"Digital transformation: Our trajectory is good, our position is imperfect."

Welcome to the big unlock podcast where we discuss digital transformation and emerging technologies in healthcare. Here are some of the most innovative thinkers and leaders in healthcare and technology talk about how they are driving change in their organizations.

Paddy: Hello again everyone and welcome back to my podcast. This is Paddy and it is my great privilege and honor to have as my special guest today John Halamka who is the former CIO of Beth Israel Deaconess and he is now the Executive Director of the Health Technology Exploration Center for Beth Israel Lahey Health. John welcome to the show. .

John: Well thanks so much. .

Paddy: You’re most welcome. So, John let’s start with this after a long career as CIO you’re now leading an exciting new initiative. Tell us how Technology Exploration Center came about and what are you focused on.

John: Sure. So, if we look at the evolution of healthcare digital health innovation throughout the world what you’re seeing is an explosion of apps, cloud services, wearables. But does anyone really know if these works or not, how to integrate them into workflow, whether they’re helpful to patients or providers or payers. The answer is not totally. And that building an exploration center inside an operational healthcare system to test out a worldwide emerging technology seem like a reasonable next step. After twenty-three years of being in an operational role I am now in a mentorship and guidance role for many young faculties and those who are part of the Beth Israel Lahey Health innovation ecosystem.

Paddy: Right. And thank you for that. You know you and I met at the Harvard Medical School Executive Education Program where you were on the faculty. You talked a lot about what you’re seeing in digital health and digital transformation of healthcare across the globe based on the work that you’ve been doing now. Healthcare in the United States is in early stages of digital transformation and the definition of digital varies from health system to health system. How do you define digital? What caused digital different from traditional IT?

John: Sure. So, how about this. We’re all on a journey to turn data into information, knowledge, and wisdom. And when I see some healthcare systems that say – Oh we’ve gone digital, we have created PDFs of paper records, we’ve digitized fax workflows. Well OK! I mean yeah that’s a digital paper but is it really possible to do machine learning on a fax. Are we able to, with a digitized PDF, do structured analytics using vocabulary controls? Well not so much. So, to me digital means codified vocabulary-controlled data that is capable of serving systems that turn it ultimately into wisdom.

Paddy: That’s a really interesting definition. I’ll come back to the question of data and advanced analytics and all of that in a bit. I just want to spend a minute again on the current state of digital in healthcare. Even the health systems that we work with and we track they are driving digital mostly as a portfolio of standalone projects. Somebody is doing telehealth, someone else is doing remote monitoring, hospital home, all kinds of things. A small handful of enterprises are taking a step back and looking at digital as an enterprise strategy. What’s your take John on the current state of digital transformation by applying the definition that you just applied? Where are health systems in this journey today?

John: Sure. So how about this our trajectory is good, our position is imperfect. And what do I mean by that. My daughter is 27. In her life she will experience coordinated care that gives her the data from each visit on her phone and I have had a mostly paper based medical record in my life. And of course, my mother has had nothing but paper right. So, you know we have gone over the course of the last 50 years to the codification of problem lists, medication lists, allergy lists, slabs and rads. We’ve got federal regulations that are encouraging us to share data across the community but it’s still a subset of the data. And so, we look at cancer in particular. My wife was a cancer patient and how easy is it to get structured cancer data out of even the most modern electronic health record today. So, it’s still tough right. Things like where is the tumor, what are the nodal involvement are there, distant metastasis. It’s mostly written in notes, its unstructured. And so I would say where we are is we’ve taken in a pareto analysis the first bold steps to put in structured data things that are most important but the tail will be very long and certainly on areas like cancer we really should move quickly so that the patients in the future have the benefit of the data from patients in the past.

Paddy: Well you talked about coordinated care and obviously that implies a certain level of coordination in the sharing of data. I can tell you from my personal experience that it’s probably the experience of most health care consumers. Data is not very portable. If I have a visit with someone who is different from my primary care physician. Those visit if it’s a wellness clinic don’t automatically make it to my MyChart and then I have a discontinuity in my records and that potentially hurts me because then my PCP has to come up to speed etc etc. He has magnified this problem across the entire healthcare system. So, let me ask you this question. You mentioned data that is fundamental to the digital transformation. My sense is and this couldn’t be a provocative statement. Much is made of the potential for advanced analytics, but the vast majority of health systems are still deeply entrenched in retrospective analytics. What is your take on that?

