AI World Series – Part 1 – Sandy Aronson – Partners Healthcare talks Implementation of AI tools for Providers
Episode Summary:
Harry’s guest Sandy Aronson argues that AI and apps are not the solution for better healthcaree; more effective care workflows *enabled* by AI and apps are the solution. Aronson is the executive director of information technology at Partners HealthCare Personalized Medicine. His team develops the IT infrastructure needed to support genetic-based personalized medicine in both patient-based and laboratory settings. This episode is the second in a two-part series on getting AI, machine learning, and analytics working in the healthcare provider setting, recorded as part of the AI World conference produced by Cambridge Innovation Institute in Boston in December 2018.
Please rate and review The Harry Glorikian Show on Apple Podcasts! Here’s how to do that from an iPhone, iPad, or iPod touch:
1. Open the Podcasts app on your iPhone, iPad, or Mac.
2. Navigate to The Harry Glorikian Show podcast. You can find it by searching for it or selecting it from your library. Just note that you’ll have to go to the series page which shows all the episodes, not just the page for a single episode.
3. Scroll down to find the subhead titled “Ratings & Reviews.”
4. Under one of the highlighted reviews, select “Write a Review.”
5. Next, select a star rating at the top — you have the option of choosing between one and five stars.
6. Using the text box at the top, write a title for your review. Then, in the lower text box, write your review. Your review can be up to 300 words long.
7. Once you’ve finished, select “Send” or “Save” in the top-right corner.
8. If you’ve never left a podcast review before, enter a nickname. Your nickname will be displayed next to any reviews you leave from here on out.
9. After selecting a nickname, tap OK. Your review may not be immediately visible.
That’s it! Thanks so much.
Transcript
Harry Glorikian: Welcome to this two episode series of Moneyball medicine on AI machine learning and analytics focused on how we get these technologies working in the provider setting. This special series was recorded as part of the AI world conference produced by Cambridge innovation Institute. This last December in Boston, Massachusetts.
I am your host Harry Glorikian in this. I will interview two of the speakers from the event and we will hear their experiences. We will dive into the challenges and opportunities they’re facing, looking at how we implement AI and machine learning into clinical practice. Welcome to Moneyball medicine.
My next guest has said that AI is not the solution and that apps are not the solution, but that more effective care workflows that apps and algorithms enable are the solution.
Today we’ll be discussing how integrating algorithmic techniques into clinical care and decision making can really work.
My next guest is the executive director of Information technology at Partners HealthCare, personalized medicine, Samuel, or Sandy Aronson. His team develops IT Infrastructure required to support the evolution and practice of genetic based, personalized medicine in both patient facing and laboratory settings.
The system ecosystem, the team maintains enables a real-time continuous learning process. This process harnesses clinical testing workflows to advance knowledge surrounding genetic variation and deliver this knowledge in near real time to clinicians treating patients. Mr. Aronson is one of the founders of GeneInsight.
He co-chairs the displaying, an integrating genetic information through the EHR initially organized under the Institute of medicine. Now, part of the F H I R foundation, Mr. Aronson also co-chairs the Emerge network, EHR integration, working group and serves on the scientific advisory board for the clinical pharmacogenomics implementation consortium.
Mr. Aronson holds a master’s in organizational behavior and a Bachelor’s in computer science from Stanford university. He also received a master’s in biology from Harvard extension school. Sandy, welcome to Moneyball Medicine.
Sandy Aronson: Thanks for having me Harry.
Harry Glorikian: So Sandy we’ve known each other. Actually, I was thinking about it for quite quite some time now. I remember when this was sort of first, you know, an idea, um, back when I was even talking to Dr. Roland’s, but can you tell me something or tell us something about your group and how it’s evolved. Say over the last 5 years.
Sandy Aronson: Sure. So, um, so my group actually going back a little bit further is, is actually a little over 15 years old.
Um, we came in because Partners Healthcare recognized that genetics and genomics had the potential to fundamentally transform the way we care for patients. Um, but at the time Partners didn’t know, um, how quickly that transformation would happen, but it knew that really significant infrastructure would need to be built and that would take time and they wanted to get started on it, not, um, not just from a kind of laboratory and infrastructure perspective, but also from an IT infrastructure perspective.
So, so we started building infrastructure. Again, a little over 15 years ago. Um, the very, very beginning was focused on wet lab infrastructure. But within the first year we realized that sort of the management of clinical interpretation of genetic data was, was where the real challenge was. So we started working on that.
