Transcript for the [ontolog] Scheduled Discussion session - August 25, 2005

• Topic: "Healthcare Informatics Landscapes, Roadmaps, and Blueprints: Towards a Business Case Strategy for Large Scale Ontology Projects"

• Moderator: Rex Brooks

• Panelists:
  • Bob Smith, Brand Niemann, David Whitten, Marc Wine, Mark Musen, Ram D. Sriram

[ontolog] 2005.08.25 Session Transcript

Transcript of Ontolog Forum Panel Discussion: Healthcare Informatics Landscape: Towards a Business Case Strategy for Large Scale Ontology Projects

• Hosted by Ontolog Forum, Courtesy of CIM Engineering, Peter P. Yim, (PY)
• Moderated by: Rex Brooks, (RB)
• Panel Members:
  • Marc Wine, GSA Office of Intergovernmental Solutions, Health IT Coordination (MW)
  • Bob Smith, Professor Emeritus, California State University (BobSmith)
  • Brand Niemann, EPA, chair of the federal Semantic Interoperability Community of Practice, CIO Council (BN)
  • David Whitten, WorldVistA, Co-founder; Public Domain Knowledge Bank (DW)
  • Mark Musen, Stanford University, Professor of Medicine and Computer Science, Director Stanford Medical Informatics (MM)
  • Ram D. Sriram, NIST, Manager, "Manufacturing Metrology and Standards for the Health Care Enterprise" program (RS)

• Introduction

We have Bob Smith. We have Marc Wine who just spoke. Mark Musen is the head of the Stanford Medical Informatics, He is also head of the Protégé effort which is one the main tools that I use and that a lot of other people use for actually creating an ontology. Then we also have... the next speaker is Ram Sriram from NIST, who's the lead of the design process group in the manufacturing systems integration division and he's manager of the manufacturing metrology and standards for healthcare enterprise program. So, I expect that between what Marc has done and what Marc said and what Ram is about to inform us about, we'll be having the, ah, public sector very well represented.

Following Ram will be David Whitten, who works as a senior computer specialist at the Department of Veterans Affairs, but he's also the founder of WorldVistA and OpenVistA and gave a remarkably excellent presentation to this forum a month ago and I'm looking forward to actually including OpenVistA in work I am doing, and I will mention that at the end of the process here.

Following Dave will be Brand Niemann, who is a computer scientist at EPA and the chair of the Semantic Interoperability Community of Practice, which along with the GSA sponsors the Collaborative Expedition Workshops at which I've given a couple of talks and which has proven to be the linch-pin as far as I can see for the advancement of cross-cutting inter-agency information and exchange and I think that perhaps more than other components is responsible for the imminent widespread adoption of the Data Reference Model of the Federal Enterprise Architecture effort.

And with that I'd like to say that one of the things we have put on the resources on the wiki page that falls in line with our initial topic of "Towards a Business Case Strategy Large Scale Ontology Projects," is last Friday's publication of a notice in the Federal Register of the Office of the National Coordinator for Health Information Technology, and I actually sent out some messages to our panelists about this, and I'm hoping that we will be able to perhaps tie into that. It's a clear where this is a landmark, a landmark in time, that gives structure to the function of the office and begins to fill it out. And gives us a place -- a point of entry -- for beginning a conversation with them. And we're hoping that this will also help us support Mark Musen's nomination to the uh what's the exact title of that, Bob?

Bob Smith: AHIC the American Health Information Committee.

RB: Right. We're hoping that Mark will be able to not only represent the Ontolog Forum but also the rest of the efforts behind the Stanford Medical Informatics

And with that I have the pleasure of introducing Marc Wine. Marc works with the General Services Administration's Office and the Office of Intergovernmental Solutions, and Marc, carry on from there...

• MarcWine

It has been only 12 weeks since I've begun my new role and responsibilities at the Office of Intergovernmental Solutions in GSA. What we are doing is exploring different opportunities for engaging in collaborative activities with state, local, federal and international governments to help guide and support the development of health information technology systems particularly as they relate to state, local and federal governments.

These are coordination activities that are going to be in line with the Nationwide Health Information Network (NHIN) Strategic Framework Goals. [ http://www.hhs.gov/healthit/strategicfrmwk.html ]

Most recently, for the last five years I've been a Program Manager for the Health IT Sharing Program over at Veteran's Health Administration and played a key role in supporting the (calvan?) of sharing Projects across federal agencies and the program there also focused on the building of health it sharing projects across state, local governments - private sector with the VA, and, of course, the history between VA and DOD in building integrated electronic health record systems and bi-directional communications electronically.

Going forward we're discussions on Regional Health Information Organizations (RHIOs)... looking at developing a strategy to engage in coordination to provide guidance and support for the development of integrated health information systems, making the most of health information exchange projects that have been implemented through funding from AHRQ [ http://www.ahrq.gov/ ] and HRSA[ http://www.hrsa.gov/] across the country. And, promoting the concepts and projects of open standards; open architectures; how do we harmonize interoperability among these approaches in both open source and the various proprietary electronic health record models systems

I want to emphasize [ ] that we look toward opportunities where there are health IT applications that could be engaged with ontology software tools that could be notable demonstrations of the use of ontologies to encourage and advance interoperability in health systems, in health enterprises and across providers offices as these networks begin to evolve.

I'm in favor of early planning in health IT projects where we can provide opportunities to demonstrate the use of ontology tools. I think some key strategic assumptions would be to consider how ontology tools, how interoperability in advanced ways can support the use of electronic health IT systems that aim at improving the situation with medical errors, improving patient quality -- the quality of patient care delivery -- and I think also, cost effectiveness and efficiency in the planning, development and implementation of health IT applications employing ontology tools. I think, assuming that we can support cost effective project development and implementation and have results that create the efficiencies in the use of electronic health record systems that must be interoperable across different kinds of models using different lexicons to harmonize; to make a common playing field - a level playing field for the end users in terms of the definitions of diagnoses that drive judgments that are used at the point of care with electronic health records.

So, I would highlight efficiency, cost effectiveness, and a goal toward using these ontology tools in health IT applications to support the improvement of quality care.

MW: There are a couple of other slides on strategy, benefits of ontology tools in health IT applications in my slides that you can look at, and I'll close with that and thank you very much for the opportunity.

• BobSmith

Bob Smith: Thank you, Rex. Thank you, Peter. And Thanks for tuning in. Rex and I have a general perspective on where Health Information strategies and policies are now and how they may evolve with and without an explicit ontology methodology. I personally have two perspectives and it is slowly dawning on me how daunting is this challenge. The first perspective is that of a consulting academic involved with Healthcare operations research, artificial intelligence, and information technology strategies for 45 years. The vast majority of this time was very frustrating in dealing with subtle resistance to changes in strategic processes. The second perspective I want to explore in today's very broad canvas is that of a consumer. A person who is sitting at the bottom of this healthcare supply-demand chain who has at least three or four major problems because they're north of the age of 65. They are on Medicare and they're facing a set of very confused doctors, nurses, practitioners, pharmacists, laboratory technicians. As you walk through this system as a consumer you realize how little these individuals understand of healthcare information technology standards. [ slide ten].

[ One additional point from the last slide, I'm also a volunteer in emergency response training in Huntington Beach, and I'm also involved in GIS GMS systems the Southern California-Tustin area with the national mapping project. So I've got a reasonable level of comfort with how the data has changed, and that we've got the tools (needed for the landscape). ]

Very briefly, I want to concentrate on Personal Health Information Exchange and Interoperability from those two perspectives. The perspective on the left (slide 10) is the top, with looking at Business Ontology Management. the Business Process, BPMN, ebXML , UBL, all those international standards ... we're looking at process, business process, business language, the analogs I'm involved with. On the right is the consumer perspective ... looking at a very fuzzy, indistinct healthcare standards organization market. We obviously need roadmaps and we need blueprints to craft healthcare standards intelligently for this huge baby boom retirement that could easily bankrupt most of the families in the US. Next slide, please.

We can move quickly on this, there is a complexity we see ... partly a lack of our understanding of the field, with improved tools and improved concepts and improved language capability ... can a lot of the complexity be simplified and the causal chains ... made much more evident?

I think Ontolog has an excellent opportunity [ I would ask the question] "What have we seen in the past twenty years about how health standards have evolved effectively and what's the actual market and the dynamics?

Gordon Bell has some excellent insights that Rex, I think, will address. [slide __ ]

I'd like to look at the landscape between 2005 and 2008 from the business standpoint of individual entrepreneurs looking for competitive advantage in solving what they see is a significant marketing opportunity that extends as we know to some questionable processes. I also see, in response, a government role in balancing conflict for this value -- public benefit ... various fail points I think can be achieved, verbalized, made explicit with business process tools, with full understanding of some of the thought leaders and entrepreneurs.

It's easy to start a list of thought leaders, looking at conferences, doing a google search ... it's incredibly popular these days, part of the representation that this is merely the business (seas?) is especially relevant. [ Slide number 5 of Bob's deck, but 13 on the combined deck. ]

Bob Smith: So, we have a framework that we're attempting to formalize and we can look at the various roles of government, industry, entrepreneur, and consumers. Today all of us are now, or in the near future, a consumer of healthcare. Overlooking the Federal government is the General Accounting Office. It's sometimes seen as a paper tiger, but some recent things that I've seen from the GAO linked with the OMB, is a global information grid, federal health architecture maturity taxonomy of five parts. [ have a url if you're interested.]

I'm playing three different roles with HIMSS, a large, twenty-thousand member community within the healthcare information (realm/arena) that could use ontology perspectives, ontology strategies. Perhaps the most exciting thing as you look at inflection points and the demand for health information exchange interoperability is Esther Dyson's public Health Information workshop in September, and I see that Brand Niemann has also referenced this outstanding conference.

I call attention to some very important domains working in the area of health information exchange that I haven't seen referenced too much in the United States, particularly openEHR.org and a excellent effort that we're going to see a lot more from, Ethidium, which is openEHR, open standards with a business model that provides extremely low cost, highly secure platforms.

Not to make this a commercial, but I think we need to identify openEHR and some of the organizations working around that standard. [ Slide fifteen]

Bob Smith: This is in general the certification process that the HIMSS group certification commission for health information technology and right there at the bottom, underneath AHIC, NIST, Dr. Brailer's office and the four contracts, is this, (chuckle) There's several ways of pronouncing ON-Kit ... but this mess down here at the bottom, we have public and private sector stakeholders... weighing on top of this there's this nice, rational box ... with connections between the boxes.

Let's take a look inside this fuzzy muddle, underneath ... It's obviously not a cloud. [ Slide sixteen]

Bob Smith: This is an extremely complex effort where we need to map a person as basically a "token" flowing through the healthcare petri net with the boxes, the squares being agents and ellipses being opportunities to employ improved coding, improved ontologies, improved languaging.

RB: Thank you, Bob.

Here I want to briefly return to slides 2 through 7 of the combined deck and I will tie together as best I can what Bob has briefly mentioned. Because the overarching concept of this panel discussion of roadmaps, blueprints and landscapes ... these are really nothing more than useful metaphors for us to get a handle on domain mapping, navigating those domains, making rational plans about how to improve them, then. begin to get a grip on how we ought to be making those blueprint and building out a much better, much more cost-effective system such as Marc Wine was alluding to. [ Slide 3 full deck]

The chief proposition is that Brailer's Office of National Coodinator for Health Information Technology is now taking form. That's where the Federal Register Notice I mentioned earlier comes into play, because this is a very important aspect. It's no longer just words. There's actually some meat on the bones of this OMCHIT fulcrum of change, and now the large IT players are motivated to participate. Because I work with OASIS, and I'm on several technical committees, I can attest to the fact that IBM, Oracle, Sun, and BEA (which just bought Plumtree), are definitely taking notice of the healthcare standards market and now with the co-mingling of HL7 and OMG [ http://www.omg.org/news/meetings/HC2005/index.htm] , we're seeing some advancement of the notion of getting more coordination in the standards.

Bob referenced Gordon Bell's Twin Peaks to underscore the window for implementing bold new standards such as our Ontology approach promises. I'd like to us to address this question: "Where do we think we are between the technical innovation and the billion dollar investments that are about to stampede right over us?"

Because, if we don't help create these standards, they will be created for us by the people who have the biggest vested interest. And those are the major IT developers. My sense is that the big tech innovation was XML, in the late nineties and the first year, or first two years of this century.

But I could be wrong, and I spoke with Bob about it, and he is of the opinion that, perhaps, the advent of RDF-OWL marks the beginning of the technical innovation. I have to think that what I'm seeing in the standards world and the involvement of the major IT players, tends to make me think that the big investment is ready to happen sooner than, perhaps, we might all wish it.

And so my question to everyone is, "Do we need to build a problem statement about how and where and when the things we need to do, need to be done?"

Next slide, please? [ Slide number four combined deck].

RB: I wanted to address the key elements of a problem statement. And I wanted to say that we don't have a lack of standards, and we don't have a lack of standards organizations. What we have is a lack of adoption, a lack of coordination, a bunch of duplication and overlap and some grave problems with discoverability and accessibility of these standards and these are things that happen in the trough between the peak of technical innovation and the advent of the investment.

And, we're in the trough, is what I'm looking at. Next slide, please?

When Bob was talking about a framework, this is one of the things that he was alluding to, and what you see is an idea of how standards, as they come along, need to take advantage of the fact that we have ontologies at work in this whether we're aware of it or not. They exist by default, rather than by intention and we probably need to start being intentional about it.

Next slide, please? I wanted to show you this so you can come back to it and think about it later. This starts to give you an idea of how the framework expand up from the bottom, Next slide.

PY: You wanted seven?

RB: Uh No, Peter. I wanted to show Roadmap Constraint. I'm sorry Peter, I did get one ahead of myself.

I put this slide in from Bob's work because it illustrates the hellish complexity at work. This is an overview of the infrastructure interdependencies. I put this in specifically because Bob's thought about this and it is something that I had not really given a lot of thought to and it illustrates, kind of, some of the technical problems and if you put this beside the web standards stack and start trying to figure some of the complexities out, you can give yourself a big headache, but the primary thing I wanted to point is, "There has to be a way to navigate this." Next slide please?

And what Bob and I have thought about is the axes of healthcare at this point in time are: threat can be anything from a hurricane to a terrorist threat; time in the sense of what time of year it is or when it's happening, but also where in the course of, say the advent of technological change and where we are with standards and where ( axis-place) this happens, you know. Ah, is it at a nexus in the communications network, or the transportation network? Is it going to cause, uh, a big chokepoint in a nexus for the economy? And, then, as it relates to you and I and as it relates to our national concerns, is how do we deal with individual privacy and security as well as national security. So these are things we do need to think about.

And I just put this slide in here to show that there's more than a three dimensional matrix and we need to find a way to start plotting how things fit within this.

• MarkMusen

And with that we will then move to Mark Musen, and his presentation.

Mark Musen (MM): I'm Mark Musen. I'm head of what's called, quote "Stanford Medical Informatics" which is an academic group at Stanford University Medical Center and, my perspective, obviously, is that of the academic. So, I'm obviously the wrong guy to talk about business cases, but I think I can address very clearly, some of the landscape issue that Rex has brought up. And, I think this is a very exciting time, and I think that the original charge, this notion of questioning, "Are we at a tipping point?" is really quite apropos. And I think if we didn't have Dr. Brailer's office and even if we didn't have all these new federal awakenings to the importance of semantic integration and ontology, there's enough happening at the grass roots (so) that I think what we're seeing to a large degree is actually a response to the feeling that people throughout the medical community have and the solutions that they're (already) seeking rather than a mandate that, necessarily, is being imposed from the top down.

So, with that, let me go the the first slide, which is to remind all of you that what is really unusual about healthcare is that, although healthcare currently, probably, is not at the forefront of semantic integration, it is probably catching up very quickly in terms of being able to deal with the issues that the Ontology community is concerned with recognizing the importance of ontology, really, when you take a step backwards, medicine was one of the first sectors of the community to recognize the importance of ontology.

What I have on this slide is a very small snippet of classification of diseases which was the first compendium of all possible causes of mortality and death that really was an outgrowth of the industrial revolution.

In the 19th century as transportation suddenly made it possible for people with infections to easily move from one locus to another, the response on the part of government was to realize that without a way of categorizing how people may be infected and sick and might infect others in other communities, there was no way for public health authorities to be able to deal with the transmission of infection across political boundaries.

And it is really quite remarkable that medicine was probably one of the first groups, if you will, to recognize the importance of ontology as the basis of trying to get work done. And the international classification of diseases has persisted for the past 150 years and it is only now that the World Health Organization maintains in various incarnations ... and right now the most prevalently used version of the ICD is ICD Version 9, and there is a version 10 that is not very well used ... in part because it doesn't really capture the world as an ontology in the way that people would like. And I will talk a little bit later about plans for ICD 11, which is more responsive to the kinds of things our community would like to see. And that the World Health Organization now recognizes as being essential.

[slide 26 of the combined deck].

