My friend Charles Boicey joins to talk practical Data Science and Artificial Intelligence. We also discuss the push to a national unique patient identifier in healthcare.
My friend Charles Boicey joins to talk practical Data Science and Artificial Intelligence. We also discuss the push to a national unique patient identifier in healthcare.
Bill Russell: 00:06 Welcome to this week in health it where we discuss the news information and emerging thought leaders from across the healthcare industry. It’s Friday, February 23rd
Bill Russell: 00:15 this week. Do we need a national unique patient identifier, AI in the clinical setting and how to get started with artificial intelligence project at your health system. This podcast is brought to you by health lyrics, a leader in digital transformation in healthcare. This is episode number seven. My name is Bill Russell, recovering healthcare CIO, writer and consultant with the previously mentioned health lyrics. Today I’m joined by a great friend of mine and a wicked smart data scientists, Charles Boicey. Wicked smart. That’s for my friends on the uh, in the Boston area. Uh, Charles is chief innovation officer for clearsense, a healthcare analytics organization specializing in bringing big data technologies to healthcare. Prior to clear sense, Charles was the enterprise analytics architect for Stony Brook Medicine. In his role, he developed, uh, the analytics infrastructure to serve the clinical and operational and quality and research needs of the organization. He was a founding member of the team that developed the health and Human Services Award winning application. Now trending to assist in early detection of disease outbreaks. So utilizing social media feeds. Charles holds an ms in technology management from Stevens Institute of Technology and is the president of the American Nursing Informatics Association. Good Morning Charles, and welcome to the show.
Charles Boicey: 01:31 Good morning.
Bill Russell: 01:33 Wow. You have so many Bios out on the web. I hope that one was current. You, you do a lot of speaking and a lot of different things. Is that, is that when pretty current?
Charles Boicey: 01:41 Yeah, it’s pretty current. With the exception, I am now past president of the American Nursing Informatics Association. I’ve got one to add for you. Um, professor at Stony Brook Medicine. I helped them develop out their applied analytics master’s program. So I developed three classes and I teach three classes there.
Bill Russell: 02:04 Wow. So, so can we start calling you professor?
Charles Boicey: 02:08 Yeah, but you’d have to use it at like an assistant level professor, not quite the full fledge.
Bill Russell: 02:15 You know, we’ve worked together for a while. In fact, we were at competing health organizations and I was so impressed with your work. I brought you into to talk to our team and then eventually to the leadership about what could be done with big data and big data analytics and, uh, uh, and I’ve tried to hire you on several occasions. You’re the hardest person to hire. You just, you’re so loyal to the people your with. Um, so I, you know, I’ve been stilted a couple times, but I still enjoy our friendship because you, uh, you really caused me to think about this, about these topics of machine learning, AI and big data. So one of the things we do is we like to ask our hosts to give us an idea of what they’re currently working on are, or what they’re excited about. So what, uh, what have you, what do you have going on these days?
Charles Boicey: 03:03 Sure. So there’s a couple of things. Um, I’ll kind of start off from, um, you know, the academic perspective or education. It applies the s and the academic environment, but a lot of the, the, the work that I do, um, you know, with Clear Sensors is actually education, if you think about where healthcare is, you know, as far as analytics, machine learning, AI and whatnot, many of us are still in the spreadsheet, stage, right? Um, some of us have progressed to data marks and so forth with some visualization, but very few have made the leap into, you know, these big data technologies, you know, advanced analytics, AI and so forth. So a lot of what I do and what I really am excited about doing is educating clients, prospective clients. And as you know, for the last eight years or so, I’ve been evangelizing this throughout the community and whatnot.
Charles Boicey: 03:58 On the professional side, on some of the things that we’re doing with our clients I’m really, really excited about are in areas of operation operations as, as well as in clinical practice on the operation side. And we’ll take a look at it like let’s say patient access, if you will. We all have a problem with keeping the clinics, you know, fall in that capacity and whatnot. So a lot of the work that we’ve done lately is developing out models that take into consideration a patient’s future or past attendance habits. We take a look at the weather, we take a look at traffic patterns. Uh, we take a look at a lot of different extraneous factors, put that all into a model and then give our, give our clients the ability, you know, at the end of the day, these are the folks who are likely not to show up tomorrow, but not just that give them a list of those that are likely to be able to fill, fill in those slots if those folks, you know, indeed don’t show.
