OHDSI Home | Forums | Wiki | Github

Prevalence of Autoimmune Disease - OMOP Script available (execution time 3 to 10 min)

General
#1

Hello! As some of you know I have been working to estimate prevalence of autoimmune diseases, individually and as a class of disease. I now have a script that generates the statistics to answer this question. It’s about 300 lines of SQL that uses OMOP. It’s currently written for MSSQL, but I would be happy to convert it to PostgreSQL or any other variant (of SQL that is).

I have watched the OHDSI community for years but have not posted code here and would like some feedback on where to post the script.

I also have a narrative describing the code, the algorithms and definitions used, and links to our website where the list of diseases is posted (currently not up to date with what’s in the code, but it will be updated soon).

All suggestions welcome, and anyone who wants to participate in the final preparation of this paper should please contact me.

Thanks!

Aaron Abend
Autoimmune Registry Inc.

#2

We can create a repository if it were to be a growing project, alternatively if its a script you would like to share, I often use github gists for that. (https://gist.github.com/)

#3

Any reason it is not in Atlas?

Can you share?

#4

Hey Christian, I have direct access to an OMOP database but not to Atlas. I used the Atlas demo system to create a definition for Lupus (based on the same algorithm I am using in SQL) but it took 10 minutes and got zero records so I cannot tell if there are no data or if my logic is wrong. Even assuming it is correct, and I could export JSON, it would take days to create the same algorithm for the other 155 diseases I have SNOMED codes for. If you know a shortcut, please let me know!

1 Like
#5
#6

@aaronabend:

I am not sure I follow. You are complaining that an Atlas cohort takes 10 minutes. How long does it take you to build and debug the 300 line SQL code? 3 hours? What does it actually do? Do you want to invite people to reverse engineer?

Let us know.

#7

This is a common lesson in @Patrick_Ryan’s cohort building classes. A cohort definition can be a cohort of 0. :wink:

Why not use @Gowtham_Rao and the Phenotype WG? Are you sure they don’t have some of the 155 disease already in their library of Phenotypes?

#8

Where is the Phenotype Library? If you mean PheKB, there are quite a few lupus algorithms there, but not even 100 in total. And I need 155 that are based on similar logic so the data are comparable.

#9

The code I wrote computes the number of people with each of the 155 diseases (numerator) and a denominator for the population.

#10

@Gowtham_Rao runs this resource via the Phenotype WG. There’s some discussion here about its evolution. There is a public ATLAS atlas-phenotype.ohdsi.org that you can get access to. I think @admin has the keys to that car. (I can’t remember where the Google form is. Sorry, Lee!)

Sounds like something for CohortIncidence or its ATLAS siblings. Maybe @Chris_Knoll has some ideas.

#11

@krfeeney @aaronabend

I’ve now created an OHDSI wiki page with the link to the atlas-phenotype.ohdsi.org web site and the link to the Google form to request access here:

https://www.ohdsi.org/web/wiki/doku.php?id=documentation:ohdsicloud:ohdsi_phenotype_library

1 Like
#12

Thanks Lee, I just submitted a request to join the Atlas project.

1 Like
t