EHRtv

This Video requires Flash Video Player

Nuance Communication is the largest provider of speech recognition in the world. Keith Belton, Senior Director of Product Marketing for Nuance Communications speaks to Dr. Eric Fishman about Nuance’s product development in the healthcare sector. The most popular product with over 150,000 physician users is Dragon Medical. The Dragon Medical software utilizes speech recognition to help physicians improve clinical documentation and communications.

Mr. Belton and Dr. Fishman talk about Nuance Healthcare’s most recent version of Dragon Medical 10.1, which has been improved to limit training time, utilize expanded medical specialty vocabularies, and recognize many different language accents and dialects. Mr. Belton further discusses new products and innovations, such as Mac Speech, the Dragon iPhone speech recognition application, and a new prototype of Dragon Medical for the iPhone that was showcased at HIMSS10.

Category: HIMSS10, Tradeshows
Date: March 11, 2010
Views:1,864 views
Information:

Dr. Eric Fishman: This is Dr. Eric Fishman: for EHRtv and today, we're discussing the industry of speech recognition with Keith Belton, the Senior Director of Product Marketing for Nuance Communications. They're aren't many people who know more about speech recognition than Keith. Keith, thank you for joining us today.

Keith Belton: Thanks, Eric, thanks, a pleasure. Thanks for having me.

Dr. Eric Fishman: I appreciate your time. So let's talk about Nuance which is obviously the leader in speech recognition industry. Tell us how it got started and what your involvement is there.

Keith Belton: Nuance is the largest provider of speech recognition in the world. So if you voice dial on your Blackberry or if you call United Airlines and get their automated attendant or if you program your TomTom or your Ford SYNC, you're using speech that's Nuance. We're in pretty much every computer experience or automated telephone experience in the business. The Nuance Healthcare business is in the business of providing speech recognition to improve the documentation and clinical communication of patient information generated by physicians and nurses.

Dr. Eric Fishman: Does the telephony speech recognition use the same underlying patents, the same intellectual properties as the, for instance, Dragon Medical and e-Scription products?

Keith Belton: It does. Nuance has about 600 speech engineers and over 1,000 patents and so much of the technology that's found in the telephony is the same similar technology. The particular speech recognition used by physicians is a much more complex technical challenge because you're dealing with vocabularies of tens of thousands of words where of course, if you call Dell computer, hopefully you're always saying six or eight things - fix my computer, pay my bill. So it's similar technology but the medical challenge is probably the most daunting one for us speech scientists.

Dr. Eric Fishman: So maybe spend just a moment discussing the dichotomy between speaker independent and speaker dependent speech recognition.

Keith Belton: Speaker dependence and independence refers to how much knowledge the system needs to know about your speech and how much the system trains. When you call United Airlines or Dell Computer or 411 using Nuance, the system is speaker independent which means it doesn't adapt to the speaker, it doesn't know who the speaker is. With speaker dependence, the systems used by physicians today adapt to each user. So if you're Dr. Fishman, you log into one of our speech recognition systems, you identify yourself to it and as you use the system, as you correct the technology, it's learning for each physician. Each physician's profiles are analyzed and stored and over time and use, the recognition gets better and better.

Dr. Eric Fishman: How much training or usage does it take to have it be usable at this point and time?

Keith Belton: With Dragon Medical, which is the main product we're featuring at HIMSS today, there's really little or no training. Dragon version 10 is accurate in the high 90s out of the box and there's a number of things that we've done to make it very, very accurate. We have about 80 medical subspecialty vocabularies so you, as an orthopedist, would choose the orthopedic vocab. We have rheumatology, cardiology, various flavors of pediatrics, surgery; so pretty much for every medical specialty, there's a vocabulary for Dragon.

Dr. Eric Fishman: As you know, I've been involved in Dragon for quite a number of years and continue to be involved in it. We're up to version 10.1 and soon there will probably be other versions. Tell us what new features have come out in Dragon Medical in the recent past and if you're able to talk about the future, I'd love to hear that as well.

Keith Belton: Version 10.1 is the current version that we're shipping. The major advances with 10.1 were the expansion of the vocabularies to really cover the waterfront in terms of physicians both in primary care and specialty care. One of the other major advances that we've announced is recognizing that about 30 percent of all physicians in this country were not born in this country. Many of them have accents. You don't have an accent because you're from Florida, I do because I'm from Boston. Dragon Medical has an Indian - you have to, of course, be politically correct these days - but as an Indian sort of Southeast Asian dialect, it's got an Asian which covers Japan, China, Korea. It's got various flavors of English accents, one for the UK, one for Australia Kiwi, it's even got a Deep South accent. We were thinking about building one for New York but we found that it was challenging just because of the five boroughs.

Dr. Eric Fishman: And a Hispanic accent.

