Thank you, everybody, for joining us today.Eva and I were waiting to see when we would be live, so you got to see a little bit of sausage making as well, and thank you for that. Let me introduce myself, and then I'll introduce our speaker today, Eva Kligsek. I am Reina Srivastava. I am based in San Francisco. Eva and I have collaborated previously as well. Eva is an expert in all things AI and health care. And if you have joined the session, you have an interest in figuring out, number one, you know, all of the AI information explosion that is happening. Two, you also know that there is a problem with health care and has been for a very long time. How is AI helping to solve that?

And at the same time, what are the risks that we need to be aware of as we move on to this fresh new world of more data that AI is bringing in and make sure we do it the right way? So welcome, Meeba. And I have my question for you. And I would love for everyone to be as interactive as possible. I'll be keeping an eye on all the questions as they come in, so let's make sure we get them answered. So, Iva, let's start with the big picture. Why is health care data still so messy, and what makes it so hard to

fix? Well, wonderful question. of all, I welcome everyone. Welcome to all of the women from all over the world joining us for I see one Polish name, so, and hello everyone, those joining us here. of all, it is a quite an exciting time. It seems like every day something changes when it comes to, AI and how we are applying this. And it is extremely interesting, especially in the health care arena, because we are in a very regulated industry. So where everyone is talking AI, there is a lot of discussion about the AI, but it's also a discussion how do we apply it in a very responsible and ethical way so we can really help, our patients and our, and and everyone, really utilize it and and get the benefits of the of the great tools.

So let's start with the data. of all, the health care data is extremely messy even right now, and there is a lot of work that is being put in to try to put it in a format that is standardized, that is harmonized, then we can make sense of it. The reason being why the data is in such a different shape that many of the consumer data, that is that is coming in from sales is because every dataset had a different purpose in use. Hospital produce their data. Pharmacies produce their data. Insurance produce their data. All of the lab, labs produce the data, and they are all having a very specific, purpose in addressing the part of the health care that we are working with. So, so while we have a lot of data, unfortunately, it also comes in in many different shapes and forms, and it's not harmonized.

And for the longest time, there was no way for us in health care to join the data until recently.

So

with AI, we there is a lot of work being done on how do we harmonize, structure, curate, and protect the information that is in the in the health care data to make sure that we can get the best insights. Cybersecurity, it has evolved. Operability is improving. But there is also one component that we really don't talk about it very often, but it's the organizational changes that needs to happen in order to be able to, to get the data organized. For example, it has to be cultural. In health care, we all grew up in a place where technology is separate, analytics is separate, research organizations are separate, clinicians are separate, lab technician, are separate, and there is the always looking regulatory and legal component of that whole system. And I think it does require cultural change knowing that everyone has to work together through proper tools.

And I think AI is the one that it helps us integrating and linking the data in appropriate way while preserving privacy and also ultimately understanding what to do with it. We in health care trying to help our patients. So if we cannot use the data to improve patient outcomes, then we have to think about it. What is the reason for it and why we're so far and why can't

So, you know, a lot of things that I heard are immensely interesting, I believe, you know, to overall everybody and to our audios in particular. Lots of data, lots of places, and a way to kind of organize that while keeping that data safe. So, Piro, clearly, you've had a lot of experience with this kind of data, and you worked closely with it. You worked with EHRs. You worked with digital data. I'm sure you're seeing something that is particularly exciting and interesting around what is something that has caught your attention in this transformation.

Yeah. Wonderful question. So of all, we are producing a lot of data, and this is what is so exciting about, but also as as as we were just discussing the challenge on bringing the data together. With AI, we are able to much more seemingly, and seamlessly, integrate the information. Creating linkage between the disparate, files or different data sources that are created, in the health care system is quite laborious. So with AI, we can really improve that process. More importantly, we can actually help with curating the data in a safe way. If you think about if any of our viewers here think about, who is the when you're going to a physician office, physician is typing in or the nurse is typing in all of your information, and they are putting information from the discussions they are having.

There's a lot of very rich information in EMR that that needs to be taken in a best way possible, in order to provide the insights. So AI is helping us bring that information forward, help us organize across the different data sources, and make sure that we, improve the harmonization. No one talks about it, but the same variables are showing up in many different places. Talk of putting it in a structured way so we can really, as analytics, professionals can make sense of it, but also allowing us to then disseminate the information in the best way possible. Claims data, great tools. EMR data, great tools. All of the lab data provides a lot of information, very rich information that without the proper tuning and understanding of the data, it's very difficult to make sense of it and ultimately use it to improve patient outcomes.

So a lot are happening, a lot of great, applications to this, and I'm and I think we always have to think about it. What's the purpose? How we organize it? How we scale up? And, ultimately, how do we preserve privacy as we are trying to improve patients' outcomes?

Which is a fantastic summary, and thank you for that, Eva. So I'm hearing cleanup. I'm hearing automation. I'm hearing protecting, you know, privacy while we're doing all of that. You know, something that is on my mind, and, you know, I'm sure the audience has a question as well. AI and automation sound amazing. They sound very promising and, but they also sound complex. And I was curious, what technologies do you think are truly worth it? They're truly something that is gonna make a difference versus all the buzzer, hype words that we don't typically hear about.

Yeah. Great word. Yeah. I mean, great question, quite frankly. We all know that AI is everywhere, and everyone is talking about it. And you can see that my even my, little tagline here is AI doesn't replace you. It it elevates you. But so everyone is with it and everyone has to come and join the board. As I am thinking from where I'm sitting, all of the large learning models, so so the generative AI, the ungenting AI, all of those tools are actually very helpful when it comes to getting the insights and operationalizing various very difficult processes that we have when it comes to data integration, abstraction, and ultimately ensuring data quality and and patient privacy.

Generative AI is for those of you who who do not know necessarily what it is, is really helping us, create prompts that we can then apply it to the data and get the appropriate information out. Very often in the notes, we are looking at, the the drug use that that the patients have when they stopped when they started. We are looking for information about some, test outcomes or some even qualitative components that physicians are putting in, like alcohol intake or our medical history, family history. Those of us who are coming from different countries which are across the different worlds, we have different way of style of living, style of food. So all of that is being documented, but often not using analytics to derive patient outcomes during clinical trials and so All of that can be very easily translated and created very informative, datasets. The other pieces, all of the the the different we are moving into world now of small language models because they are cheaper, but also can be much more is powerful when it comes to training for a very, very specific and very technical and medically informed, I guess, concepts that we require in health care.

So adopting those that these are becoming, not yet as used, but they are becoming more and more used now within the within the research as a result of the fact that we, you are really can focus on very specific components that you need. For those of you who do not know, those applications are allowing us to do much better clinical trials. All of that, the optimization of clinical trials, identifying where the patients are, identifying the sites that are important, and also at the end, driving all of the insights. of all, with all of the developments that are happening in this, applying wearables, applying all of the new toolings, applying all of the copilots, we are also able to then think about it. What is the future for them? How do we ensure that we can leverage the data that we have and create new ways of looking at at at clinical trials? One of the examples, I don't know how many of you have heard, Regeneron recently or yesterday, actually, finalized the the the purchase of '23 andMe.

Probably many of you have heard that they have a lot of genetic data. I actually think it's probably one of the best applications because it will allow and help hopefully in a very safe way move clinical trials forward. If you apply the right tooling and you create digital twins, which we all hear about, you can upload all of your information, and they can create the you. But the digital twin from the point of view genetics is very important for advancement of clinical trials and pretesting and identifying the right patients and the right, potential drugs and the and and the dosing as a result of leveraging the information and mining the data in the best way and the fastest way possible.

So this is such an interesting time and where we can be talking about whether you're using tuning that ChargeGPT has OpenAI and Gemini and all the Copilot by Microsoft and so and all of the other companies creating tools on top of this. I think the most important is to want to understand, what is your question, where are you gonna be really applying this, and what is the ultimate outcome that you're gonna improve. For me, as I was thinking about the patient and thinking about it, what can we do to move patients towards more healthy lives, get the better outcome for for survival, and ensure that we protect as much as we can with a very ethically sound frameworks, all of our privacy, and and well-being at the end.

I love that. I love that. And, you know, something that totally caught my attention, I'm sure it caught your audience's attention as well, is a digital twin, but in health care. You know? Who who have heard about the digital twin? Like, you know, I wish I had a twin twenty ten years ago.

I was supposed

to get my work done. But digital twin is something that can improve my health. It can improve prediction. It can improve analytics, and it can help others behind me, but at the same time, while keeping the data safe. I'm thoroughly excited by this idea. Now tell me, Eva, and, you know, we do have one question from, Georgia. So I'm going to go on to Georgia's question and I'll come back to mine. So, Georgia Thorne, thank you for chiming in. How do you ensure that when the data is harmonized, it is kept secure, and there is no data loss? We're all concerned about data privacy, and Georgia's question speaks right to that.

Oh, it's a it's a wonderful question. Now of all, every company needs to invest as much money as possible in the security, but it's also data governance and data provenance. So the five the primary aspect is the provenance and, understanding what data is, how are you making changes to it, recording everything. We are in a highly regulated industry, so every change needs to be high be be documented the proper way. And, also, as you're thinking about, anonymized data, and those who have actually PHI in it, you have to preserve it and you have to store it and create security levels where only certain individuals can access. It shouldn't be especially as we are in a, many companies are start ups, having that protection and making sure that everyone understands that this is your own data, this is your mother's data, this is your friend's data.

