(Part I)

GUERRA: In your report, you talk about the importance of storing data in structured formats, such as LOINC. Can you talk about that a bit?

DAVIS: If hospitals have laboratory data and they want to share that data, the best way is to put that data in the LOINC format. Now LOINC is part of SNOMED. So it’s the ability to encode all of that data, and to actually capture more data and discreet data, that allows us to do a lot of this. And so when you start to look at this, most of the solutions out there have some kind of proprietary data format. The challenge will be how many of those have actually started to use SNOMED for the encoding of that, how many people want to convert to that. So it’s just an area they need to look at and be careful with because, as you start to read through those measurements, one of the areas that they keep talking about is that summary report they want to see shared between all the care entities. So that’s going to include things like your problem lists and progress notes and medication lists and allergy lists and laboratory results. So what we’ll have to do is create that summary document using standard data formats and, of course, what they’re going to use is the CCD transaction standard at some point in time. That’s where it’s all morphing to.

So there are some challenges with controlled medical vocabularies. As you start to bring data into those clinical data repositories, how do you normalize it? Not everybody is using the same clinical department solutions from the same vendor. So if you have Sunquest for labs and you’re using Eclipsys for your EMR and you’re using maybe Siemens for your radiology and you start to bring all that information together, how do you normalize it so that, first of all, you can share it with people and, second of all, your clinical decisions support systems can understand when they should fire.  

GUERRA: Where do you think most hospitals stand with that? It’s hard to imagine smaller hospitals and critical access hospitals having a sophisticated data management group.

DAVIS: Well, I think part of that is probably going to be resolved by the Regional Extension Centers. That may be where they get a lot of guidance and direction. There will also be consultants. I think most of the consulting companies out there understand these issues and have dealt with them.

GUERRA: You wrote that hospitals must move forward in the area of nursing protocols because of all the measurements that will be required. So how do organizations get from that basic level of a simple drug-drug interaction check to where you’re saying they need to go?

DAVIS: Well, first of all, there are some vendors out there that have very good clinical support systems, Epic and Eclipsys and Cerner and people like that, McKesson has done some of that work and Siemens. Most of them set up these user panels, where the client is starting to share and exchange information about how they’re setting up those clinical decision support protocols.

In terms of adoption, if you don’t do this right and your physicians start getting all these alerts that really aren’t meaningful to them, they start ignoring them. That’s not a good thing.

GUERRA: Sometimes we talk about all these things – EBM, CDS, the creation of order sets – as if they are distinct. But isn’t this all really under the umbrella of CPOE, bringing everything to bear in that one moment for the physician?

DAVIS: It should be. Now remember, alerts are going to be little boxes that pop up as they start to enter data on that screen for a number of different tags, dosages, forms of drugs, stuff like that. Where we really want this to go, and I think this is what the government is trying to get to, is they want the physicians documenting. When you start to get to clinical documentation where the physicians are documenting their history and physical, their progress notes, the discharge summary, all those things, it’s putting guidance in there so you get evidence-based medicine. So are you using protocols, let’s say, like APACHE in ICU or InterQual?

Now, long term, what ONC is driving to, and this is very important, the reason they want to capture all this discreet data – which you’re starting to see in 2011 and you see more of in 2013 – is they are going to establish these evidence-based medicine protocols because they’re going to have a ton of data.

What the government is getting to is they’re going to have this huge repository that people start to look at and say, “Here’s best practice, here’s evidence-based medicine.” You’ll have a huge repository and a lot of really smart people looking at it and trying to figure all this stuff out. Then they can actually take that data and create a standard transaction for clinical decision support that can be implemented in all these systems.

GUERRA: I can just picture Deborah Peel reading that.

DAVIS: But you can still do this. You can still de-identify the data and do all this research.

I understand the issues of privacy, believe me, but I think the thing is you can still submit that data as de-identified data and come up with the protocols and the guidelines and best practices. So to me, that really doesn’t fit into the privacy issue. The privacy challenge to me is once you get this data, the summary data, and you start sharing it, that’s where you have to make sure it’s encrypted. You have to make sure you’ve got the appropriate people, processes and guidelines in place for exchanging the information.   

GUERRA: Let’s talk about structured versus unstructured text. Since the physician narrative probably isn’t going away, how do you deal with the important information that might be trapped in unstructured text?

DAVIS: Well, I think you can look at this from a couple of different perspectives. First of all, there’s a movement out there called Health Story. Health Story says that what physicians are trained to do and what they’re very good at is called dictation transcription. The problem with that is it can take more than 24 hours to get the information which is then put into a text log, so it’s hard to extract information. Health Story is trying to do two things.

The first step is that part of the CCD standard is the document formats. So when you create a document, it actually creates the formats for certain data. So the first step of Health Story is to let physicians do the dictation and transcription and create these structured documents so that when people are looking at those documents they know exactly where to go to find the information they want. So it’s an incremental step of improvement, and the nice thing is physicians are okay with it because it doesn’t disrupt their workflow, they understand it, and so the adoption should be pretty high.

One thing you can do to that data is called Natural Language Processing technology. That’s an engine that looks at text to extract discreet data. So retrospectively, you can take those documents around the naturalized processing and extract discreet data for reporting or research or whatever you want to do. What you really want to get to is, as the physician dictates, speech recognition is involved. They can see the text being created. You can actually dictate discreet data into certain fields to be captured and interacted with clinical decision support during that process.

Health Story already has implementation guidelines for several different documents. There is an ideal model in the future where the physicians would dictate, speech recognition would take over, would be able to capture discreet data, would be interacted with clinical systems support. In that scenario, you’d apply evidence-based medicine protocols and clinical decision support while the physician is documenting. 

Part III