Tutorial on Principles of Health Interoperability, presented at Informatics for Health Conference, Manchester 23 April 2017. Covers SNOMED CT, HL7 and FHIR and why interoperability is hard.
Call Girls Jp Nagar Just Call 7001305949 Top Class Call Girl Service Available
Interoperability, SNOMED, HL7 and FHIR
1. Principles of Interoperability: Tutorial
Tim Benson
tim.benson@r-outcomes.com
Informatics for Health Conference, Manchester
23 April 2017
2. Sources
• This tutorial is based
substantially on:
– Benson T, Grieve G.
“Principles of Health
Interoperability SNOMED CT
HL7 and FHIR” 3rd Edition,
Springer, 2016
5/2/20172
4. Approach
• Limited time, vast scope
• How and why these were developed
• Their complementary roles
• Why each is important
5/2/20174
5. PART 1 CORE CONCEPTS:
WHY INTEROPERABILITY IS HARD
5/2/20175
6. Interoperability Definition
• Interoperability is the ability of two or more systems or
components to exchange information and to use the
information that has been exchanged
– IEEE 1990
5/2/20176
8. 1. Technical Interoperability
• Move data from A to B
• Domain-independent
• Based on information theory
• Now commodity
5/2/20178
9. 2. Semantic Interoperability
• A and B understand data in the same way
• Domain specific
• What health interoperability standards focus on
• Needs codes and identifiers
5/2/20179
10. 3. Process Interoperability
• Business systems at A and B interoperate
• Business process-specific
• Requires re-engineering and staff training
• Generates all of the benefit
5/2/201710
11. 4. Clinical Interoperability
• The ability for 2 or more clinicians in different care teams to
transfer patients and provide seamless care to the patient.
14. Standards Definition
• A standard is a document, established by consensus and
approved by a recognised body, that provides, for common and
repeated use, rules, guidelines, or characteristics for activities
or their results, aimed at the optimum degree of order in a
given context
– ISO 2004
5/2/201714
15. Consensus
• General agreement, characterized by the absence of sustained
opposition to substantial issues by any important part of the
concerned interests and by a process that involves seeking to
take into account the views of all parties concerned and to
reconcile any conflicting arguments. Consensus need not imply
unanimity.
– ISO 2004
5/2/201715
16. Recognised body
• Internationally recognised standards development organization
(SDO)
• ISO
– Regional SDOs
• CEN
– National SDOs
• BSI
• ANSI
– Affiliates (e,g, HL7)
5/2/201716
17. Health Informatics SDOs
• ISO TC215
• CEN TC251
• HL7
• SNOMED International (fka IHTSDO)
• CDISC
• DICOM
• ASTM
• GS1
5/2/201717
18. Industry Consortia
• IHE
– Integrating the Health Enterprise
• Continua Alliance
• COCIR
• Open Health Tools
• OpenEHR
5/2/201718
19. User Benefits
• Right information
• Right place
• Right time
• Right person
• Safe, efficient and effective care
5/2/201719
22. Types of Communication
1. Within workgroup
2. Between departments
3. Between providers
4. Provider to payer
• Listed in increasing use of computers and decreasing order of
value to patients
5/2/201722
23. 1. Within workgroup
• Scheduling
– Appointments
• EHR
– Problems
– Medication
– History and physical
– Procedures
• Medical devices
– Vital signs
5/2/201723
24. 2. Between departments
• Orders/ Labs
– CPOE (Computer-based Physician Order Entry)
– Reports
• Medication
– Prescribe
– Dispense
– Administer
• Finance
5/2/201724
25. 3. Between Providers
• Referral
• Discharge summary
• PMR
– GP2GP (Exchange of complete medical records between primary care
providers using different systems)
5/2/201725
26. 4. Provider to payer
• Invoices
• Authorization
• Accountability
– Statistical returns
• Managed care
5/2/201726
28. Why interoperability is hard
• Translation
– A to HL7; HL7 to B
• Bit-perfect
– Computers are digital
• User language
– Each specialty has its own context-specific dialect
• Developer language
• Each vendor has its own computer-specific dialect
5/2/201728
33. Humpty Dumpty
• 'I don't know what you mean by "glory",' Alice said.
• Humpty Dumpty smiled contemptuously. 'Of course you
don't—till I tell you. I meant "there's a nice knock-down
argument for you!"'
• 'But "glory" doesn't mean "a nice knock-down argument",'
Alice objected.
• 'When I use a word,' Humpty Dumpty said, in rather a
scornful tone, 'it means just what I choose it to mean—
neither more nor less.'
