Medical Concept Representation: From Classification to

Biomedical Informatics

Mayo, CTSA, and WHO

Prespectives on Clinical Phenotype

NCBO Forum: Ontology of Clinical Phenotypes

Dallas, Sept 2008

Christopher G Chute, MD DrPH

Professor Biomedical Informatics

Mayo Clinic

ICD-11 Revision Steering Group Chair

World Health Organization


The Historical Center of the

Health Data Universe

Billable Diagnoses

Clinical Data

Billable Diagnoses

NCBO Phenotype; © 2008, Mayo Clinic

2


Copernican Healthcare

(Niklas Koppernigk)

Clinical Data

Billable Diagnoses Clinical Guidelines

Medical Literature

Scientific Literature

Clinical Data


3


Mayo: A Century-Long Tradition of

Studying Patient Outcomes

Demographics

Diagnoses

Procedures

Narratives

Laboratories

Pathology…


High-Volume

Data Storage

4


Early

Phenotype


5


From Practice-based Evidence to Evidence-based Practice

Data

Clinical &

Genomic

Registries

Data

Warehouses

& Marts

Inference

Patient

Encounters

Biomedical

Knowledge

Decision

Support

Expert

Systems

Clinical

Guidelines

Knowledge

Management

6



Mayo Clinic

Data Governance

• Enterprise-wide (Eight States)

• Data Modeling, Metadata, and Vocabulary

• High-level representation

• Senior institutional Leaders – Physician led

• Two-tier organizations

• Steering Committee, Stewardship groups

• Multidisciplinary

• Care Providers, Informatics, IT, domain experts


7


Mayo Enterprise Vocabulary Organization

Reference

Vocabularies

Mayo Thesauri Value Sets

Cross Mapping

Tables

External

SNOMED

LOINC

GO

ICDs

CPTs

… ~200

Mayo Internal

Table 22

Table 61

SNOMED mods

Mayo CPTs

WARS

… ~50

UMLS

External

WHO FIC

… ~3

Mayo Internal

Drugs

Disease

Symptoms

Pt Functioning

EDT Aggregation

… ~8

External

JACHO

NACCR Ca

NIH/NCI

… ~1000

Mayo Internal

Flow sheets

MICS apps/screens

Dept systems

Registry screens

Form questions

Inf. for your Phys.

… ~10,000

External

SNOMED↔ICD

CPT↔ICD

LOINC↔SNOMED

FDB↔NDC

… ~500

Mayo Internal

All thesauri

All value sets

Tab 22↔ICD

FDB↔Fomulary

…

… >10,000

LexGrid: Data Model, Data Store and Machinery


8


The Genomic Era

•

The genomic transformation of medicine far exceeds the introduction of antibiotics and aseptic surgery

•

The binding of genomic biology and clinical medicine will accelerate

•

The implications for shared semantics across the basic science and clinical communities are unprecedented


9


The Continuum Of Health Classification

Biology meets Clinical Medicine

Biology Medicine

10

9

5

4

3

2

1

0

8

7

6

Chasm of Semantic Despair

G ic s en om

P ro

M et ic s te om

P at hw ay s ar

M od el in g ab ol ic

M ol ec ul ol ec ul

M ar

S im

C el ul at io n lu la r

M

M ol ec od el s ul ar

A

G en ss ay s om ic

T es tin g

B io sp ec im en s

La b

D

D at

Tr is ea ia a ls se

D at

&

S a yn dr

P om

M ed ic e al

Im at ie nt

R g ag in ec or d

D

E

H

R at a

S tr uc tu re s


10


NCBO – A Bridge Across the Chasm


11


Feudal Cognition:

Intellectual Semantic Baronies

• Genetic variation – Genomics

• Haplotypes – Statistical Genomics

• Molecular – Metabolomics, Proteomics

• Binding – Molecular simulation

• Pathways – Physiology and Systems Biology

• Symptoms – Consumer Health

• Rx and Px – Clinical Medicine

• Risk – Public Health, Epidemiology

• Social impact – Sociology, Health Economics


12


Whither Phenotype?

Spans spectrum from enzymes to disease

• Pharmacogenomics – enzyme functionality

• Physiologist – cellular function

• Systems biologist – pathway circuit flow

• Sub-specialist – organ functioning

• Patient/Clinician – disease manifestation

• Public Health – population characteristics

Highly specific to use-case context


13


Can the Genotype be the Phenotype?

• If the “disease” is defined by Genotype…

• Lesch-Nyhan syndrome

• Huntington's disease

• Gene sequence as “sign” of disease

• What is the distinction between them

• Consequences for genotype to phenotype research


14


Whither Disease as Phenotype?

