OHDSI Home | Forums | Wiki | Github

Overlap between Measurement & Observation Domain


Hi Everyone - wondered if I could ask for some clarification on the Measurement table? In the CDM docs, it says that:

‘Valid Measurement Concepts (MEASUREMENT_CONCEPT_ID) belong to the ‘Measurement’ domain, but could overlap with the ‘Observation’ domain.’

Just to confirm, does this mean that it’s okay to store standard concepts from the Observation domain in the Measurement table (in measurement_concept_id)?

I am finding that certain sets of facts have concepts split between Measurement and Observation tables, and I’d prefer to store them in the same table — but keen not to breach convention.

(P.s. apologies if this has been asked elsewhere, I found some related threads - but nothing that quite answered my question.)

(Christian Reich) #2


Repetition is a good thing. :slight_smile:

Do you have examples?


Ha! Thanks, @Christian_Reich

Sure thing. So we are trying where possible to map to SNOMED codes, as that is the general direction of travel here in the UK. Then when we look at vitals for example, most of the SNOMED vitals codes are measurements - while the observable entity ‘Core body temperature’ (Concept ID = 4174894) is in the observation domain. Similarly for ‘Blood oxygen saturation’ (Concept ID = 37398626)

The story is similar for the observable entity for Glasgow Coma Score (Concept ID = 4093836) — which although perhaps more “observation-like” really does fit the measurement ‘systematic and standardised’ criterion in our acute setting. (Though admittedly in this case Concept ID = 4296538 for the ‘staging/scale’ class concept is, in fact, a measurement.)

(Eduard Korchmar) #4

Hello! ‘Core body temperature’ belongs to a branch of SNOMED’s internal attributes. It is indeed Observation, as it represents body temperature itself, not the process of taking or monitoring it.

This concept could be used with values like ‘increased’ or ‘normal’ (value_as_concept_id) to represent the state of the patient, but not the measurement of the temperature.

The best Vocabulary to use to represent measurements in CDM is LOINC. LOINC may duplicate SNOMED concepts in many senses, but it handles various measurements and tests better.

Here is a standard measurement ‘Body temperature’ from LOINC:


Hi Eduard, thanks for your reply!

As far as I can tell within the SNOMED hierarchy, that code (SCTID: 276885007) is a quantitative scale — which would perhaps fit better with a numerical value. Although I do see how body temperature could be viewed as an observation as well.

Yes, we have noticed that there are LOINC codes that fit nicely with these concepts. For the reason above (i.e. UK adoption of SNOMED), we do have a preference to use the SNOMED codes – particularly if there is flexibility to use the codes from the observation domain in the measurement table as the docs suggest there might be?

(Eduard Korchmar) #6

There is currently ongoing effort to standardize LOINC with SNOMED. That means that similar concepts from both systems will very soon be hierarchically related or directly mapped. I would suggest to use LOINC concept, since it is active standard Measurement.

According to current documentation, you could use observation concepts in measurement_source_concept_id field, while putting 0 in measurement_concept_id. It is however rather hacky and may need manual modification of custom querries, which seems unneeded given there are standard measurement concepts available.


Thanks for this, @Eduard_Korchmar