Keith Belton: And a Hispanic accent as well. So you identify the Dragon when the physician starts using it, not only what specialty you use but also what language model that's appropriate, what Regional Accent Wizard as we call it.

Dr. Eric Fishman: Have you found that to be helpful for the 30 percent of --

Keith Belton: Enormously helpful, enormously helpful because what it's really done is it's made that Indian accent speaker or that Asian accent speaker, who might previously out of the box have had maybe 80-85 percent first pass accuracy. They start in the mid to high 90s just at the same time so it makes an enormous difference.

Dr. Eric Fishman: We've all heard about front-end speech recognition, back-end speech recognition. Maybe you could describe the differences there. It would be very helpful.

Keith Belton: The basic challenge with physicians is you've got eight physicians in a room and there'll be nine different ways of documenting care. So understanding the clinical work flow - how the physician, in the course of providing care, documents where they document and what their style is. Are they using an EHR? So there are two basic work flows which Nuance supports with the speech recognition. One, as you mentioned, is back-end speech recognition. Back-end speech recognition essentially takes traditional transcription and puts it on steroids. So a physician would typically be dictating into a transcription tank. The way that the Nuance speech recognition works is after the physician hangs up the phone, instead of the voice file going to a transcriptionist somewhere either in their office or overseas in India, instead of the transcriptionist listening to the dictator and typing word for word, Nuance runs that voice file through a speech recognition system offline - Dragon - and then the transcriptionist is presented with the draft of the physician's dictation. So the transcriptionist is basically becoming an editor so that doubles the transcriptionist's productivity and can really reduce the turnaround time. So it takes a lot of cost out of it.

Dr. Eric Fishman: Does the physician even know that that's happening?

Keith Belton: Absolutely no change. That's really the big change is that it doesn't change the physician work style.

Dr. Eric Fishman: The physician will still pay usually a per line transcription.

Keith Belton: Per line transcription but it's about 50 percent of what they would pay if they were not speech enabled by Nuance. So that's a tremendous improvement.

Dr. Eric Fishman: So the transcriptionist passed through that savings?

Keith Belton: Exactly, exactly. So systems that we provide that back-end speech recognition include our traditional dictaphone Enterprise Express solution that's in use by about 2,000 hospitals. We have the hosted version of that called iChart and more recently, we acquired a company called eScription which is the leader in that space which is again, another back-end speech recognition that's been best in class by about six years. We provide both a capital based model on premise, either through dictaphone Enterprise Express and then the ASP Internet based version called eScription.

Dr. Eric Fishman: Just to bring it back, we've been discussing the back-end speech recognition where the physician may not even know that Dragon, for instance, is being used, but now we're looking at the front-end speech recognition where the physician actually talks using Dragon Medical. Explain how that works.

Keith Belton: In the case of the front-end speech recognition, the physician is logged into their EHR and they're presented with creating a patient record where they're needing to put information directly into the EHR. In that scenario, there's really only three options for the physician. They can use back-end speech recognition where they dictate into a phone and at some point the transcription is completed through speech, through the typist and then it has to be uploaded into the EHR. The challenge with that is the turn around time is basically a function of how many transcriptionists. It can be hours or it can be days and in today's fast paced medical environment, whether you're in primary or urgent care --

Dr. Eric Fishman: Hours are required.

Keith Bolton: -- hours are required. It's not an acceptable time frame. The second is if physicians can type, but you didn't go to medical school to type and most of our physicians who use Dragon don't either. The third version is really Dragon and the power of Dragon is two-old. Number one is anywhere the physician can type into the EHR, they dictate and the words appear, as you say, on the screen in real time for the physician. The other powerful version of Dragon is a lot of times physicians complain about as wonderful as EHR's are, the fact that it takes me 19 clicks to get from my HPI screen to my lab screen --

Dr. Eric Fishman: So Dragon also provides command and control?

Keith Belton: Exactly. So Dragon allows you to say show me current meds. Dictate a script and it will automatically execute in the background the mouse clicks that it requires to take you through it. We actually have physicians who say that Dragon has helped save them from carpal tunnel because it makes it easier to navigate.

Dr. Eric Fishman: I was once writing a book on carpal tunnel syndrome. I've operated on many of them and I was typing it and started developing carpal tunnel syndrome and started using Dragon for that purpose.

Keith Belton: Physician, heal thyself, yes.

Dr. Eric Fishman: It was very helpful. Maybe you can discuss how many physicians using Nuance products are using Dragon as opposed to back-end. And of the physicians using Dragon, how many of them, if you know, are using it inside an EMR, outside an EMR, how many are actually using it self-direction style et cetera. Do you have that data available?

Keith Belton: We have fairly good data. There's about 700,000 physicians in this country and about 250,000 use some flavor of Nuance speech recognition. So it's just over a third.

Dr. Eric Fishman: A third, that's quite impressive. I guess 160,000 is the most recent number I've seen for Dragon Medical.

