Approaches to Workflow Analysis in Healthcare Settings

Abstract

Attention to workflow is an important component of a comprehensive approach to designing usable information systems. In healthcare, inattention to workflow is associated with poorly accepted systems and unforeseen effects of use. How best to examine workflow for the purpose of system design is in itself the subject of scientific inquiry. Several disciplines offer approaches to the study of workflow that can be tailored to meet the needs of systems designers in healthcare settings. This paper will describe several methodologies for workflow analysis according to the major schools of thought from which they have been developed. The level of workflow they examine and their theoretical underpinnings will also be described.

Introduction

Several approaches exist to describe and represent work processes for the purpose of information systems design. Which approach to take depends on the goals of the project as well as constraints such as access to a particular setting, the type of work, workers or specific population involved and larger organizational goals impacting system use. Analysis of workflow requires a clear definition. Karsh defines workflow as occurring at three levels1. First is inter-organizational workflow which includes processes occurring across organizations such as hospitals and ambulatory clinics1. The second level describes workflow occurring within an organization which can include the flow of workers, patients, and information across space and time in a department as well as the interdependencies of tasks carried out1. Third is the individual, cognitive level which refers to a person’s flow of thoughts related to a task1. Various workflow analysis approaches offer methods for modeling workflow at different levels. The level(s) of workflow that an information system will support is a vital consideration in choosing the approach to analysis.

Information Systems/Computer Science

The fields of information systems (IS) and computer science (CS) have contributed methods for examining work processes for the purpose of IS design. The theory of petri-nets provides the basis for a CS approach that requires system use data and is typically employed to understand workflow using a workflow management system2. An example of a method based on the theory of petri-nets is process mining 3. Process mining uses system log file data to construct event-based depictions of processes carried out using an information system. It allows for examination of the order of tasks that make up a process, the identification of actors involved in tasks and their relationships. The examination of specific cases that are handled in different ways are also a focus of investigation 3. This method is beneficial for evaluating the use of a particular tool once it is put into practice or for determining how tools may be improved upon to better support desired work practices. This approach examines workflow at the organizational level. It can be used to identify and model standard and alternate paths that may be used to carry out tasks. However, the context in which the work occurs is not included in the analysis.

Another approach to understanding workflow is Contextual Design4. Contextual design provides a framework as well as techniques that can be used by designers to understand work practices5. Three phases of the design process include: 1) contextual inquiry- a type of work interview that takes place in the field with the worker as they carry out work-related tasks; 2) work modeling - creating models of work based on the results of contextual inquiry; and 3) consolidation-the creation of a single statement of the work practice 6. The focus of contextual inquiry is to uncover four aspects of work that are relevant to design: 1) the motive behind tasks; 2) patterns used in carrying out tasks; 3) the structure in place that enables task accomplishment; and 4) conceptual distinctions between aspects of work 7. Contextual design is useful for describing workflow at the organizational and individual levels. Its strength is in its provision of a specific method, contextual inquiry, which provides focal points that are relevant to design 7.

Computer-Supported Cooperative Work (CSCW)

CSCW is a field that was begun by technology developers in an effort to understand the activities of groups engaged in collaborative work activities8. Activity theory (AT) and coordination theory (CT) are two theoretical models that have been utilized in the CSCW literature to describe workflow among groups.

AT posits that human beings engage in purposeful interactions (activities) that are goal-directed and context-specific9. A central tenet is the concept of mediation which refers to the idea that all human activity is mediated by tools. Tools or artifacts are developed as a mechanism for goal-achievement and in turn they mold and influence future interactions between people and their environment10. In addition, all activity is socially and culturally determined. Effective technology development begins by describing activities, their target goals and actions, and the environment in which they take place9. Korpela described the activity analysis and development (ActAD) framework as a methodology for IS design based on AT concepts11 The ActAD framework consists of four steps. First, a description of the activity network is obtained. Second, the development of work activities over time is assessed and developmental goals for a new tool are identified. In the third step, existing goal conflicts are determined and in the fourth step a model of the future view of the activity network is developed. All steps are conducted using workshops with users11 The Activity checklist is another methodology based on AT that is meant to focus developers on those aspects of work that are relevant to design 9. The checklist consists of four categories: 1) means/ends; 2) environment; 3) learning, cognition and articulation; and 4) development 9. Means/ends focuses on the hierarchical structure of activities. Environment focuses on context. Learning/articulation describes the internal cognitive components and external actions related to activities. Anticipated changes in actions related to the use of the new technology are also described. The Activity checklist provides more specific guidance than the ActAD framework. However, both require tailoring to meet specific project needs. The Activity checklist can be used to develop guides for the purpose of interviewing prospective users or as a tool to guide direct observations of work practices.

