Peripherally inserted central catheter-related bloodstream infections (PICC-CRBSI) pose a serious threat to preterm infants. This study aimed to develop and validate an interpretable machine learning model for risk assessment of PICC-CRBSI at the time of clinical suspicion. A total of 490 preterm infants who underwent PICC insertion in a tertiary hospital NICU in China were prospectively enrolled from January 2024 to October 2025. After feature selection, prediction models were constructed using four machine learning algorithms. Model performance was evaluated using area under the receiver operating characteristic curve (AUC), calibration, and decision curve analysis. The optimal model was interpreted using Shapley additive explanations (SHAP). CRBSI occurred in 68 patients (13.88%). The Random Forest model demonstrated the best performance, with AUC values of 0.973 and 0.934 in the training and validation sets, and overall accuracy of 0.945 and 0.905. SHAP analysis revealed that C-reactive protein(CRP), white blood cell count, and respiratory rate had the most significant influence on the model’s predictive performance. The random forest model demonstrated robust performance for risk assessment of PICC-CRBSI in preterm infants at the time of clinical suspicion. These findings may support clinical risk stratification and provide hypothesis-generating insights into key factors associated with PICC-CRBSI.
Guo et al. (Thu,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: