So, we'll invite everybody to come in and sit down, and thank you for hanging around for the nightcap session of today's series. This is, I'm Gil Allen at the University of Vermont, and this is my co-collaborator, David Chong from Columbia, and we are going to be hosting and moderating the STAR Research Presentations on Sepsis Therapies. Dave's going to introduce the first speaker. Okay, so first up is the Impact of Delays in Second-Dose Beta-Lactam Antibiotics on Mortality in Sepsis and Septic Shock by John Hammer from Carolina's Medical Center. Thank you for that introduction. I think we are most pretty familiar with this graphic from Kumar and colleagues and also from the data from this study and other studies that show the importance of early antibiotics in sepsis and septic shock, but what is less well-known is the importance of second-dose antibiotics in sepsis. This brings us to a 2017 study in critical care medicine that evaluated delays in second-dose antibiotics. They found that a third of the patients had a major delay, and they found that these delays were associated with increased risk of in-hospital mortality and mechanical ventilation. They found risk factors associated with delays to be more frequent dosing intervals, boarding in the emergency department, and older age. This brings us to the purpose of our study, which is to quantify delays in second-dose beta-lactam antibiotics in sepsis and to evaluate the effect of beta-lactam delay on mortality. We chose to focus on beta-lactams due to their broad-spectrum nature, their time-dependent killing, and also their short post-antibiotic effects, which makes the importance of timely second-dose administration of the utmost importance. Moving to where our study took place, the Metro Charlotte Facilities of Atrium Health, which is a network of 12 hospitals and 10 standalone emergency departments, which encompass 347 ICU beds. Atrium Health has a well-established sepsis treatment pathway, which we refer to as our code sepsis program. In order to be activated as a code sepsis, we have to have suspicion of infection plus either hypotension after 20 mLs per kilo of IV fluids and a serum lactate greater than 4. And then across Atrium Health, we see greater than 2,300 code sepsis activations annually. Our study was a multi-center retrospective cohort analysis from September 2015 to 2017. We included patients greater than 18 years of age who presented and had a code sepsis activated in the ED and who were receiving beta-lactam antibiotics. We excluded those where the first dose of antibiotic was not administered in the ED, where the beta-lactam antibiotics were discontinued prior to the second dose, and those who were transferred between facilities for their first and second dose. Our primary outcome was to compare in-hospital mortality in those with and without a major delay in second dose beta-lactam antibiotics. Our secondary outcomes were to identify risk factors associated with major delays and when looking at the association between major delays and the duration of mechanical ventilation and hospital and ICU lengths of stay. We defined a major delay as being greater than 25% of the dosing interval. So for instance, if someone was on Q8 hour dosing, if they received their second dose more than 10 hours after the first dose, that would be a major delay. And we calculated antibiotic dosing intervals by using patient-specific creatinine clearances. We estimated that we would need over 1,800 patients, assuming that greater than 272 patients had a major delay, in order to have an 80% power to detect a 5.7% absolute difference in mortality, and that we used chi-squared Wilcoxon rank sums and logistic regressions to analyze our endpoints. Of the 3,600 eligible patients, 1,430 were excluded, primarily for either not presenting to the ED directly, for having incomplete data, or not having a second dose of a beta-lactam administered. 284 patients, or 13% of our patient population, had a major delay, and 1,887 patients did not have a major delay. When we look at our baseline characteristics, I want to draw your attention to the higher severity of the illness in the no major delay group, where they had a higher Apache-4 scores, and higher vasopressor use. I also want to bring attention that there was no differences seen in broad-spectrum antibiotics or fluid resuscitation during the first three hours. When we look at the group without a major delay, we did notice that there was more changes in beta-lactam doses between the first and the second dose, and also that our three primary antibiotics given were cefepine, ceftriaxone, and zosyn. Moving to our primary endpoint, we can see that there was around 15% mortality in each group, which leads to no difference in our unadjusted odds ratio. However, when we adjust our odds ratio for the lower Apache-4 scores and less vasopressor use in the major delay group, we see that there's a one and a half times increased risk of mortality in that population. Looking at our secondary endpoints, we'll see that there is no statistically significant differences in either the unadjusted or the adjusted relative risk. But one thing I do want to point out is that when we look at days of mechanical ventilation, ICU length of stay, and length of hospitalization, that the adjusted relative risk are all numerically higher with the group with major delay, as we might have expected. When we look at risk factors for major delays, we see that both age and frequency of beta-lactam administration were associated with major delays. And to point out, so for instance, those on a six-hour dosing interval, they are 43 times more likely to have a major delay than someone on a 24-hour dosing interval. We also found changing the beta-lactam between your first and second dose was protective against major second dose delays. Some strengths of our study were a large sample size, and it was multi-center. We also focused just on those that got the first dose in the ED, and then that were sent up to the ICU associated with that ED. And then we also focused on the beta-lactam antibiotic class, which we think is of the highest importance. Some limitations are that we had fairly strict exclusion criteria. We did not allow transfers between facilities, and once again, we did require that the first dose was given in the ED, which may limit generalizability to other populations. This was retrospective data, so we cannot assess causality, and we also cannot fully control for confounding. And then we also had no data on ED boarding times. So to begin to sum this up, outcome data is well-established on the benefit of first dose antibiotics and the timeliness of first dose antibiotic administration in sepsis. However, our study found that major delays in second dose beta-lactam administration was associated with higher adjusted in-hospital mortality. So in conclusion, we believe that institutions should try to prevent delays in second dose beta-lactam antibiotics in patients with sepsis and septic shock, and that healthcare professionals should be educated on risk factors for major delays, including older age and more frequent antibiotic dosing intervals. I'd like to acknowledge the members of my study team, and we'll take any questions. Thank you. I was just wondering, in your analysis, did you look at the first dose delay in association with the second one? Or did you control for that? We did not control for that, but I will say that when we looked at our first dose delays, and also when we look at it historically through our code sepsis database, they're pretty rare. Our first doses are within the three-hour period and well over 95% of our population, but no, we did not control for that in our analysis. Just curious in terms of the process of handoff. So obviously, while you looked at your initial antibiotics and fluids and those things, which are probably all done in the ED, is this really, I guess, how much of it is the fact that there was delay in the antibiotics, and how much of it was in that transition of care from the ED to ICU? There have been other studies, for example, second lactates associated with worse outcome. Is that really that you didn't draw the second lactate, or was it just a handoff issue that they didn't recognize this person was septic when they got to the ICU, right? So I'm just wondering if you factored in other measures of that handoff to know that that was done. Sure. The question was, how much of this was about second dose delays, and how much of this was about handoffs between the ED and the ICU, if that's correct. One thing about our code sepsis program is, especially during this time, we've had these pretty robust handoff tools, which actually checks where they are in the process of the code sepsis bundles. We still use roughly some modification of the three-hour and six-hour bundles that were historically used. And then that is, basically, our progress on those bundles are handed off from the ED to the ICU nurse. Also our ED boarding times are fairly small as well, so much shorter than the previous study that looked at this. But absolutely, in the ED, they are generally ordering one-time antibiotics, so it's usually the ICU team ordering the second dose of antibiotics. And so a lot of times, the biggest factor in that is delay in ordering. It's not necessarily delay in administering it, it's that they're not ordered. So that has been one kind of future direction that we've really worked more on. We do have continuation phase code sepsis plans, and we are educating about, obviously, these delays and really should order them sooner rather than later. Please identify yourself and where you're from, if you can. Hi, my name's Anand Kumar. I'm from Winnipeg, Canada. So I was wondering, I have a couple of concerns. One is that you basically looked at the three-hour, how many people got antibiotics in the three-hour window. The problem, as I see it, is these days, the time to antimicrobials is relatively short in ERs, like an hour. So you could easily have one group that has a median time of one hour and another group that has a median time of two and a half hours, and that would explain the differences. So it might make sense to actually time it from, you know, to actually look at the time to the first antibiotic. The other thing is, the other reason that may be relevant is that those persons who got delayed in the second dose, the same reasons that caused those delays, whatever they may be, in terms of, you know, what the patient looks like, et cetera, may have caused delays in the first dose as well. So there's some reason to think that may be possible. Sure. Well, thank you. Yeah. Yeah. So basically, to summarize the, you know, may there have been differences in, you know, antibiotic administration time in the groups if we looked at more at one hour or just mainly looked at the time to antibiotics. I think that's a great suggestion, and I believe we have the data to look at it that way also. So I do think that is a great suggestion. One thing the previous study actually found, though, is they actually found early bundle compliance to actually have an antagonistic factor on second dose delays, which would be opposite of what the study authors thought and I think what your suspicion, a lot of our suspicions would be. Thank you. We're going to have to go on to our next speaker. So I'd like to move on. Our second speaker is going to be Brian Anderson from the University of Pennsylvania. He's going to be speaking on improving time, I'm sorry, improving time to urgent antibiotic administration in critically ill patients with suspected sepsis. Thank you for that introduction, and thank you to SCCM for inviting us to present on our quality improvement project. I have NIH funding to study sepsis outcomes, but otherwise I have no disclosures. So just a bit of background. There's fairly consistent observational data linking time to antibiotics and patient outcomes in sepsis with shorter time to antibiotics being associated with less mortality in both sepsis and septic shock. Accordingly, the surviving sepsis campaign recommends that IV antibiotics be administered as soon as possible after sepsis recognition and within one hour for both sepsis and septic shock. While I suspect that there's universal agreement that once sepsis is identified, getting antibiotics in as promptly as possible, that's probably universally accepted, but the one hour mandate can be challenging and probably ends up adding a lot of burden to our bedside providers. When we started looking at this at our institution to, one, start to understand how we were performing with that metric and also trying to understand how we might be able to improve the process, we found that our median time to stat antibiotic administration was 103 minutes. When we initially started to map out the process, we envisioned this very straightforward flow chart from order to administration, but as we dove deeper into the data, realized that there was a lot of variation based on antibiotic type. Some drugs were prepared differently than others, and also by site of delivery. Our hospital has a couple different inpatient pharmacies. So it really made us recognize that we needed to do a much deeper dive into the process to have a systematic understanding of where all of these steps were. So with this study, we sought to map out the process in greater detail, evaluate some potential interventions, and then try to enact measurable change. We started with a swim lane diagram, and we were able to identify two general areas in the antibiotic supply chain that took up most of the time. One was pharmacy dispensing that took up a median of 39 minutes, and the second was antibiotic administration, which took up a median of 40 minutes. When we dove even deeper into those two areas, we found a couple of things that we thought were probably the primary drivers for those long time periods. One was that stat orders printed on the same printer as non-urgent orders. So our pharmacists were trying to be able to pick out the stat orders from all the other orders, and it was really challenging because everything came in the same pile in their queue. The second was that there was a lack of nurse awareness that a stat antibiotic had arrived to the floor. Both of those seemed to be driven by weak visual cues of the priority of the stat antibiotic. So if you look on the right side of the slide is an example of our medication label, and you can see the only thing that indicates a stat order is that one line above the patient's name there with the asterisks and the stat. So that's what everybody was relying on to be able to pick that out amongst all the other orders. So we started in October 2018 with two specific process changes. We purchased new stat printers that were solely for stat antibiotics, and we created a new distinct teal-colored label so that it was visually distinct from all the other medicines, making it easy for the pharmacist and also for the nurse and really everybody along the supply chain to know that this was a specific medication that needed to be handled differently. And separately, we trained our nurses and really all of our providers in the ICU to treat stat antibiotics like blood products. And by that I mean as soon as it arrived to the floor, either through the tube system or by delivery, that whoever got it, whether it was the unit secretary or one of the CNAs or one of the nurses, would bring it right to the bedside nurse so that it could be administered as soon as possible. To evaluate whether our changes had any impact, we did a before and after study. We looked both at process metrics and at clinical outcomes. For the process analysis, we looked at all stat antibiotic orders that were placed during the study period, which was July 2017 through November 2019. And for our patient analysis, we looked at those patients who were admitted to the medical ICUs that we implemented the project in during that same time period who received at least one stat antibiotic during their ICU stay, and we excluded their subsequent ICU admissions if they were admitted to the ICU more than once. For the process analysis, we looked at time to administration using both a Cox proportional hazard regression and restricted mean survival time. And we looked at administration within one hour using a two-sample test of proportions. For our patient analysis, we looked at ICU mortality using logistic regression and ICU length of stay using Cox proportional hazards regression. To be able to evaluate whether there were any temporal trends that could have led to or biased our results, and also to be able to look at the sustainability of our interventions, we also performed segmented regression with data aggregated in four-week blocks. Here's our consort diagram. We started out with a little over 3,600 patients and over 2,200 stat antibiotic orders. In the pre-intervention phase, we had a little over 1,200 stat antibiotic orders and over 1,000 stat antibiotic orders in the post-intervention phase. And those numbers made up our process analysis. After excluding patients on their subsequent ICU admission, we ended up with 640 patients pre-intervention and 450 patients post-intervention, and that made up our patient analysis. The baseline characteristics for our process analysis, you can see there was a pretty broad array of stat antibiotics that were ordered both pre- and post-intervention. We implemented this process change in two units simultaneously. One unit was a 24-bed ICU that was staffed by house staff, and the second was an eight-bed ICU that was staffed by advanced practice providers. And here are our main results. You can see in the graph on the right, the y-axis represents the proportion of antibiotics that were administered, and the x-axis represents the time in minutes. The red line is post-intervention, and the blue line is pre-intervention. And you can see a marked and substantial increase in the proportion of antibiotics that were administered at any time point during the follow-up. This resulted in a 130% increased likelihood of the antibiotic being administered at any time point during that period. And overall, ended up with an average time saved of about 49 minutes and a 47% absolute increase in the proportion of antibiotics that were administered within one hour. When we looked at this using our segmented regression analysis, we found similar results. On the y-axis in this graph is the median time to antibiotic administration in minutes, and on the x-axis is the weeks since the intervention. So on the left side of the graph with the white background is pre-intervention. On the right side of the graph to the right of the zero in the light blue is post-intervention. And so what you can see is before we implemented the changes, our average median time to antibiotics was about 90 minutes. Immediately after we implemented the changes, there was a significant improvement down to about 35 to 40 minutes. And you can see based on this graph that that change and that improvement was sustained for a year after we implemented the changes. Similarly on this graph on the y-axis, you can see the percent administered within one hour, and on the x-axis, again, the weeks since the intervention. In the pre-intervention period, you can see our baseline was about 25 to 30% were administered within one hour. Immediately after we implemented those changes, it rose to about 70 to 75%, and again, you can see those changes were sustained for over a year after we implemented them. When we did our patient analysis, you can see baseline characteristics here. Our ICUs take care of some pretty sick patients. Our baseline APACHE scores are in the high 80s, low 90s for an APACHE 4 score, and about two-thirds of patients were mechanically ventilated. We did find that severity of illness seemed to increase throughout the study period and was higher in the post-intervention period. But even after adjusting for covariates, we found no significant difference in ICU length of stay or ICU mortality before or after the intervention. So in summary, our QI methodology identified opportunities to improve the antibiotic supply chain, incorporating the simple visual cues that we created and default behaviors substantially improved antibiotic timeliness in our ICU. We are now doing a much larger study across our health system and across several different hospitals to try to see, do a little bit more of an evaluation of the clinical implications. This work really could not have been completed without a very large team. I want to specifically thank Bill Schweikert for his help with the QI methodology and also with operationalizing the changes, and I want to specifically thank all of our pharmacy leadership without which we never would have been able to complete this process. And thank you very much, and I'm happy to take any questions. While we're waiting for people to ask questions, I was wondering, you had mentioned that a big chunk of the delay was in pharmacy as well. Was there any intervention, I may have missed this, that the pharmacy implemented in terms of trying to optimize the creation and the gathering of the antibiotics? Yeah, so in addition to creating a distinctly colored label, we also had a brand new printer system that was solely for stat antibiotics so that routine drugs were printed out on one printer and stat antibiotics were printed out on a different printer. So that if the stat antibiotic order came through, they could stop whatever they're doing with this pile over here in their queue and work on that immediately, get it out, and then go back to what they were working on before. So the delay was mostly in knowing which ones were stat and which ones weren't stat? Yeah, so one of the big issues was that it was very challenging based on the way our label was designed to know which orders came through as stat, and particularly also as a stat antibiotic because a lot of other medications are ordered as stat other than antibiotics, but this was a printer that was solely for stat antibiotics. Hey, John Hammer, Carolinas Medical Center. Hey, I was curious, does your institution use ready-to-use antimicrobials in omni-cell or in automated dispensing cabinets, and is that something that you guys had considered in order to eliminate the order preparation time? So that's a really great question. So that was one of the things we did look at early on in our process to determine whether we could have premade antibiotics or the most common antibiotics created, have them ready in the omni-cell. One of the challenges that we came across when we started thinking about that was we have a large population of transplant patients, both solid organ and bone marrow transplant patients, that often require antibiotics beyond the scope of our normal cefepime vancomycin. So when we started thinking about that, we realized that for those patients, those complex transplant patients and also patients who develop nosocomial sepsis and they've just been on a week of cefepime and then get septic again, that we wanted to create a process that would benefit all patients and not just be able to benefit that first dose. And again, one of the limitations of our study is these were all stat antibiotics that were ordered in the medical ICU, and we recognized that a lot of stat antibiotics are given either in the emergency room before they come up to the ICU or on the floor if they're compensating and coming to the ICU. So that's one of the things I think we'll have to look at as we're going wider with this study. Brian, that's just terrific data. And knocking it down by 40 to 50% and going from 100 minutes to 50 minutes, at some point you kind of wonder, wow, no translation outcomes. But I'm always thinking about balancing measures. I'm just curious, A, to be the devil's advocate here, but did you see whether or not you had a increased failure rate to get blood cultures before antibiotics in these patients? That's one of the next things we're looking at. We haven't seen any signal for that. But I do think just for the not finding a difference in patient outcomes, I think there's probably two main reasons. One is, again, we're not talking about probably might not be the first dose. And so because these patients might be coming from the ED, might be coming from the floor, this might be subsequent septic episodes. So that may not relate the same. I thought whether you might be able to more promptly recognize sepsis in the ICU because you're getting their heart rate and blood pressure every hour. So do you find them earlier so then the signal is less? And I think the last piece is also our power. We ended up with about 1,000 people. The largest studies that we're looking in ED patients had 30,000, 40,000, 50,000 patients. And the mortality benefit that you see is very small. It's 4% or 5% or a NODS ratio of 1.04. So I think that's probably one of the main reasons. Yeah. And so I'm guessing most of these are probably not sepsis present on admission, right? They're being recognized and started on the first antibiotic in the ICU. Yeah. But probably most of these are either second episodes of sepsis or broadening because they're not doing well after they've gotten their first round of antibiotics. One of the things we're doing next is we're going to be trying to really nail down to the patients who we know were septic that got antibiotics. The cohort that I'm showing you here is anybody who got an order for one. So whether they got one and then everyone said, oh, they're not septic, they don't need any more, those people would still be included. That might wash out any of the potential benefit as well. Do you think this workflow would translate to the emergency room and on the floors as well? That's what we're doing now. We're taking this hospital-wide and system-wide. I think the main thing that I think take away from this is we found some pretty simple interventions in our health system to be able to make this change. Those interventions may not be the same at different places, but the methodology that we use to be able to identify really simple, cheap, to be honest with you, effective measures, I think is what everybody should take away. Because there may be other things that your system, you might already have a separate printer and a separate color for your stat antibiotics. But there might be other things that you can find by doing this simple methodology to find ways to improve your timed antibiotics. You guys did a really great job of picking at the drivers and the barriers and just finding the simple solutions. Thank you. Thank you very much. Next up is Seth Boyer from Cleveland Clinic. He's going to be talking about hemodynamic response to initial vasopressin dose of 0.03 versus 0.04 units per minute of vasopressin in septic shock. thank you for the opportunity to present our work that was a multi-professional and multidisciplinary effort. In the interest of full disclosure, I'll mention that I'm a consultant for Wolters Kluwer to provide drug information to their products Lexicomp and UpToDate, but I don't have any actual conflicts of interest relevant to this