Symposium Series: Dr. Edward Shortliffe – The Amplification of Informatics Challenges



and and it gives me an excuse to talk about you know you can't talk about amplification till you talk about what you have before you start amplifying so there's kind of a historic perspective to this topic you could argue that we're always amplifying what we do as new needs arise and as the field evolves and as the technologies change and so it's kind of a natural state and informatics to be expanding and amplifying the basic methods and techniques that we make use of so I will I will use the era of genomic medicine to make some of the points but I want to put this in the context of the field as a whole one of my goals today is to try to synthesize across biomedical informatics as a discipline some of you most of you have not seen much of this before but a few of you will recognize a few slides and things like that but they're part of a an effort on my part to remind you that there's a very big picture here in this field and we sometimes end up very focused on our little areas or our little projects and yet it all fits into a kind of continuum both historically and also across a very broad spectrum of areas of application it does make it hard sometimes for us to wrap our our minds around the full extent of the discipline and I'll acknowledge that and tell you a little bit about how I think you could try to get a feel for as a student what you really need to understand about the field as a whole but what you also need to be able to convey to folks that aren't in the field and are trying also to understand much more about about what informatics is and not to be a lot has happened very quickly i happen to be just grey enough that i have been around almost from the beginning of the field and it's been quite of a white quite a wild ride when I look back especially because of the explosion in the technology as well as the underlying scientific discipline right so if you go back to about nineteen eighty-four I love to use this slide because it was absolutely the first time I ever saw anything in the general medical literature that suggests that doctors and computers might go together that was an ad for an antihypertensive agent in the New England Journal of Medicine okay it wasn't an ad about computers you can tell it's pretty all just look at that computer fact i'm not sure that computer ever actually existed i no idea what it was but doesn't look like any pc i know so we went from that maximizing patient benefits that's actually a pretty positive portrayal of computers in 1984 with doctors that's at a time when not too many doctors felt very positive about this I whole idea and by 20 years later that was in businessweek the doctor had moved from the user to being actually in the machine somehow as we began to look at what all this this computing and medicine men right so there's been quite a progression and there's still a lot of uneasiness as well as recognition that it really is a force that cannot be ignored so when did it start as some of you have heard me claim that probably if you had to pick a kind of seminal event that started the field and most many of you were in it would be this article back in nineteen fifty-nine I was published in science by ledley and lusted they were two researchers wanted dentists when a physician at NIH and they wrote this article called reasoning foundations and medical diagnosis and now that you've been students here for a while and surprise you if I tell you that in fact what this article did was point out that there was a statistical basis for diagnosis and that there was this thing called Bayes theorem that could be used for doing diagnosis and also they were computers they could do the calculations for Bayes theorem and maybe we should be looking at how computers could help us do medical diagnosis and it's fun to read that article which is still widely available because it's been reproduced over and over again in books and the like because of its historic importance the reason I say it's kind of the starting point is that if you do citation analysis that means looking backwards for all the current literature to see what they cite and then go to those articles and see what they cite and they go to those articles do with a site and so forth you begin to see dominant influences where the citations just make it really clear that they kind of had an ancestral importance for the for the field and in the case of Informatics this article is like the granddaddy article of them all okay and we really began to see computers being used quite a bit in biomedicine starting in the 1960s mostly in hospitals mostly to manage patient data so these were clinical systems there was a first NIH study section looking at grants the kind of grants people wrote were very applied they were just trying to figure out what they could do with computers and managing large amounts of patient data so over time we began to see a discipline begin to emerge from these this early recognition the computing had a lot to offer to that world of healthcare the word informatics finally came along and was pretty much adopted by the mid-1980s we started hearing in the late 70s but it really became kind of accepted in the 80s it's borrow from European languages informatica is that word for computer science in French and people liked informatics because it seemed to emphasize the information and knowledge content of a discipline rather than the computers per se and I'm getting a little tired of the exploding interest in big data because it seems to me we've always had big data this is just a current buzzword one that seems to be attracting dollars from NIH agency so it's hard to ignore when that happens but if you think about it the reason they started using computers at hospitals in the 1960s was to them they had this huge amount of data that they were trying to deal with me had developed the methods that allow them to actually adapt early database technology for purposes of managing clinical information and those of us working on it and decisions support and expert systems in the 70s the complexity of those problems had to do with the incredible amount of data that we felt we were having to sort through and under stand all the clinical research systems that kind of emerged the 1980s similar issue to them that was big data and of course we have the Human Genome Project pushing into biology in the 1990s where we really were beginning to see even more impressive amounts of data that needed to be managed and so in each of those decades big data seemed to be the problem we just didn't call it big data like that today it's maybe it's bigger now I think it is we'll come back to that so around about 1985 those of us that were beginning to do training in this field said you know it's really hard to teach this field using xeroxed articles and of course that's there was no online resources for distributing articles in those days so we were handing out readers of articles in journals and finally we decided it was time to actually create a textbook it took several years it finally came out in 1990 and it was called medical informatics that term by 1990 was pretty well accepted and then it was computer applications in healthcare and I'm going to use this book to illustrate how the field is evolving because each addition we face new challenges in making changes to the book adding new topics pulling out some of the ones and so just worrying about a book like this forces you to think broadly about a field like this one because the book is intended to be the introductory course and I think you guys use it in your intro course here too right not this version I hope so that we we divided the book from the beginning into the methods and the applications and it's interesting I don't think we were as aware of that importance of that distinction right in the early days but the first part of the book is about methods and these were the only you know kind of method or any other chapters in the first edition medical decision making a methodology of that something about computing data systems how data are managed and essentials of computing and system design there was a chapter on evaluation and that was kind of it for methods and then there was a lot about applications and they're all very clinical right medical record systems hospital information systems which was still the way we refer to these systems back then we have systems pharmacy systems radiology systems clinical research systems health assessment systems the health assessment was the old multiphasic health testing when you actually put people through kind of assembly line of history taking and physical examination done by machines as well as by some clinicians and out the other end you came with a very full assessment of how you've done okay so throughout we've had to deal with the fact that we didn't know what to call the field probably the first thing people call it with simply medical computer science we had a group at Stanford when i was there called the medical computer science group it basically did what you all do here but that was the name we had for it somehow another it bothered people though including those who were using that name because it's focused on the technology when we all felt that what we're really doing was dealing with something much more fundamental in the way of information and knowledge processing and and use and good decision-making and that that's really what the core the field was now some people say well we would better get information emphasized then so let's call it medical information sciences and when we started the training program at Stanford that was actually name of the degree for many years but it has this problem that its associated with libraries and library science and the like and people got the wrong impression about what field was about with that name too and that's why we ended up by 1990 with a textbook that was called medical informatics but of course things didn't stop there during the 1990s a couple of things happen one was there were there were sub groups of people in the community that didn't like the word medical in particular it seemed to leave out non-physician health professionals and second it seemed to de-emphasize health promotion and public health as opposed to seeming to be too focused on disease so there were those who started to call the field health informatics instead of medical informatics then the other thing that happened the 1990s of course was the human genome project and the exploding explosion in biology and by the end of the decade it was clear that you could not do life science research without computers anymore into the 80s and 70s there was lots going on that didn't involve computing at all and that's been a hard transition for people who were wet than scientists starting to produce huge amounts of data with the new techniques that emerged during the Human Genome Project and not having the skillset necessary so suddenly they were looking for postdocs and PhD students from informatics programs that can work in their labs and help them actually get their arms around the data so you will have then seen the then moved to the word biomedical informatics I'm going to talk a little bit more about it in a second but basically was an attempt to say we have to acknowledge that there's this broad spectrum and it doesn't it isn't just clinical anymore there's this important biological component to the field and recently I've even heard a fair number of people refer to the entire discipline as bioinformatics wrong okay so when you hear people say oh you're doing bioinformatics and you're working on an app to be used in the clinical setting by physicians that's not bioinformatics is it and do we all agree on that but I can tell you at ASU I have run into lots of people that refer to this department is bioinformatics they don't have it straight what the difference is and I'm going to talk about that now there was a feel that emerged during the 1990s called bioinformatics these are both from early in the new century journals books and the like there's no question that informatics methods were being used in the life sciences and they called it bioinformatics but part of the promise of all this in fact the motivation for what was going on in the human genome project if you think about it was we have to start to understand how the genetics of individuals affects the way in which we're going to diagnose disease choose treatments and basically personalize the way in which they're cared for so there were tremendous potential synergies between what goes on in the genetic genomic proteomic world and what goes on in the clinical world we saw that the same methods and techniques had broad applicability certainly if we have a problem with terminology in the world of clinical medicine we also have a problem with terminology in the world of genomics and genetics and the interdependencies seemed very clear in fact everybody was working towards creating interdependent interdependencies and yet as the academic groups and the training programs began to evolve we saw this schism that occurred between the folks who were doing bioinformatics and the folks that we're doing so-called medical informatics okay now it's true that many of the people doing bioinformatics were fundamentally biologists who were motivated by the biological problem primarily and they've grown up in a life science environment maybe not even in a medical school maybe not even with patients in mind but if it were working more on the basic science end of the spectrum and not even that related to the hell Campus of their University whereas the medical informatics folks were essentially all in medical schools or nursing schools and nurses by the way I didn't like calling a medical informatics either so they tended to use health informatics for the field so that in an effort to hold the field together to keep these centripetal forces from allowing it to kind of break up there was an effort to rename the field by the early part of this century as biomedical informatics to Bridget and departments like this one have explicitly tried to make sure that you both are taught about the full spectrum and also that their faculty that have expertise across that spectrum all doing informatics but doing it in different areas of application so this is the part that some of you will have seen but not all think about this field and by the way this has become you heard I was at a mia and i fortunately