A predictive model for PICC-related thrombosis in sepsis patients using XGBoost algorithm - PubMed
4 hours ago
- #Sepsis
- #PICC
- #Thrombosis
- A predictive model for PICC-related thrombosis in sepsis patients was developed using the XGBoost algorithm.
- The study analyzed data from 8,128 ICU patients with sepsis using PICC from the MIMIC-IV 3.1 database.
- The model achieved an AUC of 0.761 in the training set and 0.766 in the validation set, indicating good performance.
- Key predictors included white blood cell count, platelet count, history of myocardial infarction, hemoglobin levels, and PICC indwelling time.
- The XGBoost model demonstrated clinical utility, outperforming treat-all/none strategies with a net benefit of 0.31 at a 20% risk threshold.
- The study highlights the potential of the XGBoost model in guiding clinical decision-making for high-risk sepsis patients.