This Week Health

Charles Boicey on How to become a Clinical Data Scientist

Charles Boicey - MS, RN-BC joins us to discuss the state of Data Science in Healthcare and the path to becoming a clinical data scientist. We also go in depth into the process of data science and how to ensure you aren't spinning your wheels. Hope you enjoy.

Transcript

This transcription is provided by artificial intelligence. We believe in technology but understand that even the smartest robots can sometimes get speech recognition wrong.

 Welcome to this Week in Health It where we discuss the news, information and emerging thought with leaders from across the healthcare industry. My name is Bill Russell, recovering Healthcare, c I o, and creator of this week in Health. It a set of podcasts and videos. Dedicated to training the next generation of health IT leaders.

Today, healthcare needs more data scientists. We take a look at the path to becoming a clinical data scientist. This podcast is brought to you by health lyrics, looking for someone to be your personal coach in healthcare technology. Let's talk visit health lyrics.com to schedule your free consultation.

Our guest today is Charles Boise. Uh, Ms. R n bbc, lots of acronyms by his name, chief Innovation Officer for Clearsense and Assistant Clinical Professor at Stonybrook University. Uh, good afternoon, Charles. Welcome to the show. Hey, bill, thanks for having me. Well, I, I love having you on the show. I, I, we don't do it enough 'cause you're all over, literally all over the world.

The last time I tried to have you on the show, you were in India, and, uh, you know, next week you're gonna be in Cleveland. You're, you're literally all over the world. It's, it's track. We ran into each other at HIMS and we started talking about this topic of, of data scientists and clinical data scientists, and I got, I got kind of excited about having you on to talk to, Uh, to talk to this specific subject because there's, there's so many doctors and nurses who are saying, Hey, I, I see the need.

I see the need to really turn data into, uh, into insights and into things that could be used to effectively to drive down the cost of healthcare, improve access and better experience, better outcomes, and, and they get excited about it. I thought it would be great to just have you and I talk about what does that path look like for them to go from practicing physician or practicing clinician, uh, to, to the data sciences role.

And, and that's what we're gonna try to do today. Sure. So, um, let's kind of start off with, um, you know, something you know, that, you know, probably people don't think about and, and they should, you know, really what is a, um, you know, what is a data scientist? What is that all about? And I'm not gonna get into, um, you know, what degrees, you know, what certifications.

I'm not gonna get into any of that at the end. We'll talk about some programs that are out there that can help you get to that point. But, um, um, I really want to, you know, kind of go over, you know, initially, you know, what is a data scientist, you know, what are the three domains, if you will, that constitute, um, you know, a data scientist, what do you need, et cetera.

And, um, and the three domains are, um, Subject matter expertise, uh, computer science, which they have the subject matter expertise, and then domain science in or domain, uh, expertise in computer science. And then what's the third? The third one is, is math. Um, stats, mathematics skills, um, critical thinking skills.

Yeah. So we're, we're gonna dive into those, but, uh, here's the thing. I know about the two of us. We get excited and we get going, so we needed sort of a framework to keep us from going off of the, we usually do Absolute, absolutely. Uh, the first thing I wanna do is, um, so there's, there's three things, three movements we're gonna do on the show.

First is current state of data science and what is data science? The second movement's gonna be a typical data science project in healthcare. And what the process looks like. 'cause I, I think some people think data science and they think the, the report writers or the people who are doing the reports.

And, and that's, um, we'll call that like the, the, the beginner, but that isn't really data science. And we'll talk about what a pro, uh, a project looks like. And I think that will open people's eyes to what's possible. And then the third movement's gonna be the path of becoming a. Uh, a clinical data scientist and, and we'll talk about the different programs and whatnot.

Sounds good. So let's start with current state of data science and healthcare. So you're doing a lot of things. What is, I, I don't even to word how to frame this question other than to say broad question. Current state, what is it? Uh, current state of data science is, we're jumping into it and we're doing as much as we can, as fast as we can.

