HCLS/ClinicalObservationsInteroperability/27-coi-minutes.html

Agenda Item 1) mailing list: Eric P. will set up a new mailing list for the COI project. Vipul action item to add email addresses to it.

We won't try to get people "invited expert" status right now. We'll wait for a need (e.g., registration for an event).

Agenda Item 2) people are encouraged to try out the ircatwork IRC

Agenda Item 3) Updates on getting anonymized patient data-

Vipul: Learned what forms need to be filled out, what the policies are at Partners. As long as pt data doesn’t leave firewall, it's easier for him to get the data. If you need deidentified data, he can still get it, but it’s harder.

Tom: Found a series of committees and steps to go through. Needs to investigate further to see if it's feasible.

Jyote: his unit head wanted specific requirements – what data, how will data be used, what’s the expected outcome. Jyote will write up something on how we’re going to use the data. This will be needed again and again as we move forward.

George: familiar with the forms he needs to fill out. Vipul will run his forms by George to see if he’s missed anything.

Agenda Item 6)- George Allen feedback on Data Requirements:

Variables of interest are data elements related to a disease, e.g., in the case of hypertension, variables of interest help us understand why the patient is hypertensive in the first place. A suggested approach is to focus on a disease and its variables of interest.

Should we narrow down our scope to focus on hypertension? Let’s explore. Alan concerned that it’s so prevalent. The assumption is patient consents will be needed to release data to us. Since hypertension is so prevalent, we might not be able to get necessary consents, and hence not much data. George clarified that at this point, we’re just using aggregated, deidentified data, so consents aren’t needed.

Would diabetes be another candidate? Joyte to speak with people at Mayo.

Is there any way to narrow focus other than by disease? George -- Cholerectal screening is a possibility. But there’s probably not a clinical trial associated with screening.

What about diagnostic processes? Are there clinical trials around diagnostic processes?

For now we’ll focus on diseases to constrain data.

Update: during the call, George found clinical studies related to screening. So it may be possible to

Agenda Item 4) -- Tom to give feedback on DCMs for Data Content Requirements

Tom has begun to go through models, marking ones he had. He'll forward to Vipul. George’s scoping/constraining will be a help here, identifying a narrowed down set of DCMs that are needed.

Question -- Is there any alignment between SDTM and CEM? Not deliberately, the intent is to see how semantic web can be used to map independently created models.

Agenda Item 5) -- Vipul update on Goals and Tasks

Vipul has begun charting tasks – at some point they need to have owners. Right now he’s looking for feedback. He went through his initial draft.

1. Identify key data elements – providers in our group need to complete this task.

2. Get anonymized data (pre consent phase, we don’t worry about consents)

3. Identify sets of DCM models -- Tom Oniki

4. Identify sdtm models

5. Design set of mappings – there are concerns here: Chimize has created problem oriented ontology, alan rector has an ontology, hl7 has models, etc. This is a very focused task. We will do these mappings specifically for the data items George identifies.

     Also, we chose these modeling paradigms because we wanted to support standards.  The Intermountain group hopes to align their DCMs with standards or become the standard.  Dan Russler has suggested the HL7 models themselves.  We should bring him in.  SDTM is the emerging standard in the clinical trials world.

     User is on the clinical trials side.  We’ll need some translation from SDTM to DCM.

     Question: Stanford has developed the epoch ontology for managing clinical trials data.  Is this applicable?  No, because we’re retrieving EMR data, not managing clin trial data.

6. Implement a patient data store

7. Implement mapping repository.

8. Implement a reasoning engine -- could be rule-based, could be OWL-based

9. Integrate

Feedback on resource commitments needed. Please let Vipul know.

Next week, Susie to coordinate call.