How to Get Speech Recognition to Work for You
Making the Most Out of Automated Speech Recognition
Automated Speech Recognition (or ASR as it is known within the medical transcription business) was once thought to be the way to eliminate most medical transcriptionists from the picture.
Although ASR has done a fairly good job at slicing away a lot of the money that medical transcriptionists (MTs) used to make, as the speech recognition technology has evolved, one thing has become very evident.
Medical transcription (at least to this point) cannot survive without MT editors.
Nevertheless, the changes in the industry put most of us in a bind, because let's face it....doing medical transcription with voice or speech recognition is a whole different ballgame than how we transcribed or typed before!
Let's take a look at some of the changes. More important, I'll give you some tips that you can start applying today that will make you better at your job and also hopefully increase your paycheck.
Understanding Speech Recognition and How it Works
In order to get better at anything, you have to understand the mechanics of what you're trying to learn. Even if you've been doing medical transcription for 35+ years as I have, medical transcription seems to be in a whole new world.
For almost all dictators now throughout the medical industry, speech recognition is the dictation tool of choice. Rather than dictating into phones somewhere in a hospital, health care providers dictate from anywhere they can, on cell phones or other mobile devices. The resulting files are almost always run through a speech engine, converting the voice information into text.
That's where the MT comes in. As smart as the speech engines are, few if any documents come out perfectly. What the speech engine perceives as "text" and what actually is supposed to be text are often two entirely different things.
Although MTs are still needed, they haven't got the benefits from this new technology that were promised to them. In the beginning, MTs were told that their line counts would double, so that the decrease in their pay per line would be made up for by the increase in the number of lines that they could produce daily by "merely" editing. This has proven to be far from true.
Another disruption caused by this new technology is that some MTs have a much harder time than others adapting to it.
Increased Critical Patient Errors
Speech recognition was supposed to be a means to eliminate human error and reduce critical patient safety errors. However, sadly, it has done just the opposite! Critical patient safety errors have been shown to be increased by this technology, even though it has now been in place for several years.
Why? The reason is very simply stated this way. Traditional medical transcription was a procedure whereby the MT listened to the voice, thought about it, then typed it onto the page. They then reviewed it quickly, going on to the next bit of spoken words. Today, that has all changed. Now the words are already on the page for the MT and it is not a matter of "hear it, type it" but rather "read it, listen to it, verify it matches what the voice dictated" and then send it through as complete.
The problem with the new method of transcription or editing is that the brain doesn't always catch up with what the MT is seeing. The brain somehow blindly will read something and not see that it doesn't make any sense, while the eye is telling the brain that indeed, that information on the page is accurate. It may be accurate in terms of spelling, grammar, etc. but in fact, it can be totally, 100% incorrect. Consequently, if the MT editor does not catch the mistake, the error has now become part of a patient record. Frightening? A little bit!
Striving to Avoid Inattention
The major drawback to this new methodology for medical transcription is that it is extremely labor intensive. The most glaring problem seems to revolve around one human attribute that can rear its head unexpectedly even in the face of the most disciplined medical transcriptionist: inattention, or lack of concentration, whether momentary or over the course of the day many times.
In the nanosecond that it takes to look away from the screen to move your mouse, you can miss a word. Likewise, in the snap of two fingers, you can add a critical patient safety error to a report because you weren't paying enough attention when your mind drifted off.
There are some techniques that can help all MTs increase their accuracy while doing their job. From being an insider in this profession, I can honestly say that none of us set about to do sloppy transcription or to make errors. Most MTs are perfectionists, and more than anyone they want to see an error-free report. However, the potential for error is so high that it seems some days you're all thumbs or that your eyes are betraying you right and left.
How can we cut down on these errors then and strive for better quality when it comes to patient records?
ASR Tips and Tricks for the Medical Transcriptionist
First and foremost, in order to do a good job, anyone performing a job needs the proper tools. Consider investing in these if you don't have them already:
- Flat screen monitor—the largest you can afford—cuts down on glare and helps you see clearly what you are editing
- USB speakers with volume control. You can't hear what you can't hear properly, so invest in the best speakers that you can with volume control and the most variability in ranges.
- Proper lighting for your work area
- A working environment that is distraction-free: reduce noise and anything else that tempts your concentration to stray.