Um, You know, basically we've, we were jump, we jumped on the horse and we're, we're, we're charging off widely in all directions so people recognize the need for it. They don't know what direction to point it in yet. Is that, is is or or are they finding ways to, to do it and they're sort of experimenting in a lot of different areas?

Well, just like anything, you know, even bringing in, you know, data visualization and analytics as we did, um, you know, there's some that are at a higher maturity level than others. And, um, what's really good is it's been recognized, um, it's legitimized. Uh, many organizations have legitimized, you know, data science in the application of, um, you know, data science project products.

So that's, that's the good news. Um, the, the, the not so good news is it's still the wild, wild west. Um, there's still lot, a lot of organizations that, um, are hiring folks, um, really without a problem to solve. I want to get into data science. I want to do predictive analytics. Let's hire a team. Um, it's kind of like the religious movement of data governance a few years back.

And we all know what, we all know what that looked like. Yeah. And you're, you're like, smacking me around here. 'cause I, I am one of those people who hired three data scientists. I put 'em in a, in a group and I said, all right, you know, we're gonna, we're gonna, and they were like, uh, we know what to do. And I checked back on 'em six months later and they were writing reports and they were buried in.

Different places of the organization and I realized, uh, you, you really do have to lead this. And so what does it look like to, what does a. A leading health system look like that is applying data science today. Sure. So that's, that is the key word that you said was applied. They understand that there's a need to a apply, uh, that, which is, you know, the products that are, you know, developed in a, in a Data science pro program.

They actually have. Um, and it's, they actually have a program in place. You remember when we started with the 30 day readmissions? Can you predict 30 day readmits? Um, the first question that I, I asked when I was asked that, I said, yeah, absolutely, but do you have a program in place to address those patients that we discover that are likely to readmit?

Because if you don't, we've got a problem. We've got an ethical problem. We've identified people that we're not gonna be able to take care of. And then, you know, secondary to that, we've just performed a statistical exercise. I'll predict 'em, we'll see if they readmit, and we'll just keep some stats. So you really, really have to think this thing through.

Um, and you have to have a plan in place, and you have to have, you know, objectives, you know, clear objectives, you know, why are you doing this? Not just because, um, you know, my healthcare organization neighbor down the street now has a data scientist. Um, I want one too. Yeah, I mean, you make a good point there.

So, um, I, I was talking to a C I O and we were talking about in the consumer area, how the chief digital officer, the c i o, the ops, the c o o and the chief strategy officer, need to almost be tied at the hip because one begets another, begets another. I mean, they're also, um, so closely linked. But in the data science world, the same thing's true in terms of operations.

Everything you're gonna be analyzing is gonna kick off something. You're gonna find something out, and it's like all. Now what do we do? We know the people that are gonna readmit. We gotta put a program together. Yeah. So it covers operations, it covers clinical practice, it covers finance. And if the organization happens to be tied to, to research, you have a a research component of of it as well.

And you're absolutely right, they all have to be, um, you know, tied at the hip. So, where does this typically report? I've, I've heard this, I've, I've, now, I, I do a fair number, fair amount of consulting still. And I, and we had this problem where I was, the C I O, uh, analytics was everywhere, right? And it should be everywhere.

It should be close to where it can be applied. But from a reporting standpoint, it literally was everywhere. And we, uh, we went through the organization and looked at everybody with a, a title of analyst. Which is typically how they hide people within a, a budget. Uh, and we found a hundred and 180 of 'em or something like that, didn't report into it.

They were just data analysts of some kind, and they were all over the place. Um, you know, what does a, what does a good program look like? Where does this report into, uh, you know, what, what's the matrix? What's the dotted line? And then, and then how do they ensure quality throughout the organization? Sure.

Um, so an organization that's mature that, um, you know, gets it, that, you know, is, is where they, where, where they need to be. Um, I usually see these teams, you know, reporting through, um, the, the chief data officer, to be honest with you, or the chief analytics officer. Uh, and these are. You know, if they're not at the end of maturity, you know, somewhere in a midpoint, um, they have their, their data needs taken care of.

And what I mean by that, there is a central, um, repository or platform, if you will, that these types of activities, um, can occur in. And, you know, it's, it. It's really interesting, bill. 'cause these types of organizations also have decentralized, you know, analytics and whatnot. That's great to have a department level, but they're all working off the same data sources and whatnot because they've gotten to the point where it is absolutely centralized.

And, and I say that because, um, Data scientists, if they're spending 80% of their time data wrangling, you're not using them correctly, you're not using them wisely. They should spend their time, um, actually, you know, applying data science to the data because the environment is such that, um, they can do that.

But usually the, um, if there is not a chief date officer, um, this is something that the, as you know, you want to, the c i o needs to keep, uh, keep tabs on because it can, you know, go, go out of control pretty quick. Yeah, that's, uh, that's one of the conclusions we came to is that the, uh, it, it function was gonna be the data wrangling, it was gonna be the interoperability, pulling all the data together and providing the tools and, uh, helping to establish the dictionary.

Uh, you know, what did the, what's, what does the terminology mean? Um, because that's the other big problem you end up with, is people generating, uh, insights and it's based off of data that may or may not have the definition they thought it. Yeah, sure. So, so the ideal state for a, you know, a data science team is absolutely to have that governance in place, to have that catalog in place, to have things such as length of stay, um, defined for the entire organization, and that could be multiple hospital entities and whatnot.

So you're, you're absolutely correct. So let's talk about a, a typical data science project. Um, so, uh, so I mean, you, you used readmissions. I don't know if you wanna use another one, but No. Yeah, I'll use, I'll use another one that's, um, um, that I think is, you know, is an important one. Um, let's look at, um, we'll use two of 'em.

We can talk about two 'cause some ofs are, Let's look at it in a, in a couple of levels. So let's say, let's say somebody came to me and said, Hey, hey, Charles. Um, we'd like you to develop a, uh, predictive model for predicting patients that are likely to become hypertensive, or let's say likely to become type two diabetics.

That's, you know, that's, that's reasonable. Um, but I would push back and say,

Here first, let's find those patients within our population that are undiagnosed diabetics or undiagnosed hypertensive patients. We all have them. So that would be step one for me. Step two would be the be the predictive model, because you know, I, as a clinician, I wanna help as many people as I can, you know, uh, Quicker.

And that's one of the benefits of having a clinician data scientist because they got the domain, they totally understand the domain and they would come to, you know, something, you know, similar to that. So, um, so, so what do we do? What does that look like? Um, you know, really first step is what I call, you know, business understanding, if you will, really what are the, um, uh, defined objectives, uh, what are the data sources that we need and.

Let's do some research. Um, I'm not gonna be the first one to build a model for, for those two, those two items. I'm sure a lot of work has been done prior to me. So let's find out the work that's been done. Um, let's, um, do the research and then let's start, um, you know, bringing in the, the data that's required for that.

Um, so let's bring in the data, explore the data, get the data quality, um, you know, work all the, the kink out there. At that point, we've got what we wanna do. We've got the data that we need to do it in, and now we do what's called, uh, uh, feature selection. All right, so in the business understanding, so you have, uh, you're gonna define the objectives.

You're gonna, you're gonna identify the data sources, and then you're gonna do, uh, some research. Where would you go for the research? I mean, do you go to. Sure this is PubMed for our purposes being in healthcare. Yeah. Um, this is peer reviewed, um, peer reviewed, um, you know, journals and. It's really important.

Um, uh, you know, Google search in this case usually doesn't cut it, although you might, you know, you might find a few things that are, you know, valid. But again, you know, PubMed and, you know, some of the peer reviewed journals and whatnot, just to kind of get a flavor of what's gone on before you, and be honest with you, it actually kickstarts, um, and get you thinking.

