A Model-Driven Approach to Create and Maintain an Executable
Transferal Management Platform
Emanuele Laurenzi
Institute for Information & Process Management, University of Applied Sciences, St. Gallen, Switzerland
1 STAGE OF THE RESEARCH
My work falls within the eHealth application domain
and it is embedded into the just started research
project Patient Radar. The project is funded by the
Swiss Commission for Technology and Innovation
(CTI) under the lead of the Institute for Information
and Process Management (IPM-FHS) - University of
Applied Sciences St. Gallen. Two regional hospitals
and a Swiss national technology provider are
involved in the project in which they contribute with
their expertise.
The Patient Radar project wants to facilitate
intersectoral collaboration within the inpatient
sector, also called “transferal management”, i.e.
between acute hospitals and rehabilitation clinics in
Switzerland.
My research aims at supporting and optimizing
such a collaboration by setting up a framework
which adopts a model-driven approach to enable the
creation and maintenance of a transferal
management platform. The model-driven approach
makes the platform highly configurable to
accommodate new clinical pathways and be easily
extendable to include additional functions to meet
future needs. All domain-specific aspects are
described declaratively in application models.
Hence, domain experts will be able to
create/use/manage application models with no
required programming skills. To provide an
executable platform, models are first specified in the
formal semantics description logics and then their
elements are mapped to corresponding elements in
an application framework. In this way, we will
ensure that executable code can be derived from all
application models. Additionally, the transferal
management platform includes reference models
from which domain experts can easily create and
adapt application models.
Research questions are established and along my
work, these will be refined more and more.
2 OUTLINE OF OBJECTIVES
Developing a transferal management platform
includes the challenge of combining the clinical
pathways of several hospitals and several
rehabilitation clinics into one coherent treatment
process. The project intends to address this
challenge by creating a transferal management
platform that will serve as a hub for post-acute care,
which enables and supports the interaction between
acute hospitals and rehabilitation clinics as well as
further actors such as nursing facilities, family
doctors and health insurances for granting cost
reimbursement. While a patient is still in the acute
hospital, information such as concerning the
patient’s health status, required rehabilitation
treatment, planned transfer date need to be available
on the platform so that rehabilitation clinics can
apply for those patients and plan ahead their
resources and determine early on therapies which are
specifically adapted to the expected patient. In
particular, the transferal management platform
should
know about clinical pathway (i.e. the patient
treatment process) of acute hospitals and partly
of the rehabilitation clinics;
monitor the progress of a patient along the
clinical pathway according to predefined
medical indicators;
give the rehabilitation clinics access to the
progress of the patients as well as further
rehabilitation-specific parameters such as age,
weight, and diagnosis; all other personal data
are anonymized until the patient is transferred.
The transferal management platform will have to
support wide variety of clinical pathways for all
kinds of acute hospitals and rehabilitation clinics.
Therefore, the platform must be highly configurable
to accommodate new pathways, and it must permit
that a given pathway slightly differs between
hospitals. Moreover, the platform should be easily
extendable to include additional functions to meet
future needs. As a consequence, we decided to adopt
11
Laurenzi E..
A Model-Driven Approach to Create and Maintain an Executable Transferal Management Platform.
Copyright
c
2014 SCITEPRESS (Science and Technology Publications, Lda.)
a model-driven approach where all domain-
specific aspects are described declaratively in an
application model. The elements in the application
model will be mapped to corresponding elements in
an application framework to obtain the executable
transferal management platform.
In order to guide and support building and
maintaining the application model we introduce
reference models that provide blueprints for
common application scenarios, such as typical
pathways with the necessary patient data, generic
models of the associated administration processes as
well as role models for typical users of the platform
with their rights.
While the use of reference models helps create
specific application models it is very difficult to
constrain the adaptations to a reference model so
that only models are created that can be mapped to
the underlying application framework, i.e. can be
made executable. Since our primary purpose is to
obtain executable platforms from the application
models losing executability must be avoided.
We therefore decided to ground the reference
model in a domain-specific language (DSL)
(Jouault F. und Bézivin 2006) (Mernik et al. 2005)
(Ranabahu et al. 2012) (van Deursen et al. 2000)
instead of a general-purpose modelling language
such as UML or BPMN. The reference models and
all application models are created using
representation constructs from the DSL. Since the
definitions of the mapping from model elements to
elements of the application framework are tied to the
DSL contructs any model expressed in the DSL can
be mapped to the framework and thus be made
executable.
3 RESEARCH PROBLEM
As outlined in the previous two sub-sections we
adopt a model-driven approach where the domain-
specific aspects of the transferal management
platform are specified by an application model from
which code fragments for the actual transferal
management platform are generated.
Without the framework given by both, the
reference model and application model, the
transferal management platform would just consist
of an ordinary model that can be modified in any
possible way, including desired ways as well as
undesired ones. In particular, our framework should
enable domain experts to adapt and extend the
platform to future needs without the help of
computer specialists; this is done by adapting
the platform through modelling instead of
programming;
ensure that changes to the model do not destroy
the code generation property.
The following three sub-sections describe our
proposed approach. The research questions are
defined subsequently.
3.1 Basic Approach
We embed our approach into a meta modelling
framework (Karagiannis und Kühn) (Laarman und
Kurtev 2010) where the application models reside on
level 1 and the DSL is defined as a meta model on
level 2, thus providing the modelling language in
which the level 1 models are represented (cf. Figure
1).
The reference models also reside on level 1 and
serve as blueprints for the application models. An
application model is derived by modifying and
adapting a reference model but not by instantiating
it. That is why reference models and application
models are on the same level. Level 3 provides the
language constructs needed to define the DSL.
By using the two-tier approach shown in Figure
1 we aim for two kinds of extensibility:
On level 1 new application models are created
and existing ones adapted to suit new needs,
e.g. to accommodate completely new or
slightly adapted clinical pathways. This is the
task of a domain expert who is guided by the
reference models and uses the domain-specific
constructs provided by the DSL. We especially
aim at enabling domain experts for this task
who do not have to be modelling experts by
providing a user interface which translates
modelling into an interactive visual paradigm.
New requirements that cannot be met by
adapting an existing model on level 1 are taken
care of on level 2 by extending the DSL to
provide the additional expressiveness needed
on level 1. This is the task of a modelling
expert.
Executability of the models on level 1 will be
provided by mappings from the DSL constructs to
corresponding elements in the application
framework. Via this approach we will ensure that all
models on level 1 can in fact be properly mapped to
the application framework. An alternative approach
without a DSL would have used a general purpose
language, such as UML, and extended or refined its
semantics by adding constraints that restrict the
range of possible changes to the model (Lodderstedt
et al. 2002)
IC3K2014-DoctoralConsortium
12
Figure 1: Our approach in a meta modelling framework.
Such an approach, however, has the disadvantage
that it increases the complexity of the modellers’
task because they have then to know all these
constraints and take them into account, or they are
hindered by the system to make certain
modifications to a model when they would violate a
constraint. Our approach using a DSL promises to be
much easier for the modellers.
3.2 The Approach in Detail
A first idea for designing the DSL might be to use a
traditional process modelling language as its basis
since a description of the clinical pathways are at the
center of the transferal management platform. In our
case, however, we do not need to describe the
pathways in detail but only certain aspects of them,
mainly from a healing progress point of view and
not from a medical treatment perspective. The DSL
must allow to represent what the main healing and
rehabilitation phases are and which conditions must
be met (called gateways) to get to the next phase.
For this we do not need a full-fledged process
modelling language. Instead, our DSL only includes
some basic process modelling elements, thus
reducing model complexity, increasing modelling
productivity (Ulrich F. 2010) and enabling a simpler
mapping to executable elements in an application
framework.
The backbone of our DSL is the domain-specific
terminology which includes the object types relevant
for our domain as meta classes. Currently we have
the following meta classes: “ClinicalPathway” for
the patient treatment process, which consists of
activities. An activity leads to gateways that stand
for certain conditions being tested. Gateways with
fulfilled conditions enable further activities (cf.
Figure 2). Activities are associated with patients and
can be specialized into various subclasses such as
treatment activities and administrative activities.
