Abstract: Objective  To systematically evaluate the risk prediction model for catheter-related bloodstream infection in hemodialysis patients and to provide references for selecting appropriate prediction models or developing new models.  Methods  We searched related literature in CINAHL, PubMed, Web of Science, Cochrane Library, Embase, Wanfang database, CNKI, VIP website, China and Chinese biomedical literature database from establishing the database to October 1, 2023. Two investigators independently screened the literature, extracted data, and analyzed the risk of bias and applicability of included literature using the prediction model risk of bias assessment tool (PROBAST).  Results  A total of seven articles were included. Age, combined diabetes, serum albumin level, hand hygiene and catheter retention time were the major predictors for catheter-related bloodstream infection. The area under the curve was 0.734 to 0.889, from which four models were calibrated. All studies were well applicable but had a risk of bias.  Conclusion  The risk prediction models of catheter-related bloodstream infection in hemodialysis patients have better prediction performance, but with higher risks of bias caused by methodological defects, such as improper processing of missing data, variable selection without appropriate methods, no mention of blindness, and others. Therefore, they cannot be directly applied in clinical practice yet. In the future, the existed models should be verified in-depth and extensively by prospective studies using large sample and multiple population to develop prediction models with excellent predictive performance and simple use.

Key words: Hemodialysis, Catheter-related infection, Prediction model, Systematic evaluation