HCLS/ClinicalObservationsInteroperability/Oct30-2007.html

COI Minutes for 30th October 2007

Scribe: Ton Oniki

• Questions on Functional Requirements Spreadsheet:
  • Kerstin expressed a concern that we won't be able to do all the various types of mapping that are being discussed. (Refer to recent email thread.) As a starting point, she suggests clarifying the various levels of models that are being discussed -- e.g., come to consensus on a common ontology for observation types, clarifying the difference between domain model and record model.
  • Alan further discussed domain vs. record level. e.g., "the valid values for this attribute are a disease". At first glance, this might appear to refer to a domain model (real diseases). But if this is actually saying "the valid values for this attribute are codes in a list of diseases applicable in the institution" then this is a record model. We need to be careful how we express things.
  • SDTM is trying to communicate what happened (doses administered, adverse events, etc.) and possibly assessments. In so doing, it addresses both the record level (is this form filled out right) and the domain level (the sensible way of talking about real world "things").
  • Eric commented that "domain" is getting fuzzy, e.g., when we talk about "organ systems", that's commonly understood to be "domain". But what about a physician, his role, etc.? Is that also "domain" or is it "record" because its peripheral to the patient?
  • Alan thinks the distinction between "domain" and "record" is not so much about focus or main statement vs. qualifiers but instead is about "real world" vs. use (although he notes that the issue of focus needs to be addressed). e.g., a domain model can very well talk about physicians and their roles and participations, since those are real people, things, and events. But when we say we have to put this physician name in this field, we're talking about a record-level model. One heuristic he uses -- consider "wrong" statements: putting a lab test where a disease is called for -- this indicates a domain model putting in two values where one is called for -- this indicates a record level model.
• Vipul reviewed his presentation for the upcoming F2F
  • General comment: remember the audience. Many will not have much clinical background, some will not have much technical background. Cater to the intersection, simplify.
• Planning:
  • Dinner is planned for Thurs night at the Zephyr in the Cambridge Hotel.