Bootstrapping Adoption of a Universal
Exchange Language for Health
Information Exchange
Speakers: Tajh L. Taylor, Lowell Vizenor
OMG SOA in Healthcare Conference
July 15, 2011
Agenda
The Health Information Sharing Problem
PCAST Report
Ontologies in Health IT
The Bootstrapping Model
Conclusions
Who We Are
2
© 2011 Washington Consulting, Inc. All rights reserved.
Agenda
The Health Information Sharing Problem
PCAST Report
Ontologies in Health IT
The Bootstrapping Model
Conclusions
Who We Are
3
© 2011 Washington Consulting, Inc. All rights reserved.
The Health Information Sharing Problem
4
© 2011 Washington Consulting, Inc. All rights reserved.
“Meaningful Use”
Meaningful use Stage 1 objectives split into
“Core” and “Menu” groups
“Core” are required
“Menu” are pick-list optional with constraints
Core objectives focus on
Information gathering
Decision support in the context of patient care
Some limited emphasis in objectives on
information sharing
Example: “Generate and transmit permissible
prescriptions electronically (eRx).”
5
© 2011 Washington Consulting, Inc. All rights reserved.
What’s missing?
The Federal government has not specified or
mandated a particular terminology set or
vocabulary for consistent representation of
medical data
The overall strategy is to “let the industry
work it out” with light-touch guidance
6
© 2011 Washington Consulting, Inc. All rights reserved.
Some Popular Controlled Vocabularies
Vocabulary
Source
Notes
HL7 CDA
HL7
V2 may be UEL candidate
UMLS
NLM
Includes “UMLS Semantic Network”
ontology
SNOMED CT
IHTSDO
Includes ontological concepts
ICD-9-CM
NCHS (CDC)
ICD-10-CM required for HIPAA in
2013, billing focus
NCPDP
standards
NCPDP
Prescription processing
LOINC
Regenstrief
Institute
Laboratory testing
CPT Codes
AMA
Medical billing focus
7
© 2011 Washington Consulting, Inc. All rights reserved.
Interconnection Hierarchy
NHIN
HIEs
RHIOs
HMOs
Providers
8
© 2011 Washington Consulting, Inc. All rights reserved.
Agenda
The Health Information Sharing Problem
PCAST Report
Ontologies in Health IT
The Bootstrapping Model
Conclusions
Who We Are
9
© 2011 Washington Consulting, Inc. All rights reserved.
PCAST Report Findings
Early stage “meaningful use” has driven
adoption of EHRs, less emphasis on broader
information sharing, risks fostering more
“stovepipe systems”
Current standards for vocabulary and
messaging are not up to the task
Market incentives are misaligned with
economic benefits from information sharing
10
© 2011 Washington Consulting, Inc. All rights reserved.
PCAST Report Findings
Recommendations:
Evolutionary transition from traditional EHRs to tagged
data element model
More rapid transition for data exchange by means of a
universal exchange language
Tagged model to include attributes for provenance,
security, other metadata
Benefits:
Avoids universal patient identifiers, centralized
databases, enables EHR structural “traps”
11
© 2011 Washington Consulting, Inc. All rights reserved.
Universal Exchange Language
Extensible, XML based
language
Information exchange
based on tagged
message fragments
Aggregated message
fragments can form an
EHR
Extensibility is key to
flexibility beyond
static EHR structures
Transition from traditional
EHR to UEL view of EHR
12
© 2011 Washington Consulting, Inc. All rights reserved.
Data Element Access Services
Data element access services (centralized
agencies) for crawling, indexing, security,
identity, authentication, authorization, and
privacy
National infrastructure to be used for locating,
protecting and transporting data, not for
storing it
13
© 2011 Washington Consulting, Inc. All rights reserved.
PCAST Criticisms
Yet another standard in a sea of standards
Focused on driving implementation of
middleware
Lack of clinical representation on panel
Government mandate to override industry
progress (no matter how slow)
Note: HIT Standards Committee recommended
CDA R2 XML headers for metadata to ONC two
weeks ago
14
© 2011 Washington Consulting, Inc. All rights reserved.
Agenda
The Health Information Sharing Problem
PCAST Report
Ontologies in Health IT
The Bootstrapping Model
Conclusions
Who We Are
15
© 2011 Washington Consulting, Inc. All rights reserved.
Why do we need Ontologies?
Ontology (information science) definition:
A structured representation of the types of entities
and relations existing in a domain (e.g. clinical) that
is designed to support the exchange and reuse of
data.
Ontologies vs. Information Models
Information models (e.g. HL7 CDM) define the
structure in which information is carried
Ontologies (e.g. SNOMED CT) define the meaning of
the content carried by those structures
16
© 2011 Washington Consulting, Inc. All rights reserved.
Why do we need Ontologies?
Ontologies
Support interaction between EHRs and Clinical
Decision Support systems
Harmonize and deconflict local terminologies and
thereby provide more effective access to and reuse
of data
Are supported by open standards such as OWL,
RDF(S), SKOS, SPARQL , etc.
