The Sheep, the Wolves and NLP. A Cautionary Tale.

At the Great ACE Tweet-up (woohoo!) in Austin, TX, we’re going to kick off our series discussing innovation for modern transcription.

We’re going to talk about the reality of keeping transcription relevant into the future – emphasis on reality.

I’m always a bit surprised at how easily our industry finds something that it thinks will be its salvation – its silver bullet. And we’re so ready to believe in a one-size-fits-all solution to all our problems.

But we’re not always so good at making that silver bullet a reality. We go through a lot of hard knocks before we get it right. Look how long it is taking many of us to adopt speech recognition effectively. Seriously – for many of us by now, speech rec should be standard operating procedure. Enough people have done it that the lessons are already learned.  So why do we have to learn the same hard lessons over and over?

That’s why I have concerns about the industry’s new focus on NLP as the means of staying viable into the future.

Just like speech recognition, NLP is not going to be the silver bullet that kills all our werewolves. It’s not a one-size-fits-all, plug it in and let it rip, solution. It will take thought. It will take planning. It will take effort.

Now don’t get me wrong – I’m not saying NLP is a bad thing! I’m saying we have to learn more about it – and more importantly– we can’t expect it to compensate for all of our inefficiencies the way we did with speech rec.

Part of what we will be doing in our sessions is – just like we’ve done with speech rec over the years – stripping away the hype and getting down to the basics of what all this change really means for the clinical documentation industry.

But here is the difference. With speech recognition, we had time. But we don’t have 10 – 15 years to figure out how to make the next saving grace a reality.

There are a lot of gaps that need to be filled before any solution, be it NLP or anything else, becomes a reality for us. There are a lot of claims being made that, I’m sorry, shame on us if we believe without question. I’m hearing some pretty fantastic claims about NLP “accuracy” that I hope we’re ready to give a good hard look at.…just like speech rec! Does anyone remember claims about speech rec in the 90s? Does anyone remember the buzz about how it was going to eliminate transcription?

Let’s ask ourselves some questions…

  • What IS the problem we are trying to solve? How will NLP solve it for us? What do we actually need NLP to do?
  • NLP is not as accurate and as easy to “plug in” as it would seem. In fact, there are MANY different NLPs out there. How do we know which to use for what purpose, and how?
  • If NLP does work for us, do we have the knowledge and skill to make use of it?
  • Do we have the workflow required to accommodate its use?
  • Let’s say NLP “codes” the document accurately and MTs validate it – then what? Is it getting to the EMR by ESP? Osmosis? Telekinesis?
  • Do we know what knowledge will be required for us to use NLP as the tool to take us from transcription to Meaningful Use and the EMR?
  • Do our MTs have that knowledge and skill? Do we?
  • And while we’re busy slashing MT line rates so that the highly skilled ones leave the industry – who will be left with the requisite knowledge to do all this validation and reconciliation?

In a nutshell – who is thinking about the reality versus the hype? Who is thinking about what we need to know to make this real? Are we relying on yet another technology to save us without focusing on what we are at the core?

That’s what is even scarier  – we’re so ready to give away the value of our expertise to the technology companies. Hey folks, technology doesn’t mean squat without the people who make it work!! Part of the reason for low adoption of the EMR is that technology vendors tried to tell doctors what doctors needed instead of listening to what the doctors really need. The people should be driving this boat – not the technology vendors!

Let’s not be sheep this time around ready to follow the first person to say they have a solution for us.

This time let’s ask the hard questions FIRST.

The moral of this story is… This time – Let’s be wolves.

-Lynn

Let’s Get Tweetin’!!

Who says transcription can’t be innovative? Not M*Modal and not AHDI!

Innovation is not a word typically associated with medical transcription. Despite many technological advances, the product delivered has not changed much over the years. Transcription has come to be seen as an expense rather than as something of value.

The situation is not helped by the fact that technology vendors often present their solutions to hospitals as the means to reduce or eliminate transcription costs. Transcription has come to be seen as a commodity and even the transcription provider seems to have forgotten its value.

