Designing queries and searches in SNOMED CT - PowerPoint Presentation

Slide1

Designing queries and searches in SNOMED CT

Principles plus hints and tips

presented by

Denise Downs

Slide2

Housekeeping

Please keep yourself on Mute during the presentationPlease use chat if you have any questions –change to Everyone so that we all see the question

There are various pauses within the presentation to ask questions BUT If you have a ‘burning question’ please feel free to take yourself off mute and ask

Quick Feedback form

Email with links after the webinar

2

Slide3

Queries/searches - definition

What do we mean by queries/searches

Identifying patients or content in records

This could be used for:

ReportingSending letters

Risk Analysis

Business protocols

QOFetc.We will refer to these as searches from this point forward

3

Slide4

What this is and isn’t

Assumes using the supplier reporting functionalityWe will be producing documentation and a recorded presentation for those with their own databases who write SQL type queries – a taste at the end

4

Slide5

Objectives

Understand how to undertake searches in SNOMED CT, what will a QOF rule look like

Understand the principles of converting existing searches in Read v2/CTV3 to SNOMED CT

Understand why converted

searches may not produce quite the same results as the

original

A brief summary of the technical considerations

5

Slide6

Read v2 / CTV3

Examples are in Read v2, this is where the changes are probably greatestThere are some differences between SNOMED CT and CTV3, but mainly to do with inactivation and modelling

6

Slide7

Safety net - Dual Coding and the Subset

Suppliers have been provided with a GP Subset, these are SNOMED CT codes for which a Read equivalent exists

All historical data will have a Read code AND SNOMED CT

New SNOMED data within the subset will also have a Read code entered by the system

This Dual coding approach will provide time for searches to be updated

7

Read v2

CTV3

SNOMED CT

Slide8

Recap on the fundamentals of

SNOMED CT

Slide9

A Concept in SNOMED CT

Irritable bowel syndrome

(disorder)

10743008

Relationships

Is a

Disorder of intestine

Finding site

Intestinal structure

Preferred Term

:

Irritable bowel syndrome

Synonym

: Irritable bowel

Synonym

: Irritable colon

Synonym

: Irritable colon syndrome

Synonym

: Membranous colitis

Synonym:

Spastic colitis

Synonym:

Mucous colitis

Synonym:

IBS - Irritable bowel

syndrome

Synonym:

IC - Irritable

colon

FSN: Fully specified name

Slide10

Concepts and descriptions

Synonyms are truly synonymous

Searches are done on the concept id

FSN is often the term provided in searches

10

Fully specified name

Preferred term

Acceptable synonyms

Preferred term is usually the description seen in the system

Slide11

NHS Digital browser

11

Select options

and then click to display Description Ids

Browser available at

https

://

termbrowser.nhs.uk

Slide12

Structure of SNOMED CT

12

19 hierarchies to organise the content of SNOMED CT

Semantic tags change within some hierarchies:

Top level hierarchy

Sub hierarchy tag

Sub hierarchy 2 tag

body structure

(body structure)

(cell)

body structure

(morphologic abnormality)

 

clinical finding

(disorder)

 

clinical finding

(finding)

 

Slide13

Navigation within a hierarchy is done via the |is a| relationship

Infective pneumonia

|is

a| child of

infection

Infective pneumonia

|is a| child of

respiratory disease

Viral pneumonia

|is a| descendant of

infection

Relationships - recap

13

Searching on attribute relationships will not initially be available in systems

Slide14

14

Bacterial

pneumonia

Infective pneumonia

Bacterial lower respiratory infection

Pneumonia

Infectious disease of lung

Disease

Lower respiratory tract infection

Lung consolidation

Disorder of lung

Disorder of thorax

Disorder of trunk

Disorder by body site

Respiratory tract infection

Infection by site

Infectious disease

Bacterial respiratory infection

Bacterial infection by site

Bacterial infectious disease

|Is a | relationships

Slide15

Consequences of multiple parents

15

These are the same concept – found in more than one place

Slide16

Writing a new Search

SNOMED CT in primary care

Slide17

Differences?

