A point-and-click electronic medical record (EMR) can be very fast, at least for simple, uncomplicated cases. However, some point-and-click EMRs try to convert the information from clicked checkboxes into English. The results, just like the corrections applied by a word processor spellchecker, or the misrecognition when you’re using speech-to-text on your phone, can be amusing. My recent favorite text was when I was at a Chinese restaurant after work and my wife, who was on her way home from something, texted me to bring home some tungsten fried rice. (It was supposed to be young chow = combination fried rice.) But my all-time favorite is what I first heard of as Ode to a Spellchecker.
Candidate for a Pullet Surprise
by Mark Eckman and Jerrold H. Zar
I have a spelling checker,
It came with my PC.
It plane lee marks four my revue
Miss steaks aye can knot sea.
I have used speech recognition for my medical charting for decades. Not all that long ago, we switched from Dragon Network Enterprise to Dragon Medical One (DMO). Overall it has been a significant improvement. DMO integrates with electronic medical record systems such as Cerner or Epic at the server level. This brings better recognition and makes it easy to use speech recognition to complete or amend charts from home.
However, one lasting complaint from my partners was the sign-in for DMO. When you start up Cerner or Epic, a separate sign-in dialog pops up. You have to put in your username, and then pick a vocabulary; every time we logged in, we had to change this from “General Medicine” to “Emergency.” Couldn’t we set this as the default and not have to change it each time? Finally, after many months, we heard that it would default to General Medicine for everyone and we should leave it there.
After this, I spoke with someone from our Nuance/Dragon support team about this. He explained it to me this way. Those different choices we had been forced to choose from? They did nothing. Absolutely nothing. It was, as he said, “a placebo button.” The vocabularies that we each were assigned, two of them, were set by Nuance when they set DMO up for us. Neither we nor his support team could affect this, and our prior choice of “Emergency” every time we logged in was totally nonfunctional: it did absolutely nothing. The IT support person told me that that button is still entirely nonfunctional, and so they asked Nuance to remove it. Nuance said that, due to the structure of the system, they couldn’t remove it.
We had discussed data pixels in another post. We also discussed anti-data pixels, things on the screen that distract you from the real data. Well, those pixels where you can pretend to choose a Dragon vocabulary but it does nothing, making you do work that does nothing? Or pixels in some other program that look like they do something but don’t? Or those that entice you into doing the wrong thing,or wandering off into dusty back hallways of the software. And are so enticing that you accidentally click them on a regular basis? Let’s call those anti-user pixels. Perhaps a more formal definition of anti-user pixels might be: “pixels on the screen that have not been removed, because they’re so tightly tied to the underlying code structure that taking them out requires major effort, or other reasons such as corporate requirements, that are not only useless and distract from data pixels but mislead and make a user’s job harder.” I challenge you to report other examples of anti-user pixels, or better definitions, in the comments section.
This relates to some other user interaction design and coding concepts that we can apply (with a little cutting and fitting) to this issue:
Information Hiding: in this context, if it’s not immediately relevant to the particular user or task, hide it so people can better see the forest rather than the trees, and are less likely to click something that will lead them astray and perhaps lost amongst the trees
Task/Work Process Analysis: in this context, analyzing users’ tasks and work processes and omitting needless pixels (like Strunk and White’s “omit needless words”), pixels that don’t contribute to the task at hand.
Encapsulation: in this context, presenting the user with a screen with the most common or best choices for the task at hand; and, hiding rarely or never used choices behind a single or at most a few less-enticing links: basically, keeping information related to the task together in one place and only presenting the user with that information. That requires knowledge and understanding of the various user tasks and work processes; see the two above bullets.
This also relates to an idea, introduced in the first edition of Alan Cooper’s groundbreaking 1995 book About face: The essentials of user interface design. It’s that a user interface should correspond with the user’s mental model rather than the coder’s/programmer’s implementation model.
The Master said: “When the noble man eats he does not try to stuff himself; at rest he does not seek perfect comfort; he is diligent in his work and careful in speech. He avails himself to people of the Way and thereby corrects himself. This is the kind of person of whom you can say, ‘he loves learning.’” [Confucius: Analects 1:14]
User interaction design is now the victim of fashion, or perhaps worse, fads. First, we had what artists call trompe-l’œil, which is creating a 3-D impression by skillful painting in 2-D. A great recent example is a painted concrete bridge in Frederick, Maryland. The computer equivalent of the Frederick Community Bridge is skeuomorphism. Some say that computer interfaces took skeuomorphism to excess. Many cite recent versions of Apple’s iOS operating system; others cite Windows 7’s “Aero” interface.
Most obviously with the Windows 8 “Metro” interface, an anti-skeuomorphic design fad took off. Some tout the “flat design” in Apple iOS 7. Others say that Apple stole the idea from Android, from the highly-lauded Windows Phone, and the Windows 8 “interface formerly known as Metro” (Microsoft removed the “Metro” label and now just calls it the Windows 8 interface; most of the rest of us still call it Metro.) UXmatters has an entire post about it.
Information Design is the art and science (or perhaps engineering) of presenting information so it can be easily interpreted without error. Sometimes it seems that the presentation of data in electronic medical record systems is the art and science of presenting information so that it is difficult to interpret and highly likely to cause error. There may be what seem on first blush to be good reasons for presenting data this way – responding to legal concerns, regulatory abreactions to specific medical errors that occurred, or that might occur – but the end result can be ugly and dangerous.
Let me give a specific real-world example from a few weeks ago, from an EMR that shall remain nameless to protect the guilty. This is the way the EMR reports a fingerstick glucose:
GLUCOSE [82947]
GLUCOSE [82947]: 331 mg/di Abnormal High (GLUCOSE NON-FASTINGRef Range:80 to 180 mg/di Critical Low:70 Critical High:400)
Fingerstick right 3rd digit. ac,rn
non fasting