AHDI and M*Modal will be collaborating in a series of conversations over the next few months to discuss these topics and more. We will be using several different venues, including webinars, articles, blogs, and online chats to talk about the following and more:

1. What is transcription and why is it still relevant?
2. Wanted: The Career-Minded MT
3. Managing for efficiency
4. Transcription Innovation in the World of Meaningful Use

Join us as we kick off these important discussions at The Great ACE Tweet-Up, hosted by Lea Sims of AHDI and Lynn Kosegi of M*Modal, at ACE 2010 in Austin, Texas, in the AHDI Lounge…Live!

Thursday, August 5, from 1 – 2 p.m. local time.

Visit http://www.ace365.org/page/the-great-ace-tweet-up for more information….

Speech Rec is here to stay…

Hello all: I wanted to take a moment to thank and send a shout out to the MTs who are following the blog and who are plugging away every day learning to become effective MT editors! It is WONDERFUL to hear from you and it is great to see you making such an effort even with the challenges facing you. THANK YOU!

I also wanted to send out a note to those who haven’t quite embraced speech recognition…

I see posts on other forums and I hear from MTs by email who still seem to think that speech recognition is an experiment which will just go away if they refuse to participate.

That’s the wrong way to go folks…hospitals and MTSOs MUST find more efficient ways to produce their work. MTSOs can NOT charge what they used to charge for a line. Hospitals do not have the transcription budgets they used to have. If they don’t use technology to become more efficient – they’ll sink.

Despite what you read on other forums, the MTSO isn’t adopting this technology because they want to pad their pockets by taking money away from the MT. Believe me – the margins grow smaller and smaller every day. The price per line the MTSO can charge today is significantly lower than it used to be. There is no evil scheme to harm MTs going on. Do the MTSOs always get it right? No – of course not. After all speech rec often is new to them too.

There is so much work being edited now – successfully – that there just isn’t a question anymore about whether or not speech rec works. Can the level of success vary? Absolutely. But it works, it is here to stay, and if you want to stay in this business MTs, you’ve got to learn how to edit.

Just like the MTs in the 80s who refused to leave their IBM Selectrics for the computer – you will be left out. You are not helping yourselves or other MTs by refusing or by insisting it doesn’t work when there are many millions of lines of transcription being produced with the help of speech rec every single day across our industry.

And you know, if you have no wish to learn to edit – that is certainly your choice! But lets please be aware that many MTs need their jobs and don’t have other options for employment. Lets stop making things more difficult for MTs who DO want to learn to edit by continuing to bash the technology and the companies who use it and the MTs who try to learn it. I am appalled by what happens on some MT forums when some poor MT dares to ask for editing advice. MTs who can not be supportive of their fellow MTs should be ashamed of themselves. They are discrediting the entire profession. How do they expect to be treated (and paid) like professionals when they don’t behave as such?

SPEECH RECOGNITION WORKS. We see many many MTs who prove this every day. Companies would not pay for the technology if it didn’t work. Lets be supportive of the MTs who must learn to edit efficiently in order to make a living and stop the negativity and lack of professionalism.

Folks, speech recognition is here to stay – it is no longer an experiment, it is no longer a question. Lets focus on addressing and correcting the challenges instead of hurting the profession for other MTs.

Transcription these days IS editing.

Thank you!
Lynn

Meaningful Use

Hello everyone: For a terrific explanation of Meaningful Use and how it relates to those of us in the transcription space, take a look at Jay Vance’s latest post on AHDI Lounge. Be sure to check out AHDI Lounge at the link below for the complete post.

NLP isn’t going to be the silver bullet that solves all of our problems – but recognizing that there are ways for transcription to stay current is tremendously important.

Thanks to Jay for keeping us all informed about this important topic!

-Lynn

Why Should Medical Transcriptionists Give A Darn About “Meaningful Use”?
Posted on Saturday, July 24, 2010 comments (4)
Labels: meaningful use, medical transcription, natural language processing, NLP,speech recognition
Writing in the New England Journal of Medicine, Dr. David Blumenthal, National Coordinator for Health Information Technology, talks about the new “meaningful use” guidelines that hospitals and doctors must meet in order to qualify for EHR incentive money. Included in the article is a table showing the core objectives that must be met as well as the criteria used to measure whether or not the objectives have been accomplished. I’ve cropped the table graphic to highlight one particular section:

See a pattern here? While much of this information is fairly standard for a typical patient encounter, stuff we hear doctors dictate all the time, every one of these measures requires that the information be recorded “as structured data.” Those three words are about to trigger a tsunami of change throughout the health information technology and healthcare documentation sectors, and the medical transcription industry is positioned squarely in its path.

