OHDSI Home | Forums | Wiki | Github

Profile data relationship in Atlas


(Vikram) #1

Hi ,

We are trying to visualize the data for a particular patient through atlas . I could see the data as concepts in the profile for a patient with a patient id. Is there any functionality in atlas which lets us see the relationship between the data for the patient.

If i have some data relationship in fact relationship , can i see them through atlas?


(Frank DeFalco) #2

Could you be more specific about the type of relationships you are trying to see?

(Vikram) #3

I have some relationship between the data in the fact relationship table, i want to know if i can see them in atlas.

(Gowtham Rao) #4

If the question is – can we see the relationship between the condition and visit i.e. what diagnosis were entered for a patient_id during a visit_occurrence_id; this is more of less possible now thru temporal correlation. The patient profile visualization, lines up events by time – and so events that are correlated temporally should be close to each other.

However, it wont say (right now) the condition_occurrence_id in the same selected visit_occurrence_id (or measurement, procedure, drug etc).

In general, there is a lack of support for FACT_RELATIONSHIP table in OHDSI community. It is considered a legacy table that is being retained for backward compatibility with poor performance associated with it. ATLAS does not support FACT_RELATIONSHIP table at this time - to my best knowledge.

(Melanie Philofsky) #5

Oh, I didn’t know this. We still actively uses this to relate data that belong together and/or are mostly useless without the other data value. i.e. systolic & diastolic blood pressure. Maybe it is time to find another solution

(Gowtham Rao) #6

In addition to the example of SBP + DBP, other examples maybe lab-panels. or even associating two procedures together.

The solution I would recommend is to use visit_occurrence in association with visit_detail. Visit_detail is designed to capture details within visit_occurrence_id, and there cannot be a visit_detail record when visit_occurrence_id is null.

Example in outpatient setting: outpatient primary care visit with a physician visit_concept_id = 8940 . For illustration, lets say the visit_occurrence_id = 123456

  • In that visit, patient had a lab encounter where a panel of tests were drawn. This lab encounter may be a visit_detail_id (say 84543) with the same visit_occurrence_id as above 123456. The visit_detail_concept_id for visit_detail would be 32036 . Each of the lab test results in the panel may now be a record in the MEASUREMENT table with visit_occurrence_id 123456 and visit_detail_id 84543.
  • Also in the visit, patient has biometric such as weight, SBP and DBP. This would be a different visit_detail_id (say 93432) with the same visit_occurrence_id as above 123456. The visit_detail_concept_id for visit_detail would be 44777714 or 32271 or a newly proposed concept_id belonging to VISIT domain (

I think this approach is scalable. visit_detail is still not supported in ATLAS, but the lift to support visit_detail is much much lower than fact_relationship IMO.

Multiple Addresses per Provider, Provider Address differs from Care Site
(Christian Reich) #7

The first half of the sentence is wrong, the second one is correct. Generally, it is not used very much, slow and kludgy: You have to see what it connects to in order to understand the meaning. But we have not found a better solution, yet.

That would only work if you have only one pair of systolic and diastolic blood pressure measurements per visit. But that’s not defined. You can have as many blood pressure measurements in each visit_detail as you want.

(Gowtham Rao) #8

You are right. We need to explicitly identify the measurement panels, and if there are repeats of the same panel within the same visit – then we need to be able to group the panels and seperate the repeats. My recommendation, doesn’t work in the case of repeat of same panel.