Topic Maps for Exploring Nosological, Lexical, Semantic and HL7 Structures for Clinical Data

pogonotomygobbleAI and Robotics

Nov 15, 2013 (3 years and 6 months ago)

145 views

Proceedings of the 12
th

International

Symposium on Health Information

Management Research


iSHIMR 2007


1




Topic Map
s

for Exploring Nosolog
ical
, Lexical,
Semantic and HL7 Structures for Clinical Data

Grace I. Paterson
1
,

Andrew M. Grant
2
,

Steven D. Soroka
3

1
Medical Informatics, Faculty of Medicine, Dalhousie University, 5849 University Avenue,
Halifax, NS, Canad
a B3H 4H7,
grace.paterson@dal.ca

2
Collaborative Research for Effective Diagnostics,
Universit
é

de
Sherbrooke, Centre de
développement des biotechnologies de Sherbrooke, Parc Biomédical du CHUS, Sherbrooke
(Québec), Canada, J1E 4K8,
andrew.grant@usherbrooke
.ca

3
D
ivision of Nephrology,
Department of Medicine, Dalhousie University,
1278 Tower Road,

Halifax, NS, Canada

B3H 2Y9,

steven.soroka@cdha.nshealth.ca

A topic map is
implemen
t
ed
for
learning about clinical

data

associated with a hospital
stay

for patient
s

diagnosed with Chronic Kidney Disease
, Diabetes and Hypertension
.
Topic maps allow us to use concepts and relations among concepts to express
statements about the way we organize subject matter.
T
he question

posed is
:
How
might a
topic map help
bridge per
spectival differences among communities of
practice

and help m
ake commensurable the different classifications they use
?
The
knowledge layer of the
topic map
wa
s
generated from existing ontological
relationships in boundary objects.
The

boundary objects

inc
lude
d

nosology systems
(
SNOMED

3.5, SNOMED CD, ICD
-
9, ICD
-
10
-
CA), lexicons (UMLS
, English, French
),
semantic indices and HL7
Clinical Document
Architecture (CDA)

markup standard
.
Discharge summaries, patient charts and clinical data warehouse entries reifi
ed the
clinical knowledge used in practice. Th
is

clinical data
wa
s normalized to HL7 CDA
and
stored i
n the Clinical Document Repository.
Each CDA entry was given a subject
identifier and linked with the topic map.
The ability of t
opic maps
to

function as t
he
infostructure “glue”
is assessed using

dimensions of
semantic interoperability

and
commensurability.

Keywords

topic map,
boundary

object,
c
hronic kidney disease
,

classification

systems,

common ground

1. Introduction

The Boundary Infostructures for Chron
ic Disease research aims to model the common
ground in the communication space.
The major prerequisite for
the emergence of a mutual
semantic foundation on which to base a common ground

are explicitly expressed concepts

understandable by humans and compute
rs

[1]
.

Boundary objects have attributes that enable
them to serve as translation devices among members of different communities of practice.
They are
an intellectual tool that is flexible enough to deal with needs and constraints of
several parties while
retaining a common identity across sites
.


Classification systems and standards are considered boundary objects because their
structure is common enough to make them function as a means of translation
[2]
.
There are
two major types of classification system
s: enumerative and analytico
-
synthetic
.
Enumerative
classification attempts to assign headings for all subjects of the past, present and
anticipatable future and enumerate them. The
International Classification of Diseases (
ICD
)

Proceedings of the 12
th

International

Symposium on Health Information

Management Research


iSHIMR 2007


2




is enumerative. The ICD has

been evaluated as a boundary object and deemed to be a
pragmatic construction. It is “tailored to the degree of granularity that can be realistically
achieved”
[
3
]
.

An a
nalytico
-
synthetic classification
approach to terminology requires the
analysis of a s
ubject domain into different facets, the analysis of a concept into its atomic
components and the synthesis of knowledge through concept definitions that depend on
relating one concept to another. The Systematized Nomenclature of Medicine
Clinical Terms
(S
NOMED

CT
) is analytico
-
synthetic.
The
core

building blocks
of SNOMED
CT
are
the
concepts table
, relationships
table
and descriptions
table.

SNOMED
CT
was deemed the best
choice of terminology for Canada’s interoperable electronic health record
, and Canada
was a charter
country for the establishment of the International Health Terminology Standards Development
Organization that acquired SNOMED CT from the College of American Pathologists

[
4
]
.