got this i did get it instilled in the ami of mindset while i was there and so this is the way that a mia the american medical informatics association which of course should be the american biomedical informatics association right but it was named in 1988 so it's got its got that history behind it this is the way they agree we should think about the field and this is the terminology that they're using okay that there is a basic science discipline then that's typically what we named apartments after is the discipline right and that's biomedical informatics it's a set of methods and techniques and theories and they grow over time and you learn new ones and create new ones in response to the growing needs of the application areas which you work but there's a whole lot of effort which we might call applied research and increasingly practice of the discipline and as I've indicated the clinical world is where most of that started clinical informatics but very soon after it got started in the 1960s we saw more and more use of informatics methods and dealing with images and radiology departments became very computer-oriented in the high-tech compared to the rest of the health science environment as the pathology departments quite early pathology often involved with imaging as well we saw bioinformatics emerge for the biology applications by life science applications during the 90s and in the last decade is when you've really seen the exploding recognition of public health and public health informatics as another area of application okay so by this diagram bioinformatics is one of the many application areas of the field of biomedical informatics and clinical informatics is meant to contain within it nursing informatics dental informatics and today medical informatics is being used pretty much to refer to disease oriented physician oriented informatics activities okay so we don't call the field medical informatics anymore week we talked about medical informatics some hospitals do they're pretty much talking about doctors and disease and certainly not public health and the like okay okay so we all agree this these are not the same thing and you now have an explanation you can give to people who use bioinformatics inappropriately to refer to the field all right so it's a broad broad field much broader than many very interdisciplinary can't work in it unless you're collaborating I I don't think I've ever seen a really good one person informatics project does it make much sense does even dissertations so molecular and cellular processes the imagers more at the level of tissues and organs systems so there's obviously radiology pathology hematology dermatology visually oriented fields Jimmy working in that area clinical dramatics the entire patient public health informatics populations and society okay now less we pretend there's no such thing as health informatics everybody I know who wants to call the field health informatics is working in applied clinical or public health informatics or there or there either researchers and practitioners in that area they do global health they want to call it health informatics so a mia and others have agreed health informatics is essentially those two boxes if you had a department of health informatics you would not get any bioinformatics faculty to join your department they would feel that this was about the other end of the spectrum here okay so these are also not synonyms all right so we did have a second edition of the textbook come out in two thousand still called it medical informatics but you'll notice that the subtitle was now not computer applications in healthcare but it was computer applications in healthcare and biomedicine because it was pretty clear that things were changing by the way in 1990 not one mention of the internet in the book 2000 every chapter okay that's how much things change in one decade so in the second edition of the textbook the new things I've underlined on these slides were new not not in the previous version and an indication of the evolving understanding of the field so the methods had started to get beefed up we'd added standards as an important issue because of all the integration interoperability concerns ethics legal issues Technology Assessment so new chapters in those areas and on the air and the case of applications things really started a change instead of medical record systems we talked about computer-based patient record systems those of you who know the early reports that introduced the concept that the Institute of Medicine may remember that the medical record systems were called computer-based patient record system CPRS okay public health chapter consumers information systems and integrated delivery networks patient care systems that used to be called nursing informatics as written by nurses but they didn't want to make nursing the point of the articles it was about patient care that chapter imaging systems new information retrieval systems MEDLINE search and all the things that followed and last chapter and a chapter introduced on bioinformatics this was in two thousand so it was written in 98 99 okay at the end of the Human Genome Project efforts now recently jamie has published this amia statement of the definition and the core competencies in the field since all many of you are being trained in the field this would be a good time to look at this article if you haven't see if you think the core competencies you're requiring are a good match with what is described in the article okay when this was an effort by a group of people working with and then finally approved by the board to define what ought to be in a graduate education program it's not for certificate programs it's not for ten by ten programs is what you should be learning if you're going to be a masters or PhD in this field and this is the desk of the definition an interdisciplinary field by the way every word in this was selected with great care it looked like three days to develop this definition a disciplinary feel that studies and pursues the effective uses of biomedical data information and knowledge for scientific inquiry for problem solving for decision making always motivated by efforts to improve human health and the emphasis on human health means that the interests of biomedical informatics in the bioinformatics area tend to be more on so-called translational bioinformatics you know there's not a lot about fruit flies going on in this field except to the extent that fruit flies allow you to develop methods that are important for human disease okay now there are four major sub-themes that they also identified in this article corollaries if you will to this definition first one is that there really are theories and methods and processes in this field the notion that biomedical informatics people carry wires through the halls string them in the ceiling and make sure you have cat5 in every office wrong okay that's not what this feels about I think everyone in this room knows that but a lot of people do not understand the difference between health information technology and computing support and the discipline that this department is based upon and I trust all of you were learning it does have although it's not in that definition I gave you but it does have an inherent connection to computing and communications it'd be pretty hard to do data information knowledge and there in the modern era without some deep connection to computer science right third that it's really broad from molecules to populations with a variety of biological systems bridging basic and clinical research and practice and the our healthcare enterprise including disease prevention and health promotion and finally this is a human oriented science the users are humans too and that means that unlike some pure computing science programs you've got to learn a lot about Social Sciences behavioral sciences that's the reason that we see cognitive science increasingly in these departments and healthcare financing and the issues of privacy security and security from a obviously there's a technical side of security but there's also a human side of security the trade-offs and the like so social educational financial organizational theory these are all work flow analysis these are increasingly viewed as key parts that of the field all right well I gave you a kind of simplistic diagram and the world has to acknowledge that there are no sharp boundaries among these areas right and many of you may feel in fact that you're working at something which crosses these boundaries it's pretty easy to think of examples by molecular imaging for example which is increasingly a big part in radiology departments is clearly got elements of bioinformatics and imaging informatics in it consider the whole burgeoning area of pharmacogenomics which clearly requires looking at both genomic and phenotypic data on the same patients and bridges clinical informatics and bioinformatics highly translational science in its orientation right consumer health has got elements of patient care it's also had elements of population health and prevention the like so it's a kind of at a boundary between some of these areas as well and you could i'm sure think of other examples in fact i'm going to give you another one later what about the education well you look at that diagram and if you're a student in this program ask yourself what have I learned about the various things in this diagram and I would hope that you've been exposed to things both at the methodology level and at the applied research level and that's you even if you're working in clinical informatics that you've had some exposure to by bioinformatics and vice versa but you're all doing projects you're doing research and research in this field requires informatics innovation certainly if you're going to get a PhD I would argue most masters can also be quite innovative at that level so if you're working on apps that's what that's the bottom of the diagram so ask yourself on every app what am i doing that's a contribution informatics the app itself probably isn't but think about what you had to do different in order to make that at work and you'll begin to tease out where your informatics contributions are and what I hope you'll make a point of writing about that's how you advance the science alright so this third edition of textbook came out in 2006 you notice the name had changed to biomedical informatics still computer applications in healthcare and buy medicine of course it had gone from 600 pages in 1992 about 1,200 pages and it like doubled in size and I can't imagine what this next one's going to be out coming to that later and the methods that were emphasized in this one new methods cognitive science natural language and test pro text processing in retrospect how did we ever not have them in there in the first place you know they're pretty crucial areas but our understanding of what needed to be included evolved over time and now it's not computer-based patient record systems anymore it's electronic health records and its public health informatics and the nhi I the national health information infrastructure consumer health telehealth bioinformatics still there okay now so that in a sense is what we're at where we are right now and this is where we need to talk about amplification in the era of genomic medicine that's our goal has always been and this diagram would have been just as apt when I started working on my first EHR in 1968 as it as it is today this notion that there are providers caring for patients and that they need to be recording their data so that the data are handled in a much more apt way and and you've all heard the stories of my paper-based records are a mismatch with the with the modern era and have been for 30 or 40 years really now you could argue that that's all you need to draw and you've got an important contribution people using electronic records putting data and being able to get them out getting test results out making better decisions because they have better access to the data that they can't get well from patient records they can communicate using it with other providers and indeed that's a really important reason for putting in EHRs but if that's all you do with your EHR the office of the National Coordinator says you are not making meaningful use of your EHR right at least not not enough so everybody's heard the meaningful use phrase and they basically says there's more all those data that are being collected in electronic health records are potentially useful for biomedical and clinical research and we need to get them in a more facile fashion into the kind of databases either regional or national or increasingly even the international the disease registries that inform us about what's going on is that there's this notion that the medical record has locked up for too long what we know from what we do and anybody who has ever done a medical research project with paper charts sitting on a desk looking through them trying to fill out data sheets all that kind of sub knows that can't be the best way in the modern era to do this kind of thing we can extract evolving standards over time for prevention treatment management in general and those can lead into the creation of protocols guidelines educational materials the kind of things that kind of summarize what we've learned from our experience which then feeds back into information systems decision support systems reminders and order entry systems and the like and actually has an impact on the way in which providers care for patients now this cycle which which we once knew we wanted to accomplish but have no idea how to do in an era of networking standards development system integration hie health record banks and the like we're beginning to see how this might actually happen right now this has been called a learning healthcare system a system that actually learns from what it's doing it's not randomized control trials although it can also feed into rcts but there's so much that we can learn through careful data analytics of big data and these are big data by the time you do it across large populations that especially in those areas where RCTs are either too expensive or take too long or the like there's what there's there's knowledge to be gleaned we could do better the Institute of madison's had a whole series of reports workshops on a learning healthcare system this diagram I don't know if the diagram has been used in them but it's the it's the notion that they're increasingly identifying