CT is concerned with task-interdependencies among workers that result in harmonious goal-achievement12. Types of task interdependencies include prerequisite or tasks in which the output of one is required for the next, tasks which require shared resources and simultaneous tasks or tasks that require synchronization12. CT suggests that uncovering task interdependencies can result in identifying new ways to manage them and thus new technologies to support them. Additionally, four processes underlying coordination and their components can be examined: 1) coordination, 2) group decision-making, 3) communication, and 4) perception of common objects 12. Although CT does not offer any specific methods for obtaining information needed to analyze a process, a study of cooperation and conflict in a hospital radiology department used a tracer method to analyze coordination of activities13. The tracer method used a process of ‘tagging’ a particular artifact, in this case a radiology request form, to map out the process followed throughout the department. Each person who came in contact with the document was asked to sign the tag. These people were then identified as interviewees who could discuss the process followed in the department13.

CT is useful for examining workflow within and across organizations. It would not be useful for understanding individual workflow.

Cognitive Science

Cognitive science is a multidisciplinary field concerned with understanding human thought processes including knowledge attainment, memory and problem solving14. Cognitive task analysis (CTA) encompasses a group of methods based on cognitive theory used to examine human task 15. While CTA methods generally are focused on individual work, the need to study complex interactions across teams, particularly in the health care realm has led to increased interest in the theory of distributed cognition16.

Two approaches used to carry out a CTA include a cognitive walkthrough (CW) and a think-aloud protocol (TA)17. A CW is carried out by a systems analyst to simulate a user’s cognitive processes as they engage in work-related tasks. Its aim is to identify goal-action sequences including cognitive or perceptive and physical actions required to reach goals17. In a TA, system users are asked verbalize their thought processes as they carry out tasks 17. Verbalizations captured are reflective of the user’s mental model of the task.17. This information can then be translated into design requirements to support individual workflows.

Distributed cognition (Dcog) is a theoretical model that highlights the collaborative nature of human cognition16. According to Dcog people and objects form collective activity systems that are in a constant interactive process rooted in social and cultural practices18. Information distribution across groups of humans and artificial agents can be described. Zhang described a conceptual framework based on Dcog that can be used to describe workflow for the purpose of information systems design. User, functional, representational and task analysis (UFuRT) consists of four phases of analysis that result in system requirements identification19. Table 1 describes these four levels. This approach can be used to examine workflow at all levels. It is well-suited to inter-organizational and organizational workflow. For individual workflow, the task analysis stage can help describe information distribution across the human mind and the external artifacts for a single user.

Table 1.

UFuRT levels of analysis.

Level of Analysis	Focus
Distributed user analysis	Identify user characteristics, describe division of labor, identify domain knowledge and skill levels, describe patterns of communication and social relationships
Distributed functional analysis	Determine high-level relationships, determine human and object limitations
Distributed task analysis	Determine task distributions across space and time, determine information flow, describe human and non-human agents
Distributed representational analysis	Determine how information can be re-distributed across human and non-human agents in new ways

Organizational Science

One additional field that has contributed a methodology useful for understanding workflow is organizational science. Pentland and Feldman identified a model of organizational routines that can be applied to the study of workflow in different settings 20. The study of organizational routines is aimed at clarifying internal structures within organizations for the purpose of influencing change and directing process re-design. According to Pentland and Feldman’s framework, organizational routines are dynamic in nature and continually adjusting in response to contextual issues 20. Two components of organizational routines can be characterized: 1) the ostensive aspect which refers to the general pattern of a routine and 2) the performative aspect which refers to the more specific actions engaged in by individual people involved in the routine under particular circumstances 20. Artifacts are viewed as physical manifestations of routines. Artifacts may include such things as computerized workflow systems, particular work spaces or layouts, or documents that describe how a procedure is to be carried out 20. Although this framework acknowledges the contribution of individual routines to the overall routine, it conceptualizes the organizational routine as the larger interconnected patterns or steps that are played out across groups of individuals. The level of analysis is the pattern or the process as opposed to each individual’s role in the process. This makes this model useful for understanding workflow at the organizational level as opposed to the individual level. Workflow across organizations interacting with one another may also be studied from an organizational routines perspective.

