The health sector uses clinical guidelines and protocols (we will refer to both
terms as “guidelines”) as help instruments for decision making. These establish
the interventions that must be carried out with regard to the current
circumstances of a patient. Guidelines can be represented using many different
approaches and notations [
In this paper, we discuss the development of an ubiquitous decision support
system (UDSS) for guidelines in a MDA context. Such system should take into
account the indications of the guideline, the current status of the patient and
should provide the necessary information to help physicians in their decisions.
Considering the characteristics of the hospital environment, we think that such
a system must be developed in an ubiquitous setting [
The manual implementation of a UDSS for guidelines would be a long and
costly endeavor due to the vast amount of guidelines, their possibility of change
over time and their dependence on the hospital which uses them. So, we have
considered more appropriate to construct such system using a model-based
development approach. We have decided, as is shown in [
It is worth noting that, to our knowledge, this is the first work to use the MDA approach for the development of a UDSS for guidelines. For this reason our approach would open a new field for the application of the MDA approach. 2
We have decided to use UML 2.0 Statecharts [
Furthermore, the implementation of the UML Statecharts allows the UDSS to show the information related to the guideline, such as patient’s states, the possible events that can take place, the clinical conditions to be fulfilled when each event is triggered, the clinical actions to be carried out, as well as the corresponding new patient’s state. In addition, this implementation helps to record a trace of the application of the guideline, such as physician’s decisions and patient’s states. This information will be stored to be part of the patient’s clinical history and to be used for obtaining a summary concerning the application of the guideline to the patient, for physician’s future decisions or legal support, etc.
As for the use of UML Statecharts for representing guidelines, we want to
highlight two contributions of our approach. 1) We have established several
representation patterns to assist in the modelization process because there is no
single criteria to follow to create a statechart representing a guideline. We do
not describe these patterns in this paper due to space reasons. 2) We have had
to define our own action language for representing guidelines, due to UML does
not have one single and precise action semantics [
In this section we discuss the process we propose to develop an ubiquitous
decision support system for a given guideline, based on a MDA approach [
Once we have modeled a guideline as a UML statechart, the next step consists of using a specific MDA tool for code generation. There are a large amount of MDA based tools, but most of them provide support for code generation based on their own particular transformation approach. In our case, we require a specific approach of model to text transformation that takes into account the specific semantics of the statecharts which represent the guidelines. Because of that, we have chosen a tool with support for customizable model to text transformation. The idea is to define only one set of transformations for all guidelines.
In order to finally obtain the UDSS for the guideline, we propose the use of the generated code from the statechart model, combined with a platform library. The aim of the platform library is to provide the standard services that are common to all guidelines. The generated code has invocations to the methods of this library in order to make use of its services.
The definition of both, the transformations and the platform library, is independent of the used guideline. So, following our approach, if the definition of the guideline is changed, it will be only necessary to modify the statechart which represents the guideline, making it easy to obtain the UDSS for that guideline, without the necessity of modifying the code manually.
Customization of the Model to Text Transformations. Among the large
amount of MDA based tools, we have chosen the MOFScript Eclipse plug-in [
In our particular case, we use the UML 2.0 metamodel and the statechart which represents the guideline as the model. The code generator is defined as an only set of transformations in the MOFScript language. There are two kinds of transformation rules: (1) those which traverse the model (taking into account the different elements in a UML statechart and their specific semantics) and collect the information in it and (2) those who generate actual code.
As we have commented previously, we have defined our own action language as part of our representation patterns for guidelines. Taking into account the possibility of adding new elements to that language in the future, we have defined a method which allows us to generate code from the elements of that action language in an easily way. We do not comment that method due to space reasons. Development of the Platform Library. The aim of using the platform library is to make the UDSS more modular. We achieve this by defining the programming interfaces of the library general enough for being used independently of the platform chosen. This library has different methods, related with the implementation of the UDSS, such as the user interface, the strategy to store the data, etc. The implementation of those methods will depend on the platform we want to use. The generated code has invocations to those methods so that, in order to change the platform, it will be necessary to modify neither the defined transformations nor the generated code. That is, we only have to modify the implementation of those methods.
In addition, we want to develop the decision support system so that it can be accessed in an ubiquitous way, through various kinds of interaction devices. So our intention is to implement the library in such a way that it can provide support for the necessary ubiquitous services, in order to obtain the UDSS with the characteristics that we have explained previously. 4
In this paper we have shown a model-based development approach to construct a UDSS for guidelines. Our approach makes it possible to automatically obtain the UDSS for a guideline represented by a statechart.
At present, we have developed a first prototype using a simple user interface, so as future work we have to modify the implementation of the methods defined in the platform library in order to use another more suitable way of interaction with the user. In addition, the development of a way for recording the information related to the application of the guideline to the patient and the determination of evolution mechanisms to manage changes in the definition of the guideline, are a goal for further development.