Um, you know, a lot of data science is really about, you know, you know, having that, you know, Self-talk or if you will, you know, that cognitive exercise of, you know, what we're really trying to do and whatnot. Um, and having, that's why a clinician's doing this, um, Have the, the discipline and understanding of certain features that are likely, I'm gonna put some likely, uh, emphasize that, um, that'll be a benefit in a model like this.

But there's also other data elements out there that we need to look at. Um, you know, such as social determinants, um, exposome data, um, other, other data elements within the, you know, the, the healthcare data set, if you will. Play a role in that. So, um, we're now doing the modeling. We've, we've, we're doing feature selection and I may, I'm oversimplifying this bill 'cause you know, this is, yeah, well you almost have to, I mean, 'cause at, at this point, you've, you've done your research, but then you have to get the data.

Now in a, in a, in a mature organization, that data is gonna be, Just available to you either in a, in A E D W or, uh, or accessible through a, a big data, uh, platform or set of APIs. Um, but generally there's an awful lot of work that goes on there of, of wrangling that data and. And cleaning it up, I would assume.

Yes. So in, in that case, um, again, remember I told you you don't want your data science, your data science guide, um, or gal to be, um, you know, doing data wrangling, right? Um, so that's where a, you know, a data engineer or a data wrangler would come into play. Um, Data scientist says, Hey, this is what I need.

Data Wrangler goes up and, you know, wrangles it up, if you will, um, and delivers it to the, the data scientists so that they, um, you know, have a, you know, have a data set that they can start working with. So they'll send a, uh, spend a lot of time in business understanding, hopefully very little in the data, uh, ingestion and acquisition.

And then they spend a, i, I, I assume the data scientist spends a fair amount of time in the modeling. Uh, yes. And, and that's, that's where they're earning their, their keep really, I would think. Yeah. That's their training. That's their, you know, that's their discipline. That's, that's that's their gig. That's what they do.

And Absolutely. Do they, do they, do you start, I mean, this is one of the things that somebody had said to me is, do you start with the, the hypothesis or do you start with the data? The data sort of tells a story. I am the Rio approach, which is let the data speak for itself. Um, others have the other approach.

Neither one is wrong, but one should consider, um, you know, the other, um, let's say a di, a diabetic model. I'd be foolish if I. Left out A one Cs right. That would make absolutely zero sense to me. Um, or B M I A B M I calculation. Um, um, so there are some features that we know that are gonna play a, a, you know, a role, but let's look at the, the rest of the data and see what comes up with it.

And tools such as, you know, linear regression and and so forth, will, um, you know, help us understand that the benefit that we have in healthcare is we have . Diabetic type two patients. We have, um, type, you know, we have hypertensive patients. Um, and if we get into, you know, predictive models for patients that are likely to crash, we have all of that.

So, um, you know, understanding those patients, those data sets, those features and whatnot are of an absolute benefit, um, for the, for the design. And then considering, you know, features outside social determinants, um, Big personally in, um, inaccuracy by using certain social determinants of health within these predictive models, especially for, um, you know, patients likely to become diabetic and hypertensive and so forth.

So you gotta kind of take it from, you know, All, all aspects. Um, and then you've gotta have the discipline. Hey, if I come up with a, a model that's, you know, 0.5, well that's a flip of a coin that's really not gonna, you know, do me any good. This is where data science in the training and the rigor around it really comes into play.

You have to be disciplined to say, okay, this model . Isn't working, it doesn't have the accuracy that I'm looking for. I gotta go onto something else, right? Um, and I'll experiment. I'll do what I need to do. I'll maybe do an ensemble model, but at the end of the day, if my accuracy is not such that, um, it, it's gonna work.