Keith Belton: Yes. We think that the Dragon number is about 150 to 160,000 and they use all the EHR systems. Everybody from our largest customer, who is the U.S. military who's got upwards of 10,000 physicians, to folks like Epic and Allscripts and Cerner but even down to the single practice physician again, the benefits of Dragon are specific to each physician. There's, as you know, 300 EHR's and I'm sure that probably each of those EHR vendors has at least a few Dragon users.

Dr. Eric Fishman: Do you find that Dragon is compatible with essentially any, I was going to say Windows-based EHR but now maybe it's any Windows or Mac-based EHR?

Keith Belton: Any Windows or Mac-based EHR.

Dr. Eric Fishman: Tell us about Mac Speech.

Keith Belton: Mac Speech is a version of Dragon that runs on the Macintosh. It has a general medical vocabulary. We're still kind of absorbing the assets of the company and meeting with their management team but obviously, our intention is to be able to provide the same vocabulary set that's currently available for the Windows-base over time to the Macintosh base. We're pretty excited because Macintosh has long been a favorite of the physicians.

Dr. Eric Fishman: On the West Coast.

Keith Belton: On the West Coast as well as the East Coast. I'll talk a bit about the relationship we have with Apple and it's some very exciting news that we're also announcing. We're showing it today.

Dr. Eric Fishman: We can get to that in just a second though. Just one more question on Mac Speech. Is it as compatible with various electronic health records as Dragon Medical is?

Keith Belton: Yes, the same compatibility exists. Basically, if you're using Mac Speech anywhere within an Apple-based EHR, you can type, you can dictate and the navigation commands work the same.

Dr. Eric Fishman: Great. Now, iPhone. I hear it'll be coming out, very exciting so let's hear about that.

Keith Belton: Yes, very exciting. Last fall, Nuance announced a version of Dragon for the iPhone for non-medical users and so that product is available. I think there's upwards of a couple million downloads on the iStore so it's been a pretty hot product.

Dr. Eric Fishman: Still giving it away for free?

Keith Belton: I believe we're still giving it away although we're figuring out kind of the business model and the capitalization model to fund the R & D which we need to keep making the products great. So we're showing at our HIMSS booth a prototype of the Dragon Medical for the iPhone.

Dr. Eric Fishman: That's an exciting product.

Keith Belton: It's very, very exciting. A number of our EHR partners like Epic and Allscripts and Cerner have iPhone-based clients and so it's very, very powerful. It has a general medical vocabulary. You can dictate, the voice file goes to our voice server and within a second or two, it comes back with the dictation. The dictation can be as long as minutes and it's been pretty powerful and we're pretty excited. We think it's going to be a lot of buzz at the show.

Dr. Eric Fishman: This isn't back-end speech recognition like you described a few moments ago in which you've got a human being editing, this is machine.

Keith Belton: Right. This is front-end, real time based speech recognition. The only difference is from an architectural standpoint. From a speed standpoint, it's front-end because you dictate and words come back. It's actually back-end technology in that the speech recognition is not taking place on the iPhone unlike the Dragon Medical desktop product where it's being done on your laptop or PC but the voice file is actually going to our voice servers and coming back in your real time.

Dr. Eric Fishman: Then it can be cut and pasted from the iPhone into Word, into an EHR?

Keith Belton: Yes, yes it can.

Dr. Eric Fishman: Then correction can be done there?

Keith Belton: Correct, that's correct. One of the challenges of the iPhone is it's not a multi-tasking system so you're basically are left with sending it as an SMS or sending it as an e-mail but we're pretty confident Apple with solve the single threaded issue soon. I hasten to point out that the product is not commercially available yet, it's in demonstration mode. We're beta-testing with some of our large customers so our hope and expectation sometime in calendar 2010 the product will be available for our physicians.

Dr. Eric Fishman: That's an exciting piece of information.

Keith Belton: Very exciting. We’re pretty excited about it.

Dr. Eric Fishman: Let's change gears to meaningful use. It's an important thought on many physicians mind, $44,000 is not a trivial amount of money and there's obviously been a lot of discussion about the physician attitude. Now obviously, December 30 the draft regulations came out and the narrative was not a meaningful part of the meaningful use regulations. What is your opinion and what's happening in that regard?

Keith Belton: Well, it's an interesting philosophical discussion. The committee stated to us that one of the reasons the narrative was not considered required in the draft regulations was because they felt that dictating a note was a required element for being a physician. So I think they felt that it probably wasn't worth the time to spend.

Dr. Eric Fishman: Yet in 2010 tens of thousands of physicians don't dictate their history.

