Hey, everyone. Welcome to LabReflex. I am here with Dr. Aakash. How you doing, man? I'm doing good, Chris Thank you for having me again. Yeah, of course, man Yeah, so today I think we've got a pretty good lineup of topics. We've got some workforce issues. We have some interesting prenatal Discussions we've got AI tools And we have new recommendations on error Diagnostic error and guidelines from from CAP. So I think we've got some pretty cool topics to talk about decent Mount of yeah, as well as pretty neat ones, too so I'll just start with the first one here that apparently, I mean It's more as an obvious topic It's just that nobody talks about it frequently enough just because it's becoming more or more and more, right? Hard to talk about. Yeah. Yeah, exactly. So what has happened is that there has been a Workforce crisis come into focus which was already there. It's just that people were turning their backs to it Yeah, AMA has adopted policy highlighting the national shortage of medical laboratory professionals. Oh, yeah what's going on in that is the fact that You know for years the laboratory has described that there has been a Workflow workforce shortage at the looming problem. Yeah now that they have Come to realization that it's not a looming problem. It is. It is currently a problem. Yeah, right Yeah, although interestingly as CP's most recent vacancy survey if I tell you showed that While the vacancy rates improved somewhat. Yeah compared to the peak staffing that was being done in kovat and all that They remain substantially higher than the pre-pandemic levels. Yeah what is Concerning is that it's kind of a two-edged sword thing going on We are hiring But right nobody realizes the fact that we are hiring because there is there are more people who are retiring Yeah, so it's not like we have a vacancy for seven more We have vacancy for eight. Yeah, but those vacancies of eight for any original vacancies were three It's just that four are have retired. Yeah, exactly. Yeah, it's like If you think about it, it's it's the vacancy is there and we are hiring But still like we have we are short of like the workforce number is short And the other thing that is is definitely also true is that the training right? They're just aren't enough slots exactly to backfill. So, you know, we might have all kinds of demand but what we don't have is schools to service that demand and We aren't always filling all of those schools And so, you know, I think that we had a field notes from last week on Friday, I think Where we talked about clever approaches to education and I think that that's what's gonna have to happen here is MLS education is gonna be something that is gonna need to be on the forefront of the field for The foreseeable future and I'm glad that the AMA has pointed it out Yeah, and as you said like The MLS and MLT programs that are closing for the past several decades now Now the recent AMA resolution which falls for investing in these educational programs as we had discussed previously as you had you just Yeah about it. I think the key point here is that it's not just the shortage of the Shortness isn't like simply because of the vacancy positions, right? It's also that we are losing good people and decades of experience to it Oh, absolutely. Yeah, right If if there's an if there's an individual or a tech that has been working in our lab for let's say 10 years or 11 or 12 years Having the same amount of expertise is becoming hard. Yeah, right, which is being compounded bad Nobody showing up Because so whoever does whoever does show up is new right then you and they're filling in the role of somebody who just retired Who just retired so experience of 15 years, right, right and getting those is becoming harder, too So you might be like closing up the gap, but are you really closing the gap? Yeah, exactly thing Yeah, no, I think I mean everyone knows this is a problem. This isn't like something new and yet Are there resources that are being allocated to fixing it? I mean, obviously we have a school here right and we're doing our part. We've got a great school here But it doesn't seem like it's likely to be sort of fixed in the short term again I think a couple weeks ago Aakash and I Suggested to our listeners that we would be happy to set up a medical school in New Mexico for five million dollars each We would also be happy to set up an MLS school somewhere for also five million dollars each I think I think pray would be okay with that, right? I mean like I'm confident Michelle would be happy if I was making a few extra million dollars So we'll set that up but yeah, I know I mean without the educational component I I don't really see a path forward. Obviously, we're gonna be leaving leaning heavily on AI tools and automation That's gonna continue to be the name of the game. Yes, but we still need good people and It's now at least getting the national attention that it deserves You may remember that we had a we had a topic A few weeks ago where I think it was out of Michigan. Somebody gave 50 million dollars Yeah for education for like the clinical year and I think that that's the kind of thing that's gonna have to happen More often in the future. We're gonna need to be we're gonna need to see I think Forward-looking people to really improve this this thing So the other the other topic which I think is a little bit More of a good news story is that cap has given? Guidance around Diagnostic error reduction. So the last time they gave something This guidance was in 2016. So now it's ten years later almost a decade. Yeah. Yeah, and and I think that the guidance actually is Oddly consistent with what they gave in 2016 What they really want to see is they want to see inter divisional and intra divisional Assessment of some of your sign-out. So, you know you is heme path for example, you you need a second person taking a look at some of your cases and then a Standardization of that process, you know, maybe every hundredth case or whatever whatever that standard is and Then a grading of that process, you know, so some form of standard evaluation, you know High low medium, you know accurate not accurate whatever and That's more or less what they gave in 2016. There's really not much of a change Yeah, so basically what they are trying to do is like it's they are like they want to reinforce this culture of Evaluating and Assessing all these cases because they are the idea is that We were we have been doing it It's just that they want to reinforce in a manner that there's a structure to that so that everybody gets has to follow this particular structure exactly and Reinforces a structure as much as they can the question is like because they found out that most of these Errors that happen are because of the interpretive errors and these basically are not actually the errors if you may think about it It's more of like as I say interpretive Judgments, right? Yeah, but they said that the bay basically are reiterating the fact that okay it needs to happen more frequently so that The judgment errors or the interpretive errors are reduced It's so that and and not only they are reduced But the results are being delivered to the patients in a timely manner Yeah, so basically structured case review peer peer review processes timely review and continuous Continuous learning from the discrepancies. Yeah. Yeah, exactly Yeah, the goal really isn't perfection and I don't think the cap is expecting that the goal is building systems that help, you know professionals catch problems and then sort of Surface solutions and then and then do that prior to reaching patients as you said, and that's that's the real key. Yeah. Yes Excellent. Yeah, so so the other one I thought was interesting was this story About AI solutions Mayo and Microsoft maybe having a partnership there. I thought that was kind of an interesting story It looks like they're interested in trying to build a healthcare AI model native to The data coming out of Mayo you'd heard about that. We talked about a little bit. Yeah, and basically, I mean we I mean the AI they are different AI models already in the in the market available, but they are basically trying to structure it with Microsoft Using the data from Mayo and they want to like structure an AI tool that is specifically Guided towards medicine. Yeah, right in general and healthcare. Yeah, so this is all about it. People have been using different all these models but For in the end their practice for different reasons, but those models are more Expanded models like an engineer can use it a healthcare professional can use it a school kid can use it Right anybody can use it right? But this would this would be Sophisticatedly structured towards healthcare only that's my understanding. Yeah, I think that that's interesting, you know Recently the story was came out and may have been resolved by the time this episode comes out but Anthropics Fable 5 was sort of shut down by the government because it was too powerful prior to that fable was being Control live I guess the Anthropic was trying to to Temper your capability of using fable because it was somewhat dangerous. They had actually not turned on fable for a while And so there's some interesting things that are happening in the space for general-purpose AI tools that I think is interesting a lot of the Smaller tools are actually Chinese tools You know a lot of those models are Chinese and and I think a lot of companies are leaning towards Trying to figure out in the era of Government shutdown of these major models trying to create their own models And as training becomes easier I think it's very likely to happen where you might even have an LLM trained on only UTMB data For example, um, because maybe our patient population is a little bit different or whatever Or you may have some sort of way to tweak a model so that it's unique to your use case I think that's kind of what's happening here I'm not terribly surprised that we're seeing this a little bit now the frontier models are so powerful that You know, you may have some limitations moving forward You may want to tweak them for your specific use case. So anyway, it seems like Microsoft is moving in that direction Microsoft had announced You know, this is six months ago now a year ago now that they're sort of strategically partnered and aligned with chat GPT It looked like now that may have changed a little bit But I'm I'm not surprised and I think we're gonna hear a lot more of this I think we're gonna hear a lot more of labs in particular partnering with some of these tools Yes, and good the good thing that you brought it up because think about from the AI's perspective, right? The AI system works on standardized information numeric measurements. Yeah Just numbers and data that straightforward stuff. Take a wild guess. Yeah has it. Yeah labs have it. Oh infinitely more than anybody else Exactly. Yeah, it's easier to them based on the numeric Datas and set points rather than like free text physician notes, right? Yeah Yeah, exactly now the laboratory medicine may become one of the primary engines for the AI and that actually opens up new paradigms for the laboratory management's where we can In the near future utilize these AI and directly incorporate data for quality monitoring maybe error detection. Yeah and utilizing all these management tools and basically, I think we would be the laboratory medicine personnel's would be more likely to implement utilize and Expand these models in the better way. Oh, yeah use proper. Yeah, and and I can even imagine a world's Aakash where Yeah, you're The lab is not only the the data layer where the models are being trained on the data, but it's also at the layer of Validation and then verification of results from the AI. So here's what I mean. You could imagine that The AI is used to predictably model how frequently a patient has disease whatever whatever disease and Then with some periodicity the laboratory is used to make sure that the AI has not drifted away from That diagnosis so maybe it's not common for us to order a follow-on test for a disease that's suspected But maybe what we do is we say, you know what some of the time it's gonna be wrong Let's periodically check and make sure that the AI model is correct The data is gonna be in the lab all that data is gonna be in the lab And so it seems like not only are we gonna be at the implementation stage But we're also gonna be at the verification of validation stage so every stage of AI development and implementation in the hospital is gonna be Intricately involved in the lab is all lab medicine. So I think that that's kind of an interesting sort of Take home for all of us The last the last thing that we wanted to mention for today's episode was this cap symposium on Sort of like re-examining the the race based prenatal screening There's been a lot of discussion about that a lot of people Naturally would think that this is maybe more of a political. Yeah angle, but it really Isn't as much as it might appear at first there's a lot of like confounding variables related to race based prenatal screening and The way that we think about race sometimes can be confusing because Generally speaking you you might think of a an ancestry as being your race, but in fact That's actually much more fluid Who is what race is? kind of actually changing, you know, yeah from a social and societal standpoint and so what we want to do is be more accurate and more precise when it comes to our prenatal screens And so we're trying to strip away some of the ambiguity with race based screening It's I think that that's actually pretty pretty clever You can only do so much But I think that if we can really get down to the quantifiable level of what risks there are in prenatal screening I think that's gonna do everybody a better service. Yeah, no, that's true. And then as you said, like the idea is to Modify some of the norms that have been norms for so many Exactly with everything changing. I think these models still need some update in some sort of way. Yeah Yeah, exactly. This is one of the caps way to like, you know updates into certain models on that. Yeah, exactly Well, anyway, it's been a great episode man. I mean we've talked about yeah, obviously Staffing AI race-based, you know prenatal screening. I think that it's been a it's been a fantastic episode I'm getting that for us from our listeners. What do they think? Oh, absolutely. Yeah. Yeah. Yeah If anybody has been using AI implementation in their labs already would love it. Yeah. Yeah, and how's their staffing situation? What's happening with the staffing I agree with you We also are gonna probably start a Little bit of a series on the upcoming conference ADLM is gonna have their yearly conference. Yep You guys will be hearing Some of the changes to the formatting of the episodes We've got deep dives coming out on Wednesdays field notes coming out on Fridays And as always if you you know, like what Akasha and I are doing, please rate and review and subscribe That really helps us quite a bit It's really motivational for us to keep getting behind the microphone As opposed to you know, I don't know going and having a coffee or something. So yeah, we'd appreciate it Anyway, well, man, it's been great talking to you buddy here. Same here. Sure. All right. See you all next week
