9/3/24 - Episode 3 - Q&A Session

Eric Snyder: Hello everyone and welcome to webinar number three. I'm Eric Snyder and I'm the executive director of the Wilmot Technology and Innovation Group here at the University of Rochester Medical Center and the Wilmot Cancer Institute.

So, in our last couple of webinars, we showcased live demos of specific technologies, this month due to some of the conference speeches and the overall busy August. We're shifting gears to a more interactive format essentially a question-and-answer style session covering a wide range of topics. I've gotten quite a number of questions ab out Product Management Leadership strategies and just the overall technology in the future submitted over the last month. So, we're going to tackle some of those. Looking ahead to September we'll return to our technical focus with a live demonstration of the EOS. This one should be really an interesting webinar as it demonstrates some of the technical sophistication that we haven't shown you yet that our team is currently developing.

But before diving into our questions, I'd like to take a minute or two to provide some background on our team, our technology and the future traction that we're taking here. So, because this should help answer or address several of the questions that we received at least at a very high level.

So, our core mission is to support the strategic vision of the Wilmot Cancer Institute from a technical perspective. That means providing robust support to Clinical Operations, Clinical Trials, Nursing, Administration, Research, Shared Resources, all with specific focus on cancer related initiatives. With this amounts to for our team splits us into a couple different domains. We have the typical keep the lights on style stuff that we handle so you're Ad Hoc reporting, your dashboards your basic task technical consultations and these are things that I sometimes talked right about but unfortunately, they don't often get that shiny treatment of the cutting-edge technology we develop. Honestly, though it should, as a large part of why this team is so successful is our ability to handle the keep the lights on type of stuff far more efficiently than it's typically seen within the industry. So, our entire technical makeup is set up to ensure extremely quick turnaround times for any Ad Hoc report request or dashboard request. Our average turnaround is less than a day for this type of stuff often amounting to hours. As for anyone listening from other healthcare groups or if you've ever been in the industry, you know that this is sort of unheard of as the turnaround is usually measured in weeks if not months.

How we do this and the technology behind it which we call the Hyperion Data Manager is actually something we are discussing for another webinar, probably in a few months from now. The Hyperion Data Manager is a little bit harder to show and webinar style format because by its very nature, most of it is PHI.

The other half of what this team does is what gets talked about a lot, it wins a lot of awards it's that shiny technology stuff, the cutting-edge stuff like Martha, Canvas, EOS, Theia, Hypergen, Cohort Builder and so on because we have the keep the lights on stuff so manageable. We're often focused on this cutting-edge tech and that is actually one of the key pillars to our success. So, I generally talk about the four key pillars to our success. This is a big one. If you're able to lower the skill floor enough on the task that you must do to keep the things running, you're able to focus more on advancing technology and achieving key strategic goals. This is also really demonstrated the power behind a solid data foundational layer. We don't just take the data from hundreds of data sources in real time, we keep them always updated, we have extensive validation scripts constantly running to ensure data quality. And in general, this helps us to really push forward. Now this is easy, obviously. The reason I give you this intro though is a lot of these submitted questions have to do with that. How did we get to this point? How do we have the people that can handle this in space traditionally lower paying without the necessary staffing? And the answer to those at least at a very high level is basically lower that skill floor enough and automate what you can so the focus can be where it needs to be.

So, without further ado, let's get to some of these questions. We had far more submitted questions than I could possibly answer here at least in the timeframe that we have. Due to that, like any geeky technology team, we had created a Python script and copy and pasted all these things on there to randomize it. So, you know, we'll answer, or I'll answer as many as I can with them within the time that we have here. Alright, so getting to the first question. I probably should have Q/A’d this before; there it, randomized scripts.

    Q1: What Technology and languages are your team using and why?

     A:   So, it's a little varied, but the majority of the applications that we have that I post on that are on the website are using an MSSQL back end, sitting on a server that we manage, there are some no SQL things that we're running, which we'll talk about just some of the innovations we haven't disclosed yet over the next few months, but for the mass majority of applications now and probably many into the future will be will be MSSQL. The other side of it is Python and flask, the vast majority are Python and flask, there's some react in there, but it's generally Python flask and MSSQL and that kind of stuff.

