Future of Global Informatics_Episode 17_Vipul Kashyap: this mp3 audio file was automatically transcribed by Sonix with the best speech-to-text algorithms. This transcript may contain errors.
TJ Southern:
Hey, y’all! Welcome to the Outcomes Rocket Network – The Future of Global Informatics Podcast, where we discuss global informatics through conversations with industry leaders and innovators so that you can understand what it is, what it does, and how it shapes the healthcare of our future. I am your host, TJ Southern.
TJ Southern:
Hey, y’all! Hey, y’all! Good day! Good day! And welcome to another episode of the Future of Global Informatics. It’s your girl, TJ Southern, and today we have Vipul. Oh, my God, so let me tell you guys right now, right? I always say at the top of my podcast, make sure that you have a pen and a piece of paper ready because our guest can drop nuggets at any time. And I promise you that he is one of those that will definitely drop us nuggets. He is definitely one of those that will drop us nuggets. I am so excited to talk to him. I was actually rummaging through his LinkedIn profile even before we talked and I was just like, oh my god, I cannot wait for this conversation. So I’m going to let him introduce himself. Vipul, tell the people about yourself, tell our listeners who you are.
Vupul Kashyap:
Hi, my name is Vipul Kashyap, and I’m currently the SVP of Clinical Informatics and Product Strategy at BUDDI.AI, a startup out of New York City. And we are using AI and machine learning to try to solve some of the most pressing healthcare problems. So healthcare is my passion, it, we … and how siloized we are across the healthcare ecosystem. You have your payers, your providers, your pharmaceuticals, your ACOs and CRO’s and god knows whatnot. And this is really, really I mean, amazing, for a field like healthcare, we should all be working together and collaborating with each other. And you know what the key for this is? The key to this is information and knowledge.
TJ Southern:
Yes.
Vupul Kashyap:
It is key to address these challenges, right? So yeah, acquisition of information, understanding the meaning and context, both from a strategic context, like how does this piece of information impact outcomes and cost versus the tactical context? Okay, what point in the clinical workflow is this information necessary and what value does it provide the actor and the workflow in the clinical workflow at that point, right? So that’s where I am. And I firmly believe that informatics is and will solve the problem. So I look forward to more feedback from you guys and see how we can all work together and make it happen.
TJ Southern:
Oh my God, I told you all that. There are other people out here in this informatics world that are, that is just as passionate and just as concerned as I am. We love informatics, and here we go, we have a guest who absolutely loves informatics. I’m telling you, it’s just something about it that just makes me, oh my god, just makes me feel all warm and fuzzy inside. Okay, so let’s go ahead and just what inspired your work in this space? What inspired your work in this AI, informatics space? Because that is still a new developing space, right? What inspired your work in that space?
Vupul Kashyap:
Yeah, so, what inspired my work in this space is I come from a computer science background, and one of the, my Ph.D. thesis is basically using ontologies and knowledge to integrate data over the web, and I wanted to make use of my Ph.D. thesis to solve a really good problem, and there can be no better problem than improving the health of everyone else. So I told myself, hey, how do I really make an impact? So one of my first places I landed up was at the National Library of Medicine.
TJ Southern:
Oh, my gosh.
Vupul Kashyap:
And that’s where I was amazed at the different kind of medical vocabularies they have, the different PubMed articles, and it was a fun time. And then I realized the big challenge of information and data and healthcare. So, hey, I was coming from a computer science background. My passion was information, semantics, and ontologies, but I wanted to apply to a real-world problem. And there you go, there was the real-world problem staring at my face at the National Library of Medicine and like, hey, this is really important, and that’s what really got me in ….
TJ Southern:
Oh, my gosh, let me tell you, that is absolutely amazing. And let me say this before we actually really get into the crux of this conversation. It is such a pleasure and an honor to even have someone like you just in that seat and, you know, and really ready. You know, a lot of times we have nurse informaticists that come on that, you know, that’s their job. I actually interviewed a nurse informaticist a little while back and she focuses on those SNOMED codes, ontology, you know, she really focuses in on that and it really takes a passion to do that, so I applaud you for that. Okay, so now what are some of the biggest challenges that you’ve seen? What are some of the biggest challenges that you’ve seen? And I know you’ve seen a lot because you’re solving problems. So what are some of the biggest challenges you face?
Vupul Kashyap:
The fundamental problem in healthcare is siloization. People are divided up and doing their own thing, right? So the problem and the really basic problem in healthcare informatics is organization across the board are very poor in acquiring data, managing the quality of data, capturing the semantics and context of data, and so on and so forth. They spend loads and loads of money trying to buy all kinds of technological things. But in reality, the problem is not a technical one. In reality, the problem is a social one where people, if they agree to collaborate, if they agree to standardize, and if they leverage the semantics embedded in the standardization, they can go a long, long way. You know, it’s basically an all-or-nothing approach to manage information. Like, they get these huge data warehouses and a whole bunch of things. What they fail to do is they don’t understand the role of information in enabling a specific value proposition. They don’t try to think, hey, will this piece of information improve outcomes? Will this reduce cost? Will this help me reduce the risk or understand risk better, right? So that’s one piece. And the second piece is they are not able to manage their knowledge. So there are so many clinical guidelines. There’s so much research, right? They’re not able to store it, capture it, and leverage it across multiple use cases. So what people, I think they theoretically understand that medicine is information and knowledge-based, but they are not able to actualize it and implement it in the real-world context. So that is a fundamental problem.
TJ Southern:
Oh my god, like my hair is on fire over here. I’m just going to be honest with you, because I’m like, yes, yes, yes. Everything that you’re saying, every nugget, every building block that you’re talking about, I mean, let’s just break down your first one. You know, the siloed, all of these disciplines being siloed and not just the disciplines, like you said, even when you talk about the technologies, they don’t want to play in the same sandbox.
Vupul Kashyap:
You got it.
TJ Southern:
And so here we are still trying to solve a problem where acorn wants to talk to acorn, but they’re two different acorns and they don’t want it. So yes, oh my god, and that is the fundamental problem. And the cool thing about it is what it does is it allows space for individuals such as yourself and myself to operate, to go in and really say, hey, guys, we really need to come to Kumbaya and hold hands and actually talk, and then we could make the world a better place.
Vupul Kashyap:
Yeah, yeah, and you know, I like Kumbaya. The informatics version of Kumbaya is let’s all agree to talk in the same language. So let me take a simple example, right? You do clinical trials and you deliver care, you’re talking about the same clinical data. How difficult it is for you to either align yourself on the same vocabularies and data models or whatever, or if not, get together and have a mapping between the two so that data can flow, right, because there’s so much data in the EMRs which can be leveraged for patient recruitment. And then there is so much data generated in the context of the clinical trial, which, remember, the clinical trial is going on people who are patients of some sort. And that information can be so useful to their PCP or their physicians in making the care better because the clinical trial people are discovering new things and sharing some of that information to the physicians may not even require FDA approval. This is like my personal care, this can help you, so that’s like opportunity, god, right there, right? So many such examples I can give you, it’s not even funny, you know, it’s like amazing.
TJ Southern:
Oh my gosh, oh my gosh. Okay, so your organization, can you talk about how your organization is currently using data analytics to improve patient outcomes?
Vupul Kashyap:
Yes, so we are proceeding in the following manner. I think so we are, approach number one, we believe that machine learning by itself will not solve world hunger.
TJ Southern:
Thank you.
Vupul Kashyap:
Because I think there’s too much of a craze on machine learning, yes.
TJ Southern:
Yes! Thank you.
Vupul Kashyap:
And we believe that, and this is what we’ve also heard from a lot of people, that it is very easy to get machine learning from, you know, performance up to 90% of accuracy, but beyond that, you really need to engineer a way through. You need to look at rules, ontologies, and stuff like that. So our approach is a combined one, so you say something ….
TJ Southern:
Good, good.
Vupul Kashyap:
Perfect, so we’re on the same page. So our approach is we combine machine learning ontologies and rules in a smart way to deliver production accuracy where, so our first problem we are solving is the ability to look at patient charts and automatically extract clinical concepts and codes for provider reimbursement. That’s the first problem we are solving, right? And we have good traction, we have some very good customers, and so on and so forth. And we are delivering 95% of accuracy for 70% of the charts. Actually, sometimes it’s even better.