Topic maps can merge the different ontological expressions of kno
wledge embedded in
classification systems and standards. Such a
topic map can facilitate comparison of different
constructions placed on the same set of clinical data by members of different communities of
practice.
Perspectival differences among communiti
es of practice arise, in part, from
differences in their ways of classifying concepts.
The aim of this research
is to show
how a
boundary infostructure might make commensurable the different classifications used by
clinicians, health informaticians, admini
strators, medical educators and patients.

One of the
principles of social learning is that learning occurs at the boundaries
[
5
]
.

An ontology is t
he formalization of a conceptualization
. A topic map consists of both an
o
ntology and an instance of that onto
logy
[
6
]
.

It

provide
s

knowledge organization services,
which
can help determine

the semantic similarity of resources for a given topic
[
7
]
.
The basic
building blocks for knowledge structures are categories, concepts and relations
[7].

An
ontology based on
the classification of entities in reality rather than the classification of
concepts in thought
may help ensure

interoperability of coding systems

[
8
]
.


Clinicians provide documentation on the medical encounter with a patient. Health
informaticians
provide

health information standards, terminology systems and

coding rules.
Administrators apply the coding rules to the clinicians’ documentation. Medical educators
work at the boundary between theory and clinical practice, and
generate prototypical cases
to tea
ch about problems and solutions. Patients are at the centre, since it is their story that is
being documented.

The topic map provides a common interface to illustrate the different classifications used by
these
stakeholders
.
T
he set of perceptions are nee
ded to get a complete picture of an
information object, such as a clinical document.
It will be differently constituted based on the
coding that is done. The idea of deliberately constructing a boundary infostructure so that it
explicitly recognizes “the d
iffering
constitution of information objects within the diverse
communities of practice that share a given infrastructure” came from Bowker and Star

[2].


The question posed is:
How might a
topic map help
bridge perspectival differences among
communities o
f practice

and help make commensurable the different classifications they
use?
The ability of topic maps to function as the infostructure “glue” is assessed using
dimensions of
semantic interoperability

and commensurability.

2
. Methods

The clinical domain
is chronic kidney disease secondary to diabetes and hypertension.
The
first step was to define the essential elements that should be captured about these patients.
This was determined from a literature search

[9, 10]
, interviews with stakeholders and revie
w
of published health information standards

[11]
.


The
second step was a semantic analysis of clinical

data
from

two different sets of hospital
data coming from

two geographical jurisdictions (Halifax, Nova Scotia and Sherbrooke,
Quebec). The recordkeeping

is in English in the former and
in
French
for

the latter.
The
analysis resulted in a set of
terms that described the subject matter.

Proceedings of the 12
th

International

Symposium on Health Information

Management Research


iSHIMR 2007


3




The third step was to constrain the vocabulary to facilitate the organization of the subject
matter separately from data
instances.
To facilitate semantic interoperability, the subject
matter was coded using a set of classification systems and information standards. These are
boundary objects.
Every classification system is a boundary object because it serves as a
common poi
nt of reference for managing knowledge across a given boundary.


The fourth step was to structure the su
bject matter
using
patterns that have nosological,
lexical, semantic and HL7 dimensions.
The terms used to describe the subject matter were
mapped to SN
OMED CT codes.
For the constrained vocabulary, there existed a one
-
to
-
one
correspondence between the SNOMED versions, CT and 3.5, and a many
-
to
-
many mapping
from SNOMED CT to ICD
-
9
-
CM.
The clinical data
was
normalized using HL7 Clinical
Document Architectu
re (CDA) and stored in the Clinical Document Repository. The
Chronic
Kidney Disease
Discharge Summary template was used to structure the three electronic
discharge summaries generated from the medical encounter described in the patient chart.
The Quick Sta
rt Guide for Simple CDA Release 2.0 Documents
[
1
2
]

was used to structure 72
entries representing the documents in the patient chart of a Halifax patient. The Quick Start
Guide for Care Record Summary Documents Using CDA Release 2.0
[
1
2
]

was used to
structu
re 18 entries
: 17 ho
spital stays
by

five
Sherbrooke
patients
in the CIRESSS (
Centre
Informatis
é

de
Recherché

É
valuative en Services et Soins de Sant
é
)

dataset
and one Halifax
patient.

T
he
fifth
step wa
s to
implement the
design

as

a topic map
and populate
it with information
about the topics and link
s

to instances in the Clinical Document Repository.
The schema for
the Chronic Kidney Disease Topic Map was created using Ontopoly
[
1
3
]
.

The
sixth
step was to
organize group work based on action research
[1
4
]
t
o seek common
ground around data collected in two jurisdictions.
The

potential user contexts
were
:
clinician
s
, health informatician
s
, administrator
s
, medical educator
s

and patient
s
.


3
. Materials

Two sites were involved with this research and provided clin
ical data for the
Chronic Kidney
Disease
Topic Map: Renal Program, Capital District Health Authority, Halifax, Nova Scotia,
Canada and

CIRESSS
,

Centre for Research and Evaluation in Diagnostics, Universit
é

de
Sherbrooke, Qu
é
bec, Canada. The

clinical data f
rom Halifax
was
a patient chart, three
discharge summaries produced from the chart and a discharge abstract. The diagnoses,
procedures, lab results and medications
were
coded for the Halifax data. The clinical data
from Quebec
was
160 patient cases encompa
ssing 302 hospital visits. Diagnosis and
procedure coding
were
available for all Quebec cases. Lab results
were
available for 5
patients, encompassing 17 hospital visits.

SNOMED CT and ICD
-
10
-
CA
were
used for diagnostic coding of Halifax data, and ICD
-
9 f
or
Quebec data.
The two
procedure
classification schemes used
we
re
: Canadian Classification
of Diagnostic, Therapeutic and Surgical Procedures (CCP)

for Quebec data;

and Canadian
Classification of Interventions (CCI)

for Halifax data. The remaining data wa
s coded manually
using SNOMED CT
.
The
UMLS Metathesaurus
played a role in mapping across schemes.


4
. Results

4.1 Domain Analysis

The clinical concepts in text were identified using terminological extraction methods. The
MetaMap Transfer program
[15]

was u
sed for automatic mapping of text to UMLS concepts, with
preference given to concepts arising from SNOMED CT, SNOMED 3.5 and ICD
-
9
-
CM sources.


Proceedings of the 12
th

International

Symposium on Health Information

Management Research


iSHIMR 2007


4




4.
2

Subjects in the
Chronic Kidney Disease

Topic Map

The
knowledge in the boundary infostructure was organized
using a
Ch
ronic Kidney Disease
Topic Map, which is depicted in Figure 1. CDA instances in the Clinical Document Repository
were linked to the Chronic Kidney Disease Topic Map as examples of subject expression.


Figure 1
Design for
Chronic Kidney Disease
T
opic Map

The subject layer
wa
s constructed from three data sources: patient chart, discharge
summaries and hospital data. There
we
re 877
cod
able

clinical concepts

in the
patient chart
and
discharge summaries
.
There
we
re 595 unique ICD
-
9 codes
and 158 uniqu
e CCP codes

in the
hospital data from CIRESSS
.

A mapping
from the

French
hospital data to the English
codable clinical concepts was conducted manually for lab results, services and specialists.
The subjects and their attributes were organized into relation
al tables.

4.3 Code Subject Matter Using Boundary Objects

A set of boundary objects was chosen as the reference terminology systems. These included
SNOMED CT, SNOMED 3.5, UMLS, ICD
-
9, ICD
-
9
-
CM, ICD
-
10
-
CA, CCP, CCI, ATC
(Anatomical Therapeutic Chemical, a c
lassification system used by Nova Scotia’s drug
formulary) and HL7. The final subject set was constrained to those topics needed to fully
define a clinical concept using a faceted classification approach
[16]
. System topics were
introduced through the use
of Ontopoly software in the creation of the topic map
[13]
. The
Clinical Document Repository functioned as a boundary object linking how the subjects were
used in clinical documents with how they were represented in terminology systems.

Proceedings of the 12
th

International

Symposium on Health Information

Management Research


iSHIMR 2007


5




4.4 Structures in t
he Chronic Kidney Disease Topic Map

The syntax for the topic map was drawn from topic map design patterns for hierarchical and
faceted classifications and thesauri
[1
7
]
,

description logic patterns in SNOMED CT and
clinical statement patterns in HL7. Togeth
er they formed the topic map ontology.

Those implementing SNOMED CT in HL7
-
based electronic health records
, such as the CDA,

are encouraged to adhere to a set of common patterns. The subject matter for diagnoses,
procedures, lab results and medications wa
s coded according to these patterns.

As an
example, t
here
we
re 84 subjects used for referent

tracking in the patient chart. The subjects
were mapped to SNOMED CT concept identifiers. Concepts

we
re categorized
by facet
according to the SNOMED nosology as: p
rocedure (N=43);
pharmaceutical/biologic
product
(N=13);
clinical finding

(N=1
6
); observable entity (N=10); and
situation with explicit context

(N=2).



Already coded data was

mapped to other terminology systems to facilitate elaboration. Each
diagnosis in

CIRESSS was coded to ICD
-
9 and had a French description.
Figure 2 gives an
example

of how this information was further elaborated by including its English description
and UMLS concept unique identifier
.




Figure 2
Topic Map Entry for Subject 403.91
.


Th
e cross
-
mapping
s

from SNOMED CT to ICD
-
9
-
CM
were

provided to both SNOMED CT
and UMLS

licensees
.
However, a

d
ifferent versio
n

of ICD
was

used for cod
ing CIRESSS
dataset from Sherbrooke, Quebec.
There are two version of ICD
-
9: the original ICD
-
9 which
was pu
blished in 1975, and ICD
-
9
-
CM (Clinical Modification) which was published in 1986 by
the National Center for Health Statistics (NCHS) and includes new categories compared to
the original version. Both versions are
maintained and
available for download from

the
Center for Disease Control website
[1
8
]
. A discrepancy between the two versions is due to
the new category "diabetes with hyperosmolarity" added in ICD
-
9
-
CM.
Table 1

shows the
different English interpretations for
Diabetes Mellitus codes in the two ve
rsions.




Proceedings of the 12
th

International

Symposium on Health Information

Management Research


iSHIMR 2007


6




Table 1
Differences in ICD9 and ICD
-
9
-
CM
versions for Diabetes Mellitus c
oding

Code
Number

ICD
-
9

ICD
-
9
-
CM

250.0

Diabetes mellitus without mention of
complication

Diabetes mellitus without mention of
complication

250.1

Diabetes with ketoacidosi
s

Diabetes with ketoacidosis

250.2

Diabetes with other coma

Diabetes with hyperosmolarity

250.3

Diabetes with renal manifestations

Diabetes with other coma

250.4

Diabetes with ophthalmic manifestations

Diabetes with renal manifestations

250.5

Diabetes
with neurological manifestations

Diabetes with ophthalmic manifestations

250.6

Diabetes with peripheral circulatory
disorders

Diabetes with neurological manifestations

250.7

Diabetes with other specified
manifestations

Diabetes with peripheral circulator
y
disorders

250.8


Diabetes with other specified
manifestations

250.9

Diabetes with unspecified complication

Diabetes with unspecified complication


The lexical category
wa
s language (e.g., English, French). The UMLS
Met
a
thesaurus

is a
lexical resource,

and UMLS function
ed

as a “switching language” for the subjects coded to
HL7, ICD
-
9, ICD
-
10
-
CA, SNOMED CT
or

SNOMED 3.5.


The nosology structure
wa
s the classification of a clinical concept in SNOMED, a
n analytico
-
synthetic
classification, and/or ICD, an e
numerative classification.
An analytic
-
synthetic
classification supports inference. The analysis process breaks the domain down into facets
and atomic components and the synthesis process defines concepts through relationships to
other concepts.

The reaso
n for a medication can be inferred from the ontological relationships among
concepts in SNOMED’s
clinical findings

and
pharmaceutical/biologic
product classes.

The
use case was to determine whether or not a patient was taking medication for hypertension.
T
he patient has been prescribed
Hytrin
, which is a tradename
.

The concept was not found in
SNOMED CT because that terminology scheme did not include tradenames. The semantic
relationships in UMLS were used to find the appropriate generic name and its associ
ated
SNOMED CT code. From the SNOMED CT ontological relationships, it was inferred that
Hytrin was a medication, and one of its uses was treating hypertension.



