medicine webcast. This webcast hosted and supported by the Society of Critical Care Medicine is part of the Journal Club Critical Care Medicine series. This webcast features two articles that appear in the May 2023 issue of Critical Care Medicine. This webcast is being recorded. The recording will be available to all registrants on demand within five business days. Simply log into MySCCM.org or MySCCM.org and navigate to the My Learning tab. Hello, and my name is Tony Gerlach. I'm a clinical pharmacist at Ohio State University Wexner Medical Center here in Columbus, Ohio, and I will be the moderator for today's webcast. Thank you for joining us. Just a few housekeeping items before we get started. There will be a question and answer or Q&A session at the conclusion of both presentations. To submit questions throughout the presentation, type into the question box located on your control panel. If you have a comment to share during the presentation, please use the question box as well. And finally, everyone joining us today for the webcast will receive a follow-up email that will include an evaluation. Please take five minutes to complete the evaluation as your feedback is greatly appreciated. Please note the disclaimer stating that the content to follow is for educational purposes only. And now I would like to introduce today's presenters. Lucy Porter is a Ph.D. candidate at the Juren Bosch Hospital and Radboud University Medical Center in the Netherlands. Since 2021, she has been conducting research focused on quality of life after critical illness and their incorporation of long-term outcomes in clinical practice. Claudia Smith is a Ph.D. researcher, neurochemistry and brain physics laboratory at the University of Cambridge. Her interests include acute brain injury, severe traumatic injury, cerebral auger regulation, intracranial pressure dynamics, and multimodality monitoring. Thank you both for joining us today. And now I'm going to turn the presentation over to Lucy. Thank you, Tony, for the introduction. And so, as Tony just mentioned, I will be talking today about quality of life after critical illness, and in this case, predicting quality of life after critical illness. Yes. So, I have no declarations. My research is just funded by the hospitals that I work at. And I would like to start by discussing with you why we should be predicting quality of life after critical illness. So, first of all, I think it's important to note that now over 80% of patients admitted to the ICU actually survive their ICU stay. But there is another side to that. A lot of these patients do have long-term physical, mental, and or cognitive symptoms that can negatively affect their quality of life after the ICU. And patients and their families are often not aware of this. And therefore, they can be over-optimistic about their long-term outcomes after the ICU. And this is concerning because quality of life and long-term outcomes in general can be a really important factor in decision-making. So, that's why we set out to develop and externally validate a prediction model for changing quality of life one year after ICU. And one of our main goals was that this model had to be practical. And the reason for that is that we know that less than 5% of the developed prediction models actually make it to the bedside. Of course, we also wanted the model to predict well. And we measured this using the explained variance, which is a measure for how much of the variability in the outcome is predicted by the model. And lastly, we wanted the model to be suitable for the general ICU patients. So, to do this, we used data from the Monitor-IC study. This is a prospective cohort study, which I will tell you a bit more about later. And we used data from patients that were admitted between July 2016, which is when the study started, and February 2020. And this we chose because we wanted to avoid the bias that could have been induced by the COVID pandemic. So, we do not have COVID patients in the study yet. We supplemented this data with data from the Dutch National Intensive Care Registry, and they have a lot of data about the admission type, the admission diagnosis, lab values, and vital parameters. So, they have all kinds of information in there. In total, the Monitor-IC includes seven hospitals. One of those is a university hospital, and there are six smaller non-university hospitals. So, let me tell you a little bit more about the Monitor-IC study. This study is focused on the long-term outcomes after ICU, and this is up to five years post-ICU. And it includes information then on patient-reported outcomes. So, patients fill in these questionnaires, or there are proxies if patients cannot do that at that time. And they answer questionnaires about anxiety, depression, and, of course, quality of life, and many more things. For this study, we used data about their health status and quality of life before the ICU, and about their quality of life one year post-ICU. So, our outcome was change in quality of life. And to measure quality of life, we used the EQ5D5L questionnaire, and this consists of five questions that ask in how limited someone is in five domains. So, they don't ask how far can you walk, but we're asking, you know, do you feel limited with walking? So, that would be the mobility domain. The other four domains are self-care, daily activities, pain, discomfort, and anxiety or depression. And by calculating the EQ5D5L score before ICU and one year post-ICU, we could calculate the change in quality of life. Then, we also made a list of candidate predictors, and we did this with multiple ICU physicians and also researchers using previous literature. And we made a list of 52 candidate predictors, and this includes information on demographics, comorbidities, pre-ICU functioning, and clinical parameters. But we made sure that all of these were available