Systematic evaluation of risk prediction model for methicillin-resistant Staphylococcus aureus infection

Objective To retrieve relevant literature on risk prediction model for methicillin-resistant Staphylococcus aureus (MRSA) infection among hospitalized patients from databases and evaluate the predictive model.

Methods The literature on risk prediction models for MRSA infection among hospitalized patients was retrieved from PubMed, Embase, Scopus, Cochrane library, China National Knowledge Infrastructure (CNKI), WanFang data, and VIP database, with a time range from the inception of the database to January 1, 2024. Two researchers independently screened the literature, extracted data. The Prediction Model Risk of Bias Assessment Tool (PROBAST) was applied to evaluate the risk of bias and applicability of the prediction model in the literature, and descriptive analysis was conducted.

Results A total of 12 articles (15 prediction models) were included in this study, with significant differences in the total sample size, the number of MRSA infection events, sample size of modeling, and sample size of validation among the studies. Common predictors in the prediction models were admission to the intensive care unit, antibiotic use, history of residence in nursing facilities, age, chronic kidney disease, and previous hospitalization history. Nine articles conducted internal validation, and three articles conducted both internal and external validation. Nine articles reported the area under the receiver operating characteristic curve, and only three articles reported the calibration of the model based on the Hosmer-Lemeshow test. PROBAST analysis showed that 10 articles were assessed as high risk bias, mainly stemming from statistical analysis.

Conclusion Most of the MRSA infection risk prediction models in the current literature have good predictive efficacy for MRSA infection, but they all have higher overall risk of bias, and only a few models have undergone external validation. Researchers should follow PROBAST standards to construct and externally validate models in the future so as to develop models suitable for clinical practice.