SCCM Resource Library-Star Research Presentations: State of the Art

Video Transcription

All right. Good afternoon, everybody. Welcome. This is the STAR Research presentation, State of the Art. My name is John Whitcomb. I am a professor and director of the Undergraduate  Nursing Program at Clemson University. Hello, everyone. And my name is Kwame Asante Kwame-Boateng. I'm the lead APP advanced practice  provider at Virginia Commonwealth University for the Department of Surgery Trauma. And my partner is John today.  We're the dynamic duel here for years. Just a couple of points here is obviously you know where the exits are right back here in the back. Restrooms are out to the left. Make  sure that your cell phones are on silence if you could please. Pretty much the speakers will disclose any conflicts if they have any. We have eight speakers. So their talk will  be around 10 minutes, 10 to 12 minutes with a short session afterwards of maybe one to two questions and then hopefully we'll have some time at the end for further questions.  The topics that we're going to have is APP residency length, violence experienced by ICU providers, unacceptable or unacceptable pseudo hyperkalemia, goals of care in patients  after critical, surviving critical illness, factors predicting kidney transplant recipient admission to the ICU, outcomes in patients with sepsis and AKI, saline versus LR and  hemorrhagic shock, and then pediatric critical care resource utilization for disaster triage. So we'll go ahead and kick off the speakers. Amy will be our first speaker and they'll  just follow in behind that. Thank you. Thank you very much. My name is Amy Abott and I'm from Emory in Atlanta, Georgia. And I would like to give a special thank you to Cajun  Zoo, Sarah Gregg, and David Carpenter for their help with the statistical and data analysis. I'd also like to thank Cindy Berg from Ohio State University, Deb Stein from Memorial  Sloan Kettering, and Heather Meissen from Emory for their participation and data collection. With our ongoing and growing ICU population and the increasing complexity of critical  disease, coupled with the continued intensivist shortage, APPs are increasingly being utilized  as part of the ICU workforce. Most new graduate advanced practice providers, APPs, lack critical  care knowledge and confidence prior to starting their first job. This has led to the development  of postgraduate training programs for advanced practice providers. Unlike our critical care physician fellowships, which are standardized to the medical specialty areas, there's a  wide variation among APP residency programs and fellowship programs. This study sought to look at the difference between a six-month and a 12-month training program for advanced  practice providers in critical care medicine. This was a retrospective study over six years, and in the beginning, the RICU exam was used pre-residency and post-residency. And the  RICU is a web-based resident evaluation curriculum and training module in critical care medicine  designed by the Society of Critical Care Medicine. But we developed our own exam utilizing the  competencies developed by the Society of Critical Care Medicine and the European Society of Intensive Care Medicine to develop our own exam, which was adopted by the participating  programs. This exam was a 50-question fill-in-the-blank test to measure critical care knowledge, and  it was scored on a 100-point basis. We also used a 10-question Likert scale to measure the trainee confidence in procedures in the ICU, such as central line placement and arterial  line placement, as well as confidence in managing conditions like ARDS and sepsis. We asked questions on the Likert scale how prepared they felt in order to measure their  confidence, and we scored it on a one-to-five basis, with one being unprepared, two minimally  prepared, three prepared, four well-prepared, and five very well-prepared. We administered this exam at the beginning of residency and again at the end of residency. The testing  encompassed three APP residencies located at the academic medical centers in Ohio, New York, and Georgia. The data that we included covered the pre-test and post-test scores  for knowledge and confidence, as well as the length of residency. We used a PAIR-T test to compare the pre- and post-test knowledge and confidence scores, and then we used analysis  of variance to compare the interaction between the length of the residency and the scores. We had 39 pre- and post-test scores that we used. Twenty-six tests were from the 12-month  group, and 13 tests were from the 6-month group. Here you see the knowledge graph represented  on the left, and the confidence graph represented on the right. The blue bar represents the pre-residency knowledge, the pre-residency scores, and the orange bar represents the  post-exam scores, the post-residency exam scores. Each group showed a statistically significant increase in both knowledge and confidence, and they are broken down into  6-month, 12-month, and overall. The overall knowledge showed a statistically significant increase going from a score of 65.5% to 76.9%. The knowledge score for the 6-month group  showed an increase from 64.4% to 71.5%, and the 12-month knowledge score went from 66%  to 79.5%. But I also want you to look at the confidence graph. There is a much larger difference  in the confidence score for the 12-month group. Looking again at the confidence, both the 6- and 12-month groups entered in with a confidence score of 2, which was feeling minimally  prepared to practice critical care medicine. And at the 6-month mark, that 6-month group scored 3 on the confidence scale, but the 12-month group scored 4, suggesting that they  felt well prepared to practice critical care medicine.  So when we compared the 6-month to the 12-month scores, there was a statistical association  between the length of residency and confidence, but not knowledge. When we looked at the interaction between these scores and the length of residency, there was a strong correlation  between the length and confidence, but not length and knowledge. So we thought about what this  could possibly mean, thought about the results and what it could mean. We do recognize the sample size limitation for the 6-month group and recognize that that part of the design is  different. Had we had more 6-month participants, perhaps the data would have shown an association  or correlation between the length and knowledge. And then also looking at these results, it appears  that the primary gain for knowledge happens in the first 6 months. And what is gained in the last  6 months of the residency program? The second half of the residency program is confidence and how well prepared our trainees feel to practice critical care medicine.  So again, comparing the 6-month residency program to the 12-month residency program, both groups entered in with similar knowledge and confidence scores. But at the end of the  residency program, the 12-month group showed a statistically higher confidence score that correlated to the length of residency as well as a numerically higher knowledge score.  So I'll be happy to take any questions.  That's a great question. And I think the answer, it depends on what you're looking for. Dave Carpenter presented some information this morning regarding the six-month and the OJT.  So if what your administration and what your team is looking for is a knowledgeable but  above average confident provider in the critical care arena, then a 12-month residency program would be ideal.  Right, that's a really good point and good question. We did not parcel out nurse practitioners and physician assistants because we would  have lost power within this study, especially in the six-month design.  It was both. It was a 10-question Likert scale for both. Questions regarding procedures, it was separate questions.  It was 10 separate questions on the 1 to 5 Likert, and it was regarding procedures. We named ART lines, central lines, intubations, but then we also asked questions about confidence  in managing critical diseases like running, being the team lead for cardiac arrest, managing a patient on the ventilator with ARDS, managing a patient with septic shock.  So we included both of those types of questions in the Likert scale.  Right, right. We used that Likert to measure their confidence. And correct.  It's a gradual patient load through our residency program. And we recognize that very, in the very beginning, we identify where they are  in their residency program and what their patient load will be. Light, intermediate, or full heavy load, and we designate that with the preceptors  and leads of each rotation that they are attending.  Thanks everybody. Thank you.  Well, my name is Yves Ouellette. I'm the third author on this research project. Brenda is unable to attend today, so she asked me to present in her place.  So we're going to talk about prevalence of violence experienced by ICU providers. It's a national survey. And we're all aware of violence in our society,  but more and more do we see reports of violence in hospitals and with health care workers and all sorts of different workplace violence.  And some people have referred it to epidemic. So although we know that violent crimes are common and are growing in the public sphere, what is going on in health care?  WHO data suggests that there's a 38% prevalence of violence experienced by health care workers. And on this graph, you can see how  it relates to other type of class of workers, retail, trade, construction, manufacturing. So health care is actually very high in the prevalence of violence.  Health care workers are four times more common to experience serious workplace violence than the private industry. So what is not known is the experience  of violence or threatening behavior by providers. So how do we feel and how do we perceive violence in the ICU, for example? So this is why the survey was created.  So this is a research survey that was, we had IRB approval. It was actually exempt. We had funding from our pediatric department. And we surveyed the SCCM database.  Basically, all members receive an email, up to 16,000 members at the time. The data was anonymous. We did send a one-month follow-up survey as a reminder.  And after two months, we closed the survey and did the data analysis.  The survey was composed of demographic data in terms of who they were, in terms of specialty, and so on in their aspects of their respective ICUs.  We asked about experience with violent behaviors, such as type of experience, outcomes, and type of perpetrators. Also, we asked about perceptions of safety and security  in their own work environment. So we had approximately 858 surveys returned. And 843 were deemed appropriate for analysis. And you have the age breakdown here.  Majority of reporters were in the 30 to 50 age group. I hope that one day I can be one of the greater than 71 years of age and still working in the ICU, or maybe not.  We had a majority of women that reported. And the majority of reporters were physicians, with nurse practitioners and RNs being the second group. So here are the results.  In terms of ICU type, as expected, the majority of the responders worked in the medical or surgical adult ICUs, with the third group being pediatric ICUs.  Number of beds, majority of the ICUs were that the responders  worked in ICUs between 11 beds and 30 beds. And most of the reports were from urban areas compared to rural. And the majority were US-based.  In terms of some of our results regarding type of assault, verbal assault was very common. 69% of responders reported having been threatened verbally,  mostly by patient and family members with non-family visitors available as well.  Sadly enough, some people reported verbal abuse from other staff and physician colleagues, and in one case, a medical director. What was the nature of the verbal threat?  Majority reported physical threat reported for physical harm, including murder threats. In terms of other, some of the interesting reports, so we had the opportunity to, responders  had choices, one, two, five, for example, and then they could enter, self-enter some data. And some of the interesting ones here, loss of employment  lawsuit, even threat to unborn child.  24% reported having been physically assaulted in the ICU environment, with the patient being the most common perpetrator. Type of injuries sustained, majority  were reported to have moderate injury bruising or injury requiring sutures. 9% reported loss of consciousness of life-threatening injury. 36% required medical treatment.  21% reported missing work.  In terms of what is seen in terms of weapons, knife and guns were the most common type of weapons seen in the ICU, with 23% of the responders reporting seeing weapons in their ICU.  In terms of security, 77% of our responders had no training to deal with any of these type of threats, physical, verbal, or sexual threats. 86% of the responders felt violence  could occur in their ICU. So there's a certain feeling of concern about violence in their own work environment. 67% felt unsafe at least once in their ICU,  and 40% felt the threat of violence was present in the day-to-day operations of their ICU. 20% felt that their ICU was very safe,  but 28% felt that their ICU were somewhat or very unsafe. Respondents felt security could somewhat or completely handle some of these actions, such as active shooter  violence without firearm and with firearm. But certainly, the feeling here is that the responders did not feel that security was adequate for the level of possible threats.  In terms of who was feeling these threats, nurses were four times more likely to experience threatened behavior, even though in our survey,  if you remember, physicians were the most common responders. Nurses are seven times more likely to be physically assaulted than physicians. In terms of gender differences, women  had higher odds of non-physical threat and are more likely to experience being verbally threatened by other staff members. Interestingly enough, this odds ratio graph here  demonstrate that nurses who were less than 30 years of age had experienced more violence than more senior nurses. So obviously, junior nurses are more at risk.  93% assaulted by a patient had also been verbally threatened.  Median ICUs had lower odds of threatening behavior than larger ICUs, which you might suspect. However, the number of security measures available  was not statistically significant in relation to risk or threat of assault. I will skip for the one aspect that was interesting in our data is actually  rural ICUs had an increased incidence of physical assaults than urban type of ICUs. So nurses are more likely to experience threats and violent behavior than physicians.  Being female increases the risk of threatened violent behavior from both patient and professional sides. Verbal threats may be a prelude to physical assault.  Most feel that security and safety could be or has been questionable. Most are concerned about safety in their own ICU. Now, the survey does have limitation.  There's bias, obviously, from responders. We had less than 10% responders. And if you have felt violence in your life, you might be more prone to respond to this survey.  Physician role was largest response rate. In terms of next step, I think it's important that security in our work environment needs to be addressed in all institutions.  Social norms and violence should change towards violence.  Self-defense for women ICU providers could be strongly considered. There might be a need for a legislative task in order to protect all health care workers.  In fact, in Wisconsin, our neighboring state has put forward legislation in regards to violence against nurses. Thank you.  It's not something we could actually statistically demonstrate.  Well, so in our survey, we didn't actually ask that specific question, or we didn't put forward scenarios, for example, however, in terms of protection for the health care workers,  there might be better institutional policies in order to protect the health care staff. Regardless of the nature of the mental status of the patient, for example.  But in terms of verbal abuse, a lot of family members were more commonly perpetrators.  Well, I'm a pediatric intensivist, so our visiting hours are 24-7. We do have some security around our, in terms of access, people have to ring in, for example.  In our adult ICUs, I believe that there's not as liberal access to patients by family members. All right. Thank you. Good afternoon. My name is Matt O'Hara.  I'm a Managing Director of Health, Economics, and Outcomes Research at Trinity Partners, based in Boston. We're a consultancy.  I'm pleased to be here to talk to you today about what we're doing at Trinity Partners. I'm going to start with a brief introduction. I'm Matt O'Hara.  I'm a Managing Director of Health, Economics, and Outcomes Research at Trinity Partners, based in Boston. We're a consultancy.  I'm pleased to be here to talk to you today about what we're doing at Trinity Partners. I'm Matt O'Hara.  I'm a Managing Director of Health, Economics, and Outcomes Research at Trinity Partners, based in Boston. We're a consultancy. My name is Matt O'Hara.  I'm a Managing Director of Health, Economics, and Outcomes Research at Trinity Partners, based in Boston. We're a consultancy.  I'm pleased to be here to talk to you today about what we're doing at Trinity Partners. I'm Matt O'Hara.  I'm a Managing Director of Health, Economics, and Outcomes Research at Trinity Partners, based in Boston. We're a consultancy.  I'm pleased to be here to talk to you today about what we're doing at Trinity Partners.  I'm a Managing Director of Health, Economics, and Outcomes Research at Trinity Partners.  So, point-of-care testing. I'm sure a lot of you in this room have interacted with a point-of-care diagnostic device at  one point or another, but just quickly, point-of-care testing, for those that don't know, allows  for the rapid diagnostic testing, close to the patient's bedside typically, not often, not always, and often employs portable or handheld devices.  The growth of point-of-care and point-of-care testing has really exploded in recent years. It's up, let's say, about 10% to 12% per year.  And the use of whole blood and point-of-care testing is widely utilized in EDs, and there are some studies that point to the benefit for both the patient and the hospital for  this type of testing. In addition, the literature has pointed that adoption of point-of-care testing has shown  decreased patient mortality, improved patient care, and reduced turnaround time. So, the reason for our study.  I think it's probably widely suspected that in vitro hemolysis could potentially pose challenges with diagnostic testing, but in particular, point-of-care testing, it was  not well understood. So, let's just talk about some sources of in vitro hemolysis in this setting.  Firstly, in vitro hemolysis can occur during the pre-analytic phase of the testing, which includes, obviously, the blood collection process, how that sample is handled, and then  how that sample is then transferred for the point-of-care testing, either into a cartridge or directly into the device.  There's also some incidence of potential contamination that can occur during this process,  say if the clinician fails to adequately purge the line with discarded blood before drawing their sample.  So, in general, the collection of blood through existing catheters, additional sample transfer tests, steps, excuse me, repeated passing of blood through a small bore needle, all  of these things can lead to in vitro hemolysis. So, why did we want to do this study? What was the main hypothesis? Well, two questions.  One was, could we quantify the degree of pseudohyperkalemia due to in vitro hemolysis in the ED? And then, could we then take that one step further and understand what the impact of  that was on potential patient care? So, the methods we used, we identified patients in the ED who had increased point-of-care test result of elevated potassium.  Just for reference, the reference range at OHSU is 3.4 to 5 millimoles per liter. And then, the unique aspect of this is we pulled together data from not only the point-of-care  device, but the core lab, as well as the EHR, to try to tell a longitudinal story about these patients after they had the initial point-of-care test.  About the data itself, we generated a representative data set. We started first with any patients in a nine-month period that had one of the Chem-8 tests, which  is on the Abbott iStat. It's a cartridge-based test. We then pulled the corresponding core lab data from that patient, as well as the EHR  data, and we looked at which one of these patients had elevated potassium. We found that about 12% of those patients in the ED had an elevated potassium of over 5.1.  We then randomly selected a third of those patients to further analyze and essentially do a chart study on.  So, we looked at those, and we automatically pulled out those patients that we felt that had a natural explanation for this hemolysis.  