Clinical Workflow Analysis: Next Steps

In the last post, I mentioned that I intended to complete a workflow paper that had languished for the last few years.  As it turns out, I made only a little progress.  This was not due to a lack of effort. It’s just that in going over the original outline and old notes, I discovered my perspectives on workflow analysis, software design, and implementation strategies had changed considerably.

A Change in Perspective, a Brief Delay
Years ago, workflow problems captured my interest when it became obvious, from personal experience and discussions with others, that workflow disruptions were the main cause of implementation failures.  Electronic Health Records, Second Edition, reflects my thoughts on the interplay between processes, product selection, and implementation success to that point in time.    Mulling over workflow issues since the book’s publication in 2008 (work on the paper began soon after),  has helped to further clarify the common elements underlying complaints about poor user interfaces, productivity losses, implementation failures, and  training times—all are fundamentally workflow problems.    Clearly, making life better for HIT users begins with tackling workflow mismatches between people and software.   Solve the workflow impedance problem in its myriad forms, and implementation becomes much less problematic.

In the last year or so, I have completed three quarters of a discrete math textbook, a few software architecture/design/programming  texts, and two books on workflow analysis using Petri nets along with many articles on all aspects of workflow.  Now, having gained a deeper appreciation of the challenges involved, and armed with more information, my take on possible solutions has changed, as has my idea of what constitutes a worthwhile paper. The paper will be delayed for a few months, but not indefinitely.

Some Good News
Even though the paper will be delayed, there is a silver lining to this story.   Every paper starts with a review of the literature.  PubMed, the Association for Computing Machinery’s Digital Library, and Google Scholar were used for background searches.   Having done the legwork, I’ve  decided to share my bounty and move ahead with a workflow resource page.  However, since I am interested in mathematical treatments of workflow in addition to clinical workflow analysis, it will take some time to review and categorize everything. Expect the initial version of the resource page within the next six weeks, maybe eight, as there are a few thousand citations to sift through.

Towards a Unified Framework for Clinical Workflow
Because of all the ways workflow issues affect HIT, there is a need for a unified framework that offers a single paradigm that cuts across boundaries (e.g., usability, product selection, training, implementation). Consider all the possible aspects of workflow that may influence the productive use of HIT. Workflow information can be used to:

  • Capture and profile clinical processes in their original form
  • Generate and refine software requirements
  • Map workflow needs directly  to software designs
  • Design usability studies
  • Set feature requirements for user interface configuration modules
  • Configure programmable WF engines

As I see it, the central challenge of HIT is this: Starting with an information management problem, what specific reproducible steps must be followed in order to create a clinical information system such that, when teamed with humans, their combined actions result in enhanced productivity, safety, and quality?    Today, this is an art, but it doesn’t have to be.

Looking at HIT implementation as the end of a complex process that begins as a human information management challenge, I am convinced that a unified framework is essential to understanding all the complex issues involved as a coherent whole.  In fact, pursuing such a framework has become a long-term research goal.

Fortunately, outside of health care, workflow, in the guise of process modeling and automation, has been a major research subject for years.  Petri nets are the perfect starting point for moving HIT implementation toward the ultimate goal of becoming a reproducible engineering task rather than the art it currently is.    Whereas a few years ago this seemed like science fiction, today, channeling Nate Silver, I’m 90% certain it is doable. A major reason for my optimism is that many of the proverbial dots all already in place; they simply need to be connected with health care in mind.   I’ll keep you posted…



Leave a Reply

Your email address will not be published. Required fields are marked *