NCHS Mortality Data

Attribute	Details
Source Name	`nchs_mortality`
Data Source	National Center for Health Statistics (NCHS) via data.cdc.gov Socrata dataset `r8kw-7aab`
Geographic Levels	`state`, `nation`
Temporal Granularity	Week, ending Saturday
Reporting Cadence	Weekly
Temporal Scope Start	2020-01-18 (2020w02 for nation, 2020w05 for state)
Temporal Scope End	Ongoing
Extra Key Columns	None
License	NCHS Data Use Agreement

This source reproduces the legacy V4 nchs-mortality source. All signal names and metrics are preserved, with the source identifier standardized from nchs-mortality to nchs_mortality. See Relationship to V4.

Overview
Indicators (Signals)
Estimation
Relationship to V4
Schema
1. Columns
Missingness & Privacy
Limitations
Lag & Backfill
Source and Licensing

Overview

The National Center for Health Statistics (NCHS) maintains the National Vital Statistics System (NVSS), which collects and tabulates death certificate data across all 50 states, the District of Columbia, and U.S. territories. Delphi ingests the provisional weekly death data published on data.cdc.gov, tracking deaths associated with COVID-19, influenza, and pneumonia, as well as total all-cause mortality and comparisons against historical baseline levels.

Unlike surveillance data that tabulates cases or deaths on the date they are announced by public health agencies, NCHS mortality counts are organized by the actual date of death occurrence. Because vital registration and medical certification take time to complete, provisional figures for recent weeks undergo continuous retrospective revision as additional death certificates are processed.

Indicators (Signals)

Indicator Name	Pathogen or Disease	Metric Type	Description
`deaths_covid_incidence_num`	COVID-19	Count	Weekly count of deaths involving confirmed or presumed COVID-19.
`deaths_covid_incidence_prop`	COVID-19	Rate per 100k	Weekly rate of deaths involving confirmed or presumed COVID-19, per 100,000 population.
`deaths_allcause_incidence_num`	All Causes	Count	Weekly count of deaths from all causes.
`deaths_allcause_incidence_prop`	All Causes	Rate per 100k	Weekly rate of deaths from all causes, per 100,000 population.
`deaths_flu_incidence_num`	Influenza	Count	Weekly count of deaths involving influenza, including those where pneumonia or COVID-19 is also listed.
`deaths_flu_incidence_prop`	Influenza	Rate per 100k	Weekly rate of deaths involving influenza, per 100,000 population.
`deaths_pneumonia_notflu_incidence_num`	Pneumonia	Count	Weekly count of deaths involving pneumonia, excluding influenza.
`deaths_pneumonia_notflu_incidence_prop`	Pneumonia	Rate per 100k	Weekly rate of deaths involving pneumonia, excluding influenza, per 100,000 population.
`deaths_covid_and_pneumonia_notflu_incidence_num`	COVID-19, Pneumonia	Count	Weekly count of deaths involving both COVID-19 and pneumonia, excluding influenza.
`deaths_covid_and_pneumonia_notflu_incidence_prop`	COVID-19, Pneumonia	Rate per 100k	Weekly rate of deaths involving both COVID-19 and pneumonia, excluding influenza, per 100,000 population.
`deaths_pneumonia_or_flu_or_covid_incidence_num`	COVID-19, Flu, Pneumonia	Count	Weekly count of deaths involving pneumonia, influenza, or COVID-19 (PIC).
`deaths_pneumonia_or_flu_or_covid_incidence_prop`	COVID-19, Flu, Pneumonia	Rate per 100k	Weekly rate of deaths involving pneumonia, influenza, or COVID-19 (PIC), per 100,000 population.
`deaths_percent_of_expected`	All Causes	Percentage	Weekly all-cause deaths expressed as a percentage of the average death count across the same week in 2017–2019.

Estimation

Metric Definition

Delphi ingests provisional weekly death counts from NCHS. Cause-specific counts include any death where the disease is listed on the death certificate, either as the primary cause or as a contributing factor.

For location \(i\) and week \(t\):

Counts (*_num) are taken directly from NCHS without adjustment.
Rates (*_prop) divide the weekly count \(Y_{it}\) by location population \(P_i\) per 100,000 people:
\[\text{Rate}_{it} = 100{,}000 \cdot \frac{Y_{it}}{P_i}\]
Percent of expected (deaths_percent_of_expected) compares weekly all-cause deaths to the 2017–2019 baseline average \(E_{it}\):
\[\text{Percent}_{it} = 100 \cdot \frac{Y_{it}}{E_{it}}\]

Smoothing

All signals are unsmoothed. Values represent discrete 7-day weekly totals without moving averages.

Uncertainty

These indicators are based on complete vital registration records rather than sample surveys.

