| OHDSI Home | Forums | Wiki | Github |
All,
The latest version of the cohort UI (to be used in CIRCE) is available at http://www.ohdsi.org/web/cohortui/index.html. This version includes all of the entities we filter on, as well as a major re factoring of the code (for any who looked at the code from last year).
Next steps will be to take the data collected by the UI and create a CDM query that will generate the cohort results.
This looks great Chris. Thanks for sharing out.
Codesets are superb. Can you throw a select all [] up atop the concept
list?
One possible inspiration for specifying cohorts and for GUI creating them is Jigsaw tool.
See demo here
Jon,
It’s possible, but it could lead to breaks (imagine selecting all with thousands of concepts). The idea behind code-sets is to enable the selection of maximum concepts with minimum concept clicks: Imagine the case where you select Depressive Disorder, but don’t want to include anything under Bipolar Disorder. So, in the UI, select _one_concept for Depressive disorder, check ‘include descendants’, then select one other concept for Bipolar Disorder, include descendants, and, voila, you’ve selected 90+ concepts.
Also, the current UI of showing selected concepts is only really good for showing up to 20 selected concepts. After that, it becomes a wall of Concept labels (the white boxes that show the concepts). Frank talked about moving that over to a tabular views and at that point, i think having a select all makes sense, but i’d really like to see users of the app understand how the concept hierarchy works and leverage that when making codesets instead of resorting to text search + select all.
Does this make sense?
-Chris
This is from the Outcomes Insights guys, who collaborate on OHDSI projects.
@Chris_Knoll Yes that makes good sense. Patrick took me through the use of descendants in Heracles which would reduce the desire to do a mass selection. It is, as you said, about helping users understand the hierarchy and how selection of items may impact the analyses.
When we built our UI for MedDRA selection, it was easier because far less depth and variety to deal with. So it worked pretty well there (for text selection), but I can see the problem with the full vocabulary.
Jon
Hi Vojtech –
Thanks for digging up our demo (from AcademyHealth last year). We are involved in the cohort working group too.
Just to clarify a little, our GUI (Jigsaw) is for building entire studies – inclusion criteria, exclusion criteria, time windows, baseline exposures and outcomes. And it is designed to work against data in CDM v4 (at the moment) to extract all of the relevant records for the study and output them into an analysis data set with documentation. It includes a library of algorithms that can be assembled (like a jigsaw puzzle) into a study. We are actively developing it for internal use at the moment, but we can demo it to anyone who is interested. The underlying language, ConceptQL is open-source and available on GitHub.