Semantic Mapping and HL7

1
Semantic Mapping and HL7

• robert.worden_at_charteris.com

2
Why Mapping?

• HL7 have built UML semantic models of healthcare
  (RIM, RMIMs, CDA)
• If everybody migrated to these models, many
  interoperability problems would be solved
• Suppliers cannot do this they have big
  commitments to other models and structures
• They could map to the HL7 models mapping is much
  easier than migrating
• So HL7 needs to publish shareable mappings to
  existing standards and data structures

3
Why Semantic Mapping? To Handle Structure Clashes
Message B
Message A
mappings

• Most mapping tools only map data values
• These are the leaves of message trees
• Structure clashes (e.g. multiplicity) arise on
  non-leaf nodes, deep in the trees
• Most mapping tools dont address them
• Therefore the translations dont work
• So people drop the tools and hand-code

4
Structure Clashes Arise from Associations
object
Class X
association
association
Class Y
property
Message A
Message B
UML Class Model

• Deep message A represents the association XgtY
  by nesting of elements
• Shallow message B represents the same
  association by shared key values
• This causes a structure clash
• You cannot translate from A to B without knowing
  how they each represent the same association
• Therefore semantic mapping is needed, to drive
  accurate translations

5
Mappings in Progress

• Pharmacy V2-V3 (Pharmacy TC)
• Various V2-V3 (ItalTBS)
• HIPAA X12 to V3 (FM TC)
• CCR CCD (various vendors)
• Relational database V3 (basic demonstrator)

6
Example CCR and CCD

• CCD and CCR are complex structures (typical
  nesting depth 6 12)
• They have many structure clashes (CCR uses shared
  keys)
• They have both been mapped to a CCD semantic
  model (a few days work for CCR)
• The mappings give accurate translations and round
  trips translating thousands of data items
• Mapping exposes the problems in CCR

7
Mappings for ARRA/HITECH
V3 RMIM
NCPDP
CDA Templates
V 2.5.1
CCR
Application Data Model
Application Database
HIPAA X12
Application Database
Application Database
8
Benefits and Issues

• Mappings reveal the problems clearly and early
• Any-to-any translations can be generated from the
  mappings (gtrapid testing)
• By mapping an application data model to CDA, you
  will get translations to and from CDA, V2.5.1,
  CCR .
• Managing templates is a big problem
• Versions are a big problem

9
Semantic Mappings - Summary

• Declarative, simple where they can be simple,
  viewable (e.g. Excel)
• A specification for semantic interoperability
• Auto-generate runnable transformations
• N(N-1) transformations from N sets of mappings
• Rapidly testable (gt test-driven mapping)
• Mapping exposes semantic problems (gaps,
  ambiguities)