Hytrin

has
-
CUI
C0591628



Hytrin

has
-
source
-
asserted
-
synonymous
-
concept
Terazosin hydrochloride



Terazosin hydrochloride

has
-
tradename

Hytrin



Terazosin hydrochloride

has
-
SNOMEDCT_code
s

318771000

and
326510000



SNOMEDCT_code
318771000

stands
-
for

Duplicate Concept



SNOMEDCT_code 318771000 same
-
as SNOMEDCT_code
326510000



SNOMEDCT_code
326510000

stands
-
for

Terazosin hydrochloride [see chap B for
generic preps]

and is
-
a
terazosin



T
erazosin

is
-
a
Alpha 1 adrenergic blocking agent
|
Alpha 1 adrenergic blocking agent

is
-
a
Alpha
-
adrenergic blocking agent
|
Alpha
-
adrenergic blocking agent

is
-
a
vasodilating
agent
|
V
asod
ilating agent is
-
a hypotensive agent



H
ypotensive agent

ha
s
-
synonym

antihypertensive

drug

The nosological categories for the 9th and 10th revisions of ICD are used in this research.
The Ninth Revision included an optional alternative method of classifying d
iagnostic
statements, including information about both an etiology for an underlying general disease
Proceedings of the 12
th

International

Symposium on Health Information

Management Research


iSHIMR 2007


7




and a manifestation in a particular organ or site. The Tenth Revision was introduced into
Nova Scotian hospitals in 2002. Both versions
we
re represented in

the data used in this
research.
There is often a mismatch between the clinician’s words and the expression in
ICD
-
10
-
CA. As an example, if a clinician states, “Chronic Renal Failure due to Diabetes”, the
health record coder would be advised to enter a dag
ger (†) and an asterisk (*) code. These
are: E11.22†, “Type 2 Diabetes Mellitus with end
-
stage renal disease (ESRD)” and N08.3*
“Glomerular disorders in diabetes mellitus (E10
-
E14† with common fourth character .2)”
(152). The dagger and asterisk system is
for dual classification for certain diagnostic
statements.

F
our semantic indices

were used to categorize content
. T
he
clinical statement patterns
we
re
16 lab results that
we
re used as markers for disease progression
[
1
1
]

and for triage. The HL7
Clinical St
atement Model provide
d

a common method for communicating
these patterns
. The
17 disease categories
we
re based on the Charlson Index, a categorization scheme used by
administrators for case mix groups

[
1
9
]
. The five

procedure categories
we
re based on CCP
ch
apter

[
20
]
.

The
six

complications categories
we
re based on definitions in the National
Diabetes Surveillance System
[
21
]
.
Figure
3

illustrates how disease category (Congestive
Heart Failure) and complications category (Heart Failure) are associated with th
e subject
428.1.



Figure
3

Disease Category and Complications Category for Same ICD
-
9 Code
.

Figure
4

il
lustrates how the semantic indices and

the
CDA entries in the
Clinical Document
Repository

were integrated using the topic map for knowledge organizati
on
.


Figure
4
Integrating Semantic Index, Subjects and Clinical Document Repository
.


Proceedings of the 12
th

International

Symposium on Health Information

Management Research


iSHIMR 2007


8




4.
5

Context in the

Chronic Kidney Disease
Topic Map

HL7 CDA documents are expressed using XML. The XML Style Language (XSL) is used to
get human
-
readable formats. The XS
L can be written to locate data fields and transform
them for other purposes. This facilitates transforming from one HL7 CDA form,
e.g.
,
Chronic
Kidney Disease

D
ischarge
S
ummary, to another,
e.g
., Care Record Summary for the
patient
’s longitudinal record
.
The set of
common
document sections
we
re: Diagnosis,
Allergies, Immunizations, Social History, Family History, Vital Signs, Lab Tests, Medications,
Procedures, and Plan of Care. The process of aligning between the
C
hronic Kidney Disease

Discharge Summary a
nd the Care Record Summary

confirmed that we could populate
a

longitudinal patient record from multiple encounters.

Figure
5

illustrates how a CDA is linked
for patient with multiple encounters.




Figure
5


Linking from Patient

Identifier to CDA Rendition for User Context
.


4.
6

Action Research for Finding Common Ground

A measure of success for the boundary infostructure will be its ability to be used to learn
about changes in health associated with care for a particular conditio
n.
The
Chronic Kidney
Disease Topic Map
organized the subject matter from the boundary infostructure. It
categorized the CDA entries along 4 dimensions (clinical statement patterns, Charlson
disease categories, procedures and complications). This helped lo
cate similar documents
and returned instances where a subject was discussed in the context of a clinical document.