Charles Boicey: 05:00 And so they’re able to make those phone calls at night to get a definitive yes or no. I’m going to be there if it’s a no when we slot somebody else in and we keep our clinics full. So that’s pretty exciting. On the operational side. On the clinical side again, uh, you know, just working on situational awareness type applications. Um, you know, loading up a machine learning environment with all the physiology, um, all of the, you know, the, the laboratory data physiology data and really identifying patients that are likely to crash in the next 30 to 60 minutes. So, um, again, when we get into the AI side of things today, we’ll talk about AI versus what I like to call intelligence. So those are the kinds of, some of the things that we’re working on that are pretty exciting. It has to be practical and, and we’ll, we’ll get into that.
Bill Russell: 05:52 Yeah. So we’re, we’re just, we’re still scratching the surface. I remember one of the first use cases you gave me was, uh, as a CIO I was sort of struggling with an EMR implementation and our doctors were, um, we’re struggling to figure out how to, how to utilize the Emr effectively. And you told me about how you use data science to collect all, for big data to collect all this information, then use machine learning and data science to determine which doctors were actually having difficulties utilizing the Emr. You were actually able to identify them by name so that you could really do targeted education and training of those doctors based on uh, how, how they were getting lost in the system, how many clicks because we have all that data, right? Every click that happens in the EMR where we’re tracking it and you just said there’s value in this data, we’ll figure out how to use it. And I thought that was an interesting use case. You still, you still using that use case
Charles Boicey: 06:50 Yea thats the HIPPA law and and so forth. And you know, an interesting, an interesting side note on that was, and we did exactly what you described, but when you tap somebody on the shoulder because you know where they are at any particular time and then you know, say, Hey, we noticed, you know, in the last couple hours you were doing these kinds of things and it looks like you were struggling a little bit and that kind of flips them out initially after they calm down. You can, you can have that intervention really kind of helped them on their way and then track them going and going forward. We had, we had somebody, we had a, we had a clinician that was actually getting 150 a lot of notices, alarms a day, um, and which is quite extraordinary. And then once sitting down with, we kind of helped them walk it through and they got much less, you know, alerts and so forth. So yeah, there’s those Hipaa laws, there’s tons of data in there.
Bill Russell: 07:43 Yeah, there absolutely is a cultural aspect of this. So we’ll get to that in the second segment when we get there. Let’s get to the news and how you, and I know you and I can have really long conversation, so I’m going to try to keep this to a half hour. I doubt that. But let’s, let’s see what we can do. So we’ll take a look at the news. Here’s what we do. Charles and I have each selected a story to discuss and I’m going to kick us off. A story I picked is from the New England Journal of Medicine. Catalyst has time come for a unique patient identifier for the u s so it’s a 30 minute show. So I’m going to, uh, not sure all the credentials, but a couple of couple of names of note. Who are the authors of this? Uh, our Pete Sued David Bates, John Halamka, Cio Beth Israel deaconess, David Bates, Cio, Brigham Women’s hospital and Aziz shake, who’s a professor in the University of Edinburgh.
Bill Russell: 08:38 Uh, here’s a little synopsis of this story. It’s time to revisit Congress’s fears about the unique patient identifiers and institute a system that will enable more complete and accurate patient records. So a unique patient identifier was proposed as part of Hipaa, but with shutdown for privacy concerns, the primary reason, their argument is that a national unique identifier, leads to better care. So here’s another quote. When accurate information is attached to the right patient, data access as timely or inappropriate care reduced and health information exchange becomes easier with an organizations as well as between. So, and they also go on to talk about how some states have already implemented this, uh, the state of Nevada and Minnesota and they say, you know, we can see how those go and scale them up. They close with this. So with billions of dollars having been spent on Ehr implementations, the healthcare system must vigorously investigate more efficient ways to connect fragmented patient data and effort that is increasingly relevant to the, as the u s moves from fee for service to value based care.
Bill Russell: 09:40 So, uh, so Charles. I’m going to go on a little bit of a rant here because I think this, I think they accurately capture the perspective of a physician, but I’m not sure that’s the right lens to be looking at this. So, uh, you know, their argument is let’s create a longitudinal patient record, uh, so that we have all the information at the point of care. Great. No one’s going to argue with that. I think you and I would both agree that, uh, truly, uh, complete longitudinal patient record would improve care. But here’s where my path sort of diverges with where they’re coming from. And I believe we should put the, the, the medical record in the hands of the patient, not the health system. Uh, if you really want to change healthcare, we have to free the data and putting the patient at the center of the equation instead of the health system or, or Pharma or payers or, or the EHR providers or even researchers is really going to do that.
Bill Russell: 10:34 So when I throw that out, I typically get three, uh, three kind of push backs. There’s, there’s the argument against giving the patient data that Judy Faulkner was, uh, was caught espousing, but she’s not the only one. There’s plenty of physicians who have said that and they essentially say, you know what, the patients wouldn’t know what to do with the data if we gave it to them. And, uh, you know, there’s just a certain level of arrogance that goes along with making that statement. Uh, I may say that to, to my five year old, I don’t have a five year old, but I may say it to a five year old, uh, but never really to a grown adult. Uh, another argument is that it will expose data to theft and uh, you know, that sort of has a level of hypocrisy to it. Cause in 2017, uh, I’m just going to give you some stats real quick.
Bill Russell: 11:21 So in 2017, healthcare had 477 breaches and 5.7, 5.6 million records were lost. And that followed a 2016 that saw 450 breaches in 27.3 million records lost. And the article I actually pulled those from, said we’re making progress because we went from 450 breaches, uh, in 2016 to 477. So we’re slowing down the rate 27.3 million records. I, you know, it’s, uh, it’s, it’s crazy. Uh, you know, I have a stack of identity offers on my desk from various health providers and, and seriously, my credit card has never been stolen from apple. Uh, we’ve seen models from really smart people, uh, who, who show that this is a bible off option, like the, uh, the blue button initiative from pod park. And I know I’m ranting here, I’ll get to in a second that really puts me over the edge on this is, you know, you just can’t have it.
Bill Russell: 12:16 And I know that Hipaa says that we can get our medical record and in most cases, you know, we’ll get it in paper or worse, we’ll get it as unstructured data. And, and the reason that’s worse is they don’t even do us the courtesy of putting it on paper and paying for the ink and the toner. Uh, they make us do it because the next healthcare provider, we have to figure out a way to get it in whatever form they give it to us, to them. And you know, we’ve talked about this, it’s not like we can’t share discrete data elements. We’ve had the technology since the 90s, and the health systems either choose to prioritize, not prioritize data sharing or they don’t have the appropriate skills or they don’t have the right incentives to get this done. Um, so, uh, you know, when I’d rather see here, I know I’m ranting on what I don’t like about it.
Bill Russell: 13:06 What I’d rather see is sort of a change in our thinking of, of uh, uh, patient centric approach, which says let’s get, let’s get the medical record in the hands of the consumer. So epic and cerner suspend your fees for developers and implementers and, and allow that data to flow out into the, uh, into devices that can actually be mobile with the consumer because the consumer is the only constant at the point of care. I’d like to see us move from, uh, from HL seven to Apis. I’d like to see a new model where we have, um, maybe a whole person profile, you know, or a health record or fitness or food or purchasing information so that people, data scientists like yourself can really do some things with it, but also that the consumer can benefit, right? So the consumer is consumer can say, I want to participate in this, uh, in this study or, or quite frankly, they can sell the information. A lot of health systems do, uh, end up selling the information either directly or indirectly through, through third parties. That’s a, that’s a long rant. But, um, and I know it’s hard to follow rant, so let’s change this up a little bit. You’re a data scientist and, uh, talk to me about how the, the patient identifier would make your life easier as a data scientist or what would you be able to do if the federal government mandated an identifier that perhaps you can’t do today?
Charles Boicey: 14:32 Couple of things. First I’m going to agree with what you had just said. Even though I hate to, we didn’t have any, we didn’t have any pre discussions on us, but, um, you know, I, you know, Apple’s, Apple’s going to leave the way we know the odds are doing it right now. They’ve been working on for quite some time. Um, the, the data absolutely. In the hands of the patient, we have technologies that allow for that, you know, blockchain being being one of them that will let the patient decide. And then in terms of potential emergencies and so forth, is there a need for a, a, you know, an identifier for data science? No, but I used to think, you know, kind of stepping back, is there a need for regional health information exchange? There actually could be a national health information exchange that we could actually do in surrogate apply Empi to, to those patients based on a whole bunch of characteristics that, you know, everybody, on the line understands how that’s done.
Charles Boicey: 15:31 Does it have to be mandated? Um, it can be done. It can be done as a surrogate. Um, so I think that, um, you know, I think you’re right too. Um, I like to see us get away from regional exchanges and then, uh, regarding, uh, you know, the data science and, you know, profiling and whatnot. So work that we’re doing at the University California who aligned with the institute of future health is exactly what you described. We’re building what we call persona goals. So that basically is your, is your profile and it’s unique not because it’s unique with a number attached to it because it’s unique in all its characteristics. That is you and that does it. Your profile is much different than anybody else’s profile. And it’s a combination of, of physiology. It’s a combination of, uh, you know, what labs have been attributed to you, your, your eating patterns or your exercise patterns. There’s a whole bunch of ways that we can identify you as you without, you know, imposing a national identifier to, so from the data science aspect, um, yeah, absolutely. Sure it would make it easier. But, um, you know, data scientists are supposed to work around issues like that. So that’s kind of, you know, how I would approach that.
Bill Russell: 16:49 That’s interesting. It sort of flies in the face of that we’re going distributed with blockchain over, everyone almost agrees, you know, over the next five years we’ll go to this distributed ledgers and records and whatnot. And this sort of flies in the face that says, hey, let’s keep it centralized, lets you know there’s a creative an index that we can utilize and what not,
Charles Boicey: 17:08 no, this is consumerism right, you know, consumerism will continue to, you know, eat into healthcare I think for the better. Um, and you know, we’ll have to do it, uh, you know, the consumers consumer ones.
Bill Russell: 17:21 Absolutely. All right, so let’s, let’s kick to the second story here. And uh, this is your story, so take it away.
Charles Boicey: 17:28 Okay. Um, this is on the topic of Ai and I think we’ll be able to get a little bit controversial here.
Bill Russell: 17:34 I’m going to bring it that last segment
Charles Boicey: 17:39 a little bit more. So, um, this, um, this article is by Michael Art, uh, January 30th and health data and AI is disrupting clinical practice. So health check is implemented matters and absolutely. And um, I’ll kind of get into it, but, um, I’ll go with a story really quick. So back in the late eighties or early nineties, at, uh, at La County USC, I’m, you know, I’m also a trauma nurse and whatnot. We did a lot of predictive models and so forth. I worked with a doctor, William Shoemaker who started the society of critical care medicine. We actually built predictive models that, uh, uh, for patients in, in trauma that would predict depending on the therapy, what the outcome would be. Uh, we made a really big mistake back then. We call it prescriptive analytics. The clinicians went nuts. They didn’t want the machine telling them what to do. And this is really what this article is all about. And this is, you know, this is almost, oh my gosh, this is almost 30 years later, right? So, um, again, with AI we can build out, you know, beautiful models, um, that I would like to say can assist, I have to call it intelligence assist. To be honest with you. I don’t like the idea of using this technology to tell somebody what to do. Um, I’d rather produce a cognitive trigger and this is what is described in the article. If I can give you a heads up that something’s going on that you may not have been aware, aware of, that’s fantastic, then you can, you know, make a clinical decision and move forward. But, um, so here’s, here’s a couple of the quotes
Charles Boicey: 19:19 and it has to do with skepticism and that’s what we, that’s what he encountered back then. You know, still plenty of physicians and clinicians have skepticism to spare if not outright hostility. A couple of other ones is really, it’s interesting. You really can’t force these issues. Now if you come up with these great models, whatnot, you really can’t cram it down anybody’s throat. You can’t say surprise, you know, here’s what the diagnosis is. Um, you know, somebody says, slam their fist down and say, Hey,
Charles Boicey: 19:50 forget you. I’m going to go back to what I’ve been doing for the last 30 years. So, um, there’s an adoption. So how do you get, how do you get that adoption? And really this article really points out that there needs to be an adoption and they’re really the way you do it, is you don’t blackbox any of this stuff. many folks out there have, you know, their models that are their proprietary models and you know, this, that, and the other thing. You’ve got to show how you’ve got to, you know, how you got to where you got, what data elements you used. Um, what ways you attributed to them. You know, what was a neural network that was employed, was a random forest. What did you go all the way through the process to get to the point that you’re at now? And how accurate is it? do you have an roc curve to show? Um, you know, the, the, you know, the various matrices so forth. What’s the precision? What’s the recall? You have to be able to demonstrate that and you’d have to be able to demonstrate with their data.
