Governance
Methodology
Everything needed to reproduce every figure on this site: the datasets, their vintages, the joins, the geography fixes we had to make, and the two sources we tried to obtain and could not. The last of those is the section worth reading.
- Datasets acquired
- 5Census births, CBP, CMS maternal health, CMS hospital general, Census boundaries
- Datasets attempted and not obtained
- 3NPDB PUF, AHRQ HCUP PSI, Census API
- Figures estimated to fill a gap
- 0and there will be none
- Counties fully joined
- 3,144100% geographic match, no fallbacks
Datasets acquired
| Source | What we take from it | Vintage | Known limits |
|---|---|---|---|
| Census Bureau — county population estimates, Vintage 2024 | Births by county, 2021–2024; county population | Vintage 2024, released 2025 | Estimates within an estimates series, revised each vintage. Births are by mother’s county of residence, not by place of delivery. Not the final NCHS natality file. |
| Census Bureau — County Business Patterns 2023 | Establishment counts, NAICS 541110 Offices of Lawyers | 2023, released May 2025 | Counts locations, not firms. Does not distinguish practice area. Employment is noise-infused at this level, so we do not publish it. |
| CMS Care Compare — Maternal Health · Hospital | Obstetric service status (SM-7) and the outcome-measure audit | 1 Jan – 31 Dec 2024 | Attestation, not audit. Covers only hospitals in CMS quality reporting. No freestanding birth centres. Outcome measures entirely withheld — see below. |
| CMS Care Compare — Hospital General Information | Ownership, facility type, Birthing Friendly designation | Current at retrieval | Designation reflects reported structure and participation, not outcomes. |
| Census Bureau — cartographic boundary files | County and state polygons | 2023, 1:20,000,000 | Generalised geometry. Territories excluded. Further simplified at build time — see the projection note below. |
Datasets we tried to get and did not
This is the section most methodology pages do not have, and the reason ours does is that a reader can check it in about five minutes.
| Source | Why it matters | What happened | Consequence on this site |
|---|---|---|---|
| NPDB Public Use File | Malpractice payment reports since 1990 by state and allegation group, including obstetrics. The highest-value dataset for this market. | Requires a request form carrying a named individual’s name, title and email, plus acceptance of a data use agreement. Not submitted on the operator’s behalf without his approval. | That page ships incomplete and says so. No substitute was used. |
| AHRQ HCUP State Inpatient Databases (Patient Safety Indicators, obstetric trauma) | The only route to hospital-level obstetric trauma indicators now that CMS has withheld its own | Licensed and priced per state-year. Not freely available. | Excluded entirely. Not approximated from any other source. |
| Census Data API (ACS and CBP endpoints) | Convenient per-variable pulls | Now returns "Missing Key". An API key is required. | Worked around using the bulk summary and CBP files, which require no key and are the same underlying data. |
The join, step by step
| # | Step | Detail |
|---|---|---|
| 1 | Base geography | All 3,144 county-equivalents at summary level 050 in the Census population estimates file. Puerto Rico is present in the source and excluded from every national figure on this site. |
| 2 | Births | Taken directly from the components-of-change columns for 2021 through 2024. No adjustment. |
| 3 | Legal establishments | County Business Patterns filtered to NAICS 541110 exactly — not the 5411 parent, which also contains title abstract and settlement offices — and joined on state plus county FIPS. |
| 4 | Counties absent from the NAICS 541110 file | Treated as zero establishments, which is what absence means in County Business Patterns. Not treated as suppressed. There are 854 of them. |
| 5 | Hospital obstetric status | From CMS measure SM-7. "Yes" and "No" both mean the hospital provides inpatient labour and delivery; "Not Applicable" means it does not; "Not Available" means unknown and is carried through as unknown everywhere. |
| 6 | Hospital-to-county match | By the facility’s own county string, normalised: accents stripped, all non-letters removed, and the words County, Parish, Borough, Census Area and Municipality dropped. |
| 7 | Residual mismatches | Resolved by an explicit alias table covering ten source typos and renamings — for example the file’s "North Slope Borouh", "Northwest Artic Borough" and "Jeffrson Davis", and Alaska’s 2019 split of Valdez-Cordova. |
| 8 | Connecticut | CMS still files against the eight counties Connecticut abolished in 2022; Census reports the nine COG planning regions. Hospitals were mapped from their town to the correct planning region. All 27 Connecticut facilities matched. |
| 9 | Territories | Facilities in PR, GU, VI, AS and MP excluded from all national figures, and the exclusion is stated rather than silent. |
| 10 | Final match rate | 100%. Every one of the 4,658 hospitals outside the territories resolved to a county. No fallback, no nearest-match, no dropped records. |
The projection
County and state boundaries are projected at build time and shipped as flat SVG paths. There is no mapping library on this site, no CDN request, and no runtime data fetch. The maps will still draw in five years.
