Solving Health Disparities With Tech

Transcript

Bryce

Chief technology officer and co-founder at Truveta. Jay has over 25 years of experience in the technology industry, recently serving as a technical fellow and corporate vice president at Microsoft. Leading engineering efforts in fraud protection, financial services, and supply chain.

Before that, he was the GM of Bing Mobile Engineering, where he led the development of Bing Mobile’s apps and platforms. Here is a little bit more about Truveta. Truveta aims to improve healthcare through data analytics.

The company has built an AI model called the Truveta Language Model to extract insights from electronic health records. As CTO, Jay leads Truveta’s engineering team and is responsible for the company’s technology strategy. Under his leadership, Truveta has made significant advancements in analyzing healthcare data using AI.

Jay, we are so excited to have you on the show today to discuss your journey in healthcare and learn a little bit more about your views on eliminating disparities using data. How are you today?

Jay  

Good. Thanks, Bryce. Thanks for having me here.

Bryce

Yes, sir. I appreciate your time. And going through the introduction, I see your background is impressive.

Having that experience with Microsoft and Bing Mobile and two huge companies in the U.S., I am so excited to learn more about Truveta and what you guys are doing. Sure. So, getting into the beginning of this journey, Jay, you’ve had an incredible career, as I mentioned, spanning 25 years in technology and key roles at Microsoft and Bing.

Could you share that pivotal moment from your career that motivated you to transform and transition into healthcare technology and ultimately find Truveta?

Jay

Yeah, Bryce. I think it starts back to five years before Truveta probably formed. And like any other family, our family had a fair share of health issues ranging from cancer to rare diseases.

And with the two immediate family members, my dad and my wife’s dad, dying of cancer. And the last one to go was my dad in January of 2020. I was thinking of taking care of my loved ones and trying to research the best path to take forward.

And what is the best clinical research out there available? Are we giving the best care for them? I realized that there are a lot of opportunities in healthcare.

That’s what triggered my interest in looking into this. I strongly believe in data platforms because, as you know, any data platform can lead to good analytics and AI, and your AI is as good as data. And I always had this belief that data saves lives. And when Terry called me in March of 2020, that intrigued me.

And one thing led to another. We founded Truveta in September 2020 with a phenomenal vision; I would say, from my point of view, saving lives with the data because that’s exactly what every Truvetan believes in.

Bryce

It’s such a necessary need and something that can easily get kind of overlooked. It’s maybe not the sexy study; it’s the studying data every day. It’s not something everybody wants to do, but the outcomes it can help achieve are incredible.

Healthcare is a complex industry. Could you share some specific challenges you guys faced at Truveta and how you overcame them initially?

Jay

Yeah, I think, as you can see, we founded Truveta in September 2020, which is the peak time of COVID-19. So, society sought good answers during the pandemic to handle the situation. What we have quickly realized is whether it is policymakers, data providers, or whoever it is, they’ll quickly realize that to be able to solve this problem, they need fresh and real-time data because, at the time in the market, the best you could do is probably six months old data.

And if you pressed for more, probably three months old data, but the pandemic was moving so fast, you needed fresh, real-time data and precise data. And because of this, what we looked at is, hey, what is the best way to get this kind of real-time data so that Truveta can provide yesterday’s care that is available today for the researchers so we can understand the trends that are going, what exactly the policymakers can do, what health systems can do to better prepare for what’s going to come tomorrow. And you know that when you want to process the data at this one-day cadence with the massive amounts of data, healthcare was used to processing this manually using manual processes.

And there’s no way you cannot make the data available with the data coming from the desperate, disparate kind of EHR systems. And that’s what we saw: hey, we need AI’s help. Do you know what I mean? To solve this problem and get this data in a way that a researcher can consume,

And this data is available for whether you call cities, government entities, healthcare providers, and everybody, even researchers, so that we can see, we can understand, and save these millions of patients in this. So that’s exactly what was the trigger for looking at it. So that’s why we did not want to be yet another data provider that does not offer any kind of differentiation.

We saw that this is needed for the industry now and in the future.

Bryce  

For sure. And getting into healthcare disparities and the social detriments of health again, I know Trivetta aims to improve healthcare through data analytics. How does today’s language model specifically address healthcare disparities, and what impacts have you seen so far?

Have you gotten any type of feedback yet?