What we found is we ended up with 100 patients as a representative set with this point-of-care potassium data set.  We found that about 60% of those patients had true hyperkalemia, so explainable true hyperkalemia from elevated potassium.  However, about 40% had pseudo-hyperkalemia, and it was directly resulting from in vitro hemolysis. There is one little sliver there, about 1%, that was due to contamination.  So, if you take that top pie chart and then break it down to the bottom, we then looked at that 40% and tried to analyze downstream impact of that potential hemolysis on patient  care. Of those, we found that about 29% of the 30-40% had no impact on patient care. It was recognized that hemolysis probably occurred and probably had an impact on the  diagnostic test. However, about 11% of those patients, it was not recognized, which led directly to repeat testing or a delay or altered management care for those patients.  So, conclusions. This study demonstrated that in vitro hemolysis is high in the ED and is often very challenging to detect.  We found that about 40% of the patients with increased potassium had, in fact, pseudo-hyperkalemia, but that was probably due to the pre-analog errors.  11% received incorrect care or delayed testing or repeat testing. I think one key point here is that since elevated potassium is common and falsely elevated potassium  is common, we think that we looked at sort of anything above the 5.1 level, but it may also be masking hyperkalemia as presenting as pseudo-normal kalemia, causing substantial  danger to patients if proper intervention is not taken. So, basically, I think our findings say that in the current state, the incidence and inability  to recognize sample hemolysis creates a level of uncertainty for providers and could have downstream impact on patients. A couple of quick caveats if I have time.  So, we really only looked at a couple of different data sets, and to pull this data together,  because there were diverse data sets, we had to look for just those patients that we clearly identified each, so that's one caveat.  And then, although our findings indicate that pseudo-hyperkalemia could substantially impact patient care, I think, obviously, additional studies would help gain insights into this  phenomenon. And then, lastly, we only evaluated patients with an elevated POC potassium. In vitro hemolysis may mask patients with hyperkalemia who show a normal potassium by  point of care test.  I think it is. I think what Steve really did is take a critical eye at as much information about these patients  that he could tell and try to, you know, take out any sort of variability that might cause – it might impact the study, which was a rarely detailed chart review for each one  of these patients. Do you look at cost as well as value? Separately from this study, we did. We did a much more expanded study before we sort of focused on this elevated potassium  issue, looking at cartridge waste at the same institution, OSSU, which was considerable, about 16 percent of cartridge waste in point-of-care testing.  We also look at inefficiencies in repeat testing and delayed diagnostic accuracy. Thank you.  All right, good afternoon, my name is Tammy Eaton, I am a PhD student at the University  of Pittsburgh School of Nursing, nurse practitioner and the co-director of the Critical Illness Recovery Center at UPMC Mercy in Pittsburgh.  I'd like to thank SCCM for providing me the opportunity to share our work from the Critical Illness Recovery Center here today.  Today I will be discussing the result of our project titled Re-Exploring Goals of Care in Patients Surviving Critical Illness.  As far as disclosures, there's no disclosures or conflicts of interest for any of the authors of this project.  So with advances in ICU care, we are seeing an increase in ICU survivors. There are approximately 5.7 million annual ICU admits in the United States, with about  80% of these patients initially surviving the ICU. But before we get too excited about this milestone, we need to consider and explore what is happening  after the ICU, and these concerns include the post-intensive care syndrome, unplanned and late hospital readmissions, new and or worsening chronic illness, and post-ICU mortality.  So ramping up over the last decade, research focused on long-term outcomes have discovered this constellation of physical, psychological, and cognitive disability, new or worsening,  named the post-intensive care syndrome. This has a dramatic effect on both ICU survivors and their families. Also in this ICU survivor population, readmits are a worry.  As reported by Hua et al. in 2015, 16.2% of patients with an ICU stay are readmitted within 30 days, 18.9% between 31 and 180 days.  And over one quarter of all re-hospitalizations involved an ICU admission. So over 30% re-hospitalized at least once in the first six months following their initial ICU stay.  New and or worsening chronic illness, obviously another concerning outcome seen in RICU survivors.  As we see here with Van Busen in 2019, ICU patients have more chronic conditions during the year before their ICU admission as compared with a population-based control group.  And ICU survivors without pre-existing chronic conditions were five-fold more likely to develop  a new chronic condition compared with survivor control patients without pre-existing chronic conditions. And a big one, post-ICU mortality.  One in five survivors, ICU survivors, will die within one year. Most events will happen within 90 days of the ICU discharge.  With advanced age and comorbidity, significant predictors of this outcome. So why are these long-term outcomes and concerns so important to our work in exploring goals  of care in the post-ICU clinic setting? If our ICU survivors are struggling with PICS, which can linger for years after their  ICU stay, if over 30% of them may be re-hospitalized at least once in those six months following  their ICU stay, if there's a five times greater risk of developing a new chronic illness and about 20% of them might die, then somebody understanding these issues need to have these  important conversations with them. Somebody that can help them gain a more realistic understanding of the trajectory of critical care survivorship.  Someone that can help them obtain perspective on what to expect in the future. Someone that can help them understand available treatment options and weigh their care options  in the context of their own goals and values. So the aim of our project was to determine the frequency with which goals of care discussions  occur and how often goals of care change among critical illness survivors evaluated in a post-ICU clinic. Methods.  We took a look at all initial patient visits to the Critical Illness Recovery Center from its inception in June of 2018 through June of 2019.  The data that we reviewed included the frequency of documented goals of care conversations,  documented changes in goals of care as a result of that conversation, and completion of written documentation outlining health care wishes.  And to give you a really quick overview of our program, the Critical Illness Recovery Center, as you can see, we see patients with an ICU length of stay about four days or more.  Some of the common diagnoses that we see are sepsis, respiratory failure, and delirium. And our visits in the clinic occur initially at two to four weeks, three months, six months,  occasionally nine months, and 12 months after the inpatient discharge. With our multidisciplinary and, more importantly, holistic approach to the ICU survivor and  their family, our team provides a PICS assessment and develops a true patient-centered, goal-directed rehabilitation plan.  And much of our conversation revolves around that expectation management normalization of the ICU recovery trajectory.  A special part of our program involves a primary palliative care intervention, which involves goals of care conversations, symptom assessment and management, and support.  As far as a snapshot of our patient demographics within that first year, our ICU that feed into our post-ICU clinic are three different ICUs.  One is a medical-surgical intensive care unit. One is a trauma-burn intensive care unit. And the other is very neuro-ICU heavy.  And you can see that reflected in the primary ICU diagnosis that we're seeing. Our median total ICU length of stay for our patients is about 10 days.  Our median invasive mechanical ventilation days are five. And they spend about 19 days in the hospital total with the population that we're seeing in the clinic.  So to get to the meat of the data, in the first year of the program, we were able to see 117 patients in the critical illness recovery center.  81.2% of those patients participated in a goals of care discussion in the clinic during their initial visit.  And these patients participating in these discussions, 66.3% of the time, there was a family member present during the discussion.  So of these patients participating in a goals of care conversation, we found that about 31.6% of them participating in this conversation reported that their previously requested or  utilized aggressive treatments were no longer consistent with what their current and future health care goals and wishes were.  So basically, about a third of our ICU survivors that we're seeing are beginning to draw that line in the sand regarding their future health care.  It's not an all or nothing discussion, but it's solid steps towards a more goal-concordant care plan.  So as you can see on the slide, about 70% of these patients are requesting a time-limited trial only.  What that means is that although they may be OK with an ICU readmission, they don't want to take a care path that's going to take them to a tracheostomy, to a permanent feeding  tube, or any path that's going to lead them towards dependency or a long-term nursing home stay. 23.3% of these patients no longer want to be intubated.  And these were all patients that were intubated previously during their initial ICU stay. And 6.7% of them don't want to go back to the ICU at all.  And I think what we find as a group most important about this data is 76.7% of patients that  are reshaping their future health care, these conversations occurred with their family member in the clinic.  So these potential future surrogate decision-makers are now hearing their loved ones talk about their firsthand experience in the ICU.  And when they're in that situation in the clinic, they're in a steady state. It's not an emotional decision anymore, as we see so many times in the ICU.  