138
Using SNOMED CT with FHIR Prepared by Harold Solbrig Heather Grain USA www.ehe.edu.au 1

Using SNOMED CT with FHIR - Global eHealth … SNOMED CT with FHIR ... Describe some of the issues that may be encountered in the process of mapping ... Appendicitis | 41000160101

Embed Size (px)

Citation preview

Using SNOMED CT with FHIR

Prepared by

Harold Solbrig

Heather Grain

USA

www.ehe.edu.au 1

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

Welcome and introductions

2

What are you hoping to get out of this afternoon?

What we expect you already know

© Global eHealth Collaborative 2016

Preface

3

This is the first time this particular material has been presented, and it it is still a work in progress

FHIR is still a DSTU and is changing

The area of terminology binding is rapidly changing

Many of the topics discussed here are still on the “cutting/bleeding edge” — there are open issues and will likely have changed even since these slides were put together early this week.

We solicit audience input

Examples in these slides range in maturity from FHIR ‘0’ through ‘3’ — they are meant strictly as examples…

Most examples are drawn from DSTU2 except when otherwise noted

© Global eHealth Collaborative 2016

Acknowledgements

4

The authors wish to thank the International Health

Terminology Standards Development Organization (IHTSDO)

and Health Level Seven International for access to their

documentation and other materials.

Graham Grieve and Lloyd McKenzie for their contribution.

© Global eHealth Collaborative 2016

What we expect you to already know

5

General knowledge of FHIR

Ability to read XML and JSON

Basics of SNOMED CT

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

Intended outcomes for today

6

Identify roles SNOMED CT plays in FHIR

Know how to find and understand SNOMED CT references in the current FHIR

specification

Describe when SNOMED CT should be used with FHIR

Recognize situations when SNOMED CT can be used “out of the box”

Describe mechanisms that define SNOMED CT value sets for FHIR profiles and extensions

Describe process to define a map between existing data and SNOMED CT concept codes

Describe some of the issues that may be encountered in the process of mapping

Describe several approaches to determining whether required concepts exist in SNOMED

CT

Describe the process for requesting new SNOMED CT concepts for use in FHIR

Introduce semantic model binding in FHIR

© Global eHealth Collaborative 2016

Goal of this session 1. To understand the aspects of SNOMED CT that are referenced in the

FHIR spec

2. To understand when and how SNOMED CT concept codes and compositional grammar expressions can be used in FHIR:

Value Set definitions

Data Instances

Profile terminology mapping (meaning)

7

© Global eHealth Collaborative 2016

Presentation Outline

Roles SNOMED CT plays in FHIR

Coded Content in FHIR

Value Set Definitions in FHIR

Issues and things to consider

Mapping

The Semantic Gap

Model bindings and FHIR

Questions and Issues

8

© Global eHealth Collaborative 2016

Roles SNOMED CT plays in FHIR

9

© Global eHealth Collaborative 2016

Roles of SNOMED CT in FHIR Representation of terminology in FHIR

Representation of SNOMED CT in FHIR

Selecting codes from a SNOMED CT value set

Mapping to and from a SNOMED CT value set

SNOMED CT Compositional Grammar and FHIR

SNOMED CT Expression Constraint Language and FHIR

10

© Global eHealth Collaborative 2016

Roles that SNOMED CT plays in FHIR

The representation of concept codes in FHIR data records

The representation of value sets in FHIR resource definitions

Associating FHIR resource definitions with semantics using

terminology

11

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

When SNOMED CT should be used in FHIR?

12

Choice of code system depends on

Use Case

why are you collecting the data?

need for computable meaning over time?

need to aggregate according to national rules? (ICD…)

What are you representing

the terminology or the information model requirement

Type of information

clinical data (SNOMED CT / LOINC)

financial data

authority data in EHR

Is the concept available as a resource in FHIR?

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

Questions to ask …. HL7 Standards Developers…

Development International

Use of SNOMED CT content is freely available – you don’t need a

license (covered by MOU)

Publication in documents and repositories

Development National Affiliate

Requires license

No licence required to convert from HL7 published SNOMED CT

content to local codes to publish local implementation guides

13

© Global eHealth Collaborative 2016

Guides available

HTA Policies and Guidance Licensing V1.0

HTA Policies and Guidance currency of value set content V1.0

14

https://www.hl7.org/special/committees/termauth/docs.cfm

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

Questions to ask ….. Users Do you have a license?

