Network adequacy is asserted in a compliance table and experienced as a drive. Nobody publishes the difference. This app measures it — on the real road network, for a named service line, against the standard that actually applies in that county.
Network-adequacy software is excellent, mature and correctly aimed — at the plan being certified. It answers does this network pass?
But there is a second party in every one of those filings. A county health officer answering an exception request. A health centre deciding where the next site goes. A legislative staffer preparing a closure hearing. They need a different sentence: how many of my residents are outside, and where are they?
Today that question is answered by a PDF, a rumour, or nothing. There is no public instrument that reconciles the compliance table with the drive.
That is the certification layer, it works, and we do not compete with it. We serve the people on the other side of the table — and the plan's own analyst doing a pre-check.
Access analysis has a gold-standard method — the two-step floating catchment area and its enhanced form, which adds distance decay inside the catchment. The literature is blunt about where it breaks: uniform access inside the catchment, fixed thresholds that create abrupt discontinuities, edge effects, and catchment rules that transfer badly between dense and sparse geographies.
Every one of those failures has the same shape as the regulatory failure: a single threshold, set once, quietly deciding everything downstream — and a population origin nobody re-examines.
So put both in the navigation. Make the threshold a rule you can drag. Make the origin policy a switch you can flip. And show the population sitting on each side.
Census tract interior points are geometric, not population-weighted. In a 449 square-mile rural tract, that point sat two and a half hours' drive from the town where everyone in it actually lives.
Measure from it and you manufacture a desert. Measure from the town and the same people are three minutes from a hospital.
Same county. Same standard. Same router. Same hour.
The protagonist is not a map. It is a coverage curve: how many people are still without access at every drive time. The statutory standard is a rule drawn across it, and the shaded mass to the right of that rule is the deficit.
Drag the rule and the count, the map and the cohort list move together. Every place on the wrong side is named, with its population, its uninsured rate and its nearest open site.
And it draws two lines. Every framework applies a geographic test and a population-to-provider ratio test; the solid line is travel alone, the dashed line adds staffing. The gap between them is the population close enough to a clinic that hasn't got enough clinicians.
Two more pages carry the rest: a place dossier with designation status and the measured route, and a closure page — the question a California county board is actually asking in 2026.
We ran the same county, the same standard, the same router and the same hour — changing only where we decided the population lives.
The app defaults to the honest policy, prints on every origin which rule placed it, and lets you flip to the bad one on purpose — because a number you cannot stress is not evidence. The staffing gate is stressed the same way: statewide, 111 of 542 service areas fail a 3,500:1 ratio, holding 3.6 million Californians.
10,961 licensed California facilities with 18 real service levels and five emergency-department levels — enough to reason about a service line, not a dot.
9,106 tracts on California's own Medical Service Study Area geography, plus 9,070 CDC PLACES tracts carrying the uninsured rate.
268 designated state shortage areas and 180 federal HPSAs — each publishing its population-per-provider ratio, which is the second of the two tests.
A keyless road-network router returning duration and distance — the two units every US standard is written in — CORS-open, no key.
Here is the finding nobody expects. We role-tagged all 2,339 California ArcGIS services in the state's own catalogue for access to care. It returned 8 care-site services — all of them emergency-management views of health — and zero health shortage designations. The state publishes health as an asset to protect, not as a service to reach.
Everything this app runs on is external: the health department, the federal workforce bureau, and the CDC. That is the finding, not a gap in the search.
They are geometric, not population-weighted. From the interior point of one Trinity tract the hospital is 152.5 minutes away. From the town inside that tract it is 1.7 minutes. The tract holds 3,879 people.
Sort them and California's largest census tract comes back as 99 sq mi. It is 982. A numeric filter doesn't mis-sort — it errors outright.
542 shortage-area IDs, 366 current geography IDs, 6 exact matches. The 2024 redefinition renumbered everything and the 2020 designation layer was never re-cut. Join spatially, print both vintages.
The only mapped birthing flag sits on a licence snapshot whose newest record expires in 2021 — and it tracks the facility, not the service line. At least 56 California maternity wards have closed since 2012.
Thirty service areas publish a population-per-provider ratio of 0. That does not mean excellent staffing — it means no providers at all. A naïve "ratio below the threshold passes" scores every one of them as fine.
The obvious poverty layer stores its percentages as integers while the values are 0–1 proportions — so they all truncate. Statewide the column averages 0.0076: every census tract in California reports 0% poverty.
Each is reproduced as a live assertion in the test suite, so the app cannot silently regress into any of them.
California publishes time-and-distance standards and grants exceptions to them at scale. A legal-aid analysis found the state approved nearly 10,000 alternative access standard requests in a single month, raising the required provider distance by an average of 31.61 miles.
Paediatric providers accounted for roughly three-quarters of approvals. Low-income areas received more approvals, with larger distance increases, than high-income ones.
So the app carries the exception as a switch beside the rule. Flip it and watch what happens to a county's deficit — without a single resident moving any closer to care.
And beside it, the question that finding really asks: who bears it? The app compares the county against the people outside it — poverty share, hardship, ethnicity, uninsurance, population-weighted — so a county can test that pattern on its own residents instead of citing someone else's study.
Every one of those 4,332 residents is reclassified covered at exactly the same drive time. Hayfork is still 39.5 minutes from a clinic. The paperwork changed; the road did not.
There is no machine-readable register of these approvals, so the app labels the switch a scenario — never a fact — unless you supply the filing.
| The category today | Who's Outside | |
|---|---|---|
| The question | Does this plan's network pass certification? | How many of these residents are outside, and where are they? |
| The customer | The plan being certified | The county, the health centre, the advocate — and the plan's own pre-check |
| The threshold | Applied, then reported | Drawn on the distribution, and draggable |
| The exception | Filed | Rendered beside the rule, as a labelled scenario |
| Deployment | SaaS, licensed data, per-seat | MIT, on-prem or sovereign, runs on Strata or ArcGIS |
A network-adequacy certification. A provider directory. Appointment availability or wait times. An ambulance response model. A legal determination. And it does not know which hospital still runs a service line — because nobody's open data does.
Its source, its vintage, and its assumption. Travel is car-only and free-flow, and the app says "a car is assumed" on the face of it. Model-based estimates ship with their confidence interval. If a county is too large to measure whole, it prints exactly how many people were left out.
And it is operable: this ends up in a public hearing, so the threshold works from the keyboard, the headline is announced to a screen reader, and no verdict is carried by colour alone — WCAG 2.2 AA.
Built by describing it — no code written by hand. The same recipe re-targets to any health system with a population layer, a facility register and a road network.