as being key the future now in the midst of all that we're seeing this exploding interest in actually beginning to use all that genomic information Peggy Hamburg they're the head of the FDA personalized medicine coalition lots of websites you go out there and you start looking around and you'll find tremendous amount of information about personalized medicine some of it is generic leeson like the ones I just showed you here but then you start finding universities and got websites this is the University of Utah and their efforts they have huge genetic databases that they can use for these purposes in Utah because of the way in which date have been collected through the mormon church for years stanford by the way i just picked a few there are many many of these out there and then I pick Mayo because they and they're not alone have chosen not to call personalized medicine but individualized medicine there's also precision medicine out there these are all basically the same notion so let's come back to this diagram and think about the implications of that kind of combining of genetic information with the rest of what we know about a patient in order to make good decisions about their care and you'll begin to see lots of new little research challenges remember things that you order on your patient or that other physicians or other practitioners have ordered on your patients go into Data Systems the orders go into the data system the data are collected and they are eventually stored in the EHR and you access that EHR when you're making decisions about your patient that's the data you want to see in order to make decisions and maybe we'll even add some decision support so that you can get help figuring out how to care for those page but we are now talking about a whole new kind of data to go into those data systems and then from them into the EHR data of a type that's markedly different from getting a complete blood count or a chemistry or even an x-ray report okay because of what's happening in the world of genomic medicine the research world out there we have to rethink what goes on in electronic health records what we store which parts of what we know we capture today what part should be left out left for the future what do we do with primary as opposed to interpreted data I'll come back to it it affects the way in which we're going to need to give decision support you know clinicians know they don't know this stuff this is getting a little complicated and most physicians have not been trained in genetic medicine and there's new results coming out all the time you don't know whether or not the cost benefits trade-offs are reasonable and so the notion that you would get some help deciding what to order and how to interpret it is increasingly being accepted in this area and much of that would be blended into the order entry environment right so the order entry systems have to reflect this new kind of knowledge the registries are going to start storing different kinds of information there are new questions about standards and the kinds of guidelines of protocols decision support fodder that you produce through them from this process get more complicated as well right so as we push to move from the bottom of the top of this diagram the digitized EHR from the es old paper charts you can look back on what's we're in the midst of right now pretty remarkable change over the last decade it's about almost a decade now since president bush announced in his State of the Union address he actually used the phrase electronic health records for people who've been working in the area for 30 or 40 years to hear the president actually utter those words was remarkable okay now he said everybody was going to have one in ten years that's next year I wouldn't hold your breath for that to happen in ten years I don't believe President Bush had any idea what would be involved in trying to get a you know a chicken in every pot EHR in every doctor's office etc but the concept kind of got popularized and he mentioned it every year in the speeches as did tommy Thompson who was the secretary of HHS and his successor and by the time we went to the 2008 elections it was pretty clear that there was bipartisan support for this idea certainly Obama was all over this issue as a candidate which is when I got this picture back before his hair turned gray and so the vertical line you see there is the passage of the Recovery Act in February march two thousand nine and what this diagram shows it's probably hard to read from the diagram itself but this is the rate of increase of jobs since the Recovery Act was put in place and as you know and you know that job market has not done all that great so purple line is the growth rate for all jobs in the US economy this is through the end of last year health jobs are in green health IT jobs all right red so I showed this the Keith Lindor and he says yeah well maybe there really are jobs out there for the graduates of these programs the demand is definitely up okay and there's a belief that although it won't happen in 2014 that the penetration of ehrs is going to be extremely high over the next five years penalties start kicking it in a few years for those who have not implemented so those who are recognizing these issues here's an article that was published by folks who held a workshop at the NIH on the integration of genetic test results into electronic medical records recognizing there are a lot of questions about how to do this when I give you an example it kind of things people are thinking about here they had seven major desiderata that they define we have to some are another separate the primary molecular data that we generate by actually doing genomic testing from the clinical interpretations of those data because we know that the interpretations are going to change so if you store it in there just as the interpretation you can't go back and relook you have to support lossless data compression from primary molecular observations to the clinically manageable subsets everybody knows what that means and the notion is you need all the data compressing but you can't lose anything you have to maintain the linkage of the observations to the lab methods that we used to generate them because the lab methods are changing and your ability to trust or believe or interpret may very much be affected in the future by how you know that data in the chart were gathered in the past measured in the past you have to support a compact representation of clinically actionable performance so the stuff that we do think is clinically actual now better be in a form in the chart that allows you to quickly generate the kind of decision support that's necessary even though you're attending to those other issues at the same time and you have to make sure these are human viewable yeah I mean there's got to be some human being and its ability to have faith in them to understand the interpretation to know why they're getting the advice they are we have to understand that in the future there will be more fundamental changes in our understanding of human molecular variation and that has to be kind of anticipated in the design and we have to recognize we're doing this not just for the care of patients but also for future discovery