TJ Southern:
That’s awesome!
Vupul Kashyap:
And now, it is proved because what we do is we get those guys to they come in, audit us, okay? They do not put us in production till they are satisfied. We have multiple rounds of auditing and so on. And then we sign contracts based on our accuracy estimates. Hey, we will stand by them, right? So, but that’s a fundamental problem. Now, the second step we want to do is, we are now working on is, okay, now if we can extract codes, we can also extract HP and see values. We can also, in for quality metrics, we can figure out whether this patient is an inclusion or exclusion for a particular quality metrics, a diabetes-related quality metric. And what we’re finding is that we are able to identify patients from the patient charts who were, who people didn’t know about, were, because through structured data, the data is not available, so we can go into the patient chart and extract it. And now we are able to show improvement in quality metrics just because of that.
TJ Southern:
Okay.
Vupul Kashyap:
The second step we want to do is, now if you’re going to do quality metrics, care gaps is not going to be far behind. So we are, our next wave of machine learning models and our hybrid approach is to extract care gaps from patient notes, not care gaps, so you never find an electronic medical record, right? So you have to go back to patient notes and stuff like that. So we have a whole, we are looking at, we also looking at literature. We’re looking at medical textbooks so that we can extract new knowledge and bring it to the point of care and stuff like that. So we also looking at clinical decision support, clinical documentation, and all the other things. That’s what I own in my organization, I own the product roadmap and the product strategy. From an informatics point of view, healthcare is informatics, I mean, I don’t know why they call it healthcare informatics. There is nothing else in healthcare other than informatics, but we should not disappoint other people. We’ll continue calling it healthcare informatics. So anyway, that’s my story, yeah.
TJ Southern:
Let me tell you, I love how your organization is so multifaceted. It’s almost like once you find one problem, right, and you have, you know, you have a solution down now you’re like, okay, let’s go to the second phase of this problem, and I absolutely love that about your organization. First, I love that your organization understands that machine learning is not the end all to be all, because for some odd reason because it’s the new buzzword, everybody’s like machine learning, machine learning. I’m like, guys, we can have machine learning all day long, but guess what? You still need us there to assess, I mean.
Vupul Kashyap:
Let’s dive deep into that. The models have to explain themselves because a lot of times you are not convinced, you know, yeah, you may have an 80-90% success rate in the market accuracy rate, but at the time, at the point of care, it is about that particular patient. And in cases where you don’t agree with the, what the model is telling you, the model has to tell you why it did that. Okay, in finance and other places, okay, if you’re underwriting your loan and they want to know whether you’re good, borrow or not, it’s fine, you don’t need to explain. But hey, in the healthcare, you will need to explain, so that’s one big problem in healthcare.
TJ Southern:
Oh my god, I love it, I love it, I love it, I love it. I even the thing that I truly, truly love is how you’re taking your background and being able to simplify all the languages across, right? Because you said you guys are looking in textbooks, you guys are looking at clinical decision measures, you guys are looking at quality measures. I love the fact that you’re taking all of those things and you’re trying to turn it into one simple language across the board, something that’s open, transparent, and shareable amongst, you know, not only providers but organizations. I love that and PSA, guys, we need more organizations that are willing to do that, that’s what he stated at the top of the call. You know, we need more organizations that are willing to play in the same sandbox, speak the same language, share their information, share their data so that we can truly improve patient outcomes.
Vupul Kashyap:
There you go.
TJ Southern:
Okay, so now in your space, do you see opportunities for informaticists? And I know some people may say, well, he just said that he informaticist, TJ, why are you asking him this question? That is because I want you to let people know that there are more than just, you know, your position. There’s other entities and there’s other facets to informatics. So that’s why I’m asking you that question.
Vupul Kashyap:
So it’s a very good question. Let me rephrase the question. There is this confusion between an informaticist and a data modeler or a data architect and.
TJ Southern:
Thank you.
