Editorial process

Data before prose

The order is not a preference, it is a control. A writer who drafts a sentence and then looks for a source will find one that nearly fits, and the citation makes the near miss look checked. So the document is read first, the fact becomes a typed field with its quote and its date, and the page renders that field. Where a page and the data disagree, the data wins and the page is wrong.

The steps

1. Find the governing document. The primary source, not a summary of it. An aggregator is recorded as an aggregator and never as the payer.
2. Read it and record the field. The value, the document title as printed, the section, a verbatim quote containing the fact, the effective date, and today's date.
3. Record the scope. Which jurisdictions the document governs, which plan types, which drugs, which conditions. A fact may only cite a document that covers it on all four.
4. Let the checks run. They are automatic and they stop a release rather than warn about it.
5. Write the page against the data. Answer first, in plain words, at a reading level somebody can manage while upset.

What the checks refuse

A fact with no citation. A citation whose quote does not contain the fact. A quote that is really a page title or a fragment of markup. A document cited for a jurisdiction, plan type, drug or condition it does not cover. A shared sentence carrying one jurisdiction's number, which is how one state's threshold ends up published as everybody's. A statistic whose denominator is not named. A number typed into a page instead of read from the data.

They also refuse the quieter failures: a check that is written but not running, and a document claiming a check is running when it is not.

Who does what

Research and sourcing are done by the founder, who reads the documents and records what they say. He is not a clinician, and the site is careful about the difference: it publishes administrative documents rather than medical facts. Where a page does make a clinical statement, it carries a named reviewer, and clinical review says which pages those are.

Updating and correcting

Facts are re-read every January and July, and a document that changes mid-cycle is picked up by a weekly check that compares each source against the copy we read. When a rule changes, the old record is kept with its end date rather than overwritten, because what changed and when is often the question a reader actually has. Corrections work the same way and are described on the corrections page.