MM: That being said, that was all the good news. The bad new is that, until very recently healthcare as an industry has been most concerned with form but not content. Content has always been something that's been avoided. Probably the most important standards organization in healthcare is HL7, which was formed in the late1980s. And it's a way of standardizing messages that are passed among distributed computing systems in healthcare environments. HL7 has emerged as the dominant standards organization in clinical care. It's important to emphasize that when HL7 was created, the goal was to standardize messages, all the effort was put on the structure of the message. Basically, what were the necessary fields that would be used to communicate between one device or computer to another. And the whole idea that one might actually care about standardizing how the content of those fields might be organized was really beyond people.

And, it is only on the last five years or so that I think HL7 has recognized that content is just as important as structure. Perhaps more important. But HL7 certainly has come from a tradition of not really being terribly concerned about ontology per se. Along those lines, in the medical community, probably one of the most important initiatives to aid decision support and to facilitate the communication of knowledge, from one healthcare organization to another, was something which is called the Arden Syntax, which is essentially a framework for representing situation-action rules.

And Arden has gotten the support of almost every vendor in the healthcare information systems arena, has gotten support from huge numbers of academics, and was quite remarkable (because) when you look at Arden you see that it's a way of standardizing processes. It's basically a programming language. And it doesn't say anything about the data that the programming language is processing, it doesn't say anything about ontology at all, and, indeed, and one of the reasons I think Arden has been such a failure is that, although everybody eagerly embraced this notion of standardizing procedures, there was no attempt to begin to standardize content. Without an ontological commitment, Arden didn't really have the power to, ah, do very much.

And I think, now people are recognizing that the real power is not in the structure but in the content. And, obviously, that's the big gap that has to be filled.

Just in terms of showing where I think the community has come from, the American Medical Informatics Association or AMIA has an annual meeting and for the past ten years, the tutorial of that meeting that routinely sells out is the one on XML, where, people, again, are focusing on how to do markup and what the markup structure is, without necessarily focusing on content. And I think that is an indication that we are ... we've been in a situation where the emphasis has not been on the ontology, but on everything you need but the ontology.

But the world is changing radically. And I think this is where the landscape is really important and being able to identify where things are going to move. [slide 27]

LOTS of things are happening... that are making it clear that things have to change.

About five or six year ago, the Institute of Medicine released its very damning report called, "To Err is Human" and although lots of people have quibbled with the arithmetic and the statistics, the conclusion was that annually between 40 and 90 thousand people every year die in the US healthcare system because of medical errors.

And the whole problem with medical errors has been largely politicized ... obviously this has been a prominent component of George Bush's State of the Union speeches for the past couple of years, and I think, in large part, was responsible for the creation of Brailer's office in the first place.

One of the things that came out of the CP ... came out of the concern about medical errors, long before Brailer came around, was that, perhaps, if there was computer-based platforms that would allow doctor's to enter their order orders automatically, that one would bypass the chain of command and the chain of people who have to transcribe poorly written hand orders that we would eliminate a lot of those errors. And there has been this enormous momentum behind computer-based physician order entry or CPOE.

So that, just as we enter into ATMs our requests for money, doctors would enter into the computer their requests for what drugs would be given to patients, and, presumably, more effectively and with less errors. But, to do that, requires controlled terminologies if not frank ontologies of the drugs: of how they're to be given; about the indications for those drugs; and I think ... not only are those ontologies a prerequisite for CPOE, but the fact that state legislatures around the country have required that hospitals adopt CPOE means that those ontologies are necessarily going to follow. And I think that really changes the landscape considerably.

The other thing that became important in the public eye is the notion of ontology as the basis for facilitating public health. And, what I think was really most remarkable about the SARS epidemic a few years ago is that folks in Hong Kong were able to quickly identify the causative agent and the nature of the disease largely because all of the hospital information systems in Hong Kong used the same ontology for representing patient findings and patient illnesses. Obviously, having single-payer and, indeed, having such top down control over healthcare makes a big difference.

But what was remarkable, in my mind, was really how quickly the virus was identified and how quickly the virus was sequenced and how quickly the public health agencies were able to respond. And a lot of that credit, people have realized, goes the fact that a common language for talking about healthcare findings where the virus first originated had a tremendous effect on how the public health people were able to respond. And now, as we enter increasing concerns about other infections such as avian flu, but more importantly, the potential for bioterrorism, there's this real, palpable, understanding that we need to be able to integrate electronic patient records across institutions and, obviously, the only way to do that is going to be through ontologies.

The third thing which has changed the landscape in biomedicine iin the past five years, really has been the sequencing of the human genome and the ubiquity of high throughput techniques which have revolutionized biology. And when you look around at what's happening in biomedicine, you see increasingly biology is less and less a wet bench science and more and more a computational science. And the real problem is being able to handle the myriad data that come from high-throughput experiments, is having the right ontology to describe what those data really are.

And it is really quite... almost humorous, but it is really quite remarkable to see biologists suddenly asking for ontologies to help them make sense out of all their data, and recognizing that without the right ontologies, they won't be able to interpret their data, and the ability to take biological results and transform them into healthcare benefits will be stymied.

So these are sort of our three anecdotal areas where the people in the trenches are clamoring for ontologies, and regardless of what's happening in business, or regardless o f what's happening in government, that people are sensitized to this kind of need. Next slide.

And there are lots of very positive signs ... again, all of these predating what's going on in Dr. Brailer's office, and its quite remarkable to see companies like Apelon which market controlled terminologies, actually using the words description logics in their marketing materials and telling healthcare organizations that not only do they need ontologies, but they need ontologies with particular semantics. I never thought I would ever see that in my lifetime.

We're seeing very large scale ontology initiatives, tens of millions of dollars coming out of the National Cancer Institute for example, to bring ontologies, if not to the masses, at least to all the nation's cancer centers with hundreds of people around the country now working furiously on building the ontologies that are going to be necessary for the nation's war on cancer.

And, indeed, what we're seeing at NIH is that every institute seems to want to have its own ontology for its own set of diseases, of its own set of biological findings, that help inform the treatment of and understanding of those diseases. And I think that it is absolutely ubiquitous now that the landscaping is changing, where people what to see ontologies in use. Of course, as Rex said, that there are grave problems just trying to find out what ontologies even exist, trying to find them and trying to understand what people have already done, but it's really quite amazing that even without a mandate from the top, people in the trenches are asking for this kind of stuff, and demanding it, and of course, now the problem is how do you create the structures that will make those ontologies available, how do you make it possible for people to find out what exists, how do you make it possible for people to put to use the ontologies that will be useful to them.

That raises a whole set of policy issues that I'm sure we'll get to later in the discussion.[ slide 29 ]

MM: Just to reinforce that this becoming, if not a for profit business, this is becoming a large enterprise, the National Cancer Institute is building internally one the largest biomedical ontologies, which currently now has about thirty thousand concepts. I don't have time walk you through all the boxes and arrows here, but NCI has been perfecting, over the past decade, a very wide-scale initiative for distributed editors with particular areas of expertise to make contributions to what is one of the largest biomedical ontologies, one which defines very specific concepts related to oncology and cancer biology, and the clinical care of patients who have cancer.

And this is the kind of business, if you will, that will soon be replicated throughout NIH, and, obviously, will have great ramifications for the rest of the world in medicine soon.

The problem with the NIH approach, of course, is that it's highly dependent on very (highly) trained content specialists who are employed directly by NCI in order to do this kind of work, and there's been a lot of discussion particularly if one wants to go, not to ontologies of thirty thousand concepts, but, perhaps, ontologies of three hundred thousand concepts or three million concepts of how we can open this up in a way so we can have ontologies that will scale even better than what we already constructed. [ slide 30].

MM: And what's interesting, I think, there has been a lot of discussions around in the past year or two, of other approaches. A lot of people have looked at the Open Directory Project, for example, that was started by Netscape and has been picked up by number of vendors in the software community where now we have thousands of volunteers around the world who are actively editing their small piece of the open directory that drives a large number of web engines, and other software applications.

And I guess there's a lot of discussion that may not lead anywhere, but it is very intriguing for the panel to think about to what degree can biomedicine harness the intellect and the entrepreneurship of individuals who may not employees of any particular organization, but who have a vested interest in fleshing out ontological distinctions and may be able to be recruited to create the very large scale ontologies needed and that medicine is ultimately going to require.

And, obviously, right now, there is not an Open Directory Project for medicine, and, indeed when you look at ODP, what it says about medicine is pretty impoverished, but it raise the question of to what degree can we leverage the enthusiasm and the spiritedness of the people who care about these things to work on large ontologies for data integration in a very big way. [ slide 31]

MM: And, so... just to wrap up, there are clearly lots of places right ow where it's widely recognized that ontologies are needed that don't exist and there's a lot of hand-wringing about how to bring those ontologies to fruition. For example, there are not good ontologies for patient problems. The International Classification of Diseases, for example, is just that. It talks about disease. But healthcare works need to be able to talk about patient problems that much more vague and are not really specific diseases. People have chest pain and in fact chest pain can be caused by lots of different things. Shortness of breath can be caused by lots of different things.

