OHDSI Home | Forums | Wiki | Github

Making network studies easier (teach yourself to create a study) (skeleton idea)

(Vojtech Huser) #1

Last week there was a great discussion about how to make network studies easier.
Our group needs to teach an R expert some basics of creating a study.
I wanted to share some notes on the forum.
We had trouble following the skeleton for comparative study posted earlier.
Instead, we used an endometriosis study as example.

Some notes I ended up explaining (that apply to many studies) were:

  1. to create your cohorts, you use .SQL. You can add those to your package manually or use R code. This would be extras folder and in file package maintenance

You don’t have to use cohort.csv file. You can simply cut and paste from public Atlas outputed SQL for a cohort.

  1. to test your package, (outside readme snippets), you can inspect code in ‘codeToRun.R’ Like here https://github.com/molliemckillop/Endometriosis-Phenotype-Characterization/blob/master/extras/CodeToRun.R

  2. You end up picking a name for your cohort table (e.g., HIVCOHORTS) and also your cohortIDs.

  3. You can also assume a cohort is already created for you and start using other features right away (e.g., follow this vignete to produce table 1 http://ohdsi.github.io/FeatureExtraction/articles/UsingFeatureExtraction.html#creating-a-table-1 )

If the SimpleSQLSkeleton github repo and guidance is ever created (as discussed last week), I am happy to review it for currency (working after any package updates) every 6 months. We also hope to have a series of HIV studies that will be in increasing complexity. (starting with some descriptive studies first) (since our funding mandate for an HIV theme is informatics-ish in nature)

(Kristin Kostka, MPH) #2

+1 on how awesome this is! Nice to see this content making its way out of heads and into the Forums.

Yes, @Vojtech_Huser – I’m actually working on a couple of archetypes for the Skeleton repo. Not ready for prime time, give me a couple more weeks :blush:. I promise to keep you looped in as a reviewer when content is published.

(Vojtech Huser) #3

@krfeeney, any progress on the Skeleton repo? I think this would be a great resource. Seeing more use cases where FeatureExtraction is used (table 1 and more advanced use of it) for characterization would be a great skeleton use case.

(Kristin Kostka, MPH) #4

Hi @Vojtech_Huser - slowly we’re getting there :wink: Competing demands on the documentation front.

We created a bunch content at the Case F2F for the Book of OHDSI around Characterization that involves the tooling in ATLAS, which is powered by FeatureExtraction. @thomasfalconer is ring fencing the inputs into something we’ll put into the book.

The study skeleton will be a sister to this but is behind in the queue. @Elle has a great CKD characterization question she’d like to bring to the community. It would be a good unit test of a skeleton but it won’t happen for a bit of time while she’s crafting her study question.


(Vojtech Huser) #5

Any update on the CKD use case (from @Elle) ?