Charles Boicey: 20:45 And um, you really can’t make statements like, you know, uh, you know, this model will work everywhere because they won’t, they’re very geospecific well, it works in southern California. Isn’t going to work as well in Sarasota, Florida. It’s going to need some tweaking because of the demographic nature and even some of the external factors. So I think, you know, Mike put a, put a really nice package together, you know, saying yes, you know, AI is important but we need to kind of go through it in the adoptive kind of way and not just, um, you know, kind of throw it out there and you know, where does it, where does it fit in best, you know, don’t try to make a problem and then solve it. You know, it’s trying to solve problems that are already out there.
Bill Russell: 21:30 This is now the job of the informatics leader or the data science leader or chief information officer. It’s cultural change. And um, it’s really interesting cause you know, you sort of mentioned, so let’s, first of all, let’s give this really, it’s an article from a healthcare it news. So it is a plug for the machine learning and Ai Sessions at himss in Vegas in a couple of weeks. March 5th, uh, the, uh, project manager, they quote the bunch, and this is Jeff acts as project manager and system analysts in the it department of the hospital for special special care in new Britain, Connecticut. And, uh, he does say, you know, if you go into a department and say, surprise, this is the diagnosis from a machine and you’re just, I mean, you’re just going to, it is a visceral reaction, but I think that’s also why someone like yourself has been successful.
Bill Russell: 22:29 You know, you have that, that clinical background and, and, and being in the Er and, and really understanding how, uh, how these things play out and how the technology is adopted. We almost need more clinicians to get into this space so that they can, they can help people to make those transitions of, of saying, you know, hey, I understand that, that the AI model isn’t perfect, but neither is a human. A human isn’t perfect either. Right? And so if the two can figure out how to help each other, and as you say, you know, those, those cognitive, uh, the triggers that, that help both become better, that the clinicians are training the AI to be better in the AI is helping the clinician who’s busy running from patient to patient to see something that maybe they didn’t see. And, and that, that transition is going to be interesting. Uh, so anything else you want to say about this article that we’re going to jump right back into AI in the next segment, but anything else you want to say about this article?
Charles Boicey: 23:31 No, that’s fine. And then we’ll continue. We’ll continue. And I’d like to kind of bring in, um, you know, my students take on it as well in the next segment.
Bill Russell: 23:39 Oh, that’d be interesting. All right. So, uh, the second segment, we typically talk about leadership tech talk and clearly we’re going to jump into Ai. And, um, so, uh, you know, give us a couple more use cases around AI in healthcare. What do you, well actually, let me, let me step back. Uh, are you doing work outside of the u s with Ai at this point?
Charles Boicey: 24:03 Yup. So we’re working in the, in the UK was in the, in the mental health arena. So, uh, in, in the UK, the number one cause of death for males under 50 is suicide. And, uh, we’re taking a little bit different approach and the concept is to identify those at risk. Uh, but you know, we’re never going to be able to, and I don’t believe we’ll be able to actually determine where lightning is going to strike. We’re going to take, uh, you know, we’ll let you know where the thunderstorms are and then those that might be affected by those thunder storms. And you know, you can, you know, take the necessary action. So the idea is to really understand some of the factors that are involved in and you know, somebody, you know, making a choice like that. So, um, this is where, you know, big data and, um, and you know, the data science comes into play because we got to bring in social media for that.
Charles Boicey: 25:00 We’ve got to bring in the, you know, the temperature patterns we’ve got to bring in past, you know, suicide patterns. We’ve got, uh, uh, with consent, bring in, you know, the various patients that, you know, social, social feeds and whatnot. So you think about bringing all that in and then letting the folks that are following, you know, those patients, you know, giving them a heads up saying, Hey, these are the folks that are, you know, you know, susceptible at any particular point in time and that really changes, you know, as the days go by. So enough information so that they can, you know, reach out and, you know, make sure folks are okay and whatnot. Um, that is a little bit different approach.