Yes – use away….

No

Are you using SNOMED CT in a country which is a national member of

IHTSDO?

Yes

Get a license from your National Release Centre and use away

No

Get a license (from IHTSDO which may have a charge) and use away

15

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

Which Version of SNOMED CT? HL7 International Specifications

IHTSDO International Release most current version

If content you need is not there?

If it is in a national extension

Request permission to use the national extension as the basis

Inform HL7 international Terminology Authority (HTA) to request

this content in the international release.

If not – Ask HL7 International HTA to request content

16

© Global eHealth Collaborative 2016

SNOMED CT IP

17

© Global eHealth Collaborative 2016

HL7 Affiliate in a National Member Country of IHTSDO

Affiliate is required to have a license

License is obtained from your National Release Centre

With a license the affiliate may

Publish standards for use in that nation

All uses of the standard must have a license

Content requests

Go to the National Release Centre

18

© Global eHealth Collaborative 2016

HL7 Affiliate in a Non IHTSDO Member Country

If publishing content which includes SNOMED CT

Must be done under license obtained from IHTSDO

In which case operation is as per an IHTSDO Member Country

If not publishing – replace with local content

Method used to develop local content and the relationship to

international content is up the HL7 Affiliate

19

© Global eHealth Collaborative 2016

Jointly Developed Standards

Standards developed by groups of HL7 Affiliates (IHTSDO

Member and Non Members)

Awaiting advice from IHTSDO

20

© Global eHealth Collaborative 2016

Translations The HL7 Affiliate is responsible for quality control of translated

content

Translation

If the term is in your national release translation – use the translation

If the term required is not yet translated – make a request directly to your national release centre to discuss the process to be used.

If the language you require is not available contact HTA

21

© Global eHealth Collaborative 2016

Maps Published by SNOMED CT

May be used by license holders, but are not available to others

Development for local use

managed through collaboration between IHTSDO National

Release Centres and the HL7 Affiliate

Where there is no national release centre managed by the affiliate

– with great care!

22

© Global eHealth Collaborative 2016

Coded Content in FHIR

Using SNOMED CT with FHIR

23

© Global eHealth Collaborative 2016

Coded concepts in FHIR data records

FHIR Coded concept Data Types

Code - stand-alone code

Coding - system, version, code, display, etc.

CodeableConcept - collection of codings + plain text

Quantity - (special case - not further discussed)

http://www.hl7.org/FHIR/terminologies.html

24

© Global eHealth Collaborative 2016

FHIR code data type in a resource definition

25

© Global eHealth Collaborative 2016

FHIR Coding data type

28

© Global eHealth Collaborative 2016

FHIR Coding data type

http://www.hl7.org/FHIR/snomedct.html

29

© Global eHealth Collaborative 2016

SNOMED CT Version Identifier

30

http://www.hl7.org/FHIR/snomedct.html

© Global eHealth Collaborative 2016

SNOMED CT Version Identifier For Extensions

31 http://www.hl7.org/FHIR/snomedct.html

Distribution / Edition Identifier:

http://snomed.info/sct/900000000000207008 — SNOMED CT International Edition

http://snomed.info/sct/32506021000036107 — SNOMED CT Australian Distribution

http://snomed.info/sct/11000160102 — SNOMED CT CIMI Extension

© Global eHealth Collaborative 2016

SNOMED CT Version Identifier for Versions

32 http://www.hl7.org/FHIR/snomedct.html

Full versioned identifier:

http://snomed.info/sct/900000000000207008/version/20160131 — SNOMED CT International Edition,

January 2016 release

© Global eHealth Collaborative 2016

FHIR Coding data types in a resource definition

33 http://www.hl7.org/FHIR/imagingstudy.html

© Global eHealth Collaborative 2016 36

Something to think about…

© Global eHealth Collaborative 2016

FHIR CodeableConcept data type

37

© Global eHealth Collaborative 2016

FHIR CodeableConcept data type

Internal Codes

Any LOINC Code

Any code (no recommendation) 38

© Global eHealth Collaborative 2016 41

FHIR CodeableConcept data type example

http://www.hl7.org/FHIR/medicationexample2.json.html

© Global eHealth Collaborative 2016 42

FHIR CodeableConcept data type example

http://www.hl7.org/FHIR/medicationexample2.json.html

Questions:

1. Is this asserting that the two codes are

synonyms?

