At the M*Modal exhibit at HIMSS10, Dr. Eric Fishman speaks to Dr. Nick van Terheyden, Chief Medical Officer of M*Modal. Dr. van Terheyden discusses M*Modal’s innovative speech understanding capabilities. He explains how this level of speech understanding originated, and how it emerged as an important service in medicine, especially as the adoption of EHRs continues to grow. With an EHR, the speech understanding technology can help physicians input and then encode data with extremely high accuracy. Dr. van Terheyden closes by discussing the Health Story project, which was developed among a consortium of partners working to retain narratives and add value to EHR systems.
Dr. Eric Fishman: This is Dr. Eric Fishman with EHRtv and today we have the pleasure of discussing M*Modal with Dr. Nick van Terheyden. Dr. van Terheyden is the Chief Medical Officer of M*Modal. Nick, I thank you for taking your time today at the end of a very busy three days I'm sure.
Dr. Nick van Terheyden: It certainly has been.
Dr. Eric Fishman: Tell us about M*Modal briefly, tell us about speech understanding.
Dr. Nick van Terheyden: M*Modal is an innovative company that has taken speech to a whole new level of use. It was founded as a direct result of some work with the NSA listening in to telephone conversations overseas. What's interesting about that group of people is that you can't influence their behavior. You can't call them up and say, please say it this way, we need this pronunciation or we'd like the punctuation put in there.
Dr. Eric Fishman: So you get the raw data?
Dr. Nick van Terheyden: In terms of adapting that process, you have to turn the whole concept of speech on it's head and say rather than trying to influence the behavior let's have the technology adapt to the user, apply that to the physician setting. So radical thought here instead of asking physicians to change their behavior, we're going to say we'll apply the technology and we'll adapt to what you're trying to do. You don't have to do anything just continue to dictate in the way that you normally do.
Dr. Eric Fishman: Let me stop you there for two seconds. Does that mean it could be on a digital recorder, it could be on a telephone, it could be on a microphone in any of those modalities?
Dr. Nick van Terheyden: Again, one of the great advantages of coming from that heritage is, of course, the telephone is the poorest quality audio device on this planet.
Dr. Eric Fishman: Right.
Dr. Nick van Terheyden: Since we came from that heritage, we can pretty much take audio from any environment and use it no matter where it came from so absolutely.
Dr. Eric Fishman: Okay, continue.
Dr. Nick van Terheyden: As we looked to that, we took a whole different approach and said to physicians, dictate the way that you normally do. So as of today, about 96 percent of physicians that use our service are completely oblivious to that fact that they use M*Modal.
Dr. Eric Fishman: I was going to go into that because I had not heard about it very much and we'll go there in just a moment
Dr. Nick van Terheyden: So they're unaware of it, it's integrated into our partner's applications and in fact, what we have is a partner model very much like Intel, it's the Intel inside with a speech understanding in many of our partner's solutions. For many of our partners, they present it as if it's their solution and we're proud of that. We're delighted that they do so and for the physicians they use it and it's coming from that partner.
Dr. Eric Fishman: It's invisible to them. I saw Greenway up there, so obviously that's a well-know EHR product. Tell us about that relationship.
Dr. Nick van Terheyden: That's correct. Greenway relatively recent. In fact, we started out in the medical transcription marketplace. That was an interesting place to start because we could actually do this without physicians being involved. They were completely unaware of it and in fact, we're in 8 out of the top 10 medical transcription service organizations delivering value added service. So we improve the efficiency, they're able to save money and for a long time that was a really well kept secret. In part, because they didn't want to share that information with their customers and there's a lot of price pressures in that marketplace. But as we looked at it, that's obviously a changing industry and we want to move out into a closer relationship with the physicians because we are the bridge to the divide for physician adoption of EMRs. The biggest challenge for an EMR is data capture. There isn't a physician - and you're a physician, you love an EMR solution but you want to access the data, you don't want to put it in. Essentially, we solved that problem. Not just with the narrative, which from a physician perspective, essential.
Dr. Eric Fishman: It's critical.