Simply put, the term “structured data” refers to information in electronic form which is broken down into discrete, searchable data elements. In the case of a typical patient encounter, that means the sections of the report would be electronically “tagged” to identify them as representing the chief complaint, history of present illness, medication list, allergies, assessment, plan, etc. Furthermore, within each of those tagged sections, data can be further categorized so that individual words and phrases can be marked as representing a specific condition, vital statistic, medication, allergy, etc. The purpose of all this electronic tagging and sorting is to allow the relevant information to be captured and stored in structured form in an individual electronic medical record (EMR). The information in those individual EMRs would then become part of a larger electronic health record (EHR) representing the medical information of every patient in the United States. By having all of this information available in a structured electronic format, health data can be aggregated, mined, researched and analyzed in a variety of ways in order to, among other things, identify best treatment practices in order to realize best outcomes. (What other uses this information might be used for, benign or or otherwise, is a topic best left for another day and another venue!)…

Be sure to check out the rest of Jay’s post on this important topic…

MT Compensation

Last time we talked about the “percent gain” and how that metric is over-used and misused.

Today let’s talk about compensation.

Devising an MT editing compensation plan based on one across-the-board expected increase in productivity (often a 100% expected increase with 50% reduction in line rates) does not benefit the MT and does not make sense for the MTSO. MTs when typing have an incredibly wide range of productivities..how do we expect to pick ONE number upon which to base our comp plans when transitioning to speech recognition?

Here is the worst part…

Cutting line rates according to one expected percent gain in productivity will reward your lowest producing MTs, and will hurt your high producers! And your company will suffer in the long run.

Why?

Facts –

MTs who are low producers when typing will see a higher percent gain when editing. They have much more room to improve. They will see a pay increase because they are very likely to hit that magical “% percent gain” number.

MTs who are high producers when typing will see a lower percent gain when editing. They are already fast – and there is only so much that you can speed up an audio file and still produce quality work. We expect them to listen to the entire audio file, don’t we? If they don’t meet that expected percent increase, they will lose money.

Reality Check –

There is a number we should all use as a reality check point when determining if an MT can really achieve the productivity she’ll need for her pay to stay whole as she transitions to speech recognition.

At M*Modal we call it the Edit RT.

How many minutes does it take an MT to type one minute of dictated audio? An RT of 1 means the MT is typing or editing as fast as the doctor is speaking. She is editing in “real time.”

Average for traditional transcription tends to be around 4 to 4.5 for U.S. based MTs. That’s approximately 135 to 150 lph.

An MT who can transcribe from scratch at a rate of 300 lph (and they are out there) is already typing at approximately an RT of 2. For her to double her productivity, she has to get to an RT of 1. This means that you expect her to edit AS FAST AS THE DOCTOR SPEAKS for every job if you cut her line rate in half.

Do you want your MTs to listen to the entire audio?

You will hear talk about MTs averaging 600-700 lph when editing. Sure it’s possible…we see it every day. But this kind of speed usually means one of a couple things:

1. A good percentage of the text in these documents is introduced through Normals/Standards or template text. This means there are lines in the document for which there IS NO AUDIO. (think about this when you hear productivity claims from speech rec vendors)

2. The doctors are repetitive enough that the MT has created a good number of her own macros.

3. The MT is NOT playing back the whole audio and your quality is at risk.

HOWEVER, the MTs who are high producers with LOWER percent gain when editing are going to produce more lines, faster. Taking care of THESE MTs will allow your company to significantly increase its OUTPUT without increasing MT head count – and all of the subsequent support personnel and expenses.

Please remember this – your company is paid for its OUTPUT, not for that high percent gain!

This industry has seen an alarming increase in MTs who are not interested in transcription as a full-time career. What is happening to the career MT? How will the profession attract the quality-minded, dedicated professional if MTs can’t make any money?

It won’t – it will attract the person interested in making a bit of extra money working from home – the “hobby MT” as opposed to the “career” MT.

An MTSO who pays MTs for what they are – skilled professionals – will attract the career-oriented, professional MT. The MTSO who pays its MTs based on some arbitrary metric useful primarily for marketing for speech rec vendors will lose its professionals and will be forced to hire higher numbers of lesser skilled, less productive “hobby” MTs.

