SPS9-01 Comparing Care Patterns and Outcomes for Close-Aged Patients With Sepsis in Pediatric vs. Adult ICUs: A Fuzzy Regression Discontinuity Design Study of Adolescents and Young Adults

Introduction

Adolescents and young adults (AYAs) with sepsis may be admitted to pediatric or adult ICUs based on arbitrary age cut-offs and receive variable care in siloed settings. The extent and impact of these treatment disparities remain unknown.

Methods

AYAs (14-22-years-old) with explicit ICD-10 diagnostic codes for sepsis admitted to a pediatric (PICU) or adult ICU were identified in the PINC-AI database, which covers 25% of US all-payer hospitalizations. We defined care environment as pediatric or adult ICU using accommodation charges along with patient age distributions consistent with epidemiological standards. Host characteristics and treatment patterns were compared across care settings. We estimated local average treatment effect of PICU (vs adult ICU) admission on mortality using a Fuzzy Regression Discontinuity Design (FRDD). We modeled mortality and PICU admission over age using local linear regression with uniform kernels across the age cutoff and robust bias-corrected standard errors and confidence intervals. We accounted for baseline covariates including patient-level (sex, ethnicity, pre-existing comorbidities, acute illness severity) and hospital-level factors (bed size, teaching status, urbanicity, technological capabilities). Sensitivity analyses using narrower age bandwidths tested stability of point estimates.

Results

Of 19,915 14-22-year-olds with sepsis admitted across 856 US hospitals, 18.4% and 81.6% were admitted to pediatric and adult ICUs respectively. PICU (vs adult ICU) patients had higher medical complexity classified by the Pediatric Complex Chronic Condition system with 80.8% vs 70.0% meeting criteria for a least one domain and 44.1% vs 26.2% with three or more. Patients admitted to PICUs (vs adult ICUs) were less likely to receive mechanical ventilation (38.8% vs 43.9%). There was higher proportional use of epinephrine as first-line vasopressor in PICUs (24.3% vs 6.3%), whereas norepinephrine predominated in adult settings (78.6% compared to 41.8% in PICUs). In the RDD analysis, we observed no statistically significant treatment effect of admission to PICU (vs adult ICU) on mortality risk for AYAs with sepsis (Figure 1). These results were consistent across bandwidths (2, 3 and 4 years surrounding the age threshold of 18) and present with and without accounting for baseline covariates.

Conclusions

In this large cohort of critically ill AYAs pseudo-randomized by an arbitrary age cut-off, risk-adjusted mortality outcomes for sepsis were similar in pediatric vs adult ICUs and robust to disparate practice patterns between these settings. Findings suggests equipoise for cluster randomized trials comparing prevailing care differences for sepsis in close-aged patients across pediatric and adult ICU settings to further optimize outcomes.

This abstract is funded by: None