Some additional outcomes of these goals of care discussions. Seven patients that we spoke with reported that they were questioning or reconsidering future aggressive therapies.  Fourteen patients completed written documentation of their future health care wishes. These are 14 patients that did not have any previous written documentation and not only  wanted us to be able to document it in the medical chart, but wanted to also in writing document their future health care wishes.  And 10 patients formally changed their code status to DNR. So this slide is more of a thinking slide than a presentation slide.  So when we look at and discuss future directions of this type of work, I think it's going to be important to consider how important family member engagement and these goals of care  discussions within the post-ICU clinic setting can help future potential surrogate decision-making situations.  Will these changes in future health care that they verbalized in a clinic stand the test of time, or at least through the next rehospitalization?  And timing of goals of care discussion after surviving critical illness, when is the right time? So I wanted to share this slide of our critical illness recovery team.  And in the middle, I want to say a special thank you to our ICU hero, who was honored this morning in the opening session.  Having her as a stakeholder in our vision for ICU recovery care has been a blessing to our entire group. So thank you, Connie. And a shameless plug.  Please feel free to follow all of our group and see all the exciting things that we're doing.  I got some of my group laughing about this. Following all the work that we're doing in the SIRC clinic at UPMC Mercy. Thank you. Is there any questions?  Yeah. So after we have these documents within the medical records, so not including the patients  that are taking that written documentation with them back to their PCP, which we're instructing  them to do, we actually send a copy of our note over to the primary care physician office. If they're in a skilled nursing facility, we send it to the skilled nursing facility.  If they're in the state of Pennsylvania, we have a post form. So we actually will send that original along with them as well.  As our clinic has been open now for about a year and a half, we're building these relationships  with primary care physicians so that we can make a phone call or we can call the APP that's staffing the skilled nursing facility and share the conversations with them.  We are. And that could be a whole other presentation. So as far as the psychiatric evaluation that we're doing, we do a HADS. So we're screening for anxiety and depression.  We're doing a PCL-5 that screens for PTSD. And we are also doing a Zaret's caregiver burden assessment for the caregiver.  And along with that for the caregiver, we are also doing anxiety, depression, and PTSD screening in the caregiver as well.  It's a wonderful question. We just started. So our ICU hero has helped us create the ICU journal program at our hospital. So we just started.  We have not been able to use that in the clinic yet, but the plan is to go ahead and have them bring that to the clinic so we can re-explore and revisit the ICU stay.  And we're pretty excited about it. Stephanie? Yeah. Thank you. I'm sorry. Can you repeat the question?  Yeah. So of course, you're reading my mind. So that's an analysis I'm actually doing right now is looking at predictors of, or influencers of what is changing these schools of care.  So stay tuned. And I do have, I'm looking at a handful of predictors.  Okay, thanks. Thank you.  Hello. Good evening, everyone. My name is Nitin Abrol. I am from Mayo Clinic's transplant unit. I will be talking about our study on the factors predicting kidney transplant  recipient's admission to the ICU. That's a retrospective study over a period of 10 years.  The study was supported by a financial grant from Mayo Clinic's critical care research subcommittee to help with the data extraction and statistical analysis.  We all know that kidney transplant recipients are quite different from other general surgical patients. They have end-organ disease that puts them in ASA class 3.  Many of our recipients are on dialysis before receiving a kidney, and they suffer fluid shifts in the perioperative period and some electrolyte imbalance.  The recipients receiving kidney from a deceased donor often have delayed graft function,  might require dialysis even after kidney transplantation. Moreover, they are highly immunosuppressed, and kidney transplantation is a major surgery that is associated with the  catabolic state. And diabetes has a major cause of renal disease in the United States.  Long-standing diabetes and uremia has  effect on the cardiovascular function, and many of our patients have poor cardiovascular function  at the time of their surgery. Unfortunately, not much is studied in this patient population.  All the scoring systems and risk classification we have with the general surgery patients  unfortunately cannot be applied to the kidney transplant patient population.  The first study looking into this patient population came from University of Pittsburgh Medical Center in 1992. They studied their patients over a period of one year and found that  nearly 42% patients required ICU care after kidney transplantation. And mortality rate in this patient population in that era was nearly 11%. Subsequently, there have been  few other reports from various parts of the world investigating this particular  patient population. However, all these studies are limited by studying very heterogeneous patient  population. They included patients with combined organ transplantation, but we believe that when  a kidney is transplanted with another organ like heart, liver, or pancreas, it's the second organ that drives the risk, not the kidney transplantation itself. So it's not reasonable  to combine multi-organ transplantation patients when we study the risk imposed by kidney transplantation surgery alone. Moreover, the risk of kidney transplant patients is different  depending on at what stage they are from the time of kidney transplantation. In the immediate  post-operative period, the risk is mainly related to surgery and their duration of uremia. However,  as they recover from surgery, then the complications because of immunosuppression kick in.  After six months, when they are weaned off from prophylactic antivirals, then CMV infections and  other opportunistic infections kick in. And further down the timeline, after two years,  they develop complications related to long-standing effects of immunosuppression like secondary  malignancies and problems related to coronary artery disease. And all these studies had very  small sample size. So we were interested to study this particular patient population. We tried to  control all the confounders. And we tried to include very homogeneous cohort of patients.  We included patients from January 2007 till December 2016. Nearly 1,878 patients received  kidney transplantation in this 10-year period. After excluding children, patients who didn't  give any consent, and patients who received multi-organ transplantation, there were 1,527 eligible patients included for this study. We divided this cohort into two groups, one who  required ICU transfer for high acuity care within 24 hours after surgery. That was total 196. And  another group who didn't require high acuity care in the immediate post-operative period.  Unfortunately, one patient had wrong data entered. The cold ischemia time was entered as negative.  That is not possible. Time is always a positive variable. We had to include that patient from the  analysis. So broadly, these are the various indications for requiring high acuity care after kidney transplantation surgery. Majority of patients were transferred for cardiovascular  indications. That included hypertension, hypotension, intraoperative, ST changes, ECG changes. A handful of patients were just transferred for monitoring purpose. So I would  like to share a disclaimer. We don't have a step-down unit in our department. So many of the  high-risk patients who are anesthesia colleagues felt that they would require monitoring immediately  after the surgery. Those were also transferred to the ICU, just because we don't have a step-down  high-dependency unit. So looking at the baseline characteristics, patients in the ICU group  had higher BMI compared to patients who didn't require ICU admission. There was no significant  difference in the age. Patients in the ICU group required transplantation from a diseased donor.  Higher number of patients required transplantation from a diseased donor. More patients were on the  dialysis before transplant. More patients had another surgical procedure combined with transplantation surgery. And majority of patients in this group had induction received from  thymoglobulin compared to LMT-Zumab or basaliximab. Coming on to the intraoperative parameters, only difference that was statistically significant was requirement of crystalloids  during the surgery, even though this was statistically significant. We don't believe it's clinically significant difference of just 200 ml between ICU and non-ICU group.  In terms of hemodynamical parameters, the mean diastolic and systolic pressure was lower in the ICU group compared to the non-ICU group. So these are all intraoperative parameters.  We did a univariate analysis to find various factors that can predict transfer to the ICU  after surgery. BMI, age, higher Charleston score, pre-transplant, dialysis, age of the donor,  type of the donor, induction, ischemia time, blood transfusion during the surgery,  lower systolic and diastolic pressures, and combined procedure was a significant variable in our univariate analysis. However, when we controlled for other factors and did the  multivariate analysis, only BMI, higher BMI, higher age, induction with thymoglobulin compared to  basal liximab, total anesthesia duration, and combined procedure were the only significant factors those could predict initial ICU transfer after the surgery.  Out of these, anesthesia duration and thymoglobulin were the strongest factor. We found that  many of these factors are correlated, like combined procedure and anesthesia duration are actually correlated. When we have another procedure combined with kidney transplantation,  that prolongs anesthesia time. So we excluded those patients. There were about 70 patients who  had combined procedure like allograft nephrectomy at the time of transplantation or a hernia repair  at the time of transplantation. We see many patients with polycystic kidney disease, and we have a policy of removing those kidneys in a symptomatic patient at the time of  transplantation. So we excluded these patients and did sensitivity analysis, and still anesthesia  duration, induction agents, age of the recipient, BMI, and lower systolic pressure during surgery were significant factors predicting admission to the ICU.  