Risk prediction of pediatric intensive care unit admission in children with respiratory syncytial virus infection using interpretable machine learning - Summary - MDSpire
Advertisement
Risk prediction of pediatric intensive care unit admission in children with respiratory syncytial virus infection using interpretable machine learning
To develop and validate an interpretable machine learning model for predicting PICU admission in hospitalized children with RSV.
Approach:
Data Collection: Included children from January 2023 to March 2025, with a focus on predicting PICU admission within 48 hours of hospital admission.
Key Findings:
Among 1,606 children, 209 required PICU admission.
Final predictors included dyspnea, serum ferritin, wheezing, immunoglobulin G, interleukin-6, preterm birth, and personal history of wheezing.
Random forest model achieved AUROC of 0.94 in internal test set and 0.92 in external validation cohort.
Interpretation:
The model demonstrated acceptable calibration and interpretability, supporting early risk stratification for PICU admission in children with RSV.
Limitations:
Development estimates may be optimistic without bootstrap-based optimism correction.
The small size of the temporal external validation cohort limits precision.
Further validation is needed across different geographical and healthcare settings.
Conclusion:
An interpretable machine learning model was developed to predict PICU transfer in children with RSV, showing good performance but requiring further validation.
