SCCM Resource Library-Star Research Presentations: Epidemiology

Star Research Presentations: Epidemiology

Video Transcription

Good afternoon, everyone, and thanks for the opportunity to present today. We'll be discussing whether minority race ethnicity children are proportionately represented  in critical care and emergency research. I don't have any conflicts of interest to report. My funding is listed here, and I'll list it more on the acknowledgment slide.  So, while children of minority race and ethnicity may suffer a higher burden of mortality and morbidity from critical illness, it's challenging  to enroll acutely ill children in critical care and emergency research. Under-enrollment, however, threatens study validity and generalizability, risks an unfair distribution  of benefits and burdens of research, and threatens premature study termination. Over-enrollment, however, is equally problematic  as it risks exploitation of some groups with exposure to risks at the expense of others. And there's prior conflicting data on the enrollment  of children into emergency care and critical care research. In one large study in the general population, it was found that African-American children were  overrepresented across all research, while white children were underrepresented. In contrast, in a review of the RESTORE study data, it was shown that non-Hispanic black children  and Hispanic children were underrepresented in that trial when compared to white children. Enrollment into biorepository or precision medicine-based studies,  in particular, bears some importance as we move towards more precision medicine-based initiatives for critical care. To that end, we conducted a study looking at enrollment  into a pediatric critical care biorepository at our institution. This study enrolled children into a biorepository using a single-page opt-in consent form.  And we evaluated sociodemographic factors, including race and ethnicity, that were associated with missed opportunities for enrollment. There were two time points  at which there would potentially be missed opportunities to enroll. The first was whether the families returned to the completed consent form, which we constituted a missed approach  if we never got their response. And the second was their response on the consent form, whether they accepted or refused. This slide shows the odds of missed enrollment due  to incomplete consents by race and ethnicity. On the y-axis is the odds of missed enrollment, and on the x-axis is race. For odds higher than one, this reflects a higher odds  of a missed enrollment for this population. You can see that all non-white race ethnicities had a higher odds of missed enrollment compared to white race.  When we looked at consent refusal, we found a similar trend with all individuals of non-white, non-Hispanic white race ethnicity having higher odds of missed enrollment.  Based on the previous literature in this data, we then hypothesized that acutely ill children of minority race and ethnicity are underrepresented  in some types of research and overrepresented in others. We undertook to study this in this particular project using pediatric and emergency critical care research networks,  comparing participation in studies to institution-based population targets. We looked at eight studies across two large critical care research networks,  the PCARN Research Network and the CAPCORN Network. The purple studies here reflect critical care. The blue studies are emergency studies. There were more than 60,000 participants  across these eight studies. You can see that there are a variety of both interventional and observational studies and a variety of whether  or not consent was required across the studies. And this will become important as we look at the data further. We characterized participation using what we called a  participation proportion. This was the number enrolled in the study by race and ethnicity defined by the available denominator.  Before I proceed to the data, I want to just explain a moment about denominator complexity. So we looked at this data retrospectively using publicly available data sets.  That limited the denominator data that we had available to us. You can see there are a variety of ways you can look at how many children are able to participate in research,  from the general population down to the most specific, those who are ideally the ones you would want to look at who are eligible for the study.  Unfortunately, we did not have this granularity of data available to us. So instead, we were able to look at the general population at pediatric research institutions,  network institutions, and I'm not going to share this data today, but we also will look at it by institution based on the diagnosis of interest in the particular study.  Today we'll focus on the network institution data. This data was obtained for the denominators using the core data registries for the PCORN and CAPCORN research networks.  We analyzed the data by looking at the population distributions by race and ethnicity in the numerator or the study populations, comparing those  to the populations in the denominator or the research network institutions. We assessed the comparison between these population distributions using a chi squared goodness of fit.  If there was a significant difference in the population distribution, we then applied one sample test of proportions by race so that we could identify  which race skewed the population distribution. This slide shows the total ED studies that we looked at. On the y-axis is race, on the x-axis is the participation proportion.  The hatched bars show the study participation, while the black bars show the institutional participation. You can see across all four emergency department studies,  the chi squared goodness of fit test was significant, indicating a difference in distribution across all of these studies. I'll now show you how that varied by race and ethnicity.  So for ED studies, on these following slides, we're going to look at the difference of proportions between the study and institutional population by race. The x-axis shows race.  The y-axis shows the difference between populations. A positive deflection from zero means that there were more participants in the study  relative to the institution or over-representation, while a negative deflection from zero reflects the opposite or under-representation.  Across ED studies, we found that non-Hispanic blacks and those reported as other race and ethnicity were under-represented. Meanwhile, Asian Pacific Islanders  and non-Hispanic whites were over-represented. This slide shows similar data for the Capcorn Research Network. We can see, again, across all four studies,  there were statistically significant differences in the population distributions between the study participation proportion and institutional representation.  Breaking down the critical care studies, we found over-representation of Asian Pacific Islanders and non-Hispanic blacks in critical care studies and under-representation  of the non-Hispanic white population and other population. We next looked at a few study characteristics to see how these varied across both research networks.  In observational studies across both networks, we found under-representation of black and other populations and over-representation of Asian Pacific Islander and white populations.  In intervention studies, however, we found the opposite. For African American children here, we found over-representation across intervention studies  with, again, under-representation of the other population. Finally, we looked at studies with an exception from informed consent. We found this particular population of interest,  as this reflects study team interactions that don't have to do about whether a family is willing to participate since consent was not required for participation in these studies.  Across these studies, we found that blacks, Hispanics, and others were under-represented in exception from informed consent, while Asian Pacific Islanders  and white populations were over-represented. And this may reflect how we differentially approach these populations for research. So, in summary, we found that participation  by minority children remains still incompletely characterized, but appears disproportionate to the population of pediatric research network institutions.  Summarize the data across all race ethnicity groups here that we just reviewed. Importantly, we found both over and under-representation,  and as I mentioned, these are both ethically problematic and deserve further investigation. We know that knowledge gap remains. To begin, the mechanisms underlying disparities  enrollment are really still poorly understood in these acute care settings. Moreover, socioeconomic characteristics beyond race and ethnicity are important,  but are infrequently collected and reported. And race and ethnicity often acts as a proxy for these factors, which might have differential influence  in how we can change participation. So, our next steps will be, again, to continue with our retrospective data to assess some of these mechanisms  that I've just described for the data that we have available retrospectively. We also anticipate a prospective evaluation to better characterize where disparities exist in order  to identify targets for intervention. With that, I'd like to acknowledge my funding support, my K-12 advisory committee, those who participated  in the study that supported this data, and I'm available to take questions. Thank you.  So, going back to the interventional studies, the two interventional studies were one, a  study in bronchiolitis, and that study looked at the use of oral dexamethasone in that study. And the second was the use of steroids in critical illness.  So they were both medication drug studies. These were all multicenter studies. So we analyzed the aggregate data across the different research institutions.  Unfortunately with the public use data sets, the sites submit with maintenance of anonymity. So we are not able to assess that retrospectively.  We are comparing also to the FIS database as a second denominator set, just to try to validate the data. And we're restricting by looking at diagnoses as well.  We do have in the core data registry the denominator and ICD-9 codes. So we're in the process of looking at that to try to get a little bit more granularity to the data.  Craig Smith from Lurie Children's Hospital. So what's the answer then? Do we have to create quotas for research protocols that are making us cognizant of  this because of our own biases that we're bringing to the table? Do we have to concomitantly educate minority families about the importance of representing  and participating in research? It seems like a complicated web that has deficiencies and knowledge gaps on both the researcher side and the enrollee side. Yes.  Thank you for that question. So I think it's a combination of things.  I think one is figuring out how we identify our targets for enrollment. I think general population targets are not necessarily the best target depending on what you're researching.  So I think that is one question to address. The second is, as you mentioned, how do we understand the rich complexity that goes into decisions for people to participate in research?  So there's a logic model here that we've envisioned looking at enrollment and retention, different  influencers, which include the participant institution, the study team, the parent, drivers of those different influencers, and then the factors impacting them.  Just to take, for example, the parent, you can see the multitude of factors that might overlie a single decision to participate or not in research.  And then if you add on to this all of the other factors in terms of where trials exist, how study teams interact with families, it's really a complex process to understand how  we can better target our interventions. And we're working to understand how all of these factors play together to develop what  we're calling an adaptive engagement model, where we can assess these factors and vary our enrollment approaches based on the ones that are of highest priority.  So those are going to be sort of the next stages of the research as we move forward. Thank you very much. Bye-bye.  All right, thank you everyone. My name is Colleen Badke. I'm also from Lurie Children's Hospital and I'll be talking to you about heart rate variability  and its association with organ dysfunction and mortality. Multiple organ dysfunction syndrome or MODS is defined as dysfunction of two or more major  organ systems in a critically ill patient. It occurs frequently and is a common pathway for morbidity and mortality in this population.  However, as shown in the graph on the right, 85% of MODS is already present on day one of ICU admission or at recognition of critical illness.  And therefore, MODS is a limited outcome measure to discriminate which critically ill children are at highest risk for morbidity and mortality.  New or progressive MODS, NP-MODS, is an alternative outcome measure for critically ill children.  New MODS describes patients with zero or one organ dysfunction at admission who subsequently develop two or more organ dysfunctions.  While progressive MODS describes patients with MODS at admission who go on to develop  at least one additional concurrent organ dysfunction or worsening of an existing organ dysfunction.  In comparison to the traditional outcome measure of MODS, NP-MODS is more strongly associated with morbidity and mortality in critically ill children.  However, early identification of critically ill children who are at highest risk of NP-MODS and at highest risk of death is a challenge in the PICU.  Measuring the autonomic nervous system may be one way to identify these children. The autonomic nervous system regulates functions in nearly every organ system as demonstrated  in the schematic on the right. It maintains homeostasis when confronted with stress such as injury or infection.  However, autonomic nervous system dysregulation, ANSD, can occur in critical illness. One way to measure the autonomic nervous system is through heart rate variability or the normal  beat to beat changes in heart rate as heart rate is controlled by the sympathetic and parasympathetic systems.  When autonomic nervous system dysregulation occurs, there is imbalance to the autonomic input to many organ systems including the heart.  Increased heart rate variability, a marker of autonomic nervous system dysregulation, has been shown to be inversely correlated with MODS and with worse long-term outcomes.  So why should we look at heart rate variability? Heart rate variability is innovative. It's characterized by invasive, continuous assessment of autonomic nervous system function  to determine which critically ill children are at highest risk of developing NP-MODS. It has not been previously studied as a predictor of NP-MODS in critically ill children.  But identifying changes in heart rate variability may lead to earlier identification of children  at risk as some studies show that heart rate variability can decrease up to 18 hours prior to the diagnosis of shock.  And lastly, this could have therapeutic and prognostic implications, perhaps leading to earlier initiation of therapies which could in turn lead to fewer organ dysfunctions and  decrease mortality. Therefore, the objective of this research was to determine whether autonomic nervous  system dysregulation measured via heart rate variability is associated with NP-MODS or mortality in critically ill children.  This was a retrospective study of pediatric patients admitted to our PICU from January 2012 to December 2016.  Patients were included in the study if they had adequate bedside data monitor data recordings of their heart rate for at least the first 24 hours of PICU admission.  Heart rate variability was measured using the integer heart rate variability or HRVI which takes the heart rate measured every one second and then calculates the standard  deviation over five minute windows. Since heart rate and heart rate variability vary with age, we calculated age normalized heart rate variability integer values.  The final 24 hours of PICU stay were used as a proxy for homeostasis within this population. This was a reference to calculate median normalized HRVI measurements.  We then stratified by age to adjust for baseline age dependent variability in heart rate.  The age normalized measurements were then used to develop a heart rate variability dysfunction score.  We calculated this by taking the inverse of the normalized values multiplying by 10 and truncating at zero.  So for an example, if the normalized HRVI was minus one, we took the inverse becoming plus one and multiplied by 10 so that the HRVD score would be 10.  We felt this would be easier to compare patients. Such that a higher HRVD score was equivalent to lower heart rate variability, both of which  were associated with worse autonomic dysregulation. We assessed two outcome measures in this study.  NPMODs was computed using the Pediatric Sequential Organ Failure Assessment or PSOFA score. An increase in PSOFA of greater than or equal to four points was considered positive for  NPMODs. And patients were assessed for NPMODs after the first 24 hours of admission through day seven of PICU admission. Mortality was defined as 28-day hospital mortality.  We examined the HRVD scores in the first 12 and 24 hours after admission. And these included median, maximum, and standard deviations.  We compared these measures between patients with and without NPMODs and between those who died versus survived. NWTU tests were used to perform univariate analyses.  And we also performed logistic regression to adjust for confounders. There were 8,922 patients admitted to our PICU during the time period.  A little over 5,000 of them had adequate HRV data. The patients on the left are those who developed NPMODs.  You can see 215 patients, or about 4% of our population, developed NPMODs after day one, of which 75 of those patients, about a third, died.  The patients without NPMODs are on the right. And only 34 children, or 1% of that population, died by 28 days. This table displays some demographic data of our population.  The second and third columns show demographics for patients with and without NPMODs. And the fourth and fifth columns show the demographics between non-survivors and survivors.  Over 50% of patients with NPMODs were mechanically ventilated. And 60% of patients who died were on mechanical ventilation. 17% of patients with NPMODs were on catecholamines.  And 39% of patients who died received catecholamines. For patients with NPMODs, the most common organ dysfunctions were respiratory, neurologic, and hematologic.  And there was a similar distribution for patients who died. The severity of illness calculated using the PRISM-3 score was high for patients with NPMODs  and even higher for those who died. In the entire population, the median 24-hour HRVD score was 1.1, with an interquartile range of 0 to 4.3.  These box plots display the distribution data for this most significant measure that we found, the 24-hour median HRVD score.  And on the y-axis is the median with interquartile range. The blue boxes here represent NPMODs.  Patients who developed NPMODs had a median HRVD score of 4.7, while those without NPMODs had a median score of 1.1. And the red boxes represent survival.  The median HRVD score for patients who died was 7.1 versus 1.1 for those patients who survived.  There was a statistically significant difference in the distribution between patients with and without NPMODs and between those who survived versus died.  This ROC curve displays the discrimination of NPMODs based on that 24-hour median HRVD score. The area under the curve was 0.67.  And this one shows the ROC curve based on discrimination of death with an area under the curve of 0.79.  Since the area under the curve of NPMODs was only 0.67, we decided to look at the areas under the curve for individual organ system dysfunction.  Here organ dysfunction was defined as an increase in PSOFA of two or more points within the first week of PICU admission.  We found that cardiovascular dysfunction was the most predictive of a higher 24-hour median HRVD score, followed by renal, liver, and hematologic.  This table displays the adjusted odds ratios after controlling for confounders of the autonomic nervous system.  For every one-point increase in the unadjusted 24-hour median HRVD score, we see a 25% increase in the odds of developing NPMODs.  This remains significant after controlling for severity of illness using PRISM-3, mechanical  ventilation, and exposure to medications like catecholamines, dexmedetomidine, or albuterol. We observed a 46% increase in the odds of death for every one-point increase in the  unadjusted 24-hour median HRVD score. And again, this remains significant after controlling for potential autonomic nervous system confounders.  There are several limitations to our study. Our patient population was heterogeneous, and we have not yet performed any subgroup analyses based on admission diagnosis.  In addition, we have not yet performed external validation of HRVD as a prognostic physiologic biomarker.  And lastly, it's unclear if using HRVD as an additional vital sign at the bedside is practical, if it improves clinical decision-making, and most importantly, if it improves patient  outcomes. The ICU is full of massive amounts of data, most of which we don't utilize on a daily basis at the bedside.  Heart rate variability, or the HRVD score, is an age-normalized continuous physiomarker that reflects autonomic nervous system function.  Its degrees of change have the power to predict outcomes in critically ill children, such as NPMODs and mortality, independent of many confounders.  An HRVD may be able to be used as an additional vital sign, as risk stratification, or as  a predictive marker to earlier identify children who are most at risk for NPMODs or mortality.  Our future research goals are to prospectively collect other measures of autonomic nervous system function in critically ill children and assess their prognostic value.  We hope to establish new modalities to assess autonomic nervous system dysregulation in critically ill children, which may inform our ability to predict which children are  at highest risk for morbidity and mortality. I'd like to thank my research team, and I'd be happy to take any questions.  Burkett from Stanford. Just was wondering, you guys looked at median for 24 hours prior to event.  Is there any chance of looking at trends, changes in heart rate variability within X amount of time as also a predictor for new mods? Yeah, that's a great question.  So we looked at a few different measures. The median was the one that stood out the most in terms of its discrimination.  But we looked at maximum HRVD scores within different time periods, the first 12 hours and the first 24 hours of admission.  And we also looked at the change, excuse me, or delta in the first 12 and 24 hours after admission. We actually found the least correlation with the change, surprisingly.  So, when we split out new MODS versus progressive MODS in the first 24 hours, about 50% of the patients with NPMODS had new MODS and about 50% had progressive MODS.  However, the signal for mortality was much stronger for progressive MODS. About 60% of those patients ended up dying versus only 8% for those with new MODS.  Heart rate variability itself has been used in adult medicine. The HRVD score itself is something we just calculated, but essentially the outcome is the same.  And in adults it also is inversely correlated to multiple organ dysfunction and to death. I haven't seen any comparisons between adults versus pediatrics  to see if it's more predictive in one population versus the other. Excuse me.  Thank you.  Can everyone hear me okay? All right, thank you. So my name is Andy Janislaw. I'm a new faculty member in pediatric critical care at Columbia University in New York,  and thank you for giving me the opportunity to speak with you guys today. So this is actually the culmination of the work that I did as a fellow that we're working  on publishing right now. I've been interested for a long time in what happens to our children in the ICU after they recover, for a long time  after they recover, potentially years and years down the line. So I think that the adult world of critical care has been kind of ahead of the curve on this for a while,  although as we saw from one of the plenary addresses today, we're catching up in pediatrics. The study that's on the left is looking at adults  with shock and or mechanical ventilation. This is probably something that was published when I was a resident that really opened my eyes to this issue.  And they looked at essentially a healthy population, or at least healthy in terms of cognitive function, and they found that up to a year after they were discharged  from the ICU, something between a third to a quarter of those patients had a degree of cognitive impairment similar in scale to traumatic brain injury.  The study that's on the right was a Danish epidemiologic study by Hannah Wunsch's group where they looked at unplanned mechanical ventilation admissions  to the ICU, and they found a significant incidence and hazard of new mental health diagnoses after those patients went home that was persistent.  Also, persistent use of psychotropic medications afterwards, whether or not there are actual diagnoses. We know a lot less about children, partially,  at least largely, because it's harder to study. We do have a sense that mechanical ventilation might be something that actually does matter for them.  The reasons for that, there's sort of myriad reasons. It probably involves sort of a bundle of different exposures that might be bad for cognition, bad for mental health.  Hypoxemia, exposure to sedation, as our anesthesia colleagues have been looking at for a long time, immobility, as well as just degree of illness.  And it's something that I'm interested in looking at, both because it's sort of one of the most common bread and butter things that we do within critical care in general,  and knowing that it's such a common problem is something that's causing our patients to have an issue, or at least associated with an issue. I would like to know about that.  But studying children does come with a number of different problems. First off, you have a baseline problem. There's a lot of heterogeneity in the pediatric intensive care unit.  There's a lot of chronic illness, and there are a lot of patients who have comorbid diagnoses, such as genetic diagnoses that if they wouldn't sort of doom you  to an altered neurodevelopmental outcome, they'd at least predispose you to one. And so you have to make sure to adjust for those types of things,  or at least take those into account.  And really, comorbidity adjustment is sort of what goes into both of these issues, about heterogeneity as well as altered neurodevelopment,  which really leads me to ask the question, and whenever I'm looking at epidemiologic studies or thinking of doing one, I sort of think to myself,  well, what's the patient population I actually do want to study, and what's my actual question? So for me, it's really the healthy kid who comes  into the intensive care unit with a common problem, supposedly recovers, gets mechanically ventilated, supposedly recovers, what happens to them  years down the line, and what's the best way of looking at them? Which is what led us to look at the use of administrative data sets for these patients.  You lose some of the granularity of clinical data, but you do get a long, long period of follow-up, and you can look at sort of more common problems  than you otherwise might be able to. So our hypothesis, and I'll get into the data source in a moment, was that children who receive  mechanical ventilation for respiratory illnesses, meaning common illnesses in the intensive care unit, are going to have more mental health problems,  mental health diagnoses in the years after they leave. The groups that we included, we looked at children with primary respiratory illness, not infection,  and we tried to exclude patients, and I'll go into this in some detail, who had other comorbid diagnoses that we thought would alter their neurodevelopment.  We had two exposure groups. The first was children who were in the ICU and did not receive mechanical ventilation, and the second was children who did receive  mechanical ventilation in the ICU, and we compared them to a general hospitalized cohort. Again, all of these patients who had  primary respiratory illness, no other primary illnesses, and our outcome was mental disorder diagnoses as coded by ICD-9 codes based on the error that we were using.  I should mention also that all of these children who received invasive mechanical ventilation or ICU, it was their first time receiving it. We didn't include children  who had received it multiple times. So the design of the actual study, we used what's called Medicaid max files, so this is files furnished by CMS  that allow you to link at the individual patient level from year to year, so you can form a longitudinal data set, and we identified, ours were from Texas  over the course of about 12 years was the max, and we looked, we identified our outcomes based upon ICD-9 and procedural codes as well as sort of billing codes for accommodation  to show where the patients were actually located. Our study design was as below, where we looked at patients, basically once they had an index admission,  we followed them until either they got a diagnosis or they fell out of the data set, but we had a certain period of baseline that was required to make sure that we knew enough  about their comorbid illnesses going in so we could either exclude them or adjust for it. So the actual outcomes that we looked at,  this is from a previously described sort of composite, and our primary outcome was a composite of this group of mental health diagnoses, so developmental delay, autism, ADHD,  intellectual deficit, as you go down that line, and then as a surrogate, because everyone probably knows, at least anyone who's taking care of children in a general clinic,  not everyone that has a diagnosis gets medicated, so we looked at medication use or medication prescription as a surrogate marker for a mental problem  or for mental health diagnosis.  Comorbidity is a big issue, like I alluded to, and this was a huge part of the work that I'm gonna gloss a little bit just in the interest of time,  but we used Chris Feudner's complex chronic conditions and we looked through them and tried to find ones that we thought were going to very realistically  predispose children to having altered neurodevelopment and exclude those children. The ones that weren't going to necessitate altered neurodevelopment, we included in a model  and we adjusted for them based on propensity scores.  So we started with just over two million admissions to the hospital, not for unique patients, for admissions, and after we whittled out  about a million and a half neonatal admissions, which we were not looking at, as well as other invalid hospitalizations and primary diagnoses,  we were left with 170,000-odd primary respiratory infections or respiratory illnesses. Then we weeded out patients based upon comorbidity or the lack of sufficient information  to exclude comorbidity, and we were left with 115,000 patients of whom 1,400 received invasive mechanical ventilation, about 8,000 were in an ICU without mechanical ventilation,  and just over 100,000 were in the ICU, were in a general hospital.  Then we had to actually compare apples to apples. So we used a propensity score model  where we used sort of a machine learning technique that's called generalized boosted regression with the covariates that are in the chart to your right,  looking at demographics, resource utilization, and then a number of different comorbidity markers to design a propensity score or to estimate propensity scores  to make sure that our matched groups and perform matching were similar to each other. And so this is a busy plot, I apologize, but it's called a love plot.  The horizontal axis is looking at standardized mean differences and the y-axis is looking at all of the different covariates that we included. And what it's essentially showing,  the red plot, the red line is pre-matching and the blue line is post-matching. The takeaway is that anything that's to the left of that dashed line  is less than 0.1 standardized mean difference. So all of the covariates after matching were essentially, you can't say significant, it's not a significant test,  but the differences were relatively trivial.  So onto the actual results, there's a lot of preamble. So this is a Kaplan-Meier curve looking at PICU patients versus the general hospitalized cohort and looking at their incidence  of new mental health diagnoses. And although it is statistically significant by log-rank test, what you can see is there's really not a lot of daylight between those two plots.  Whereas when you compare that to the invasive mechanical ventilation group versus the general hospitalized cohort, statistically significant, you can see a real difference.  I hesitate to draw conclusions as the two lines converge towards the rightward end of the plot, just because there are so few patients at that point  that I think that it's hard to really say much about them. I think an easier way to look at it is probably via hazard ratios. So this is just looking at the hazard  for new mental health diagnosis for the IMV group versus controls  as compared to the PICU group versus controls. And you can see that although they're both significant, the mechanically ventilated group has more illness.  When we looked at psychotropic medication, you see very, very similar results.  One criticism as I was trying to criticize myself for this that you can come up with is it could be that just the patients who are receiving these diagnoses  haven't been diagnosed yet at the time of their hospitalization. And so we looked at children who were older, who were age two or more at the time of hospitalization.  The left plot is just the same plot you saw before rescaled, and the right plot is only the older children. The results were robust. They were still the same thing.  So pulling this back to the individual patient level, which you have to be careful with with that, which you have to be careful with with epidemiologic studies.  But to me, what it means is that our patients who are relatively healthy coming in, first-time mechanical ventilation, are at risk, 43% increased risk  for new mental health diagnoses years and years down the line. That was robust when we looked at psychotropic medications, when we looked at older children. And so to me,  I think that we need to be careful with these children. I don't wanna scare all of their parents and say, yes, your child's gonna have a developmental problem  after you're mechanically ventilated. But I do think that those children merit follow-up.  And I think it's on us, and this study does not begin to address that, to think about what aspects of mechanical ventilation and of PICU care might be modifiable  so that we can figure out  what is it exactly that's associated with this risk, whether or not it's causative, just so we actually know what's going into it. So I want to thank my research team  and everyone at Columbia that's helped me out with this. And I'm very happy to take any questions. Thank you.  Northeastern University, great talk. Two questions. One, do you have any data on delirium in the ICU and whether or not that impacts the hazard ratios that  you're seeing? And two, are there any specific mental disorders that were driving this or was it a generalized increase? So for the delirium  question, I would love to have that. I actually looked. So I looked across this entire data set for all the times delirium was was coded by by billers,  eight times. So it's not, at least in this data set, it's not. It could be that in a data set now that people are more attuned to this that perhaps that's  being more more readily billed and diagnosed. But for this, I can't tell. For the second question about which diagnoses are driving it. So it depends  on how you pair it out. So if you just look at incident diagnoses, meaning which which were in each group, the developmental delays were by far the  most prevalent. And that's probably because this is a very young cohort. I didn't show it in the interest of time, but the patients were tended to be  around less than age one when they were at their index hospitalization. So that was the highest incident diagnosis. But when you actually look at the hazard, as  in like comparing the two groups, intellectual intellectual disability, autism, ADHD, I think were three of the highest individual diagnoses. Things like  depression, schizophrenia, PTSD, those are all much lower. And my hypothesis is that probably because the patients were very young, whether or not that other things go into it.  I'm Jason Falvey, Yale. Great talk. One of the other questions I had is having worked with claims data. I know changes from 1999 all the way up through 2012,  there's been really big changes in how we paid attention to long-term outcomes. I mean, even though it's not documented in PEDS, it's been, you know, there's been a lot more  awareness of these long-term problems. And did you look at the differences across some of these different time periods to see if people that were treated earlier  were maybe under-recognized and now were doing a better job or recognizing or doing a better job just kind of staving off these mental health diagnoses with the later years?  That's a great question. I didn't look at it in any kind of granular way. And I didn't say, okay, well, let's look at the later, the later epic versus the earlier epic  and see kind of what our diagnostic rates were. We did include in the model, I can't imagine you couldn't see in that giant white list of covariates,  the actual era in which the patient was included. So we should have adjusted for that to some extent. But I didn't look piece by piece.  Pradeep Kamath from Emory University, great talk. Is there a way we should look at whether a patient on high-frequency ventilation versus a mode like bi-vent,  will you think that could make a difference? Yeah, I mean, so there's sort of three ways to think about this to me. There's none of it matters.  The patients are doomed to their diagnosis. We just haven't picked it up by the time that happens. Or the invasive mechanical ventilation,  whether it's high-frequency ECMO, et cetera, is a marker for the severity of illness. And that might be what really matters. Or it's something about what we're actually doing  during this that might matter. To me, the answer is probably yes to all three of those. I think they all do matter. Probably, in this dataset,  you can't look at the type of mechanical ventilation that was actually provided. But I suspect that patients who are much sicker are gonna be more predisposed to a bad outcome.  And I think it's borne out. Most of the studies on cognitive impairment in children have been looking at cardiac arrest, where you're much more likely,  your pretest probability of having a problem afterwards is much higher.  I'm Leslie Durbin from Seattle. Do you have any information about what the patients were ventilated for?  My question is whether there's a possibility that they were intubated in some part for a neurologic injury and whether that then drives later psychiatric diagnoses.  Yeah, so we look – so the primary diagnosis is supposed to – caveat, supposed to – just be the main diagnosis for why the patient was actually in the hospital.  There's obviously things that can kind of – there's some cross-contamination there. We did exclude patients who had a pre-existing neurologic diagnosis, so anything that was  like a bad seizure disorder, they were not in the data set. All of the diagnoses, the primary ones, I think the top three for each of the groups  was either RSV, bronchiolitis, NOS, and then for the PICU group was just acute respiratory failure. It's possible, certainly, that it was for a neurologic – for a logic illness.  I think we did as much as we could to weed that out, but there's probably some cross-contamination in there. Okay.  All right, hi, my name is Megan Land. I'm a third year critical care fellow at Emory University in Children's Healthcare of Atlanta. And I'm here today to talk to you a little bit  about the opioid crisis and its impact on pediatric patients. There's a QR code in the bottom right. We recently published a manuscript on this same data.  So if you'd like to take a little deeper dive, you can open the camera on your device and your phone will do the rest. So I think we're all aware that the U.S. has a serious problem  with opioids that's been plaguing us since the early 2000s. In 2017, there were actually 70,000 overdoses that were recorded, and 70% of those were due to opioids.  And in 2017, you were also more likely to die from an opioid overdose than you were in a motor vehicle accident on the road. It's pretty alarming. So this continued rapid growth  of the opioid epidemic has placed a significant strain on the healthcare system, and that includes ICUs. While this impact has been pretty well described  within the adult population, it's been less described amongst pediatric patients. There was a really nice study done by Dr. Jason Kane  at the University of Chicago in 2018 that looked at the FIS database in order to identify opioid ingestions. They looked at 2004 to 2015 and demonstrated that PICU admissions  for opioid ingestions were actually increasing at a faster rate than admissions overall. But this data is based on billing data.  And so there remained a knowledge gap regarding clinical impact, the type of drug ingested, the age of the patient, and the reason for ingestion.  In terms of our primary and secondary outcomes, our primary outcome was really to determine if there was an increase in the proportion of children admitted  to PICUs for opioid poisonings from 2005 to 2018. And our secondary objectives were to determine whether there was an increase in any clinical interventions,  ICU-level interventions, and then mortality. In terms of our methods, we used the National Poisoning Database, or the NPDS. And this is a database that is put together  by the poison control centers nationwide. So all of them contribute in real time to this really beautifully constructed database. We looked at a population that was right  around a million patients. We took out the patients that were more than 19 years old. And we were left with about one in four of all the opioid poisonings that were reported  to the poison control center being related to children. We then took out opioids as a secondary exposure. And we also took out patients where the exposure was considered irrelevant.  And we were left with 207,543 kids that we studied within our cohort.  In terms of demographics, we saw that there was a bimodal distribution in terms of age groups, which is pretty similar to other ingestions in the NPDS database.  So we saw about 51% of our patients were in the zero to four toddler preschool age group. And then about 31% were in the teenage 15 to 19 year age group.  We also saw about an even split between sex in our cohort. So in terms of our results, we actually saw that there was a declining proportion of children that were being called  into the poison control centers since about 2010. We saw that the opioids that were primarily ingested were typical opioid medications that are pill based.  So primarily things like hydrocodone, oxycodone, codeine, and tramadol. All of those, so oxycodone, codeine, tramadol were increasing over the epochs.  Hydrocodone was the only one that didn't continue to increase over the period that we studied. So the NPDS defines clinical effects in their database as minor, moderate, or major.  Minor symptoms are considered just that, mild. They're things like cough, drowsiness, and minor heart rate changes. Things that are considered moderate are more pronounced  or systemic, and they're things like desaturations requiring oxygen, a seizure that required treatment, or hypotension that was responsive to fluids or medications.  Major effects were life threatening, so it was things like respiratory arrest requiring intubation, status epilepticus, or cardiac arrest requiring CPR.  What we saw within our cohort was that the moderate effects have risen over the course of the eras that we studied at a pretty high rate,  a significant rate, and we also saw that the major effects in death rose also. In terms of hospitalization, we saw that children being admitted  to the general ward and children being admitted to the ICU rose over the eras, but the trend that was most alarming to us was the children being admitted  to psychiatric hospitals rose at a much faster rate, and mirroring that, the NPDS is able to record intents and saw that suspected suicide  as an intent rose dramatically from 2010 forward. Other intents, like therapeutic error, remained about the same, and misuse and abuse actually downtrended over the eras.  So, in terms of resource utilization for kids in the ICU, about 14% of our, of kids that required an intervention required a PICU level intervention.  So, we saw that mechanical ventilation and that vasopressor use actually remained stable over time. Mechanical ventilation right about 11%,  and vasopressor use right at about 3%, but we did see that two things, naloxone and CPR, increased over the eras. So, naloxone use went from about 42% in the initial epoch  to about 50% in the final epoch. This is likely as a result of better education and availability of naloxone in the hands of first responders  and the families of patients that are known opioid users. We also saw that CPR more than doubled, went from about 1.3% in the initial epoch to about 3.1% in the final epoch.  We think that that's related probably a function of the severity of the illness increasing in these kids. So, we also did a multivariate logistic regression to look  at factors that were associated with requiring an ICU intervention. And what we saw were that the synthetic opioids, including fentanyl and methadone,  and then the illicit heroin were the opioids that were associated with requiring a critical care intervention. We also saw that an intentional intent,  so suspected suicide specifically, also put you at a higher odds of requiring an ICU level intervention. In terms of regional trends for pediatric opioid poisonings,  we saw that over the eras, in the 2010 to 2014 era, there were more cases per 100,000 population and then it decreased after that. The Northeast was the least affected, and we saw  that most states that worsened improved, but there were a few that remained the worst and stayed that way over time. So, to summarize a little bit, we know that children being  admitted to the ICU as a result of opioid poisonings are increasing. And we know that these children are requiring more naloxone and CPR than they ever were before.  And we also know that you require, that children with illicit and synthetic opioid ingestions required more PICU interventions, and that suicidal intent also placed you  at a higher risk for requiring those interventions. So, to conclude, I think that this really requires a two-pronged approach.  One is that we need to increase mental health resources for kids to address adolescent suicidality. And secondly, we need to decrease access to opioids  in the hands of pediatric patients by decreasing prescribing and then also getting those that are unused out of the homes. So, once again, that's the QR code if you'd  like to access the manuscript that was just published last month, and I'd be happy to take any questions at this point. Thank you.  So we, those were the opioid pills, so they weren't the things like fentanyl and heroin that are injected, so we saw that it was like the same ones.  It was oxycodone, hydrocodone, and codeine were the ones most associated.  And more of that is in the paper too.  Thanks, that was really great.  I don't know if you had access to this data, but states that have more of an opioid problem,  were they the ones that were having more deaths or more hospitalizations due to opioid use? So we were really curious about that too.  The way that the Poison Control Center reported it out to us was kind of difficult to tell,  but it does seem like states like West Virginia, if you looked at this map,  oh, I messed this up somehow.  So states like West Virginia, we know have a big opioid problem, remain the deepest blue throughout.  The Northeast is known to have a really terrible heroin problem, but within the pediatric population, it wasn't as severe.  Hi, Jim Murray from Indianapolis. Did you do any, I guess you would say,  correction for the decrease in pediatric calls that have been coming into the Poison Center database?  Yeah, so all of our data, the way we reported it out was proportion, because of that exact issue. So we know that there's less calls overall coming into the Poison Control Centers,  and so we tried to just divide it out based on the number of calls that came in. And so everything that I reported out was not raw number, but rather was proportion of calls.  Okay, good. Thanks.  Thanks.  Hi, thank you. My name is Elizabeth Killeen. I'm from Seattle Children's Hospital, and I'll be talking about the components  of health-related quality of life that are most affected following pediatric critical illness.  So to start with some background on health-related quality of life in general,  it measures the impact of someone's health status on a variety of domains, including their physical, mental, emotional, and social functioning.  It's increasingly being used as a comprehensive measure of health outcomes in both observational and interventional studies, and has been identified by both families as well  as providers as the most important outcome to assess after pediatric critical care other than mortality.  Most studies have shown that between a quarter and a third of PICU survivors experience impairments in their overall health-related quality of life in the first few months to years  after their hospitalization.  And that's regardless of the population studied.  But what hasn't been described is which individual elements of an overall health-related quality of life are most affected.  So the aims of this project were twofold. First was to evaluate how frequently patients experienced declines in the individual  questions comprising a health-related quality of life measure, and then to determine the risk factors for deterioration in each of those questions  with the hope that this information could ultimately lead to development of interventions targeted towards those components that are most affected.  So to accomplish this, we conducted a retrospective cohort study of patients admitted to the PICU and CICU at Seattle Children's Hospital over a seven-year period.  Health-related quality of life assessments had been collected on a sample of this population during that time period, mostly via parental proxy report.  Families were asked to complete a baseline assessment of their child's quality of life status based on a recall of the month prior to their hospitalization, and then to complete  assessments at the time of admission and follow up a median of six weeks after discharge.  Oops.  We use these data to assess the change for an individual between their baseline and their follow-up status for each question.  The data that I'm going to show today represent data from the PEDSQL health-related quality  of life measure, which was administered to children without severe developmental disabilities or chronic oncologic conditions.  The PEDSQL has two main versions, one for infants less than two years of age that includes 36 questions, and one for children ages two to 18 years old that includes 23 questions.  And the questions in those two versions don't overlap, and thus we analyze the results separately from those two versions.  But for both, we assessed the highest prevalence of decline for each question between baseline  and follow-up, and then used multivariable Poisson regression to determine the risk factors that were associated with decline in each of those questions.  So what did we find?  Our cohort included 539 patients who had completed all three PEDSQL assessments. And this group differed from our general PICU and CICU population in a few important ways.  They were younger.  They more commonly had normal baseline function.  They were more commonly admitted after scheduled surgeries. But they had similar median prism and scores on length of stay to the rest of the population.  Similarly to prior studies, about a quarter of our cohort experienced a decline in their overall health-related quality of life that was considered clinically significant.  So this first graph is just going to be an overview of all the questions that we evaluated  for the infant scales to start, divided into their general domains represented by different colors.  And they're in descending order of the proportion of patients who experienced a decline in that question.  So first, the questions in the emotional domain were by far the most commonly affected for infants,  followed by physical, then social, and finally cognitive.  Taking a closer look at the questions that were most frequently reported, this is for families who endorsed at least 20% of families endorsed declines in these questions.  By far the most common area of decline was in fussing or crying, with 28% of patients endorsing declines in that area, followed by 25% experiencing difficulty sleeping at night,  and then difficulty with soothing, feeling tired, afraid, or cranky, and difficulty napping.  When we look at the same graph for the generic core scales, the 23 questions there, we found that for the older children and teens, declines in emotional function were the second  most common, whereas physical impairments were by far the most common for this age group,  followed by social and school function. Followed by social and school function.  For the older children and teens, nearly half of the questions were endorsed as declining for at least 20% of the population. This is 11 of the 23 questions.  Most commonly, participating in play or exercising, pain, lifting, keeping up with things that other children can do, and then feeling angry was the emotional area most  commonly affected.  Then when we looked at our multivariable analysis of the different factors that were associated with declines in each questions, we came up with a number of findings.  I'm just going to show you some of the most notable ones here. Let's start with demographic factors for age.  There were no significant associations for the younger group, but in the 2 to 18-year-olds, older age was associated with a higher risk of difficulty with physical activity and  ambulation, and lower risk of experiencing anger with increasing age.  There were some interesting associations that emerged with race.  The families of white infants were more likely to endorse symptoms across a range of areas compared to non-white infants.  Oops, excuse me. While black children, older children and teens experienced more pain compared to white children.  And then finally, I found very interestingly for both infants and older children, not having a complex chronic condition was associated with persistent fear after hospitalization.  For illness severity, there were multiple questions that were associated with illness  increasing illness severity for the infants, but none for the older population.  Whereas for diagnosis, there were no associations for the infants, but multiple for the older children. Just a few of them are represented here.  But most notably, the patients with new oncologic conditions experienced quite a bit more decline or higher risk of decline in multiple questions, and particularly anger and pain.  