People come up to doctors complaining of abnormal skin lesions without having a definitive diagnosis. And we don't have good ontologioes that allow us to talk about the ways in which patients want to describe their problems, and the way doctors want to describe problems that are not yet defined and agree not yet diagnosed but obviously need to have some sort of a name.

Again, following up on the IOM Report on Patient Safety, there's been an enormous bandwagon to try to develop automated systems that will facilitate or encourage best practices and enforce clinical guidelines to optimize patient care. And yet, we don't have any recognized set of ontologies that define how to represent those best practices and how to encode guidelines. And until we have those guidelines represented within ontologies, we won't be able to share them across institutions, and obviously that's an area where there's an enormous amount of hand-wringing but there also needs to be in the future a robust solution which currently does not exist.

All of his are aware of what has happened in the past couple of years, over patient suits for unsafe drugs, and the Merck Settlement is in everyone's mind, that there's an obvious need, in fact there's a mandate now to be able to store clinical trial results including those trials which are not published in the literature because the results are negative or those results for trials which are stopped because there are adverse reactions, so that later we can go back and actually learn about what's important about those trials even though they may not have led to something that the major journals will want to see in their pages.

And the idea is important. The idea has enormous appeal. What we don't have an ontology that everyone's agreed to that will allow us to do that and to have automated agents to do the necessary queries on those trial results. Ida Sim at UCSF has done a lot of work in this area, but what she has done is not regarded as comprehensive and certainly is not a standard at this point. ( http://rctbank.ucsf.edu/sim/pub.html )

And then, what everyone talks about ... the whole reason for the Human Genome Project was to be able to understand the relationship between differences in the genomic makeup of one person versus another to be able to predict how those people will respond to different drugs. The reason for how I might take a drug and do well and you might take the same drug and have a severe reaction is obviously a function of our genome and what we don't have are the standards and the ontologies that will allow us to take advantage of genomics in the practice of medicine. And this represents an area where clearly there's enormous support at the grass roots ... The ontologies don't exist, and if the right policies were in place and the right business practices were in place to enable us to create those ontologies, it would obviously be very important and have great benefits for society.

So, I'll stop there, and basically, I'll say that I think the landscape is really quite impressive right now. Obviously the fact that all of us are on this call, in some sense, is a marker for how the landscape is changing. And I think from the perspective of an academic, that the opportunities are tremendous, and I think that with the appropriate support of government, and with the right kinds of industries, we can see just enormous benefits with healthcare.

• RamSriram

Ram D. Sriram (RS): This is Ram here, I'm next.

I think Mark had a very comprehensive story of the whole thing, so I'll add a couple of things to this ... Again, I have a tough time in terms of doing the business case, so what I'm going to do is talk briefly about some of the activities we (NIST) are doing and how we should be doing next [ Slide 33 cDeck]

What I would view in this is the following, people are talking about interoperability of all the systems and then we have the ONCHIT initiative. Now what people really see is the final end result. Okay, this is interoperating as long as you can smoothly send ... seamlessly send data or information from one system to another system that's okay, and that's what people see. But there's a lot of stuff underneath as such. For example you need such things as terminologies and terminologies are the crucial thing for that to happen. And that we don't see.

And that's one of the problems we are going to face in the future, in terms of what you don't see, is what people don't understand, in a sense, okay? Because some of this kind of stuff is invisible to the external world, in terms of policy makers and things like that. So, somehow or other, we need to convince them that this is important.

I borrowed this slide from Chris Chute to show the importance of ontologies. So that's one level, and then you have one more layer below that. And that layer is the layer in terms of ... like ontologies and terminologies are one layer, and then you go one layer below and that's actually trying to understand the complexities of these things and trying to make sure you have metrics to measure these things as such: What is a good ontology? How do you test this ontology? Etc.

So we really have to work at these three levels as such. The first level is something that is demonstrable, and the second and third levels are something about which we need to educate people quite a lot. People do not understand. Now we, as a group, are people who can talk about ontologies and we all understand what we mean and things like that, so it's like preaching to the choir. But a lot of he policy makers may have a touch time understanding in terms of semantics and ontologies and things like that. They still think ontologies are something that has to do with being and philosophy and so on and so forth, okay? And especially, when you study the metrics and when you try to study the IT metrics associated with these ontologies you kind of ... many times come against walls as such.

And I believe that in terms of this field, we are where the Romans were with their aqueducts, about two thousand years ago, they kind of built all these aqueducts without understanding the science behind this whole thing. They just built them, and a few of them fell down on someone else's heads, I guess, and then they rebuilt them, and that's all ... that's the way things worked. And the science came around, some time like the 16 to 17 hundreds or so in terms of the strength of materials. And similarly, right now we are just building a few things and putting them together, and people do not clearly understand what's the science behind this.

And we really need to put that in front of our priority list, in terms of your landscape thing.

Now, I'll just add a few things to what Mark has already said, in terms of ... things like ICD 9 they don't really have, truly reflect the disease state as such. For example this 185, when you do it, use it for coding could be a whole lot of things. Like use it for both the metastatic disease and then with people who just have clinically indolent disease described as TURP , and so the problem is that even that is incomplete as such.

Next slide, please?

So you have to look at several layers around here. So in terms of our interest in the clinical informatics area, the program which I head is called manufacturing metrology and standards for the healthcare enterprise. And there are two aspects of this. One is the informatics aspect. Another is the device aspect.

In terms of the informatics aspect, there are a number of areas we are looking into like bioinformatics, health informatics, pharmaceuticals and so on. And the slide that I showed you shows our interest in the clinical informatics area where our primary focus is on electronic medical records, vocabularies and evidence-based medicine ... and I think from the previous slide and the slide Mark showed before, the case for these kind of things is clear in terms of vocabularies and evidence-based medicine and I think the formalization of these things and developing proper standards and so on.

And, of course, the problem here is that there are so many vocabularies floating around, how do you harmonize all these things?

And in terms of the electronic medical record, and that is one of the focus areas for the ONCHIT, there are two aspects, and one is actually building the electronic medical record. Second is testing the software. How does the software actually work? [ slide 36].

These are some of the things we think are useful, like using systems engineering techniques for mapping requirements to an appropriate architecture. There are a lot of languages for expressing semantics and different things can be expressed in different kinds of languages. For example, processes can be very effectively expressed in the Process Specification Language, and I expect you know about OWL and those sorts of things. And there is debate in terms of expressivity of languages and things like that which academics are looking into.

One of the major activities we do at NIST is this notion of, apart from understanding the semantics of the vocabulary, the notion of this testbed for testing both the software and the concepts. And Lisa Carnahan is working closely with HL7 on conformance testing and we have a performance website, and Steve Ray could probably tell you more on this and the various tools we have for testbeds for testing interoperability and conformance from the manufacturing engineering laboratory perspective. Do you have something you would like to add Steve? Is Steve there?

Steve Ray: Yes, I am, but I think I will just let you go ahead, and if there's discussion later, we can talk.

RS: Okay, then, finally we have these metrics that one needs to understand is measuring and calculating the information and understanding the expressiveness and the computability of the case. But that is more of a technical level as such. [ Next slide.]

Okay this is the summary slide, and I won't go into that. But the message I would like to give you is that we have about three levels we have to work with here.

One is the top level when you are interfacing one system with another system, and to support that we need the terminologies and ontologies, and, again, one has to make sure that these ontologies are indeed correct and appropriately tested and things like that. And we have to work carefully at all three levels and that is the message I want to put forward in your landscape.

I know I haven't gone into the business cases and things like that, but I guess we have other speakers who will go into that.

RB: I have to point out that the difficulty has only been in this process of trying to winnow what we can address.

• DavidWhitten

DW: Thank you, Rex. Just to give you an idea, I am here wearing my VA hat WorldVistA hat too, since I'm here on my lunch hour.

Basically, the last time we talked, we went into the details, about how the VistA system works and the various ways that and how the ontologies interact with the VistA systems.

PY: Before you start, just let me tell the other audience who are not accessing the VNC server, that if you refresh your wiki page, David Whitten's slides are posted, and you can click on that link and get the slides to run locally on your machine.

DW: I spoke to Marc, and you all recall Marc was involved with the Office of IT Sharing at the VA, the Veterans Administration. So we agreed that I would put a little bit of energy into talking about community, of the stuff going on with WorldVistA and the attempt to try to organize the community outside the VA in terms of the VistA system. And, just a real overview of some of the ideas that WorldVistA is thinking about, about these issues.

