Summary:
The integration of augmented intelligence (AI) in clinical settings is revolutionizing healthcare, but it also brings its own set of challenges. In this episode, Maria Granzotti, MD, MBA, CPE, CHCQM, FABQAURP, FACHE, dives into the current landscape of clinical augmented intelligence (CAI). She defines CAI, discusses its primary purposes and benefits, and examines its potential to reduce healthcare costs.
You cannot read a newspaper or watch a stock market report without the term AI being included. As a physician friend told me, I am not sure where we are going with AI, but we are on our way.
This transcript of their discussion has been edited for clarity and length.
Michael Sacopulos: Today's episode, we will explore the difference between AI and CAI. Prepare to discover where clinicians are heading with CAI, next, on SoundPractice.
Maria Granzotti is a physician with more than 20 years of emergency medicine experience. Dr. Granzotti has served as a chief medical officer, she is an expert in current trends and technology issues and medicine. Maria Granzotti, welcome to SoundPractice.
Maria Granzotti, MD, MBA, CPE, CHCQM, FABQAURP, FACHE: Thanks for having me, I really appreciate it.
Sacopulos: It is my pleasure. We have a number of interesting topics to get into today, but before we do that, could you please tell me a little bit about your career path?
Granzotti: Well, it is a bit circuitous, but I have always desired to be able to connect the dots in various spaces in healthcare, and I really think the seed for my passion for healthcare was from my grandmother. She was an Iowa farm wife, she was a nurse for over 50 years, and she worked at night and worked the farm during the day with my grandfather and my uncle. So, I had a good example set in front of me. So, I went to medical school at UT Southwestern in Dallas, and after having gone to undergraduate at the University of Notre Dame, which, my passion, I have to say that. And then, after that, did an internal medicine residency in Chicago at Rush Presbyterian St. Luke's, and that really opened my eyes to, again, connecting the dots. And I loved so many aspects of medicine that, after that internal medicine residency, I actually practiced in the emergency room as a teaching attending there, teaching residents for over three and a half years, and subsequent to that, went on a journey 20 years in emergency medicine in Illinois, majority of that in Illinois.
After that, about 2009-2010, knowing that there was just so much more to do, and I really had an interest in quality and safety and compliance and regulatory issues, that I became, that was my first chief medical officer role with the health system, and that expanded to divisional roles, etc., got my MBA. Now, it gets a little boring. And then, my last large role was as chief medical officer for Ascension in the entire Texas market, so over 14-15 hospitals at the time. Subsequent to that, my husband passed away suddenly in 2020, so I really took a stop and tried to listen to myself and what the industry was doing and where my interests lied, and where I wanted to continue to grow. So, I started consulting, and also was a chief medical officer for a national company, and decided the tech space is just so rich and ripe for further innovation and application in the clinical space that that is what I am doing now, I am consulting in various areas, and one of my very passionate interests is the digital and technology space.
Sacopulos: Excellent, well, that takes us right into our first question, and it is about CAI. Can you tell me what CAI is?
Granzotti: CAI is clinical augmented intelligence, and it is not a platform, it is not the everyday alerts that the majority of physicians have gotten used to, such as, you are trying to order a penicillin derivative antibiotic, an alert pops up. Those are good in the right space and time, but this is more of an augmentative type of support, evidence-based support, for clinicians, whether it is a nurse, whether it is a technician, or whether it is a physician. Any type of provider could partake in this, and it is building supportive assistance in the clinical workflow within the current platform, within the current management of information, but again, building upon that, bringing forth more of the standards of practice to facilitate the delivery of care. That is what CAI is, clinical augmented intelligence.
Given the plethora of information, and you hear about newness and evidence, and physicians have a tough time keeping up with, what is the latest and best for individual patients? And that is where that individual uniqueness of this support can be applied. It really builds on the experience of clinical decision-making. I know there is a lot of words in there, but as a clinician, if You are in a workflow in decision-making, you want the best information possible, not only for that individual patient and what you are treating, but any therapeutics, any diagnostics that can be facilitated with that augmented information in your workflow. So that is what CAI is. Again, that is a lot, and it is going to continue to grow, but it is very exciting.
Sacopulos: How does it differ from simple AI, which is not so simple?
Granzotti: No, that's a great question, because CAI is the interface, it's the evidence of the interface of computing in clinical activities, whereas AI, if you think of artificial intelligence, regardless of what space it is in, healthcare or otherwise, it is the programmatic achievement of a given result. It is not requiring human interaction to get to that result, it is algorithms, it is a programmatic end result that is delivered, whereas the difference with CAI, clinical augmented intelligence, it is facilitating the product of organic improvement. It is dependent on the human interaction. There is an interface for information, but it is requiring the human interaction to decipher, how does this apply in this clinical setting? How does this apply to this individual patient? So that is the difference, one is more one-size-fits-all, let's say, AI, whereas clinical augmented intelligence has the ability, it is not a one-size-fits-all, it can be very specific.
Good examples right now for AI to continue that, the differentiation, for AI is in diagnostics, in mammography, in results, in reading those images, that is very powerful, also with precision medicine, in the right space and time, whereas augmented would take even more of the aspects of that individual that you are caring for in your care delivery, bringing up maybe some past safety concerns with that individual. It is the confluence of that data, that information, which can support your clinical decision-making. It is more of a, not only diagnostic, as AI is right now facilitating, let's again, using the example of mammography and radiology studies, but it is also therapeutic, we are bringing in the realm of therapeutic delivery and therapeutic uniqueness for each individual patient. So those are the big differences.
Sacopulos: And is the human interaction at the clinician level, or is it at the patient level, or both?
Granzotti: That is a great question, and it is facilitating both, because what clinical augmented AI does is that it opens that space for the clinician to be able to absorb, sense, interact more with that individual patient in their care delivery versus poking on a keyboard, searching for past results, searching for past information that can be facilitated with the interaction with the technical aspects of CAI.
Sacopulos: So, as you know, medicine is both an art and a science, and it seems like there may still be a little art left when using CAI. Is that right?
Granzotti: That is the most exciting part for me, and that is interesting you got right there. CAI has the ability to bring value and resilience into existing workflows, existing interactions, but it only heightens that human essence of that interaction. That interaction is so sacred between a clinician and a patient, and CAI has the ability to promote and build and continually learn from, that is the other thing, continually learn from every interaction between that physician and that patient.
Sacopulos: Who is learning clinician behavior, as well, is that correct?
Granzotti: Absolutely.
Sacopulos: So, it would be customized to or preferential to different clinicians' behavior or practice?
Granzotti: No, that is a great point, because if I am interacting with a patient, taking a history, dependent on what they are presenting with their signs and symptoms, their past medical history, I may go down various orders or various paths. There is some structure to it, but you can go down various routes to get more to the information that you as a clinician are used to assimilating, you as an individual. If the system, if the technical interaction can be cleaned up, let's say, and be more supportive and augmentive to what you are asking and what you are doing, it is going to make that interaction that much richer, and guess what? You may be learning a lot more than you would have otherwise, wasting time searching and digging for information that is relevant to that visit.
Sacopulos: Does it also help to reduce unconscious bias in clinicians?
Granzotti: Yes, yes, I do think it really has that potential. Now, people talk about hallucinations in AI, where if the information, if the data integrity is not rich and clean, let's say, there can be a propensity for hallucinations in AI. This really can help to facilitate cleaning, again, I keep using the word cleaning, but scrubbing that, scrubbing that. I think it is also that you bring up the bias, because in medicine we have learned that the social determinants of health really have an impact on people's care delivery and longevity and their disease experiences. That infusion, that is a good example of an infusion of information that could really help the clinician in their approach to their individual patient, and then, not only learning from those individual interactions, but the system, again, can learn to population, can learn to different type of physicians, all supportive and all augmentive.
Sacopulos: You talked about this from the medical point of view or from the clinician's point of view, what about from the patient's point of view? How do patients feel about this? Because I could imagine that it might be a little scary to certain patients.
Granzotti: Yeah, patients right now, I mean, historically, patients have been aware and have sensed the dependency on the technical, whether it is the documentation, whether it is registration, whatever it is, this is really trying to take that away, trying to take away the clinician having to spend so much time at the computer, at the device, whatever it is. Obviously, there are necessary times for that, but it just opens the time and space and interactive capabilities of a clinician to the patient, and vice versa. Patients are very smart to try to learn and engage more with their clinicians, with their ecosystem of care, with their caregivers, that is always a better environment for patients in their care.
In this way, patients can ask more, can relay more, it just gives more time and space for that, rather than having to continually spend so much time on the keyboard. The other thing about clinical augmented intelligence is that there are capabilities for speech. I mean, people hear about speech documentation, there is a lot out there on record keeping, but in this way, if I am the clinician, the ability to say, I need the EKG from two years ago, or I need the CBC, whatever, blood counts from a year ago, or show me the trends X, Y, Z, instead of manually having to configure that, there's capability to be able to bring that up audibly.
Sacopulos: Well, that is exciting, because as I am sure you know, we hear all the time about how medical records are so much more difficult now that they are electronic, and how that slows down providers. So maybe we are getting to the point where there is actually an increase in usable information.
Granzotti: Yes, yes.
Sacopulos: That is exciting. How far away are we from this being implemented in significant ways?
Granzotti: That is a great question because there are so many layers to it, and I am not going to pretend to be the technical architect here, but at its foundation, there must be data integrity. There has got to be data integrity, and security alongside with that, and we are building that as we speak, the security aspect, in order to build the correct data elements in that space, and how we can utilize that space in current platforms, in current EMRs, EHRs, their electronic medical records.
Again, this is not a new platform, this is not a new hardware, because health systems and clinicians have spent so much time and effort and really good products, foundational products, but we really need to start to consider further, what is that interaction with the clinician, what is the interaction with the patient, and what is the delivery of the care that we provide? With clinical augmented AI, with CAI, we can increase, there is no question, we can increase the quality and the safety and any type of regulatory elements, or let's say, pay for performance elements that have to be captured in patient care, that some of the administrative tasks, it has the ability to help clean that up and just reduce the timeframes of that.
Sacopulos: Does it have the ability to help educate new providers, younger providers?
Granzotti: Yes, it is interesting, and I have thought of that, how this could help with the education of new providers, or after residency, you are out there, you are taking care of patients one-on-one and in teams. There is definitely that potential, depending on the inputs, it all depends on the inputs and who is utilizing the system, obviously. But the team interaction, and you are hinting to that, the team interaction is very exciting, because if I can know more easily, let's say it's a hospitalized patient in the ICU, what my care team is doing for my patient and how that is impacting their stay and their day-to-day progress, that is powerful.
Sacopulos: Absolutely. Any idea on increased efficiency on a percentage basis? Is it too early to know? I mean, is the clinician 10% more efficient?
Granzotti: Yes, it is a great question, it is just too early to say. I do not want to even throw out a number, but if it was myself, thinking back to my over 20-year history in the ER and clinical workflow, oh my gosh, I would be shocked, again, for myself, just speaking for myself, if this did not help me decrease my administrative burdens and deliver better care by at least 30%. I mean, I will just say just a low number, but that is, again, a guess for me in my workflow.
Sacopulos: Do you think it will help with burnout?
Granzotti: Great question, yes. When physicians talk about it, it is multiple factors, it is not one thing. It is not only where you are working, but it is how you are working, the team support you have, the equipment, you have, the regulatory pieces, etc. There is no question this can help reduce burnout and delivery, because you are getting back to the human essence of care delivery to that patient-clinician interaction. That is the whole point of CAI.
Sacopulos: What are the training requirements to use this as a clinician? Is this something that would take me a couple of hours to begin to get into, or is this something that I need to take a course and spend many days?
Granzotti: Yeah, I think it is somewhere in between. Again, we are still in the build, we are still in the, how does this impact workflows, how does this look on particular workflows? And the beauty of it, also, is that it does not have to be an all or nothing, it can be a, I only want this piece when I am addressing this interaction or this workflow. The choice is there for the clinician, and everybody is different in how they interact with technical aspects. So, it's hard to say, in my eye, it's not going to take a course, it's going to be, how does the system continue to learn me, how does the system grow with me, and how am I applying the information that it's giving me?
Sacopulos: Are there any downside potentials that you worry about?
Granzotti: That is where you must get back to the data integrity and the security with it. Obviously, you are going to need to be able to have certain user access rights to the flow. And people are used to this in healthcare, they think about a physician access versus a nurse access versus, let's say, a CNA access. There are just certain pieces of information you must be careful with. Also, if there are security or privacy aspects too, let's say it is a behavioral health history, or let's say it is a sexually transmitted disease history, some aspects like that, again, from the security aspects, those access rights or not can also be assured. But those are some of the obvious things that you must be careful with, and that is more around the governance of the system, also. Right now, there is a lot of confusion and a lot of grayness on AI governance, and we are thinking of that in the approach, we are absolutely taking that into consideration with this approach.
Sacopulos: When you talk about grayness, that strikes me from a legal background, that so often, technology outpaces the law. Do you know, are there organizations that are coming up with guidelines or best practices to try to direct this, so we do not have problems?
Granzotti: Yeah, there are several out there that are coming out with standards and approaches to this, but this is just my personal opinion, I do think that this got out of the box a little early, the anticipation of this was a little earlier than what was readily available for structure and standards. But there are several out there that have been dipping the toe, and also, some more stringent information that people need to account for, and I do believe a lot of that is based on, again, the security. You must account for high trust standards, accreditations, there is a lot there, but it is good to see that people are really, really taking it seriously, because the application is quite remarkable.
Sacopulos: Absolutely, so if there are members of our audience now, which I am sure there are, that would like to be following this and know more about this topic, are there sources that you would recommend to keep them up to speed?
Granzotti: Yes, absolutely, our website is digitalstrategiesgroup.com, and that is plural, digitalstrategiesgroup.com. You can reach out to us in our group. In addition, I am on LinkedIn, you can find me, Maria Granzotti, on LinkedIn. You connect with me, I would be happy to connect with you, any questions you may have or further information you may need. It is exciting to see, though, that people are taking this seriously, that it is not, I did want to add, it is not a replacement, it is not a replacement for individuals. The whole point of clinical augmented intelligence is as a support to promote the human essence in healthcare.
Sacopulos: Brilliant, Dr. Granzotti, thank you so much for your time and your willingness to be on SoundPractice.
Granzotti: Oh, thank you so much, it has been great, it has been fun.
Sacopulos: My thanks to Maria Granzotti for her time and instruction on CAI. Dr. Granzotti is helping open a new world of possibilities for physicians and patients. My thanks also to the American Association for Physician Leadership for making this podcast possible. Please join me next time on SoundPractice, we release a new episode every other Wednesday.
Topics
Environmental Influences
Technology Integration
Quality Improvement
Related
Seven Practice AssessmentsHandling Litigation — How to Live (Well) with a LawsuitEmployers Haven’t a Clue How Their Drug Benefits Are ManagedRecommended Reading
Quality and Risk
Seven Practice Assessments
Quality and Risk
Handling Litigation — How to Live (Well) with a Lawsuit
Strategy and Innovation
Employers Haven’t a Clue How Their Drug Benefits Are Managed
Strategy and Innovation
Minimizing the Potential Impact of a Crisis: Preparedness
Operations and Policy
Surviving (and Finding Ways to Thrive) With Difficult Leader Phenotypes
Operations and Policy
Shifting from Star Performer to Star Manager