presentation. As you know, vasopressin is suggested by the surviving sepsis campaign guidelines at a dose up to 0.03 units per minute. As an adjunct to norepinephrine for patients with septic shock. Most clinicians think of vasopressin in two different ways, either as endocrine replacement therapy or as a vasopressor for patients with septic shock with corresponding dosing regimens to match these intents. The dose of vasopressin 0.04 units per minute likely comes from the original studies that evaluated this therapy in sepsis from Landry and colleagues that use fixed dose vasopressin as an endocrine replacement therapy. However, the largest study that's evaluated vasopressin for patients with septic shock, the VAST trial, titrated blinded vasopressin at a dose between 0.01 and 0.03 units per minute. And this likely informs the surviving sepsis campaign guideline recommendation for this dose. However, when pharmacists were asked what dose of vasopressin do they recommend for their patients, 200 pharmacists were queried and responded to this survey. And this data found that 86% of pharmacists recommended fixed dose vasopressin. And there was a near equal split between a dose of 0.03 units per minute and 0.04 units per minute with very few respondents indicating that they recommend to titrate vasopressin. And to our knowledge, these two doses have not been evaluated before. So we sought to evaluate the hemodynamic response incidence and clinical effect of patients receiving these two specific doses. And secondarily, to describe the epidemiology of vasopressin dosing in our large health system to see if it corresponded to what we are seeing from pharmacists' respondent data. We designed a retrospective multicenter cohort study that included eight of the 12 Cleveland Clinic Health System hospitals. We included patients with septic shock who received a first-line catecholamine and adjunctive vasopressin. We excluded patients who had vasopressin titrated within the first six hours of therapy or an initial dose other than those study doses. And then we developed cohorts based on this initial vasopressin dose. We defined septic shock based on the CDC adult sepsis event criteria with the new organ dysfunction defined as a new vasopressor infusion with a lactate above 2 millimoles per liter. And our primary outcome was hemodynamic response, which was defined as a decrease in norepinephrine-equivalent catecholamine dose at six hours after the initiation of vasopressin, which has been independently associated with lower mortality in patients with septic shock. From a statistical approach, we developed a propensity score for which vasopressin dose the patients received. And then combined the propensity score with inverse probability of treatment waiting and multivariable logistic regression with adjustment for confounders. I'll make a quick note about APTW because I was not familiar with this approach before starting on the project. This approach weights the observed data based on the inverse of the probability of receipt of the actual therapy the patients received based on the propensity score to create a pseudosample of patients where you can measure the average treatment effect between groups. Our consort diagram starts at the top with over 10,000 patients identified through our informatics department of patients meeting the CDC definition for sepsis and ordered vasopressors. We didn't include most patients because they didn't have hyperlactatemia or didn't receive vasopressin as treatment of their septic shock. We were able to match patients from our informatics report to our outcomes data set, and those patients were excluded. After inclusion and exclusion criteria were applied, we had 1,389 patients included with 53% of the patients receiving a dose of 0.03 units per minute and 47% receiving an initial vasopressin dose of 0.04 units per minute. I'll pause here to just comment that we excluded very few patients because they received an initial dose that was different than one of these two doses, and very few patients had vasopressin titrated within their first six hours of therapy which corresponds to the pharmacist respondent data that I previously mentioned. When we looked at where vasopressin dosing was being used, we saw some very intriguing results. The first was based on hospital location. We found that 96% of the use of vasopressin 0.03 units per minute was from our main campus academic medical center which is a 1,400-bed hospital that is quaternary care versus 0.04 units per minute was equally used at the main campus hospital as well as regional hospitals which are a blend of community and academic teaching hospitals that range in bed size from 150 beds to about 500 beds. Secondarily, we found that the preponderance of patients receiving 0.03 units per minute were in a medical ICU versus patients who were in the other dosing group was 0.04 units per minute, had a near equal spread between medical-surgical and mixed medical-surgical ICUs. At baseline, our patients were quite ill and I'll direct your attention to the observed data. Our mean APACHE-3 scores were above 100 in both groups and SOFA scores were in the teens for both groups with about half of the patients mechanically ventilated. The observed data showed large differences between groups with absolute standardized differences that are much above a comfort threshold. However, in our weighted sample, the absolute standardized differences became much more reasonable and approached what most clinicians would consider significant at about 10%. In regards to our primary outcome of hemodynamic response, we did not find evidence to support a difference in the incidence of hemodynamic response with a vasopressin dose of 0.04 units per minute. And after adjustment, vasopressin dose of 0.04 units per minute was not independently associated with higher odds of hemodynamic response. Our conclusions are twofold. The first is that initial vasopressin dosing varied by hospital and unit culture. And an initial vasopressin dose of 0.04 units per minute was not associated with higher odds for early hemodynamic response to vasopressin. At this point, I will conclude and answer any questions you may have. Thank you. Great results, a lot of data. I'm just curious, it almost looks like a randomized trial in the 0.04. It's almost 50-50. How do you explain the split in the regional center? And obviously, it was more like 70% at the main hospital and almost 90% at the other hospitals. That's a great question. And we were surprised by that data. We actually also found that the use of the higher dose increased over time, which we're evaluating further. But we didn't design the study to answer that question of why there was a difference. But what I suspect is that our main campus medical ICU has