=Paper=
{{Paper
|id=Vol-410/paper-17
|storemode=property
|title=A Methodology for Encoding Problem Lists with SNOMED CT in General Practice
|pdfUrl=https://ceur-ws.org/Vol-410/Paper17.pdf
|volume=Vol-410
|dblpUrl=https://dblp.org/rec/conf/krmed/LauSL08
}}
==A Methodology for Encoding Problem Lists with SNOMED CT in General Practice==
https://ceur-ws.org/Vol-410/Paper17.pdf
Representing and sharing knowledge using SNOMED
Proceedings of the 3rd international conference on Knowledge Representation in Medicine (KR-MED 2008)
R. Cornet, K.A. Spackman (Eds)
A Methodology for Encoding Problem Lists with SNOMED CT
in General Practice
Francis Lau, Ph.D., Ray Simkus, M.D., Dennis Lee, M.Sc.
School of Health Information Science, University of Victoria, Victoria, B.C., Canada
fylau@uvic.ca, ray@wmt.ca, dlhk@uvic.ca
ABSTRACT study of diagnosis and problem lists in a
computerized physician order entry system,
This paper describes a methodology for encoding Wasserman [9] reported that 88.4% of their 8,378
problem lists used in general practice with SNOMED clinical terms were found in SNOMED CT. With the
CT. Our intent is to help general practitioners to addition of 145 site-specific terms they were able to
incorporate SNOMED CT into their existing achieve 98.5% overall content coverage. With the
Electronic Medical Record (EMR) systems with formation of the International Health Terminology
minimal disruption as a first step, thus allowing them Standards Organization (IHTSDO), the historical
to assess its impact prior to full-scale conversion. We barriers to SNOMED CT related to cost and the
started with 1,713 original unique terms that made proprietary nature of the product have now been
up the problem lists from the general practice EMR removed, and national initiatives related to EMR’s
used in the study. We ended with 1,468 unique are emerging to use SNOMED CT as a clinical
concepts after two cycles of matching and revisions terminology in several countries around the world.
that led to 1,347 or ~92% successful matches. The
remaining terms were revised to tease out modifiers Despite such impressive development, the effort to
or secondary concepts that could be used to provide adopt SNOMED CT in Canada has been minimal to
equivalency through post-coordination. While date. There continues to be a concern especially in
skeptics of reference terminology systems often balk the primary care setting where most general practices
at their unwieldy size and complexity for local are made up of small groups of practitioners, of
adoption, this study has demonstrated that, using our whom few are equipped with an EMR. Critics often
methodology, it is possible to create a manageable balk at the enormous size and complexity of
subset of SNOMED concepts for problem lists used SNOMED CT, considering it as too unwieldy and
in general practice with immediate tangible value. costly for local adoption and use. But a review of
data collected from several sites by one author
INTRODUCTION showed the number of codes needed to cover
disorders of at least 1:100,000 occurrence would be
The problem list is the keystone of the medical under 5,000 [10]. Work is underway with IHTSDO
record. In general practice settings, the type of and the WICC group of WONCA to finalize this list
problems presented by patients can be quite diverse. as a potential primary care SNOMED subset [11].
Examples range from non-specific symptoms such as
headaches with unknown cause, to a diagnosis of In this paper, we describe a methodology that we
coronary disease that can be expressed in different have developed based on an ongoing study to encode
ways such as heart attack and myocardial infarction. problem lists using SNOMED CT (July 2007 release)
The choice of terms used in problem lists becomes an for a local general practice in Canada. The intent of
important design issue for the electronic medical this methodology is to enable general practitioners to
record (EMR), since the level of granularity selected incorporate SNOMED CT into their existing EMR
for defining the problems and the actual terms systems within minimal disruption as a first step, thus
entered into the system can affect one’s ability to allowing them to assess its potential impact prior to
retrieve the information afterwards, thus impacting full-scale conversion.
the overall quality of the EMR system.
