| OHDSI Home | Forums | Wiki | Github |
Hi,
I noticed both C90 (Multiple myeloma and malignant plasma cell neoplasms) and the subcode C90.0 (Multiple myeloma) map to the same SNOMED standard concept (437233). This does not happen to other ICD-10 codes, as far as I am concerned. What could be the reason?
Thanks in advance.
Good question.
Well, it happens a lot that two ICD codes that are parent and child in the hierarchy go to the same SNOMED concept. Usually, when the child is a “Not otherwise specified” version of the parent. But here that is not the case.
The problem is that C90 is a classification, rather than a concrete disease concept. And classifications don’t map well, because they have different hierarchical relationships to their constituents.
You are right, C90 should probably be mapped to Plasma cell neoplasm, rather than to myeloma. But if you look what’s underneath them they are quite different:
| ICD10 C90 | SNOMED 443743 |
|---|---|
| Multiple myeloma | Multiple myeloma |
| Solitary plasmacytoma | Plasmacytoma |
| Plasma cell leukaemia | Plasmablastic lymphoma |
| Extramedullary plasmacytoma | Heavy chain disease |
| Monoclonal B-cell lymphocytosis | |
| Monoclonal gammopathy of uncertain significance | |
| Necrobiotic xanthogranuloma with paraproteinemia | |
| Non-amyloid monoclonal immunoglobulin deposition disease | |
| Plasma cell dyscrasia with polyneuropathy | |
| Primary amyloidosis of light chain type |
What is your use case? Do you have data containing C90 you want to map for a myeloma study? Or do you need to summarize various diseases into classifications?
Thank you for replying, @Christian_Reich:
That explains a lot. I have the first case, working on a myeloma study. My source diagnosis-related data has both C90 and C90.0. So when I apply the transformation, both end up with SNOMED 443743, which is not entirely correct.
Yeah, that is the plight of ICD cancer codes. We call them “shallow”, because they really are insufficiently describing the disease. In solid cancers, they lack any useful histology (e.g. C16 “Malignant neoplasm of stomach”), and in hem cancers all they provide is the cell type (your C90 or e.g. C84.2 “T-zone lymphoma”). What’s particularly missing is any sense of stage or advancement of the disease, which of course is essential for prognosis and treatment, and hence our research.
You will need to rely on the data the diagnostic workup (usually at the beginning after detection of the cancer) produces: clinical diagnosis in the EHR, pathological diagnosis in the path lab, imaging results. Do you have those?
Hello @a.licona,
Is your data sourced from an EHR? If yes, does the EHR utilize the IMO interface terminology?
My data comes from an EHR, and we have test metadata, but not the pathological or imaging diagnosis data needed for the full diagnostic workup.
Hi @MPhilofsky,
It is from an EHR. I am not sure about the IMO part. In this case, we only received the data dump, which contained ICD-10 codes but not IMO codes.
I could ask for IMO codes. Do you think having those will help with the mapping?
@MPhilofsky is on a fishing expedition trying to find out if your EHR has better codes than ICD10. Those won’t do the job in cancer research most of the time.
IMO is often used in the US, but maybe also outside, we don’t know. Other coding systems EHRs have are ICDO3 and SNOMED, but also coding systems for cancer registries (in North America often NAACCR) or path system (in the US often CAP eCC).
Do you know if you have anything else but ICD10?
Hi again. I just confirmed we do not have an alternative to ICD10. Our data is from Turkey, so there is little that we can do regarding other vocabularies here.
You may have to go to the actual reports and NLP or LLM them.