The ability of this topic map to help learners from multiple contexts reach a common ground
needs to be assessed.
The planned method for the
evaluation of learners is the action
research method
[1
4
]
. The participants
are

from the Halifax and Sherbrooke jurisdictions
, and
will explore the topic map from five user contexts:

clinicians, health informaticians,
administrators, medical educators and
patients.

The problem is supporting multiple user perspectives and achieving common ground through
a discussion of differences. The topic map is proposed as a tool to facilitate the discussion.
The domain is chronic disease management for patients with
chronic kidney disease,
diabetes and hypertension. Participants will be asked to navigate the topic map, explore how
concepts are represented,
clarify

differences

in how subject matter is classified in the two
jurisdictions and identify common ground
.
With

respect to their different scientific
backgrounds, they will be asked if they perceive the topic map as flexible enough to
Proceedings of the 12
th

International

Symposium on Health Information

Management Research


iSHIMR 2007


9




represent the way they organize their knowledge.
Participants will be asked to engage in
reflecti
o
n to assess the experience gained
through
group work via
a web conference
.
The
social context is
the group’s
shared work with standards organizations
. The construction of
the topic map by health informaticians using available ontologies will be evaluated for its
usability in a learning sit
uation
.

A survey will be used to query what are the essential elements of a boundary infostructure,
what issues might prevent such an infostructure from being developed, what are the major
advantages and disadvantages of using a topic map to represent suc
h an infostructure, and
what advice do they have for improving the topic map. L
essons that arise from this activity
should be generalizable as new knowledge that can be shared with our colleagues in
standards development.

5
.
Conclusions

5
.1
Semantic Inter
operability

Topic maps
we
re proposed as a solution to the semantic interoperability problem. They
enable
d

cross mappings among terminology systems and pragmatic enhancement of clinical
concepts through
information structuring
. They
were
used for inference,

as was illustrated for
determination of the reason for a medication
.

The topic map addressed deficiencies in
individual boundary objects. For example, neither SNOMED 3.5 nor SNOMED CT included
medication brand names in their terminology system. This defic
it was addressed through a
merger with the terms from UMLS, where both the
trade
name and the generic name of a
medication are linked
, albeit with using different semantic relations for each instance
.


The topic map
provide
d

a store of information that c
o
uld

be
navigated to gain an
understanding of knowledge organization from multiple perspectives.
By ensuring that the
electronic health information that is needed is expressed the same way for all patients, we
can achieve semantic interoperability among the

community
-
based physicians making
referrals, and the clinicians acting on the referrals. Administrators, such as the booking
clerks, act on the same information in scheduling appointments.


The Ontopoly application was
able to configure

the topic map

by a
dapting

existing
ontology

but required

mostly manual methods

in its construction
. The Omnigator application
, also from
Ontopia,

supported navigation of the topic map from multiple perspectives.


Further work is required to automatically
load the topic map
from information stored in flat
files. Concepts, Relationships and Descriptions are flat files that SNOMED CT provides to
licensees. The data in these files
wa
s used to generate the description
logics that link
ed

SNOMED CT concepts to each other.

5
.
2

Comme
nsurability

The
topic map

addresse
d

the commensurability of different classifications used by
stakeholders: clinicians, health informaticians, administrators, medical educators and
patients
.

The UMLS was able to function as a switching language
for the sub
jects coded to
ICD
-
9,
ICD
-
10
-
CA, SNOMED CT
,
SNOMED 3.5

and HL7, since all these are source
vocabularies in UMLS
.

The topic map was useful for identifying that there were different
interpretations for the same ICD code number, 250.3, in ICD
-
9 and ICD
-
9
-
CM.
SNOMED CT
was also able to function as a switching language.
Subjects coded in ICD
-
9 or ICD
-
10
-
CA
were mapped to concept descriptions from SNOMED CT. The SNOMED nosology system
provided concept descriptions which defined a concept in terms of other concept
s. Th
e
web
conference with users

is designed to
illustrate the potential for the

topic map
to

mediate
learning at the boundary between and among different communities of practice.

Its

objective
Proceedings of the 12
th

International

Symposium on Health Information

Management Research


iSHIMR 2007


10




is to find

common ground through clarification of the differen
ces between data collected in
two jurisdictions. Th
e

feedback will help determine

how useful

and usable
the topic map is for
bridging perspectival differences among different communities of practice
.

