Methodology and coverage

A source-navigation edition.

This edition provides original WHO and CDC resource links, dates, and coverage notes. It does not yet contain human-approved event records, a medically reviewed clinical comparison, or approved fatality interpretation cards.

AI assistance was used to assemble resource metadata and prepare private candidates. An access check is not a human fact-check or a medical review. Candidate content is excluded from public pages and data files.

A selected collection of public reports.

The scope is publicly documented human plague events, with clinical types kept separate. It is not comprehensive global surveillance, a case-reporting service, or a real-time risk map.

No matching event records means that our collection does not contain a published match. It must not be interpreted as an absence of disease. Animal monitoring and historical annual totals are not combined with event counts.

Map markers show only a location precision supported by the source. An administrative representative point is not a patient’s address or proof of an exposure location. Basemap boundaries are for orientation and do not express a position on political status.

Different dates answer different questions.

• Source publication: when the original material was published.
• First report / latest development: source-supported dates used for event browsing.
• Statistical cut-off: the point up to which a figure applies.
• Accessed: when the resource was retrieved; it does not imply review.
• Human review: when a named reviewer checked a specific version.
• Content updated: when a substantive change to this website was made.

Recent records use a 90-day source-date window. An old record checked today stays historical. Month- or year-only dates retain that precision; ambiguous dates do not get invented days or times.

Confirmation and progress are separate.

Confirmation labels describe what a source supports: unclassified, suspected cases reported, confirmed cases reported, or plague ruled out. Progress labels describe the event: investigation ongoing, under monitoring, or closed.

A confirmed event may also be closed. Time without a new report does not automatically close or rule out an event. Mixed case definitions remain visible; an event with some confirmed cases does not make every reported case confirmed.

Unknown is not zero.

We preserve five value states: reported, unknown, not reported, not applicable, and suppressed. Only an explicit reported zero is displayed as zero.

Repeated reports do not add cases. Cumulative snapshots are not added together. A subset is not added to its parent total, and numbers with unknown overlap are not combined. Conflicting sources remain separate until reviewed.

Fatality cards retain the source population, clinical type, case definition, period, treatment context, and follow-up limits. A recomputation is permitted only for a matching, approved numerator and denominator. The tool does not infer a personal outcome from a published percentage.

Review is tied to a specific version.

The publishing sequence is candidate preparation, schema and citation checks, human editorial review, any required medical review, a frozen edition, and publication. Review records bind to a content hash: changing a fact makes the old approval inapplicable.

AI cannot approve its own extracted records. Automated checks catch structural errors but cannot establish that a source supports a conclusion. Editorial review and medical review are separately recorded, and no expert identity is invented.

When clinical review is unavailable, the affected content stays private. Official source navigation can still be published with a transparent status.

One edition across every view.

The HTML, list, map, timeline, and data cards use the same frozen edition. Sharing a filter changes the view; it does not create a new publication or a new case count.

Substantive corrections produce a new edition with an explanation. A failed build leaves the prior edition intact. Withdrawn content must also be removed from retained public data versions before those versions can be used again.

Use the correction template to describe a problem and provide an original source. The website does not accept patient details or personal medical histories.