ED visits with a discharge diagnosis of RSV.
pct_ed_visits_combined Combined Percentage Share of ED visits with a discharge diagnosis of COVID-19, influenza, or RSV.
pct_ed_visits_ari ARI Percentage Share of ED visits with a discharge diagnosis of acute respiratory illness.
smoothed_pct_ed_visits_covid COVID-19 Percentage, 3-week mean Trailing 3-week mean of pct_ed_visits_covid.
smoothed_pct_ed_visits_influenza Influenza Percentage, 3-week mean Trailing 3-week mean of pct_ed_visits_influenza.
smoothed_pct_ed_visits_rsv RSV Percentage, 3-week mean Trailing 3-week mean of pct_ed_visits_rsv.
smoothed_pct_ed_visits_combined Combined Percentage, 3-week mean Trailing 3-week mean of pct_ed_visits_combined.

Estimation

Metric Definition

Each raw signal represents the percentage of total emergency department (ED) visits diagnosed with the specified condition.

Smoothing

The smoothed_* signals are a trailing 3-week mean (the average of the reference week and the two preceding weeks) computed by the CDC and passed through unchanged.

Temporal Handling

Each value covers a 7-day week (Sunday through Saturday), labelled by its Saturday week-ending date.

Geographic Handling

The CDC reports values natively for the nation (nation), states (state), counties (county), and NCI-modified Health Service Areas (hsa_nci), which Delphi reads directly. Within the dataset, state values are aggregated by the CDC, and are served as published.

Delphi derives the remaining levels as a population-weighted mean of native values. State records aggregate into Department of Health and Human Services (HHS) regions, census divisions, and census regions. County records aggregate into Hospital Referral Regions (HRRs) and Metropolitan Statistical Areas (MSAs). For target geography \(g\) composed of sub-units \(c\) with 2020 US Census population \(w_c\) and reported percentage \(p_c\),

\[\hat{p}_g = \frac{\sum_{c \in g} w_c\, p_c}{\sum_{c \in g} w_c}.\]

Because values are percentages rather than counts, missing sub-units are handled explicitly through fill_method:

  • fill_zero: missing sub-units contribute \(p_c = 0\) while retaining their population weight in the denominator.
  • fill_ave: missing sub-units are excluded from both numerator and denominator.
  • source: native reported values.

Population-weighted aggregation is a standard spatial upscaling method used to aggregate rates across regional boundaries from survey and surveillance sub-units. For practical background on spatial aggregation choices in health surveillance, see Lee et al. (2022). For foundational spatial epidemiology principles, see Lawson (2006).

Relationship to V4

The V4 nssp source published ED visit percentages for nation, state, county, HSA, HRR, MSA, and HHS regions. The signal names that overlap remain the same in the V4 to V5 transition.

V4 Signal V5 Signal Notes
pct_ed_visits_covid pct_ed_visits_covid Exact match
pct_ed_visits_influenza pct_ed_visits_influenza Exact match
pct_ed_visits_rsv pct_ed_visits_rsv Exact match
pct_ed_visits_combined pct_ed_visits_combined Exact match
smoothed_pct_ed_visits_covid smoothed_pct_ed_visits_covid Exact match
smoothed_pct_ed_visits_influenza smoothed_pct_ed_visits_influenza Exact match
smoothed_pct_ed_visits_rsv smoothed_pct_ed_visits_rsv Exact match
smoothed_pct_ed_visits_combined smoothed_pct_ed_visits_combined Exact match
(not available) pct_ed_visits_ari New in V5

What changed in V5:

  • Expanded Geographies. V5 adds census division and census region levels (aggregated from state records).
  • Aggregation Choice. V4 published derived aggregations with a single fixed treatment of missing sub-units. V5 computes both methods and lets the caller choose via fill_method. See Geographic Handling.
  • New Signal. V5 adds pct_ed_visits_ari. The COVID-19, influenza, RSV, and combined pairs are unchanged.
  • Revision History. V5 exposes the full revision history through /archive/ and point-in-time reads through /snapshot/.

Schema

Columns

Column Key Type Data Type Description
signal Primary Key string The name of the requested indicator.
report_time Primary Key date The publication or release date (YYYY-MM-DD).
geo_type Primary Key string Geographic level (county, hsa_nci, hrr, msa, state, hhs, census_division, census_region, nation).
geo_value Primary Key string Unique code for the location (e.g., tx for Texas, 06001 for Alameda County).
fill_method Primary Key string Imputation method used during geographic aggregation (source, fill_zero, or fill_ave).
reference_time Primary Key date The date or surveillance period represented by the observation, labeled by Saturday week-ending date (YYYY-MM-DD).
value Value Column float The recorded percentage of ED visits.

Missingness & Privacy

The CDC suppresses facility and county values with low visit volumes to protect patient privacy. Suppressed values appear as missing in the raw feed and are treated as null during ingestion.

Unobserved values arise from uneven facility participation and state reporting restrictions. Nationwide, NSSP captures data from approximately 78 percent of emergency departments. Many rural counties contain no emergency departments or only non-participating facilities. Coverage gaps are most pronounced in California, Colorado, Missouri, Oklahoma, and Virginia.

State reporting operates independently from county reporting. Some facilities are included in state aggregates but withheld from local data. Fifteen states report no data at the county level: AK, AL, AR, AZ, CA, CO, CT, FL, MO, ND, NH, NJ, OH, SD, and WA.

At the state level, South Dakota, Missouri, and US territories have historically reported no data through NSSP. Missouri reports neither state nor county records, making it the only completely non-reporting state.

Wyoming facilities report literal zeros because of an administrative reporting artifact. During ingestion, Delphi converts zero values for Wyoming to null to prevent distortion of rates.

Missing sub-units during geographic aggregation are handled according to the selected fill_method, as described under Geographic Handling.

Limitations

Percentages reflect visits at facilities reporting to NSSP rather than all facilities in a region. Counties without reporting emergency departments have no native data.

County-level values published by the CDC are approximations inherited from their parent Health Service Area (HSA). The CDC defines these clusters using NCI-modified Health Service Areas. Because rates are calculated at the HSA level, every county within the same HSA receives an identical percentage.

Diagnoses reflect clinical diagnostic codes assigned at discharge without mandatory laboratory confirmation. Because emergency departments do not test every patient for respiratory viruses, infections can be missed and visit percentages may be biased downward.

Delphi aggregates local records into derived geographic levels using 2020 Census population weights. This weighting assumes that emergency department visit volumes scale proportionally with resident population across counties. In practice, urban areas tend to have higher per capita emergency department utilization, and rural residents frequently travel to urban centers for acute care. If urban residents seek emergency care at higher rates per capita, population weighting can overrepresent rural counties in derived regional values.

Low-population counties occasionally report outlier percentages such as 33 percent, 50 percent, or 100 percent. These spikes arise by chance from very small total visit counts in a given week rather than widespread transmission.

Lag & Backfill

The CDC publishes the weekly dataset on Friday mornings, covering the 7-day week that ended the previous Saturday. This creates an initial lag of 6 days between the end of the observation period and the initial release.

Historical weeks experience frequent backfill, primarily when newly enrolled emergency departments join the NSSP reporting network. When a new facility begins reporting, the CDC incorporates its historical data into the dataset.

Adding facility history alters past estimates for every geographic level that includes the facility, from county to nation. Because broader geographic levels aggregate data across many facilities, national, regional, and state series revise more frequently than local ones.

Revisions can alter historical series for up to 2 years after initial publication.

Source and Licensing

This dataset originates from the CDC National Syndromic Surveillance Program (NSSP) and is published under Public Domain U.S. Government terms.