Fulfilling Big Data’s Promise of Providing a Granular Picture of a Patient’s Journey
Atlanta-based OMNY Health has expanded its data network built for real-world data insights at scale by adding more than 300 clinical assessment measures. The company said these measures provide a more granular, real-world view of patient health, enhancing the ability to monitor disease progression and treatment impact. They build on the company’s data strategy in dermatology, where it first demonstrated the value of leveraging disease-specific measures.
Healthcare Innovation recently spoke with OMNY CEO and co-founder Mitesh Rao, M.D., along with Alvin Liu, M.D., inaugural director of the James P. Gills Jr. MD and Heather Gills Artificial Intelligence Innovation Center at Johns Hopkins Medicine.
Healthcare Innovation: Dr. Rao, could you talk about your background and the founding of OMNY?
Rao: I’m an emergency medicine physician. I spent most of my career as a health system executive. Prior to starting OMNY, I led safety and quality at Stanford, where I'm still on faculty, and before that I was at Northwestern, where I led innovation. I have always been focused on the data side, in particular how we can do research across industries. I kept having pharma, med device companies, early AI companies all wanting to partner on really deep, best-in-class information, which we were generating within our four walls, but there wasn't a way to bring it out at scale.
The consistent challenge that we have in healthcare is that in the incumbent tech infrastructure, the EMRs, the way we store data doesn't really allow for data to move between constituents and to really enable collaboration. So we built OMNY with an idea of creating infrastructure to be able to connect that data on a national scale. We always joke that the easy part is getting access to the data. The hard part is making that data usable.
HCI: Does OMNY use a federated model, where the data stays where it is and you send questions to it, or is it building a repository? Or is it something different?
Rao: We actually physically allow data to flow between constituents after it's been made secure, compliant, de-identified, and tokenized. I always crack the joke that if we were selling cookies, what I get out of a health system is the flour. It takes a lot to really turn it into something usable, right? Once that data is curated, de-identified, secure and compliant, and protects patients on both privacy and the compliance side, then we make that foundational data available. We effectively serve as a common language of collaboration between large health system providers as well as the broader world of biotech. So when you think about rare disease research, you think about safety and efficacy research, where that type of not just timely and deep data, but comprehensive information can really be a game changer.
HCI: Dr. Liu, could you talk about your role leading this new AI innovation center at Johns Hopkins?
Liu: I'm a retinal surgeon by training. I did the vast majority of my clinical training at Johns Hopkins and I stayed on as faculty afterwards. Right now at Hopkins, I have three different roles. The first one is seeing patients. Second, I'm heavily involved in translational AI research, specifically when it comes to ophthalmology. The AI innovation center that you mentioned became possible from a very generous donation of $10 million a few months ago. This is the first endowed AI center at the Johns Hopkins School of Medicine. My third role is involvement in AI on the health system level. In that role, my purview goes beyond ophthalmology and involves really everything AI-related.
A few months ago, the leadership at Johns Hopkins Medicine recognized the need for a more formal governance structure, so they put together a leadership team of eight people across the entire health system for everything AI-related, both clinical and operational, and I'm one of the eight.
HCI: OMNY made this announcement that it has expanded its network to include many more clinical assessment measures. Dr. Rao, could you talk about the significance of that and what that might lead to?
Rao: A lot of the real insight in healthcare data is often buried in the unstructured data. It’s buried in clinical notes. It's buried deep within the repositories of information where it's very difficult to actually curate and find. OMNY has gone that additional mile now and for all these therapeutic areas, we're actually curating out disease-specific measures that include things like clinical severity indicators, surveys, questionnaires. An example here is in the inflammatory bowel space with the Crohn's disease activity index. Or the Harvey Bradshaw index for gastroenterology. It’s very specific. What that allows you to do is not just understand clinical staging and outcomes on a very deep level within a patient journey, but also understand disease progression. We can take this messy, real-world data, and actually distill it down into key insights that can lead to the next generation of drug discovery, device development, you name it. This is the first time you're really going to see that level of scale and depth tied to such a large national repository of data.
HCI: We’ve always heard that there's all this good information in the unstructured data in the EHR, but it was hard to extract. So what’s the key to making it happen?
Rao: A lot of blood, sweat and tears, and then, frankly, access at scale. So earlier this year, we announced that our repository hit about 4 billion unstructured notes and that continues to scale exponentially on our side. Because we've done the hard work of pulling together all those data, going through and effectively curating it, distilling it down and pulling those pieces out, we’re finally at a stage now where a lot of the promise of data technology is starting to actually show. Historically, you'd need thousands of chart abstractors sitting there manually combing through data trying to find things. Now you have large language models that can actually do some of this at scale and can enable finding those pieces. So some of it is technology, some of it is timing, but frankly, also it's just the right alignment of both a national network as well as the right volume of data behind it.
HCI: From your perspective, Dr. Liu, could you talk about how getting access to the more specific clinical assessment measures could help researchers at Johns Hopkins?
HCI: Sure. My main focus is in artificial intelligence, which, of course, is heavily dependent on the quality of data. Without good data, whatever AI you build on top of it is complete useless. And if you look at the trajectory of big data, people have been trying to do that for a very long time, and it's never been satisfactory in two different ways. One is something that Mitesh mentioned, which is the structured versus unstructured data. Before the rise of large language models, the only useful data was structured data, so that was a fundamental limitation. Now Mitesh is able to leverage large language models to really unlock that data source. That's great. The second sort of limitation, if you look at most of the existing big data sets, they are either very deep or very broad. And of course, you can get different kinds of insights depending on which one you're going for, but you're really missing out either way. So I think what's really exciting about this data set is that it's both broad and deep.
If I were to summarize what this kind of data set could do, it is to provide a very granular patient journey on a patient level, both in terms of clinical outcome — what happened to them — the quality of life, and also the cost journey of a particular patient. If you are a clinical researcher, you care about clinical outcomes. It is also going to be very helpful if you are a healthcare executive who cares about outcomes, but also cares about costs. The combination of costs and outcomes on an individual level, in a longitudinal manner, it just makes it so powerful.
HCI: Dr. Rao, could you talk about how you built this big network for health system data? Are the health systems themselves customers of the data, as well as as providing access to the data?
Rao: I mentioned that I started OMNY out of frustration, because most organizations that are in the data space are either focused on providing some type of solution on top of data, or, frankly, trying to get access to it. And historically, we had companies come through that offer platforms, but on the back end they take rights to the data. What I wanted was to create a more democratic approach, where everyone could be a stakeholder by thinking of it as pure infrastructure. So we designed this to really be an empowerment platform to help any organization that was connected to data, that was accessing data, that was storing data, to be able to bring that forward. Today, the platform is over 85 million patients and will be north of 100 million patients in relatively short order. Many of the largest nonprofit academic medical centers all over the country leverage the data layer to be their foundational front door for data.