within the first 24 hours of ICU admission, because as I said earlier, we wanted this model to be practical and actually be able to use this in the ICU in a timely fashion. So, to build a prediction model, we used data from the university hospital, the largest hospital also, and we then performed variable selection and internal validation, all with this development data set. Then, we externally validated the prediction model using data from six non-university hospitals. And I don't want to spend a lot of time on our statistical analysis, but I think there's a few things that are important to mention. And one of this is that, of course, we did have missing data. In most cases, when it was missing at random, we used multiple imputation by chained equations, which is called MICE, to minimize the bias caused by this missing data. Then, the variable selection is an important step, and we chose to use LASSO here, and this is a regularization method. It does both regularization and variable selection, which is meant to reduce overfitting of the model. And finally, to internally validate the model, we use a method called bootstrapping, which is a resampling method. So, let me quickly take you through our flowchart. As you can see up here above, we had about 12,000 patients admitted in the time frame we were looking at to seven hospitals. Patients were mainly excluded because they died before informed consent was obtained, or the ICU stay was less than 12 hours, or they did not speak the Dutch language. Due to this, we had about 9,000 eligible patients, and in the end of those 9,000, over 4,000 of these patients gave informed consent and also completed the baseline questionnaire about their pre-ICU health status. Of these 4,000 patients, we did see that over 1,000, so 1,400 of these were cardiothoracic surgery patients, and we did choose to exclude these for this model, mainly because this is a very select patient group that we do not see in all hospitals in the Netherlands. It is usually limited to the larger university hospitals, and we wanted to develop a model that was suitable for the general ICU patients. So, that's why we have almost 3,000 general ICU patients in this study. Of course, not all of them were able to fill in the one-year follow-up questionnaire, and a small amount of patients were excluded because the baseline or outcome data was really insufficient, with the majority of data missing. So, in the end, we included 1,800 patients, of which 1,000 were in the development dataset and just over 700 in the external validation set. So, let me tell you a little bit about our study population. So, we see that the median age in our study population is 64 years old. The majority is male, and we have about 10% of patients that have COPD and diabetes. We also measured frailty, and the median frailty scale was three in our population, with a median length of stay of about two days. And we did see that the majority of patients just were admitted for medical reason, and about half, then, of the patients for surgical reason, which could be either acute or planned surgical. We did look at the differences between the development and the external validation dataset. So, we see an ethnicity. We do not see any differences, but we do see differences, among others, in the amount of patients with patients with diabetes and COPD, with the development dataset having less comorbidities. And we also see differences in mission type, with the development dataset having more patients admitted for a planned surgical reason. And finally, the development dataset also had a lower APOCH severity score, and also a shorter ICU length of stay. So, as I said earlier, we had constructed a list of potential predictors. Of those, a few, then, were not included in the model because of correlations. So, we ended up with 49 variables that we entered into a multivariable linear regression model. And this gave us an explained variance of 56.6%. But 49 variables, of course, is not very practical. And that was our main goal. So, that is why we performed variable selection. And that gave us a final model with just three predictors that had an explained variance of 52.5%. And considering that our outcome is subjective, and 365 days in the future, that is actually pretty good. So, the three predictors included in the final model are quality of life before ICU, the admission type, and the Glasgow Coma Scale in the first 24 hours. So, this model was then externally validated, and we saw good predictive power there. And this plot here also shows that we saw good calibration in the external validation. So, I think there are a few things that it's good to notice here. So, I think the first one is that the predictive value of our final model is actually not much lower than that of the full model. And that is probably because baseline quality of life alone accounts for over 50% of the explained variance. So, that shows us that baseline quality of life is a very important predictor for quality of life one year after the ICU. The second is that the explained variance was actually higher in the external validation dataset than in the model development dataset. And this might be because our external validation dataset had less planned surgical patients and might be more homogeneous because of that. And finally, the majority of our patients actually have a maximum Glasgow Coma Scale of 15. So, the percentage that had a lower score, apparently, it had a big effect on their change in quality of life, also showing that that is an important predictor there. So, of course, we do have limitations. And I think the first two are kind of part of the deal when you're performing research like this. And that is recall bias, first of all, because we do not know which patients are going to be in the ICU tomorrow. So, when we want to know what their quality of life and their frailty and