Bill Russell: 25:35 That’s fascinating. Actually, pretty relevant given the, uh, the Florida school shooting that and I actually wrote this down and you know, lightning and thunder storms, we’re probably not going to be able to protect that this student at this time, is going to go into this school and do this, this action. But, but the whole idea of thunderstorms, there’s enough activity going on that you might want to take shelter or you might want to look into something. So data science is, isn’t that the point of saying, you know, it’s this person at this time, but it is at the point of saying, hey, there’s there, there’s a storm forming over here. We might want to, might want to get in front of that. Is there a difference in the UK versus the US in terms of adoption? I mean, are they more prone in the UK or are we more prone in the u s to be adopting AI type models?
Charles Boicey: 26:27 I think it’s pretty much the same meets with the initial skepticism, which is important because it keeps us on our toes. Uh, so it’s not, you know, tell me, show me, prove it to me.
Bill Russell: 26:42 So real, give me three AI healthcare models in healthcare that, that, that you’ve seen that are, that are effective. So you’re, that you gave us a mental health. Give us, give us a couple more real quick.
Charles Boicey: 26:54 Sure. I think, um, uh, we did mental health, the patient deterioration that I talked about earlier. Patients that are likely to, you know, crash, uh, patients that are likely to enter a sepsis pathway. So treatment can be, you know, be gone earlier. Um, outside of that, we are looking at, you know, deep learning, machine learning, finding those, those patients that are likely to be, um, uh, and we’re doing this for a screen in particular, um, that are likely to have an opiate problem. So if you think about all the different data points that you can bring in and do those triangulations and whatnot, you know, identifying, I think even from a data science, not necessarily AI but it data science, you know, finding those, those patients within our populations that are pre, pre diabetic or hypertensive, yet undiagnosed. Um, the other UK projects is, you know, identifying patients that are, have a fib that are not being treated. They’re at risk for stroke. So there’s a lot we can do with the data that we have. You know, you know, initially I think that’s how, you know, data science makes its initial wins. Um, and yes,
Bill Russell: 28:03 so I mean, so you have, you have Watson then and we’ve seen some, uh, some crash and burn type scenarios with, with Watson specifically MD Anderson was a, was a crash and burn kind of thing. And it, when we talked about it, when I’ve talked about it with others, they said, you know, it was the quality of the data. We couldn’t get the data to the point of actually being able to do the things we wanted to do. It was that because there is a data quality problem or is that because we’re not choosing the right use cases?
Charles Boicey: 28:30 Sure. So Ibm, there’s some wonderful things in AI watson it in particular was, uh, was a system developed to answer questions.
Bill Russell: 28:40 Yeah.
Charles Boicey: 28:41 Yeah. I’ll keep it. I’ll keep it as simple as that. So you can ask you a question if the, if the answer is within the confines of, of Watson, if it’s not there, then there’s no response. Yeah, there’s a lot more, there’s a lot more. Again, this is where, um, I, I really think that it’s really important that we think about how these technologies can assist us in intelligence.
Bill Russell: 29:09 So what, what’s, give us an idea of some of the, some of the good data sources that you’re utilizing. I assume, you know, bedside data is pretty consistent, right? It’s, it’s sending you the, uh, I mean, that data is very consistent and, and you’re consuming probably as much of that bedside data as you possibly can. Are there other sources of high quality data that you’re utilizing?
Charles Boicey: 29:33 Yeah, sure. So, um, anything off of physiological monitor, ventilators, smart pumps, anesthesia machines, they all are accurate, pretty accurate. And so there are accurate, like you stated, um, you know, every now and then you’ll get a weird signal. Somebody will stop talking, they get a CVP of 13, it jumps up to 310 or an arterial line is, is included. And you know, that jumps out. Those can be taken care of within, within our system. Um, you know, laboratory data coming in, pathology, uh, all of the universe source system, you know, you know, very plain. Um, especially, you know, if it’s been ontologically right size, but we can make, we can, we can fix that. But it really is, you know, what, what day do you need to do whatever it is you’re going to do with it. Um, if you try to chase 100%, you’re going to be in trouble. And within the realm of data science we can make, we can make, we can help out if things aren’t totally perfect. There’s things that we can do. There’s methods to account for missing data or data that’s outlier or out of range or not expected
Bill Russell: 30:42 And at the risk of going a little bit over on this episode. How does it, how does the health system get started with their AI
Charles Boicey: 30:49 sure, sure. So I’m a commitment to do the education process to really understand what it is you want to do before you jumped into it. I see healthcare organizations that hire data science teams, they really don’t know what a data scientist is. And I’m going to give you my definition. I say I participate in data science. I do not call myself a data scientist. And I’ll tell you why. The, the rigor of a phd program prepares you for the rigor that’s required for data science. It’s absolutely essential because you can get sloppy, you can get lazy and you can jump off the, you know, the long track and you can really lead an organization, you know, down the wrong path. Um, uh, folks like myself are very assistive to a data scientist. Data scientist on my team are phd folks. I work with them, we get to where we need to go. Um, so I think that before you go out there and hire the team, make sure it is, you understand what your, hiring, what you want them for and then maybe bring in some folks to help you, you know, put that team together. Because if you don’t know what data science is really all about, well what a data scientist is, you can be six months down the road and you go, oh, and I lost six months. So that’s kind of the spend as much time as you cannot on education.
