Our guest this week is Shahid Shah. Shahid is CEO at Netspective, and writes three blogs: Health Care Guy, Shahid Shah, and HitSphere - the Healthcare IT Supersite. During the program, Kyle recommended a talk from the 2014 MIT Sloan CIO Symposium entitled Transforming "Digital Silos" to "Digital Care Enterprise" which was hosted by our guest Shahid Shah. In addition to his work in Healthcare IT, he also the chairperson for Open Source Electronic Health Record Alliance, an non-profit organization that, amongst other activities, is hosting an upcoming conference. The 3rd annual OSEHRA Open Source Summit: Global Collaboration in Healthcare IT , which will be taking place September 3-5, 2014 in Washington DC. For our benevolent recommendation, Shahid suggested listeners may benefit from taking the time to read books on leadership for the insights they provide. For our self-serving recommendation, Shahid recommended listeners check out his company Netspective , if you are working with a company looking for help getting started building software utilizing next generation technologies.
Data Skeptic
Data in Healthcare IT with Shahid Shah
[MUSIC] >> The Data Skeptic Podcast features conversations with researchers and professionals working on problems or projects related to data science. >> Welcome back to the Data Skeptic Podcast. I'm here today with my guest Shahid Shah. How are you doing, Shahid? >> Doing great. Thanks for the invite. Looking forward to talking about big data in healthcare. >> Yeah, me too. I'll give you a quick intro. You are currently the CEO of Netspective Communications. Do I have that right? >> That is correct. >> In addition to a long career of many sea level and executive type positions at a lot of places, you're also currently writing three blogs, chahid.shah.org, healthcareguy.com, and hitsfear.com as well. So you're like a pretty busy guy. >> In fact, I am, and I do enjoy writing a lot these days, sharing that knowledge, so much going on in general IT, where I've spent, of course, a lot of my time. But now in healthcare IT, pharma IT, bio IT, this area of biological, pharmaceutical, and healthcare technology is just starting to explode. They're going to see the kind of growth that we saw in financial and in retail, internet commerce, etc. We've been in this world from general IT perspective for a while, but we ain't seen nothing yet as far as big data is concerned when you look at healthcare. >> That's excellent. Yeah, so one of the ways I first found out about you is I enjoyed watching a panel that you moderated at the 2014 MIT Sloan CIO Symposium, and I'll put a link in the show notes for anyone who wants to go check that out, I'd highly recommend it. The title of that was transforming digital silos into digital care enterprise. So for me, the idea of liberating those silos and enabling the aggregation and potential data mining of a lot of healthcare related databases and information is really exciting, and I would like to see where my medical practitioners could see my complete medical history, and it seems like there's some potential for exciting innovations there. So I was wondering what benefits to both patients and providers do you see coming out of some of the growth that we're anticipating in healthcare IT? >> I think, as I mentioned, it's just starting now. So we're like in the first innings of a game where everybody's starting to fill each other out, they're getting the plans ready and things like that. But where we've noticed a lot of effort in the past 30 to 40 years of healthcare, and it's not commonly known, but right after the military, healthcare was one of the first users of computers back in the '50s to crunch numbers like figuring out what insurance costs are and things like that. So when you look at the payer community, such as insurance and other areas, they've been using IT for a while to try to integrate and figure out how to connect within their insurance clients and us as patients, of course. So administratively things have been around for a while. Then we've seen administratively people managing financial data, things like our basic medical records like demographics, claims against insurance, again, around for a long time, maybe 30 to 40 years. Last 10 or 15 years, though, is where it's starting to get exciting, is we're starting to see clinical data coming in. And the big news for patients is going to be what's called patient-generated data comes in. And that is things like we're connected to a Fitbit. We get on to a Wi-Things Wi-Fi scale, et cetera. This is all generating a bunch of stuff today that is more fitness-related. We might start to be putting in things like what are our foods that we like, et cetera. However, the next piece that we're going to see in the next few years are things that are attached to our body and are doing things like pulse oximetry, figuring out what our temperature is at the beginning of the day versus the end of the day, like body temperature, et cetera. So getting a little bit more internal data that will then be used, can be used by technologies today or in the future to help do better diagnosis. So where we're seeing the most use in the present and going forward is going to be better and better diagnosis because we have more and more information. And so moving away from financial and administrative, et cetera, to clinical, and then basic patient-generated data for remote access, being able to see, for example, your physician on Skype or on a Google Hangout versus having to go into the office for routine kinds of things, obviously not complex things, that's all starting to happen now. The big thing, and this is where data comes into play, is that bioinformatics, pharmaceutical IT. These are roughly new concepts within the last five to seven years, and we're not quite sure how to handle these. Do we put them into our electronic health records and use them for diagnosis? Do we use them for drug discovery or use it from point of care? So there's a lot of discovery going on if big data scientists and data scientists in general should see this as a major area for them to be focusing on. While we're on the topic of patient-generated health data, I'll confess I wear a Basie swatch, which is one of the mini wearables. I try and be a vendor-neutral podcast that just happens to be the one I've worn, but there's Fitbit and plenty others. So it's been fun for me to track all that data and collect it and kind of look at it, but I don't think my doctor would appreciate me emailing him a spreadsheet of all my data. Yet there's probably something useful there. And I don't know if it's an engineer's choice to say what is or isn't important to save, and the idea of saving everything is probably a day-to-day luge. So do you have any thoughts on where we meet in the middle where researchers can come and say what's important to save and how it should be saved versus what patients are looking to provide and what the infrastructure we need to build to support that is? Yeah, that's a great question because when we think about data, depending on who's asking the question and why we're talking about it, we always have a lens that we're looking at it from. So you and I today are looking at it from a patient lens. But if we were to move to the physician lens, physicians don't really care about that data as much right now because they're not paid to manage that data for us. But you can imagine a world in which if we're paying insurance companies some money to pay for bills as they pile up, it's possible that some of that insurance money through reimbursements could go to the doctor to say, hey, this is a diabetic, and therefore you should, for this much money per month, spend time looking at the A1C levels for this diabetic patient, or this is a congestive heart failure patient, you need to keep track of blood pressure and other things. So in general, doctors don't care and it is a data delusion, it's not that important. But when you look at disease specific conditions, the world starts to change. And when you look at over the last few years, we've gotten enough data now, the insurers have this data, providers have the data, where primary monitoring of at-risk patients has enough value that you can monetize it. And once you can monetize it and you can pay physicians, then you don't have to worry because if your physician is getting paid to help either your chronic condition or something from becoming a chronic condition, that's perfectly fine. For example, today you have a next generation business model such as what's called an accountable care organization, where an insurance company or Medicare don't pay for every time you go to the doctor. But instead, they pay a fixed fee to a doctor to say, hey, for Shahid and for Kyle in this community, you are being offered $10,000 a year, your job to do whatever monitoring you need to do, keep them out of the hospital, get them to do diagnostic and imaging when you need to. So when new business models catch up with new data, we're going to see a grand new fostering of innovation. Until then, you're right. Nobody cares about the data that we capture, and you definitely don't want to be sending that spreadsheet to your doctor. So that's one lens, right? The other lens is, if you're a pharmaceutical company looking to do drug discovery, you may say, hey, whatever data you have, send it to me. I'll pay you for that data so that I can use it to find new patients that I want to recruit or other things. Like, there are many times I live close to NIH. So all the time, we see advertising on buses or other walls would say, hey, do you suffer from disease so and so, please sign up for this kind of clinical trial. Well, if you've got the data, you don't have to ask anybody. You know who's suffering from what because the data is coming in, and some patients may want to give that freely because they care about health in general. Some patients say, don't contact me at all, but I think we're going to start to see new business models coming out. And that's really what the cool thing is. We have a lot of data. The technology is there. We're starting to figure out how to aggregate it. We're not quite sure how to pay for it yet, and that's what I'm looking at happening over the next few years. It's even more interesting. Yeah, it seems like there's a great opportunity there where, as you point out, something I had never thought of. The doctor's being paid for monitoring, and the insurance company sees the potential to reduce their long-term costs by maybe a short-term expense for some relatively inexpensive piece of hardware that a user would wear and submit data with. There's a win-win on all three fronts. Absolutely. And that's really the thing to look at, is that don't look at data from one lens. Look at each view of that data and say, when the patient generates it, there's only one place you can actually get the data. It is from the patient. And if you think about it, for thousands of years, patient data has always been available. I mean, even the earliest folks, what did they do? They ask you a question. They observe, and then they write things down. That's patient-generated data. Now, observations, which are the time-honored, tried-and-too tradition of how you get data from patients, is fraught with danger, right? It can only give you so much. This other non-observational, but quantified kind of data, structured data that's coming out of these devices, they are, in general, high volume, but pretty trustable. Because there's no way to get them wrong per se. Now, because they're high volume, this is where we need extra technology to figure out, where's the needle in the haystack? What do I need to care about? For example, we worked on a project a couple of years ago on an epilepsy-specific electronic health record solution, which had an embedded device that was surgically implanted into epilepsy patients. And then there was a wireless monitor that took the implanted device, put it onto the wireless, and then this was a couple of years ago, so the USB stick was used to pull data off of that wireless device, put it into your computer, and then send it to your physician. Now, the physicians, you know, the neurologist who would be looking at this, has no clue what to do with continuous data coming from days and days worth of your brain waves. What's he supposed to do? Now, so the project we worked on was, you had to get new visualization models out, they had to put new data analysis techniques in to figure out, and this was measuring seizures, and sort of trying to do seizure prediction models to say, if certain things were happening during the day, what is likelihood of you having a seizure because epilepsy patients have seizures that are not predictable at all in most cases? So if you think about that problem, that's a real problem for millions of patients going through all kinds of these kind of conditions. That's a big data problem. We're dealing with gigabytes per hour. It doesn't get any bigger than that, right? Now, and that's from one patient. Imagine tens of thousands of patients, a gigabytes per hour, the numbers add up. So think about that scenario is, if you're a data scientist, what's unique here? Enormous volumes of data, it's coming in pretty rapidly, but more importantly, we don't know what to do with it. So we've got all this data sitting there. I mean, for months we sat around saying, what do we do? How do we visualize this in a way to give this to a physician? And that's really what's most interesting here. You know, when you're working with retail data or financial data, the quantities just don't add up as fast. They are big data when you do it in the aggregate, but for any single individual, their consumer data, what their buying patterns are, et cetera, are tiny. But for any single patient in healthcare, their data is monstrous if we were to go to bioinformatics or pharmaceutical kind of data. And we need as many smart brains as we can get to come into this field. >> Yeah, it's one of the biggest of the big data problems, I think. >> It is, absolutely. So there's a part of me that would like to, you know, as you were mentioning, you see promotions asking people to come contribute data and participate in studies. There's a part of me that would love to volunteer whatever data I have available if it could advance medical science or, you know, maybe help prevent a disease that one of my ancestors died from. But the flip side of that is privacy. You know, privacy is a concern in every industry. Healthcare is certainly no exception there. And it might be fair to claim that some of the privacy we have in healthcare could be due to security through obscurity in that systems today are a lot of our paper