 Live from Cambridge, Massachusetts, it's theCUBE at the MIT Chief Data Officer and Information Quality Symposium with hosts Dave Vellante and Paul Gillin. Hi everybody, welcome back to Cambridge, Massachusetts. We're at the MIT Chief Data Officer Forum, MIT IQ Information Quality. Behind us is all the logos of the participants in this, a lot of MIT collaborators and of course theCUBE is here for the second year in a row. I'm Dave Vellante with Paul Gillin. John Holomka is here, he's the CIO of Beth Israel Deaconess. John, thanks for taking time out and coming to theCUBE. You're very happy to be here. So, you know, I've, I've from a distance followed your career for a while. You've always been a pretty, viewed as a pretty innovative CIO technology leader. We were talking off camera about what you call SMAC. Right. What is SMAC? So if you look at healthcare in 2014, the data is no longer just what doctors are entering into in the health record. It's patient data, it's telemetry from your Fitbit. It's really kind of a social way of gathering data. So, S of SMAC, of course, is social media. You know, your mood might be more reflected on your Facebook page than in your medical record. We need to mine structured and unstructured data, bringing it together from a multitude of sources in kind of a social media way. Mobile, 80% of all of the Beth Israel Deaconess medical information is accessed by patients via mobile devices. The laptop is pretty much dying. The desktop is just dead. And so, our doctors, our nurses, our patients, they've all gravitated to mobile platforms, whether that's the iPhone, the iPad, the Android device, Google Glass, this is the way people are consuming information. So, mobile, very important. What percent of the docs are wearing Google Glass? Well, our whole emergency department is Google Glass-based. So, since December, we have been live on Google Glass. A doc walks into the room wearing Glass and sees the data about you and spends time looking at you instead of looking at their keyboard. So, that's been actually a great improvement in patient satisfaction. Big data and analytics, the A, of course, we have three petabytes of data. I'm not sure if that's big, but it is challenging to turn data into wisdom. And so, analytic platforms that enable us to query, what were the last 10,000 patients like you? How were they treated? What were their outcomes? That's increasingly important. And the cloud, well, I believe that the role of the CIO, the role of the IT department has fundamentally changed, that I am now the enabler of services. I don't necessarily need to provision the services. It really doesn't matter where they are hosted. I have many of my services offered from the West Coast. The doctors don't know that. I just enable them. So, cloud is actually a big part of what we're doing today. Oh, let me ask you a question on cloud. So, let me say cloud. We're talking public cloud, private cloud, combination. Is public cloud a bad word in your world? And of course, the answer is both private and public. So, I host 300 docs, electronic health records, and my private cloud. And those cloud technologies very similar to what an Amazon or others would use. It's just I host them, and they're in three data centers, geographically dispersed. Did that in 2008, because back then, public cloud, oh, that just sounds evil and unstable. And I said, no, no, no, it's not a cloud. It's just a multi-tenancy off-site hosted model. Oh, that sounds fine then. We'll take one of those. Yeah, whereas the public cloud is actually being used in ways you may not even know. When you have a disease, the Centers for Disease Control gets information about your symptoms. And where is that hosted? Amazon. Yes, the Centers for Disease Control of the United States puts 100% of the syndromic surveillance data for our country in the Amazon cloud. And that's okay. CIA's doing it, why not? Why not? Yes, the answer is we use public cloud technologies for a variety of applications, including public health. It should just pick the right technology which has the right reliability and security, business associate agreements, policies, and that sort of thing. It's fine. So earlier, actually yesterday, one of your colleagues, we talked about this a little bit, Jim Noga from Partners was on. And we were asking about the role of the CIO. So we want to ask you, very innovative. The scene is an innovative CIO, somebody who is a mover and shaker in that world. He basically put forth the premise that the CIO role in 10 years might disappear and essentially morph into a COO, CTO, CDO type of structure. What do you think about that? And so I would agree with Jim. So I have been the CIO at Beth Israel Degnes since 1996. And in 1996, I wrote code. I was doing architecture. I was doing scalability, reliability, security. I have an MIT degree which really helped write in the code. Do you think CIOs in 2014 are writing code? Probably not. Right, so what you're doing is budgets, strategy, getting the business owners engaged, governance. But I look at my role today and I oversee 22,000 customers, a million patient records, 83 different locations. It's actually getting to the point where one person can't actually do everything. So you actually break the role onto, well who's the guy doing strategy? Who's the person doing data curation? Who's the person who's making sure the services are reliable enough? Who's doing security? So you could imagine this office of the, what do you call it, the CIO, or technology services or whatever that's broken into these multiple component parts. Very reasonable assertion. So