Let's take a closer look at the diagnoses, starting with the infants.  This is divided up into different diagnosis groups. And the questions, the top five questions that experienced the most decline in each  diagnostic category are represented with a red asterisk, representing questions for which patients in that diagnostic group experienced more common decline than patients who didn't  have that diagnosis.  So a few of the interesting findings that I wanted to point out.  One is that while surgical patients did not endorse, or families of surgical patients did not endorse persistent pain, they did endorse a number of findings, including fussing,  cranking, crankiness, crying, and other symptoms that could represent undiagnosed pain. And then also that fear was a very common area of decline across diagnoses.  For the older children, a few of the notable findings. One is that while the surgical patients experienced multiple physical impairments, as might be  expected, nearly a quarter of surgical patients also endorsed difficulties with attention after their hospitalization.  Whereas patients with infections, which were primarily respiratory infections, very commonly endorsed pain after their hospitalization.  Nearly 40% of patients with congenital heart disease had persistent fear after their hospitalization.  And while the sample size is small, because these are only new oncologic diagnoses, the proportion of patients with these new oncologic diagnoses who experienced declines across  the entire range of symptoms is quite notable.  So in conclusion, there were several limitations to this work, including that we only had a  small sample of the overall ICU population at our institution, with lower medical complexity in general.  Our baseline assessments were based on recall, which is inherent in all of our clinical assessments, were based on recall, which is inherently subject to bias, and we had  both short and variable time to follow up.  But despite this, we did find that the most commonly affected components of health-related quality of life do vary based on both patient and illness-related factors, that patients  with medical diagnoses frequently experienced pain and physical impairment, not just surgical  patients, and that both fear and sleep disturbances were extremely common across ages and diagnoses.  And again, our hope is that identification of some of these principal contributors can be a first step towards development of targeted interventions to actually start improving  patient outcomes after their ICU hospitalizations.  I'd like to thank my collaborators and open it up for questions.  That's a great question. So I think what was interesting in this population is that we definitely found a different distribution  of the general domains by age in our group, again, with emotional domains by far most commonly affected for infants, whereas physical domains were by far most affected for the  older children. And that is, I would imagine, in part based on how the different abilities that patients of those ages are able to manifest problems.  So crying, for example, might be much more common in an infant population in general compared to difficulty walking.  It's obviously something that's mostly going to be seen in an older population. I know that in the adult literature, it obviously includes multiple different areas of impairment.  And like we just heard, even for pediatrics, there's impairments across different types of spectrums, including mental health, physical health.  I believe that a lot of the adult literature shows quite a bit of physical impairment after hospitalization.  And there's increasing awareness of the mental health and emotional components for adults as well that I would imagine track fairly well with our teenage population.  The one thing that I skipped over a little bit too quickly at the beginning was talking about the types of health-related quality of life components in children compared to adults.  Some of the areas that are differently assessed are the school impact, the family, and the  peer elements that are obviously more represented in the pediatric population than adults, which is why the results can't be compared directly between the populations.  Okay, thank you. Okay, I know you are all probably tired. It's 5 o'clock and you've all been here all day. We're going to switch over from pediatric to adult now.  My name is Matthew Dupre. I'm a pharmacist and a PhD candidate at Northeastern University in Boston, Massachusetts. I do not deal with children at all. I deal with adult patients.  And today I'll be talking about an association between haloperidol exposure and mortality in critically ill adults. If you are on Twitter and you're tweeting about this,  you can use my Twitter handle, PharmDupre. Conflicts of interest. So the parent study that we used for secondary data analysis  was funded by the Good Medicine Use Program from the Netherlands national government. And I am supported by an F31 research grant.  So we know that delirium occurs in up to 50% of our ICU patients. The symptoms of delirium include things like agitation, fear, and hallucination.  And we know as clinicians that this is something that is distressing to patients, to their families, but also to our caregivers. Our nurses, our physicians, our pharmacy staff  all experience these symptoms, especially in our hyperactive delirium patients. And this frequently is driving clinicians to use antipsychotic therapy,  regardless of the data that currently exists in the literature. Haloperidol is the antipsychotic that is most commonly used in these situations,  especially for symptomatic control of delirium symptoms. The current evidence that we have does not support the routine use of haloperidol  for either delirium prevention or delirium treatment. There have been several studies that have come out in the last few years, and the PADIS guidelines that were released from SECM  in the last two years also do not support the routine use of haloperidol. But we do have some conflicting data that is out there.  So delirium is hypothesized, or haloperidol is hypothesized, or, excuse me, delirium is hypothesized to include an inflammatory mediated pathway in the brain.  And we have some animal models that demonstrate that haloperidol has potent anti-inflammatory effects through inhibition of cytosine secretion, specifically IL-1,  and enhanced IL-1 receptor antagonist production. We know from some small ICU cohort studies that haloperidol use is associated with lower mortality,  specifically the study by Milbrandt and colleagues, which showed this nice separation between patients that got haloperidol and no haloperidol.  But there are some potential issues with those studies, with residual confounding and dose-related relationships. So the relationship between the use of haloperidol  to treat delirium symptoms in the ICU and mortality still remains unclear. So the objective of my study was to explore the association between treatment haloperidol exposure,  so that is haloperidol for use in patients that have a diagnosis of delirium, and mortality in a population of critically ill adults  who are without delirium at the time of ICU admission. So, as I said, this is a secondary analysis of the REDUCE trial that was published about two years ago in JAMA.  That was a multicenter, randomized controlled trial conducted in 21 ICUs in the Netherlands, which comprised 1,789 critically ill adults  who were free from delirium at the time of admission to the ICU. All these patients were randomized to receive either haloperidol 1-milligram IV,  haloperidol 2-milligrams IV, or placebo three times daily until day 28 ICU discharge or delirium onset. If delirium occurred, patients were then allowed  to be initiated on open-label symptom-driven IV haloperidol via a treatment protocol, where they were started at haloperidol 1-milligram IV three times daily  and increased by a 1-milligram three times daily increment if symptoms persisted up to a maximum of five milligrams IV three times daily or 15 milligrams per day.  Haloperidol was then weaned off when delirium and its symptoms had abated. Use of other antipsychotics for control of these symptoms  was not allowed in the study, and haloperidol use was not allowed in patients when delirium was not present. All patients were managed with established pain, sedation,  and non-pharmacologic delirium reduction protocols  while they were in the ICU, these 21 ICUs. So that brings us to my study. So this is a post-hoc cohort analysis of that reduced trial. We included data from 14 of the 21 Dutch ICUs  that were in the original study. These were ICUs where all delirium and coma assessments were completed. These were done using the CAM ICU and the Richmond Agitation  Sedation Scale, and those were evaluated three times daily in all of these hospitals. This gives us an analytic cohort of 1,495 patients,  or about 84% of the total cohort in the initial RCT. Mortality was evaluated for up to 90 days after ICU admission, and we did statistical analysis using a Cox time-varying hazards  model for both 28-day and 90-day mortality. Our model controlled for days spent with delirium or coma, patient age, baseline severity of illness as assessed  by the APACHE-2 score, presence of sepsis, requirement for mechanical ventilation, and the length of ICU stay. So our cohort, as you can see, is fairly old.  It has a mean age of 66 years. It was primarily male, about 62% male. APACHE-2 scores were, on average, 19.2, lower than we're used to seeing in the U.S.,  but consistent with European standards for ICU admission. Sepsis was present in about two-thirds of patients, and mechanical ventilation was required in three-quarters  of patients in this study. The ICU length of stay was a median of four days, and the predeliric score, which is a score that measures  a patient's risk for developing delirium at the time of ICU admission, was 26, which means about a 26% risk of developing delirium during the ICU stay for those patients.  36% of our patients became delirium positive at some point during their ICU stay. Of those who were delirium positive, their average length of delirium was a median of four days.  62% of the cohort did have a positive coma screen on a day where they did not have delirium, and their time spent with coma was about two days.  When you combine those two, it was a median of about two days spent in either delirium or coma for the patients in the study. Reintubation rates were about 10%,  as well as ICU readmission. So the results of our Cox time-varying analysis show that each additional milligram of haloperidol administered per day decreases the risk  or the hazard of mortality at 28 days by 7.3%, and decreases the hazard of mortality at 90 days by 3.2%. I've also presented the slope coefficient for you,  because we did have that time interaction between haloperidol and time. So we find that as time increases in the ICU, the effectiveness of that haloperidol dose decreases.  So haloperidol doses given later in the ICU stay were not as strongly associated with decreases in mortality. And as you look down the rest of the slide, you can see that all  of the things that we controlled for that are associated with mortality were still associated with increases in mortality. Patients with higher age, patients that spent more time  with either delirium or coma, higher APACHE-2 scores, those patients were more likely to die. We also included their initial randomization arm  as a sensitivity analysis in our study, and we found that even when we control for prophylactic haloperidol that patients received and their random treatment allocation,  our results still held. We still had a 7.3% decrease in mortality at 28 days, and we still had a 3.2% decrease at 90 days. Obviously, our study has some limitations.  We were not able to control for all time-varying factors. Specifically, we did not have daily changing severity of illness with something like SOFA.  We only had that baseline severity of illness with APACHE-2, and we also did not control for competing neurologic risks in a time-varying fashion,  using something like a Markov model. We also could not control for pharmacologic and non-pharmacologic strategies that are known to reduce delirium,  but we did note that there was 88% compliance with the delirium prevention methods in the original randomized control trial.  Also, we don't have any data on delirium subtype and the type and severity of delirium symptoms, so we could not do anything like hyperactive versus hypoactive delirium assessments.  So, in conclusion, haloperidol exposure is associated with a lower 28-day and 90-day mortality in a dose-dependent, time-dependent manner when we look at haloperidol for treatment.  There are questions, is this potentially causal? We are showing temporality. There's a dose-dependent effect here, and based on the data  that we've seen in other studies and in animal models, there's potential biological plausibility here. But we do know that future prospective controlled studies  need to confirm these results. I'd like to thank all of my co-authors, including my collaborators in the Netherlands, as well as those at Northeastern University  who helped me with this. And with that, I will take any questions. Thank you.  Yeah, so the question was we know that 90-day mortality is more related to comorbidities. We did not have that in our model. I would have to go back and double-check our database.  We might have something like a Charleston comorbidity index that we could potentially explore as another sensitivity analysis.  But that's why we think that the 90-day mortality, we're not seeing as strong of an association as we do with 28-day mortality.  So the question was, did we see any differences in our groups that had sepsis? We didn't do a specific sepsis versus non-sepsis analysis, but we did see that patients  that had sepsis were about twice as likely to die. I think it was the hazard ratio was 1.8 in all of our different analyses, and that was fairly consistent.  