Then we divided these patients into five different time zones to investigate if there has been any change in the policy in our department. Number of transplant patients  in these time zones has fairly remained stable except for an increase in the last one in 2015  and 16. Similarly, the patient population has remained fairly stable in terms of age, BMI,  and 24-hour ICU admissions. However, there was a trend of decreasing use of thymoglobulin  and increasing use of CAMPATH over a period of time. Recently, we are using less thymoglobulin  and more CAMPATH for our patients. This is a trend over 10 years in terms of percentage of patients requiring ICU admission that hasn't really changed much.  So based on the statistical analysis, only two slides left. So we did ROC curves, and we  did try to find out the cutoff values. And we found that age around 50 and higher BMI of around  30, induction with thymoglobulin, anesthesia duration of about five years, and lower mean  systolic pressure less than 120 were significant predictors. However, on running the sensitivity  analysis, we found that this model had low sensitivity, but it was highly specific and very high negative predictive value. So transplant recipients less than 49 years of age,  BMI less than 30, and not receiving thymoglobulin, pressure of more than 120. With this model, we can fairly predict that who will not require ICU admission.  So to summarize the finding of this study, overall, there are 13 percent patients require high acuity care in the first 24 hours after surgery. Another surgical procedure can be  combined with transplantation without significantly increasing the risk.  Lower mean systolic pressure is the only significant interoperative hemodynamical parameter. Thymoglobulin induction and anesthesia duration are most important,  but non-modifiable parameters. Older age and higher BMI are only patient-related factors  that showed significance in this analysis. And comorbidities and donor type are not important  in predicting ICU admission in immediate post-op period. Overall, the model accurately predicts who would not require high acuity care after transplantation surgery. Thank you.  I'm happy to take questions.  Yeah, so each center has different policies in terms of transferring patients to the floor or to the ICU. It depends on the volume of transplants done at their centers.  There are many centers not doing high-volume transplant surgery, so their policy is to transfer the patients routinely to the intensive care unit for initial monitoring.  Based on this study, only 13% of patients require ICU transfer. So we can reasonably and confidently tell who would not require ICU transfer, and we  can transfer those patients to the general floor. And I believe this can cut down the cost of transplantation.  In last critical care medicine meeting, we presented a study on outcome of these patients,  and we found that patients who required transfer to ICU from the general floor had higher mortality rate and had longer duration of stay in the ICU.  So we believe that probably there's a window that we can miss if we transfer those patients to the floor initially.  So those patients can be transferred to ICU electively for monitoring purpose. And once they're stable, it's just the overnight stay. They can be transferred back to the floor.  The problem with first studying the transplant patient is there are so many confounding variables  It's very difficult to control those variables, like induction with thymoglobulin. That's important finding in the study, but we don't have really any answer or any explanation  why patient with thymoglobulin would require more ICU transfer. Maybe it's related to cytokine release, or maybe related to institutional policy, because  thymoglobulin can be given to center line, and many of patients who do not have center line require center line placement before surgery, and that's anesthesia time.  So whether these two variables are interrelated, it's really hard to control these two variables.  Probably that's why there was low sensitivity in this model, but with the high specificity and active predictive value, we can confidently say who will not require ICU transfer, and  that patient can be transferred to general floor. Thank you. Thank you.  Good afternoon. My name is Sebastien Rachouin. Over the next 10 minutes, I'll be talking to you about the impact of gender and ethnicity on outcomes of patients with sepsis and AKI.  As we all know, sepsis is bad, and sepsis and AKI is worse. Unfortunately, over the last few decades, we know much more about sepsis, AKI, but we  don't have a ton of great modalities to mitigate the outcomes. One thing we wanted to do was to look at risk factors that would identify high-risk patients  and high-risk groups and see if we can maybe in the future tailor studies to target those groups.  We used the National Inpatient Sample, which is a very large database that is accessible to everyone. It has more than 7 million annual visits.  We did a retrospective study, obviously, between 2005-2014. We looked at ICD-9 codes for AKI, ATN, sepsis, severe sepsis, and septic shock.  The outcomes of our study were mortality and dialysis, and we included the Charleston score to adjust for comorbid conditions.  For those of you who don't know, the Charleston score is 19 comorbid conditions plus age, and that has been shown to be a reliable way to assess severity of illness in a big  data set and big administrative data studies. So we had 65 million records that we reviewed, and here are the demographics and differences  between the patients that were hospitalized and the ones that were hospitalized with sepsis. Of note, when I refer to sepsis, here it's sepsis, septic shock, and severe sepsis all  bundled together. Differences between the groups were that septic patients were older, and they were much less female, a little bit more African-American.  They were sicker, as shown by the Charleston index. They had more patients with AKI, more mortality, and less dialysis.  So over the 10 years of our study, we saw that incidence of sepsis went from about 2.5% to more than 6.3%. This is just looking at the patients in the general population.  You can see that AKI started increasing starting in 2006, went up to 13%, whereas mortality and new dialysis were roughly stable over the same time period.  If you look just at patients with sepsis, AKI also increased more spectacularly, actually, from 29% up to 41% of patients, but at the same time, mortality plummeted from 23% to  13%, and the new dialysis also went down from 11% to 7.4% in this subset of patients. We then looked specifically at the question of gender, ethnicity, and age.  We found that female gender was actually associated with better outcomes for all the different categories and combination of categories we looked at.  African-American ethnicity was associated with worse outcomes, and age was worse, and this is by year. So every year of age adds 1.03 in the risk factor.  We then combined those into some subgroups. So we found some literature that says that estrogen protects against bad outcomes. So we arbitrarily defined 50 as a cutoff.  We said above 50 would be postmenopausal, before 50 would probably be premenopausal, and then we combined into some groups.  And there was pretty much a lot of concordance between the groups, except for African-American males less than 50 who had worse outcomes when they had sepsis to get new dialysis,  and for non-African-American females more than 50 years old had some protective effect for new dialysis and for mortality when they get dialysis.  Obviously, this is a retrospective study and has all the limitations of a retrospective study, and it's based on administrative data, so also this is a big limitation.  We could not distinguish the etiologies of AKI. What we also decided to do in this study, we decided to exclude all the patients that  had any diagnosis of chronic kidney disease or chronic renal failure so we don't muddy it with some patients with ESRD.  My conclusion is that it seems that women have a lower incidence of dialysis, sepsis, and death than men.  Black ethnicity has worse outcomes overall, and increasing age generally has less favorable outcomes.  And hopefully, it will help us to design future trials to analyze the impacts of therapies. That's pretty much it for me.  Questions? Thank you.  The idea is maybe that in the future, we figure out why. And we mainly focused initially on gender and ethnicity. There's some biological plausibility for the gender.  It seems that the ethnicity is mostly associated with poor insurance status. And maybe there's a gene that's called APOLI1 that could be associated. The age, we had to include it.  I'm not sure how much I can explain.  Absolutely. All right. Thank you. Can you look at the federalism titles, the source of the substance itself? No, we could not. All right. Thanks.  Good evening. My name is Mitch Daly and I'm a pharmacist in the Surgical Trauma ICU at Del Seton Medical Center at the University of Texas.  And today I'm honored to share with you a research project on the safety of normal saline versus lactated ringer solution in hemorrhagic shock from trauma, a secondary analysis of  the proper trial. The investigators, including myself, have no conflicts of interest to disclose related to this subject material.  Hemorrhagic shock is one of the most common forms of shock that we're faced with in clinical practice, particularly in a Surgical Trauma ICU.  Hemorrhagic shock is defined by a loss of intravascular volume that impairs the delivery of oxygen to the tissues.  Now although there are a variety of causes of hemorrhagic shock, the most common is trauma and you can see that here on this slide.  In the United States, it's estimated that about 50,000 people lose their life to trauma and bleeding each year and globally this is just under 1.5 million.  The American Trauma Life Support, or ATLS, has developed this classification of hemorrhage to help estimate percent blood loss and guide initial resuscitation efforts.  