Sanjog Aul [00:00:06]:
Hello and welcome to CTN. To learn more about the show, please visit ciotalknetwork.com. Our topic today is Strategic AI Deployment in Healthcare, Exploring Ethical Considerations and Operational Challenges in Predictive Care. So we know healthcare has already been a tough cookie as an industry, where most leaders in technology and business both are trying to get their arms around it and there have been many operational hurdles. There has been the business of healthcare, which has always been challenging, and of course the technology. So here comes artificial intelligence. You are trying to get the organization to benefit as a business but also help the patient. And here we are also throwing in this predictive care paradigm, which means you perhaps can figure out a way to predict what is likely to become an ailment or something really serious which will require a lot of damage control. If you could predict to prevent it, then that would save a lot of patients agony and their family members agony.

Sanjog Aul [00:01:13]:
Plus also it’ll give brownie points to the healthcare as an industry, or that healthcare system which is providing that service. Now AI, as it’s coming, the healthcare industry is trying to use it to see how it can further that cause. But at the same time we have got the privacy related issues, you got compliance which we have to deal with. And while we are still grappling with HIPAA, which is not new, how does an organization go about dealing with artificial intelligence, dealing with HIPAA, and at the end of the day give something what the patient wants but also stay healthy as a business? So that’s quite a bit. We’ll try to unpack this during a conversation. I have with me Cherodeep Goswani. He’s the Chief Information and Digital Officer for University of Wisconsin Health.

Sanjog Aul [00:02:03]:
Hey Cherodeep, how are you, sir?

Cherodeep Goswani [00:02:05]:
It’s a Monday, it’s a great day. Thanks for having me on the show here, Sanjog.

Sanjog Aul [00:02:09]:
Pleasure to have you. So as I have given a loaded introduction and what’s been going on, because of course we as media have been covering it for last 21 years and we see it has morphed, but there are still some things which have not changed much. And now you throw AI in this mix, and at the same time you’re also trying to become almost having a crystal ball with predictive care so you prevent things from happening. Give me your firsthand reaction to how I painted the picture. Do you agree with it, or would you have said anything different in describing what the state of affairs is in healthcare?

Cherodeep Goswani [00:02:47]:
Well, first, out of curiosity, I would like to know, did AI generate your opening dialogue or did you write that on your own? Because it was very good, the way you introduced AI out there. But the fact that 21 years, I think you said 21 years you’ve been doing AI in this area. AI, to your point, is not new. We’ve been doing this for a long time. The mass media has just got onto it in the last 18 months. And healthcare is always evolving. It’s always evolving because people are always falling sick and there is not a lot of people that are coming into healthcare. So AI doesn’t mean the same thing to a lot of people.

Cherodeep Goswani [00:03:24]:
And what we have been trying to do is to generate awareness that AI is actually a how, not necessarily a what of what we do. It’s how we do things that has augmented in the last few months, and happy to take the next 45 minutes or so to share examples of where we have seen it work, how we make it work. Because at the end of the day we all work in the healthcare industry and take care of people, not necessarily roll out the coolest of technologies, which often becomes a distraction.

Sanjog Aul [00:03:56]:
So I like to build upon this what versus how. So how is when we knew what we need to get done, it was just a matter of time. And maybe this is one of the mechanisms through which we could get it done realistically and at a reasonable pace. So it is practical. But then many things which we spoke about and or envisioned in healthcare were almost Star Trek for many, if I were to go 15 to 20 years back. And now if you were able to move from what to how, would you attribute anything to the confidence or even a sense of possibility that it inculcated in the minds of the healthcare technology and business leaders? That yes, with AI now I can go past what at all I could do, and now let’s focus on the how.

Cherodeep Goswani [00:04:45]:
Yes. So let me give an analogy for your non-healthcare users, using non-healthcare examples, so I can draw the correlation and causation over here. So all of us have been using spell check on our phones and our computers for years. In some ways that is a basic form of AI. And you also have self driving cars that use AI. It is a question of individual preference and also a question of individual comfort, in that probably most people will take the spell check option at a heartbeat’s notice versus think 100 times over before getting into a self driving car. Healthcare AI is very similar to that. You can use AI for a lot of things.

