MEDICAL TERMINOLOGY BROWSERS
How Usable are them for Describing Clinical Archetypes?
Jesús Cáceres Tello
1
, Miguel-Angel Sicilia
2
, Adolfo Muñoz Carrero
1
Carlos Rodriguez-Solano Nuzzi
2
and Juan-José Sicilia
3
1
Carlos III Health Institute, Telemedicine and e-Health Research Area, Madrid, Spain
2
Department of Computer Science, University of Alcalá, Alcalá de Henares, Madrid, Spain
3
Department of Internal Medicine, Henares Hospital, Coslada, Madrid, Spain
Keywords: SNOMED CT, Archetypes, Usability.
Abstract: Clinical terminologies are a major concern in medical informatics, as they are key to provide medical
systems with higher levels of interoperability. Large terminologies as SNOMED CT are gaining presence in
practical applications. In a related but different direction, archetypes or data type templates are becoming
widespread as interchange mechanisms for medical information. Archetypes support mapping to
terminologies, in a process that is typically done by the experts developing or revising the archetype. It has
been argued that terminology browsers are not appropriate for the task of helping clinical experts in the
mapping process. This paper reports usability studies on two widely used SNOMED CT browsers when
used as tools for mapping archetypes.
1 INTRODUCTION
The archetype formalism has been proposed for the
specification of models of electronic healthcare
records as a means of achieving interoperability
between systems. Archetype-based systems have
attracted an increasing attention as a relevant
technology for the interoperability of heterogeneous
systems (Wollersheim et al., 2009). However, some
sort of mapping of archetype data elements with
shared terminologies is required to guarantee a level
of common semantics across archetypes and also
with existing clinical systems. SNOMED CT
(Systematized Nomenclature of Medicine Clinical
Terms) is a large-scale comprehensive clinical
healthcare terminology that is gaining widespread
use and can be used for that binding.
The mapping process of archetypes with
SNOMED terms consists essentially on associating a
data element identified according to the formal
Archetype Description Language (ADL) to a
SNOMED element or expression.
These difficulties include the following:
Experts need to decide on the archetype elements
to be mapped; due to the hierarchical nature of
archetype expressions.
SNOMED CT represents different kinds of
entities, and often a lexical matching is not enough
to provide a correct mapping, as there is a need to
understand the hierarchy to which the term belongs
and the kind of entity that the archetype author
intended to capture.
In many cases, the concrete entities are not
directly available in SNOMED CT as concepts, but
they can be expressed through coordinated post-
coordinated expressions, which use a combination of
two or more concepts combined with qualifiers. The
use of post-coordinated expressions is justified since
SNOMED CT does not cover all potential clinical
concepts and ideas explicitly, as this would be
practically unfeasible. For example, "emergency
appendectomy" can be expressed by combining the
concepts "appendectomy" and "urgency" that are
included in SNOMED CT by the expression:
"80146002|apendiceptomía|:260870009|prioridad|=1
03391001|urgente|"
However, the language for using those expressions
requires some training beforehand for the experts to
master it.
This paper presents an initial exploratory study
to gather insights on the usability of SNOMED
browsers for the particular task of archetype
413
Cáceres Tello J., Sicilia M., Muñoz Carrero A., Rodriguez-Solano Nuzzi C. and Sicilia J..
MEDICAL TERMINOLOGY BROWSERS - How Usable are them for Describing Clinical Archetypes?.
DOI: 10.5220/0003789304130418
In Proceedings of the International Conference on Health Informatics (HEALTHINF-2012), pages 413-418
ISBN: 978-989-8425-88-1
Copyright
c
2012 SCITEPRESS (Science and Technology Publications, Lda.)
mapping. The aim of this research is obtaining
insights that can be used as input for devising
integrated SNOMED-archetype editors or browsers
for doing the mapping.
There is an increasing number of SNOMED
browsers available, either commercial or free. And
they differ significantly in the way of searching
elements and presenting the structure of
relationships inside SNOMED to users. Rogers and
Bodenreider (2008) inspected 17 different browsers
out of 23 identified and extracted a common set of
characteristics. For this study, Minnow and
CliniClue were selected due to their free availability
and the fact that they are complementary in the ways
they structure search interfaces (graphical in the first
case and text-based in the second).
The guiding exploratory hypotheses of the
present study are the following:
Are SNOMED browsers usable for supporting
the task of archetype mapping?
How are SNOMED browser users approaching
the search and navigation process for the concrete
task of searching for terms mapping particular
archetype entities?
