RE: tabulator with clinical data

Tim,

Thanks for your response... Here is the link to the public data which  
with Stephen Dobson (Pfizer) has kindly provided us...
http://esw.w3.org/topic/HCLSIG/Drug_Safety_and_Efficacy? 
action=AttachFile&do=get&target=DummyDataSDTM.zip

In it are six (4 main) tabular files (as excel docs):
1. demography (dm) - trial subjects
2. experimental design (ex) (along with treatment elemnts and  
treatment arms); drug administration can change depednign on the  
design arm.
3. vital signs (vs) (taken during clinic visits)
4. Adverse events (ae); negative responses to drugs in trial

The data in CDISC's SDTM format (Study Data Tabulation Model), which  
is a tabular form that is supposed to be converted to an XML form  
called ODM (not too well implemented). By using RDF, there may be  
possible saving for a lot of folks (pharmacueticals, clinics, and  
regualtory agencies) in avoiding XML parser-writing grief...

The three key ids used are USUBJID, studyid, visit date (these  
probably should all be URIs, and visit date as well); In addition, we  
think each "Observation" (AE, VS, or Intervention) should also be a  
URI, so that annotations and analyses on them can easily refer to any  
study component.

This would also allow looking at the data in different ways: by  
person over many visits, across all subjects with similar adverse  
events, treatment (arm) type and count of adverse events, etc. Here's  
where a good table viewer is VERY important !

We're trying to see how well RDF-OWL works when applied to the STDM  
model, and see what additional functionality would now be possible  
(e.g., mechanisms of action or toxicity, patient biomarker and  
genetic inclusion, classes of drugs and their target proteins). This  
kind of information would play a critical role in looking at multiple  
studies longitudinally. The other kind of lat/long could be assigned  
to clinics where studies are done (subject privacy policies may  
constrain doing this with patient info).

-- but be sure to check out CDC's epispider (www.epispider.org) which  
seems to use geo locations info in localizing disease surveillance info.


Eric


-----Original Message-----
From: Tim Berners-Lee [ mailto:timbl@w3.org]
Sent: Sat 3/31/2007 12:25 PM
To: Eric Neumann
Cc: public-hcls-dse@w3.org
Subject: Re: tabulator with clinical data

(Apologies for the delay, my email filters have been misbehaving.)

On 2007-03 -11, at 17:16, Eric Neumann wrote:

 >
 > Tim,
 >
 > As part of our work in the HCLS drug safety and efficacy task force
 > ( http://esw.w3.org/topic/HCLSIG/Drug_Safety_and_Efficacy), we're
 > thinking of using your Tabulator for a proposed demo since the
 > clinical SDTM model (ala CDISC) is pretty much based on a table
 > system.
 >
 > We have been working refining CDISC's SDTM to take better advantage
 > of RDF-OWL, but many practitioners out there are used to also
 > seeing clinical trial data in rows and columns, which would seem to
 > be a good match using Tabulator. Pfizer has already provided us
 > with some *manufactured data*, which we're using to get a first
 > pass RDF model working.

Is the manufactured data public?

 > This data is (pretty closely) organized as follows:
 >
 > - Clinical Trial Study
 >     - Patients
 >         - Clinic Visits
 >             - Observations (treatments, findings, interpretations)
 >                 - Vital signs and biomarkers
Great!   Clinical trial data is one of the silos of HCLS field which
I mention in each talk about HCLS integration.

What about the drug/treatment involved? Is that constant for each
study, or do people track administrations of the drug?
 > Since many clinical researchers are used to seeing this data in
 > tabular form (along with statistical analyses), we thought
 > tabulator might be a straightforward way of visualizing the RDF
 > formatted data, and possibly using maps and timleine views as well.
 > I will try a first pass, and see if the others find it useful as
 > well, but your group may be able to give us pointers of we get stuck.

At the moment, it is term time, we have Jim Hollenbach working
couple of afternoons a week, currently making a Ffox  extension, to
give us more power to do stuff.  I haven't touched it for a while but
I would like to clear time to do it.

I wonder also how we could show it linking with other data might
work.  Do you have lat/long for patients?

When it comes to statistical analysis, it would be  good to add that
to Tabulator.  There was an open source AJAX toolkit which I noticed
a few moths back, but lost the pointer to it. It would have made a
neat addition wrapped up as a tabulator view.

 > Will let you know how this proceeds...
Do.  It sounds exciting.

 > cheers,
 > Eric
 >
 >
 >