Abstract
Background/Objectives: Bloodstream infection (BSI) is a significant cause of mortality. The availability of a convenient tool for predicting the risk of BSI at the early stage would be beneficial for clinicians, allowing them to improve the outcomes of BSI and avoid antibiotic overuse. Methods: A multivariate prediction model was constructed based on conventional laboratory test results and novel serum inflammatory markers in a cohort of patients with suspected BSI over a one-year period using least absolute shrinkage and selection operator (LASSO) and logistic regression. Results: BSI was confirmed in 99 (32.0%) of the 309 enrolled patients. Five readily available markers were identified as independent predictors: the presence of local infection, platelet count, and C-reactive protein, procalcitonin (PCT), and CD64 levels. A nomogram based on these five variables achieved an area under the receiver operating characteristic curve of 0.85 in predicting the risk of BSI. The nomogram was superior to PCT alone in terms of the net clinical benefits obtained in a rather wide range of threshold probabilities. Conclusions: The simple five-variable nomogram developed in this study is useful for timely prediction of individuals at high risk of BSI. It may be used in clinical practice to facilitate timely decision-making on antimicrobial treatment and avoid inappropriate overuse of antibiotics.