If a patient presents with class 1 or class 2 hemorrhage, which is typically less than 30% total blood volume loss, then they would potentially be managed initially with crystalloid  based resuscitation. But certainly crystalloids have a role across all of these different classifications to expand intravascular volume and increase tissue perfusion.  When we talk about crystalloids in the modern era of resuscitation fluids, we now really focus on normal saline versus the balanced crystalloids, being lactated ringers, or LR,  as well as plasmolyte. And there's some key differences to be aware of. Particularly lactated ringer and plasmolyte, the balanced solutions, are far more physiologic.  The electrolyte composition more closely resembles the electrolyte composition of our extracellular fluid.  And you can see that with chloride, ranging anywhere from around 100 to 110, compared to normal saline, which is very hyperchloremic.  In theory, and what has been shown in some studies, is the hyperchloremic metabolic acidosis that NS can cause is actually associated with renal arterial afferent vasoconstriction,  and that can cause acute kidney injury. And so there's more and more research ongoing regarding this, and many, many hospitals are now using these balanced solutions.  The problem with the balanced solutions is the sodium content, particularly in the setting of trauma when you have patients who have traumatic brain injury.  You can see lactated ringers is relatively hypotonic, and there's always a concern about giving that to someone with traumatic brain injury and precipitating cerebral edema.  So the clinical evidence that supports this, many of you, I'm sure, are familiar with the SMART trial.  It was presented here a few years ago and published in the New England Journal of Medicine in 2018.  This was a single-centered, randomized controlled trial done at Vanderbilt that essentially compared the balanced crystalloids versus saline, and they found that exposure to saline  was associated with more acute kidney events within 30 days, 15% versus 14%. When we talk about trauma specifically, about 20% of patients enrolled in this trial did  experience trauma, and when they looked at rates of AKI, there was no difference within that subgroup. So why not use LR based on that in the setting of trauma?  This seems to be very promising. There is a question of safety of LR in trauma, and you can see that here. This was a secondary analysis of the PROMET trial.  Essentially what they did here is they looked at pre-hospital administration, or they looked at a secondary analysis of patients who were exposed to LR versus NS pre-hospital.  And in the subgroup of patients with a head AIS greater than or equal to 3, which represents severe traumatic brain injury, they found that those exposed to lactated ringers had  increased mortality, and it's again thought to be the hypotonic sodium content of LR. So the purpose of the current study that I'm presenting today is to determine if LR reduces  the risk of AKI compared to normal saline for initial resuscitation of hemorrhagic shock from trauma.  The hypothesis is that there's no difference between groups in the development of AKI. Study design, as mentioned, this is a secondary analysis of the Pragmatic Randomized Optimal  Platelet and Plasma Ratio Study, otherwise known as the PROPER trial, which was published in JAMA in 2015.  This trial originally was designed to include severely injured trauma patients predicted to require a massive transfusion at 12 level 1 trauma centers in North America.  In this trial, patients were randomized to a 1 to 1 to 1 versus a 1 to 1 to 2 resuscitation ratio of plasma to platelets to erythrocytes.  And in this study, crystalloids were allowed at the discretion of the treating providers, and they were not standardized. So our trial design is a secondary analysis of this.  We included patients who were exposed to NS or LR as the exclusive crystalloid within a 24-hour period on admission.  We excluded those who died within 24 hours, received colloids, and received multiple types of crystalloids, including hypertonic saline.  And ultimately, we found 50 patients who received LR as their exclusive crystalloid and 133 who received NS.  The primary endpoint was rate of acute kidney injury at 30 days, and secondary endpoints were clinical outcomes, including length of stay in the ICU as well as hospital, duration  of mechanical ventilation, and mortality. For statistical analysis for categorical data, we utilized the chi-square with Yates correction.  Continuous data we analyzed with the student's t-test or rank-sum test, depending on if the data was parametric or nonparametric.  And finally, we did a stepwise logistic regression to identify predictors of acute kidney injury. Baseline characteristics are listed here.  What I will point out is the group of patients who received LR generally received a greater amount of fluid, around 9 liters in the first 24 hours, compared to normal saline, which  was around 7 liters in the first 24 hours. Additionally, patients who received normal saline were more likely to have lower systolic  blood pressures as well as an elevated respiratory rate, suggesting that maybe there was an acuity of illness difference between these groups.  From an injury perspective, those who received LR were more severely injured with an injury severity score, or ISS, of 31, compared to normal saline, which was 25.  And when you looked at where these injuries occurred, the one that is the largest difference is extremity injuries, which occurred in 56% of patients who received LR versus 38% in  the normal saline group. So the primary endpoint of acute kidney injury occurred in 32% who received LR, compared to 19% who received normal saline.  This was not statistically significant, and when we did a stepwise logistic regression controlling for age, gender, ethnicity, blood pressure, pulse, respiratory rate, and injury  patterns, we found exposure to NS was protective for the development of acute kidney injury. Other clinical outcomes are listed here.  There was no difference in hospital or ICU length of stay, duration of mechanical ventilation, or mortality.  So from this study, in hemorrhagic shock from trauma, NS does not appear to increase the risk of acute kidney injury.  And when thinking of why that might be the case, I think I have two potential explanations. First, in hemorrhagic shock, crystalloids have a role to expand intravascular volume,  but clearly blood product resuscitation is the cornerstone towards management. When you look at current guidelines, such as the European guidelines on management of  traumatic hemorrhage, they recommend crystalloids in a very restricted way, generally less than 1.5 liters is the example listed.  And although patients in this study did receive more than that, the administration of crystalloid  fluids for this shock state is very different than that of sepsis, where patients may require large volumes of crystalloid resuscitation.  Additionally, there may be a physiologic rationale. Not only are these patients having tissue perfusion problems, but they're also losing  their hemoglobin and oxygen carrying capacity. And so the rates of acute kidney injury may be more influenced by that and more influenced  by the resuscitation efforts of blood products. Taking this into account that LR has been shown to worsen outcomes in traumatic brain  injury, NS may still be the preferred crystalloid for initial resuscitation of hemorrhagic shock from trauma.  Certainly, balanced solutions may be preferred when brain injury is excluded. In that type of setting, that data would be largely driven on the randomized control trial,  which was SMART, the SMART trial, which does require validation. This study has a number of limitations.  It is a secondary review of randomized clinical trial that was specifically not designed to evaluate acute kidney injury as the primary endpoint.  Additionally, the administration of crystalloid was dependent on the treating physicians, and there was no clearly defined protocol. So this limits the generalizability.  Additionally, there's an imbalance between groups regarding size, a sample size, the number of injuries, as well as potential risks for acuity of illness.  Sample size was limited for the power of the logistic regression. And then finally, you may be wondering, plasma light has a sodium content that is around  140, so it would potentially be safer in the setting of traumatic brain injury. But I don't find that plasma light is used widely in practice, and it was not used in  the proper trial to be able to evaluate for that. So in conclusion, in this study, normal saline was not associated with increased risk of  acute kidney injury and hemorrhagic shock from trauma. Due to safety concerns with LR, NS may still be the preferred initial crystalloid for resuscitation and trauma.  Once you identify and diagnose injuries, then you can further tailor that fluid. And then further prospective studies are needed regarding normal saline compared to plasma light.  Overall, normal saline is the most studied crystalloid fluid, and that's particularly true in trauma. So here are my co-investigators.  I'd like to thank you for your time and open it up to any questions.  Yeah, and I do think that, although I am a believer in the risk for acute kidney injury with normal saline, I do think that I've seen so many single center trials fail validation  trials, and due to the safety concerns, I find our trauma group is hesitant to fully embrace LR.  Yeah, it's a great question. So the question was if we looked at the time of onset of acute kidney injury and  Unfortunately because it was a secondary analysis the proper trial  They didn't report that data in their database. And so we weren't able to evaluate time onset of acute kidney injury  Yeah. That's a great question. So the question was the amount of crystalloid that patients received in this trial versus the SMART trial.  So in this trial, patients received between seven to nine liters, which is actually kind of more than you would want in the setting of an acute trauma.  Certainly you're at risk for dilutional coagulopathy and the risks that crystalloids bring. In the SMART trial, the average volume of fluid in that trial was like one liter, which  is not a lot. And they still found a significant difference. And a lot of that seemed to be driven by the sepsis group.  So there was like a mortality reduction in the LR portion if they had sepsis. And so we did have a substantial amount of fluid, even relative to the SMART trial, and  we still did not find that NS seemed to cause acute kidney injury.  Yeah. Thank you.  