Beyond the ABCs

Hey everybody, this is Sarah Cracker. I'm an emergency physician and intensiveist at UCLA, and this is the ICUEDU podcast. First I wanted to mention that there is a new version of the reframing shock part 2 talk where we talk about the three pressures model of shock physiology, and happily the new and improved version is actually shorter than the old version which can only be a good thing. So take a look at that if you haven't heard the new version, but today we are going to talk about something very different. Today we are going to talk about approach to clinical decision making and iterative hypothesis testing. So this came up for me the other day when I was doing some cases with some residents, and I give them the case stem, and it kind of seems like no matter what case stem I give them, everybody reflexively says, "ABCDE" and/or "IVO2 monitor." And that's fine. I mean that's what we're taught to do. That's what we're taught to do for the oral boards, and that's great. But I got to thinking that why do I hate it when they do that? And I think the reason that I hate it when residents reflexively do that is because there's no thinking involved. And in the same way that you know ACLS and ATLS, I see as a place to start, but not a place to end, I see this sort of check the ABCs on every patient, and IVO2 monitor on every patient. Yeah, that's a place to start, and when you're a medical student, when you're an intern, and a sick patient comes in and you freeze, yeah, it's good to have, okay, I just start with the ABCs. But that's a place to start, not a place to end. Now I think what we get right about ACLS, ATLS, and teaching people like ABCD is the need to be systematic. But I think what we get wrong, especially when we're trying to train people as resuscitationists at a more advanced level, what we get wrong is conflating systematic with algorithmic, because you can be systematic without having to say, ABCDE, IVO2 monitor every single time on every single case. So I think that when we're doing a challenging complicated sick patient, just because the patient's sick, just because we have to think better and faster and do all of this under pressure, to me, isn't an excuse to switch from thinking to algorithms. So this is going to be about how do you do a systematic approach to assessing the critically ill, acutely ill, decompensating patient without relying on an algorithmic approach that applies to all patients and rather practice a thought pattern rather than reciting some letters of the alphabet in order. One of the big issues, and I think one of the themes that comes up, when I think about how I approach these patients, as a signal versus noise issue, because there's a whole bunch of noise going on in the background. You have a bunch of data, a bunch of abnormal vital signs, a bunch of labs coming back, and then you also have literal noise where you have a bunch of people screaming in the background. In fact, you may have somebody screaming, what are the ABCs in the background literally? But how do you block out all the noise literally and figuratively, and be able to practice a thought pattern that lets you lift out of that noise the relevant signal to be thoughtful about how you're addressing this patient? So I'm going to break this up into one of the initial assessment phase and two, the iterative hypothesis testing process. And I think this applies to patients, you know, when they first present to the ED, but I think it also applies to patients when I'm called as a consultant from the ICU, maybe when I'm first seeing the patient in the ED, maybe when I get called to the floor, and there's a patient decompensating on the floor. So I think this overall strategy to me applies to any acutely sick patient that I'm being asked to evaluate. So let's start with initial assessment. For the initial assessment of one of these patients, at a higher level where yes, I've learned the ABCs, let's take it beyond a level really, step number one for my assistant assessment is just assess and address immediate life threats. Do I think this patient has a imminent circulatory or respiratory collapse? And if so, I need to work on that. But the problem when we talk about the ABCs, we talk about that like we can act on them immediately without getting any further information. And sometimes we can, but sometimes we can't. Sometimes ABCs in the right order, but sometimes it isn't. And so as a more sort of experienced higher level assessment, do I think that this person has such imminent circulatory or respiratory collapse that I just need to act first and think later right this second? That's the case sometimes, but not very often. Like most things, much of the time, not all the time, but much of the time you have at least a minute, two minutes, five minutes to just take a breath and think. Step number two, interpret your initial data. And to me, what this means is context and vitals. So this is just your first past contact. So what's the context for this patient? What is their age? Are they in the range of 90 or in the range of 19? How's your eyeball test? Are they getting like a B minus or a D plus on the eyeball test? And what is the context for this presentation? And then vitals. So vitals, like yes, of course, you're supposed