Related Experiment Video
Updated: Sep 1, 2025
Author Spotlight: Exploring Venous Waveforms in Porcine Models to Tackle Volume Overload in Medicine
Published on: January 12, 2024
Clinical implications of discrepancies in predicting pediatric mortality between Pediatric Index of Mortality 3 and
Eui Jun Lee1, Bongjin Lee2, You Sun Kim3
1Department of Emergency Medicine, Seoul National University College of Medicine, Seoul, Korea.
Insights
The Pediatric Index of Mortality 3 (PIM 3) and Pediatric Logistic Organ Dysfunction-2 (PELOD-2) tools predict pediatric mortality. Factors like oxygen saturation and specific conditions influence their differing predictions, requiring careful interpretation.
Area of Science:
- Pediatric critical care medicine
- Clinical informatics
- Biostatistics
Background:
- Pediatric Index of Mortality 3 (PIM 3) and Pediatric Logistic Organ Dysfunction-2 (PELOD-2) are established tools for predicting pediatric mortality.
- Existing research indicates potential variations in their predictive performance based on patient characteristics.
- This study aimed to identify factors contributing to discrepancies in mortality predictions between PIM 3 and PELOD-2.
Purpose of the Study:
- To investigate the clinical factors associated with differences in mortality prediction between PIM 3 and PELOD-2 in pediatric intensive care unit (PICU) patients.
- To analyze the impact of specific patient conditions and physiological parameters on the divergence of these two mortality prediction tools.
Main Methods:
- A retrospective analysis of 945 pediatric patients (<18 years) admitted to a PICU between July 2017 and May 2019.
- Defined "difference in mortality prediction" as PIM 3 predicted mortality minus PELOD-2 predicted mortality.
- Utilized multivariable regression analysis to identify significant clinical characteristics and employed Hosmer-Lemeshow tests and AUROC curves to assess predictive performance.
Main Results:
- Both PIM 3 (AUROC=0.889) and PELOD-2 (AUROC=0.731) demonstrated good predictive performance.
- Multivariable analysis identified oxygen saturation, CO2 partial pressure, base excess, platelet count, and blood urea nitrogen as significant physiological factors.
- Specific conditions including cardiac bypass surgery, seizures, cardiomyopathy, necrotizing enterocolitis, cardiac arrest, leukemia/lymphoma, bone marrow transplant, and liver failure were significantly associated with prediction differences (P<0.001).
Conclusions:
- While both PIM 3 and PELOD-2 effectively predict observed mortality in critically ill children, their predictions can differ.
- Interpretation of predictions from these tools requires consideration of specific clinical characteristics and patient conditions.
- Further research may refine the application of these tools in diverse pediatric populations.
Background:
Pediatric Index of Mortality 3 (PIM 3) and Pediatric Logistic Organ Dysfunction-2 (PELOD-2) are validated tools for predicting mortality in children. Research suggests that these tools may have different predictive performance depending on patient group characteristics. Therefore, we designed this study to identify the factors that make the mortality rates predicted by the tools different.
Methods:
This retrospective study included patients (<18 years) who were admitted to a pediatric intensive care unit from July 2017 to May 2019. After defining the predicted mortality of PIM 3 minus the predicted mortality rate of PELOD-2 as "difference in mortality prediction," the clinical characteristics significantly related to this were analyzed using multivariable regression analysis. Predictive performance was analyzed through the Hosmer-Lemeshow test and area under the receiver operating characteristic curve (AUROC).
Results:
In total, 945 patients (median [interquartile range] age, 3.0 [0.0-8.0] years; girls, 44.7%) were analyzed. The Hosmer-Lemeshow test revealed AUROCs of 0.889 (χ2=10.187, P=0.313) and 0.731 (χ2=6.220, P=0.183) of PIM 3 and PELOD-2, respectively. Multivariable linear regression analysis revealed that oxygen saturation, partial pressure of CO2, base excess, platelet counts, and blood urea nitrogen levels were significant factors. Patient condition-related factors such as cardiac bypass surgery, seizures, cardiomyopathy or myocarditis, necrotizing enterocolitis, cardiac arrest, leukemia or lymphoma after the first induction, bone marrow transplantation, and liver failure were significantly related (P<0.001).
Conclusions:
Both tools predicted observed mortality well; however, caution is needed in interpretation as they may show different prediction results in relation to specific clinical characteristics.
Related Concept Videos
01:26Pharmacokinetic Models: Comparison and Selection Criterion
Physiological models take a detailed approach by considering specific molecular processes. They can predict drug distribution, metabolism, and elimination changes, providing a comprehensive understanding of how drugs interact with the body.
01:20Comparing the Survival Analysis of Two or More Groups