Bill Russell: 32:09 Yeah, I made a, I made some mistakes there. I brought in some data scientists and they were immediately gobbled up in the and push back down into the day to day
Charles Boicey: 32:20 Yup.
Bill Russell: 32:23 Uh, it was, it was an awful waste. It took me about six months to, to extract them out of some of those projects and get them to focus on it a little higher level things. So,
Charles Boicey: 32:36 and bill and from an interview perspective, just little tidbit, if you’re, if you’re, um, your interview in a data scientist and you don’t understand what the heck they’re saying, then you probably are interviewing a data scientist.
Bill Russell: 32:54 The distinction I found was they don’t answer questions. They tell you which questions you should be asking. And it was correct. They, they look at the data and the data informs them and it’s amazing the number of tools they have good ones have in there in their bag. Uh, so it’s time to close the show. Favorite social media posts for the week. I’ll start it off. This is from William Walders. Um, and, uh, it’s, it’s a comic strip and it has a gentleman sitting across from a receptionist and the receptionist saying to him, you cannot list your iPhone as your primary care physician. So Charles, to you, what’s your favorite social media posts?
Charles Boicey: 33:36 Sure. My social media posts today comes from a colleague of mine, Brian, Doris, because we’ve worked together for years. Um, his Twitter handle is geek_nurse and he put out this week fellow nurses, we need to elevate our seat at the machine learning AI tables, bring that clinical digital act, you know, to bear. I think we need a nursing coalition of nursing data scientists, folks driving the digital age with a shout out to myself and Judy Murphy at IBM.
Bill Russell: 34:08 Awesome. So, um, that’s, I assume you’re going to be at HIMSS, you’re going to be at HIMSS.
Charles Boicey: 34:14 Oh, I’ll definitely be there as well as all my students will be there. And they kind of shouting part, they cited on it. Mostly millennials, they sided on the side of Ai are, you know, the, the system’s telling people what to do versus the assistance assist the people. I just wanted to get that out to them.
Bill Russell: 34:33 Yeah, I tried to, but I trust the machine more than I trust the person had a oh wait, could do that. I would love to, I’m gonna be at himss as well. And uh, I would love to catch up with you and your students. That would be a, a lot of fun. So, uh, that’s all for now. You can follow Charles @N2informaticsRN on Twitter and me @thepatientsCio. And don’t forget to follow the show on Twitter as well @thisweekinHit and check out our new website. And Charles, where are you calling? You’re calling in from Jacksonville?
Charles Boicey: 35:05 Yeah, call me from Jacksonville, Florida. Clearsense at clearsense.com.
Bill Russell: 35:10 All right, so I will, uh, one of the things we do on the, uh, on the website you’ll notice is, uh, the image is usually a skyline of where the guest is calling in from. So we will have a skyline of Jacksonville. Jacksonville has a skyline, I assume?
Charles Boicey: 35:26 Yeah.
Bill Russell: 35:29 Or maybe just a beach shot. That’s probably a good way to go. Yeah. Yeah. Well, if you like the show, please take a few seconds, give us a review on iTunes and Google play a, that’s all for now. Please come back every Friday, uh, where we will do this again with another great thought leader does the next week, Dr. David Bensema. It’s going to be here two weeks. Uh, David Baker, CIO for Pacific dental. I think that’ll be an interesting conversation because that’s awesome. That’s all for now.