Figure 2: Part of the domain-specific language.
All the classes used in an application model are
instances of a meta class in the DSL. We have
already created an application model for knee
replacement surgery. It contains a sequence of
activities with gateways between two activities.
Most important are the gateways since they are the
indicators how soon a patient can be released to a
rehabilitation clinic. Each activity has two duration
attributes: “typical duration” says how long the
activity usually lasts, “actual duration” says how
long the activity has already lasted. From these
attributes the transferal management platform can
derive the expected transferal date. Each gateway
has an attribute which says if the condition the
gateway stands for has been fulfilled, not been
fulfilled or not yet been tested.
Besides their duration the activities themselves
are often not of interest and are then represented by
the class “SomePostSurgeryTreatment”. In cases of
complications a second surgery or even a stay in an
intensive care unit might become necessary. When
covering such cases in the model it might be
necessary to represent these treatment activities
explicitly so that a rehabilitation clinic can see the
reason why a patient will be released later than
originally expected.
Reference models are on the same level in the
meta-model hierarchy as application models (cf.
Figure 1) so that their classes are instances of meta
classes of the DSL as well. The difference to an
application model is that a reference model is more
abstract and thus can serve as a blueprint for many
application models. We have also created an
example of a reference model for knee surgery,
which is more abstract than the application model
(above described) and only includes the absolute
necessary activities and gateways. There might also
be a reference model for surgery, which would be
even more generic. In fact, many different reference
models might be created over time, some of them
MetaMeta Model
MetaModel
ReferenceModels
Application Models
instanceof
instanceof
Domain‐Specific
Language
defines
expressed
in
Level3
Level2
Level1
AModel-DrivenApproachtoCreateandMaintainanExecutableTransferalManagementPlatform
13
specializing or generalizing already existing ones.
Thus, for example, even an application model might
be regarded as a reference model, which would then
still have to be refined to reflect the specific
procedures of different hospitals. Summarising, if a
model is an application model or a reference model
depends on its degree of generality and on the
purpose it needs to fulfil and is thus much a matter
of viewpoint.
3.3 Mapping Application Models to an
Application Framework
The application models are not just created to
describe an application domain but primarily to be
translated into some specific behaviour in the Patient
Radar transferal management platform. To achieve
this, application models will be mapped to
corresponding elements of the application
framework that is the basis for the transferal
management platform (cf. the similar approach in
(Reimer et al. 2008)). In our project we use the
application framework Vivates by the Swiss Post
(http://www.vivates.ch/). The mapping rules will be
defined in the DSL where they are attached to each
meta class.
For example, certain classes such as
“SurgeryPatient” will be mapped to corresponding
object types in Vivates. The objects belonging to
these object types reside in the runtime environment
of the application framework and are conceptually
instances of the corresponding classes in the
application model.
Other classes, such as “Gateway” will be mapped
to states in Vivates. A sequence of gateways
(through intermediate activities) will be mapped to
rules which reflect that order by specifying which
gateways have to be reached before another gateway
can be reached. In the case of parallel gateways a
rule is generated that requires both gateways to be
fulfilled in order to enable the subsequent activity.
The duration attributes of each activity class will
be used in a web service of Vivates to compute the
expected time until transferal to the rehabilitation
clinics is possible and makes this value available on
the transferal management platform.
3.4 Research Questions
The development of reference/application models by
adopting the model-driven approach pose the
following research questions:
How can DSL constructs be mapped to
corresponding elements of the application
framework to enable and preserve executability
in the platform?
How much variability can the application
model and/or the reference model allow
without losing its executability property for the
platform?
How can an application model and/or a
reference model take advantage from a
domain-specific modelling language?
How do reference models and application
models differentiate each other to enhance the
domain expert understanding of managing the
work space platform environment?
How can the employment of a description logic
with formal semantics and reasoning
capabilities enhance the definition of a DSL in
a meta modelling approach?
How and in which level of our proposed meta
modelling framework (cf. Figure 1) does the
integration of the description logics take place
in order to assure the dynamic behaviour of the
transferal management platform?
