"Delivering clinical decision support services: there is nothing so practical as a good theory" J Fox, D Glasspool, T Hope, V Patkar, M Austin, L Black, M South, D Robertson, C Vincent to appear in Journal of Biomedical Informatics
Abstract: Sittig et al (2008) set out ten “grand challenges” for the successful development and deployment of Clinical Decision Support services in order to “inspire stakeholders in a position to advance the state of CDS technology and practice”. The 10 challenges were identified and prioritized on the basis of empirical experience and with the expectation that overcoming these challenges will depend heavily on practical problem solving and finding out “what works” in clinical use. As active participants in this field we take the position that a pragmatic approach to design must be accompanied by sound theoretical principles and safe engineering methods. We illustrate this with an approach to CDS design that starts with a formal model of knowledge-based decision-making, clinical processes and distributed care services, identifying four key pillars of theory, and relationships between them. Sittig et al’s challenges are reviewed to consider how such a framework can facilitate application design and implementation, clinical use, service interoperability etc. We do not claim a formal approach to design is an alternative to empirical evolution of clinical services but is a foundation on which practical experience can be understood, shared and built upon.
"Sharing Choreographies in OpenKnowledge: A Novel Approach to Interoperability" P Besana, V Patkar, A Barker, D Robertson, D Glasspool To appear in Journal of Software, Special Issue on Semantic Extensions to Middleware, August 2009
Abstract: Software systems are becoming ever more complex, and one source of complexity lies in integrating heterogeneous subsystems. Service Oriented Architectures are part of the answer: they decouple the components of the system. However normally service oriented architectures are still designed and enacted from a centralised perspective: a single process invokes remote services, unaware of being part of a larger, more complex workflow. We claim that the orchestration-based approach does not scale well with increasing complexity and heterogeneity of the components. We propose a choreography-based approach, where choreographies are the shared contracts that actors agree to follow. We first present the OpenKnowledge framework, designed according to choreography-based principles. We then demonstrate how the implementation of a complex, distributed model such as the triple assessment, used for the diagnosis of patients suspected with breast cancer, can benefit from this framework.
"Goal-based decisions for dynamic planning" E.Black, D. Glasspool, M. A. Grando, V. Patkar and J. Fox In Proceedings of the Twelfth Conference on Artificial Intelligence in Medicine (AIME09, Verona, Italy), LNAI 5651, pp. 96-100, 2009. Springer-Verlag (In press).
Abstract: The need for clinical guidelines to be implemented at different sites, to adapt to rapidly changing environments, and to be carried out by distributed clinical teams, implies a degree of flexibility beyond that of current guideline languages. We propose an extension to the PROforma language allowing hierarchical goal-based plans. Sub-plans to achieve goals are proposed at runtime so that changing circumstances may be flexibly accommodated without redefining the workflow.
"Distributed decision support in healthcare" A Barker, P Besana, M South and J Fox (internal report)
Abstract: Task based modelling of clinical processes has been clearly demonstrated to improve the delivery and safety of healthcare services. This paper focuses on the modelling and execution of computer interpretable clinical guidelines (CIGs) in a decentralised environment, where all services, knowledge and data may be distributed. Through a motivating scenario taken from the field of cancer research, we argue against complete centralisation and towards an open, decentralised architecture, allowing domain experts to curate and maintain their own processes and data sets. As a solution, we propose an architecture based on Tallis, a framework for describing and enacting clinical guidelines and OpenKnowledge, an enabling technology for choreography based services. We claim this architecture, although more complex to initially model, scales with increasing complexity, is more flexible and reliable than architectures which rely on centralisation. A mini roadmap is presented which outlines our key objectives in advancing this field.
"Distributed agent-based services for decision support" Adam Barker, Paolo Besana and John Fox (to appear in Brahanm (ed) Multi-agent systems in healthcare, forthcoming)
Abstract: The sheer quantity and complexity of medical information, even within a single speciality, is beyond the power of one person to comprehend. Clinical decision support (CDS) systems have been clearly demonstrated to improve practice by removing complexity and aiding the decision making process for clinicians. However, the speci c pieces of information most relevant to a particular clinical decision are typically scattered over a wide range of databases, applications, journals and written notes. Centralisation of knowledge is becoming less practical as the volume and complexity of data increases. Through a motivating scenario taken from the eld of cancer research, we argue against complete centralisation and towards an open, decentralised architecture, allowing domain experts to curate and maintain their own processes and data sets. We introduce the UK-based Safe and Sound project and propose an architecture based on PROforma, a formal language for describing CDS systems and OpenKnowledge, an enabling technology for decentralised agent-based systems. We demonstrate that although more complex to initially model, our architecture scales with increasing complexity, is more flexible and reliable than architectures which rely on centralisation.
"Goal-based design patterns for delegation of work in health care teams" Grando A, Peleg M & Glasspool D (submitted to MedInfo)
Abstract: We show how a domain and language independent design pattern, defined as networks of tasks and goals, can be used to formally specify the transfer of responsibility and accountability when tasks are delegated in healthcare teams. The pattern is general enough to be applied unchanged across a broad range of different healthcare situations.
"A goal-oriented framework for specifying clinical guidelines and handling medical errors" Grando A, Peleg M & Glasspool D (submitted to Journal of Biomedical Informatics)
Abstract: Computer-interpretable guidelines (CIGs) aim to improve patient care and reduce medical errors. Although CIGs implement evidence-based recommendations they cannot prevent exceptional behavior from happening. Therefore, to promote patient safety, we developed a framework that can monitor, detect, and handle such exceptions that occur during normal CIG execution and prevent them from developing into medical errors. Our framework, which is independent of a CIG formalism, enables specifying the goals of a guideline and linking them with recommended tasks that could satisfy the goals. Exceptions are linked with goals that manage them, which can be realized by tasks or plans. To achieve a link between the tasks, plans, goals, monitored effects, and exceptions, our definition of goals and exceptions is state-based. We demonstrate our approach using a generic plan for management of a chronic disease and a particular instantiation for hypertension management.