And as you mentioned, um, we built, um, in addition to, you know, systems to integrate the, kind of the, the diverse types of instruments that are involved in genetic testing, also a system or a suite of applications called GeneInsight, that really focused on, um, managing the report generation process, the knowledge management surrounding specific variants that are found in patients and making sure, that if new information emerged on one of those variants, um, you know, we had a mechanism for, um, integrating with the EHR when possible and delivering that new information back to clinicians.
Um, so if, if that information for patients, you know, where variants had previously been found, if that new information could materially improve their care. So we took that infrastructure, um, and it really did create, as you said, an instance of, of a true continuous learning environment, that was IT-mediated not like machine learning. It was manually, you know, facilitated learning, but it, it learned as a result of, um, clinical transactions through the system. So we recognize that its ability to evolve would be limited based on our own internal clinical genomic testing volume. So we registered it as a medical device.
We began distributing it outside of Partners and we network the different instances together. Um, so that labs could learn from each other and then little over two and a half years ago. We sold this infrastructure to, um, to Sunquest. And the reason we did that was because we, um, we wanted to make sure that in Sunquest, I believe, had the largest number of, or at least at the time had the largest number of, um, installations in pathology departments.
And, and, and we, you know, we’re, we’re fundamentally not a healthcare software, you know, distributor and, um, we, you know, we wanted to make sure that, that the system could grow, um, to the greatest extent possible. And after that sailed, the vast majority of my team went to Sunquest and the team that stayed behind there were four of us.
Um, and we started looking at what we do next and what we wound up focusing on was, um, really this problem of how do you get innovation? Exactly what you were saying at the beginning. Harry, as you know, the world is starting to shift as, as these digital technologies really enable us to, to think about care pathways differently, you know, how do you interject?
How do you enable clinicians to truly in the real world take advantage of those technologies? And shift their clinical process flows in ways that will be more effective.
Harry Glorikian: Well, you’ve said before that, you know, that, the system is primed for radical disruption. And so what you say that, what do you mean by that?
Is that, is that a, um, AI machine learning sort of a revolution is that, um, you know, uh, how the payment system is working or where care will be provided. What is, what is it when you look at it from the lens you’re looking at, what is the radical disruption mean to you?
Sandy Aronson: Yeah, so the thing that I feel I’ve learned over the last two and a half years, is that, I think that the, I think the most profound thing that’s happening is what I’d call algorithm, um, enhanced care, algorithmically, enhanced care, where that can be machine learning algorithm, but it certainly doesn’t have to be, but it’s some type of algorithm that may operate on a new type of data that wasn’t previously available or may not, but that assists decision-making in some way.
And the, you know, when we started this process, we felt that the change that we were going to, you know, be seeking to bring about was interjecting these capabilities kind of surgically into clinical processes to give clinicians the ability to use new types of data, new algorithms to improve their decision-making.
But what we’ve actually found is while, that’s useful and, and I think, you know, a very meaningful thing to do, there’s a broader opportunity, which is once you have these capabilities, you can actually look at how the clinical process itself could be redesigned around them. Um, and once you start doing that, then everything comes, you know, onto the table.
You know, the way you engage with patients, the frequency with which you engage with patients, I do believe, as you mentioned, Harry, that that payment models could, could wind up shifting on this. It could enable sort of new ways for, for new players to take risk and, and, and by taking risks enable even more significant changes in clinical processes. So, so, so like algorithmic care is the fulcrum and it enables you to pivot, you know, in ways that go beyond just what the algorithm itself is doing.
Harry Glorikian: Yeah. It’s interesting that you say, you know, um, I was just also speaking to Dr. Patel, who was also at the same meeting we were at and we were- because he has, he’s not just a physician, he’s an engineer so slightly, um, you know, interesting background on, on, on that. But we were talking about how, you know, is it the implementation of the machine learning or AI into the existing workflow that has the power, or is it rethinking the workflow to really maximize the benefit of the machine learning or AI system. And what I’ve seen in other industries is, is, you know, the latter, not the former. And what you just said is plays right into that is, uh, you know, how, how are we going to make that shift? How are you, uh, working with clinicians right now to implement that sort of workflow shift, mindset shift, you know, educational understanding at, at Partners.
Sandy Aronson: Yeah. So, so I do believe, I, I do believe that it’s the latter. I think that it’s like these complete workflow redefinitions that, you know, that have the most power to, you know, improve things. Um, I think that, but at the same time that said it is a continuum. Right. So, so any intervention that you do that works is going to cause people to do things differently. And you can look at that as a workflow change. It’s a question of sort of how much of the workflow is changing. So in terms of some specific examples, you know, one of the amazing, you know, privileges of, of working where, where, you know, I work is that you, you have these amazing clinicians who come up with these incredible ideas.
Um, and you know, it’s, it’s, it’s not me or my team that comes up with the transformations that are possible. We get to work with clinicians who, you know, who have that idea who typically have started to work on the new workflow and, and, you know, and then the types of IT support that we can bring to the table can enable them to really accelerate, um, the, the, you know, bringing that idea to fruition and, and having an effect as many patients as possible. So some specific examples, if we, if we go with a, a, you know, a very focused example, we have, um, we’ve built a solution that’s deployed now we’re gathering data on it, that enables a blood bank technician to um, to select the best unit of platelets for an individual patient, based on, you know, the HLA profiles of donors who have platelet donors who have units of platelets and inventory, and, um, the platelet recipient where, you know, previously the way that this process worked is about 15% of the time when we gave platelets to bone marrow transplant patients, the bone marrow transplant patients, white blood cells would immediately destroy the platelets because of a lack of NHLA match between the donor and the recipient, which, you know, there’s an economic cost to that, there’s a clinical cost of that, there’s human cost to that. Um, And it would often take, you know, three to seven tries before we, we recognize that the patient we were dealing with was particularly hard to match and we’d have to do something special. Now what we do is by marshaling data in a whole bunch of different systems, we actually bring them to a women’s hospital, the technician goes to, to take a unit out of inventory and as opposed to just taking the oldest unit, they can take the best unit, um, for that patient. And we think that this is going to have, you know, a really positive effect on sort of the whole process of administering platelets to bone marrow transplant patients.
Harry Glorikian: So it’s interesting you say that it’s the person doing the work, but it’s interesting that I, I look at it and say, well, they it’s the it’s the system that is a sub selecting based on data and offering it up to them. Uh, they may have a choice of like three or four, but it’s offering those three or four as opposed to the potential hundreds or thousands that, that were a different option. So it’s, it’s a machine human interaction that I think that that’s sort of interesting to get to a better level of care.
Sandy Aronson: Yeah. So what we’re doing is, and I should say these, these units of platelets are. They’re precious. So, so you don’t, you don’t necessarily have you, you have a limited number of them in inventory at any given time, um, and what the system does. It, it, it does a lot of things in terms of giving them information on, you know, how, how hard this patient appears to be, to match, um, you know, historical platelet values for the patient, things like that.
Harry Glorikian: So, um, Sandy you’ve said in the past that that interjecting new types of data is critical. So, uh, how do you, how do you inject new types of data and algorithmic support into care delivery process so that you can, I don’t know, help improve, clinical decision-making whether it’s from the physician, whether it’s from a nurse or anybody who’s sort of managing that patient.
Sandy Aronson: Yeah. So all of this comes down to really understanding what the current workflow is and, and, and what the goal is. You know, how are we attempting to perturb the workflow? So that either a better decision is made or a better form of care is delivered or care is, is delivered in a way that’s more efficient for both the system and the patient.
So it’s really understanding, you know, the change that we’re trying to make and then working with the clinicians to identify what the best way is to really work through iteratively. Usually, um, What type of IT support can be used as an intervention to get us to the new, more powerful workflow. And then, you know, as you’re figuring that out at the higher level, then there’s always, you know, a huge amount of work associated with, um, with data access, with UI design with information security, um, often with, you know, figuring out how you represent algorithms robustly, um, and you know, and maintain them over time. Um, there’s a, there’s a whole bunch that goes in behind the scenes, but at the highest level, it’s that partnership between the, the team involved in the patient’s clinical care, um, and the IT team that brings it about.
Harry Glorikian: So just, but just out of curiosity, when you say interjecting new types of data, what, what would be some new types of data? I mean, I’m, I keep thinking about, um, you know, papers like Joel Dudley wrote in, in June. And I keep referring back to this over a, you know, again on this podcast, but, you know, bringing in multiple types of data that then point us to the, what could be the driver of a disease or something. And when you say new types of data, what do you mean by that?
Sandy Aronson: Yeah, it’s a great question. So, so I was speaking a little bit ago about, um, this, this application that we built for, um, for platelet management and the central, you know, new type of data. And I do that in air quotes.
Is the HLA values for the patient and the donors, which, you know, we always, so we’ve always, you know, um, tested the patients. We always generated HLA profiles for the bone marrow recipients, because we need to, in order to match, match the bone marrow, um, and we have, there’s actually a relatively limited number of incredibly altruistic people who come into, you know, sit for the two hours while their blood is circulated outside of their body to donate platelets that make the bone marrow transplants possible.
So, because it’s a relatively small number of people, you can type them as well. Um, And then once you have those two types, then you bring them together in an algorithmic way. So, so it was really the platelet management use case was really about HLA data. A great deal of our work today is with Callum MacRae’s team, um, at BWH Cardiology, Innovation.
That really focuses on, you know, completely changing the way that, um, we manage lipids, hypertension and heart failure, and that works starts with, um, the really looking at the process redesign itself, um, which is made possible by instantiating um, Algorithms, or guidelines into algorithms and then putting in place a whole process for, you know, for leveraging them to improve care.
Um, but once you have that framework for, for managing a specific disease, then you can look at how you bring in orthogonal data and how you manage that. So that started for us with things like, you know, integrating home blood pressure cuff data into the process, but you can easily see how many additional forms of data will be useful.
One of the ones that, that seems most compelling to me is, um, you know, the geospatial data that can be gathered through watches, et cetera, and, you know, and seeing how it seems likely that there would be signal there that would help with understanding how, um, you know, heart disease progresses. Um, so, and then there’s, you know, there’s other types of data in other areas that, um, is potentially potentially useful.
Harry Glorikian: Yeah. I mean, there’s some, there’s some amazing things that are, that are coming up. Um, when you look at the input of the data, but then the, the ability to actually crunch that data. And when, I mean crunch it’s, you know, call it AI machine learning, call it whatever name you like. Um, uh, plus all the processing power and the ability to store much cheaper, um, and easier than we ever have before and starting to see some incredibly interesting things that we never were able to see before, which I find fascinating.
I mean, how do you see the, how do you see care changing based on your prediction of the future? How do you see the business model itself being affected by this, uh, and where you and I sit from these technologies and these data sources.
Sandy Aronson: Yeah. It’s interesting. And a bunch of what I’m about to say. I learned from really working with, with Calum MacRae and his team. But I think that, I think what’s going on here is for, for certain classes of patients, there are guidelines which when you implement them robustly, you know, improves that patients, you know, those patients outcomes. And so there’s, there’s a need, but those guidelines are often difficult to implement in the setting of modern healthcare because of the level of interaction that’s required between the patient and the person administering their care.
And that creates a world where if you can interject new capabilities, for particularly, um, you know, interactions based on, you know, phone and text, um, with, you know, not only clinicians, but also people like patient navigators who can really work closely with patients to optimize their care. Um, you know, when you start putting in place these capabilities there, you know, fundamentally you can think of these guidelines as algorithms that can get better over time.
In order to robustly use an algorithm in the course of modern healthcare, you have to robustly obtain the data that the algorithm operates on. So in order to put all of this into action, You need to transition to a new process and support that process that enables you to, to robustly collect this data over time.
Once you’re robustly collecting that data, then you’ve got new types of opportunities for machine learning based on data that’s potentially has more signaling in it then the data that was collected before you instituted the new process, it also enables you to, to under, you know, very careful using very careful, you know, um, interventions, add new types of data to, to the data being collected and see whether it’s potentially helpful or not.
So there’s, there’s really a continuous improvement cycle that can be set up here where essentially by implementing the algorithms, you have to standardize data collection and improve data collection. And then as a result of that, you gain the data required to continuously improve the algorithms as well as assess, you know, whether new forms of data as they come in, alot online are actually helpful or not, and integrate those new forms of data they are.
Harry Glorikian: So it’s, it sounds like it’s, it’s interesting, right? I mean, I think if, if you go back in time, there was a lot of sort of trial and error that went on to you know, develop like best practice, but now we’re going to, uh, it seems like a new era of trial and error because of the power that these capabilities now bring to the table to enter that next phase of, of continuous process improvement.
Um, I mean, I know with healthcare that’s, that’s everybody goes, uh, not work, but, but, but doesn’t like that necessarily because they, they want something absolutely proven, but it seems that we’re in that era of, you know, you just handed somebody, you know, the super computers, you got to, you got to try them out.
You got to see, you know, what could completely change the outcome for these patients. And I think the sensors that, you know, we’re, we’re seeing deployed in, in the field. I mean, just looking at, you know, the Apple watch on my, on my arm is it’s got to be a complete new paradigm to how care will be managed. And, um, you know, patients can be actually influenced right in their own home, as opposed to that one time a year when they come in and see the physician.
Sandy Aronson: Right. Right. So, so in today’s world, there is a great deal of trial and error in medicine, right? And in fact, some of these guidelines actually in a way call for that.
So they call for, you know, starting with an initial drug, seeing whether it has the desired effect. If it doesn’t have the desired effect, you know, titrating up the dose of that drug. Until, you know, you hit side effects or you hit the desired effect, you hit side effects, then you, then you go to the next drug or you try a combination of drugs or you titrate back.
Um, so there is maybe trial and errors, the wrong, um, way to describe it. But there’s an iterative process associated with arriving at the best treatment for the patient. And so, so that’s a process that, you know, we want to get, we want to make sure is as efficient as possible. You just look at it in terms of instituting the guideline you want to, you want to perform that optimization as quickly as possible, because we know that that’s, that that’s good for patients, but that’s good for mortality.
So-
Harry Glorikian: I guess when I say that I mean, I almost feel like this is when I would, you know, and I don’t do this anymore is sort of, you know, work on my engine in my car when I was younger is, you know, you, you now you’ve got the ability to see this constant stream of data coming from multiple sources. Um, and, and you, I’m not going to say you can fine tune it in real time, but it, you almost feel like you can start to head in that direction.
Sandy Aronson: Yeah, so, right. So you’ve got, you’ve got a couple of different modes here. You know, one mode is, you know, really trying to figure out are texts sent at 9 am more effective than text set and 10:00 am in terms of hitting a particular outcome. So, so understanding sort of fine tuning, you know, that aspect of the, of the, of the process.
You’ve got another mode where, you know, looking at data that you don’t know whether it’s relevant or not, but is associated with this process that, you know, you, that, that you obtain you know, in concert with the patient, you know, with a fully informed patient, um, to, to try to ascertain whether there’s signal in this data in terms of it being able to, you know, help the care delivery process, sort of collecting data in parallel with the existing process. And then, you know, you ultimately, you have, you progressed to where you actually want to do something that looks like a trial and you want to try, you know, things two different ways, prospectively and see which way works better.
And it-one of the nice things about the infrastructure that’s being created is it should make those trials much, much more efficient to do than then they can be done today.
Harry Glorikian: So if, if you had to like, I don’t know, I don’t want to say pick your pet peeves or, or think about the impediments or the big, your biggest challenges in what you’re trying to do.
And then the implementation of it for those people are, you know, who are listening and want to hear what, what really are the things they need to focus on when either getting the information or, or getting, getting the output to the right people. What, what would they be? What would be the key factors?
Sandy Aronson: Yeah, that’s a great question. And I guess it depends on the context. So if the question is, you know, you’re operating within, if the context is you’re operating within, you know, a, in an academic medical center and your goal is to work with the clinician who is, who both has the source of funding and has a strong idea for how they want to perturb their process flow.
So, so on a high level, the project is, you know, the project objectives are well-defined the funding’s well-defined. Then the biggest challenge is um, become, um, things that are highly related to the technical process. So obtaining the data, winds up being, you know, a significant part of the project, um, actually doing the user interface, design, designing the new workflow iteratively, also a big part of the project.
And you know, where, where. Perhaps, and of course, you know, quality assurance being critical where one of the biggest risks that you face is that you’ll build something that maybe even, you know, tests well, but doesn’t truly get adopted. It doesn’t truly get adopted in a way that that improves care. So to protect against that you wind up needing to get so deep into process design with the clinicians and, and, and really focus in that way.
But that’s, if your context is, you know, within the AMC, within a defined project, if your context is sort of pre that, where you have an idea how to fundamentally transform a process, but you don’t have funding yet. And you know, and this isn’t an internal sort of, you know, quality process improvement project. This is, this is something that’s, you know, that, that you’re going to try to convince people to, um, to, to adopt and, you know, in the broader world, I think then the biggest challenge becomes understanding how this will fit with how the change that you want to Institute will fit with reimbursement models.
Once it has been proven to have the impact that you’re looking for it to have. And whether you actually, in order to Institute the change that you’re going after, whether you actually have to fundamentally introduce an entirely new reimbursement or, or risk management model, um, or whether you can fit within the current economic system.
Harry Glorikian: Yeah, it’s a, in some ways quite daunting, depending on how are you going to put one of these things? Sometimes I always think that doing it a completely different way is always easier, but I, you know, you’re living within the context of, of the institution that you’re in, which is, um, you know, what many of the people that we, uh, we saw at the conference are sort of, uh, living in is how to, how to make change and implement these capabilities in the institutions that they’re, they’re working in, uh, which is always not easy because there’s an existing status quo and a process to doing it.
Sandy Aronson: Yeah. Now that said, you know, when we work internally, our intention is to open source our work and to distribute it as broadly as we can. Um, You know, the other thing that is likely that we will do is, you know, seek to use these capabilities, to enable ourselves, to provide care to more people than we can currently, um, provide care to.
So our ultimate goal is for the infrastructure we’re building to, you know, to, to have a much broader, much more positive impact than we could achieve. Just through, you know, internal use for our existing patient population. But that said, I do think it is easy to underestimate the resources and capabilities required to affect these transformations.
Um, you know, a lot of times folks feel that. If they just could, if they could just train this machine learning model or if they could just build this app that, you know, that, that, that that’s going to affect the change that they’re looking for. And the reality is that to affect change, you have to change what people are doing on the ground.
You have to be effecting you know, clinical processes. And once you’re doing that, you’re into something that I think is much larger in scope than folks realize. And therefore it does favor, you know, actors who can put a lot of resources behind what they’re trying to do. It becomes harder for a smaller actor to actually be able, even if they have an awesome idea to, to, to really be able to get it off the ground.
Harry Glorikian: Well, we were, you know, when I was talking to, uh, to Dr. Patel, it was saying that, you know, we need to be trained physicians and, and, um, think about it almost like a process engineering course that everybody needs to take so that they it’s much easier to implement these things in the future.
Sandy Aronson: Yeah. And, and I do think in my view, this is going to happen process by process.
You know, we’re going to, we’re going to learn things we’re going to become better and better as, as time goes on. But I do think that this is an exercise and this isn’t like there’s some Meta, you know, high level like application that we’re going to build. That’s going to transform all of medicine. No, this is, this is ultimately taking a look at every clinical process and saying, you know, if this clinical process was rebuilt today, in the context of all of the digital and algorithmic capabilities that we have, how would we rebuild it?
And how much better would that process be than what we’re using now? And if the answer is a lot, then the question becomes what types of intervention, what types of IT support or other support is needed in order to leverage, you know, that old process, you know, over to the, to, to, to the new improved process.
Harry Glorikian: Yeah. And, and, you know, I think that there’s going to be the groups that do and survive and thrive in the future. And I think there’s the one either don’t or don’t do it as swiftly. And that, that, that, I’m not sure where it leaves a system like that, uh, based on where we’re headed.
Sandy Aronson: I hundred percent agree. I think that, um, it’s exactly like, you know, we were looking at 15 years ago with genetics and genomics, you know, we know, at these kinds of capabilities are going to be incredibly disruptive in a good way. We, but we don’t know when, you know, this fundamental shift begins to happen and maths, but we do- but at the same time, we do know that there’s a whole, whole new infrastructure that needs to be built to enable all of this. And that infrastructure is going to take time to build. So if you wait too long, you can wind up really behind the eight ball here. You know, where, where it’s it’s, it’s the type of thing where it takes time.
To establish support for these new processes. So you can’t react on a dime, it’s just not possible to react on a dime. So you have to, you know, figure out where you proactively want to make your bets, where you can make the most difference. And, um, you know, I, I don’t know that reactive works in this model.
Harry Glorikian: No, I don’t think so. I think that’s going to be almost impossible.
Harry Glorikian: Well, let’s hope that most of these, the institutions, um, you know, at least in this country, if not worldwide, sort of make the proactive changes they need to make. Um, I, you know, Thank you so much for joining us on, on the show today. I’m sure the listeners got a lot out of the, you know, the, the wisdom that you’ve accumulated over all these years of, uh, of trying to make all these things happen.
Um, and I, and I, and I wish you great success, uh, going forward.
Sandy Aronson: Thanks, Harry. Really appreciate it.
Harry Glorikian: Thank you.
Sandy Aronson: Take care.
Harry Glorikian: And that’s it for this episode. If you enjoyed Moneyball matters. Please head over to iTunes to subscribe, rate, and leave a review. It is greatly appreciated. Hope you join us next time until then farewell.