Needs help on chief complaints mapping

[sumnemo]

Hi, everyone.

In these days, I’m working on mapping the chief complaint(c/c) data to SNOMED vocabulary.
Currently, my hospital’s c/c data is loaded in condition table, so I am trying to map the data as “condition domain” with “clinical finding class,” but it is hard to map ‘within’ the condition domain.

If I look for the certain vocabulary on condition domain with clinical finding class,
most of the mapped concepts seem to be the name of disease, not the complaints from the patients.

For example, if the patient comes to the hospital cause of tuberculosis exposure,
the chief complaint seems to be "exposure to tuberculosis(43530675) to me, which belongs to
observation domain not the condition.

There are many patients who already have diagnosed the disease and come to the bigger hospital
just for the procedure. On those cases, the chief complaints are written as the name of the procedure.
I even cannot guess why the patient needs the procedure at all, so I have to map the data as
the concept in ‘procedure’ domain.

To make story short, it seems like the chief complaints data may include various kinds of domains.
It could be the name of disease, the name of procedure or patients’ observational status.

However, as you know, since the data is in one certain table, the concepts have to belong to the
same domain, so I don’t know what to do right now.

Please tell me the rightful guideline for mapping c/c data.

[Eduard_Korchmar]

Hello!

since the data is in one certain table, the concepts have to belong to the
same domain

This is not necessarily true. In my experience, things that are similar in clinical settings, like injection of contrast media and injection of a vaccine may be on neighbouring rows in source data, but end up in different CDM tables (device_exposure and drug_exposure respectively) and will have to be queried differently.

"exposure to tuberculosis(43530675) to me, which belongs to
observation domain not the condition

That is absolutely intended. Exposure to tuberculosis is an event that does not necessarily reflect changes in patient’s biology. Researches looking for patients with exposure to infectious disease would query OBSERVATION table. Similar groups of concepts that are Observations are personal histories of diseases and suspicion of diseases. Providers may code them in the same fields as actual Conditions, because they require similar set of tests or procedures done when reported, but they are not Conditions in OMOP.

I even cannot guess why the patient needs the procedure at all, so I have to map the data as
the concept in ‘procedure’ domain.

That is also correct approach.

My advice would be to map all c/c concepts to whatever Standard concepts for them are in Athena and put them in Standardized Clinical Data Tables later with respect to which Domain target concept belongs to.

If you have massive data set that has concepts from mixed domains, you could start by automatically creating draft mapping with USAGI with filtering by Domain left unchecked. It’s not without it’s quirks, so you will still have to check resulting mappings manually, but usually source data with higher occurrence is also cleaner lexically (containing less acronyms and typos), so it’s reasonable to expect good results on high counts.

[Laurentiu (St)]

Hello @sumnemo
I am not sure of this will help you, but the type of variables and values you need to include in the set looks to be more close to HL7 FHIR standards. Please follow the link below and you will find the “Chief complaint” values under the name of “Condition” according to this standard.

http://hl7.org/fhir/condition.html

[sumnemo]

Thank you for the reply.

I assume that you’re suggesting is mapping all the c/c data to whatever concepts not considering the domain.
Then after mapping the concepts, I need to put the standard concepts to each target table which domain belongs to.

Am I understanding right???

[sumnemo]

I’ve never heard of HL7 FHIR standards before.
Thank you for letting me know now.

[LeileifromUK (Leilei Zhu (Clinical Data Standards Lead, UCLH, UK))]

hi ,

Derivation of SNOMED CT dataset for Emergency Care should have been handled at a national level rather than down to individual organisations. One, it is duplicated efforts; Two, it is unlikely going to achieve consistency; three, It is the national body’s responsibility anyway.
Also, national team should really drive the organisation to adopt data standard (SNOMED CT as one of them) at the source data level.

In the UK, we have the Emergency Care Data Set in SNOMED CT, which covers a wide range of elements, e.g. Chief complaint, diagnosis, investigations and treatment etc. It has a lot of issues but at least, we are seeing some data standards efforts at the national level.

Chief complaint covers a range of things that are falling within at least two top level concepts: finding (including Disorder) and procedures, e.g. wound care, for example. I think also it depends on the healthcare system in the country. In the UK, patient doesn’t necessarily come to ED unless they have “something wrong”. so if have an exposure to TB, but I didn’t show any symptom, it is unlikely I would come to ED. My GP might refer me to do a lab test but we will go through the outpatient clinic route.

However, I believe it is mainly falling within Finding.