METHODS
There have been many studies on the design and use
of controlled terminology to encode the problem lists
in EMR systems and their impact on practice [1-8]. Design and Setting
Most of these studies are focused on large institutions For this study, we included all the problem list (PL)
involving a substantive number of clinical terms in terms from the commercial EMR system used by a
order to accommodate the needs of a wide range of local general practice in British Columbia, Canada.
clinicians in the institution. For example in their This setting is typical of many general practices
97
Representing and sharing knowledge using SNOMED
Proceedings of the 3rd international conference on Knowledge Representation in Medicine (KR-MED 2008)
R. Cornet, K.A. Spackman (Eds)
across the country, which are made up of small
groups of general practitioners working in a private No Step Example
medical office, mostly on a fee-for-service basis. The 1 Remove Hodgkin’s disease, NOS o Hodgkin
genitive diseases, NOS
medical office in this study has four general
2 Remove stop Hodgkin diseases, NOS o Hodgkin
practitioners who have worked as a group for 30 words diseases,
years in a township with a population of 100,000 3 Convert to Hodgkin diseases, o hodgkin diseases,
located east of Vancouver, British Columbia. The lowercase
practice has had 8 years of experience using an EMR. 4 Strip hodgkin diseases, o hodgkin diseases
punctuation
At least two of the practitioners record all of the
5 Uninflect hodgkin diseases o hodgkin disease
information on their patients on a daily basis at the phrase
time of encounter or shortly thereafter. Laboratory 6 Sort hodgkin disease o disease hodgkin
and imaging results and consult reports from external words
sources – both electronic and on paper – are entered Table 2a. UMLS normalization steps [8, slide20]
into the EMR either by the practitioners themselves
or the medical office assistant. Matching PL Terms
The process of matching the PL terms involved
Matching Algorithms cycling through the matching algorithms one at a
We applied four matching algorithms used in an time to find the best candidate SNOMED CT
earlier SNOMED CT to ICD-10 mapping project to concepts. For each algorithm we always began with
find matching SNOMED concepts for each of the PL the original terms, then the UMLS normalized terms,
terms [12]. Three are lexical techniques for exact- followed by the stemmed terms. During each cycle,
match, match-all and partial-match. The fourth is we would review the candidate concepts found to
semantic matching that involves retrieving the determine if it was a match, and if so, what type of
current concepts based on historical relationships if match it was based on the algorithm applied. When
the initial SNOMED concepts found were inactive. no matching concepts were found, we would label
These algorithms are summarized in Table 1. the term as unmatched. Our experience with the
matching algorithms had been that, the sooner we
Algorithm Explanation could find a match in the cycle, the greater
1. Exact match Exact string match where all words are confidence we would have that the candidate concept
same and in same sequence, including
punctuation
is appropriate. The preferred order of matching
2. Match all String match where all words are same but selected is always exact first, then all, followed by
not necessary in same order; additional partial. For exact-match and match-all if only
words allowed inactive concepts are found then a semantic-match is
3. Partial match String match where one or more words is done to find their corresponding current concepts
found
4. Semantic match For inactive concepts use historical
through the historical relationships.
relationships Was-A, Same-As, May-Be-A,
Replaced-By to find current concepts Step-5 Explanation
5. Unmatched Assigned when no match is found Stop Frequent short words that do not affect the phrase:
Table 1. Matching algorithms used in this study words and, by, for, in, of, on, the, to, with, no, (nos)
Exclude Words that may change meaning of the word but if
words ignored help to find a term otherwise missed: about,
Normalization Steps alongside, an, anything, around, as, at, because,
In addition to applying the matching algorithms to before, being, both, cannot, chronically, consists,
the original PL terms, we reran the algorithms after covered, does, during, every, find, from, instead,
into, more, must, no, not, only, or, properly, side,
we normalized the PL and SNOMED terms to sided, some, something, specific, than, that, things,
remove “noise” using the Unified Medical Language this, throughout, up, using, usually, when, while
System (UMLS 2007 version) normalization steps, SNOMED [X] – concepts with ICD-10 codes not in ICD-9
shown in Table 2a [13,14]. To improve matching, we Prefixes [D] - concepts in ICD-9 XVI and ICD-10 SVII
[M] – morphology of neoplasm concepts in ICD-O