It should be noted that the cognitive task required for
mapping an archetype term requires an
understanding of the context of the archetype
element selected, and this in turn requires
understanding the position of SNOMED terms
inside the taxonomy.
The rest of this paper is structured as follows.
Section 2 provides background information on the
archetype approach and the role of terminology
mappings in providing semantics to archetypes,
along with a review on studies related to the
usability of SNOMED or similar tools. Then, the
methods for the exploratory study are described in
Section 3. Results and discussion are provided in
Section 4 and finally, Section 5 is devoted to
conclusions and outlook.
2 BACKGROUND
The archetype approach to the interoperability of
healthcare systems is based on the notion of “two-
level” modelling (Beale, 2002).
This concept is centred in the idea of obtaining a
complete ontological separation between the
information model and the model of knowledge. On
the one hand data from standards such as UNE-EN
ISO13606 or terminology databases of different
nature as SNOMED CT or LOINC, and on the other
hand, the knowledge generated by this information
in healthcare.
Using the UNE-EN ISO13606 we can see that
the information object per excellence in the norm is
called “extract” while the knowledge is represented
by the model of archetypes (Muñoz et al., 2007).
The challenge to achieve interoperability in this
sense is able to represent all possible data structures
properly, in terms of health records are concerned. A
feature in this context is the variability and
complexity of clinical data sets, templates and
ultimately in the way of representing data. The
archetypes emerge as a possible formal definition of
possible compositions of model components for
each clinical service. In this sense, an specific
archetype or restricts the hierarchy of subclasses of a
record component within a larger structure such as
the extract defines names, optionality, cardinality,
data types and ranges and even defaults for some of
its components.
3 METHODS
This study was carried out in response to three
specific objectives:
a. Developing a usability study of each of the tools
proposed
b. Gather feedback on usability using the thinking
aloud protocol.
c. Getting feedback on the feasibility of using
SNOMED CT browsers by professionals for the task
of defining archetype mappings.
A questionnaire was used to collect satisfaction
feedback from the users together with basic data of
the participants, relating not only to their profile,
including nationality, sex or year of birth, and career
and background facts. In this context, potential users
have different profiles, e.g. primary care doctors,
hospital doctors, nurses, pharmacists or documentary
staff. Previous experience with medical terminology
was not considered, as the study is aiming at
exploring the use of the tools for people not having a
specific training on medical terminologies.
The study evaluated the usability of the two
applications by the participants. Usability is an
important aspect of any software. Factors such as
ease of installation or further learning and use are
very important indicators when assessing the quality
of a particular software application. In this sense we
analysed different types of questionnaires, including
SUS (Brooke, 1996), QUIS (Harper and Norman,
1993), CSUQ (JR Lewis, 1995) and SUMI (van
HEALTHINF 2012 - International Conference on Health Informatics
414
Veenendaal, 1998). Finally, it was decided to
conduct the study using the SUMI questionnaire
(Software Usability Measurement Inventory). This
type of questionnaires was developed by the Human
Factors Research Group (HFRG), University
College Cork, in 1986 and allow complementing the
evaluation of usability with the perceptions and
attitudes of users, resulting in reliable indicators in
five key areas (Kirakowski and Corbett, 1993),
namely Efficiency, Affect, Helpfulness, Control and
Learnability.
The experience study consisted of the search of
SNOMED codes for the users to identify both the
archetypes and also the data fields contained in
them. In this sense, a first methodological issue was
that of the source of the archetypes. The CKM
repository maintained by the OpenEHR foundation
was used, as it is the more mature platform and it is
also open and not restricted to some particular
institution’s viewpoint. Another reason for choosing
the repository maintained by the openEHR was that
the archetype language of OpenEHR includes five
kinds of entities that come from ontological analysis
(Beale and Heard, 2007), namely Observation,
Evaluation, Instruction, Action and Administrative.
They are representing different types of care entries,
and may be hypothesized to bring different
requirements for the mappings.
In order to make the test short for a single
session, it was decided to choose only two of the
aforementioned archetypes kinds. With this premise,
the selected were “observation” (concretely, the
indirect_oximetry.v1 archetype) and “action”
(concretely, the medication.v1 archetype)
archetypes, considered as the most important types
of archetypes to carry out the study. To increase the
reliability of the study, different combinations of
archetypes and software used were used for the
different users as shown in Table 1.
These combinations covered all possible options
to prevent order biases in the results. The sequence
began again with user number nine.
