Development and validation of a risk prediction model for hospital mortality in adult patients with Staphylococcus aureus bacteremia.

BACKGROUND

Staphylococcus aureus bacteremia (SAB) is a life-threatening infection with high mortality. Early identification of high-risk patients remains challenging. This study aimed to develop a prediction model for in-hospital mortality in SAB patients and compare its performance with the SOFA score.

METHODS

This single-center retrospective study enrolled 160 patients with SAB (2019-2024). Patients were categorized into survival (n=87) and non-survival (n=73) groups. Variable selection was performed using best subset, stepwise, and LASSO regression. A multivariable logistic regression model was developed and internally validated using bootstrap resampling. A nomogram incorporating six predictors (PaO2/FiO2 ratio, urea, albumin, RDW, MV, and platelet count) was constructed to provide individualized risk prediction. Model performance was compared with the SOFA score using ROC curves, calibration plots, and decision curve analysis.

RESULTS

The in-hospital mortality rate was 45.6%. Independent risk factors included invasive mechanical ventilation (OR 4.738), elevated RDW (OR 1.252), and elevated urea (OR 1.065), while higher platelet count (OR 0.991), albumin (OR 0.825), and PaO2/FiO2 ratio (OR 0.988) were protective (all P < 0.05). The model showed excellent discrimination (AUC 0.948) and calibration (Hosmer-Lemeshow P = 0.658). The SOFA score showed comparable discrimination (AUC 0.959) but suboptimal calibration (P = 0.026). No significant difference in AUC was observed (P = 0.565).

CONCLUSIONS

Invasive mechanical ventilation, elevated RDW, and elevated urea are independent risk factors for mortality in SAB, while higher platelet count, albumin, and PaO2/FiO2 ratio are protective. The proposed nomogram provides excellent predictive performance comparable to the SOFA score and may aid early identification of high-risk patients.