Improve adoption of translational medicine.
17
© 2011 Washington Consulting, Inc. All rights reserved.
Agenda
The Health Information Sharing Problem
PCAST Report
Ontologies in Health IT
The Bootstrapping Model
Conclusions
Who We Are
18
© 2011 Washington Consulting, Inc. All rights reserved.
Bootstrapping Philosophy
Combine a small, lightweight core of universally
understood metadata elements with community
owned, systems-based development efforts to
support incremental, community based
development of EHR components
Leverage existing standards
Look to similar success stories in other domains
(e.g. the National Information Exchange Model).
Ontologies can be built from component
messaging models
19
© 2011 Washington Consulting, Inc. All rights reserved.
Bootstrapping Philosophy
The pace of change is historically slow, so use
simple use cases to show quick benefits
Use ontologies and semantic technologies to
bridge between existing standards
Adopt the use of a small set of universally shared
and understood terms (i.e. a universal core) from
which more community or application specific
efforts can extend (see NIEM and UCore).
White House CTO Aneesh Chopra:
Eschew “top down” approach in favor of collaboration
Follow Open Government principles
20
© 2011 Washington Consulting, Inc. All rights reserved.
The Bootstrapping Model
1. Select use cases
2. Assess existing information models
3. Synthesize ontology fragments from existing
vocabularies for use cases
4. Create interaction models
5. Apply semantic technology
21
© 2011 Washington Consulting, Inc. All rights reserved.
Step 1: Select Use Cases
We will examine specific instances of the
following use cases:
Direct patient data collection
Medication and laboratory test management
Public health incident analysis
For each use case, we show the ontology
fragments and interaction models
22
© 2011 Washington Consulting, Inc. All rights reserved.
Use Case: Direct Patient Data Collection
Collect data from patients outside of the
clinical setting
Can include active and passive data collection
Active: patient or caretaker data entry at intervals or
upon specified events
Passive: monitoring devices automatically sending data
Involves patients more directly in their own
healthcare
Increases the range and fidelity of diagnostic
information
23
© 2011 Washington Consulting, Inc. All rights reserved.
Use Case: Direct Patient Data Collection
Step 2: Assess Existing Information Models
For the active data collection case: patients don’t
know medical vocabularies, but they do know
natural language and can work with graphic
depictions and iconography
Data collection approaches must balance structure
and expressiveness
Data collected this way have a different level of
fidelity and accuracy than data entered by a
medical professional
Probabilistic representations and inference are
likely needed to execute useful decision support
24
© 2011 Washington Consulting, Inc. All rights reserved.
Use Case: Direct Patient Data Collection
Step 3: Synthesize Ontology Fragments
Note: these examples use SNOMED CT concepts in lieu of an
defined UEL representation
Here, SNOMED CT lacks the means to include information
about the fidelity and accuracy of the pain score
25
© 2011 Washington Consulting, Inc. All rights reserved.
Use Case: Direct Patient Data Collection
Step 4: Create Interaction Models
:Pa tie nt
:Hom e C om pute r
:Practice EMR DB
:Physicia n
Enter Sym ptom
Se nd Sym ptom
Store Sym ptom
C he ck Sym ptom s
Ente r R ecom m e ndation
Send Re com m enda tion
C he ck R ecom m enda tion
Straightforward case
26
© 2011 Washington Consulting, Inc. All rights reserved.
Use Case: Direct Patient Data Collection
Step 4: Create Interaction Models
:Pa tie nt
:Hom e Com pute r
:Pe rsona l EHR Store
pdb1:Pra ctice EMR DB
phy1:Physicia n
pdb2:Pra ctice EMR DB
p hy2:Physicia n
Ente r Sym ptom
Store Sym pto m
Se nd Sym p tom
Store Sym ptom
Che ck Sym ptom
Ente r R e com m e nda tion
Se nd Re co m m e nd a tion
C he ck R e com m e nda tio n
R e trie ve R e com m e nda tion
R e que st 2nd O pinio n
R e que st O pinion
Se nd Sym ptom
C he ck Sym ptom
Ente r Re com m e nda tion
C he ck R e com m e nda tio n
R e trie ve R e com m e nda tion
PHR plus multiple providers case
27
© 2011 Washington Consulting, Inc. All rights reserved.
Use Case: Medication and Lab Test Management
Discover, track and manage medications and
lab test results across multiple clinical and
non-clinical settings
Better identify medication and test history
Detect and manage medication interactions
and conflicts
Avoid unnecessary and duplicate tests
28
© 2011 Washington Consulting, Inc. All rights reserved.
Use Case: Medication and Lab Test Management
Step 2: Assess Existing Information Models
Current, widely adopted vocabulary standards
such as LOINC are well suited in terms of
expressiveness
These standards lack the transport and
messaging protocols to support the
information sharing use case
Information ownership (read: control) issues
impede the execution of this use case, and
probably require impartial (federal?)
mediation
29
© 2011 Washington Consulting, Inc. All rights reserved.