2. Can an application safely use whichever

code it is equipped to deal with?

© Global eHealth Collaborative 2016 43

FHIR CodeableConcept data type examples

http://www.hl7.org/FHIR/medicationexample2.json.html

Questions (continued):

3. What is going on here?

© Global eHealth Collaborative 2016

What of version?

44

© Global eHealth Collaborative 2016

What of version?

http://www.hl7.org/FHIR/datatypes.html#Coding.

45

© Global eHealth Collaborative 2016

US Edition

SNOMED

CT

Internation

al

74400008 | Appendicitis |

46

Versions / Editions

CIMI

SNOMED CT

International

74400008 | Appendicitis |

41000160101 | Good Concept |

1072931000119109 | Non-traumatic partial tear of right

rotator cuff (disorder) |

© Global eHealth Collaborative 2016

Concept Identifiers

47

SNOMED International Code

48

© Global eHealth Collaborative 2016 49

Compositional expressions as values

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

SNOMED CT Compositional Grammar

50

A formal language, specific to SNOMED CT, that is used to represent formal

logic description of the concept referent.

One way of representing the formal (description logic) aspect of a SNOMED

CT concept description

Can be interpreted as representing the set of entities “in the real (clinical)

world” that meet the defining criteria.

Does NOT represent a set of SNOMED CT concept identifiers (!) …

… in fact, there may be no SNOMED CT concept identifier defined whose

referents meet a given CG description

http://ihtsdo.org/fileadmin/user_upload/doc/download/doc_CompositionalGrammarSpecificationAndGuide_Current-en-US_INT_20150708.pdf?ok

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

Compositional Grammar and “coordination”

51

Pre-coordinated — a compositional grammar expression that has is associated

with a SNOMED CT Concept Identifier

Post-coordinated — a compositional grammar that is not necessarily

associated with a SNOMED CT Concept identifier

© Global eHealth Collaborative 2016 52

SNOMED CT Compositional Grammar Examples

The set of all right hip joints:

The set of all excision procedures having a direct site of the appendix structure

and a priority of emergency. Any emergency excision on the appendix.

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

SNOMED CT Compositional Expressions

53

Need to consider:

• Semantics — how do you constrain the set of expressions that are allowed

• Logic aspect — in DL, there is a huge difference between “the set of all things

that have the following properties” and “Specimen X has property Y”. It may

be blurred in SNOMED CT, but we still need to consider the ramifications

• Search, query and granularity — how do you search for “laterality left” when

it is contained (or inferred from!) a compositional expression

© Global eHealth Collaborative 2016 54

Navigation - Resource to Value Sets

© Global eHealth Collaborative 2016 55

Navigation - Resource to Value Sets (cont)

© Global eHealth Collaborative 2016 56

Navigation - Resource to Value Set (cont)

© Global eHealth Collaborative 2016 57

Navigation — Resource To Value Set (cont)

© Global eHealth Collaborative 2016 58

Navigation — Value Sets

© Global eHealth Collaborative 2016 59

Navigation - Code Systems

© Global eHealth Collaborative 2016 60

Navigation — Code Systems

© Global eHealth Collaborative 2016

Value Sets

How much do we “mean it”?

61

© Global eHealth Collaborative 2016

Value Set Strengths in the Core FHIR Spec

Value set references

(total / unique / includes SNOMED CT / core includes SNOMED CT)

Required: 798 / 281 / 2 / 1

Extensible: 199 / 89 / 13 / 2

Preferred: 140 / 101 / 43 / 8

Example: 587 / 271 / 96 / 87

62

© Global eHealth Collaborative 2016 63

Core SNOMED CT Required codes

© Global eHealth Collaborative 2016 64

One way to look up concept codes

© Global eHealth Collaborative 2016 65

Possible encounter reason codes

© Global eHealth Collaborative 2016 66

Possible encounter reason codes

© Global eHealth Collaborative 2016

Value Set Definitions in FHIR

67

© Global eHealth Collaborative 2016 68

Defining Value Sets in FHIR - Data Element Sheet

Type Link to entry on Bindings worksheet

This slide shows an example entry in a FHIR resource definition spreadsheet.

See: http://wiki.hl7.org/index.php?title=FHIR_Spreadsheet_Profile_Authoring) for more detail.

