AI Leadership Insights: Applying AI to Health Care

Mike Vizard: Hello and welcome to the latest edition of the Techstrong.ai video series. I’m your host, Mike Vizard, and today we’re with Robert Poolman, who’s Senior Vice President for the life sciences and health care products portfolio at Clarivate. And we’re going to be talking about how AI gets applied to health care because, well, there’s a lot going on in this space and maybe some amazing things are about to be happening. Robert, welcome to the show.

Robert Poolman: Thanks, Mike. It’s a pleasure to be here.

Mike Vizard: A lot of issues have perplexed us for a long time and I think the one people probably are most familiar with are things like cancer clusters that are unexplained. Somewhere in the data there’s an answer, but we can’t seem to parse it just yet. And so is that one example of where we might see some advances in health care that we have not seen so far or that have alluded us, and what else can we look forward to?

Robert Poolman: Yeah, no, listen, it’s a great question and let me take you through a few examples of where I think you’re going to see … already actually you are seeing and you will see even more going into the future, the applicability of AI and generative AI. So it’s really helped help spur health care, spur the life sciences industry. So if I take you to three main challenges, Mike, that the industry is facing. So firstly it’s around R&D productivity. To your point about leveraging data, leveraging that genomic data as you called it. Clinical trial optimization will be the second challenge I would point you to. And then thirdly, the increasing regulatory safety hurdles that the actual industry needs to overcome.
And it’s interesting, because if you look at all three of them, AI, artificial intelligence, and generative AI specifically, can actually play a role in all of them. And we’re starting to see some of that already. So for example, if I take you to your question around R&D productivity and leveraging that genomic data, there is already technology out there to bring different data assets together, whether that’s publicly available data, [inaudible 00:02:06] actually with a company’s data themselves. And then to be looking at that and suggesting potential new drug targets that a particular company can go after and try to actually target if you like, to treat a particular disease. So that’s something that you’re already starting to see in the market.
And in addition, another follow on from that in that R&D productivity space is the fact that now we’re starting to see AI play a role in suggesting drug candidates, suggesting new drugs that can actually be used to treat patients with a particular disease, and also importantly to screen out drug candidates as well. We know one of the biggest challenges that the industry faces is the fact that it takes a lot of new ideas, new drugs, to go through the R&D process to actually get to a successful drug that comes out at the end. The success rate is really, unfortunately, it’s quite low. You’re talking about 88% of drugs that go into the clinical trial phase actually fail. So if we can increase that screening and remove a lot of those failures early on, it’s going to help the industry be more productive, be more efficient, and it’s going to help patients. So we’re going to get drugs to patients faster.

Mike Vizard: Can we shrink the time it takes for those breakthroughs to come to market? Because sometimes I feel like we’ll announce that there’s been a breakthrough, but by the time we get through all the trials, a decade can go by and a lot of people can be adversely affected in that time. So can we shrink all this to something that feels a little more accelerate?

Robert Poolman: Yeah, I mean, listen, I think we can shrink elements of it. At the end of the day, you know as well as I do that at the moment, if you look at the length it takes for a particular drug, to go from that idea to get into a patient for the first time, to treat them with a particular disease, it’s on average around 10 to 15 years. It just depends on the particular therapy. There are elements in there that the likes of artificial intelligence and gen AI can really help, can really shrink. Some of the earlier stages around that R&D, the what we call the more drug discovery process, like we’ve just been talking about, they can be shrunk for some of the pieces we just talked to. There’s elements of the regulatory process, which can sometimes take a while to actually get the relevant documentation into the regulators, like the FDA, for example.
Again, there’s opportunities there to shrink that in terms of the actual applicability of using generative AI to support some of that documentation. That’s going to come with time. That’s not going to happen now or immediately. But as we go forward in the future, I can see an opportunity there to really help that particular process. The thing that takes a lot of the time, as you know, is taking a drug into the clinic and into patients and getting the right readout of those different phases of clinical trial testing. That piece, there’s going to be some opportunities to make that more efficient. For example, selection of patients, the right patients that can benefit from a particular drug or a particular asset. By looking at, for example, their genomic data, their makeup, if you like. And selecting them based on where they’re going to have the greatest chance for success, let’s put it that way.
But at the end of the day, the drug will still need to be tested. It needs to be made sure that it’s safe, it’s efficacious. The standard criteria that a drug has to meet before obviously it will be approved by the regulators. And that is still going to take the time. It’s still going to have to go through the different phases of clinical trial testing to make sure that the drug meets those criteria. So there are elements, like I said, that can be shrunk. Elements, which I think we’re going to have to face reality, will still stay probably the same as they have been to date.