to look at them, but interpret them. Don't just look at them, you know, like, okay, this patient is hypotensive, but they're hard. It's only 67. They're inappropriately created. Cardiac for that degree of hypotension. Or wow, this patient is really, really hyper intensive. Is that the sort of chicken egg? Is that the driver? Is that a secondary effect? Because they're so sympathetically activated. So start to interpret your vital signs and use them to start to form some initial hypotheses. Then you need to initiate your initial management. Now this is really happening in parallel to interpreting your initial vital signs. And this right here is really an exercise in anticipating and prioritizing resource utilization. Because, you know, our categories of IVO2 monitor are not wrong. They just don't apply literally in that order, necessarily to every single patient. So as I'm initiating initial management, I am trying to sort of anticipate what I think is going to go wrong. If my initial interpretation of their initial sort of context in vital sign data is like, okay, I think we're about to have a respiratory problem here. Then I'm probably prioritizing figuring out my oxygen support. Am I sort of calling for high phlonial cannula? Am I causing for bypass? Am I having them set up for intubation? Versus if I think they're decompensating from a hemodynamic standpoint. And I'm like, we got to get some really good IVs. One is not enough. And then in terms of hemodynamic monitoring, I mean, depending on the patient, am I like, okay, I want that blood pressure right now versus, no, no, can you please get them on telemetry versus that SPO2 waveform is not good. Can we double check that that's right? And then finally, vascular access and the kind of vascular access I want again depends on, okay, is this patient in front of me shooting blood out of their femoral artery versus is this patient in front of me old sick volume overloaded. And I think I probably just need to put them on bypass and maybe give them a little slug of lasix. So my initial management is happening in parallel to my initial interpretation of my vital signs and context, but I'm prioritizing it based on what the anticipated course thus far is. And that allows me to prioritize if I don't have 17 nurses, it was just me and one nurse. Step number next, okay, now that I have assessed and addressed immediate life threats, and I've sort of interpreted my initial data and initiated initial management based on that data. Step three, now I get to take a breath and gather some additional patient data. And really my focus here is soliciting information if any at this point that I need to guide immediate management next steps. For example, if the most salient thing right now is how hypotensive my patient is, okay, my next management step is do I give this patient IV fluids or not? So my next number of questions are going to be focused on trying to elicit. Do I think that this is hypervolumic shock, distributive shock when I should give a bolus or do I think this might be cardiogenic shock in which case that would be a bad idea with my backup option if I have no idea, start pressers and then figure it out. So my initial data gathering strategy is really focused on guiding my immediate management next steps. Now in terms of information sources, obviously it can be the patient, sometimes if they're in such a condition they can tell you anything, it can be the family, but utilize your EMS people, utilize your paramedics, utilize your bedside RN and often utilize the chart. I think that we sort of often do this as a last step, but I have made the mistake so many times of you know I'm assessing a patient for hypotension and I'm getting called to the bedside and then it turns out this morning they had a CT scan that showed they were bleeding into their abdomen and nobody followed up that result or whatever it is. So don't under utilize the chart in real time it can give you some very very useful information. And your information gathering strategy because you have a lot of time here right, this is a sick decompensating patient, this is not the time to be like let me go and order the way I was taught in medical school, what is the quality of your pain, what are exacerbating and relieving factors and go all the way through down through family history, social history, the name of all of their pets, that is not going to work. You need to have a information gathering strategy. And to me that information gathering strategy is sort of three steps one, what is the real reason they're here? Like how did they actually end up here? In a patient in the ED that's often what is their actual chief complaint? Like when they say dizziness do they mean light headedness or vertigo? When they say chest pain but they're kind of pointed to their abdomen are they really having chest pain or is it abdominal pain? Now sometimes if the patient is not in shape to tell you this it's the nurse concerned. Like if you're being called to the bedside in the middle of the night on a patient on the floor, why did the nurse actually call you? You know, was it that they were desatting or was it just that the patient didn't look so good and yeah, they were desatting but then they couldn't get a blood pressure or if EMS is there and they're bringing the patient in from home and the patient can't give you history, what actually happened at home? And this information is very useful very early because it's what's going to allow you to focus and hone in on your next set of questions. Because once you have the real reason for presentation, is it vertigo, is it hypotension, is it abdominal pain, whatever it is, the next questions are trying to hone in on your initial differential diagnosis. So if their true chief complaint is vertigo, the questions that I'm going to use to hone in on my true differential diagnosis are all about that differential diagnosis and those questions to differentiate in this example, central from peripheral vertigo. So I don't know what questions to prioritize till I have that actual chief complaint and my next subset of questions are all about honing in on that initial key differential diagnosis and then once I have that, okay, central versus peripheral vertigo, now I can hone in on the precise questions for my history and findings on my physical exam to try and help me dictate my next management steps. Now when I'm doing this exercise with residents and even for myself as a sort of thought exercise to practice the right thought patterns and prioritization, I make myself sort of say, okay, if I could only ask this patient three questions and only do three exam findings, what would they be and why? Now obviously in a real patient, I can ask them as many things as I want or I can until they further decompensate or the next trauma comes in. So I think it's a useful exercise when you're practicing this thought strategy. Alright, so that's step number three, gathering some additional patient data. Now based on that patient data that you just gathered, step number four is now you're going to gather some initial lab and imaging data. And I find it helpful to break this lab and imaging data down into categories. And to me there's three different categories of data that I'm going to gather. One, hypothesis directed data, two, what I call logistic data and three, what I call phishing expedition data. So hypothesis directed data. This is the data that you're going to prioritize and this is the data that you get based on what your differential diagnosis is at this point for the patient. And you're going to prioritize data with management pathway commitment implications. For example, you know, am I going to call the cath lab? Okay, I really need to get an ABG for that. You know, am I going to give Heparin? Well, I'm worried about their brain. Do I need to get a CT scan? And so things that have immediate management pathway commitment implications like, you know, cattle, cath lab, call consulting service, give TPA, and debate the patient. That's what you're going to prioritize. But overall your hypothesis directed data is going to help you sort of support or refute your current leading hypotheses as to their differential diagnosis. But we work in systems. And you know, it's not just about you and the patient, it's about the system you work in. So this is what I call the logistic data part. So for example, I know perfectly well that I don't need a white count to diagnose appendicitis. But I know that if I'm going to call a surgeon later to be like, I think this patient has appendicitis, they're going to want that piece of data. So I'm going to anticipate the subsequent clinical course. And it's okay to gather logistic data that you don't really need for your hypothesis direct date exercise, but you know that you're going to need logistically for the next steps down the road. And last, my phishing expedition data. So this data isn't really hypothesis directed. Sometimes you just go on a phishing expedition and you just like pan-send everything. Now I wouldn't recommend this strategy in general because the problem with a phishing expedition is that you're now going to get a whole bunch of additional data. And the best way to hide useful data is to put it in the bunch of wholly useless data. So I think a phishing expedition overall is not a great strategy. Not being said, if I have a really, really sick patient and I am really just not sure after my essential assessment, what are my initial hypotheses? Do I have a strong initial hypothesis? If I have a sick patient and a weak initial hypothesis, I might go on a phishing expedition. Because the sicker the patient is, the higher pre-test probability that I have for finding a true positive rather than a false positive in this patient, the less time I have to figure it out. And if I'm really uncertain, sometimes I'll just go on a phishing expedition and do things like, I'm just going to send every lab I can think of, I'm going to pan-scan the patient. So again, I wouldn't recommend that as a general strategy. But if my patient is really, really sick and based on my initial assessment, I still don't really have a good hypothesis as to what's going on. Sometimes I'll go on an early phishing expedition. Now, in terms of gathering that initial lab and imaging data, as a thought exercise practicing the right thought patterns, I like to do the same thing. When I'm doing this with my residents or with myself to practice this, I like to say, okay, you can only get three tests, what are they, and why? Now, of course, in real life, you can get a million tests if you feel like it, but as a thought exercise, three tests, and why. Useful to practice the right thought patterns. And then the last step, so you've got your initial patient data, you've got your initial lab and imaging data, and now finally, you're going to summarize. This part is really, really important. You're going to take a pause, take a mental break, and say, okay, I'm going to summarize for both myself and out loud for my team where we are now. And my initial summary consists of the context, the patient's age, their sort of overall protoplasm, the acuity of what's going on, and why they're here. Then I'm going to follow that with what is my current working hypothesis. And then my management thus far has been x, and their response to said management has been y, and then finally concluding with some next steps in the form of if then statements that should include your reassessment plan. And this initial summary, I think, is really important. One for your team so that you can make sure everybody's on the same page, but two, forcing myself to do that, I think is a really useful thing to do, because it's like anything else. Something that seems like it makes sense in your brain, when you force yourself to say it out loud, sometimes you're like, that doesn't make any sense at all, that's a really dumb idea. Or even just talking it through can be very helpful. So following with that initial summary for both yourself and to make sure that you and your team are all on the same page. So that's my initial assessment process for these patients. You start with assessing and addressing immediate life threats. Then two, you go on to interpreting initial data and initiating that initial critical management. Then three, you're going to gather your additional patient data from the history and physical. And then gather your initial lab and imaging data, bring it all together finally with your initial summary. After you've done this initial assessment, now we're going to go into iterative hypothesis testing mode. And iterative hypothesis testing mode, the iterative part means that you might go through these next set of steps, ones or twice or many times until you fix the patient. So this is my iterative hypothesis testing next steps. All right. So we've summarized what's currently going on with the patient and our reassessment plan. All right. Let's go back and now let's reassess. So one, you're going to reassess for the development of any immediate life threats. And that's either can be from progression of whatever previous process they had going on, like they were in respiratory distress. And initially they weren't on the brink of imminent respiratory collapse, but now they kind of are so maybe we should intubate them. So did their previous process progress or did they develop a new process? Maybe my intern put in a central line and in punctured the long and now we have attention in both our ex and now we've developed a new immediate life threat. So step one of iterative hypothesis testing reassess for the development of immediate life threats, whether there are progression or new. Two. And this is the fun part, or at least what I see as the fun part, because step two is now you're going to go back and you're going to revise your working hypothesis based on new data. Because hopefully by now you've gotten back some of that lab and imaging data that you send out during your initial assessment. And I again like to put this data in categories, because often what happens in these super sick patients, you get this pile of data back and you don't know what to do with it. And so I find it helpful to put this new data that I'm getting into categories. Everyone does my new data support my working hypothesis. That's one possibility. Does this data refute my working hypothesis? Then you can also have data that's just a red herring, right? Like if I everything else about this patient's presentation, it's like it looks like a duck, walks like a duck, talks like a duck, and then we have a photo bar my bedillo. Okay, if everything is pointing to duck and then an armadillo sticks their head in the mix, it's probably still the duck. So sometimes it's just a red herring and you need to learn to sort of select those red herring pieces of data out and identify them. The really useful data is sometimes what I call the new lens data. Because sometimes you have all of this initial data and you're like, I don't quite know how to make sense of this. Are they septic? What's going on? Is the heart the problem? And then sometimes you get a piece of data back that shows you all the previous data through a new lens. So your new lens data is basically actually suggesting a new hypothesis by revealing an alternative interpretation. of previous data. And it's not all that often that happens, but when it does, if you miss that new lens data, then you often really miss some big things. So sometimes there's new lens data. And sometimes you have to experiment with multiple different lenses. You know, if you're like, "Oh, yeah, "I see this data through the lens of, "this patient has a cute pancreatitis, "but wait a minute, maybe those LFTs "are actually because the RV is failing, new lens." And then last category of new data, sometimes they've just developed a new problem, right? I mean, maybe you were doing kind of okay, you're resuscitating them, but now they're in real failure. Or now all of a sudden, they have developed a metabolic acidosis. Or again, maybe they've gotten attention to a new month or X from your central line placement. So some of your data may relate to a new problem. So I'm gonna revise my working hypothesis based on my new data. Some of that data may support my working hypothesis. Other data may refute it. Then there might be some red-haired data thrown in there. And valuable new lens data, if you can see it that way, another data might just be telling me a new problem. All right, step three. Based on my hypothesis revision, I'm now gonna gather some additional data. And this is back to our initial data gathering strategy. Again, we have our hypothesis directed data, our logistic data, and our phishing expedition data. So my hypothesis directed data, at this point is data to definitively confirm or maybe refine my working hypothesis. Alternatively, if I need to start completely over due to the new data having refuted or maybe to suggest it a new lens, maybe I just need to start all over and gather some additional data. Logistic data. This can be a very important thing at this point. So maybe to this point, I've gotten that, okay, this actually is something surgical. And now I've confirmed the hypothesis that my patient has a bowel perforation. Okay, now I need some additional logistic data. Now that I know that they need to go to the OR, I need to order some coax. I need to order a type in screen. So this can be an important time to order logistic data. And then phishing expedition. Because if you're in your first round of iterative hypothesis testing, and your patient's not responding to current management, you're really not feeling like you're honing in on a hypothesis. I start to be saying, okay, maybe I have a lower threshold for going on a phishing expedition. And again, my willingness to go on that phishing expedition is guided by the sum of how sick the patient is, how far am I progressing towards actually confirming my working hypothesis, and is my patient actually getting better with current management? The later the cycle, the lower the threshold I have, because I don't want to miss my window in a deteriorating patient to figure out what's going on. So sometimes if I don't have some kind of sense of what's going on at this one, I'll go on a phishing expedition data. And then lastly, when I'm gathering additional data, okay, I need just to gather reassessment data. Do I need to repeat EKG? Do I need to repeat, focus them? Do I need to repeat glucose, ABG, whatever it is? Do I need to get some reassessment data? All right, so now based on where I am in terms of my current hypothesis revision process, I'm gonna go to step number four, which is adjusting management. So after I've integrated all of my new data, refined my hypothesis, I reassess where am I now? Do I need to adjust my current treatments? Do I need to adjust my vent settings or antibiotics? Do I need to adjust my IV fluids? Do I need to adjust my vasoactive trips? Or do I need to backpedal? Am I like, "Nh, I should have not given them all those fluids, "maybe I need to give less X?" Do I need to additionally new treatments because I didn't realize they had a bowel perforation, so I haven't even given antibiotics. And then I need to adjust my management next steps logistically. So they're going to the OR. I need to make them NPO. Or actually know they're renal failurely very profound. And they didn't respond to the interventions I used to try and get them peeing. So maybe I need to place a dialysis catheter, whatever it is. So now I'm gonna adjust management. And then finally, my last step in my iterative hypothesis testing, I'm gonna summarize again. And I'm gonna summarize again, both for myself and for my team. I'm gonna recapitulate the context, age, protoplasm, acuity, and reason for initial presentation. Then I'm gonna give my current working hypothesis and specify if it's changed from previous, I'm gonna update the management thus far in response, then my next steps with if-then statements and my reassessment versus dispo plan. And doing this for myself and for my team, especially emphasizing changes in direction. If we thought, oh yeah, definitely septic shock, but then based on our iterative hypothesis testing, we've had to revise our working hypothesis to just kidding, they are actually in cardiogenic shock or have a massive PE, we need to emphasize that in our iterative summary. So our iterative hypothesis testing process, we're going to reassess for the development of life threats, revise our working hypothesis based on new data, gather additional data, adjust our management accordingly, and then summarize, and then we rinse and repeat. So that is me trying to break down and back calculate my approach to clinical decision making. And I know it sounds a lot more complicated than AVC or IVO2 monitor, but I have found it useful to try and be explicit about what I'm actually doing at the bedside and to practice a systematic thought pattern because to me, I found that I'm less likely to miss things if I'm systematic but not algorithmic. Thanks for listening.