expanded step-2 to remove both “stop words” and [SO] – concepts in OPCS-4 chapter Z in CTV3
“exclude words” and SNOMED prefixes, shown in [Q] – temporary qualifying terms from CTV3
Table 2b. For step-5 we included the lookup and [V] – concepts in ICD-9 and ICD-10 on factors
stemming methods to uninflect the phrase. The influencing health status and contact with health
services (V-codes and Z-codes)
lookup method uses the UMLS SPECIALIST
Lexicon’s inflection table with ~1 million entries, Table 2b. Expanded UMLS normalization step-2
whereas the stemming method is a computational
technique that reduces word variants to a single Encoding the Problem Lists
canonical form [15,16]. The process of encoding the problem lists extracted
from the EMR followed these steps: (a) tabulating the
frequency of occurrences for all of the original PL
98
Representing and sharing knowledge using SNOMED
Proceedings of the 3rd international conference on Knowledge Representation in Medicine (KR-MED 2008)
R. Cornet, K.A. Spackman (Eds)
terms; (b) cataloguing all of the unique words across partial-matches we selected the SNOMED terms that
the PL terms present; (c) examining all unique words were closest to the PL concept involved, such as
and PL terms to identify and revise for acronyms, GERD gastro-esophageal reflux disease. For
abbreviations, spelling variants and errors; (d) semantic matches we looked up the current concepts
matching the PL terms to SNOMED CT concepts of the matched but inactive SNOMED terms through
using matching algorithms described earlier; (e) their historical relationships, such as Cirrhosis. For
producing detailed and summary outputs to show the post-coordination we added qualifier and refinement
type of matches found; (f) reviewing/verifying the terms to SNOMED concepts or combined those that
matched concepts one term at a time for accuracy; (g) are lexically closest to the original PL terms, such as
repeating steps (c-f) until no further matches could be Atrial fibrillation+Chronic, Kidney disease+Chronic,
found; (h) examine remaining partial-matches for and Headache+Migraine. After the second cycle any
post-coordination; (i) create an index table of all PL remaining partial-matches were treated as
and matched SNOMED terms. As part of this study, unmatched. Initially there were eight PL terms not
we also explored navigating within the SNOMED found in SNOMED CT. Five were spelling errors
hierarchy to examine how the super-types and and were revised for the second cycle (e.g.
relations could be used to improve the quality of hepatomegally o hepatomegaly); three were
recall using the matched SNOMED concepts. legitimate missing terms – vasculopath, pyocystitis
and hypotestosteronemia, where we had to modify
RESULTS the PL term or tag as local extensions. Using these
outputs we created an index table to link the PL
Summary of PL Terms and Matches
terms to their matched SNOMED terms, shown in
A total of 7,833 PL entries were extracted from the
Table 5. Each row contains the PL-termId, conceptId,
EMR for this study. The majority of these entries
descriptionId, relationship-typeId match-type, and
were recorded by one practitioner over a 7-year
post-coordination-sequenceId.
period. Of these entries, there were 1,713 unique PL
terms present. Based on the frequency distribution of Description Frequency
the entries, the top 10 PL terms were hypertension, No. of patients 2,894
hypercholesterolemia, diabetes mellitus, hypothyroid, Total PL entries 7,833
asthma, atrial fibrillation, gastroesophageal reflux, Total words in PL terms 16,455
depression, congestive heart failure and chronic Unique words 1,764
Longest word Hypercholesterolemia, 20 characters
kidney disease. After the second cycle we had 1,296 Median length 8 characters
(88.23%) exact-matches where the PL terms are Most common word Hypertension, 585 times
exactly the same as the SNOMED terms found. Matching Initial Cycle 2nd Cycle
There were 51 (3.47%) match-all where all the words Algorithm Frequency (%) Frequency (%)
in the PL terms are present in the SNOMED terms Exact-Match 905 (52.83%) 1,296 (88.23%)
Match-All 167 (9.75%) 52 (3.47%)
but not necessarily in the same sequence. There were
Partial-Match 633 (36.95%) 120 (8.17%)
120 (8.17%) partial-matches where one or more Semantic-Match 49 (2.86%) 20 (1.42%)
words matched the SNOMED terms. Another 20 Unmatched 8 (0.47%) 2 (0.14%)
(1.42%) SNOMED terms were found with semantic Post-coordination Not done In-progress
matches. Between the two cycles partially-matched Total unique PL terms 1,713 1,468
terms were revised to tease out qualifiers and Table 3. Summary of PL terms and matches. For
secondary concepts if present in order to explore frequency %, once a match has been found it is not
post-coordination. A summary of the PL terms and included as part of the next matching algorithm
the SNOMED matches found is shown in Table 3.
