Hi everyone! I’ve got a new problem about how to record family history in to OMOP.We have NLPed the family history and got some relatives as well as their disease.Some patients have several young sisters or older sisters(just a example) and they have different diseases.OMOP CDM only have yonger sister CONCEPT_ID and older sister CONCEPT_ID.When we record them into the OMOP,we found that we can’t distinct which sister gets which disease in the Atlas.For example,the patient’s oldest sister gets a disease and another older sister gets another disease.We can’t figure out who gets the first disease and who gets the other disease.
So how can l solve this problem?
What’s the use case? Why do you need to distinguish which sibling? Is there a scientific question where this is relevant?
Maybe I asked this question in a wrong way.In my data base,there are some patients who have several sisters for example,and their sisters got the same disease.But the data resource is:patient’s three sisters get Diabetes,and our NLP result is “three sister”-“Diabetes”.This is gonna be one record,but actully it should be three records.That means we will lose two records.
Another question is we can not distinguish which relatives are the father side and which relatives are the mother side.
So, nothing to do with old / younger sister? Agree with @Christian_Reich that there’s no difference.
Clinically signifacant is what percentage of siblings have this disease.
hm,
sister has diabetes
sister has diabetes
sister has diabetes
looks like a duplicate of the same row.
I think to have one row ‘sister has diabetes’ is the best we can have now.
see here, “Children” block
https://browser.ihtsdotools.org/?perspective=full&conceptId1=57177007&edition=MAIN/2019-07-31&release=&languages=en
it is SNOMED browser, you can find these concepts in Athena as well.
If something is missing from your list, you can create custom concepts.
Thanks a lot.So my further questions come.If a patient’s three sisters get three different diseases like diabetes 、depression diorder and hypertension.And we will get :
sisrer has diabetes
sisrer has depression diorder
sisrer has hypertension
We will not be able to distinguish whether one sister gets three different diseases or three sisters get three different diseases in the OMOP.And that’s what makes me painful.
I really like your commitment to save as much as possible data during ETL.
Usually we find some balance between losing data and overcomplicating your CDM.
Not sure whether there will be any analytical use case where we need to know that one sister has diabetes AND depression, but another has hypertension.
But if you really want this (I think nobody did this before in CDM):
create person_id for these sisters, put them in PERSON table with year_of_birth = 0 and fill CONDITION_OCCURRENCE with diabetes for this person(you still need to put some date of event).
and then connect given patient’s person_id and their sisters’ person_id using FACT_RELATIONSHIP.
Looks cumbersome, right?
I would recomend to stick to 3 unrelated entries for
and return to this question once you’ll have analytical use case.
Please let me know if it makes sense.
I agree with @Dymshyts. Usually, in medical practice this kind of detail is neither collected, nor relevant. All you want to know if there is a genetic predisposition to certain diseases. So, don’t stress too much if you can’t distinguish between the three sisters.
If, however, you want to do real pedigree family studies you should let us know, and we figure out if and what kind of modeling we should provide.
Makes sense?
Thanks a lot.What you said makes me more clear about this question.Why I ask this question is because we don’t have relatives’ person_id in our data base.We got family history from patient report.So we can’t distinguish different relatives by person_id.This is our problem.
Yes,you have provide a practical method.Maybe we will give it a shot in some day if we need.At that time I will get back to you and tell you the results.
Thanks again!
Got it.Thanks!