Cherodeep Goswani [00:05:23]:
Including for the right reasons, helping with scheduling appointments, trying to find out certain areas where the impact to the human life or the risk is minimal. Those are the tasks that we say it leads to greater efficiency and are easy wins using AI. If you go further up or down, depending on which way in the pyramid of complexity, now you have AI that is capable of completing a provider’s task for the provider, giving them much needed relief and capacity. One, to rest themselves, and two, to take on greater patients. And then you go even further out. There is AI out there, or just augmentation out there, that can allow you to do better predictive modeling or predictive analytics as we often call it. All of this comes back to a level of what is the risk and what is the impact if I go wrong. Not saying that everything we do today is right, but technology should be held to a higher bar than a human is when dealing with the human life.

Cherodeep Goswani [00:06:27]:
So those are the three examples of the categories in which at least we try to generate awareness of AI. And in the first case we say take it and run. In the second case we say let’s talk through this and understand what’s the risk. And the third one we say go slow to go fast, Sanjog, in those three domains.

Sanjog Aul [00:06:45]:
If I were to take these three domains or three lanes, and once you do your guardrail creation, maybe awareness creation, does the implementation take care of itself?

Cherodeep Goswani [00:06:57]:
No, implementation takes care of itself. One of the buzzwords that I absolutely detest in our industry is plug and play. Nothing works by plug and play. Especially in the world of healthcare, we use the word ethics. I will use another couple of words here. Empathy. Empathy is very important. There has been technology around for years, some form of very rudimentary AI that could collect your results after a test, do all the augmentation and say, Sanjog, you or your loved one has tested positive, positive being bad, to a certain test.

Cherodeep Goswani [00:07:34]:
This is an example. Technology can do that. It has been doing that for 25 years. But does it replace the human being that calls Sanjog and gives the bad news and shares the empathy of that individual and says, now let me tell you, what are the next 48 hours going to look like, or the three weeks or the four weeks? The human has to augment and support and manage the technology, and that stays part of the implementation side of the house. Technology, left to itself, can often be a disruptor. It needs to be managed because healthcare deals with emotions, and which is why I use the word empathy.

Sanjog Aul [00:08:09]:
So imagine if we were to deconstruct any domain. Let’s take healthcare for example, because that’s the topic here, or that’s the domain we are covering. And I could deconstruct it into a bunch of business functions and use cases within those functions, or use cases at the business domain level itself. Traditionally, healthcare has been around before even technology was around. Healthcare has been around since then. So what we tried to do based on those use cases, we tried to do them manually, use our best intellect, best faculties, best leverage we had with our arms and legs, and try to make healthcare happen earlier. Then comes some technology of some sort which gave us the mechanical leverage.

Sanjog Aul [00:08:52]:
Then we got something where technology brought some leverage in terms of information technology as a leverage. Now we are talking about AI. I’m assuming that it is being looked at in the same light, that it’s going to give us specific leverage so that it furthers our ability to create value for our business and for the patient. So if you were to peg on this word leverage, can you inventory the kind of areas where AI specifically is going to give us leverage, or at least expected to give us leverage?

Cherodeep Goswani [00:09:24]:
I would prefer to use the word complementary leverage versus specific leverage, because that’s the difference of the how versus the what. Let me again articulate with an example over here. If you think about it, for over 100 years, a patient has walked into a physician’s office and had the conversation, and the physician has done something to record the encounter, so that three weeks later or three years later, when the patient comes back, there is a history of that. That’s what we call a chart. Fifty years ago, those charts were managed in paper. And if a physician retired and a child of the physician became the family physician, those charts went with the human element out there. Then came computers and EMRs. And then we decided to record everything.

Cherodeep Goswani [00:10:13]:
And the recording happens whether you type on a keyboard, whether you dictate and you capture it, or whether you hire what we call a virtual scribe, sitting 50 to 5,000 miles away listening to the conversation. And in the last three years, we are talking about ambient listening. It’s five different hows of doing the same what: capturing an interaction and recording it for further use. AI in the form of ambient provides the specific benefits, that it has actually given the physician time back to look at you, the patient, and do eye contact versus looking at the keyboard and covering it. It also has the capability to augment the provider’s thought process and highlight certain things from 25 years ago on the patient’s history, to say maybe you should look at this and cover this a little bit. And at the end of the day it allows the physician to capture a complete note under the circumstances, doing a lot more than before and ensuring that the patient is in a better state when they left the room than they were when they came into the room. Now, has this replaced anything? It has replaced or augmented the different hows, versus typing versus speaking into a microphone. But it has not replaced the fact that the physician is listening to the customer or the patients and looking beyond the 30 minutes of the encounter, and having the depth of technology to look after the 30 years or before the 30 years and various other factors in creating care.