I've got, I've got a call at that point. Um, I'm not gonna go bend features and, you know, do manipulations and so forth so that I get a area under the curve that's, you know, in the high nineties, it is what it is. Um, maybe in some models, you know, Point seven five would, will, will be acceptable and others may be 0.88.

Uh, you know, 0.9, uh, will be acceptable, you know, and that's where the communication comes in, bill. Everybody's involved in one of these projects, um, has to be there for, for all of this. Um, this, none of this can be black box. They have to see the data sets, they have to see the feature selection, they have to see the math, they have to see the output and whatnot.

That's the only way you're gonna get buy-in as well. Alright, so just a little tangent, just because you said that, uh, we've been talking a lot about ai, uh, recently on the show and we've talked to different people about, about ai and one of the things is, you know, you have these vendors who come in and say, well, we have these algorithms and they're, they're sort of behind the curtain.

Um, and it sounds like what you're saying is there's no such thing as behind the curtain when you're talking about clinical, um, you know, efficacy within clinical environment. So you have to. They almost have to reveal some of that, some of the algorithms, how the math is done, and how they come up with their, uh, their findings.

Otherwise, it's gonna be really hard for us to, to just implement some of these things within the clinical environment. Yeah, absolutely. And you know, I'm gonna, I'm gonna tell you that, um, you know, a certain model in, uh, you know, in Irvine, California. May not be applicable in Sarasota, Florida in a healthcare organization that the predominance of patients is 65 and above.

Um, these models have to fit the, um, fit the population and that's kind of where we go, you know, awry kind of in healthcare where. You know, a lot of our best practices and models and so forth, were built on a population that, you know, may or may not still exist or not exist in, you know, in the, in the, you know, geolocation that we're gonna try to apply it.

Um, and this is not a one and done, you know, as we go through the process, yeah, we have to operationalize it, but it, there's a lot more than operationalizing. Uh, we gotta continue to test and validate. You can't just dump the back black box off at the door, you know, walk away from it. Um, you gotta be really monitoring that over time.

Not only is there, is it accurate, but for crying out loud, did it actually affect any change? Was anybody paying attention to it? So, um, these products not in a workflow, um, you know, there's gonna be some issues there, especially on, you know, the adoption side. So when it, when it goes into operationalize, when you're operationalizing it, the data scientist is still, uh, pretty, pretty heavily involved in terms of, uh, testing, validation, and keeping the, uh, making sure that, uh, that the model is working out the way that, that everyone expects it to work out.

I mean, their, their job is to sort of, Uh, act as the scorekeeper and say, Hey, you know, the model's still holding, uh, the results are good, and those kind of things. I, I mean, am I, am I seeing that right or am I saying that? Yeah, abs, absolutely. So let's take the, um, let's take the, you know, the, you know, diabetes or the, you know, hypertension, you know, did those patients that you know were identified in fact, you know, You know, become hypertensive or even, you know, looking at, you know, patient deterioration type models, you know, you know, every month looking at the rapid response team responses, you know, looking at the code blues and so forth.

And actually going back to, you know, the, you know, the algorithm and whatnot, were these scored. Did anybody take action on it? Did we miss anything? You know, what were our false positive? What were our false negatives? Um, yeah, this is not, this is something that you really have to, um, you know, pay a good bit of attention to, um, you know, ongoing, um, you know, especially in the, you know, the first stages of, of deployment and whatnot.

But you've gotta go back and, you know, you gotta go back and look at this stuff now. It, so does a data scientist handle like two or three of these kinds of projects at the same time? Or are they typically focused in on one? I. Sure. Um, so from a, my personal is I like to do one. Um, there are folks that can do, you know, two or more at a given time.

That's okay when you're, um, kind of doing the data wrangling and whatnot. But when you're really have to be concentrated and focused, uh, you know, for me, um, I really have to be into that one until I get to the, you know, the conclusion of I've, you know, got a model that I feel is, is fit and ready for deployment.

