OHDSI Home | Forums | Wiki | Github

Metadata and Annotations WG


(Ajit Londhe) #22

Hi all,

Just got back from vacation and am sick, so I’m re-scheduling our next call for May 16 at 2 pm est.


(Hanieh Razzaghi) #23

@Ajit_Londhe I think I didn’t realize the meetings have already started. Was there an invite that was sent out? If there wasn’t, no problem - just want to make sure that my name is added to an email distribution list if something official went out. Thanks!

(Ajit Londhe) #24

No worries, we met on April 27 and kicked off the goals and scope. The meeting dates are posted in the Wiki:

I’ll be posting the meeting minutes from our first meeting by this Monday.

(Hanieh Razzaghi) #25

oh, yikes, I’m sorry I missed this… will be sure to join the next one. Thanks!

(Ajit Londhe) #26

Meeting slides are now posted on the Wiki:


(Ajit Londhe) #27

All – Due to Memorial Day weekend and early office closings, I have rescheduled tomorrow’s call to Wednesday May 30 at 2 pm est. Details are on the Wiki.

(Vojtech Huser) #28

I missed the initial meetings and wanted to provide one comment after studying the slides.

  • Another type of metadata are historical facts about organization. E.g., XeiserFenente transitioned from EHR A to EHR B on May 7th, 2001

  • Or, XeiserFenente acquired 1 new big hospital on June 1, 2007 - hence more admissions are recorded. (assuming encounter locations are not in the data)

  • There was epidemic of Ebola starting on date xyz. (which annotates the data) (100% change in monthly trend is explained by this - it is not an error)

  • Since June 7, 2016 - we link mothers the babies and we use this EHR mechanism or this Payor-based data mechanism

So statements about a single concept (new death type introduced), about trends, but also about the institution (or location or environment) which generated the CDM data.

(Vojtech Huser) #29

Here is an elaboration on the use cases I proposed on today’s call

  1. authoring study protocol or result publication
  2. determining study feasibility

I proposed to take few existing studies (e.g., canagliflozin study), and look at metadata needed for use cases above.

For 1, this page 18 from the protocol shows dataset level metadata (“description”). Existing METADATA table may support that.

second example is this table in manuscript

(Vojtech Huser) #30

Jeff Brown just presented metadata on Sentinel.

Some highlights are

full document at

(Vojtech Huser) #31

another great candidate for metadata from Martijn https://github.com/OHDSI/SkeletonComparativeEffectStudy/blob/master/extras/CodeToRun.R#L26

(Vojtech Huser) #32

Please review and provide feedback on Themis proposal for metadata slight change in convention.
Ideally, we would need a concept_id for it. https://github.com/OHDSI/Themis/issues/46

(Vojtech Huser) #33

I wish the workgroup would resume work again in 2020 or 2021. We now have more experience with network studies. We had high ambitions in 2018 and maybe we aimed “too high”.
Can we have fields for ID, Name and Description. From here https://data.ohdsi.org/PhenotypeLibrary/#shiny-tab-databaseInformation

We have a nice metadata table. All we need is either nice concepts (may take a while) or simply use same name in this table https://ohdsi.github.io/CommonDataModel/cdm531.html#metadata (just 2 columns: name and value_as_string).

See examples.
A text analysis of the description would also give us some nice guidance and what goes into PUBLIC database description. (description always safe to share anywhere).

Many are longer than max allowed 250 chars in current METADATA table