RESUMO
This study was designed to determine the utility of procalcitonin (PCT) and C-reactive protein (CRP) as predictors of Gram-negative bloodstream infection (GN-BSI) in hematological febrile outpatients at the time of the emergency unit admission. Overall, 286 febrile episodes, which included 42 GN-BSI (16%), were considered. PCT levels at patient admission were statistically higher in GNB-BSI when compared to Gram-positive bacteria BSI (median 4.06 ng/ml (range 1.10-25.04) vs 0.88 ng/ml (0.42-10), p<0.03) and to all other fever etiologies. For CRP, differences within fever etiologies were less profound but statistically significant, except for GN-BSIs vs GP BSIs (p=0.4). ROC analysis of PCT showed that an AUC of 0.85 (95%CI 0.79-0.95) discriminated GN-BSI from all other fever etiologies, with a best cut-off of 0.5 ng/ml, a negative predictive value (NPV) of 98%, and a negative likelihood ratio (negLR) of 0.1. ROC analysis of CRP showed an AUC of 0.67 (95%CI 0.53-0.81) with a best cut-off of 6.64 mg/dl, a NPV of 94%, and a negLR of 0.33. This study confirms that 0.5 ng/ml represents the PCT best cut-off to differentiate the cause of fever and rule out a GN-BSI in febrile hematologic outpatients at the time of the emergency unit admission. Therefore, introducing PCT testing could be a valid measure in order to tailor a more precise prompt antimicrobial therapy to the febrile outpatient while waiting for blood culture results.