HCLSIG/PharmaOntology/Meetings/2010-05-05 Conference Call

From W3C Wiki
* Call with Marshfield Clinic
* Present :  luke  chris  conor  bosse eric elgar
* This was an excellent meeting.  Extremely insightful
* Discussions :   (see summary at the bottom)
* Algorithms are central and from eMERGE - They are applied separately among institutions in network
* Longitudinal Data - Marshfield has longitudinal data extending back into the early 70s or 60s!- this makes marshfield unique
* Personalized Medicine Research Project - this describes patients/data involved 
* CPT codes
* Snomed and RxNORM   - in the process of trying to map
* Doing their best to try to standardize their data
* Site-specific codes taken into warehouse ON SITE - all internal - even have a research side to system
* Are companies involved(outside)?  It appears that most of what is being done is being done by employees of Marshfield 
* Data can be queried  as a whole detail depends on the query - If drugs - stored NDC from 2000 not before - Mixture of issues ICD codes vary - specialist physician may be interested in why type of cataracts rather than a primary physician who may only be concerned with the fact that someone has cataracts - this makes a massive difference in the detail and linkage we need. ICD will suffice for the primary care setting but need deeper tools and representation for more advanced query.
* Often many things are only documented and in text rather than coded for control of query.  These need to be extracted - discussed ways of doing this
* Marshfield is in the process of coding many different things to enable this to happen.  Medications are a big source of concern and current work
* Summary of Events :
* There are two big ways we can work together
* 1)Discuss a SPARQL endpoint and RDF for implementation over some of marshfield data. Luke and Conor
* 2)How can we translate the most basic of a spectrum of clinical queries into an answer from the marshfield data using the approach in #1 and a proposed user interface?  More importantly, how can a simple physician/researcher apply meaningful concepts into an interface that will relay this to LUKE and retrieve an answer?  How can this be done in the future using this interface without requiring LUKE to manually perform operations to obtain answers? 
* We tried to refine the day's work into only two select realistic goals:
* A)LUKE and CONOR to discuss endpoints and RDF as a basic step - Implementation time
* B)BOSSE CHRIS ERIC ELGAR to oversee and develop a small list of questions on the far end of a continued Spectrum of questions to provide to luke to then trial over the data.  Keeping in mind that development of an interface should accomodate the most basic of individuals (the clinician- often with NO computer literacy) - perhaps a simple query page as a start to force the query into a digestible format for luke will facilitate this process.
* The point :
* We must learn to crawl with the data before we move on to genomes and enable basic clinicians to access medical record data. 
* We may already know the answers to some of these, however they require basic answers: ie. out of all of the patients in the data base presenting with a fractured neck of femur, how many experienced a pulmonary embolism during the admission and discharge from hospital? If so, was it the cause of death?
* How many people with a diagnosis of COPD or pulmonary fibrosis had work that may be related to noxious substance exposure mentioned somewhere in their medical records?  What are the prevelance and incidence?
* More to come, but you get the picture - Its not only the complex questions regarding genotype-phenotype correlations that change practice