their anxiety symptoms before ICU are, we are asking them afterwards or their proxies after admission to the ICU, you know, how was the situation beforehand, which can induce recall bias. The second is, of course, the last two follow-up. We did try to minimise data or minimise bias by performing multiple amputations, but there are patients that were completely lost to follow-up and that can induce bias as well. And lastly, it could definitely be that there are other predictors that were not included in this study. And one of those is possibly delirium that we did not include. So, for the last part of this presentation, I do want to talk to you about use and practice. As I said at the beginning, we really wanted to develop a practical model. And we do think that this model can support healthcare professionals in preparing patients for life after the ICU. And this is important because realistic patient expectations have actually been shown to contribute to a higher level of patient satisfaction and better patient-reported outcomes in the end. So, what we are doing now is we are evaluating the impact of this prediction model, and we're doing that in a multi-centred, randomised controlled trial. And in that, we are using the prediction model in family meetings in the ICU and discussing the expected change in quality of life with patients and their families. And you can see here the graph that we used for that. We showed them their quality of life before ICU, and we showed them the predicted quality of life one year after ICU. And to evaluate the impact, we will look at different things. We are looking at families, patients and clinicians. And for clinicians, we are looking at the ethical decision-making climate, so their experience. And we're doing that both before and after starting the intervention. As far as patients, we are looking at both their experiences and also their outcomes, in particularly quality of life and anxiety and depression. And that is really the same for their family members. We are looking at what their experiences will be and how their outcomes are. So, to conclude, this is the first validated prediction model for quality of life one year after ICU, and it is appealing for clinical use due to the small number of predictors. And lastly, I would like to thank you guys for your attention, and also, of course, thank all the patients and healthcare professionals of the participating hospitals. And then I would like to give it back to Claudia Smith, who will present her research. Thank you. Thank you so much, Lucy. Great presentation. Thank you. So, I will be discussing the research that's coming out of our ABC group, our African Brain Child Group in Cape Town, South Africa. And we're looking at cerebrovascular pressure reactivity in a large cohort of children with severe traumatic brain injury. So, firstly, to start off, I have nothing to disclose. As you may know, as clinicians and basic scientists, severe traumatic brain injury is really a fundamental global neurosurgical problem, and it accounts for the large neurosurgical problem, especially funding and looking at our research that goes into this disease. Now, we know that more than 80 percent of severe traumatic brain research comes from high-income countries. However, there is a disconnect with where the injury is actually mostly experienced. We see that approximately 90 percent of head injury-related mortalities occurs in lower-middle-income countries. So, there's this disconnection between where we see the problem and where we are looking for solutions. And this is of particular concern in pediatric severe traumatic brain injury. Now, as you may know, we have a very heterogeneous population in pediatric severe traumatic brain injury. We have a very large age range, varying from newborn child up until pre-adolescence and adolescent patients. And it's particularly difficult to study because of this reason, and many of our centres see more adult severe TBI compared to child severe TBI. And this is of particular concern in lower-middle-income countries because of this disconnection between research, but also because of the disease burden, as well as lack of resources. And so, we here in Cape Town are looking to show our research that hopefully demonstrates the things that we can achieve when we have the infrastructure set up in our lower-middle-income country resource setting. Now, severe traumatic brain injury, as many of you may know, we have the primary injury, which is largely regarded as untreatable. So, as clinicians, they focus on the secondary injury, which may be treatable and possibly avoidable. And hopefully, by directing targets at the secondary injury processes, we can hopefully positively impact outcome. And one such second injury process is the loss of cerebral autoregulation. Now, cerebral autoregulation is the brain's ability to maintain adequate cerebral blood flow despite changes in perfusion pressure and cerebral perfusion pressure. And this is often looked at as changes in systemic blood pressures. Now, as you can imagine, if we're trying to maintain an adequate cerebral blood flow, if we have the loss of this mechanism, the loss of cerebral autoregulation, the brain can therefore be at risk at secondary injuries. And so, this matters to a clinician because we have two examples here. The first patient has intact autoregulation, whereas another patient has impaired autoregulation. And in both cases, as part of clinical management, we would have blood pressure manipulation. So, we can see here that if you increase the blood pressure, the intracranial pressure will have very different effects, depending on whether the patient has intact or impaired autoregulation. And so, from this little example, you can see how it would be important for the clinicians to understand what's happening at the autoregulatory level in the patient when