Demel Ji Mehta

00:00:05 Speaker: welcome to targeted Oncology's Treating together series. I have with me today, two esteemed oncologists and we'll be talking about the topic of information overload in oncology. I am Doctor Paul Mehta. I am the medical director of Reimagine Care. I'm also a practicing medical oncologist. next week will be twenty years as a breast cancer oncologist, and I'm happy to have Doctor Kurt Demel and Doctor Yan Ji introduce themselves. my name is Kurt Demel I'm a medical oncologist and a hematologist. I work in the Twin Cities. I have been in the field for about as long as you. So we probably did our fellowship, around the same time. I work at a large multi-specialty group in Minnesota. we started when I was, a young recruit, like with six people in the department. And now we kind of, straddle both Minneapolis and Saint Paul. And the group is at last count between twenty five and thirty oncologists. So it's really expanded in the time that I've been there. I oftentimes will, refer to my practice as kind of like the Kaiser Permanente of Minnesota were both an insurer and a medical practice. So, the whole issue of value based care and it's very, very appropriate for my group because, we're looking at both outcomes from a micro and macro level. I'm Doctor Yan Ji G. I'm a medical oncologist and practicing a little bit less than you, So I've been in practice for ten years and, I'm practice general Hemonc and also in Midwest. I'm also very fortunate to hold, uh, some research position. We have metro Minnesota and corp and here at Twin City. I'm there. contact PI I'm also the medical research director for our institute. we started to use, um, AI to alleviate a lot of like our burden for clinical practice and also start to thinking using AI in clinical research. I think the topic we're going to talk about today, even over the last few years, things have changed. But certainly I think we all remember our training and the volume of of trials. And we thought it was overwhelming back then. I think, we see the trainees now and what they have to, what they have to deal with. it's, it truly is overload, I think is the perfect term. so why don't we start with this idea of of of the overload, what is information overload? But, but when in your mind does the learning become overload? Like what is that, that threshold point at which it's, it's really almost impossible to keep up. And how would you define it? I'll start with Kurt. one of the reasons I went into oncology is because it's not a stagnant field. I love learning. I love interacting with, my younger colleagues and fellows. And again, I really get engaged and kind of jazzed up with that. I'm not, inhibited by the fact that, you know, this is a constantly changing field and it's evolving and I find that exciting. What I find overwhelming is just like the, kind of like the whole, like you said, I mean, at some point the learning becomes information overload. And I'm sure that we'll get to this topic. But, just like the vast number of trials that are coming out and the data points that are coming out and how we're even interpreting that data, like, you know, not only is it an overall survival kind of question that we now have to think about, but it's the FS one and the FS two. having all of this kind of like new data points that we're analyzing and kind of assessing in terms of like, what's the right step for that patient sitting right in front of you? I think that becomes somewhat, difficult and it's difficult to digest sometimes. And it's even difficult for the quote unquote, experts in the field, I think, to, digest at times. what I found, I guess my unifying kind of like statement to what is information overload is when I feel so overwhelmed that I can't make a decision. I think that with some of the tools that we'll be talking about, I hope that that doesn't become a consistent thing, but I guess my, the difference between learning and, respect for the discipline and the issue of information overload. in my mind, it's like the inability to digest it and interpret it and make a decision at the end of the patient interaction. that's an interesting point and I say this to patients, and I'm sure you do too, that, you know, more information is not always good information. And it's not always necessary to make a decision. and you're right, when we almost taken too much, it is hard to then kind of triangulate into what we really should be doing, because there's always a contrarian position that that we find. I always tell my colleagues. you can always find a study out there that like, refutes what you're going to do, at the end of the day, we have to kind of assimilate all like all of this information Yeah. and Jan, maybe, you can give an example Has there been a time in your practice when you felt like that, when you felt that there was just, there was a decision to make and, and you're going down the rabbit hole of data on PubMed and open evidence and everything else. And you really felt like you had a tough time making that decision. I call myself that I have this memory battery, right. So, so it depends on what, what time of day depends on, you know, what kind of situation. I always feel like sometimes after I did my thorough research, I still very confused, you know? And just because you're trying to put the dots together, sometimes the dots just doesn't get connected. And sometimes it depends on, you know, what trial you read or where the trial is conducted. what kind of patient was enrolled in the trial they gave you totally complete, you know, completely different conclusion. I think as a breast cancer, I think we I remember we had this killer study, you know, it did not show any benefit, you know, doing the chemo for the er, positive breast cancer. But recently there are another presentation from China. They did show. so how are we going to synthesize the data together when we see the patient like this, and we have to give them a recommendation whether I think you're going to benefit from chemo or not. Right. So I think definitely I feel sometimes that it is really, make me, exhausted, not not physically exhausted, just mentally exhausted. And then, you know, that, let me realize that actually, the limit of my knowledge now also pointed me that I need to reach out to some other expert or my colleagues to, to have multidisciplinary discussion. So that actually comes very neat with, you know, now merging AI, you know, how we can use AI, you know, to tease out some of those noise and information. I like the memory battery. it's this we talk in oncology about various types of fatigue. Compassion fatigue is something that I know in, in fields where we deal a lot with serious illness, we have. But memory fatigue, I think is a good one. Or maybe before three years ago before ChatGPT came out and could collate information. it's, the idea of trying to put everything in our brains to then make it. Yeah. It's almost, it's really not possible. Except for those, those few exceptions that you see on, CBS news that have these tremendous memories for everything. we actually have at our practice, we have, an email chain. I'm a breast cancer oncologist, but we have about ten of us. And at tumor boards, as you know, a lot of those are surgical cases. And so we don't get a metastatic tumor board, so we just have an email, chain that we just present cases in all of our hive mind comes together. And you're right, I think reaching out to colleagues, is clearly as valuable as, as doing your own, your own research. but something you, I think you both brought up was this idea of the, there's a, the general oncology, you do general oncology and then there's specialists. So I think in breast cancer I still can't keep up and I just do breast cancer. I had a tumor board this week where one of our younger oncologists was just spouting off every new data point and new trial that I just I didn't even know some of them. how do you think this would impact a general oncologist, potentially even a community oncologist practicing, let's say in a rural setting or in a, underserved area versus someone who's practicing a specific specialty at, at an academic center. Kurt, what do you think? first off, I would echo your email chain, kind of like, accolade. my group has a little bit of a WhatsApp message group. So we're always bouncing questions, day in and day in and day out. Ian's in it and some other colleagues are in it. I totally agree with having a, posse of friends that you can kind of bounce cases off of That's very, very helpful. Now, in terms of like specialists versus general oncologists, I think both are grappling with the same thing. I mean, I practice in both an urban center and in a rural community. And I know for my patients in a rural community, like I'm the default oncologist for everything, I'm going to have to lead them through, all the decisions. But I don't do that alone. again, I still use my, friends and colleagues for, for those cases in which I need some help with, I also think that I feel good about like in most cases about like first line and second line treatment options. I think once you get past like first and second line and, the patients in an appropriate performance status. And you're thinking about third line options. You know, that's, that's when you're starting to dial a friend. And I think that the general oncologist is, a little bit of a dying breed. I think especially with the new recruits that we've had even in my practice, they're trending towards subspecialty oncology, which I think, that's one way of coping with a little bit of the information overload and all the data points that are coming. But even, I mean, I'm sure that you could attest to this, even for the breast oncologist, it's tough. I mean, there's patients that there's people that are only doing Her2 positive breast cancer and there's people that are only doing adjuvant breast cancer decision making. So, I don't think there is a one size fits all. I think we're all grappling with the same thing of like, just like. Information overload and ways to deal with that. I think, you know, I wonder how much of this is is driven by patients as well. as they have become savvier about disease and, and, and understanding, especially in certain diseases that are more public, like breast cancer, colon cancer that has a more public face that, that they have more information. And so they expect somebody who has a potential specialty in something because it's almost like versus, you know, twenty, thirty years ago when, when it was just assumed an oncologist was an oncologist knew everything, right? There's more, more demand, I think for for sure. I mean, every day a patient comes in with like, well, my, Facebook user group wanted me to ask you about this or, I tend to see it like in the younger demographics too, but like in even my older patients like it's the grandkid that comes in and says, well, you know, what did you think? What do you think about this? Or, ChatGPT says, you're in the right ballpark. But did you think about this option? So I mean, it's good. It's empowering for patients and lets us kind of like digest and have a good conversation. But, it's, it's a lot, right? I mean, that, that thirty minute appointment that you might have with the patient to go over their ChatGPT findings, it's, it takes time. Right, right. Jan, you mentioned, AI and so it's actually a great segue into this. And it's true. It's not uncommon now that we see patients print out their ChatGPT questionnaire. And honestly, I'll say versus the days of Google when it just felt like a, you know, diarrhea of information. I think if, if appropriately asked the questions that that, generative AI seems to give patients are seem like they're more, more reasonable to me. but Jan, what do you think as far as kind of broadly speaking, with AI and how, how that has helped us as, as oncologists and maybe where it has, it has been made our days a little more challenging. I am cautiously optimistic about the AI because, you know, we start, we already start using AI with our, note taking. and in our clinical practice, you know, it does help me. And, you know, a lot of times after we see the whole the patient for the whole day. We don't remember, you know, the first patient I talked to him, you know. Exactly. We went through you know, there are five points. I probably remember three. You know, I missed two. So now I don't have to worry about it. Right. So because the the AI actually summarized that for me. So I do think that's helpful. and as you mentioned that the AI, the patient used it, they become a little bit more informative about the disease, about the prognosis diagnosis, even about the treatment. Sometimes the patient brought me the treatment algorithm. I was like, wow. I think that's a good start, and I think that AI, because it's, it's basically pattern recognition, right? So it has a really good use, for us to use in the medical image, recognition, explanation, you know, as a medical oncologist, we rely on the radiology, the pathology so much. I think that AI actually is going to help them to become better radiology, more effective radiologists and and also the pathology, right? Because we can, because then they can do so much more because we have so many, so much more technology and they will make them more effective, more efficient. So AI for us, you know, for me, I think also, like you mentioned, open evidence and all that, right? So it make my life easier because, you know, when I'm trying to find some evidence before I have to, you know, PubMed and trying to, you know, teasing out each abstract to see which one I really, you know, think it's useful now the open evidence, you know, with the reference. So I think that's really important because you have to know where those data is from. Right. Because I think that's, that's so important because AI sometimes make can make some delusional summaries, you know, some summary. So we as a physician, that's, really dangerous thing because you read it. You think that's true. But I think that, you know, at least open evidence gave you reference and then read the summary. I go back to the reference, you know, make, I think, make my life. Oh, definitely more effective. Now I talk about the research, AI, I think it will be a play, a much bigger role in, patient screening because we have such a hard time to screen the patient despite the best effort, we still enroll like less than ten percent patient into our clinical trial. Right? So that's a really sad number. we all want to enroll the patient to the good trial, and a really good match trial. So I really think that AI is going to play a bigger role. So I think that's all the good thing. The bad thing is that, as I said, some of us, including me sometimes can be like, misled by the AI because it makes things so easy. I'm especially worried that some of the young oncologists is going to use AI so much. They actually they actually forgot. You have to still know the basic, guideline, memorize that and then, and then memorize the practicing change clinical trial that has to be in your brain. You can't every time search the AI I think the, the brain is like a muscle. You have to train it, right? Because you cannot just use the AI as your, brain and just using that as a data seeking reservoir. So I think that part I a little bit worried. And the second is, you know, how the patient is going to use it because a lot of times my patient came in, you know, they search something on the website, you know, Gemini all that they feel like that's really the thing. It's true. Right? So they did not go back to the reference because remember, the AI is really just seeking all the data, you know, from what they can find, right? So I think in that part is I'm a little bit worried because a lot of times I spend a lot of time to explain why this is not true, why we cannot do it. And then the third I worry is that we use AI to give our the recommendation of the treatment and ignore. So all the treatment, is from all the clinical trial and all that, right. But the patient is different than the trial patient, right? So the patient, they have different comorbidities, they have age, they're from different area. So all the things has to be added to the decision making. So I think that part, I also think that we need to do a better job and to educate ourselves, to educate our younger oncologists. Educate the patient. Right? So that's why the AI cannot replace oncologists. I'm pretty optimistic about that. I'm with you on that. I'm an AI optimist, at least for now. Although we'll see every month there's, there's some new, uh, trick that, you know, these different agents are playing on their masters and eventually taking over. But, um, but yeah, you bring up so many, so many good points around, um, you know, particularly around the thinking, right. I do think about that a lot that, that when, and I have three children, I have a twenty four, twenty two and a seventeen year old. and it's interesting just generationally, even though the ages aren't that different, they have different comfort levels with, some of this and, and the idea of just having a second to, think about something and not knowing the answer. I think that alone has value when you're thinking. And for our trainees and Season. Residents and fellows. This immediacy of the answer in front of you, sometimes taking time to just put stuff together and then looking down and finding things I think is a valuable is really valuable for our trainees. And they're not getting it. I was just going to mention that it's like, we want immediacy, in everything in life, we want an instant response to our texts. We want an instant response to our email. The patients want that, every time they send me a MyChart, I feel guilty if I'm not like responding to it, like right away. But like sometimes life, you know, you need to digest some of the, some of the intricacies of, the question to come up with a good response. And one thing that I do like when I run with the fellows again, again, everyone has different comfort levels with AI. I'm learning as well. But I say you're not allowed looking at open evidence until like you can kind of come up with at least response here, like, tell me what you would do with your, like with your experience to date. What would you do or what would the research tell us to do? I don't, and I try to tell them, you know, it's okay to be wrong. It's okay not to use open evidence until the end of rounds when we're like going back and like thinking about our, thought process. But you have to have a little bit of a repertoire, to be able to like, have spontaneous discussions with people. you can't always have open evidence in the palm of your hand and say, oh, well, let me check what open evidence would say, right? It's like, you have to have some comfort. And to what Jen was saying, I mean, AI will never replace an oncologist. You need that humanity. You need that human, that human front. And that's never going to be replaced by our devices or by, ChatGPT or whatever, whatever comes next. I was thinking about what John was saying around, AI doesn't have that that nuanced information. now, you know, an AI company can push back and say, well, just enter the nuanced information and we'll give you an even better answer. But ultimately, like you said, there's this frontier of humanity. And one day maybe it'll hack that as well. But for now, yeah, as an oncologist, when you think about what our value is to the patient, it's true that, you know, just spouting off clinical trial data and quickly making a decision for people is not the is not the value that I mean, it's part of the value, but it's not really, especially in this new world, what we're bringing, it's kind of navigating and guiding people through some difficult, journeys for sure. And it's all about how to even ask the right question in AI, right? It's not even like that's giving its output. It's all output, but the input is just as important.