SCCM Resource Library-Hospital-Specific Contribution to Mortality in Children With Severe Sepsis and Head Injury

Video Transcription

specific contribution to mortality in children with severe sepsis and head injury. My name is  Nader Yahya. I'm giving this talk on behalf of Hannah Mitchell, who unfortunately wasn't able to  come to SCCM. These are my disclosures. So we know that there's significant disparities in pediatric acute care with substantial regional variation, and it's important to get at the  mechanisms underlying these disparities in order to be able to target interventions.  We hypothesized for this specific study that hospital level variation is a substantial contributor toward disparities in pediatric critical care. You can tell Hannah should be  giving this talk because pediatric is spelled incorrectly as if I was British, but I'm not.  This is a population-based retrospective cohort study using the 2016 submission of the KID inpatient database using patients between 0 and 21 years of age. We separately assess two  diagnostic groups which are represented in critical care, but which are not necessarily specific to critical care, severe sepsis and traumatic brain injury. We perform multi-level  logistic regression clustered by hospital, and we created a few novel hospital characteristics in order to fully adjust for our model with hospital-level characteristics.  Finally, we derived hospital-specific odds for mortality and then compared median odds ratios to assess for hospital-level effects. We'll talk about some of that in a few minutes.  The median odds ratio can be thought of as a summary metric comparing the hospital-specific  odds of mortality between a lower-performing and a higher-performing hospital. If you take two random hospitals on this caterpillar plot, which is hospitals ranked according to  their predictive probability of mortality either on a probability scale or on an odds ratio scale,  and you compare a lower-performing one to a higher-performing one, then there's an  attributable risk of mortality, which after multivariable adjustment for known confounders that were included in the model, you can decompose and extract the hospital-specific  contribution to mortality. This can be ranked and then compared. The median odds ratio is the median  of the odds ratios for the increased risk of mortality between a lower and a higher-performing  hospital. When you actually get to our study, the sepsis cohort had almost 11,000 children,  the TBI cohort had approximately 31,000 children, of which 15% of the septic kids died, 4% of the  TBI kids died, and the major comorbidity difference between the two populations was that sepsis had 73% of subjects with a complex chronic condition present.  We found and confirmed racial disparities with higher mortality in black children compared to white in both sepsis and traumatic brain injury, with trans but not significant  increased mortality in the other category for both subjects.  This is a caterpillar plot of the MT model without any variables, just assessing the  hospital-specific variation of mortality in sepsis, for example. And this is the bulk of our results.  When you just focus on the sepsis column, the MT model had a median odds ratio of 1.66, suggesting  that for septic patients, simply moving from one hospital lower-performing to a higher-performing  hospital was associated with a 66% higher odds of mortality. When we included individual-level  characteristics, then this odds ratio was attenuated to 1.53. When we included admission-level  characteristics, then this was further attenuated to 1.46. And in our full model, in which we also  had hospital-level characteristics, this median odds ratio is 1.34, which was still significant,  suggesting that after adjusting for everything available in KID, which was a plausible confounder,  there's still a 34% residual increased odds of mortality as you go from a lower to a higher  performing hospital. We found the exact same pattern in TBI, where the inclusion of all the  available variables in the full model partly, but did not completely eliminate the hospital-level  variation. We looked at the best-performing and the worst-performing hospitals that were shared  between sepsis and TBI, and we didn't find any obvious characteristics that were particularly different. The percent publicly insured was higher in the best-performing hospitals,  by like approximately 7%, but the clinical significance of that is not clear. This is not particularly surprising, because all of these variables that were available for  comparison were included in the model as hospital-level variables.  The limitations of our study are that there's no severity of illness data, there's no information  on the precise location of the hospital and neighborhood characteristics, and there's no information on actual treatments received in the hospital, such as time to antibiotics,  or availability of neurosurgery, for example, as things which could impact our outcomes.  Overall, we conclude that there's substantial inter-hospital variation in mortality risk, and there's only modest attenuation by individual and hospital-level  characteristics when they're included. Further research was warranted, particularly in the  future, to see whether there are better, more granular, or just different variables which can  capture some of the differences between the hospitals. This kind of analysis, although it  needs the numbers of large databases, it may actually be better served by having qualitative  and systems-level analyses, which look at whether there's different protocols in place or different  approaches that these hospitals take towards septic and TBI patients.  This is our reference list, and with that, we thank you.

Video Summary

A recent study examined the contribution of hospital-level variation to disparities in pediatric critical care outcomes. The researchers used a population-based retrospective cohort study and analyzed data from the KID inpatient database. They focused on two specific diagnostic groups: severe sepsis and traumatic brain injury. The findings showed that there is significant inter-hospital variation in mortality risk for these conditions. Even after adjusting for individual and hospital-level characteristics, there was still a residual increased odds of mortality when comparing lower-performing hospitals to higher-performing hospitals. The study highlights the need for further research to better understand the factors contributing to these disparities and to improve outcomes for pediatric critical care patients.

Asset Subtitle

Sepsis, Quality and Patient Safety, Pediatrics, 2022

Asset Caption

INTRODUCTION: Sepsis and head injury are common diagnoses in pediatric critical care, with data showing disparities by race/ethnicity for both conditions. We investigated the role of individual hospital-level variation in contributing to these disparities. METHODS: We used the most recent (2016) Healthcare Cost and Utilization Project (HCUP) Kids’ Inpatient Database (KID) including 80% of pediatric discharges from 4,200 U.S. hospitals across 47 states. Using multilevel causal logistic regression, clustered by hospital, race/ethnicity was tested for association with hospital mortality adjusting sequentially for individual-level (age, sex, insurance type, presence of chronic complex condition), admission-level (admission day, quarter, admitted through ED, transferred in, elective admission) and hospital-level (hospital region, size, location, control, teaching status, percent of patients who are Black and percent of patients who are uninsured) characteristics. Hospital-specific odds for mortality and median odds ratio for hospital effects were calculated for sequential models. RESULTS: 10,973 children with severe sepsis and 31,582 children with head injury were included. Mortality was 14.9% for severe sepsis and 4.1% for head injury. Mortality was higher among Black children with severe sepsis and head injury (crude OR 1.41 [95% CI: 1.20-1.65] and 1.27 [95% CI: 1.09-1.49] respectively). Hospital-specific ORs varied from 0.18 to 3.20 for severe sepsis and 0.27 to 3.24 for head injury, suggesting mortality effects dependent upon being admitted to a given hospital. The median odds ratio for the unadjusted model was 1.66 (95% CI: 1.50-1.81) for severe sepsis and 1.63 (95% CI: 1.49-1.77) for head injury, suggesting an average 1.6-fold increased mortality odds when comparing a better- to a worse-performing hospital. On sequential addition of all available patient-, admission- and hospital-level characteristics, this decreased to 1.34 (95% CI: 1.20-1.48) for severe sepsis and 1.43 (95% CI: 1.26-1.59) for head injury. CONCLUSIONS: Disparities exist in mortality from sepsis and head injury. Our study showed substantial hospital-level variability in patient mortality with only modest attenuation by individual and hospital level adjustments.

Meta Tag

Content Type Presentation

Knowledge Area Sepsis

Knowledge Area Quality and Patient Safety

Knowledge Area Pediatrics

Knowledge Level Advanced

Membership Level Select

Tag Sepsis

Tag Mortality

Tag Pediatrics

Tag Healthcare Delivery

Year 2022

Keywords

hospital-level variation

disparities

pediatric critical care outcomes

severe sepsis

traumatic brain injury

	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.