How this site is built

DxCodebook is generated by a program from the files listed on the sources page. No page text about a code is written by hand or by an AI model: code text is copied from the files, and our own words are counts and labels computed from them.

Checks on every build

Before a build is published, a separate checking program re-reads the NCHS and CMS archives with its own parser and compares every printed code title, note, 7th-character meaning and Index entry with the source, character by character after collapsing spaces. It also checks that each page carries the edition, its dates and the source line, that counts printed on pages match the files, and that no licensed content appears.

Which pages exist

Every chapter, every block of two or more categories and every category has a page. A code or a subcategory gets a page of its own when the NCHS files give it at least 2 facts of its own (Index entries that lead to it, or notes printed at it); a subcategory also needs 2 or more codes below it. In this first wave 54 such pages exist, for frequently used codes; the other codes that meet the rule follow in the next wave. Until then they are rows of their category's table and cards of the search box. A category page lists all its codes with their status and CMS flags; a code that takes a 7th character is one row with its characters, and each full code has a card in the search box.

Updates

ICD-10-CM changes every October 1, and NCHS sometimes publishes an April 1 update. When the new files appear the site is rebuilt from them; a build that still shows last year's edition after October 1 is stopped by its own check. Page dates show when what a page says (a code title, note, status, CMS flag or Index entry) last changed, not when the build ran or the design changed.

What we leave out

No clinical descriptions, no advice on which code to choose, no summaries of the classification in our own words, and no synonyms from commercial databases: only the words NCHS and CMS publish.