Transcript Video From Coma to Clarity: The Art and Science of Neuroprognostication Hi everyone, um, thank you for that, uh, very kind introduction. Um, so as she said, I'm, I'm David I'm a neuro intensivist, um, and my, um, my interest is in disorders of consciousness and neuroprognostication. Uh, I love talking about it and BD wanted me to talk about it, so that's what I'm doing, um, uh, so I'm gonna go through, um, these are my financial disclosures, let's see if this will work. OK, these are my financial disclosures. Um, I'm gonna talk a little bit about some of the challenges that we're currently facing in neuroprognostication, and I'm gonna talk about a systematic framework for approaching this problem. And then I'm gonna in that context gonna talk about you know the state of the literature and the guidelines around neuroprognostication. So to be clear, when I'm talking about disorders of consciousness, I'm not talking about any one specific ideology of brain injury. I'm talking about any sort of severe acute brain injury that results in this final common pathway of a disorder of consciousness and as many of you know, one of the biggest questions that arises in that context is that of neural prognostication. Are they gonna wake up again? Are they gonna regain acceptable neurologic function or not? And it's a really important determination because it often influences whether or not life sustaining treatments are continued or withdrawn. And yet as many of you also know it's also really hard um and this has been, um, shown in the literature as well this is one study that I think highlights the challenges particularly well. It was a study of over 4000 patients who survived their initial cardiac arrest. This is across 151 different hospitals. And the first thing that they showed is that if you look at the causes of death in the following days, far and away the most common cause of death represented by the purple bars was the withdrawal of life sustaining treatment due to a poor neurologic prognosis. Far more patients die because of that than die because of a poor medical prognosis, medical instability, etc. So there's a lot of mortality that's attributable to these neuroprognostication decisions. And yet we're also highly variable in how we make these decisions. If you compare any two of these 151 hospitals, the chances that life sustaining treatment would be withdrawn were about 60% different. And that's after controlling for many of the clinical variables that influences this decision. And you might worry that if we are so variable in how we're making these determinations, that we could be getting it wrong a certain portion of the time. And specifically might be withdrawing life sustaining treatment from patients who could have otherwise had an opportunity to recover. Now that's a difficult error to estimate because if we withdraw life support, these patients die, we don't get to know how they, how they would have done. But this study tried to estimate it, essentially by looking at the patients who had their life sustaining treatment continued, looking at the variables that predicted those outcomes, then extrapolating to the patients who had their life sustaining treatment withdrawn. And what they estimated was that if we hadn't if we hadn't withdrawn on life sustaining treatment, about 16% of them would have been able to walk independently by hospital discharge. Now this is not a perfect approach, but I think highlights the very real concern, which is that if we are so variable in how we're making these decisions that we might be getting it wrong a certain portion of the time and may be contributing to excess morbidity and mortality. I think one of the ways that we can try to help mitigate these risks is by trying to approach neuroprognostication in a little bit more of a systematic way. And I'm kind of a visual guy, so I have like a visual framework for how I think through neural prognostication, which I'm gonna share with you guys. So you can kind of imagine neurologic function on the Y axis and time on the X-axis. And if you think about how these severe acute brain injuries typically work, you have someone who's humming along at a certain degree of neurologic function. And then suddenly and unexpectedly, some acute brain injury happens, cardiac arrest, traumatic brain injury, etc. And suddenly there's this precipitous fall in their level of neurologic function. And often that kind of boils down to the question of, you know, you, you wanna know how far they've fallen on this curve which really kind of boils down to the question of how conscious are they now, right? And, and that makes sense if you want to know how much consciousness someone's likely to recover, figuring out how conscious they are now is a reasonable starting point. So how do we figure out how conscious a patient is now? Well, we think about consciousness in these two parts, right? There's arousal and or wakefulness and there's awareness, kind of the content of experience, and we think that awareness depends upon arousal and that you need to be awake in order to be aware of the world. We think that awareness depends more upon the coordinated activity of the cortex, whereas arousal depends on these, uh, nuclei in the brain stem. How do you know when someone is awake? Well, it's often as easy as seeing are their eyes open. And you may also potentially see sleep-wake cycles on EEG. How do you know if someone's aware? That's a little bit trickier, but essentially you're looking for purposeful behaviors. Are they following simple commands? Are they, uh, giving yes no responses? Are they verbalizing intelligibly reaching for objects or following a target with their eyes? These are the types of behaviors that we have conventionally felt reflect awareness. And these different components of consciousness give us our different diagnostic categories when it comes to disorders of consciousness. So there's the minimally conscious state where patients are awake, but only minimally or intermittently aware. There's the vegetative state where patients are awake but show no evidence of awareness, and then there's coma where patients are neither awake nor aware. And this has been kind of our conventional model for for many years now, but recently the question has been asked is behavior enough to evaluate a person's level of consciousness? And that's where some of these newer functional MRI or FMRI assessments come in. Now for those who aren't familiar with FMRI, this is a technique that uses the same MRI scanner that we normally use. And like our regular MRI's, it looks at the brain as a series of three-dimensional pixels or voxels. But instead of looking at signals within each of those voxels that tell us about the structure of the brain, we instead look at signals that tell us about the function of the brain. Now we don't measure that brain activity directly, we measure it indirectly through this phenomenon of neurovascular coupling. The idea is that as neurons become more active, we see a dilation in the afferent arterios, an influx of oxygenated hemoglobin relative to deoxygenated hemoglobin. And that's, uh, results in this increase in the blood oxygen level dependent signal or bolt signal. So in essence, an FMRI, you're measuring bolt signal as a proxy for brain activity. Now, classically the way we've used this is you put someone in the scanner, turn it on, and then you give them a task or command. Like you might say, imagine you're moving your hand, and if they do that, then you'll see this increase in bolt signal in the motor parts of the brain and it might give you a map that looks something like this. So in 2006 there was this seminal study where they took a, a healthy person, they put them in an FMRI scanner they asked them to either imagine they were playing tennis or imagine they're walking through their house. When they did the former they saw activity in the supplementary motor area. When they did the latter, they saw activation in the parietal cortex. And there was a patient who behaviorally was in a vegetative state, no purposeful behavior, who could willfully modulate their brain activity in the same way. And so this gave rise to this phenomenon that we call covert consciousness or cognitive motor dissociation. It turns out that it's not particularly rare, so there was this recent study that looked at over 300 patients across 6 different sites, and they used both FMRI as well as EEG to look for the willful modulation of brain activity in patients who are behaviorally unresponsive, and they found that 25% of them were able to show this evidence of covert consciousness or cognitive motor dissociation. So this is causing us to rethink a little bit how we categorize these disorders and move away from this kind of conventional model of a single dimensional hierarchy of behavior from coma to the minimally conscious state and we're trying to figure out as a field kind of how to categorize these disorders now. There's been a suggestion that maybe we should just have, you know, multiple axes for behavior or technology or maybe we should be collapsing behavior and technology into a single axis. We're kind of working through this. But kind of suffice it to say that you collect as much data as you can to evaluate a patient's current level of consciousness. And right now the convention is still behavior, but the hope is that as these technologies become more widespread, we might be able to supplement that with some of these techniques. Once you evaluate a patient's current level of consciousness, the next question is to figure out what their likely recovery trajectory is going to be, right? This is at the heart of neural prognostication. And this is where all of our tests come in, right? And there's lots of tests, there's a lot of data points. I could have put, I put them into 3 main buckets. So there are tests that tell us about the functional integrity of the brain. There are tests that tell us about the structural integrity of the brain, and then there's all the other data points that can influence a patient's recovery potential, like their age, pre-morbid function, comorbidities, and, and social supports. So I'm gonna go through each of these things, each of these tests, and talk about their prognostic value. We'll start by talking about the neurologic examination. I already talked about how behaviors can reflect the patient's current level of consciousness, but those same behaviors also have prognostic value. So patients who start off in a minimally conscious state have better long term survival than patients who start off in a vegetative state, and they're more likely to have better long term functional outcomes as, uh, indicated by the GOI score. It's important to note though that these disorders of consciousness as a rule fluctuate over time. Patients, these are dynamic states, patients can move between the vegetative state and minimally conscious state from hour to hour. And so if you just do a single neurologic assessment, you're you have a high risk of missing these signs of awareness and this is one study that showed that if you do just one assessment, there's about a 40% misdiagnosis rate, about a 40% risk that you're gonna underestimate that patient's level of consciousness, whereas the more assessments you do, the more likely you are to detect these sometimes fleeting signs of awareness. So kind of serial exams is really important, and this applies across all ideologies of brain injury. For the rest of these tests I'm gonna focus on three specific ideologies of brain injury hypoxic ischemic brain injury after cardiac arrest, traumatic brain injury, and intracerebral hemorrhage. And that's for a few reasons. Uh, one is that they're relatively common causes of disorders of consciousness and common reasons for neuroprognostication. And the second is that these are the three ideologies of brain injury for which we have guidelines from the neurocritical Care Society that kind of summarize the literature on what tests are considered valid in each of these scenarios. So I'm gonna focus in on these. And so kind of continuing along that line of uh thinking about the neurologic exam. We'll start with hypoxic ischemic brain injury. Kind of interestingly, the literature has evolved a little bit pre and post targeted temperature management or TTM. Before TTM, the guidelines suggested that myoclonic status, absent brain stem reflexes, or absent or extensor motor responses were all associated with the poor outcome. After the advent of TTM, uh, the, the literature changed it kind of evolved to say that absent brain stem reflexes greater than 72 hours after, uh, after arrest is specific but insensitive for a poor outcome, whereas absent or extensive motor responses are sensitive but not specific for a poor outcome and you know it's not clear if this is related to TTM per se or just to evolution in the research methods, but the way the neurocritical Care Society guidelines have boiled this down. It's to say that of all these exam features, what's probably most reliable is absent pupillary reflexes greater than 72 hours from Rosk. That's associated with a poor outcome, whereas localization or withdrawal greater than 72 hours from Rosk or rewarming is associated with a good outcome. Those are kind of the two neurologic exam features that were felt to be most reliable based on the literature. In traumatic brain injury, uh, many different, uh, clinical variables were looked at by some in this room. Thank you, Susannah. And of all of those. Different clinical variables. The only variable that was felt to have prognostic value was in fact an aspect of the neurologic exam, the absence of pupillary reflexes. So like in hypoxic ischemic brain injury, the absence of brain stem reflexes is felt to be the most reliable of all of these clinical variables that was assessed in predicting a poor outcome. In intracerebral hemorrhage, uh, the guidelines did this very kind of rigorous systematic evaluation of the literature and concluded that there were actually no clinical variables including aspects of the neurologic exam that reliably predicted outcomes. So, this is what the guidelines currently say about how the exam might be used in prognostication for these ideologies of brain injury. Returning back to our figure here. Our next step is, uh, EEG. In hypoxic ischemic brain injury, suppression and birth suppression, what's in other uh contexts considered uh highly malignant features. Uh, have been endorsed as being associated with a poor outcome. And just to point out, there are lots of other EEG features that have been studied in neuroprognostication for cardiac arrest, things like reactivity, for example, but those, um, those EEG elements were not felt to be sufficiently reliable to be endorsed by these clinical guidelines. And while similar elements of the EEG have been studied in prognostication from these other ideologies, uh, those were not felt to be sufficiently robust to be included in the guidelines either. What about somatosensory evoke potentials? So as a reminder, this is how somatosensory evoke potentials work. The technologist comes in and delivers stimulation to the bilateral median nerves, and then we watch that sensory signal ascend through the brachial plexus at Ub's point, through the cervical cord, and then ultimately to the somatosensory cortex. That signal at the somatosensory cortex is called an N20 response because it's a negative deflection 20 milliseconds after the stimulation. And so what we want to know is, is that sensory response, uh, retained or not? And to be clear, the, the absence of that response is only interpretable if the signal has an opportunity to make it to the brain. So if that N20 response is missing, but the signal was undetectable at URB's point or at the cervical cord, then the absence of the N20 response is meaningless because the brain never had a chance. So in hypoxic ischemic brain injury, bilaterally absent N20 responses, greater than 48 hours from Brosk has been associated with a poor outcome. I'll qualify that to say that I think there was a point historically where it was felt like this was like the perfect test, you know, if you had absent N20s and cardiac arrest that was like 100% specific for a poor outcome. It turns out it's not quite so perfect. There are some patients who, despite bilaterally absent N20 responses, can actually do well, but it's still very specific. Um, it's also worth pointing out that N20 amplitudes, greater than 4 microvolts, has actually been associated with a good outcome, and that's important to point out this is based on a relatively recent literature, but I think there's less awareness of this and I think in a lot of institutions we don't even report N20 amplitudes. So if you're at an institution where you're not seeing those amplitudes reported, it might be worth suggesting that that's included in the report. And while SSCPs have been studied in these other contexts, it wasn't felt to be reliable enough to be endorsed by these clinical guidelines. All right, so let's move over to our structural assessments. We'll start with the head CT. So interestingly, guidelines that um from the American Academy of Neurology prior to the advent of TTM felt like there were actually were too few studies at that point to uh endorse it for uh endorseET CT for prognostic use. But after TTM, there are more recent guidelines, uh, have felt like there are enough studies to to endorse it. So if you see diffuse loss of gray white differentiation and sul effacement greater than 48 hours from Brosk, uh, that's associated with a poor outcome. So if you have a head CT like this one all the way on the right, uh, that is pretty reliably associated with a poor outcome. I'll make a couple of points about that. The first is that, you know, this 48 hours piece matters, and that's because this injury can sometimes often takes a couple of days to show up. So those early head CTs that are collected in the emergency department, for example, are often very insensitive, even for very severe injury. Another point to make is that again if you see a head CT like the one on the right that's pretty reliable but if you get a head CT where the report is, you know, mild anoxic brain injury or you're squinting at it and you're like, is there a loss of gray white differentiation? just know that there is a very low integrator reliability for that level of injury. So, uh, be cautious in concluding anything about prognosis based on more mild forms of injury. In traumatic brain injury, head CT can be useful, but with, with the caveat that, uh, it only has been felt to be reliable in the context of these scores. So there are kind of these, uh, head CT scores where like isolated head CT scores that were not felt to be reliable, but these broader scores like the crash score and the impact score which can include elements from a CT have been felt to be at least moderately reliable. So this is the crash score. It includes things like age, Glasgow coma score, pupils, and then has elements from the CT scan, including the presence of particular hemorrhages, obliteration of the third ventricle or basal cisterns, subarachnoid bleeding, midline shift, or non-evacuated hematoma. The impact score is similar, you know, also includes things like aged pupils and from the CT includes things like traumatic subarachnoid hemorrhage or epidural mass on CT. So at least those CT elements within the context of these scores has been felt to be somewhat reliable in predicting recovery. And so if you have a head CT that looks like this one on the right where there's some subarachnoid blood, maybe some epidural blood, that patient might do worse than the patient on the left again within the context of this score. Intracerebral hemorrhage is interesting, uh, because on the one hand, it's kind of obvious, right, that the patient on the right is probably gonna do worse than the patient on the left. But it's precisely because that's so obvious that the literature hasn't really been felt to be that reliable. In other words, we are so much more likely to withdraw life sustaining treatment from the patient on the right that the literature on these uh on these CT aspects have been felt to be kind of too biased by this self fulfilling prophecy, uh, bias and so ultimately. These CT elements, whether it's hematoma volume or location, have not been felt to be reliable in predicting recovery from the guidelines. How about MRI? So again, like CT there was felt to be too few studies prior to TTM to, to really speak to what the prognostic value would be, but these recent guidelines have felt that MRI, there's enough literature now to support the use of MRI. And basically what they say is that an MRI within 2 to 7 days after Rosk, if you see diffuse diffusion restriction that's associated with a poor outcome, whereas if you see little to no diffusion restriction, that's associated with a good outcome. So these are DWI sequences. If you see a scan like the one on the right, that patient is likely to do worse than the patient, uh, on the left. Now this kind of 2 to 7 day window is interesting. And it kind of reflects the fact that we don't really understand where this diffusion restriction is coming from, right? It's not like acute ischemic stroke where the diffusion restriction occurs within minutes of the stroke for some reason this takes like a couple of days to show up. And then kind of like disappears again after like 5 to 7 days. So that's why, uh, this 2 to 7 day window matters because that's when you're most likely to pick up on the signal but again speaks to the fact that there's a lot more work to be done to understand what's actually causing the signal abnormality and maybe it's intervenable. In TBI there's been a lot of work to try to figure out if MRI can help predict outcomes. So for example, uh, MRI can help reveal evidence of diffuse axonal injury or DAI. There's grade 1 DAI where you have these lesions like microhemorrhages at the gray white junction in the cortex, grade 2 DAI where you see these lesions in the corpus callosum, and grade 3 DAI where you see these lesions in the brain stem. And there have been some studies kind of associating this with with prognosis, but ultimately it wasn't felt to be a robust enough literature to be included in the guidelines. And similarly for intracerebral hemorrhage, there are small studies that suggest that the degree of diffusion restriction around the hematoma might have prognostic relevance, but again those were small studies that were not felt to be reliable enough. What about serologic biomarkers like neurons specific emolase? So all three guidelines concluded that this was not reliable. Again, this is, uh, for those who don't know, neurons specific inolase it's a protein contained within neuronal cell bodies, and the idea is the more of these neurons that die, the more this protein gets spilled into the blood. So the higher the level, the worse the outcome. So none of these guidelines felt like this was reliable for these ideologies of brain injury. In the case of hypoxic ischemic brain injury, even though there are lots of studies that have shown this association between neurons specific inlas and outcome. One of the biggest problems was that all of these studies identified different thresholds for what level would distinguish a good outcome from a bad one and so on the basis of lacking that kind of uh clear threshold, they didn't felt that they didn't feel it was reliable enough but I think if you think about neurons specific inlays as a continuous measure rather than as a binary measure, there's a very strong literature to support that association. And it's worth pointing out that there are 2 other major guidelines for hypoxic ischemic brain injury from the American Heart Association and the European Resuscitation Council that do endorse collecting this. I'll say that locally we do collect it, you know, from a cost benefit standpoint this is way cheaper than any of the other tests we're talking about, and there are some occasions where when data are conflicting this can be uh an additional helpful data point. Um, there are lots of other serologic biomarkers that are in development, uh, so for example, neurofilament light chain, a protein contained within the axons, uh, may actually have better sensitivity and specificity than than neurons specific laces and predicting outcome. Uh, we have actually started collecting that, uh, locally, even though that is not included in the guidelines has not been particularly helpful for us because it takes forever to come back. But there are occasions where it might be helpful, so you know, neurons specific enolase, despite the name, is present in other cell types too, like red blood cells and platelets. And so patients who have hemolysis that can cause a false elevation in your neurons specific enolace. It's helpful to be aware of, particularly for patients on ECMO, for example, and in cases like that, sometimes the neurofilament light chain can help clarify whether it's a true elevation or a false positive. All right, so those are our conventional tests that are have been specifically, uh, endorsed by the neurocritical Care Society guidelines, but there are lots of other tests that are in development, including functional MRI which we talked about, which you can use in different contexts, right? We talked about task-based FMRI where you're looking for responses to command, but you can also look for stimulation-based FMRI where you're looking at the responses to language or resting state FMRI which we'll talk more about in a bit. There's pet that looks at the degree of uh cerebral metabolism. And there are more advanced ways of assessing EEG quantitative techniques and looking at the response of EEG to transcranial magnetic stimulation, for example. On the structural side, uh, there's a lot of work being done to try to quantitate our imaging a little bit more effectively, so looking at specific ratios to measure gray white differentiation in our, uh, CT scans for hypoxic ischemic brain injury patients, we're looking at specific ADC thresholds for patients on MRI for patients with hypoxic ischemic brain injury. There's also, uh, new, uh, more advanced sequences like uh diffusion MRI like diffusion tensor imaging MRI that can look at more detail at the structural integrity of the white matter in these patients. And as I mentioned, there's a lot of work being done to identify more effective serologic biomarkers. Our fields in a little bit of a gray area right now with respect to these new technologies because on the one hand there's a growing literature to suggest that a lot of these emerging techniques are pretty effective in predicting the outcome in some cases even more effective than some of our conventional tools but and and and that literature has inspired the American Academy of Neurology and the European Academy of Neurology to endorse several of these tests like functional MRI for example. But there's still a lot of unanswered questions about how these tools should be implemented into clinical practice, and they're not routinely covered by insurance so we're still trying to work through this as a field as to how we roll out these techniques locally we we do collect them, uh, or at least several of them through a research protocol, but then for the more validated sequences we share those results with, uh, families and clinical teams. But this is, again, a moving target, a dynamic field, and it'll likely look a lot different even 5 years from now. Just to give you an example of how, you know, what these techniques look like in real life, so I'll talk a little bit more about resting state FMRI. In resting state FMRI, you're measuring the integrity of these functional brain networks, right? So these are networks you may have heard of, like the limbic network, the default mode network, frontal parietal network, attention network, so at a motor network, and visual network. If you put a healthy individual in a in an FMRI scanner and you just measure the brain activity at rest, you're gonna see these fluctuations, these kind of spontaneous fluctuations in brain activity, and you'll notice that these spontaneous flu fluctuations in brain activity tend to synchronize within regions of each of these networks and you can look at the degree to which that synchronization is retained. As an indicator of whether they might recover or not. So for example, this is a patient who we saw in our neuro ICU. It was a woman who presented with this severe intraventricular hemorrhage, and she looked horrible. She was comatose. She had absent pupillary reflexes. Neurosurgery was a little reluctant to even put in an EBD they were thinking, you know, it might, she might be too far gone, but ultimately they did. And we got an MRI with with fMRI with resting state FMRI. So to give you some context, this is what, uh, these networks might look like in a healthy individual, right? So you can see the default mode network, the left frontoparietal network, the right frontoparietal network, and the sensory motor network where warmer colors reflect a greater degree of synchronization. In a patient with a disorder of consciousness who never woke up again. This is what these networks looks like kind of just looks looks like static. But our patient with the intraventricular hemorrhage had networks that were very similar to those of the healthy control. A few days later, her pupillary reflexes came back. A week or two after that, she started following commands again. And we follow her longitudinally in clinic and now she has a conversation with us now she still has some pretty profound motor deficits that are associated with a very prolonged ICU and hospital stay, but cognitively she's made a great recovery and that was a recovery that was, uh, first indicated by this, uh, this, this technique here. All right. So suffice to say, you try to collect as much data as you can, and you synthesize it together to try to get a sense for what their recovery trajectory is going to look like. And unfortunately, it's not kind of so easy as like choosing one of these three options, right? Instead, it kind of often feels more like this, right? You have a, a dis a range of possible recovery trajectories, some of which might be more likely represented by the darker shades, some which you might consider less likely represented by the lighter shades. And once you have kind of in your mind's eye what that. Probability distribution might look like. You kind of superimpose onto it a sense of the patient's goals of care. Now the the term goals of care has become like a little bit loaded like you say if you say, you know, go talk to the family about goals of care, it's kind of become a euphemism for go see if we should transition them to comfort focused care but if you think about what the term actually means, you know, goals of care are the objectives of our intervention, right? What are we trying to accomplish with our life sustaining treatment? And in the context of these disorders of consciousness and a neuroprognostication, that essentially boils down to the question of what is the minimum level of neurologic function that would make life worth living? The minimum level of neurologic function that would make life worth living. That's our goal. If we can get a patient back to that level with life sustaining treatments, then maybe we should. If we can't, maybe we shouldn't. Now, that's easier said than done, right? If you ask me what my goals of care are, I'm gonna have a hard time telling you, let alone trying to figure this out for a patient who can't speak for themselves. So how do we actually think about how we identify these goals of care? Right, and I'll point out that we can actually represent this bar on the schematic, right, because we're talking about a minimum level of function, we can actually plot it where the height of that bar reflects what that minimum acceptable level of function is, and you'll notice that it has not only a a height to it but also a width to it. And that's intended to reflect the fact that not every patient would be willing to wait indefinitely to reach that minimum level of function, right? They might not want to, uh, withstand a long ICU stay and hospital stay and LA stay and sniff stay, right? So that width helps kind of reflect that that deadline by which they would need to get back to that minimum level of functioning. So how do we assess that height and width for these patients? I think there's a few kind of guiding principles that we can use in thinking through this. The the the the most important thing is that we wanna think about what the patient in the future is gonna want. The patient's future interest is the priority, right? Because that's that future patient is the one who's gonna have to bear the brunt of the consequences of continuing life sustaining treatment. So we have to think through what are they gonna want now obviously we can't know that they're in the future. So we have to try to estimate it through these proxies. One of the proxies that we use a lot is what the patient has said in the past, right? Have they left us in advanced directive and this is kind of convenient, right, because oftentimes those advanced directives are very concrete and you can feel reasonably reassured that their incentives are gonna be aligned with that of the future patients, right? It's the same patient just kind of disseminated across time. But there are a lot of drawbacks to these advanced directives too. First of all, only like 10% of patients who come in with an acute disorder of consciousness actually have an advanced directive. Nationally And uh, and even if they do, these advanced directives are very crude, right? It's like they would or would not want a trach, they would or would not want to live as a vegetable, but they rarely talk anything about the minimum level of acceptable neurologic function. And they might be out of date. So that's why we why we rely a lot on families and surrogate decision makers to speak on the patient's behalf. And there are a lot of advantages here, right? The families often have a much more nuanced and up to-date sense of what the patient would want. They may be able to actually speak to what that minimum level of acceptable function actually is, but there are some drawbacks too, even with family or surrogate decision makers, their interests may not be totally aligned. With that of the future patients. And you can think about like dramatic scenarios where family wants kind of the patient out of the picture for one reason or another, but that's like pretty, that's pretty rare, right? Much more commonly is a scenario like this where the family says, you know, I kind of feel like. The person, the patient wouldn't want that prognosis you're telling me, but I, I just can't let them go. I can't do that. Right? That's a misalignment in interest. A third proxy that we don't often use that much, but I think maybe we should, we were talking about this just last night is actually using population data, right? Actual data from patients with brain injury with neurologic deficits to see what their quality of what their quality of life actually is. Right, and the major advantage here is that this helps us circumvent this really pervasive bias that we have right this this ablest bias that we have as clinicians that families have that past patients have when they're formulating their advanced directive that typically overestimates how much function you actually need to have a life worth living. Right, that's this what's been called this disability paradox right this this surprising paradoxical finding that actually you can be happy even if you're in a wheelchair or even if you're, you know, at a nursing home it's possible, right? And the and the studies do do suggest that for a lot of patients and so there's a lot of advantages to using this kind of data to help inform these decisions because they actually reflect unbiased, uh, or less biased I should say opinions from patients who. Have actually experienced this kind of disability before. But of course, the big challenge is, will that extrapolate to the patient in front of you? Maybe this really is the patient who needs to golf and grill in order to have an acceptable quality of life. You do your best. You know, you collect as much data as you can to try to estimate what these goals of care are, again, both in terms of height and width. But once you do that, then you can partition this range of outcomes. Into those that are good, where the patient recovers to this minimum level of acceptable function and within kind of an acceptable time frame, and those, uh, that don't. And once you partition this range of outcomes, now you can start to think in terms of probability. The probability of a good outcome is the proportion of all trajectories that are considered good. And the probability of a bad outcome is the range of all trajectories that would be considered bad. Now we still don't have quite enough information yet. Not every family when confronted with a 60% chance of a bad outcome is gonna choose to withdraw life sustained treatments, right? There are gonna be some families that want to give that 40% chance a shot. So how do we capture that into our, into our framework? That's where these like value modifiers come in, right? So a good outcome is by definition good. A bad outcome is by definition bad. But how good a good outcome is and how bad a bad outcome is is gonna vary a lot from patient to patient, right? And you can even actually think about quantifying this on a scale from 0 to 1 where like a good outcome value modifier of 0 is like relative apathy towards a good outcome. One would be it's like really important to get this good outcome. Like, so like someone with a 1 would be like. Maybe a young person, they have their whole life ahead of them, would really want that good outcome, whereas someone maybe closer to a zero would maybe be someone who had a lot of comorbidities already had kind of a middling quality of life to begin with and kind of returning to that quality of life would comparatively be less important for them. And similarly for the bad outcome value modifier we can we can use this scale to capture how important is it to avoid that bad outcome, right? Someone with a bad outcome value modifier closer to zero might be the patient whose family is telling you like, yeah, they said they wouldn't want this, but they've adapted to these adverse life circumstances in the past. We think they might be able to cope with it, whereas something closer to one might be a family that's telling you, oh gosh, yeah, they already kind of. We're struggling with life the way it was and we're very clear that any more disability would be a a life worse than death. Once you have all this data, now we can think about putting it together. Right, so you have a good outcome term, which is the probability of a good outcome multiplied by how good that good outcome would be. And you have a bad outcome term which is the probability of a bad outcome multiplied by how bad that bad outcome would be. And now you just kind of see which is bigger, right? You subtract one from the other if it's, uh, if it's and you get this what we call predicted life quality term, right? If that PLQ term is positive, it means the good outcome term is higher. If the predicted life quality term is negative, it means the bad outcome term is higher, and we can use the values of PLQ to figure out what to do with life sustaining treatment. If PLQ is positive, it means either the probability of a good outcome was high enough or the importance of achieving that good outcome was high enough that it makes sense to continue life sustaining treatment. If the PLQ term is very negative, it means the probability of a bad outcome was high enough or the importance of avoiding that bad outcome was high enough that we should think about potentially withdrawing life sustaining treatment. And if it's only slightly negative, maybe we like wait a little longer, right? Because that's where the error is, that's where the risk is, right? The bad outcome term might slightly outweigh the good outcome term, but if you, but there could be error there and if we stop life sustaining treatment, the patient dies, we can't take that back. So this is the full kind of framework, right? And I just wanna kind of point out that um this is like a, this is kind of crazy, right? This is like a little bit crazy. Um, it's complicated. But I, but I want to point out a few things. One is to say, this is what we're already doing. Every time we approach these neuroprognostication decisions, we're doing this risk benefit analysis. And it is this complicated. It actually is this complicated. This is what we're doing trying to do in our heads every time we approach this problem. This is just making those elements more explicit. I also want to point out that we rarely, if ever. Have precise numbers that we can plug in for any of these variables. But what this framework is intended to do is to remind us what the important elements are. And to help figure out what the relationship between those concepts is. If we don't even have a ballpark sense for each of these variables, we might not have enough information, right? So for example, the, the the next family that tells us that they're hoping for a miracle. You know, before we just jump straight to a trick and peg, do we have a good sense of their bad outcome value modifier? In, in other words, do we have a sense for how bad it would be if that miracle didn't happen? Right, I think that's important to factor into the, the decision for what to do next. I think also another benefit to this is it gives us just kind of a vocabulary for talking about the problem, you know, if you're on a clinical team and people are disagreeing about how aggressive to be. It can be hard to know, are we dealing with the difference in opinion about the recovery trajectories, about the patient's goals, about the value modifiers? This gives us a way of communicating about the problem. And lastly I think this is ultimately what we should be striving for in our research, right? Are the literature takes this very kind of simple approach to kind of favorable and unfavorable outcomes that really doesn't help us reason through this like we want to. And so I think it's worth thinking about how we design our studies to try to uh plug more precise numbers into this equation. All right, so in summary, neuroprognostication is important, but it's prone to error. Evaluating the level of consciousness is evolving beyond behavior to these technological supplements. Predicting recovery potential requires multimodal assessment, and I've summarized here what the neurocritical Care Society guidelines say about each of these ideologies of brain injury. That emerging techniques like FMRI may help us in predicting these recoveries more accurately. And that even if we had the perfect test for predicting exactly what the recovery is going to be, that would still be insufficient. These, these need to be integrated with a a clear sense of a patient's goals and values so that we can help counsel families on next steps. So thanks so much um, anyone who's interested in learning more, we have, we, as was mentioned earlier, we do have a specialized neuroprognostication program called the Recover program that's a QR code on the left for anyone who's interested in learning more about that and anyone who might be interested in developing their own kind of neuroprognostication program, uh, we have organized this consortium to try to help coordinate our efforts and just form a community around this. We call it the Consciousness Prognostication and recovery Consortium. There's a QR code for our website there on the right. And I just want to advertise the fact that at 4 p.m. today in just 3 hours we're gonna be hosting our first in-person event for this consortium. So if you're interested in learning more about it or potentially enjoying it, please feel free to to come to that and then I have my contact information below for anyone who wants to reach out and chat. Thank you so much. Created by
