Dr. Suchi Saria, the Founder and CEO of Bayesian Health, was featured on a recent episode of Becker’s Healthcare Podcast. In her conversation with Scott Becker, Dr. Saria dives into discussion about how artificial intelligence and machine learning (AI/MR) in health systems can revolutionize care delivery and improve outcomes. Dr. Saria emphasizes that with EMR, health systems are now in an advantageous position to leverage AI/ML to deliver accurate and actionable clinical signals to physicians and nurses, helping them anticipate patient needs and deliver safe, quality care. Listen to the podcast here or read the lightly edited transcription below to learn more about Dr. Saria’s work and the future of health AI/ML.
Transcript
Scott Becker:
This is Scott Becker with the Becker’s Healthcare Podcast. I’m thrilled today to be joined by a very special guest. We’re joined by Dr. Suchi Saria. Dr. Saria is the founder of a company called Bayesian Health, but far more than Bayesian Health she’s also the Director of Research and Technical Strategy at the Johns Hopkins University in the Malone center for Engineering in Healthcare. She’s also a Stanford Ph.D., a Harvard University fellow, an Amherst graduate. She’s got a master’s from Stanford, who is sort of in the middle of this world of artificial intelligence today. It’s just a different level than anybody else I get a chance to visit with. Dr. Saria, could you take a moment to introduce yourself and tell us a little bit about Bayesian Health?
Dr. Suchi Saria:
Absolutely. Scott, it’s great to be here. Thank you for inviting me. So, my background is in deep AI healthcare research. I’ve been doing work in AI for almost two decades now building technology that delivers safe, reliable, unbiased results, working at large academic medical institutions like Stanford, Hopkins, and Harvard, and with large governmental organizations like DARPA and the FDA. And Bayesian Health in particular, what we’re doing is bringing best in class AI to the bedside to augment physician decision-making, helping them deliver care that is safer, higher quality.
And just a little bit more about how it works: so, Bayesian is a platform that integrates within the EMR, delivering clinical signals to physicians and nurses within their workflow. And the models that are powering these workflows, you know, we sort of spend a lot of time thinking through things like potential issues, bias, and so on so that these signals are very precise, almost 10 to 15 times more performant than typically what they’re used to. They come with insights and context that make it very easy to use at the point of care. And then we think a lot also about not just the tech itself, but also ‘how do you implement it in a way so that you get sustained adoption?’
And finally, one thing that digital health has suffered from is — a lot of times we introduce tech and you don’t really know if any of it is working. And so, you know, coming from a deep research background, I think extremely hard about evaluations but also infrastructure to evaluate on an ongoing basis what is working, what isn’t working, and very rigorously measuring evaluations so that we transparently understand, how it’s benefiting our patients, how it’s benefiting our providers and the health system overall.
Becker:
Thank you very, very much. And talk for a moment about – you’re this brilliant person, a leader at Johns Hopkins, also formerly Harvard, Stanford. How do you then start a company? How does a company get started out of that research? How do you begin a company? Tell us a bit of the genesis there.
Dr. Saria:
So, interestingly, I used to advise companies a lot as a researcher, but what I was creating in the lab was pretty effective, pushing the technology to be able to get to really effective AI tools that were catching critical complications early. In fact, some of my very early work, we wrote one of the first papers in the field to show that you could use machine learning and AI to identify sepsis early and it was a cover article in Science Translational Medicine.
But I had a personal moment in my life, my family’s life, where my nephew in India, he developed sepsis and unfortunately, he passed away. And it was a tough moment for me, a huge loss for my family. And what was hard was in that moment, I remember when my mum called me about this, and my family called me because they saw me as someone who had done a bunch of research on sepsis and cutting-edge research on sepsis, and I realized I couldn’t do anything to help it and none of my research was actually at the bedside. None of the state-of-the-art research was actually making it to the bedside. So that’s sort of what prompted a partnership with system leaders to be able to figure out a way to deliver this technology within our EMRs to physicians who could use it and make it really easy to use. So that’s exactly what we’re doing with Bayesian. It’s bringing this mentality of outcome-first technology to health systems to bring and in part, bringing back the joy of practicing medicine where they don’t see technology as making lives harder but easier and, in the process, making outcomes better.
Becker:
Take a moment and tell us a little bit about, you know, you’re building this company, brilliant research and technology behind it. Take a moment and tell us about the future of artificial intelligence in healthcare. What does that look like in the future? Are we going to, we’ll still have doctors, we’ll still have procedures, we’ll still have nurses, we’ll still have all of that, but how much will their work be helped by artificial intelligence?
Dr. Saria:
I think one thing that’s really interesting is health AI has matured a lot in the last five years. I think what we’re seeing today is in most sectors outside of healthcare, some form of AI is driving core value in almost every function. In healthcare, historically, marketing had outpaced results. It was a little bit hard for a non-expert to tell what’s good from bad. But in the last five years, I think health AI has matured quite a bit. And I think one thing that’s interesting is most of the early impact of health AI in healthcare has been, to date, I would say mostly in the fringes right now like automating back office. They were seeing good success. I see massive opportunities in impacting sort of the core business of healthcare going forward which is care delivery innovation, right? So, when you look at problems like systems struggling with hacks and highs like 200,000 preventable deaths a year, where there’s an opportunity you come in, move from reactive to proactive guess so that’s higher, safer quality care. When you’re looking at complications like sepsis, where a single lawsuit can cost you easily a hundred to 200 million dollars, then there’s an opportunity to come. And many times when that happens, when you look at the individual case, the keys were there, this patient could have been identified early, but they were missed. So again, by basically moving to the use of these kinds of AI-driven proactive technologies that allows you to anticipate patients at risk, with proactive workflows in place that allows you to not only deliver higher, safer, quality care, but also then manage risk. And manage risks, not just from a malpractice standpoint, but also confidently take on more risk than you were thinking about value-based contracts as you’re increasing emphasis on patient experience overall.
Becker:
And you had mentioned earlier in some of your comments, Dr. Saria, this issue of clients, it might be health systems, really over the period of time, I think has had a very hard time differentiating what’s real and what’s not. So, there’s been an explosion of AI-driven, you know, efforts in sales pitches, and marketing. I mean, this one’s backed by one of the brightest leaders I know, that I’ve ever met, but how do health systems differentiate what’s real and what’s not? You know, what’s really going to be effective and useful and what is just smoke and mirrors? How do people actually differentiate that in symptoms. What advice would you give them?
Dr. Saria:
Absolutely. Great question, Scott. And I think this is essential. So, one big part of it is again emphasis on results and emphasis on outcomes. Most tools deployed today, they really don’t understand if anything is working, right? There’s no measurement of like efficacy, utility. Think about how drugs work, right? We have a very robust infrastructure for putting trials in place and measuring efficacy and understanding what is and isn’t working, including post-marketing surveillance. We need the same kind of discipline in this space wherever you’re trying to make clinical impact using AI and machine learning-driven technology. So to identify what’s good from bad, the number one priority needs to be sort of a team that has deep expertise in this space that really understands why the data is messy, how it’s this messy, have a proven track record of publishing in this space and the top journals, they’re not just using off-the-shelf technology or have superficial knowledge of the space because they’re not going to be able to demonstrate and deeply tackle some of the issues with healthcare data. So that’s number one.
And number two is actual studies showing the efficacy of the software that’s deployed. One very interesting thing I discovered in doing this work is for sepsis, there’s a plethora of companies attempting to tackle it. Interestingly, when you go back and look at it, fundamentally, there are very few that have actually shown results with the tools that it actually works. They have not been able to show adoption, like ‘are physicians actually adopting this?’ Do you actually have provider engagement because you can’t get outcomes if providers don’t adopt, and then do you have any kind of study showing prospective benefit in terms of outcomes like mortality, morbidity. But this discipline of really rigorous studies is also very important. So, deep expertise in the actual technology itself because this is a very fast-moving field and you want to make sure the team really understands the issues around how to apply it effectively, tackle issues around bias. And then also deep expertise around evaluation and studies showing it works.
Becker:
Dr. Saria, you’ve had this amazing career, literally, crazily successful at a huge institution, also as an entrepreneur, and a real brilliant thinker and researcher. What advice would you give to a person trying to have an impactful career? What advice do you tell somebody trying to have a great career, trying to have an impact in the world?
Dr. Saria:
So, first of all, Scott, very, very kind of you to say that. It’s a very tough question. I often think about it myself and what I’ve learned from the people who’ve mentored me. I think one thing I often think about is great leaders need to be very comfortable with taking risks and taking risks, not reckless risks, calculated risks, educated risks. But it’s very easy to do something that someone’s already done before, right? But in healthcare, like we’re in a very interesting time, there are a lot, there’s so much opportunity for pushing the needle because, we went from no data to data, EMR, we have the infrastructure, you’ve invested hundreds of millions of dollars, billions of dollars in this space. And now there’s really an opportunity to make it all work for us, but great leaders have to get comfortable with the idea of identifying opportunities where they’re uniquely able to be from their own backgrounds, take calculated, educated risk, and prioritize big initiatives that they can uniquely champion and make time for solving.
Becker:
What a great perspective, because it really does take this mentality of being able to take big risks, but also being safe. And that’s a hard thing for people to balance. How do I measure safety, professional safety, personal safety, financial safety, family safety, plus taking some big risks? And that’s hard for a lot of us to do, and Dr. Saria, finally, what are you most excited about this year? What are you most focused on and excited about this year?
Dr. Saria:
So, we are seeing from the studies we’ve done over the last several years as we’ve deployed, partnered with health systems, really exciting data. And the use of our technology, it’s very easy to implement, deploy, and we’re partnered with the leading EMR vendors to be able to actually show mortality and outcome benefits. So, the thing that we’re most excited about is partnering with health systems where we look at their outcome data, their publicly reported outcome data, identify opportunity areas for them very concretely both for clinical benefit but also financial benefits and pick clinical areas. One common mistake people make is they sort of get, they start thinking, ‘okay, we want to use that AI,’ but the question is ‘where do you start?’ And so, we’re excited to continue to partner with systems to identify areas that they can benefit based on their outcome data. And then continue to build studies, continue to deploy and demonstrate the efficacy of our solutions. And then most importantly, for me personally the way we measure internally successes — is how many lives have we saved? And I want that number to be going up.
Becker:
But what a magnificent way to measure success, isn’t it? I mean, if you can measure success in how many ways, how many people are we saving, is anything better than that? I mean, what a way to measure success.
Dr. Saria:
Thank you, Scott.
Becker:
Dr. Saria, what a pleasure to visit with you. I can’t wait to continue to watch Bayesian Health evolve and grow and what you’re doing. What a pleasure to visit with you, an amazing, you know, one-of-a-kind leader and person. Dr. Saria, thank you so much for joining us today on the Becker’s Healthcare Podcast.
Dr. Saria:
Thank you so much, Scott, such a pleasure.