Temporal Handling

Observations cover 7-day epidemiological weeks defined by the CDC’s Morbidity and Mortality Weekly Report (MMWR). An MMWR week runs from Sunday through Saturday, standardizing public health reporting across calendar years.

reference_time is the Saturday week-ending date of the surveillance week when deaths occurred.
report_time is the release date when Delphi ingested the weekly NCHS snapshot. Because mortality data undergoes substantial backfill, each new report date revises counts for earlier weeks.

Geographic Handling

Data is published at two geographic resolutions:

Nation (nation). National totals are labelled as us.
State (state). Standard two-letter postal codes. New York City is reported separately in raw NCHS data, but Delphi aggregates New York City counts into New York State (ny) before calculating rates.

Because Delphi ingests counts directly without geographic imputation, fill_method is always source.

Relationship to V4

This source reproduces the legacy V4 nchs-mortality data source. All 13 signal names are retained with identical definitions.

V4 Signal	V5 Signal	Notes
`deaths_covid_incidence_num`	`deaths_covid_incidence_num`	Exact match
`deaths_covid_incidence_prop`	`deaths_covid_incidence_prop`	Exact match
`deaths_allcause_incidence_num`	`deaths_allcause_incidence_num`	Exact match
`deaths_allcause_incidence_prop`	`deaths_allcause_incidence_prop`	Exact match
`deaths_flu_incidence_num`	`deaths_flu_incidence_num`	Exact match
`deaths_flu_incidence_prop`	`deaths_flu_incidence_prop`	Exact match
`deaths_pneumonia_notflu_incidence_num`	`deaths_pneumonia_notflu_incidence_num`	Exact match
`deaths_pneumonia_notflu_incidence_prop`	`deaths_pneumonia_notflu_incidence_prop`	Exact match
`deaths_covid_and_pneumonia_notflu_incidence_num`	`deaths_covid_and_pneumonia_notflu_incidence_num`	Exact match
`deaths_covid_and_pneumonia_notflu_incidence_prop`	`deaths_covid_and_pneumonia_notflu_incidence_prop`	Exact match
`deaths_pneumonia_or_flu_or_covid_incidence_num`	`deaths_pneumonia_or_flu_or_covid_incidence_num`	Exact match
`deaths_pneumonia_or_flu_or_covid_incidence_prop`	`deaths_pneumonia_or_flu_or_covid_incidence_prop`	Exact match
`deaths_percent_of_expected`	`deaths_percent_of_expected`	Exact match

What changed in V5:

Source Identifier. Renamed from nchs-mortality (hyphenated) to nchs_mortality (underscore) to match V5 source naming conventions.
Revision History. Users can query past releases using snapshot_date or the /archive/ endpoint.

Schema

Columns

Column	Key Type	Data Type	Description
`signal`	Primary Key	string	The name of the requested indicator.
`report_time`	Primary Key	date	The release date on which Delphi ingested the NCHS publication (`YYYY-MM-DD`).
`geo_type`	Primary Key	string	Geographic level (`state` or `nation`).
`geo_value`	Primary Key	string	Location code (`us` for national, 2-letter state code).
`fill_method`	Primary Key	string	Always `source`.
`reference_time`	Primary Key	date	The Saturday week-ending date of the surveillance week (`YYYY-MM-DD`).
`value`	Value Column	float	The recorded measurement.

Missingness & Privacy

NCHS suppresses values under two conditions:

Counts between 1 and 9 are suppressed to protect privacy. These values appear as null.
Weeks where provisional death counts fall below 50% of historical expected deaths are withheld due to low completeness.

Delphi does not impute suppressed or missing values.

Limitations

Deaths are tabulated by date of death rather than date of report. Recent weeks undercount actual mortality until all death certificates are submitted and processed.
Reporting speed varies by state. Some jurisdictions submit records within days, while others take several weeks. Comparisons across states during the most recent weeks often reflect reporting delays rather than true differences in mortality.
New York City is included in New York State totals and cannot be queried separately.
Deaths pending investigation or toxicology results can cause delayed ICD-10 coding.

Lag & Backfill

Completing, submitting, and coding death certificates creates delays between when a death occurs and when it appears in the data:

Initial data for a given week is published 11 to 17 days after the week ends.
Counts for past weeks increase over time as additional certificates are processed. Provisional all-cause counts typically take 6 to 8 weeks to reach 99% completeness (Spencer et al., 2022). Specific causes may take longer.
Weekly updates revise past weeks. Historical revisions are available through the /archive/ endpoint or by querying with snapshot_date.

Source and Licensing

This data was originally published by the National Center for Health Statistics (NCHS) on data.cdc.gov, and is made available here as a convenience to the forecasting community under the terms of the original license. The NCHS places restrictions on how this dataset may be used: you may not attempt to identify any individual included in the data, whether by itself or through linking to other individually identifiable data; you may only use the dataset for statistical reporting and analysis. The full text of the NCHS Data Use Agreement is available on their website.