NIH | National Cancer Institute | NCI Wiki  

  • Created by Unknown User (endlecm), last modified on Jan 06, 2017

 Attendees

NameRolePresent
Wright, Larry NIH/NCI  x
Fragoso, Gilberto NIH/NCI     

Safran, Tracy

NIH/NCI [C]x
Ong, Kim L
ISx
Wynne, Robert    NIH/NCI [C]x
Lucas, JasonNIH/NCI [C]x
Bauer, Scott  Mayox
Stancl, Craig
Mayox
Endle,  CoryMayox

 

Agenda

Goals:  

  • Better understand the struggles and issues associated to loading value sets.   Once those are identified, understand the actual requirements.
  • Once requirements are agreed upon, propose solutions to satisfy them.  
Additional requirements
  • Value set resolution
  • Search and manipulation

Discussion Points: 

Decision Points:

Review Value Set workflow process (Tracy and Rob)

  • The goal is to identify the areas of the workflow that cause the greatest issues. (for example, long load time for 700 value sets)

Discussion Points: 

  • What are the requirements for resolving value sets?
    • Performance parameters?
  • Rob mentioned there are 760 resolved value sets - loaded over the weekend.  Last one took about 2.5 days.
    • This is run once a month.
    • Is there a simpler/quicker way to resolve these value sets?
    • Better logging would help to determine if there were any failures for resolving the value sets.
    • Rob suggested that if the script ran over night, this would be acceptable.
  • Scott mentioned we could make the value sets transitive.  This would make loading resolved values sets perform better.
  • Rob questioned if we can load multiple values in parallel.
  • Kim asked if we need to resolve all of these value sets every month.
    • Produce a large OWL file that contains entities with the value sets. 
    • This file needs value set membership information.
    • Can search against this large file.
    • This is all in the LexEVS environment.
    • Kim would need a convenient API method to generate this output file.
    • There are a few value sets that are not from NCI Thesaurus.
  • We will need to verify that the value sets are consistent from the current method of generating them to any new method of creating them.
  • Tracy mentioned that most value sets (about 700 of them) are concept in subset.  The hierarchy of the targets are about 2 levels deep.
  • Most of the value sets can be built off from the information in the NCI Thesaurus.

Decision Points:

  • Requirements
    • Ability to audit and simplify the workflow.
    • Better logging to help determine if there were any failures for resolving the 760 value sets.
    • Rob suggested that if the script ran over night, this would be acceptable.
    • Ability to quickly resolve value sets on the fly.

Discuss and capture issues for further review.

Discussion Points: 

Decision Points:

 

REFERENCES

Value Set Issues Identified during 2016.12 Technical Face-To-Face

PriorityIssue
1Investigate ability to be able to determine if Resolved VS coding scheme has changed.
1Investigate ability to update as needed (not have to load all 700 at the same time). Decrease time it takes to publish/deploy.
2

Investigate ability to be able to determine if Value Set Definition has changed.

2Investigate ability to capture "Hierarchal Value Sets" as coding schemes with hierarchy. (inside value set)

 




  • No labels