Acknowledgements

Initial research was supported by a tr
aining grant from Canadian Institutes for Health
Research (CIHR) Strategic Training Program in Health Informatics.

Refe
r
ences

1.

Gal U, Yoo Y, Boland RJ. The dynamics of boundary objects, social infrastructures and social
identities. Sprouts: working papers
on information environments, systems, and organizations.
2004;4(4):193
-
20
6.

Available from:
http://sprouts.case.edu/2004/040411.pdf.

2.

Bowker GC, Star SL. Sorting things out:
c
lassification and its consequences. Cambridge, Mass:
MIT Press; 1999.

3.

Bowker G, S
tar SL. Situations vs. standards in long
-
term, wide
-
scale decision
-
making: the case of
the International Classification of Diseases. IEEE proceedings of the twenty
-
fourth annual Hawaii
international conference on system sciences; 1991 January 8
-
11.

4.

Canada’
s EHR initiatives to benefit from SNOMED CT terminology standard. [updated April 26,
2007]. Available from: http://www.infoway
-
inforoute.ca/en/News
-
Events/InTheNews_long.aspx?UID=267

5.

Wenger E. Cultiv
ating communities of practice: a

guide to managing knowle
dge. Boston
, Mass
:
Harvard University Press; 2002.

6.

Durusau P, Newcomb S, Barta R. Topic maps

reference model, 13250
-
5 version 6.0. 2006 Mar
ch
1. Available from:
http://www.isotopicmaps.org/TMRM/TMRM
-
6.0/TMRM
-
6.0.pdf

7.

Park J, Hunting S
,

editors. XML topic ma
ps: creating and using topic maps for the Web. Boston,

Mass:

Addison
-
Wesley; 2002.

8.

Ceusters W, Smith B. Strategies for referent tracking in electronic health records. J biomed

inform

2006;39(3):362
-
378.

9.

Clase CM, Garg AX, Kiberd BA. Classifying kidney prob
lems: can we avoid framing risks as
diseases? BMJ 2004 Oct 16;329(7471):912
-
915.

10.


Jindal K, Chan CT, Deziel C, Hirsch D, Soroka SD, Tonelli M. Hemodialysis clinical practice
guidelines for the Canadian Society of Nephrology. J.Am.Soc.Nephrol. 2006;17(3 Sup
pl 1):S1
-
27.

11.


Western Health Information Collaborative. Western Canada Chronic Disease Management
Infostructure Initiative. Standards Document Guide. Government of Alberta. 2006 Jan; (1).

12.


Alschuler Associates Inc. CDA support: Quick Start Guides.

[update
d January 2006].

Available
from:
http://www.alschulerassociates.com/cda/?topic=quick
-
start
-
guides

13.


Ontopia Solutions for Managing Knowledge and Information. Onto
pol
y
.

[copyright 2005
]
. Available
from:
http://www.ontopia.net/solutions/ontopoly.html

14.


Lau F,
Straub D, Hayward R. Fostering an Internet
-
Based Work Group in Clommunity Health
Through Action Research. Journal of Healthcare Information Management. 2001;15(3):207
-
221

15.


Aronson AR. Effective mapping of biomedical text to the UMLS Metathesaurus: the Meta
Map
program. Proc.AMIA.Symp. 2001:17
-
21
.


16.


Denton W. How to Make a Faceted Classification and Put It On the Web [updated Nov. 2003].
Available from:
http://www.miskatonic.org/library/facet
-
web
-
howto.html

17.


Ahmed K. Topic
m
ap
d
esign
p
atterns
f
or
i
nformation
a
rchitecture. XML Europe, Londra 2003:5
-
8.

18.


National Center for Health Statistics Publications and Information Products. Available from:
http://ftp.cdc.gov/pub/Health_Statistics/NCHS/Publications/

19.


Charlson ME, Pompei P, Ales KL, McKenzie CR.
A new method
of classifying prognostic
comorbidity in longitudinal studies: development and validation. J Chron Dis,
1987;40(5):
373
-
383
.

20.


Lalonde A, Taylor E. Medical classification systems in Canada: moving toward the year 2000
.
CMAJ, 19
97
;157(11):1561
-
1565
.

21.


Pearce M
. National diabetes surveillance system reference document (V2.1).
Public Health
Agency of Canada (PHAC) 2005 March 31.