We’re already doing the first issue. The question is when or do we tackle the second issue.
I guess what I’d do is wait and see based upon customer demand, - how much stuff people wanted us to do for computable phenotype work that was solvable via the PCORI CDM for PPRNs and other investigators - how much we’d benefit from executing that computable phenotype query at a non-i2b2 site like Vanderbilt that has the full CDM. If the answers were a) there were very few questions from investigators that could be answered with the CDM because it lacks meds and labs and registries b) none of the non-i2b2 CDRNs were interested in supporting an inbound cohort request based on an i2b2 query, then I’d wait on the second issue work till after the CDM evolves. Likely need to wait a while anyway because we need to get our data sharing agreements solidified which is a higher focus for me. If a) and b) were false, after we’ve got data sharing signed I’d probably see if Nathan (plus Phillip Reeder and others in the GPC who are working to create the bitbucket standardized ETL code from i2b2->CDM working that we can use repeatably across our network) could get a consult with Shawn’s team about their thoughts and then estimate effort. If this is a tractable problem (1~2 months coding work estimate), I’d probably see if we could do it either out of current effort or spend down my commitment here. If it was really critical, my guess is the constraint may be coding chunks time from someone who’s up to speed; not funding. Would probably know after a week or so if it’s solvable. There’s no GUI work… it’s all just “transformation” of SQL and mapping ;) But, I don’t think we need to think too much about it over the next 3-6 months; need to focus on getting our standardized approach to filling the CDM tables deployed: raw data -> local i2b2 terms -> GPC compliant i2b2 ontologies -> GPC/PCORI CDM i2b2 ontology -> run standardized extract code into -> CDM Still have work to do to finish off that pipeline for all the data the CDM requires and the data we need in GPC compliant ontologies for things like cancer registries, meds, labs, and vitals/flowsheets/observables. Russ On Sep 4, 2014, at 4:34 PM, Aaron A Sorensen <[email protected]<mailto:[email protected]>> wrote: Russ, I think we are actually on the same page. Let's say you had the answers to all your questions below and had formulated a clear plan of attack as to how you would go about accomplishing the popmednet-i2b2 you had envisioned. The next question would, "Who will do all this work?" Based on the conversation during the last call, Shawn and Jeff would say, "Not us, unless we get more funding." At which point, Maryan would say, "I might be able to help you with that if you can give me a compelling abstract and concrete specific aims." So I think, what you wrote below gives the group a pretty good idea of the "flavor" of your preferred specific aims. Aaron Aaron Sorensen Director of Informatics Temple University School of Medicine Office: 215-707-8079 Mobile: 215-341-5033 ________________________________ From: Russ Waitman<mailto:[email protected]> Sent: 9/4/2014 17:02 To: 'Maryan Zirkle'<mailto:[email protected]>; Aaron A Sorensen<mailto:[email protected]> Cc: [email protected]<mailto:[email protected]>; Morris, Michele<mailto:[email protected]>; Shirey, Bill<mailto:[email protected]>; Borromeo, Charles<mailto:[email protected]>; [email protected]<mailto:[email protected]>; Dan Connolly<mailto:[email protected]>; Nathan Graham<mailto:[email protected]>; Espino, Jeremy Umali<mailto:[email protected]>; Mignogna, Linda Kathryn<mailto:[email protected]>; Murphy, Shawn N.<mailto:[email protected]>; Jeff Brown<mailto:[email protected]>;Thompson, Helga<mailto:[email protected]>; [email protected]<mailto:[email protected]>; Wehbe, Firas<mailto:[email protected]>; Denny, Joshua Charles ([email protected])<mailto:[email protected]>; Harris, Paul ([email protected])<mailto:[email protected]>; Basford, Melissa ([email protected])<mailto:[email protected]> Subject: RE: i2b2 discussion Hi All, I was on vacation last week but unless I am missing something there’s a difference between data representation and end user intuitive query interfaces. Not sure the gain with the approach outlined by Aaron and my thoughts are below: Translating between the first is a straightforward though an attention-to-detail set of work while the second can be unbounded in effort or requires constraints that artificially limit its value to users (eg. 3 booleans). I think the initial point of the call was to get people talking and get agreement on the question can you translate data from i2b2 to the CDM? - Joe/Maryan/Jeff/CC: Yes, that is allowable. - Nathan’s GPC code to do it is here: o https://bitbucket.org/njgraham/pcori-annotated-data-dictionary o We now need to standardize across our network the mapping file work (heron_to_pcori.csv); as Shawn says… the devil is in the mapping with i2b2. We want to change this from a csv to something managed all within i2b2 concept paths from local site to GPC standard to CDM standard But for the “query”, right now we’re basically giving Jeff data tables and he’s free to write arbitrarily complex queries in SQL and I assume SAS. I think there might even be challenges with getting simple SQL statements running across different SQL database unless it’s always wrapped by SAS (syntax of Oracle versus Microsoft versus MySQL versus PostGres…). If the query spec was bounded by ANSI SQL, someone might be able to code a “translator” that would convert code designed to work against the CDM to run directly against the same data in an i2b2 schema. That might be interesting but not the same as a “query” that can be created by an end user via the i2b2 interface. So the first issue is would the functional PCORNet Coordinating Center “basic” queries be constrained by or a subset of i2b2’s user functionality? Which version of i2b2? - If yes, then it might be worth thinking about. If the first query from the coordinating center wants to apply a MAX(), no. - I think the answer is likely no. Jeff and if I was a programmer would desire greater flexibility to write code that uses the full data representation of the CDM and not have artificial constraints imposed. - Now we may learn that the management overhead of creating