Abstract / Summary
Conventional bacterial cultures are time-consuming and lack sensitivity for early-stage infections, often delaying targeted therapy and promoting antibiotic misuse. This study aimed to evaluate a rapid biomarker panel combining cytokine profiles and routine markers for early etiological differentiation in febrile patients. This study enrolled febrile patients admitted to Sun Yat-sen Memorial Hospital, Sun Yat-sen University, between 2021 and 2023. Based on final clinical diagnoses and microbiological results, 457 eligible patients were classified into bacterial, viral, or fungal infection groups. A panel of cytokines (IL2R, IL-6, IL-8, IL-10, TNF-α, IL-1β) and routine markers (PCT, CRP, NLR) was assessed. Receiver operating characteristic (ROC) curve analysis was performed to determine optimal cut-off values and diagnostic performance. Among the 457 febrile patients, the biomarker panel demonstrated good discriminative ability across multiple clinical scenarios. For distinguishing bacterial from viral infections, a model combining IL-6/IL-10 ratio > 1.13, IL-8 > 28.70 pg/mL, and NLR > 2.96 achieved an AUC of 0.80 (95% CI: 0.74–0.87). For discriminating Gram-negative (G-) versus Gram-positive (G+) bacterial infections, site-specific models were required; notably, in bloodstream infections, a model incorporating IL-6 > 44.91 pg/mL and PCT > 2.30 ng/mL yielded an AUC of 0.82 (95% CI: 0.70–0.93). Furthermore, for identifying sepsis among bacterial infections, the combination of IL-8 > 72.45 pg/mL and PCT > 1.60 ng/mL showed an AUC of 0.80 (95% CI: 0.72–0.89). This study established a practical, hierarchical diagnostic pathway utilizing routine immunological biomarkers. This strategy enables the sequential determination of infection etiology, pathogen subtype, and severity, achieving AUC values of 0.80–0.82 across key diagnostic steps. It provides a cost-effective, rapid decision-support tool to facilitate early clinical decisions and antibiotic stewardship, particularly for medical institutions with limited diagnostic resources.