Is Speech Recognition the answer to all your problems?

There are business practices that make sense to the MTSO regardless of speech recognition – but sometimes it is undergoing a big change like a speech rec conversion that uncovers existing strengths and weaknesses.

Lets start by listing some Truths about Speech Rec that many speech rec vendors won’t tell you – these are some of the things we’ll be talking about in more detail in coming posts.

1. Speech recognition will not compensate for poor work habits, inefficient management practices, insufficient training, poor quality, illogical pricing practices, unfair MT compensation, or weaknesses in the transcription platform.

2. Assuming you can use speech recognition to hire unskilled MTs at lower line rates will increase your head count, decrease your quality, and increase expenses – even if your “percent productivity increase” between typing and editing goes through the ceiling.

3. “Percent productivity gained” is not the correct metric to use to measure the results of speech recognition. It is not the correct metric to use to determine MT compensation.

4. Your highest producing MTs are those most likely to see the lower productivity percent gains. But keeping your highest producing typists out of speech does not benefit you. It does not benefit the MT. It only benefits the speech recognition vendor who can now use your high percent gain claims in their marketing materials.

5. For a global workforce, do not compare productivity for MTs expected to reach quality sufficient to deliver to the customer to productivity for MTs who are expected only to produce quality sufficient to send to QA. Again, this number is only meaningful for speech recognition company marketing materials.

6. High productivity means nothing if utilization (how much of your total volume you are editing) is low.

7. Don’t engage in business practices that cause you to lose your best MTs. Without them, you’re just another commodity. Given a choice, would you pay more for electricity produced by one electric company than you would for electricity produced by another? Who cares about the electric company as long as the lights are on? You all know which companies(y) will survive if you all become the electric company.

8. Speech recognition will never be perfect. Never. No matter how well “the engine learns” and how many times MTs make every correction. If speech recognition could ever be perfect, there would be no more medical transcription.

9. It is incorrect to measure effectiveness of speech recognition by using a “word accuracy” score. A draft that is 100% word accurate can require a lot of editing. Ask any knowledgeable MT if she’d feel comfortable delivering a 100% word accurate document to a doctor and she’ll laugh at you.

10. Errors in final delivered documents are not the result of speech recognition. These are MT errors. Addressing quality problems by removing work from speech recognition is just, well, silly. If an MT doing typing from scratch delivers errors to your customer, do you remove the work from transcription?

11. Willingness to fulfill every individual preference for format and style is not a differentiation point for you. If your value proposition to your customers is providing custom formats for every doctor, you will lose them to the next company to offer the same at a lower price. Anyone can manipulate document format. Not everyone understands how to deliver valuable documentation.

12. You will not see the same results when speech recognition is kept invisible as you do when processes are changed to accommodate speech recognition. Do not pay your MTs the same line rates for these two scenarios. Do not charge your customers the same line rates for these two scenarios.

13. The “You get what you say” – “type what you hear” method of implementing speech recognition is not “standardization” or “optimization” for documentation. It is optimization for the speech recognition technology. It is the furthest thing there is from useful standards that will drive quality, compliance, and usability.

14. Know what your customer wants, know what you want to sell. Price accordingly. Pay MTs accordingly. Every other industry in the world understands this concept.

Here’s an example.

I love my Jetta. Having a Porsche would mean nothing to me – so a car dealer who sells nothing but Porsches will never get my business. And a car dealer who sells Jettas at Porsche prices will be out of business fast.

On the other hand, if I ever break down and buy that screamingly beautiful Plum Crazy purple Challenger SRT8 with the v8 hemi from the dealership down the street – the car dealer darn well better not deliver me a Jetta.



Is Medical Transcription Still Relevant?

I found my favorite business article on, titled Dee Hock on Management .

A direct quote from Dee Hock:

“Substance is enduring, form is ephemeral. Failure to distinguish clearly between the two is ruinous. Success follows those adept at preserving the substance of the past by clothing it in the forms of the future. Preserve substance; modify form; know the difference. The closest thing to a law of nature in business is that form has an affinity for expense, while substance has an affinity for income.”1

What does this have to do with medical transcription?

It struck me as I read this article for about the hundredth time that this statement has special relevance to clinical information.

In all the debates about healthcare these days – about health insurance and costs and the EMR and meaningful use criteria – we all but forget what healthcare is all about, and what it is about our segment of the healthcare industry that is important.

Healthcare is about the interaction between doctors and patients. Everything that is done in healthcare revolves around the central fact that healthcare is purely and simply about doctors caring for patients.

We have a tiny piece of this of this interaction because we are involved in the creation of the documentation which helps to facilitate it. So in the broad spectrum of what we do – what is the “substance” and what is “form”?

We take information provided by the doctor and help to turn it into useful documentation.

But do we create the information? No. The doctor does. The doctor decides which information is relevant to the care of the patient.

Clinical information has endured from the time doctors began writing down what they observed about the patient, what they did to treat the patient, and how the patient responded. This information is the “substance” in this equation.

How we capture, communicate, store, and disseminate that information. Ah, now THAT changes all the time. That is the “form” in this equation.

Which of these documentation solutions preserves the substance of the information while “clothing it in the forms of the future?”1

EMR and templating systems –
• Often change the information by steering doctors through a series of text boxes and drop-down menus rather than allowing them to enter information according to intuition and knowledge. They may even change the examination process to fit the template.
• Doctors may limit the amount of information captured because of how much time it takes to do it.
• But the information is available for use in the EMR.

So yes, the EMR puts the information into the form of the future, but no, I don’t think we can say it preserves the substance.

Front-end speech recognition –
• Doctors may limit the information provided not wanting to take the time to correctly dictate (“typing with their tongues”) and then edit a long document.
• Behavior modification comes into play again as doctors change what they speak to words they know will be recognized. If the system gets a word wrong often enough, doctors will find a way not to say it.
• The speech recognition might be a type that can be used to point, click, and dictate directly into the fields of an EMR, achieving the goal of having the information available in the EMR.

So yes, front-end speech recognition may help to put the information into the form of the future – but the doctor falls prey to the same problems – limiting and changing the information he provides according to the structure of the EMR, and he might actually change the substance of the information to make the speech recognition work better.

Dictation and back-end speech recognition with editing by skilled medical transcriptionists –
• With the right speech recognition system and implementation, doctors can provide information the same way they have for years, without change.
• Doctors may provide whatever information they find useful, without being limited by the technology.
• Doctors’ time is not negatively affected since skilled MT editors will take care of turning the draft text into a useful document.
• Acceptance of standards such as those proposed by the Health Story ( allow the information to be captured, shared, and used in an electronic environment – putting the information into the form of the future.

So let’s see – which of these methods preserves the substance, which is the valuable information provided by the doctor, while “clothing it in the form of the future?”1

I only see one – how about you?

Hey, let’s look at another piece of this quote –

“The closest thing to a law of nature in business is that form has an affinity for expense, while substance has an affinity for income.”1

Since the combination of transcription and technology IS preserving substance while modifying form, and our technology-only “eliminate-your-transcription-costs” competition is preserving form while modifying substance, isn’t THAT fun! But that is a topic for another day….

1 – Waldrop, M. Mitchell. (October 31, 1996). “Dee Hock on management.” Fast Company. Retrieved from , December 18, 2007

%d bloggers like this: