چارچوب برای یک سیستم پشتیبانی تصمیم گیری هوشمند:یک مورد آسیب شناسی مرتب سازی آزمون

Decision context, knowledge management, decision makers, and decision strategy are fundamental components for understanding decision support systems (DSSs). This paper describes the specific case of designing a framework for an intelligent DSS in the context of pathology test ordering by general practitioners (GPs). In doing so it illustrates the processes of discovering practical and relevant knowledge from pathology request data generated and stored in a professional pathology company, investigates and understands the decision makers (GPs) through a survey about their current practices in test ordering and their requirements for decision support, and finally proposes an intelligent decision support framework as the decision strategy to support GPs in ordering pathology tests more effectively and appropriately. The process and framework developed through this case contributes effective guidance for practitioners and theoretical understanding concerning intelligent decision support in a complex environment.

Ordering of pathology tests by general practitioners (GPs) contributes significantly to the rising costs of health care [48]. Over the past decade Australia has witnessed a considerable rise in the number of and expenditure on pathology requests by GPs. This increase is the consequence of: improved communication between patients and GPs; government incentives for longer consultations; the shift of health services to a community environment; increased concern about medical litigation; and/or increased patient expectations [11], [29] and [48]. Other external factors include the introduction of new Medicare Benefits Schedule (MBS) items and increased computerization [12]. Globally there is a perception that pathology tests are not used appropriately [49], [65], [67], [75], [76] and [80], although there are concerns with the rigor in some studies and associated weak supporting evidence [45], [64] and [74]. Evidence-based medicine indicates that tools like computerized clinical decision support systems (DSS) can improve the quality and effectiveness of clinicians' decisions [16], [18], [28] and [34]. For Australian GPs, although government promotion and incentives have resulted in increased use of “medical desktop” software as the referral point for primary care during patient consultations, particularly for prescribing medications (98%), checking for drug–drug interactions (88%), ordering laboratory tests (85%), running recall systems (78%) and recording progress notes (64%), the current application of computerized clinical DSSs is limited [33], [47] and [78]. For example, with respect to pathology requests, the most common practices involve ordering laboratory tests (85%), receiving or storing pathology test results (79%) and running the recall system for routine tests (78%) rather than investigating the suitability of available options. Evidence from some studies show that a high percentage of real-time clinical decision support suggestions are being over-ridden or ignored due to disruptions to workflow, time restraints and a perceived lack of relevant suggestions [52], [69] and [77]. Hence, in designing a pertinent DSS, it is crucial to take account of contextual factors. The aim of this paper is to develop a framework for a DSS that can assist GPs in ordering pathology tests more effectively and appropriately. In so doing we establish the merit of an integrated approach that combines knowledge discovery and case-based reasoning (CBR) mechanisms to capture the contextual requirements for an evidence-based, situationally relevant, flexible and interactive DSS, which we call an intelligent DSS. The contributions of this study are three-fold. Firstly, by discovering and extracting practical and relevant knowledge from past pathology request data, we provide fresh understanding about the use of pathology tests from both patient-centric and clinical situation-centric perspectives. Secondly, results from our online survey provide comprehensive understanding about the appropriateness of GPs' ordering behavior as well as their needs of and requirements for intelligent decision support. Finally, this study shows how an integrated approach can be used to create an evidence-based, situationally relevant, flexible and interactive DSS that suits complex environments. The remainder of the paper is structured as follows. Section 2 discusses the limitations of existing support for GPs in ordering pathology tests, while the components involved in decision making in the context of pathology ordering are outlined in Section 3. Section 4 describes the processes deployed to discover and extract useful knowledge/evidence from past ordering behavior that can be used to inform the decision making process, while Section 5 reviews the decision makers' (i.e. GPs') needs of and requirements for support in ordering pathology tests. These ideas are synthesized in Section 6 where the research proposes and reviews a framework for an intelligent DSS as a strategy to support GPs in ordering pathology tests more effectively and appropriately. In Section 7 we highlight implications for future research and present our conclusions.

The aim of our study was to design a framework for an intelligent DSS to optimize GPs' practices in pathology test ordering. Here the optimal (or appropriate) choice involves ordering the maximum number of “necessary” tests and the minimum number of “unnecessary” and “unsuitable” tests as indicated by clinical guidelines within constraints including patient pressure, limited consultation times and cost considerations. Given that the existing knowledge sources for GPs were found to be difficult to access, we mined patient and clinical data contained in past pathology requests to extract patient- and clinical situation-centric knowledge. Then, through an online survey and literature review, we established GPs' needs for a DSS that is evidence-based, situationally relevant, flexible and interactive. Such a DSS offers the “right” knowledge in the “right” format at the “right” time to GPs at the point of test ordering. Methodologically, the advantage of the proposed integrated approach is its ability to use the available information at the cluster level to form a generalized case based on a set of similar cases. This presents a new perspective on the use of prototypes through case aggregation — one of the current trends of medical CBR systems according to a recent overview of medical CBR systems and system development [51]. Adopting this perspective better equips the designers of a DSS to address the most challenging task for the CBR method — namely adaptation. In medical applications it is almost impossible to generate adaptation rules that consider all possible important differences between current and past similar cases. Therefore, some adaptation solutions have been developed that are rather typical for medical domains [59], one being to generalize from single cases into abstracted prototypes or classes, since a problem for adaptation is the extreme specificity of single cases. This can be achieved by retrieving past cases at the cluster level as described in the “case retrieval” step detailed in this study. It is noteworthy that the generic nature of the proposed approach provides enough flexibility to customize it to other domains including, for example, customer relationship management and customer care. This and the applicability and useability of our DSS for other complex settings and countries, constitute interesting and promising directions for future research. Moreover, further research could explore external evaluation of the framework with experts and users. In conclusion, our framework for the proposed intelligent DSS draws together a new operative and robust methodology that can be used to generate the required evidence to support GPs' decision making and achieve more effective and appropriate pathology test ordering. Further, our process and framework contribute effective guidance for practitioners as well as theoretical understanding concerning intelligent decision support in a complex environment.