• 'The question is,' said Alice, 'whether you can make words
mean so many different things.'
• 'The question is,' said Humpty Dumpty, 'which is to be
master—that's all.'
33
34. Desiderata for clinical
terminology
7. Reject NEC
8. Multiple granularities
9. Multiple consistent views
10. Context representation
11. Evolve gracefully
12. Recognize redundancy
Source J Cimino 1997
34
1. Content completeness
2. Concept orientation
3. Concept permanence
4. Meaningless identifiers
5. Polyhierarchy
6. Formal definitions
An excellent set of requirements, which was influential in the design of SNOMED CT
35. Exploding Bicycle Accidents
Source: Rector and Rogers (7 x 4 x 4 x 4 =448)
35
Structured Data Entry
File Edit Help
What you hit
Your Role
Activity
Location
Cycling Accident
36. LOINC Scope
• Clinical Laboratory Tests
– The test not the result
• Clinical observables
– “Eye color”, not “Blue eyes”
• Form headings
• Document types
• Assessment scales
37. Origins of LOINC
• Designed for use in interoperability
• Clem McDonald
– ASTM 1238 (1988)
– Standard Specification for Transferring Clinical Laboratory Data Between Independent
Computer Systems
– HL7 V2.0 OBX Segment
• EU EUCLIDES and OpenLabs projects
39. LOINC Codes
• LOINC Code
– Consecutive number + Check digit
– e.g. 12345-9
• Short convenient name
• Long common name
• Pre-coordinated to common concept model
• Six dimensions
– slightly different from EUCLIDES
40. LOINC Dimensions
• Component
– what is being measured
• Property
– kind of property measured
• Timing
– point in time or period
• System
– sample or body part
• Scale
– nominal, ordinal, interval or ratio
• Method
– procedure used to produce the result or other observation
41. Problems of LOINC
• Complex
• Many similar codes with only small differences
– mass concentration v molecular concentration
– anatomical position of pulse or BP
• Dimensions are not hierarchical so limits subsumption testing
42. Good aspects of LOINC
• Free
• Does what it claims to do (but no more)
• Fit for purpose
• Easy and quick to add new codes
44. Origins of SNOMED CT (1)
• SNOMED
– CAP committee 1955
– SNOP 1965
• topography, morphology, etiology, function (physiology)
– SNOMED 1975
– SNOMED III 1993
– SNOMED RT (Reference Terminology) 1997
45. Origins of SNOMED CT (2)
• Read Codes (UK)
– Coded terms in EHR e.g. “Blue Eyes”
– Developed 1983-1986
• by James Read and Tim Benson for use in GP EHR
– Purchased by DH in 1990
– Version 3 (CTV3) 1992-96
• Used by all GPs in UK (100%) and NZ
– Basis of payment
– Unsuccessful in any other specialty
46. Read Code (v1,2) Chapters
• Diagnoses
– ICD Chapters A-Z
• Medication
– BNF Chapters a-z
• History and physical
– Occupations (0), history and symptoms (1), physical examination findings (2)
• Procedures
– Diagnostic (3), lab (4), imaging (5), prevention (6), therapy (7), surgery (8), admin
(9)
47. SNOMED CT
• Merger of Read Codes and SNOMED RT
• Reference Terminology
– Defined by reference to other concepts
– All concepts are in hierarchies (19)
• Concepts
• Descriptions (terms)
• Relationships
– Defining, qualifying, sub-type, association
• Description Logic
50. Concepts
• A concept is a clinical idea
• Each concept has a unique identifier
– SCTID
• Each concept is in a hierarchy
– 19 hierarchies
• Concepts can have more than one parent
• Concepts are defined by reference to others
5/2/201750
53. Relationship
• Description logic
– “concept: attribute = value”
• Defining
– fully defined or primitive
• Qualifying
– SNOMED concept model
– Post-coordination
• Historical and additional
5/2/201753
54. Refsets (subsets)
• Language
– e.g. UK English
• Realm
– e.g. UK general practice
• Navigation
– e.g. pick list of common problems
5/2/201754
55. Problems of SNOMED CT
• Legacy baggage
• Lack of transparency
– Lack of free access outside member countries
– Initial lack of good web-based tools
• Complexity
– Post-coordination is not a practical proposition
– Slow rate of adoption
• Undefined boundaries
– Over-sold as answer to all terminology problems
56. Benefits of SNOMED CT
• Future-proof structure
• Can be improved and made fit for purpose
• Inherently multi-lingual
• Broad coverage
57. The Education Problem
• Few people are competent in both LOINC and SNOMED
• Clinical terminology is not taught well
– One of 186 topics in AMIA Core Curriculum for Clinical Informatics (JAMIA, 2009)
• Formal training is essential
– 93% of all those who rated themselves as competent had had more than 3 days
formal training in SNOMED CT
• (n=177, Report for DH, 2011)
58. Need for blended education
• Learn by doing (assignments and examples)
• Face-to-face presentations
• Web-based presentations and videos
• User guides and books
• Pick it up from colleagues
59. LOINC + SNOMED CT together
• Joint collaboration
• All clinical statements comprise:
– Narrative text
– Context (who, when, where, etc)
– Observable
– Finding
• (Rector A. What’s in a code, MEDINFO 2007)
60. Observable
• “Observables” are qualities of patients that are present in all
patients
– and whose values or states are determined by observation
• This is what LOINC is designed to code and does it well
61. Finding
• Information specific to a particular patient
• Finding = Observable + Value
• Value may be various data types
– Physical quantity, code, date, text etc
• SNOMED CT is good for Observation Value Codes
62. Conclusions
• LOINC and SNOMED are complementary
• You need to understand know both
• Both are complex!