Hi. I'm Elizabeth Killeen. I'm from Seattle Children's Hospital, and I'll be presenting on Prediction of Pediatric Critical Care Resource Utilization for Disaster Triage.  To start with some background about resource use during disasters, we can expect that in the event of a major epidemic or mass casualty event,  that there will be limited ICU resources available to care for a surge of critically ill patients.  Typically, up to about 30 percent of hospitalized disaster victims require intensive care. While that proportion is similar between adults and children, the resources available per  capita for children at baseline are much fewer. The Pediatric Emergency Mass Critical Care Task Force has recommended that hospitals that normally  provide pediatric critical care should be able to triple their normal PICU capacity for at least 10 days without external assistance. However, most hospitals are unable to stockpile  enough resources such as mechanical ventilators to be able to satisfy those guidelines. Unless there's really a need for an ethically sounds,  well-validated, and widely accepted approach to allocation of scarce resources for children during disasters.  Most adult protocols are based on the SOFA score and preferentially direct resources towards patients with moderate scores who will likely recover if given that resource.  However, only a very small proportion of children would ever be expected to reach a mortality probability threshold in  any triage protocol that would suggest limitation of resources. Thus, this task force has recommended that a pediatric protocol be developed that incorporates expected duration of  critical care needs as well as mortality.  Thus, we aimed to determine whether existing pediatric illness severity scores can be used to predict a patient's need for  and duration of their pediatric critical care resource use. Thus, that could be incorporated into an ultimate triage algorithm that includes both mortality and expected duration of  resources for use during a disaster. So to accomplish this, we evaluated all patients admitted to the PICU at Seattle Children's Hospital for  at least 12 hours over a two-year period. We included patients regardless of their diagnosis, reason for admission, or whether their admission was a repeat PICU admission.  We excluded patients who were transferred from another ICU or who left against medical advice in order to establish their true duration of need for that resource.  With this, we ended up with a cohort of just over 3,200 patients.  We compared three illness severity or organ dysfunction scores, the PRISM, the PILOD, and the PIM, and evaluated five different resources,  mechanical ventilation, non-invasive positive pressure ventilation, vasoactive infusions, CRT, and ECMO, as well as length of stay.  We developed models that uses exposures, the admission PRISM at 12 hours, admission PIM score at one hour, and the PILOD score obtained at 6, 12, and 24 hours.  Then our outcomes were both the need for and total duration of any resource as well as each individual resource within three, seven, and 14 days of admission.  So for example, if a patient was mechanically ventilated from days five through 10, their three-day resource utilization would be zero, their seven-day use would be two,  and their 14-day use would be five days. We had that data down to the minute, not by the full day. We included only covariance that were easily obtained at the time of  admission for use in a triage type algorithm, and we intentionally did not include diagnosis in order to keep the models as flexible as possible.  We made the models using logistic and linear regression, and evaluated the fit of the models based on the area under the ROC curve for need for each resource,  and adjusted R-squared values for the duration of each resource. Because this was a predictive analysis, we developed the models in a random two-thirds of our cohort,  and then tested them in the remaining one-third. We then performed a secondary test of fit in patients who are admitted during weeks of a respiratory surge, so to speak.  In order to evaluate model performance during times with a surge of a single illness type, to best approximate what a disaster situation might look like.  Here's an example of a graph of how that looks at our institution, with periodic surges in viral respiratory disease during the two-year time period.  We considered all patients who are admitted during the weeks in which more than 25 percent of the census tested positive for  a viral respiratory illness as our respiratory season cohort.  For a graphical representation of this model development process, we started with developing base models comparing each exposure with each resource within seven days of admission.  Again, that's need for, it's a binary, or duration of as a continuous outcome. We then selected the single best exposure that had the best fit, and added various combinations of  covariates to develop final models, again, within seven days of admission. We then tested those final models in our validation cohort, within three days, seven days,  and 14 days of admission, and tested them within that respiratory season cohort as well. So what did we find? The illness severity and  organ dysfunction scores that we used had overall low median values, but quite large ranges. Overall, just under half of our cohort  needed any of the resources within the first week of admission, including about 25 percent who required mechanical ventilation. The median length of stay was two days,  and mortality was 2.1 percent. So to start with the base models. This just shows each of our exposures with any resource,  and we did the same process for every single individual resource as well. But what we found is that PILOD calculated at 12 hours after admission,  had the best fitting model for need for the resource, and the second best after 24-hour PILOD for the duration of the resource.  The results were very similar for mechanical ventilation. Overall, all exposures had less goodness of fit for the other resources evaluated.  Non-invasive ventilation had such poor fit, there was no correlation at all between any of the illness severity scores and the need for a duration of non-invasive ventilation,  that we actually took that out of our final models. So based on the balancing the best of fit with the utility during a disaster type situation and need to use it in triage,  we selected the PILOD obtained at 12 hours as our single best exposure.  Then we used that and added covariates for our final model development. For the need for each resource, the final model with the best fit was the 12-hour PILOD plus age,  pediatric overall performance category score, PCPC, the pediatric cerebral performance category score, need for chronic mechanical ventilation, and postoperative status.  And with that, we had an area under the ROC curve of 0.8722. When we tested this in the validation cohort,  what we found is that the area under the ROC curve actually increased up to 0.8808 with a positive predictive value of just about 80% and a negative predictive value of 72%.  And it performed similarly in the respiratory season cohort. This graph here shows that the predicted probability  of requiring any resource increased steadily with higher PILOD scores up to a threshold effect at about a PILOD score of 10 with greater  than 95% predicted probability at that level. For mechanical ventilation, the graph looks almost identical with actually higher predictive accuracy than for any resource  with a positive predictive value of 80% and a negative predictive value of almost 90%. This is the graphs for the other resources evaluated.  They overall have slightly different shapes and are less well-fitting than any resource or mechanical ventilation.  Now, if we look at the final model development for duration of each resource, the same set of covariates ended up included as the best-fitting model.  And with that, our adjusted R-square value is 0.2993. When we tested that in the validation cohort, we ended up with better accuracy  in both the validation cohort here and the respiratory cohort. And the model was able to accurately predict whether patients would require  at least 12 hours of any resource for two-thirds of the cohort and at least 24 hours of a resource for 80%. This graph demonstrates the mean predicted duration  within the first seven days of admission. And what you can see is that it steadily increases to a peak of 18,  a POD score of 18 before starting to decrease likely due to mortality in that group. And once again, the mechanical ventilation graph looks very similar  with even higher predictive accuracy with 12-hour accuracy of 77% and 24-hour accuracy of 85%. For all of these analyses, the three-day window had higher predictive accuracy  and the 14-day window had lower predictive accuracy for all resources and both need-for and duration of resources.  The models for the other durations of resources similarly to need-for had less goodness of fit with the ECMO and length of stay graphs showing somewhat similar patterns  of an increase followed by a decrease. And interestingly, the vasoactive infusions just continuing to increase with increasing POD score.  So, in conclusion, there are several limitations to this work including that it was only at a single center and only with internal validation at this point  and that our exposures were existing scores rather than developing new scores that might be included elements that are more predictive when combined together  than any of these existing scores.  However, we did find that in at least non-crisis situations, we can relatively well predict PICU resource utilization using these existing scores  for resources in general and in particular for mechanical ventilation. And it seems like a PLOD score calculated at 12 hours after admission balances the predictive accuracy  with practical use during triage decision making. This type of algorithm could have both individual and facility level utility.  On an individual level, it could be used to help determine whether a given patient should have a particular resource provided or withheld.  Whereas on a facility level, it could be used in a regional sense to allocate resources to hospitals based on their current ICU populations such as ventilators.  But obviously, any final algorithm will need external validation, inclusion of likelihood of survival as well as resource utilization  and extensive consideration from an ethical standpoint before adoption into practice. I'd like to thank my collaborators on this project and open it up for questions.  Thank you.