•

Notions of disease are inextricably bound to an underpinning Knowledge Base (rabies)

•

Disease manifestation occurs across multiple planes of human understanding

•

Boundaries among clinical observational elements * are relativistic, rapidly changing, and probably serve little practical purpose

*

Observations, findings, syndromes, diseases, genotype, haplotype, epigenomic change…


15


Practical Application and

Consequences

Biomedical Informatics eMERGE: e lectronic M edical R ecord and GE nomics

• The eMERGE Network is a national consortium formed to develop, disseminate, and apply approaches to research that combine DNA biorepositories with electronic medical record

(EMR) systems for large-scale, high-throughput genetic research.

• Reproducible phenotype from EMRs

17



EMR Phenotypes and

Community Engaged Genomic Associations

• NHGRI U01 HG004599; Mayo Clinic

• CG Chute, PI

• Genome Wide Association Study (GWAS)

• Two Clinical “conditions”

• Myocardial Infarction

• Peripheral Arterial Disease (PAD)

• Balance with ethical and community engagement studies

18



Standardized diagnostic algorithm

Biomarker ECG

Abnormal Evolving Diag

Equivocal Evolving Diag

Incomplete Evolving Diag

Diagnostic

Evolving ST-T

Diagnostic

Evolving ST-T

Equivocal

Absent

Suspect MI

Suspect MI

No MI

No MI

No MI

No MI


Diagnostic

Evolving ST-T

Diagnostic

Evolving ST-T

Equivocal

Absent

No MI

No MI

No MI

No MI

No MI

No MI

19


• Increased troponin

Troponin

• a measurement exceeding 0.05 ng/ml

• Raising or falling pattern for serial troponin

• 0.05 ng/ml

• Historical overlap with CK-MB biomarker

• false positive issues - exclusions

• Ablation, cardiac surgery, CHF-acute, CPR, critically ill, crush/ contusion/ bruising, drug toxicity, embolectomy, endomyocardial biopsy, external defibrillation, hyperthermia, hypotension, hypothyroidism, infiltrative disease, inflamm. dis-cardiac, internal defibrillation, neurological disease, pericarditis, pulmonary embolism, transplant vasculopthy, uncontrolled hypertension, rhabdomyolysis, sepsis

20


CHF-acute

CPR critically ill crush contusion bruising drug toxicity hembolectomy endomyocardial biopsy external defibrillation hyperthermia hypotension hypothyroidism infiltrative disease inflammatory dis-cardiac internal defibrillation neurological disease


Exclusion Criteria for Troponin ablation cardiac surgery

33250, 33251, 33254, 33255, 33256, 33257, 33258, 33259, 33261, 33265, 33266,

33120, 33130, 33300, 33305, 33310, 33315, 33320, 33321, 33322, 33330, 33332, 33400, 33401, 33403, 33404, 33405, 33406, 33410, 33411, 33412, 33413, 33414, 33415, 33416, 33417, 33420, 33422, 33425,

33426, 33427, 33430, 33460, 33463, 33464, 33465, 33468, 33470, 33471, 33472, 33474, 33475, 33476, 33478, 33496, 33600, 33602, 33860, 33861, 33863, 33508, 33510, 33511, 33512, 33513, 33514, 33516,

33517, 33518, 33519, 33521, 33522, 33523, 33533, 33534, 33535, 33536, 33542, 33545, 33572, 33600, 33602, 33606, 33608, 33610, 33611, 33612, 33615, 33617, 33619, 33641, 33645, 33647, 33660, 33665,

33670, 33675, 33676, 33677, 33681, 33684, 33688, 33690, 33692, 33694,33692, 33694, 33697, 33702, 33710, 33720, 33722, 33724, 33726, 33730, 33732, 33735, 33736, 33737, 33750, 33755, 33762, 33764,

33766, 33767, 33768, 33770, 33771, 33774, 33775, 33776, 33777, 33778, 33779, 33780, 33781, 33786, 33788, 33800, 33802, 33803, 33813, 33814, 33820, 33822, 33824, 33840, 33845, 33851, 33852, 33853,

33860, 33861, 33863, 33864, 33945, 93501, 93503, 93505, 93508, 93510, 93511, 93514, 93524, 93526, 93527, 93528, 93529, 93530, 93531, 93532, 93533,

428.0, 428.1, 402.01, 402.11, 402.91, 402.91

92950

411.81, 411.89, 434.11, 434.91, 570, 584.5, 584.6, 584.7, 584.8, 584.9

861.00, 861.01, 929.9

861.01, 924.9

861.01, 924.9

977.9

33910, 33915, 33916, 33917, 33920, 35875, 35876, 92973,

93505

92960

780.6

458.9

244.9

135 [425.8], 164.1, 212.7, 429.1, 277.39, 429.71, 413.1, 425.1, 425.4, 428.9, 429.0, 427.89, 429.89

