Dan:
Judy. Welcome to the podcast.

Judy:
Thank you.

Dan:
So Judy, you've worked in a number of places from healthcare policy to informatics to IT, and now you're at IBM. Can you tell us a little bit about your background and what you're currently doing at IBM as a nurse?

Judy:
Yeah. So it's probably helpful to trace just a bit of that history. I grew up, if you will, at Aurora Health Care in Wisconsin. Worked as a staff nurse, worked in administration of nursing and then, moved into the inservice education department. And through that, I got involved in computers because I was responsible for doing some of the training of the new mainframe applications that we had at the time, which was the 1980s. And got real interested in the computerization part of it and pitched the job to the director of IT at that time. The role of CIO and some of the infrastructure kinds of things that we have now actually didn't exist. Although he didn't hire me on the spot, he hired me about six months later. And I will tell you there were only 27 people in the IT department at that time.

Judy:
And fast forward then. I stayed in the IT department at Aurora for about 25 years and by the time I left, there were 750 people in the department. Again, you can imagine all the different kinds of transitions that happened in the 80s, the 90s and into the 2000s, et cetera. But one of the things toward the end of my time at Aurora that I was involved with was The Meaningful Use Program. Some of you may recall who are listening to this, I was one of the nurses that was on the Health IT Advisory Committee, the Federal Health IT Advisory Committee. So I got to know the Office of the National Coordinator and what they were doing and got really interested in getting out of my little world, if you will, and doing something more on a nationwide scale.

Judy:
When I had the opportunity to apply for the deputy national coordinator position under First Admiral [inaudible 00:03:49], I snurfed that up and applied for that job and did get it and stayed at ONC for three years.

Judy:
I actually thought that was going to be lots different than my role at Aurora. But it turned out kind of the same, actually, because I was still working on electronic health records and sort of that first rollout of things. So when I began to think about transitioning out of the federal role and doing something else, that's when I really thought, "I want to do something beyond the electronic health record." I was doing that for almost 30 years of my life at that point. And I thought, "You know, if we're going to really change things in healthcare, it's going to be a company like IBM that's going to make a difference." It may not be those incremental changes that were small with some of the other health IT technology that we're using. So making a long story short, I ended up getting interviewed and applied and hired at IBM to work on the healthcare specific industry things that IBM was doing.

Dan:
That's great. So IBM has a huge portfolio of technologies. Where did you start focusing once you started there?

Judy:
So they were moving from brand strategies to industry strategies right around the time I was hired, which now, is about five and a half years ago. They were looking for industry expertise in 15 different identified industries and healthcare was one of those 15. So extremely interested in thinking about how the IBM assets, could in the brand things that we offered, could be made more specific to work well in an industry.

Judy:
So for example, analytics. We had people that knew data strategy and data analytics inside and out. But what they didn't know was healthcare. And so, that's where my expertise was going to be helpful. Helping with, not only product development and consulting and thinking how we use our consulting force, but also in the sales area. How do we want to position our products so that folks will understand the unique things that we have to offer that match up with what that industry needs?

Dan:
Yeah. That sounds very familiar. You're like that translator and that meaning maker for all of the awesome technology and how does it translate into the huge complexities of a massive industry like healthcare?

Judy:
Absolutely. And needless to say, we had a lot of folks at IBM who didn't know industry. And so, a big part of my role was also what we refer to as enablement. And enablement means education and training specifically related to that particular industry. So I built some programs and did some web X presentations and in person presentations helping people understand what was all going on in the industry so that they could adapt it in their sales and development place.

Dan:
Got it. That's great. That's similar to the role I'm playing here at Trusted too, which is kind of that industry translator and what I had done in the past too. And it's always interesting to have a nurse leading that translation. 'Cause I feel like that's what we do on the clinical side, and now we're doing it in the tech world, and it seems to come naturally. As for me, it seemed to come very naturally on how to translate complex things to people and be that coordinator. I don't know. Do you feel similar?

Judy:
Oh. I absolutely agree. For a long time, I've been saying nurses are the best project managers when it comes to managing any kind of a project, whether it's IT or not. Because it's like the nursing process. Right? You gather data, you create a plan, you execute the plan and evaluate the plan. So same thing here. We're so used to explaining things to people in a way that they can understand, that I think, as you say, it just quote unquote comes naturally to us.

