OHDSI Home | Forums | Wiki | Github

Mother (Father?) and baby link

(Ian Wong) #1

Dear All,

I am interested in finding out what databases are available for pre-natal drug exposure study. I know CPRD and Hong Kong CDARS databases has mother and baby link.


Ian Wong

(Vojtech Huser) #2

Mother baby link would be a great think to add to OHDSI CDM specs. So that all CDM datasets have it linked the same way. (and can be queried the same way)

I was interested in such link about 4 years back.
In EHR dataset
a nice link may be possible with EHR data.
Usually the baby’s record essentially lives in the mother’s record until birth.

With claims data:
A non-biologic (so not always valid) link can sometime be extracted from Health Plan (Insurance) data (in USA) where one family member (mother or father) may provide healthcare coverage for whole family. They share the same insurance and somehow link several PERSON_IDs into a “family group”.

(David Madigan) #3

FYI, per the CPRD:

There were 782,170 females and 1,132,250 live deliveries in the March 2015
version of the mother-baby link.

(Jose Posada) #4

Reviving this old thread. How Is the best way to link mother and baby in the CDM? Through the fact relationship table? If yes, by linking two person_id’s ? I was also thinking in linking a gestational age measurement for the baby with a delivery from a mom, but not sure if it is a good move?
Any advice @MPhilofsky @Vojtech_Huser @Christian_Reich @David_Madigan @tduarte

(Christian Reich) #5


If you want to capture the duration of the pregnancy create a record in CONDITION_OCCURRENCE.

(Jose Posada) #6

Thanks a lot for the reply. We actually do have a lot of records in the condition_occurrence table. However, those codes indicate the status of the pregnancy at the time of the visit and not at delivery time. That is the reason I was thinking of using a measurement concept like this and populate value_as_number with the actual days. Also, I was looking into the relationship vocabulary and the closest I could find was this, but it does not look entirely right. Should we create a new concept with a relationship like children of or something along those lines to have such linkage at the person_id level?

Thanks again @Christian_Reich

(Christian Reich) #7


Ouch. Never put in age or another time-changing number. Plus, it’s a Condition. You have to pick a concept from the Condition domain and put it in CONDITION_OCCURRENCE: Pregnant, with the start date = conception date and end_date = birth. The relationship should be declared as Parent of and Child of.

(Jose Posada) #8

This is great @Christian_Reich thanks a lot!

(Melanie Philofsky) #9


The concept you suggested has a domain_id = ‘Observation’. How would this linkage?

I would use a standard concept with domain_id = ‘Metadata’ and concept_class = ‘Relationship’ to link the two Person.person_ids together using the Fact Relationship table. Except we don’t have the correct standard concepts to identify the relationship.

(Vojtech Huser) #10

If you know that one parent delivered (mother), I would strongly prefer a biological mother of (rather than parent of) (in latin: mater certa).

See extensive discussion on this here: https://github.com/OHDSI/Themis/issues/38

(Christian Reich) #11

I don’t care. Since the mother is female we know who is the mother. Unless the data are wrong, but then the mother/father assignment can be wrong as well. So, by database good practices don’t put the same information in twice in different ways - you will certainly run into trouble.

(Jose Posada) #12

@Vojtech_Huser was this change ever incorporated?

And for anyone with the same question, I would like to say that I will go this route, and wait for the vocab team to fix the domain of this concept.

@MPhilofsky Where should I submit the request for the change?

(Vojtech Huser) #13

It was not incorporated. It received Themis feedback that was incorporated. No community feedback. And then Themis got less and less activity. And then the v6 phenomena happened…

I don’t like that the relationship does not say ‘biological’ parent. For true science, just mother and just father is not good enough. It should be biological.

The proposal should include an example. So if person 123 is father of person 456, can you provide the exact row and all columns (and of what table) you plan to represent it. (FACT_RELATIONSHIP?). My themis proposal tried to do all that (including Achilles Heel rules since when you write law, you must also be writing the “Achilles Heel police” rules).
Back in 2018 - I put a lot of effort into that Themis proposal (and in Themis in general) and was kind of “not happy” that it went “somewhat nowhere”.

(Melanie Philofsky) #14

Hi @jposada,

@Christian_Reich beat us to it and already submitted a vocabulary change request based on @Vojtech_Huser Themis discussion. Looks like all bases are covered!