PlainHealth

A Complete 19-Year Grid: 9,690 State-Cause-Year Mortality Records (1999-2017)

PlainHealth's mortality table is a complete grid: 51 jurisdictions, 10 leading causes of death, and 19 years (1999-2017) from CDC NCHS, with deaths and age-adjusted rates filled in every one of the 9,690 cells.

Research period:

Research Question

How complete is PlainHealth's CDC mortality dataset — how many years, how many causes, how many state-cause-year records, and how many cells are actually filled?

Methodology

We queried PlainHealth's mortality, states, and causes_of_death tables to count distinct state_fips values, distinct icd_chapter values, the MIN/MAX year, and the total row count. We then compared the observed row count against the theoretical grid (states x causes x years) to measure completeness.

Findings

9,690 records, zero gaps across 51 jurisdictions, 10 causes, 19 years

The mortality table holds 9,690 state-cause-year records drawn from the CDC NCHS Leading Causes of Death series.CDC NCHS Leading Causes of Death: United States, Socrata dataset bi63-dtpu, 1999-2017 That figure is not an estimate. It is exactly 51 jurisdictions (the 50 states plus the District of Columbia) times 10 cause chapters times 19 years, so every cell in the grid is filled. There is no suppression and no missing year: each record carries both a death count and an age-adjusted rate.CDC WONDER Online Database, Underlying Cause of Death, 1999-2017

State pages at /states/ read directly from this grid, and cause pages at /causes/ rank all 51 jurisdictions on a single cause. Because the grid is complete, a comparison like Oklahoma's 2017 age-adjusted heart-disease rate of 237.2 per 100,000 against Alaska's far lower figure rests on the same denominator everywhere, with no holes to interpolate around.

What are the 10 causes the dataset tracks?

The causes_of_death table mirrors the CDC NCHS leading-cause groupings exactly: Heart disease (I00-I09), Cancer (C00-C97), Unintentional injuries (V01-X59), Chronic lower respiratory disease (J40-J47), Stroke (I60-I69), Alzheimer's disease (G30), Diabetes (E10-E14), Influenza and pneumonia (J09-J18), Kidney disease (N00-N07), and Suicide (X60-X84).CDC NCHS Leading Causes of Death: United States, Socrata dataset bi63-dtpu, 1999-2017 Together these accounted for 2,081,531 deaths in 2017 — heart disease and cancer alone made up roughly six in ten of them.

  • Heart disease (I00-I09) and Cancer (C00-C97) are the two largest chapters every year in the series.
  • Unintentional injuries (V01-X59) is the leading external cause and rises sharply over the period.
  • Suicide (X60-X84) is the smallest tracked chapter nationally but varies widely between states.
  • Every chapter is present for all 51 jurisdictions in all 19 years — the completeness is what makes year-over-year and state-to-state trends trustworthy.

Why does the series begin in 1999 and end in 2017?

The start year is a coding boundary. The United States adopted ICD-10 for death certificates in 1999, and earlier ICD-9 data classifies disease differently enough that splicing the two together would distort trends without NCHS comparability ratios. The end year reflects the CDC NCHS Leading Causes of Death series itself, which publishes this clean, age-adjusted, state-level grid through 2017.CDC National Center for Health Statistics, National Vital Statistics System Provisional later years exist but lack the complete age-adjusted rates this dataset depends on, so they are excluded rather than shown half-filled.

The result is a dataset whose scope is easy to state and easy to verify: 1999 to 2017, the 50 states and DC, the 10 leading causes, and a death count plus an age-adjusted rate in every one of the 9,690 cells.

What this analysis cannot tell us

PlainHealth tracks the CDC NCHS Leading Causes of Death series, which covers the 10 leading causes at the state level from 1999 to 2017. Years before 1999 use ICD-9 codes that are not directly comparable to the ICD-10 era and are excluded. Final CDC mortality data lags the calendar by roughly two years, and provisional later years were excluded because they lacked complete age-adjusted rates; extending the series past 2017 with comparable rates is planned. CDC data comes from death-certificate records, where cause-of-death miscoding is estimated at 1-5% in major CDC audits.

Discussion: the ICD-10 boundary at 1999 and why completeness matters

The series begins in 1999 for a concrete reason: that is the first year the Centers for Disease Control and Prevention NCHS adopted the World Health Organization's ICD-10 coding standard for US death certificates. Pre-1999 NCHS data uses ICD-9, and the two systems classify diseases differently enough that direct comparison across the 1998-1999 boundary requires NCHS-published comparability ratios — typically a chapter-specific factor between 0.93 and 1.07. Anyone extending the series back into the ICD-9 era should consult the NCHS comparability study series (Vital and Health Statistics Series 2, Number 130) before drawing trend conclusions.

The 10-cause framework — heart disease (I00-I09), cancer (C00-C97), unintentional injuries (V01-X59), chronic lower respiratory disease (J40-J47), stroke (I60-I69), Alzheimer's (G30), diabetes (E10-E14), influenza and pneumonia (J09-J18), kidney disease (N00-N07), and suicide (X60-X84) — is the NCHS "leading causes" rollup. It is coarser than the full ICD-10 code set, but that is the point: a small, stable set of chapters minimizes coding-practice variation between states and makes cross-state comparison defensible.

The 51-jurisdiction count is the 50 states plus DC, matching NCHS's state-level publication practice; US territories file records separately and are not directly comparable, so they are excluded. Because every state, cause, and year is present, the 19-year window can be read as a continuous trend rather than a patchwork — it spans the late-1990s baseline, the 2000s rise in chronic disease, and the run-up to the opioid era, without the gaps that force interpolation.

Glossary of NCHS Mortality Coverage Terms

National Vital Statistics System (NVSS)
The federal-state cooperative program through which NCHS receives and processes death, birth, fetal death, and marriage records from US state vital records jurisdictions. Coverage extends back to 1900 for births and 1933 for deaths.
Comparability Ratio
An NCHS-published multiplier that adjusts pre-1999 ICD-9-coded mortality counts to approximate equivalent ICD-10 codes, allowing limited time-series analysis across the coding-system boundary. Heart disease has a comparability ratio of approximately 0.99; suicide approximately 1.04.
Provisional Estimate
NCHS mortality data released within 18 months of the calendar year, before all late-arriving death certificates are processed. Provisional rates typically shift by less than 0.5% between provisional and final release.
Suppression Rule
The CDC WONDER convention that masks subnational mortality cells with fewer than 10 deaths in a year-cause-geography combination. It does not affect this dataset: at the level of 10 broad leading-cause chapters by state, every cell clears the threshold, so PlainHealth's grid is fully populated.
Leading Causes Rollup
The 10-chapter cause-of-death categorization used in NCHS's annual leading-causes publication, designed to minimize state-to-state variation in detailed code selection while still capturing the major mortality drivers.
Crude vs Age-Adjusted Rate
Crude rates divide deaths by population without adjustment, reflecting actual mortality experience but conflating age effects. Age-adjusted rates standardize to the 2000 US standard population to enable fair geographic comparison.

Leading causes of death by total deaths, 2017

National deaths across the 10 CDC NCHS leading-cause chapters

1. Heart disease647,4572. Cancer599,1083. Unintentional injuries169,9364. CLRD160,2015. Stroke146,3836. Alzheimer's121,4047. Diabetes83,5648. Influenza/pneumonia55,6729. Kidney disease50,63310. Suicide47,173

National deaths across the 10 leading causes, 1999-2017

Total tracked deaths per year, all 51 jurisdictions combined

19991,905,82620031,912,11520071,846,30120111,869,32120152,013,01720172,081,531

Related data

Sources