Dan:
So one of the products that IBM has is obviously Watson. It has a lot of press about it. Have you worked a lot with the Watson capabilities and tried to translate them into healthcare?

Judy:
Absolutely. It's one of the industry areas that we have been very focused on. Realizing that this is an area that can benefit greatly. Now, I have to say IBM's history and legacy is in things like security analytics. Now Cloud. Certainly we have a big legacy in hardware itself and still sell hardware, but hybrid cloud, public cloud, red hat, open source, all those kinds of things really lay the infrastructure. Because a lot of what we're doing in Watson is things that we are selling by the sip, if you will. So there's things that are offered as a service and we sell that service. So without that kind of IBM first infrastructure stuff, we wouldn't even be able to have our Watson product line. But in our Watson product line and in healthcare specifically, there's all sorts of things. I don't know if you want to delve into a few of them?

Dan:
Yeah, I'd love to hear some of the top use cases that Watson's impacting healthcare with.

Judy:
So oncology has been an area that we've been focused on since we first started this. In the oncology area specifically, we're looking at clinical trial matching. So the ability to take and find, use electronic health record and other clinical data to find patients who qualify for clinical trials. And there's sort of two ways to approach that. One way would be to go to a provider and look at the patients that they serve, and look at the data that they have about them in the electronic health record.

Judy:
The second way to do that is to work with the life sciences or pharma companies that are offering the trials that don't have the same kind of access to that information, but they want to be able to outreach to patients to see if they'd be interested in being involved in those clinical trials. And so, we can play that middleman, of course, with all of the security and privacy concerns taken well into account. But actually, outreaching to people that might be candidates for clinical trials that otherwise wouldn't know those clinical trials are available.

Judy:
So if I'm not at a Mayo, I'm not at a Memorial Sloan Kettering or an MD Anderson, but I'm interested in getting involved in a clinical trial, how could I find out about those? So there's really both of those use cases that help actually find the patients.

Judy:
The second area is in the actual care of the patient. There's constant research being done and constant publication of new ways to treat things and the results that have been obtained. I don't remember the exact statistics, but you and I could not keep up with reading all of those articles to be able to treat patients up to the latest and greatest evidence. And so, Watson for Oncology is the product name. Watson for Oncology helps providers know what the latest and greatest is based on a particular patient and then can have that conversation with that patient.

Judy:
Because what it incorporates in is the probability of success and the probability of side effects and the impact that it's going to have on the term of life. So it gives a lot of that kind of information as a discussion document. It is not replacing the physician. It is giving information to the physician to work with the patient. So a discussion document, as I pointed out.

Judy:
And then, last but not least, the third sort of leg of that stool is Watson for Genomics. And Watson for Genomics is looking at the genetic makeup of the tumors themselves and making sure that we're applying the exact appropriate in evidence-based therapy to that individual tumor based on the research as well.

Dan:
Great. And it sounds like the theme through all of that is taking some of that transactional work or sifting through big pieces of data or articles or big datasets and automating that. And so, the theme of the series we're doing with this podcast is around the automation of tasks like that. Is Watson simply an automation engine or is it more than that? Is it providing machine learning insights and those types of things?

Judy:
Definitely applying machine learning instincts. But let me make a comment on what you just said because it's extremely important. We've got sort of a little phrase we use around IBM a lot, and we say "There's no AI without IA." So the AI, of course, is artificial intelligence, and the IA is information architecture.

Dan:
Okay.

Judy:
This stuff is not magic.

Dan:
Right.

Judy:
You've got to structure your data appropriately. You have to have the appropriate types of data sources. And so, what we often find implementing, whether it's the oncology products or any of the AI analytics products that we offer, the architecture is not complete or adequate to actually serve the AI. So you have to think about it, we always, again, use the term "a ladder". So you have to have all the data, but then you have to architect it in a way that is being accessed appropriately, organized correctly, if you will. And then, do the analysis on the data. And once you're doing the analysis, that's when you infuse the AI. So it's really kind of a four step process. And that becomes extremely important because somehow, people think AI is sort of magical. Like you can just start doing it. You can't just start doing it without that information architecture.

Dan:
And in healthcare it's been pretty interesting to see how much structured data there is. I mean, early on it was all on paper and written narratives and those type of things. And as we moved to electronic medical records, some of it got more structured. I would say more of the physician workflow got structured and nursing was still kind of this nebulous thing. We're getting more and more into structured data within nursing, but there's still these narratives and things. So as you look at that data, how do you take into account that unstructured or natural language processing kind of piece of the narratives that add a lot of the context to this more structured data?

Judy:
Yeah. Real important point that in healthcare we automatically think of the electronic health record. And so, I want to address right up front, the electronic health record documents our encounters with the healthcare organization. Though it could be ambulatory, it could be acute, it is a small percentage of the information that we actually need to understand health, healthcare and the behaviors of the individual. And we've all been talking about migrating away from the sort of facility centric concept of healthcare and moving to a more continuum based concept of life care. So the data that becomes important isn't just about structured or unstructured. But it's also the sources of the data itself. Family history and genetics, personal behaviors, environmental and social factors. Those all become much more important than what we actually have in our electronic health record today. We've been talking about something like maybe we capture 10 to 20 percent of the data that we actually need in our acute ambulatory facilities today in an electronic health record.

Judy:
And then, there's that 80 or 90 percent of all those other kinds of things like genetics and behaviors and social determinants of health. Well, it's become a little bit harder to capture whether it's coded or unstructured. We have come a long way in taking unstructured and making it understandable or taking unstructured and actually codifying it. And so, that's helpful. But I want to bring up this idea that we have to start looking at datasets outside of our normal world of this health IT concept that we have, and start thinking about zip code data that we can get from the government. Information that we can get from social media, information that we can get from things like the weather channel or other datasets that can inform peoples' behaviors and how they're living their lives and where they're shopping and all those kinds of things. And again, I'm going to bring up we have to respect security and privacy. All that kind of stuff. I'm not talking about know looking at data.

Dan:
Right.

Judy:
I am saying that that data becomes really important. So with the right structures in place, we need to start adding that in so we can understand how to impact peoples' health and impact their healthcare in a way that we haven't done in our past.

Dan:
Yeah. No. That's a great point. And that contextual data of zip codes and styles of living and access to food and all those things that the social determinants... I think we're finding more and more data and literature is saying that that's even more impactful than the one percent of time you go visit your physician. And so, how do you take that and actually add that in? I'm interested to hear about as you're out talking with different clinicians, what are they talking about or what are they asking for or even using when it comes to automation and this insightful data?

Judy:
Yeah. I think everybody is talking about how can we appropriately use AI. So earlier, I mentioned a few product names in our oncology space. But I think what people are starting to realize is that there are APIs, application programming interfaces, for example, that incorporate AI capabilities. So it's not the big, it's been productized and packaged all together, but it's got some clinical decision support capabilities in it as compared to being productized. And you can take those capabilities and you can do things with those capabilities. And that can be classifying, that can be tone analyzing, you can take speeches and you can digest them down that people give and what does that tell you. So there's some of that subset of AI things that folks are starting to talk about. But more importantly, how do we make this work for us?

Judy:
So we don't have to necessarily buy a product that does it, but we can incorporate it into everything we do. So just some examples. IBM happens to own Micromedex. And Micromedex, many of you who are listening to this have probably used it and had a hierarchical lookup. You would go onto the webpage and you would type in the name of the medication that you were looking at. And then, it would take you to the page that had that medication. And if you were looking for normal dosing for an infant or normal dosing for an adult or the different forms that it came in, you had to navigate through maybe five, six, 10 pages of information to find what you were looking for. Imagine today, we incorporated an AI look up into that, and that allows you to use a natural language question to find the information that you want.

Judy:
So I can type in, "What is the normal adult dose of Dejaxon?" in a little window and that'll take me directly to the Dejaxon page and the spot on the page where the normal dose for an adult is actually listed. See how much time that [inaudible 00:19:07] practitioner and how it changes their workflow. And that's AI. It's making a product that's seeking a capability and building it into something. So that's what I think folks are excited about.