Implications for Systems Development

The study of workflow is important in systems development because it allows for a clear definition of the problem space to be addressed. Systems developers have looked to various disciplines for systematic approaches to understand the features of workflow that are important to design. These approaches guide designers in various directions depending on how work activities are conceptualized. Several theoretical frameworks highlight the interplay between people and systems working together in context. These frameworks keep the focus of the design process on the actual work activities to be supported allowing design features to emerge from data. Table 2 summarizes these approaches, their level of analysis and their associated methods.

Table 2.

Workflow analysis approaches and associated methods.

Approach	Workflow Level Analyzed			Associated Methods
	Inter-Organizational	Organizational	Individual
CS/IS
Process Mining		√		Data mining
Contextual Design	√	√	√	Contextual inquiry, work modeling, consolidation
CSCW
Activity Theory	√	√	√	Direct observation, interviews
Coordination Theory	√	√		Direct observation, tracer method, interviews
Cog Science
CTA			√	CW,TA
Dcog	√	√	√	Direct observation, group and individual interviews
OS
Organizational routines	√	√		Interview, direct observation

CS/IS: computer science/information system; CSCW: computer-supported cooperative work; cog science: cognitive science; CTA: cognitive task analysis; CW: cognitive walkthrough; TA: think-aloud; Dcog: distributed cognition; OS: organizational science

Methods employed to collect data often include user participation. The need for participatory design processes is widely recognized as critical to the development of successful systems21. Contextual design, for example utilizes contextual inquiry. Methods of cognitive task analysis also emphasize the role of users. In general, user participation is encouraged and the incorporation of qualitative methods such as interviews and direct observation are utilized to gather needed data.

Several common concepts appear to have important implications for design. Both AT and contextual design recommend identification of motives that are related to specific work activities. High level goals provide the basis for design. CTA also uncovers goal structures that are important in identifying specific design features.

Another common concept is context. Context refers to the environment in which the work takes place. This includes the physical environment, people present, their social relationships, artifacts and environmental constraints. AT, Dcog, contextual design, and organizational routines all emphasize the importance of context in identifying appropriate design. Task inter-dependencies and information flow are also a focus of Dcog, CT, and process mining. Organizational routines and contextual design stress descriptions of patterns of activity while AT refers to task hierarchies.

All of these approaches can be useful in analyzing workflow for the purpose of design. The type of project or phase of design that a project is focused on may guide researchers toward an appropriate approach. When the system is created from scratch with the intent of improving workflow and efficiency then more comprehensive methodologies may be needed. AT and contextual design offer broad approaches that can highlight areas in need of improvement. If system updates are needed, contextual design and Dcog may be helpful in understanding how to implement changes without unnecessary disruptions to existing workflows. CTA methods are also useful to describe how updates may support individual users. System evaluations can be informed through process mining combined with CTA and observational methods. This approach would allow for examination of system use in context where process mining alone or with CTA only would not take context into account.

Conclusions

The frameworks presented are representative of the major schools of thought related to studying workflow that have implications for systems design. Each of these offers a conceptualization of workflow that can help identify design features needed to support work processes. Understanding goal structures for example can help to determine the focus of a system. The individual level of workflow is not always considered in these approaches. It is important to evaluate individual components of tasks because these may impact group goals. A comprehensive understanding of workflow provides input to develop, refine and evaluate systems that can support the objectives of workers and organizations. The use of systematic methodologies can focus attention on those aspects of workflow that have particular implications for design. In this way, design can be approached in an efficient manner leading to well-developed systems that avoid pitfalls encountered when less focused approaches are used.