© Global eHealth Collaborative 2016 69

Defining Value Sets in FHIR

FHIR Bindings Worksheet

Name - globally unique Type - one of:

• code list - (enumeration) defined in spreadsheet

• value set - name of a FHIR value set resource instance in same

directory

• reference - reference to external standard

• Required

• Preferred

• Extensible

• Example

© Global eHealth Collaborative 2016 70

Defining Value Sets in FHIR

The CodeList tab

Defining Value Sets in FHIR

71

Type Link to entry on Bindings worksheet

© Global eHealth Collaborative 2016 72

Defining Value Sets in FHIR

FHIR Bindings Worksheet

External value set reference

© Global eHealth Collaborative 2016 73

Substance-code value set

© Global eHealth Collaborative 2016 74

Ways to define SNOMED CT Value Sets Lists of specific codes (similar to Coding)

By subsumption (all ‘isA’ descendants + the root concept)

By reference set — the identifier of a SNOMED CT Reference Set

As an Expression Constraint

Set operations on the above

… (there are several other alternatives in the ValueSet Resource which we

choose to ignore)

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

SNOMED CT Expression Constraint Language

75

A formal language that describes a set of SNOMED CT Concept Identifiers

• The interpretation of an Expression Constraint Expression depends on the

release / version / extension (“substrate”) to which it is applied

• One of many possible value set definition languages that can be used in the

FHIR environment.

http://ihtsdo.org/fileadmin/user_upload/doc/download/doc_ExpressionConstraintLanguageSpecificationAndGuide_Current-en-US_INT_20150820.pdf?ok

© Global eHealth Collaborative 2016 76

SNOMED CT Expression Constraint Language

Examples

77400008 | Appendicitis | OR 90176007 | Tonsillitis |

The concept identifiers for appendicitis or tonsillitis:

All concept identifiers that represent descendants of disorder of lung and

have an associated morphology of edema

All concept identifiers that represent descendants of clinical finding,

have a finding site of pulmonary valve structure or one of its descendants and

have an associated morphology of stenosis or one of its descendants

have an associated morphology of edema

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

SNOMED CT Expression Constraint Grammar and FHIR

77

SNOMED Expression Constraint Grammar can be used to define:

“Extensional” SNOMED CT value sets — (lists of concepts)

“Intensional” SNOMED CT value sets using a grammar that is very expressive

in the SNOMED CT context

The contents of a SNOMED CT Simple Reference Sets — when interpreted

against a given substrate

The contents of a SNOMED CT Query Specification Reference Set (or

equivalent), which can be interpreted against any substrate

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

Value Set Definition Issues in SNOMED CT

78

1. You can use the FHIR value set definition grammar directly

2. You can use the SNOMED CT Expression Constraint Grammar within the

FHIR value set definition

3. You can use the URI of a SNOMED CT Query Specification Refset, which

resolves to a Value Set

4. You can use the URI of a SNOMED CT Simple Refset, which is (arguably)

the result of a resolution

5. Some mixture of the above

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

SNOMED CT / FHIR Value Set decision points

79

SNOMED CT Expression Constraint Language and the FHIR Value Set Resource have considerable

overlap. The choice of which to use and when is not easy.

• Are your value set definitions specific to a given application or to be generally available?

• Are your value sets exclusively for use in the FHIR environment or will they have other

applications?

• Do you need the full expressivity of the SNOMED Model (e.g. role groups, site, etiology,

laterality, etc.?)

• Are you allowed to use all of SNOMED?

• Are you able to mint your SNOMED Value Set (Concept) Identifiers?

• Do you have a workflow that enables the submission of new concept identifiers and their

descriptions to a national or international body?

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

Additional things to consider with FHIR value sets

80

Should your profile really allow more than one “Coding” in a CodedEntry? If

so, what do multiple elements mean?

They should NOT represent compositional semantics

Should they include information that is explicit elsewhere in the document?

Should you leave the coding open-ended (e.g. any Finding or Procedure) or

should you lock it down to an agreed-upon set of codes?

Should you lock in a specific extension / release / version?

© Global eHealth Collaborative 2016

Maps and Mapping Issues

81

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

Known Issues

Increasing use of maps

Need for HL7 to publish (reference) maps to

Support conversion to improved value set content and

harmonisation across products (V2, FHIR, CDA)

Support conversion to nationally required representations

Risks of map use – dependent on use case

82

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

What is mapping? process of defining

a relationship between

concepts in one coding system to concepts in another coding system

in accordance with a documented rationale,

for a given purpose.

