Behind the Scenes: Mayo Clinic's Data Strategy for Transforming Patient Care

Phil Sobol: Today, we're pleased to welcome to the CereCore Podcast, Ajai Sehgal, the Chief Data and Analytics Officer at Mayo Clinic. Ajai is Mayo Clinic's first Chief Data and Analytics Officer, joining them in December of 2020. He oversees data and analytics across the Mayo Clinic, product teams for software and artificial intelligence, machine learning, data governance, and literacy programs. He has also established Mayo Clinic software as a medical device quality management system and regulatory controls.  

Ajai, welcome to the CereCore podcast.  

Ajai Sehgal: It is a pleasure to be here, Phil.  

Phil Sobol: Wonderful. Well, we always like to get started with a little bit of background and I left quite a bit of your background out of that introduction. So, if you would not mind, tell us a little bit about your work and life before joining Mayo Clinic.  

Ajai Sehgal: Well, I started off my career at a very young age, joining the Canadian Armed Forces when I was 16 years old. I went to the Canadian equivalent of West Point, the Royal Military College, the French version, which was Collège Mid-Terre Royale. After graduation, I had a fairly lengthy career, probably my longest career stint to date with the military. And I loved it. I loved every minute of it. The last five years was with imagery intelligence based in Ottawa. And imagery has been a passion of mine for quite some time. Digital imagery, I am very heavily into astronomy and digital imagery.  

So, as a result of my master's thesis, which the military paid for, I did my first startup. And it was obviously in an area that my professor at the time said, you are going to do a master's thesis, do it on something you are really passionate about, because you are going to spend a ton of time on it. That is right.  

And so, it was on cleaning up the Hubble Space Telescope images, because back then it had been launched into space with an aberrated mirror and they put the call out to universities, hey, can we make useful science out of this data? And that is where Hidden Image was born from.  

And my partner, Doug George and I did the user experience. I did all of the back-end technology. And we sold it. And it was our first startup. And we used to sell it for $1,000 a pop with a dongle on the back and made enough money to stay in toys. So, it gave me my first real taste of the commercial world outside the military.  

Well, that is how I ended up getting into Microsoft. My software got their attention, got the attention of a couple of folks at Microsoft who were amateur astronomers. One was Paul Allen, and the other was Nathan Myhrvold. And when Microsoft opened an imaging division, I got an invitation to apply. And that's how my non-military career started. 

Phil Sobol: Wow. That is fantastic. And certainly, for those of us who have been around a little bit, we do remember the Hubble fiasco when it got up there and they went, whoops, everything's fuzzy. So, I am interested, and I am sure others are too, just what was your transition like in working in the commercial technology companies and then moving over into health care technology? I can only imagine that there are a number of lessons that can be learned from industries outside of health care. 

Ajai Sehgal: Yes, I do not hesitate to say it was a little bit of a culture shock because the way healthcare operates is fundamentally different than a technology company, especially when you are going into a not-for-profit healthcare facility of the scale that Mayo Clinic is.  

But there are many things that are the same. And we are not for profit, and not for revenue. You still have to optimize resources. So, there is a lot that is the same.  

But one of the key things in healthcare is to focus everything on the patient. And this is particularly true of Mayo Clinic, where our motto is the needs of the patient come first. And that goes all the way back to the founding fathers of Mayo Clinic, the Mayo brothers, and every employee at Mayo Clinic takes that into the core of their being.  

And so, regardless of the complexities around healthcare, as long as people keep that first in mind, you always end up doing the right thing. And very few commercial companies have that same type of guiding principle. 

Phil Sobol: That is very, very true. It is for those of us that have been in other industries in the past and then come here, I think it's certainly that mission-driven aspect to the role. It really just makes what we do meaningful on a day- in and day- out basis. More so, than just trying to do something a little bit faster through a typical supply chain.  

Ajai Sehgal: Healthcare also has a series of complications, right? You are dealing with data that is highly protected by HIPAA, and dealing with data that has nuances in it. So, partnership within the institution is absolutely critical to getting things done and getting buy-in from physicians and researchers is critical to getting things done. So, things might move just a little bit more slowly in healthcare, but that deliberate pace is necessary to ensure that we do the right thing. 

Phil Sobol: Well, you have touched on data, and during your tenure at Mayo, you have seen data and analytics really explode as generative AI has really become the buzzword and common phrase now, and it continues to evolve very quickly.  If you would not mind, just tell us a little bit more about your approach to leading through that change and maybe some examples of some of the work that you are doing at Mayo. 