What does this mean to an MTSO’s finances?

What expenses do you have that are the same for every MT, regardless of her productivity?

How much does your company pay for:

• Turnover
o Training per every new hire?
o QA for every new hire?
o Support for every new hire?
• Supervisors and managers for every x number of MTs?
• Technical support people for every x number of MTs?
• Technology costs, foot pedals, software licenses, workstations, internet connectivity for every x number of MTs?
• Paid time off and health and welfare benefits for every x number of MTs?
• Administrative, HR, payroll personnel for every x number of MTs?
• Managers, directors, and senior managers for every x number of MTs and the requisite QA personnel, trainers, administrative personnel, technical support staff…..

Which MT is more expensive? Your low producer or your high producer?
• Which group has the highest turnover and subsequent costs?
• Who sends more work to QA?
• Who consistently delivers high quality work to your customers?
• Who requires more training, software support, and assistance?
• Who can you most rely on to work scheduled hours, to work extra when needed, and to help you meet your turn-around time requirements?

Your high producing MT is your biggest cost efficiency. Be creative – think beyond that MT line rate when you consider cost reductions.

Cut the line rates by a reasonable number – but use incentives or bonuses to keep your high producers whole! Use the transition to speech recognition to get rid of your dead wood – NOT YOUR MOST HIGHLY SKILLED EMPLOYEES!

Think about where that high-producing MT can save you money elsewhere in your organization.

Let me leave you here with one thought about what you’ll need to do as a company if you intend to stay in business as the dust from meaningful use settles and everyone finally comes to terms with the fact that there is no value in a paper document.

I once heard an MTSO executive say that his company would reduce costs by cutting line rates, allowing the high-producing MTs to leave, and then hiring a bunch of lesser skilled MTs at lower line rates.

Great idea – if you want to support three times the number of MTs you need to produce your volume, significantly increase your turnover and the cost of turnover, and if you have no plans to stay in business into the future.

We have all been talking about Meaningful Use and the EMR and the role transcription plays in that discussion.

The MTSO of the future will need to transition from a company that produces typed documents to a company that produces and VALIDATES meaningful clinical content.

I have heard a number of comments that MTs do not have the skill necessary to validate content in a meaningful way. I just heard that yesterday!

All I have to say is that a professional career MT certainly has an excellent knowledge of medical terminology, anatomy and physiology, and understands the subtleties of the meanings of words based on the prefix or suffix used, and the context in which they are used.

Come on folks – how many of us used to test applicants for MT positions who were nurses and coders and turned them DOWN because they did not have the wealth and breadth of medical language knowledge that a good MT has? But is an $8 an hour job going to attract a person with that level of skill?

Cutting your compensation rates such that you lose your highly skilled MTs may save you a few dollars now. But you will not be able to provide the level of service required of you into the future.

Thoughts? Would love to hear from you…

Lynn

When Metrics Mean Nothing – The Myth of the “Percent Gain”

I am a proponent of using metrics to measure results. My team and I encourage our partners to “manage by data, not by emotional response to change” every day. But sometimes a metric used gives a partial or even inaccurate picture of what’s happening.

An example of this is the use of the “percent of productivity gained” in transitioning from typing to editing.

I don’t object to monitoring the change in productivity – after all, isn’t that the point? And my objection is not that “percent gained” is used.

My objection is that “percent gained” is used as the driver for everything from measuring success to planning compensation. This is only one of many metrics that should be used – and it is not the most important one!

MTSOs must measure OUTPUT. “Percent gain” in many situations is NOT an indicator of output.

If your goal is a high percent gain I’ll tell you how to get your numbers. Keep your high-producing traditional MTs out of speech recognition.

You’ll see “percent gain” figures that would make any speech rec marketing person jump for joy. And what we all want is to make our speech rec vendor’s marketing materials look great, right? I hope so.

BECAUSE IT IS THE SPEECH REC VENDOR WHO BENEFITS MOST FROM ALL THE EMPHASIS ON THE “PERCENT OF PRODUCTIVITY GAINED” – NOT YOU!