Cherodeep Goswani [00:11:46]:
I hope that example helps in articulating the how.

Sanjog Aul [00:11:50]:
Absolutely. Now let’s see, since the topic is around predictive care, before we get into the ethical and operational challenges, et cetera, what’s the current state in your view? You live in healthcare. In this case of predictive, is it still a crystal ball syndrome there? Are we making any progress, given that predictive care concept itself is not new? People have been toying with that for some time, and people may have done something to attempt and go in that direction, but it doesn’t seem to have taken on to the degree that everyone would ask, at least as patients would like it to happen the way we want. But where does it stand today? And take a 360 view if you could.

Cherodeep Goswani [00:12:38]:
Yes. So given my background, that I didn’t grow up in healthcare, for the years I stayed outside healthcare I always thought healthcare was 50 years behind in technology. I’ll be the first to admit we are still 20 years behind in some cases. But I’ve also had the humility to understand that healthcare deals with an emotional aspect of a human being that very few other industries do. And the degree, or the margin of error, is very small. You use predictive analytics to say what time will the flight land, and if you’re off by two hours, so what? You take predictive analytics to predict the lifespan of your child. If you’re off by five years, you have some emotions to repair out here. I want to start with that domain, to say the margin of error is what drives the higher bar for bias and accuracy in these areas.

Cherodeep Goswani [00:13:28]:
Second thing I will say is if you compare technology between healthcare and other industries, technology uses the word standard, standard, standard, reduce variations. And when you allow analytics on patterns, it works very well. A physician leader in my very early days introduced me to the determinant and said, Cherodeep, I will take all standardization. Give me two patients with the same DNA. You produce that, Sanjog. Tomorrow I will produce predictive analytics two years from there. That’s a friendly challenge. The predictive analytics, beyond the emotional aspect, becomes complex because of genomics and everything else.

Cherodeep Goswani [00:14:06]:
Having said that, we can only look back 20 years to see how far predictive analytics has come, how far science has helped us to progress, that we could actually contain, if not solve, the greatest pandemic in our life history two years ago, and we’ve just forgotten. That was a lot of predictive analytics that actually led to sequencing and identifying the vaccine, but also helping day to day operations, to say how bad will day after tomorrow look like given the circumstances in which we have over there. So call me a convert, but I have come so far to say that predictive analytics is moving faster than ever before. The challenge is you never know what the finish line is. So you’re only looking how far you have to go and forgotten how far you’ve come in this race.

Sanjog Aul [00:14:59]:
So when it comes to predictive care, if you were to put that on a scale of level of whether implementation, adoption and value creation sitting on the patient side. Are we waiting and watching still, or there is some value created in that regard, in maybe very unique or very few cases, but at least is there a start?

Cherodeep Goswani [00:15:24]:
I would say it’s a very segmented, targeted population where you start looking at it. When you talk about value based care, let’s also face the fact that incentives are not always aligned the way the economics of healthcare works. Not every organization gets paid if I keep you well enough that you don’t ever visit the hospital. Not every organization gets there, what we call the capitated model. So in a world where we have shortage of resources, a provider is going through in a 20 minute encounter and trying to make sure that, are you doing well for the care that you wanted to come in or treat over here? How much time do you spend in looking through and saying, for the next 10 years you need to do this, you need to do that or whatever? And sometimes the patient, again, this is the maturity of the patient also coming into play, to say how much can I consume that information. If you think of what we as an industry has done in managing and reducing breast cancer for middle aged women across the country through predictive care and preventive care. Those two words are often exchanged, but predictive also leads to preventive. There is significant benefit if you look at pediatrics and what you can do with science in the last trimester for a complex OBGYN case. It has led to much healthier babies and shorter length of stays in the NICU, where neonates end up, and led to better quality of life and longer lifespans for those kids.

Cherodeep Goswani [00:17:03]:
A lot of that comes from predictive analytics, except that the media doesn’t catch on to that one often as much as I would like to.

Sanjog Aul [00:17:13]:
Now that is from the patient side. And as you mentioned, or at least alluded to, the very economics of healthcare, if given a choice, are there business models being discovered or created, so that with predictive care it’s going to be a dream come true for every patient? And as we are trying to get creative in getting predictive care implemented for patients, are we also looking at the economic models which will make an incentive for us to work harder on this? Because eventually, if the healthcare systems work proactively and invest to the degree that they invest in other types of technologies and paradigms, if they get the incentive to invest enough, we will see less and less patients in a hospital, or I would not lose my loved one. Imagine the value creation there. But are we looking at one is to say, oh, it’s very tough. Well, things are tough. So have we done something to that regard to simplify the economic model or create an incentive for us to work on predictive care?

Cherodeep Goswani [00:18:28]:
So I would say yes. And if it was easy, it would have been done. People are not dumb and stupid. People are smart enough, and who hasn’t been a patient at some point in their lives? We all go through this. Now if you start back and you look at economics and policies, and then what I would call operationalization, where provider systems like ours come in, then you look at payers that come in in terms of payment models, and then you look at pharma. Together, all of this brings up the ecosystem. If I were to come and charge you, you specifically, and say, I’m making this up, okay, if I were to come and charge you and say, for $1,000 more, I will do predictive analytics for you and let you know every year out. I’m sure you can probably afford and pay for that, but a lot of our folks cannot pay for that. And if you really think about it, till maybe about 10 years ago, computational power did not exist beyond a predictive analytics of the next few days or a few months, or maybe a couple of years at the most.

Cherodeep Goswani [00:19:28]:
So you always have to look at, how are the economic models? One, coming in, who’s paying for it. As a business person, I always ask the question, who’s paying for it? Because if no one’s paying for it, it won’t get done. Okay. Number two, there is also regulation that drives what you can and cannot start predicting, because the only one time you’re wrong, you will be accused of playing with something wrong. Number three is not every patient wants to know what’s going to happen to them day after tomorrow as much as we think we want them to know. All of this comes together, and what we call about value based care is to come back and say, here’s the population that we need to keep them out of the hospital. And with this science and technology and with the resources at our disposal, we are going to work on those routes. And I would say for the better part, some data shows that it is working. Is it a straight line to success? No.

Sanjog Aul [00:20:26]:
Could I take the GRC aspect, the privacy which could be part of IT, and standards or standardization aspect, and operational aspects, and literally draw out a playbook for predictive care and be ready, audit it, poke holes in it, experiment with it, and then all along, on a second track, I continue to work on an economic model to make it a success? One should not wait for the other. But at least, and maybe I’m ignorant, I’m not sure if I’ve seen a playbook created for predictive care to that degree where every I is dotted, every T is crossed, assuming that it has an incentive. Or is there one?

Cherodeep Goswani [00:21:18]:
No, I would say playbooks do exist. This is how you work into value based care, and you look at certain care populations, whether it’s dealing with oncology or it’s dealing with one of the other ologies. But to your key point, every I dotted and every T crossed. I’ll put it this way, the I’s and T’s that we know are dotted and crossed in some cases. The challenge, again the uniqueness of the industry, is you don’t know how many I’s and T’s exist that you don’t know. That’s where the example comes in. But I don’t think there are. If you think of what was called, still is called, the accountable care models, ACOs, 15 years in the making, started with 60 odd, 70 odd organizations 15 years ago. Now the numbers grow. But in some cases we are hedging our future.

Cherodeep Goswani [00:22:02]:
We are betting our future by taking on those care models, using data to the best possible ability in our domains, and making sure that we take very good care of our patients in those models, because we lose money if we fail to take care of them outside the four walls, outside the clinics, treat them where they are, in their homes. So if you look at that number, of course it’s changing. But what we don’t know, which is where you’re guessing the future, is think of the long COVID effect six years ago. I can tell you, and you’re smart enough to know this, I can tell you no predictive model in 2019 looked at a long COVID effect, because those three words did not exist. So in some ways, come 2023, the question now becomes how many predictive models do we have to go back and redefine, redesign, redeploy, taking those five variables that we did not know. It’s never a finished product. It is always in the evolution of the product, that the tomorrow becomes a little better than yesterday. Progress, not perfection.

Sanjog Aul [00:23:11]:
So EMR, EHR, patient centric care, population health, all of these terms have been floating around for quite some time, but operationally we haven’t seen a beautiful looking, the way things are running. Sometimes there is disconnect between the data providers, etc. And that I know going back at least 10 years, personally having facilitated sessions, those challenges have still been lingering.

Cherodeep Goswani [00:23:37]:
I was going to say, I don’t think I’ve ever heard the word beautiful in healthcare. One needs to look beautiful, they don’t come to a hospital. So beautiful is not the right objective. Efficient, if you allow me to use the word efficient, that should be the driver, not beautiful. And in the spirit of efficiency, I’ll be the first to admit, which is, I came into healthcare because of a not so pleasant experience with a loved one, and today it’s to make it more efficient. Does it reduce the provider’s burden on entering data? No, in fact we have made them glorified typists. But if the provider doesn’t enter that, then in many cases we don’t get paid. That’s a policy decision, a regulation decision.

Cherodeep Goswani [00:24:21]:
I’ll give you another example of a benefit that I saw firsthand. It was a firsthand experience where I noticed a lady on a flight have an adverse outcome on a flight, and there was a provider on the flight who was asked to help, and at some point the provider turns around and asks the lady’s daughter, I think I know what to do, but I’m not sure what her allergies are. And I’m sitting behind, because I’m not a provider but I work in healthcare, I’m helping this gentleman, and I asked the daughter, I said, do you have a MyChart account for your mother? And that will show the exam, this thing? And they looked at me as if I was a genius. No, I wasn’t a genius, I’ve deployed EMRs to say that is one of the reasons you do it. Now for that patient.

Cherodeep Goswani [00:25:09]:
If you turn and say, has EMRs helped? I think she will be sending flowers to me for a long time, so to speak. But again, it’s in the eye of the beholder, in avoiding the worst possible outcome, in which case is death or a bad disease or something like that. We made significant progress. Has it been easy? Not at all.

Sanjog Aul [00:25:35]:
If you were to come back to the predictive care, and we know the evolution is still on, the economic model is being created, we are trying to find use cases, etc. But if you were to take, like you said, value based care, can you take that operational model and literally apply that to predictive care? Is it going to work, or are there unique operational challenges you anticipate with predictive care?

Cherodeep Goswani [00:26:01]:
So it’s a question of and, not or, in my perspective. It is being allowed. Value based care is based a lot on those predictive analytics, because you’re assuming the risk and you’re doing that based on certain models that you have run, and fingers crossed, your numbers are right and then whatever happens. So it is working on that. As the complexity of the variables grow, the model changes every day. There are people that work on these ACOs every month, if not every day, and look at these models. These are the models that actually end up doing the negotiations or the contracting between the payer and the provider that we go through out there.

Cherodeep Goswani [00:26:42]:
So I don’t want your audience to think that nothing is being done. There’s a lot more to be done. We didn’t touch on the technology aspect of this, but also the way, especially quantum computing, which is the new buzzword right now. I shouldn’t say a buzzword, it’s a reality. But the computational power that is coming up in the last five to seven years has made this possible, to take in more and more and more variables and make it more accurate. Most people don’t realize the human body has over 200 bones and 600 muscles. Do the permutation combination of those, add the organs, and think of the number of variations that you have to run.

Cherodeep Goswani [00:27:25]:
This was not going to a restaurant and picking from an a la carte menu of 25 things.

Sanjog Aul [00:27:31]:
Now, if you were to inventory some of the areas where ethical concerns would arise as a result of people trying to subscribe to the predictive care, if you’re able to crack the code of how to make it happen, what would those be, and how would you safeguard or work to minimize the impacts?

Cherodeep Goswani [00:27:55]:
So I’ll add another word over here, and let’s take ethics, and let’s take equity, ethics and equity together, and look at this. This is a very important topic. So again, in healthcare, we are trusted with one of the most sensitive and personal information that a person owns in their life, their own private information. And so we are entrusted with that private data, and it is very important that we never compromise the privacy of that information. This is where HIPAA comes into play and so on and so forth. Now, if you ask me the economics, again, this is the danger that we all face, because the more data we collect, the more variables we run, the more dangers are that that data will be hacked by some unknown source or inadvertently used by a known source for unknown reasons. So the ethical side of this is very rarely… Well, you have the hackers who come in with a bad purpose, but there’s also an ethical side of it where we inadvertently may have to compromise that data because somebody did not know when they send it to another vendor, let’s just say, to use their predictive analytics model. So that’s the ethical side that we absolutely need to manage. The second part of that is the equity side of it. There is what we term the digital divide that happens across any society, any organization, any community, any country.

Cherodeep Goswani [00:29:27]:
We also have to make sure that the models that are grown, that are developed with certain kind of individuals, certain kinds of population, are not then grown and covering the entire universe. It’s just where we call national data models and local data models. At the end of the day, I can say this to you. You and I look similar, you and I probably have similar genetic backgrounds, you and I will probably have certain diseases and probably not be prone to certain diseases. But to say 100% of the people, based on the N of 2 on this call, are applicable to that would be a pretty dangerous thing, which is where the equity side of that information comes up.

Sanjog Aul [00:30:08]:
And is this equity stemming, or a lack of equity or the threat to equity comes from the bias which somebody might have developed through where they come from?

Cherodeep Goswani [00:30:20]:
Very well stated. Humans, when we are born, I think we have a bias, and at the end of the day, humans write the algorithms until the point where algorithms write algorithms. So bias is inherited. And believe it or not, now there are models in the market that actually measure the bias of an AI model. There are actually organizations and companies producing products out there. So bias is a living risk to everything we do.

Sanjog Aul [00:30:53]:
So if we are looking at the ethics side of it, which you mentioned, and it is stemming out of bias, and that’s a social bias, as you said, people are born with it. So is it like a leave it to God problem, or can we do something about this?

Cherodeep Goswani [00:31:08]:
Well, I don’t know if that’s a solution, leaving it to God. So what we do within the variables that are at our disposal is you start with awareness, then literacy, and then governance. You never start with governance, because if you don’t know what you’re governing, then we have a bigger problem. So all the conversation that we had for the last 10 minutes is tied into, at least in the organization that I work on, and I know most of my peers do some form of this, is to generate awareness and drive literacy to those individuals that have access to this data, that are writing these models, that are consuming these models. Once you have strengthened the human factor in that ecosystem, you have probably negated a significant part of that risk. Because yes, again, the bad actors may come with a malicious intent of stealing your data.

Cherodeep Goswani [00:32:02]:
But I truly believe nobody in healthcare comes in to say I am going to do something unethical to my patients. So that awareness and literacy builds the guardrails, and on top of that, when you put governance, you sort of contain and manage the probability that somebody will do something wrong. Ethics is just as much an art as it is a science.

Sanjog Aul [00:32:24]:
So while we are talking about this, and you mentioned that healthcare is something which has got an emotional component of it, when it is practiced, it can create a lot better results versus us having the sterility in the whole process. But then at the same time, there is a school of thought that if you try to get too emotionally attached with the subject, then you will lose your objectivity as well. Now, when you’re bringing AI into the picture and more and more technology into the picture, and you’re trying to do more data driven stuff and you do not have a soothing hand, if you will. If that has given any value in healthcare as a business, or in terms of what the patient is expecting, then where is it taking us? Are we becoming a bunch of robots, while we may still look like humans and trying to practice medicine? Or we will try to train AI in making sure that when it gives us the decision, it is not just data driven analytics that is driving it, but also what it could do to a person’s ability to think positive about what’s the outcome that’s going to be for them, and that itself has a long term impact. I’m getting into an area which could be fuzzy, but we cannot ignore it, because we are trying to supplement humans with technology, but we are going to reduce the sum total of emotions that are required in healthcare for you to really take care of the patient.

Cherodeep Goswani [00:33:58]:
Yes. So I’ve probably said this over a dozen times in various such interviews and podcasts. I hope in my lifetime we never have technology replace humans 100% in doing a medical diagnosis. I hope it doesn’t happen. To me technology makes the provider a better provider, and I would say, as many of my colleagues have said, it’s not that a provider cannot write a summation note or cannot write an email response. The enemy is not the provider or their intelligence. The enemy is the person’s time.

Cherodeep Goswani [00:34:34]:
Time is the enemy. We all have 168 hours, and where do we spend that every week is the challenge. So wherever we use this kind of AI, again, I use the words use AI and automation for the things that people don’t want to do, use it for things that people cannot see, use it for things that people cannot understand. Use it for those three things that makes the person a better person, a full person, and then help in the interaction with the patient out there. So to the best of my knowledge, 100% of the use cases that we are using in a clinical setting using some form of AI, a human being is still the final actor in that supply chain of actions that puts their stamp of approval to say, this is my diagnosis, my note, etc. Because that’s our way of saying you came to be treated by a doctor, taken care by a nurse, taken care by a pharmacist. You did not come here because we ran a great technology. That’s my opinion. That’s the way I lead my teams.

Sanjog Aul [00:35:43]:
When it comes to, you use the word lead. So I’m going to ask you this one last question about leadership. So given predictive care as the construct within which we are trying to have this discussion, we spoke about operational challenges, ethical challenges, then privacy, and many other things.

Sanjog Aul [00:36:02]:
You got technology leaders, you got clinical leaders, and you got business leaders. Three constituents who are, in a way, running this organization one provider at a time. Let’s start with one provider at a time. What would be the appropriate, not best, not anything else, because it could be relative, but what would be the most appropriate mindset to wear as a leader? What would be the most appropriate communication style to use as a leader? And what level of collaboration would be most appropriate to create most value for predictive care? So the ethics taken care of, operational efficiencies there, patient gets what they want, and you get to a point where the business side of it truly flourishes so you can create more and more value. What would that look like?

Cherodeep Goswani [00:37:00]:
Yes, if I repeat this for my own memory. So, technical leader, business leader, clinical leader. Clinical leaders, that’s the way you said, right? Good organizations do this all the time. It’s easy to be one of those three. Truly the word leader is one who does at least two out of those three. Those with a clinical background do all three, which is where you cross pollinate the greatest minds and train the future minds to come together. It’s not a question again of either or.

Cherodeep Goswani [00:37:28]:
Which is why we say technology in many ways is the bridge between the mission that we all signed up to do and the margin that we have to create, that further augments the mission for the next day, month, year, decade, so on and so forth. So when you start with any form of technology, we’ve been talking about AI, but say the same thing about cloud. Cloud is storage. Does a patient go to a hospital saying, hey, what’s your cloud? No, they come in because they want the best provider or the best care, and we, the technology leaders, become the translators in ensuring that the margin and the mission are always aligned on the same page out there, and making it happen. A senior leadership cabinet in any organization constitutes of domains of all of those three. But each of us can walk in someone else’s shoes within the limitations and say, here’s what we agreed to do and here’s where we disagree to do so.

Sanjog Aul [00:38:25]:
So if you were to leave a message for the leaders out there, if they could do better than today, what would be your appeal, so that we achieve this goal of hopefully one day crack the code fully on the predictive care and everybody sings Kumbaya, including patients and the provider.

Cherodeep Goswani [00:38:44]:
So I would leave by saying, don’t let perfect be the enemy of good, because there is no such thing as fully. As long as we move the needle and make tomorrow better than yesterday, that is progress. Two, for technology leaders, I would say if you’re in the business of healthcare, learn the business of healthcare. You cannot design an EMR sitting in a glass house. You have to see the EMR being used at the ED on a Friday night shift. You have to see the EMR being used during a transplant case, where behind five happy families there is a very sad family. That’s where we design better technology and create outcomes, versus just writing code for the sake of code, and I’m a programmer that’s telling you this.

Cherodeep Goswani [00:39:30]:
So that’s the healthcare leadership, technology leadership in healthcare, that we also have to identify. And lastly, any leader in healthcare needs to understand economics. This is a zero sum game.

Sanjog Aul [00:39:46]:
On behalf of the show and our audience, thanks so much Cherodeep for sharing your insights about this rather fascinating area of predictive care. And we did cover quite a bit of ground beyond predictive care, and we spoke about operational challenges, ethics, leadership and all that. And my sincere hope, working along with leaders like you, to keep spreading the word about doing the right things when even no one is looking, so that we have a better society, healthier people and thriving businesses. So thank you.