Terminological Systems in Medicine

1
Terminological Systems in Medicine

• Ronald Cornet, PhD
• Dept. of Medical Informatics
• Academic Medical Center University of Amsterdam
• r.cornet_at_amc.uva.nl

2
Overview

• Part I The role of Coding, Classification and
  Terminology in Registration of Patient Data
• Part II SNOMED CT

3
Registration of Patient Data The role of
Coding, Classification and Terminology
4
Outline

• Context
• Coding Classification
• Coding Systems Overview
• Coding in Practice
• Coding Systems, the next generation
• Types of systems Requirements

5
Starting point E-record

• Cost of care
• Quality of care

6
Motivation cost of care

• Digitizing medical records in the U.S. could save
  the health care industry as much as81 billion a
  year and help medical practitioners avoid
  mistakes
• The study found that electronic medical records
  systems save money by reducing redundant care,
  speeding patient treatment and improving safety.

Can Electronic Medical Record Systems Transform
Healthcare? An Assessment of Potential Health
Benefits, Savings, and Costs - Sep. 14,
2005 Hillestad R, Bigelow J, Bower A, Girosi F,
Meili R, Scoville R, and Taylor R (Rand
Corp.) Health Affairs, Vol. 24, No. 5.
7
Motivation for Clinical Terminology

• Costs
• Terminology use benefits entire health system
• Save as much as 5 of total healthcare costs
• up to 100 Billion per year in US
• Source - Walker J et al., Market Watch
  200519th January10-18

8
Motivation quality of care

• Multiple registration of patient information
• Free text detailed description for daily care
• Less detailed descriptions
• DRG (Diagnosis-Related Group)
• ICD code
• Limited integration of patient information
• Hampered consistency

9
Use of Patient Data

• Documentation in the EPR/EHR
• Decision support
• Clinical audit
• Reporting
• Summaries
• Administrative management information
• Epidemiology
• Billing
• Resource management

National Health Services United Kingdom
10
Documentation of Patient Data

• Free text
• Expressive, Maximal freedom, Precise
• Ambiguous
• Hardly computer-processable
• Coded
• Limited expressiveness
• Potentially less ambiguous
• Computer-processable

11
Outline

• Context
• Coding Classification
• Coding Systems Overview
• Coding in Practice
• Coding Systems, the next generation
• Types of systems Requirements

12
Why using codes

• Data reduction
• Standardization
• Avoiding problems with natural language
• Acute heart attack
• Acute myocardial infarct
• Acute myocardial infarction
• Myocardial infarction acuta
• Acute coronary thrombosis
• ? Solution 410 Acute myocardial infarction

13
Coding

• A code is a sequence of symbols which refers to a
  concept and which can be used for identification
  and selection

14
Coding

• Example coding gender
• Male m
• Female f

15
Principles for defining Codes

• Explicit eligibility criteria (definitions)
• E.g. genotypic, phenotypic gender
• Disjoint categories
• male, female
• Exhaustive categories
• male, female, other, unknown
• Reasonable
• Klinefelter's syndrome (XXY) ?

16
Types of Codes

• Significant
• Mnemonic
• Juxtaposition
• Hierarchical
• Non-significant / context free
• Random
• Sequential

17
Mnemonic Codes

• Formed from one or more of the characters of its
  related class
• M Male, F Female
• KL204 KLM flight 204

18
Juxtaposition Codes

• Composite codes consisting of segments
• Room J-1B-115

building
floor
location
19
Hierarchical Codes

• Example from ICD-9-cm
• 003 Other Salmonella Infections
• 003.0 Salmonella Gastroenteritis
• 003.1 Salmonella Septicemia
• 003.2 Localized Salmonella Infections
• 003.20 Localized Salmonella Infection,
  unspecified
• 003.21 Salmonella Meningitis
• 003.22 Salmonella Pneumonia
• Aggregation, retrieval on different levels

20
Reasons for using significant codes

• Codes can be remembered
• Meaning can be derived from code
• (Juxtaposition and hierarchical) codes can be
  used for aggregation