And you can justify that cost. And, and I, I've had this conversation with executives, you can justify that cost because, uh, you know, some of the outcomes dramatically improve outcomes or dramatically improve efficiencies to the point where you're, you're freeing up millions of dollars. Um, and you're keeping people, uh, healthy and I mean, there's the, there's significant benefits from, uh, from these projects generally.

There's a, there's significant benefit from the project. There's not single benefits from hiring individuals because if you don't have that complete program in place, then I don't care how many accurate models you've built, um, if they're not deployed properly, If they're not, you know, you get you, I won't, I won't reiterate it, but You get the picture.

Yeah, you get the picture. I mean the, the system, uh, the system has to have the ability to, uh, to act on the data once they see the data. Um, alright, so let's, let's take a look at, uh, let's take a look at becoming, uh, clinical data scientists. We had, uh, We've had a couple shows now where, where people have talked about the fact, um, you know, I mean, Lil Lee Milligan's been on the show and he's, uh, you know, he's a physician who became A C M I O, who's now become a C I O, uh, who now oversees their, their data practice.

And we're seeing a lot of clinicians move in this direction. You know, what does that look like? What does it look like for a clinician? Um, do they generally, the, the, I mean, for Lee, the, the path was, you know, essentially, uh, you know, a a somebody who really understood the data and understood the E M R and then they said, Hey, why don't you become the C M I O and help us to implement this?

He implemented it. And then he became, Hey, why don't we, we'll put you over analytics 'cause you seem to get analytics. And then he just kept learning and, and growing into the role. Do you see that, I mean, is that generally what has happened and do you see that transforming into a more, uh, disciplined path where people actually get degrees and, and move into it?

Sure. So a couple of things. One, um, that was an organic progression. That's, you know, my story as well, really. Um, you know, it happened organically. Um, um, when. When I talk to clinicians now, the first thing I ask 'em, I ask 'em a question. Are you running from something or are you running to something?

Because if you're running from something, this probably isn't gonna work for you. But if you're running to it because you really want to embrace it, then you know, let's keep talking. Uh, So, you know, kind of the pathway, the, the domain knowledge is there. Um, you can be a physician assistant, you can be a, uh, a nurse, a nurse practitioner, a pharmacist, uh, uh, dietician, physician, um, you clinician, you have that domain knowledge.

You know how a hospital works, you know, the ambulatory environment works. You absolutely have that. That's gold. That is, you know, that will get you a majority of the way there. A lot of the tools we use are, are develop, are based on, um, having some programming knowledge. So you've got Python, you've got R, there's a few others out there.

You have to be able to, you know, um, Amp those skills up. Um, and on the math side, we're all pretty well versed in that. If you think about, um, you know, clinicians', education pretty good with math, pretty good on the statistics side. Yes, there'll be a need to, to brush up on that, but I would say that, um, The computer science side is really the, uh, area that we need the most focus on.

There's some pretty good programs, bill, that you can audit within the Coursera environment. Uh, and that's where I would start. And the reason I say that to folks is before you start, you know, I. Spending, um, you know, tuition, dollars and so forth, um, take a look at, you know, what you're getting involved in.

Um, and so you can make a really good choice, you know, based on your interest and, and so forth. This isn't gonna be something that you can, you know, jump in. It's not gonna be easy. It's gonna take some discipline and, and rigor. Um, but you know, the transition is absolutely there and I'm not, I'm not gonna get into the MS versus PhD argument.

Okay. . Um, because that could be a whole nother show and, you know, I may not do a, the benefit of both. Um, I will say that those that have gone through a, a doctoral program have a pretty good, um, sense of rigor and that is beneficial in, you know, this kind of environment. Uh, the last thing you want are, are models that, um, were created to get to a certain goal as opposed to.

It is what it is. You know what I mean? Yep. Uh, it's interesting to me that you say, uh, R and Python. 'cause uh, you know, for people who feel like, you know, uh, programming languages change very rapidly and I'm gonna have to learn something new in three to five years. R and Python have been the sort of bedrock for this for.

