External validation of the Pediatric Extracorporeal Membrane Oxygenation Prediction model for risk adjusting...

Area of Science:

• Pediatric Critical Care Medicine
• Extracorporeal Membrane Oxygenation
• Biostatistics

Background:

• The Pediatric Extracorporeal Membrane Oxygenation Prediction (PEP) model aids risk stratification for pediatric patients on extracorporeal life support (ECLS).
• External validation is crucial to assess the PEP model's generalizability and performance in diverse clinical settings.

Purpose of the Study:

• To externally validate the PEP model using a large, contemporaneous cohort from the Extracorporeal Life Support Organization (ELSO) registry.
• To evaluate the PEP model's discrimination and calibration in predicting outcomes for pediatric patients requiring ECLS.

Main Methods:

• A multicenter, retrospective analysis of pediatric patients (<19 years) undergoing their initial ECLS run (January 2012–September 2014).
• Utilized the ELSO registry data, with median activated partial thromboplastin time and international normalized ratio from the BATE dataset as surrogates for missing values.
• Assessed model discrimination via area under the receiver operating characteristic curve (AUC) and goodness-of-fit using the Hosmer-Lemeshow test.

Main Results:

• Compared 4,342 patients from the ELSO registry with 514 from the BATE development dataset; overall mortality was similar (42% vs. 45%).
• The PEP model's c-statistic decreased from 0.75 (development) to 0.64 (validation).
• Model calibration notably decreased, particularly for the highest-risk patient deciles.

Conclusions:

• The PEP model demonstrates modest discrimination after external validation in the largest pediatric ECLS cohort to date.
• While overestimating mortality in high-risk patients, the PEP model remains the sole prediction tool for both neonates and children across all ECLS indications.
• The model retains potential utility for research and quality benchmarking in pediatric extracorporeal life support.