Download Controlled Medical Vocabulary in NAHLN

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Veterinary physician wikipedia , lookup

Transcript
Controlled Medical
Nomenclature for NAHLN
NAHLN
January 2005
Las Vegas, NV
Qualifications
There’s “retentive” – if that’s not enough:
There’s “anal retentive” – if that’s not
enough:
There’s “retentive enough to create and
maintain a nomenclature”
Lists of words…
Nomenclature

The system or set of names for things, etc., commonly employed
by a person or community (Petchamp, SNVDO, SNOMED)
Vocabulary

A collection or list of words with explanations of their meanings
(SNOMED)
Classification

The result of classifying; a systematic distribution, allocation, or
arrangement, in a class or classes; esp. of things which form the
subject-matter of a science or of a methodic inquiry. (SNOMED)
What do we need?
Nomenclature ONLY

Provides a simple list for data entry
Vocabulary / Classification



We can be CERTAIN that the “term” (description in
SNOMED) means what we think it means.
We can develop rules that allow us to combine
concepts to express ideas more complicated than
those contained in the nomenclature.
We can use the knowledge base supported by the
vocabulary/classification to search, retrieve and
analyze our data.
NAHLN Nomenclature needs
Is controlled nomenclature (1 or more)
essential – YES
Is LOINC essential – Yes (or something
close)
Is SNOMED essential – NO, but you need
standard microorganisms, breed/species,
etc.
Can SNOMED provide what’s needed
outside of laboratory tests - YES
Why a controlled nomenclature?
Aggregation of text-based content from
multiple sources


Multiple individuals
Multiple institutions
Any time you rely on a computer to
manipulate language and “meaning” is
critical.

Procedures list (for billing)
Why a controlled vocabulary /
nomenclature?
Bovine respiratory disease complex =
Shipping fever = BRDC
How does one automate recognition of the
synonymy?
Why did “we” pick SNOMED?
A shared nomenclature must be
maintained.
SNOMED is the ONLY actively maintained
nomenclature that has veterinary content.
AVMA had a long-term investment
Veterinary medicine CANNOT afford:


To build it’s own competent nomenclature
To continue to live without a competent
nomenclature
Characteristics of a controlled
vocabulary / nomenclature.
Cimino, JJ. Desiderata for Controlled Medical Vocabularies in the TwentyFirst Century Methods of Information in Medicine. 1998 Nov, 37(4-5): 394403
• THE Review of ideal characteristics of a controlled
medical vocabulary / nomenclature.
• Perhaps the best READING review of medical
vocabulary / nomenclature available.
Content, content, content
First criticism of ANY nomenclature (by
users) is lack of content
Content, content, content
Approaches

Enumerate all possible concepts (simple and
complex)
Pneumonia
Pneumonia of left lung
Pneumonia of left caudal lung
This nomenclature will grow quickly. It will
get VERY large VERY fast.
Content, content, content
Approaches


Enumerate all possible concepts (simple and
complex)
Provide all necessary “atoms,” create effective syntax,
“teach” the syntax to users (or systems).
Pneumonia
Left lung
Left Caudal lung
Pneumonia
Pneumonia, body part (left lung)
Pneumonia, body part (left caudal lung)
Probably the theoretical ideal. I have not the first idea how!
Content, content, content
Approaches



Enumerate all possible concepts (simple and complex)
Provide all necessary “atoms,” create effective syntax, “teach”
the syntax to users (or systems).
Drastically limit the scope of the nomenclature.
May actually improve usability of the nomenclature for
a limited purpose, however:
1.
2.
3.
The next project and the next interaction require development
of a new nomenclature
Connections between projects require “mapping”.
Users will not be happy.
Might be easier to build and use a large single
nomenclature.
Concept Orientation
Concept – an embodiment of a particular
meaning



Concepts must correspond to at least one
meaning (non-vagueness)
Concepts must correspond to no more than
one meaning (non-ambiguity)
Meanings correspond to no more than one
concept (non-redundancy)
Concept permanence
Once created, the meaning of a concept is
inviolable.
Concept identifiers can NEVER be reused.

