Archetypes

1
Design Principles for the EHR Thomas Beale
Ocean Informatics (Mooloolah, Australia) http//w
ww.gehr.org/openEHR http//www.openEHR.org http//
www.deepthought.com.au/it/archetypes.html
2
Design Principle 0 Standards based on open,
distributed systems
3
Distributed Health Information Environment
4
Structure of Standards

• Separate standards for each focus area (EHR,
  terminology, demographic, guidelines)
• Define the relationship between standards at the
  technical, functional and knowledge levels
• Each standard potentially consists of
• Reference Model
• Archetype Model
• API/Query Interface
• Normative/Informative archetypes
• Subdomain Terminologies

5
Relationships Between Standards
6
Design Principle 1Two-level Modelling
7
Challenges

• High rate of change of informational concepts
• Domain experts and users have no control over
  their systems
• Systems need to be able to communicate
  knowledge-level structures, not just data
• Terminology does not solve all problems
  pre-/post-coordination problem need for
  standardised models of volatile concepts

8
Classical Design Approach
9
Problems Inflexible(Biggest problem)

• Encodes only today's requirements addition of
  new concepts -gt rebuild / retest / re-deploy (V.
  expensive)

10
Problems Size of Model

• Size of model Model is large because all domain
  concepts are included -gt
• Unwieldy for modelling (too many concepts)
• Difficult to complete (probably won't happen)
• Difficult to standardise - too much for humans to
  agree on
• Difficult for interoperability - too much for
  machines to agree on
• May not be oriented to any particular use
• Examples early HL7v3 RIM most Health Department
  models / data-sets today

11
Problems - Management

• Two types of concepts in model
• Small number of generic concepts - the grammar
  of the domain" - understood (?) by IT people
• Large number of domain concepts - understood by
  domain experts
• ... two types of people, two types of process
  needed but only one formalism and one process
  available...
• ? Classic user/developer collaboration problem

12
2-Level Methodology

• Tomorrows standards and systems must be based on
  a future-proof methodology...

13
Software Meta-architecture
14
Software Meta-architecture

• Two models
• Reference model (RM) concrete model from which
  software can be built, and of which HIS data are
  instances.
• Archetype model (AM) a model whose instances are
  domain concepts Archetypes - which are directly
  processable by health information systems.
• Software
• Archetype Editor a GUI application for creating
  new domain concept definitions, based on the AM.
• Validator any component which creates or
  manipulates valid data using constraint models.
  This is based on the RM and AM classes.
• Browser a generic browser can be built, based
  solely on the RM, although a smarter browser can
  be built using the AM.

15
Consequences

• Knowledge-level Interoperability explicit Domain
  Concept Models
• Domain empowerment models created by Domain
  people/organisations
• Future-proof software (now software is a
  LEGO-builder) and information (now information
  is like LEGO). Models introduced post- system
  deployment
• Enables Decision Support DSS can make explicit
  assumptions about structure of information
• Intelligent Querying Information is
  transparent
• Solves Pre-/Post-coordination Terminology Problem

16
Language Analogy
17
Finding Archetypes

• Need to understand Knowledge structure of domain
  by using ontologies
• In each ontology there are
• Concepts coherent descriptions of entities in
  the domain, which are separately identified by
  domain users, and used in a self-contained way to
  communicate information
• Relationships compositional (various semantic),
  specialisation, time-related versions

18
Design Principle 2Multi-level Ontologies
19
Ontologies for Health(non-normative)
20
Design Principle 3Context Theory of Information
Acquisition
21
Information Contexts
22
Meaning of Contexts

• Values leaf values of observed or intended
  phenomena
• Semantic semantic context of values based on
  knowledge structure
• Temporal time of occurrence of structured values
• Information-generating fine-grained clinical
  context
• Organising arrangement of information into
  clinical investigation structure
• Clinical session business activity of seeing
  patient or doing test
• EHR system interaction act of committing info to
  EHR

23
Discussion