Vupul Kashyap:
Not, yeah, and a lot of people don’t understand that an informaticist, a healthcare informaticist brings much more. First of all, he brings a deeper, you know, knowledge about the data, the meaning, and the context. Someone coming from the finance world will not be able to understand the difference on day one, right? He also brings a deep domain knowledge about healthcare. He understands at what point in the workflow is this piece of information going to be useful. Does it, is it really important for me to spend thousands of dollars of extracting the speed of data where I know in my workflows, you know, it can wait a little bit? It’s not, right, so that, he understands how it impacts cost, right? And if a particular piece of information is not available, he has the domain knowledge to go figure out proxy information to drive, you know, if I don’t have X, I can replace X by A and B and I can get a reasonable estimate of the outcomes or risk or cost or whatever you’re looking for, right? And finally, I think a real, real savvy clinical or medical informaticist understand, has knowledge about clinical guidelines and all the other …, and he’s able to distill them and bring them to a point where they can be implemented at the point of care as a CDS to enable better outcomes in cost. And I’m not saying he understands the clinical guideline to the extent that he can solve cancer, but he has the knowledge to walk through an NCCN clinical protocol, cancer care protocol. Of course, he’ll get stuck in some places because he’s not a physician, but he can go and connect with the physician. You know, a data architect in the same place would be able to do the same, but he would ask ten times more questions, will be a much more friction-filled process. So it’s a very subtle point, and people, it’ll take a while for people to get through, you know. And people think that they can data science their way through it, but eventually, the smart organizations realize that you need informatics. And, you know, I’ve had, I’ve worked for some startups, startups in the Bay Area, and they’re beginning to get it some of them, but quite a few of them are still technology and data-driven.
TJ Southern:
Oh my God, so how do I get you on my team? Because I told you all at the top of the call, have your pen and your piece of paper ready because he literally just broke down what a healthcare informaticist, clinical informaticist, nurse informaticist, physician informaticist, yes, he broke down exactly what it is that we serve, the value that we add to a lot of these projects. And the thing that I love so much is you’ve even taken the time to point out the difference between data science and the informaticists, and a lot of organizations think that they can just get away with a data science, you know, person. And I understand because data scientists are hot now, right? And a lot of people are wanting them because they have a lot of analytical skills. But at the end of the day, you still have to have that informaticist to tell you the context of the data in which it is being used, right? So how do I get you on my team? How do I get you on my team? So I hope you all wrote that down and I hope you all got that clarification because that was textbook 101 what he said. So I know you and I were kind of having a little bit of discussion before we started the cast, and one of the things that we talked about was, you know, informaticists, you know, a lot of people still don’t know about it. So how do you think individuals such as yourself and myself can inform, you know, not only the world, but help organizations understand the value that we bring?
Vupul Kashyap:
Yeah, so that’s a tricky question. Here’s one insight I’ve seen in the informatics world. It’s a cross-disciplinary field. So, for example, Tonja, you are a nurse informaticist. So you have your nursing skills and your informatics skills. I come from computer science, so I have my computer science algorithm and data structure skills, and I have my informatics skills. So I, given the nascent nature of the field, I doubt there’s any individual out there who’s just purely in informatics. He has some other skill which plays into it, and that’s where you start. Because if you are a nurse, let’s talk about nurse informaticists, right? You are embedded in the clinical workflows, both internal and I mean both inpatient, and outpatient, right? More valuable in the inpatient context. You build care plans, right? And you, that is your opportunity to show the value. You can say that, hey, I’m building these care plans, but I’m not going to keep building a care plan for every new patient which comes through. I’m going to start building cohorts, I’m going to start template using the care plans, I’m going to start identifying the data elements used in the context of the care plans. And then I will ask you to go and create a database for me. Don’t go and spend your billions of dollars creating these humongous data warehouses which fail and they provide no value to anyone, right? So let’s just watch, so I think the first thing I would do is leverage your, say, nursing skills in your case and identify value, clear situations of value. And believe me, there are huge. And then propose a quick pilot and prove the value. If you can say, hey, give me three months, let’s do a quick pilot, and let’s show how this improved outcomes, and then that gives you the credibility to start pushing the ideas more and making suggestions and stuff like that.
TJ Southern:
Love it, love it. Okay, so what are you most excited about as it relates to the future of what you and your organization is doing?