A1

A1234

SNOMED CT

Slide18

All patients who have had …

How might I write a search to find those patients who have had an appendectomy…

Slide19

Search for clinical codes:

Type what you want to say …

Slide20

Appendectomy and all its descendants…

Slide21

That’s it ….

21

The rest of this presentation is about dealing with anomalies with Read

Slide22

Comparison with Read v2 and CTV3

SNOMED CT in primary care

Slide23

Some term text differences from Read

SNOMED CT uses the Body Structure hierarchy to specify body parts

Terms beginning [SO] do not exist

7N890

00 [SO]Cervical lymph node maps to 81105003 |Cervical lymph node structure

(body structure) |

SNOMED CT uses the Morphologic abnormality sub hierarchy to specify

morphologyTerms beginning [M] do not existBBEJ

. 00 [M]Intradermal

naevus

maps to 112681002

| Intradermal

naevus

(morphologic abnormality) |

Slide24

Some term text differences from Read

Clinic A / Clinic B codes are inactive

These will map but to inactive codes; they are not interoperable, never went through

GP2GP

Bill/fee paid don’t exist in SNOMED CT

It is ambiguous, do have Bill paid and Fee paid

See

Data Quality Guidance for more information, update due in July24

Slide25

NEC / NOS

SNOMED CT does not

allow concepts such as NOS, NEC

‘other’

NOS, NEC, other F52z. 00 Otitis media NOS maps to 65363002 | Otitis media |

Q4z.. 15 Stillbirth NEC maps to 237364002 | Stillbirth |

25

Asthma unspecified maps to Asthma – as it maps to its true meaning

Slide26

SNOMED CT default context

Clinical Findings and procedures have a default meaning:

For

a clinical finding this means that:

the finding has actually occurred (vs. being absent or not found)it is occurring to the subject of record (the patient)it is occurring currently or at a stated past time.

 

For

a Procedure this means that:the procedure was completedit was performed on the subject of record (the patient)it was done in the present time or at a stated past time.So procedure declined is not in the procedure hierarchy but in situation with explicit context (tag is situation)

26

Slide27

Some model differences

SNOMED CT model requires that all descendants of a concept have same

characteristics

Opposites are not listed under the same concept

Constipated and not constipated

are listed

under

the same parent in Read v2 but in SNOMED CT they have different parent concepts27

Slide28

Other differences

SNOMED CT does not contain duplicates

F004. 00 Meningitis – tuberculous maps to 58437007 | tuberculosis of meninges|

A130. 00 Tuberculous meningitis maps to 58437007 | tuberculosis of meninges|

SNOMED CT

concept IDs don’t mean anything

Users are not expected to remember any of the concept IDs but to search on the words

391281002 | Mental health assessment |

703257008 |

Assessment of cause of psychotic and

behavioural symptoms |

28

This is a child of 391281002 but the IDs are not related.

Users are not expected to remember SNOMED CT IDs

Slide29

Not ambiguous

In Read, some terms were ‘overloaded’ i.e. there were used in multiple waysExample:

29

Assessment Name

Score

Procedure

Observable

Generalized anxiety disorder 7 item

Generalized anxiety disorder 7 item scale (assessment scale)

Assessment using generalized anxiety disorder 7 item score (procedure)

Generalized anxiety disorder 7 item score (observable entity)

Generalised anxiety disorder 2

Generalised anxiety disorder 2 scale (assessment scale)

Assessment using generalised anxiety disorder 2 scale (procedure)

Generalised anxiety disorder 2 scale score (observable entity)

Alcohol use disorders identification test

Alcohol use disorders identification test (assessment scale)

Assessment using alcohol use disorders identification test (procedure)

Alcohol use disorders identification test score (observable entity)

Alcohol use disorder identification test consumption questionnaire

Alcohol use disorder identification test consumption questionnaire (assessment scale)

MISSING

MISSING

Slide30

‘Grouper Terms’

30

O/E – weight was a heading for the different findings of the patient’s weight in Read, but over the years has been used with a value.

