CDISC models clarification and context of use case for EHR - Clinical Research Interoperability

For those interested in the Clinical Observations Interoperability
activities, I would like to generate some discussion on this topic as I
attempt to outline a use-case these next 2 weeks.... So I will share my
thoughts on the problem context so far:

 

I am trying to determine the over-arching use-case context for this
Clinical Observations Interoperability demo, in order to determine the
appropriate CDISC model for this pilot project. I think a broad context
(e.g., EHR --> screening for trial eligibility, or EHR --> Phase1 trial
adverse events) needs to be defined here first; and then specific
examples, such as the breast cancer domain, and a project strategy (OWL,
RDF, etc) can easily be generated.  However, I am having considerable
difficulty in identifying a context and justification for the use case
and wanted to solicit feedback from the group before our next call on
8/28.

 

I had originally thought that a clear and easy context, one that would
be relevant both to investigator-initiated and industry-sponsored
studies, would be the use of EHR data to estimate study population sizes
or to identify persons to participate in trials. But, this would not be
a "regulated" use of the clinical data, so I do not even see CDISC
standards (at least current ones) logically fitting into that scenario.

 

The CDISC Study Data Tabulation Model (SDTM) is the standard for the
regulatory *submission* of clinical trials data (implying that the trial
has already been conducted). The SDTM standard is designed only for the
regulatory submission, and is not necessarily an operational standard.
I assume individual pharmaceutical companies store their data in a
variety of different models (for which none are "standard") and then
they convert the data into SDTM before submission. So, if the group
wants to use SDTM as part of this demo, a use case based upon sharing
EHR data for a phase 1 or 2 trial would seem more appropriate. However,
I am not convinced that EHR data is collected in a structured fashion or
at time points that would make it useful in that scenario (e.g., Do all
study visits necessarily collide with a clinical visit? If so, does the
clinician need a formal association with study or training in order to
collect the data with rigor required by the research protocol?)  

 

CDASH is still early in development and focused on determining a set of
(minimum) required elements for data collection (forms) in certain
areas, such as adverse events, medical history, demographics, etc. There
is significant potential for overlap in scope/activity of CDASH with the
SDTM activities, and the CDISC organization has just recently developed
processes to make sure that the terminology suggested by CDASH does not
conflict with what are becoming close-to-final standards for SDTM
terminology.  Since CDASH is really a data capture standard, it seems to
be more appropriate to use as a target model in this project, but it is
really quite new and I am not sure when it will become a finalized CDISC
standard.

 

We could assume that SDTM represents (or will represent) a (FDA)
standardized data structure, terminology, and code sets, and that CDASH
will be harmonized with it some day (soon I hope). In that sense we
could justify its use, but it would be nice to find some business
justification for EHR-Clinical Trials interoperability that could be
appreciated by a wide audience.  [[What about post-market surveillance
for AEs? Or populating the EHR with clinical trial data? Or
retrospective studies that abstract data from EHR, but these are not
typically FDA-regulated, are they?]]

 

What is very clear to me, as most of you know and we discussed on the
phone last week, is that Lainden Baines and others from CDISC should be
very involved in identifying the use case, and certainly he will be able
to correct any of my mis-understandings of the CDISC standards. Either
Vipul or I hope to be in contact with him next week. I hope that those
in the group with understanding of regulated research environment or
CDISC standards will join in this thread as well.

 

Once a context is fleshed out, then I think it will be quite easy for me
to find examples of protocols that require certain information types
(e.g., lab, diagnosis, imaging, genetic) that we would like to
demonstrate interoperability between the EHR (via Stan Huff's Clinical
Data Model structure) and the appropriate CDISC model. 

 

The use-case and context are critical to the success of the project. I
hope that those of you on this list serve with relationships/experience
with pharmaceutical industry can help focus me on this, because I am
struggling with it. Meanwhile, I am still thinking hard on this....
Have a great weekend.

Thanks,

Rachel

 

p.s. - the documents at the bottom of the wiki
(http://esw.w3.org/topic/HCLS/OntologyTaskForce/BIONTDSEDCM ) provide
critical background on this topic.

 

 

Rachel Richesson, PhD, MPH
Informaticist and Assistant Professor
Pediatrics Epidemiology Center
USF College of Medicine, Department of Pediatrics
3650 Spectrum Blvd., Suite 100
Tampa, FL 33612
Office: (813) 396-9522
Fax: (813) 396-9601 or (813) 910-5922
Email: richesrl@epi.usf.edu