We are trying to understand the distinction between the OBSERVATION and CONDITION_OCCURRENCE tables in OMOP CDM, particularly for provider-observed signs or structural abnormalities found during procedures.
OMOP definitions:
OBSERVATION: captures clinical facts about a person obtained in the context of examination, questioning, or a procedure.
CONDITION_OCCURRENCE: contains events suggesting the presence of a disease or medical condition, stated as a diagnosis, sign, or symptom, observed by a provider or reported by the patient.
Examples causing confusion:
Hepatomegaly present – observed during examination. Seems like an OBSERVATION, but it’s also a sign of a condition, which could suggest CONDITION_OCCURRENCE.
Lower end CBD stricture, discovered during an ERCP procedure – a structural abnormality. Should this go in CONDITION_OCCURRENCE as a diagnosed abnormality, or OBSERVATION because it was noted during a procedure?
Question: Are there general guidelines or examples for deciding when a provider-observed clinical finding or procedure-discovered abnormality should be recorded in OBSERVATION versus CONDITION_OCCURRENCE?
We have a definition of Conditions and Observations, as you cited in your question, but no clear guidelines to unambiguously decide whether a finding belongs to one or the other. It is a hot topic in the Vocabularies and people have strong opinions on the matter.
Over the years, we have had many debates. The most recent decisions are
Disturb domains as little as possible (if something is assigned a domain, let it stick, we do not want many changes in domain between releases, so the phenotypes stay stable).
Use SNOMED tags more often for domain assignment (previously we worked a lot on the hierarchies trying to find the most suitable ancestor to assign the right domain to all of the descendants - didn’t work out).
If the clinical fact is important and you know that you will be looking for it, it makes a lot of sense to put it in a ‘cleaner’ Condition domain.
You may want to check this topic. I also invite you to come to the Vocabulary WG sessions in our Teams, where we can discuss cases you have.
According to current state, Hepatomegaly is a condition (tagged ‘Disorder’ in SNOMED, we followed the logic), as well as a Common bile duct stricture. Most conditions are diagnosed with procedures, so the fact of the procedure performed does not rule out a condition domain. Eg. you performed an ECG, found some ST-elevations and concluded that patient had a myocardial infarction. Keep it as AMI in a condition domain and you are fine.
Thank you so much @zhuk for this clear and insightful response! It’s reassuring to know this is a known ‘hot topic’ and not just a gap in our understanding. The guiding principles you outlined are exactly what we were looking for. The logic of relying on SNOMED tags and considering the analytical importance of the finding makes a lot of sense. This clarifies our immediate questions regarding Hepatomegaly and CBD stricture and gives us a much better framework for making these decisions moving forward.
We really appreciate the invitation to the Vocabulary WG sessions and would love to take you up on that. Could you please point us to where we can find the schedule or a link to the Teams channel?
Absolutely, here is the list of upcoming WG calls. Search for Common Data Model WG- VOCABULARY Subgroup. In fact, we also have office hours to answer your questions.