So we know that sepsis patients were more likely to die. I did not look at an interaction between sepsis and heliperidol, though.  Okay. Well, thank you guys all for coming and giving me the opportunity to present this work.  I am going to be a little bit different because I don't touch patients in a hospital, and I'm not a physician, and I'm not going to talk anything about drugs.  But what I am going to talk about is something that is probably relatively unfamiliar to most of you, it's home healthcare rehabilitation for older ICU survivors.  And it's, you know, from my clinical experience and now my research experience, I think it's an important topic.  I do receive some royalties for teaching, related content for geriatric physical therapists.  I think the most important piece of this talk is that it kind of takes place in the post-acute care realm, and post-acute care occurs in this 30-day, 60-day,  90-day window that is a critical period of vulnerability for our older adult patients. Failure to make good functional improvements or gains  in those early windows is pretty prognostic of poor long-term outcomes.  This is particularly important for older adults who have high vulnerability to functional decline for myriad reasons, increasing skeletal muscle loss,  higher levels of pre-hospital vulnerability,  and other social support factors that may influence outcomes.  Rehabilitation services like physical, occupational, and speech therapy are thought to be essential parts of the post-ICU recovery continuum.  They're in the NIECE guidelines in the United Kingdom. Multiple stakeholder groups recommend it here in the United States.  But we really have no empirical data on how much rehabilitation these patients are receiving.  And I want to caution you that discharge to a rehabilitation setting does not mean that patients are receiving rehabilitation.  There are codes in both skilled nursing facility and in home care where people can be admitted and only receive nursing care services.  So being discharged to a skilled nursing or home health care after hospitalization does not guarantee that you're receiving rehabilitation.  And I think you'll see that in some of the data. So really, at the end of the day, we don't know how much people are participating  in rehabilitation across the post-hospital continuum. And that may be a reason why functional outcomes are so poor.  Now, home health care, more broadly, is nursing, rehabilitation, social work, and home care aids all paid for under a comprehensive bundle by Medicare Part A.  And there's really no clear rules or guidelines for how these services need to be provided to somebody recovering from critical illness.  The only statutory requirements for Medicare are that patients require skilled care and that they are homebound.  So they all have physical, disability, or medical complexities such that they can't leave home without difficult and taxing effort.  And this says to us that they would all benefit from at least a screening or consultation from rehabilitation if they're having that much difficulty leaving home without assistance.  And that puts a lot of burden on caregivers and a lot of other medical issues that come up with patients impact that as well.  Now, post-acute care, if you look at the continuum, 30% upwards of patients after an ICU stay receive home health care as their first discharge setting.  And that probably underestimates the overall use because skilled nursing facilities and inpatient rehab facilities often discharge patients  to home care as well as part of the continuum.  So it's an important area for us to study, especially for Medicare beneficiaries.  So our research question was to look at the amount and type of rehabilitation these patients are receiving just to get an idea of the scope  of what we're looking at and look at factors associated with this variability. So we identify patients in home care and identify who's getting rehabilitation  and what medical and social factors might be interrupting or facilitating that access so we know who to target.  So we looked at older Medicare data that we had from a project I started during my PhD training at University of Colorado.  And we took Medicare hospitalization and home care claims files and put them into a composite so we could identify older ICU survivors who stayed at least 24 hours  in the ICU and survived 30 days at home after discharge. This told us that we were really looking for rehabilitation candidates,  people who are medically stable and probably able to participate in rehabilitation and decrease some of the bias for patients who may have discharged to hospice  or died in those early post-hospital days.  Now the advantage of using home care data is it includes a lot of things of interest to us as clinicians.  It includes functional ability, which is not included in a lot of other Medicare claims files.  It also has burden of symptoms, things like dyspnea, pain, and anxiety  that certainly limit participation in physical activity and exercise  or other rehabilitation activities.  And we also have measures of social support, whether or not patients live alone and whether they live with caregivers.  And also resource needs, unmet needs for ADL assistance and other in-home concerns. And there's even medication and high-risk medication data which we don't use  but certainly of value to some of the other research I've heard about here.  Now because this is count data, we used a negative binomial regression to evaluate the distribution of these counts of physical occupational speech therapy visits.  And in the adjusted models, we adjusted for patient factors, which are important, but also availability of home health care services.  So number of home health agencies per 100,000 people in a county. And ownership characteristics, like for-profit agencies, for example, might be more willing to provide therapy  because they make more money off of those visits. So for us, that kind of really told us that we'd get a good granular detail of how much therapy these people were getting  and what the factors related to that use was.  In terms of the results, I think this is striking that on average, combined between all the therapies, patients are only receiving about one visit of therapy  for a week for a time of about 30 days, okay? So that's four total visits, which to us is not a very high dose.  And one in three received zero rehabilitation visits, despite all being homebound for physical or medical complexity and unable to leave home without significant assistance.  That was striking to us.  In terms of how much rehab people are receiving and what types, most are receiving physical therapy.  Very few, one in four, are receiving any occupational therapy, which would be more of your bathing and dressing.  These important tasks that patients tend to value, and only 5% received a consultation or visit from a speech-language pathologist,  despite 15% being mechanically ventilated in the ICU.  That suggests to us that we're probably missing people that may be at risk for dysphagia or vocal cord injury that's pretty common after an ICU stay.  And that is very related to market availability of speech therapists in home care, which is not good.  We also wanted to see what factors influenced rehabilitation use. And we broke it down into these four categories from my clinical experience  and the clinical experience of my team at Yale to really figure out which of these factors are driving rehab use.  And not surprisingly, we found that those with higher levels of disability were receiving more rehabilitation.  And in the highest tertiary, they were receiving about 30% more visits.  That's good news. We're happy about that. We're targeting the right people.  Medical complexity was a different story.  Highest Elixauser scores, the highest tertial of this multi-morbidity score, those patients were receiving 11% fewer visits at home compared to the lowest tertial.  And it's hard to say whether therapists are uncomfortable pushing or working with medically complex patients, or these patients refuse participation,  shut the door in the therapist's face, don't let them in. It's really hard to tell the difference there. And I've had both of those things happen to me.  So, but it's just something to be aware of.  Symptom burden, which we were happy about, did not have a negative influence. So, patients with high anxiety and high pain did not have a difference in rehabilitation use.  And actually, patients that had shortness of breath identified at evaluation got more therapy visits. It tended to be a trigger. And that may be because that is a quality metric  on the home health care publicly reported data. So, that's probably something agencies are particularly aware of.  However, I think the most striking finding was the sociodemographic factors that really drove utilization of therapy in a negative way.  Living alone, social isolation was associated with between 11 and 15% fewer therapy visits.  And living in a rural area was associated between 6 and 10% fewer visits. And these are happening over a 30-day window and probably highlight the importance  that social determinants of health have in utilization of services and may point to a mechanism why patients that have these factors may have worse functional recovery.  And I'll tell you that that rural finding has been consistent across other diagnoses that we've looked at, including joint replacement and dementia.  So, it is a pretty robust finding and one that we're exploring a little bit further.  Now, I think it's, the key points of this article, which if you want to read more, has been accepted in American Geriatric Society Journal like three days ago.  So, I don't have a QR code like one of the other one of you had. But we identified that these patients are receiving low doses of rehabilitation in home health care settings.  And perhaps doses low enough that to not be physiologically adequate to overcome the deconditioning and aerobic capacity concerns that these patients have.  Four visits over 30 days for older adults is, in my opinion, inadequate.  Social and geographic factors explain a larger portion of variability than we would have expected.  What I didn't show was some of the U.S. variability across geographic regions, which was also very striking.  But those factors explaining so much variability suggest that we have to pay more attention to those maybe in the time of discharge from hospital to home.  And I think it's really important to remember that this is happening when therapists are coming to a patient's home and driving and doing all of the leg work.  These disparities are probably worse when we're asking patients to get in the car and drive to a physical therapist. Voluntarily exercising is difficult for most of us.  But for very sick and very frail patients, it typically doesn't happen. I think we already know from post-ICU clinic data that attendance is pretty poor for some  of these reasons, and I'm going to guess that these disparities that we're seeing in home rehabilitation are occurring in the outpatient setting as well.  So future directions, obviously, this rural-urban disparity issue, while somewhat small in magnitude, certainly may hint at further disparities.  You think about regionalization of ICU care when you have very concentrated ICU beds in these rural communities, and then they're discharging two and three hours away  and coordinating resources, and things may be more challenging. So we may look at that in terms of these patient-centered outcomes  where patients may find rehabilitation more optional than following up with surgeons or other specialists.  And then obviously look at the implications of this low use on patient-centered outcomes,  which in our other diagnoses that we've studied have been strongly associated up to a point.  So people that receive fewer than five visits of rehab typically make less functional gains than those who receive a more moderate to high dose.  So I want to acknowledge my team at Yale, my mentors, Lauren Ferrante, who many of you probably know, who's a critical care physician, and this data came  from my PhD mentor at University of Colorado, and funding from my physical therapy foundation and an NIA training grant, and the data was acquired  through the Department of Veterans Affairs.  So I appreciate your time, and thank you so much. I'll take any questions.  Thank you very much. Thank you.  Yeah. I find it striking. And, you know, if we look at total joint replacement patients, so in the same data set, we have done this prior.  And they're receiving eight visits over 17 days. These patients are receiving three or four visits over 30 days. So the difference in dose used is, it was really striking to us.  I think part of it points to the fact that therapists don't know what to do with these patients. There's not a protocol. There's not a joint replacement protocol  where every week there's exercises. There's a laid out, you know, timeline and goals and expectations for where your knee should be bending at every point in time.  For these patients, they have global widespread deficits, weakness in every muscle group, widespread fatigue and symptom burden.  And we just have no real guidelines to put therapists and say these are the milestones. And when therapists have patients that don't have much potential for improvement,  there's a tendency to discharge patients if they don't think they're going to be good gain, make good functional gains, despite the fact that rehabilitation can be used palliatively  to address other symptoms or to address community level issues, car transfers, things that aren't measured on this scale but are really important.  So we found it striking and certainly something where we're raising attention to in our community.  and maybe that follows on to better for healthcare reasons? Yeah. That's a great question. And only one Medicare inpatient file has what actually was  delivered in terms of rehabilitation. The MedPAR file, which is the more common one people use, only has rehabilitation charges. And charge data, as you guys probably expect,  is pretty widely variable across hospitals. There is a colleague of mine at Northern Arizona that is doing that kind of work within hospital rehabilitation.  