Objective: To systematically review the performance and predictors of risk prediction models for central venous catheter-related bloodstream infections (CRBSI) in Continuous Blood Purification(CBP) patients.Methods: We searched multiple databases from inception to July 14, 2025 for studies developing or validating CRBSI prediction models in hemodialysis patients.Two reviewers independently screened studies, extracted data, and assessed risk of bias using PROBAST.Meta-analysis of AUC values and predictors was performed.Results: Twenty-three studies involving 5,998 patients were included.CRBSI incidence ranged from 8.5% to 41.15%.The 28 developed models showed AUC values from 0.734 to 0.962, with pooled AUC of 0.879 (95% CI: 0.851-0.906).PROBAST assessment indicated high risk of bias but good applicability.Significant predictors included diabetes history, catheter insertion site, age, albumin levels, hemoglobin levels, smoking, antibiotic use, catheter reinsertion, dialysis vintage, catheter care frequency, catheter type, and previous catheter infection (all p < 0.05). Conclusion:Existing CRBSI prediction models demonstrate good discriminatory ability and clinical applicability.Identified risk factors can inform clinical early warning systems.However, methodological limitations and lack of external validation constrain current evidence.Future research should improve model development methods, standardize reporting, and conduct rigorous validation to develop more robust prediction tools.
Liu et al. (Fri,) studied this question.