Revision of PL Terms
Characteristics of Encoded PL Terms Manual revisions were done on the 1,713 unique PL
In Table 4 we have examples of the frequently used terms after the initial cycle. By selecting the PL terms
PL terms with their SNOMED terms found by exact, that were not matched in SNOMED CT, we were
all and semantic matches. Also shown are the able to identify entries that were misspelled,
matches after revision and post-coordination of the idiosyncratic local terms or ambiguous concepts. A
original and partially-matched PL terms. For most number of spelling mistakes were corrected. The
exact-matches we selected the preferred terms from CliniClue Browser [17] was used to find matches for
SNOMED CT as they are identical or closest to the each term. A few terms were found in our problem
original PL terms, such as Atrial fibrillation. In some lists but not in SNOMED CT. Some were local terms
cases we chose the synonym terms, such as that needed to be reconsidered but there were also
Hypertension instead of the preferred term which is terms that would be submitted for inclusion in
Hypertensive disorder. For match-all and some SNOMED CT. One example is “chronic kidney
99
Representing and sharing knowledge using SNOMED
Proceedings of the 3rd international conference on Knowledge Representation in Medicine (KR-MED 2008)
R. Cornet, K.A. Spackman (Eds)
disease” which seems to be the preferred term in 3. Examine all unique words and PL terms to
common usage. Yet the closest SNOMED term is identify and revise for acronyms, abbreviations,
“chronic renal failure.” In this revision we also noted spelling variants and errors;
parts of some PL terms could be removed as 4. Match the PL terms to SNOMED concepts using
qualifiers or modifiers, thus increasing the number of the matching algorithms outlined in this paper
exact matches found. Examples include left, right, (contact authors for copies of the algorithms);
lower, midline, chronic, recurring, active, query and 5. Create detailed and summary outputs to show the
multiple. These modifiers seemed to be clustered exact, all, partial and semantic matches found;
around the concepts of time course, number, location 6. Review matched SNOMED terms for accuracy;
and severity. We found 313 such instances in our PL remove successful exact-match and match-all
terms. In another 89 instances we found post- terms from further matching cycles;
coordination of two SNOMED concepts produced a 7. Repeat steps 3 through 6 for remaining partial
good match. matches until no further matches found;
8. Post-coordinate remaining PL terms with
Navigating the SNOMED Hierarchy qualifier, refinement and combined concepts;
As part of this study, we explored ways to navigate 9. Create a pruned PL hierarchy tree showing all
the SNOMED hierarchy to determine if it could concepts with positive frequency counts and
improve one’s ability to retrieve related concepts. Of immediate super-type concepts;
the 1,296 exact matches found for the 1,468 unique 10. Create index table containing unique identifiers
PL terms present, we selected a subset of 32 PL for the PL and matched SNOMED terms.
terms related to cardiovascular disorders for this
analysis. First, we did frequency counts of these PL Implications
terms to show how often they were present in the Post-coordination is thought to be a feature that is
EMR system. For each PL term present, we difficult to implement. Yet based on the small
navigated up the hierarchy until we reached the number of SNOMED concepts used in this study to
super-type “49601007|Disorder of cardiovascular post-coordinate our PL terms, it seems feasible to
system.” We then pruned the tree to include only achieve. We did note the use of pre-coordination in
those concepts with a positive frequency count, but SNOMED CT is unpredictable, and it seems common
left their immediate super-types intact. This partly- to include acronyms within SNOMED descriptions.
instantiated cardiovascular disorder hierarchy is Careful use of modifiers such as laterality, chronicity
shown in Figure 1. The value of this tree is that it and severity should be considered. Further studies are
shows the SNOMED concepts that are actually needed.
present in the EMR and how often they occur via the
frequency counts based on the PL terms recorded. Critics often balk at the unwieldy size and
This tree can aide in the retrieval of relevant concepts complexity of SNOMED CT as too impractical for
recorded using different PL terms. For instance, by local use. In Canada the vendor and general practice
specifying the concept “56265001|Heart disease” in communities, which are often small in size, are
the query, one should expect to retrieve all sub-types reluctant to adopt SNOMED CT, questioning their
under “5754005|Acute myocardial infarction” and return on value for the effort required. From this
“12026006|Paroxysmal tachycardia.” On the other study, we have shown it is feasible to incorporate
hand, by specifying the concept “57054005|Acute SNOMED CT into EMR in the general practice
myocardial infarction” in the query, the sibling setting. The methodology we have outlined is
concept “12026006|Paroxysmal tachycardia” should practical even for small medical offices with an EMR
automatically be excluded. in place. We have also shown the potential use of
SNOMED CT to improve the quality of recall from
DISCUSSION its hierarchy. The ability to demonstrate return on
value, as in our encoding of problem lists with
SNOMED CT to improve recall, is an important first
A proposed Methodology step for practitioners to consider before full-scale
Drawing on the lessons learned from this study, we conversion of their EMR.
propose the following steps for general practitioners
to encode problem lists from their EMR in SNOMED
Limitations
as a first step for review before full-scale conversion:
There are several limitations to this study. First, the
PL terms used have been established over the years
1. Extract all PL entries from the EMR and tabulate
mainly by one practitioner from a single setting,
the frequency of the PL terms present;
which are likely to vary between practices. Second,
2. Catalogue all unique words across the PL terms;
our current matching algorithms do not take into
100
Representing and sharing knowledge using SNOMED
Proceedings of the 3rd international conference on Knowledge Representation in Medicine (KR-MED 2008)
R. Cornet, K.A. Spackman (Eds)
account subtype hierarchy to limit searches, which instance, we need to expand the use of SNOMED
could otherwise restrict unlikely choices such as terms to other parts of the EMR such as procedures,
Physical Object and Substance. Third, the evaluation medications and billing. We also need to refine our
of this methodology is incomplete to date; the full encoding methodology to take into account specific
extent of the post-coordination effort required to contexts such as past/family history and health risks,
encode the entire set of PL terms in this EMR should and to use subtype hierarchy to improve search
be further examined and reported. Fourth, the use of precision. The inclusion of frequency statistics on the
our partly instantiated hierarchy tree to improve distribution of matched SNOMED CT terms across
recall quality, while promising, requires more the hierarchies would be useful to validate the results.
thorough investigation into its utility with more These efforts should aid in the eventual creation of a
complex real-life cases. Its design should also be primary care SNOMED subset, and eventually a
aligned with the existing SNOMED navigation concept model in the primary care domain. But most
hierarchy feature that is already in place as part of the important, we should continue to exploit ways by
new RefSet release. which the use of SNOMED CT in the EMR can
actually enhance patient care.
Next Steps
We are developing a Web-based mapping tool made ACKNOWLEDGMENTS
up of the matching algorithms described earlier to Funding support for this project has been provided
allow the matching of clinical terms to SNOMED CT by the Canadian Institutes for Health Research
in an interactive or batch mode. With our focus Strategic Training Initiative.
continued to be on general practice EMR systems,
there are several steps ahead to be considered. For
Original PL Term Type of Identifier Id SNOMED Term Descn Descn
Match Type Type Status
Atrial Fibrillation Exact 49436004 C Atrial fibrillation (disorder) F 0
82343012 D Atrial fibrillation P 0
Hypertension Exact 38341003 C Hypertensive disorder, systemic arterial (disorder) F 0
1215744012 D Hypertensive disorder P 0
64176011 D Hypertension S 0
Gastroesophageal Reflux All 235595009 C Gastroesophageal reflux disease (disorder) F 0
- GERD 2535970019 D GERD – Gastro-esophageal reflux disease S 0
Cirrhosis Semantic 155809006 C Cirrhosis U 4
19943007 C Cirrhosis of liver (disorder) F 0
33568015 D Cirrhosis of liver P 0
Atrial Fibrillation Post, Exact 82343012 D Atrial fibrillation P 0
- Chronic 288524001 C Courses (qualifier value) F 0
428182017 D Courses P 0
90734009 C Chronic (qualifier value) F 0
150360019 D Chronic P 0
Chronic Kidney Disease Post 90708001 C Kidney disease (disorder) F 0
- CKD 150315015 D Kidney disease P 0
263502005 C Clinical course (attribute) F 0
391753013 D Clinical course P 0
90734009 C Chronic (qualifier value) F 0
150360019 D Chronic P 0
Headache Migraine Post, Exact 37796009 C Migraine (disorder) F 0
63055014 D Migraine P 0
246090004 C Associated finding (attribute) F 0
367802015 D Associated finding P 0
25064002 C Headache (finding) F 0
41990019 D Headache P 0
Table 4. Examples of matched PL and SNOMED terms by exact, all, semantic and post-coordinated matches.
