Artificial Intelligence (AI) in Health Professions Education

Morning everybody so lovely to see you all and your names I wish that we could all be sharing um um or not sharing a cup of coffee but we send you all coffee and we could be having coffee together or tea whatever your preference is um welcome welcome welcome on this beautiful morning at this point let's hope it stays this way um you all are in for a treat this morning thank you all for getting up early joining us um this is the second this is take two of this this presentation um Dr Russell presented for us last night um from 5: to 6 and we had a great crowd over at the um med school but this morning uh I don't think you will regret being here early we've got um Dr Regina Russell here presenting to us on artificial intelligence and medical education now Dr Regina Russell is director for assessment and evaluation at the Vanderbilt University School of Nursing an associate professor in the health promotion populations and Health Systems Community she's a social scientist who focuses on the description and measurement of performance to support learning and educational Improvement Dr Russell holds graduate degrees in higher education sociology and public administration and her professional experience includes Social Work nursing and medical education Dr Russell is active in the School of Nursing Center for research development and scholarship in the Vanderbilt University data Science Institute her recent collaborative research includes studies of Competency Based assessment artificial intelligence competencies for for clinicians experiences of first generation college students in medicine and the application of large language model to Grant evaluation activities that's a lot that's a mouthful you'll see how all of her experience comes together in a one hour presentation um I'm going to turn it over right to you Dr Russell and I will see you all in a few excellent so let me share my screen um good morning everyone and thanks for um getting up early for this I this is maybe the early bird crew can you hear me well can I get a thumbs up is that okay excellent so let me get my oh you don't need my email let me get this screen share all right how's that so um sure I can see here excellent so um thanks for that introduction and thanks for the invitation to come speak with you all here um and it's been my first um visit here so I'm learning a lot about your institution and also your wonderful city um thanks to the continuing professional development team that has been organizing my trip and helping me uh get ready for all of these um discussions I would like to acknowledge that I expect there are a number of people in this um call that have experience with artificial intelligence or um and definitely Health Professions education so I would love to hear your feedback I'll save time at the end for questions and I hope that you will um write some notes maybe um about what you're hearing and I will say that a lot of what I'm saying um won't be necessarily new some of it is research that I'm sharing that's new but um some of it is really um connecting some dots I think between what we've been doing in Health Professions education and what's coming uh and what's here now and what's coming with artificial intelligence so um if this uh I know that we're recording so if you're watching this in the future it's March 2024 early March now and the talk is artificial intelligence and Health Professions education and um this is the teach Grand rounds at at Virginia Tech Cillian school of medicine and I would just like to say hello to my mom and dad and Arkansas in case you are watching this um oops get the going forward so this is a just sharing some disclosures about funding that I received for some of the research that I'm going to share later in the talk um there was a collaboration between Vanderbilt University Medical Center and IBM Watson Health where they were interested in um looking at artificial intelligence and health care and so that was a large multi-year collaborative and there were small grants off of that and I was part of a small research Grant um the pi is Dr Bonnie Miller and I received 10% effort to um work on this uh research project and I would say I was invited because of my background in medical education and Competency Based education so um I am not a computer scientist I've learned a lot about artificial intelligence during this um project and I hope to share some of that with you uh this study was conducted that I'll talk about later in 2020 and 2021 and this is an interesting article from The New York Times that came out in 2021 I Watson Health which was the company we were collaborating with closed at that time and um IBM Watson is still around but not the um incarnation of IBM Watson health so you might be interested to read this article about the backstory of sort of tech development um in healthcare and and some of this early um efforts that they made uh also I want to point out that the graphics that I'm using in this talk I generated with Adobe Firefly which is a a generative AI U creates images based on prompts so I would say you know I want some futuristic um uh Pharmacy bottles and some old textbooks and uh the background to look a certain way and you can um Adobe since it's a design tool allows you to choose certain Design Elements so I just wanted to let you know that this is a contribution of generative AI to the talk um and I hope that it helps us think think through some of the implications on many levels so the learning objectives for today are um what hope that you get out of the talk and that we do together is that you would be able to describe the expanding influence of artificial intelligence-based Technologies in the broader Society in Health Care Systems and particularly in Health Professions education because this is a teach Grand rounds and we're going to review the proposed competencies that our team developed through this research project and so that you would leave with um some awareness of what that competency set looks like and how um that might um be incorporated ated into Health Professions education so the structure of the talk today is I want to talk a little bit about um foundations and Futures I want to focus our attention on some things that we already do well and we know in Health Professions education are foundational to our work before we start talking about potential futures um sort of what's coming down the road for us um I will talk a little bit about the collaborative research that we did and um specifically the competencies for the use of AI based tools by health professionals which was the outcome um of that research but throughout the talk um and particularly as we get to the end I'd like to have more discussion about the role of Educators and education systems in this change so a lot of been has been driven by computing computational power but now it's in all of our domain and um so uh part of what I hope to to share and this is I was asked by the team to come up with a tagline for the talk um if you could walk away with one thing what would it be and um this is it I would say um I hope everybody walks away with an awareness that education is necessary for the responsible use of aib based tools in the Health Professions and that's