I think that the power of this talk we're having here today is going to be the questions and answering the questions, so I've only got five slides, as I recall.

In the first slide, I review the fact that WorldVistA is a 501(c)(3) Not For Profit Public Benefit Corporation. We try to get a wide variety of people involved. VistA is an open-source/public domain effort. The original code is in the public domain because it has been paid for by tax dollars and U.S. Citizens. The code that is going to be purely developed by WorldVistA has been agreed that it is going to be released under a GPL license. That's not on the slide, but just so people know.

Looking on to the second slide, [slide 39 of cDeck] we get a profile of what WorldVistA is trying to do.

We have over a hundred members. We've increased about twenty percent for each one of our meetings. We have between two and four meetings each year. There's a mailing list if anyone is interested. We have a series of mailing lists related to the various kinds of aspects of this issue, from the technical aspects of what makes OpenVistA software, where the VistA software works, to different ones where we're talking about people are trying to adopt it, about what issues they are running into about options. And, then, as we know that the Health and Human Services, through the Iowa Foundation is going to be releasing a version of VistA intended for smaller doctors' offices known as VistAOffice EHR. And there's a separate mailing list for that group as well.

I encourage anyone who is interested to involved with any of the electronic discussions. If you subscribe to the Hardhats mailing list, you'll see where every week we give the announcement of the conference call. We have an hour conference call on Friday, where we're free form usually, where we discuss what is going on with VistA and the experiences various people have. So we try for in person meetings, we have electronic conversations and then we have the weekly conference calls as well.

Looking at slide two, uh sorry, slide three, we have just a simple description of the things I've done in the past as long as I've been in this type of field.

The next slides is strategic challenges. [ Slide __ ] These are some of the issues we know are happening. These are not VistA-specific issues. Some of these are crises that are happening in the field we are choosing to work in. We've got a medical care crisis currently going on, most people understand. We have increasing numbers of uninsured people. The cost of medical care and medical insurance. Associated expenses are increasingly a larger and larger part of our gross national product. In the early nineties, the medical percentage actually went higher than the donation to the Department of Defense.

I'm not sure if those numbers are still true now, but the fact is that it is still a significant portion of our gross national product. Over ten percent, last I heard. And the last I heard was in the early nineties.

We're dealing with a software crisis that has to be developed in some way because of increasing complexity. We're trying to shift things from software into hardware but there's also a corresponding effort not to have everything in hardware, for microcode and various kinds of software solutions, so that there is a lot of flexibility in our current system.

One of the problems is that we're getting to the point where the complexity of our code systems is so great that we are not able to hold the whole process in our head. This is certainly true for VistA, but I would suspect that it's true for most of the operating systems out there. Any operating system that has any functionality that we would consider a necessity, nowadays, the way that it does things is becoming increasingly difficult to keep track of. So we have an increasing cost of software and hardware in terms of complexity and in terms of how do we manage these processes.

The Chaos Study [ Standish Group ] shows that although we have ways of dealing with things using examples and patterns, but we don't really have a true understanding of how to solve it. In terms of medical informatics, I think Mark Musen gave an excellent overview of many things going on. There was a study by ACBE [ ?? ] that showed that medical informatics, if you compare the amount of money that's been put into it, versus the amount of impact and change that has effected the entire economy, it's been very disappointing.

A lot of money has been spent, some of it on proprietary solutions, but some of them were just on conflicting solutions. So if you look at the ratio of the amount of money put in and the total benefit, it is actually one of the most backwards sectors of our U.S. Economy.

The VistA system is generally recognized to be an exception to this. But, there's not a good explanation why.

Going on to the next slide, [ ] we give you an idea of why we think the VistA system has been widely accepted as an exception to the problems we've seen in medical informatics. One aspect is that if you're going to have a quality environment, you're going to have to take advantage of the "net" effect, the idea that as you have more people involved, they feed off each other and they're able to work together more effectively because they don't all hold the same amount of information in their heads, but (where) a single person may not be able to understand the complexity of the system, perhaps a group of people would be able to understand it, whether that system happens to be a computer program or whether that system happens to be a medical delivery system. The second idea here is that we see an increasing view from people who love to formalize to think that this is actually a product, that developing a computer system for medicine is a product, that you can put it on a shelf, and take it off the shelf and it does everything you want. I believe Marc will agree with me that there is a lot of efforts that have to be done to maintain and to keep alive ontology-based systems. And this is true no matter what kind of medical system we have.

Medicine is a rapidly changing field, and it's no surprise that the computer system that supports the medical system has to rapidly change as well.

So we deploy a process. We are not putting forth the concept of doing a product solution. We think that the product enables the medical care delivery system, but it must be flexible, it must be changing, it must be a continuing, sustainable approach to developing the software system, to developing the ontology that is the foundation behind it.

We believe that it's not simply a matter of just teaching somebody in school for few days and then they now are a medical informaticist. We feel that the whole culture of medicine is such that you really need to be involved in the culture itself, if you are going to be involved in the medical information system.

Finally, we feel that the software improvement process will continue only if we're looking toward an evidence-based, user-driven, and continuous improvement.

RB: Amen.

DW: The final issue is just that if you're going to have to have high quality, you're going to have to have people looking at it, no matter what the case may be, no matter if it's running a gasoline station and always pumping a gallon of gas whenever anyone is pumping gas, there has to be certification checks and we believe that's a necessity for medical information systems.

• BrandNiemann

RB: Thank you, David, and we want to move on quickly to Brand's ...

I appreciate being asked to participate. I've enjoyed the discussion, the presentations, and I appreciate coming last because I've been making notes, and updated my presentation based on some of the things I've heard.

I want to talk about the four things on my overview slide. [ Slide 2 ] First a bit of history, or actually even preface that with saying that I think we have been trying to work on the business case for over a year now. And, I'm going to provide a little report out on some suggestions on what should go next.

Little did SICoP and Ontolog know that when they partnered about a year and half ago, that we'd be taking on this grander challenge, but thank goodness we are. I think it culminated recently in our meeting with ONCHIT and while we didn't come away with a funded project in hand, I think we came away with maybe something even better in the longer run, namely they didn't say, "We don't want you to do that," they just said, "We're not ready to do that." And that's typically been the reaction, in my experience now, across the government, the last year or so, with the semantic or ontological approach to enterprise architecture and interoperability.

So, I'm greatly encouraged that we, at least, are able to work our way up to the meeting. And , I think, that indicates that we do have some business case in the minds of those people, or we wouldn't have even gotten an audience without some business case and some traction.

I think the point I left with them, uh wanted to leave with them, is in a slide you do not see , but that basically, that this is getting more institutionalized now... in the GSA has hired Marc Wine and that we continue to do pilots and expose them in our workshops and public forums.

So that would be the first request I would make, keep feeding us the wonderful pilots you have, because we need to keep populating those monthly and quarterly meetings with that, to keep that in front of people.

The second point I would make is what I would describe as a marketing point, while ontology is really catching on, we've come a long way from where people were reticent to use the "O" word, I think it's really "interoperability" that we really want to lead with. So my second point here is, we first want to explain the difference between integration and interoperability. I think the Pollock-Hodgson book does an excellent job of it, and it really is brought out in the Australian article that was sent around recently. And secondly, I think that our marketing phrase should be something like "Ontology provides agile, adaptive interoperability," so lead with interoperability and that ontology is really the enabler for the kind of interoperability we're trying to achieve.

Another point I would make with standards is that we met with Steve Ray's people recently because we're working to standardize the Data Reference Model where we are working standardize it in terms of tools and alternate expressions, et cetera, and Steve Ray's people made the point that, really, it's not standards per se that sail, it's the implementation of them, particularly in the government.

And that leads me to my third point, that ZapThink recently has come the attention recently of the Federal Enterprise Architecture particularly because it is considered to be the Gartner of SOA, the Service-Oriented Architecture, and they've made a point to us recently that applies to this area as well, that we need tomove, just as they realize in SOA, they need to move beyond talking continually about what it is, and start focusing on "how-to-do-it."

And I think that has been our emphasis with the pilots. We have to lead by example: how-to-do-it and then, of course, what is the best, most effective way to do that, and we're trying to do that in the Data Reference Model, you'll see in point three. Ontology, I believe, we have raised to a significant part of the Data Reference Model. You can see that will be part of almost, if not all, six pilots that will be given at our second public forum for the DRM, September 14th, which will attract at least 300 people and we'll have to turn away any more beyond that for space limitations.