Dr. Nick van Terheyden: Let's not take the narrative out of it. Let's include the narrative for in addition to that we want to encode that information. So we do that as part of the speech understanding process. We tag that data against a variety of controlled medical vocabularies, Rx norm, radlex for radiology, Snowmed CT, ICD-9. Not a coding engine but a tagging engine that allows you to create structure from that free form interaction that you and I go through when we interact with a patient. It's part of our diagnostic process. We enable the use of that but then deliver the structure that the EMRs demand and need.
Dr. Eric Fishman: You're stating that if I as a physician were to say something like patient is a 24-year old white female who was involved in a motor vehicle accident at the intersection of Marietta Street and Beach Street that that would become data and searchable as data instead of just being free text narrative?
Dr. Nick van Terheyden: Exactly, but then, I as a physician would want to read what you wrote because that would actually give me information.
Dr. Eric Fishman: Of course.
Dr. Nick van Terheyden: But the computer system wants to know that it's a 24-year old that drives a whole series of activities, motor vehicle accident, may be drives some activities in terms of investigations, imaging and so forth or maybe alerts and all of that becomes tied together but we don't lose the narrative, but we convert it into data. To take Dr. Blumenthal's example from today when he talked about sulphur and the sulphur allergy that was a great example. The one challenge with that is you had to have that as a data point in the system and he had to enter that he was prescribing sulphur drug to be able to tie those two together and say that it was a problem. We can't do that with narrative but that’s what we do with speech understanding. So when he said sulphur drug, we would tie that back to Rx norm, it would look at the EHR system and say, hey there's a conflict here, let's provide that alert and that provides exactly what physicians want. Ease of use and entry but then giving them all the value add from the EHR from those alerts and that decision making and evidence based medicine.
Dr. Eric Fishman: Clearly, the EHRs need all the data they can get. Nick, just talk about the quality of the understanding. I walk up to one of the computers in your booth here and in a female voice in moment just started to do an orthopedic note and to be quite honest, I was shocked at the accuracy. Has there been any recent increase in the accuracy or is this how it's been in 2009 and 2008?
Dr. Nick van Terheyden: Shocked I hope in a good way.
Dr. Eric Fishman: It was phenomenal, phenomenal I thought. With zero training enrolled in an opposite gender voice profile.
Dr. Nick van Terheyden: Right.
Dr. Eric Fishman: And it was accurate.
Dr. Nick van Terheyden: I think there are some very good reasons for that. If you look back at our heritage, what we did was we delivered this as a service. So we've been creating this since 2001 when we formed this company and started to roll this out to the medical transcription space, what happened was we've got all of these editors, hundreds, thousands of editors around the U.S. that are correcting and editing this information every single time. We're not just correcting the text, but also the structure, the tagging of that information. Every time we correct and we edit this, we learn from that.
Dr. Eric Fishman: It improves the understanding for all the physicians.
Dr. Nick van Terheyden: For all physicians. A physician from L.A. contributes and a physician from New York contributes, a physician from Atlanta. So what that means is when you come to use the system, first of all, we've got this massive corpus of knowledge that we can apply immediately so you can get very quickly up to speed and you saw that in real life. The other thing that we had to do because we didn't know who we were listening in to overseas, we had to try and be intelligent in the application of the right profile. The same thing happens in the medical setting. I don't know what domain you're in but it does try and listen and say, oh, it's a male voice so I should adjust. So even though you picked a female profile, it has some real intelligence behind the scenes. That's really part of the value that's coming from this knowledge of this big corpus that we can apply so you get very quickly up to speed. But then as you correct, if you train your own profile, it's going to already improve. Imagine how good it was there and then if you apply some customization to your individual profile, it's even better.
Dr. Eric Fishman: Do you have any published numbers that you're willing to discuss, 98 percent accurate, if you're doing front end speech recognition that's not going to be corrected by anyone.
Dr. Nick van Terheyden: I love this accuracy question a lot because physicians have been trained for the last 20 years to believe that they dictate in coherent structured sentences. The reality is that medical transcription has been editing that as a black box behind the scenes and converting that into the structure.
