Good morning, everyone. Thank you for joining us today for Insights into Shock. I would like to start with, to give the talk about our work on aggressive versus conservative hydration, acute pancreatitis. I want to specifically thank Omar Sroor, he's one of the Palimcrit fellows at Henry Ford Hospital, and Dr. Juneeva Tatum for their help throughout this project. My name is Al-Muthanna Shadeed. I'm one of the third-year internal medicine residents at Henry Ford Hospital. I have no financial disclosures. Our objective is to compare the efficacy as well as the safety of aggressive hydration versus conservative hydration in the treatment of acute pancreatitis. Just a little bit of a background, a little bit of a background, like pancreatitis is a common disease that we encounter in the general wards as well as in intensive care units. Its mortality is significant and it can increase exponentially with local as well as systemic complications and manifestation. Supportive treatment has been the mainstay, including pain control and aggressive hydration. The American College of Gastroenterology, in the 2013 guidelines, they recommend aggressive hydration, which is defined as 250 to 500 mLs per hour, especially in the first 24 hours, in those who do not have any cardiovascular or renal comorbidities. The American Gastroenterological Association in the 2018 guidelines, they suggest using the goal direct medical therapy similar to that of sepsis. With the thought process that these both diseases have similar physiologies, so this could be potentially beneficial, with the goal to titrate to the specific parameters such as the BUN, the hematocrit, the creatinine. However, the GDMT in pancreatitis did not show any improvement in mortality nor the prevention of local complications such as pancreatic necrosis. In the recent years, there has been more trends and suggestions towards having, like evaluating the restrictive approach versus the conservative approach, which brought up the idea of our project. So in our project, we did a meta-analysis of only randomized controlled trials in English, and we defined aggressive hydration as having 200 mLs per kilogram per bolus followed by 300 mLs per kilogram per hour, or an infusion rate of more than 10 milliliters per kilogram per hour. This was defined as the aggressive hydration. The conservative hydration was half of that. Our primary outcomes that we were looking for is clinical improvement, which is defined as improvement in the BUN, the hematocrit, the creatinine, or subjective resolution of like epigastric pain, as well as the ability to tolerate oral nutrition. Other primary outcomes we're looking for is the organ failure, persistence of systemic inflammatory response syndrome, and other secondary outcomes we were looking for is the length of hospital stay, as well as the mortality. For each of these outcomes, we plotted a forest plot, having a 95 percent confidence intervals, and having the p-value of less than 0.05 as significant. So for our outcomes, for the primary outcomes, three studies evaluated these. The clinical improvement, which includes some subjective and objective measures, and there was no one way or the other favor, like favor, basically no significant difference between both groups. We did have similar outcomes in people, the persistence of SERS beyond the 48 hours mark, and there was, between the two groups, there was no difference. In terms of organ failure, although some individual studies did notice some difference, but overall running the meta-analysis, there was no difference between the aggressive versus the conservative groups. In terms of the mortality, our study does suggest that conservative hydration, people who have received conservative hydration can potentially have mortality benefit due to the complications and risks of having this aggressive hydration. We did also find one of the other secondary outcomes that the length of hospital stay can be lower in people who received conservative hydration, as opposed to those who received the aggressive hydration approach. So some conclusion and takeaway points that we can see is that the efficacy of aggressive and conservative hydration appears to be similar, at least from our data, and aggressive hydration can have some morbidity, mortality, and increased hospital stay. However, there's a few things to keep in mind, is that severity of pancreatitis is something that needs to be considered when suggesting aggressive or conservative hydration. There is definitely still more room for research to evaluate comparing both approaches. Thank you. Good morning, everybody. My name is Asfand Yar. I'm one of the New Parliamentary Critical Care Fellows at the University of Toledo, and today I'll be presenting my residency QI project, discussing the safety and efficacy of peripheral vasopressor protocol in early shock in a selected group of patients who we will discuss. I have no financial disclosures. Neither me nor my team had anything to gain out of it except for research. The major objectives of today's presentation are going to be discussed, what the current guidelines