Question about multiple related observations (e.g. BP)

Implementers
1

What is the best practice for importing multiple related observations/measurements into OMOP (for example blood pressure)? Should they just be uploaded as separate observations/measurements? Is there a way to indicate that these measurements/observations are related? Is BP a measurement or observation? Lab Panels?

Is there some one that can volunteer as a Sherpa to help guide me through some of this?

Thanks,
John

2

Hello @greshje.gmail ,

It depends which version of the CDM you are using. What version of the OMOP CDM are you using? Are you using the Oncology extension?

BP and Labs are Measurements. The Book of OHDSI explains how source data map to the CDM.

3

Thanks for the help Melanie, We’re using CDM 5.4 (not using oncology extension, using files from https://github.com/OHDSI/CommonDataModel/tree/master/inst/ddl/5.4/oracle)

4

Thanks for the information that BP and Labs are measurements. I’m currently mapping from FHIR data where everything seems to be an observation. Is there a list/mapping or set of rules that defines what is a measurement v what is an observation (e.g. what are height, weight, bmi, level of pain on a scale from 1 to 10, etc.)?

5

You’re welcome!

Yes, The Book of OHDSI explains the rules. In short, you map your source value to a source concept_id, then map the source concept_id to a standard concept_id and the domain_id of the standard concept_id will tell you where it lives.

The conventions for CDM v5.4 will give you a description for each table will give you a general overview of what is expected in a CDM table.

To link clinical events, i.e. systolic BP to diastolic BP, you can use the Fact Relationship table or the measurement_event_id and the meas_event_field_concept_id in the Measurement table.

6

Also, as Oleg mentioned on another post, check out the FHIR to OMOP WG. They might have more specific resources for FHIR data

7

Thanks, the conventions for CDM v5.4 link seems to have what I’m looking for!

8

As an update, I’m still not seeing anything other than visit_occurrence that will allow for the grouping of observations (e.g. how are systolic and diastolic blood pressures marked as related, or associate the weight and height used to calculate a bmi, or group lab results of a panel of tests together).

9

Hello @greshje.gmail,

Just like Measurement, Observation table has similar fields observation_event_id / obs_event_field_concept_id that can be used to link Observation record to another clinical event (possibly from another domain).

As previously mentioned by @MPhilofsky, another option is to use Fact Relationship table. Compared to the first method, this allows to reflect the context/nature of relationship by choosing a certain relationship_concept_id:

‘Systolic to diastolic blood pressure measurement’ and reverse

possibly ‘Parent to Child Measurement’ and ‘Child to Parent Measurement’ for reverse record.

1 Like
10

Nice! Thanks @rookie_crewkie