science research right so we have seen some big changes in the last decade or so as the computing technology that was kind of in its infancy for so long finally began to really take off in healthcare environment so there are enough people here who are physicians who have been around a while that you can remember when this was the picture you've got a piece of x-ray film and you held it up to the light and put it up on the screen box and that's what you looked at x-rays and that is pretty rare now we're in an era now where almost everything is digitized and those big folders of films are starting to go away and we're now seeing new kind of viewing mode allottees and capabilities that are changing the way in which people practice and I use radiologists here because they've tended to be the quickest to adopt the technology among all the specialties so the question going forward of course is yes we're still going to have doctors taking care of patients and doing bedside care but are they going to in fact be looking not only at their traditional lab results but also at their genetic test results trying to figure out how to take that into account in the way in which they select drugs or advise that regarding prognosis and the like all these kinds of information are going to require new decision support functionalities and there is a big opportunity for new research for people that are interested in decision support in the world of genetics and genomics there are also big challenges for the privacy advocates in this world it's been recognized for the last decade or so do you want your genetic information out there there are people that have had their genetic testing done for research purposes but don't want to look at the results themselves thank you very much I don't want to know what I what my risks aren't various things of course the challenge compared to our traditional concerns about data privacy or data confidentiality patient privacy is that we're worried about not about what's true now you know that you saw a shrink or that you've got an HIV positive blood test but what might happen in the future maybe not even to you but to your children that's the privacy issue that now is and it clearly affects the systems the way in which you share data what it means to do anonymization of data sets alright so let's acknowledge that there is a big exploding interest in Big Data and medicine is often used as one of the prime examples especially the genetic end because the numbers there are truly remarkable so what do we mean a lot of people are written about it i've been reading several things this is an interesting report it tries to do some definitions IBM published it and they said in short the term big data applies to information that campy processor analyzed using traditional processes or tools which kaiser invites the question how do you define traditional because traditional changes all the time so for a little more clarity there are three characteristics that define big data volume variety and velocity and that these are all taking off tremendously and that's the reason we are now identifying this as a new kind of problem but i would say volume variety in potentially velocity we're always issues that have been challenging at each era as I described earlier rather than confining the idea velocity to the growth rates associated with your data we suggest you apply the definition to data in motion it's actually the speed at which they move around now in this world not just the rate at which they accumulate that is referred to by velocity and for those of you who don't know what Hadoop is because I sure didn't that's what it is but it's the it's the IBM solution and I have nothing to do with IBM this just happened to be a useful source but since I showed you a slide that has Hadoop on and I'm going to tell you what it is how many of you knew what Hadoop was already there are a few all right I'm impressed this is a week old two weeks old right I mean this is really getting to be a big topic and now we've got a request for information about how to do training in response to the need to take these big data and derive new knowledge from the kinds of data that are exploding in the healthcare environment so it's an NIH initiative all right I said I talked about the fourth edition of the textbook this is what I've been working on for the last few years there are changes you can go to my website to see this if you want but I won't go down all the changes I want to emphasize these ones at the bottom translational bioinformatics new chapter and clinical research informatics another new chapter okay there are few other new chapters just as well there's a chapter on health policy and I T which is probably overdue certainly a big topic in the last decade all right so trans so there's still the bioinformatics chapter so there's bioinformatics in this translational bioinformatics so why is this suddenly taken off the clinical translational science awards that the NIH introduced about seven eight years ago now have transformed the ways of which medical schools and others are thinking about the role of informatics they required informatics if you wanted to get one of these grants these grants are big money there's kind of Club you're in if your school has one of these awards there was an effort by the University of Arizona with a couple of rounds to get them but it was collaborative and it included this department and those and those proposals pretty hard to do a statewide CTSA and I think experiment just looked a little too risky to NIH so we don't have a CTSA in the state of Arizona but Mayo does at least based in Rochester and these have had a big impact on the recognition that biomedical informatics is here to stay it's a crucial part of the core competencies of a clinical translational science organization within a medical research environment and it's not just to support clinical research this is an interesting article that argues that in fact you need to be able to look at some of the human health databases in order to do good basic research with mice and alike okay say these are human disease and data resources that that complement the research activities of those working in these basic biological problems so they go back to my earlier diagram translational Sciences is actually across the entire bottom of the diagram it goes into public health community based research imaging and basic biology and there's a meeting amia does some of you know about this one it's two weeks but it's been every year for several years it's called the joint summits on translational science it's an informatics meeting it's three days of each topic translational bioinformatics three days of Clinical Research informatics they overlap one day in the middle because there's a fuzzy boundary and that one day everybody is there for some people stay all five days anyway but the point is is this continuum as we move from the more basic bio bio informatics components aimed towards human disease its prognosis its diagnosis its management pharmacogenomics obviously as part of this field and then clinical research informatics when you begin to get into things that support the clinical research endeavor clinical database management systems and eventually