21
Problems with significant codes

• Mnemonics

07.45 KL 1124 Copenhagen Arrived 07.4107.45 KQ
1124 Copenhagen Arrived 07.4107.45 MH 9264
Copenhagen Arrived 07.4107.45 NW 8400 Copenhagen
Arrived 07.41
1 flight, 4 codes!
22
Problems with significant codes

• Hierarchical codes
• 003 Other Salmonella Infections
• 003.0 Salmonella Gastroenteritis
• 003.1 Salmonella Septicemia
• 003.2 Localized Salmonella Infections
• 003.20 Localized Salmonella Infection,
  unspecified
• 003.21 Salmonella Meningitis
• 003.22 Salmonella Pneumonia
• No other aggregation than Salmonella
  infections, e.g. Meningeal infections

23
Non-significant codes

• Random pick any (unique) number
• Sequential number consecutively, e.g., start by
  1 and increase
• Such meaningless codes should NOT be presented to
  users

24
Outline

• Context
• Coding Classification
• Coding Systems Overview
• Coding in Practice
• Coding Systems, the next generation
• Types of systems Requirements

25
Classification

• Classifying
• Designing a classification
• Assigning a class to an object

26
Classes of Objects

• How many classes do you see below?
• Eagle
• Elephant
• Shark
• Telephone
• Television
• Videocamera

27
Classes of Objects

• How many classes do you see below?

28
Classification Principles

• Aristoteles (384BC - 322BC) definitio per genus
  proximum et differentia specifica(definition by
  the nearest higher class and differentiating
  properties).
• Classes fulfill criteria of superclasses
• Classes are more specific than superclasses

29
Classification example Biology

• African Elephant Taxonomy
• Kingdom Animal
• Phylum Chordata / Craniata / Vertebrata
• Class Mammalia / Theria / Eutheria / Afrotheria
• Order Proboscidea
• Family Elephantidae
• Genus Loxodonta (African elephants)
• Species Loxodonta africana

Full lineage is over 20 levels! http//www.ncbi.nl
m.nih.gov/Taxonomy/taxonomyhome.html/
30
Example ICD-10

• Certain Infectious and parasitic diseases
• Viral infections of central nervous system
• A87 Viral meningitis
• A87.2 Lymphocytic choriomeningitis

Systems such as ICD-10 typically contain 10.000s
to 100.000s of terms (codes)
31
Classification Chapters ICD-10 (1)

• Certain Infectious and parasitic diseases
• Neoplasms
• Diseases of the blood and blood forming organs
  and certain disorders involving the immune
  mechanism
• Endocrine, nutritional and metabolic diseases
• Mental and behavioural disorders
• Diseases of the nervous system
• Diseases of the eye and adnexa

32
Classification Chapters ICD-10 (2)

• Diseases of the ear and mastoid process
• Diseases of the circulatory system
• Diseases of the respiratory system
• Diseases of the digestive system
• Diseases of the skin and subcutaneous tissue
• Diseases of the musculoskeletal system and
  connective system
• Diseases of the genitourinary system

33
Classification Chapters ICD-10 (3)

• Pregnancy, childbirth and the puerperium
• Certain conditions originating in the perinatal
  period
• Congenital malformations, deformations and
  chromosomal abnormalities
• Symptoms, signs and abnormal clinical and
  laboratory findings, n.e.c.
• Injury, poisoning and certain other consequences
  of external causes
• External causes of mortality
• Factors influencing health status and contact
  with health services

34
Single ordering (monohierarchy)

• Pros
• Categories are mutually exclusive (disjoint)
• ? No double counts
• Straightforward, understandable
• Cons
• Only 1 supported categorization
• Disjointness often artificial

35
Multiple Ordering (polyhierarchy)

• Pros
• Multiple aspects (axes) for ordering,
  e.g.Anatomic location, Etiology, Morphology
• Multiple paths to items
• Cons
• Double counts (e.g. Viral Meningitis is both
  Infectious disease and Meningeal disease)
• More complex

36
Single or Multiple Classification?

• Documentation in the EPR/EHR
• Decision support
• Clinical audit
• Reporting
• Summaries
• Administrative management information
• Epidemiology
• Billing
• Resource management

Coding
Classification
National Health Services United Kingdom
37
Coding which information?