In addition to the SUMI questionnaire and the
recording of the sessions using screen recording
software, a "thinking aloud" protocol (Lewis, 1982)
was used during the execution of a task (concurrent)
following the indications by (Ericsson and Simon,
1993).
The study was limited to free browsers having at
least the core SNOMED CT data browsing features
identified by Rogers and Bodenreider (2008). After
a review of existing free ones, the selection resulted
in CliniClue and Minnow.
Table 1: Distribution of tasks among the participants in the
study.
Users
Archetypes Software
Observ. Action CliniClue Minnow
1
x x
x x
2
x x
x x
3
x x
x x
4
x x
x x
5
x x
x x
6
x x
x x
7
x x
x x
8
x x
x x
CliniClue browser was found by Elhanan et al.
(2010) the most popular tool used to access SCT
(54%, exceeds 100%), as well as the most preferred
one (29%). In contrast, Minnow is a relatively new
browser. Minnow was selected due to their wide
differences in response time (as they are using
Lucene indexes inside) and the use of more agile
hierarchical browsing representations using
visualizations of hierarchical trees.
Participants in the user test were asked to
verbalize their intentions while performing the
mapping tasks, while being observed by a
researcher. Prompts were only provided if the user
was unable to proceed. This was in combination
with screen recording for subsequent analysis of the
actions taken and the notes of the researcher. Users
are asked to formulate any question or comments for
discussion with the researcher after the event.
Each task was presented as a task of assigning
codes SNOMED CT as the name itself of the above
archetypes as well as the data fields they contain. In
this case, the breakdown in subtasks was user
initiated, as they were given freedom to decide from
where to start and what archetype elements to map.
The users were exposed directly to the user interface
of the browser, and let alone to guess what features
they should use (the search for concepts, for
example). The openEHR CKM was not showed or
explained, and the concept of archetype was
simplified and users were not exposed to the
complexities of the ADL formal language.
Concretely, the archetypes were explained using
plain text without any reference to concrete
standards, languages or mappings. This avoided
confusions or difficulties that are external to the
central task of finding SNOMED concepts that map
MEDICAL TERMINOLOGY BROWSERS - How Usable are them for Describing Clinical Archetypes?
415
to archetypes or their components.
Finally it was decided not to ask for post-
coordinated expressions as a possible solution in the
SNOMED CT code assignment, as this would
require prior training and requires some knowledge
of the SNOMED formal language. Instead, it was
asked to users to informally express the
identification of an archetype or part of it as a
combination of SNOMED CT terms rather than a
single term.
4 RESULTS AND DISCUSSION
A room with several computers with the two
browsers installed (CliniClue and Minnow) was
used for the text. In addition, each team had an
Internet connection to allow the user to each of the
two online surveys, the part of usability by entering
the personal and professional data and the SUMI
questionnaire of 50 questions for further automatic
processing. Each participant was accompanied by an
observer to collect interesting data using the "Think
Aloud" protocol.
4.1 Participants
Users who participated in this study were all active
professionals coming from different health
institutions including the University Hospital of
Fuenlabrada, Clinic Hospital San Carlos, 12 de
Octubre Hospital and Henares Hospital in Coslada,
all of them in the Madrid region.
The study involved a total of 14 participants (8
women and 6 men). As the cost-benefit ratio is
considered lower applying the tests 3 to 5 users
(Nielsen, 1993), no additional users were considered
for this initial exploratory study.
The ages of the participants ranged from 34 to 52
years the average age being 40 years, all of Spanish
nationality.
As for the experience of participants in the health
context, the average was 15 years and his previous
experience with medical terminologies did not reach
33%.
As for the time spent in completing this test, it
varied according to the previous experience of
participants with clinical terminologies as well as
browsers of this type. In mid-range, it was situated at
72 minutes in the realization of global test.
Interestingly, in this case it was that less time spent
in conducting the test with the browser Minnow than
the browser CliniClue.
In all cases, participants expressed the
importance of the linguistic barriers (the browsers
are in English) that they found on search the
concepts in other language than their own. While
some had experience with clinical terminologies,
they had always worked with the Spanish version of
SNOMED CT.
4.2 Usability of the Browsers
Regarding the usability of the browser, after
collecting data from users in the SUMI questionnaire
on each browser, results were compiled and
descriptive statistics applied.
In the case of CliniClue, the highest levels were
obtained in the Efficiency subscale, while the lowest
corresponded to the subscales of Helpfulness and
Control, i.e., users perceived a lack of friendliness in
the search interface and of control in the search
process, instead, assessed the performance aspects of
the program as well as the level of satisfaction in
their management (affected).