An impact on 30-day outcomes and discharge home for stroke. What they found is higher levels of rehabilitation delivered in the hospital have an effect,  a pretty monotonic relationship to a point with better post-discharge outcomes. So, but there's a selection bias issue. Obviously, healthier patients are going to be able  to participate more, so.  Sure. I mean, those were all controlled for in the analysis. We didn't, there wasn't any actual gender or racial differences in the data. But they were accounted for.  So, we were heartened to see that. But there's, you know, Medicare population is almost overwhelmingly 85% white. So, there's a pretty small number of racial minorities in the data.  And it just may not have been enough to see the signal.  Yeah? Any small studies that look at what the outcome of various studies post-discharge in terms  of it's not, you know, obviously it's not one-time. I don't know if that should be. Yeah. I mean, there's so few studies that have both the sample size and the adherence  to actually participating. Rehab studies are notorious for dropout in this population. There's one study in Thailand or another country  that has a massive insurance database where they looked at 18 and over critical illness survivors. And their optimal dose, they were saying a dose  or an episode of care was at least six visits. And people who received one or, you know, two or more of those episodes of six visits, or 12 to 18 visits had lower mortality.  But this was 18 and older. And it was a very large cohort of insured patients in that other country. And I'm not, they didn't really describe what they were doing  in therapy and what the, and what the content of those interventions were. So I think it's still a very open question as to what the best interventions, dose,  and frequency are for rehabilitation.  Thank you. Thank you.  Hello, everyone. And thanks for staying late to listen to us this evening. My name is Mohamed Al-Rawashdi. I am a postdoc fellow at Harvard Department of Population Medicine.  And today I will talk about the epidemiology outcomes and trends of sepsis in patients with opioid use disorders in the U.S. hospitals. I have no financial disclosures.  So apparently this is consistent with one of the previous presentations. Many of us know that the U.S., in the midst of an opioid epidemic, actually the drug overdose deaths tripled  between 1999 and 2015. And in 2017, there were two-thirds of the drug overdose related deaths were involving opioids. And these include both prescription and illicit opioids.  Besides mortality also, these patients are at high risk for transmission of viral disease, such as HIV and hepatitis C. And they're at higher risk  for bacterial and fungal infections that may lead to sepsis, such as bacteremia and endocarditis, due to immunodulatory effects. There is very little data in the literature  about the epidemiology of sepsis in patients with opioid use disorders. This is hindered by lack of accurate sepsis surveillance methods due to various reasons,  such as the subjectivity in diagnosis, variable definitions and changing diagnosis and coding practices over time. Luckily, in 2018, the CDC released the adult sepsis event  surveillance definition based on the epicenter's lead work that was done by our team at the Department of Population Medicine, which proved  to be a highly sensitive method with a large or high positive predictive value. Utilizing this definition allows us now to have more confidence  in estimates of sepsis epidemiology and its trends. So in having that in mind, the aim of our study was to characterize the epidemiology outcomes and the trends  of sepsis in patients with opioid use disorders. To answer this question, we did a retrospective cohort study in which we analyzed administrative  and electronic health record data for hospitalized patients from three large databases, including the Cerner Health Facts, the Institute of Health Metrics and the HCA Healthcare.  And these databases cover the years between 2009 and 2015. And overall, we had over 6.5 million inpatient encounters from 373 hospitals. We used the CDC adult sepsis event criteria  to define sepsis here by two factors, presumed serious infection plus concurrent organ dysfunction. So for the first one, was confirmed  by blood culture draw plus at least four days of antibiotics. And the second one was confirmed by organ dysfunction within plus minus two days of the blood culture draw.  For the opioid use, we defined it by all acute and the chronic poisoning ICD-9 codes. We used all discharge codes here. Or by the inpatient medication prescriptions.  Namely here, we're talking about Suboxone. We also collected data about the presumed body site infections, such as pneumonia, bacteremia, endocarditis and others.  We also looked at blood stream pathogens associated with sepsis based on the microbiology data. For the data analysis, we used a generalized mixed models.  In which, in order to look at the adjusted impact of opioid use on sepsis associated hospital mortality. We had random effects for hospitals.  And the results were compiled using the study level meta-analysis. Because our databases could not be combined due to hyperregulations. We looked at temporal trends  that they were modeled using the generalized estimating equations to fit Poisson regression models adjusting for hospital characteristics. Our sample, and that included  about 6.7 million hospitalizations. And the bars here show the count of cases per 1,000 for different age groups. And here, like while the red line here shows the opioid use.  On average, the opioid use rate was about 1.9%. While the opioid use rate was higher among younger patients as clear here, like across, like ages.  For, this figure shows us the sepsis incidence by age and opioid use status. It is clear that sepsis rate is higher among hospitalized patients with opioid use.  Now we're switching gears here to focus only on the opioid use patients. And we divided them here based on their sepsis status. Those who had sepsis only accounted for 6.7%.  And these patients compared to no sepsis patients were more likely to be older, comorbid, higher, at higher rate, have higher rate of ICU admission,  longer ICU and hospital length of stay. And more positive blood cultures.  Here we looked at the discharge disposition for opioid use, users by presence of sepsis. And as you can see here, that no sepsis patients were more likely to be discharged home.  However, like sepsis patients were at higher risk for mortality compared to no sepsis patients. When we ask the question, how many opioid use patients who die have sepsis?  Among opioid users who die, this was about 51%. Compare that to the opioid non-users who died. That was about only 36%. Now we're switching gears for a second time here.  And we're only focusing on the sepsis patients. And here we're dividing them by their opioid use status. The opioid use patients accounted for only 2.3% of this group.  And these patients were more likely to be younger and of white race, healthier or less comorbid, at higher rate of ICU admission, have longer hospital length of stay,  and less hospital onset sepsis. When we looked at their key comorbidities, we found that opioid use patients having higher rates of hepatitis C and HIV.  But lower rates of key comorbidities such as CHF, DM, DM with complications, hypertension with complications, tumor and renal failure.  However, they still have higher rates of comorbidities that we expect to see in this subgroup, such as chronic lung disease, liver disease, alcohol abuse,  drug abuse, and psychological disorders. When we looked at the infections they are prone to, we found that opioid use group has higher rates of epidural abscess, osteomyelitis,  septic arthritis, and endocarditis.  Looking also at the sources of infection for sepsis, we found that the opioid use patients had higher rates of bone and joint infections, skin and soft tissue,  and pulmonary infections. This makes sense, as we know, like they use more non-sterile needles. But they had less intra-abdominal,  genitourinary, and septicemia and bacteria infection. Looking at the blood culture pathogens, we found that opioid group has higher rates of Staph aureus yeast infections,  but least rates of Escherichia. And in general, they had, the opioid users, they had higher rates of gram-positive and fungus infections, but lesser amount  of gram-negative infections. And they also had higher rates of polymicrobial profiles. It was interesting to see that the mortality was lower among the sepsis patients  with opioid use, and the odds ratio was .6. But when we controlled for other confounders, such as demographics, that changes to .75. When we controlled for clinical characteristics,  that adjusted, the adjusted odds ratio was .65. But when we added to the model also comorbidities, the adjusted odds ratio was .73.  So in general, like with the adjusted and non-adjusted dose with, the mortality was lower among the sepsis patients with opioid use.  And here, when we ask the question, how many sepsis patients who die are opioid users, the answer was 1.5% among all sepsis patients who died. But among healthy sepsis patients  who died was 2.7%, and among young sepsis patients who died was 5.7%. This graph shows the trends of, and the prevalence of opioid use disorders in sepsis hospitalizations  between 2009 and 2015. The top line, the black line here, this is the, based on our definition that we talked about. And we saw a 77.3% increase in, in the rate  between the years 2009 to 2015. But also, the different lines down there, they looked at the sub-definitions of opioid, dissecting the opioid use.  Like all acute and chronic poisoning, acute, all acute poisoning, acute poisoning from NLG6 and opioid meds prescription. So, these were either increasing or sustained over time.  Our study has few limitations. That the sample was, that included only hospitalized patients. And our results can be generalized only to hospitalized patients.  However, our sample was representative of the U.S. hospitals. And the use of administration data to identify patients with opioid use was reported previously that more  than 50% of drug use cases may be unrecorded using ICD codes. Our definition included inpatient opioid medications in addition to ICD-9 codes.  Which increased the, increased case counts. But net sensitivity is unknown. There was lack of consistent reporters over time. However, models adjusted, were model adjusted  for hospital characteristics. In conclusion, sepsis is a previously underappreciated cause of morbidity and mortality in patients with opioid use disorder.  Occurs in one in 15 hospitalizations and have deaths among opioid users. Opioid use disorders only contribute to a small fraction of sepsis cases overall.  But disproportionately affects younger and otherwise healthier hospitalized patients. And clinically important differences in sepsis patients  with opioid use disorders are noted here. A few implications here. Such as empirical treatment. Imaging, implications for imaging and source control. And the aggressiveness of care.  And we also saw an increasing incidence of hospitalization with sepsis and opioid use over time. So in addition, finally in addition to ongoing efforts to combat opioid crisis.  Future public health interventions should focus on increasing awareness, recognition and aggressive treatment of sepsis in this population. Thank you.  I would open it now like for questions.  Yeah, so there was a study before that was a smaller, to use a smaller sample size. But in that study, they found that opioid patients have more infections.  But it wasn't as comprehensive as the database we have and the analysis we conducted here. Yes.  Yeah, so as I stated in the conclusion, like, you know, previous study shows that ICD coding might miss about more than 50% of actual sepsis cases.  But here, like, let me show you, like, you know, in our definition, we use the adult sepsis criteria. Oops. Okay. Yeah. So it's this one here.  So we used the adult sepsis event criteria. It was developed by my mentor, Chanu Ray at Harvard Medical School. And that was funded by the CDC. And the CDC adopted that.  So that was confirmed by blood culture draw and more than four days of antibiotics. The second condition was organ dysfunction. That should happen within two calendar days  of the blood culture withdrawal. So this we're trying to have higher sensitivity and positive predictive value compared to the implicit  and explicit codes that are used for the claims data. So this was one of the strengths in our study, that it wasn't only reliant on administrative data,  but also we used electronic health records data. Yeah. I wonder, like, the sepsis definition, I didn't have much information on my area,  but the substance abuse definition, the opioid. The opioid. Yeah. I wonder how well, you know, if your patients are just not self-reporting that to you,  are they in the other group when they're related to the opioid group? That's very good question. Like, you know, we don't have any, like,  other Gordel standards to compare with right now, which is, yeah, one of the limitations of our study. We try to, like, limit that by including also medications,  like the prescription of Suboxone. In fact, like, you know, our initial analysis, the one we reported for the initial abstract, we included methadone also, prescription.  But our collaborators from the CDC, like, and our opioid experts, like, they suggested it might be not that sensitive, since,  like, as you know, methadone can be used for either, like, you know, chronic pain, like, you know, treatment, for example, or can be used, like, you know, for the opioid use.  So, yeah, I totally agree. But this is, yeah, one of the limitations of the study. Any other questions?  Okay. Okay. Thank you.

