- From: David Booth <david@dbooth.org>
- Date: Wed, 14 May 2014 18:09:40 -0400
- To: "Richards, Rafael M." <Rafael.Richards@va.gov>, 'Vipul Kashyap' <kashyap.vipul@gmail.com>, 'Michel Dumontier' <michel.dumontier@gmail.com>
- CC: 'Anthony Mallia' <amallia@edmondsci.com>, "'Prud'hommeaux, Eric'" <eric@w3.org>, 'Claude Nanjo' <cnanjo@cognitivemedicine.com>, 'w3c semweb HCLS' <public-semweb-lifesci@w3.org>
Hi Rafael, On 05/13/2014 05:29 PM, Richards, Rafael M. wrote: > What problem is RDF solving that HL7 is not? Great question! RDF complements and supports information standardization activities. The problem is that each information representation standard becomes its own interoperability silo. Each one attempts to achieve interoperability by expecting the world to conform *its* data format or *its* data model or *its* vocabulary. If the entire world did, then indeed the interoperability problem would be solved. But for many reasons, the entire world does not -- and never will -- including: - Political or business reasons. For example, proprietary information representations are still a sad fact of life in healthcare. - Systems cannot all suddenly change at once, in one synchronous big bang, to use a single new standard. At least until we reach nirvana, when all healthcare systems speak the same language with the same meanings, we MUST accommodate a diversity of standards and versions. - The world is continually changing, with new medical techniques and technology continually requiring new concepts that must be standardized. - An individual standard cannot cover all use cases -- at least not well -- or the standardization committee would never finish its work! Each standard typically tries to address a particular range of use cases, excluding or poorly addressing others. But there is no end to the use cases, because healthcare connects with everything else. - Different use cases require different kinds and granularities of information. A blood pressure measurement that fails to indicate the patient's position (standing or sitting) may be good enough for some purposes but inadequate for others. When an automated blood pressure machine is used, details about its make and model may be irrelevant to most use cases but may be critical for a few that are looking for subtle trends and thus may be sensitive to systemic differences between models. Fine grained information is great when it is needed, but harmful when it is not needed, because it adds complexity. - The data formats, models and terms chosen in a particular standard may be distasteful or hard for some parties to use. The designers of a particular standard may not have always made the best choices, though the notion of "best" is very subjective and changes over time. For these fundamental reasons (and probably several others too), the end result of standardization efforts is often this: http://xkcd.com/927/ This is NOT to disparage any particular standards effort, nor is it to imply that standardization is pointless or unnecessary. Standardization *is* needed. But at the same time it is important to recognize the inherent challenges and limitations of standardization. So, returning to your question of what problem RDF solves that HL7 does not, RDF addresses this standardization problem at its root, by: - being completely independent of data format, data model and domain vocabulary. RDF allows the process of standardizing data models and vocabularies to be *decoupled* from the underlying semantic representation. Standardization can continue in parallel. - providing a common semantic foundation that can span across all healthcare information representation standards. RDF allows multiple standards to be mixed and matched. This will also make standardization much easier, because data models can be standardized in much smaller pieces. Divide and conquer! - providing a common abstract information representation that: (a) can losslessly represent any structured information; and (b) can act as an abstract lingua franca for defining and sharing model transformations, via ontologies and rules, for use when they are needed. And model transformations will *always* be needed to some extent . . . at least until we reach nirvana. ;) These model transformations used to be done inside proprietary black box integration engines. But RDF allows them to be represented explicitly and visibly, enabling them to be shared and standardized. In short, RDF does not enable semantic interoperability by itself, but it facilitates the process of achieving semantic interoperability, both by supporting and complementing standardization efforts and by facilitating model transformations. Sorry for such a long answer, but I hope this helps. It was a great question. David
Received on Wednesday, 14 May 2014 22:10:11 UTC