Some organisations have high level objectives to meet but need a way of knowing how. We present here an application of a previously published method to a medical case study. This method consists in assigning sub-goals of a high-level goal to the actors of an organisation in order to guarantee the satisfaction of the high-level goal. The application is organised in a modelling session with a domain expert and produces goal models. Feedback from the domain expert on the method is proposed.
patient’s journey and (
In this section, we briefly introduce the method presented in [ 1] and adapted from [2]. We start the method with an HLG and a set of actors from the organisation. All the actors in the organisation want to satisfy the organisation’s goals and are looking for a way to do so. Then, the HLG is translated into a goal and assigned to an actor in the set. The goal is analysed with respects to the skills of the actor and then refined into two sub-goals: which contains the part of the goal that the actor can satisfy, and , which contains the rest, such that satisfying and induces satisfying the initial goal. The goal is held by the actor and is labelled satisfiable , meaning that the actor can satisfy it. The goal is delegated (i.e. given) to another actor of the set. This new actor is now solely responsible for satisfying . Thus, is examined with respects to the skills of this new actor, and the process is continued until all goals are labelled satisfied. All actions (refinement, delegation, labelling) are decided by the experts who are part of the decision-making entity Global Manager. This method was previously applied to three HLGs of an aeronautical company.
We organised the method application with the domain expert in three workshops conducted by videoconference. The first session lasted 30 minutes. The aim of this session was to unfold the algorithm with the domain expert acting as Global Manager. During the session the model expert used a free hand drawing tool and the domain expert did not have access to this model. At the end of this session, a first model was built. The second session, also 30 minutes long, was aimed at reviewing and consolidating the model obtained in the first session. The focus was on improving the wording of the objectives and clarifying some medical domain aspects. At the end of this session, the initial model was complete and the goals were clear and unambiguous for all participants. The third and final session of 10 minutes was dedicated to the final validation of the model. This session provided an opportunity to reach consensus on the final model, taking into account any suggestions for improvement made by the domain expert. During the last two sessions, the domain expert had access to the model, but modifications were only made by the model expert.
Between sessions, questions were exchanged by written message. At the end of the three sessions, the domain expert was interviewed to gather her impressions of the models obtained.
The final goal model is presented on Figure 1. The initial goal is B0 : optimal management of stroke in adult patients. A stroke is an interruption of blood flow to part of the brain and can cause irreversible damage to the brain, leading to motor, cognitive impairment and even death.
At the beginning of the method, the Global Manager decides to assign this first objective to the Relatives actor. Relatives does not have the skills to fully satisfy the goal, so the Global Manager chooses to refine it into two goals: B1a: alert that Relatives can satisfy, so it is labelled satisfiable and put in grey in the model. The second objective is B1b: rescue. As B1b contains elements of B0 that cannot be fulfilled by Relatives, B1b is delegated to another actor, Emergency Service, chosen by the Global Manager. The algorithm is pursued. At the end of the method, five actors have received at least one goal : Rehabilitation service, Relatives, Emergency service, Social Workers and Other medical teams. Twenty-eight actions have been done including labelling satisfiable for twelve goals.
The domain expert appreciated the method concepts and more specifically the delegation mechanism. In her opinion, delegation highlights the importance of each actor in the care pathway, while emphasising the notion of continuity and collaboration in their role. In addition, delegation emphasises the transfer of responsibility for achieving objectives between diferent actors. In this application, that responsibility is patient care. These two aspects of delegation were seen by the expert as strengths of the modelling approach used. However, the domain expert points out that the method “segments things that are not so segmented in real life”. This segmentation is particularly significant for goal B4b": prevention which is performed in real life by both Other medical teams and Rehabilitation service. In order to express this very strong collaboration between the two actors, we have added a collaboration link in the final diagram.
Both objectives of the domain expert are satisfied at the end of the application. About
objective (
This application allowed us to test our algorithm in a medical context, with a real organisation and an expert in the field studied. It was a completely diferent domain from our previous aeronautical applications. Nevertheless, the method was still globally adapted and allowed us to build a goal model of the situation under study. The objective (i) consisting in testing the method on a diferent domain can thus be considered achieved. The final models are validated by the domain expert who wants to use them professionally. Objective (ii) is therefore met.
However, there were some deviations from the original algorithm over the course of the sessions. For example, some refinements were made in more than two goals, or they followed each other without any delegation. It was also sometimes dificult for the domain expert to distinguish between refinement and delegation, and the division of time. These points need further investigation and could contribute to the development of our method and algorithm.
The authors would like to express their thanks to Barbara Chan, Stéphanie Roussel and Anthony Fernandes Pires for their support and valuable suggestions.