Reference: ISO TR 12300

83

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

Reasons to map

Data re-use

Statistical

Research

Planning and policy

Public health reporting

Interoperability (clinical use and data sharing)

Migration of ‘old’ system/data

Historical analytics

84

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

What mapping is NOT A low risk, low cost approach to interoperability

A one off exercise (except for migration)

A simple task

Mapping skills and skills in source and target are necessary

85

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

Map quality measures in development ISO work item

with collaboration of HL7 International and IHTSDO

Technical Specification – to establish and test conformance

Define quality requirements for map sets to Establish quality determinants of a map

Assess quality of a map for a purpose

Guide decision makers in map project requirements and processes

Establish pathways to improvement

Based on ISO TR12300 – Principles of mapping between terminological systems

86

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

Determinants include Terminology resource capacity

Equivalence achieved

Development Methodology

Team skills

Implementation functionality

87

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

Example – of measures

Type 1 – Level of quality

0 = best - high quality

4 = worst – low quality

Ability to sum measures to obtain overall score with low

being the highest quality.

Type 2 – outlier acceptability

Define acceptable quality limit for purpose

Define % outside acceptable limit which can be tolerated

88

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

Measures – Row Equivalence 0 = equivalent meaning

1 = source is totally included in target

2 = source is partly included in target

3 = source is mapped but there were many options and it is a best comparison rather than an actual correspondence

4 = no map possible

Method - Collect equivalence for every concept mapped

Measure 1 - Average equivalence for the Map Set

Measure 2 % X or over

89

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

Build Measure – Documentation and Process

0 1 2 3 4

Use Case Clear, specific

,documented

Clear, general,

documented

Clear,

documented

Documented Not Documented

Rationale of

decisions

Consistent,

Documented

Applied

Consistent

Documented

Documented

Applied

Documented Not Documented

Methodology Documented

Equivalence measured at

0-4 minimum

Process for review

Equivalence

measured

Documented

Process for review

Documented

Process for

review

Documented Not Documented

90

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

Build Measure – Validation

0 1 2 3 4

% of validation 100% 80% or more 60% or more 40% or more Less than 40%

Method of validation Blind double mapping

Manual review of

results

Validation for purpose

Double mapping

Manual review of

results

Validation for

purpose

Manual review

of results

Validation for

purpose

Manual review

of results

No validation

methodology

91

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

Build Measure – Tools

0 1 2 3 4

Level of

comparison

Lexical – offering

alternatives,

Semantic model applied

Lexical – exact

match

Semantic model

applied

Lexical – offering

alternatives

Lexical exact Not applied

Rules applied Data normalization

assessment

Data normalization

undertaken

Source/Target rules

integrated

Data normalization

undertaken

Source / Target

rules integrated

Data

normalization

undertaken

Basic rules

applied (i.e. age)

No rules or data

normalization.

92

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

Build Measure – Skills

0 1 2 3 4

Target Skills Examined recent

knowledge of Target and

experienced in application

in use case

Knowledgeable about

use and experienced in

application n use case

Experienced in

application in

use case

Practical but not

extensive

experience in use

case.

Basic or little

knowledge or

experience

Source Skills Examined recent

knowledge of source and

experienced in application

in use case

Knowledgeable about

use and experienced in

application n use case

Experienced in

application in

use case

Practical but not

extensive

experience in use

case.

Basic or little

knowledge or

experience

Mapping process

skills

Examined recent

knowledge of map

processes

Prepare documentation

on rationale for decision

Knowledge of map

processes

Prepare documentation

on rationale for decision

Examined

recent

knowledge of

map processes

Prepare

documentation on

rationale for

decision

No experience

or training

93

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

Governance Measure – Composition

0 1 2 3 4

Composition •Data Source

•Result Users

•Health Informatics Data

specialists

•Terminology Resource

•Map process

•Sponsor

5 of these 4 of these 3 of these 2 or less of these

94

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

Governance Measure – Process 0 1 2 3 4

Process Documentation of

decisions

Decisions applied across

the map.

Equity of decision making

to support use case

Responsibility for final

decision

Documentation of

decisions

Decisions applied

across the map.