what do you think about that? Well what has happened to your responsibilities, the way you allocate your time since then? Presumably in 96 you hosted everything on site. You provisioned all the apps yourself. Now so much of that has moved to the cloud is not your problem anymore. How has that changed the way you spend your time? So oddly enough, most of my time is spent on governance. And that is no matter what your budget is in IT, it's too low. And so that implies you cannot possibly do all the things that are requested of you. And so what you have to do is gather all your stakeholders together and say, guys, here's the fixed amount of resource. Here is the fixed amount of time. Our only lever is scope. And I can't be the guy that decides on scope. It has to be the business owners of the institution. And so I have governance committees over various domains, lab, radiology, ICU, and an overall governance committee that helps weigh all these various requests. So in effect, my job is to keep customer satisfaction high by balancing priorities and constantly doing mid-course correction. Now if bad stuff happens in the cloud, I am held accountable. If a privacy breach occurs, I am held accountable. So I am the one throat to choke. But much of my time is spent on customer service, priority setting, and strategy, not technology. Now these days in many ways, in many areas, customers can provision their own services if they want. I'm sure not in some of your more specialized disciplines. But what governance principles do you have in mind? Do you have in place for that, for whether someone could just go out and provision their service in the cloud without even telling you? Right, so we do have very strict policy, which says it's not that it's a bad thing to provision your own services in the cloud, but if you were to do it at some, oh, I'll make this up, xyz.com. It's hosted in Beijing, it's great. Maybe that's not a place you're gonna put patient-identified data. So what our policy states is you will not utilize one of these external services unless a business associate agreement has been signed and approval has been granted by the institution for that service, not your instance of the service, but for the general use of that service. Give you an example, Dropbox. Dropbox is a fine company. It just turns out that Dropbox will not sign a business associate agreement to guarantee the privacy and integrity of patient data placed on Dropbox. Box, however, will. So what I've said to our community is, you know, you need to use Box? Totally fine, do what you want. We've already put the protections in place. Dropbox, well, it's fine for some things, but don't ever put person-identified data there because it doesn't meet our federal and state requirements. And are you answerable to the compliance authorities for that? I mean, is that something that you have to approve with a regulator? We do. So in fact, it could be argued that the placement of patient-identified data in Dropbox is actually a violation of federal law because at the moment they're not signing these business associate agreements and therefore HIPAA says that if we're using these third parties to host whatever services, storage, data exchange, et cetera, there really needs to be those legal protections, roles and responsibilities in place. So yes, I am responsible to the Office of Civil Rights and the Attorney General for keeping data integrity and safety, respecting patient privacy preference. So you have to have the policy in place, but you also have to enforce that policy. So how do you do that? Does technology help you do that? And so we have a variety of data loss prevention technologies. So suppose I want to email my medical record to you on your Gmail account. Well, you know, if this was actually a patient-consented action, that would actually, from a hip-hop perspective, be okay. But we don't know that. So we actually put these DLP or data loss prevention technologies in place that say, oh, patient-identified data, leaving secure network on its way to Gmail. Stop. Morning, Will Robinson. Exactly. So yes, I have had to employ a multi-layered defense to contain the data to the institution and ensure it flows to trusted parties. I want to talk about data standards. It's a question that's come up a lot, a conversation we've had a lot this week. What's going on in healthcare? What are you doing within your organization to promote data standards? Talk about that challenge and what the future looks like. Sure. So I don't know if you guys have ever worked in standards development organizations, but there are people who see the forest. There are people who see the trees. Standards, folks, in general, I love them, are bark people, right? We could spend hours debating on the relative value of ampersand versus a comma. And so the challenge is, it's hard to get consensus on any standard for any purpose. So the country over the last several years at a national level through what's called the HIT Standards Committee has brought together all the stakeholders in a public forum to set priorities and to work through what are the standards most applicable for a given purpose. If I want to send content, if I want to use a vocabulary or ontology, if I want to use a transport mechanism, what is the best standard for that purpose? And then those are written into federal regulation. So the Meaningful Use Regulation, the Obama Stimulus Program, actually mandates the use of very specific standards for very specific purposes. So it made a lot of traction over the last couple years. Yeah, so traction in terms of within your