Jay

Yeah, for starters, let me quickly explain what a language model is, right? You know what I mean? A language model is an AI construct that mimics how humans understand or comprehend language.

So what it does is the AI learns this skill by looking at several thousands or millions of human language artifacts. Okay, and now, with this knowledge, when it looks at a new document, it can comprehend concepts like, okay, this person is talking about a name. This is a movie theater.

This is a place. This is a city, right? So, that’s exactly what a language model does.

Now you extrapolate it. Trivetta tries to specialize in medical data by using this language model. So we make our language model run through a bunch of de-identified patient artifacts, whether it’s all the diagnosis as medications, doctor notes, radiology notes, and everything so that now it can start to understand these medical documents that have very high accuracy, whether it is your diagnosis, medications, and everything.

Before we came up with why we thought this was needed, I explained previously right that AI is needed to solve this in an automated fashion with high accuracy. At the time, if you look at what was available for most researchers, it was claims or your medical billing data. Your medical billing data says nothing about your diagnosis or the doctor’s feelings. So what are the side effects, or what medications did you take, and things like that?

So that’s why some AI is used here, specifically in medical coding, but it is always optimized for billing purposes. So because of this, it does not carry the clinical accuracy that a researcher needs to be able to perform their medical research, okay? So that’s where we focused quite a lot.

We apply AI for this structured, semi-structured, and unstructured data to provide an unbiased view. If a researcher goes to a single hospital claims data and tries to do research based on how that hospital system processes things, they could be biased in the data. So, the outcomes you’re driving for will have a bias.

So that’s why we wanted to ensure that when we discover things for a population, we have the kind of available signals that carry the clinical accuracy and the medical disparities, and you think about it. And that’s what makes it powerful. So, the researcher can look at this data as a whole, in a 360-degree view, and analyze it in different dimensions.

And for that, you need a bunch of capabilities in the platform and a whole nine yards. So that’s exactly what we bring to the researcher. So, the researcher does not need to worry about the cleanliness of the data, how structured it is, the bias of the data, and whether they can process this million of our patient records in a secure fashion.

That’s where we come in, and it solves several problems. I can talk for the whole day about how talking to a licensed or pharma customer or a medical researcher opens the door for the enormous amount of clinical research that can drive the wellness of patients.

Bryce

Right. Yeah, and I’ve, and, interestingly, you touched on AI and medical billing. Cause I had, that’s all I’ve ever seen it for.

I have some experience with a company called Expert Docs, and they did AI medical billing. And you’re 100% correct where you don’t see AI being delivered or used in the outcome-based or research-based way or how we can make things better, not how we can be more efficient on how we charge our patients. So that is amazing.

That is an amazing accomplishment that Truvett has overcome. Kind of getting into the impact of technology, kind of the health equity, how do you see technology playing a crucial role in achieving health equity? Maybe more specifically in underserved communities and communities that don’t have access that some other communities do.

Jay

Yeah, I think if I have to go to the fundamental issue here, probably for any researcher to be able to research in a dimension, as I said, whether it’s underserved community or underserved population, what are the dimensions of the available data? It instructs them on the correlation or causality of the data to this. So, most of the time, I think the inequity in this research comes from researchers being unaware of this data in the data platform.

So I say that because health systems or wherever the data comes from, they’re not transmitting this data to a researcher. That’s what I think the first problem is. And the second thing is, you can see that most of the time, it is not just the treatment, right?

There are other factors. Do you know what I mean? The non-medical factor is the environment. Do you know what I mean? You can see a lot of previous articles about lead poisoning or lead exposure to people, what caused some kind of medical issues, and things like, if you go in such kind of cases, you see that a drug or treatment may not be the focus, it is the other environmental factors could be a focus.

In this way, you can look at the causality and the correlation of the data and see other factors that can help, either by helping the patients or even preventing them from having these things. That’s the second point. And if I think third, and now that we know how to mitigate, we can give this data, and researchers come in, now any research is expensive from the researcher’s point of view if a licensed pharma is running it too, right?

So the treatments could be very expensive, okay? Where a specific population cannot afford it, you know what I mean? That’s when you know what I mean.