Consider the following:

• You, the MTSO, are not paid by percent of productivity gained. You are paid for output – in lines, reports, VBCs, or minutes of audio – whatever. You’re paid for what you deliver to them.
• Increased output leads to organization cost efficiencies.
• Hospitals decrease costs by increasing output produced at a lower human resource cost. They sometimes reduce the work outsourced to YOU because they increase the OUTPUT of their existing in-house MTs!
• Don’t hospitals using speech in-house measure “percent gained”? Yes – but more tightly controlled in-house MTs tend to be employees – not “hobby MTs”. Everything from work assignment to scheduling is more controlled. The “percent gain” has a more direct correlation to output because there are fewer variables.
• “Percent gained” for an MTSO often does not have a direct correlation to production because:
o Variable schedules mean variable output.
o A high percent gain for an unsupervised MT with poor work habits still will not result in high output.
o The least productive MTs are those who see the largest percent gains, but the high producing MT with the low percent gain still creates the highest output.

In the figure below, let’s assume 150 lph is the average productivity for MTs typing for an MTSO:

• MTE 1 has seen a 125% productivity gain, and is producing the work equivalent to 1.5 traditional transcriptionists.
• MTE 2 produces the work of THREE average transcriptionists though she has only seen a 50% gain in productivity – because she was a highly productive typist.

And yet time after time, MTSO managers will be excited over MTE 1, and will ask what’s wrong with MTE 2 because they only look at the “percent gain.”

Over the course of a full-time year, MTE 2 with her 50% gain will produce >450,000 more lines than MTE 1 with her 125% gain!

Consider the costs associated with the number of MTs supported by your organization, including:
1. Supervisors and managers
2. Technical support staff
3. Trainers and training programs
4. QA personnel and QA’d lines
5. HR, payroll, and other administrative staff
6. Upper level management and supporting staff
7. Physical facilities, office space, furniture, equipment for all the managers, administrative and technical staff
8. Technology, software licenses, internet and telephone service
9. Paid time off and health and welfare benefits
10. Worker’s compensation and taxes

How much will your MTSO pay to support turnover as the highly-productive MT leaves and must be replaced by a higher number of less productive, less reliable MTs?

Hasn’t efficiency gained through productivity by the fewest required workers long been a must-have for any industry’s economic growth? Why is transcription any different?

This might sound as though I’m proposing that we WANT MTs to lose jobs through increased output. I’m not. Remember that “hobby MT”?

Imagine if we can gain enough financial benefit through efficient use of technology – meaning that we adopt efficient business practices to optimize the benefits – to pay the dedicated Career MT a better wage because we’ve decreased costs associated with supporting a high number of unproductive “hobby MTs”!

The MTSO that uses speech recognition technology to become lean enterprise-wide instead of LOOKING ONLY TO THE MT for the financial efficiency will be the one to exist into the future.

Next – We debunk the myth that compensation for MTs who transition from typing to editing must be driven by the “percent gain”….stay tuned.

‘Til next time,
-Lynn

Transcription world-wide

Hello all: I’ve spent the last couple of days talking to our transcription counterparts in London. Just a few items I thought might be of interest…

1. They don’t have a dedicated MT position – transcription is done by medical secretaries. The secretaries do transcription and perform many other functions including helping patients to navigate the NHS (National Health System) for the public hospitals, and performing scheduling and other support for physicians in the private hospitals.
2. They don’t do nearly the amount of documentation through dictation and transcription that we do. Some assume it is because the push for the EMR has been more successful in Europe than in the U.S. In actuality, in the U.K. they complete much of their documentation on paper forms including discharge summaries and operative notes. They would very much love to allow the doctors to dictate these types of documents if they could do so in such a way as to get the dictation into the EMR. Sound familiar?
3. They do a lot of letters and correspondence since patients are often referred to specialists through the NHS. Each patient has a GP (counterpart of a PCP in the U.S.) who refers patients to specialists when needed. Letters and correspondence are used by these physicians to keep each other informed.
4. Doctors like having the medical secretaries and the outsourcing of transcription is not a popular topic.
5. They don’t have a widely accepted and well-organized group like AHDI to represent them – but there are two groups AMSPAR and BSMSA which aim to serve the needs of medical secretaries, to define qualifications, and to support the profession. Take a look! They aren’t as widely known or accepted as AHDI – perhaps we should reach out across the pond and meet them!

-Lynn

%d bloggers like this: