Run this for your condition, your state, your people

1 · Another condition

One object in data/conditions.json, pointing at a row that already exists in the price table. No figure lives in the conditions file, so adding a condition cannot introduce a number nobody checked.

# 1. find or add the published figure, in data/prices.json
#    a row needs the file it came from, the year, the population it describes
#    and its basis. Nothing else may carry a dollar amount.

# 2. prove the row reproduces from the file it cites
python3 data/verify_price_table.py         # exits 1 if your row does not

# 3. name the condition, in data/conditions.json
{
  "id": "sickle-cell",
  "label": "Sickle cell disease",
  "icd10cm": "D57.1",
  "icd10cm_source": "CDC/NCHS, ICD-10-CM code descriptions, FY2026",
  "price_row_id": "your-new-row-id",     # or null, and the product says so
  "figure_kind": "condition_attributed", # or "excess" — they are not the same number
  "note": "…"
}

# 4. nothing else. The panel renders whatever that row says.

When there is no published figure. Set price_row_id to null. The product then says, in its own voice, that no federal source publishes an annual figure for that condition, and sends the person to the gap so the absence is counted. An absence you count is data. An absence you fill with the nearest number is a fabrication.

2 · Another state

Nothing to edit. All 109 CMS payment localities ship with the tool, so a person in any state sees their own locality’s figure and the drawer shows the three geographic indices that produced it. What you change is what you build on top: hand the API your state and it reprices every line without the interface.

# every state is already priced. To check one:
grep '^cms-99213,.*,TX-31,' public/data/locality-prices.csv

# to add a service, add a row to data/prices.json with its CPT/HCPCS code,
# then re-derive every locality figure from the CMS files:
python3 data/build_state_prices.py <dir-with-PPRRVU2026_Jul_nonQPP.csv-and-GPCI2026.csv>
node    scripts/gen-locality-table.mjs      # recomputes all 5,668, exits 1 on a cent of drift

One curl returns your state’s figure with the arithmetic: GET /api/table?locality=IA-00, or /api/localities/IA-00 for every figure published for one place. A locality we cannot honour is refused with a sentence, never answered with the national number.

The generator is also the audit: it recomputes every one of the 5,668 figures from the RVUs and the geographic indices on its own row and exits non-zero if a single one is off by a cent. A published number that drifts from the formula printed beside it cannot leave this repository.

3 · Another population

lib/fit.ts holds the whole rule in one function: given a published row and a person, does this figure describe them. It chooses between published figures and never computes one.

// lib/fit.ts — fitOf() is the whole rule, in one function.
// Add the coverage to COVERAGE_OPTIONS, then add its branch:

case 'tricare':
  return base('REFERENCE PRICE',
    'TRICARE pays its own rates and they are not in this table. '
    + 'This is the federal reference figure for the same service.');

// A population with no published figure returns NOT DESCRIBED with
// offerGap: true. That is the honest answer and it is counted at /gap,
// never filled with the nearest number.

The API answers the same question as the screen, from the same module: POST /api/price with coverage and state returns the fitted figure and the verdict per line. See the API.

4 · Your own community, counted separately

The burden instrument takes a channel. Give your organisation a slug and every response through your link carries it, in the public export as well as in your own reading of it.

https://waypointledger.org/survey?c=your-org-slug

# the slug is validated /^[a-z0-9-]{1,24}$/ and published as the channel
# column of /api/export/survey.csv, so your community's responses are
# separable from everyone else's — by you, and by anyone reading the export.

Small cells are suppressed before anything is published. The instrument, its exact wording and every response option are in dictionary.csv, so a reviewer can read the questions without running the site.