Representing and Processing Dynamic Healthcare Workflow

Healthcare workflows are complex and highly variable. Healthcare workflow execution can be affected by any participant in a process, including clinicians, the patient, and the patient's family, as well as environmental factors such as clinician, staff, facility and equipment availability, and patient clinical status. Attempts to document healthcare workflows result in highly detailed descriptions (often incomplete) with many possible pathways to achieve the goals. A variety of graphical methods and languages have emerged to support the documentation of workflow and computer-based execution. However, only a few solutions exist that enable workflow to address the full complexity and variability of healthcare processes. We have re-conceptualized workflow and developed a new workflow representation and execution framework based on software engineering, inference engine, and database concepts, which has freed workflow representation from the constraints of past methods. We have reported our work on context-aware workflow in a separate paper. Although much has been done on the representation of workflow in business settings, the representation of workflow in highly dynamic settings is still a focus of research. Complex dynamic environments are characteristic of healthcare. They typically involve considerable human interaction resulting in a high degree of variability in scenario outcome. Healthcare settings have many decision makers, kinds of decisions, events and a multitude of reactive, subsidiary workflows that often require a quick revision of the course of action (See Table 1). Operational and treatment protocols attempt to regularize workflow, but the needs of care, the great variety of situations and individuals, the exigencies of the moment (such as equipment failure), and the nature of human beings frustrate attempts at regularization, often resulting in protocols being labeled as "rigid" and hence being abandoned. While event sequences in healthcare processes may abide by loose constraints, they are largely non-deterministic. Therefore, it is difficult, if not impossible, to prescribe fully healthcare workflow. Instead, workflow must be dynamic, self-adapting and evolving at execution-time to match the dynamicity of the environment. Traditional workflow technology by its very static nature supports a finite set of scenarios. In fact, traditional workflow is understood to support, at best, the union of atomic workflow patterns described by van der Aalst1. Available workflow platforms that support these patterns are often incomplete, unsatisfactory, or even non-existent. In fact, no single commercial product supports all listed patterns.1 We have developed a new way of representing healthcare processes that is able to address human-machine interaction and complexity. It converges service oriented architectures concepts with decision support techniques.