On February 16th, the Dutch RIMBAA group will meet at Portavita's office in Amsterdam [1]. During this meeting, we will give a presentation about the following clinical update problem: When new information becomes available, how do you update the structured clinical data repository (CDR)?

The new information might come from a medical communication server or a user who just submitted a form in an EHR application. The CDR could be an HL7v3 RIM-based relational database. The question now becomes, “How can we update the CDR in such a way that…

1) the maximum amount of the new information is stored in the CDR,
2) no inconsistencies are introduced, and
3) no security policies are violated?”

Point 1 emphasizes that we must try to apply as much as possible of the new information to the CDR. Point 2 means that we want to make sure that the new information and the CDR as a whole are coherent after the update. This still sounds vague, and we must make it exact. Point 3 specifies that only authorized medical professionals can update patient information, and that the result of an update or merge is subject to the same security policies (access / authorization) as the original source documents.

“Why not simply use the SQL UPDATE command?” When updating clinical information in an abstract information model like the HL7v3 RIM, the pieces of information are almost always spread amongst sets of related rows in multiple tables. So, an update of information is not the update of a single row in a table, but rather a set of rows that belong together in such a way that if we take the individual pieces, it no longer forms a coherent whole. If the individual rows of the tables are the atoms, then our pieces of information are the molecules. To properly query and update the CDR, the question then becomes, “What are the molecules of information?”

To define the “no inconsistencies” clause of point 2, we must answer the following questions: What is overlapping information? Can two authors – medical doctors – claim conflicting information about the same patient? How do you deal with authorless information? How do you deal with partial overlaps in time? Is the overlap dependent on the observation code? Can there be vague overlaps?

This issue has been discussed in several past RIMBAA meetings. [2,3] Our talk will illustrate the problem with an example that supplies information about allergies, adverse reactions & intolerances [4,5,6]. We will also try to identify how questions match with topics from computer science and philosophical logic, such as data integration, knowledge representation and epistemic logic.

The other talks will cover real projects, with a lot of best-practice information and hands-on experience from experts. It promises to be an interesting day!

[1]
RIMBAA meeting February 16, 2011, Amsterdam

[2]
Kramer, E. (2010)
CIM Based CRUD Storage.
RIMBAA WGM held in Rome on September 15/16 2010.

[3]
Several related RIMBAA pages on the HL7 wiki, accessed February 4, 2011
Safe querying of a RIM-based data model
Object nets and object trees
SMIRF
Context SMIRFs for RIM based relational databases

[4]
Zel, M. van der, Goossen, W. (2010)
Bridging the gap between software developers and healthcare professionals.
In Hospital Information & Technology Europe Summer, Vol 3 No 2. Campden Publishing Limited.

[5]
HL7v3 Allergies & Intolerances Topic
HL7 Version 3 Standard: Care Provision; Allergy and Intolerance, Release 1
Last Ballot: DSTU Ballot 3 - September 2007

[6]
Implementation Guide for CDA Release 2, p29, 48,49.
IHE Patient Care Coordination (PPC)