Pediatric suicides in Turks and Caicos Islands.

Every page on this site was looked at by me, one at a time, before it was published. This is exactly what that involved, and what it did not.

What I checked

• The numbers, year by year: deaths and the population they are counted against, from the file the page is built from.
• That the arithmetic agrees with itself. A rate equals its deaths over its person-time, every interval contains the rate it belongs to, the sexes add up to the combined figure, and no year appears twice or goes missing. These are checked by machine on every build, and I read what it reports.
• The case definition and the age band, against what the source publishes, and whether either changes part-way through the series.
• The source, its link and the date it was retrieved.
• Every chart view the page offers, drawn, and read.
• WHO's own grade for the country's cause-of-death data, where there is one, and whether the page reads more confidently than that grade allows.
• The prose, against the numbers it describes.

What I did not

• I did not re-obtain the primary source documents or re-key the data from them. Where a national statistics office publishes a figure, this site carries that figure.
• This is not a re-analysis. I have not attempted to correct, adjust or model around a country's counting; I have shown what it published and said what I know about how it was counted.
• Being reviewed does not mean a series is comparable with any other. It means I have looked at it and said what it is.

Automated checks ran over every page and I read their output. A language model was used on some pages to suggest things worth a second look; nothing it produced is published, and nothing it said stands as a finding on its own.

A review is a point in time. If the data behind a page changes afterwards, the page does not ship again until it has been looked at again — that rule is enforced when the site is deployed, not left to memory.