Fallon Clinic’s Dr. Larry Garber on EHR + voice recognition (transcript)
This is a transcript of my recent podcast interview with Dr. Larry Garber of the Fallon Clinic.David Williams: This is David E. Williams, co-founder of MedPharma Partners and author of the Health Business Blog. I'm speaking today with Dr. Larry Garber. He is medical director for Informatics at Fallon Clinic. Dr. Garber, thanks for you time today.Dr. Larry Garber: It's my pleasure.Williams: Dr. Garber, what is the Fallon Clinic and what's your role there?Garber: The Fallon Clinic is a multi-specialty group practice located in Central Massachusetts. We have 250 physicians and another 80 advanced practitioners seeing over one million patients a year here. We have physicians representing approximately 30 different specialties.I'm Medical Director for Informatics and also a practicing internist. Part of my job is to help implement information technology throughout the organization.Williams: What's Fallon's history with electronic medical records?Garber: We've been using homegrown systems for many years. Back in 1992 we wrote a homegrown system to store information. We stored lab results and radiology reports and medication histories and things like that. We realized around 2002 that a homegrown system would fall short on workflows and decision supports that are crucial to electronic medical records. We subsequently implemented Epic’s electronic health records throughout the organization; we completed implementation at the end of 2007.Williams: I understand that a year or so after that that you launched a pilot with Dragon Medical. Why?Garber: When we first started to roll out Epic, we knew that we had to pay for it. There was no stimulus funding to help us and so we looked at what it was going to take to get a return on our investment.Most of the savings were coming from two places. One is the reduction of medical records staff, since we wouldn't have to be shuffling paper throughout Central Massachusetts anymore. The other savings that we were looking for was a reduction in the transcription cost. We were spending several million dollars a year with doctors dictating into recorders and having those sent overseas electronically to be typed at ten cents a line, so we budgeted a certain amount of savings.Epic comes with tools to help physicians and staff create notes so that they don't have to dictate. We were pretty sure at the beginning that those tools, while good, were probably not good enough to completely eliminate dictation. So our goal had actually been to drop the lines of dictation and transcription by about 75%.We knew about Dragon Naturally Speaking back at the very beginning of our implementation and gave it a try (they were on version 7 at the time) to see if we could use that to assist with creating the notes to avoid sending our transcriptions overseas. At that time (2005), the study showed that it wasn't good enough for production use by physicians. So we went ahead and rolled out Epic.We found that we only dropped lines of transcription by about 35%, which wasn't good enough. So we came back in 2008 and took a look at Dragon again. At that point they were up to version 9.5 and we had heard good things about it. We approached the folks at Nuance, the makers of Dragon and said that we would like to do a formal study to evaluate Dragon with EMR in place and to see if it really works.Williams: What's the scope of the study: how many physicians, what objectives or endpoints?Garber: We wanted to do what we call a “real world” study, reproducible both within and outside of our organization. We picked nine physicians and one physician assistant. We looked for varying computer skills, varying speech clarity, varying ages, varying specialties; we had a very good mix. It ranged from people who used Epic for 100 percent of their documentation to those who used Epic for none of their documentation and were doing all dictation. We had some people who could speak very clearly, some with a very strong Indian accent and one with an Israeli accent. We had medical specialists, surgical specialist, internists, pediatricians. We wanted to identify the time impact to create notes, the quality of notes and the financial impact.These were all doctors and a PA who had been using Epic for about a year. We did a baseline analysis for one week. We watched their patterns in terms of how long it took them to create their notes and we made copies of all the notes that they were creating and we did analysis on productivity.When we let them use Dragon we trained them for about an hour and a half and then had them use the medical version of Dragon for two months. After two months we came back and reanalyzed the practice and also surveyed them to see how they were doing. What we found, which I guess was a little surprise, was that the time it took for the notes to be finalized in our electronic health record dropped from an average of almost four days prior to Dragon down to 46 minutes after using Dragon.It makes sense. They were just doing the notes right then and there and they were done as opposed to dictating it, sending it overseas, having it come back, having to sign them. Something that we didn't really think about was the implications of this drop in turnaround time was. It actually changed a lot of our workflows.So for instance, it's not unusual for us to see a patient in the office who needs to be seen by a specialist. We would call over to the specialist and say, “You need to see this guy this afternoon. He's got an incredible rash, this is the history that I've taken, this is what I know,” and then we would send the patient. That was a doctor-to-doctor communication, which took up a lot of time.Now we can dictate the note right into Epic, then arrange to have the patient see the secretary to get the appointment. Then when the patient goes over, the dermatologist just looks in Epic and sees the whole note with all of the history. Now we don't have to pull doctors out of rooms to try to get two people on the phone at the same time.Similarly, we found patients would sometimes call back in the afternoon and say, “I saw the doctor this morning and he said something and I don't remember exactly what he wanted me to do. Can you go ask him what he wanted?”They'd have to hang up on the patient, go find the doctor, call the patient back and do the communication. Now what happens is that the nurse who is on the phone actually just looks into the Epic electronic health record and sees the note right there. She can tell the patient immediately what it was that the doctor wanted done. It's really made us much more efficient and it's been great in terms of quality of care that we're giving.There are a few other findings. We asked doctors how satisfied they were with creating notes prior to using Dragon. We used a five point scale, where 5 is high. The average was about three. After we implemented Dragon and surveyed them again, the average was over four. So there was a dramatic improvement