OHDSI Home | Forums | Wiki | Github

How to load hospital attributes into CDM?

(Qi Yang) #1

Our data asset have many hospital attributes. Some examples are shown below:

Hospital bed size range (200-299, 300-499, 500+ etc.)
Teaching hospital indicator
Rural-urban indicator
Full catheterization laboratory indicator
Diagnostic catheterization laboratory indicator

Do we want to load these information into CDM? If yes, do we have relevant concepts in our vocabulary? Even if we have these concepts, how do we load into CDM? Let’s say there is an observation concept on a hospital attribute, but loading into Observation table requires a person_id. Since there are many patients who went to the same hospital in our source data, how should we pick a person_id? Do we create a dummy person_id?

(qiongwang) #2

Not sure about OMOP CDM, database in my hands doing this way:
Use another table for hosptials information. Every patient’s visit should contain the hospital ID he/she went.

(Melanie Philofsky) #3

Hi @QI_omop,

Hospitals are care_sites. And their attributes should be attributed to a care_site_id, not a person_id.

The use case will inform the ETL. What’s the use case for this data?

(Qi Yang) #4

Hello @MPhilofsky

We don’t have any use case. And that is my question as well. I would like to ask the community if there are any use cases for these hospital attributes. If not, we will simply not load them into CDM.

By the way, I searched literature and it seems some study do use number of bed and teaching status attributes.

But I guess if you search literature, you will probably find anything. So how important are these hospital attributes? I will leave it to the community to answer.

(qiongwang) #5

Hi @MPhilofsky, care_sites is a table right? every piece of patient’s encounter information has a care_site_id, right?
Thank you

(Melanie Philofsky) #6

Correct, Care Site is one of the CDM tables.

“The CARE_SITE table contains a list of uniquely identified institutional (physical or organizational) units where healthcare delivery is practiced (offices, wards, hospitals, clinics, etc.).”

Here’s the link to the documentation.

(Christian Reich) #7

As @MPhilofsky said: The hospital is defined in CARE_SITE. But that is really only an address, and no other attributes. Those we capture it in different places:

  • Hospital bed size range (200-299, 300-499, 500+ etc.) - not captured
  • Teaching hospital indicator - not captured
  • Rural-urban indicator - not captured, but inferrable from the location_id
  • Full catheterization laboratory indicator - this is specialty information. Those are captured in the PROVIDER or PROCEDURE table. However, that requires that an actual catherization conducted on a patient. So, if you want to identify Care Sites that do catherizations you follow the logical chain Procedure - Visit - Care Site.
  • Diagnostic catheterization laboratory indicator - same thing, except you need to pick different Procedure codes

If you want put in a change request to the CDM with these use cases.

(qiongwang) #8

Got that, thank you @MPhilofsky
I found the care_site_id in VISIT_DETAIL .

(Qi Yang) #9


The question is not just about that we don’t have relevant concepts in our vocabulary but mainly our current model does not support adding these data elements even if there are relevant concepts in our vocabulary. This is because our model is person centric, so you always need a person_id, say in Observation table, and no care_site id in that table. So for a hospital where thousand of people visited, whose person_id we are going to use in Observation table? And how to tie that observation record to the care_site?

I am proposing to add a new table to the CDM to resolve this.

This may also help to resolve a recent suggestion of adding Region_concept_id column to the Location table.