Garth Sundem  00:00

Welcome to this episode of the Medical Affairs Professional Society podcast series, Elevate, gathering the voices of Medical Affairs thought leaders and stakeholders to explore current trends define best practices and empower the medical affairs function. I’m your host, Garth Sundem, communications director at MAPS. And today we’re speaking about natural language processing with Hywel Evans, director of the IQVIA Natural Language Processing Insights Hub, with a mission to discover previously unseen insights, drive smarter decisions, and unleash new opportunities. So how first, I love the term insights hub. Can you tell us a little about it? And what is your role there?

Hywel Evans  00:46

Absolutely. Hi, Garth. And thanks for having me today. Yes. So um, yes. Now, the insights hub really is all about pulling together information and insights from rich but often difficult to deal with data sources. And so my role within the insights hub, is to drive the functionality as well as work with our customers and our clients on how to use this technology best to help their teams. So it’s things like taking the outputs from natural language processing, rather than the sort of underlying technology taking those outputs, which is really the most important part, and putting those into the hands of different teams. And within medical affairs, specifically, of course, to help them cut through that noisy world of unstructured data and turn that into actionable insights.

Garth Sundem  01:41

Well, so how is your background? Is it data science? Or is it on the business side of things? What is your background?

Hywel Evans  01:49

It’s a bit of both, actually, yeah, I originally started out in data, and data science, and analytics. But I’ve also had roles in and around consulting and working with different groups around using the outputs from data to do things like change strategy, developing processes, things like that. And prior to coming on to the insights, hub role, I was part of IQ v as business in Asia pack. I’m currently based in London, but I was previously based in Singapore, where I lead our analytics Center of Excellence there and diverse and digital initiatives as well. But I’ve always worked in and around data and the application of different technologies and analytical methods to that data.

Garth Sundem  02:35

Okay. So, you know, in terms of insights, I think the medical affairs function thinks about msls, bringing insights back from the field, you know, and so this is sort of like a human applying their lens to what an insight might be. What are the data sources you’re working with? Are we still talking about NSLs generating the data of insights? Or are we talking about totally new sources of data?

Hywel Evans  03:05

Hmm, I think there’s essentially a need to do both. refocus on core data sources that are, you know, with us and will be with us for the foreseeable future things like the information from MSL discussions and interactions, but also looking more broadly and bringing in additional insights from the wealth of data sources available out in the wild as it were. And so we have a lot of effort placed on things like scientific literature, and both established sources as well as kind of emerging sources around things like preprints. And then we talk a little bit about additional sources from things like social media, and isolating key data points from sources that are really quite hard to navigate, you know, at first attempt and so on, there’s an opportunity to apply this type of technology across quite different sources. But all at all, with the aim of bringing together a coherent set of insights in and around let’s say, therapy area, and, and really empowering people like me cells to do more with that.

Garth Sundem  04:21

Well, that that’s interesting that you could apply this same methodology to something like a scientific journal and also to Twitter. You know, how in the world does that work? You know, are you looking for the same kinds of I don’t even know how this works as a keywords or how does natural language processing work with scientific journals and with social media.

Hywel Evans  04:48

So the key part about natural language processing is it’s really looking at sentence based information, or information is presented as words assembled, usually by people. So whether it’s people altering scientific information in a somewhat more formal style, whether it’s a patient’s talking to another patient, or on a forum, for example, sharing some of their experiences, maybe some of their challenges in managing a disease perhaps. And all of that information is presented as sentences, sometimes shortened to the point other times long form within a document. The key to natural language processing is the ability to cut up those sentences. And make associations between different parts of those sentences, as well as identify terminology, as you said, sometimes it can be as simple as perhaps a keyword. But often it’s more looking at things like synonyms for concepts. Because of the variety, as you rightly point out the variety of ways that the same thing can be written or in different styles, whether it’s more formal, or it’s more informal styles of writing, and natural language processing can be configured in different ways at to have to adapt to those different authors, as well as those authors styles. Well, I’m glad I’m glad you’re doing this, and not me, is what I got from that.

