OHDSI Home | Forums | Wiki | Github

Phenotype Workgroup

Hello,

We are choosing to switch the mode for this collaboration to emails and forum posts rather than calls for now. So this will server as the official thread of the group where updates will be presented. If anybody has a something that is highly relevant to the group and will prefer to hold a call, we can accommodate on a case-by-case basis.

For the people that want to join the group and need to get caught up here is the wiki page of the group http://www.ohdsi.org/web/wiki/doku.php?id=projects:workgroups:phenotyping

Current updates for this week have been outlined during Tuesday’s call and are listed here:

APHRODITE released 1.2 on November 30th, 2015.

  • Anchors incorporate on version 1.2 - November 30th, 2015.
  • Fully Oracle/Postgres/MSSQL server compliant as of version 1.1.
  • 6 sites have attempted building a model- 2 sites failed because of data in CMD v4 version issues (Columbia and Vojtech)
  • 3 sites successfully finished all tests (Stanford, Janssen and Regenstrief - as they have data in the observations and conditions table regardless of the source)
  • 1 site has successfully executed the Anchors code. NYU but they have only 10K patients, making
    it difficult to have enough patients to build a full model.

2015 accomplishments:

  • four models built- used in patient-provider partnership (find fh)
  • anchors code has been fully incorporated.

2015 Issues:

  • need place for notes / terms in CDM
  • currently depends on notes data being placed in the observations table.
  • need partners with data in CDM v5

Plans for 2016:

  • get three sites to run APHRODITE
  • adding extra functions for more granular cohort selection.
  • get 20 or so phenotype models built

dependencies:

  • NLP WG for CMD extension
  • Repository for phenotypes.

We have devised a creative way to build and validate models which will save sites work and effort, so we will be pinging the sites that have data/have run aphrodite individually for this. If your site has CDMv5 data and want to join in, please send us a direct message to get you up to speed.

Thanks!

t