Responsibility for

final decision

Documentation of

decisions

Responsibility for

final decision

Responsibility for

final decision

No documented

process

Maintenance Process established

Identified currency

requirements for update

Maintenance team

established and resourced

Maintenance follows

build process

Procedures

established

Identified currency

Maintenance team

established and

resourced

Procedures

established

Identified currency

Procedures

established

No process

established

95

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

Implementation Measure 0 1 2 3 4

Version Map, target and source

versions maintained in

the original data and

any messages

associated with the

data

Map, target and

source versions

maintained in the

original data

Target and source

versions maintained

in the original data

Details not maintained

in original data

Equivalence

maintained in the

map

Equivalence known

and communicated

Equivalence known Equivalence of mapped

data not identified

Communication Notification to user

that data was mapped

from original and the

map etc used.

Notification to the

user that the data

was mapped from

original

No notice of mapped

data

96

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

Rules Measure

0 1 2 3 4

Source Rules required for accurate

representation for use case – rules

applied in map

Or

No rules required

Rules required – some

rules applied

Rules required – not

applied

Target Rules required for accurate

representation for use case – rules

applied in map

Or

No rules required

Rules required – some

rules applied

Rules required – not

applied

97

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

Use of measures

For a given use case

Identify acceptable minimum required measures

Undertake quality measure review of the map

Compare results and map acceptability for purpose

Intended uses

Establish vendor conformance requirements

Establish ongoing improvement pathways

Provide quality and safety benchmarks to industry

98

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

Use Case - changes measure required for

conformance

Include examples such as

Direct Clinical Use E.g.: equivalence 0 with max 1% outliers)

Administration / Financial

Public Health

Research

Service Planning E.g.: Equivalence 1 with max 10% outliers

99

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

Mapping and FHIR

100

Mapping a resource to a local code system

Mapping from a local code system to a resource

May be SNOMED CT but could be other code systems

Publication method – being considered – not yet determined

Need to establish the metadata required and how this will be maintained

Possibly one off from old representation to new at the point of change

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

Measuring quality

101

Document the purpose of the map

Clearly indicate if clinical use is part of that purpose or not

If rules have been used associated with how the code system is used (not the meaning but the coding rules). Eg X is not coded in case Y cause we can’t claim it.

Consider maintenance requirement and frequency

Consider equivalence required for purpose

For all or for some of the content

Publication details – include all of this metadata as well as the actual map which should include equivalence clearly defined for each individual map in the set.

© Global eHealth Collaborative 2016

Solving the semantic gap

© Global eHealth Collaborative 2016

Semantic interoperability – the holy grail

Not just sharing data

Sharing and ‘understanding’ meaning

103

© Global eHealth Collaborative 2016

Once data is in the record – it’s there for life Ability use computers to find clinical information for

Analysis

Clinical Decision Support

Automation of reporting

But….

104

© Global eHealth Collaborative 2016

Technology solutions

Complex clinical terminologies

Terminology servers

Terminology governance in the organisation for implementation

Terminology specification for shared data

Information Model use of terminology

Post-coordination

105

© Global eHealth Collaborative 2016

This has led to…

106

© Global eHealth Collaborative 2016 107

What we need is….

The semantic line…..

© Global eHealth Collaborative 2016 108

Hypertension as a diagnosis

38341003 Hypertensive

disorder, systemic arterial

© Global eHealth Collaborative 2016 109

Hypertension as a risk factor

38341003 Hypertensive

disorder, systemic arterial

© Global eHealth Collaborative 2016 110

Hypertension – excluded – not a diagnosis

38341003 Hypertensive

disorder, systemic arterial

© Global eHealth Collaborative 2016 111

Hypertension in Family History

38341003 Hypertensive

disorder, systemic arterial

© Global eHealth Collaborative 2016 112

In the terminology

© Global eHealth Collaborative 2016 113

Finding the line…

Terminology

Information Model

Semantic Gap

Ref: terminology, clinical models and queries combined,

Karlsson, Heather Leslie, Sundvall

© Global eHealth Collaborative 2016 114

Where do the pieces fit….. 18336000 Open Fracture of

upper limb

© Global eHealth Collaborative 2016 115

Where do the pieces fit….. 439987009 Open Fracture of

Bone

371195002 Bone structure of

upper limb

© Global eHealth Collaborative 2016 116

Hypertension in Family History

303071001

Person in the

family

439401001

Diagnosis

© Global eHealth Collaborative 2016 117

Working Together

373572006 Clinical finding

absent

439401001

Diagnosis

38341003 Hypertensive

disorder, systemic arterial

© Global eHealth Collaborative 2016 118

Problems to resolve Equivalence testing (work in HL7 and IHTSDO)

Pre-post coordination

– where to use or not to use

Management of terminology and knowledge resources

© Global eHealth Collaborative 2016

Boundary Problem

• consistency is hard

• use case changes the practical solution

• Clinical need

• Recording

• Re-use

119

© Global eHealth Collaborative 2016

HL7 TermInfo,

Defined many of the use cases

Suggestion - qualifiers in the model

works sometimes

Solutions emerging in

HL7 CIMI, IHTSDO and OpenEHR

120

© Global eHealth Collaborative 2016

Experience and learning will..

