Construction and validation of a predictive model for mortality risk in patients with Staphylococcus aureus bloodstream infection

PURPOSE: To establish and validate a predictive model for the risk of death in patients with Staphylococcus aureus (S. aureus) bloodstream infection (BSI) to support clinical decision-making and patient management. METHODS: This study included demographic and clinical data from 206 patients with S. aureus BSI in China from January 2020 to June 2025. Variable selection was performed using Least Absolute Shrinkage and Selection Operator (LASSO) regression. Independent risk factors were then identified by multivariate Cox regression analysis. A prognostic model and corresponding nomogram were constructed. The models were evaluated using bootstrap, the area under the curve (AUC) of Receiver Operating Characteristic (ROC), decision curve analysis (DCA), and calibration curves. Finally, data from 60 patients with S. aureus BSI from other centers were used for external validation of the model. RESULTS: Based on the results of LASSO regression, the low red blood cell count (RBC), increased age, elevated C-reactive protein (CRP), elevated blood urea nitrogen (BUN), and low platelet counts (PLT) were used to construct the prognostic model. Among the aforementioned factors, the low RBC (hazard ratio [HR] of 0.41; 95% confidence interval [CI],0.22–0.76) and increased age (HR,1.04; 95%CI,1.00–1.07) were found to be independent risk factors for death in patients with S. aureus BSI. The results of bootstrap showed that the model’s bias and C-index were 0.003 and 0.735, respectively. The ROC curve shows that AUC values across three cohorts ranged from 0.724 to 0.831. These three calibration curves show that at 7, 14, and 28 days, the curves fluctuate around the 45°diagonal line. This indicates a good correlation between the actual risk and the predicted risk, demonstrating a high degree of calibration. The DCA curves showed that the model yielded relatively stable clinical net benefits for 28-day mortality risk prediction within the risk threshold range of 10 ~ 25%. CONCLUSION: RBC and age are independent risk factors for 28-day mortality in patients with S. aureus BSI. When combined with CRP, BUN, and PLT, they show certain prognostic predictive value. In patients with S. aureus BSI, our model could facilitate close clinical monitoring, prompt intervention, and improvement of patient prognosis.