According to Google, “clinical workflow analysis” is the most popular search term that brings visitors to EHR Science. I am not surprised. Workflow disruptions are increasingly being recognized as workarounds, usability issues, safety concerns, and CDS problems. The first step to solving any problem is recognizing that it exists. In the case of HIT, this means acknowledging that clinical care consists of a series of specialized workflows. Clinical care consists of directed sequences of tasks that use, generate, or share information and that involve one or more individuals or machines. If the task sequence is wrong, the information incorrect or unavailable, or the wrong people, software or equipment involved, problems occur.
EHR system-clinical work impedance
The automation of clinical care with current EHR systems has resulted in numerous complaints from clinical professionals who are fed up and discouraged by systems that make their jobs harder to do. The number of workflow disruptions that occur as a result of EHR use should surprise no one. Disruptions were to be expected because EHR systems are archival systems that do not contain models of clinical work. Making matters worse is the fact that EHR systems have their own internal workflows. Consequently, a good portion of EHR training is spent helping EHR users learn to adapt their workflows to those of the software. Thus, training times are one hint of impending EHR system-clinical work impedance and attendant clinician misery.
The problem with current clinical workflow approaches
Beyond training, entire organizations have to adapt their processes to match the hard-coded workflows of EHR systems. Recent demands for improved system usability indicate the amount of misery that EHR-clinical work impedance has caused. However, the solutions proposed do not seem to grasp just how fundamentally important workflow issues are.
Understanding exactly how clinical work is disrupted or enhanced, whether using software or not, requires the ability to precisely model clinical work. And here is where the first major problem arises. Most modeling approaches used in clinical settings are based on flowcharts and swim-lane diagrams. Neither of these tools offers a means to capture all information movements, resource interactions and complex task sequences in one unified model. Unfortunately, attempts to improve the fidelity of workflow models often involve adding UML (e.g. state, activity) and other diagrams to the mix such as data flow diagrams. The key to better modeling is not more diagrams, but rather a single framework that allows for representation of all workflow concepts.
Obviously, detail is required to determine how the information needs of a nurse doing a patient intake are best mapped to his/her physical actions. This leads to the second problem: current workflow modeling efforts are usually insufficiently detailed for the processes they represent. This lack of detail is evident in workflow modeling training materials (see Clinical Workflow Analysis: The Value of Task-Level Detail ).
Workflow modelers seeking to add detail to their diagrams have another problem: there are no formal standards for encoding clinical work such as task names, step increments, information requirements, or notation symbols. Lacking such standards, models of the same workflows from different modelers will likely differ significantly. Such variation impedes learning and the progress of clinical informatics as it concerns understanding the interplay of clinical work and clinical care software (see Modeling Clinical Workflows and Processes). Fortunately, there has been a significant amount of workflow research in the last 20 years, and all of it can be applied to health care.
Workflow research outside of clinical care
Automation of business processes has been a major focus of workflow research in the computer science, engineering and business communities. While automating processes is a worthwhile goal, workflow research — especially that focused on creating process models — can also be used to study clinical care activities at the lowest levels and build abstract models for analysis. Automating a process before fully understanding it can be disastrous. If one is losing money because of bad billing processes, automating those processes “as is” will result in bigger losses. Thus, some degree of analysis and modeling are essential before any attempts at automation. Clinical organizations that have been successful in implementing clinical software get this point.
Workflow patterns have the ability to express every major aspect of clinical workflows. Yes, they must be adapted to clinical use, but the basic concepts stand. Even better, workflow patterns and Petri nets are based in mathematics, which assures that precise meanings and notations are possible. Again, some changes and adjustments are required for clinical care, but the required changes are not rocket science. Aside from workflow patterns and Petri nets, I will go out on a limb (not far) and state that every clinical workflow (e.g., task sequence, information movements, and resource interactions) can be represented by a combination of common mathematical objects–logic, sets, functions/relations, and graphs. It is time to take the rich legacy of workflow research from computer science, engineering and business and apply it to understanding how clinical care happens.
Workflow disruptions by any other name…
When caring for a patient, information is important. When and/or where in a care process information is required, collected, saved or shared are not simply usability issues, or safety problems, or human factors concerns, they are basic workflow issues as well. Addressing information needs requires close attention to workflow tasks and how they are sequenced. Who or what will participate in or complete a task are likewise workflow issues.
Given the importance of workflow to clinical care, the ability to expertly conduct workflow analyses should be right up there with being able to normalize a database or understand basic programming concepts as a training curriculum objective. In light of the fact that so many aspects of software selection/implementation, safety, usability, and CDS are workflow-based, clinical informaticists should be experts in analyzing and modeling clinical processes.
Clinical Workflow Center
While working on the EHR project at the 1917 Clinic, I ran into workflow issues head-on. First, I had the problem of mapping out clinical processes. This proved to be much harder than expected because there was no “best practices” guide to consult for clinical workflow modeling. The second problem was that I could not figure out a way to encode configurable workflows in an EHR. These problems gnawed at me for years until I found the work of Wil van Der Aalst. His research has profoundly changed my view of clinical care, clinical processes, and software design. Over the last few years, I have shared my intellectual journey with you. And now, having become convinced of the value offered by a formal approach to clinical work in terms of quality of care and clinician productivity, I hope to share even more.
As you know, I attempted to set up a workflow community on EHR Science, which did not work out, to put it mildly. I still think it is a good idea to have a place where everyone can share and discuss workflow issues, projects, and challenges. To that end, I have decided to set up a site dedicated to everything clinical workflow.
Aside from providing a community space, I have a few specific goals for the site. At the top of the list are 1) teaching clinical workflow analysis methods and 2) demonstrating the mathematical properties of clinical workflows and how those properties can be used to solve real-world problems. Since I consider workflow analysis to be a fundamental informatics skill, there will be tutorials, and if there is sufficient interest, a book (I have an outline). I am also keen to demonstrate and discuss the many practical uses of workflow analyses in software design, selection, and implementation. After all, use cases are workflows too.
At the outset, the site will provide a Q&A forum, resource page, and links to EHR Science. Long-range plans are more ambitious. At some point, the new site will accept outside content from those who have a passion for workflow topics, something that EHR Science doesn’t permit.
The resource page will have a wider variety of content than is seen on EHR Science. For example, conference and event-related announcements will be welcomed. Finally, I hope to engender more dialog between practitioners, researchers, and those who offer workflow products and services, so brief product announcements and reviews will eventually make their way on to the site. Community members may also submit content suggestions.
Of course, how the site evolves will depend very much on what those of you who visit have to say. So, please visit often and share whatever. The working name for the site is “Clinical Workflow Center.” Final testing and design are going on now. Look for it soon!