Legend: Identifier (contains ConceptId or DescriptionID depending on Id-Type); Id Type (C- Concept, D-
Description); Descn-Type (P-preferred, S-synonym, F-fully specified name, U-undefined); Descn-Status (0-
current, 4-ambiguous); note that all selected SNOMED terms are shaded and in bold
101
Representing and sharing knowledge using SNOMED
Proceedings of the 3rd international conference on Knowledge Representation in Medicine (KR-MED 2008)
R. Cornet, K.A. Spackman (Eds)
Rec PL-Id PL-Term ConceptId DescriptionId Match AttributeId SequenceId
1 160 Atrial Fibrillation 49436004 83243013 Exact 0 0
2 789 Hypertension 38341003 64176011 Exact 0 0
3 685 Gastroesophageal Reflux - GERD 235595009 2535970019 All 0 0
4 32666 Chronic Kidney Disease CKD 90708001 150315015 Post 0 0
5 32666 Chronic Kidney Disease CKD 90734009 150360019 Post 263502005 1
6 431 Cirrhosis 19943007 33568015 Semantic 0 0
7 1044 Headache Migraine 37796009 63055014 Post, Exact 0 0
8 1044 Headache Migraine 25064002 41990019 Post, Exact 246090004 1
Table 5. Examples of the index table linking the original PL terms to matched SNOMED terms.
Legend: SequenceId indicates the relative ordering of the post-coordinated records
Two sets of post-coordinated terms shown above
49601007 Disorder of cardiovascular system (disorder) - 1
128487001 Acute disease of cardiovascular system (disorder)
127337006 Acute heart disease (disorder)
57054005 Acute myocardial infarction (disorder)
70211005 Acute myocardial infarction of anterolateral wall (disorder) - 1
73795002 Acute myocardial infarction of inferior wall (disorder) - 5
307140009 Acute non-Q wave infarction (disorder) - 5
12026006 Paroxysmal tachycardia (disorder) - 1
9904008 Congenital anomaly of cardiovascular system (disorder)
363028003 Congenital anomaly of cardiovascular structure of trunk (disorder)
13213009 Congenital heart disease (disorder) - 1
10818008 Congenital malposition of heart (disorder)
27637000 Dextrocardia (disorder) - 1
27550009 Disorder of blood vessel (disorder)
359557001 Disorder of artery (disorder)
72092001 Arteriosclerotic vascular disease (disorder)
53741008 Coronary arteriosclerosis (disorder) - 9
414024009 Disorder of coronary artery (disorder)
53741008 Coronary arteriosclerosis (disorder) - 9
55855009 Disorder of pericardium (disorder)
3238004 Pericarditis (disorder) - 2
15555002 Acute pericarditis (disorder) - 1
56265001 Heart disease (disorder) - 1
127337006 Acute heart disease (disorder)
57054005 Acute myocardial infarction (disorder)
70211005 Acute myocardial infarction of anterolateral wall (disorder) - 1
73795002 Acute myocardial infarction of inferior wall (disorder) - 5
307140009 Acute non-Q wave infarction (disorder) - 5
12026006 Paroxysmal tachycardia (disorder) - 1
PL-Id Original PL Term Concept Id Fully Specified Name
4435 Dextrocardia 27637000 Dextrocardia (disorder)
10086 Heart Disease 56265001 Heart disease (disorder)
10087 Heart Disease Congenital 13213009 Congenital heart disease (disorder)
1035 MI Inferior Myocardial Infarction 73795002 Acute myocardial infarction of inferior wall (disorder)
12653 Myocardial Infarction Anterolateral 70211005 Acute anterolateral myocardial infarction (disorder)
1591 Myocardial Infarction Subendocardial (Non Q wave) 307140009 Acute non-Q wave infarction (disorder)
1202 Pericarditis 3238004 Pericarditis (disorder)
13641 Pericarditis Acute 15555002 Acute pericarditis (disorder)
15976 Tachycardia Paroxysmal 12026006 Tachycardia paroxysmal (disorder)
Figure 1. A partial SNOMED hierarchy for cardiovascular disorders derived from a set of original PL terms. The
upper figure portion shows the partial SNOMED hierarchy for cardiovascular disorders; the lower figure
portion shows the original PL terms with the matched SNOMED concepts and their fully specified names. In
the hierarchy, concepts that are bold and italicized are exact matches for the PL terms, followed by the
frequency of how often they appeared in the EMR.
