I'm Jake Jenser from Mayo Clinic. I am a cardiac intensivist by training, and I'm here to talk to you about mortality risk stratification in the cardiac ICU, including SOFA, Apache, and I'll try to convince you the importance of the SkyShock classification for this purpose. So myself and others have clearly demonstrated that the same mortality risk scores, severity of illness risk scores that are used in other ICU populations are very effective for mortality risk stratification in the cardiac ICU. This includes the SOFA score, which was designed for sepsis, the various permutations of the Apache score, as well as the OASIS, which is a simplified Apache derivative. And in general, these all perform very well in unselected cardiac ICU cohorts, but they don't work very well for cardiogenic shock populations specifically. They generally have poor discrimination for mortality. In particular, they are very ineffective for identifying low-risk patients. And one of the reasons is that shock severity is not well characterized in Apache and OASIS, although it is included in SOFA. So as an example from my database, I took patients and just separated them based on their SOFA score, and I looked at the cardiogenic shock patients and the patients without. And the patients without cardiogenic shock, the blue line, you can see a very dramatic difference in mortality between the low-risk patients and the high-risk patients. This has got to be a 30-fold difference. But when you look at the patients with cardiogenic shock in orange, the high-risk and low-risk patients, it's really more of a three-fold difference. And so this suggests that there's room for improvement for this purpose. Why do we care about risk stratification? Well, there are many reasons. The first that I always think about is triage and transfer. If you have a patient who you know is very high-risk or very low-risk, that can give you insights about what to do with them. If they're very low-risk, they may not necessarily need to be in the ICU, or at least they might be a candidate for early ICU discharge. If they're very high-risk, that can inform prognostic discussions and help you think about where they might best be served. You can look at outcomes, particularly benchmarking your outcomes between different centers and over time. So you can look at the observed-to-expected mortality, for example, and see how that trends. But the real goal is to match the intensity of your therapy to the severity of illness. Sicker people maybe get more therapy. But when we look at generic severity of illness scores, that doesn't really work because the sickest patients, the ones who are most likely to die, often have non-modifiable risk factors and more aggressive therapy won't necessarily change their prognosis. So what you really need is a disease-specific severity score where you can talk about the intensity of therapy for that disease and match that to the severity of that disease. For example, in patients with shock, those who have more severe shock might be able to get more intensive therapy for shock. And that is what the SKAI classification can do for us. As Dr. Barron had already explained, patients who have acute cardiovascular disease are at risk of cardiogenic shock but don't yet have it. Those who are beginning to develop hemodynamic instability but do not yet have hypoperfusion do not yet have shock, but they're beginning to develop shock. Patients who have hypoperfusion that requires an intervention, they have shock, and we can grade them in C, which is the initial stage where they need something beyond fluids to help reverse their hypoperfusion, to stage D where that initial intervention is failing, to stage E or extremis, which are patients with refractory shock and circulatory collapse. So again, pre-shock is stage B. Shock is C, D, and E, escalating severity. Now when I try to think about assigning this, you can see how complicated it can get. The criteria that I've used in my papers, there's like a dozen criteria that help define this. You're not going to do that at the bedside, although you could use an electronic medical record. So let's simplify this. The first question to assign SKAI stage is, does the patient have hypoperfusion? And what you do is you use the data that's available to you at that moment. This could be clinical signs, delayed capillary refill model extremities, or this could be laboratory findings, elevated lactate's an obvious one. And then if the patient does not have any evidence of hypoperfusion, then you can say they don't have shock. And the question is, do they have pre-shock? And so you go by, do they have hemodynamic instability? This extends beyond just vital sign abnormalities. For example, a patient who maybe needs to be on a little, you know, some vasoactive drugs but did not have hypoperfusion, or someone who has poor hemodynamics might be classified as pre-shock. On the other hand, once patients have hypoperfusion, they have shock, and the question is, how bad is it? Realizing that, yes, there are stages, but there are gradations within these stages. So if the patient needs an initial therapy, vasoactive drugs, mechanical circulatory support, you call them stage C, recognizing that not everyone in stage C is actually hypotensive. If that first therapy fails and the patient either doesn't get better or actually gets worse, then you say they're stage D. They have worsening shock. Usually this is your rising lactate or escalating vasopressor requirements. The patient in stage E is actually the hardest to define for research, but the easiest to define clinically. This is the patient with the crash cart outside their room. You can recognize them. Everybody knows what's going on. They're maybe getting bolus vasopressors. They're pericode. They're on high doses, multiple drips. And this is a patient who has, you know, impending circulatory collapse. Now, in research, there are at least six or seven definitions floating around in the literature, and all of them work really, really well. This is a summary of all the different SkyShox validation studies that look at mortality as a function of the SkyShox classification, and you can see this impressive stepwise increase in mortality as the SkyShox stage increases. All the way on the left, you see cardiac ICU cohorts where this works really, really well. In the middle, you see cardiogenic shock cohorts where this works very well, but you can also see that it works in out-of-hospital cardiac arrest patients and patients with sepsis over on the right. So, it's very clear that the SkyShox classification is effective for predicting mortality, but we know that as shock gets worse, so do other things. Patients in higher SkyShox stages are sicker. They have higher severity of illness scores. And so, on the left, this is the Apache 4 predicted mortality. It goes up very dramatically with the SkyShox stage, and of course, so does the mortality. Likewise, the presence of non-cardiovascular organ failure or the non-cardiovascular SOFA score goes up. But, you know, if I'm just telling you that people with shock are sicker people, how does that help you? Well, it turns out that if you adjust statistically that these are two separate independent variables. They're both strongly associated with mortality, even when they're considered together. And here's a visual representation. So when we take patients and break them down by SkyShox stage and then look at Apache scores, patients with higher Apache scores are more likely to die, but so are patients with higher SkyStage. And if you look at them together, you get this gradient in mortality. The same is true in cardiogenic shock patients when we're looking at organ failure defined by the SOFA, but specifically non-cardiovascular organ failure. More organ failure, worse outcomes. Worse shock, worse outcomes. And again, they create this marked gradient. So in addition to the separations between the SkyShox stages, I really want to reemphasize that gradations of severity exist. In my studies, I've used different criteria to define hypotension and hypoperfusion. And the more criteria you have, the worse you are. It makes sense. The more hypotensive you are, the worse your prognosis might be. And that's clearly shown here on the left, where you can see this stepwise increase in mortality as either hypotension or hypoperfusion or particularly together. The vasoactive drug requirements, that's a very simple way that people have been defining shock severity for many years, and again, it integrates with the SkyShox classification. If you need more vasoactive drugs, you're more likely to die regardless of your SkyShox stage. So this gets me to this concept of risk modifiers. And essentially, these are independent risk factors for mortality that exist across the spectrum of shock severity. So regardless of how severe your shock is, these are things that are going to predict that you're more or less likely to have a good outcome. The two most established ones are cardiac arrest with coma. So patients who maybe had a brief event, you know, a VF, shocked once, totally awake, not sure that matters. But the patients that are comatose, who are going to end up going on to have an oxyc brain injury or post-cardiac arrest syndrome, that really modifies their prognosis regardless of how much shock they have. And of course, older patients we know are always at higher risk of adverse outcomes because of all the multitude of factors that go along with aging. And so I'm going to show you those data here. So these are patients with cardiogenic shock. And you can see this dramatic difference in the prognosis between younger patients in the front blue bars, who have less severe shock on the left side of the figure, versus older patients, the orange bars in the back, with more severe shock on the right side of the picture. But there are many other well-established and emerging risk factors. Some of the ones that we think are best established are the presence of right ventricular failure. And that can be defined in multiple different ways. Poor systemic hemodynamics, again, lots of different markers that have been integrated for that. Patients with systemic inflammation, again, defined in multiple ways. This concept of hemometabolic shock, which I'll get to in a little bit, which is a bit nebulous, but really emphasizes that multi-organ failure and severe acidosis are bad things. And then patients with worsening shock. You saw Dr. Barron's data looking at patients whose sky shock stage increased over time, for example. Now, when we think about phenotypes of cardiogenic shock, these are ways to define different groups of patients who are similar to each other, but different from others. There's lots of ways to do this. A very basic and simplistic way is to look at their pattern of ventricular function on echocardiography. So we look at patients with moderate or greater ventricular dysfunction. Is it single ventricle, biventricular? And you can see this stepwise increase in mortality in each sky shock stage as the degree of ventricular dysfunction increases. In particular, patients with biventricular and RV dysfunction do worse. But there's other ways to define phenotypes. For example, based on biomarkers. And the Cardiogenic Shock Working Group did this paper, which I think is a really important paper, and they took admission laboratory values, just the basic stuff that we all get for all the patients. And they used unsupervised machine learning to cluster them based on the six most important prognostic variables. And they found three different groups within the cardiogenic shock population. And these groups differed not only in their biomarker profiles, but more importantly, in their clinical profiles as well as their hemodynamics. And so they found a group that they called the non-congested group, which had fairly limited hemodynamic compromise. They were younger. They were less sick. Didn't really have much organ dysfunction. There's another group that was very, very different. They called the Cardiorenal Group. These patients were older with more comorbidities. Their hemodynamics were not particularly bad either, although they did have left-sided congestion. But importantly, they had very severe renal dysfunction. And then the Hemometabolic Group, which is patients who have very severe hemodynamic compromise, right-sided congestion, and multi-organ dysfunction with severe acidosis. And so based on my descriptions, you know these groups are going to have different outcomes and different prognosis. And this is what it looks like. But more importantly, in each SkyShock stage, each of these different phenotypes had a different prognosis. These are not just sick people, not sick people. These are different subgroups that can be identified within the cardiogenic shock population. Are these patients different stages of the same disease or different disease trajectories? It's hard to know at this point, but this is, I think, an important avenue for research. So when we put this all together, we create this three-axis model of cardiogenic shock, which is in the new SkyShock classification statement. So this can be a little complicated, but I'm going to try to break it down. Essentially, the concept is there are three important domains when you are considering a cardiogenic shock patient, either for prognostication or for clinical decision-making. And there are multiple components in each domain, and there's a gradient of severity in each domain, but each of the domains is distinct. So the first one, the most obvious, is shock severity. So again, this is the SkyShock stage, but also these other things that I've alluded to, vasopressor requirements, hemodynamics, hypoperfusion, response to therapy, these all are integrated together to get an integrated concept of how bad is this person's shock. Then there's the phenotype. And I alluded to a couple things, like their ventricular function or their biomarker profile, but also the etiology and the acuity. These are things that are different. They're very distinct from the shock severity. They may interact, but they are not the same. And then finally, these risk modifiers. These are your non-modifiable risk factors for adverse outcomes, essentially saying, is this patient, pound for pound, at the same level of shock severity, going to do better or worse? And so I alluded to cardiac arrest with brain injury. I alluded to age and several others. And so you can almost imagine this as a three-dimensional space with three axes, an X, Y, and Z axis. And the farther the patient is from the origin, the sicker they are and the more likely they are to die. But what really matters is clinical decision making. So yes, I'm an outcomes researcher, but I really am a nearly full-time clinician, and that's what I care about. How am I going to actually use this model to take care of my patients? Well, the answer is, each of these domains points you in a certain direction for patient care with shock severity. This helps you decide if the patient has severe shock that might need mechanical circulatory support. If a patient has mild shock and they're clinically doing well and they're responding well to low doses of vasoactive drugs, it's not clear to me that escalating to mechanical circulatory support is always necessary. Additionally, once you've decided that mechanical circulatory support is appropriate, what degree? You know, mechanical circulatory support comes in a wide variety of flavors ranging from relatively low-risk, weak interventions like the balloon pump to high-grade interventions with potential high-risk of complications like ECMO. The phenotype helps you decide what type of device might be matched to your patient. A patient with isolated LV failure might do well with a percutaneous LVAD, but a person who has biventricular failure won't, particularly if they also have lung disease, in which case they may need ECMO. And finally, these risk modifiers tell us, is this patient going to die whether or not we fix their shock? If I have a patient who's going to die of anoxic brain injury, I'm not sure that it matters how well I treat their shock if their death is inevitable because of a non-recoverable brain injury, for example. Or another way to look at this is, if I salvage this patient but they don't totally recover, will they eventually go on to LVAD or transplant? This is crucial because we know that a substantial percentage of patients with cardiogenic shock end up needing these advanced therapies because of cardiac non-recovery. And these risk modifiers are going to help you pin that down to some extent. So you can think about all these things separately but integrating them together. So in summary, the SkyShock classification grades the severity of shock in a straightforward manner. You don't need all these fancy criteria. You can do it based on your gestalt. If you want to do it in a more systematic way, at your institution, you can agree upon what criteria you use. You may agree that the lactate has to be some number to count. That's not hard and you can tailor it to your needs. There are gradations in shock severity. We make it sound like it's these different stages that are totally discrete. They blur together both in terms of the severity of shock and the prognosis. But you can do this at bedside. You can teach your EMS crews to do this. You can teach your ER docs to do this. You can teach your interns to do this. This is not hard. Importantly, shock severity is a very, very powerful predictor of mortality. And it is complementary to and independent from standard predictors that are in the well-known risk scores. And you can integrate the SkyShock classification with many, many, many, many different prognostic variables to create this nuanced prognosis for your patient. Although sometimes, honestly, that's a bit more complicated than what I'm personally willing to do at bedside. But this three-axis model has value because it allows you to put these different components into three simple bins, each of which translates directly into a way to do clinical decision-making in terms of, does this patient need support? What type of support might be appropriate? And is this patient really a candidate for all that? So I thank you very much for your attention. And I'm really excited to hear our final speaker.

mortality risk stratification

cardiac ICU

severity of illness risk scores

cardiogenic shock populations

SkyShock classification