I reviewed an abstract for an article in JAMIA1 evaluating the use of dashboards as clinical decision support tools. The short story is that they generally demonstrated little or no benefit when used as standalone tools, and mixed results when they were included with other automation tools depending on the disease state.
I found myself thinking about my own experience with dashboards and what makes a dashboard likely to be useful. I have seen some that seem to be quite useful, while others not so much, sometimes to my surprise. Admittedly, most of my experience has been with dashboards as operational tools rather than as clinical tools, but I suspect that what makes a dashboard successful is pretty much the same in both environments.
I find it worth noting that there really are two kinds of dashboards:
- Public dashboards that are intended to display useful information, typically on a large screen. These are not interactive; they provide actionable semaphores to a population of observers. An example might be a dashboard of operating statistics for a work area that permit the users to readily understand the work ahead of them and indications that work may not be progressing.
- Personal dashboards that are intended to display summary information to a user and provide the ability to "drill down" into the detail as the user may find appropriate.
What follows is a list, in no particular order, of features of a dashboard that I think make them useful:
1) Readability - While this may seem obvious, I have seen far too many dashboards that were crammed with detail that required the user to do a lot of reading to figure out what was going on. In my experience, nobody has the time or interest to pour over an immense amount of text looking for something useful. Rather, a dashboard tends to be useful when it provides summary information that is readily discerned with very little reading. For this reason, I tend to favor graphic representations over text, especially for public dashboards.
2) Scope - A dashboard must present information that is useful for its intended audience. I have seen dashboards fail because their scope was too general and the population for which they were intended was unable to discern what displayed information applied to them, and what did not. In my early years in the workflow software business, I made the mistake of making a dashboard that was not sufficiently focused in scope. My customers replaced the dashboard with a display of the pharmacist checking queue; it told them at a glance whether there was work to do, and actually reduced the time between completion of work and checking.
3) Actionable - The information displayed needs to display information that is useful to the observer, which means that the information must be actionable. The observer should not have to infer what action needs to be taken by digesting a plethora of information. This, too, causes me to favor graphic presentations that can use size, position, and color to draw the observer's attention to places where action might be required. Ideally, for personal dashboards, the ability to navigate to individual items that require action, and to take the appropriate action within the dashboard create the most actionable dashboards.
4) Realistic place in the workflow - dashboards are generally not working tools. They cannot take the place of systems that are used to drive routine work. Rather, they are management tools intended to permit a manager to identify whether work is proceeding appropriately or falling behind. So, for example, a dashboard might signal that there are ADC stockouts that have gone unattended for too long, but it is unrealistic to expect that someone tasked with refilling functions would regularly consult the dashboard for stockouts that needed refilling. Other workflow tools should drive that process; people doing heads-down work are unlikely to regularly review a dashboard for what to do next.
5) Attractive presentation - this is taking readability to the next level; making the dashboard something that people like to use.
For public (read-only) dashboards, this likely means graphics that are fun, interesting, and immediately informative. In most cases, these dashboards communicate graphically, and users learn to gestalt2 the dashboard rather than read it. Changes in color, shape, or layout are noticed even before they are understood, and draw the user's attention to the detail, and, because it is fun, their eyes tend to be drawn to the dashboard more frequently.
For personal dashboards, this means that graphics are interactive, display multiple layers of information at a glance, and are easily navigated while maintaining context which can make these dashboards fun to use. Try looking at d3js.org for a host of interesting graphic presentations.
I once participated in an experiment where we presented an entire formulary in a bubble graph which displayed the formulary as a "heap" of bubbles whose color, size and position in the heap provided semaphores for which items should be of interest, and hovering over a bubble provided a plethora of information about that item (such as rate of expiration, current average unit cost, links to where it was located in the enterprise, etc.). What was fascinating was that people would look at it and immediately express feeling overwhelmed but would then mouse over the graphic and learn to navigate it in less than a minute, because it turned out to be fun. Some refer to this as gamification3.
I don't know what, if any of this, would have changed the outcome of the dashboards reported in the JAMIA article, but I have experience that dashboards that are readable, actionable, properly scoped and fun to use tend to get more use than those that do not.
It has also been my experience that there can be too much of a good thing, so white space is important.
What has your experience been? Have you had successes, or failures with dashboards that could contribute to this discussion? What do you think?
As always, the ideas expressed within this blog are my own, and are not necessarily those of ASHP or of my employer, BD.
Dennis A. Tribble, PharmD, FASHP
Ormond Beach, FL
1 Charis Xuan Xie, Qiuzhe Chen, Cesar A Hincapié, Léonie Hofstetter, Chris G Maher, Gustavo C Machado, Effectiveness of clinical dashboards as audit and feedback or clinical decision support tools on medication use and test ordering: a systematic review of randomized controlled trials, Journal of the American Medical Informatics Association, 2022; ocac094, https://doi.org/10.1093/jamia/ocac094
2 From the Oxford English Dictionary, gestalt is defined as an "organized whole that is perceived to be more than the sum of its parts". When used as a verb (as in this case) it refers to perceiving and interpreting an object as a whole rather than looking at its parts and reacting to changes in the whole image.