John: And of course, you have to look at different use cases. So, if the use case is – Oh, I am working on reimbursement and will reimbursement take many forms fee-for-service, value-based purchasing. But your retrospective analysis of performance and quality might work you know as a short term first effort and you are correct that prospective at point of care, workflow integrated clinical decision, support based on machine learning, care plans and algorithms is still very much a work in process. And, where do I have hope the FHIR CDS hooks specification which basically says EHR will be able to consume cloud services provided by external entrepreneurs inside the workflow so that you have more prospective guidance and best practices. It’s early it’s coming. So yeah, you’re right for reimbursement purposes a lot of retrospective analysis today but for safety quality total medical expense control FHIR CDS hooks and prospective Point of Care Decision Support coming in the next year or two.

Paddy: Alright, now you mentioned FHIR, Fast Health Interoperability Resources, for those who may not be familiar with the acronym. We have to talk about therefore the I word, interoperability. And I know you’ve been deeply involved in efforts to standardize APIs as part of your work with HIMSS with the Argonaut project. Tell us a little bit about the state of the union on data interoperability.

John: Well sure so if we do the very quick history of interoperability remember in 1989, we said oh what we need is something that sends demographics and orders and results around HL 7 V2. And actually, I mean we’ve seen pretty good success of being able to knit together ancillary systems and core electronic health records with HL 7 V2 but that by no means is interoperability of your entire lifetime record or sufficient for many other use cases. So, in the Bush administration and the Obama administration we moved to XML forms and although those were good at providing a summary record of an encounter, they were filled with optionality you never quite knew what data you were going to get. And they were challenging to get say a subset. I just want to see the EKG or just the lab because they were a document right. They read the entire visit. So, the next logical step after the CCDA was an API where you are able to query a resource and say give me just this bit of data in a highly structured schema for this patient. And that notion is very empowering because it means that entrepreneurs without a huge amount of healthcare IT expertise if given a spec for an API can say oh well all I need is the allergy information and here’s how I call it and now I’ll get the substance, the reaction, the level of certainty, the observer in a date timestamp with great consistency I don’t need to Parse XML. I don’t need a PHD in informatics. So, we’ve got our EHR our vendors going from V2 to CCDA and now to APIs a federal regulatory requirement is likely to be on the books soon requiring these Argonaut APIs for patients to easily access their core medical data. So yeah, I think that trajectory is pretty good.

Paddy: Yeah and that will be hugely empowering for patients as well. So, we certainly hope that happens sooner than later because it’s been a vexed issue for several years running. So, John we do something called a lightning round in my podcast where I ask for top of mind thoughts, I ask my guests. The top of mind thoughts on a handful of emerging technologies in health IT so it’s a perfect segue from our immediate last discussion. AI and machine learning what are your thoughts?

John: Sure, the challenges is you need to be careful about what use cases you pursue. And let me give you just two quick examples. So, you ever flown a 737 MAX airplane. How do you feel about a machine learning algorithm based on a single sensor having closed loop control of flight control? Oh, that sounds very bad. So, when somebody says no, I’ll use machine learning to control ventilators, drug dosing, make diagnoses. You say you know maybe what we want to start with are use cases that augment clinical decision making by helping say narrow the scope of what our potential actions but ultimately give those access to humans and all the machine learning work I’m doing is about augmenting and sifting through data. I’m not building any closed loop use cases for now.

Paddy: Right and when you say close loop use cases, you’re referring to use cases where the machine pretty much makes the decision on your behalf. Is that what you mean?

John: Right and may make a ventilator change, a drug change, a diagnosis you know the machine is in fact without human intervention. Creating an action that potentially could result in harm.

Paddy: Right. I don’t think the public is ready for that either. I kind of agree with you on that. All right. Next one on the list voice recognition, Alexa, Siri. What do you think of those?

John: So if you look at the history of our industry and I’m 57. So, I was there for the beginning of the PC revolution. I built an Altair 8800. Wow we didn’t need a mainframe. We could have something in our homes and then we went from that to the web. And from that to mobile. Well to me the next great natural extension of computing is Ambient Computing. And that is a quick example for you. My mom, nearly 80, lives in Southern California and I have with her permission created a comprehensive Google home environment. So, she simply walks around her home and says oh I’d like to watch this program, I’d like to listen to that music could you give me information on this author, where do I buy this book? And she doesn’t even think about it as being a computer. It is simply an ambient part of her environment. And certainly, I believe that this is the next direction for much of what we need to do in health care is we just whether we’re a clinician creating a church or a patient trying to navigate the health care system. We’re using these ambient tools and not typing.