And it is becoming part of the tool discussion for the DRM, that we're having with the vendor community and, ah, as well as the government community, et cetera. But that leads me to the next point I'd like to make, that we have to make strategic alliances to pull this off. Beyond NIST, the ones that appear to be the most obvious are the Markle Foundation, Esther Dyson ... I was interested in finding as I went through my folder for this that there was the article that was sent around that mentioned this fellow at Perot Systems who was favoring the open source approach. We might have overlooked trying to partner with hi, in fact, Rex, in relation the Red Hat Licenses as that was mentioned iu that article. In Government Health IT it says, "HHS keeps options open" article quoting Richard Pico, chief medical and technology officer at Perot Systems, Healthcare Division, wanted to keep open source in the mix, et cetera. [ see the July 6th Government Health IT on "HHS Keeps Health Options Open."]

Oracle, of course, since they've announced support for Ontology in 11g, but that will bring me to my last point, I really think that a point may have been overlooked in the Australian article, was the one that relates to the final point: "How Do We Make the Business Case?" I will quote it, it says:

"Larger, more complex organizations should do nothing that cannot be done as well and as cost-effectively by smaller, less complex organizations."

I think that's very consistent with three principles we have been following in our workshops:

One is what Tim Berners-Lee gave us at the SWANS conference where he talked about the constant tension between working locally and working globally. That refers to harmonizing vocabularies; and,

The second is what we hit upon early in the Ontolog-SICoP pilot, was that hierarchy of ontologies, it takes large and small organizations working to connect that hierarchy of ontologies all the way from the top down which is the government enterprise architecture to things like HL7 all the way down to individual doctor applications, that is clinical applications, et cetera; and, then lastly supports the principle we're following, in the government on this, a development of lines of business or profiles.

We might want to think about, while there is a line of business for health architecture, beginning to think we might want espouse a profile. in the sense that we have several other profiles in the federal enterprise architecture ... we have profiles for geospatial data, for records management and for security and privacy ... I was thinking we should have a profile for health information and data interoperability.

We might float this because it's really not being done anywhere in the federal health architecture yet... and I'm not even sure the federal health architecture is broad enough to deal with it when you're talking about personal health information systems. So I toss that out to think about that as we make our business case.

I think our business case should make the point that's made in the Australian article, that there's a role not only for big organizations, but smaller, simpler organizations and that's really what we are. ... That there is the need to work both locally and globally ... and the Semantic Web and its standards enables that ... and that we're on the right track with hierarchy of linking ontologies vertically and we really are in a position to do a profile or what's called a profile for this area and that might be the best expression of the business case that we could come up with.

RB: Thank you, Brand.

• Open Discussion

RB: I think our audience now has more than enough to think about, while we move on to our discussion.

And since we haven't had people asking questions during the course of the discussion, we do want to encourage you to ask while you can, because we only have a half hour left in the time we set aside for this, and I want to start with the question: "Is it now important to proselytize the business value of ontology more effectively?"

This picks up right where Brand left off, I think its kind of apropos ... and turn that out to the floor ... Should we be proselytizing more effectively, and how should we do that?

DW: I'm an advocate of the idea that when you're trying to convince somebody to go a new way, or adding more effort to their life, you need to show them some of the benefits of it. You know in terms of medicine, as was mentioned earlier by Mark Musen, medicine has done some things with ontologies and controlled vocabularies and various kinds of taxonomies ... for years. So the issue, I think, for medicine, is not clearly the case that there aren't things out there that can do it, rather it's trying to understand that medicine is a huge field that it has a lot of complexity, and that the ontologies are almost necessary to make it, actually, a manageable field.

So, if we're trying to proselytize the idea of using ontologies, then we need to have a strong focus on the fact of how these things can really change the practice of medicine ... and... or the practice of information systems. I'd like to hear Mark Musen's ideas about this as well.

MM: Yes, that's really an important point. And I think it resonates well with what Ram said, that we have worked on ontologies in medicine for a very long time. (However), those ontologies have always been disconnected from the kinds of activities that are actually going to bring value to patients most directly, and that really means decision support, and those kinds of activities. I think it's very interesting. I just had a conversation last week with Bedirhan Üstün from the World Health Organization who is responsible for the next generation of ICD 11. ( http://www.un.org.tr/who/bustuncveng.htm )

With ICD 11, they've realized that ICD 10 needs to be replaced quickly. And what is fascinating is that his business case to WHO is that ICD 11 is needed so that the terminology is not going to be standard or link directly to decision support services, to systems that will enforce clinical practice guidelines and best practices that will inform providers and patients about the relationships between different diseases. And, so, what I think is really very exciting now is that everyone in ontology business so to speak, now recognizes that the next win is going to be, not in creating new ontologies, but creating ontologies that are linked to other services that themselves are valuable.

RS: Yes, I think that's a very good point.

RB: So we want to be able to show and demonstrate the value of linking ontologies to practice?

MM: Absolutely, I think Ontology is still and obscure kind of word. I resonated with what Brand said, that we have to be focusing on what the solution buys, rather than on the solution itself. And I think that if we view ontologies as the means by which we can finally allow best practices to be enforced within healthcare organizations, by which we can allow integration of medical information across healthcare institutions, by which we can finally allow a patient's medical record to follow him or her from one institution to another. Those are the kinds of things that people will become excited about, and there's really only one way to achieve this from my perspective.

DW: Dave Whitten again, and I think that one of the ways in which ontologies can come into practical play are in some of the ways in which just verifying information, and it's already stored in the database. Looking for cross-dependencies for fields that are there and recognizing the fact that if you have stored information about locations, perhaps, but that may have an impact on their medical care ... Looking for issues of just making sure that the information has been entered appropriately ... where it doesn't make sense to have somebody in with an inappropriate age, perhaps, or someone else has an age that doesn't make any sense ... Those thing look like they don't have any impact on care, but the fact is that all the normal ranges for laboratory information is all dependent on age, the normal values change depending on the gender, if the race is correct, or the age is correct. These are obvious ways that ...

RB: So we could say that ontologies are ... do better data validation?

DW: Absolutely. Ontologies are one way of doing better data validation. These are one way, as Mark mentioned, of just finding the information you need. You know it's not just the case that we want to find information about a particular patient, but we also need to find information about the particular kind of care scenario that the patient's in. You know, in the process of your selecting the symptoms for someone, if you have a system which is indexed properly, you can try to take some of these symptoms, some of these problems that the patient is reporting with, you can use that to give suggestions to the care providers. I mean, one of the things VistA does, is these clinical reminders.

Again, we're not trying to tell anybody how to practice medicine, but the field has drastically changed over the last few years, to the point where, if you don't have some system to inform somebody about what's going on, it's not gonna happen. They're not going to know about the study that was done six months ago in Massachusetts ... double blind, showing the impact of that particular study on the care of this particular patient.

I'm not suggesting that they spend all their time being computer people, doing research and not dealing with the patient, I'm suggesting that the computer has a role in terms of supporting them in the practice of medicine, in the delivery of adequate and valuable care.

RB: Well, I was writing as fast as I could and I got down that we have dependencies, we can improve data validation, we can improve quality of service, and we can improve the outcomes and we can reduce costs.

DW: We can also provide a supportive system, a decision support system to the actual clinical providers.

RB: Decision Support?

Bob Smith: Yes. This is Bob, and I'd like to urge that we go one step further down into the market, and not just provide the providers, but the consumers. I would like, ideally, to have maybe five questions that I could ask of a potential provider to make sure that my data is going to be handled correctly, safely, and effectively. And that's the results of a good ontology. And, it would be great to have a diagnostic, a differential diagnostic, that would indicate what steps, and how mature they are, and what they can't do yet, and that they're promising to do. There's a lot of purchasers of healthcare, that I think, with the national health information network, at the talking level, and particularly, Esther Dyson's movement, seems to be tapping into a tremendous volume (of interest) and the thought leaders, to be able to simply have for decision support, a finite set of differential questions that would be useful for the consumer, for illustrating just how sophisticated or ineffective their health providers are. [1]

RB: Okay.

DW: Related to what Bob is saying, is the stuff done by Octo Barnett ( http://citeseer.ist.psu.edu/shortliffe96intermed.html ) right now, over at University of Massachusetts General Hospital that is very powerful, but it's basically providing the differential diagnosis kind of information he's talking about. It's actively being provided by them, but the problem is, it takes a lot of updates. It takes an awful lot of people paying attention to puting information in, then, of course, there's the presentation which is targetted toward medical providers so they use the true jargon. They don't try to make it accessible to the end user.

RB: Okay, the root question there, we went from assessing the business value of ontology strategies to thinking about agenda items, for how, you know, we can enable consumers to identify whether or not their health providers are taking care of their information well. That actually covered several of the areas we had questions for, and, so I'd like to move on while we have time, to one of the next two questions. So I will read them both and you guys can decide which one you want to jump on.

One is: "What viewpoints and what critical assumptions need to be considered and honed for our business model, for our business value proposition? And ...

