Desiderata for Controlled Medical Vocabularies in the Twenty-First Century

1
Desiderata for Controlled Medical Vocabularies in
the Twenty-First Century

• James J. Cimino

2
Overview

• Define need for a controlled vocabulary
• Disclaimer
• Desiderata I-XII
• Future Directions
• Predictions

3
Defining the Need for a Controlled Vocabulary

• Application developers don't want to create or
  maintain them
• Sharing data
• Multipurpose
• Maintenance

4
Disclaimer

• I searched the literature as best I could
• I interpreted what I found as best I could
• I tried to generalize and unify opinions,
  theories and assertions
• I welcome additions/corrections to desiderata
  list
• and references

5
Desideratum I Content

• Must seek to be provide breadth and depth
• Atoms versus molecules
• A formal methodology is needed
• Structure must not limit size

6
Desideratum II Concept Orientation

• Concepts, not terms
• One meaning (nonvague)
• No more than one meaning (nonambiguous)
• Higher-level ¹ ambiguous
• One concept per meaning (nonredundant)
• May have multiple context-dependent meanings

7
Desideratum IIIConcept Permanence

• Old concepts can't be deleted (Randy Miller's
  "emeritus terms")
• Example non-A-non-B hepatitis
• Names can be changed as long as meaning doesn't
  change (retronyms)
• Example transvenous pacemaker

8
Desideratum IVNonsemantic Concept Identifiers

• Don't use the name
• Don't use a code that will run out of room
• Don't use a hierarchical code
• Meaningless integer (/- check digit)

9
Desideratum V Polyhierarchy

• Needed for tree walking
• Needed for inferencing
• Needed for "essence"
• Example diseases of the liver which also involve
  the kidney

10
Desideratum VIFormal Definitions

• Support understanding
• Support maintenance
• Structured and controlled (not narrative)
• Represented through relationships within the
  vocabulary
• Defintional versus assertional knowledge
• Additional effort minimal and will pay off

11
NEC
12
NEC
13
Desideratum VII Reject "Not Elsewhere
Classified"

• Can never have a formal definition
• Vocabulary changes induce semantic drift

14
Desideratum VIIIMultiple Granularities

• Different levels for different purposes
• Uncertainty is allowed, imprecision is not
• We must be precise about our uncertainty

15
Desideratum IXMultiple Consistent Views

• Multiple views for multiple purposes
• Must not lead to inconsistency

16
A
C
B
E
D
F
H
G
I
17
A
C
B
E
F
D
(G,H)
(I)
18
A
E
D
F
H
G
I
19
A
C
B
E
D
F
H
G
I
20
A
C
B
D
21
A
C
B
E
D
F
E
H
G
I
22
A
C
B
E
D
F
E
H
G
I
23
Desideratum XRepresenting Context

• Needed a grammar to show usage
• "What is sensible to say"
• Consider modeling "events"

24
Controlled Medical Vocabulary
Definitional Knowledge
(how concepts define each other)
Finding
Pathologic Function
Body Site
is -a
has- site
is -a
is -a
Joint
Joint Swelling
Swelling
is -a
has- abnormality
Right Knee
25
Controlled Medical Vocabulary
Assertional Knowledge
Definitional Knowledge
(how concepts combine)
(how concepts define each other)
Finding
Pathologic Function
Body Site
modifies
is -a
has- site
Finding
Body Site
is -a
is -a
Joint
Joint Swelling
Swelling
is -a
has- abnormality
Right Knee
26
Controlled Medical Vocabulary
Assertional Knowledge
Contextual Knowledge
Definitional Knowledge
(how concepts combine)
(how concepts are used)
(how concepts define each other)
Finding
Pathologic Function
Body Site
part-of
Physical Exam
Finding
modifies
is -a
has- site
Finding
Body Site
part -of
is -a
is -a
Joint
Joint Swelling
Hospital
Progress
Swelling
is -a
Admission
Note
part-of
has- abnormality
Right Knee
27
Controlled Medical Vocabulary
Assertional Knowledge
Contextual Knowledge
Definitional Knowledge
(how concepts combine)
(how concepts are used)
(how concepts define each other)
Finding
Pathologic Function
Body Site
part-of
Physical Exam
Finding
modifies
is -a
has- site
Finding
Body Site
part -of
is -a
is -a
Joint
Joint Swelling
Hospital
Progress
Swelling
is -a
Admission
Note
part-of
has- abnormality
Right Knee
Hospital Admission
Progress Note
Physical Exam
Progress Note
Progress Note
Joint Swelling Right Knee
Electronic Medical Record
28
Desideratum XIGraceful Evolution

• Will always need to fix mistakes
• Medical knowledge will grow
• Bad reasons
• Good reasons

29
Desideratum XIGraceful Evolution

• Bad reasons
• - Redundancy
• - Major name changes
• - Code reuse
• - Changed codes

30
Desideratum XIGraceful Evolution

• Good reasons
• - Simple addition
• - Refinement
• - Precoordination
• - Disambiguation
• - Obsolescence
• - Discovered redundancy
• - Minor name changes

31
Desideratum XIIRecognize Redundancy

• Synonyms are good
• Redundant concepts are bad
• Redundant expressions are inevitable
• Example
• - Year 1 "Pneumonia", "Left Lower Lobe"
• - Year 2 "Left Lower Lobe Pneumonia"

32
Desideratum XIIRecognize Redundancy
33
Desideratum XIIRecognize Redundancy
Left Lower Lobe Pneumonia
34
Desideratum XIIRecognize Redundancy
Left Lower Lobe Pneumonia is-a Pneumonia

35
Desideratum XIIRecognize Redundancy
Left Lower Lobe Pneumonia is-a Pneumonia
has-site Left Lower Lobe
36
Desideratum XIIRecognize Redundancy
Left Lower Lobe Pneumonia is-a Pneumonia
has-site Left Lower Lobe participates-in
Finding
37
Desideratum XIIRecognize Redundancy
Left Lower Lobe Pneumonia is-a Pneumonia
has-site Left Lower Lobe participates-in
Finding
38
Future Directions

• Sharable
  vocabulary
• multipurpose
• high quality
• Some desiderata are easy
• Some are tractable
• Some will require philosophical shifts

39
Easy Desiderata

• Nonsemantic concept identifiers
• Polyhierarchy
• Multiple granularities
• Multiple consistent views

40
Tractable Desiderata

• Content
• Recognize redundancy
• Graceful evolution

41
Philosophical Shifts

• Concept orientation
• Concept permanence
• Formal definitions
• Reject "Not Elsewhere Classified"
• Representing context

42
Predictions What will Happen Soon?

• Vocabularies will become concept-oriented
• Nonsemantic identifiers
• Semantic networks and multiple hierarchies
• Standard notation (conceptual graphs?)
• Maintenance will become tenable

43
Predictions What will Take Longer?

• Formal definitions
• Multiple consistent views
• Recognition of redundancy
• But...
• ... implementation-specific
  solutions will be possible

44
Discussion

• Desiderata too many or too few?
• Arguments for and against
• Citations accurate? complete?
• -