Prediction of moderate to severe bleeding risk in pediatric immune thrombocytopenia using machine learning.

Abstract

This study aimed to develop and validate a risk prediction model for moderate to severe bleeding in children with immune thrombocytopenia (ITP). Data from 286 ITP patients were prospectively collected and randomly split into training (80%) and test (20%) sets. Least Absolute Shrinkage and Selection Operator (LASSO) regression was used for feature selection. Among seven machine learning algorithms, the eXtreme Gradient Boosting (XGBoost) model demonstrated the best performance (AUC = 0.886, 95% CI: 0.790-0.982) and was selected as the optimal model. Shapley Additive Explanations (SHAP) were used for model interpretation, identifying child age, age at diagnosis, and initial platelet count as key predictors of moderate to severe bleeding risk.

Conclusion: The XGBoost-based prediction model shows strong predictive performance and could assist healthcare providers in identifying high-risk ITP patients, supporting appropriate clinical decision-making.

Trial registration number: ChiCTR2100054216, December 11, 2021 What is Known: • Current clinical practice relies solely on platelet counts to guide hospitalization and treatment in ITP children, often overlooking bleeding manifestations, leading to delayed or inappropriate treatment. Existing severe bleeding risk prediction models are primarily designed for adults and lack applicability to children.

What is new:  • This study prospectively collected data, enhancing accuracy. A novel machine learning-based prediction model was developed to assess moderate to severe bleeding risk in pediatric ITP patients.