are, what the current standard of care is regarding using vasopressors through a central line, reviewing my own intervention and going through some other outcomes and parameters that we'll discuss, and then analyze how it could be applicable to general use in multiple hospitals throughout the nation and the world. Some background to the study. The administration of peripheral vasopressors has always been associated with a theoretical risk of extravasation, infiltration, skin necrosis, and in very severe cases, limb ischemia. But there have been multiple studies over the last two decades that have proved that they're very safe and effective if used in the right manner. Off-note, this study actually started off because we started having a severe increase in collapse rate through COVID. The use of peripheral vasopressors did end up providing a practical alternative in multiple researches before this, so we thought of using it as our own. The aim of our study was to evaluate the impact that it had on central line usage, collapse rate, and other complications in critically ill patients in multiple areas of the hospital. It was a pretty simple study design where we did a pre-post analysis of the intervention throughout a 12-year mark. Sorry, a 12-month mark. It was just one year pre and post. Sorry about that. We did a very superficial analysis before starting, so we did a chart review before, and then we came up with a standardized order set through EPIC. Off-note, we actually removed every other way of ordering vasopressors in the hospital, so anybody who needed a vasopressor needed to use this order set to make sure that we actually have accurate data. The order set did have multiple access points, so you could start the vasopressors directly from a central line or a peripheral line, but you had to choose that order set so that we could just gain our data and standardize the practice throughout. The intervention also included educating the nursing aides, nursing staff, physicians, residents, attendings included through multiple areas of the hospital, including the emergency room, the sick queue, the MICU, and the PACU. The data collected included etiologies of shock, ICU length of stay, Venn days, central line space, extravasation rate, and then the mortality rate. The inclusion and exclusion criteria were pretty straightforward. Anybody who had shock requiring norepinephrine, epinephrine, or phenylephrine as their first pressure of choice, not requiring vasopressin or dopamine as their first pressure of choice, and anybody who was pregnant was out. Anybody who had hemorrhagic shock was out. Now, because we did start this intervention during peak COVID, we had to take COVID out of the equation to make sure that it's not a confounding factor. The good thing was that in our post-analysis we only had 20, 25 patients to compare to a few hundred in the pre, so we just removed all of them to make sure that this does not come up. The results were pretty significant. There was a straightforward 53% reduction in the central line placement. There was a statistically significant reduction in the proportion of central lines used versus peripheral line used, and then there were potential benefits that we will discuss in the later slide. So our CLABSI rate dropped from 1.7% to 1%. Then we also tried corroborating our study with the Nishan data that the hospital independently provides to the state to make sure that they were accurate and collaborated, which I'm happy to inform that they were almost exactly the same. We did come up with the Nishan standard as infection ratio did also significantly come down from 2 to 0.75. The sad part is that the goal was 0.5. We still did not meet that goal for the hospital, but for our study we were still pretty good. The extravasation rate, which was to be expected, did go up from 0.2 to 1.5%. There were no long-term morbidity mortality. There were no problems with skin necrosis. There were no problems with limb ischemia. We had a backup that the pharmacy came up with, sub-Q fentolamine, five injections at the injection site that proved effective every single time we used it. We did have more extravasations beyond that, but we had to exclude them because they were either using more than the recommended dose that we set an upper limit to or in emergent situations where they just added a second vasopressor through the same peripheral access site. Again, even including those, there was no long-term morbidity mortality. There was no issues with skin necrosis and there was no other problems. The only thing was that those patients did have more pain. This is more in the graphical form. Again, from 2020 to 2021, there was a steep increase, which was associated with COVID as well as increase in the critically ill patients throughout the country. 