Figure 1: Quantitative measurements of the usability of the
CliniClue browser.
In the case of Minnow, the levels are clearly
higher than CliniClue. The highest levels
corresponded to the subscales of friendliness and
satisfaction in their use. Participants generally
viewed the multimedia component of this browser
and the arrangement of windows in the interface
very useful and explanatory. Lower levels in this
case corresponded to the lack of control and
learnability.
Figure 2: Quantitative measurements of the usability of the
Minnow browser.
The following graph shows an overview of the
HEALTHINF 2012 - International Conference on Health Informatics
416
results. It is clear that participants in this study
clearly preferred Minnow.
Figure 3: Comparative results of measurements obtained
in the usability study.
As for the protocol "Think aloud" some
conclusions could be considered quite significant. It
is important to note first that The indicators explored
were:
Complexity of the test.
Comments on the proposed archetype and all the
data it contains.
Observations on the coding of these data.
Regarding the test itself, the perception of most
participants (84%) was that it had not much
complexity, also they agree in stating that they had
found the test very complete. Regarding the duration
of the same they all considered too long, answering
the question asked at the end of the test by the
observer.
In regard to the proposed archetypes, most of the
comments received were related to the inability to
find some of the proposed concepts. In some cases
this lack of results in the searches occurred in both
browsers, at other times produced results in only one
of them. The following chart shows the total number
of concepts located in both a browser and in another
by all members of the study.
Was also seen that the encoding proposed by the
users varied markedly. The level of agreement did
not reach 10% for participants in the study. Possible
causes are the difference in the search engines of the
two software or differences in the visualization of
concepts as which they could infer dramatically in
the final choice of concept. A source of differences
found was concepts that are related lexically but
refer to different kinds of entities as defined in
SNOMED CT high level categories.
Although in all cases there was this mismatch of
proposed coding, all participants agreed in their
positive attitude towards the test performed and their
willingness to use of terminology standards as a
necessary step to achieve semantic interoperability
of medical systems.
5 CONCLUSIONS
A usability evaluation of the browsers tested is
plausible and appears as a relevant technique to
gather more information in devising SNOMED CT
browsers. This study provides a better understanding
of the usefulness of SNOMED CT terminology
standard and their use in identifying the concepts of
archetypes. In this sense it has been shown that
using SNOMED CT codes for the identification of
the concepts that make up a openEHR archetype is
feasible, but entails some difficulties from the side
of the user that deserve further attention if reliable
and consistent coding is sought.
This study has also proven the effectiveness of
the protocol "think aloud" in the participant's
registration information with the figure of an
observer.
Future work will extend the exploratory usability
studies to a wider range of SNOMED CT browsers,
and to search and browsing tasks following a finer
grained approach in task descriptions.
Anyway this study is certainly relevant to
government agencies and institutions as well as
companies that are committed to interoperability of
medical systems using teminologies standards such
as SNOMED CT with clinical standards UNE/EN
ISO13606.
ACKNOWLEDGEMENTS
This work has been supported by the project
“Métodos y Técnicas para la Integración Semántica
de Información Clínica mediante Arquetipos y
Terminologías (METISAT)”, code CCG10-
UAH/TIC-5922, funded by the Community of
Madrid and the University of Alcalá.
Likewise, this research work has been partially
supported by projects PI09-90110 - PITES -
from "Fondo de Investigación Sanitaria (FIS) Plan
Nacional de I+D+i", "Plataforma en Red para el
Desarrollo de la Telemedicina en España" and FIS
08-1148 "Representación de información clínica en
cáncer de mama mediante arquetipos".
REFERENCES
Arh T. Blazic B. J., (2008). A Case Study of Usability
Testing - The SUMI Evaluation Approach of the
EducaNext Portal. In Wseas Transactions On
Information Science & Applications. Issue 2, Volume
5, p. 175 – 181. ISSN: 1790-0832.
MEDICAL TERMINOLOGY BROWSERS - How Usable are them for Describing Clinical Archetypes?
417
Balbo, S., Steinkrug, J., Lee, L., Scheidt, S., Hullin, C. and
Gogler, Janetter, (2008). The Importance of Including
Users in Clinical Software Evaluation: What Usability
Can Offer in Home Monitoring. In: Grain, Heather
(Editor). HIC 2008 Conference: Australia's Health
Informatics Conference, August 31 - September 2,
2008 Melbourne Convention Centre
Beale T., (2002). Archetypes: Constraint-based Domain
Models for Future-proof Information Systems
Beale, T. and Heard, S., (2007). Anontology-based model
of clinical information. MEDINFO 2007: Proceedings
of the 12th World Congress on Health (Medical)
Informatics: Building Sustainable Health Systems:
760-764.