• Shortly after dinner on the day before admission
  to the hospital, this 48-year-old obese woman
  developed a cramping, epigastric pain that
  radiated to the back, followed by nausea and
  vomiting. The pain was not relieved by position
  or antacids. The pain persisted, and 24 hours
  after onset, the patient sought medical
  consultation. The patient was admitted to the
  hospital with a diagnosis of acute pancreatitis.
  Radiological findings included widening of the
  duodenal C loop and blurring of the left psoas
  muscle margin. Serum amylase was 1120 units per
  liter. The day after admission, the patient
  seemed to improve. However, that evening she
  became disoriented, restless, and hypotensive.
  Despite intravenous fluids and norepinephrine,
  the patient remained hypotensive and died 8 hours
  later.

Lu TH, Shih TP, Lee MC, Chou MC, Lin
CK. Diversity in death certification a case
vignette approach. J Clin Epidemiol. 2001
Nov54(11)1086-93.
38
Classification cause of death?

• Shortly after dinner on the day before admission
  to the hospital, this 48-year-old obese woman
  developed a cramping, epigastric pain that
  radiated to the back, followed by nausea and
  vomiting. The pain was not relieved by position
  or antacids. The pain persisted, and 24 hours
  after onset, the patient sought medical
  consultation. The patient was admitted to the
  hospital with a diagnosis of acute pancreatitis.
  Radiological findings included widening of the
  duodenal C loop and blurring of the left psoas
  muscle margin. Serum amylase was 1120 units per
  liter. The day after admission, the patient
  seemed to improve. However, that evening she
  became disoriented, restless, and hypotensive.
  Despite intravenous fluids and norepinephrine,
  the patient remained hypotensive and died 8 hours
  later.

Lu TH, Shih TP, Lee MC, Chou MC, Lin
CK. Diversity in death certification a case
vignette approach. J Clin Epidemiol. 2001
Nov54(11)1086-93.
39
Outline

• Context
• Coding Classification
• Coding Systems Overview
• Coding in Practice
• Coding Systems, the next generation
• Types of systems Requirements

40
Overview of Coding Systems

• Large number of systems
• Unified Medical Language System (UMLS)
  Metathesaurus 1 includes over 100 systems,
  totaling more than 1.000.000 medical concepts
• Systems are large
• Number of concepts has increased from 100 to gt
  100.000

1 http//www.nlm.nih.gov/pubs/factsheets/umlsmeta.
html
41
Overview of Coding Systems

• Diseases
• ICD
• Specialties
• Anatomy
• Literature
• Genomics

42
ICD

• London Bills of Mortality(16th century)
• 60 disease categories
• Collected by parish clerks
• International List of Causesof Death (19th
  century) ICD
• International Classificationof Diseases (20th
  century) ICD-10, tenth revision of ICD

43
Overview of Coding Systems

• Diseases
• SNOMED Systemized Nomenclature of Medicine
• Specialties
• Anatomy
• Literature
• Genomics

44
SNOMED CT
1965 SNOP 1974 SNOMED 1998 SNOMED Version
3.5 2000 SNOMED RT ( 3.5 READ) 2002 SNOMED CT

• Aims at coding of detailed informationfirst
  episode of severe, acute E-coli pneumonia with
  sudden onset
• Formal definitions provide multiple
  classifications

45
Overview of Coding Systems

• Diseases
• Specialties
• DSM Mental Health
• ICPC Primary Care
• Anatomy
• Literature
• Genomics

46
Overview of Coding Systems

• Diseases
• Specialties
• Anatomy
• Terminologia Anatomica
• FMA Foundational Model of Anatomy
• Literature
• Genomics

47
Overview of Coding Systems

• Diseases
• Specialties
• Anatomy
• Literature
• MeSH Medical Subject Headings
• Genomics

48
Overview of Coding Systems

• Diseases
• Specialties
• Anatomy
• Literature
• Genomics
• GO Gene Ontology

49
Overview of Coding Systems

• And many, many more

CDT5
CPT
WHOART
ICD-9-CM
NANDA
ICD10
NIC
DSM4
MedDRA
Loinc
NOC
Omaha
COSTAR
UltraSTAR
OMIM
http//www.nlm.nih.gov/research/umls/sources_by_ca
tegories.html
50
Outline

