OHDSI Home | Forums | Wiki | Github

Frailty in All of Us and OHDSI

Working to figure out how the frailty concept was devised. ID #4086506, CODE: 248279007, Condition domain, Standard Clinical Finding, Frailty. Want to use it for sensitivity training of a phenotype. Specifically, I’m looking to understand the history of this code. For example, is it based on ICD or other type of claims documentation? A little birdie told me that @Christian_Reich or @Alexdavv might know something about this??
Thanks for any help anyone has!!


Not sure I understand the question. This concept is a SNOMED concept. It was added 31-Jan-2002. We don’t know if they just decided one sunny morning in January that this would be a useful concept, or if they got the inspiration from another vocabulary.

Why would that matter? What are you trying to achieve?

AoU can have problem list data from EHR. Look in source value column. Run Achilles on AoU (or similar approach) and you can find out if it is mapped from some other code or more “direct”.

Concept having the date of January of 2002 places it at the very first SNOMED release, meaning it was directly inherited from Read3. It also did not change much since then.

As for concept meaning, SNOMED is very strict in this sense: whatever is the meaning of the concept, it is completely and exhaustively contained in its name. It’s just frailty.

Thanks @Christian_Reich @Vojtech_Huser and @Eduard_Korchmar for your replies. We are devising a frailty index similar to Rockwood and Dae Kim FIs using All of Us. But we need to do a sensitivity analysis to be sure we are actually capturing frailty in a meaningful way. To that end, the frailty concept exists but since I’m new to all of this I’m not sure what goes into that concept. or is it just ICD and doc notes identifying people as having frailty.

@Vojtech_Huser I’m interested in your idea to run Achilles on AoU but honestly have no idea what that entails. Would be interested in a further discussion if you have time? Thanks!

It’s just that. Somebody encoded frailty, using a standard code or by scribbling into the EHR. If you want to introduce a scoring instrument that would be a new concept.

1 Like

You can use the following query against your source data to find the provenance and original code:

SELECT condition_source_concept_id, condition_type_concept_id, count (*)
FROM condition_occurrence

Since this is US EHR data, you might find more granular data in source Flowsheets. Many indexes/scales are present along with their individual components and scores in Flowsheets.

ok perfect thanks so much for the help!

@MPhilofsky Thank you for the code! we will apply it and see what we get.
Much appreciated!!

1 Like