2021 was the time when we actually implemented the project and then you can see it coming down. So hopefully this year we might just reach that benchmark. The usage, as expected, also came down significantly and it continues to come down till date without any effect on the patients. This, we corroborated, as I said, with Nishin as well and we were about to reach our goal, but we had one unfortunate fallacy that we were still up. But the good thing is the standardized utilization ratio was significantly lower than the benchmark so we met that threshold. Discussing some other results, the mortality, the length of stay, and the Venn days, I was very happy to inform that there was no statistically significant difference in them. There was a theoretical risk that if the patient already requires a central line and would need a central line later on, it might actually increase the length of stay in the ICU or in the hospital, which did not prove to be true. This actually could have significant implications on how we manage early shock. There should be a significant reduction in using central lines, especially in emergent situations if you have good peripheral access sites. Not just that, also that it's not the insertion of the central line, it's also the nursing burden that comes in with maintaining a central line that once it's placed, the cleaning and the sanitization of the central line every single day, sometimes two times a day, or if the patient's oozing out of the site, three or four times a day, that can be huge and it has to be done in a sterile fashion every single way. So every time you remove that band-aid, there is a chance of actually introducing infection. Now I come from, this project was actually done in a small rural level one trauma center with around 300 to 400 beds, so we couldn't come up with an RCT style study, so it does come in with its own selection bias, it being retrospective. The data quality relied heavily on the EMR, and to make sure that we were getting accurate data, we removed every other way for people to actually use vasopressors, which may be incomplete or inaccurate. So the incomplete forms were where we actually wanted to see where people transitioned in a 72-hour period from peripheral to a central line if they required one, but because we were unable to calculate that accurately, we did not add it to our study. There were other confounding factors that we did not account for, other chronic illnesses that we could not account for, except for COVID-19, because that was the beast of the day. We still could not accurately say about temporality, because it's not, again, an RCT style study, and to confirm our findings, we would need a much larger randomized control trial to say for sure that this is purely safe and nothing else. But this should be a stepping stone towards moving that. The implementation of the vasopressor protocol does provide a significant benefit, not just to the patients, but to the nursing staff, as well as to the hospital, without actually causing any harm to the patients. It should be becoming the new standard of care in certain patients who are eligible for it. Thank you. These are the references that we used before we started our study. One other thing that I would like to mention, all these studies did show an extravasation rate of 2 or more percent. We were still at 1.5. And last but not least, I would like to thank my mentor, Dr. Ayaz, Dr. Mahmood, Tom, my statistical analyst. I do not know how he comes up with these numbers. Sam, our pharmacist of the ICU, who helped come up with safe dosages and practices, and Amanda and Julia, who actually helped me make this presentation. I could not do it without them. Thank you. Hey, everybody. I'm Jake Jenser from Mayo Clinic in Rochester, Minnesota, and I'm here to present this abstract on shock severity classification and mortality in unselected curriculal adults on behalf of my author group, including several of my good friends and mentors. So for those of you who aren't already familiar with it, the SCI, or Society for Cardiovascular Angiography and Intervention, shock classification was developed in 2019 as a way to describe the severity of cardiogenic shock and to emphasize that cardiogenic shock exists on a spectrum, including patients who have acute cardiovascular disease but do not yet have shock. And so this is a five-stage grading system from A through E of escalating severity. Patients in stage A are those at-risk patients who have acute cardiovascular disease but do not currently have shock. Patients who are beginning to develop hemodynamic instability but do not yet have overt shock, defined as hypoperfusion, are labeled as stage B. Patients who have developed hypoperfusion requiring an intervention are defined as stage C for classic shock. When that initial intervention in a stage C patient fails to stabilize them and they are deteriorating, we label them as stage D. And for patients who progress to refractory shock and extremis, we're labeled as stage E. And so essentially there is a pre-shock stage, which is B, and three stages of shock of escalating severity, C, D, and E. So the question and the whole purpose of this research study was to determine whether this SCI shock classification could apply to other forms of shock and other patient populations besides those just with cardiac disease, specifically patients with medical, neurologic, and surgical critical illness. And so we did a historical cohort study involving all comers to four different ICUs at Mayo Clinic, the Cardiac Intensive Care Unit, Medical Intensive Care Unit, which accounted for the largest group of patients, the Neurosciences Intensive Care Unit, and the Surgical Intensive Care Unit. We didn't want to include cardiac surgery ICU patients because they're a little bit different and have a lot more going on. And so we defined the SCI shock classification on multiple time points every four hours using data from that four-hour period over the first 24 hours. And then we said, okay, what was the highest level that they achieved during that first 24 hours? And we focused on in-hospital mortality, which we analyzed using logistic regression. We did a fairly sophisticated multivariable adjustment, including a lot of relevant variables, including the predicted risk of in-hospital mortality estimated by the Apache 4. So there were a lot of different clinical variables that were used to define the SCI shock classification. And to do so, we defined four different constructs. We defined hypotension tachycardia, which was hemodynamic instability for the purposes of defining who is in stage B. We defined hypoperfusion, which was also synonymous with shock and was a required factor to be in C, D, or E. We defined deterioration, which was used to define stage E. And we defined refractory shock for stage E. And so this was a hierarchical classification based on all these different data points. And so the breakdown of the maximum SCI shock stage for the population shows that the most common stage that patients met as their worst stage was stage C, followed by B and A, and then a smaller number were in D or E. And you can see that this was fairly similar in each of the ICUs, but the prevalence of shock was higher, for example, in medical and surgical versus lower in neurosciences when compared to the cardiac ICU. Now a key factor that I want to mention is that within the cardiac ICU population, as well as patients with cardiogenic shock, acute myocardial infarction, et cetera, there's been a very strong association between the SCI shock classification and mortality, and that's been well established and, again, was part of the motivation for this analysis. And so as you will see here, we saw the same exact thing in this mixed unselected ICU cohort. We saw this clear stepwise increase in the risk of both in-hospital mortality, as well as the ICU mortality, which was part of in-hospital mortality across this whole population. And after adjustment, each higher SCI shock stage, when we thought about it as a continuous variable, was associated with an increase in the odds of in-hospital mortality of about a 30% increased odds per stage. But what you see from this graph is that the patients in stage E were dramatically higher risk than the rest, and that was borne out on our multivariable logistic regression, and these patients in stage E had more than threefold odds of in-hospital mortality compared to the at-risk patients in stage A. And these findings were consistent across the different ICU cohorts, although there were some important differences, and so this is kind of my key slide. So overall, each higher SCI stage was about a twofold odds of higher mortality, and this was the case in the cardiac ICU, where, again, this has been previously validated, and you can see that the performance in terms of discrimination, which we define using the area under the receiver operating curve, AUC, or the C-statistic, was highest in that cardiac ICU population. But you see really the same shape of a relationship in all of the other ICUs. It's most clear, I think, in the medical ICU, and maybe a little bit less robust in the other two ICUs. And I want to sort of highlight the neurosciences and surgical ICU cohorts. Now these have a lot of elective operation patients who presumably are gonna have lower risk because they're sort of critically ill because of a surgery, but they're expected to recover. You can see that borne out by their overall mortality. and of course, in the neurosciences ICU, patients are not usually admitted for shock, they're usually admitted for other things. But nonetheless, you can see the same type of a stepwise increase in mortality across the SkyShock classification. You can see the performance in the surgical ICU really isn't that bad in terms of the discrimination, although it certainly is not as good in the neurosciences ICU. So this, I think, clearly shows the same stepwise relationship between the SkyShock classification and mortality in all these cohorts, even as has been shown in cardiac patients. Now this, of course, this is not a perfect study. There are limitations to any historical cohort. It's very difficult to define cause and effect relationships, and issues related to missing data could have affected the assignment of SkyShock classification, particularly because we did it in four-hour blocks, and instead of looking at the totality of the first 24 hours, we looked at individual four-hour blocks and then took the worst. And this can be particularly relevant for some intermittently recorded data, such as, for example, urine output. And so in conclusion, though, the severity of shock is a crucial factor for predicting outcomes in critically ill patients. And using the SkyShock classification can provide incremental mortality risk stratification on top of established risk predictors, specifically the APACHE-4 predicted in-hospital mortality, which is currently used for benchmarking outcomes. And this was present across the spectrum of medical and surgical critical illness. And so my conclusion is that the SkyShock classification is useful for mortality risk stratification, even outside of its original intent, which was for patients with acute cardiovascular disease and cardiogenic shock. And it can be applied more broadly in unselected critically ill populations. And so I would propose that this can be a universal language for shock that we can use when communicating with other providers and thinking about escalations of therapy, triage and transfer, and mortality risk stratification. So I thank you very much for your attention. This paper has been published online simultaneously with the presentation in Mayo Clinic Proceedings. And if you're interested, you can get it at mayoclinicproceedings.org. Thank you very much. Thank you. Excited to present my research. I'm Ram from Central Minnesota. I'm gonna be presenting the Comparison of Effect of Passive Leg Rays on Pulse Project Variation in Tidal Carbon Dioxide and Velocity Time Integral for Fluid Responsiveness in Patients with Moderately Severe ARDS, who are on Mechanical Ventilator and Vasodilatory Shock. This is a prospective observational study. No financial disclosures, no funding received. IRB approved at the site. Again, it's an observational prospective study. Little background. We know that inappropriate use of vasopressors, especially in uncorrected hypovolemia, can lead to organ dysfunction. At the other end of the spectrum, hypervolemia and excess fluid administration is linked to mortality. And based on the recent literature, it's only 50% of hemodynamically unstable critically ill patients are actually voluminous responsive. So dynamic assessment of preload responsiveness has been proposed to identify fluid responsiveness. These are my abbreviations. Velocity Time Integral, Pulse-Pressure Variation, Internal Carbon Dioxide. So we all know that passively grazed is one of the great methods to assess fluid responsiveness. And it requires a real-time monitoring of cardiac output. Even thermodilution is not validated for passively grazed. So Velocity Time Integral is one of the methods that can be assessed using echocardiogram to identify any change in passively grazed on the cardiac output. So what I did was I compared three different modalities to monitor cardiac output with passively grazed. One was Velocity Time Integral on the echocardiogram. The second was Pulse-Pressure Variation. We know that Pulse-Pressure Variation is inversely correlated to cardiac output. And then Internal Carbon Dioxide, as long as the ventilation is constant, the Internal Carbon Dioxide is correlated with cardiac output. So by doing a passively grazed, I compared the changes on VTI on echo with changes in the PPV and changes in Internal Carbon Dioxide. This is a study protocol. All patients with a new need of vasopressors were first screened with Bed-Stuydeco. Baseline Velocity Time Integral was noted. And tidal volume might temporarily increase to eight mils per kilograms, just for temporary, only for that PLR maneuver. And we had Philips Bed-Stuydeco monitor and activated PPV on the Bed-Stuydeco monitor. And we always monitor Internal Carbon Dioxide, so I made a note of baseline PPV and baseline Internal Carbon Dioxide. And then we did a passively grazed maneuver without stimulating the patient. And then we rechecked VTI. Fluid responsiveness was considered if the VTI was increased by more than 10%, and the baseline VTI was less than 17. These people were given 500 cc of Nomosol, and then I did the VTI back again. If it increased to more than 18 centimeters, we considered that as fluid responsiveness. And again, we noted the change in PPV and change in Internal Carbon Dioxide. So any new onset of hypertension, MAP less than 65, or escalating need for levophedose of more than 0.03 micrograms per kilogram per minute, or especially if we wanted a high MAP goal because of decreased urine output or decreased lactate clearance, those were enrolled. Again, ARDS with PF ratio less than 150, and who are paralyzed or deep sedation. These were my exclusion criteria. Anyone with arrhythmias, chest tube, intra-abdominal hypertension, RBSP greater than 45, pre-MARB, FEV1 less than 1.5 liters, histopseopathy. There were about one or two eligible patients, and there were 23 fluid non-responders, and the rest of them were fluid responders. Of the remaining eligible patients, 17 patients did not have a great echo window for the VTI. Six patients were excluded from the study because despite doing one liter of fluid bolus, they did need vasopressors. Otherwise, four patients were excluded because there was tachycardia during