• Context
• Coding Classification
• Coding Systems Overview
• Coding in Practice
• Coding Systems, the next generation
• Types of systems Requirements

51
Coding in Practice, scenario 1

• Clinician records items as codes
• By entering (recollected) codes
• e.g. gender
• By using a pick list
• By searching for phrases
• e.g. meningo

52
Coding in Practice, scenario 2

• Clinician records free text
• Clinical Coders derive codes
• Different codes are needed for different purposes
  (billing, mortality, )

53
Patient trajectory
task terminological phrase
admission operation for lower third rectum cancer
scheduling abdominoperineal amputation of rectum
reporting low anterior resection of rectum with double stapling technique
discharge other anterior resection of rectum, ICD-9-CM 48.63
reimbursement operation for rectum cancer, DRG 147
cost analysis anterior resection of the rectum with double stapling technique
quality assurance low anterior resection of rectum without temporary colostomy and operation for lower third rectum cancer
Rossi Mori
54
ICD Code accuracy

• Main error sources along the patient
  trajectory include amount and quality of
  information at admission, communication among
  patients and providers, the clinicians knowledge
  and experience with the illness, and the
  clinicians attention to detail.
• Main error sources along the paper trail
  include variance in the electronic and written
  records, coder training and experience, facility
  quality-control efforts, and unintentional and
  intentional coder errors, such as
  misspecification, unbundling, and upcoding.

O'malley KJ, et al.Measuring Diagnoses ICD Code
Accuracy. Health Serv Res. 2005 Oct40(5 Pt
2)1620-39.
55
Coding consequences

• Codes often lack detail that is clinically
  necessary
• Coding can be time-consuming
• Searching the correct code
• Searching the correct description

56
Data in Patient Records

• Patient data are attributes values
• In a record, these data are stored as record
  items having a value.
• Male patient gender male
• Kidney patient disease renal failure

57
Coding practice

• Patient record item
• Blood pressure
• Hemoglobin level
• Disease

• Patient data/code values
• 120/80
• Normal
• Serum Hepatitis

Definitions are crucial!
58
Coding Quiz (1) Rh positive

• What is the record item, and what is the code
  value, for a statement that the patient is Rh
  positive?

Item Value
Blood type Rh positive
Rh D antigen status Positive
Rh blood type D positive
Blood bank test result Rh positive

Example from Kent Spackman
59
Coding Quiz (2) MRSA

• What is the record item, and what is the code
  value, for a test on blood culture of Staph.
  Aureus for methicillin resistance?

Item Value
Blood Cult MRSA
Blood Cult Methicillin S.Aureus Resistant
MRSA Cult Positive

Example from Jim Cimino
60
Coding practice Current status

• ICD-9(CM) and ICD-10
• used globally, both for mortality and morbidity
• SNOMED CT
• Licensed to 9 countries, Canada and UK as
  frontrunners

61
Outline

• Context
• Coding Classification
• Coding Systems Overview
• Coding in Practice
• Coding Systems, the next generation
• Types of systems Requirements

62
3 generations of coding systems

• First-generation systems, (e.g. ICD, MeSH)
• fixed organization (typically hierarchical)
• simple representation such as a systematic list
  that is alphabetically indexed
• Second-generation (e.g. MedDRA, SNOMED Int)
• dynamic organization (i.e. provide multiple
  hierarchies)
• compositional, combining the simple list
  representation of concepts with a knowledge base
  to define and extend these concepts
• Third-generation (e.g. SNOMED CT, Gene Ontology)
• based on formal models providing symbols denoting
  concepts and a set of formal rules to manipulate
  them

63
First generation Problems

• One fixed hierarchy
• Lung diseases
• Inflammation of lungs
• Lung tumor
• Liver diseases
• Inflammation of liver
• Liver tumor