That data is extremely needs to be protected with the highest quality of it and the highest importance. It's very important. The thing which I which is, you know, we're talking about harmonization and bringing the information in. It's it's a it's and structuring it. We all know that the data is very often extremely unstructured. But defining exactly how, how the different units, how the different, terminologies I work with a lot of data across multiple health systems. Believe it or not, every health system have their own way of presenting the data or way of coding the data. So it's a really big aspect of creating a uniform way of pulling that information out out of there and storing it in a way that then you can connect all of the different pieces together.

AI is helping us really with that component and helping us create the the ability to understand what is different between one health, health source versus another, how each side or each health care system, provides that data. But it also allows us to understand what is the leakage when it comes to PHI. It's understands to to track that information and and ensure there is there is broadening in a way that we are not causing any harm. One thing I wanna mention is the more data you combine, the more data you bring in, you are increasing the ability to de identify a patient. If you are in a place where you can hold PHI, it's great. But if you are in a place where you cannot, you have to work with experts who can tell you what to do with data privacy, and how do you need to store it, how do you secure it, and what are the newest technologies to ensure that there is no offensive there is no, hacking when it comes to your your services.

And the final thing is your IT technology individuals need to be testing your systems continuously with with the with the AI technology. Was were very helpful. It can also cause a big issue, moving forward for us.

You know, Eva, what I find so fascinating, and I just put a poll in for our audience in this, is that, you know, you are from health care. I'm from tech. But all of these ideas, all of these needs are becoming so universal across all these different industries. The human plus AI collaboration, the fact that there is data in different formats, in different ways, across different companies, that is such a common thing in tech, in financial services, in payments, and in health care. So the whole idea that AI can help, I believe, is truly so universal. And I'm very glad that, look, you know, we are taking this beyond just, AI will replace us all to AI will help us all. So, Ewa, we are the last four minutes, and this has been fascinating.

What do you think is a real world example of, AI in here that you've seen work particularly well? I'm sure you've had a situation where, you know, something just clicked. What was that?

Oh, absolutely. There is there is so many of them, but I think I wanna just go, talk about, from the operational component. Because we we work with real world data, which is extremely important for optimizing clinical trials and providing insights and providing the remote evidence, it's also very important to ensure that AI and the human in the loop work together. And we didn't talk here about the human in the loop. But in the health care is the is the individual who is the expert in the medical field or the expert in a particular therapeutic area who can help, validate the outcomes, the output of the AI models, as well as can help fine tune them so they can produce the appropriate, insights.

As we are thinking about where where we are headed in the future, and for those of you who are younger, who are listening to this, I would say upscale yourself in a way that you understand what what AI does. But, also, I was just reading before the session article, a great article that is not only understanding AI and being technique savvy. It's understanding the data. Understand what are the pitfalls of the data, how data trends are changing, how the models that are underlying, the AI can ultimately drift. If you have that structure and that baseline for yourself and then you bring the the law, which is lacking behind us, you understand the regulatory aspects of this, and you can work across, many different individuals in a, cross functional team, I think you can be very successful.

Because just knowing the technology, I mean, we all can, I you know, I don't code anymore as I not not necessarily, but I can quickly go in and get that from ChargeGPT a great code? But if I don't understand how it works, I'm gonna apply that and ultimately not get the right outcome. So understanding the technicality of it, understanding the foundational aspect, and also learn what the data is, how the data can drift, what the data collects, what doesn't collect, where the biases. We, in clinical trial, think about it, the the representation, because we all know that representation is very it's it's not very good, or it's not representing the communities. So if you are projecting or predicting something to a minority group that are not represented in your data, you probably need to think about it some other ways of applying models or getting other data that can inform some of your decisions. So we talked a lot about AI, but I do wanna mention that AI will not be successful in a vacuum, and it's not a learn it and leave it and let it run.

It's AI is the human in the loop that is extremely important to be successful for the future.

I think that's an amazing insight for everybody. Knowing that there are so many headlines around, like I was saying, AI will replace us, the CEO panic, you know, the layoff announcements, etcetera, etcetera. I just saw something on another conference where the spend has gone down to zero because of AI in the picture. I believe your insight is far more important that it's important to upscale. It's important to learn. It's important to kind of keep the human in the loop because otherwise, AI is not successful in a vacuum. Eva, this was wonderful. You know, the twenty minutes have gone by way too fast. If you guys have any questions, we're more than happy to answer them. Please find us on LinkedIn.

We're happy to continue the conversation. And, well, thank you, Aviva, for joining today.

Thank you, everyone.