a vasopressor treatment algorithm suggestion guideline that actually suggests an initial dose of vasopressin at 0.03 units per minute. And that likely drove the preponderance of use at the main campus medical ICU. Do you have a critical care pharmacist rounding in that unit on a regular basis as opposed to your regional hospitals? We have critical care pharmacists embedded in each of those hospitals. Each of those ICUs in each of those hospitals. It honestly may be just differences of intents and approaches in those different units, but I can't answer that. I'm gathering that there is no system-based protocol for the use of vasopressin. Obviously, it's one of the reasons why you did this. That's correct. We don't have a health system-based approach, and that's one of the reasons why we sought to investigate this. Monica A. Pee Pee from Colorado. Did you happen to look at your incidence of hyponatremia on that higher dose? It's a great question. We haven't looked at that. It's something that we have the data for that we should evaluate. That's a good question. Any other side effects? We didn't systematically evaluate for it, but we can with our data set. It'll be something we need to look into. Great question. And I guess my other question would be that you had average APACHE scores, but were you able to sort of put them into quartiles of severity of illness to see if there's any particular effect that was different from the two doses? That's a great question. We didn't bin our APACHE scores, but we can look at that piece. We did include the APACHE score in our propensity score that the IPTW accounted for, but we can look at it in different ways to see if that indicated some difference. The reason I bring that up, I mean, even if to split it like they did in the VAZ trial. So as I recall in that trial, the FDA actually forced the investigators to split the severity of norepinephrine between 15 mikes and above. They didn't think that that was going to be useful, and it turned out that in the results, the people with less pressers actually benefited more than the people that were higher. So it would be interesting to see if you can see that also in your data, just to split it at least from the initiation of the norepinephrine at up to 15 or less to break that up. It's a great suggestion, and I'll mention that one of my colleagues was presenting data in that relation at this Congress. Thank you very much. Thank you. So our last speaker this evening is going to be Gretchen Satcha, PharmD from Cleveland Clinic. She's going to be speaking to us today about evaluation of the initiation timing of stress dose steroids in adult patients with septic shock. Thank you for the introduction. And if you were here for my earlier session, you might notice that timing is kind of a passion of mine. So for the second time, we're going to talk about the timing of initiation of a therapy, this one, stress dose steroids, in adult patients with septic shock. So in patients with septic shock, we know early intervention and diagnosis is key, as well as implementation of our therapeutic options, including antibiotics, vasopressors, fluids. And corticosteroids, specifically hydrocortisone, are also utilized in some patients for their anti-inflammatory properties, as well as supplementation in what is known to be an adrenally suppressed state. To date, there are four large randomized control trials that have compared the utilization of hydrocortisone to control. And they all, unfortunately, have, not for us to apply this literature, they have very variable results and conflicting results. And there's a couple of factors that have been teased out that have been thought to really drive the difference that has been shown in these clinical trials. First is the actual regimen that is utilized. And then secondly, the severity of illness of the included population in each of these studies. Now, these, again, are the two factors that are most commonly cited and referenced when they're talking about why there might be a difference between the trials. But I would argue that there might also be one key component that's not really talked about very much that might have driven the results or the disparities between trials. And that's the timing of hydrocortisone initiation. So we look at the early 2002 Annan study. So those patients receive steroids within four hours after their vasopressor was initiated or within eight hours after shock onset. And they found a benefit with its use. So maybe early indicating that early therapy or early initiation provides the benefit. Comparing this to the 2008 Cortica study, it's actually unknown when they initiated hydrocortisone in this patient population. But we do know that they randomized patients up to 72 hours after shock onset. And they did not find a benefit with their global overall use. And maybe this is because they initiated it later in their shock course. We don't really know. Now, when we talk about our 2018 studies, Approaches and Adrenal, you can see Approaches, we also unfortunately don't know exactly when they initiated steroids. But we do know that it was noted that they were initiated at least within 24 hours. We don't really know within the spread of that 24 hours. They did see a benefit overall with their use. And then Adrenal in 2018. Again, they didn't really specify when it was initiated, but it was noted that patients were randomized within 21 hours from shock onset. But when they received the drug from randomization is something we don't really know. So I can't exactly explain why the difference in the timing, but maybe Approaches started steroids a little earlier, maybe Adrenal a little later. We don't really know. But regardless of this, there's enough signal for me to at least want to look into this a little bit further, especially knowing that in spite of the conflicting literature, they are still, the surviving sepsis campaign guidelines still do recommend utilizing hydrocortisone in patients that are not able to achieve hemodynamic stability with fluids and vasopressors alone. Which led to this study goal, is to determine the time at which hydrocortisone should be initiated in patients with septic shock after shock onset, as maybe there's a signal that there's benefit with early initiation, but we don't really know at this point. The primary objective was to compare vasopressor duration between patients initiated on hydrocortisone based on different timing cohorts after shock onset. And then secondarily, to compare the mortality rates between those timing cohorts. And then to determine the association of timing of hydrocortisone initiation on vasopressor duration and mortality after adjustment for confounders. This was a retrospective single center observational database study. It was utilized the same