64
First generation Problems

• More detail requires many combinations
• Infective Pneumonia
• Severity severe, mild, moderate, fatal
• Course acute, subacute, chronic, cyclic
• Cause virus, bacterium, fungus
• ?at least 44348 combinations

65
Second-generation systems(e.g. LOINC, SNOMED
International)

• Dynamic organization (i.e. provide multiple
  hierarchies)
• Compositional, combining the simple list
  representation of concepts with a knowledge base
  to define and extend these concepts
• Published electronically

66
Example SNOMED International

• T-28000 Lungs
• T-62000 Liver
• M-40000 Inflammation
• M-8FFFF Neoplasm
• D2-0007F Pneumonia T-28000 M-40000
• D5-F150F Liver tumor T-62000 M-8FFFF

Topography
Morphology
67
Second generation Problems

• Different codes for one concept
• D2-00004 Infective pneumonia
• DE-00000T-28000 Infectious Disease Lungs
• M-40000L-00110T-28000 Inflamm. Infectious
  Agent Lungs
• ? A problem for selection and grouping, e.g.
  based on T-28000
• Impossible combinations
• M-12000T-28000 Fracture Lungs

68
Third-generation systems(e.g. SNOMED CT, GALEN)

• Formal models providing symbols denoting concepts
• and a set of formal rules to manipulate them
• Published electronically, requiring dedicated
  software

69
Example SNOMED CT

• infective pneumonia ?

70
Third generation problems

• Harder to comprehend
• Harder to implement
• Harder to maintain

71
Vision for the Future

• Patient data are registered
• Detailed, for everyday clinical care
• Structured, for automated processing
• Once, to reduce duplicate efforts
• Patient data are used
• For a wide range of purposes
• For automated support

72
Outline

• Context
• Coding Classification
• Coding Systems Overview
• Coding in Practice
• Coding Systems, the next generation
• Types of systems Requirements

73
Types of systems

• Coding system, classification, vocabulary
• Many names for (often) the same thing!
• We prefer Terminological System as an umbrella
  term

74
Terminology

• a list of terms

75
Thesaurus

• order list of terms, synonyms

enterovirus type 72 HAV - Hepatitis A
virus hepatitis A virus human enterovirus
72 human enterovirus serotype 72 human hepatitis
A virus infectious hepatitis virus
76
Vocabulary

• Definitions!
• Infective pneumonia, free text definition
• Inflammation of the lung parenchyma
  characterized by consolidation of the affected
  part, the alveolar air spaces being filled with
  exudate, inflammatory cells, and fibrin.
• Formal definition

77
Classification

• hierarchical order

78
Coding system

• Codes!

79
Nomenclature

• rules for combinationsor the result of
  applyingthese rules

80
Terminological Systems

• Terminology
• Thesaurus
• Classification
• Vocabulary
• Nomenclature
• Coding System

? list of terms ? ordered terms/synonyms ?
member_of arrangement ? definitions ? composition
rules ? codes as designators
81
Examples

• SNOMED-CT Terminology, Thesaurus,
  Classification, Vocabulary, Nomenclature, Coding
  System
• ICD-10 Terminology, (Thesaurus), Classification,
  Vocabulary, Nomenclature, Coding System

82
Requirements of TS (1)

• Terminology to adequately describe patients
  health problems and the care process
• Requirements
• Domain completeness
• Post-coordination
• Context-free codes
• Non-ambiguity
• Synonyms
• Multiple languages

83
Requirements of TS (2)

• Structure that supports aggregation of
  homogeneous groups
• Requirements
• Domain completeness
• Post-coordination
• Context free codes
• Definitions
• Non-ambiguity
• Non-redundancy
• Multi-classification
• Explicit relations
• Crossmapping

84
Summary

• Documentation in the EPR/EHR
• Decision support
• Clinical audit
• Reporting
• Summaries
• Administrative management information
• Epidemiology
• Billing
• Resource management

National Health Services United Kingdom
85
Summary

• Terminological systems
• can support a broad range of use
• come in various types, each with their own
  characteristics
• still need to be further developed and researched

86
Steps towards use of coded data

1. Electronic Patient Record
2. Structure data items
3. Structure data values
4. Determine/develop appropriate Term.Systems
5. Integrate Terminological Systems into EPR
6. Record detailed coded data as soon as possible
7. Use coded data and term. system for analyses

87
The 81.000.000.000 question

• What is the best terminological system?