We think that the government or a nonprofit is not solving that problem. Where can a pharma make this treatment available for underserved communities and things like that? But the fact that this particular treatment option helps this condition, then the government or policymakers and nonprofit organizations like the Gates Foundation and people like that can step in and help, okay? So that’s how we think because, in the end, I know that because I’m a big believer in data; data drives knowledge. The knowledge gives the power to help not only one community, you know, kind of diverse communities and their problem to be able to go from, you know, how to help to even prevent.

This is where we have seen some customers looking at our data. We get our SDH data, very deep SDH data, from our partnership with LexisNexis, and things like that; companies like Pfizer, Moderna, Boston Scientific, and even recently CDC is also using the Truetta data to be able to do; you know, that kind of studies that they could not have done before because we take these SDH data and combine with the de-inferred patient record with high accuracy so that now they can understand the correlations and causalities. They can analyze this data in multiple dimensions they could have never dreamt of.

Bryce  

And it’s just; it’s the adage where it’s like, you don’t know what you don’t know type situation where, you know, you might have the very best, you know, the very best reasons you’re trying to do something or the very best intentions is what I was looking for. But if you just don’t have that data to understand it or to understand how to overcome whatever that data is showing, it is so key. Getting, I know, you know, you were back in your past as somebody who was deeply involved in fraud protection, financial services, and supply chain technology at Microsoft.

How has that expertise translated into creating some solutions that, as we just discussed, bridge those disparities and help improve quality care?

Jay  

Wow, this question is so dear to my heart because many people ask me this: hey, how did your, always our experience instructs, you know, our thinking currently, right? So I would say that because I always, whether in Bing, whether in, you know, the financial systems fraud protection and supply chain, I was exposed to millions and millions of, you know, data points, okay? And how can you create these outcomes that could be, you know, better than when you research them in silos, okay?

So, the first thing I would say is it taught me the power of data gravity platforms. That is when you bring data from multiple parties, and these multiple parties are automatically incentivized by the platform’s power to give more data, okay? That’s what we call consortium principles. Do you know what I mean?

I know I can rattle off, but, you know, first of all, any consortium has the data you need to have, the first thing that you need to build is a network effect where the value of the consortium grows as more parties come in. That is, N plus one is greater than N, the value you are getting out of, you know, the party. The second thing is approved use.

This is one of the biggest things in healthcare. You know what I mean? Do you want to get this consortium data and use it to serve ads on patients?

That’s what we say in True Data Values; we never use it for such purposes. This data is used to solve health equity and healthcare disparities. Do you know what I mean? And patient care and saving lives.

So that’s the second thing. The third thing we want to look at is privacy; security is so important. One of the core principles of the consortium that we wanted to uphold was the interest of patients and businesses.

There are patient interests. For example, when I get a health system record, and there is some way the health systems are doing a particular case or anything, we do not want to; we do want to preserve the privacy of the patient and also the business interests of this so that they can continue to operate in the way they are doing. And the last thing I would say, what I’ve learned in doing this is consistency.

You know what I mean? When people come to, instead of going for a single party, they’re coming to this conglomerate, which is actually aggregating this data; you need consistency in use. And that’s what we do in our data model and through the language model that we consistently process this data to a common model so that regardless of where the data is coming from, which health system, whether it is coming from Florida versus diagonally opposite Seattle, it looks the same.

And even if it’s coming internationally, a researcher can research a patient on any disease anywhere in the world. And these are the things that actually, I could formulate this vision for Truvata in our platform because of the earlier experiences, because we have, you know if you are doing for the first time, you know, for the first timers, you’re there to always be learning, but it helped me start at a place where we have created a, what I say, commercially viable product at Microsoft at scale that helped instruct what are the kind of a common pitfalls that we should avoid and what are the best principles of the consortium that we need to put in place so that we can create this thriving community that’s going to help patients on our Truvata platform.

Bryce

Yeah, for sure. And, you know, getting into how that data is shaping healthcare. Under you, I know Truvata has advanced and has advanced in healthcare data analysis with AI.

Could you provide maybe a couple of examples of some data-driven insights that have positively impacted outcomes that you have seen results in that say, hey, this research has produced X, Y, and Z as far as outcomes that you guys have helped create?

Jay

I think Truvata data and, you know, what we call a Truvata Studio, a platform where researchers come in and do their analysis and everything. We are trying to change, you know, the age-old narrative in the healthcare industry takes almost 17 years. That’s what they say: 17 years from clinical research to patient care.

