Jim M If the observation facts are standardized at the CONCEPT_CD to reference terminologies (or, I suppose any network code - but why circumvent the S&I architecture proposed by ONC?) then additional metadata layers supporting different browsing perspectives can be added with little cost other than curation. I expect we will have 2 or more i2b2 workbench views that support different use cases for i2b2. The reference terminologies are required for interoperation beyond our network but are still not well organized for concept selection by the clinical or research user.
Jim C ________________________________ From: Meeks-Johnson, Jim [[email protected]] Sent: Monday, September 28, 2015 8:32 AM To: Campbell, James R; Wanta Keith M; 'Dan Connolly'; [email protected] Cc: McClay, James C; Campbell, Walter S Subject: RE: GPC and ACT synergy Thanks everyone who replied to this query. You’ve been very helpful. We were hoping something like the UTSW or UNMC solutions would work, and it is good to get some validation. It sounds like one data set with multiple ontologies might be the best fit for us (we do have a third local instance to support that I left out of the original query), but we are not yet committed, and need to detail out the larger systems ecology first. I have to admit I don’t quite understand the database swap approach at UW Health, but I don’t see the drivers Keith mentioned as paramount here. I thought I might even hear about some shared tools, but no such luck ;) Thanks again, Jim Meeks-Johnson IU PS. Heh, there seem to be a lot of Jims in this group. I can’t help that, but I have been known to go by JMJ in such situations. From: Campbell, James R [mailto:[email protected]] Sent: Thursday, September 24, 2015 11:55 PM To: Wanta Keith M; 'Dan Connolly'; Meeks-Johnson, Jim; [email protected] Cc: McClay, James C; Campbell, Walter S Subject: RE: GPC and ACT synergy Jim (Meeks) Following the lead from Kansas we implemented identified and de-identified copies of i2b2 extracted from Epic and other data sources including tumor registry. We found that we had much more reference standard data in SNOMED CT, LOINC, RXNORM and HIPAA transaction standards than some GPC sites and expanded the i2b2metadata to support those ONC reference terminologies in i2b2. We found that deployment of polyhierarchy ontologies in i2b2 metadata required some creativity and revision of the i2b2 algorithms for hierarchical aggregation. >From the star schema i2b2 data extraction we create the CDM view of data >tables to support PCORnet research queries. This week at the Data Standards (DSSNI) call we learned that we will be required with CDM V3 to support those data as SAS datasets which means, I assume, that we will be using SAS ODBC to support views of the CDM V3 tables accessible in a SAS application environment by (SAS) code received via Popmednet. >From PCORI point of view this reduces the number of versions of net queries >they issue, but it does impose some unanticipated (from my point of view) >development requirements on our network (query) interfaces. Networked i2b2 >client queries are not part of PCORI's game plan although they may be viable >within collaborations with closely aligned metadata deployment. This latest >pronouncement from DSSNI seems to destroy my expectation that SQL employing >CDM data model would be the basis for interoperation of research data queries >within PCORnet Jim Campbell ________________________________ From: [email protected]<mailto:[email protected]> [[email protected]] on behalf of Wanta Keith M [[email protected]] Sent: Thursday, September 24, 2015 1:16 PM To: 'Dan Connolly'; Meeks-Johnson, Jim; [email protected]<mailto:[email protected]> Subject: RE: GPC and ACT synergy Jim, If you prefer to keep isolated i2b2 systems, you could implement database swaps as a solution. We’ve recently installed SHRINE with other sites in Wisconsin and the systems are a bit different, so we use database swaps now at UW Health (as of a month ago). We chose this strategy not because it’s better, but because the requirements vary, our i2b2 versions vary slightly, data governance reasons, and other reasons. Do you spend more energy on complicating the ontologies and/or data? Or do you spend more energy on complicating the ETL code? Or do you spend more energy on complicating the release process? There’s a fine line of when you have more risks. These risks will change overtime with a health care organization depending on the politics, but right now, database swaps work well for us. Keith Wanta UW Health From: [email protected]<mailto:[email protected]> [mailto:[email protected]] On Behalf Of Dan Connolly Sent: Thursday, September 24, 2015 10:01 AM To: Meeks-Johnson, Jim; [email protected]<mailto:[email protected]> Subject: RE: GPC and ACT synergy Good question. KUMC hasn't crossed into ACT and SHRINE yet, though it's on our horizon. I'm pretty sure some GPC sites have, though; I'll let them chime in for themselves. My impression is that no, a separate isolated i2b2 is not needed. We ask that sites share as much of our ETL processes as is manageable, though; we recommend HERON<https://informatics.gpcnetwork.org/trac/Project/wiki/DevTeams#heron-kumc>, or at least pieces such as TumorRegistry<https://informatics.kumc.edu/work/wiki/TumorRegistry>. We're also using RXNorm, ICD9, and LOINC in GPC, though there's more than one way to integrate those into i2b2. Some parts of our design are more stable and uniform across GPC than others (DataRepositoryManagement<https://informatics.gpcnetwork.org/trac/Project/wiki/DataRepositoryManagement> is probaby the best summary and/or starting place). I see you have an account on babel<http://babel.gpcnetwork.org/>, so you can poke around whenever you like. It'll be great to have stuff from your site any time; ACT style stuff is more than welcome. Another opportunity for comparison is a demo in one of our weekly gotomeetings<https://informatics.gpcnetwork.org/trac/Project/wiki/SoftwareDev#next-agenda>. We've had many of the other GPC DevTeams<https://informatics.gpcnetwork.org/trac/Project/wiki/DevTeams> do that. (We even have recordings of a few, though they're not terribly handy; here's hoping for time to index them on that DevTeams page.) We plan to have UNMC talk about their CDM ETL on the 29th, but any week after that might work fine. -- Dan ________________________________ From: Meeks-Johnson, Jim [[email protected]] Sent: Thursday, September 24, 2015 9:38 AM To: [email protected]<mailto:[email protected]>; Dan Connolly Subject: GPC and ACT synergy Hi, all, Indiana University is pleased to be joining GPC as part of phase II! We are already part of the SHRINE-based ACT consortium, as I believe some of you are also. We build a separate instance of I2B2 using RXNORM, ICD9 and LOINC from our HIE for ACT. We wondered if you had any suggestions or guidance on best practice for leveraging this into GPC/CDM, or if a separate isolated I2B2 and ETL is needed for GPC. What are the rest of you doing about this? This may be explained in the rich technical documentation you have... if so could you point me to the right resources? Thanks, Jim Meeks-Johnson Principle Engineer The information in this e-mail may be privileged and confidential, intended only for the use of the addressee(s) above. Any unauthorized use or disclosure of this information is prohibited. If you have received this e-mail by mistake, please delete it and immediately contact the sender.
_______________________________________________ Gpc-dev mailing list [email protected] http://listserv.kumc.edu/mailman/listinfo/gpc-dev
