Oncology Data-survey based

Implementers
1

Oncology data has a common code called ICD03 which is not mapped to SNOMED. This code is very important for analysis purpose and research based on cancer types and localization. How can this be dealt with in 4.5 version of CDM?

Secondly, there is a field named intent for surgery. This field is currently mapped within procedure occurrence Would it make sense to make a new field called “Intent” and map to snomed codes?

2

One quick solution for the ICD03–>SNOMED would be using USAGI. I don’t know much about ICD03 so I don’t know if this mapping would be difficult or not for USAGI to produce.
http://www.ohdsi.org/web/wiki/doku.php?id=documentation:software:usagi

3

I have spent a lot of time on this. We are working on adapting the SEER data for OMOP v5, but it should not be different for v4.

ICDO3 cannot be mapped easily to snomed. ICDO3 has up to 4 terminologies within it – location, histology, behavior, and grade. So, SNOMED does not have a single term to which all 4 pieces (or even 3 – ignore grade) can be mapped.

The best solution is to use the ICD10 (or 9) code that is the closest available code, and put that in the condition occurrence table. That will allow automated tools to find it (e.g., Achilles). But all of the cancer-specific information goes in the observation table (maybe also in the measurements table in v5).

The problem with cancer data is that solid tumors tend to be classified by location, and hematological tumors tend to be classified by histology. So using ICD10 codes handles this, more or less.

This page has some crosswalks that might be helpful. I know NCI is working on updating their ICDO3 to ICD10-CM crosswalk, but it is not ready yet.

1 Like
4

@Mark_Danese:

One solution could be to “append” more granular ICD-10-CM (I assume you are looking at the American version, are you not?) concepts to the more simple SNOMED concepts. This would extend the hierarchy down.

Thoughts?

5

@Christian_Reich

Sorry – could you clarify what you mean? I was going to use the ICD-10-CM (American) to SNOMED mapping that is being developed. Are you suggesting also to map to multiple SNOMED concepts?

There is a “site recode” in SEER that divides the world into about 80 cancers. This is what most people use to pull in the right group (e.g., lung cancer). Then they go to more specific pieces to get the right subgroup (e.g., using histology to get to adenocarcinoma). But SNOMED could not address the “not otherwise specified” or “all other” groupings in that set of codes (e.g., “all other mouth”).