These questions are intended to be answered
within our application domain. In this way, insights
gained in the application domain will be valuable
and will be a contribution to a general understanding
of how to make use of application models, reference
models, DSLs, description logics and model-driven
development.
4 STATE OF THE ART
There are various approaches that combine a DSL
with a model-driven approach. For example, (Nunes
und Schwabe 2006) describes their HyperDe system
as an environment to support the rapid prototyping
of Web applications by combining model-driven
development with the use of DSLs. This
combination allows the developer to create code by
manipulating models that specify the application.
HyperDe environment supports designing Web
application through a meta model instantiation.
Furthermore, HyperDe extends the Ruby on Rails
framework into a DSL, allowing direct manipulation
within Ruby scripts of both the model and the meta
model.
Similarly, (Cadavid et al. 2009) introduces a
DSL into a model-driven software development
process to reduce the complexity of Web
applications development. The work describes the
elements used in the process of transforming a UML
domain model into a deployable Web application. In
this way they demonstrate that models can be
IC3K2014-DoctoralConsortium
14
transformed and executed for the automatic
generation of applications.
Different to existing approaches we not only
employ a DSL but also include reference models in
our approach to provide modelling guidance. Thus
the reference models will make it easier for domain
experts to create and adapt application models
because they do not have to start from scratch but
only need to adapt already existing meaningful
model fragments from the reference models.
Among the several meta modelling frameworks
available today (Kern et al. 2011), we adopt the one
provided by ADOxx. ADOxx features a three-step
modelling hierarchy with a rich meta-meta model
(Kern 2008), i.e. the ADOxx hierarchy (Karagiannis
und Visic 2011). This hierarchical approach suits
our approach of first defining a DSL and then a
reference model which is created using the DSL.
Additionally, in contrast to most other
approaches, which are e.g. based on UML, we will
employ a description logic for defining the DSL.
We will utilize the model-theoretic semantics of the
description logic to obtain a sound semantic
foundation of the DSL. This will allow for
terminological inferences and integrity control. The
semantics will specify in which ways the language
constructs can be combined and will enable the
modelling tools to support the users in creating
consistent and meaningful models, e.g. by
prohibiting inconsistent combination of constructs
(see e.g. (Staab et al. 2010)).
The application models are not just created to
describe an application domain but primarily to be
translated into some specific behaviour in the Patient
Radar transferal management platform. To achieve
this, application models are mapped to
corresponding elements of the application
framework that is the basis for the transferal
management platform (cf. the similar approach in
(Reimer et al. 2008). In our project we use the
application framework Vivates by the Swiss Post
(http://www.vivates.ch/). The mapping rules are
defined in the DSL where they are attached to each
meta class.
5 METHODOLOGY
Since my research work is about developing a novel
kind of artefact, the design science research
methodology has been adopted to successfully carry
out this dissertation. The design science research
allows building a sound knowledge base through
cycles of artefacts construction and subsequent
evaluation. In particular, my work will be based on
the general design cycle (GDC) (cf. Figure 3)
elaborated by (Vaishnavi and Kuechler 2007) whom
applied it into design science research.
Figure 3: Reasoning in the general design (GDC) adapted
from (Vaishnavi and Kuechler 2007).
The cycle depicted in Figure 3 shows that the
design begins with Awareness of Problem. In it, the
problem is identified and defined. More specifically,
a case study will be created and approved among the
participant of the Patient Radar project. The research
strategy “case study” thus provides a concrete
contribution to better understand the application
domain, i.e.
Which terminology needs to be developed
within the application domain
To which general degree a reference model and
an application model need to be designed,
Which elements of an application model need
to be mapped to corresponding elements of the
give application framework.
Techniques to be used in creating the case study
are literature review, expert interviews involved in
the Patient Radar project such as Physician, Nurses,
eHealth professionals.
Next, in the Suggestion phase, a problem
solution is abductively developed and a tentative
design is given. Collected data in the Awareness of
Problem together with further literature review will
help elaborate the conceptual model related to the
domain-specific terminology, reference model and
application models. Additionally, the mapping rules
in the DSL where they are attached to each meta
class will be defined.
Then in the Development phase, the design is
further refined and an actual artefact is produced
through many iterations.