Ajai Sehgal: So, when we started the Center for Digital Health, roughly four years ago, I was one of the first employees hired into the Center for Digital Health. We had a mission that started off with getting Mayo Clinic's data in shape. And by getting Mayo Clinic's data in shape, I mean applying good governance, de-siloing the data, and making it accessible in a centralized location.  

Mayo Clinic had already done a deal with Google to develop Mayo Clinic Cloud, which is a private instance of GCP. So, it gave us a place to put the data. Now we had to put it there using the principles of very good data management and governance. We needed to build the literacy programs.  

Mayo Clinic data comes from all over the institution, and you cannot have a central organization to manage it. So, we had to build a stewardship organization so that the individual creators of the data could steward that data.  

So, this is all done before the explosion of generative AI. And part of the reason we were doing it was there was a tremendous amount of research at Mayo Clinic on predictive AI.  

And one of the recurring themes was, how do I get to the data? Where is the data? How do I get enough data to train my algorithms?  

Our physician researchers had great ideas, but they could not get to the data. But building that accessible data stack, it is more than just a data lake or a data warehouse. It is an interconnected system of resources that allow people to easily access validated data with the full context.  

So, when generative AI hit about 18 months ago at scale out of Google, Microsoft, and open AI, we were super well positioned to take advantage of it because we had got our data house in order.  

And I love quoting Cris Ross, our CIO, who at one meeting was scratching his head and saying, you know, I would hate to think where we would be now if we had not gotten our data house in order. And that was, you know, people who do not have their data house in order and try to take advantage of generative AI are going to do it badly. 

Phil Sobol: I would go so far as to say it is impossible. So, that makes a lot of sense. And as you talk to the clinicians there at Mayo Clinic, they face all sorts of challenges. You mentioned physicians asking about the data. What are some of those issues that they are trying to solve with data and AI on a go-forward basis? 

Ajai Sehgal: So, now that our data is readily available, there are a myriad of problems that are being tried to solve. At last count, I think we had over 160 different AI algorithms within our software as a medical device pipeline being validated for whether or not they need FDA clearance or not, and if they do need FDA clearance to get all of the paperwork in order. So, the themes are very widespread, but they distill down to two main areas.  

One is, how do we improve healthcare for patients with serious and complex illness? And that is what Mayo Clinic's focus is. We focus on the illnesses that nobody else can solve.  

And the second area is supporting our workforce. Because as we all know, there's a massive shortage of healthcare providers, not only in the United States, but globally. And we are not going to be able to train up enough people to meet the demand of an aging population.  

So, what we need to do is optimize the workforce that we have, and we call it relieving the administrative burden on physicians. And there's a lot of focus in that area as well.  

For disease, early detection of disease is a way of both curing patients and delivering better healthcare, but also scaling our physicians. Because if you catch the disease early, you can actually treat it much cheaper and more efficiently with less physician intervention. And the ability to do that is all hidden within the data.  

And the new tools that we have today with generative AI are helping us to unlock some of that early detection of disease. A really good example of that is the detection of heart disease. And this was using predictive analytics, predictive AI. And our clinicians were able to discover a ripple in a regular 12 lead electrocardiogram, ECG. That was indicative of what is commonly known as a weak heart pump or in medical terminology, left ventricular flow restriction. This normally goes undetected in men, and we are not very good about going to the doctor. So, how many of us have actually gone and had a stress echo or a cardiac CT? Not many, and that is the only way to detect this prior to this algorithm.  

Well, usually the first symptom is you drop dead, and that does not help anybody. Well, they can now, in a routine 12-lead ECG, detect the ripple with high confidence and then refer that patient for further tests.  

Mayo Clinic is now taking that algorithm and partnering with a company called Anumana to make it commercially available to Philips, GE, and others to build in electrocardiogram machines. And that is a really good example of how medicine's getting advanced through data.  

Phil Sobol: That is fantastic.  

Ajai Sehgal: I recently came back from presenting at a conference in Morocco and one of our physicians who was Nigerian-born took the same algorithm and has adapted it for a single lead ECG because it can now detect peripartum cardiomyopathy, which is super common in Nigerian women.  

It is a genetic trait that causes them to have massive heart attacks while pregnant or shortly after delivery. And that algorithm is now being built into a single lead ECG stethoscope by a company called Eko. And that can be distributed to NGOs. They can go out into villages and we can actually take health care globally. All because we have our data house in order. 

Phil Sobol: That is incredible. And it is just amazing to see what is, and I think it is really just the tip of the iceberg, right? To see what is possible as we really start getting the house in order data-wise and then beginning to apply this.  

And, certainly, when you think about how do we make this all happen, we have got to lean on the clinicians, right, in that expertise. I mean, it is like one of the core tenets of healthcare technology and one of the ways in which we differ a lot of times from technology organizations outside of healthcare. So, tell me a little bit about your philosophy and how you incorporate clinicians into the whole role of data analytics AI there at Mayo Clinic. 

Ajai Sehgal: So, Mayo Clinic is possibly unique, well, it's unique in my experience, in how the relationship between administration and the practice operate. Every administrator at Mayo Clinic has a dyad partner who is a clinician. My dyad partner, I have a few, I have two.  

One is Dr. D.J. Kor. He's a renowned anesthesiologist. And my second dyad partner is Dr. A.J. Forte, who is a hand surgeon in Florida. And both those gentlemen are intimately involved in everything that we do in data and analytics. We work everything together. And the benefit of this, especially to someone who's coming into Mayo Clinic from outside from the technology world, is you get a perspective that you otherwise would not have. And you make better decisions.  

Our chief administrative officer is partnered with our CEO, who is a physician. And that ripples all the way down the administrative chain at Mayo Clinic. So our clinician, we are a physician-run organization. We do not do anything without consulting our physicians.  

And by doing that, we maintain the focus on the patient. And finance has become somewhat of a secondary, important consideration, but more of a secondary consideration. And we can maintain focus on not only the needs of the patient, but also the needs of our clinical practitioners. 

Phil Sobol: Yeah, I think that's just absolutely fantastic. And quite frankly, critical in our space, having that perspective and not just in having mentor peers is certainly one thing. But at the same point, you've done a nice job of organizationally tying that together into the fabric of your being. And I think that really translates well into what you all are doing. So, I appreciate you sharing that for sure.  

We'll get a little bit now into kind of like some of the advice that you have, right? So, conversations that healthcare leaders should be having across the continuum, whether it be IT operations, clinical, in order to prepare their organizations and guide their strategies to leverage data. What advice do you have there? 

Ajai Sehgal: Well, you know, the effective use of data analytics to improve patient care and outcomes requires a multifaceted approach with healthcare leaders engaging in critical conversations, but across several key areas.  

Number one, there has to be a convergence of strategic goals. So, you can't have just one organization within the institution saying, hey, this is what we want to do. You have to ensure that the strategy is shared across the organization. And that's difficult in an organization like Mayo Clinic, where we have three shields. We have a research shield, a practice shield, and an education shield. And then we have all of the shared services around that.  

And we try to build a consensus on the strategic direction amongst everyone. So sometimes things take time. Data governance. I said at the beginning of the conversation, I'd hate to think where we'd be today if we didn't have our data house in order. Getting data well-governed, well-stewarded, well-defined. And it's not only the data, you need to understand the workflow that went and created that data so that you have the full context of the data. That is critical to success with data and analytics.  

You can produce all the analytics you want, but you know, as in, I think it was Theodore Roosevelt who said there are three kinds of lies, big lies, damned lies, and statistics, right?  

Phil Sobol: Exactly.  

Ajai Sehgal: Someone can ask you, what does the data say? And my answer to that would be, what do you want it to say? We can make it say anything. But if you have well-governed data, you are going to get the right answer from the data. And it won't just say anything because you've got the full context.  

So, there are other areas, you know, data literacy mindsets are really, really important. And I think one of the benefits of generative AI is that people now universally have a bias for good data governance.  

We don't have to sell the story anymore. The story was sold for us because. Part of the fact that generative AI can hallucinate from time to time and produce bad data really illustrates the point of what bad data can do. So, you know, data literacy and stewardship, et cetera, has become easier to sell. So, let's sell it. 

Phil Sobol: Yep. No, indeed. You made mention of the fact that, yes, people are on board, but I would say that there's still a contingent of healthcare, whether it be in the community space or whatnot, that might need some nudging, right? Might need someone to come in and almost prove out the value, right? Because there's a lot of tight margins right now throughout the industry. And have you ever had to prove out that I don't really want to say business case per se, but the value of heading down this path with standardization of data governance and data analytics. And if so, what advice would you give for folks that are in that position? 

Ajai Sehgal: Absolutely true. If you do not tie the expenses that you are incurring for management of data, storage of data, utilization of data. You don't tie that to a business value outcome. You're not going to sell anybody on it. And that value has to be measurable. And it might be short-term value. It might be long-term value. But being able to tell that value chain story is an essential skill set for any chief data analytics officer or any data analytics person. What's even better is if you can actually deliver that value and have someone else in the organization trumpet the value that you generated for them. And that's the approach that I try to take is we want to deliver excellence for the people that we're delivering for so that they stand up and say, I can't live without this. And this is why it's saving me money. It's saving me time. And I can quantify that. And that's happened.  

One great example is in hospital optimization through throughput dashboards. We can now predict when a bed is going to become free within an ICU so that we can schedule a surgery just in time so that when that patient comes into the OR, that bed is free. ICU beds sometimes limit surgical throughput. Well, if you can optimize that through data, you are actually generating revenue for the institution. And so, value creation is essential and you need to be able to tie what you are doing to value creation. This is going to become even more important because operating the data with generative AI is expensive. And that those expenses need to be justified. 

Phil Sobol: Yep, agreed. And you know, I think we already kind of introduced the concept of you've got those large organizations that do have some capital to invest in these things. And you've got the smaller ones that have to be very, very mindful. And I think you made a point earlier about IT, leadership, operations, and clinicians all being on the same page together with this. And I think as they move forward, particularly for those organizations that might be smaller, community hospitals, et cetera, what process should they potentially walk through to identify, what should we be focused on? Maybe at the community level where you've got 50 beds in a hospital, they're probably not looking and thinking Gen AI today because it almost seems out of reach. Where should they start? What should that process look like? 

Ajai Sehgal: Well, from my perspective, the principles that apply to startups also apply to small community hospitals. So, if you want to take advantage of data analytics, start small, fail fast, and scale gradually as you learn. That should be the very first tenet.  

Prioritize data security and privacy up front, because it is less expensive to prioritize it up front. Make sure your information security perimeter is sound, especially in today's world where foreign national actors are attacking hospitals on a daily basis.  

Invest in the engagement with healthcare professionals and physicians. Do not leave clinicians out of the loop. It's not every hospital that has a physician-run organization like Mayo Clinic. Many are administrative run for profit organizations. But make sure physicians are engaged from the get-go, because that's how you are going to generate value.  

And then prioritize value-based care. Everybody struggles with economic constraints, particularly smaller health systems. But data and analytics can play a very, very critical role in early detection of disease in optimizing the dollars spent within the hospital.  

Make sure that you govern your data well so that you are able to do that. And this landscape changes by the day, not even by the week. So, stay informed. Engage external professionals if you do not have the capability internally. 

Phil Sobol: You need to stay informed. Now, sound advice for sure. Well, we always like to wrap up the podcast, Ajai, with just a call for any final words of wisdom that we didn't happen to get to in the conversation.  

Ajai Sehgal: Well, you know, healthcare evolves continuously. And today's world, it's evolving so rapidly, it spins your head. And one of the things that people aren't very good at is embracing change.  

I think embracing the change is something that you need to do in today's world if you are going to stay relevant. Embrace the change, but always prioritize a patient-centric approach.  

Patients come first, especially when you are making decisions, investment decisions. If you put the patient first, you will end up doing the right thing from an ethical perspective.  

Invest in people. Technology is changing rapidly. You need to invest in your people to keep them up to date so that they understand how to make best use of the tools.  

Stay curious and committed to lifelong learning because the technology landscape is changing by the day. A year ago, there were very few experts in generative AI, and now there's a whole plethora of experts in generative AI.  

And I guess probably the most important thing is remember why you are here. Four years ago, I almost retired for good, and Mayo Clinic gave me a reason to never retire. Our history of 150 years of data and innovation and the ability to do something special for humanity, leveraging stuff that I had learned for many years was way too compelling to ignore. And so, remember why you are here. 

Phil Sobol: That's fantastic. Well, Ajai, thank you so much for taking the time today to be on the podcast to share your little bit of your background, a little bit of your story, a lot of exciting information about what you are doing there at Mayo. And hey, we look forward to continuing to hear from you and hear what you and the Mayo Clinic are doing in the industry. and ultimately for patient care. So, thank you very much.  

Ajai Sehgal: Thank you, Phil, for having me. Appreciate it.