Article Preview
TopIntroduction
The patient treatment process is increasingly evolving from isolated treatment episodes towards a continuous process incorporating multiple organizationally independent institutions and different professions. The focus of the medical supply chain in Germany is the patient, who is treated primarily by office-based physicians (primary care sector). Secondary care includes hospitals, laboratories, pharmacies, and ancillary medical institutions, all participating in the medical supply chain. Effective treatment of unclear symptoms or multimorbid patients increasingly leads to the need for information transparency along the medical supply chain. In particular, establishing and managing dynamic teams of cooperating specialists becomes more and more important. Approaches supporting cooperation and coordination among the various cooperating participants need to cope with the heterogeneity of the systems at different sites.
Many sites have their own Electronic Medical Records (EMR) (Powell & Buchan, 2005) for storing patient related information, which typically can be extracted on demand. Yet, it is unclear how these systems scale and how direct communication between institutions can be effectively supported in large-scale scenarios.
Independent Electronic Health Records (EHR) are discussed as a basis for inter-organizational cooperation. Yet, despite of existing standards like openEHR, reality is far from the vision of seamless record exchange, and IT-support for inter-organizational patient treatment processes is an open issue. The message-based approach to system integration used within single organizations is unsuitable for ad-hoc cooperation between independent organizations. A conceptual change from messages to documents is provided by HL7 v3 Clinical Document Architecture (CDA). CDA provides a framework for XML-structured medical documents. Such self-describing electronic documents can be stored independently from the originating system. Document content specifications like the Continuity of Care Document (CCD) have been developed, a U.S.-specific standard, which is a based on HL7 and focuses on document-oriented medical content types. Such content-oriented standards, however, do not consider process history or coordination information, their primary function is standardized semantic tagging of information within electronic documents—not the infrastructure for managing such documents.
In order to minimize the initial effort for establishing an information exchange between healthcare professionals willing to cooperate, we are looking for an evolutionary and decentralized approach to support ad-hoc processes that are emergent in inter-institutional medical supply chains. The traditional approach to manage such healthcare processes is based on paper documents with a dedicated semantics, such as a referral or a discharge letter. We adopt and extend this interaction paradigm to support more complex cooperation scenarios.
The basic idea in distributed document-oriented process management (dDPM) is to use electronic documents as self-contained units of information interchange which also contain process related information. The notion of dDPM is the conceptual foundation of α-Flow, a general framework for document based ad-hoc coordination. α-Flow aims to provide a dDPM process model using active documents.
Traditional Workflow Approaches and Unsolved Issues
The dominant approach for formal workflow models is the activity-oriented workflow paradigm, e.g., Petri Nets with states and transitions or Business Process Modeling Notation with actors and activities. The characterization of tasks or actions by pre-conditions, post-conditions, and possible exceptions is dominant. In contrast, the content-oriented workflow paradigms (Cohn & Hull, 2009; Müller, Reichert, & Herbst, 2007) implement coordination based on state-changes of an artifact life-cycle model. The characterization of artifact states is dominant, for example editing states like “draft”, “submitted”, “revised”, and “final” in a publishing scenario. Both approaches require predefined semantically rich schemas.
To put workflows into practice, prospective conceptualization of the activities is necessary, and task automation must be provided through service implementation. Enactment engines provide process control and monitoring in order to direct firmly rather than to assist wisely. Because workflow systems are server-centric, decentralized workflows are still a challenge. Ad-hoc workflows are not traditionally considered, and initially unknown sets of actors, states, and transitions are not supported.