those that take you out into the community now some of you may have seen this diagram that I've borrowed from edited a little bit Bob greenness that is but it's this cycle of community feeding back into our kind of prioritization of the projects we want to where we need new new knowledge and new understanding disease leading to the identification of promising opportunities and then their eventual application in specific projects and taking those into the clinical setting for validation that's called t1 translation translation number one bench to bedside research t through and t3 translation have to do with moving it out into the community and then making its standard and accept it okay this is what the ctsas are supporting and the NIH and it's wisdom and I think Elias aronia was the director of NIH earlier during the bush years gets a lot of credit for that recognize that you need to have informatics to do any literally any element in that but it's all over our ability to do modern science and therefore needs to be part of any active program and that's how you achieve personalized medicine it's in the context of that kind of cycle and the role of informatics so I'm going to close just by pointing out that there was a report in 2009 from the National Academies on basically trying to inspire computer scientists to find great projects that they could work on his computer scientists in the world of medicine and there were many examples I've emphasized the first that they emphasized in their report which was we need much more patients than her cognitive support we need to do cognitive science in the context of these systems they did a lot of site visits a lot of the people on the committee or computer scientists but there were many informatics people and physicians as well and of course down here data management at scale it's part of the Big Data story so I showed you this diagram many times the notion of the methods at the top of the diagram the applications at the bottom all of us in this field are interested in some aspect of applied informatics they're trying to solve problems that we see in the clinical or biomedical domain that interests us and when we find those problems they inspire us to work on a solution and occasionally the solution is really simple it's a good applied research project okay but often there is no easy solution it's a complex world out there in biomedicine and you often have to go back to the drawing board the top of this diagram and innovate and when you innovate you draw on whatever components science you feel has a potential solution to the problem you've identified and sometimes it's computer science and so we learned about computer science science as biomedical informatics students and faculty and like researchers and occasionally the solution that you come up with solves a great problem at the bottom of this diagram but it's also great computer science and you publish in the computer science generally make a contribution to computer science so you can do that as a biomedical informatics researcher if you're well trained in computer science and you understand computing and you recognize the generalizability of the solution that you've that you've come up with but not all the problems are computer science and that's why we don't want to call this field medical computer science because its medical lots of things it's biomedical lots of things right it's the decision sciences statistics decision analysis cognitive science obviously talk about that a lot information science is more generically management sciences live students want to take courses in business schools when they study this field not just start a company lots of reasons management and other component sciences organizational theory I mentioned earlier okay so we need more people few programs like this get around the country aren't producing enough we need clinicians to know more about it so trying to get this into the medical nursing curriculum we need informatics people who really can function effectively as scholarly collaborators with clinicians and life scientists that means you've got to be part of the culture of health care and biomedicine which is why computer scientists who haven't had the exposure to biomedical informatics sometimes find it difficult working in hospitals or biomedical settings they haven't got that acculturation to the field they may come up with very clever computing solutions that are insensitive to the realities of the way in which practice occurs we need to recognize as this integrative role of informatics it touches all areas of biomedicine there's a large workforce requirement that I suggested on that one diagram as you look at the need for people out there as we implement all these electronic health record NH IT systems there are implications for the education people in this field and that is you need a broad exposure to the content areas and the informatics applications work across the entire biomedical field even if you are purchased personally focusing in one area and as I've suggested you should always be asking how you can demonstrate broad applicability of a solution that you've developed in your own work what's it what what other problem would it be good for think about its its range of applicability that's how you make a contribution beyond the specific project that you worked and as I've suggested as a field grows as this one has it tends to splinter and if you have the big picture perspective you can try to wrap your arms around it and keep that from happening and departments that that cover the entire Ranger clearly explicitly try to do that to keep those synergies in place because there are synergies you can avoid reinventing things that one community is doing the other one doesn't know about and my favorite example of that is the gene ontology which was developed without any understanding of all that we knew about clinical terminology development and the problems that we encountered many of which were rediscovered in the in the bio area and you can resize that you really are working on shared goals and motivations and you can design your departments so that they are inherently about partnerships and about coherence across a discipline and identification of those methods service to the field but at the same time being innovative and trying to make sure the field has the ultimate impact so those are comments I wanted to make a run over five minutes but I will be happy to answer questions thank you hi so my name is Valentin dinner assistant professor in the department and I have a question regarding the integration potential challenges in integrating genomic and clinical data and i'll start by saying but i teach a translational by informatics class here at the AC on the spring and just a few weeks ago we're just doing a demo of 23andme website that stores genetic information and it was interesting to see what 23 and we could see about me just from a sample of saliva that I have blue eyes curly dark hair hopefully still dark but also for example some interesting pharmacogenomic information I would be a regular respond better to work for it and a student asked head your head you send in saliva it was oh I see so you had your own data in on 23 exactly okay oh yeah so a student