I will implement the conceptual model by using
the ADOxx Modelling Toolkit Platform. Hence,
AModel-DrivenApproachtoCreateandMaintainanExecutableTransferalManagementPlatform
15
guidelines and best practices implemented in
ADOxx will be taken into consideration. Moreover,
further literature will be sought aimed at
implementing the mapping rule which then allow to
establish and maintain executability of the transferal
management platform.
Subsequently, we reach the Evaluation phase.
Here the artefact is evaluated by using the well-
known Technology Acceptance Model (TAM)
which provides a valid and reliable measure to
predict acceptance or adoption of new technologies
by end users (Davis et al. 1989; Davis 1989).
Additionally, TAM is a commonly used model to
measure technology acceptance (King and He 2006).
Finally, the Conclusion phase, in which insights
gained from the work are reported and future
research will be addressed.
6 EXPECTED OUTCOME
The work aims at adopting a model-driven approach
to create and maintain a transferal management
platform for supporting the collaboration between
acute hospitals and rehabilitation clinics to optimize
transferal management. Thus, the expected outcome
is a framework which allows the development of the
transferal management platform being highly
configurable, e.g. to accommodate new clinical
pathways, and to permit a given pathway to
(slightly) differ between hospitals. Moreover, the
platform should be easily extendable to include
additional functions to meet future needs.
All domain-specific aspects will be described
declaratively in a reference or application model.
The elements in the application model will be
mapped to corresponding elements in an application
framework to obtain the executable platform. The
mapping will be enabled by mapping rules that are
defined on the constructs of a domain-specific
language (DSL). In this way, it can be assured that
executable code derived from all
reference/application models expressed with the
DSL. The mappings from DSL constructs to
application framework elements will be specified
using the description logic semantics to enable the
modelling tools supporting the users in creating
consistent and meaningful models.
REFERENCES
Cadavid, J. J.; Quintero, J. B.; Lopez, D. E.; Hincapié, J.
A. (2009): A Domain Specific Language to Generate
Web Applications. In: Antonio Brogi, João Araújo und
Raquel Anaya (Hg.): Memorias de la XII Conferencia
Iberoamericana de Software Engineering (CIbSE
2009), Medell\’ın, Colombia, Abril 13-17, 2009, S.
139–144.
Davis, F. D. (1989): Perceived Usefulness, Perceived Ease
of Use, and User Acceptance of Information
Technology. In: MIS Q 13 (3), S. 319–340. Online
verfügbar unter http://dx.doi.org/10.2307/249008.
Davis, F. D.; Bagozzi, R. P.; Warshaw, P. R. (1989): User
Acceptance of Computer Technology: A Comparison
of Two Theoretical Models. In: Management Science
35 (8), S. 982–1003. Online verfügbar unter
http://EconPapers.repec.org/RePEc:inm:ormnsc:v:35:y
:1989:i:8:p:982-1003.
Jouault F.; Bézivin, J. (2006): Km3: a dsl for metamodel
specification. In: In proc. of 8th FMOODS, LNCS
4037: Springer, S. 171–185.
Karagiannis, D.; Kühn, H.: Metamodelling Platforms. In:
In Proceedings of the 3rd International Conference
EC-Web 2002 – Dexa 2002, Aix-en-Provence, France,
2002, LNCS 2455: Springer-Verlag, S. 182.
Karagiannis, D.; Visic, N. (2011): Next Generation of
Modelling Platforms. In: Janis Grabis und Marite
Kirikova (Hg.): Perspectives in Business Informatics
Research, Bd. 90: Springer Berlin Heidelberg (Lecture
Notes in Business Information Processing), S. 19–28.
Online verfügbar unter http://dx.doi.org/10.1007/978-
3-642-24511-4_2.
Kern, H. (2008): The Interchange of (Meta)Models
between MetaEdit+ and Eclipse EMF Using M3-
Level-Based Bridges. In: Jeff Gray, Jonathan Sprinkle,
Juha-Pekka Tolvanen und Matti Rossi (Hg.): 8th
OOPSLA Workshop on Domain-Specific Modeling at
OOPSLA 2008: University of Alabama at
Birmingham, S. 14–19.