What models and meta-models are available to guide our policy level debates in appropriate forums? These two are tied together basically because a ot of what Bob was saying, that if we are going to enable the conversation across all levels, in terms of getting support ... at least what I took from what Bob said was that we need to do to drive the market is to empower the consumer to be able to demand more efficient health IT from their providers.

And, you know, demand drives supply. Most of the time, as a business case. So do we want to talk about viewpoints and assumptions or models and meta-models?

PY: Can I go back and ask Bob what were your five questions ... from a consumers perspective?

Bob Smith: I don't have the five questions. Dr. Musen?

MM: Yes?

Bob Smith: How do you select your physician? To make sure that they're ontologically sound, have an ontologically sound infrastructure, that they know what they're talking about, that they're keeping current with the literature, that they're using knowledge management portals, effectively? Have you formalized your criteria?

MM: That's a great question, Bob. And the issue is that there is no way of getting that information. I think the scenario where most of my colleagues who practice medicine and which happens routinely, is that a patient with a problem will march into the physician's office with printouts of thousands of web pages and ask the provider what he thinks about all this.

MM: It's very scary because physicians are increasingly finding themselves less informed than their patients whom they care for. But if you ask me how do I find a physician who is going to be able to respond appropriately, who is well connected, who keeps up with the literature, who's invested in the technology, that's really hard. There's no card-carrying members of that kind of a club.

On the other hand, I think physicians, because they're scared of the fact that the Internet has been an incredible leveler of knowledge, I mean basically what has happened with the Internet is that the guild of netizens, if you will, no longer has control over its own knowledge, that there's been increasing pressure from physicians on the part of ... or rather that physicians to have the healthcare organizations that employ them, to provide the kind of technology that allows them to stay one step ahead. I guess it's another example of where pressure from the grass roots is actually going to change the business case.

Bob Smith: Yeah, I can ask my doctors and health group what IT system they're moving towards. They purchased Epic ( http://www.epic.org/ ). they're planning a roll out in two months to three months. The people that I've talked to are terrified. My primary physician is in his residency at a Veterans Hospital, and he describes how just tremendously impressed he was with VistA.

So as a consumer I can start asking questions about their health infrastructure: what vendors (do they use)? what models (are the vendors using for their application)? and (shouldn't be able to) kick the tires?

MM: Yeah, I think that what's always been amusing to me is this has been a driver at Stanford, our house staff prefer to do their rotations at the VA where they have access to VistA, rather than working at Stanford Hospital where they use a commercial information system whose name I won't mention.

(Chuckles.)

RS: Mark on that note, there's a lot of physicians who, in my opinion from the physicians whom I have contact with, and most them don't use any (word or acronym unclear on recording--?CHAIR-Coordinated Health Administration Information Resource?) at all.

MM: That's actually right.

RS: Especially, you know, in rural areas, even a place like Baltimore, they don't even know what there is, well, they know what there is, but they... I just think they'e too intimidated by everything, so ...

I think unless, somehow, in terms of from policy level, there's some mechanism to incentivize people to use this, they're not going to use it. If you look at it carefully, for every (patient visit) they get about $60 dollars or so, for about half an hour of consultation, and if you the math thing, let's say you work about 8 hours a day , you see eight times twelve, you see about 800, 900 dollars, but if take all your overhead off, you make about 400 dollars a day. And with that, they're busy the whole time, so where's the time to input information into the computer and things like that?

MM: The point is well taken, Dr. Brailer has suggested that there be pressure put on the Center for Medicare and Medicaid so that somehow reimbursement to providers is better if they use an electronic patient record, in a sense compensating them for their time. I guess I'm skeptical that simply manipulating market forces is all that's necessary, but that's certainly the attitude that the administration is taking.

Bob Smith: My sense is (that) the people at (?CHIME?) the HIMSS CIO team are getting increasingly sophisticated, and looking beyond simply the hardware and the software, as they make their decision between (?SERNER?) or VistA or EPIC in that there's a lot people in HIMSS, there's twenty thousand or so healthcare providers, who have bought the $110 dollar study by Blackford Middleton are basing their assumptions on what it's going to take to convert from a paper-based system to a level two, to a level three, to a level four which is fully interoperable... The problem with Jan Walker and Eric Pan and Blacford Middleton reports, is that they are silent on ontology. I talked with both Jan, briefly, and with Eric on the phone, and they are very interested in understanding ontology, so I see a milestone of being able to present a strong elevator speech and look at their literature, of where and how the cost of what they expect, $200 billion dollars to implement Dr. Brailer's plan and realizing a $278 billion net result, leaving, if their math is correct, a $78 billion dollar surplus, by making the changes over a five year period.

So I see those reports and digging into those reports, exposing their awareness of certain things, and lack of awareness of certain things, as a good agenda item for discussing in the near future.

RB: That brings up a specific point, since we're now down to ten minutes, and I wanted to move on to a couple of other things. You guys actually did hit the strategic assumptions that need to be considered... and honed, and a little bit about the meta-models. But the fact is, that in the course of this exercise, we've identified a lot of the landmarks in the healthcare informatics landscape.

And one of the things we haven't touched on a lot is that, while there are a large number of standards, and there are quite a lot of standards organizations, neither the standards nor the organizations are terribly useful because we don't have the ontologies that can map the standards to one another, and can identify for people which particular standards they actually need, when you have two standards that address different aspects of the same, uh, field, for lack of a better word. So, one of the things I put into the first part of the slideshow was the fact that blueprints will be needed... that we don't have many blueprints, so this discussion sort of needs to come to some conclusions about what we need moving forward. But, before we get to that, I would like to say that on Peter's service, I'm going to be building a public service preparedness portal that will have a registry within it for emergency management health information and there's no reason why that can't be expanded. So, I just wanted to let you know that there's a testbed on its way that we'd be happy to provide. or at least one part of it.

Pat Cassidy: Okay, that would be my cue to come in and comment. This is Pat Cassidy, and your comment about there not being any coordination standards is precisely the issue that's been of concern to me for a long time.

As most of you who know me know, I think the solution is that we really need some kind of wide agreement on a common upper ontology that can serve as the defining vocabulary for all the domain ontologies people will be building, building independently using the same fundamental defining vocabulary, you will achieve the interoperability we need.

For me this is not just a desideratum, and I know a lot people are skeptical, but having worried about it for quite a while, I haven't seen anything that comes close to approaching the functionality of a common defining vocabulary. And no plausible alternative. People talk about mapping. I've never seen a mapping that's the slightest bit of use to anybody. And if anybody knows of one, I'd like to see it.

Bob Smith: Uh, Pat? Can I ask you a question?

Pat: Yes.

Bob Smith: Looking at Steve Ray's work at NIST, on PSL (Process Specification Language), and Michael Gruninger's work on PSL, how does that map up to your criteria?

Pat: Well, it's interesting, of course. You know they do PSL, and they do good work on formalizing processes, which is a part of an upper ontology, and I would consider PSL to be a part of any good standard upper ontology. And Mike Gruninger has a paper in the Spring 2005 issue of AI Magazine, titled "Semantic Integration Through Invariants" in which he talks about, hmmn, interoperability. It was a special series on Interoperability, and rather than talking about an upper ontology, he says that you want to map all your ontologies to an "InterLingua." So I got a chance to talk to Mike Gruninger, after I read that article, and I said, "Mike, what's the difference between an "Inter-Lingua" and an upper ontology?" And he says, "Well ... " (and you know Mike is very concerned about having very careful axiomatic, precise, definitions and minimizing the number of possible models), says, " . . . in an interlingua you have a sufficiently rich axiomatization that only the intended models you want are the ones that are actually represented."

Okay, so Mike doesn't like an upper ontology, but likes an interlingua, which seems to me like an upper ontology on steroids, perhaps with extensions. So you look around and there's a lot of people who say the same thing, that to get interoperability you need a mapping to some kind of common basis, whether you call it an Inter-Lingua or an upper ontology. I like to use the word these days, "Common Semantic Model," but the big question is how to get there.

I have no doubt it can be done, but the difficulty is in trying to organize a coordinated program to do this. I'm hoping that with the kind of money and resources going into the NHIN ... any little portion of that that could be set aside to build this Inter-Lingua or Common Upper Ontology or whatever you want to call it would be a big help ... that's what I hope.

Bob Smith: Hmmn, that's a good seque to Brand Niemann...

BN: In the interest of time, I want to jus to pose two questions for quick answers.