Borovicka M., Laschitz L. et al., (2010). Chaos or
Content: Towards Generic Structuring of Medical
Documents by Utilisation of Archetype-Classes, In:
Schreier G, Hayn D, Ammenwerth E. (Eds.):
eHealth2010 - Health Informatics meets eHealth:
Tagungsband eHealth2010 & eHealth Benchmarking
2010, in OCG Books Band 264. Wien: OCG. S. 47-54.
Duo Wei, B. S., Yue Wang, Yehoshua Perl, Junchuan Xu,
Michael Halper, and Kent A. Spackman. Complexity
Measures to Track the Evolution of a SNOMED
Hierarchy. Proceedings of the American Medical
Informatics Association Symposium, 2008: 778-782
Elhanan, G., Perl, Y. and Geller, J., (2010) A Survey of
Direct Users and Uses of SNOMED CT: 2010 Status
AMIA AnnuSympProc. 2010; 2010: 207–211.
Ericsson, K. and Simon, H. A., (1980). Verbal reports as
data. Psychological Review, 87, 215-251
Ericsson, K and Simon, H. A. (1993). Protocol Analysis.
Cambridge: The MIT Press.
Gilb T. (1987) Principals of Software Engineering
Management, Addison-Wesley, Reading, Mass.
Harper, B. D. and Norman, K. L., (1993). Improving User
Satisfaction: The Questionnaire for User Interaction
Satisfaction Version 5.5. Proceedings of the 1st
Annual Mid-Atlantic Human Factors Conference, (pp.
224-228), Virginia Beach, VA.
Jaspers, M. W., (2009). Comparison of usability methods
for testing interactive health technologies:
Methodological aspects and empirical evidence,
International Journal of Medical Informatics, 78(5),
pp. 340-353
Kirakowski, J., Corbett, M., (1993). SUMI: The Software
Usability Measurement Inventory. British Journal of
Educational Technology, 24(3), pp. 210-212.
Lewis, C., (1982). Using the thinking-aloud method in
cognitive interface design. IBM Research Report RC
9265, Yorktown Heights, NY, 1982.
Lewis, J. R., (1995): IBM Computer Usability Satisfaction
Questionnaires: Psychometric Evaluation and
Instructions for Use. International Journal of Human
Computer Interaction. 7, 57-78.
Muñoz, A., Somolinos, R., Pascual, M., Fragua, J.,
González, M., Monteagudo J. L., Hernández, C.,
(2007). Proof-of -concept Design and Development of
an EN13606-based Electronic Health Care Record
Service. Journal American Medical Information
Association (2007;14:118 –129. DOI 10.1197/jamia.
M2058.)
Nielsen J., (1993). Usability Engineering. AP
Professional. Boston (MA), USA,
Nielsen, J and Levy, J (1994). Measuring usability:
preference vs. performance. Communications of the
ACM, Vol. 37, No. 4. (April 1994), pp. 66-
75. doi:10.1145/175276.175282 Key: citeulike:
296786
Nilsson, O. y Olsen, J., (1995). Measuring Consumer
Retail Store LEALTADalty. European Advances in
Consumer Research, 2, pp. 289-297.
Rogers, J. and Bodenreider, O., (2008) SNOMED CT:
Browsing the Browsers. Proceedings of the 3rd
international conference on Knowledge
Shackel, B., (1991). Usability – context, framework,
design and evaluation. In Shackel, B. and Richardson,
S. (eds.). Human Factors for Informatics Usability.
Cambridge University Press, Cambridge, 21-38.
Tullis, T. S. and Stetson, J. N., (2004, June 7-11). A
Comparison of Questionnaires for Assessing Website
Usability, Usability Professionals Association (UPA)
2004 Conference, Minneapolis, USA.
Van Veenendaal, E., (1998). Questionnaire based usability
testing. In Proceeding of European Software Quality
Week, Brussels.
Wollersheim D., Sari A., Rahayu W., (2009). Archetype-
based electronic health records: a literature review and
evaluation of their applicability to health data
interoperability and access. HIM J 38: 7–17.
HEALTHINF 2012 - International Conference on Health Informatics
418