What is the clinical evidence from this study?

Study design: Observational. Population: pediatric ICU patients (n=8629). Intervention: Stacking ensemble machine learning model vs. Data subdivision approaches. Primary outcome: all-cause mortality (AUC-ROC 0.88).

June 17, 2026Open Access

Data subdivision approach enhances machine learning-based mortality prediction in pediatric ICU patients

Key Result

An ensemble stacking machine learning model demonstrated superior performance for predicting all-cause mortality in pediatric ICU patients compared to data subdivision, achieving an AUC-ROC of 0.88.

Key Points

The aim is to evaluate machine learning models for predicting all-cause mortality in pediatric ICU patients using biochemical data.
Retrospective analysis of 8,629 pediatric ICU patients
Applied Multiple Imputation for missing data and Lasso regression for feature selection
Model performance assessed with 5-fold cross-validation and SMOTE for class imbalance handling.
476 non-survivors (5.5%) identified among included patients.
Stacking ensemble achieved best AUC-ROC of 0.88; data subdivisions reached up to 0.83.
Calibration analysis revealed stacking model had the lowest Brier score at 0.04, showing better probabilistic accuracy.

Study Design

Type

Observational (n=8,629)

Structured PICO

Does a stacking ensemble machine learning model improve all-cause mortality prediction compared to single-model classifiers and data subdivision in pediatric ICU patients?

Population

8,629 pediatric ICU patients aged 28 days to 18 years, evaluated for all-cause mortality prediction using biochemical panels.

Exposure

Machine learning-based models (stacking ensemble and data subdivision strategies) using 22 biochemical variables measured on the first PICU day

Comparator

Single-model classifiers (Random Forest, CatBoost, Extra Trees, Logistic Regression)

Outcome

All-cause mortality predictionhard clinical

Ensemble stacking is a more effective strategy than data subdivision for addressing class imbalance in machine learning-based PICU mortality prediction.

Main Result

Effect estimate: AUC-ROC 0.88

Abstract

Objective To evaluate machine learning–based models for predicting all-cause mortality in pediatric ICU patients using comprehensive biochemical panels, with a focus on addressing missing data and class imbalance. Materials and methods A retrospective analysis was performed on a publicly available PICU dataset comprising 8,629 patients aged 28 days to 18 years. Twenty-two biochemical variables measured on the first PICU day were analyzed. Missing values were addressed using Multiple Imputation. Lasso regression was applied for feature selection. Models were trained using 5-fold cross-validation with the Synthetic Minority Oversampling Technique (SMOTE). Two ML-based imbalance-handling strategies, stacking ensemble and data subdivision were evaluated. Pairwise DeLong tests were used to compare AUC performance across models. Results Among the 8,629 included patients, there were 476 non-survivors (5.5 percent). Multiple imputation followed by SMOTE improved model performance across all algorithms. The single-model classifiers achieved AUC-ROC values of 0.82 (Random Forest), 0.79 (CatBoost), 0.83 (Extra Trees), and 0.79 (Logistic Regression). The stacking ensemble demonstrated the best overall performance, with an AUC-ROC of 0.88 and an AUC-PRC of 0.45. The data subdivision approaches also produced strong discriminative performance, achieving AUC-ROC value up to 0.83 for three-subdivision, and 0.82 for five-subdivision strategies. Calibration analysis showed that the stacking model achieved the lowest Brier score (0.04), indicating superior probabilistic accuracy compared with individual classifiers. Feature importance analyses across all MI-based models consistently highlighted coagulation markers (D-dimer, reference TT, PTT, INR), electrolytes (chloride, potassium, sodium), and metabolic and organ-dysfunction indicators (AST, ALT, creatinine) as key predictors of mortality. Conclusions This study demonstrates that ensemble stacking is a more effective strategy than data subdivision for addressing class imbalance in PICU mortality prediction.