How to Tell a Story About Your Data
INTERVIEWEE: Ridge Foster
Without the help of technology, transforming the output from engagement sessions into actionable insights can be a barrier to fast and precise decision-making. In this episode of the MAPS Elevate podcast series sponsored by Within3, you’ll learn how emerging technology can help Medical Affairs teams quickly understand the results of their engagements. Experts will demonstrate how to implement AI-powered natural language processing to gather, interpret and integrate insights to uncover the story in your data.
Following is an automated transcription provided by otter.ai. Please excuse inaccuracies.
Garth Sundem 00:00
Welcome to this episode of the Medical Affairs Professional Society podcast series: “Elevate”. I’m your host, Garth Sundem, communications director at MAPS. And today we’re speaking about AI powered natural language processing to uncover the story of your data with Lance Hill CEO Within3, and Ridge Foster, Global Customer Success Manager at Within3. This episode is made possible by Within3 whose Insights Management Platform helps life science companies identify the right people actively engage them and get answers that informed strategic decisions. So, Lance, we’ve talked before and in previous episodes, we’ve chatted about kale discovery and engagement. And today, we’re talking about what to do with the data that comes from these interactions. And so, I was hoping maybe you could get us started by describing this challenge. Why is it so, hard, especially now, to know what our data means?
Lance Hill 01:05
Yeah. Hey, Garth, great to speak with you again. Yeah, it’s, it’s actually an increasingly more difficult challenge. So, one of the advantages of how Medical Affairs has grown over the years is Medical Affairs gathers much more information from many more sources than ever before. From one-on-one interactions with HCPs, which were always going on traditionally, but now are typically either done or it can be done electronically. Or if done in person, the feedback or the output of those interactions is oftentimes stored electronically. Information coming in, through, through call centers, information coming through interactions around congresses, whether it be you know, science things happening online and digitally, in what you know, all the way to population health sorts of understanding and insights of what’s happening within a disease community at large. And so, with all of those inputs, at the end of the day, someone has to determine what it all means and what to do about it. And how it affects strategy. And a lot of those inputs from someone posting a tweet to an MSL saying I just came out of a meeting with a physician. And here’s a big takeaway, it ends up being free text. So, it’s not structured data, like here’s a database, and it’s always, you know, numeric in cell number three, and I can just do the do the additions. It’s a lot of free text coming in with different contexts and different regions. And the volume of it is higher than has ever been the case in the history of life sciences. And so, trying to pour through that, in real time with just people and trying to do that in a way where you’re, you’re being very objective about what you’re pouring through, can be really hard, bordering on possible to do at scale. And so, that’s where that’s where technology comes in is to help teams do that.
Garth Sundem 03:02
So, we’re not just talking about engagements today, we’re talking about NLP to survey the wider ecosystem, you know, of the conversation that’s, that’s going on about your data. Lance, is that does that sound right to you? I mean, are we using NLP not just to figure out what the heck are MSLs are bringing back as insights but wider than that?
Lance Hill 03:24
Yeah, and natural language processing is really a technology. Again, it’s used in all different industries in all different ways. But the investments made in the last several years in life sciences, in particular, and Medical Affairs, more specifically, have taken it from an interesting tool to something that’s becoming almost vital. And really, what natural language processing is doing is an input, free information, free text input from whatever the source can analyze it, pull out what matters, pull out what’s trending, evaluate sentiment, a variety of things. And, you know, gather all that together so, that someone trying to understand what’s happening within a medical strategy, what to do next, can quickly make sense of it. So, NLP as a technology is not just limited to I had a an MSL, enter call note into my CRM system or something like that and evaluating it. It really is a technology that can be applied to any sort of input from any channel that serves Medical Affairs and helping understand the voice of the customer.
Garth Sundem 04:31
Okay, cool. So, Ridge, let’s jump to the far other side of what we’re talking about here. What are we trying to get? So, we’re trying to get the story of our data? What do we mean by the story of our data?
Ridge Foster 04:44
Yeah, really, Garth, that comes down to understanding as we like to think about it outside the four walls of your organization. So, what are the conversations the topics that are there trending? What are the experts in the field saying whether it’s from a scientific perspective Have a clinical perspective, a research perspective, across those different range of interactions as Lance just went through. So, having your pulse on those conversations, gives you a better ability to curate your strategies, calibrate them, and ultimately drive impact for your, your product or patients within the marketplace. And it’s the stage or therapeutic areas you’re trying to target.
Garth Sundem 05:24
Yeah, I’ve heard people on the marketing a commercial side talk about, you know, telling their product story. And in some ways, it seems like one of the goals of Medical Affairs is to tell the scientific story of an asset or even an organization. You know, do you both? Do you think that this sort of listening allows Medical Affairs to not just sort of tell the story of data, but to look at a wider picture of what the organization’s scientific message is? And should be?
Lance Hill 06:01
Yeah, you know, what’s interesting about what you’re saying Garth, there, I think, and this is maybe semantics, but I think, yeah, really, Medical Affairs is more about two-way communication than telling the scientific story. So, what I mean by that is telling the scientific story means me talking, and I’ve told it doesn’t mean you believe it, Garth, or that you heard or understood it. And I think really, ultimately, what Medical Affairs is trying to do is to determine if Garth is someone who is important in making decisions around my therapeutic area, you know, determining the best treatment for patients or, or talking about in scientifically, what matters. My real goal is to try to understand how can I get Garth, to understand what I’m talking about? How can I listen to what Garth thinks, and bring that back. So, I can say, you know, what, Garth is really coming at this from a different point of view than I suspected, we should alter the work we’re doing at Medical Affairs to better communicate with folks like Garth. And that’s that, that ability to hear your point of view, to understand where you’re coming from how different is then then maybe within the four walls, as Ridge just said, within our institution, how we feel about the scientific narrative, as an example, and what it takes to bridge that gap. Or maybe what else is needed from an educational perspective to help you understand, you know, why our therapy has advantages in different situations, or whatever the case might be. So, it’s that really, a lot of times on the commercial side, you’ll hear words that are more about like reach and pushing the message out and getting the message in front of people telling what makes medical to me so, special as medical exit communicates in both directions. And NLP as a technology and AI, when the volume of communication coming back to medical becomes very, very high, and very, very intense. This technology helps you simplify it all. So, you can work with it in real time and make the right decisions versus combing through hundreds or 1000s of pages of notes, and trying to correlate what might it all mean? And what context?
Garth Sundem 08:18
So, first of all, I find this so, empowering that that you’re speaking to me, Lance as a decision maker in the in the healthcare space. But
Lance Hill 08:27
You’ve got quite some clout, Garth. So, we’re gonna use the example for the rest of the segment.
Garth Sundem 08:32
Well, good light, good, good luck creating this understanding that you’re trying to create in me, so, I wish you I wish you all the best. But so, Ridge, how do we do it? So, I mean, NLP I think most of our audience has familiarity with the term we know is natural language processing. And we know we’re trying to create meaning from unstructured data, meaning that there’s all these words out there, people are saying all these words, and they’re typing all these words, how does it work? How does NLP actually create the story of our data?
Ridge Foster 09:03
Yeah, well, it there’s, there’s quite a few moving pieces to it. But I think that the nice thing is, as you look at the industry over the last couple of years, it’s really evolved to be much more approachable and packaging the ball to be deployable to two teams, particularly for Medical Affairs, where the uniqueness of the solutions have risen to a degree that they’re competitive, and in many cases, overlapping kind of the more kind of historical solutions or general solutions. And so, what it really comes down to is, you know, to two pieces primarily, you need to have your NLP model, and you need to have a way of getting it into the hands of the folks who are going to be capturing those interactions and discussions on the NLP model front. You want to be able to have something that is tailor fitted for the discussions that are life science based. So, really understanding the terminology, the words that are actually being used so, that it can differentiate between, you know, what is important and non-important, again, detecting the signal from the noise, being able to understand the context of those discussions. So, in certain discussions the competitors have may have a different kind of tone or a different understanding to it than other contexts or other discussions. So, you need to have a model that’s really trained to understand those conversations. From a scientific perspective, Medical Affairs perspective, this, the second piece, then is the implementation of that model. And that typically comes with the productization of that model, and getting it into the hands of your field teams who are actually engaged in these discussions. And in that case, there’s been, like I said, a lot of investments of more modern solutions that really are built with the process of Medical Affairs in mind. So, that wherever you go to deploy them, it handles both the kind of process aspects of insight gathering, and the nature of kind of the interaction as well as then the model behind it to help power the concepts or the insights that are coming out from those discussions.
Garth Sundem 11:16
Okay, so, if I’m a Medical Affairs team, and I realize that I’m overwhelmed by the volume of data, can I implement an NLP solution on to this massive lake of data? And say, you know, please NLP, tell me what all of this means? Or do I need to restructure the entire way? I’m getting data representing data? How hard is this for a Medical Affairs team to implement?
Lance Hill 11:45
Yeah, probably the coolest thing, Garth is you don’t have to do any debt restructuring. So, literally, the tools just ingest the data you have. And that’s what the tools are for us, then analyze that data and do all that structuring for you. So, if you’re implementing a tool that might, there might be some upfront where you say, hey, I’m really interested in these sorts of concepts. I want to come back to the idea of concept in a second. And as it pertains to NLP, okay. I’m interested in adherence, and I’m interested in the mechanism of action about my therapy, and I’m interested in, you know, a couple of regional things I’m interested in on that need, there might be some things that you know, based on where you are, that you say, when you look, when you apply NLP to all the data have historically plus what’s what will be coming in, analyze it for all of these sorts of concepts and come back and tell me, do they exist? Are they trending up? Are they trending down is the sentiment positive or negative, it helps you group it all. And so, the great thing about these technologies is you do not have to rework your whole data lake or your data warehouse or your CRM or anything, all you really need to do is feed the data in the thing about concepts and by the way, that that that’s that typically is very lightweight, and does not take that much time to do. And so, that’s where a lot of companies can kind of go from, I’m not doing any of this at all, what you guys are talking about sounds crazy to me, like futuristic, oh, we’re doing it at a really high level, because it’s not that invasive in your environment. The thing with concept is really, I want to build on something that I talked about, when we talk about how you want NLP trained for Life Sciences. What we mean, is it No, there is no NLP, there’s something in the middle, that is called tagging, which basically says, I don’t know what efficacy is as a concept. So, what I’m going to do is manually say, when you see this keyword, when I see the word unresolved, I’m going to say that goes under the category of efficacy. And then when I say, you know, when I see the word cure, that is going to manually go into the category of efficacy. So, so, kind of the middle is companies trying to build out these really complicated tagging systems to try to in advance imagine every word that might attach to a concept that they care about, and then run. Yeah. And that’s important, but kind of where the technology is advanced to now is out of the box. Some of these NLP solutions have the ability to understand the concept of efficacy, using various advanced AI technology, so, that it can automatically look at the data. And you already know things like efficacy, adherence, mechanism of action, product names, things that are very life sciences specific. You’re not sitting in a room defining all of those with a blank slate. Yeah, it already exists. And that lets you again, even if you without customization kind of run those analyses across all of your data, and immediately have very robust reporting. And that’s what’s exciting about where the industry has gone last couple of years is it used to be a very generic sort of approach that ended up being quite manual behind the scenes with humans tagging things. Yeah, it is not an any longer.
Garth Sundem 14:57
Okay, well, that’s it. I was gonna ask about that so, If I was a company and I was interested in certain things, I could adjust this tagging, is it almost like building a taxonomy where you’re looking at, okay, these things are birds and these things are, you know, raptors, and then we have different kinds of hawks below that or whatever. It but if you are interested about specific things, you could adjust that kind of tagging taxonomy…
Lance Hill 15:23
Exactly. Yeah, Ridge, you maybe want to talk a little about that, but exactly, yes.
Ridge Foster 15:27
Yeah. And that’s the more kind of antiquated approach, right where that is manually built by customer, they’re kind of experience in this space, and looking at the conversations that have happened historically. But with NLP in concert with AI, and it’s really the AI piece that helps inform the model ongoing basis, as those discussions are occurring on a continual basis, it helps kind of inform that taxonomy all by itself by understanding those conversations. So, you’re really talking about a entirely manual approach with almost an entirely automated approach.
Garth Sundem 16:03
Okay. So, the past was an entirely manual approach where you had to give the NLP the taxonomy, yeah, you know, the right present is that the taxonomy? And I know, I’m probably misusing that word, but it’s stuck in my head. The taxonomy evolves based on AI, you know, it can learn what is the future of NLP to make meaning of data?
Lance Hill 16:30
That is a great question. Yeah, I’d love to take the first crack on it, this is one of the things we’re most excited for. I’m personally excited for and I think, you know, the broader Within3 team is, to me, it’s, it starts to bring in what we were just calling kind of outside the four walls, bring that into combining with inside the four walls. So, how do you leverage other models that are trained for other discussions that are happening, say, through your clinical trial management system, or through your adverse events reporting system, and combine that with what you’re hearing in the field, so, that way, you’re now getting a holistic 360 view of the trends in conversation, not just conversations, but trends, key items, kind of topics, that are surging from holistically across a given therapeutic area or disease state, and not just from outside the walls that, you know, your fields may be engaged in. And that’s, you know, it’s a little bit different models, and a little bit of a different kind of productization around it, but it gets at that holistic view of you know, what is really happening with this product with this, this, you know, scientific molecules in the space, what is truly happening from all these various angles that we want to be able to, you know, dissect and slice and understand from
Garth Sundem 17:48
Okay, and Lance, where do you hope NLP will bring Medical Affairs teams? What do you what do you want NLP to deliver for teams,
Lance Hill 17:59
You know, what I really want is NLP to make healthcare smaller. And what I mean by that is, the more inputs and touchpoints that Medical Affairs can realize, because only so, many hours in the day, the more real outside voice of the customer and even inside the four walls of the bridges are talking about that can be kind of gathered up and synthesized, so, that more voices are included within healthcare, the better for everybody, the more informed our HCPs are, the more informed patients are, the better for everyone. And one of the barriers to that is there’s only so, many hours in the day, and without tools to help deal with the volume and separate the wheat from the chaff. It’s just really not possible to go too far beyond the traditional, you know, opinions of the high-end KOL. And some of those things, it just becomes too difficult, especially for smaller teams, to really take a wider, more connected view, a more democratized view of the participants in healthcare into their plans. And that, to me is the most exciting thing about these technologies they are they’re like having, you know, a team of assistants, you’re your medical strategy to help you understand what’s happening and synthesize it for you and bring it back to you. So, you can look at it and make a decision. And you know, how many of us wish we had a team of assistants to help us in our in our busy lives. That’s what this technology really does. It’s not it doesn’t replace the need for you to make those judgments at the end of the day. But it sure makes your life a whole lot easier and lets you make judgments off of a much, much broader data set in more real time than if you’re trying to do it off with spreadsheets or those sorts of technologies.
Garth Sundem 19:41
So, we want to appreciate the complexity. We want to embrace the diverse and complex voices that make up, increasingly make up health care, but we also want to then simplify it. So, alright, let’s leave it there for today. Thanks, Lance and Ridge. To learn more about how your organization can partner with Within3, visit Within3.com. MAPS members don’t forget to subscribe and we hope you enjoyed this episode of the Medical Affairs Professional Society podcast series: “Elevate”.