we're looking at blood pressure manipulations. And so, if we're looking at severe traumatic brain injury, many cases and in many patients, they experience a loss of cerebral autoregulation, and this leads to worse patient outcomes and oftentimes mortality. Now, we can look at pressure reactivity or cerebrovascular pressure reactivity, which is a distinct component of cerebral autoregulation. And we look at cerebrovascular pressure reactivity by something called PRX. Now, PRX is simply the correlation coefficient between the intracranial pressure and the mean arterial pressure of the patient with a negative PRX being intact pressure reactivity or suggesting intact pressure reactivity and a positive PRX suggesting impaired pressure reactivity. And so, the simple correlation coefficient has been related to mortality and patient outcomes in the sense that we identify in adult literature that a positive PRX, in most cases, we also see worse patient outcomes. And this has become a very growing field in adult severe traumatic brain injury, so much so that PRX has actually been implemented into the treatment protocol. However, the same cannot be said for pediatric severe traumatic brain injury. We don't really understand how PRX is what is happening to PRX in our pediatric population. And so, we aimed to look at and characterize PRX in a large cohort of children with severe traumatic brain injury. And we aimed to look at PRX association with clinical and physiological variables, as well as between PRX and outcome, both as looked at via outcome via dichotomized outcome, favorable and unfavorable, as well as PRX and mortality at six months. We looked at our children with severe traumatic brain injury with the inclusion criteria being a GCS score of less than eight, so a Glasgow Coma score, which gives us an indication of consciousness of the patients as they come into the hospital. And we looked at continuous recordings of intracranial pressure and mean arterial pressure within 24 cumulative hours of the three days of monitoring for which PRX can be calculated. We calculated PRX in ICM+, which is the integrative software that was developed by Cambridge University. And we made PRX the center of our study, and we looked at its relationship with clinical and physiological variables, GCS score, age, intracranial pressure, mean arterial pressure, and cerebral perfusion pressure. Cerebral perfusion pressure is the difference, the mathematical difference, between the mean arterial pressure and the intracranial pressure. And then we had a look at PRX and outcome, so mortality as well as this GOSE PEED score, as well as our functional outcome, which is when we looked at the patients that survived only. And we did multivariable logistic regression analysis, as well as our receiver operating characteristic analysis. And just to summarize our results, we had 196 children with severe traumatic brain injury, with our age ranging between four days and 14 years old. Now, just to give you an indication of what's currently present in the literature, the next biggest study after this one was including 58 patients. So this is really the largest study that's been done to date. And we see that we have a 10.7% mortality rate in our population. And we see consistently that the intracranial pressure and the PRX is consistently higher in patients that died and patients that had worse outcomes compared to their favorable counterparts. And this little diagram will just summarize our correlation coefficients with other clinical variables. So we saw that PRX had a moderate positive correlation with intracranial pressure, a negative moderate relationship with our cerebral perfusion pressure. And we looked at a negative amino arterial pressure relationship with PRX. But importantly here, we can see that there was no correlation between our PRX and our GCS score of the patient. Now, this is really important when we think about the fact that oftentimes GCS scores are used as a proxy for injury severity. And so we can see if there's no relationship between the GCS score of the patient and the PRX status. This may suggest that our PRX value is giving us more information over and above just how severely injured the brain is when the patient comes in. So this is already giving us some idea that PRX could be important in our patient population. PRX was consistently higher in patients with poor outcome. So you can see over here, we've got the first 10 days of monitoring plotted. And this is because we see a lack of data after day 10. So we just did the first 10 days. And you can see here that PRX is much higher and consistently higher over the monitoring time in patients who did not survive. And in addition to this, we can see that PRX has a strong and independent association with mortality. So when we looked at PRX and mortality, we can see that we have an odds ratio of 1.87. However, when we include other co-variables that are known to influence patient outcome, such as the intracranial pressure, the cerebral perfusion pressure, the GCS and the age, we still see that PRX has a robust relationship with mortality. So it's independent of the other intracranial variables that we already look at in our patient population. When we're looking at receiver operating characteristic analysis, we're really looking for a large area under the curve. So for those of you who are maybe not familiar with ROC analysis, we're looking at this area under the curve here when you're looking at a variable and its ability to predict whether a patient dies or survives. And so we see a very high area under the curve when we're looking at our median PRX for our patients. We see an AUC of 0.91. And just comparatively, we know that intracranial pressure has a very strong relationship with mortality and outcome. And the ICP AUC was 0.86. And that's compared as well to our median CPP AUC of 