OHDSI Home | Forums | Wiki | Github

Including deprivation indexes into OMOP Vocabulary

For working with SDoH (Social Determinants of Health) data especially aggregate measures, it would be great to add some commonly used deprivation indices into OMOP vocabulary.

For now, there’s a Townsend Deprivation Index already exists in vocabulary (Athena).

Is it possible to add these two popular indexes into vocabulary as well?

Also the Townsend Deprivation Index is in measurement domain which I thought it would come from observation domain. Any insights on which domain those indexes belong to are really appreciated.


@cukarthik FYI

Hi @xj2193

We can consider adding new concepts to the OMOP Extension the same way we did for the 1st one.

When we added the first one, the most concern was that these indexes are supposed to characterize the place of living, not the actual patients, while OMOP is a person-centric model.

How do these data end up on the patient-level records?

Thanks @Alexdavv ! Yes those indexes are area based characterization but not from individual patient’s data. The ideal world is to have individual SDoH collected from EHR which is an on-going work. For now, it’s a proxy of a person’s socioeconomic status in this neighborhood, such as education, household income, access to phone etc.

If we have the patient geocoded location (lat, long) in OMOP 5.4, we are able to link these indexes to patient-level record. And they could be used to estimate the deprivation status of a patient.

If these indexes are not linked to the patients, how do you record them to OMOP? What is the person_id of the respective Observation/Measurement record?

I think it makes more sense if we link the ADI and SVI not on person_id in person table but on location table with geocoded location.

@xj2193 The problem is OMOP is person-centric. That is why we generally do not allow concepts that can’t be assigned to specific patients. In your case they could (even though you describe it as ongoing work), that’s why we will add them to the OMOP Extension vocabulary.

But what you described further is a sort of OMOP hacking that we can’t expect other users will do.
I hope ongoing work will be finished soon so you’ll be able to use these concepts properly. :slightly_smiling_face:

@Alexdavv yes you are right. These indexes are not patient-centric but location-centric. Adding them into OMOP extension sounds good.

@Alexdavv @xj2193 Maybe I’m missing something, but using the patient address (number, street, city, state, zip), a lon/lat geocode can be created using a geocoding service API, which then can be converted to FIPS code using the census block API, and based on the FIPS code, the ADI index can be located in the Neighborhood Atlas. So there should be no problem to link ADI to a person, thoughts?


For this to happen, three things need to be accomplished:

  1. You need an OMOP Expansion or Extension (extra table or early testing table later to be added to the CDM), which has to go through @clairblacketer’s CDM WG.
  2. You need to get the ADI and SVI added to the OMOP Vocabularies. Though it is not clear to me how you turn percentiles or deciles into concepts, except creating concepts like “47th percentile” and use it that way. But I am sure that can be done.
  3. You need to Github, test and describe the address mapping to long/lat and that to the ADI API.

You also need to tweak the methods so they will use the quantitative measures as opposed to dichotomous covariates. But I guess you could live with that shortcoming for a while by bucketing the index values.

And a use case showing everything soup to nuts would be nice, as well.