base, not everything's federated or talking to each other. And it seems like there's a trend moving away from that where we're going to have more accessible data and more referential integrity between different systems and different offices. But it seems like at the same time, IT organizations who are helping to provide that infrastructure are gonna face new challenges regarding privacy. Do you think we're poised to face these? Or are there ways that people need to take a careful look at that? - Yeah, I think we all do definitely need to be careful in this era where ultimately we, the moment something is electronic, you have no control over it. And data spills are fairly routine and fairly common in general and not even in healthcare they are. So privacy, again, if you think about it, privacy writ large, we are all scared. But if you think about it in a little bit narrow sense and say, well, what is it about the privacy that I'm most worried about? In the case of mental health, for example, there are things that we still as a, even as an enlightened community in America and around the world, certainly the first-class world, you see that we still care if someone went to the doctor for mental health. Like if I broke my leg and my leg got fixed, I'm done. What do I care who knew I broke my leg? But if somebody finds out that I went to go see a mental health professional, just that word, just that statement alone, not knowing why you went, et cetera, now all of a sudden a potential employer, somebody looking to give you a loan, et cetera, starts to think twice. So you have to say to yourself, what kind of physician am I visiting? What kind of data are they capturing? And before you choose a physician, and this is how you tell whether we actually care about privacy or are we just using it as talking points is, if you do not choose a physician, because of the way that they manage and safeguard your records, then shame on us, right? If we, so like today, if I go and say that I'm gonna choose Dr. Smith, am I talking to Dr. Smith? You know, how are you storing your data? Do you keep it in a file cabinet that's unlocked? We never ask these questions. So in essence, in the written record world, we didn't care. We never asked these kind of questions. We just assumed that there was privacy because we never thought there wouldn't be, but never in the digital world. And this is not about big data or anything. In the digital world, you never assume privacy 'cause it's so dangerous. What you want to do though is you want to hold people responsible for the kinds of data you're giving them that you think should not get out. Try not to just say, well, I wanna protect all my data. If I don't get all my data protected, then I'm not gonna share anything. And that actually puts you in greater harm. It puts us all in greater harm because there's a lot of harmless data that we should not mind sharing, but there's a lot of harm full data. It used to be, for example, 10 years ago that AIDS records were not as confidential as they are today, but today they are. Same thing with mental health records. And this is why you see, for example, especially the veterans who are coming back with post-traumatic stress. And you see PTSD being changed from PTSD to PTSD. Why? Because there's a idea that a disorder never gets fixed. But a syndrome, post-traumatic stress itself, you came in, you got fixed, just like a broken leg. There's no broken leg syndrome forever. It's just broken leg, you get a fix and you're done. So there are things that we have to fix in medicine, in healthcare, that have nothing to do with privacy of data, et cetera, that will ease the privacy concerns. If mental health has continued to be a stigma in society and the release of medical data around mental health causes privacy concerns, then all we're gonna do is never share anything with our mental health professionals, right? Electronically, which is bad, which is absolutely bad. So that's the kind of danger. Let's not treat privacy as a black and white. There are tons of shades of gray. And let's figure out which areas are the most damaging and the most concerning and treat those with an enormous amount of concern, but other areas with kid gloves so that medicine can move forward. So there are multiple camps in here. There are schools of thought that say, all electronic data is bad and be very, very careful, or all electronic data is good and keep it going. But as we know as engineers, you make a bomb, sometimes it's good, sometimes it's bad. You make electronics, sometimes it's good, sometimes it's bad. This is one of those areas. Tons and tons of data is coming in. Some good, some can be used for harm, but the only way to move forward is to not treat it all the same. - Yeah, yeah. I was considering for a while, wouldn't it be great if there was some sort of universal system where patients could control their data? So perhaps the fact that I went in for mental health or an AIDS test, I want my GP to be able to see that, but not necessarily my optometrist. And wouldn't it be great if I could control my own privacy on sort of a deeply accessibility level? But at the same time, I'm not sure that I'm knowledgeable or the general public has the time and the knowledge to manage that. So it's almost like there's no owner of who should arbitrate what's private and what's not, who has access to what. And I don't know where the responsibility lies. How do you see it? Is it a doctor's responsibility to set at the time that data's created some sort of protection policy around it? Or is it IT's responsibility to be putting up the right firewalls? - I'd say it's a patient's responsibility to know what your records are, who did you give it to, where did you leave it, et cetera. So it's always, ultimately, our responsibility for our records. Now, I'll give you something that you won't commonly think about when you walk into your physician's office and tell him your first name, last name, insurance information, et cetera, that's not just our medical record, right? That's his business transactional record that he has to charge insurance for so he can get billed for the service that he performed to us. So what we have to think about and say, the medical record that belongs to me is mine, it's my responsibility. And there are things, for example, you can use Microsoft Health Vault or related tools to manage your online record. But when you go see a doctor, part of that is your medical record, but the other part is his business record. Do we as patients have a right to be able to tell Amazon how it maintains our purchase history? In some ways, yes, in some ways, no. Same thing with our doctor, how can we tell him, this is how you need to store my data that you tracked on my behalf for services. It seems easy, right? To say, hey, it's my record, you just do what I say. But that's not true, 'cause it's his record too. In a digital world, in a paper world, I could say, give me my paper, you got nothing, you can't make copies of it, et cetera. What do we do in a digital world there when there is no such thing as one record? And this is where the debate really falls down flat is, we haven't thought through what this means going forward. What is a business record of the medical side? What is the clinical record? What is the bioinformatics part of this? What is the drug discovery part of this? What