Kern, H.; Hummel, A.; Kühne, S. (2011): Towards a
Comparative Analysis of Meta-Metamodels. In:
Proceedings of 11th Workshop on Domain-Specific
Modeling (DSM’11). Online verfügbar unter
http://www.dsmforum.org/events/DSM11/Papers/kern.
pdf.
King, W. R.; He, J. (2006): A Meta-analysis of the
Technology Acceptance Model. In: Inf. Manage. 43
(6), S. 740–755. Online verfügbar unter
http://dx.doi.org/10.1016/j.im.2006.05.003.
Laarman, A.; Kurtev, I. (2010): Ontological
Metamodeling with Explicit Instantiation. In: Mark
Brand, Dragan Gašević und Jeff Gray (Hg.): Software
Language Engineering, Bd. 5969: Springer Berlin
Heidelberg (Lecture Notes in Computer Science), S.
174–183. Online verfügbar unter http://dx.doi.org/
10.1007/978-3-642-12107-4_14.
Lodderstedt, T.; Basin, D.; Doser, J. (2002): SecureUML:
A UML-Based Modeling Language for Model-Driven
Security. In: Springer, S. 426–441.
Mernik, M.; Heering, J.; Sloane, A. M. (2005): When and
How to Develop Domain-specific Languages. In:
ACM Comput. Surv. 37 (4), S. 316–344. Online
IC3K2014-DoctoralConsortium
16
verfügbar unter http://doi.acm.org/10.1145/1118890.
1118892.
Nunes, D. A.; Schwabe, D. (2006): Rapid Prototyping of
Web Applications Combining Domain Specific
Languages and Model Driven Design. In: Proceedings
of the 6th International Conference on Web
Engineering. New York, NY, USA: ACM (ICWE
’06), S. 153–160. Online verfügbar unter
http://doi.acm.org/10.1145/1145581.1145616.
Ranabahu, A.; Sheth, A.; Manjunatha, A.; Thirunarayan,
K. (2012): Towards Cloud Mobile Hybrid Application
Generation Using Semantically Enriched Domain
Specific Languages. In: Martin Gris und Guang Yang
(Hg.): Mobile Computing, Applications, and Services,
Bd. 76: Springer Berlin Heidelberg (Lecture Notes of
the Institute for Computer Sciences, Social Informatics
and Telecommunications Engineering), S. 349–360.
Online verfügbar unter http://dx.doi.org/10.1007/978-
3-642-29336-8_24.
Reimer, U.; Heck, U.; Streit, S. (2008): Collaboration-
Oriented Knowledge Management Using Interaction
Patterns. In: Takahira Yamaguchi (Hg.): Practical
Aspects of Knowledge Management, Bd. 5345:
Springer Berlin Heidelberg (Lecture Notes in
Computer Science), S. 26–37. Online verfügbar unter
http://dx.doi.org/10.1007/978-3-540-89447-6_5.
Staab, S.; Walter, T.; Gröner, G.; Parreiras, F. S. (2010):
Model Driven Engineering with Ontology
Technologies. In: Proceedings of the 6th International
Conference on Semantic Technologies for Software
Engineering. Berlin, Heidelberg: Springer-Verlag
(ReasoningWeb’10), S. 62–98. Online verfügbar unter
http://dl.acm.org/citation.cfm?id=1886135.1886138.
Ulrich F. (2010): Outline of a method for designing
domain-specific modelling languages. University of
Duisburg Essen (42). Online verfügbar unter
http://hdl.handle.net/10419/58163.
Vaishnavi, V. K.; Kuechler, W. (2007): Design Science
Research Methods and Patterns: Innovating
Information and Communication Technology. 1st.
Boston, MA, USA: Auerbach Publications.
van Deursen, A.; Klint, P.; Visser, J. (2000): Domain-
specific Languages: An Annotated Bibliography. In:
SIGPLAN Not 35 (6), S. 26–36. Online verfügbar
unter http://doi.acm.org/10.1145/352029.352035.
AModel-DrivenApproachtoCreateandMaintainanExecutableTransferalManagementPlatform
17