0.76. Now, interestingly here, we wanted to look at the PRX versus the cerebral perfusion pressure in our pediatric population. So if we look into adult data, there's really a lot of research going into CPP OPT. And so many of you may have heard this term CPP OPT. And basically what that is, just to summarize, it's the cerebral perfusion pressure that is deemed to be optimal, meaning that it is the cerebral perfusion pressure at which PRX is the lowest. So if you recall, a low PRX is favorable. It indicates intact cerebrovascular pressure reactivity. And so in adult patients, researchers are discovering that adult patients have a U-shaped curve when you plot PRX versus CPP. And so we did something similar here. The study was not powered to do a CPP OPT calculation or anything related to that, but we thought we'd just investigate it as we have the largest dataset. And so you can see here in pediatrics that we do not experience this U-shaped curve that one may see in adult populations. We see here that a high PRX is found when we have very low cerebral perfusion pressures. So that makes sense to us. We understand that a CPP 20 and 30 is probably not ideal for the patient. And so we can see here a dysfunctional cerebral vascular reactivity. However, when we start to increase our CPP, we see a more favorable negative PRX. And so this may suggest that even at high CPPs, we can still experience intact vascular reactivity in pediatric populations. And that may be different to what we see in adult populations. Some important things to note here. As I said, the study wasn't powered to calculate CPP OPT or to look at CPPs, but we just thought we would plot the data. And so this may be one of the reasons we don't see that U-shaped curve. Importantly, this curve is for the entire population where individual curves may look different. And our center, our Red Cross Children's Hospital, does not target the upper threshold of cerebral perfusion pressures. And so this relationship between PRX and CPP may be very center dependent dependent on your treatment protocols. And lastly, as I mentioned, we have a very large age range for our population here. We went from four days old to 14 years old. And so we need further age related analysis. We did in our analysis separate our age groups. However, we had low patient numbers there in each age group. So no further conclusions could be drawn. So just some take home points to summarize. This was the largest known PRX study in pediatric severe traumatic brain injury with 196 patients. And PRX was consistently higher in patients with poor outcome and had a strong and independent association with mortality. And this was over and above our already measured intracranial variables such as intracranial pressure and cerebral perfusion pressure. Given these results and the potential for avoiding secondary injury, the data suggests that PRX could be useful in directing clinical care going forward. And so further investigations are needed in order to assess its clinical utility. A very big thank you to the African Brain Child Group in Cape Town, South Africa. And thank you so much for having us today and for giving the opportunity to present our research. Thank you. Thank you, Lucy and Claudia for some wonderful presentations. And just a reminder for everyone in the audience, if you would like to type a question in, just use the question box and type your question and I'll get to it. I'm gonna ask Lucy the first question. Why do you think only three of the 49 variables made it into your final analysis with your quality of life? Yes, thank you. So we performed a pretty strict variable selection. We went through two options and one did have a larger model. And we chose the stricter model with only three because we wanted a very practical model. I think the reason it was possible to create such a small model is that baseline quality of life alone is a very important predictor. And then these two other predictors, admission type and Glasgow Coma Scale actually make a difference in the predictive value, whereas the other 50 variables we have only add on a few percentages of predictive power and don't give us a lot extra. So I think we show that these are three very strong predictors of quality of life. Thank you very much. Now, the first question for you, Claudia, it's really with your U-shaped curve of the PRX and the CPP. And I think you kind of mentioned that the geriatric population and adults and kids are not always the same, especially with your age range. How would you, one, think that teenagers might be different than a five to seven year old versus those in the first two years of life before their skulls are fused? Yes, thank you so much. So yes, I think that the first really important point here is that we cannot really treat children as small adults. We kind of enter the point of dangerous territory when we try and extrapolate data. And I think that this CPP-APT or CPP versus PRX curve could be an example of that. We could very much easily see differences in our PRX behavior in children compared to adults. And as you said, that could be age related because we have different cerebral blood flow and different blood pressure targets as well as different response to stimuli in children compared to adults. So asking regarding the age groups, we did actually separate in our analysis, we did have a look at our age group. So we saw toddlers versus two years old to eight years old versus eight years old and above. And we did see a kind of a U-shaped curve in toddlers, which kind of makes sense as well. They may be more sensitive to cerebral blood flow changes. However, we had such a paucity of data that we can't really draw any conclusions from that. There was a relatively robust lack of the U-shaped curve in our preteens and teens. And I do imagine that this will be vastly different to adults and to the geriatric