the shadow CDM is really bad but we can cross that bridge later for now. We know we have to write this i2b2->CDM translator anyway The second issue would be is there a place for i2b2 as a non-programmer means to build a query against a PCORNet compliant ontology that could then be shared first amongst i2b2-using PCORI sites and then perhaps equally important, as a means for users to build PCORNet complaint queries? - I think the answer might be yes. I think there’s also a PopMedNet GUI but my sense is more investigators are familiar with i2b2 and it has increasingly rich functionality that’s battle tested by many sites. - An i2b2 query is at its heart translated into a SQL statement that runs against tables. - If the i2b2 sites all have the same PCORNet CDM ontology implemented, it’s the foundation of cross site queries and we can copy our queries between one another and eventually use things like SHRINE - But, if someone could write code that translated the i2b2 query into something that hit the PCORNet CDM, you could have users build queries with i2b2 and then have that get executed against sites who hold data in the CDM but don’t have the i2b2 schema and application running. o Shawn and other i2b2 gurus would need to think if this is possible but my guess is it’s a tractable problem (meaning once written, it’s going to work across the CDM spec) o That would help people author things in i2b2 and share back with PEDSNet and Vandy (was talking today with Vanderbilt regarding computable phenotype). o That said, I don’t think this is top priority at this point since we’ve got to get the first leg nailed down Russ From: Maryan Zirkle [mailto:[email protected]] Sent: Thursday, September 04, 2014 4:18 AM To: Aaron A Sorensen Cc: [email protected]<mailto:[email protected]>; Morris, Michele; Shirey, Bill; Borromeo, Charles; [email protected]<mailto:[email protected]>; Russ Waitman; Dan Connolly; Nathan Graham; Espino, Jeremy Umali; Mignogna, Linda Kathryn; Murphy, Shawn N.; Jeff Brown; Thompson, Helga; [email protected]<mailto:[email protected]> Subject: Re: i2b2 discussion Aaron, This is perfect! Yes, specific aims such as these are exactly what is needed on my end; not to mention the direction they can help give the overall efforts of the group. I agree with your interpretation of the consensus after our last discussion, but welcome the perspective of others along with their suggested aims as you mention. Once we agree on the aims, I would like to determine the who, what, when of each so that I can create a timeline of our efforts and outputs for my leadership. This can also help to give shape to the manuscript you speak about in #5--as to who will own discussion of what. Eventually, I will be finding a consistent time (weekly or bi-monthly) for us to meet, but the calendars aren't aligning well enough until November for that. Best, Maryan Zirkle MD, MS, MA Program Officer, CER Methods and Infrastructure Program Patient Centered Outcomes Research Institute (PCORI) Sent from my mobile office On Sep 4, 2014, at 4:52 AM, "Aaron A Sorensen" <[email protected]<mailto:[email protected]>> wrote: Maryan, Assuming that we are all in agreement that the end result of last call was general consensus that a basic level of i2b2{for PCORnetCDM}/PopMedNet{for PCORnetCDM} interoperability is desirable and completely feasible but just not funded, I would propose that on the next call, a faux grant abstract and specific aims would start to take form so that we could give you something concrete to take back to the PCORnet/CTSA leadership. If people like this idea, maybe anyone who has an opinion could send in three to five sample specific aims e.g., 1. Boarder cases will be explored in order to come up with a crisp definition of “simple query” e.g., no more than three Booleans and no more than two stratifying dimensions 2. Within the confines of the simple-query space, a proof-of-concept (PoC) will be executed first individually (and manually) in both systems, and subsequently using an automated algorithmic transform that takes PopMedNet-Query XML and translates it to the corresponding i2b2 query and then does the reverse translation with the query’s results set. 3. Based on specific aims #1 and #2, identify boarder use cases to guide the efforts of making the PoC production ready 4. Ten development sprints of two weeks each (with end-user testing after each sprint) to harden the PoC code using the use cases identified in #3. 5. Drafting of a manuscript describing the effort to be submitted to JAMIA or other peer-reviewed journal. Aaron Aaron Sorensen Director of Informatics Temple University School of Medicine Office: +1.215.707.8079 Mobile: +1.215.341.5033 -----Original Appointment----- From: Maryan Zirkle [mailto:[email protected]] Sent: Wednesday, September 3, 2014 14:19 To: Maryan Zirkle; [email protected]<mailto:[email protected]>; Morris, Michele; Shirey, Bill; Aaron A Sorensen; Borromeo, Charles; [email protected]<mailto:[email protected]>; Russ Waitman; Dan Connolly; [email protected]<mailto:[email protected]>; Espino, Jeremy Umali; Mignogna, Linda Kathryn; Murphy, Shawn N.; Jeff Brown; Thompson, Helga;[email protected]<mailto:[email protected]> Subject: i2b2 discussion When: Monday, September 8, 2014 13:30-14:00 (UTC-05:00) Eastern Time (US & Canada). Where: Toll-free dial-in number (U.S. and Canada): (866) 802-2104 Conference code: 6225772781 Hi all, We will continue our discussion from last week. Please send me any specific topics for discussion via email and we will quickly go through them like we did on the last call. I know 30 minutes isn't ideal, but it is all that we have with the busy schedules! Best, MZ Toll-free dial-in number (U.S. and Canada): (866) 802-2104 Conference code: 6225772781 Russ Waitman, PhD Director of Medical Informatics Assistant Vice Chancellor for Enterprise Analytics Associate Professor, Department of Internal Medicine University of Kansas Medical Center, Kansas City, Kansas 913-945-7087 (office) [email protected]<mailto:[email protected]> http://www.kumc.edu/ea-mi/ http://informatics.kumc.edu<http://informatics.kumc.edu/> http://informatics.gpcnetwork.org – a PCORNet collaborative
_______________________________________________ Gpc-dev mailing list [email protected] http://listserv.kumc.edu/mailman/listinfo/gpc-dev