Judy:
We've been talking about using it for example, in HR. HR with nursing in particular. We have turnover and it's above places that are maybe higher than we really want it to be. So can we use AI to analyze the people that have left in the last two years? Understand what they have in common so that we can look forward and say, "Well, if all the people who quit in the past had this in common, maybe I should be looking for people that have that thing so I can intervene and prevent them from leaving us in the future." So we're trying to really, I think, open up everybody's eyes to the different ways you can think about using AI.

Dan:
Yeah. No. Across all the different spectrums, I think people jump right into clinical decision support as the use case. But like you said, there's so much opportunity to help leaders understand the 150 people that report to them and insights into the workforce and culture and all those other things that go along that context around care. So that sounds really great. What, in your opinion, what is one of the most innovative applications you've seen using AI or automation out in the clinical space?

Judy:
So what I'm getting excited about is in the imaging space. For a while now, IBM and others have been looking at trying to help AI be able to actually look at images and understand what they're seeing. And of course, there's a huge training factor there. I said this wasn't magic. So you want Watson or any other AI product to help you identify a pulmonary embolia. You have to train it to what a pulmonary embolia looks like on an image. So that whole training, we're in our second, almost the end of our second year, and we now think that we're going to be able to actually release some product about that at the end of this year. But in doing that, looking at radiology, this workflow and how they do what they do, we've discovered that radiologists, in an ideal world, want information from the electronic health record that helps them read the image.

Dan:
Yeah.

Judy:
That history helps. So one of the things I get excited about is the ability of AI, for example, to delve through an electronic health record and find things that are important and relevant to the type of x-ray that's being read by the radiologist at any given time and bring that forward in a very smooth way to that radiologist and present it. Now, they could have logged into a different PC, logged into the electronic health record and looked for the information themselves. They don't do that the majority of the time. Would they want to? If they have the time. Yes. They want the information.

Judy:
So imagine again, if you can pull it and present it. So kind of like an AI synthesizer if you're doing a chest x-ray, knowing that these are the things that might be helpful to look at a chest x-ray. If it's an abdominal series and you're looking for, I don't know, an ileus, here's the kind of information that you want to pull forward that might help or inform that radiologist. So that's kind of how I get excited, is how do we augment decisions, not just replace the decisions that you make?

Dan:
Yeah, I've heard some people call it, "Giving clinicians superpowers" because it gives them data they've never had access to before to use their expertise in even new ways. So just some really interesting ways that you can give physicians and nurses superpowers in their clinical decision making. I think on the other side of that, and this is something I've been exploring on the academic side, is we're not training our future clinicians in how to take AI insights and incorporate it into clinical decision making. And so, I think there's that gap needs to be filled as well. Where academia needs to start preparing our future nurses and physicians and care providers in how do they trust a computer generated insight? How do they assess that it's a good thing? And then, also how do they incorporate it and not just take it for blanket a recommendation, just follow it, but actually put it into their decision-making process that they do on a daily basis.

Judy:
Oh Dan. That's a really, really, really good point. I think we've had all this fear factor about it replacing the clinician, and it's never going to replace the clinician. But can it augment? Can it help? Can it give them superpowers? Oh my goodness. Yes.

Dan:
So AI is the new trending. I think if we looked at the Gartner Hype Cycle, it's probably up there in the hype area, just because I think people see a lot of opportunity with it. There's definitely use cases which we talked about today that are in practice, actually impacting care. What advice would you give listeners that are leading health systems to even begin the process of incorporating AI even in a pilot or to consider it as part of a workflow that they'd like to improve over time?

Judy:
Well, I think everybody's got partners that they work with and I would suggest starting to open up these conversations with their partners. And whether they use that term AI, 'cause I said there's so many different types of it, or whether they just talk about augmenting decision making, they should start having those conversations. What are their electronic health record vendors doing around that? What are their radiology vendors doing around that? And if you don't work at a provider organization, you know the payors in the plans are also looking at some of this kind of stuff. So I think at least opening your mind to the fact that it's a broader context than just a product. That it can be a service that gets incorporated into a workflow that you currently have today. And I think once you open your mind up to that and start exploring that a little bit, you might find some easy ways to begin. Well, nothing's easy. I shouldn't even use...

Dan:
Nothing's easy in healthcare, Judy!

