Dealing with Radiology ‘Orphan Issues’ in Voice Recognition Software

Regular readers of this blog know I am not the biggest cheerleader of voice recognition software. I resisted using it and when the places I worked at stopped using human transcriptionists, my last stand was typing my own reports. This was possible because A) I was pretty fast from having learned to keyboard at a ridiculously early age, and B) the rads working around me were less industrious so I could still outproduce them.

That didn’t hold up when I switched to per-click teleradiology. One of the unhappy adjustments I had to make was keeping a constant eye on the dictation screen so I could catch all the stupid errors the software made. I swiftly learned that I would notice/correct more of them if I did so as I generated my report rather than waiting till my report was done and trying to proofread the entire page(s).

Over a decade later, I have gotten somewhat comfier with the latter option, even if means more dumb mistakes slip past. We now live in a world where everybody uses voice rec to some extent, even if just for instant messaging on their cellphones. You don’t have to be a radiologist to see a transcription error as a sign of flawed technology rather than a dumb/careless user.

However, I still catch plenty of stuff. This past week, I caught a routine offender, one I have been seeing since I first caved in to use voice rec in 2011. I was dictating about some anatomic structure that was “less” than previously (I don’t call the size, density/intensity) and the software unhelpfully wrote “LEs.”

With the two capital letters, I always assumed that was some kind of abbreviation, but had no idea for what. Further, I figured pronouncing such a thing would be more like “ell ease,” not “less.” It would have made more sense to me if it transcribed the name “Les” (as in Lesley). But multiple different voice rec platforms have always insisted on the LE capitalization.

Once or thrice, I tried punching the letters into Internet searches to educate myself, but never came up with anything. The vast majority of the time, dictating “less” works fine, and some platforms let me make a “replace” rule so the software will substitute “less” whenever it wants to write “LEs.”

Still, it nagged at me. I hadn’t previously risked showing my ignorance by asking other rads what the heck LEs means, but this time I decided I needed to know before I retire or die. I put the query on our group messaging app and nobody else knew. (Nobody had experienced the LEs voice rec issue, either, but that is another mystery entirely.)

All of which had me thinking that if zero percent of an admittedly small group of surveyed rads know of this pseudo word, and I never found an explanation for it on the Internet, why the hell is it in multiple voice rec lexicons? Who put it there?

In the absence of a “parent” to take ownership/responsibility, the pseudo word is kind of an orphan, and it is not alone. There are more than a few orphan issues in our field: These are things whose origins and function can’t be explained and which, once identified, somehow manage to persist even though nobody wants them nor can think of anyone who ever did.

A lot of things seem to be orphan issues but are not. They might have no utility in the eyes of those they routinely impact — radiologist for example — but do some digging and you find out that the issues were born on behalf of others. Why are we doing imaging procedure X in a given scenario when we all agree Y is more appropriate? The surgeons or ER docs wanted it this way.

The deeper you have to dig to find the issue’s parents, the more likely you might be to wrongly conclude it is an orphan. Suppose the procedure X vs Y matter was determined 20 years ago, before Y got better, faster, and generally more feasible, but nobody ever revisited it. The issue’s original parents may be long gone and if you ask, “Why are we doing X,” the best you might get is “That’s the protocol.” Organizational inertia might still resist any effort you make to change things.

Differentiating those pseudo-orphan issues from the real ones might not always be possible, but some things ring truer than others. I get a mixture of entertainment and intellectual stimulation from rooting out the real orphans. I imagine evolutionary biologists might feel the same way when they try to figure out how the heck things like platypuses (yeah, I know, “platypi” sounds better, but it is wrong) came about.

Things don’t just happen, even if they got started for no good reason or by complete accident. There had to be some sort of causation. With orphan issues, pinpointing the origin might be impossible, but it’s fun to think about how one person’s stupid mistake or self-serving behavior somehow managed to propagate itself through the years and become entrenched in the minds of other people and entire organizations.

A more idealistic or energetic individual might take an interest in all of this for the purpose of improving things. If manpower or other resources are being wasted on policies/procedures that nobody can justify, what is the argument for going on with such wastage?

I have got enough on my professional plate and sufficient use for my downtime that I am not looking to tilt at any extra windmills. It might otherwise be amusing to inquire with the various voice rec vendors I have used over the years to ask them: What is this LEs for?

I know what would almost certainly not happen. The industry would realize that it has nonsense in its systems, purge the offending entry, and maybe launch a genuine effort to annually review their lexicons to remove anything that didn’t belong within these systems.

Instead, if they got back to me at all, it would be some low-level support type advising me how to do the “replace” rule I have already been using, or to delete LEs from the lexicon on my system. In other words, they would do nothing at all to address the orphan issue, which would continue to survive for the rest of the world while I, personally, used a workaround to avoid seeing it.