63. PART 3 HL7 V2 AND V3
SYNTAX FOR INTEROPERABILITY
5/2/201763
74. Example lab report
• Header
– type, origin and date time of the message
• Single patient with
– ID number, name, sex, date of birth, address and GP identifier
• Specimen details
– laboratory accession number (ID), source, body site, time of collection and
requester
• Test results
– including the test name and result and abnormality flag.
5/2/201774
75. Example
• The abstract syntax:
– MSH Message header
– PID Patient Identification Details
– PV1 Patient Visit
– OBR Results header
– {OBX} Results detail (repeats)
• All segments are required.
5/2/201775
78. Example message rendering
Report from: Lab123457, 15:30 14-Aug-2008, Ref 123456789
Patient: MICKEY MOUSE, DoB: 14-Jan-1962, M
Address: 14 Disney Rd, Disneyland, MM1 9DL
Specimen: Swab, FOOT, Right, Requested By: C987654,
Location: 5N
Patients GP: Dr Smith (G123456)
Organism: STAU
Susceptibility: AMP R
SXT S
CIP S
5/2/201778
79. Why V3?
• “When you have seen one implementation of V2, you have
seen one implementation; every one is different”
• Need for a stringent, model-based approach
• Work on V3 began in 1992.
5/2/201779
80. V3 RIM
• Reference Information Model
• The “grammar” for V3
– nouns, verbs etc
– structural attributes
5/2/201780
89. V3 Participation
• Links Role to Act
• typeCode
– subject
– author
– performer
– location
– active ingredient
5/2/201789
90. V3 Entity
• Any living or non-living thing
• Living things
– person, animal, plant, organism
• Non-living
– manufactured item, chemical substance,
– physical place
– organisation
• Determiner code
– instance, quantified kind or kind
5/2/201790
92. V3 Entity-Role association
• Entity plays role
– person plays patient
– person plays practitioner
– chemical plays therapeutic agent
• Entity scopes role
– organisation scopes patient
– organisation scopes (makes) therapeutic agent
5/2/201792
93. V3 Role
• A competency of an Entity
• Associations
– Participation
– Entity scope and play
– RoleLink
5/2/201793
94. V3 Role
• Person
– Patient
– Practitioner
– Employee
• Place
– Hospital
– Home
– Clinic
• Item
– Therapeutic agent
– Instrument
– Computer system
• Organization
– Provider
– Payer
– Employer
– Supplier
• Responsibility
– Parent
– Employer
– Manufacturer
5/2/201794
95. V3 data types
• Basic data types
• Codes and identifiers
• Date/time
• Name and address
• Generic collections
5/2/201795
96. V3 Basic data types
• BL Boolean
• BIN Binary
• ST String
• ED Encapsulated data
• INT Integer
• REAL Real number
• PQ Physical quantity
• MO Money
5/2/201796
98. V3 Code data types
• CS simple code value (defined by HL7)
• CV code value, includes coding scheme identifier
• CE code with equivalents
– used with local and national codes
• CD concept descriptor allows qualifiers
5/2/201798
99. V3 CD example
• Compression fracture of neck of femur
<code code="71620000" codeSystem="2.16.840.1.113883.6.96"
displayName="fracture of femur">
<qualifier>
<name code="363698007" displayName="finding site"/>
<value code="29627003" displayName="structure of neck of femur"/>
</qualifier>
<qualifier>
<name code="116676008" displayName="associated morphology"/>
<value code="21947006" displayName="compression fracture"/>
</qualifier>
</code>
5/2/201799
100. V3 Date/Time
• TS Point in Time
• IVL <TS> Interval of Time
• PIVL Periodic interval of time
• EIVL Event-related periodic interval of time
• GTS General Timing Specification
• The time format based on ISO 8601
– (e.g. YYYYMMDDhhmmss).