429.89, 429.0, 422.90, 422.92, 422.93, 422.89, 424.90, 424.99, 424.91, 423.1, 423.2, 423.8, 423.9

92961

340, 341.0, 341.1, 341.20, 341.21, 341.22, 341.8, 341.9, 349.9 342.00, 342.01, 342.01, 342.10, 342.11, 342.12, 342.80, 342.81 , 342.82, 342.90, 342.91, 342.92, 343.0, 343.1, 343.2, 343.3, 343.4, 343.8, 343.9,

344.01, 344.02, 344.03, 344.04, 344.09, 344.1, 344.2, 344.30, 344.31, 344.32, 344.40, 344.41, 344.42, 344.5, 344.60, 344.61, 344.81, 344.9, 345.00, 345.01, 345.10, 345.11, 345.20, 345.21 345.3, 345.40,

345.41, 345.50, 345.51, 345.60, 345.61, 345.70, 345.71, 345.80, 345.90, 345.91,346.00, 346.01, 346.00, 346.01, 346.10, 346.11, 346.20, 346.21, 346.80, 346.81, 346.90, 346.91 pericarditis pulmonary embolism

423.9, 420.90

415.19 transplant vasculopthy 996.80, 996.83

uncontrolled hypertension 401.9 rhabdomyolysis 728.88 sepsis

21


Cardiac Pain

• NLP of the text from EMR

• Can distinguish negated terms

• Search Term:

• Chest pain, thoracic pain, precordial pain, pain chest, chest discomfort, atypical chest pain, exertional chest pain, chest pain exertion, anginal chest pain, anginal pain, chest distress, acute chest pain, chest aching, aching chest, typical chest pain, chest pain, jaw pain, pain jaw, throat pain, pain throat, neck pain and pain neck.

• HPI in 72 hour window around presentation


22


Positive

Predictive

Value

Concordance MI Symptoms

Human vs. NLP

0.90

Nurse Abstractor

Yes No

NLP

Yes 590

No 63


35

80

23


Mayo Clinic Cancer Center

• Informatics Infrastructure

• $3.2M caBIG funding in 2008 (Chute, PI)

• LexBIG development

• Vocabulary Knowledge Center

• BRIDG development and leadership

• Clinical Trials infrastructure

• Common Data Elements

• Consistent and comparable

• Inclusion, exclusion criteria

• Co-morbidities


24


Cancer Phenotype

• Increasingly dependent on genomic characteristics

• Blend with pathology, imaging, laboratory

• Extent of disease measures are crucial

• Comparable and consistent data ultimately depends upon common ontologies

• BiomedGT

• SNOMED CT

• ICD-O

• HUGO

• GO

• Adverse

Events


• Drugs Rx

• Radiation RT

• Surgury Px

25


BRIDG Data Models

• BRIDG is the HL7 RCRIM domain model

• Collaboration among CDISC, HL7, FDA, NCI

• Historically expressed as UML model

• Evolving translation into OWL

• Preserve OWL and UML modes interoperably

• Assert that UML and OWL are isomorphic

• Span the information-terminology model boundary

• Enable reasoner support to create coherent model

• Preserve domain expert access to model in UML

26



Content vs. Structure

Contest or Synergy? Computer Equivalent?

Family History of Breast Cancer

Family History of Heart Disease

Family History of Stroke

Terminologic

Model

Family History

Breast Cancer

Heart Disease

Stroke

Information

Model

Equivalent

Content

27



Mayo CTSA

• Catalog ongoing projects, grants, studies

• Scientific description of research

• Administrative categories (human, animal, …)

• Regulatory (ClinTrials.gov, FDA, etc)

• Patient cohort identification

• Clinical trial recruitment

• Case-report form generation


28


So what does Mayo want?

• Ability to record patient data using common models and vocabulary content

• 100 year tradition

• Approximating reality in electronic modes

• Ability to create overarching warehouse

• Now on 3 rd generation “data trust”

• To support “understanding”

• Research, knowledge discovery, practice insight

• Practice improvement, quality, safety

• Decision support, guidelines, best evidence


29


How are we doing it?

• Department of Health Sciences Research

• Epi, Biostat, Informatics, Health Policy

• Enterprise wide (3 campus presence)

• About 4,000 studies, reports, retrievals per year

• Turnkey “phenotype” templates

• Diabetes (Type I, Type II, metabolic syndrome)

• CAD, MI, PAD,

• Stroke (embolic, aneurysm)