| Element | Detail |
|---|---|
| Projection | Albers equal-area conic, composite, matching the standard d3.geoAlbersUsa parameters: lower 48 at parallels 29.5° and 45.5° rotated 96°, Alaska at 55° and 65° rotated 154° at 0.35 scale, Hawaii at 8° and 18° rotated 157°. |
| Layout | Scale 1300, translate (487.5, 305), logical viewBox 975 × 610. Alaska and Hawaii insets clipped to the standard frames. |
| Simplification | Ramer–Douglas–Peucker at a 0.3 pixel tolerance. At the published width one pixel is roughly 4.6 km on the ground, so the tolerance is below anything a reader can resolve. |
| Encoding | Integer relative-delta paths on a 2× grid — about a third of the bytes of absolute coordinates at sub-half-pixel accuracy. |
| Verification | Projected centroids checked against known positions for Los Angeles, New York, Cook, Harris, Miami-Dade, King, Anchorage and Honolulu before publication. |
The suppression rule
| Code | CMS definition | How we treat it |
|---|---|---|
| 4 | Data suppressed by CMS for one or more quarters. | Withheld. Hatched. Never coloured, never zero. |
| 5 | Results are not available for this reporting period. | Withheld. Hatched. |
| 19 | Data are shown only for hospitals that participate in the Inpatient Quality Reporting (IQR) and Outpatient Quality Reporting (OQR) programs. | Noted; affects which facilities appear, not their values. |
| 1 | The number of cases/patients is too few to report. | Withheld. Would be hatched if it appeared in the measures we use. |
Where we model, we say modelled — in the headline, not the footnote
Nothing in the intelligence layer is modelled. The only modelled figures on this site are in the tools, they are driven entirely by inputs you supply, and they are labelled modelled in the result panel itself. If you find a modelled figure on this site presented as a measurement, that is a correction and it goes in the log.
Sources
- U.S. Census Bureau — County Population Totals and Components of Change, Vintage 2024 — births by county, 2021–2024
- U.S. Census Bureau — County Business Patterns, 2023 — NAICS 541110, Offices of Lawyers; establishment counts by county
- CMS Care Compare — Maternal Health · Hospital — PC-02, PC-07a, PC-07b, SM-7; 2024 reporting period
- CMS Care Compare — Hospital General Information — ownership, facility type, Birthing Friendly designation
- CMS Care Compare — Footnote Crosswalk — definitions for suppression codes
- U.S. Census Bureau — Cartographic Boundary Files, 2023 (1:20,000,000) — county and state polygons; projected to Albers equal-area at build time
- National Practitioner Data Bank — Public Use File — malpractice payment reports; acquisition pending, see Methodology
The layer itself → · The standard → · The log → · Ask for a custom data cut →
Authorship & review
- Author
- Kevin Schwaner, Founder; Chief Marketing Officer. Credentials.
- Last reviewed
- Corrections
- Found an error? Tell us and it goes in the dated corrections log, whether or not it flatters us.