5/2/2017100
101. V3 Name and address
• TN Trivial name (unstructured)
• ON Organisation name
• EN Entity name (any thing)
• PN Person name
– a sequence of name parts such as family name, given name(s), prefix and suffix, together with name use
(legal, maiden name, former name, alias)
• AD Postal Address
– a sequence of address parts, such as house, street, city, postal code, country and the address use (home,
temporary etc)
• TEL Telecommunication Address
– Universal Resource Locator (URL), which covers telephone numbers (voice, fax or mobile) e-mail addresses
and web pages.
5/2/2017101
102. V3 Generic Collections
• SET is an unordered collection of values without any repeats
• LIST is a sequence, containing values in a defined sequence (repeats not
allowed)
• BAG is an unordered collection of values, which are allowed to repeat
• IVL is a set of consecutive values of an ordered base type such as time
or physical quantity.
5/2/2017102
103. V3 Root attributes
• Optional attributes in all classes
– nullFlavour
– realmCode
– typeId
– templateId
5/2/2017103
104. Constrained Information Models (CIM)
• All uses of V3 are constraints on the V3 RIM
• Types of constrained model
– DMIM
– RMIM
– CMET
– HMD
– MT
– ITS
5/2/2017104
105. Constraint Types
• Omission
• Cloning
• Multiplicity
• Data type constraint
• Code binding
– CNE or CWE
5/2/2017105
106. Vocabulary and value sets
• Vocabulary is list of all possible codes
• Value set is set of codes permitted in a constrained model
• HL7 Vocabulary
– Hierarchical structure
• External Vocabularies
5/2/2017106
107. V3 Artifact naming
• Subsection
• Domain
• Type
• Numeric Id
• Realm
• Version
• Type
– Application Role
– DMIM
– HMD
– Interaction
– Message Type
– RMIM
– Storyboard
– Trigger event
5/2/2017107
120. CDA Releases and Levels
• Most widely used application of V3
– Single RMIM
• Release 1
– 2000
– Level 1 Document plus metadata
• Release 2
– 2005
– Level 2 Multiple documents plus metadata
– Level 3 Structured data
5/2/2017120
121. CDA Structure
• Header
– who, what, where, when and why
• Body
– Human readable
– XML Body
– Non-XML Body
• Entry
– Clinical statements
5/2/2017121
122. Constraining CDA
• CDA Templates
– Narrative
– Schematron assertion
– static model (RMIM)
• Multiple templates can apply to same part of a CDA document
• CDA Profile is set of templates that apply to a document type
5/2/2017122
123. Continuity of Care
• CCD
– CDA implementation of CCR
• CCR
– Developed by ASTM for patient summaries
• GP2GP
– Used in NHS to transfer life-long electronic medical records between
GPs using different systems
5/2/2017123
126. XDS Registry
• Contains metadata needed to find documents
– Patient ID
– Time of service
– Document class
– Type of event
– Facility type
• Based on OASIS ebXML Registry
– ISO 15000 parts 3 and 4
5/2/2017
127. XDS Repository
• Can be any number
• Stores document permanently
• Allows retrieval
• Registers documents
5/2/2017127
128. XDS Consumer
• Queries Registry
• Gets list of URLs
• Retrieves documents from repository
• Mainly for human readable documents
– CDA
– PDF
– Images
5/2/2017128
130. FHIR
• Fast Healthcare Interoperability Resources
Pronounced “Fire”
• New kid on the block
131. The “FHIR” Acronym
• F – Fast (to design & to implement)
“Fast” is relative – no technology can make integration as fast as we’d like
• H – Health
• I – Interoperable
• R – Resources (Building blocks)
132. What is FHIR?
• A set of modular components called “Resources”
• Resources refer to each other using URLs
– Build a web to support healthcare process
• Exchange resources between systems
– Using a RESTful API (e.g. web approach)
– As a bundle of resources (messages, documents)
133. Genesis of FHIR
• There has been a need to share healthcare information
electronically for a long time
– HL7 v2 is over 25 years old
• Increasing pressure to broaden scope of sharing
– Across organizations, disciplines, even borders
– Mobile & cloud-based applications
– Faster – integration in days or weeks, not months or years
134. Genesis of FHIR
• HL7 undertook a “Fresh look”
– What would healthcare exchange look like if we started from scratch
using modern approaches?