Use Case: Medication and Lab Test Management
Step 3: Synthesize Ontology Fragments
The “fragment” approach endorsed by PCAST
assists with the information ownership issues
30
© 2011 Washington Consulting, Inc. All rights reserved.
Use Case: Medication and Lab Test Management
Step 4: Create Interaction Models
:ER Physicia n
:ER EMR DB
:Q ue ry Brok e r
pdb1:Pra ctice EMR DB
pdb2:Pra ctice EMR DB
:Pe rsona l EHR Store
R e trie ve m e ds a nd la bs
R e que st m e ds
R e que st scripts
R e que st O TC s
R e que st la bs
R e que st la b te st re sults
Federated search for medication history and lab
results
31
© 2011 Washington Consulting, Inc. All rights reserved.
Use Case: Public Health Incident Analysis
Public health incident analysis – Extraction
and aggregation in real-time
32
© 2011 Washington Consulting, Inc. All rights reserved.
Use Case: Public Health Incident Analysis
Step 2: Assess Existing Information Models
Existing standards are heavily focused around
the patient record
This example focuses on a small fragment of
information that will be aggregated and
analyzed
Extracting the relevant data for transmission is
key and should be lightweight
33
© 2011 Washington Consulting, Inc. All rights reserved.
Use Case: Public Health Incident Analysis
Step 3: Synthesize Ontology Fragments
Patient privacy is preserved by sending only
relevant de-identified information
34
© 2011 Washington Consulting, Inc. All rights reserved.
Use Case: Public Health Incident Analysis
Step 4: Create Interaction Models
p1:Pa tie nt
phy1:P hysicia n
pdb1:P ra ctice EMR DB
p2:Pa tient
phy2:Physicia n
pdb2:P ractice EMR DB
sidb:PH Incide nce DB
fidb:PH Incidence DB
:R e gulator
R eport Sym ptom
Ente r Sym ptom
R e port Incide nt
Re port Sym ptom
Ente r Sym ptom
R e port Incide nt
Re port Aggre ga te Incidents
Ana lyze Incide nts
Incident reporting and aggregation
35
© 2011 Washington Consulting, Inc. All rights reserved.
Step 5: Apply Semantic Technology
Use existing semantic technology tools to
implement the transformation from the stored
representations at either end of the
transaction
36
© 2011 Washington Consulting, Inc. All rights reserved.
Agenda
The Health Information Sharing Problem
PCAST Report
Ontologies in Health IT
The Bootstrapping Model
Conclusions
Who We Are
37
© 2011 Washington Consulting, Inc. All rights reserved.
Conclusions
There are interesting use cases that can drive
the “messaging” orientation recommendation
of the PCAST report
There may not be a need for a newly created
universal exchange language to implement
these examples
Ontological representation and translation can
serve as the “glue” between systems
Success can be had by starting small and going
large
38
© 2011 Washington Consulting, Inc. All rights reserved.
Agenda
The Health Information Sharing Problem
PCAST Report
Ontologies in Health IT
The Bootstrapping Model
Conclusions
Who We Are
39
© 2011 Washington Consulting, Inc. All rights reserved.
Who We Are: The Authors
Tajh L. Taylor
Lowell Vizenor
Senior Manager
Washington Consulting, Inc.,
a division of Alion Science
and Technology
tltaylor@washingtonconsulti
ng.com
Ontology and Semantic
Technology Practice Lead
Alion Science and Technology
lvizenor@alionscience.com
40
© 2011 Washington Consulting, Inc. All rights reserved.
Who We Are: Washington Consulting, Inc. and Alion
Alion Science and Technology
Corporation is an employeeowned technology solutions
company delivering technical
expertise and operational
support to the Department of
Defense, civilian government
agencies and commercial
customers.
Washington Consulting, Inc.,
an Alion Science and
Technology Company, is a
management and information
technology consulting
organization. Headquartered
in the Washington
Metropolitan region, we serve
premier clients in the
commercial, public and notfor-profit sectors. We
leverage our experience and
expertise across commercial,
nonprofit and public sectors
to deploy the best resources
and solutions for our clients.
41
© 2011 Washington Consulting, Inc. All rights reserved.
References
Aneesh Chopra blog. “Sending Health Data Safely And Securely Over the Internet.” February 3,
2011. http://www.whitehouse.gov/blog/2011/02/03/sending-health-data-safely-andsecurely-over-internet
SNOMED Clinical Terms User Guide. July 2008 International Release.
http://www.ihtsdo.org/fileadmin/user_upload/Docs_01/SNOMED_CT_Publications/SNOMED_
CT_User_Guide_20080731.pdf
President’s Council of Advisors on Science and Technology Report: “Realizing the Full
Potential of Health Information Technology to Improve Healthcare for Americans: The Path
Forward.” December 8, 2010.
http://www.whitehouse.gov/sites/default/files/microsites/ostp/pcast-health-it-report.pdf
42
© 2011 Washington Consulting, Inc. All rights reserved.