OHDSI Home | Forums | Wiki | Github

ICD-0 Versions

So ICD-O 3.2 came out 1-Jan-2021.Perhaps this has been discussed before. The way ICD-O works is that the preferred terms updates in a given version are for diseases diagnosed after the update comes out. There are significant changes to 3.2. Has any thought been given to tracking what version of ICD-O a term comes from?

So for example under 3.1 the the description is Urothelial papilloma, inverted. Under 3.2 the description is Sinonasal papilloma, exophytic. The description depends on when the diagnosis occurred.

Also why are some ICD-O3 concepts standard and others non-standard even thought are both current ICD-O3 codes?

There is no good way to track temporal code changes in OMOP. We generally assume that various descriptions provided by ICDO are synonymous names of the same morphology. So no matter when does the diagnosis occurs, semantical meaning should not change. To be fair, there are vocabularies, like NDC, where this is not always the case. However, ICDO3 does not really seem to have this problem. In case of Sinonasal papilloma, exophytic, it still keeps alternative description of “Urothelial papilloma, inverted” with Related status in 3.2:

ICDO3.2 Level Term
8121/0 Preferred Sinonasal papilloma, exophytic
8121/0 Synonym Schneiderian papilloma, NOS
8121/0 Synonym Sinonasal papilloma, NOS
8121/0 Synonym Sinonasal papilloma, fungiform
8121/0 Related Urothelial papilloma, inverted
8121/0 Related Transitional papilloma, inverted, NOS
8121/0 Synonym Transitional cell papilloma, inverted, NOS

(taken from here)

We do not preserve “Related” concepts as synonyms in OMOP, because they are likely subcategories of their “Preferred” terms, and it may be incorrect to fully equate them.

For Condition domain, we try to align everything to SNOMED. The main reason for this decision is SNOMED’s multi-axial hierarchy. For example, SNOMED allows user to create concept sets for all malignant neoplasms, tumors of skin regardless of malignancy or lymphoproliferative disorders with one button. ICDO does not have that.

ICDO concepts are made non-standard only in cases where their meaning fully aligns to existing SNOMED concepts with defined attribute set, and are mapped to the latter. Standard ICDO concepts don’t have exact equivalent in SNOMED and are instead assigned ancestors among SNOMED concepts, as close as possible (example, see ‘Is a’ relationships). In both cases, concept sets created from SNOMED concepts will have all intended ICDO concepts included.

There is a consideration for OMOP to have own extension of SNOMED, so that necessary concepts could be created to support mappings from currently standard ICDO concept, but there is no ETA for this yet.