database that I mentioned in one of my earlier talks just about an hour ago. Patients were included if they were admitted to the Cleveland Clinic between 2011 and 2017. They had to be adults, meet the CDC adult sepsis event definition, and reside in our medical, surgical, or neuro ICU. They had to have received hydrocortisone for at least two consecutive doses to have considered receiving that drug. We excluded patients who had multiple steroids administered at the same time with the exception of flugocortisone. And then you can see the timing cohorts listed for you on the bottom of the table. So we separated patients into whether or not they received steroids zero to six hours after shock onset, six to 12 hours, 12 to 24, 24 to 48, and then over 48 hours after shock onset. About 2,000 patients were screened for inclusion into this study. After appropriate exclusion, we were able to include 1,470 patients. You can see that about 40% received steroids within zero to six hours, and then around 15 to 17% received steroids within all of the other four timing cohorts. At baseline, you can see looking at the total population just to characterize the population as a whole, you could see the average age was around 61, split pretty evenly between males and females, primarily a Caucasian patient population with majority residing in the medical ICU. And then to point out the severity of illness markers that we have, which is the APACHE-3 score, as well as the lactate at the time of hydrocortisone initiation. What's interesting about this is that you can see that patients in the earlier timing cohorts actually appeared to be more severely ill than those in the later timing cohorts, as evidenced by those APACHE-3 scores and their lactate scores. And this is gonna be important when we talk about our results that we showed. Looking at the vasopressor characteristics, on average, patients were initiated on hydrocortisone when their norepinephrine dose was around 23 mics per minute. Very few patients received a loading dose of hydrocortisone. And then looking at the timing from norepinephrine to hydrocortisone initiation, you could see that on average it was around 26 hours. Looking at our clinical outcomes for vasopressor duration, it actually was shortest in the earliest timing cohorts and then increased up to the longest duration within the highest timing cohort. But what's interesting about this is you might say, okay, maybe that means they were sicker, so maybe they died earlier. But what we actually found was that, looking at vasopressor duration, we found that with ICU and hospital mortality, there was overall no difference in mortality, both ICU and in-hospital mortality within our timing cohorts. But if you look, there's a numeric difference that did numerically increase with the lowest rates seen in the lowest timing cohorts and the highest rates of mortality seen in the highest timing cohorts. So to me, even though the patients that had hydrocortisone initiated earlier were sicker at that point, they actually did not have higher rates of mortality because of this. Looking at our multivariable models, first looking at our linear model for association with vasopressor duration, you can see that we actually found an association with timing of hydrocortisone that met significance for all of our timing cohorts, except for the six to 12-hour versus zero to six. And what we found was that these actually increased up until the maximum association when we compared having hydrocortisone initiated more than 48 hours after norepinephrine initiation. This was associated with a 130-hour increase in their vasopressor duration. And then looking at ICU mortality overall, there was also a significant impact or association. Just when we compared having hydrocortisone initiated more than 48 hours after norepinephrine initiation, this was associated with a 70% increase in the odds of ICU mortality. And the factors that we adjusted for in these models are listed at the bottom of the table. Overall, what we found is that the timing of hydrocortisone initiation after shock onset really just varies at our institution and there's no standard for when we start it. But what we found based on the results of this evaluation was that if initiating hydrocortisone, the benefit does possibly appear to reside with early initiation, typically within the first 12 hours after vasopressors are initiated. However, further prospective data, because this is a retrospective evaluation, are warranted comparing early versus late initiation of hydrocortisone at this time. Thank you for your attention and I will take any questions you might have. Thank you. Just a quick question in terms of your metrics to other sepsis treatment algorithms. So in terms of the fluids and antibiotics and all those other metrics and not just the steroid? Yeah, so good question. So for this evaluation, we actually did not have the opportunity to tease out those factors, but with our inclusion criteria, patients had to have been on antibiotics to even be included. So I wouldn't anticipate there should be a difference. But the fluid component is definitely something that we have the data for, haven't teased out, but is something we can look into. Hi, Michael Shishadi from the University of Pennsylvania. And hopefully I have not exhausted my questions yet that I'm allowed to ask you. I'm wondering, how do we know that the late hydrocortisone is not just a marker of prolonged vasopressor use? Like, you know, hey, this person's been on vasopressors for five days, all right, let's go ahead and start them on hydrocortisone. Right, and actually, I kind of think that that might be what is showing, is that maybe early up we're saying, okay, these patients are sick, these are the patients, we get it. But maybe later, we're kind of just throwing it on and it's not really doing anything. I don't know if there's a way for me to be able to actually tease that out of the data, but I do think that that might be what's happening. Right, I mean, I think the issue there is just, obviously, once again, getting to what do you do in clinical practice, and when you're making the decision, do I initiate hydrocortisone early, or do I initiate it late? Is it necessarily bad to initiate it late if you say it's the vasopressors that are causing the hydrocortisone initiation late? Right, and I think, it's a great question, it really is. I mean, you always have very great questions for me. Thank you. But I think that what it means is, I think if you're looking at a patient early on and you're saying, oh, maybe I should start it, I think what this is saying is just start it. If you're gonna do it earlier is better, don't wait because by then, it's probably not gonna add much benefit. But I don't know how to apply that other than if you're thinking about it, maybe you should just do it. Mike, it almost seems like you have to subtract the entry phase of timing and look at what proportional increase in length of vasopressor timing was there added to the patient's vasopressor use when you started it at 48 hours, right? Because you gotta take away those first 48 hours because it becomes a self-fulfilling prophecy. It does seem like it was a much longer period of time that those patients were vasopressor dependent if you waited 24 to 48 hours proportionately, but it's a big confounder. Yeah, and I'm also just kind of wondering what's going on with those patients because it's greater than 48 hours, but the median time in that group was really long. Very long, yes. What was it, 110 hours or something like that? Yes. You almost wish you could look at, tease this apart provider by provider and see which providers were waiting this long, which providers were giving it early, what else were they doing? I'm not gonna do that personally, Mike. Sounds like you need a randomized trial. Yeah, right, absolutely. Another one. Another one. Hi, Anand Kumar from Winnipeg. Hey, I think timing is really cool too, by the way. It's just fascinating. So I was gonna bring up the same thing, but I might have a solution for you. If this was done over a long enough period of time that the data was covering, how many years was it? 2011 to 2017. Okay, so you're talking about 2011 to 2017. Okay, that might work. So the reason I say that is earlier on when Anand's data came out in particular, then what was going on is that people were using low-dose steroids basically as a supplement on everybody kind of early. Later on when the data started kind of shifting towards the issue of pressers, when the studies came out that were kind of negative on survival but positive on pressers, people started using it later because of the deterioration in hemodynamics. So if you see a temporal kind of arrangement, that might help you a little bit in terms of trying to determine what that is. The other thing I wanna tell you is on the flip side, on the good side, is I actually did a study looking in animals at the relationship of survival in septic shock. If you gave the steroids before or after the antibiotics, just like the meningitis data, and at least in the animals, it actually made a significant difference which way you gave it. I haven't published the data yet, but it's there. That's great. Thank you for those comments. Yeah, it would be very easy for us too to look at if there is a temporal relationship to something that's a great thought. Hi, I have a quick question, I promise. I'm Molly, a pharmacist at the Memphis VA Medical Center. My main question is, you didn't report any of this data, but I was just curious if in any of these patients, fluid decortisone was used as well as hydrocortisone. Because as you mentioned, the two anon trials are the only two trials that use that combination. And I know in my ICU, there's always this discussion of whether we add that fluid decortisone treatment on whether that's making any difference. So I was just wondering if that, if any of your patients had that fluid decortisone, if those patients were excluded. Yeah, great question. Those patients were not excluded. We did include them. I didn't, I have the data. I didn't, it didn't make it into the small time point that I have. But it was also because it was a very low amount. It was, if in groups, it was probably less than five patients each. We don't frequently utilize the combination and we can see that by the results of the study. It was there, but not even enough to be able to tease out those patients to see if they had any extra benefit or not extra benefit. And that would be interesting. I just did a study as well, comparing hydrocortisone to the combination head to head, just in a single center study. But also thinking about the timing of that as well. And if you were going to use fluid decortisone, when that would be initiated. Right, yeah. In addition to the hydrocortisone. Yeah, I think that was great. It was something we were hoping we might be able to tease out. We just didn't have enough use. Thank you. Also, I think that this may be a separate study, but to ask clinicians that you're working in those centers, what they think of steroids as whether rescue therapy or early on, you may get a flavor of what the sentiment is in terms of steroids for patients in shock. Yeah, that's great, yeah. Hi, my name is Tutao. I'm a pharmacist at Community Regional Medical Center. I just had a quick question in terms of when you guys added steroids. Because usually our practice has always been to add levol, vaso, and then the steroids. Do you guys do levol, steroids, then vaso, or kind of that weird combination? Great question. How do you define vasopressor refractory shock? Yeah, yeah. We actually don't have a protocolized approach to the order in which, and it is really variable, too. I think we have that data, but it's kind of across the board on whether it was steroids before vaso, steroids after vaso, steroids at the same time. So it's variable. It's almost like there's three people out there. There's people that use steroids right up front, and some that are waiting for that vasopressor refractory shock, and then some are just using it. And were the doses the same? Was it the 50 or the 100 Q8? It was very often the 50 Q6. We saw a little bit more 100 Q8 in our surgical ICU population, but across the board, the majority was 50. 50 Q8, 50 Q6. Yeah, 50 Q6. And you may have said this already, just one simple question. How did you define onset of timing of shock? Norepinephrine initiation. So it's the initiation. Correct. It wasn't, okay. Yeah, yeah, we didn't have the abilities to look into the actual time that the patient met. So it's the onset of the bacteria. Correct. Now, that in itself could be driving some of this, right? Correct. Yeah. Hi, Hassan Al-Sharif from Saudi Arabia. I have just one question. Why didn't consider adding flugocortisone to your study, or just focusing on steroids? The flugocortisone component, yeah. So what we found was, we were gonna tease it out, but we found we just had such small, little utilization of it. That we weren't able to really pull out any association or impact with its addition. So that's why we weren't able to. Thank you very much. Thanks to everybody. Really enjoyed it. Thanks for hanging around for the NICAP session. I just wanna say, we're one minute early, so.

septic shock

delays in second-dose antibiotics

mortality

mechanical ventilation

risk factors

vasopressin

hemodynamic response

hydrocortisone initiation

ICU mortality

Sepsis

Antibiotics

Shock

Vasoactive Agents

Presentation

Select

Professional

Associate

2020

Advanced

Sepsis Resources