the PLR maneuver. Tachycardia is one of those things that PLR is not validated if there's tachycardia. And two patients were excluded because the VTI did not return to baseline. Sample size was calculated based on the predicted AUC threshold of 75, and we came up with 44. Once we enrolled 50 patients, we stopped further enrollment. We used MedCalc for statistical analysis, and I did statistics based on three parameters, reliability, correlation, and agreement. Reliability, I used receiver operating curve approach with cutoff identified by urine index and correlation with Pearson, correlation with scatter plot. Agreement was based on Bland-Altman plot with a paired t-test. So these were the results between responders and non-responders. No major difference between Charlton comorbidity, SOFA, Apache, PF ratio, PEEP, except for VTI. And this is the ROC curve. For the VTI, it was sensitivity of 100% because that was the definition of fluid responder. And for the PPV, for the delta PPV, the sensitivity was 93%, and the associated criterion was 3.5, which means, for example, at the baseline, if the PPV was 11.5, when we do the PLR, if the PPV drops to eight, which is a difference by 3.5, that indicates fluid responsiveness with a sensitivity of 93% and specificity of 78%. Whereas with internal carbon dioxide, the sensitivity was 86.3, specificity was 84.6, with a change of 5% change in the internal carbon dioxide. These are the scatter plot. Delta VTI versus delta PPV. As you can see, there is good correlation with line of identity. The correlation coefficient is 0.9. I converted to z-scores on the VTI. On the PPV, you can see on the right side, it's almost pretty close to line of identity, although it's not exactly the same. This is a scatter plot for delta VTI and delta internal carbon dioxide. The correlation coefficient is 0.93. On the right side is a z-score, almost close to line of identity. Just for fun, I compared delta PPV and delta internal carbon dioxide to see if there's a correlation or not. It's not great, but the trend is validated. On the right side, you can see the mountain plot between delta VTI, delta PPV, and delta internal carbon dioxide. Because there was good correlation, I wanted to check if there is agreement between delta VTI and delta PPV and delta internal carbon dioxide. I did the bland Altman analysis. You can see that the slope is between the two standard deviations within the limits of agreement. But as you can see, the slope is almost similar to proportional bias. I did the regression equation, and the agreement was validated as long as the VTI was greater than 15%, which means the delta PPV is validated if the VTI was increased more than 15%. And similarly, for the comparison of velocity time integral to internal carbon dioxide, agreement was validated only if the change of VTI was more than 18%. And this is, again, for fun. I compared PPV and internal carbon dioxide. There's one outlier there, so I did not do the regression here. And this is a paid t-test. You can see as, on the left side, the increase or the change in VTI is proportional to the change in PPV. And on the right side is the z-score, almost identical. And again, VTI versus internal carbon dioxide, the increase in VTI was proportional to increase in internal carbon dioxide on the right side is the z-score. Again, conclusions, changes in the velocity time integral, pulse pressure variation, internal carbon dioxide appear to be reliable bedside surrogates to monitor cardiac output for estimating preload responsiveness during a PLR study. And the best cutoff point, delta PPV is 3.9, with a sensitivity of 93%, specificity of 78%. And the delta internal carbon dioxide of change of 5%, with a sensitivity of 86% and specificity of 84%. This is a good correlation between delta VTI, delta PPV, and delta internal carbon dioxide, although it's not completely linear. It's a good agreement between delta VTI and delta PPV, as long as the VTI was greater than 15%, and a good agreement between delta VTI and delta internal carbon dioxide, with a VTI greater than 18%. Clinical implications, in patients with moderate to severe ARDS and shock, preload assessment may begin with simple bedside tools, such as PLR-induced changes in the PPV and internal carbon dioxide. And if positive, fluid bolus can be considered. If negative, additional bedside assessment of cardiac output, such as echo, should be considered. These are my references, and thank you. And lastly, before we have some time for questions, we'd like to ask Dr. Viglianti to come talk to us about intensivist prognostication, late onset shock, and acute hypoxic risk pre-failure. Thank you. Do I need to exit? I want to say good morning, but it's good afternoon on the East Coast, which is where I'm from. So good afternoon to those of you who just flew in overnight as well. And I'm excited to be here with all of you. It's slightly different than the talks we've heard, but I would love discussion if possible. So I have no financial issues or conflicts to disclose, and because I work at the VA, I have to tell you that this work does not necessarily represent the views of the US government or the Department of Veteran Affairs. And of course, I believe team science, you have to give acknowledgments to everyone that's helped support you, whether mentorship, collaboration, sponsors, and so all of them have played pivotal roles in the work I'm gonna be sharing with all of you today. But let's start with a little bit of background. So I think everyone in this room probably agrees with the statement that physician prognostication in the ICU influences and guides how and when information is shared with patients and surrogate decision makers. But our accuracy varies. And so what do we know about this? This work actually was first published back in 2014 by Dr. Brad Barnard in PLOS One, and we found in this study that physician prognostication really varies, and it depends on the outcome. And in this study, they were looking at inpatient mortality. They looked at 2,400 general medicine admission patients, and they asked, is this patient gonna die in the hospital? They also asked the nurses. And when you've looked at the patients and the providers that took care of them, as physicians, we did okay. Our AROC curve, which you've heard quite a bit about today, is around 0.7. But when they asked the nurses and the physicians and they both agreed, that's when we did really well. And the AROC curve went over closer to 0.9. And the predictions were better. So when we have concurrent on decisions from other providers as well as physicians for within hospital mortality, we do pretty good. The next study that we have about physician prognostication, though, starts to take a further step back from when they were in the hospital. And this was looking at six-month survivorship as well as physical and cognitive function for these patients. Similar setup, though. We talked to physicians, we talked to nurses, asked them on the patients that they're caring for, what's the likelihood that they're gonna be alive in six months, and what's their cognition and functional status gonna be? And here again, we have some really interesting studies and results, excuse me. So for physicians, we were actually pretty decent for six-month mortality. And interestingly, we were really good at toileting, which I haven't quite figured out how we got to be so good at toileting. Nurses, on the other hand, were really good at in-hospital mortality, toileting, and going up and downstairs. So there's clearly a gap in our knowledge, though. We have in-hospital mortality, and we have six-month outcomes. But the question that really bears is how do we do for within-ICU outcomes in our prognostication? And if we're good, can we leverage it? Can we leverage it from a standpoint of trying to understand which of these patients are going to identify late-onset organ failure, specifically shock and acute hypoxic respiratory failure? And then this etiology that I'm gonna call persistent critical illness, and I'll define it for you in a couple seconds. That was the foundation of the study, and that's what we set out to go do. How did we do it? Oh, my colleagues. So we did a prospective longitudinal survey using Qualtrics, and we started the study in 2020 and completed it in January of this year. All medical ICU attendings were invited to participate, and because we're all colleagues, of course they all participated in my study. Surveys were electronically sent on ICU day of admission, and then if the patient and the provider were still on service, they were resurveyed again on ICU day three. We had two outcomes we were interested in. Did the patient develop late-onset shock and acute hypoxic respiratory failure? And did they develop this phenomenon we call persistent critical illness? This is how we define them. So late-onset shock was the need for mechanical ventilation, non-invasive ventilation, or heated high-flow nasal cannula on after ICU day four through ICU day 10, and you can see the P to F and the SpO2 to FiO2 ratios, but they couldn't have had anything really on ICU day three. And we picked that ICU day three intentionally because we wanted to distinguish those that came in with this, the initial presentation to the ICU, as compared to those that acquired it later in their ICU stay. Late-onset shock was the administration of any vasopressors on or after ICU day four through ICU day 10, with no vasopressor administration on ICU day three. And then persistent critical illness is defined as an ICU length of stay of at least 10 days. The analysis was simple. This was a two-by-two table, looking at test characteristics of sensitivity, specificity, negative predictive value, positive predictive value, and then that area under the ROC curve. What did we find? So in this study, we had 1,292 surveys that were completed of the 1,471 that were administered over this time period, which gives us a response rate of nearly 88%. It encompassed 960 unique ICU admissions, which happens to be 819 ICU patients. There were some that seemed to come back to us. And 21 attending intensivists. The demographics are on the left side of my colleagues, and it should come as no big surprise. They were predominantly male, between the ages of 36 to 45. We're kind of a younger group. Almost half of us completed fellowship within the last five years, but we do a bit of service between five to 10 weeks. This table probably haunts all of us from our med school days, the two-by-two table, with the disease state on the top and the diagnostic test on the left. I share this with you, because I'm gonna show you a couple of tables for the two primary outcomes. But the disease state in this was chart review. So that's your gold standard. What am I comparing physicians to? And the physicians are gonna be my diagnostic test. Sensitivity, specificity, positive predictive value, negative predictive value. You can see how they're calculated. At the bottom of each of the slides, you're also gonna see an area under the ROC curve. For those of you that tend to fall asleep like I do sometimes, and don't recall, an area under the ROC curve of .5, you're as good as chance. Might as well flip a coin, because you're no better than that. If you really wanna be good, you're at one, and that's when we can really tell the true negatives from the true positives. That's what every test, you want it to be as close to .1 as possible. So what did we find? Here are the two-by-two tables for the first outcome of late onset shock and acute hypoxic respiratory failure. On the left side, you have ICU day one. On the right side, you have ICU day three. Certainly, our sensitivity and specificity suck. I think what I'll also highlight is that the area under the ROC curve also sucks. It's .53 on ICU day one, which if you recall, that's as good as chance just about. And on day three, it's .55. And I highlight the point, the day three, because that's with time, which anecdotally, almost all of you will tell me, I just need time to know how this patient is doing, and yet our prognostication doesn't get better with time, at least for this part. This is for persistent critical illness. Again, similar setup, day one and day three. Our sensitivity and specificity do improve here, which is kind of a relief. But the area under the ROC curve, while better, it again does not change much between ICU day one and day three. It goes from .70 to .71. And truthfully, .7, you're getting a good, and none of us are good. We all want that star and the A+, so we're not really seeing much improvement in our area under the ROC curve for this outcome either, but it is better than at least the late-onset organ failures. So what are the implications from this work? I think as a profession, we really begin, we really need to start to argue and accept the uncertainty surrounding our prognostication early within the ICU stay, of who are going to develop these late-onset organ failures and who are going to get stuck in the ICU for a prolonged period of time. And I haven't quite figured out why, but time doesn't seem to help us in this setting. The other part that's disconcerting to think about is that this survey was done with attending physicians. Most of these conversations at night are not happening by myself. They're happening by my trainees, by my interns, and by my residents. And so if I'm this bad, how are they doing when guiding our patients and the families and the surrogate decision makers on these conversations? And so it gives me pause of what that study would look like. And then lastly, I bring up a new concept of time-limited trials, a concept that's certainly gaining momentum within critical care as something we're supposed to use as a conversation tool with surrogate decision makers and patients when there's uncertainty in their diagnosis and their trajectory of care in order to hopefully align care with their values. And if our prognostication isn't changing over time, how does that fit with time-limited trials when it's supposed to give us time to figure out how we think this patient is going to do? And it just makes me feel a little creepy seeing this data as of right now. So what are my conclusions from this study thus far? Physician prognostication for within-ICU events, it continues to vary, and it really does depend on the outcome. Physicians, we cannot discriminate early in an ICU admission between who will develop late-onset shock and acute hypoxic respiratory failure, but at least we're slightly better at discriminating between who will develop persisting critical illness and who will not. Time in the ICU does not improve physician prognostication, and that's a real big question mark in terms of why. I think the future work and the work that we're doing right now is trying to understand what information physicians are using to guide their prognostication, to try to distill this between the implicit biases we may have, the cognitive biases we may have, and the heuristics we're using early in these ICU stays to help provide guidance to our patients and their surrogate decision makers. And with that, I will stop. So thank you. Thank you.

critical care medicine

acute pancreatitis

aggressive hydration

conservative hydration

peripheral vasopressor protocols

early shock management

physician prognostication

ICU outcomes