us um and so the Educators really need to be in this discussion and thinking about how um we're going to evolve this future together um this is just a little bit of grounding uh this is Dean uh Dr Bonnie uh she was the senior Associated de Dean for Vanderbilt University School of Medicine when I worked there for about 15 years and she really helped me understand the landscape of Health Professions education and where Competency Based education fits in there and how we can uh think about Futures in Health Professions education where people can come in and out and um move into different learning based on what they need based on the competency develop that they're development that they're working on Within in a framework of required competencies right so this is the institution that I'm coming from I'm at Vanderbilt University and in the foreground um this is the old campus this is old you know University buildings all these old brick um kind of uh Academia I guess you might call um and in the back this is our Medical Center so uh Vanderbilt University Medical Center all these shiny buildings and the cranes that are coming up very fast when I worked at the school of medicine I worked over in this environment and it moves very fast and that was something that I learned I'm I come from higher ed my whole career um but working with health care providers seeing um just the urgency of that space they are moving much more quickly into using artificial intelligence-based tools and were already there um when we were doing these interviews this is before the generative AI boom um but I would say that is really was sensitizing to me to be aware of how fast um paced the environment is uh you know really people's lives on the line and people running fast trying to figure out how to use these tools to the best effect now I would say Academia is a little bit more careful and a little bit more you know thinking about what are the implications of this going forward and how do we want to use it and my sense is we've really got to bridge those two conversations and people in Health Professions education are doing that all the time so um I'm going to next move into some conversation about what I think um we already have built in Health Professions education that will be helpful for us moving forward into these different alternative Futures and this is a shout out to Dean Miller I would say also she was one of the first um female surgery residents at Vanderbilt when she went there so is really a Trailblazer and has been thinking for a long time about how we can do education more effectively going into this future where all the information is available to us but it's about um critical assessment understanding what is good information what has qu quality evidence behind it so um the first foundation and I've already given a clue about this I think is Competency Based education and we've been moving down that path ever since I've been in medical education it's been a priority and it's been one of the areas that I've focused my energies on um is thinking about and of course I I am uh reminded that the residency programs in gme had started in Competency Based education and Medicine um way earlier than I showed up and then um we went about trying to build Competency Based assessment for our undergraduate our MD students so that as they moved along their pathway they would align with what was expected of them in residency they would understand the language of the competencies and they would be able to drive much of their own learning because they knew what they needed to work on so um those of you who aren't as familiar with Competency Based education the idea is we're working from professional competencies knowledge skills and attitudes that are necessary for practice and there can be basic competencies that everybody needs and then there can be specialized competencies um that you might need if you're an anesthesiologist compared to a geriatric care provider um it is in these professional communities that competency requirements are determined so um what filters into Health Professions education is part of what's expected for people that will be working in those fields can consider Healthcare ethics one of those areas that everybody has to learn about before they can be licensed as a healthare provider because the implications are so profound of the possibilities of um harm for individuals um so why do Competency Based education one is to take it out of the time-based idea everybody may not need four years to get an MD degree some people may need longer some people may need may need shorter it's not the time that somebody spends it's whether they can do the things that are expected for them to do in practice are they safe can you trust them um Competency Based education as I noted is very learning centered which allows you to also the learner to Center on what it is they need to learn and good assessments as the priority so good feedback for people about how they're developing um and including in clinical uh really good clinical feedback um we see some of this work and assessment coming through in Milestones I've done some of that work and in trustable professional activities how do we know when people are ready and how they're progressing um supervision luring continuing education continuing professional development are all part of a competency based um framework for uh lifelong Learning and Development so that's one Foundation another one I think is interprofessional teamwork um interprofessionalism and Healthcare is just required it's part of how the I don't have to tell you all this but it's how it works I would say also from a patient perspective we see all the different people that work on care and um think of them as a team and the development of tools and Technologies using artificial intelligence um also needs to be co-development co-developed between teams of people that um you know really know how to do the code know how to build the tools but also the people who are going to be applying this in practice what's useful what's appropriate what's not appropriate um what's ethical not ethical and I would say that all of the professions in healthcare are facing similar challenges of Rapid change this is all happening to everyone at the same time so I've recently um moved to the School of Nursing and we are facing the same technological environment um Artificial Intelligence coming in through um many different Avenues in healthcare and education and then the third Foundation I would like us to focus on is the idea of evidence-based practice and this applies across the health Health Professions also in education the expectation that there is some evidence for the ways that we do things we don't just sort of make it up as we go along but we work in a community of Science and communities of practice that are working on best practices and sharing that um so whenever we start thinking about new tools or technologies that we might use I always think it's a good question to ask what's the evidentiary basis to evaluate these tools how do we know if they're high quality if they're doing what they're supposed to do um but also how can the new tools contribute to our evidence-based practice um what new ways might we be able to learn with these new um capacities and then relevant to the question that we were asking what do Frontline clinical users need to know to evaluate what's happening with these tools and if they're being used effectively appropriately and ethically so moving from the um idea of what we've already built what we know to the Future what's coming with artificial intelligence I would emphasize that um EXC oh I'm sorry let go backwards um I would emphasize that um artificial intelligence is not a new thing uh it's been around longer than I've been alive their computer scientists have been working on this challenge of how to get um computers to um think more like the way humans think and computers don't think they calculate um but to be more functional in providing decision support that's um the language that many of you will be familiar with from um working in health systems but how can we uh get the computers to do more of the work that would help humans and do it in ways um that are more intelligent that they can take on more of um the work and we can guide that work so artificial intelligence really at its base is um a bunch of computers Network together calculating off of a lot of data that um gives Clues and insights about um what is the expected next thing to do so it isn't um sometimes people will say you know like it's it's intimidating and I don't understand it and there's a lot of words used like machine learning and neural networks deep learning a lot of tools and approaches that are used in artificial intelligence that can seem overwhelming and um you know if you're not a computer scientist feel like that's not really your place but at this point it's in all of our um workflows in our in our lives so understanding sort of a little bit about what's underneath is helpful um much of what uh we've seen in healthcare are things like Risk scores um algorithms that provide prediction predictive capabilities probabilities of patients having certain um outcomes or disease States um and then providing some support for clinicians about this might be some um option that you would take there have been a ton of applications for artificial intelligence in engineering um you know industrial uh you know factories having robots building cars all of these things have been developing for a long time but they've been um I would say more specific use cases and not as publicly available so now we're seeing them in our cell phones where the um you know you'll look for a pair of shoes and then your phone will Amazon will say hey I you were looking for some shoes here's some other shoes that you might like that look like those shoes so there are so many applications of it and it's coming in so many ways in our lives many ways that are incredibly helpful and um reducing work but in other ways that raise flags and concerns about do we understand the implications of that so I would point out that there's ongoing research and development in computer science communities um informatics data science just a huge amount of development around how we use all of this data that's coming in at this point to train uh machine models and a distinction that's helpful I think is between when we talk about artificial intelligence it can mean a lot of things and I always like to say artificial intelligences because it's not one thing it's a lot of different tools that's why we use the the term tools um with artificial intelligence Technologies embedded but a lot in healthcare um we're talking about specific use tools so tools for monitoring patients um diabetes and a you know a very specific use case um and a lot of what maybe is more in the public is the idea of kind of a general intelligence the Sci-Fi kind of idea we're not there um I I I think there's a lot of people who would like to get there um but we have a lot of ground to cover um in terms of what that might look like what are the policies and the guard rails and so I'm really talking a lot about um specific use tools here but generative AI is kind of a general tool um that it's not a general intelligence um like you know F or something but it does start people thinking down that path and talking down that path and so it's helpful for us to think about what's now and what's coming and how how we need to fill the Gap in between or shape the future of what we want to see coming so on Q anytime I start talking about artificial intelligence I want to raise the flag about ethical challenges related to this and um they're large there's a lot of concerns around the you the increasing use of artificial intelligence in society um in healthare and in education so a few of those things um around safety particularly in healthcare do know that the um risk predictions are of high quality depends on the model that you have depends on the underlying data that was used to construct that model for predictions um that those data sets can misrepresent populations so you can train your model to make predictions on one population set and when you apply it in another group it doesn't work for them and that can be a huge problem because um a tool is is sold saying this makes good predictions for breast cancer um but if it's in a primarily white population and then it's put in another population it may or may not work the same um I've labeled that as training data misrepresentation um but representation or misrepresentation the sourcing of data where does all this data come from is it ethical does everybody agree that they can be tracked all the time um with their cell phone you know everywhere that they go and all the um questions that they ask in Google um we do agree because we want Google to tell us the answer so there is a bit of back and forth of about those things um definitely bias and there's a lot of conversation if you haven't you know been hearing about bias in AI um welcome to the party uh but there's a lot of concerns about how you know not just the underlying um data sets that's a huge source of bias but everything that's built by humans has embedded bias in it about how we plan to use it how it works in systems and so thinking and talking about those things at the front end is really important privacy uh also like I said you know what you can be tracked on your cell phone there's now um sort of uh uh teddy bears and different types of uh dolls that patients can families of patients can put in hospital rooms to track healthc care providers what they're you know how often they're coming what kind of things they're saying so there's not just privacy for patients but also for um whole systems in terms of tracking and monitoring performance intellectual property I always tell people I'm from Nashville and you can't just scoop up all the songs and not worry about how how the um creative people are going to be paid so you know who owns the copyrights and the intellectual property that's an ongoing uh issue of Ethics in um use of artificial intelligence but also there's just large human impacts as these as our systems change radically economic changes and environmental costs I was talking with someone the other day about artificial intelligence I said you know um artificial intelligence is very resource intensive it takes a lot of energy to run these data stores to run um these analytical models the large language models are costly and that means costly on the planet and environmental resources so we I think have to talk about this in a balanced way um and how we're going to choose to use these powerful resources there's a great book um by Kate Crawford that came out in 2021 on yel press she's a professor at Stanford and um it's called the atlas of AI power politics and the planetary cost of artificial intelligence I would recommend that to you if you're interested in sort of some of the deeper policy issues and some of the larger um economic and and um environmental consequences of of these growths so the work that I did was in 20 uh the research that we did was in 2020 and 2021 and so that was before chat GPT exploded on the world um so I want to you know call out this new um set of tools that we call generative AR artificial intelligence artificial intelligence that's able to create which is different than many of the tools that have been in the past that are um you know even before artificial intelligence Much More rule-based Much More constrained so now we're talking about tools that can create can develop new sets of workflows uh text generate um whole papers on their own you've probably seen these things so I wanted to point out where that's coming from uh there have been huge competitional advancements in the last decade uh around natural language processing and so that's the ability of computers to read text just like your Siri on your phone or whatever you know if you use these tools talk to the computer and the computer understands our language um and can uh process based on just and I don't have to learn code I just can speak in my own language now that is a huge advance and that's what's enabled um this you know chat GPT World also processing power of chips is greatly advanced and the storage Capac capacity having these large data centers um there's some that you can see from outer space so um that has been great um technological advancements that have created this infrastructure that's different now than we've had in the past and so I like to think about that as we're in 21st century digital and social environments and that's different than what people have had in the past and we need to think about that pretty hard and I think in the professions come up with um competencies guardrails guidelines expectations and ethics around these new tools uh the type of generative capabilities are incredible um many of us have used the chat GPT for text um these image generation I asked um Firefly for this picture show me some of the ways that generative AI is being used and this is what it came up with these are you know different in the back this is what the robots are are are generating but um generative AI can also generate audio video audio video together um 3D we have now have 3D printing so you can you know tell a generative Model come up with something you know a um a picture or a a 3D image of something and then print that in plastic uh so we are now in multimodal world with generative artificial intelligence and this is now accessible to lay users um I have a nephew with a 3D printer in his garage right he can make all kind of things so um there are extensive ramifications for this individual at the individual level what should I use what shouldn't I use um processes in institutions policies and thinking about the whole organizational landscape the way we've always done things the way we think things ought to be done including in our institutions Academia healthc care but also the professions what are the guidelines in the professions and of course policy related to governments local state and federal um I have two slides here talking about Futures one related to healthc care and the next one I'm going to talk about Health Professions education I've kind of align them the same way so that um we can uh talk through some of the current uses and where it's likely to continue growing so it's say administrative teams are already using artificial intelligence a lot built into electronic health record finance and billing systems workforce management clinicians are already monitoring care with certain um types of artificial intelligence tools giving you know heads up and alerts when things are going in directions that need um attention a lot of uh potential for our Diagnostics and use of artificial intelligence for Di diagnostic purposes just looking up information keeping up to date with um information and communicating uh you know there's a lot of um communication that come goes between clinicians and patients and so how can that be streamlined how can artificial intelligence help with some of that Burden Burden of documentation um and then patients also are using artificial intelligence-based tools many people have uh Health tracking uh Technologies watches keeping track of their blood pressure um their heart uh heart function Etc and they also use artificial intelligence to seek information about diagnosis and management of their own conditions um some people find this uh frightening and some people find this as a a way to continue growing um individual patients own knowledge and awareness of their health and how to collaborate uh with their care providers to um continue to be uh healthier so uh a number of the tools um down here at the bottom of natural language processing is huge uh being able to use natural language processing on electronic health records and notes many of you I'm sure have been involved in that research or seen that research the integration of large language models that's happening everywhere and will continue to computer vision and I I always think it's um helpful to share that computer vision has two different meanings um in the computer science World computer vision means that the computers are able to use video to see what's going on to create to take that input in order to um you know provide feedback or decisions or information so the computers can see computer vision but the um the eye doctors the optometrists call computer vision that condition when your eyes are very blurry and tired um because you've stared at a computer screen for too long so I think there's a lot um of even development of the concepts and the health Concepts around um all the computering that is involved um with these new technologies robotic surgery requires computer vision um for the computers to be able the robotic arms to be able to get that input about where you are in space and what needs to happen next um carebots is a whole area of development I have a friend who's working in that area in the future uh I hope that when I'm if I'm older and I need um care in my home that I might be able to have a care bot that connects with the health care System um in ethical and appropriate ways um to make sure that I'm on track and help me continue uh managing my health predictive modeling as we discussed um and talk about that a little bit more um risk scoring uh data that comes from electronic health records and even at Vanderbilt we have a huge genetic database that's connected to the electronic health records so the type of information that we can find out about individuals and population groups using these um artificial intelligence-based machine learning and predictive modeling tools it's very powerful and does require a lot of thought and guard rails policies around how that information can be used and again Diagnostics and imaging Imaging is an area that has really um just boomed with artificial intelligence there's a radi ology and AI Journal already um that's just talking about how artificial intelligence-based tools can be used in Reading um images and um and providing information about what's in them often as a second reader or a third reader to a human but they are very good at being able to pick up anomalies that's actually one of um the great strengths of these types of tools is pattern pattern finding and an picking out anomalous um information so now now that we're thinking about you know this is already in healthcare it's already happening and um it's going to continue to grow that's what I'll talk about in just a few minutes um the the interviews that we conducted but that means it's also in Health Professions education because we're Preparing People for their careers um to work in these health systems that have a lot of embedded artificial intelligence so even before chat GPT exploded in healthcare and um the world and education um Health Professions education really um needs to think and focus on how these tools um are going to be used by the students who will become uh providers care caregivers in the future so already same administrative teams things like enrollment management evaluation systems Improvement um people are already using artificial intelligence Bas tools or testing them for these purposes educators are already thinking about content development course support assessment um related to these tools or how these tools might be helpful in those processes um and Learners same as patients are trying to um monitor their own progress and information seeking about what they need to know next and these tools are incredibly helpful for them sometimes same natural language processing uh large language model integration into education um we might think about ways that education um has benefited from and will continue to benefit from automation uh like carebots we might have Coach Bots where every student has has a guide um in their machine that's helping them know what's the next what's the next experience what's the next thing that you need to learn to to build your competency in areas that you need to develop predictive modeling based on student data mining is the same concept developing algorithms about uh which students might struggle in different um contexts or over time and how we might help them or how students might be able to move F faster through programs if we give them um more guidance more information uh assessment of students in simulation using uh computer vision V virtual re reality those are all um happening now people are testing those now so that was a long introduction to the research that I'm going to share with you but I wanted to really ground Us in what we already know and what these tools are before I shared with you the study that we did so this um was a research collaborative between IBM Watson Health Vanderbilt University Medical Center and I was the representative from Vanderbilt University University and our question was what are the competencies necessary for the ethical and effective use of artificial intelligence-based Tools in clinical settings so there were already some competency statements about what do informaticists need you know what competencies do they need about artificial intelligence that were a much higher level um that involved you know knowing how to uh understand the underlying code but this is really about every single person who works in a health system and provides care for other humans what do they need to know about how these tools um might impact their decision-making impact um the flow of care and um helping to make good decisions about how to use these in healthare so this is a picture of um just the timeline of our study and uh in Spring 2020 we conducted a scoping review and I'll share that um slide with you in a minute and in the fall of 2020 after looking at that um all of those articles we um developed an interview script about you know asking inter asking experts um across the Health Professions but particularly people who are have been doing a lot of work publishing research and development and artificial intelligence to ask us what do you see now what do you see coming 10 years from now five years from now um and what do you think that um clinicians are going to need to be prepared for what competencies do they need um after we interviewed those folks and I'll share that with you more in a minute we drafted a set of competencies and then we sent those back to our experts and said did we get these right give us feedback and we adjusted those competencies and then we published those um they came out in the spring of 20123 this is and the citations are here if you're interested in following these um links I think this presentation will be available after um but this was the first um work that we did was a scoping review and we were looking really at um what type of competencies have been written about what type of expectations for education and learning for humans are necessary um when we're looking at machine learning models and Healthcare and um you can read this paper that um has our entire search strategy but I'm just showing the one picture um that we started based on our search strategy with 3,400 records but after we went through the process of reviewing and selecting just the articles that focused on competencies or some type of statement of human learning uh we came up with only four and so that's very concerning for there to be 33400 articles about machine learning and the use of machine learning in healthcare and most of them are really just about how the tool was developed or what the tool can do um and sometimes it's about how the tool might um be used in conjunction with with a human but without discussion about what the human needs to know to manage the tool to understand if the tool is working correctly um to be part of that process so we felt after we did this we were on the right track and we wanted to um go forward with the interviews and ask um well let me let me say before we did the interviews this was our second paper which was really just a statement um a commentary that what we saw in the literature was a lot of AI related Health Innovations risk scoring image analysis clinical decision support lot of patient generated data but that this connect between clinician competencies and all of these Innovations there was a big gap there and so we wanted to focus on that Gap how can we build in learning activities and assessments for for clinicians both General competencies for artificial intelligence and specific for a a type of tool like robotic surgery you need very specific competencies as the human surgeon to run robotic surgery so this is just kind of a graphic representation of what we see as a big gap in education and in our education conversations about preparing future clinicians these are the experts that we interviewed we interviewed 15 and we call them subject matter subject matter experts 15 smmes and the professions that they came from and their areas of expertise every person on this list was working at the front end of um Health informatics bioinformatics nursing informatics Pharmacy informatics um and have been thinking about this for a long time and so all of these people are very positive about what the potential is for artificial intelligence but they also gave us a lot of things to think about in terms of how the larger Community should move forward together this is the paper I think that was shared in advance of this talk competencies for the use of artificial intelligence-based tools by Healthcare professionals was published last year in academic medicine and it provides this is our table of um or our picture of the high level competency domains that we expect every healthc care professional will need so I'm going to talk about these in a little bit more detail but to show that under each of these domains are specific competency statements that our um subject matter experts and our our group's combined process of um sorting through what they said and developing competencies we came up with these set of statements and I always again highlight the social and ethical implications because I think um Healthcare and education both um sit in that space a lot um thinking about the social implications and the ethical implications and so while workflow analysis and um many of these other elements are critically important uh we we also need to spend a lot of time here in going back and forth because as these tools continue to evolve the implications social and ethical will continue to change and evolve as as well so I'm not going to read all this to you thank goodness I'm just going to um highlight the highlevel competency domain and then you can see the statements below these are in the paper and this presentation will also be available to you if you want to dig deeper so the first is everybody's got to know have some basic knowledge of what artificial intelligence is so that competency would look like be able to explain what artificial intelligence is and describe its Healthcare applications that doesn't mean look under the hood you know it's more like driving a car than fixing a car or building a car but um these are the specific statements there um for social and ethical implications the the large framing domain is to be able to explain how social economic and political systems influence AI based tools they don't they're not in a vacuum and they're not bias free and they're not um free of sort of Economic and political interests um and so understanding or having some awareness about how these relationships between technology and human systems impact Justice equity and ethics is really important for everyone as we continue to um adopt these tools workflow analysis this is already happening uh every clinician um is probably aware that the electronic health record changed their workflow in many uh positive ways and some frustrating ways um but the electronic health record is something that um is a massive Boon for artificial intelligence because all of the data that's available there so um understanding the implications for workflows and what we do in our jobs education included um so be able to analyze and adapt to the changes in teams roles responsibilities and workflows that are resulting from the implementation of these tools it's happening now and the expectation of our experts is that it will continue to grow in a very um Rapid Way AI enhanced clinical encounters um Carry Out AI enhanced clinical encounters that integrate diverse sources of information to create patient centered care plans well as we think that most of um you know patient data currently lives in um electronic health record systems um what about all this other data that patients will be collecting and carrying around who owns the data how do we combine those data sources to to help make good decisions um in collaborative patient centered Healthcare um I like this one evidence-based evaluation because I like to think about evaluation that's part of my job but every person who has some uh responsibility for use of these tools has to understand how to evaluate the quality accuracy safety appropriateness and biases of these tools and their underlying data sets when they're providing care to patients and populations this is hard because many of the tools are not um easy to understand what's going on um how the decisions are being made what information is being used some tools are proprietary and they don't want to share that information because that's sort of their business model so um every person that is taking a prediction and applying that to a patient case needs to understand where that comes from and so whether that is the individual person that's reading a what is called a Model facts sheet kind of like a dietary label I apologize for the balloons that happens on occasion um everyone will need to have some understanding of that and even if you're not doing it yourself do you have a group in your unit that you can go to and say hey how does this thing work it doesn't seem to be giving me answers that make sense in my context so um putting this evidence-based evaluation in the hands of every user is really important but also a very heavy lift for understanding these tools and so that's where this final domain comes in practice based learning Improvement participate in continuing professional development and practice-based Improvement activities related to the use of AI Tools in healthcare and I would say education are are you um continuing to learn I would say this is a great example of that for education and the Health Professions um is continuing to be in this conversation learning um what are the things that are happening and what is my responsibility in that um development and I would also emphasize that it's not going to be the same for everyone there will be these basic skills that everybody needs to have but if you're working in anesthesiology the tools that are available to you and that are will be expected to be used in your specialty are different like radiology and I would also say that some of these tools and I work in a very high-tech environment but some of these tools just are not going to be available in uh low resource setting areas if you don't have electricity you don't have artificial intelligence so we also need to be considering the implications across Ross the field um across professional Fields And across many contexts if we teach everybody how to do Healthcare with using their AI bot to help them what does that mean if you don't have an AI bot in that moment um one last slide um on Health Care ethics specifically and some quotes from the people that we interviewed one that healthc care is about humans first and continuing to prioritize humans in this technological artificial intelligence um future is really important and one of um there's some uh papers about trustworthy AI but the number one is that humans are driving humans are deciding um one of I will say this um warning proceed with caution is just a summary of what we heard from these experts they want to perceive these tools are very powerful they have great potential to do good in the world but they also have potential to do harm and um the professional communities are going to have to get their arms around what that means in each context and proceed with caution this is a quote um that I always like to share we have seen what AI does when it ignores equity and inclusion in criminal justice in education in housing and you name it we should not have done it in those fields and we certainly cannot do it in healthcare so this person just really prioritizing that the healthcare environment with vulnerable humans who are trusting um sick people um are trusting that we are using these tools for the best intention for for the care of them individually and for populations that were responsible to help this is the final paper that we published out of this work it's called clinical use of artificial intelligence requires AI capable organizations and I'll give credit to Dr Lori Novak here who's an anthropologist working in in biomedical informatics which is a fascinating combination of sort of you know peeking peeking over peeking over at what um the computer scientists are doing from an anthrop anthropological lens um and this was published in Jamia this is the Journal of American medical informatics Association open in last year in 2023 and what we're emphasizing here is we did all this work in this space that I'm circling here about competencies individuals needing to understand what is artificial intelligence and how is it imp impacting my job and my profession that's super important we need that but competent but individual competencies individual people work in systems and organizations need to be capable they need to be AI tuned um there need to be infrastructures that are basically outside of organizations laws uh pervasive enabling resources that allow organizational capabilities to be up to speed on the AI based tools that are Incorporated so that every individual can use use their own knowledge and competencies and ethical um understandings to be effective in those organizations if we don't have the infrastructures if we don't have the organizational capabilities you can individually be competent but our systems will not be competent and we will not put out successful um outputs that we're expecting from Healthcare and education so I would I would emphasize while this is all Healthcare focused we need the same thing in education competencies are system dependent so an individual educator is working inside of a health education system that also needs to be a AI competent so this slide I just want to draw your attention to all of the policy and Regulatory implications uh we talked some about these safety privacy high level intellectual property cyber security um all of this data needs to be protected um you know the more we pile data into data sets the more more vulnerable that information is to theft and misuse um as a discussed infrastructure and costs um are really important to consider but what we're talking about today is professional communities and I think um education professional communities health education professional communities really need to be um thinking in this space about guiding with ethics which contexts are appropriate and not appropriate what are the guidelines for um responsible use how will we monitor over time and I've got this super highlighted and bolded but education how will we continue to teach about about this so that the Next Generation and the next Generations continue to be better prepared to use these tools to the best effect and make sure that they aren't being used um generating unintentional harms sometimes people have the best intentions but systems work differently than we um might expect and the outcome in a system is different than was anticipated so back to the start here um I hope that um you've gotten from this discussion um the ability to describe some of these changes that are coming in Health Care Systems and Health Professions education we talked about the competency set that my team developed and then again the big tagline here education is necessary for the responsible use of AI based tools in the Health Professions this is not just for the computer scientists anymore I always tell people now the computers speak my language and so I'm um going to speak about this as much as I can and I hope that you all um will as well be interested in continuing this conversation so I will stop sharing now and be really happy to take any of your questions or comments um about what you you've seen or what you are interested to talk about more my email address is here I'm always happy to take questions I tell people my email box is always open so if you don't haven't thought of your question right now I'm happy to take it at another time so I I did chat yes okay nope go ahead s here I was just I I hadn't had a chance to to look at what was going on in the chat so please if there are looks like people sharing papers links there so awesome yes our team is awesome when you mention a link go resource incredible you have an incredible team here keeping you up there is no doubt about that so what type of questions or comments do you all have this is so much information really just um scratching the surface getting our minds flowing in lots of different ways um [Music] um I have a question okay I do have one um what do you see as the role for um these new AI models for documentation where they kind of listen to the patient as they talk and then put in their their information directly into a note so it's not even done by the provider at all a great question so I'm not a provider I'm not a clinician and my feeling is that these tools need to be um brought into systems with partnership with I mean clinicians need to lead these conversations so if it reduces documentation and it helps in that conversation between the patient and the health care provider then I think there might be a space for that um are do you have concerns about how that information is gathered and how that information might be used yes what are your concerns and how will those be communicated in your system that's my my um emphasis is there's a thousand tools and there's a thousand ways that people are thinking about using them so my question is how will your system um evaluate whether they're being useful or harmful who gets input in that um and I would say that patients are going to continue to ask for more input about that as well because they see the power of these tools too so some of this is a power change and that's uncomfortable too or it's unclear what the implications of that are going to be if patients have more power over their information and clinicians are used to writing notes in particular ways that have meaning in their shared communities does that disrupt that practice that workflow of how do I code something you know when I write it I'm not a clinician but if I were if I wrote a note um there's certain words that mean certain things that patients aren't going to write in there because they don't know so I would think of it as a um an adjunct um an additional way that information might be coming into the system rather than a replacement at least in the short term but I will say there are a number of these tools that people thought would be ad junk um to workflow processes that eventually are much better and um takes that out of the the um expectation that a human has to do it and in many um I would say outside of healthcare I hear a lot more concerns about my job's going to go away because AI can do all of these things um those of you who have been working with AI realize that it's not anywhere close to having for most things is not doing at that level but um in healthcare What I Hear most of the time is not my job's going to go away it's like oh my gosh there's all these jobs we can never get to that take so much energy or I have to keep repeating this same you know data entry or this same Etc and if we could build some tools to offload some of that burden we could do more of the things and I would say same in education more of the things that you hope to be able to do but that depends on these tools working in ways that we're not constantly cleaning up a mess right so um James to your your comment um I think that um this is a Brave New World as somebody said to me before and um the clinicians are really going to have to be in the conversation about whether that's appropriate or not are you a clinician James I am yes I work in the emergency department so I I like the idea but I'm also worried about um taking things out of context is my biggest fear for that environment so I love for people to say this seems interesting and I'm there's possibilities about that but this is my biggest concern I love for clinicians to be able to say this is what I'm seeing as like really potential great potential but here's the implications that people maybe work in computer science haven't thought about right as an emergency room physician the patient coming in with all of their information absolutely great Insight thank you so much yeah yeah and you talked about um Dr Russell clinicians being at the table and um information technologists being at the table and all these different groups but even when you think about clinicians you want to have the variable perspectives right because emergency medicine is going to use it very differently than a Pediatrics um provider um it looks like uh Elena has a question she has her hand raised over there yeah um I was wondering about education because uh some of the students are learning and studying using Chap jpt and that's freaks me out because it's helpful I I use it sometimes but when I know the content and then I can see mistakes so what about somebody learning something using CH jpt so your question is Right On Target it's um I think where the educators are thinking right now um at my own institution um we have set up um expectations about use and education that leaves it to the faculty member so just as Sher said um faculty are teaching in different disciplines and some disciplines grab this and they're like let's run let's see what we can do with this um and then other disciplines are like no we're teaching writing and if you use chat GPT to write your paper you're not actually learning how to do that so I think there's going to be a lot of variability about how it's used I'm as noted um earlier in my bio I'm part of the data Science Institute at Vanderbilt and I went to a um like a a day symos where all of the students that are part of data Science Institute were showing how they had trained their chat bots so they have each different you know like little instantiations and they're training them to do different things it was incredible what these students had come up with and there was a group of three women who had built um using chat GPT a whole Suite of tools to help each other learn so it was you know putting together notes from their classes and um generating quiz questions so it's at you know maybe a more like we're talking about Coach Bots kind of level you still need supervision and oversight of you know the faculty member and the educator but these May provide and people who have you know need to practice more you know we have simulations where people can do this simulation multiple times so I agree with you and same with James there are big concerns and we should be talking about those concerns but we also should see how are they using it and where in what ways can it sort of Advance our knowledge Advance our educational practice but in conversation with what's appropriate and what's ethical right not just oh my gosh this thing can you know create anything um you've probably seen you know like the image generators there's many image generators that are just pulling up copyrighted images well that you know that's not a functional model and we need to be training students that you can't just take people's intellectual property and call it your own um but there are potentials I think to really Advance learning with these tools I'm I'm trying to learn myself how to better use large language models in ways that are bounded and appropriate excellent there are a couple of questions right in the chat um one is uh from Dr Carter the literature review excluded 48 papers from outside the US is it possible that insightful information from other countries may have been missed in in the review absolutely um I would say we talked about that quite a bit we just had to bound our search um to English language um but the and we did only interview um American uh I think maybe it was a Canadian as well um so our popula you know it's a smaller population I would say most of those people are at the front of the conversation globally so they were talking about you know what they knew as well I think that was included but yes this is an international it's an international uh thing we've got going on here artificial intelligence is you know all around the world so very important Point um we talked a little bit about that last night um too so in the topic of Ethics the intent of using AI I believe is an essential question have you seen what are the key questions to establish the foundations of an AI project so um there are several you know competing ideas about this and I think it depends I agree completely what is the intent but intent is not the only thing and anybody who studies implementation science would point out that we can have great intent but the way it actually plays out in the world is not the same so there are like I said there's um some uh um work about what is trustworthy AI I would point you down that direction um seven principles of you know setting up AI in a trustworthy way and so that means does it work in systems the way that you expect it to work does it have the guardrails and the boundaries so I would look I would look there um thinking about trustworthy AI as giving you a set of of guardrails or key I wanted to say one last thing I forgot to say in the talk if I can I think this is a really helpful um one of the subject matter experts said to us um that every clinician knows something about pharmacology they're required to learn that um and he really wants that everybody should learn something about AI so that they know if somebody is trying to sell them some he called it magic be that you would know what kind of questions to ask and so I think this point about what are the foundational um principles are really important and they will vary depending on your context Education Health Care car manufacturing so I hope that is um somewhat helpful and again thanks for showing up this morning so early and having your questions and I'll Stick Around yes if anybody wants to stick around for a few minutes to ask more questions that's great I realize it is 8 o'clock and we are ready to get rolling with our day I hope you all have a wonderful day um again if anybody has any questions feel free to stay on for the next few minutes.