Deep Dive: The DMT Defined (Part 2 of 6)

Hello, everyone. This is our second conversation in a multi-conversation series on diagnostic management teams with Dr. Michael Appasada. Thank you again for coming, sir. Thank you. And so, I'd like to really just go ahead and get started. We are going to talk today about defining what a DMT is and what it is not. And I think that's a challenge for a lot of people because, in particular, many things kind of look like DMTs. They sort of smell like DMTs. They walk and talk like DMTs. And so, I think it would be valuable for us to kind of outline what it is that we mean when we talk about this kind of structure. So, really, with that, when you hear DMT used today, what do you think is most misunderstood about that? Oh, that's such an important question. Really central to this discussion. So, let me tell you that in 1984, when it was apparent to me that the people who were patient-facing providers needed help picking the right tests and knowing what the results meant, there was no plan to understand how we were going to deliver this information. So, some years passed and I had moved from Penn and now I'm at the Mass General as Director of Clinical MGH, even though there were 2,000 who actually knew and understood coagulation. So, we had a rehabilitation hospital, Spalding Rehabilitation Hospital, across the street from the MGH. And they thought that maybe I would be willing to go across the street and see their patients and provide a consult. So, here I was sort of a hybrid pathologist and sort of an internal medicine person because I saw patients with bleeding and clotting problems. So, I went over to Spalding and started seeing patients and I realized that I could do a good job at this and it would be just like an internal medicine consult. But boy, was it going to be expensive in time. I might spend the whole morning figuring out how to provide something that's essential because everything was written. There was nothing electronic. So, I had to read charts, pages and pages of charts to be able to take somebody who had a stroke and maybe fallen as now in a rehab hospital and talk about using anticoagulants and what the risk of bleeding or thrombosis is. So, I realized the quality of my note had to be as good as what a cardiologist would provide about a heart or a rheumatologist would provide about joints. So, as I went back, I thought I could do this much more efficiently and I'm working in an area which is dominant test information over valuations from examining a patient. So, I can make a diagnosis of a clinical problem in coagulation, bleeding or thrombosis by looking at the lab test primarily and asking the patient a few questions about history. So, understanding that, I thought I could do this quickly and I don't have to spend my whole morning seeing this patient. So, the information about a patient became electronic. I could read it from my office. I could see what the history was and then I could do the two essential things that I have said are the critical ingredients of a DMT. You have to say what tests are necessary and you have to provide a diagnosis for that patient, patient specific in the clinical context of that patient. And it can't be a cookie cutter thing. It has to have the same quality and the same amount of information that is helpful to whoever is ever managing that patient just like any other consult and it's not the usual consult because I'm not going over there doing the same sort of evaluation. I don't need to look at the patient's knee. I can read what they say about the patient's knee. I can even see the film if I'm reading the report. So, I realized that the diagnostic management team as we implemented it in the mid-90s had to have two major things to it. We're going to recommend tests. We're going to say what the diagnosis is and we're not going to do what lots of laboratories do and provide a list of things it could be. So, it's not very helpful to the doctor who gets a cookie cutter response that every time the PTT is elevated it says the same thing. That every time that somebody has a factor 5 Leiden, that it says the same thing about inheritance and could produce thrombosis. It's too general. So, it had to be patient specific. Obviously, if you have a Leiden and you're 26 and you're on oral contraceptives, you write a different story than if you're 96 and you just had a stroke. So, the patient's specificity was key. It had to be as valuable the way I was doing it when we started the DMTs electronically where I could do 10 in an hour as it was if I went there and spent the entire morning looking at the patient asking a few questions. So, this was a very different thing. It meant I could take something that might be endocrinology where there were some signs and symptoms that I could see were present. But really, it was very dependent upon what was the TSH level in this patient who may have thyroid disease. So, that's the way it started. So, lots of people said that's really valuable and I want to take an even shorter shortcut. So, somebody figured who was remembering the days when they made a lot of money as a clinical pathologist, when they got paid for interpretations that said something as simple as the glucose is high, which is not very informative when you have glucose and doing the same thing and saying, oh, now I have a DMT. They put a name on it. So, for example, there was a hospital in Florida where the ER doc was friends with a pathologist and they said, you know, I'd love to get back the money I used to get for getting paid for interpretations of testing, not just providing the test because that all goes to the hospital. So, they said, okay, I'll write a note. We'll follow the rules. I'll say, please provide a consult for this elevated troponin. Well, you know, two days later, they would write something like, this is consistent with a heart attack. So, did they ask for it? Yes. And was it relevant because it was abnormal? Yes. So, can you get paid for that? Well, it's totally useless. They've already figured out if the guy had a heart attack. So, there have been many ways to manipulate this and say it's valuable. Unless you're providing it at a time when it's needed for clinical decision-making, then it's not a DMT. So, a lot of people try to say, okay, I did everything. I've met all the regulations and I deserve to be paid for it. I'm calling it a DMT. Humatopathology says, we're already a DMT. We already do the morphology and the genetics, but they come in in pieces. And what happens is the hepatopathologist puts a couple of pieces together, maybe the flow and the morphology and says, okay, this is a DMT. Well, wait a minute. Cytogenetics or FISH or something genetics, they're all coming in and you're going to provide those as little bits of information. The DMT has to unite all the information. It's an integrative report. So, if all you're doing is saying the value is high, that's not helping anything. And it doesn't provide information that somebody can make a clinical decision about it. So, I think the key elements of which tests do I need means it could go back and forth. It's not just, I wrote this and I'm done, please pay me, but it's valuable. And what it means with a diagnosis and providing one or maybe two diagnosis, this is lupus and it might be rheumatoid arthritis, but I'm not writing 16 different autoimmune diseases. So, the idea here is it's as specific as possible. And that means it takes time. It really takes time. It makes me, there's sort of a movement in the clinical space to almost proceduralize the diagnosis of very simple things. For example, Amazon One, right? Or I think it's One. Amazon Health, I don't remember. What they do is they basically diagnose only very simple diseases and they don't really look holistically at the patient. And so, in a way, the DMT kind of represents a movement in that direction where what we're trying to do is avoid a completely holistic assessment because it's really disease kind of category specific. In this case, it would be like thrombotic risk, but then completely encapsulating that. Whereas in previous eras, what we were doing was giving an interpretation for a test, one test, maybe even two tests, but it was only the test. It wasn't the disease profile that we were sort of trying to characterize. That's a key point, Chris. I think being able to say more than the glucose is elevated and it can take the whole picture of a diabetic and talk about the status of renal function and some other things in a very concise note. And I just came from signing out a case and goodness, it was a woman who has all kinds of autoimmune problems and is pregnant. And she was being evaluated for anti-phospholipid syndrome. And there's a controversy about whether or not a particular test, IGA, makes the woman at higher risk for thrombosis during pregnancy. And in that room, we opened up Open Evidence AI and it has all the papers linked to it. And we clicked on the two papers and we found one from 2001 that said definite thrombotic risk. 2023 said, no, it's not a risk at all. And then we pulled up the other one, a large language model, and it said controversial. So we were able to put that in there and educate the patient-facing provider that maybe their knowledge is 2001. Yeah, 20, 25 years. Yeah. So if you were to define what a true DMT is in one to two sentences, how do you think you could do that? It is a consult that is expert driven and patient specific in the context of the patient's illness. And its goal is to conclusively establish a single diagnosis. I like that. Yeah, that's pretty succinct. I do kind of wonder how you perceive the difference between a DMT, say, for example, and like a utilization management structure, right? There's a lot of these sorts of things on, you know, in hospital administration. We talked about this last time about how there are these guidelines groups and utilization groups and a bunch of different structures. How do you see the difference from the DMT to that? That's really an important distinction that everybody has to understand. DMT is thought to be something progressive. People say, oh, we're doing DMTs. Really, when you're talking about utilization management, we do too many CBCs. Yeah. Are the same people who do the DMT able to look at that data? Well, it's kind of related. You see cases and you could say, oh, that's something we could reduce. You could say platelet function tests, a more sophisticated thing are overused and you could have your DMT group assess that because they have content knowledge about, in this case, coagulation. Yeah, of course. But remember what we just said about a definition for the DMT? We said that it not only tells you what test to use, but it tells you a diagnosis. Utilization management has nothing to do with the diagnosis. That is a separate thing. Is there a little overlap with the DMT? These are the tests that are useful. Sure, it does, but that's in a group utilization management. When we give a DMT report, that's about a very specific patient and we talk about whether or not this test is useful or that test isn't. So utilization management, maybe a little overlap, but it's groupthink and it doesn't say anything about the diagnosis. Yeah. Okay. So then it seems to me like, what about more, uh, at a lower level things? For example, you might have algorithms, these sort of protocols, order sets where maybe it gets you to the right endpoint relatively quickly, you know, playing the game of Plinko with lab testing. Okay. How is a DMT different fundamentally than that kind of structure? So the DMT utilizes the concept of reflex testing. So I'm providing a case report and somebody has a prolonged PTT, very simple. And if you have a prolonged PTT, you want to do a PTT mixing study and then based upon that you can do a lupus anticoagulant or factor assays. So you see, I used reflex testing and I use that information to provide the recommendations for testing. So I would say we want to do a PTT mixing study and I would say that in my narrative. So I use that information, but reflex testing is not the DMT because reflex testing says to figure out the diagnosis, you would want to do this test first based upon that you do this test or that test. If I said that to a doctor who has a patient in front of him who's seven years old awaiting a tonsillectomy, the doctor would say, I don't want to know the reflex testing panel, just please tell me the PTT is long, can I do the tonsillectomy on this child? So reflex testing is a tool, but it's not the diagnostic management team because we're not establishing a diagnosis, we're telling you how we think about it to get one. Yeah, I think a lot of times the analogy that makes it salient for me is that there is such a thing as an internal combustion engine and then there's such a thing as a car and oftentimes the one uses the other, but I could imagine that you could design or speak about the internal combustion engine as though it helped propel vehicles along, right? But it's the car that we're talking about when we are talking about the DMT, it's the whole thing where we're delivering a result as opposed to the individual component parts, the wheels and the engine and the roof and the windshield. And so yeah, we can use a bunch of things, but the resultant interpretation and what to do next is recommendation is what matters for the DMT. It's a great paradigm and I've never used it in all the talks I've given. That's good. Well, you can borrow it. Okay, thank you. So one quick question. So one of the big pushbacks on doing DMTs in general, you've mentioned it before, is the effort that it takes for people to do the DMT. So is a DMT, is real time support required for a DMT or is it kind of case specific? I would say real time support is important, but you have to understand what real time is. So if somebody has had a blood clot and what we're asked to do in the DMT is figure out what's the likely explanation for the blood clot, we're not going to change anything in an hour. We don't do a DMT, for example, on why somebody has a blood alcohol level that's way higher than the level of intoxication, because most of those people are showing up at three in the morning and we can't really do a good job with that. And there would be a lot of things to say about that patient if we could do DMTs 24 hours a day, seven days a week in real time. But somebody has to make a decision about what to do. Is this chronic alcoholism? Is this acute? What if it's too high? Should I use x, y, or z? So we have picked areas where we can make an impact. And so the real time part is an absolute necessity. As far as what the critical parts are for a DMT, it has to provide information, diagnostic information at a time when clinical decisions are being made. So if you've made it 10 minutes after some clinical decision was made about treatment, it's too late. It's too late. You can't say, I can't get through all the cases today. And somebody's already bleeding. You can't think that it has value. It's really kind of common sense. Is this valuable as a consult that you get from internal medicine? You have to compare it to a good consult from internal medicine. Not when it's late, not when it's uninformed, but something that can move the treatment along faster. So it has to be patient specific. You can't have a cookie cutter thing that you just whip in there for everything. It has to be expert driven. And the expert driven doesn't mean I have to write it. I can take advantage of all these large language models and give it the information and then it creates paragraphs. And then I as the expert can modify them, saved a lot of time from writing the original paragraph. But I am involved in that case and saying this is the answer. And so expert driven, patient specific, and then the question of how good does it have to be. So I use the target of 99.9% is our goal for getting them right. Now that's a lofty level. That means we're only making mistakes one out of a thousand times. Can we do that? And we absolutely can, especially now with all the abilities we have looking put the same case into seven different large language models and compare them all. See what they get. They all give the same thing. It's pretty hard to say that one's going to be wrong. This is the way they used to type the original books that were available for print. Before anything electronic, they would ask four assistants, secretaries at the time to type the book and they made errors, but they made different errors. So this way there was no error in the final version of a book printed four thousand times. And so if we can do that with the large language models and set it in, boom, boom, boom, boom, boom. I recently looked at a case yesterday and they said, well, here is the perspective of somebody who's conservative and here's the perspective of somebody who's aggressive and this person is in the middle. So wow, we can get that kind of information about the success of this. So negotiable and non-negotiable got to be patient specific expert driven. We can make it accurate ninety nine point nine percent of the time. What do you say? What do you say to people who are trying, I think in good faith, are trying to improve the quality of medical care for our patients, but that the majority of cases don't require this level of investment and that maybe it's more like a case conference that maybe what we can do is only the interesting cases or the complex cases go to the DMT and everything else is just kind of put into a bargain bin and they just kind of churn through that. How do you respond to that? Yeah, I think we've overestimated what health care practitioners know and we have to be realistic about it. Remember that people like nurse practitioners and physicians assistants are the only experts that are seen by a patient who's sick. In a lot of cases across America, start from there, maybe that's the beginning part. So I have somebody in the family who's a nurse practitioner and she said, so I'm going to tell you, I really don't know what a TSH is or does. Yeah. So that's a pretty common thing, right? Shouldn't I know like what the test selection should be for somebody who might be hyper or hypothyroidism, a patient? And I said, so that's unfamiliar to you. I said, what would I, what would you tell me about a patient who has an ANA of 1 to 360? And her answer was, I don't even know what an ANA is. So the bottom line is, give us your test results. You look at the patient, we're going to give you an answer. This is hyperthyroidism and it might be this particular underlying disease. That's one of many different things that makes you hyperthyroid. And we start from there. Now, are there physicians in private practice or clinical practice in an academic medical center for 40 years and there are specialists in thyroid disease? Oh, sure. And then we're not going to be benefited by this. But you have to think of who are the people we trust as providers who are seeing most of the patients in America and they are not the experts at all. Yeah. I think it's right that it's a challenge to expect that your intuition is the right driver. You've only seen what you've talked about last time about this sort of unknown unknowns concept, right? And so it seems to me like that's the first component that you really just don't have any way to follow your sort of gut and understand when there's any problem with your diagnostic process. And the second component that I think is a little bit bizarre outside of healthcare, this notion that there's such a thing as an interesting case is actually a very bizarre concept. People don't like being characterized as interesting. What they want is everyone to get a standard of care that is as high as possible. Just because you have some rare disease doesn't mean that you deserve good health care and someone who has a more, you know, run-of-the-mill disease doesn't deserve the good health care. And you might have to wait a month to be the patient in the case conference at which point everything is all too late. So this idea that people are interesting, that's the main focus, not diseases are interesting. And so like if everyone is interesting, then that means that everyone deserves this level of care. We shouldn't preserve it for a special view. So here is the problem that leads to the interesting case conference. It means I was at the Mass General for a while and I went to many, maybe most community hospitals in the whole state of Massachusetts during the 19 years that I was there. They would bring me in and say, here's four cases for coagulation. And I would do the interesting case conference. It might have been a year old, but really you couldn't expect that community hospital to have 25 people who could cover 25 different areas of diagnostic medicine and be local experts. So the real challenge here that I think we are still facing, but we have some ideas about, is how do we get that expert care for somebody in a small town in Texas and the person who lives next to a major academic medical center. And we have a few ideas. So we're now reality based to know you can't have it in a way that covers most of the areas of diagnostic medicine even at a big academic medical center. So maybe what we do is we build a regional diagnostic center and we made reference to this in a previous conversation, but this means that we could have experts in 25 different areas. And if we're all electronically connected, the one expert could be in Cleveland and the other expert is in Texas. And so we have all these people listed. The case comes in, we triage it just like it comes to an emergency room and we say, okay, that is a thyroid case. The doc in Cleveland, he or she is the person who's going to get this case. And we have all standardized, much like McDonald's standardizes everything they do with food, we standardize the way the case material comes in, how it gets reported, what we consider, all the AI tools, and then it comes out. And then we have a southeast regional diagnostic center. And I suspect that it would grow quickly into something where there's a northeast one and a northwest one, because it's probably going to take maybe 10 diagnostic centers in America. But you know, if we had that, I bet you everybody who had a prolonged PT or PTT, lots of common things, right, that they could get evaluated. In that, so I, of course, I think that this is the only way that we can distribute some of the expectations from the DMT. But in that model, we live in a society that is often litigious. And who do you think then ought to own the diagnosis in that model? That's a really good question. I got to say it's one we haven't figured out. And there are a number of things about the DMT that we haven't figured out that are mostly related to legal and business. You have to start by saying, how do you deliver the right answer? Yeah, right. Boy, we've worked 99% of the time on getting to the right answer. The next question is, you know, how do we bill for it? Here's somebody who has had Plavix, but they had an MI, and then you put them on Plavix and they had a second MI. And nobody wants to do pharmacogenomics because the insurance doesn't pay for it in my state. Right. Like, well, okay, I didn't figure that out. I thought we were just trying to stop this guy from getting a second heart attack. So the reality is that it's tough to practice medicine in America with all of these challenges. And it is much easier to practice in a system where we say, what does the patient need and how do I get it to them? And how can I do it and save money? Obviously, if you prevent a second heart attack, that's a lot cheaper than worrying about how much the pharmacogenomics testing is. So we're in a complex state of doing this. And I think what we're trying to do is establish one model. If we could get one diagnostic center that was either academia-based or private practice-based, and we could show people how it works, then let all the business people and the lawyers get at how to make it work. And I think that's their portion of this. But what we want to do is to show we actually can take all the people in a region and do all the coagulation cases for like the whole state of Texas. Yeah, just to put a bit of a bounding box on what we think though. Right now, the way the DMT, who owns the diagnosis, is the signing out pathologist in the DMT. True, true, true. And what we're considering are alternatives to that model. For example, if we're working with an entity outside of our state, for example, is it the diagnosis is owned by the consulting physician or the consultant physician? Who does the sign out? And so it seems to me like there is a solution. It's not that you're saying that there's not a solution. It's that we're not sure what solution is the easiest to implement and the most useful. So you bring up a really important point. We've been doing this in anatomic pathology for a long time. Right. So my brother-in-law had a mass that was impacting his spinal cord. So the question is, what is it? Is it a sarcoma? Is it something more benign than a sarcoma? So the doctor in Alabama in private practice knows somebody from his residency and she's practicing in Ohio. So he sends the slides to somebody in Columbus, Ohio. And what she says is, this is what I think it is, who owns that diagnosis. Actually, hers is a recommendation. So she sends it to the pathologist in Huntsville, Alabama. And then he says, okay, I think that's what it is. I agree with that. He's responsible for it. That's right. He agrees with it. He pays his former colleague in Columbus a fee that she charges for that. But she is not the responsible physician. And I think that's the model that is certainly going to work. The interesting part is that was really never laid out legally. It just seemed like the right thing to do. And then it became part of practice. And the question is, will they do it differently if I ask a colleague in Columbus, Ohio, what do I do with this total cholesterol of 800? Yeah. And the reason that it made the most sense is because of our sort of federalist approach to government. This is like unlikely that this is how it would work in other countries. Correct. So we're sort of like backing into a solution because of our legal and insurance restrictions. And we're trying to solve a problem as best we can. It's a good solution. It may not be the perfect solution, but it's a good solution certainly for the United States. So I think that's an important point to make. This really is totally focused on how do we help the most people in the biggest way. It's not like taking some political side of anything. We don't care. Whoever can help us help the most people, that's what we want to do. And we want to be able to help them and make it available across the board. It should not be limited because you don't have the funds to figure out why you're coughing and you're doing it for six months. Yeah. Because you can't afford this. Yeah. To kind of go back to the sort of ideological framing of a DMT. How do you think, what are the warning signs that a DMT has kind of drifted away from its purpose? So you could see what if we put somebody in charge who really didn't know that topic. When I was at the MGH, there was a person who was in a position of influence who said, you know, if you're not here, I'll do the coagulation DMT. I'm just signing them out. I said, you didn't even know what a lupus anticoagulant is. I said, what are the possibilities for a diagnosis if you have a prolonged PTT? No idea. He said, really, I can just leave it at the very high level. This patient may end up with a bleeding problem. Really? That's all you're going to say? You're not going to say what the diagnosis is? Nah, we don't need that. So if you don't value it, getting the specific diagnosis, doing it in time, making it patient specific, you shouldn't be in charge. So anybody who's saying we're doing DMTs and the person who's in charge of the DMT has absolutely no background or experience in it, it has slipped. If you're starting to do DMTs and you say, well, you know, I don't have time today, so we'll do them tomorrow. Well, that doesn't work either. I had a friend who said we tried to do DMTs, but I only have the residents for an hour. I said, so if you have 25 cases and it takes three hours, you have no resident to present the cases and organize your report? No. I mean, you need more than one person. It can't just be you. So whoever that person is, they have to have some knowledge. They have to be invested in it. So there were all kinds of things thrown in and you can see, okay, the residents only stay for an hour. Well, that's slippage too. Somebody who really doesn't know it, that's slippage too. So it's easy. If there were some inspection, think ahead 10 years, where we inspect so many processes in the lab, what have you inspected the DMTs? And then you looked at somebody and they just walked in here and they were from some other area and they didn't have clinical knowledge and they didn't know what CHF stood for. Okay, that slipped. Boom, boom, boom. No, no, no. You don't get to bill for this. Yeah. I'm curious, you know, as institutions may or may not want to implement a DMP, what do you think would be valuable for them to understand kind of going into the process that you've learned now where it's hard to implement a DMT because of A, B and C institutional reasons? Yeah. So I would say this has really evolved very recently and it's with the advent of all these AI tools because it takes a lot of time to write the report. So you're right. I gave a talk at one of the med schools in Philadelphia and without all the AI stuff, they would be spending half an hour at least writing the final report, maybe with an error or two here and there because it's just them. It's not seven large language models looking at the whole body of literature ever written. Right. And so you can't do that. I mean, you have too many other responsibilities in your day and it's hard to hire people. The payment for a DMT, member of the business and legal aspects lag. So they say, well, I can't pay her for, you know, doing 20 cases a day because it only brings in $2,000 and she isn't going to make her salary with that. Wow. I mean, there's so many things that are in the way of just providing patients the right answer when they really need it and while we can. So I think the best way today, this day in the middle of January in 2026 is to help us build a diagnostic center. What I say when I give a talk now is I say, who's an expert and what? And as we evolve in this, can we call on you? Will you be the expert in this? So this is brand new really weeks, but I think we're overcoming the structural problems that prevented us from doing it for everybody. And that is the goal. Could we, are we taking care of the patients here at our institution? Yeah, better than most because they're getting our interpretation. The trouble is there are medical institutions in Houston, a city that is the fourth largest city in America. They don't have this. Yeah, right. Yeah. It makes perfect sense. Yeah. The regional sort of diagnostic center concept makes it to where many of the institutional hurdles just go away because you just kind of plug into this process. Yeah. I just, we're getting kind of long on time and I didn't mean to take so long. I appreciate it. I did have one last question for you. To really define a DMT, a lot of people have come to us and said things and you've alluded to this before that they in fact already do a DMT, but they're not maybe sure or we're not sure that they actually have a DMT. So what would you say to somebody who thinks they might have something on their hands, but they're not sure? We've seen that when we had our three national DMT meetings here in Galveston. Yeah. And it was easy. You just go in and say, what are you trying to do? And then ask the questions, is it timely? Who's doing the work as the expert? Are you considering the patient context or are you just looking at the number and saying it's high or low? So really, there are probably eight or nine things on a checklist. And if any of them are a no, you're not doing it. Yeah. It's not, it's not something you're doing. Yeah. It makes sense. Well, sir, thank you very much. I appreciate it. It was fun as always. Yeah. Thank you, Chris. Yes, sir.