OHDSI Home | Forums | Wiki | Github

Feature Analysis/Cohort Characterization Tool in ATLAS


(Diana) #1

I am working on a project trying to answer the question of COVID mortality rate of the hospitalized among the racial groups. I selected race/ethnicity through the already built-in defaulted demographics race/ethnicity on the ‘feature analysis’ section of the cohort characterization in Atlas. As I tried to run this cohort characterization through Atlas, I noticed on the output that the total percentage of the racial groups does not equal 100%. So, this tells me that the built-in defaulted demographics race/ethnicity selected on the feature analysis did not include all possible values (eg. Null, unknown, others…etc). So I was wondering if there was a way I can edit or create a race/ethnicity for the feature analysis selection?

@schillil @krfeeney

(Kristin Kostka, MPH) #2

Hi Diana!

Sorry to be so slow! Holiday week. You did the right thing bugging me on the forum to ask this. :smile:

Couple of things:

  1. I consulted some ATLAS developers. Some of this is a quirk of WebAPI. WebAPI is designed to filter out stats that are < 1% prevalence – this is for performance purposes (lots of empty cells). This is problematic for demographics because it hides some information.

  2. Also, OMOP standard concepts do not use the flavors of null. We see things at “0 - No Matching Concept” or it’s a valid standard concept…

I’ll tag in @anthonysena, our guru on Characterization, if he has any advice on how to get to the numbers you’re looking for.

Do you have database level access?


(Lisa Schilling) #3

@dir0417 @krfeeney @anthonysena
Kirstin- thanks for the great info!
We’re looking for a guide or instructions on to create new features. There is another post about how to so by Chris Knoll, but I think it’s a bit outdated and it led to an issue being created in github (How to create own features - Atlas Characterization - Summary Statistics). We searched the Book of OHDSI but didn’t find anything. Has anyone made an internal cheat-sheet or any guide? Thanks! Lisa