Video Summary

The study aimed to develop a triage algorithm for pediatric patients during disasters by assessing existing pediatric illness severity scores and their ability to predict resource utilization. The study included over 3,200 patients admitted to the pediatric intensive care unit (PICU) at Seattle Children's Hospital. Three illness severity scores were evaluated, and five resources were assessed, including mechanical ventilation, non-invasive positive pressure ventilation, vasoactive infusions, continuous renal replacement therapy (CRRT), and extracorporeal membrane oxygenation (ECMO).<br /><br />The results showed that about 50% of patients required any resource within the first week, with mechanical ventilation being the most common resource needed. The median length of stay was two days, and the mortality rate was 2.1%. The PILOD score at 12 hours had the best fit for predicting resource utilization, while the PILOD score at 24 hours was the second best for predicting resource duration.<br /><br />The final models developed for predicting resource utilization included the PILOD score at 12 hours, age, pediatric overall performance category score, need for chronic mechanical ventilation, and postoperative status. These models had good accuracy in predicting resource utilization.<br /><br />The study concluded that existing pediatric illness severity scores can be used to predict resource utilization in pediatric disaster triage. However, further validation and ethical considerations are needed before implementing these algorithms in practice.

Asset Subtitle

Professional Development and Education, Sepsis, 2020

Asset Caption

The Society of Critical Care Medicine's Critical Care Congress features internationally renowned faculty and content sessions highlighting the most up-to-date, evidence-based developments in critical care medicine. This is a presentation from the 2020 Critical Care Congress held February 16-19, 2020, in Orlando, Florida, USA.

57-Differences in Advance Practice Provider Residency Length: 12-Month Versus 6-Month Training

Aimee M. Abide

58-Prevalence of Violence Experienced by ICU Providers: A National Survey

Yves Ouellette

59-Unacceptable Pseudohyperkalemia due to Sample Hemolysis in Emergency Medicine Point-of-Care Testing

Matthew O'Hara

60-Reexploring Goals of Care in Patients Surviving Critical Illness

Tammy L Eaton

64-Factors Predicting Kidney Transplant Recipient Admission to the ICU: A 10-Year Single-Center Study

Nitin Abrol

62-Impact of Gender and Ethnicity on Outcomes in Patients With Sepsis and AKI

Jean-Sebastien Rachoin

63-Safety of Saline Versus Lactated Ringer Solution in Hemorrhagic Shock From Trauma: PROPPR Trial

Mitchell Daley

61-Prediction of Pediatric Critical Care Resource Utilization for Disaster Triage

Elizabeth Y. Killien

Meta Tag

Content Type Presentation

Knowledge Area Professional Development and Education

Knowledge Area Sepsis

Knowledge Level Advanced

Membership Level Select

Membership Level Professional

Membership Level Associate

Tag Fellowship Training

Tag Shock

Year 2020

Keywords

triage algorithm

pediatric patients

disasters

pediatric illness severity scores

resource utilization

PICU

mechanical ventilation

CRRT

ECMO

PILOD score

predicting resource utilization

models

validation

ethical considerations

	Society of Critical Care Medicine 500 Midway Drive Mount Prospect, IL 60056 USA Phone: +1 847 827-6888 Fax: +1 847 439-7226 Email: support@sccm.org	Contact Us About SCCM Newsroom Advertising & Sponsorship DONATE	MySCCM LearnICU Patients & Families Surviving Sepsis Campaign Critical Care Societies Collaborative	GET OUR NEWSLETTER
© Society of Critical Care Medicine. All rights reserved. \| Privacy Statement \| Terms & Conditions The Society of Critical Care Medicine, SCCM, and Critical Care Congress are registered trademarks of the Society of Critical Care Medicine.