Abstract / Summary
Abstract Background Bacteremia is a serious and potentially life-threatening complication in patients with cancer, requiring prompt recognition and early initiation of treatment. However, early identification of bacteremia in febrile cancer patients presenting to the emergency department (ED) remains challenging due to nonspecific clinical presentations. This study aimed to develop and internally validate a simplified prediction model for bacteremia in febrile patients with solid tumors in the ED. Methods We conducted a retrospective study of adult patients with solid tumors who presented with fever to the ED of a tertiary academic hospital between January and December 2024. Bacteremia was defined by blood culture. Candidate predictors were evaluated using least absolute shrinkage and selection operator (LASSO) regression followed by multivariable logistic regression. Model performance was assessed by discrimination, calibration, decision curve analysis, and internal 10-fold cross-validation. Results A total of 466 febrile ED presentations were included, of which 82 (17.6%) had bacteremia. The final four-variable model comprised procalcitonin, pulse rate, neutrophil-to-lymphocyte ratio, and albumin. It showed strong discrimination (AUC 0.948; 10-fold cross-validated AUC 0.947), compared with an AUC of 0.922 for procalcitonin alone. At the optimal cutoff, sensitivity was 0.819 and specificity was 0.923. Cross-validated calibration showed an intercept of − 0.144 and a slope of 0.913. Conclusions This simplified model demonstrated strong internal performance and may provide a promising approach to early bacteremia risk stratification in febrile patients with solid tumors. Independent external validation and prospective evaluation of clinical utility are required before routine clinical implementation. Clinical trial number Not applicable.