Methodology

One page, one question

A page exists because a reader has a specific question, not because a phrase gets searched. If a proposed page would only restate what an existing page already answers, it does not get made. If a page would exist only to catch a keyword, it does not get made either.

That rule is why this site has no city pages, no clinic pages, no week-by-week pages, and no automatically generated comparisons.

How a page is built

1. Define the question. One reader intent, stated plainly.
2. Read the sources. Primary sources in full, in the order set out in the editorial standards. Search snippets and secondary coverage are used to find sources, never as sources.
3. Record the mapping. Every consequential claim is recorded against the source that supports it, along with what that source cannot support.
4. Draft. The answer goes near the top. Specialist terms are defined on first use. Detail that does not improve understanding is left out, which includes graphic procedural detail.
5. Check the boundaries. No dosing, no eligibility judgement, no triage, no reassurance, no claim that a process is complete, no assigned emotion.
6. Check the language. Terminology contract, dates, banned punctuation, no developer-facing text, and no unfinished copy.
7. Publish, with publication date, substantive update date, sources, and a change history.

What has to be true before a page is listed in search results

A page is only listed in search results when all of the following hold:

• the content is complete rather than partial,
• the sources are visible on the page and resolve to a live document,
• all applicable editorial checks are recorded and current,
• the page canonicalises to itself on the bare production domain,
• the accessibility checks pass, and
• no consequential fact on the page is past its verification window.

If any of those fails, the page is kept out of search listings while remaining available to anyone who visits it. Withholding a page from search results is not the same as hiding it.

Handling data

When this site describes a dataset:

• the data year and the publication date are both shown, and they are not the same thing,
• the reporting areas included are named, and the ones missing are named,
• the definition and denominator are stated,
• suppression and uncertainty in the source are explained,
• the exact report or dataset is linked,
• residence and occurrence are distinguished, and
• units are preserved, with any transformation documented.

Datasets from different sources are not combined into one figure without a methodology note. A chart is never described as current merely because the site was rebuilt. Government data is never presented as though this site produced it.

Every chart is accompanied by a prose summary and an accessible data table carrying the same information.

Monitoring sources

Sources are grouped by how quickly they change.

• Weekly: regulatory material on medication abortion, and the sources behind emergency wording.
• Monthly: global clinical guidance.
• Quarterly: professional society material, terminology guidance, and surveillance methodology.
• Every release: hotlines, helplines, and provider routing services.

A scheduled check records whether each source still resolves and whether its content fingerprint has changed. A change opens a task. It never edits a page, never republishes anything, and never alters a clinical or legal conclusion on its own.

The distinction between a temporary network failure and a genuinely removed source is made by a person, because treating the first as the second produces false alarms and treating the second as the first hides real problems.

Law and policy

Legal material follows stricter rules than clinical material, because it changes faster and because a stale legal summary can cause real harm. Those rules are set out in the law and policy section: a summary is never reduced to one unqualified word, it carries a verification time stamp with a time zone, and it is withdrawn automatically once it passes its verification window rather than sitting there looking current.

Translation

Translated pages are professionally translated and reviewed, and where the page is clinical they receive medical review in the target language as well. A machine translation is never published as a finished page, and a translated page is not listed in search results until its English source, its translator record, and its reviews are all current.

Corrections

The corrections process covers how errors are reported, triaged, and published.