Dr. Eric Fishman: And coming up no higher than 98 percent accuracy.
Dr. Nick van Terheyden: Well, actually they're already better but what that tuned us to believe was that's the way we dictated. If I gave you 100 percent accurate rendition of what a physician dictated, they'd be unhappy with it. So actually, accuracy is a very poor indicator.
Dr. Eric Fishman: I'm a surgeon, don't give me what I ask for but give me what I need.
Dr. Nick van Terheyden: Give me what I thought I said. In fact, what we try and do is we have something called a Q Score which tries to give you an indication of --
Dr. Eric Fishman: The quality.
Dr. Nick van Terheyden: --the quality and it's not quality in terms of accuracy it's how much effort. What's the effort required to make any changes? It varies widely. Clearly, if you're a non-native English speaker they're still some challenges, but we'll still learn. That's the beauty of this technology. It's no longer singly to radiologists and clear speakers and people without accents. It now applies across the board, but those individuals will be picked up very quickly and have very little requirement for that additional training. Somebody that's a non native speaker takes a little bit longer but they'll still get up to that point where it requires very little effort.
Dr. Eric Fishman: It's wonderful. Nick, what do you see likely to be happening in the next 12 months?
Dr. Nick van Terheyden: Well, I think clearly part of the meaningful use, and I apologize, everybody's using that term but we've been using that term for a long time. We see meaningful clinical documents as a significant thing in this whole equation. The divide for EHRs has been capturing that information so that retain the narrative but continue to feed those database elements that are contained within the structured EHR. As we move towards meaningful use, one of things and we talked about this before we got together, the meaningful standards do not include narrative.
Dr. Eric Fishman: Specifically excluded it.
Dr. Nick van Terheyden: Right. I firmly believe that there isn't a physician on this planet that would accept receiving a medical record that is pure check boxes and "select from this list". Number one, it dumbs down the medical record and number two from a physician perspective, they can't communicate effectively and you're not going to get sufficient information. So even if it's not in that meaningful use criteria --
Dr. Eric Fishman: It needs to be undertaken.
Dr. Nick van Terheyden: I think it will be included and I think the key to the success is the ability to drive the two together but use the source as narrative. The narrative should be - not everywhere, the blood pressure doesn't need to be narrative.
Dr. Eric Fishman: The history.
Dr. Nick van Terheyden: But in the history and physical, I want to hear the story, I want to hear that but then extract the meaning from that, tag it so that these clinical systems can use it. That's essentially what we do. We do it as a CDA compliant. Everything that we do is a CDA compliant and we were one of the co-founders of the Health Story project that actually contains this.
Dr. Eric Fishman: I was just about to ask you about Health Story as the last thing. So describe that a little bit.
Dr. Nick van Terheyden: Health Story was founded about two or three years ago a consortium with AHMIA, MTIA, AHDI, Alschuler Associates and M*Modal. We believed in CDA as a concept for containing all this information, retaining the narrative but adding value to it. That's the beauty of CDA is that you can impose additional layers of information while not losing that original content. As we looked at that information and said, hey how do we do this? We formed an open consortium, it's a non-profit organization. We have a booth over in the connect-a-thon. I gave a presentation and in fact, we've had significant uptake of that. There's been some members that have signed up both announcing at the show and also signing up at the show when they've heard some of the talk. What it does is it retains the narrative, but includes the structure and allows for the sharing of that information effectively and efficiently between any system that's able to consume Health Story CDA compliant documents. My favorite sort of analogy for this relates to Paul Harvey and "now here's the rest of the story". Who doesn't want the narrative? That's part of the Health Story concept. Include that narrative but include also all the rich text and the data that's included with that.
Dr. Eric Fishman: Absolutely needed, absolutely. Nick, it's been a pleasure.
Dr. Nick van Terheyden: Thank you very much.
Dr. Eric Fishman: This is Dr. Eric Fishman for EHRtv. We've been speaking with Dr. Nick van Terheyden, Chief Medical Officer of M*Modal. Thank you.