Evolution from Read Codes to SNOMED CT

1
Evolution from Read Codes to SNOMED CT

• Tim Benson
• tim.benson_at_abies.co.uk

2
Outline

• SNOMED CT evolved from Read Codes
• Understanding origins of Read Codes can explain
  some of the challenges in implementing SNOMED CT

3
Why it is Hard

• Scale and diversity of use
• People and machines have different needs
• Clinical pragmatics (fit into daily tasks)
• Clinical linguistics and formal concept
  representation
• Clinical conventions do not fit logical paradigms
• Formal knowledge representation is hard
• Achieving clinical consensus is hard
• Existing schemes are idiosynchratic
• Consistency with electronic records
• Change must be managed
• Rector A. Clinical Terminology Why is it so
  hard. 1999
• http//www.cs.man.ac.uk/7Erector/aim-bio-paper-re
  ctor.pdf

4
Spackmans Rules

• First Rule of Coding
• yesterdays data should be usable today
• First Rule of Data Quality
• the quality of data collected is directly
  proportional to the care with which options are
  presented to the user
• Kent Spackman, Chicago, 2006

5
Fundamental Dichotomy
6
First Rule of Coding

• Yesterdays data should be usable today
• We have to live with Read Codes and similar code
  systems
• SNOMED CT was designed to simplify migration
• by preserving the mistakes of its ancestors

7
Timeline
SNOMED CT
2000
SNOMEDRT
READ 3 (CTV3)
1990
READ 2
SNOMED 3
Read 4 byte
1980
SNOMED 2
1970
SNOP
8
Origins of Read Codes

• Commercial motivation
• to sell systems to individual GPs
• commoditisation
• keep it simple
• Quick to set up a practice system
• Quick to enter data speed writing
• Quick to analyse standard reports

9
History
10
Design Constraints
11
Components

• Codes (4/5 char)
• Position dependent hierarchy
• Index strings 4/5 char
• Look up terms
• Allow synonyms and homonyms
• Text descriptor (30/60 char)
• Text is definition, with clues from position in
  taxonomy

12
Content
13
Read Codes

• 4 byte version (1986)
• 32,187 codes
• Version 2 (1989)
• 61,762 codes
• CTV3 (1994)
• 230,000 concepts
• SNOMED CT (2008)
• 370,000 concepts

14
Read Codes (Diseases)
15
Exploding Bicycle

• 10 things to hit
• Pedestrian / cycle / motorbike / car / HGV /
  train / unpowered vehicle / a tree / other
• 5 roles for the injured
• Driving / passenger / cyclist / getting in /
  other
• 5 activities when injured
• resting / at work / sporting / at leisure / other
• 2 contexts
• In traffic / not in traffic
• V12.24 Pedal cyclist injured in collision with
  two- or three-wheeled motor vehicle, unspecified
  pedal cyclist, non-traffic accident, while
  resting, sleeping, eating or engaging in other
  vital activities
• Alan Rector 2006

16
Fundamental Differences
17
Code-based hierarchy

• Fixed for ever
• impossible to maintain
• Single hierarchy
• medicine is multidimensional
• Only one way to analyse data easily
• users expect to retrieve data in the same
  groupings that they use to enter it

18
Reference Terminology

• Everything is an expression
• description logic
• pre-coordination
• post-coordination
• Less than 20 of SNOMED CT is sufficiently
  defined
• the rest is primitive
• Defines what is always true
• it is not an ontology

19
Reference Terminology Example
Laparoscopic appendicectomy
Laparoscopicprocedure on appendix
Appendix
Appendicectomy
Laparoscopic excision
20
Expression Handling

• Quite complex concepts
• Close to user
• Stated
• Normative
• Transitive Closure

21
Expression (Single Level)
22
Nested Expression
23
First Rule of Data Quality

• The quality of data collected is directly
  proportional to the care with which options are
  presented to the user

24
Read Codes

• Original Version
• 4 levels top to bottom
• One of the reasons for shallow hierarchies
• 60 sub-nodes (0-1, A-Z, a-z, ., !, _at_ etc)
• Hand-crafted to facilitate data entry
• Original UI showed up to 10 items only

25
Subtype hierarchy Looking from leaf to root
Root
Clinical finding
Finding by site
Disorder
Disorder by body site
Finding of body region
Disorder of body system
Finding of limb structure
Disorder of cardiovascular system
Disorder of extremity
Finding of lower limb
Vascular disease
Disorder of lower extremity
Thrombotic disorder
Disease of vein
Peripheral vascular disease
Venous thrombosis
Vascular disorder of lower extremity
Deep venous thrombosis
Thrombosis of vein of lower limb
Deep venous thrombosis of lower extremity
Deep vein thrombosis of leg related to air travel
26
SNOMED CT

• Complex machine-built hierachies
• Necessity for navigation sub-sets/reference sets
• Navigation hierarchies not provided out of the box

27
Conclusions

• Original Read codes were simple, straightforward
  and understandable out of the box
• SNOMED CT designed to solve problems of Read
  Codes
• Read Codes expansion between versions 1 and 3
• Need to provide this human interface
• Cannot leave this to implementers

28
Questions?