population as well because of how cerebral blood flow changes with age. So yes, I don't think we were too surprised when we looked at this because we definitely need to treat pediatrics as their own population compared to adults. But further investigation is definitely required. Thank you very much. Lucy, I find it very interesting that your cardiothoracic surgery patients were left out. And I was wondering, is that mostly because of the health system that might be different than the Netherlands and the United States, where what we see here in the United States is a lot of emergent CT surgeries where you guys might see more of an elective one? Yeah, I think that's definitely true. The majority of our CT patients are elective and the cardiothoracic surgery is only done in I think about 10 or 12 hospitals in the Netherlands. So if we were to have a large portion of CT patients in our dataset, that model would not be applicable for the majority of hospitals in the Netherlands that don't have any patients after cardiothoracic surgery. It's very centralized in the Netherlands to a few large hospitals. So that was our main reason for leaving them out. We will be developing a prediction model for those patients specifically for quality of life after cardiothoracic surgery. So it's not that we didn't want to include them, it's just we are keeping them separate from the general ICU population. Well, and I think that also brings up a good point working in a surgical ICU, that people don't always understand that these elective cases and versus emergency cases, that can be completely different, especially I think with your generalized wellbeing. Because oftentimes we're trying to do things beforehand with nutrition care or pain management and stuff, and there might be differences just like how peds might be a little bit different than the geriatric patients. So it'll be interesting to see what differences unfold in the future with that. So great work. My next question is for Claudia. What are the next steps to really look about what we should be doing with this PRX, especially in its utility and the care of children with severe TBIs? Yeah, that's a great question. And I think one that we should all be asking after we finish a piece of research. I think that while this is a very large study in the field, really the next step is to get a multicenter trial or a multicenter observational study rather, rather than a trial, that would be first. Primarily because we understand that the treatment protocols differ so vastly from center to center. In addition, we have difference of injury mechanisms. So we understand that a pedestrian in a car accident versus physical abuse versus shaken baby syndrome, all of these things will have different effects on cerebral autoregulation and PRX. And so we really need more data. And this needs to come from different centers globally, especially a collaboration between low and middle income countries and high resource settings. And I think after that, once we have more information, we can start looking at how PRX can be used at the bedside. For example, like the example that I gave earlier, PRX can really help give us an idea of whether the patient is autoregulating or not. And based on that, we can look at blood pressure manipulation. So is it actually a good idea to manipulate this patient's blood pressure if they have dysfunctional autoregulation? Because we may be putting the patient at risk for further injury, further secondary injury. And if we don't know the cerebral autoregulation status of the patient, we could be doing that without understanding what we're doing exactly. Thank you very much. Now, here's another one for you, Lucy. How do you think healthcare coverage plays a role in prediction models, especially in places such as the United States where medical care is a leading cause of financial hardship? Yeah, I think that's definitely a good one. I think it's hard to answer that question because there are two previously developed prediction models for quality of life, but those were also from the Netherlands and Belgium. We are lucky in the Netherlands that we do not have the issue of people going bankrupt based on medical finances. So in our prediction model, stuff like education level and income did not play a large role, but I can imagine that would be different in the US, that those would be predictors of quality of life one year after the ICU. So I would be very interested to see if we do that research in the US, how a prediction model like this would change. And that is definitely an important step. Now, it sounds like some more research collaborations in the future, definitely. Definitely, yes, that would be great. Yep. Next for you, Claudia. Did you guys notice any procedures or drugs that might've caused artifact with your manual data cleaning? And was there a fair amount of data cleaning that was required? Yes, for sure. I think the manual artifact cleaning is a very important methodological consideration in a study like this, especially because it can be so personal. So depending on the researcher doing it, it can vary the results quite a bit, but that's something that can definitely be controlled for. But as far as any kind of clinical interventions that made artifacts, I think particularly important was the disconnection. So the artifacts that we cleaned and removed were mainly just to do with lack of signal or dysfunctional signal. So it wasn't anything physiological. If there was a physiological artifact, we would actually leave it in the data because it really gives us insight into how PRX reacts to that procedure or reacts to that medicine. So for example, if you gave a blood pressure agent and we wanted to look at the intracranial pressure, we would leave that artifact in, and quote unquote artifact, because it's really helpful physiological insights into the intracranial