• Web search for success markers led to RESTful based APIs
• Drafted a healthcare exchange API based on this approach
135. FHIR Development Progress
• July 2011 – Conception
• Aug/Sept 2012 – First Draft Ballot
• Sept 2012 – First Connectathon
• Aug/Sept 2013 – First DSTU ballot
DSTU = Draft Standard For Trial Use
• 2014 – DSTU 1
• 2016 – DSTU 2
• 2017 – DSTU 3
136. FHIR Manifesto
• Focus on Implementers
• Target support for common scenarios
• Leverage cross-industry web technologies
• Require human readability as base level of interoperability
• Make content freely available
• Support multiple paradigms & architectures
• Demonstrate best practice governance
137. Implementer Focus
• Specification is written for one target audience: implementers
– Rationale, modeling approaches, etc. kept elsewhere
• Multiple reference implementations from day 1
• Publicly available test servers
• Starter APIs published with spec
– C#, Java, Pascal, ObjectiveC – more to come
• Connectathons to verify specification approaches
• Instances you can read and understand
• Lots of validated examples
using HL7.Fhir.Instance.Model;
using HL7.Fhir.Instance.Parsers;
using HL7.Fhir.Instance.Support;
XmlReader xr = XmlReader.Create(
new StreamRead
IFhirReader r = new XmlFhirReader
// JsonTextReader jr = new JsonTe
// new StreamRead
// IFhirReader r = new JsonFhirRe
ErrorList errors = new ErrorList(
LabReport rep = (LabReport)Resour
Assert.IsTrue(errors.Count() == 0
138. Support “Common” Scenarios
• Focus on scenarios implementers ask for
• Inclusion of content in core specification is based on core
content rule
– “We only include data elements if we are confident that most normal
implementations using that resource will make use of the element”
– Other content in extensions (more on this later)
– Easy to say, governance challenge to achieve
• Resources are simple and easy to understand and use
139. Web technologies
• Instances shared using XML & JSON
• Collections represented using ATOM
– Same technology that gives you your daily news summary
– Out-of-the-box publish/subscribe
• Web calls work the same way they do for Facebook & Twitter
• Rely on HTTPS, OAuth, etc. for security functions
140. REST
• “Representational state transfer” – an
architecture for how to connect systems
• Outcomes
– Simple stable interfaces
– High Performance / Scalability
– Visible Process (e.g. can debug)
– Portability
– Reliability (resistance to failure)
141. Human Readable
• Clinical Documents have both narrative and data
• The data / narrative dynamic exists throughout the process
• In FHIR, every resource has a
human-readable expression
– Can be direct rendering or human entered
143. FHIR & Cost of Integration
• These factors will drive down the cost of integration and interoperability
– Easier to Develop
– Easier to Troubleshoot
– Easier to Leverage in production
– More people to do the work (less expensive consultants)
• Competing approaches will have to match the cost, or disappear – effect
is already being felt
144. Future impact of FHIR
• Impact of FHIR on the market:
– Drive interoperability prices down
– Higher Expectations
– Increased spend on integration (N x 2!)
• Overall Market focus
– PHR on the web
– Healthcare repositories (MHD+)
– Device Data management
– Retooling existing connections (will happen slowly)
145. REST Operations
REST is based on CRUD(E):
• Create – create a new instance of data
• Read – get the content (state) of an instance
of data
• Update – change the content of an instance
of data
• Delete – remove the instance of data
• Execute – get the instance of data (?) to do
something for you
146. Resources
• “Resources” are:
– Small logically discrete units of exchange
– Defined behaviour and meaning
– Known identity / location
– Smallest unit of transaction
– “of interest” to healthcare
– V2: Sort of like Segments
– V3: Sort of like CMETs
147. What’s a Resource?
Examples
• Administrative
– Patient, Practitioner, Organization,
Location, Coverage, Invoice
• Clinical Concepts
– Allergy, Condition, Family History, Care
Plan
• Infrastructure
– Document, Message, Profile, Conformance
Non-examples
• Gender
– Too small
• Electronic Health Record
– Too big
• Blood Pressure
– Too specific
• Intervention
– Too broad
Goal: 100-150 total
151. FHIR Information
• FHIR:
http://hl7.org/fhir
• Development team wiki home:
http://wiki.hl7.org/index.php?title=FHIR
• Twitter:
https://twitter.com/search?q=%23FHIR
• Stack Overflow:
http://stackoverflow.com/questions/tagged/hl7-fhir