Dan and Clinical Observations Interoperability Group members:
Thanks for the interesting discussion on current approaches for the
Clinical Obs Interoperability Use Case (Protocol Eligibility Screening).
I uploaded the latest use case summary (no changes since 9-25-07) and a
word doc with sample protocols to the Resources section of the wiki at:
http://esw.w3.org/topic/HCLS/OntologyTaskForce/BIONTDSEDCM
Even in the 10 protocols that I sampled from
http://www.clinicaltrials.gov <http://www.clinicaltrials.gov/> , which
has its standards for posting structured protocol information on the
web, you can see that there is significant variation in the content,
detail, and structuring of eligibility criteria across protocols. I
believe that these sample protocols are necessary to review to
conceptualize a strategy and to understand the nature of the protocol
data source. Also, these might be a handy reference to ground us
throughout the development phase that we are now approaching. This
variation and non-standardization and multiple content areas (e.g.,
clinical findings, lab results, medication history, demographics, etc.)
will be a big challenge.
You raised an excellent point on the phone that I have been thinking
about this past week. Although there is currently no standard
Information Model for protocol descriptives, there eventually will be (I
hope :-) ), and so maybe that could be our starting point. A
demonstration from one standard (clinical research) information model to
a single healthcare information model (Dr. Huff's detailed clinical
models, DCM) could at least demonstrate the value of a standard clinical
research information model and might even facilitate its adoption...
Either way, I believe that an early task for our group should be to
identify what information model is most likely to represent Protocol
Eligibility in a standardized way. I do not believe that the CDISC's
SDTM is the best model, but I could be wrong. I am fairly certain that
*all* the CDISC models will eventually become harmonized with the BRIDG
model (https://cabig.nci.nih.gov/News_Folder/BRIDG_v1.0_Release/ ) ,
which is planned to harmonize w HL7 models as well.
So, I wonder if one strategy might be to use the BRIDG as a starting
point - map sample protocols to BRIDG as case exemplars, and then map
the appropriate BIRDG constructs to the DCM models in some general,
re-usable way ....
I look forward to talking on this topic tomorrow. Meanwhile, another
outstanding task for me is to think of definitions/measures of "success"
for this project. If anyone has suggestions, please direct them to me.
Rachel