Research Informatics.org

It seems that in the biomedical informatics community, we have a tendency to focus more on the technology that can be employed to improve a particular domain, rather than the “process” that we are trying to address. If we look to the publications and best practices that are available in the business and social science communities, there is a plethora of methods and approaches that can be used to support the study and optimization of such “processes” - often referred to as “workflow.” So the question is - why don’t we see this intersection being leveraged more frequently? Based upon some recent pilot projects that I have been involved with here at Ohio State University, I would suggest there are a couple of major reasons: The basic meaning of what constitutes “workflow” is not consistently understood or shared throughout the informatics domain. For those of us with more computationally-centric backgrounds, the term is used to refer to the orchestration of multiple computational processes, agents, or components. In contrast, for those of us with social sciences-centric backgrounds, the term is used to refer to a holistic understanding of the major players, artifacts, and activities that exist in a domain, and their relationships to one-another in a particular problem area. The tools available to support the observation, modeling, and analysis/optimization of workflow are diverse, and not always easy to evaluate or adopt. As with many “standards”, the real benefit of most workflow analysis approaches is that if you don’t like one of them, there are always numerous others to chose from. The literature is rich with descriptions of ethnographic studies, time-motion studies, key participant interviews, UML activity diagrams, “swim-lane” diagrams, and workflow simulation studies. Unfortunately, most of these reports assume a relative familiarity with their methodological approaches, making it difficult to quickly ascertain their applicability to ones own projects without extensive correlative research on those methods (if you can first find the correct resources from which that research can be extracted). One of the biggest issues with this lack of consistency and explanation is that methods and results from potentially complimentary studies are rarely compatible or interchangeable with each other. Imagine if everyone used a different way of calculating and presenting statistical significance in their publications (well, in all honesty, this is sort of the case, but I am not a statistician, so I will hold my comments on that topic for a later post)? That is sort of what reading the literature in the workflow analysis domain is like. Combinations of qualitative and quantitative approaches to workflow analysis seem to be rare, and in many cases, discouraged. The communities who frequently engage in qualitative workflow analyses often argue that quantitating such models will only lead to erroneous assumptions and conclusions. Similarly, the communities who frequently engage in quantitative workflow analyses seem to be solely focused on quantitating a domain, with little regard to the complex interplay between actors, artifacts, and activities. It seems to me that we need to find a middle-ground, such that these approaches can be combined and systematized. In all honesty, a fully qualitative workflow model is of little assistance on its own if we wish to instrument, optimized, or further evaluate a domain. Developing a better understanding of how qualitative workflow models can be linked in a efficacious manner to quantitative measurements of the domain under study should be a high priority goal for our field. And finally - how does this even relate to the domain of translational informatics? Well, from my perspective, it seems unlikely that we can be effective or successful in developing and deploying informatics platforms intended to address the needs of the translational research community if we do not first understand: 1) what the current workflow of our end users is; 2) where there are opportunities to optimize that workflow using informatics interventions; and 3) how we can design such interventions to achieve the targeted workflow optimization without disrupting other areas of the contributing workflow. Such an approach is consistent with about every well validated software engineering approach known - all of which aim to ensure that new software platforms are accepted and well utilized. The real challenge is not deciding to use such workflow or participant-centric approaches, but rather, the actual “nuts and bolts” of doing so in the context of complex biomedical workflows. The alternative approach is the “if we build it, they will come” model, which I think goes without saying and being less than ideal!