asked how about integrating all this nice interesting data and they give you a lot of information about many diseases and what will it take things a great base data with your each other with your cleaning with your doctor and actually one of the student the student was his question is still here and I don't think I had a very good answer to her and that would be very interested to see what you think our is your opinion on base integrating so I mean the the damn Macy's article that I showed you was about the technical challenges and issues in doing that but if that was based on the assumption that you would in fact have the genetic data for more and more patients and and in fact genetic testing being done as part of your medical care rather than because you did your own swab and presented in the mail to 23andme is likely to start raising the issues that then Macy's and his crew wrote about in that article when you start talking about what happens when patients are starting to collect data on themselves from various sources you could argue is that really any different from collecting data about yourselves from multiple providers and you know that's in a sense what we've all done if you've changed positions are moved or all those those things this notion that maybe patients or at least those who are capable of it and not all patients are but that patients need to be much more in charge of the gathering combining managing providing access to their own data that suggests some kind of central mechanism by which they can do it it's not Microsoft health fault or or Google Health it's definitely not Google Health but I'm increasingly a believer in the notion of patient controlled health record banks that centralized data there was an early effort to get one started in Arizona didn't go anywhere i know but i think the politics has not been obviously optimal but we may well see there are ways that those kinds of problems can be attacked the technical issues are identify with the ones that damage is raised it's just that suddenly you're you're now talking about what happens when it's the patient in some sense some more regional resource that is the place that's trying to decide what to do with all your genetic data from 23andme rather than because it was ordered in your honor in an institution where you're being cared for watch JAMA next week you'll see an article on this thing I and a couple of other people wrote lines penalty no more during this when you think about integrating this genomic data you think about like I get myself sequence and I've got a terabyte of genomic data that it somehow gets associated with my electronic health record but I can receive a point in the future where physicians don't make their diagnosis based on signs and symptoms they make it based on the translational your transcriptome at that point in time what genes are you expressing and how much and if on multiple visits you start acquiring these huge chunks of data based on your transcript I'm at that point in time that could just far outweigh your one-time here's my genome your 1 times you don't well today I didn't mention it but the mesas article talks about how you may need your genome for specific cell lines increasingly to be especially patients with cancer because the whole point is that their cancer cells have a different genetic makeup the or Africa at least as expressed than the normal cells do so it you know the simplistic notion is well they're all just another lab tests it's just that they're huge data sets now some people would respond I know computer scientists respond and say look story just getting so cheap and it's true I mean when you start getting you know USB drives that are terabyte you begin to believe that this really is getting to be an incredibly cheap commodity but it's clearly the data analytics of all those data so that you can do realistic things when you're actually seeing a patient that becomes the challenge how do you extract and it was in those those seven points that were made but in the basis article you know how to extract from the transcript own on a given visit the information that you need in order to make reasonable decisions and we and who knows I mean I don't think I said it here today but I've often compared where we are today to roughly commercial aviation in nineteen thirty now there was commercial aviation in nineteen thirty if you were getting on airplanes and flying around the country and they were even transatlantic flights but an end it was a lot better than the Wright brothers right but we all know in retrospect that that was pretty palsied compared to modern air travel the technology and its penetration and its use and all that and I think that analogy with medical records today is quite apt I mean we know or nowhere near where we're going to be with medical records in the future but it doesn't mean we shouldn't be using them there are too many good reasons for being part of the process that in fact allows them to evolve properly but questions like that we are so far away from doing per visit transcriptome analysis and incorporation and decision support but I wouldn't be surprised if it happens in the day we don't know how to do it yet race because of it in one of the things I remember well from again from macys article and others is that we are currently having problem even extracting clinical data vocation use which are not stored in a way that's easily retrievable you're not in relational format any form now you've got the biologic this new new data thrown in that's going to give additional challenge for storing that kind of data in a form that not only could be extended but combined in any way or even use itself so the whole area of storing data and a form that is little trooper unusable will be a huge challenge of ways represented would be my challenge I'll be dealing with free to move forward you know well a lot of vendors would argue that they're storing data in a way that's nicely retrievable as long as you practice in the institution that has that system there okay haven't helped you if you practice seen across town unless you have to be in Indianapolis where they actually have hie working such a you know so we're still evolving I admit but within a closed environment we're pretty good now at allowing access to patient data you really can't ask a patient of patient record something you could never do with a paper record which was as this patient ever had a new of X vaccine and those three volume charts you did not want to have to answer that question with the paper chart you do yeah there was no way you're going to spend the time necessary to read all three volumes trying to find out if they were had a new of X so you'd order another one so waste and now it's easy to ask a question like that in most EHRs right so that's why I say we're a lot better off but it's still 1930s vintage ok and and certainly all the hie the health information exchange issues are far from from being dealt with and we don't know how to do the genetic parts and in a way that it can be retrieved in a way that's really useful alright any additional questions all right thank you again the doctor

Be First to Comment

Leave a Reply

Your email address will not be published. Required fields are marked *