Slide31

Expressing searches – Clare Burgon

SNOMED CT in primary care

Slide32

Language of SNOMED CT specifications

Operators in SNOMED CT

-

Expression Constraint Language (ECL)

conceptid

just this code

<

conceptid the descendants of this code<<conceptid

this

code and all its

descendants

MINUS

conceptid

exclude

this code

MINUS <

conceptid

exclude the descendants of this code

MINUS <<

conceptid

exclude this code and all its descendants

^refsetId

members of refset

32

Slide33

Reference set example

33

Slide34

QOF refsets

34

Slide35

Language of SNOMED CT concepts

Concept can be referred to in

computer and human

language, the syntax used is to express the concept in terms of its id and either preferred term or fully specified name.

ID

|term|

233838001

| Acute posterior myocardial infarction (disorder) |

35

Slide36

Expressing what we want to search on

36

We could express a search in two ways:

Use the

hierarchy structure to state:

Concept and all of its descendants

<<

233838001 | Acute posterior myocardial infarction (disorder)

|

Cherry pick the codes

List each concept seperately

233838001

| Acute posterior myocardial infarction (disorder) |

70998009 | Acute myocardial infarction of

posterobasal

wall (disorder) |

15713201000119105 | Acute ST segment elevation myocardial infarction of

posterobasal

wall (disorder) |

Although these produce the same result we’ll see that using the hierarchy is a better way to write the searches

Slide37

Example – Diabetes mellitus

Identify in SNOMED CT everyone with Diabetes mellitus but not during pregnancy

Use browser to find the concepts needed

Selecting each concept separately would be time consuming and potentially prone to human error

So… select the concept and all descendants

<< 73211009 | Diabetes mellitus (disorder) |

37

This concept has 16 children

Many of the child concepts have children too identified by >

Slide38

Example – Diabetes mellitus continued

Say we decided that we did not want to include

diabetes mellitus

that occurs during pregnancy

<<

73211009 | Diabetes mellitus (disorder)

|

MINUS<<199223000 |Diabetes mellitus during pregnancy, childbirth and the puerperium (disorder)|

This search expression would exclude any concepts listed under

concept 199223000

38

Exclude this concept and all descendants

Slide39

Searches originally written in

ReadUnderstanding search results once converted to SNOMED CT

Slide40

Moving to SNOMED CT

Two ways to approach migrating a search from Read v2/CTV3 to SNOMED CT

Take

the previous

search definition

and write a new SNOMED CT search from

scratch

This may include SNOMED CT concepts that do not have a Read v2/CTV3 equivalent

Take the previous

search codes

and map these to SNOMED CT

This

may need extending and

reviewing in future as SNOMED CT concepts with out a Read v2/CTV3 equivalent are used

Slide41

Forward mapping tables to convert to SNOMED CT

For each Read v2 / CTV3 code, a FORWARD map to SNOMED CT is provided where appropriate

There are two mapping tables one for Read v2 and one for CTV3

Maps have been clinically assured, and are designed to work in one direction only

You system supplier should be using the maps to map historical Read v2/CTV3 codes to SNOMED CT

Your system supplier will also be converting national system reports to SNOMED CT

41

Slide42

Role of the mapping tables

Take the previous

search codes

and map these to SNOMED CT

Following example will illustrate why some results may be different once data entry moves to SNOMED CT

42

How can this be transformed into SNOMED CT?

Slide43

Not fully automatic

Cannot not be converted automatically as

contains content not in Read

, this needs a human to

checkExample:

C10..%

Diabetes

mellitus TO << 73211009 | Diabetes mellitus (disorder) |May be okExample

: H33..% Asthma TO

<<

195967001 | Asthma (disorder) |

For example include Exercise induced Asthma , which may not wish to include

43

Slide44

An example

44

Expanded to

73 Read

codes

Which

map to

67 SNOMED CT concepts

Notice two different Read codes go to the same SNOMED CT concept

Slide45

Why searches might vary

Slide46

Introducing the ‘Round trip’

In simple terms we have performed the following:

We have only considered the maps for Read v2 codes that were in our original specification

Are there any other Read v2 codes that map to our list of SNOMED CT concepts?

