In the last decade, the health sector has used clinical guidelines to help make decisions. In this context, there have been a growing interest and active research in computerizing clinical guidelines mainly to facilitate their effective implementation in complex clinical environments. Computerized guidelines substitute the ambiguous and hard to understand hard copy documents in which guidelines are normally represented. It enhances, at the same time, the advantages of guidelines, by aiding timely decision making and easing quality assessment and monitoring. There currently exist various computer models and tools for the representation, management and handling of guidelines (see e.g. GLIF, Asbru, EON, etc). Nevertheless, such proposals give partial solutions to the overall problems of the representation, quality improvement and implementation of the guidelines we cover in this thesis. Additionally, in the majority of cases, these tools do not take into account the possible significance of being able to register and trace doctors' decisions, not only in terms of patient health care, but also when faced with legal issues.
When dealing with the computerization of clinical guidelines, several factors must be taken into account such as the vast amount of clinical guidelines, the possibility of their changing over time and their dependence on the hospital which uses them. These features make the manual implementation of every guideline into a software system a long, cumbersome and costly endeavour, a significant challenge to implementing or defining tools to support decision making for guidelines.
Regarding the characteristics that a decision support system for guidelines must have, we believe that the data generated during the application of guidelines to patients constitutes a key factor which can provide great benefits to several health care fields. Nevertheless, this aspect has been neglected in the literature and there are limited works that consider clinical guidelines' traceability.
On the other hand, it is clear that any such implementation has to be made in guidelines without semantic errors or inconsistencies in their definition. Nevertheless, many works in the literature have assessed the quality of clinical guidelines finding that most guidelines in use lack quality, and require re-evaluation and improvement.
Taking into account the partial solutions that offer a definition of decision support systems for clinical guidelines, we propose the development of a framework which will allow us to develop and manage such decision support systems. In other words, we propose to develop a framework called Decision Facts Management System (DFMS) which carries out such development and management. In particular, we propose that the DFMS uses, in contrast to other works, Model Driven Development (MDD) techniques mainly to automatically generate the decision support system for a guideline.
Furthermore, we propose to provide each decision support system with traceability properties in such a way that the information generated during the application of the guideline is recorded automatically. For such a task, in contrast to other works, we will automatically generate the mechanisms and devices which allow us to manage the storage of the decisions taken by the physicians, as well as the patient's states during the guideline's application. We propose the combination of MDD techniques with database schema mappings for metadata management, so that the DFMS Framework can automatically generate these storage mechanisms and devices.
Another characteristic that we propose for the DFMS Framework is to include formal techniques for verifying requirements in the guidelines, so that we can find semantic errors and inconsistencies in their definitions, ensuring in this way that the decision support systems are developed from quality guidelines. Nevertheless, every verification process requires the formal specification of the requirements to be checked. We have developed a pattern-based approach for defining commonly occurring requirements in guidelines in order to help non-experts in with their formal specification. Additionally, this pattern-based approach is general enough to be also used in non-clinical contexts. In order to reach this validation and verification goal, we propose to combine MDD techniques with formal verification techniques, particularly with model checkers. Such techniques have been integrated into the DFMS Framework in such a way that it allows us to verify specific requirements in guidelines, which have been defined using our patterns.
The overall DFMS Framework has been developed based on a proposal given by the research group to which the PhD belongs and which consists of using UML statecharts as a method to model the dynamics of clinical guidelines. Based on such proposal, the DFMS Framework takes the statechart model that represents the dynamics of a clinical guideline as starting point to carry out both the verification and development of GBDSS processes of the guideline.
To sum up, the proposed DFMS Framework includes the possibility of both using verification techniques to check guidelines against definition errors or inconsistencies, and automatically generating decision support systems for guidelines (with traceability properties). It is therefore a comprehensive tool which will facilitate improvements in the representation, quality and application of clinical guidelines in daily clinical practice.