Video Summary

Summary:<br /><br />The first study analyzed the impact of the opioid epidemic on pediatric ICU admissions for opioid poisonings. The study found a decreasing proportion of children being admitted to the ICU for opioid poisonings since 2010. The opioids primarily ingested were hydrocodone, oxycodone, codeine, and tramadol. The study also found an increase in moderate and major effects in children who ingested opioids, as well as an increase in hospitalizations to general and psychiatric hospitals. Naloxone use increased, indicating better education and accessibility. Synthetic opioids, intentional intent, and illicit opioids were associated with requiring ICU intervention.<br /><br />The second study aimed to characterize sepsis in patients with opioid use disorders in US hospitals. The study found that the rate of opioid use among hospitalized patients was 1.9%, with higher rates among younger patients. Patients with opioid use disorders who had sepsis had higher rates of comorbidities, ICU admission, longer length of stay, and positive blood cultures. However, they had lower mortality rates compared to those without opioid use disorders. The study also found an increasing incidence of hospitalizations with sepsis and opioid use disorders over time.<br /><br />Both studies highlight the need for increased awareness, recognition, and treatment for opioid-related issues. The first study recommends increasing mental health resources for children and decreasing access to opioids. The second study suggests focusing on aggressive treatment of sepsis in patients with opioid use disorders. However, the studies are limited by the use of administrative and electronic health record data.

Asset Subtitle

Worldwide Data, Research, 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.

49-Are Racial/Ethnic Minorities Over- or Underrepresented in Research Enrolling Seriously Ill Children?

Erin Paquette

50-Heart Rate Variability Dysfunction Is Associated With Outcomes in Critically Ill Children

Colleen Badke

51-Increased Risk of Mental Disorders Associated With Childhood Invasive Mechanical Ventilation

Andrew Geneslaw

52-Analysis of 207,543 Children With Acute Opioid Ingestions From the National Poison Data System

Megan Land

53-Components of Health-Related Quality of Life Most Affected Following Pediatric Critical Illness

Elizabeth Y. Killien

54-Association of Cumulative Haloperidol Dose With 28- and 90-Day Mortality in Critically ill Adults

Matthew Duprey

55-Factors Associated With Home Health Rehabilitation Utilization Among Older ICU Survivors

Jason Falvey

56-Epidemiology, Outcomes, and Trends of Sepsis in Patients With Opioid Use Disorders in U.S. Hospitals

Mohammad Alrawashdeh

Meta Tag

Content Type Presentation

Knowledge Area Research

Knowledge Area Worldwide Data

Knowledge Level Advanced

Membership Level Select

Membership Level Professional

Membership Level Associate

Tag Clinical Research Design

Tag Outcomes Research

Tag Epidemiology Outcomes

Year 2020

Keywords

opioid epidemic

pediatric ICU admissions

opioid poisonings

hydrocodone

oxycodone

codeine

tramadol

synthetic opioids

illicit opioids

sepsis

opioid use disorders

comorbidities

ICU admission

mortality rates

mental health resources

	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.