88
The answer(s)

• There is no such system
• It depends on your needs
• It depends on your possibilities(e.g. technical,
  financial)
• The terminological system that satisfies your
  needs and is properly implemented and used

89
Round-up of Part I
90
Part II SNOMED CT
91
SNOMED CT

• What it is
• What it isnt
• Why (a system like) SNOMED?
• How to use it
• Challenges

92
About SNOMED

• In development more than 40 years
• 1965 SNOP (Systematized Nomenclature of
  Pathology)
• 1974 SNOMED (Systematized Nomenclature of
  Medicine)
• 2000 SNOMED RT (Reference Terminology)
• 2002 SNOMED CT (Clinical Terminology) RT
  Clinical Terms V3 (a.k.a. Read codes)

93
Owner Maintenance

• Until 2006 SNOMED Organization
• Fully owned by CAP (College of American
  Pathologists)
• Since 2007 IHTSDO
• International Health Terminology Standards
  Development Organization
• 9 Member states USA, Canada, Australia, New
  Zealand, UK, Sweden, Denmark, Lithuania,
  Netherlands

94
Figures (July 2007 release)

• 376,046 concepts

95
Figures (July 2007 release)

• 376,046 concepts
• 62 types of relations 1,359,435 instances
• Is a
• Part of
• Causative agent
• Associated morphology
• Laterality

96
Figures (July 2007 release)

• 376,046 concepts
• 62 types of relations 1,359,435 instances
• 1,060,424 English descriptions
• US and UK English
• Also Spanish translation
• Danish and Swedish translations underway
• No Dutch translation, only Latin alphabets

97
What is SNOMED?

• Terminological system
• Codes identify concepts
• Relations between concepts
• Definitions based on relations
• Terms describe concepts and relations
• Rules to compose concepts

98
Wat is SNOMED?
99
Wat is SNOMED?
Definitions based on relations
Composition rules
Codes
Terms
100
What is SNOMED?

• NO Software!
• NOT Perfect
• NO Silver bullet
• NO Total solution
• Part of the solution
• With numerous new challenges

101
Why SNOMED?

• Information is recorded and searched for
• In many different ways
• Acute pneumococcal bronchitis
• Pneumococcal bronchitis course acute
• Acute bronchitis cause pneumococcus
• At many different levels
• Lower respiratory tract infection
• Acute inflammatory disease

102
(No Transcript)
103
Coding versus classification

• SNOMED CT enables recording data in detail, and
  abstracting based on these data
• Classifications (e.g., ICD, DRG) aim at
  aggregation i.e., putting exactly one label on
  patients

104
Applications

• Registration
• Exchange of patient data
• Decision support
• Checking protocols and guidelines
• Creating homogeneous groups of patients
• Healthcare evaluation

105
How to use SNOMED CT

• Invisible
• Coding data items and data values using
  SNOMED CT
• Gender Male 263495000 248153007

106
How to use SNOMED CT

• Invisible
• Coding data items and data values using
  SNOMED CT
• (semi-)automatic conversion of free text to
  SNOMED CT concepts

107
How to use SNOMED CT

• Visible
• Pick lists with SNOMED CT terms/concepts
• Support composition

108
Challenges

• Quality correctness and completeness
• Implementation
• Registration using SNOMED CT
• Exchange of patient data
• Decision support
• Checking protocols and guidelines
• Creating homogeneous groups of patients
• Healthcare evaluation
• (Semi-)automatic classification (e.g., cause of
  death)
• Creating appropriate subsets

109
Suppose you want a terminology

• Determine what you want it for
• E.g., what users, which domain
• Perform a requirements analysis
• Perform a content coverage study
• Take sample from collected data

110
More information

• Online SNOMED CT browsers
• http//snomed.vetmed.vt.edu/sct/menu.cfm
• http//www.jdet.com/
• International Health Terminology Standards
  Development Organisation
• www.ihtsdo.org
• r.cornet_at_amc.uva.nl