Do we know what the status of our abstract submission is to the November conference? And, if it looks like we're going to do it, then I would suggest we lay out a series of at least two or three more calls in which I will volunteer to put forward a business case strategy strawperson, but it would essentially follow the model for business cases in the government and would draw on each of you to provide me with what we started to do in the start of the discussion, namely that this is where ontologies could be used, and this where' they could be used, et cetera. ... that I would need specific examples of those, as well as, if possible, the ROI, the Return On Investment for those because if we want to ... we ultimately want to do a business case, and the first iteration is a strategy we roll out in this panel at the conference, and I think that's all well and good. And my inclination is to follow the pattern for such business cases that have been developed in the government. So I will help do that. I think we need to come up with an outline of what our business case strategy would be, and we would parcel out the components of that to each of you, especially the participants of the panel and even if we don't get it accepted for the November conference, I think we're on the right track to put forward a business strategy on the wy to a business case.

BN: Well we've already met the deadline for the submission, but I haven't seen when we're going to hear, when we're going to be notified.

Bob Smith: Okay, that's the November Conference you were talking about.

BN: Anyway, that's an excellent opportunity for us to say that in our panel we're going to roll out our strategy on our way to a business case. And that's where you'll hear ...

BN: I am at least going to borrow from another community of practice we're partnering with at the Enterprise Architecture Conference, the one Pat Plunkett leads on Federal IT Performance Measurement and basically you've got three categories and a methodology for doing that, namely, working to demonstrate the following: reducing redundant IT applications, increasing interoperability, and the facilitating of business improvement.

I'm not suggesting that by November we can actually calculate dollars, but I'm talking about addressing what we've done in these three categories, that have been decided to be performance measures for IT investments ... qualitatively, what we've been able to accomplish. Hopefully, that's just an interview or a thought process.

DW. Okay, I just want to be sure that we have a measure that makes sense for all of the healthcare systems available to us. Since I'm advocating OpenVistA and VistA as a system, it's primarily used in public health in that kind of environment, it has different characteristics to it than those that are used solely in business ... hmmn ... non-governments, non-public-health type environments.

RB. Depending on who has, who among our audience, which has been very very quiet, and patient I must say, considering we are now overtime, that there are probably some other people that could, ah,

contact us and provide us with information that would give us multiple inputs for that.

DW: Exactly.

PY: I have a question to some of the members of the audience, that as Rex said has been quiet, particularly to the NIST people, if they have like quantitative measurements and studies that have already been done, that would be helpful.

Steve Ray: I was going to defer to RS on the healthcare side.

RS: I think the question is more general than the healthcare kind of thing, I believe...

Steve Ray: Okay, well. Certainly, for a general case, we have done a number of economic studies, both on the, hmmn, missed opportunity costs, if you like, or (costs incurred) through the lack of interoperability. Those kinds of numbers go across several communities and range typically in the billions of dollars per year, whether it's in electronics or manufacturing, or construction in the capital equipments industry, I'm sure the numbers are at least as large if not significantly larger in the healthcare industry as well.

We haven't commissioned a study like that for the, if you like, the wasted costs in the healthcare industry due to lack of interoperability. But that, actually, is an interesting consideration. They typically take couple of years to do.

RB: What we can, if you have some examples of that, is cite those examples and say that there's sufficient similarity between the, ah, you know, fields, to make some assumptions about what ROI in terms of reduced costs would be.

Steve Ray: Absolutely, and I can provide both the studies which are called perspectives and studies in terms of the wasted expense and some ROI work on things where a given standard has saved the country x millions or hundreds of millions of dollars.

RB: Yeah, and I know there are some excellent statistics on that from the Interior.

PY: And then, on our recommendation, we can say that these are studies related to other areas, but that we would recommend setting aside some funds to do both a specific qualitative and quantitative economic study on what kind of results we can expect.

DW: Let me comment here, I believe I was told that the Markle Foundation did studies of questions related to interoperability , but I don't have a specific reference. They're www.markle.org.

BN: They've released their report and we've looked at it. I also want to mention that I have access to the government agencies' business cases submitted to OMB and while I can't reveal all of the specifics, especially the numbers, we can essentially say that we've looked at those that relate to health IT investment. We can talk about them in generalities. And we have some specific data we can refer to here.

DW: This information is not public, you're saying?

BN: It's in FEAMS, the Federal Enterprise Architecture Management System whose access is tightly controlled.

BN: But I have access to it, so I'm looking to hear from everybody who's on the phone, if either they've have an experience with the system, or they can tell us what its been and if there's been a role for ontology or if there is something that they are thinking about doing and what they think the benefit would be. That's the thing I'm looking at initially. So we go to a summary coming out of this conference call, to an outline for a strategy for doing the business case where everybody inputs something, whether they've had actual experience or they're thinking about doing something that they hope would have an ROI, so that then, if we get that far, to really build a more rigorous business case that's what OMB is trying to do.

Michelle Raymond: This is Michelle Raymond from Industry and one of the Return On Investments I'd like to see is the middleware. So, if we're providing systems in centers how do I take it to my management that says, "It is a really good idea for us to be involved in this ontology work. It's a really good idea for us to be in the forefront of ..." This is something I'm convinced (of) but how do I convince my management?

RB: We need to marshal our arguments there.

Michelle Raymond: Yes, yes, but I think that needs to be when you're talking about Return On Investment, you're not just talking about the Return On Investment to the caregiver, to the government, but to the industries that are supporting medical specifically, but also those of us who are actually going to be actually creating the systems. And if you don't have us involved, you're not going to get anything into the market.

RB. Yeah, that's what we just asked for and I'm pretty sure we can guarantee some response.

BN: Let me mention something that's very exciting, that happened recently, where you could help us, Digital Harbor rose to the top of the 40 vendors in the SWANS Conference, supporting RDF and OWL. We then launched into a series of pilots. They do what are called composite applications, with multiple ontologies, et cetera, and we've launched into a pilot with them for four use cases, two of which have been delivered. Mike Daconta, Dick Burk, Kim Nelson and all the important people in the government, have now seen it, and Mike Daconata has said "this looks like its the killer application for the Data Reference Model."

And you'll see it again at the September 14th (Data Reference Model) Public Forum, but we are looking for a medical use case now along with several others, so, I would ask that you look at the Digital Harbor presentation ... it's posted at web-services.gov ... it was just given Monday to the Data Reference Model Working Group at the insistence of Mike Daconta and others. So look at it and suggest to us a medical application because it handles multiple ontologies, multiple rules, multiple events. It's a very amazing platform, one which I didn't think would exist for another five or ten years. So that's an existing development that could just accelerate the adoption of an ontology approach in government.

RB: Okay so now we have two things, the business case, for which we also want use cases, and the Digital Harbor composite applications, and we have the argument we want to making that will cite some studies that have been done and that similar studies need to be done to be funded for this arena, for health IT. And we want to get as specific as we can about asking for input from the audience here.

And please come back to regular calls and help us out with this on places in the economy where these advantages and benefits can be, you know, realized.

PY: Can we slip in one more issue?

PY: I would like to ask those who are working on standards, are the healthcare standards open enough? In terms of the participation in the development of those standards to the adoption?

RB: When it comes to HL& and OMG, to the people who are going to engineer the toolsets, and applications, that's a pricey club to get into. OASIS, on the other hand, tends to take the more international approach, in the International Health Continuum TC, which I can sat something about, is due to be re-invigorated or renewed, and there are several ongoing projects that should be coming out of that. Quite a lot of the steam went out of that when the European Union had its constitutional crisis. But that's sort of becoming background noise now and not dampening everyone's enthusiasm about the TC and things moving forward in healthcare.

DW: This is David Whitten again, one issue which I didn't hear addressed was the idea that Pat was putting forward about type approach. the tie between the specialized ontology of medicine and a more general ontology, the upper ontology. Do we feel that an upper ontology approach will help our case? Or do we have any opinion about that at all?

BN: That was in our presentation to ONCHIT.

PY: Right, I mean, all along based on the three and half years of conversation, there's sort of general consensus of the ontolog community that some sort of common upper ontology is a better approach.

Of course, it still boils down to specific applications, but yes, the answer is yes.

Transcript Revision History

• Live! (current) editable version above (on this wiki page).
• 2005.10.21: this page created. The merged version (RexBrooks's original, Pat Cassidy edits & Bob Smith edits) is posted above. {GY3}
  • Participants review and enhancements solicited.
• 2005.10.18: Bob Smith sent in his reviewed and edited (cleaned up) version. - with BobSmith's's edits
  • editor's note: my intent is to convert Rex's transcription into a short working document with links to slides; and then to position both Panel content into a useable/useful format documenting costs and benefits of ontology in large projects. (Without boiling the entire Pacific Ocean) --BobSmith/2005.10.18
• 2005.09.09: Pat Cassidy sent in his edits (mainly cleaning up the part where he spoke.) - with PatCassidy's edits
• 2005.09.09: Rex Brooks finished and send in the full transcript of the entire session. - Rex's first cut/original