سیستم ویرایش دانش مبتنی بر الگو برای ساختن سیستم های پشتیبانی تصمیم گیری بالینی

Decision support in medicine is being more and more configured as an innovative and valuable way for promoting more consistent, effective, and efficient medical practices. The broad acceptance and efficient application of Decision Support Systems to medical settings strongly require some mechanisms to conveniently update and handle these systems with respect to medical progress or adaptation in the treatment of individual diseases. In this respect, this paper proposes a pattern-based knowledge editing system to guide and assist the creation and formalization of condition-action clinical recommendations to be used in knowledge-based Decision Support Systems (in the following, DSSs). This system has been devised with the aim of: (i) offering a set of patterns for easily inserting and editing such clinical recommendations; (ii) synergistically combining multiple knowledge representation techniques to instantiate these patterns within hybrid knowledge bases (KBs), made of if–then rules built on the top of ontological vocabularies; (iii) reducing the complexity of the formalization process, by graphically guiding the definition of hybrid KBs that could be functional in the context of clinical DSSs and enabling their automatic encoding into machine executable languages. A usability evaluation has been carried out, showing a good satisfaction of medical users with respect to the system implemented, and, thus, proving both the feasibility and usefulness of the approach proposed.

In the last years, decision support in medicine is being more and more configured as an innovative and valuable way for providing clinicians or patients with clinical knowledge and patient related information, intelligently filtered or presented at appropriate times, in order to enhance the overall quality of care. Clinical knowledge of interest could range from simple facts and relationship to best practices for managing patients with specific disease states, new medical knowledge from clinical research and other types of information [23]. Several recent studies have suggested that decision support in medicine without advanced systems designed to help doctors to make decisions by providing motivated suggestions [30] may not provide the promised improvements in patient safety or quality of care [13] and [18]. In this respect, a Decision Support System (in the following, DSS) for healthcare applications can be defined as an active knowledge resource that uses patient data to generate case-specific advice which supports decision making about individual patients by health professionals, the patients themselves or other concerned about them [19]. Recent implementations of DSSs in medicine, known as knowledge-based, encode clinical practice guidelines into a logical formalism for simulating the process followed by the physicians [8] and [25]. The knowledge base (KB) is their key element, since it includes the corpus of relevant knowledge, coming from the clinical recommendations. This kind of DSS is being more and more widely adopted, since it is expected to promote more consistent, effective, and efficient medical practices and improve health outcomes when used [26] and [35]. With respect to the typology of clinical guideline encoded, up to now, several DSSs have been focused on condition-action clinical rules rather than time-oriented guidelines [2], [4] and [22]. Condition-action clinical rules represent elementary, isolated care recommendations, which specify one or at most a few conditions which are linked to specific actions [29]. It is interesting to note that the most diagnostic and therapeutic clinical guidelines can be distilled in terms of a set of condition-action clinical rules, although this discards the control flow structure [24]. Building a DSS strictly based on condition-action clinical rules mainly requires their collection, systematization and technical formalization within the KB. Typically, clinicians are not supposed to directly access the clinical recommendations encoded in the KB, but they can only ask for the assistance of the DSS, which can then decide to use the KB for its decision making process. It means that, the KB is not accessible and editable directly by clinicians and it can be altered and updated only by means of an intervention made by technicians. However, to perform such an intervention, knowledge of both medicine and formal languages must be combined to create a valid and medically useful DSS. Thus, all the process absolutely requires the cooperation of both clinical experts and experts in medical informatics [27]. Nevertheless, it is extremely worth highlighting that the need of technical experts for editing and upgrading clinical recommendations in a knowledge-based DSS is a strong limitation for medical users. As a matter of fact, actually, one prerequisite for the broad acceptance of such DSSs and their efficient application to medical settings is the guarantee of a high level of upgradability and maintainability, (i) to change clinical rules according to their evolution to implement medical progress in the treatment of individual diseases, or (ii) to adapt generic, site-independent clinical rules to a patient to be treated [9]. Since updating the KB can require a continuous intervention, it is unthinkable that it cannot be done directly by doctors when needed. Also, by providing a direct access to the KB, doctors are encouraged to use clinical DSSs built on the top of it, since mostly entrusted with the suggestions generated starting from their expertise, especially if encoded by them. In contrast to the intensive efforts made to develop knowledge-based DSSs, the issue of providing solutions for easily editing and upgrading condition-action clinical rules into their KB has been widely neglected thus far. In this respect, this paper proposes a pattern-based knowledge editing system to guide and assist the creation and formalization of condition-action clinical recommendations to be used in knowledge-based DSSs. This system has been devised with the aim of: (i) offering a set of patterns for easily inserting and editing such clinical recommendations; (ii) synergistically combining multiple knowledge representation techniques to instantiate these patterns within hybrid knowledge bases (KBs), made of if–then rules built on the top of ontological vocabularies; (iii) reducing the complexity of the formalization process, by graphically guiding the definition of hybrid KBs that could be functional in the context of clinical DSSs and enabling their automatic encoding into machine executable languages. The rest of the paper is organized as follows. Section 2 introduces an overview of the state-of-the-art solutions for building KBs within clinical DSSs and addresses the motivations underlying the development of the proposed approach. Section 3 depicts the pattern-based approach as well as the methodology chosen for representing medical knowledge and linking it to the defined patterns. Section 4 outlines the results achieved, in terms of the editing system implemented, and reports some illustrative examples about how to use the system for inserting clinical rules according to the patterns defined. Section 5 details an evaluation of usability with respect to the system implemented, Finally, Section 6 concludes the work.

The presented system offers an innovative and valuable way to guide and assist the creation and formalization of condition-action recommendations to be used in a clinical knowledge-based DSS, with the aim of being mainly oriented to medical users. The key issues of the systems are: (i) a set of patterns for easily inserting and editing clinical recommendations; (ii) the combination of multiple knowledge representation techniques to instantiate these patterns within hybrid KBs, made of if–then rules built on the top of ontological vocabularies; (iii) the simplification of the formalization process, by graphically guiding the definition of hybrid KBs that could be functional in the context of clinical DSSs and enabling their automatic encoding into machine executable languages. Since both the patterns defined and their instantiation within a hybrid knowledge-based model have a general basis, the system is undoubtedly applicable to many medical scenarios. The encouraging results given by the usability evaluation suggest that the pattern-based system could be simply and proficiently utilized by clinicians, especially if equipped with a positive attitude towards the use of clinical DSSs, to directly author and encode their knowledge, Such a way, the remarkable aims of conveniently updating and handling knowledge-based DSSs with respect to medical progress and adaptation in the treatment of individual diseases can be achieved, so as to improve their acceptance and efficient application in real medical settings. Finally, future work will regard the improvement of the described editing system by means of new intuitive and user-friendly facilities to handle vagueness in condition-action clinical recommendations when doctors’ decision-making model results to be pervaded by uncertainty and vagueness. Indeed, in many situations, clinical rules the doctors have in mind could be vague in nature, since they are used to formulating their expertise at a high level of abstraction, in form of smooth linguistic labels, rather than as expressions with clear-cut boundaries. Next steps in this direction will regard the application of Fuzzy Logic [36] to model vagueness in clinical recommendations, similarly to other approaches already proposed in literature (G. Kong et al.), but with a particular emphasis on the possibility of building fuzzy rules on the top of ontological concepts and properties.