102
Representing and sharing knowledge using SNOMED
Proceedings of the 3rd international conference on Knowledge Representation in Medicine (KR-MED 2008)
R. Cornet, K.A. Spackman (Eds)
9. Wasserman H, Wang J. An applied evaluation of
REFERENCES SNOMED CT as a clinical vocabulary for the
computerized diagnosis and problem list.
1. Chute CG, Elkin PL, Fenton SH, Atkin GE. A Proceedings AMIA Symposium 2003, 699-703.
clinical terminology in the post modern era: 10. Comment: one author reviewed datasets
pragmatic problem list development. provided by colleagues from hospitals in Buenos
Proceedings AMIA Ann Symposium 1998; 795-9. Aires in Argentina, Kaiser Permanente in United
2. Warren JJ, Collins J, Sorrentino C, Campbell JR. States, and Sherbrooke in Canada during Fall
Just-in-time coding of the problem list in a 2007.
clinical environment. Proceedings AMIA Annual 11. Comment: WONCA is the World Organization
Symposium 1998; 280-4. of Family Doctors, and WICC is the WONCA
3. Petersson H, Gunnar N, Strender LE, Ahlfeldt H. International Classification Committee. URL
The connection between terms used in medical http://www.globalfamilydoctor.com/; Jan20/08.
records and coding system: a study on Swedish 12. Lee DHK. Reverse Mapping ICD-10-CA to
primary health care data. Medical Informatics SNOMED CT. UVic Master of Science research
2001; 26(2):87-99. project report, Oct 2007. Unpublished.
4. Wang SJ, Bates DW, Chueh HC, Karson AS, 13. Wang Y, Patrick J, Miller G, O’Halleran.
Maviglia SM, Greim JA, Frost JP, Kuperman Linguistic mapping of terminologies to
GJ. Automated coded ambulatory problem lists: SNOMED CT. Semantic Mining Conference on
evaluation of a vocabulary and a data entry tool. SNOMED CT Oct 2006, Copenhagen, Denmark.
International Journal of Medical Informatics 14. Kleinsorge R, Willis J, et al. UMLS Overview –
2003; 72, 17-28. Tutorial T12. AMIA Annual Symposium 2006.
5. Fabry P, Baud R, Ruch P, Despont-Gros C, http://165.112.6.70/research/umls/pdf/AMIA_T1
Lovis C. Methodology to ease the construction 2_2006_UMLS.pdf. Jan15/08.
of a terminology of problems. International 15. National Library of Medicine. The SPECIALIST
Journal of Medical Informatics 2006;75:624-32. Lexicon.
6. Meystre S, Haug PJ. Automation of a problem http://lexsr3.nlm.nih.gov/LexSysGroup/Projects/
list using natural language processing. BMC Summary/lexicon.html. Jan15/2008.
Medical Informatics and Decision Making 16. Goldsmith JA, Higgins D, Soglasnova S.
2005;5(30):1472-6947/5/30. Automatic Language-specific Stemming in
7. Elkin PL, Brown SH, Husser CS, et al. Information Retrieval. Springer-Verlag Berlin
Evaluation of the content coverage of SNOMED Heidelberg 2001.
CT: ability of SNOMED clinical terms to 17. CliniClue. The Clinical Information
represent clinical problem lists. Mayo Clinic Consultancy, Ltd., UK.
Proceedings 2006;81(6):741-8. http://www.cliniclue.com/software. available for
8. O’Halloran J, Miller GC, Britt H. Defining download. Jan22/08.
chronic conditions for primary care with ICPC-
2. Family Practice 2004;21(4):381-6.
103