Paddy: That’s fascinating use case. Do you ever have privacy concerns around this pervasive use of the technology in the home and this constant stream of information that is being gathered up and servers at the backend somewhere?

John: Right. And of course, every one of us has a different privacy preference. And in my particular case I have made the decision that you know I run a 70 acre farm the entire farm is Ambient Computing enabled the convenience and efficiency of having Ambient Computing for me to get through my day outweighs what I perceive as my privacy risks. But everyone may feel differently and in fact on the ambient computing devices on the farm there is an on off switch. So those who feel uncomfortable simply turn it off and that’s OK.

Paddy: OK all right. The next one on the list. This is something that you have become quite an expert on blockchain.

John: Right. So again, with blockchain you need to pick your use cases carefully. Is block chain a data analytic platform? No. Is it interoperability platform? No. And is it a management platform? No. Is it fast? No. Is it easy to use? No. But things like ensuring data integrity I can prove that a medical record was not altered or a public ledger that enables me to find information about doctors location or credentials or the notion that I as a patient might declare my privacy preferences and consents in a public ledger for general reuse. These are some use cases that absolutely have promise. So, it will not solve all problems, but it certainly is handy to solve a few.

Paddy: Right. And you mentioned the provider directory and alliances like the Synoptic Health Alliance are trying to solve just that problem. What really is a critical success factor for a blockchain initiative to gain widespread adoption. Is it the network effect? Is it something else?

John: And the answer of course it’s value alignment of incentives. So, for example as a physician I have over 1000 insurance companies that want my provider directory information and my credentialing data and things of that nature. It’s a nightmare. So, what you said was we now have a who cares is blockchain. We have a workflow that is going to radically reduce complexity in your dealing with insurance companies. Great. Anything that saves me time reduces costs a prevents public embarrassment whatever you know that will be a critical success factor for adoption.

Paddy: Right. OK. Last one in the lighting round, extended reality – AR, VR.

John: Sure, so many of my surgical colleagues feel the notion of taking imaging studies and overlaying those in a field of view as one is doing an operation provides a level of guidance and safety. Certainly, the notion of training using a VR is important in many of the sim centers do that fairly well. So, as a physician who trained in a very old fashion dissect a cadaver. And by the way the tissue is largely degraded and very hard to visualize. I absolutely see the value of both augmentation and training using these techniques.

Paddy: I’m actually familiar with that use case. It’s a startup that I’m aware of that uses haptics to simulate the whole cadaver experience that you just described. Fascinating. Let’s switch back to digital health. Now this week has been a big week for digital health. Two big IPOs – Health Catalyst, Livongo – hard to escape you know all the attention that’s going into those two. So, 10 billion in VC money every year give or take. Most startups are struggling and after a long draught out we have three major health IPO this year. What’s your take on the current state of digital health startups and digital health adoption? Are we turning the corner here or are we going to continue to struggle for a little longer? What’s your take?

John: Sure. So, I think there is some worldwide societal problems to address like the aging society, lack of access to appropriate specialists. And so, when I look at some of these digital health startups which are doing Internet of Things, telemedicine, AI, machine learning they are addressing some of these major societal problems. So, as John Kotter at the Harvard Business School tells us nervous is going to change unless there’s an urgency to change. And having health care at 18 or 19 percent of the GDP aging society a low birth rate and not enough care is an urgency to change. So, it seems to me that the next couple of years we’ll see these digital health startups become an essential part of our health care system and potentially could even reduce some costs or if nothing else bend the cost curve.

Paddy: Right. OK. We’ve pretty much coming up to the end of our time here too. And I have one last question for you. Your career 23 years as a CIO and more you successfully reinvented yourselves yourself a few times over. What advice do you have for those coming into CIO roles in health systems today?

John: I know I am often asked to predict the future and I say I can look ahead to six quarters. Beyond that who could predict right, who would have predicted the internet’s impact in 1993. So, I think the answer is just be agile, recognize your time horizons are short, and take risks because you don’t quite know what technology is going to triumph. So, try a lot of them, fail fast and eventually one will hit. And as you point out their reinvention looks like it was planned well to be honest sometimes just good luck.

Paddy: That’s a candid assessment. John it’s been such a pleasure speaking with you. Thank you so much for coming on the podcast and look forward to following all of your writings in the Geek Doctor and the blockchain newsletter. Among the other things that you are doing. Thank you once again.

John: Well absolutely glad to be here and thank you.

We hope you enjoyed this podcast subscribe to our podcast series at www.thebigunlock.com and write to us at info@thebigunlock.com