Reid Conant, M.D., wears several hats these days. He practices half-time as an emergency department (ED) physician, and in that role, he  is also CMIO of the Tri-City Emergency Medical Group, a 23-doctor emergency physician group in Oceanside, Calif. Tri-City provides emergency physician coverage at Tri-City Medical Center, a 400-bed community hospital in Oceanside.  Conant also consults privately through his firm, Conant and Associates, where he focuses on the optimization and deployment of physician documentation solutions.

At Tri-City Emergency Physicians and Tri-City Medical Center, Conant has been leading the implementation and optimization of physician documentation, including through the Dragon Medical speech recognition solution from the Burlington, Mass.-based Nuance. He also has a business relationship with Nuance through his consulting work. Conant spoke recently with HCI Editor-in-Chief Mark Hagland regarding his and his colleagues’ experiences with physician documentation and speech recognition. Below are excerpts from that interview.

In your view, are the requirements of physician documentation in the ED as onerous as before?

There’s been significant improvement in adoption, primarily because of new technologies, but also in the understanding of how to train and re-train physicians to adopt these technologies. One thing that we encountered at my facility was an initial reluctance to document electronically, because of all the pointing and clicking. Now, since we’ve added speech recognition as an element, we’ve made it so that it’s no longer all points and clicks; and that has enhanced adoption. The documents were more meaningful to the hospitalists or ICU nurses; they were more meaningful when we added Dragon to it, because we were able to add more to the narrative. I’m a decent typist, but I’m nowhere able to get near the level of efficiency that’s possible using speech recognition solutions.

I do think that we’re in a transitional period right now as an industry, including medical informatics and clinical practice in total. And the reason for that is that we have requirements for problem list management and for diagnosis list management, as well as core measures and other regulatory requirements. And it’s not just completing these, but documenting them thoroughly as well. So there has been an increased burden for providers not only to deliver care consistent with clinical guidelines, but also to document that one has done so. And we’re in a transitional period in which the technologies are catching up with the requirements, but they haven’t entirely done so yet.

Reid Conant, M.D.

Can you speak a bit more specifically to the transitional aspect of this?

Yes. For a while, we’ve been working in an environment with fairly regimented formats, which was necessary because the technology was not yet there to capture data from unstructured text. Now, we can get structured data out of narrative text, because of natural language processing. In an ideal world, the physician would be able to create a document, guided by a framework or template consistent with a presenting condition or care plan, but they’re able to add patient-specific narrative that could then be mined and accessed to create discrete data elements. That’s kind of the best of both worlds—the necessary flexibility to make providers efficient—along with the ability for the technology to capture discrete data elements.

How frustrated are most emergency physicians right now, with having to move forward in the electronic documentation world?

Well, there’s a bell curve in that regard. Those sites that have deployed documentation solutions in a strategic manner, with the right tools, and with the tools optimally configured, can make that transition quite seamlessly. I’ve also seen sites struggle, unfortunately; and we were one of those, when we went up six or seven years ago. But the addition of speech recognition, and the optimal configuration of the electronic health record, have helped.

So for example, the creation of commands within Dragon to help speech-enable steps that we repeatedly use within the electronic medical record: to add an order, to sign a note, many other examples. But we can also build content into Dragon that can facilitate and streamline or work as well; for example, if I have a code status discussion, such an advance directive or end-of-life direction, discussion, with a patient. In that instance, there are multiple items I would cover in that discussion on a regular basis; so why should I repeat that instruction over and over, when I can rely on a pre-created element? There are many other examples, such as  operative report details, procedure notes, assessments.

How complicated is it to build speech recognition elements into these templates?

It’s something that we’re able to train our physicians to do, and can be done on an individual-user level, or on an organizational level. It’s very doable at the user level. As consultants, my colleagues and I have also put together a bundle of about 2,000 different starter commands over 50 subspecialties.

Could you provide an example of a bundle of starter sets?

Code status discussion, smoking cessation discussion, procedure notes, critical care notes. It could really be any form of documentation that can be used repeatedly. And those provide a starting point for users, where they can go in and customize those for themselves. The beauty of it is that when you take tools like those, and apply natural language processing to them, we’re then able to dredge data out of otherwise-unstructured text. For example, Cerner and Nuance have taken Nuance’s natural language processing solution and have put it into a tool that will provide real-time feedback to physicians on the quality and appropriateness of their documentation, related to necessary ICD-10 elements. As you know, there’s a necessary increase in specificity and acuity documentation related to ICD-10. That’s going to be a major challenge for us both in the ED, and everywhere in medicine. But when we have tools like NLP embedded into the EMR, as is the case now with Cerner, that’s an advance.

What would your advice be for CIOs, CMIOs, and other healthcare IT leaders, at this moment in time?

October 2014 is the deadline for compliance with ICD-10 documentation, and it’s coming at us like a freight train. And the interesting thing is, in order to deploy these technologies to achieve those standards, providers must be documenting electronically. So with that said, we need to be focusing now on getting the docs into the record and documenting electronically; so now, in the next six months, is the time to focus on getting them into adoption in the EMR and in physician documentation.

Is there anything else you’d like to add?

So that is key right now, getting them in, increasing adoption rates, but also planning to apply these recently developed and developing technologies to assist the providers with regard to the increased scrutiny that will be applied relative to ICD-10, because as physicians, we’re not going to be able to do it alone. We already get constant feedback from coders; and the stakes will be higher when ICD-10 comes, so it’s time to put those tools into place, in order to meet those regulatory elements.