is the administrative part? When we start adding sophistication, and this is where data scientists actually do a good job, they can add a level of sophistication to our debate so that it's more meaningful. Just that one introduction of a business record versus a medical record now can say, okay, two smart people could sit down and say, all right, my name is part of the business record, but my glucose level is not, okay, and that's okay. Unless the glucose record was submitted to the insurance company as a diabetic patient in order to get a different reimbursement. So it gets very tricky, very fast, and you're right, it is our responsibility, but we have to recognize that when we're doing business with someone, part of that data is now their business record, not just our medical record. - Yeah, absolutely. From what I understand, there's a wide spectrum of size and capabilities in different healthcare providers. On one extreme end, you have the most state-of-the-art hospitals with the best staff and the largest staff and the most specialists, and presumably the latest and greatest technology. And on maybe the other end, the long tail of the distribution, you've got a single doctor in a rural office that serves a very small remote town and probably doesn't have all the right access and technology there. So what strategies for technology acquisition and rollout, would you recommend for someone making decisions in the smaller provider space, who maybe doesn't have the same budgets as the big world-class hospital, but is concerned about doing the best service for their patients? - Yeah, what we're seeing is that there's a lot of attention being paid to that primary care provider. In fact, the vast majority of care happens in those small-class clinics and small practices out of the hospital is where the most of the care happens. So cloud technologies, mobile, and the availability of these cheaper devices that doctors can give out on a reimbursed model are all good ways for them to start. So when you start and say, well, what are the cheapest ways for me to do it? Don't build something, for example, that you can only buy that has to be installed on a so-called data center inside your office. So the cloud has done for physicians exactly what it's done for everybody else. They just haven't moved as fast because in the cloud, there are rules around HIPAA and other areas which are still catching up from a privacy perspective. So that's one area. The second area is look at common tools before you look at specialized tools. So often in the medical community, people go after specialized tools before they look at common tools. But what we notice is that if you look at the vast availability of software as a service capabilities, whether it's a project management tool or a customer relationship management tool and accounting tool, et cetera, many physicians believe that the special purpose tools in their industries will do a better job. And when you're looking at the top cream of the crop of software, which is very expensive, that's true. But the guys on the lower end, you can actually say, I can use Microsoft Word for my medical records if I'm going to store them with box.com, which is a HIPAA compliant file sharing with some DICOM cloud-based DICOM imaging environments for my lab studies and things like that. Would you, if you step out from a specialty driven medical community specific software world and say, how can I use the common tool sets that are available, which are going to have better user interfaces, larger amounts of users, potentially better security than the specialty providers, 'cause if you only have a thousand customers versus a million customers, you'll have a different level of security. So if you're small, my general recommendation is absolutely look for those very well-known cloud providers or if you have a local provider that is exceptional service, like they come out, they pick up phone calls, that's great, go after them. But for everything else, see what common tools across the general internet work for you before putting into place those specialty tools for the same reasons we mentioned. - As we talk about this spectrum of different tools available and also the exciting new opportunities of what might be coming in the future, there's a question of are we going to consolidate or are we going to fragment in terms of what's out there? And on one hand, consolidation sounds really good, right? Because it means there's more interoperability of systems, there's probably better access to data and more referential integrity. And I certainly like the idea that in 30 years, when I go in for a major surgery, that that surgeon has full access to anything that's ever happened to me, that they feel might be relevant and they want to look at. But on the other hand, full consolidation means less competition, probably there's not as much innovation or price control. So I feel like there maybe needs to be a balance between consolidating and having standards and leaving enough room for innovation and new people into the marketplace. How do you, what do you think is the right balance there? And what are the opportunities for someone to make a splash in healthcare IT? - Yeah, so in healthcare IT, one of the most common things that people see are, what are referred to as EHRs or electronic health record solutions, this is a group of software, which primarily acts as a document repository for medical records in its simplest form. If you look at this market, there are about 500 vendors in this space, very, very fragmented, lots and lots of competition today. What happened a few years ago was that the federal government put down a minimal set of standards, those are called meaningful use criteria, to establish and say, you know what? Everybody doing this, if you want to get reimbursed from the government for your medical services, you have to use this kind of software. And that software must do these 15, 18, 20 things at a minimum. Now, this was an industry that had never seen regulatory requirements in this fashion before, so they were now saying, okay, if I sell this, I have to have now a certified piece of software. So the first thing it did was, there's a general excitement that, hey, we might not have some standardization here, that didn't really happen, the standardization around the kinds of things you were hoping for, which is data standardization, data consolidation, better normalization of data, better harmonization of data, none of that really happened. Instead, what happened was the number of providers who could meet certification dwindled a bit, new entrants came into this field because they saw a lot of money being put in by the government for these reimbursements. And now we saw a push and pull. So more vendors came in, so where we thought their be consolidation didn't actually happen, it may happen still in the future. So consolidation hasn't yet happened, but innovation has dropped like a rock in the EHR world because the EHR vendors are so focused on just meeting minimum government regulations, which now, of course, the physicians have to follow, they have to follow a set of procedures to use that software. So innovation has definitely been harmed in the EHR space, but that innovation has now gone in other spaces, like remote care, remote diagnostics, these new devices, what you're gonna see with