Vupul Kashyap:
Oh, I could talk about a general as well. What excites me about the future is two things, and this is of course applies to my organization, but given that we are all in the spirit of sharing and collaboration, I would really like the industry to work as one team because otherwise, this will not happen. Leaving that aside, there has to be some mutual benefit, win-wins, which we’ll have to identify, but that’s a different topic for another day. The two things are really, really, really, this is the promise of informatics. So if you proceed from the hypothesis that medicine is information and knowledge-based, then informatics provides an opportunity to rethink medicine, the practice of medicine, and even the knowledge of medicine. So that is the goal, the Holy Grail we are looking for. We should align ourselves and say, hey, let’s rethink, do we really need to, do really, do all patients, do they really need to go to physician’s office or only the most serious ones? Or can the patients do something ahead of time so that they don’t have to sit in the clinic and enter the same damn information every time they go in? And those kind of things, simple things like that. The second thing is, I think we’re still incredibly slow and, of translating from research to practice, and that is where I’m really, really excited about. Like if there’s another COVID which comes by, can be quickly figure out existing knowledge. Can AI, machine learning algorithms pull out for specific use cases around COVID? Number one. Number two, as people are posting new and new information and new and new insights because nobody sets publications here. So one source of knowledge is the research publications, the PubMeds of the world, but the other source of knowledge is the patient notes, because they are people being treated on the front lines and if there is a knowledge embedded, then people don’t get that. If he can come if you can in real-time, you know, figure out clinical protocols, clinical decision support rules, and quickly make them actionable, you know, and executable both in an IT sense and even in a manual sense. If I just hear from a lot of physicians that … is really doing well for COVID. Okay, I can put it in my order, set, create a … or whatever. And, you know, I know I may run into some FDA kind of things there, but I’m just talking about a concept here, right? So that’s, those are the two big, those are the two things which informatics has the potential and the promise, and I would love to push that boundary a little bit.
TJ Southern:
I love that. I love it. I love it. I love it. Okay, so now, I love everything that you have said, and I promise y’all, I pray that y’all was tuned in and plugged in because I’m going to have to go back and listen to this episode because this thing right here is so juicy. I’m just letting y’all know, so, do you have any parting words of wisdom for our guests?
Vupul Kashyap:
Yeah, so I have three words, two words of information, information of advice, information and knowledge. Whatever you do, and I would drop in the word ecosystem as well. Whether you’re a nurse, whether you’re an informaticist, or whether you’re a computer scientist. Try to understand the context of the information and the knowledge where it’s coming from. Try to link it to, okay, why is this? How was the information created? What is the value of this information? How can I use it to improve outcomes or reduce cost or, you know, better engage with my patients? Keep that in the back of your mind. And I think what happens is that will drive changes in the way you do your daily work. You will think differently, your program for computer scientists, you’ll write programs in a different way. For nurses, you will probably create, for nurse informatics to create better care plans, more outcomes-focused goal-driven care plans as opposed to general, you know, I did X, Y, and Z kind of care plans. So that’s my advice, be information centered, centered around information and knowledge. Be truly information-centric, it requires a change in thinking, and it’s almost behavioral to some extent, but yes, that is what will help you.
TJ Southern:
Oh my god, so change your thinking, change your view. Change your thinking, and it will change your view. Love it, love it, love it. Well, thank you so much for being on our show today. It was an honor and it was a pleasure. And I definitely, definitely pray that we can snatch you to get you back on to pick your brain some more. I love it when I’m able to talk to other informaticists who have the same fire and passion that I have about the world of informatics.
TJ Southern:
Hey, y’all! Thanks for joining us today for another episode of the Outcomes Rocket Network – The Future of Global Informatics Podcast. If your organization is looking for informatics talent, go to www.Beryllus.net. That is www. B E R Y L L U S .net, and we can assist you in finding some of the best nursing informatics talent this continent has to offer. We’ll talk to you later! Have a great day! See ya!
Sonix has many features that you’d love including automatic transcription software, share transcripts, advanced search, world-class support, and easily transcribe your Zoom meetings. Try Sonix for free today.