First of all, we are changing that. One other example is the time when we started the pandemic. You know, you can see that the trends of infectious disease are going so fast that you need a platform that acts much faster and gives this data results back to it.

So, I’ll also look at what is happening with this ongoing respiratory virus. Do you know what I mean? How exactly can you monitor and let the, you know, let the hospitals be more prepared and also be so that we can, how can they be more prepared to treat patients now and understand the trends and things like that? And we have even helped one of the vaccine makers, you know, by being able to provide surveillance of how exactly post-vaccine the side effects are.

How are the things? How are the patients doing? Are they getting COVID again? And things like that.

There are a bunch of things that helped an industry. The second thing I can think about is that we have done a real pulmonary embolism study sponsored by Boston Scientific. This is, you know, comparing the Inarifluorotrivir versus Boston Scientific Zikos devices.

And it was, it is actually, very illuminating. It was shedding some light on the patient outcomes, okay? So doctors can make more informed decisions about which device to use for patients. Do you know what I mean?

And solve this case of embolism. The other thing is that we recently published this GLP-1 comparative effectiveness study. Actually, some of the results that we published will be available.

Of course, there is a study that’s going on, which is called Surmount, which is going to give the study in a year or two from now. But we have given, you know, how exactly patients are happening because we have the most real-time patient data. We’re able to give much ahead of the time.

So this is the fundamental differentiation, Bryce, if I will tell, is because of the near real-timeliness of our platform, thanks to our, you know, 30-plus health systems that are working hand-in-hand to be able to provide this plan, we could not have done this alone. Because if a health system says, hey, I cannot provide you this data on a daily grain, we cannot do this. So it is how they are accommodating, providing this day-to-day data at the daily grain, and how we can process it at the speed of making yesterday’s data available today.

He’s enabling so many cases. We have over 50 customers, you know what I mean, that are leveraging the platform. And I’m not saying we are perfect.

There are places where we find things because a researcher is looking at a specific place, but we can quickly fix them. And the pace at which we can fix them, this researcher is not used to, because they say that when there is a data problem, they say that, oh, come back in six months, I’ll fix it. But Truett, I can fix it.

We can fix it by fixing this telemetry from the health system. We can pick up a phone call to our health system and say, hey, this is where, you know, there are some issues coming with data, and we fix it. And the researcher gets it in a short order.

And that is the differentiation.

Bryce  

So that’s so huge for them to, for these research projects, not to take four or five years. Well, some of them need to and by design, but being able to have that data, turn it around, and get it into the hands of the researchers is so key. I love that.

I know you spoke earlier; considering some of healthcare ethics and data analytics, how does TruettBeta ensure responsible and equitable data capture and use?

Jay  

Yeah, I think this is the second thing that is very dear to my heart because having worked a lot with AI in my past life, and you know, Microsoft has this notion of responsible AI, and I happen to contribute quite a lot in that area. This is such a near and dear topic. When I go to the fundamental kind of building blocks for kind of an ethical use for me, it means very concretely privacy, you know, and also business interests, as I said, in the consortium, the business interest, security, and being unbiased or fair, okay?

So whatever you produce must be fair for the population you’re serving, okay? So, these are the fundamental building blocks. For TruettBeta, the platform is built from day one, and these things are built as capabilities, not as a bolt-on.

We brought the data together and then thought about, oh, how do I ensure privacy? How do I ensure security? We fundamentally talk to each other even before data arrives. What are the principles of the platform that it needs to serve?

Okay, that’s the one thing. Out of these three or four things I mentioned, if you look at it, I’ll go one by one. If we take privacy very seriously, we take it very seriously because this is very sensitive data.

So before data is used for any research purposes, it is de-identified. And this de-identification method is certified by experts. Do you know what I mean? Hey, you provide enough power, or the data utility, without compromising the user’s privacy.

Okay, that’s the number one. We take it very seriously. The next thing is the security aspects and preserving the business interest.

We have a very high bar for security and preserving this business interest. You can see the certifications that we have received on our website. We have ISO and SOC 2. We are going for, you know, establishing our quality management system, and you are going for, you know, high trust.

These are the kinds of things that tell our commitment to upholding the best secure platform. Our investment is because a typical startup looks at them after they make what is called fit to market, and they can do the revenue. Oh, by the way, let me make sure of this.