in physician satisfaction with the note generation process.We also looked at the notes and we blinded them. In other words we changed their formatting so that you couldn't tell whether this was done with Dragon, without Dragon, with dictation or just using Epic tools. Then we reviewed those from before Dragon and after Dragon. We gave them to a few different people to analyze. We gave them to a physician who reviewed the notes for how well the notes communicated medical information and how well they supported defensible medicine if we were ever to go to court. With Dragon there was a dramatic improvement in the quality, the medical/legal quality of the notes.We also had our coding department take a look at the notes to see how well the notes supported the billing that we do. One of the quirks with the current health care system is that you only get paid for what you write down in your note, regardless of what you actually did. We found that people could justify higher levels of billing when using Dragon.This is kind of interesting, because if you think about it we're taking some people who were dictating --they had a tape recorder and they could say anything they wanted-- and now we're giving them a microphone where they could say anything they wanted. And you ask how could dictating into one and changing it to another improve the quality of the notes and allow them to bill at higher levels?It turns out when you're dictating into a tape recorder, that's it. That's your whole note. You dictated the note and whatever you said, that's it. When you're using speech recognition, you're dictating directly into the electronic health record. That allows you to take a hybrid approach. There are some things that the electronic health record is very good at and very fast at such as speaking into your note what the med list is, the allergy list, the past medical history, the smoking status and things like that very quickly. With a few clicks you can pull all of that into your notes and then you can dictate the things when you were talking about the history of how they slipped and fell on the banana peel and hurt themselves and what you're thinking about the differential diagnosis; is it a break or a sprain or whatever. Those kinds of things you would still dictate, but you would also use the EHR for what it's powerful at so you would end up with a more robust note than you would have if you had just been using the EHR or if you had just been using the taped dictation. So that's why everybody won on that.Williams: That's very interesting.Garber: What we did found out is that overall it took eight minutes per day per doctor longer to use Dragon. That was interesting because when doctors speak into a tape recorder they can speak at about 120 words per minute. We found that using Dragon they were speaking closer to 86 or 90 words a minute. So as a result, if you look overall in the course of the day it took about eight minutes longer for each doctor to get their work done.Williams: Does that have to do with the technical limitations of the software or is that just more a style of how they speak when they're dictating to the computer?Garber: I think it might be a little bit of both. I also used Dragon (not as part of the study), but I use it now and I can speak perfectly normally and clearly. I probably speak about the same that I would have if I was using the tape, but I know that some docs do speak faster on tape and slower on Dragon. Some of it is also that we're still trying to get docs used to dictating to Dragon in the sense that you can't watch it.What we found was that if you're dictating and watch the words coming up, you sit there a little bit in awe that it's doing it and double checking. Then what happens is you actually don't speak normally, whereas Dragon is much more efficient if you just sit there and concentrate on what you want to say and say it clearly. It does a remarkable job at getting it right and then afterwards you can go and take a look. My accuracy rate is 99 percent. It's unbelievable how good it is. It's just something you have to learn to do, which is just dictate and trust it.Williams: The Fallon Clinic I'm sure has some particular characteristics to it. What do you think about the applicability of the results of your study to other physician organizations?Garber: I think that what we did is applicable to any physician’s office. The reality is, I can't imagine any physician’s office not using speech recognition software like Dragon. I'm a fast typist and have been using computers for 30 years. In high school I had my typing course and I can type probably 80 words a minute, which for a physician is pretty good.However I find that with Dragon will consistently type at a faster rate and more accurately than I can. So even with the younger physicians coming out of school right now who love technology and can type at great speed, they really still can't type as fast as you can when you're using Dragon.You have to use the medical version. We had one doc experimenting with the regular version and it was not good for health care. There were so many medical terms it didn't know whereas Dragon Medical clearly picks all of those up and is highly productive.Williams: If we look down the track a few years, what kind of EHR enhancements would you expect, whether those are speech recognition or just other technologies that will make a difference in that time frame?Garber: Well, do you want me to be realistic or a dreamer?Williams: You could be realistic but use a long time frame!Garber: Okay, I'll give you two scenarios. One, which I think is happening already is we're all getting smart phones. The problem with the smart phones is that the entry into them with keyboards is suboptimal. I think that more and more we're going to see higher quality applications that allow us --when I'm on call, sitting in a restaurant or at my kids sports games or whatever-- if I need to look something up or create a note or whatever then I'll be able to do that using my smart phone with medical vocabulary.I think we'll see more and more of that mobile computing with visiting nurses going into patients’ houses. They'll be able to use speech recognition with these very portable devices.My ultimate dream is I go into an exam room, I talk to the patient, as I examine the patient I say, “It sounds like you have a one to two over six systolic murmur of the mid left border radiating to your maxilla,” and the whole conversation gets recorded and indexed and metadata is pulled out and I didn't have to touch a keyboard through the entire history of the exam.So that's my dream. Whether it will happen, who knows.Williams: That sounds good. I guess you could take it one step further and just have what you think recorded. That could be slightly dangerous, not for you personally but maybe for others.I've been speaking today with Dr. Larry Garber. He's Medical Director for Informatics at Fallon Clinic. Dr. Garber, thanks so much.Garber: It's my pleasure. Thanks for having me.