• Asthma, COPD,

• Cancers


30


NLP

• Six year operation of fully coded clinical notes

• About 24M clinical notes

• Added Pathology, Radiology in process

• 18-stage UMIA pipeline of annotators

• Tokenization, part of speech, entity recognition

• Code noun phrases into

• Sign and symptoms – SNOMED

• Diagnoses – SNOMED

• Drugs – RxNorm

• Primary source for Phenotype retrieval


31


From Local to Global

The ICD-11 Project

• 150 year legacy of the ICDs

• World Health Organization leadership

• Explicit definitions of classifications from

Terminology

• Scientific consensus of clinical phenotype

• Distributed authoring model


32


•

ICD11 Use Cases

Scientific consensus of clinical phenotype

•

Public Health Surveillance

• Mortality

• Public Health Morbidity

•

Clinical data aggregation

• Metrics of clinical activity

• Quality management

• Patient Safety

• Financial administration

• Case mix

• Resource allocation

33


Some Premises for ICD-11 development

• Rubrics defined by Aggregation Logics from terminologies

• Human language definitions will be explicit

• “core” representation will be in description logic based ontology

• A linear serialization will be derived as a view to maintain longitudinal consistency

• May require corresponding “rules” for practical use

34



Familiar Points Along Continuum

Modern Health Vocabularies

• Nomenclature – Highly Detailed Descriptions (SNOMED)

• Classification – Organized Aggregation of Descriptions into a Rubric

(ICDs)

• Groupings – High Level Categories of Rubrics (DRGs)

Aggregation Groupers

Nomenclature

Detailed

Classification Groups

Grouped

35



Traditional Hierarchical System

ICD-10 and family


36


Addition of structured attributes to concepts

Concept name

 Definition

 Language translations

 Preferred string


 Synonyms


 Index Terms

37



Addition of semantic arcs - Ontology

Relationships

 Logical Definitions

 Etiology

 Genomic

 Location

 Laterality

 Histology

 Severity

 Acuity


38


Serialization of “the cloud”

Algorithmic Derivation


39


Linear views may serve multiple use-cases

Morbidity, Mortality, Quality, …


40


Proposed Hierarchy of Wiki Authority

by ICD Domain (not implemented)

0 Revision Steering Committee

1 Revision Domain/Topic Working Groups &

WHO-FIC Network

2 Accredited Experts

• Designated by Working Group Members &

WHO-FIC Network

3 Accredited Persons

• Designated by Experts

4 Registered Interested Persons (Public)


41


Discussions with IHTSDO

International Health Terminology (IHT)

• IHT (SNOMED) will require high-level nodes that aggregate more granular data

• Use-cases include mutually exclusive, exhaustive,…

• Sounds a lot like ICD

• ICD-11 will require lower level terminology for aggregation logic definitions

• Detailed terminological underpinning

• Sounds a lot like SNOMED

42



ICD-11

Potential Future States

SNOMED

Ghost ICD

Ghost SNOMED


43


ICD-11

ICD-

Joint

Alternate Future

Effort SNOMED


44


Advantages to Collaboration

• Both organizations avoid “Ghost” emulations

• Both organizations leverage expertise and content

• More resources brought to the table

• Both organizations retain independent intellectual property and derivatives (e.g. Linear formats of ICD-11)

• Mappings become moot

• Aggregation of SNOMED is definitional to ICD

45



Caveat

ICD and IHTSDO

• No agreements have been finalized

• Intellectual property sharing is expected

• Shared tooling is being discussed

• Harmonization Board has been proposed


46

Medical Concept Representation: From Classification to

Mayo, CTSA, and WHO

Prespectives on Clinical Phenotype

Billable Diagnoses

Billable Diagnoses

Clinical Data

Clinical Data

Inference

Decision

Support

The Genomic Era

The genomic transformation of medicine far exceeds the introduction of antibiotics and aseptic surgery

The binding of genomic biology and clinical medicine will accelerate

The implications for shared semantics across the basic science and clinical communities are unprecedented

•

•

•

*

Practical Application and

Consequences

Standardized diagnostic algorithm

Terminologic

Model

Family History

Breast Cancer

Heart Disease

Stroke

Information

Model

Equivalent

Content

Scientific consensus of clinical phenotype

Public Health Surveillance

Clinical data aggregation

Related documents

Products

Support

Medical Concept Representation: From Classification to

Mayo, CTSA, and WHO

Prespectives on Clinical Phenotype

Billable Diagnoses

Billable Diagnoses

Clinical Data

Clinical Data

Inference

Decision

Support

The Genomic Era

The genomic transformation of medicine far exceeds the introduction of antibiotics and aseptic surgery

The binding of genomic biology and clinical medicine will accelerate

The implications for shared semantics across the basic science and clinical communities are unprecedented

•

•

•

*

Practical Application and

Consequences

Standardized diagnostic algorithm

Terminologic

Model

Family History

Breast Cancer

Heart Disease

Stroke

Information

Model

Equivalent

Content

Scientific consensus of clinical phenotype

Public Health Surveillance

Clinical data aggregation

Related documents

Add this document to collection(s)

Add this document to saved

Suggest us how to improve StudyLib