• decrease the semantic gap

• identify patterns for common solutions

• improve problem identification and specification

121

© Global eHealth Collaborative 2016

Model Bindings in FHIR

Semantics for FHIR modeling artifacts

122

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

Semantics for FHIR Modeling Artifacts

123

Connecting the “meaning” of various artifacts with the defining components

that describe what they are about.

Semantic binding has always been on the FHIR roadmap, but it is just

beginning to solidify

Bridging the “Semantic Gap”

124

363787002 | Observable entity (observable entity) |

704327008 | Direct site (attribute) |

246501002 | Technique (attribute) |

424226004 | Using device (attribute) |

https://hl7-fhir.github.io/observation.html

© Global eHealth Collaborative 2016 125

246501002 | Technique (attribute) |

https://hl7-fhir.github.io/vitalsigns.html

75367002 | Blood pressure (observable entity) |

70665002 | Blood pressure cuff, device

37931006 | Auscultation (procedure) |

424226004 | Using device (attribute) |

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

Bindings

126

Provides semantic specialization

Allows model validation / alignment / discovery

It isn’t always obvious, for instance when something is a Condition vs. an

Observation in FHIR…

… also allows for alignment with other models

Validates value sets

Value set is a (sub) set of concept identifiers that classify under the target

126

© Global eHealth Collaborative 2016

Semantics for FHIR modeling artifacts Metadata

StructureDefinition.code

StructureDefinition.mapping

ElementDefinition.code

ElementDefinition.mapping

127

© Global eHealth Collaborative 2016 128

StructuredDefinition

© Global eHealth Collaborative 2016

StructuredDefinition

129

© Global eHealth Collaborative 2016

StructureDefinition code

Value Set

130

© Global eHealth Collaborative 2016

Possible StructureDefinition Bindings

131

“code”: {

“system”: “http://snomed.info/sct”,

“code”: “363787002”,

“display”: “Observable entity”

},

{

“identity”: “about”,

“uri”: “http://snomed.info/id/36378002”,

“name”: “Observable entity”

},

- or -

© Global eHealth Collaborative 2016 © Global eHealth Collaborative 2016

StructureDefinition and ElementDefinition code

132

Lloyd McKenzie (email - April 18): “StructureDefinition.code is

semantically identical to the

StructureDefinition.snapshot.element[1].code.” [or differential]

Map entries are different though. They represent relationships to

external artifacts for multiple elements as opposed to defining the

meaning of the overall model or a particular element within it. For

example, if StructureDefinition.code is set to blood pressure, your’e

saying “this profile *is* a blood pressure”, while, if you do a mapping,

you can say “this structure definition is somewhat similar to this

openEHR model of blood pressure and corresponds as follows.

© Global eHealth Collaborative 2016

Elements within a StructureDefinition

133

© Global eHealth Collaborative 2016 134

ElementDefinition.code

© Global eHealth Collaborative 2016 135

Observation Elements as ElementDefinition

“code”: {

“system”: “http://snomed.info/sct”,

“code”: “363787002”,

“display”: “Observable entity”

},

{

“identity”: “about”,

“uri”: “http://snomed.info/id/36378002”,

“name”: “Observable entity”

},

- or -

© Global eHealth Collaborative 2016 136

Observation.device Element

“code”: {

“system”: “http://snomed.info/sct”,

“code”: “424226004”,

“display”: “Using device (attribute)”

},

What of: 70665002 | Blood pressure cuff, device | ?

© Global eHealth Collaborative 2016

Questions…… Issues

137

© Global eHealth Collaborative 2016

Who is GeHCo?

Global eHealth Collaborative

Not for profit

Sharing and exposing expertise

Building solutions to share

www.gehco.org - more to come – be part of the solution!

138