OHDSI Home | Forums | Wiki | Github

CDM4 to CDM5 Conversion Issues

CDM Builders
#1

Hi all:

I’m having some issues with a CDM4 to CDM5 conversion and would appreciate some guidance on the following:

Organization: I realize this table disappeared from CDM4 but what happened to the care_site relationship table in CDM5? How can I map an organization from 4 to 5?

relevant_condtion_concept_id: This column has been removed from the procedure_occurrence, observation and drug_exposure tables. Why? Will it be added back into CDM5?

I appreciate your time.

Jorge

#2

@Jorge:

ORGANIZATION table: It’s all now in CARE_SITE. There is no difference any more between the Care Site and the Organization. It was never clear which one is which. So, if you have a hospital, and in the hospital you have a surgical department, both are Care Sites, and between them you can create a relationship in the FACT_RELATIONSHIP table.

For the relevant_condition_concept_id: Same thing. Use FACT_RELATIONSHIP and link to the record in the CONDITION_OCCURRENCE table.

Hope this helps.
C

#3

That helps a lot. Can you give me some guidelines as to how to populate the relationship concept id column?

Where in the meta data should I look? I don’t see anything for “relevant concept”?

Thanks

Jorge

#4

Good point. We should establish conventions for these. You could use:

  • Between Care Sites: 4023263 ‘Is a’
  • Between conditions and observations: 4091835 ‘Associated observation’
  • Between conditions and drugs: 4132667 ‘Indication for’
  • Between conditions and procedures: 4179699 ‘Associated procedure’

However, they are not really symmetrical. We should create those. Can you list all the relationships you need, and then we vet them with the community a bit and create them?

#5

I will need the following:

drug_exposure->relevant_condition_concept_id
procedure->relevant_condition_concept_id
observation->relevant_condition_concept_id

procedure_cost->disease_class_concept_id

drug_era->drug_type_concept_id
condition_era->condition_type_concept_id

Jorge

#6

Hi Christian:
@rimma had asked about how to link 2 blood pressure measurements in a different thread. This is likely to be a common occurrence within measurements and within observations. For example, systolic and diastolic blood pressure could be linked with an “associated measurement” code. Is there an “associated measurement” code? I didn’t see one but I am not necessarily searching in the optimal way.

Mark

#7

We should just make a concept “Systolic to diastolic blood pressure measurement” and “Diastolic to systolic blood pressure measurement”. In other words, instead of trying to find in SNOMED something that fits, let’s just make them. They aren’t that many. SNOMED has a good set of family relationships, though. We would summarize all those conventions in the specs.

#8

@Jorge

Can you remind me what these are? What is disease_class_concept_id? DRUG_ERA and CONDITION_ERA already have a _type_concept_id field.

C

#9

Christian:

Disease_class_concept_id is defined in CDM4 for the procedure_cost table. There is no corresponding column in the CDM5 architecture so I thought I would have to map it to the fact_relationship table.

But I see from a posting, this data would go in the observation table. Can you tell me how?

Thanks

Jorge

#10

Ah. The DRGs. We took them out of the cost and put them into the OBSERVATION table. All DRGs are in the Concept table (but careful, they are time dependent, i.e. the same code will turn into differen concept_id values over time, check the valid_start_date and valid_end_date). So, when you see a DRG in a claim, write a record into the OBERVATION table with the concept that corresponds to the DRG.

#11

Thanks Christian.

This information should be shared with whoever is working on the CDM4 to CDM5 conversion script because there is no SQL that I can see that is taking the disease class concept id from 4 and putting it into the observation table for 5.

Jorge

#12

Jorge: We’ve posted a new version of the conversion script and it now handles the disease class concepts. Please take a look and reach out with any questions.

Anthony

#13

Anthony:

Great stuff.

Here are some more things needed that I noticed:

Convert organization data into care_site and fact_relationship.

Convert relevant_condition_concept_id from drug_exposure to fact_relationship.

Convert drug_type_concept_id from drug_era to fact_relationship.

Convert condition_type_concept_id from condition_era to fact_relationship.

Thanks

Jorge

#14

Jorge, those would be great things for you to add if you need them. We
don’t use the fact relationship table in our work, but I can see how you
could use it for your purposes. The type fields for the era tables aren’t
really necessary, as I’m not aware of anyone building multiple types of
eras (we had only done it once for purposes of running specific omop
experiment several years ago).

#15

Anthony, Great presentation on the 11th, thanks for postponing for Taylor.
Were are the conversion scripts posted, I’m having trouble locating them.
thanks,
Lisa

#16

Lisa - apologies for my delay in responding but I was out of the office. Thanks for the note and glad you enjoyed the presentation.

So the conversion scripts are housed under the CommonDataModel project on GitHub: https://github.com/OHDSI/CommonDataModel. There is a folder for the V4 to V5 conversion scripts here: https://github.com/OHDSI/CommonDataModel/tree/master/Version4%20To%20Version5%20Conversion. I’ve included documentation but please reach out if anything is unclear. Thanks!

#17

Thanks!

Lisa Schilling, MD, MSPH
Professor of Medicine
ph 303- 724-5138 (Adult and Child Center for Outcome Research and Delivery Science-ACCORDS)
ph 303- 724- 2254 (General Internal Medicine)

[Description: http://www.ucdenver.edu/about/departments/oiuc/brand/downloads/branddownloads/branddocuments/E-mail%20Signatures/emailSig_Anschutz.png]

t