HealthKit and other things coming from Apple later this year is another interesting aspect. With respect to consolidation, given how local healthcare is, until we see consolidation in the business side, you will see consolidation in the software side. And here's why. So it'd be like saying, well, why is it possible for us to be able to consolidate in other industries before you consolidate vendor offerings, you often see consolidation in the consumers that are using that software. So if we saw a lot of hospitals buying each other out, which of course there are happening, if we see lots and lots of smaller providers being bought out by larger providers, that consolidation would happen in a natural course because larger companies don't wanna see every one of their participants using different software. So consolidation isn't going to happen for the reasons we're thinking, which we were thinking. In fact, they're happening because the government regulations are getting more and more stringent. So some of the smaller vendors, who potentially could have been the most innovative, are going out of the market or they're getting bought out, the larger vendors will not see much consolidation until their customers consolidate. So if you see large customers consolidating, then the larger vendors of those customers would consolidate as well. And that will be dangerous. Consolidation will be dangerous for two reasons. What we just talked about is innovation will dry up in the EHR space, it's gonna have to go elsewhere. And number two, pricing will start to get a little bit difficult because like even today, the largest vendors that sell software are in the tens of millions or hundreds of millions of dollars. I mean, some of these smaller hospitals, I wish donors to hospitals, but there's one lesson, if there's anybody that donates to a hospital, you should ask that hospital to say, why aren't you using an open source, the electronic health record solution, of which there are many offerings, instead of spending tens of millions of dollars? It'd be like, today we use R instead of SaaS when we can't afford it, right? - Right. - However, and there are tools available like that in EHRs, except that world is not well known. So you have EHR vendors, which actually offer open source for free languishing, while they're super expensive counterparts are being bought by hospitals and physician offices, because they don't know that the others are available or they don't trust that the service and support will be there. So this is not something that the industry can change, this is something that buyers have to do. So consolidation, we saw in other industries happening because open source came in and caused a slight upheaval in the enterprise market. Unfortunately, we don't see that, that's why we're not gonna see vendor driven consolidation as much as we see provider driven consolidation driving the vendors. - So yeah, it sounds like maybe there's a good opportunity for an open source EHR evangelist to come in here and help convince some of those hospitals to take up these tools. Do you think that's a fair statement? - Absolutely, I mean, I'm actually the chairperson of an organization called Osira, the open source electronic health record agent. We are a nonprofit who pulls together different open source providers, we're in fact having a conference next month about that is how do you find out about these things? How do you instead of spending $50 million on software, spend 50 million on services that would pull together lots of different software and hardware and devices? So open source evangelists in healthcare IT are sorely needed, anyone who wants to write an article, jump in, talk about it, you can come to our events or send me some requests so I can put your blog posts up on my blog as well. We absolutely would love to have more of that. - Yeah, tell me a little bit more about that conference for anyone who might want to attend. - Yeah, so this conference is taking place right after Labor Day in Washington, DC. It's the Osira third annual summit and Osira again, open source electronic health record agent, O-S-E-H-R-A dot O-R-G. Osira dot O-R-G basically has taken an enormous amount of software from the government which was built for about $15 billion over the last 20 years and this is from the Veterans Affairs Department. Veterans Affairs or VA gave the software out to Osira. Osira now is the agent of the software and can pull in open source from anywhere, certifies it, packages it up, tests it, et cetera and sets up the process by which all this happens so that the Veterans Affairs Administration can now use this software going forward. This way it saves money because open source, of course, can save money. But more than money, it's really Osira wasn't created because of money, it was created because of innovation. When you do something inside the government or inside exceptionally large institutions, a lot of cruft starts to settle and things get older than stale and that's what was happening to some of this brilliant EHR software that was sitting inside the government. So the idea was through folks like Roger Baker and Peter Levin and others at VA from a few years ago, they put this out into the open and it's now available to be used, updated, et cetera, completely free by any hospital who wants to use it and there's a large service community that we're helping to build. So I'm doing three lectures, for example, in that conference about what are the business opportunities, how do you use some of that data, how do you begin to use this in a realistic environment, but we are so, even though all this software is available, we don't have enough help IT evangelists so that can promote it and that's where I think anybody interested should definitely check out that conference. >> Yeah, that's really exciting for me personally. I think, and this is just my own bias, but open source is the way to go, not just for enabling it and creating widespread use and standardization, but it gives me a certain level of comfort that code's available and if I'm really concerned about something, I can go review it and potentially even improve on it. >> Exactly, yep. >> So one of the, and I'm gonna talk a little bit about software for a minute, a big innovation that went on, I guess it's been 20 years ago is when we moved, well, maybe more recent, but we moved from the box product where you put software on a disk and you put it in a box and that went in a store to the more SaaS model where your software is on demand. So one of the great innovations that's come along there is we don't have what's classically been called the waterfall model of design. We're in, we can live in this agile world, as they call it, that I'm a big fan of, that I can make small changes, measure user reaction, maybe do a survey, get some feedback, or even just simple A/B testing to improve whatever my product is along the way. And if I'm doing some sort of consumer or client facing web application that's pretty easy, I would guess that in IT we might still be stuck in the something like the waterfall model where you're developing some software that might have to go on a specialized device or hopefully maybe it's mobile, but that'll go into a doctor's office that someone has to be trained on and there needs to be a user acceptance and you don't always have access to