As long as I can remember. I mean, yeah, in the last, in the last year and a half or so, um, Python has actually eclipsed r and for the, um, for the clinician, that's pretty, pretty cool because, um, Python is more of a language, natural language, if you will. Um, where r has a little bit more programming language to it.

There's a little bit more, you know, syntax that is, that's learned. Um, again, you know, I'm at, you know, Stony Brook University and I. I flipped my data science program to center around, um, a Python. Now I do teach a little bit of r but it's primarily Python driven. And the students have been, um, have had better success within the, uh, Python environment.

And my son's at one of the big four firms and they, uh, they required them to, uh, to learn Python because that's, I mean, it's, it, it's now becoming one of those things. If you're going through a computer science program at a college or university, I would assume you're gonna be picking up Python as a. As a language as opposed to when you and I would've gone through it, it was, uh, you know, it was c and it was basic and some other crazy things.

So, but there's data science programs springing up all over the place, and you were, um, why don't talk a little bit about those and then, and then you and, uh, Clearsense were involved in one. I think that was pretty interesting. Uh, give us an idea of what those programs look like. Sure. So you've got a complete master's program.

Uh, you've got a PhD route as well. Uh, many clinicians, you know, You know, at the, you know, BSS or you know, the MS or you know, PhD level. So, you know, pick your entry point. There's also some really good certificate programs. Um, I would stay with organizations that are academic medical centers. Uh, I'm not gonna name names, but you know, there, there are several.

Um, and they are the ones that are really, you want one that's focused in healthcare. Uh, you really don't want one that's focused outside of healthcare because, It doesn't really make any sense. You're gonna do a hell of a lot of, you know, adapting the, um, the curriculum to, to, to meet your needs. Um, I would definitely do that Coursera thing for, you know, audit purposes.

Um, and then make the decision whether you wanna, um, you know, do the certificate program, um, and you know, from that, you know, MS. Or PhD, uh, you know, Healthcare is employing data scientists, and there's two things that are happening. One, organizations are hiring folks not knowing what a data scientist really needs to do or what one really is, and then six months later they figured out that they find out that they didn't hire what they thought they hired.

So really, all that front work and putting the program in place is, is, you know, really, really essential. Um, and is. Hiring data scientists from domains outside of healthcare and I have, you just have to work with them and, you know, put a data wrangler or, you know, an informaticist alongside of 'em to work side by side.

'cause they can go off in weird tangents, if you will, because of not having the domain knowledge. So let's talk about that. So I'm, I'm gonna build that out. I'm gonna build the program out within my, uh, System. Am I better off taking some clinicians and putting them next to. Taking some clinicians and getting 'em the skills they need to start to play with the data.

Or am I better off, uh, bringing in data scientists and, and matching 'em with clinicians? I mean, what, which, which tends to work better? What do I see that's been the ideal? Um, you've got some really good clinician data wranglers. You've got some nurse informaticists on, on your team. Uh, many of us have, um, marrying them up with a data scientist from outside of healthcare will do two things.

That knowledge transfer of domain will absolutely occur. What's what's. I don't know, maybe more important, that spark of working with a data scientist and seeing what can be done, you know, with that data will, will absolutely occur. And that'll propel, you know, that clinician to, you know, further along their, you know, training and studies and whatnot.

That's what I've experienced. Um, and that's what I've seen that's worked, um, you know, worked really well. You know, I, uh, uh, I mean, there's a lot of exciting things in healthcare. I love the cloud because it brings new levels of efficiency. Uh, consumer trends are going to, um, you know, really help take the friction outta healthcare, and I think that's exciting.

IoT is gonna allow us to, uh, reach beyond our walls. But, uh, data clearly has the ability to transform healthcare if applied correctly in the right places. Uh, it, it, it's one of those, it's one of the things that's gonna affect outcomes costs. Quality access experience. It, it can, it can impact all, uh, all aspects.