Contributes to stability of legacy information.
Non-semantic identifier
Unique name


Using the name inhibits (prohibits?) improvements in
concept “name”
Harder to cope with synonymy.
Hierarchical designators (codes with meaning)
inhibit classification


Monohierarchies provide inadequate classification
capability for retrieval purposes.
Polyhierarchies cannot be supported.
(A concept can’t be in its multiple logical locations).
Polyhierarchies
Medical concepts often classified in
multiple ways
Acute bacterial pneumonia




It is a bacterial (infectious) disease
It is a pulmonary disease
It is an inflammatory condition
It is an acute condition
Formal definitions
Collection of relationships to other
concepts in the vocabulary
Acute Bacterial pneumonia




Causative agent = bacteria
Has location = lung
Has morphology = inflammation
Has course = acute
Reject “not elsewhere classified”
The fundamental problem is “meaning
drift”
Example

Year one nomenclature has
Pneumonia
Bronchitis
Respiratory disease NEC
Reject “not elsewhere classified”
The fundamental problem is “meaning
drift”
Example

Year TWO the nomenclature has
Pneumonia
Bronchitis
Tracheobronchitis
Respiratory disease NEC
Last year tracheobronchitis was “respiratory disease, NEC”
Reject “not elsewhere classified”
(NEC)
NEC ≠ NOS (Not Otherwise Specified)



NOS = “I’ve said all I’m going to say”.
NOS is not considered to be inappropriate.
Current preference is to exclude NOS. The
concept means what it means, no more no
less.
Multiple granularities
A large standardizing nomenclature (like
SNOMED) must provide concepts at an
appropriate level of detail for every user
class.
Not an issue with a purpose-built
nomenclature.
Recognize Context
Must provide an ability to distinguish between
uses of the concept…



Patient has acute bacterial pneumonia
Patient has history of acute bacterial pneumonia
Patients littermate died of acute bacterial pneumonia
System design and perhaps the message
system should cope with acquiring context
Nomenclature may be required to cope with
context in messages, or may provide context
concepts for the message.
Evolve Gracefully
A nomenclature must have a strategy for coping
with new content.

The bane of home grown nomenclatures
(Who’s looking after your list of concepts?)


Nomenclature maintenance is not everyone’s cup of
tea.
A controlled nomenclature serves as an arbiter of
sorts.
A nomenclature used by more than one
organization (d-lab) must have update and
sharing mechanisms in place.
Recognize redundancy
Redundancy must be avoided in the
concept list
Redundancy must be provided for using a
synonym mechanism.
Egos and dispersed training insure that
there exist multiple ways to say almost
anything.

They’re still talkin’ about the same thing.
SNOMED history / future
1965
2000
Reduce storage size Reduce Storage
size
No longer relevant
Categorize
information
Multiple code-based
hierarchies
Poly-hierarchical
categorization
Pathology content
“All Medicine”
Veterinary content
separate, then
integrated
Integrated content
“Computability” for
retrieval.
Natural language,
artificial intelligence,
decision support
Development history
SNOP

Morphologies
SNOMED

Morphologies, Etiologies, Locations
SNOVET


Same structure as SNOMED
Mix of existing SNOMED, additional veterinary content
SNOMED III


Disorders, Morphology, Living organisms, social context
Veterinary content re-integrated
SNOMED RT

Logic based approach to SNOMED. Axes became hierarchies. Most
significantly, the poly-hierarchic approach to classification.
SNOMED CT

SNOMED RT on steroids. Post merger with CTv3.
SNOMED III (3.5)
Axes
A – Physical agents
C – Chemicals
D – Disease
F – Biological function
G – General qualifiers
J - Jobs
L – Living organisms
M – Morphology
P – Procedures
S – Social context
T – Topography
X – Pharmaceutical
companies
SNOMED CT
Disease
Finding / observation /
clinical history
Procedure
Body structure normal
and abnormal
Organism
Physical Force
Substance
Specimen
Social Context
Attributes
Context dependent
categories
Physical object
Events
Environments and
geographical location
Observable entity
Qualifier value
Staging and Scales
Complaints about SNOMED
It’s too…



Big
Complicated
Expensive
Yes but…



We can make it smaller (sort of), and use small
pieces (for most purposes).
Use it in simple and straightforward ways
But nothin, it’s a bit expensive.
Not the license fees, the cost of making it work.
Is the expense worth it?
IF the long-range goal is useful…

The selected standards adhere to design
specifications that have developed through
hard experience in the medical profession.
Essential / desirable features have been
documented.

The selected standards represent
extraordinary functionality, produced and
maintained at great cost to the medical
profession.
We can leverage these standards for 10¢ / $1.00
Bovine reportable
disease system.
Bovine breeds
Bovine “occupation” and husbandry
concepts
Brief list of reportable diseases
Lab tests that support disease list
Place to put the concepts in the message
structures

Lab to network
Bovine medical record
Bovine Breeds
Bovine lab tests, all not just reportable
All applicable disorders, findings,
procedures
Message structures



lab to clinic
clinic to lab
clinic to clinic
Subsets of standards
Bovine Reportable
Bovine disorders
Bovine practice
SNOMED-CT, HL-7,
LOINC
Mixed practice
Disease reporting system
LOINC1
SNOMED1
HL72
1 = three independent subsets
2 = one subset of necessary messages
Rabies
WNV
FMD