But that’s not what we did. At the get-go, we started investing in them. And the third one is, yeah, foundation.

And the last one is the AI models, right? We take so much care in our AI models to not introduce bias, okay? We have invested a ton in measuring the quality of the models, our benchmark data set, so AI models can be used in the future. Any revision of the models can uphold our unbiased principles in our app. And all in all, regardless of all these things, we are also very, very, very transparent about how we do things with the industry and the community. And our approach is that we have been very transparent, and I know you know that transparency leads to trust.

Also, our approach of providing these certificates, getting this certification, investing in them, and fully documenting the exact results of the certification on the website. It gives that kind of peace of mind for anybody, either contributing the data to the Twitter platform or leveraging it, that we uphold the ethical principles of using this data. So, yeah.

Yeah.

Bryce  

And kind of wrapping up here, I would, you know, the expertise that you guys focus on and being able to not only just get data into the hands of researchers that are innovating and pushing our healthcare space forward, it’s also, which is one of the biggest thing non-bias in allowing them to do not only their job but their job correctly, their job in a way that is going to be reliable information that they come up with.

And then the security aspect of it, which is so important to so many Americans and people worldwide, is incredible, Jay. This is the last question I have, and I would love to know if you can share it. I know you; some things must be kept close to the chest. Are there any future projects that Srivetta is working on that you can give us a quick glimpse of or a quick preview of?

Jay  

So, as I said, we are driven by our promise towards health equity and being able to provide the best clinical research and do this, not only for the United States but also internationally. Do you know what I mean? So, we are looking at many opportunities because diseases have no boundaries. Why should a platform have boundaries?

Diseases also collaborate. Two diseases coexist in a body. Why can’t two different partners coexist in a platform?

Okay. So these are the things that we fundamentally think, and we are. Unfortunately, there are, I cannot mention names, but you will see that in the next upcoming month or two, we are going to kind of publish some great research that we can do either with academic medical centers or with our, you know, partners and, you know, Lifesense and Pharma that is going to show the power of this data and platform. And let me tell you, every day, every true word they want to see is how my work contributes back to saving lives with the data. And this is what I think: the more I, you know, in two to three years from now, the kind of place we want to be, oh my God, this company lived their vision.

Okay. They have created a significant amount of research in many areas that is helping patients and things like that. It’s a very noble cause also, you know, it’s not a nonprofit; we are doing it for profit, but being able to do it at this pace so that you are removing all these blocks so that people can do it economically so that all these research costs don’t go exponentially high.

That’s another advantage that True Data provides: you don’t need to go to 10 different parties. You can come to one party, and you can do this. Exactly.

So, I know I’m not asking a specific thing. Still, so much will come out on our website, our posts, and other things because we do most of our work with our partners. We combinedly announce this work, but keep a lookout for some of the things that we’re going to announce about the True Data Language Model and how True Data Language Model is going to enable more and more, you know, kind of needed scenarios for patients and providers and pharma license and pharma.

Bryce  

And just as an ending, I know somebody who is a type 1 diabetic. I was diagnosed with juvenile diabetes when I was 10. And I know the research, money, and work going into researching diabetes and helping overcome this challenge.

And just from me personally, not the other millions and millions of diabetics, I thank you, Jay. I thank True Data for allowing these researchers to try to help eliminate some chronic illnesses damaging to this country and the world. So, just a quick little heartfelt thank you from me for helping provide that data, which is, you know, so important and affects people like me and everyday people who might not even realize it.

But that does it directly affect someone like me, where if, you know, you can’t get this data, you can’t get unbiased, concrete, good data. How do you research something? How do you overcome a problem that is plaguing millions and millions and killing millions and millions?

So, thank you so much, Jay, for joining us today. It was great to discuss how data-driven insights can revolutionize the care landscape, especially in promoting health equity and eliminating those healthcare disparities. It requires a robust solution and a robust mind, which you have, and so, so excited to have connected with Truvia and look forward to seeing what’s next from you guys.

And please, everybody listening, please do yourself a favor, check out True Data, and see if they can help your research or if they can help something your company is trying to overcome regarding data-driven analytics. But thank you so much, Jay. Thank you so much for your time, and we look forward to talking with you again soon.

Jay  

Okay. Thanks, Bryce. It was a pleasure.

Bryce 

Yes, sir. Thank you.