those users. It seems to me that it would be very difficult to really drive innovation and if you're a product developer wanting to make a good product that's useful to your doctors, you might not have access to them. Doctors and other healthcare providers. Do you see that as a challenge? And if so, how are some of the ways we could maybe break down some of those walls? >> Yeah, Kyle, I think you're right. It is a challenge. Unfortunately, or just by happenstance, most software in health IT and if you think about it, any regulated space, and we think about financials, think about medical devices, these are exceptionally regulated spaces. In regulated spaces, waterfall has been the norm primarily because it was a de-risking strategy. So if you de-risk using a waterfall approach and say, well, the more I know, the more I won't go beyond the regulations and the more likely it is for me to release my software, the idea of consumers and what they feel like about it, et cetera, just don't matter. What's happening though is that as the consumer grade, and let's talk about it in the most difficult areas. So think about a patient monitor and the amount of software and testing and hardware that goes into creating a patient monitor that sits inside a hospital. It took years, so any patient monitor would usually be in the lifespan of the hospital for seven to nine years. It took three to four years to build portions of it, et cetera. And this is simple things like your pulse oximetry, just finding out your glucose level, et cetera. Now, if you think about the enterprise world and how the enterprise software has been disrupted by consumer software, like Basecamp disrupting Microsoft project, not because Basecamp is better, but it's good enough, that piece right there, this is the thing that's going to help us most is that these small devices that we see working on the consumer side, like pretty soon. I mean, you have an entire weather station on your Samsung devices today that we didn't have five years ago. Same thing is gonna happen with HealthKit and the future phones are gonna have more and more of these kinds of devices, which means at consumer scale, we'll be providing better chipsets, better integration components that the consumer side can use. So what will happen is that just like enterprise software was disrupted by consumer software, enterprise health IT, enterprise FDA-regulated medical devices, enterprise pharmaceutical IT is going to be disrupted by consumer grade devices that are good enough, and they're going to eat their way in, and that's what's going to create the agility. It's really not going to be that, I mean, it might be one or two companies here and there that are big enough and recognize innovation and agile that will put it in, but they're going to be disrupted out of their approach. Not going to be a cajole data there. I mean, it's not gonna be that it's a, here's this is a good thing and your consumers matter. It's only when they see profits dropping and other areas, they're just gonna buy some of those companies and build them in. So it's going to happen, but not happen in any much different way than in any other enterprise setting. And that's one thing we have to remember is, whether it's a doctor or a hospital, it's enterprise software we're talking about, not consumer grade. And that's why it takes a little bit longer to move from the agile to from waterfall to agile. - In addition to that, what other exciting things do you think are gonna be disruptive in the next five to 10 years? - Yeah, I think the idea of a patient-generated healthcare data, the so-called PGHD that we've seen, is going to be the nexus around what a lot of things are going to happen. So I'll talk to you about, there are cases when pharmaceutical companies today, today any typical clinical trial that you might run for any complex piece of pharmaceutical, a new drug that's being done, you might have 8,000, 4,000, 5,000, 10,000. I mean, numbers are tiny. People test and understand side effects. We think that the FDA, and obviously FDA does a great job in comparison to all other regulatory agencies, to keep our drug safe. But imagine that drug safety for some of the most widely prescribed drugs were done with a 4,000, 5,000, 10,000 person trial. Now imagine that that trial, what the kind of data that we captured for that trial can now easily be pulled in from millions of patient-generated health data records, what would that world look like, right? So that to me is the most exciting, is that there's gonna be an abundance of data. What's gonna be alarming is that, and it is still alarming to me as somebody in the medical field is, most health IT companies, most pharma IT companies, bio IT companies, they think that that data is not trustable because it's coming from patients. Because for thousands of years, we've been taught to not trust patient data, right? You ask a person a question, you can't just trust it, you have to look at and say, what is your temperature? Let me hit you on your knee and make sure that there's a reflex. I can't just trust you for it. What we're seeing is that it may take the next generation of doctors to come in and say, of course that's trustable. I can test sentiment analysis. Like for example, there's a very cool study that's being done to say, what can mental health professionals do just by watching how people interact on Facebook? Well, on one side, people are like alarmed saying, oh my God, that's crazy, but on another side, if patients, if your patients happen to give you access and make you friends and they say, sure, you can watch me, why wouldn't we use that data? So there's a ton of consumer generated, patient generated help data that needs to come in to FDA regulated device verification, FDA regulated, drug verifications, safety and efficacy checks. How do you know that the side effects that were in a 4,000 person trial are not going to be different when you go to 4 million people? We don't know. We need some of that data. So I'm most excited about that. In fact, I think that the fact that we have a ton of this data over the next 10, 15, 20 years is not disrupted by itself, but the innovators that use that to displace some of the, what we call a doctor friend of mine once told me that medicine is eminence driven rather than evidence driven, which is true. We don't have enough evidence for a lot of things, but we say, hey, doctor so-and-so from Harvard said this, so it must be true, I trained under Harvard, you trained under Johns Hopkins, so let's trust this. Well, of course, with enough data coming from the field as we were, just like for example, Amazon doesn't ask its executives or its customers or its vendors, how should it sell stuff? It says, I'm gonna put this out for sale and see what sells. I have so much data that I can tell my book suppliers better about what will sell than they can tell me. That's going to happen in five or 10 or 15 years, but it may take a generational shift for us to trust that data and put them in business models and approaches, and this is where data scientists today, data workers of any source, not just data scientists, but people consuming the data, pushing the data, should really keep an eye on that and say, how could they benefit from it or improve the health IT and pharma IT and bio IT universe because of it? Yeah, I'm as well very excited