Mike Vizard: Mm-hmm. Do we have enough skills in the health care sector to apply AI? Because it’s not the only sector looking at AI. And sometimes a lot of the organizations that are in this sector don’t have the same resources that, say, a financial services firm has. So can we get the right people in the right place to focus on this?

Robert Poolman: Yeah, that’s another great question. I think the answer is it’s obviously been recognized as an area that is a growth area for the industry and therefore there’s been a lot of investment made across the entire industry within Clarivate but also in actual traditional life science and healthcare companies, your traditional pharma companies, around investing in talent, especially in this space. And if you look at some good examples, like for example, Novartis are very much focused around leveraging data, being a data company and being able to really maximize the applicability of, not just artificial intelligence, but just data per se.
And therefore they’re investing in their talent to really bring in people that really can help them in this space. So I think what you’ll start to see and already are seeing is that there’s investment in talent within these companies around data, the use of data science, et cetera, and artificial intelligence. You’ll also see companies partnering as well and collaborating with external organizations that are really the experts in this space and can really help them. And already we started to see collaborations and partnerships ongoing within the industry to bridge that gap, if you like. So I think there’ll be a bit of both as we go forward from here.

Mike Vizard: Do you think there’ll be more rigor applied to the way we manage data in the health care sector in the age of AI? Because people are starting to realize that these models are highly dependent upon just how much quality there is in the data to begin with.

Robert Poolman: Yeah, no, absolutely. I mean, that is already there and I think it’s going to be a critical thing going forth from here in terms of … Let me break it down slightly for you, if you don’t mind. So if I look at what we’ve done within Clarivate, the key thing that we’ve taken here is that we’ve taken our quality data, and this is one of the things I will emphasize, that quality data in gives you quality data out. We spend a lot of time here at Clarivate bringing together billions of data points from all around the world from different sources. And we make sure that we normalize that data, we standardize that data. So it’s really structured and it really is indexed in a very structured and clear manner. Why? Because then we can feed that data into these models. And I think it’s that standardization piece, that rigor, as you were saying, behind the actual data, that’s going to make the difference with regards to the reproducibility of these models and how accurate the actual data is that comes out of them at the end.
So that’s a key element there, which is quality data in equals quality data out. I think the other thing you’ll see as well is that as we work with, I mentioned the regulators earlier, there’s going to be some time for the regulators to get used to obviously the applicability here of AI and the applicability of the data that comes out of some of these models as well that will support the actual submissions by the traditional life science pharma companies for their new drugs. And the regulators need to become very comfortable that and they need to provide guidance as well with regards to what’s relevant here, where does that … what they call provenance of data. Because they need to link the data back to the actual initial results. So how is that going to work when it comes to these models that are hopefully going to, in the future, be able to expedite the process to be able to maybe simulate some of this. But again, it’s going to need governance and it’s going to need that provenance, if you like.

Mike Vizard: Will we see more collaboration in the healthcare sector because we will have these AI models and while they take time and effort to do the research around the model itself, but it seems like there’s always been a concern about redundant efforts between organizations and so can we marshal our resources more effectively?

Robert Poolman: Simple answer, yet again, it’s got to be a big yes to that. I think you are already starting to see it. If you look at particular areas around, for example, drug safety and what we call pharmacovigilance, which is where you look at the adverse effects that a drug … the side effects, as we call it, of a particular drug [inaudible 00:09:45] take them. And what we’re starting to see there is the industry come together and look at how they can leverage their data across the board so that they can learn from each other and learn from the particular classical side effects that you would see if you target a particular drug target for a particular disease.
So the answer to your question is it’s already happening to some degree. And I think it will happen even more as we go forward from here in particular areas. Obviously there’s always that competitive nature of the industry. They’re a business at the end of the day, so there will be some degree of them holding onto some of that proprietary and novel data that sits within just their IP, if you like. But there are definitely opportunities for the industry to come together. And I think what we’ve started to see is a consortia as well, where you get groups of companies come together and really focus in on a particular area. And I mentioned that drug safety and pharmacovigilance is a good example of that.

Mike Vizard: And a lot of that effort is going to be sometimes … It’s just non-differentiated value. It’s not necessarily something that is going to be the thing that you’re going to sell the profit. It’s just an enabling tech that everybody needs to do, so why not work on it together?

Robert Poolman: Right. Exactly.

Mike Vizard: All right. Do you have any advice for folks in terms of best practices for how to go forward here? I mean, a lot of organizations are looking at this and they’re excited, but it’s not clear to me they know where to start. Sometimes they’re intimidated by the whole thing. So what have you seen people do that’s working?

