Platform standard

Research and page methodology

How a consumer question becomes a scoped, sourced, reviewable record, and what stops a record from being published.

Every record follows the same five stages before it is published. First, a brief fixes what a reader is actually trying to decide and draws the line between education and diagnosis for that specific topic. Second, editors build a set of source materials: identifying controlling sources, recording the exact span each claim relies on, and writing down where sources disagree or evidence is simply missing, before any prose exists. Third, required structural fields (a direct answer, applicability, claim-level evidence, limitations and a concrete next action) are completed before anything is edited for readability. Fourth, every claim is checked against its cited source, though clinical interpretation by a named licensed reviewer is a separate, higher bar this library has not yet cleared. Fifth, material facts carry a review interval and corrections propagate to every dependent record, not only the page where an error was reported.

Automated checks, not editorial memory, enforce link integrity, the no-price-or-outcome-figure rule, sitemap completeness and structured-data consistency across the whole built site.

1. Page methodology: resolve the question before writing anything

Every record begins with a brief that fixes what the page is answering and, more importantly, what it is not. That means naming the user job in the reader's own words, identifying the clinical or financial entity precisely, deciding whether geography is relevant, and drawing the line between education and diagnosis for this specific topic. The line moves by subject: a page about an annual maximum can be almost entirely definitive, while a page about facial swelling has to be built around urgency and referral because nearly every specific answer requires an examination.

The job: what a reader is trying to decide, phrased the way they would phrase it
The entity: the exact treatment, condition, plan term or provider type, distinguished from adjacent ones people confuse it with
The boundary: which parts of the answer an examination, a plan document or a written quote must settle
The failure mode: how this page could mislead someone, written down before drafting so the draft can defend against it

2. Build the set of source materials

Sources are collected before prose, not retrofitted to justify it. For each record, editors identify the controlling sources for the topic, record the specific span of each source that supports the claims being made, note usage rights, and, the step most often skipped elsewhere, write down where credible sources disagree and where the evidence simply does not exist. A gap identified at this stage becomes a visible limitation on the page. A gap discovered after drafting usually becomes a quiet hand-wave instead, which is why the order matters.

3. Structure before prose

Required fields are completed before anything is edited for readability. A record that cannot fill them is not a record yet. This is what keeps the library consistent enough that a reader who learns to use one treatment page can use all of them, and it is what makes the corpus machine-readable without a separate summarization layer.

A direct answer that stands alone if nothing else is read
Applicability: who or what the answer does and does not cover
The evidence behind each factual claim, at the claim level instead of a source list at the foot
Limitations, including what an examination must resolve
A concrete next action the reader can take today
Sources, an update date and a route to report a correction

4. Fact-check and indexation

Every claim is checked against the cited primary or professional source before a record is published. That is the bar records are actually held to, and it is enforced by the same automated checks that block a fabricated statistic or an undisclosed price. Clinical interpretation by a named, licensed reviewer is a separate, higher bar this library has not yet cleared, and no page here claims otherwise: no reviewer name, review date or medical-review badge appears anywhere. Records that are thin, duplicative, stale or unsourced are held back regardless of indexation status. See the clinical review status page for exactly what has and has not happened.

5. Monitor, correct and propagate

Material facts carry a review interval appropriate to how fast they change: regulatory and device information is checked more often than anatomy. Corrections are logged instead of silently patched, and, the part that is easy to get wrong, a correction is propagated to every dependent record, not only the page where it was reported. A term redefined in the glossary changes every page that links to it. The correction is not closed until those pages agree with each other.

What is checked automatically on every build

Editorial rules that rely on someone remembering them do not hold. The ones that matter here are enforced by tests that run against the fully built site and fail the build, not reporting a warning. Each check prints a count instead of a verdict. Each was deliberately broken once when it was written to confirm it actually fires.

Every internal link on every page resolves: the crawler follows them all instead of sampling
No page carries an average or estimated treatment price, a clinical success or survival rate, or a percentage describing an outcome; fees appear only when quoted from a named public fee schedule
Every question rendered on a page appears in that page's structured data, and nothing appears in the structured data that is not rendered
Every indexable page is in the sitemap, and no noindex page is
Every clinical record carries a scope boundary and a rendered sources section
No clinic, provider, address, telephone, review or rating markup appears anywhere in structured data
Titles, descriptions and canonical URLs are unique and self-consistent across the whole site