about wearables and patient generated data. There's a part of me that really wants to have the chance to opt into studies, some passive monitoring or some pharmacological study for which I happen to be qualified that I can just go to some website and say approve, approve, you can grab all the data I have here if it'll benefit the research that's being conducted and which will long term help me and the rest of humankind. But the flip side of that that I worry about is when we get, and it's just gonna have to take once, we get one data disaster that data gets de-anonymized or someone gets hacked or something like that and we lose a cultural trust for medical sharing of data. So I think that the risk and the burden lies widely across both researchers, patients and IT, but where do you think we come together on that? And what are the standards that need to be adopted to make sure that patients are able to contribute data that's useful while still getting a sense of protection that maybe that data is not gonna be sold to a third party like their car insurance provider and say, oh, you're a stressed out guy, we're gonna raise your rates. - Yeah, that's a great question. I think what we're seeing now is that there's a level of sophistication in certain fields. Like today, we simply have organizational trust that the data that we give to our partners won't be shared. The problem is that those organizations, as much as they'd like to protect our data, don't really have the tools and technologies available, standards aside, it's very difficult to do data provenance to figure out where did a piece of data come from. So we in the data science, data community, in various capacities, have to say, if I care about privacy of data and if I care about making sure the data is trusted, I have to work on provenance. How can I carry data so I know about its original data from wherever it comes from? Like for example, in finance, a dollar is a dollar, it doesn't really matter where it came from. But when you're tracking bank accounts, now it matters all of a sudden. In medicine, it's much more granular and nuanced. You can say, I have a temperature, it was carried in Fahrenheit, but now I converted it to Celsius. What was the original? What did I convert it to? What were the original units and what were they converted into? That's proven, it's knowing where something came from, what were the units of measure, and we're really not good at that in the data community today. So we talk about normalization, we talk about schemas and we talk about how cool it would be to be able to do auto discovery, but if you don't get provenance right, you can never know where something came from. So that's very, very important. Second thing is discovery. Not sharing everything everywhere, federating in a way and allowing discovery to occur to say, I would like to discover who has this kind of data. So once you have bit provenance, you've got the good structure, you've defined it well, metadata is there. And the good news is 80 or 90% of what we need is available in standards. The 10 or 20% on the provenance is still open. But if you had federated discovery, then you wouldn't have to send data everywhere. You leave it in places and say, okay, I trust this organization. So I'm just gonna run a federated query to say, do you have Kyle's data? It returns yes and email goes at the Kyle and say, hey Kyle, looks like your data is at Johns Hopkins, would you mind Harvard Medical getting a copy of it? And you say, no, of course not, because I trust both of them and I hit the button. That kind of stuff just doesn't exist today. And the reason it doesn't exist isn't so much the technology, it's who's gonna pay for this? Because when you're talking about one piece of data that's my business transaction sitting in my accounting system, I don't mind paying for that. 'Cause nobody's asking me to share my accounting system with a bank, right? Now all of a sudden, let's assume that the bank wanted to reach into your accounting system and say, hey, instead of you telling me and me doing reconciliation, why don't I just look into your accounting system, pull data out whenever we have shared customers? Can you imagine how crazy that would sound? I'm a Bank of America. I go to a hospital who has accounting records. If I could think that Bank of America could grab my data from the hospital, I'd be scared. Or vice versa. When we're talking about financial data, we never think about that. But think about it in healthcare. Multiple hospitals are competitors of each other. How can we simply ask them to just share this data without knowing where things are coming from? So I think some level of data discovery with federation plus provenance gets the foundations in there so that when we think of doing like a sparkle query or the distributive queries that we would do with RDF and other things on the internet world, that works, right? RDF would work in the medical community. It would work really well. But everybody's afraid because why should I share this with my potential competitor? I mean, there are two primary care physicians on the same block. Those aren't just two doctors sharing data. Those are two competitors asking for patients to come to each other. And that's the kind of stuff that we don't think about. We think of healthcare as needing to be pretty much free as we might want it, you take care of my medical problems for me, you take care of my healthcare records for me until we take ownership, until we take responsibility, this stuff is gonna be really hard. - Yeah, but that distributed federated infrastructure you described is really exciting to me. And I'm familiar to pretty much anyone who knows a little bit about big data, Hadoop file system, this sort of thing. It's right in line with the philosophies that are going on. How do we get there? 'Cause as you lay it out, I see the challenges, but I definitely hope that comes to fruition. Is that gonna come from a consumer push or the an insurance push? Where will that finally start? - I think the first place where it's going to come is as we see the business models changing, when we move from what's referred to as fee-for-service or FFS says like you go into a doctor, he sees you, you pay him 200 bucks, you get your tasks done, you pay them for the services that they are doing, everybody seems like they're a competitor, right? If my imaging service costs $500 to do an MRI, yours costs $400, I'm not gonna share anything with you. But if it turns out we can do, and you heard me mention these accountable care organizations, this is one model that the government is experimenting with in what they refer to as a shared savings plan. And that is, if you and I got together, if I'm the doctor and you're the imaging supplier, why don't we get together and take ownership of the patient? I won't ask for an imaging study if you just did one yesterday, because you're gonna send me the data immediately, instead of me having to call you up 10 times and say could you send me the data. So when we see a business model, which basically rewards data sharing, some of this will happen. Technically, all of us as technicians and engineers, we need to continue working on data provenance and RDF and all these other cool things that we're working on. But foundationally, it's going to require a business shift in the same way that we saw, for