HL7 TermInfo A standard for using SNOMED CT with HL7

Presentation copyright - The Clinical Information Consultancy (www.clininfo.co.uk)SNOMED Clinical Terms is a copyright work of the College of American Pathologists (www.snomed.org)

HL7 TermInfoA standard for using SNOMED CT with HL7

David MarkwellPrincipal Consultant - The Clinical Information Consultancy Ltd

Chair of SNOMED International Concept Model Working Group

Co-Editor – HL7 TermInfo Guide to use of SNOMED CT

2

Overview

What is TermInfo? Where did TermInfo come from? Why does TermInfo matter? What does TermInfo say? Where is TermInfo going?

3

What is TermInfo?

TermInfo is an HL7 Project developing standard guidance on use of Terminologies within Information models

The Information model focus is on HL7 Version 3 models including:

• HL7 Reference Information Model (RIM); and• Constrained Information Models (e.g. DIMs, RMIMs etc.)

• Particularly the Clinical Statement pattern

TermInfo Work Package 2 is the most active part of the project The Terminology focus of WP2 is SNOMED CT

4

A project launched by NASA in July 2004 To enable effective coding and structuring of

astronaut health records To decide how to use SNOMED CT in

information models derived from the HL7 Reference Information Model

Where did TermInfo come from?

Clinical semantics, the final frontier.

These are the voyages of the HL7 project TermInfo.

Its five-year mission: To explore strange old words.

To impart meaning to structured information.

To boldly code where no code

has gone before.

6

HL7 TermInfo coming down to earth

TermInfo was adopted by the HL7 Vocabulary TC in September 2004

Supported by the SNOMED International Standards Board since October 2004

7

Why does TermInfo matter?

The HL7 RIM is an information model It provides a structure that can be populated with

instances of information represented using a coded vocabulary

SNOMED CT is a reference terminology Terminology provides a coded vocabulary to

populate an information model with processable concepts

The objective of both HL7 and SNOMED is semantic interoperability Meaning depends on how the vocabulary and

structure fit together

8

Roger Tranquility,

we copy you

on the ground.

You got a bunch of guys

about to turn blue here.

We're breathing again.

Thanks a lot.

Meaning depends on the context of words within a sentence

9

Meaning depends on structure & vocabularyNotes on an English lesson

Jack where Jill had had had had had had had had had had been the correct answer Ten consecutive uses of the word “had” seems like nonsense Surprisingly punctuation can make it meaningful Question: Who was correct, Jack or Jill?

Jack, where Jill had had “had”, had had “had had”; “had had” had been the correct answer. Jack was correct!

Jack, where Jill had had “had had”, had had “had”; “had had” had been the correct answer. Jill was correct!

Meaning depends on structure

10

Meaning depends on structure & vocabularyMore notes on an English lesson

Even with the correct structure this is far from clear Jack, where Jill had had “had had”, had had “had”;

“had had” had been the correct answer. More specific vocabulary makes the meaning clearer

Jack where Jill had written “had had”, had written “had”, “had had” had been the correct answer.

Changing the structure makes the meaning unambiguous Where Jack had written “had”, Jill had written “had

had”. The correct answer had been “had had”. Meaning depends on structure and vocabulary

11

Structure and vocabularyA clinical example

Consider the meaning of the following similar looking sentences The result suggests exclusion of thyrotoxicosis The result suggests that thyrotoxicosis can be excluded The result suggests that thyrotoxicosis should be excluded The result suggests that thyrotoxicosis needs to be excluded

Can you be sure what each sentence is intended to mean?

12

Structure and vocabularyAn unconstrained HL7v3 / SNOMED CT example

Asthma Subject Family member

Family History of Asthma Subject Record target

ComponentFamily History Asthma

ComponentFamily History Family History of Asthma

ComponentFamily History Asthma value=true

ComponentHistory

Asthma

Subject Family member

Family History value=Asthma

Family History of Asthma

Family History of Asthma Subject Family member

13

Structure and vocabularyAn unconstrained HL7v3 / SNOMED CT example

Asthma Subject Family member

ComponentFamily History Asthma

ComponentFamily History Family History of Asthma

ComponentFamily History Asthma value=true

ComponentFamily History No Family History of Asthma

ComponentFamily History Asthma value=false

ComponentFamily History Asthma negationInd=true

Component negationInd=trueFamily History Asthma

No Family History of Asthma

Family History of Asthma

Family History of Asthma Subject Family member

Presentation copyright - The Clinical Information Consultancy (www.clininfo.co.uk)SNOMED Clinical Terms is a copyright work of the College of American Pathologists (www.snomed.org)

What does TermInfo say?