Garth Sundem  06:12

So most of your clients at IQVIA, are they trying to build sites around a disease landscape or around their product or around their organization? What sort of insights you see clients looking for?

Hywel Evans  06:29

I think, yeah, of course, a lot of our customers are organized around or focused around therapy areas, and tend to have products then associated with those therapy areas. So the therapy and all products, so the lens is a very, very common one. Absolutely. However, we do see other cases where an organization might be looking at an inherited database of documents. So if for example, you acquire an asset, as part of a commercial endeavor, you acquire an asset and all the different documents, regulatory and scientific as well as marketing and various other documents that will have come with that asset. And you may want to mine that as a whole corpus, and really get under the skin of what do you have one of the key key artifacts, and what can be perhaps ignored as noise, etc. So really depends on the team involved, what they’re trying to achieve. And we would usually start by framing that problem, as it were with our customers, in order to really understand Where’s NLP going to be useful and valuable in where it’s applied?

Garth Sundem  07:40

That’s interesting. So if you’ve grown up with a new product, through your own r&d, and in your own strategic planning, maybe you know a bit about the landscape, but if you’ve inherited, maybe you don’t know that landscape, and I generated a new. So I, I hear a lot of people talking about the pace of change right now in medical affairs. And it does seem like insights is is top of the list of pace of change. You know, do you see current trends in medical affairs, driving the need for natural language processing?

Hywel Evans  08:21

Absolutely. I think one area of change is certainly the demand for strategic input. So being able to put medical information and insights from the field insights from real world evidence, being able to put those data points and the conclusions and the insights drawn from those data points into a strategic context, and then play that role of the that medical voice within the strategic discussion of the wider organization. So there’s a need to elevate the conversation and bring that expertise in to that strategic conversation across the organization. And natural language processing can be a provider of those additional data points, or those contextual data points, or those types of discussions. And then at the more working level, the day to day level. Certainly expectations are increasing around the ability of metaphors, professionals, and cells to really be on top of what’s happening in the market. And so arming those individuals with the most up to date information. And allowing them to kind of cast their eyes further and outside of the kind of normal sources of information to get that extra piece of information is challenging. And so technology can be a way to enable and empower those people to be better informed without giving them you know, a ton more work to do. And really, that’s, you know, there was always a sort of efficiency as well as kind of richness and benefits to applying these types of technologies. You’re always looking to make life easier and better, as it were, you know, you’re not just adding complexity and more and more data, you’re trying to be selective about adding that data in so that you can then, you know, add it in without burdening those individuals, and give them that much richer set of data.

Garth Sundem  10:18

Well, you know, I hear a lot about the developing strategic role of medical affairs, and you bring that up in the context of insights. And it seems like, one thing I’ve heard from our members, is that it’s been possible for a long time to generate a significant amount, number of insights. But then it’s difficult to decide how those insights should drive strategic actions. So once you have natural language processing, and and it gives you I imagined a menu of insights. What do you do with those? How do you prioritize those? How do you understand them? How do you use them to drive strategic action?

Hywel Evans  11:03

I think two things spring to mind, especially around prioritization. And I think the ability to prioritize key data points will partly come, of course, from the expertise of medical professionals themselves, you know, there’s no point do we talk about technology kind of stepping in and replacing the role or the activity of professionals in this space, this is all about giving those those same professionals, better tools, sharper tools, if you will, if you will, to do to do what they do today in a more efficient and original manner. And so the one of the ways that you can kind of decide on focus areas of insights and where to where to prioritize is to really understand the context in which some of these insights are generated. So knowing what’s happening with patients, knowing what’s happening with your competitors, or competitive products, alternative therapies, knowing what’s happening by geographic market, these are all important pieces that can then lead you to the right set of priority insights for a strategic conversation.

Garth Sundem  12:21

So you know, yeah, well, so natural language processing can provide the insights. And then the humans involved in this provide the context Is that still the, that’s the case?

Hywel Evans  12:32

The product context, and are able to then feed back into this type of technology, their own priorities. And so the way we assemble our insights hubs is in a workflow manner. So we would start with what is the background or core information that somebody needs? What is the backdrop against which they’re operating? So one of the latest publications happening? One of the key themes or topics of conversations, the patients are having one a Keith van themes and topics that ACP is talking about? And what does that look like? And how is that changing over time, and then giving the user the ability to then drill down underneath those summary pieces of information so that they can really understand how are things changing? What don’t we know? Or what are other people talking about that we need to be talking about. And so really helping individuals to cut through that noise in a way that is aligned to their thinking and aligned to their workflows is really important. And so we we co create our insights firms with our customers, they’re not prescriptive things. They’re an application of technology, where we would work with different teams to make sure that we’re understanding how they think about their therapy area or their key priority analysis. How do they think about it? What is the sequence of insight that they need, from a high level summary right down to very detailed pieces of information extracted from the data? What do they need? And how do we assemble the entire staff so that they can intuitively get that information? So that’s a really, really important part of this process, if you will, is to understand how people work today, one of the extra pieces of information and insight that they need, and how can that be presented without burdening that team?

Garth Sundem  14:27

You know, it’s such a, I mean, obviously, it’s such a technology based system, but what you’re talking about is designing around people, you know, making making it intuitive, so that you have, you know, a tool that is not just a tool, but it’s a tool that provides value. So where do you see this going, you know, what is on the horizon for natural language processing?

Hywel Evans  14:52

I think, you know, the, one of the key data points that we look at is, is data generation, you know, where, whereas information being created, how much of it is being created. And so, you know, a lot of organizations and a lot of different technologies talk to the ability to deal with that deluge that that increase in information. As far as natural language is concerned, it, of course, needs to focus on information that is, as I said, at the beginning of conversation, created by people so authored, as it were. And so sources like the ever growing amount of literature and science being generated, and looking at things like preprints, to get out to get hold of that information as quickly as possible, is one area. And certainly in terms of patient forums and social media, we see that being an ever growing data source as well. And so, you know, those two are important, but you know, stretching a bit beyond that, machine, human interactions are continuously evolving and changing. We were all familiar with chatbots. And voice bots, for example, when it comes to things like consumer experiences. So if you’re booking a flight, or you’re seeking support, for you know, for, for for a broken appliance or something like that, it’s very common to be faced with a chatbot, or an automated interaction, at least, to begin with. And, in, in, in medical affairs, in life science, in healthcare, it’s much, much harder to do that at scale. And these are much more complex topics, and it’s a much more human interaction. So over time, I think tools like natural language processing will be embedded within some of these interaction systems. Again, maybe not replacing that human interaction, but augmenting it or helping those individuals involved to to be better understood, or to get the right information. And for what they’re asking for, to be fed quickly into processes where they can be acted upon. If patients, for example, are confused by something, and they want more information, that’s a vital piece of data that needs to be fed into processes that will, you know, close that gap will create better materials for patients, or better communications, and better guidance through hcps, etc. So I think, you know, looking to the future, it’s a really important cog within many, many processes, I think, well, it’s a better understanding of human systems, you know, from authored material you say, so, and almost becoming more human in its ability to understand, you know, meaning and, you know, in this ever expanding data world that we live in, so, nearly Absolutely.

Garth Sundem  17:52

Let’s leave it at that. Thanks, Hywel!I you know, it’s a fascinating overview in a world that I don’t know anything about and would like to know more so to learn more about how your organization can partner with IQ via to look inside language for actionable insights, visit iq via.com. That’s IQVI a.com maps members. You can continue the conversation at the maps Connect app. And don’t forget to subscribe. We hope you enjoyed this episode of the medical affairs professional society podcast series. Elevate