dynamics of the patient. So the artifacts that we mainly removed were things to do with signal disconnections and signal crossovers. So really things that were not supposed to be recording and things that would skew the data that was not necessarily physiological. And there was a lot of data cleaning involved, but mainly because it was a large data set. So not so much because of the nature of the artifacts, but just because of how much data we had. And this is definitely something that is currently being worked on. Currently working with the Cambridge group doing my PhD, and we definitely spend a lot of time cleaning the data. So yeah, definite methodological consideration. Thank you very much. One last question for you, Lucy, besides delirium and our United States healthcare system, were there any other factors that you think that should be studied, but that were not in your prediction model? Or something you would like to take a closer look at? I think what's mainly interesting to take a closer look at is this was the first time we identified Glasgow Coma Scale as a predictor of quality of life. And we did not have a lot of other information on what caused the decrease in Glasgow Coma Scale. So why these patients did not have a maximum Glasgow Coma Scale coming into the ICU. So delirium could be one of those reasons, but I think including other factors such as traumatic brain injury or having a stroke, what those individual factors would do in predicting quality of life would definitely be interesting. And yes, the US healthcare system, or at least, yeah, bringing this research to the US would I think be the second one that's really important to do. Well, thank you very much. And finally, one last question for you, Claudia. Thinking in my mind, especially as a clinical pharmacist, typically if you have a high ICPs, you wanna do some sort of intervention to decrease it. So if you have a high PRX, such as those greater than 0.25, what interventions can you use to actually decrease it if this is becoming more and more clinically relevant? Yes, thank you. Thank you so much, Tony. That's a great question. And I think a lot of the clinicians will have that question on their mind. And to be completely honest, the blatant truth is that we don't actually know. We don't actually know what interventions will apply to cerebral autoregulation because we don't fully understand the molecular mechanisms that underlie the control of cerebral blood flow. And so it's incredibly difficult to think about interventions when we don't fully understand kind of the basics of how cerebral blood flow is controlled in a healthy individual, nevermind in a severe traumatic brain injury or in any kind of disease where we have cerebral blood flow control problems. And so really the steps working towards clinical utility will involve kind of safety and feasibility before we can even look at interventions. But I think we will go towards a place where PRX is more incorporated into the knowledge base of the clinician. So as we mentioned before, like just understanding whether the patient is cerebral autoregulating or not, that already gives us more information that we currently have into the cerebral dynamics of the patient. And so it's not so much gonna be a question of if the PRX is greater than 0.25, what drug can we give? It's more, we look at PRX at the bedside. We see that it's greater than 0.25. What else can we change? And what else is wrong? And what can we direct treatment at in those cases so we can protect the cerebrovascular functioning of the patient? So I think it's more a question of this indication being kind of a warning sign that the patient may not be autoregulating. And so you need to be very careful with what drugs you do administer and with what blood pressure manipulations you do change because that could put the patient at further risk of ischemia and raise intracranial pressure, for example. So I think it's going to be more of an index that we put into our knowledge base when you go forward with making clinical decisions rather than an intervention based on this PRX value alone, if that makes sense. No, no, I think it makes sense. I think it sounds like it's just one piece of the whole puzzle and it's another thing we need to look at to look and maybe this is the early warning system getting more information too. Yeah, exactly right. And I think that it's really important to note here that to calculate PRX at the bedside is actually incredibly easy. If you're already in severe traumatic brain injury, you already have the ICP monitor and you're looking at mean arterial pressure as well. All you need to do is do this correlation coefficient at the bedside. And with software systems like ICM+, it's actually really accessible and really easy. And so to get an indication of the cerebrovascular pressure reactivity in real time at the bedside is a lot more accessible than I think people think. And so a move towards multimodality monitoring is really the way to go in order to identify what's happening with PRX in various patient populations. Well, thank you very much. And thank you again to both Lucy and Claudia. I personally have learned a lot today and that concludes our question and answer session. And thank you yet again to our presenters and you the audience for attending today. Again, everyone who joined us for today's webcast will receive a follow-up email that will include an evaluation. Please take five minutes or so to complete the evaluation. Your feedback is greatly appreciated. And on a final note, please join us for our next Journal Club, Critical Care Medicine on Thursday, June 22nd. And this concludes today's presentation. Thank you.

Critical Care Medicine

quality of life

prediction model

baseline quality

admission type

Glasgow Coma Scale

cerebrovascular pressure reactivity

pediatric severe traumatic brain injury