And in order to do that effectively, uh, an organization, I, I just, I feel like, uh, and this is the message that we're really trying to get out there. An organization has to say, you know, we have to be, uh, an, you know, we could be a B in infrastructure, we can be a B in some other things, but we have to be an A in our use of data.

And, uh, And you, and, and what you're seeing out there is, uh, health systems are gonna start to differentiate themselves on the ability to use data. And you're starting to see some of those articles where, uh, where you know, the headline will read, you know, they, uh, impacted this, uh, outcome, or they impacted this, um, uh, these financial metrics or, or whatnot.

And, and you see. You can see some distance starting to generate or to separate these. And I think that the use of data much more so than some of these other skills, is really going to, to be, to be the driver. I, I mean, that's been my experience and I assume that's your experience as well. Yeah, and it, you know, it both from my operation financial perspective, it's survivability and sustainability.

You know, two letters, ma. You're gonna become an acquisition if you're not, um, if you don't have the skillset, and, you know, we can't fly operationally, financially by the seat of our, our pants anymore, it's, it's not working out well. Um, and you're subject to, you're subject to being acquired. Um, and on the clinical side, uh, this, you know, application of data science, you know, I look at it from a, um, an intelligent assist, you know, perspective.

Hey, look it, we're not telling anybody what to do. We're really assisting you, right? These are the patients or patients that we found in your population that are likely undiagnosed hypertensive patients and or, um, you know, diabetic patients. Hey, Let's take a look at them. Um, um, these are the ones that are likely to, these are the patients that are likely deteriorate.

These are patients that are likely to enter a sepsis pathway. Um, again, using this technology to, to help not to. Somebody up over the head or tell 'em what to do or to, um, you know, say you have to do this therapy, that therapy, um, really, but just kind of being their eyes and ears out there for them. And machines do a great job of that.

So, um, yeah, absolutely. Well, Charles, always great to have you on the show. Uh, is there a, an an easy way for people to follow you on social media? Sure. Um, you know, the, the Twitter handle is n to informatics rn. Um, and that, you know, that's, that pretty much covers it. You know, one of the other things I, I wanted to, to touch on, but, uh, we, we do have a bunch of college, uh, professors and people within academic institutions following the show.

Um, you know, those who are interested in, in developing this kind of program are, are, are, are, are those the kind of people that you generally have conversations with on what they would do to develop a program at their college or university? Yeah, and I've helped, um, I've helped several do it. Um, you know, Meharry Medical College, uh, just briefly, uh, we now have a program in place where they're, um, their, their medical students when they, when they graduate this July, they'll graduate with a certificate in data science.

Uh, the next. A couple of classes, ms. And then three years out we're looking at a, a PhD program. So, uh, basically physician data scientists and I have helped, um, you know, Georgetown, I've helped other, um, organizations with their curriculum and whatnot. I'd be more than happy to help, um, an organization with that.

And again, my stance on this is of an applied nature, um, a little bit of theory. But a lot of application, um, which is, you know, kind of my thing. I'd rather apply it than talk about, you know, what we could do. You know what I mean? Yep. And my, my two people to talk to whenever that comes up is you and Dale Sanders.

'cause I think you two absolutely. Very focused on, on making stuff happen, so. Yep. He's the guy. Absolutely. Absolutely. So I appreciate you coming on the show. This show is a production of this week in Health It. For more great content, check out the website at this week in health it.com or the YouTube channel at this week, health it.com/video.

Thanks for listening. That's all for now.

Contributors

Want to tune in on your favorite listening platform? Don't forget to subscribe!

Thank You to Our Show Sponsors

Our Shows

Keynote - This Week HealthSolution Showcase This Week Health
Newsday - This Week HealthToday in Health IT - This Week Health

Related Content

1 2 3 242
Transform Healthcare - One Connection at a Time

© Copyright 2023 Health Lyrics All rights reserved