example, 10 or 20 years ago, as value added network started coming in, they started coming in because multiple businesses said, it's more valuable for me to share the data, than it is for me to keep it to myself. That has yet to happen in healthcare. And you would think that healthcare would happen first, because this is all useful data. They haven't seen it, it's hard to do. So it's not like they're just sitting on their thumbs and they don't know what they're doing. It's hard. No other industry has the depth of information about me as a person, you as a patient. It's just mind-boggling the amount of information. Anyone who says, well, they've worked in big data, if they've not spent any time in bio-IT, like that example that I gave you with the brain, they just don't know. They don't know what real big data is, until they've worked in this area. That said, we're behind, right? Healthcare is behind in it, because our business models don't support data sharing, just yet. We haven't gotten to that level of sophistication. So if you're wondering, when is this going to happen, let's promote more and more accountable care organizations. Any politicians that say, let's work on shared savings plans, let's work on making sure that data sharing is good business and good medicine, and then we'll see a rapid migration, because good technology will follow the money pretty much any day. Yeah, and that's definitely a platform I'd like to hear out of more political candidates as well. Yeah. So we covered, or I mentioned earlier, all three of your blogs, and I'll be sure to get those in the show notes as well, and I recommend listeners go check those out and become readers. Also, the conference, I'll put in some links in reference to that in the show notes. Is there anything else, any other good resources you'd recommend for people interested in the field? Yeah, I would say that there is a lot of good bioinformatics and other classes around health IT being given in data science circles. So you look at Hopkins, Stanford. You can, I believe, have some good ones. UPMC has some good ones. There are a lot of data science types of classes and online, especially, they're mostly free, that have a bioinformatics component start there, because if you understand big data, or think you do, and then you come into the bioinformatics world in which you can see the volume and velocity is just so much greater. I mean, there's a customer we're working with in which they're working with 300 patients. The amount of data that we've already captured on 300 patients, just for basic analysis, is going to be roughly on the order of 90 to 100 terabytes, 300 patients, right? And that's gonna be used for specific specialty area kinds of biomarker development and reviews and other things. So it's just a different class of problem, and we just need new guys. So definitely, as you're learning about data science, see if you can take the specialty courses around bio and pharma and pharmaceuticals and those areas, so you get the big learnings about big data, but you do it in a way that can help humanity, that'd be great. - Yeah, I think that's a fantastic opportunity for someone who's eager to get into those spaces. So, I'd like to wind up my podcast by asking for two recommendations from my guests. The first I call the benevolent recommendation, which is it can be anything, a blog, a book, a website, a movie that have no necessarily relationship to, but think is of value and might recommend people check out. And the second, what I call the self-serving recommendation, which is something that directly benefits you and hopefully gets you some publicity and viewers for your participation in the show. So if you wouldn't mind sharing two recommendations? - Absolutely, yeah, my first recommendation is I've been reading a lot of books on the subject of leadership, because that's really what I think we're lacking in a lot of what we're doing in both medicine and in some cases, MIT. So my first recommendation is I'm reading four or five books all together, so just if you go to Amazon, look on leadership books in general. And when you look at leadership, try and say to yourself, what can I do to change what's broken right now in healthcare? Healthcare is a huge industry, 2.3 trillion to 2.5 trillion dollars, but it has a lot of nooks and crannies. When something is big, we often think, or there's nothing I can do about it. But what leadership shows us is that if you look at a problem in a way that you're trying to solve for a purpose other than what your industry is, like you and I are in IT, if we look at it from the IT angle, that will create bad leadership, 'cause we're gonna say, okay, well, we just create better tools that doctors can use. Instead, let's lead and say, what is the doctor's problem in understanding what they do that prevents them from using the tools that I create, that model of leadership and understanding how companies should be created and tools should be created. Like if you're an open source, don't just create the tool 'cause it scratches an itch for you, but it scratches an itch in healthcare for the potential consumer. That's why we see open source being used a lot in our industry is they're all scratching our itches. They're not scratching someone else's itches outside of our industry. So that I would say, just read a lot about leadership and focus on how you can help from the outside rather than the inside. - Makes sense. - Yeah, and the more benevolent one, the less benevolent one, I wouldn't say, is visiting two of our sites out, netspective.com. We're a company that focuses on helping companies get started, building tools, technologies, et cetera. We service medical device companies. We service data companies. We service health IT companies, who are building their software and want to do it in a way that accommodates all the stuff that we talked about, next generation patient generated health data, next generation devices. How do you bring all of this in there? And the other, of course, lastly is my blog at healthcareguy.com. I love getting guest articles or other things. You can go to healthcareguy.com, click on request guest articles. It's a little video of me that explains what are the kinds of things that other entrepreneurs and other innovators and other engineers look for in their tools. And so my blog is one of those where I call a operationalizing trends, or operationalizing the implications of trends blog. I like to be very, very pragmatic, and I'd love to get people to say, "Hey, I'd love to jump in, "and here's how I solve this problem in healthcare," or in another industry that might work in healthcare. So that's my self-serving one. - Yeah, all great resources. Well, this has been a great conversation for me, really informative and rich. I'm sure my listeners are gonna enjoy as well. Thanks again for coming on the show. - I appreciate the invitation, and I look forward to attending in the future. - Yeah, best of luck with the COF coming conference. - Thanks. - Take care. - Bye-bye. (upbeat music) - Thank you for listening to the Data Skeptic Podcast. For show notes or other information related to the show, please visit our website at www.datasciptic.com. Follow us on Twitter @datasciptic. If you enjoyed the program, leave us an iTunes review and help others find us. (upbeat music) (gentle music) [BLANK_AUDIO]