Robert Poolman: So I would say a few things there, which is one, identify where the AI or the generative AI can be most applicable. If you look across the industry and you look at where it’s been applied to date, a lot of that is in what I call more of the operations side of things. It’s making things more efficient, in effect. And that’s not just within life sciences, that’s across many industries. So there’s opportunities there, and they’re what I call the lower hanging fruit. They’re not easy, don’t get me wrong please, but they are the lower hanging fruit. And that’s where I think you can make progress relatively quick. I think what’s going to be critical for the future though, in terms of the real power of AI and generative AI specifically, is that data. We talked about it earlier. It’s making sure that you have the data in the right format in the right place.
And also it’s that … go back to that quality message again. It’s that quality data in equals quality data out piece. What we could face if we’re not careful is a topic of what we call bias, which is where … For example, most of the publications, the literature that comes out from our life science customers and the industry in terms of patent documents as well, this is where things are successful, where things have worked. What we don’t often see is obviously the negative data where things fail and things don’t work. And for any model, any large language model or any artificial intelligence based model, it needs to be trained on data that’s non-biased. So it has to have that negative data as well as that positive data. And that’s something that the industry’s going to have to become more comfortable with in terms of being able to train their models on both data sets.
And actually to your point before, your question about coming together around sharing some of that data that … the failures, if you like, the stuff that doesn’t work. Because that’s where we will all benefit. One, as an industry, but two, as a society. And also, at the end of the day, every life science company really has a key mantra, which is really to make patients better, patient lives better. So this is only going to come to fruition if there is that collaboration and there is that propensity and that willingness, if you like, to share that data and to be able to learn from each other in terms of things that haven’t worked in the past.

Mike Vizard: We talked a lot about how AI will be used for clinical stuff, but a lot of folks, they just find the whole healthcare experience frustrating. The workflows are somewhat complex and no two groups seem to use the same nomenclature to describe anything. And you move from one agency to the other and you start all over again. And a lot of people just spend a lot of time filling out either mobile applications or paperwork or whatever we happen to be. Can the whole experience just get better? Can I have my own digital health care buddy that will just walk me through all this stuff?

Robert Poolman: Yeah, listen, that’s a great question and I think the answer is it’s coming. I think we’re already starting to see good examples of this. And I’ll give you an example. In the health care space actually [inaudible 00:14:14] more at a general practitioner level, your doctor that you would go to see. They’re already starting to … There’s already applications out there that they can use to actually start to take some of that data, save some of that interaction with you as a patient to be able to bring that into a particular model, to learn from that, and then leverage that model as well in terms of what we call electronic health records data as well, so that you can see more … an app, if you like. A model that can actually help the practitioner, the doctor, actually really leverage the data that comes from the interaction with you as a patient, but also from the data that sits within an electronic health record.
So they’ve already got their little assistant, if you like, or it’s coming anyway very soon. As a consumer like yourself, Mike and myself, I think we’re already starting to see the use of … If you take ChatGPT, just as an example, the use of ChatGPT is something that will become more and more prehensive across all different types of data. It’s already there to some degree. And that’s going to help you and me as a consumer, as we take on that data and as we become more … Well, we already are inquisitive about particular health remedies or particular indications, et cetera, and you are looking for that type of data on the internet, for example. It’s making sure that the data that you are receiving there is accurate, if you like. And these type of models, as long as they’re trained on that right data, can really help you with that.

Mike Vizard: Do we need to pay more attention to security in this conversation in the health care sector? There’s a lot of sensitive data floating around.

Robert Poolman: Absolutely.

Mike Vizard: What is the state of the art in that regard?

Robert Poolman: Yeah, listen, let me comment from a Clarivate perspective because we take this extremely serious. Okay, so first of all, all of the data that we are taking … using within our particular large language models here that are feeding our new and enhanced search tool is all from our own internal data sets, first of all. So we’re making sure that we’re bringing the right data in at the right time. I go back to my mantra of quality data in, quality data out. The second piece there is that that is within a secure, closed environment. And these are things that are going to be critical as we go forward from here.
And I think if you look at patient data, your data, my data, for example, there’s a real awareness in the industry about the importance of that with regulations from governments, et cetera. So there’s already restrictions out there to some degree. I would say pretty much all companies are very much aware of that and very much have to obviously comply because there’s obviously significant retribution if you don’t. So I think it’s already there to some degree. And I think most companies now are taking this extremely seriously. And again, we definitely are it Clarivate, given what we’ve done with our data.

Mike Vizard: All right folks, well, you heard it here. There’s much to be gained when it comes to AI and health care and cross your fingers, we’ll all be around to see how it all turns out at the end of the day. But I, for one, am a lot more optimistic. There are causes for concern, but as always, we’ll figure out ways to address it. Hey Robert, thanks for being on the show.

Robert Poolman: Thank you, Michael, it’s a pleasure.

Mike Vizard: Thank you all for watching the latest edition of the Techstrong.ai series. You can find this episode and others on our website. We invite you to check them all out. Until then, we’ll see you next time.