Some examples of the recommendations from the initial ballot draft

… plus a few updates …

15

Family history recommendation

Use SNOMED CT expression to explicitly represent a condition of a family member rather than relying on an HL7 “participation” association to convey this

Use the HL7 participation association if necessary to specify a particular person

Family History of Asthma SubjectFamily member

Family History of Asthma

16

General recommendation

Semantic interoperability requires a shared “Model of Meaning”

Information is often collected and stored based on a “Model of Use” This view needs to be transformed into a

normalized “Model of Meaning” for comparison A simplistic view of the “Model of Use” may prevent

reliable transformation to a common “Model of Meaning”

17

Recommendations on semantic overlaps (1)

Avoid use of optional HL7 attributes that restate meaning that can be clearly represented in a SNOMED CT expression

This applies to Act

• negationInd• certaintyCode

Observation and Procedure• methodCode• bodySiteCode

18

Recommendations on semantic overlaps (2)

Ensure consistency of mandatory HL7 attributes with any explicit context stated in a SNOMED CT expression This applies to

• HL7 moodCode and the ways it interacts with …• SNOMED CT “finding context” and “procedure context”

A table specifies which values of these are regarded as consistent with one another

19

Recommendations on semantic overlaps (3)

Apply any context implied by mandatory HL7 attributes to SNOMED CT expressions that do not include an explicit conflicting context For example, the HL7 moodCode “RQO” applied to

a procedure concept (without explicit context) replaces the default context “done” with “requested”

A table specifies the default finding and procedure contexts for each moodCode value

20

Recommendations on semantic overlaps (4)

Do not treat HL7 attributes and SNOMED CT concepts in a manner that creates new combinatorial contexts E.g. moodCode=“INT” (intent) with SNOMED CT

procedure context “requested” Does not mean• “intent to request …[procedure]”

• The “intent” applies to the procedure not the request.

It means• “requested …[procedure]”

• The “requested” merely refines the “intent”.

21

Recommendations of use HL7 Observation class to represent “nominalized statements” (1)

Nominalized statements are assertions about a clinical finding that cannot be consistently subdivided into a “code” for what was done; and a “value” for the result

Acceptable pattern Observation.code – an HL7 code meaning “Assertion” Observation.value – an expression that when normalized is

subsumed by one of the following• situation with explicit context (ConceptId 243796009)

• clinical finding (ConceptId 404684003)

Note: Change from current draft based on recent MnM ruling

22

Recommendations of use HL7 Observation class to represent “nominalized statements” (2)

Other representations of nominalized statements are deprecated but may be permitted for backward compatibility. These forms are only permitted where there is a reproducible

transformation that can be reliably applied to the convert them to the preferred acceptable pattern without loss of meaning

Example of a permitted deprecated pattern Observation.code – an expression that when normalized is

subsumed by one of the following• situation with explicit context (ConceptId 243796009) • clinical finding (ConceptId 404684003)

Observation.value (OMITTED)

Note: For backward compatibility this pattern regarded as equivalent to the acceptable pattern on the previous slide

24

Where is TermInfo going?

Ballot on Draft Standard for Trial Use (DSTU) in April 2006 Large number of comments reviewed and decisions

made on significant changes

Revision of the document now in progress Next ballot cycle - end 2006 / early 2007

25

Join the HL7 TermInfo Project at www.hl7.org

26

SNOMED CT – the future

Representatives of five Potential Charter Members (Australia, Denmark, Lithuania, New Zealand, and the United Kingdom) met with the College of American Pathologists (CAP) in Copenhagen to discuss the creation of a standards development organization to maintain and promote the SNOMED CT clinical terminology.

By the end of the two-day meeting, which took place from October 4th and 5th, all the Potential Charter members (including the Canada and the United States) and the CAP agreed in principle to move forward together toward their common goal of founding the SNOMED Standards Development Organization (SNOMED SDO).

www.snomed.org

With the emergence of the SNOMED SDO, TermInfo has become part of an international effort to assemble the pieces needed for clinical semantic interoperability

Observation

Procedure

Person

Role

Com

pone

nt

SupplyActRef

OrganizationDevice

Aut

hor

Subject

Is NASA interested in

TermInfo because of

the model in space ?