- Proper desk equipment including keyboards and chairs: you want to be comfortable when working for long stretches
ASR Techniques, Tips, and Tricks
Once you have the proper work environment, now you're ready to explore some of the tricks and tips that will keep you focused on your reports and hopefully also increase your productivity while decreasing your error rates.
- Synchronize the text cursor to the audible voice file. This means that the cursor should be following along from left to right as the voice file is playing in your headphones.
TIP: Do not take your eyes off the cursor for any reason - if you do, stop the voice file and resume when you are ready. Follow the cursor from left to right like the bouncing ball on Sing Along With Mitch from the 1950s. This is the best (and in my opinion only) way to not miss dictation! Most critical patient safety errors are errors of omission or not catching an incorrect word substitution. If you're watching word for word, that can't happen.
- Scan through the document before even putting your foot on the pedal and check such things as document type, dictator, patient information, dates, etc. Have a quick look through the document to see what may need to be changed and even change it before you start, such as headings, etc.
- If there are expansions or templates needed, add them before you begin to avoid forgetting to do so later. (But have a care that you only add what was spoken and remove anything else in templates, expansions, etc - this is another error prone area with the new technology)
- Concentrate, concentrate, concentrate. This type of work requires focused concentration to match the auditory file with the printed version and is the only way to eliminate errors. If your mind is wandering, you will make serious errors.
- A good rule of thumb for researching blanks is 3 minutes per blank. Take the time to review samples or use the Internet to research phrases. Build a library of difficult dictators or procedures and have them close at hand for references.
- Spellcheck! If it is not automatic and even if it is, do it anyway! There is no room in this day and age for misspelled words but especially in a legal document.
- Before you decide to send your report off into the EHR world, have a care and review the document before you submit it. Make sure one final time that you have produced the cleanest, most accurate document possible!
Practice Makes Perfect When It Comes to ASR
I recently read an article in Health Data Matrix about speech recognition technology that fascinated me. Here's a novel idea. The author, Kirk Calabrese, challenges all of us in the profession to begin a daily routine of practicing for one full hour to get better at ASR. How?
By learning keyboard shortcuts to eliminate or cut down on the use of the mouse. This not only increases your speed, but cuts down on the potential for errors as noted above. Every time you look away from your screen for a second and disengage from the text, you create the potential for an error.
It might seem cumbersome and like just one more thing to add to a long day, but in the end, if it increased your production and made your job easier (it is much easier on hands, wrists and arms to use keyboard shortcuts), wouldn't it be worth it?
Check out the Microsoft Word keyboard shortcuts I've found most helpful in transcription in my other article How To Increase Productivity, and practice until you can literally do these without looking. Keyboard shortcuts that help most are:
- Navigation (moving you around in the document without touching your mouse)
Keys like Home, End, Arrow
- Text selection (for moving text with keyboard commands, again mouseless)
Keys like Ctrl+End or Ctrl+Right
- Deleting or moving (to copy and paste without touching the mouse)
Keys like Ctrl+A or Ctrl+V
- Undo and redo - these will save you lots of time and keystrokes
Ctrl+Z and Ctrl+Y
Making Automated Speech Technology Work for You
In short, there are many things we still have to learn about ASR. There are many dictators for instance, who are just not cut out for voice recognition due to their dictating styles or their accents.
However, in most cases, with some effort, the speech engines can be trained to remember most dictators and achieve an end result report that is cleaner and cleaner as time goes by. How?
Because the MT editor by repetitively correcting errors that the speech engine made is actually training it to not make those errors in the future. But the rub is that we have to catch all the errors and do that over many, many reports to see an end result.
Many MTs feel beaten down by the evolution of speech recognition technology and feel MT is a dying business. On the contrary: at the present time, it is a thriving business, albeit somewhat frustrating. MTs are not seeing their wages go up but instead, they are seeing them decrease. However, with some of these tips and perhaps daily practice, you can make the system work for you and increase production.
The key is to maintain a positive attitude and be open to changing technology. And take total pride in every report that you do and refuse to turn in slipshod or sloppy transcription. In order to do that, you have to concentrate on each and every report 100% of the time and not rush through reports trying to meet production quotas.
Keeping in mind that the most important part of medical transcription centers around one entity—the patient—should help us all stay focused on our job, from the first moment we sign on to the last report we grab before we sign out.
Check out the videos below to see how speech recognition technology works. It is very interesting and I have no doubt that the technology will continue to improve.