Judy:
It was sort of a joke. There may be some ways that you can begin to incorporate that into the way you're providing your care.

Dan:
Sure. Yeah, that makes a lot of sense. You kind of start small with the people you know, and then, you can expand from there. And I think what you mentioned also is it doesn't always have to be a clinical application. Sometimes, there's a lot more hurdles to that application of any technology. So you may be able to start within your own HR department or sifting through records from a research perspective rather than diving right into informing patient care decisions. So I think that's another good piece of advice.

Judy:
In fact, another good area, actually, to begin looking at is call centers. That's another area where you can use some upfront triaging and give some good answers back through chatbots and things. And chatbots are not AI, but chatbots can support AI. And they give you that 24 seven coverage, too. Where people in the middle of the night, they want answers. They can find some answers. So I think beginning to think about incorporating it into the call center work might be another interesting area.

Dan:
Yeah. It's always interesting with chatbots. People get this visceral reaction of, "Well, I don't want to talk to a robot. I'm calling a healthcare person. I want to talk to a nurse or a doctor." And I think the thing that chatbots can do, as augmented by AI and insights is, they route you to the right person faster. So you don't have to give your spiel 15 times to 15 people to figure out where you need to go in the system. But they can actually help triage you to the right appropriate care level instantly. So hopefully, it's actually speeding up your access to the real person, not replacing it.

Judy:
Absolutely. And one more comment on that. What you said is true. You'll hear people say, "I want to talk to a real person." And they're hitting that zero button-

Dan:
Right. Yeah. Yeah.

Judy:
To get to that person. The flip side of that is we've got an increasing number of people who love self-service.

Dan:
Yep.

Judy:
They don't want to make the phone call. They want to get the information online using their own fingers. You got to balance those self-services getting to be really popular.

Dan:
Yep. Yep. Totally agree. Yeah. And actually, the one of the models we have for our company here is more of a self-service model, too. Where nurses are given information, and they can make their decisions. They're not hounded by recruiters and those types of things. So we're finding that in even finding your next job, you don't want to have to go through the talent acquisition side of things. You want to kind of scope it out yourself, get all the data and move it forward. So I think it's going to impact multiple industries.

Dan:
So we're kind of wrapping this up. I would love to hear what you're most excited about in this field. And what's your dream scenario in the next 10 to 20 years with all this amazing technology that IBM and others are building related to healthcare insights, automation, AI?

Judy:
So my dream scenario is that care actually does move out of the facilities that we see it being provided in today. We've been saying this for years. It hasn't really happened. And when we think about "care", and I use that term in quotes, I'm going to use the term "health" really to replace that. When I think about what we do to take care of ourselves as people and as providers, how we take care of the people that we serve. It doesn't happen in the hospital. It doesn't happen in the clinic. It happens in our everyday lives.

Judy:
And so, my dream scenario is that we're supporting people in promoting their own personal health, and understanding where they are on the health continuum by themselves. That they can use things online, that they can use applications, that they have responsibility of understanding the importance that they play in their own lives. And then, also of course, understanding when they need to pull people in that are the more educated and things. So I believe that health IT is going to be our path to that tomorrow where we're really talking about life care, and we're not talking about the care that's provided in healthcare facilities.

Dan:
Well I love that dream. That sounds like a great reality where care happens everywhere and it's not just within the four walls of facilities anymore.

Dan:
Judy, where can we find you if we want to get more information about you and the work you're doing at IBM or some of the use cases you shared today? Where is the best place to find you?

Judy:
LinkedIn.

Dan:
Yeah. That's the place to go.

Judy:
That's going to be a great way to get through to me, no matter where I am and what I'm doing. So yeah. That's the best way.

Dan:
Great. So find Judy on LinkedIn at Judy Murphy. And Judy, we really appreciate you being on the podcast today. Some great insights for our healthcare leaders to understand automation and AI and the future of data and how it might impact their entire organization from clinical care all the way through the administration and workforce data. So really appreciate your insights.

Judy:
You got it. It's been fun, Dan.

Dan:
Thank you so much for tuning in to The Handoff. If you like what you heard today, please consider writing us a review on iTunes or wherever you listen to podcasts. This is Dr. Nurse Dan. See you next time.