46

Slide47

Back to the example of Diabetes mellitus

When

checking the

forward mapping tables for the 67 concepts in the converted search we find there are other Read codes not in the original specification

Original Read v2 specification: 73 codesConverted SNOMED CT specification: 67 concepts

‘Round trip’ Read v2 specification:

138

codes47

Slide48

Other mapping features

One Read v2/CTV3 code may go to different SNOMED CT concepts as synonyms were not always true synonyms in Read v2/CTV3

48

Slide49

Why is it important

to know about the maps?

Historical

data stored in Read

will have a SNOMED CT code mapped to it (using the forward maps from Read to SNOMED CT)

Dual coding means that future data entered in SNOMED

CT

will have a Read v2/CTV3 code stored alongside it(using the GP subset from SNOMED CT to Read)The maps mean that running a search in the original Read v2/CVT3 compared to the newly converted SNOMED CT they may produce different results

49

Slide50

So what does this mean

Maybe the original report was missing a code that should have been includedA duplicate Read code was not added to the original report when it should have been

The SNOMED CT report (restricted by the map to Read) could be a more accurate representation….

…. However, it does not use the full list of SNOMED CT concepts. It is not a full SNOMED CT search

50

Slide51

Mapping summary

All searches are expected to be executed at concept level

This is true for:

SNOMED CT

Read v2/CTV3However the maps created for mapping from Read v2 /CTV3 to SNOMED CT have been created at synonym level to allow for the fact that not all Read v2/CTV3 synonyms were true synonyms

Resulting in codes not previously selected being brought into the specification once mapped to SNOMED CT

51

Slide52

Look-up

for Read to SNOMED

52

Browsers

Slide53

Supplier systems

Suppliers will convert their national searches to full SNOMED CT searchesIf you have refined their search, then this will pick up the new search

If you have amended their search, this will need to be reviewed

53

Slide54

Searches housekeeping

SNOMED CT in primary care

Slide55

Impact of SNOMED CT releases

SNOMED CT is updated twice a year, in the UK the files are released 1st

April and 1

st

OctoberVersioning mechanism allows new content to be created and current content to be made inactivateNew content If its relevant to your search, most likely the new concept will have been placed in the SNOMED CT hierarchy as a descendant of one of your search items so your search will automatically pick this up

On rare occasions, this new concept may not be placed where you expect and would not automatically be added to your search

55

Slide56

Searches housekeeping

Meaning searches at some point will

likely:

Need

to be periodically checked to see if new content needs to be added manually Need to be periodically checked to see if content automatically added to your searches is

correct

Inactive

concepts are ok to remain in a search if you search on historical data56

Slide57

Summary

One concept has many descriptionsBuild

searches

on

concept IDs not description IDsA concept can have more than one is-a relationship (parent)Designing a search in SNOMED CT from scratch requires a browser or tools to find the content needed

When designing a new search, if its over historical data, just reflect on any anomalies from Read that you need to account for

SNOMED CT hierarchy structure means searches will generally pick up all the ‘child’ concepts needed if you select a concept and all descendants

Converting an existing search from Read v2/CTV3 to SNOMED CT requires the forward mapping tables but only uses SNOMED CT concepts where there is a map from Read v2/CVT3 – needs extending to all SNOMED CTA converted search may produce different results if run in SNOMED CT to Read v2/CVT3 due to the mapping table structure

57

Slide58

Supplier approaches

ALL suppliers wish to minimise the impact and provide time to adjust searchesOn SNOMED Day 1, all existing searches will still work

They will continue to work while users enter data that has a Read equivalent – suppliers solutions will help

New codes have to use,

eg, new immunisation, would have done new report anywayPrioritise reports and plan migrating them over the next year, possibly 2 years

58

Slide59

Own solutions?

Tools based on Read codes will need ‘updating’If via in-house solutions, need to:

Download SNOMED

CT

files from TRUDCreate some additional tablesWrite new searches based on SNOMED CT hierarchy

Understand some of the nuances

wrt

Read to SNOMED CT mapsSNOMED CT concepts can be up to 18 digits long, set excel sheets as text59

Slide60