SCCM Resource Library-Admission PaO2 and Mortality Among Pediatric ICU Patients and Critically Ill Subgroups

Video Transcription

Hi, everyone. My name is Kara Holton. I'm a third-year PICU fellow at Children's Mercy Hospital in Kansas City, and today I'd  like to share with you my research project looking at admission PaO2 and mortality in pediatric ICU patients.  This is a hyperoxia study, and so hyperoxia can be defined as any superphysiologic oxygen tension.  We know that hyperoxia can lead to increased formation of reactive oxygen species, which can cause cell damage, cell death, and inflammation.  Multiple adult studies have suggested an association between hyperoxia and mortality. The pediatric studies are slightly more limited due to smaller sample sizes and just looking  at general PICU populations, although I do put asterisks here because there's a few exceptions to this, and more studies are being produced every day.  So the objectives for my study were to evaluate whether hyperoxia on admission to the PICU is associated with increased mortality, and specifically, I wanted to look at a large  multicenter data set. Secondly, I wanted to evaluate the relationship between hyperoxia on admission and mortality in specific diagnostic subgroups.  So for my study design, this was a retrospective multicenter observational cohort study. I utilized a VPS database, which, if you're not familiar with it, is a fantastic clinical  pediatric critical care database with over 135 participating hospitals. They have very robust data quality control as well.  I included all non-cardiac patients less than or equal to 18 years of age who had a PaO2  reported in the first hour of their ICU admission, and the primary outcome I was examining was ICU mortality. I used PIM-3 scores from VPS, the Pediatric Index of Mortality.  It's a way to sort of estimate illness severity on admission. The PIM-3 score involves a PaO2 term, and so to avoid any confounding, I created a modified  PIM-3 score, which is done by just simply excluding the PaO2 term from the PIM-3 equation.  And then I was able to use VPS admission data and ICD codes to identify specific diagnostic subgroups, and you can see the six that I've selected to share with you today.  So getting into the results, the total sample size was 13,071 patient encounters over a five-year period.  I specifically looked at 2015 to 2019. The overall mortality in this group was 13.5%, and so right away you can see that this is  a sicker cohort of patients than your typical ICU population, and that's because, again, we're only looking at those who had a PaO2 obtained on admission.  Fortunately, we found that hyperoxia is fairly rare, depending on how you define it. So a quarter of our patients had a PaO2 above 200, but only 7.8% had a PaO2 above 300, and  just about 1% of the patients in the study had a PaO2 above 500 on admission, so that's excellent news.  So now we're looking at the entire cohort, and I used these PaO2 bins, so on the x-axis is the PaO2, and then the y-axis is just the raw mortality.  The number inside each gold bar is the mortality for that specific PaO2 bin, and so right away  you can see this distinct U-shape to the data that many other studies have suggested as well with the highest mortality being seen in patients who have a low PaO2 and those  who have a very high PaO2. Next we performed an unadjusted logistic regression with the PaO2 modeled as a quadratic term.  The blue bars here in the graph are the observed mortality, and the black line represents the logistic regression, and you can see that a quadratic relation fits the data remarkably  well with really minimal standard error. And then I wanted to look at standardized mortality ratios.  One could argue that the patients who have a higher PaO2 are just sicker in general, and that's why they have higher mortality, and so I used a standardized mortality ratio,  which is calculated as observed divided by expected mortality, and again, the expected mortality I calculated using my modified PIM-3 equation.  And so this line is at an SMR of 1, and so you could see that the highest standardized mortality ratios were seen in the hypoxic and hyperoxic patients.  Again, it's a little subtle, but there is still that U-shape to the data. And then for the next several slides, I'm just going to go through the different diagnostic  subgroups that I looked at, and so there were 2,700 trauma patients, and you can see there's this clear U-shape to the data here.  There were about 1,800 patients admitted after experiencing head trauma, and again, the U-shape holds.  Similarly, for sepsis patients, just over 1,200 patients with sepsis were included. Just under 1,000 patients with renal failure, and again, there's that very clear U-shape  to the data. Now we're getting into some smaller sample sizes, so there were 322 patients with hemorrhagic  shock, and interestingly, this group also had a very clear U-shape to the data, and in fact, the highest mortality was seen in the patients with the PO2 above 300.  And this is interesting because in my mind, hemorrhagic shock is one of the few situations in the ICU where I may turn up the oxygen, even if the SATs are normal, again, in order  to optimize your oxygen delivery equation. But in this retrospective study, patients who had a higher PAO2 had a much higher mortality than their cohorts.  And then here's where it gets really interesting. So there were 1,500 patients admitted after cardiac arrest, and you can see the overall  mortality for this group is quite high, not unsurprisingly, 63% mortality, and there's not really any clear association with the data.  You can see there's really, it seems like, for the most part, no difference in mortality based on your PAO2, and so I thought this was certainly a pretty interesting finding  considering how reliable all the other graphs were with the U-shaped curve. I just looked at standardized mortality ratio in the post-arrest patients just to see if  any relationships emerge, and again, you can see that there's not really any clear correlation between your outcome and your PAO2 in the post-arrest group.  So to summarize, then, I performed a large multicenter pediatric cohort study examining hyperoxia.  We found that the admission PAO2 was associated with mortality among the entire cohort, and then that relationship was preserved in some, but notably not all, diagnostic subgroups,  specifically the post-arrest group.  And so to me, this suggests that the pathophysiology of certain disease states may modify the hyperoxia association.  Additionally, I think I've demonstrated that future predictive models, such as future iterations of the PIM score, may benefit from including the hyperoxia status of the patient.  And then, like any study, there are some reasonable limitations, so this was an observational study only, so we cannot infer any causality.  And then the generalizability of it is limited as well, because, again, this was only looking  at PICU patients who have an admission PAO2, and certainly not all patients who come into the ICU are sick enough or happen to have an ART line and are able to get that data.  So moving forward, I think there's a couple of hypotheses that could be generated, and this is certainly…these have already been discussed in the literature as well, but one  hypothesis is that hyperoxia directly contributes to the mortality risk via perhaps reactive oxygen species or some other molecular mechanism.  Another…and again, this would need sort of a randomized controlled trial to further investigate.  Another hypothesis is that hyperoxia is just a marker of patients who are receiving aggressive  resuscitation, and then perhaps when you have sicker cohorts, like the post-arrest group where all the patients are being aggressively resuscitated, hyperoxia kind of loses the  ability to discriminate between outcomes. And then another question I have moving forward is, does the timing of the hyperoxia exposure matter?  Certainly thinking about the post-arrest cohort, does your PAO2 an hour after ROSC, does that make more or less of an impact on your outcome compared to your PAO2 12 hours  after ROSC? And that's something I wasn't able to look at in this study, but certainly future work could investigate. So thank you so much for your time.  I'm happy to take any questions or feedback.

Video Summary

In this video, Kara Holton, a PICU fellow, discusses her research project on the association between admission PaO2 levels and mortality in pediatric ICU patients. Hyperoxia, defined as abnormally high levels of oxygen, can lead to cell damage, inflammation, and increased mortality. The study utilized a large multicenter database and included non-cardiac patients under 18 years old. The results showed a U-shaped relationship between PaO2 levels and mortality, with the highest mortality observed at both low and high PaO2 levels. However, this relationship was not consistent among all diagnostic subgroups, especially in patients admitted after cardiac arrest. The study suggests that the pathophysiology of certain disease states may modify the association between hyperoxia and mortality.

Asset Subtitle

Resuscitation, Quality and Patient Safety, Pediatrics, 2022

Asset Caption

INTRODUCTION: Studies analyzing the association between hyperoxia and mortality in critically ill pediatric patients and those admitted after cardiac arrest are conflicting and limited by small sample sizes. The aim of this study is to evaluate the association between PaO2 on admission to the pediatric ICU and mortality in a heterogeneous, multicenter cohort and among subgroups of critically ill patients. METHODS: A retrospective cohort study was conducted using data from Virtual Pediatric Systems (VPS) database. VPS is a clinical pediatric critical care database with over 135 participating hospitals in North America. All VPS patients 18 years or younger admitted between 2015 and 2019 with an admission PaO2 were included. Cardiac patients and those admitted to a cardiac ICU were excluded. Patients were stratified into 50-point PaO2 bins based on admission PaO2. Modified PIM3 scores were calculated for all patients by excluding the PaO2 term from the PIM3 equation. ICD-9, ICD-10 codes, VPS admission data and PRISM3 results were used to classify patients into the following diagnostic subgroups: trauma, head trauma, sepsis, renal failure, hemorrhagic shock, and those admitted after cardiac arrest (post-arrest). RESULTS: 13,071 patient encounters were included with an overall mortality of 13.5%. The relationship between admission PaO2 and mortality demonstrated a U-shaped quadratic curve, with higher mortality seen in hypoxic (PaO2 < 50 mm Hg, mortality 28.5%) and hyperoxic (PaO2 > 300 mm Hg, mortality 25.4%) patients. This relationship persisted after adjustment for illness severity using modified PIM3 scores. Similarly, a quadratic relationship was demonstrated among trauma, head trauma, sepsis, renal failure and hemorrhagic shock patients. However, among the 1,500 post-arrest patients, there was no correlation between admission PaO2 and mortality, even after controlling for modified PIM3 scores. CONCLUSIONS: In a large multicenter pediatric cohort, admission PaO2 demonstrates a clear U-shaped quadratic relationship with mortality. The persistence of this quadratic relationship in some but not all diagnostic subgroups suggests the pathophysiology of certain disease states may modify the hyperoxia effect. Further studies are necessary to determine whether this association is causal or merely incidental.

Meta Tag

Content Type Presentation

Knowledge Area Resuscitation

Knowledge Area Quality and Patient Safety

Knowledge Area Pediatrics

Knowledge Level Advanced

Membership Level Select

Tag Cardiac Arrest

Tag Mortality

Tag Pediatrics

Tag Hypoxia

Tag Oxygenation

Tag Cardiopulmonary Resuscitation CPR

Year 2022

Keywords

Kara Holton

PICU fellow

admission PaO2 levels

mortality

pediatric ICU patients

	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.