OHDSI Home | Forums | Wiki | Github

AllofUs and OMOP

This thread is for discussion AllofUs use of OMOP.

I heard that the workbench should be released later this month (May 2020).

It would be interesting to hear AllofUs team (perhaps invite a speaker from their team to our OMOP weekly meeting) to describe the custom tables and changes they did to the OMOP tables.

Here is their published dictionary of those

It is listed from this page

Column suppression of PROVIDER_ID in all tables is interesting.

Lots of ds_xxxx tables.
Use of Survey table.

and some search supporting tables. I wish table level description for those would be available.

I will probably even start a similar thread for N3C and OMOP. Palantir platform is apparently up as of now.

AoU workbench is running and DUA process is doable.

I was able to query some data in AllofUs workbench. It seems the SQL dialect to use is big query.

For example part of the code looks like

library(bigrquery)

#some code here

data<- bq_table_download(bq_dataset_query(Sys.getenv("WORKSPACE_CDR"),
		sql_code,
		billing=Sys.getenv("GOOGLE_PROJECT")), bigint="integer64")

relevant links https://bigrquery.r-dbi.org/ and https://bigrquery.r-dbi.org/reference/bq_table_download.html

yes, we are still working on getting the OHDSI tools to work. specifically SQLRenderer, which requires rJava. I’m not sure around the timeline for it.

I am pleased to report that package dependency (XML package) is gone and SqlRender now installs inside AoU Workbench just fine.

If other OHDSI-ers have login to AoU workbench - can you please email me (or reply here) (or just like/heart this post. I would like to join forces with other users with such login - as we make standard OHDSI tools work inside AoU Workbench. There is community forum inside workbench but it is not very user friendly (and has limited access options).

Is there interest to start a mini group around AllOfUs?

I am putting together a set of code snippets that work inside AoU workbench and I am looking for other AoU workbench users that also read this OHDSI forum.

For example, if you want to execute an Atlas cohort inside AoU workbench - that is one goal to work on.

Or simply to compare notes and lessons learned. Just today - I submitted a support ticket to support at researchallofus.org about not able to run bq_table_create() and how temp tables or result schema is possible inside AoU.
All interested OHDSIers-and-workbench-users out there, please message me ! I am sure we are all working on exactly the same initial problems.

Btw, recent upgrade details are here https://aousupporthelp.zendesk.com/hc/en-us/articles/360048822531-Jupyter-Notebook-Upgrade-August-26th-2020 (link not public unfortunately). It says

The All of Us Researcher Workbench’s Jupyter Notebook server was updated on August 26, 2020 to include an upgrade of R from version 3.6 to 4.0 as well as some Python package and R library changes.

This is a public reply to AoU. I am re-posting links that I provided on non-public AoU forum so that I can refer to them in email.

Hi Vojtech,
Yes, I’m interested! My students @justinl and Irene Chen both have login to AoU workbench, ad we’ve run up against the same issues you mentioned. Our first step was to get our new Python-based PLP package running on AoU. You can get the link here: https://github.com/justinlimkz/omop-learn-aou
I’d be very happy to discuss further! I’ll talk about this as well in my presentation at the PLP working group meeting on Jan 7th.
Cheers,
David

This problem here Duplicate concepts is evident in AllofUs. Half of sites maps their data to HCPCS modifiers and the other half to the CPT4 modifiers. (the same concept present twice in Vocab tables). The only way to fix this is to stick to the axiom of no duplicate standard concepts. These have IDENTICAL concept name so such are super easy to find out.

HI everyone, did anyone manage to get OHDSI R packages installed and running on the AoU workbench? I am having a lot of difficulties with it. Thanks.

Quinton,

we also can’t run OMOP tools because of lack of connectionDetails. This is in tier2: registered. tier1 is public, tier 2 is registered.

Our workaround for achilles (new script called achilles2) and other work is posted here: https://github.com/lhncbc/CRI/tree/master/AoU/

I heard somewhere that tier3 (restricted) (to be coming in near future?) may address some of that.

Thanks for the information. I am keeping an eye on this. Perhaps we can get to using OHDSI R packages with the All of Us database in future.

t