Transcript Information Decomposition at NCI

Information Decomposition at
NCI
Jean Duteau
HL7 UK RIMBAA Conference, November 4, 2010
gpi
gordon point informatics
www.gpinformatics.com
gpi
Jean Duteau






Has been involved with HL7 implementations since
2000, both v2.4/2.5 and v3
Worked with Canada Health Infoway as a Standards
Subject Matter Expert (SME) and a Jurisdictional SME.
Current Modeling facilitator for Pharmacy, Past
Publishing facilitator for Patient Care
Co-chair of Modeling & Methodology, Implementable
Technology Specifications, and Tooling workgroups
Currently involved in HL7 activities in Canada, United
States, and Europe
Currently working at NCI as the Information
Analyst/Architect – in charge of producing all of the
information models
Nov 4, 2010
© 2010, Gordon Point Informatics Ltd.
2
gpi
Outline
1.
2.
Introduction to the caCIS
project
Information Models
1.
2.
3.
4.
Constrain the DAM
Service Composition
Current challenges
Wrap up / Q&A
Nov 4, 2010
© 2010, Gordon Point Informatics Ltd.
3
gpi
WHAT IS THE CACIS PROJECT?
Nov 4, 2010
© 2010, Gordon Point Informatics Ltd.
4
gpi
caCIS
 Originated to produce a set of services to
extend traditional EMR into the realm of
Oncology (Cancer EHR)
 Focused on a few main categories of services:
Referral Management
 Outcomes Management
 Patient Trial Finder
 Chemotherapy Management
 Order/Request Brokering

Nov 4, 2010
© 2010, Gordon Point Informatics Ltd.
5
gpi
HOW WE PRODUCE INFORMATION
MODELS
Nov 4, 2010
© 2010, Gordon Point Informatics Ltd.
6
gpi
Domain Analysis Model (DAM)

Domain Analysis Model is produced

Uses as input:




Supplementary Information


Existing HL7 UV Domain models
NCI’s BRIDG (Biomedical Research Integrated Domain Group)
DAM
Other relevant specifications – CCD, Imaging models, etc.
Patients, Performers, Material
Main Information




Treatment Plan
Order, incl. Referral, Document Exchange, Laboratory,
Procedure
Outcomes
Allergies, Medications, Immunizations, etc.
Nov 4, 2010
© 2010, Gordon Point Informatics Ltd.
7
gpi
Information Models (RMIMs)
 Once the DAM concepts have been created
and the Architecture team determines the
service specifications (which tends to work
hand-in-hand), HL7 RMIMs are created to
enable the services
 We are creating payloads that are used as
parameters in the service operations
 These HL7 RMIMs have a direct relationship
(traceability) to the DAM and are used to
generate PSM artifacts such as Schemas
Nov 4, 2010
© 2010, Gordon Point Informatics Ltd.
8
gpi
Proper Abstraction Level
Initial RMIMs are simple constrained versions of UV material
or cloned versions of Canadian material
 They will contain a lot of their context:



Working with the architects, the breakdown of the full
payload into appropriate parameters is done:




Patients, Record Targets, Providers, etc.
Patient becomes a separate parameter
Various providers are either service context or separate
parameters
Actual information payload becomes just the required information
Finally, the context of the service operation, eg. Create
referral, is made explicit in the structural semantics of the
RMIM

Consequence: There are no models that will have multiple mood
codes.
Nov 4, 2010
© 2010, Gordon Point Informatics Ltd.
9
gpi
SERVICE OPERATION DECOMPOSITION
Nov 4, 2010
© 2010, Gordon Point Informatics Ltd.
10
gpi
Service Operation Decomposition
Naïve Service implementations in NCI have had
two services: do and get
 We are building an Interoperability Specification
that describes the orchestration between the
services
 This requires a much more fine-grained service
decomposition than do or get!
 Constraints we put on the information model also
require this decomposition:


Fixed mood codes and/or status codes means that we
need separate operations for create and update
Nov 4, 2010
© 2010, Gordon Point Informatics Ltd.
11
gpi
CURRENT CHALLENGES
Nov 4, 2010
© 2010, Gordon Point Informatics Ltd.
12
gpi
Query-By-Example
 Our architecture provides for a query-by-
example (QBE) model
 Given that the current state of queries in HL7
are all query-by-parameter (QBP), this is new
ground
 My experience has been in QBP so my QBEs
look an awful lot like a bunch of parameters.
 Still exploring what a QBE actually represents
Provides the query parameters
 Indicates the format that the query results
should be returned in

Nov 4, 2010
© 2010, Gordon Point Informatics Ltd.
13
gpi
Transport Wrapper and Control Act





Services started out simple: return doSomething(parameter,
parameter, …)
As we considered handling faults, we realized that every
‘doSomething’ would need to include managed faults as a
parameter and would need to return detected faults
To provide context for how to process downstream services,
we need to provide some ‘infrastructure’ information in each
operation
We are now struggling with the fact that our service
operations are starting to mimic a message interaction
Need to find the right balance between a context-free service
operation and a context-rich message interaction
Nov 4, 2010
© 2010, Gordon Point Informatics Ltd.
14
gpi
WRAP UP & Q&A
Nov 4, 2010
© 2010, Gordon Point Informatics Ltd.
15
gpi
Summary
 Summary points…
Nov 4, 2010
© 2010, Gordon Point Informatics Ltd.
16
gpi
Thanks!
gpi
March 3, 2010
Jean Duteau
Gordon Point Informatics Ltd.
[email protected]
© 2010, Gordon
Point Informatics Ltd.
17
gpi