Does this data make me look fat? - The Aesthetics of Raw Information

Technology can only take us so far. At a certain point, technology becomes powerful enough that a user will simply be unable to keep up with the outflow of information and data (it's hard enough keeping up with my Twitter stream). Without parallel development in data presentation techniques and visualization, it is easy to become overwhelmed by raw data without knowing what it actually means. Furthermore, when you add the human element to the mix, it becomes even more complicated.

Because most hospitals, in the interest of getting reimbursed, require certain elements (eg., review of systems, labs, etc.) to be recorded in the patient history and physical, most electronic health records (EHRs) are able to pull in objective data from other parts of the EHR and coalesce it into the H&P. Therein, however, lies a dilemma. Much like the alert fatigue phenomenon, there may also be "data fatigue." Sometimes, a progress note contains so much raw information, it paradoxically obscures the overall picture of what is going on with the patient.

For example, take a look at this sample template of a H&P; the template alone is 281 lines! Now imagine that being populated with data. This is all before the provider even types a single letter into it. For a patient who presented to the ED with a CC of hypotension and the A/P  consisted of "dehydrated; gave IVF and D/Ced", it becomes silly why the progress note would look like a senior thesis. Of course, in an era where most progress notes don't include enough information, this observation may seem absurd to some. It becomes pertinent, though, because it seems there is an inverse relationship where healthcare providers write less the more an EHR imports data. As I alluded to in my other post, many providers could use a copy of the Elements of Style along with their Pocket Medicine reference. I'm not exactly expecting Pride and Predjudice, but basic correct spelling, grammar, and coherence should not be regarded as a luxury. As the amount of raw information increases, people tend to rely on it as the "meat" of their note. An intelligent physician once told me that the progress note should tell the "story" behind the patient. This necessitates that we improve ways to deliver information.

Even in hospitals which are considered "high tech", I suspect they lack variety in data visualization and presentation. This is potentially an area where health IT may learn a thing or two from innovators in the field of data visualization research. Chris Harrison from Carnegie Mellon University offers various cools ways to visualize huge amounts of raw data (eg., 100GB of Google search terms). Also well known is Edward Tuft, who is an expert in information design and visual literacy. His musings in his essay "The Cognitive Style of PowerPoint" is fascinating. The techniques discussed have limitless possibilities in healthcare. Looking at Probiotics.org's infographic on antibiotic resistance may also give inspiration to the possibility of producing hospital antibiograms in a more visual manner. Most antibiograms now are probably in an Excel format.

To theoretically apply this to clinical practice, let's look at the challenge of characterizing a patient's hospital course in real time. Often, we want to know their vitals, labs, how they're feeling, when they are getting a procedure, etc. Currently, the only way to obtain this information is to read their chart. The data visualization is solely in plain text format.

To mitigate this, I propose a mockup utilizing an old data visualization technique perfected by Charles Joseph Minard. Charles Joseph Minard, in 1869, produced one of the first flow maps which illustrated Napoleon's army during the Russian Campaign. It is highly regarded by many, including Edward Tuft; he calls it one of the best statistical graphics ever. The following is a flow map that can be incorporated into an EHR coversheet and characterizes the patient's length of stay. Following the map, it is a two dimensional plot from a patient's admission to discharge. Along the way, interactions with hospital staff are plotted as intersectional nodes which bisect the patient's course. The utility is that it is scaled with regards to time and whenever providers consult each other or review labs, that is logged as well. This gives a "macroscopic view" of the entire hospital course. Clicking any of the "intersections" would bring up the progress note associated with that interaction. (Click to enlarge image).

Possibilities like this is why I am excited to see what is ultimately submitted for the National Library of Medicine App Challenge. With a rich repository of different APIs and data, there's got to be some cool ways to visualize everything.

Alright, now who wants to program this?

The above map was created with Stanford University's free Flow Map program. Read their accompanying paper here.

Posted via email from RxINFORMATICA