organization, within your industry. Oh, it's a country. So we talk about Meaningful Use, talk about, I think, of adoption of these technologies. So where are we with Meaningful Use? Again, generally in the industry and specifically within your organization. Sure, so across the country, we've moved from 20% adoption of electronic health records to 80% adoption of electronic health records. And by definition, those are certified electronic health records, which means they include certain types of standards baked right in. So, wow, over the course of this whole stimulus period, we now have such things as labs are recorded such that when you say hematocrit, it actually means the same thing and it's comparable. So we talk about big data analytics, we're actually now using control vocabularies on problems, meds, labs, allergies, and other terms. So across different vendor systems and different hospitals, they're finally comparable. So that's actually good. Are we done? Of course not. But have we laid a foundation that now enables the kind of analytics across vendors and hospital systems? Yes, 2014's up. Looking pretty good. And how has that affected the quality of patient care and how do you believe it will affect the quality of patient care as organizations, as he's moving to the mainstream and organizations will learn how to use EHR? Well, let me give you some case examples. So two Mondays ago, my father-in-law, who's 83, had a stroke. Now he has records like many of us, scattered all over multiple hospitals and multiple doctor's offices. He came into the emergency department at Needham Hospital with the blood pressure of 180,000. And the doctor said, God, I wonder if this is normal for him. So because of the standards-based electronic health records we have in the Boston community, he was able to reach across to the primary care provider system and actually see the last six blood pressure measurements for the last six months of visits. Say, oh, well, he's normally 120 over 80s. Today's 180 over 90. This is totally strange for him. That level of care coordination was enabled by the meaningful use standards. Would not have been possible a year or two ago. I want to ask you about a couple of promising technologies that big data-related that could radically change your profession. And one is genome sequencing, which has fallen from $2 million per sequence to about $500, I believe, in the last eight years. And so pretty soon everyone will have their genome sequence and you'll be storing that information presumably or that data will be stored in such a way that you can use it. How is that going to change what you do? Sure, well, let me give you the case example. Since I'm the second human sequenced in the human genome project, I know my three billion base pairs and the challenge is, well, I was in the higher end of the cost of sequencing. Today with the $500 genome, it's not hard to sequence. It's just devilishly difficult to interpret. It's still a million dollar interpretation. What do you do with those base pairs once you've got them? So I have researchers throughout the world on a daily basis looking at my genome since it's in the public domain and looking at what disease states I am likely to develop. Now, how have I used this personally? So when I will go to my primary care doc, I actually am aware of various biomarkers and risks and probabilities of disease. So we can actually make decisions on diagnostic tests to be done on me based on my likelihood of developing a disease. My doc would have no use for the three billion base pairs. It's just a bunch of ATGs and Cs. But if I say, you know, I'm twice as likely as the average male to develop prostate cancer. Well, while we were a PSA today, that's probably a good thing to do. So what will happen is we'll store biomarkers, probably not the raw data. The raw data is 750 megs per patient. That is just the interpretation that's gonna give you decision support and effect care. But doesn't the real leverage come when you combine a very large number of genomes and you can begin to look for patterns that are not clearly evident? Certain characteristics of people of red hair are more likely to develop something in it. I mean, that's a big data issue. Well, it is. We're not quite there yet. I would tell you, I mean, SNP analysis of saying you have red hair and you have diabetes. Therefore, all people with red hair have diabetes. Probably not a valid thing to do. If you look at 100 million genomes. It's a little iffy. Causality is hard to infer. Whereas if you say, ah, I have a mutation. This T was replaced by a G. Ah, the protein folds funny. Ah, I know now why you have that disease today. You got causality. So you are right. What could happen in the future is that we could use the experience of many patients' genomes to customize the care, sort of ultimately personalize the chemotherapy regimen. Or how are you likely to be cured over the course of a disease because of this genome and evidence we have as to how people respond to treatment with that genome. So that is true. That's coming. But again, it's still probably gonna be at the level of biomarkers because three billion base pairs doesn't give you a whole lot of wisdom. Right. So if you get to that point where you can infer it's still very, very expensive. It is. We're just still in the point of building those tools that are going to allow doctors to the point of care to say, oh, your genome was sequenced. Your tumor was sequenced. And here's the therapy we're gonna order because on the last 10,000 patients, it worked well. Not quite there. Another technology I wanna ask