Keith Belton: That's right. Because physicians may be so focused on ensuring that they're cutting and pasting or point and clicking the key elements of their meds list and the problem list and ordering, there's a concern I think in the medical community, at least in some chapters, that the quality of the note could be somewhat compromised. We’ve had many of our physicians who have actually written to their congressman saying you're really kind of missing the boat. That as wonderful as an EHR is, if you rely on the point and click templates, the notes start to look like each other. They get to be cookie cutters.

So we did a survey a few months ago that showed something like over 90 percent of physicians were concerned that the EHR templates would kind of squeeze out the uniqueness of the story and telling the patient story is not something you can do in a template.

Dr. Eric Fishman: But the government is very focused on data and there's clearly a dichotomy between dictating an office note, at least a history and capturing data. Can you discuss that dichotomy?

Keith Belton: Yes. There's always this tension I think between using the structure of the EHR to capture the information about the specific data elements that meaningful use requires and then balancing. I think the beauty of Dragon is it allows the physician to, with no additional time and in fact less time than they would typing or dictating traditionally, to document the key pieces of the note that are critical. For example, the history, the present illness, the assessment and the plan. Those are the pieces that our 100,000 physicians say they are going to continue to use with Dragon regardless or whether or not the templates. There's also an argument that says a lot of the meaningful use data is typically collected by a nurse during triage and that in fact in many cases, the data is collected anyway and so physicians say it's much ado about nothing.

Dr. Eric Fishman: There are still a lot of doctors who are going to put in their own words what the physician says.

Keith Belton: Yes, absolutely.

Dr. Eric Fishman: Are you expecting, if I can ask, that there will be any changes in the meaningful use standards when the final ones come out?

Keith Belton: Nuance has asked our physicians to weigh in and we think that there may be some mention - I think we're probably 30 days away from it - there may be some additional mention of the value of narrative text. We're a big believer in allowing the physicians to speak what they want to say. I don't expect you'll have significant changes to the regulations but I think you may see some commentary about the value of the narrative and there may be a mention of typing or speech recognition as one input modality that makes it possible.

Dr. Eric Fishman: Lastly, the HITECH Act, the American Recovery and Reinvestment Act. Has that changed Nuance's focus in any way, shape or form?

Keith Belton: I don't think so really. Our mission is pretty much to help physicians document care, spend more time with patients, get the notes done more quickly, cost effectively so I think from our perspective there's a vehicle for funding that's helping physicians get EHRs and we're kind of riding that curve but for us --

Dr. Eric Fishman: Business as usual.

Keith Belton: Business as usual. You can't use the money to buy speech recognition. We really want to be neutral in that respect. I would say one of the exciting things that relates to meaningful use is something we're talking about in the show which is the ability to provide the physicians the Holy Grail. Over time, our vision is that physicians will be able to dictate the way they want in an unstructured format and have some intelligent technology behind the speech recognition parse the dictation that they've done and pull the key elements out that satisfy meaningful use.

Dr. Eric Fishman: And we should be discussing natural language files in just a moment so discuss that which is the direction I think you're going with this parse thing.

Keith Belton: Clearly, the physicians really want that freedom of dictation and have some other intelligent agent take care of populating the EHR database with a meds list and a problems list and vital signs and all the things that meaningful use requires. We announced an acquisition I think today of a company called Language and Computing, which is one of the leaders in --

Dr. Eric Fishman: NLP's?

Keith Belton: --natural language processing or some of us call natural language understanding. So no new product announcements today but I think it's a clear signal for us as a leader in the industry that our product road map of being able to combine the freedom and flexibility of Dragon with this powerful processing intelligent agent is going to be something that people can look for.

Dr. Eric Fishman: Here it is March 2010. Can a physician say allergies, Penicillin and have any EHR record understand that that's a data point rather than a narrative?

Keith Belton: Some elements of the note, if it's structured correctly, there is some early capabilities that we're seeing. We're still evaluating I think everything that Language and Computing has but I think over time, it's hard to say the time period, but certainly in years rather than months, there'll be some announcements in terms of being able to populate the meds list, populate the problems list. It's clearly a direction for us. We typically aren't in the position to talk about specific timelines because some of it is rocket science, some of it is very much Manhattan Project-like capabilities.

Dr. Eric Fishman: I've been with Dragon since Dragon Systems and some of the smartest people I've ever met in my entire life are back there.

Keith Belton: We've got 600 speech engineers and so I think it's safe to say with the natural language processing what we spend on R & D and speech recognition it's probably 10 times any competitor's revenue so I think it's safe to say that we've got good minds on it and our hope and expectation is we'll come to the market with something soon.

Dr. Eric Fishman: Been an exciting time to be in this industry.

Keith Belton: It very much is.

Dr. Eric Fishman: This is Eric Fishman. We've been speaking with Keith Belton, the Senior Director, Product Marketing for Nuance Communications and Keith, thank you very much.

Keith Belton: Thanks, Dr. Fishman.

Dr. Eric Fishman: Pleasure, Sir.