The why portion of why we did this, you know, six years ago when we started this team, and I was talking to one of the architects, we really thought about this question a lot  and a lot of this, the answer to why we chose Python and MSSQL, for example, is has to do with the original vision for Hyperion itself was to be sort of a plug and play system, an ecosystem, if you will, that allowed any research or any clinician code to be able to code out what they needed and attach to Hyperion.  Hyperion handles the data governance, the ticketing, you know, all of the data upkeep and all that kind of stuff. But what I realized, you know, as we went on as people, people didn't really want to do that. There are a couple people who do use Hyperion in that way. But most people just wanted us to design the app so we ended up designing the apps, but along those lines we wanted to pick technologies that were most commonly known within this industry, that being MSSQL for database and back end and then for programming languages, Python, R and SaaS, or whatever, is where your most common three.  R was probably more common in the research community than Python, but we settled on Python because it kind of crossed both worlds more developers with no Python, or if we felt with no Python, at least from the statistics out there. So, we settled on those, and that's how we came to our tech stack. Ok, so next question.

    Q2:  What do you think are the largest issues facing Healthcare IT right now?

     A: So, you know, this is a tough question. I think that my opinion on this hasn't changed in the last six years and certainly beyond those as a consultant and all that. I still think it's ultimately the data foundational layer issues in healthcare. So, data in healthcare is different than everywhere else. It's messy. It's obviously it's not 100% accurate. There's a lot of subjective data. So, you know, my pain score is different than your pain score of five type deal. There's a distinct lack of definitions for data and what people are asking for, which makes things even more difficult sometimes within the same organization, certainly outside of organizations and ultimately data interoperability is really, really challenging because those definitions are different and how people handle things, and the technical sophistication needed sometimes to do things like something like Fire. You know, your small organizations aren't going to easily be able to do that so there's these flaws in every approach and so that is what I would say is still the largest issue facing and you can see this anywhere, including the NIH.

I was just on a webinar for our Cancer NIH, a webinar focused on cancer where it was the second time I heard of them moving away from something like Fire into their own data models and all these types of things. It's also why we created Hyperion. I mean, it's, the base root of everything that we did. You know, having that no code always updated with all these validation scripts being taken care of so that people can just focus on what they need to do. It was an incredibly large reason for Hyperion. You know, and I can go on and on about that, but I still think it's data because I think without that foundational layer everyone's focused on AI,  for example you're not going to get too far with AI if your data is not validated and cleaned and all that kind of stuff, and then healthcare just doesn't. So that's what I would go with for that question.

     Q3: How exactly are you able to produce reports so quickly? What is the technical reason?

   A: I kind of talked at a high level about this at the start here but I guess for this question, I'll delve deeper into that, and this is a question that I could give an answer in that I've given speeches on for hours. So I’ll peck something to sort of delve deeper on and the answer to that is the Hyperion Data Manager, it is sort of a different way of thinking, so let’s pick on Epic for a second since it's so big and every and most people know it. Epic's database, there's many ways to attach this many different types of ways to deal with this but let's just talk about Clarity for a second.

Clarity has something like I forget what it is, it's a hundred thousand tables or something insane to say. You know, hundreds of thousands of fields in those fields aren't made like a Clinician, a Nurse, MD,  researcher whatever would think of them, and they're often named not like a programmer would think of them either. And so, what happens is to design seemingly simple reports, you, often have a lot more time then to design those things then but it would generally take in other industries. You combine this with a lot of times health care is employing technical leaning clinical individuals that might have to write these reports and not necessarily industry level technical issues who studied this and know these deep, deep sequel terms. And you get these reports that takes weeks or months like I mentioned at the start.

So, what we did was we decided to create our own data model to redefine all these semantically redefine everything into terms that made sense so, you know, patient name in medication. labs and so on and so forth or they're all named exactly as you would expect them where they're not, and these EHR's, and they're all combined and normalized and, because again in Clarity, you could get their normal form sometimes you might get something else it's just all over the map. The idea being that we would lower the skill for enough that taking someone just out of college for example that new sequel off the streets and giving them, you know, a query that would normally take a week for most people, they should be able to do in a few hours without even really training as long as they understood the basic terminology of medications, labs, you know, zip codes, whatever the case may be. And that's, and that's the start of how we did obviously there's an incredible amount of validation and there's probably 100 other things that I could mention for this question. But that kind of gives you an idea of how we start to lower the skills for with Hyperion Data Manager, which, you know, and hopefully we can, we can talk about it in a future webinar.

    Q4:  Where do you think the future is with Health IT?

    A:   So, I think I'm supposed to say AI and to be honest, you know, a lot of the future is with AI, even though I do think a lot of AI in healthcare is hype and everyone's inputting LLMs into everything they do or like search with human text and all that kind of stuff and so are we. But I think the future is still in someone or not someone. I mean, we've figured it out and I know there's a couple other people who have gone down this road as well and you see within the NIH to digress a little bit. I think the future is in finding that data foundational layer that is interoperable to everyone. So, Germany's done this as a nation state. I don't know that without laws in place that you could, you could really do it. I think you would have to do it slowly then and get a few organizations to agree to these things, but I really think it's in that data foundational layer getting a true interoperable system between systems and all that kind of stuff which would need people to work together and it's a tough one. But once you have that, we've proven at least as a team how fast you can iterate development of systems.

So, with the Hyperion Data Manager, we were able to think we had to take a couple steps back and spend you know six months making that work but after we did that, we took quite a leap forward. I think the same thing holds true with AI. When you have that, we are now entering at least in our team the phase of dealing with Hyperion AI and all those types of things which I've posted about recently. We held off for those first eight months or whatever that everyone's been pushing out AI systems. We did more traditionally AI but now we can push out AI systems faster, most likely than anyone, because we spent so much time on that foundational layer and worry about that fear of missing out type stuff. So I think the future is certainly things like AI, but I think to really move this industry ahead, there is going to have to be some kind of stepping a few steps back and figuring out, you know, these data issues I mean it's extreme as a consultant in my past life I'd been in 40 or 50 of the top institutions in America, everyone has the same problem, everyone. So, you know, there's a lot of a lot of publications and there are a lot of presentations and press about, you know, company or health co-organization acts implemented this AI but, you know, and then people have the fear of missing out, but I would question what that implementation is, at least from like an industry developer perspective because I haven't seen anything that is too crazy amazing that we didn't already know about.

    Q5: I hear health tech is years behind. Is it and why?

      A:   Oh yeah, so I mean, I think it probably is years behind, but I don't know if that's necessarily always the complete fault of health care and all of that. You know, I think that when you get into looking at like how health care runs, it's a lot more challenging than other industries in fairness. Like that example I gave of pain scores, if you were asked to, you know, plot a chart of patients’ pain scores with, you know, that were on this drug, you would have to know an awful lot of knowledge to understand how to sort of tease that out. But you also have data and health care that's just going to be wrong because you're relying on patients to tell you certain things and they're going to change the way that they say it and how it gets entered so, you know, we've taken a lot of steps to, for example, to correct ways, to make sure that we're getting not correct it, but to make sure that we're getting the correct race, sex, ethnicity, gender and so on and so forth, but ultimately, you're reliant on whatever information that works on the paper or, or what they tell you and that's not always easy. Obviously building, like things like AI offer that becomes really challenging and other routes but, is it be years behind it probably is you know there.

The other half of this is they health care generally does not hire highly technical industry people there are always some. But a lot of the systems are being run by clinical folks with technical leaning, which you would absolutely need because that's who should be determining what's getting done to the technical folks and all that, but I think we did some research as well. And I know we did some research as well; on the top 200 organizations and what their technical leadership looked like. It was something like less than half of them were actually technical people in like CIO and CTO positions which I know is probably a little controversial, but I think they mostly had MBA’s, and this doesn't occur in other industries and because we also looked at other industries the top 200 like finance,  top 200 in law,  and for example and it's almost 98% of them have technical backgrounds of some sort, but healthcare doesn't do that for whatever reason. I think that that also sort of hurts the healthcare being behind phase because there are a lot of things that that I hear in conferences and stuff that I could tell you aren't really technically accurate. Now we're in the real controversial territory. But I think, I think those are some of the reasons that’s the case. It's starting to turn around though you see a lot of organizations building in innovation groups machine learning groups and stuff like that, so I think the industry as a whole has started to sort of realize that.

We are way over so I think I will end it there. Maybe we will do another one at some point but hopefully, those quests, those answers make sense and look forward to seeing everyone in September for EOS which should be pretty fun. Thank you!