you is IBM Watson, which is being, IBM is heavily pushing it toward medical applications, this idea of combing through very large numbers of records and looking for patterns that humans can't spot. Do you see a lot of potential in that? So I'll use Watson as a class of technologies. So natural language processing with subject predicated, object analysis is a reasonable idea. So what if I can look at the unstructured text in an electronic health record and say, you know, I have just discovered that in my two million patients that actually there are a hundred thousand people who have depression that's untreated. Because you know the word sad, moody, down actually appears in these unstructured texts. So I think we all believe that the future is gonna actually find more rich data in unstructured text analysis than the highly structured text we enter in problems, meds, allergies, and other places. And this is something that, historically, that's how data has been captured, right? So about 50% of our record is unstructured text. But of course there's another way to look at it, which is Facebook is unstructured text and your email is unstructured text and your tweets are unstructured text. And maybe I can actually learn more about your activities of daily living from your Twitter stream than from your medical record. Are you working on that problem or are you waiting for others to solve that problem and you apply it? Oh, I am working on that problem. I have already looked to a third party firm in Palo Alto to do the natural language processing on our unstructured text as it's written. And then come back with knowledge, metadata around that unstructured text. What are you doing? One phenomenon I hear from docs all the time is the patients come in now and they're very well informed. They're researching everything on the internet. They may not know the right information but they think they know the right information. So you're dealing with patients now who come in with much more specific questions, often a higher level of knowledge about what afflicts them. Has this been a good thing or a problem? Well, we're level. Having patient engagement is great. And so one of the things we've done is shared every note written about you with you. So our medical record is totally transparent to the patient, electronically available, online as it's being generated. So that works well. We try to curate knowledge resources because if you search on the internet, there is no editor. There is no curation. You're as likely to get the, say, Bulgarian Journal of Irreproducible Results as the New England Journal of Medicine. So we use a variety. If you subscribe to the Bulgarian Journal. Oh, it's a fabulous journal. B-J-I-O. Yeah. So we tend to say, here is your information and here is curated knowledge from experts about it. That's a lot more effective than a blank internet search. So curation is a service that you guys are providing? We is. It's a human service? Well, we have licensed content from third party providers who specialize in this area, as well as my medical librarians. We rename the whole department the Department of Knowledge Services. So they're not actually just retrieving papers and books and that sort of thing. They're actually organizing knowledge and ontologies to deliver it to patients in a form they can understand. And then inserting those into radical records so the patient can see them and access them. That's right. Do you have an initiative to become a publisher, to become an information provider, such as what Mayo Clinic has done, with a vast library of podcasts and documents? We do not have a specific initiative of that nature. But the idea of using a constellation of services we build and services we buy to help that cloud of knowledge of decision support that informs decision making at the point of care is absolutely something we've been building. And is that a marketing initiative? To safety initiatives, because doctors like me are trained as apprentices. I learned 25 years ago what antibiotic you need. Now that may be a totally wrong assertion, because remember, 50% of what I learned in medical school is wrong. I just don't know which 50%. So that's patient safety and quality should be informed by evidence. And cloud-hosted decision support systems provide that evidence. So John, we promised we'd have you out of here by the end of the hour. Pass that a little bit. You're still OK for your next one. My last question. You're excited about a lot of things. So could have summarized that excitement. What's exciting you these days? You travel around a lot. You probably spent a lot of time in Silicon Valley. Right. And so what are you excited about? So the next project, which I blogged about yesterday, is that as more and more individuals are wearing telemetry, devices in the home. We find out more about your activities of daily living, your diet, your sleep patterns. We could even discover glaucoma, blood pressure, or other body parameters. That needs to inform the medical record between doctor visits. Health is continuous. Not, oh, I see a doctor once a year and I have data points every 365 days. I'm now being paid, as we are in Boston, based on wellness rather than sickness. And therefore, we need these novel sources of data, especially patient-generated data, in order to keep you healthy. That's the next body work. I'm excited about that, too. That's a great vision. All right, John, thanks very much for coming on the queue. Really a pleasure. Happy to be here. Thanks so much. Thank you. Keep it right there, everybody. Absolutely. Back to wrap up right after this.