Mining of a Clinical Database: The Interpretation of Intense Serial Procalcitonin in the Prediction for Bloodstream Infection

Background: Procalcitonin (PCT) is a promising biomarker for predicting infection. Bloodstream infection (BSI) is usually a deteriorating stage of sepsis. The purpose of this study was to explore the predictive value of intense serial PCT assays for BSI in the intensive care unit (ICU). Methods: This study was a retrospective study based on a clinical database. We analyzed the data of critically ill patients from February 2016 to May 2020. The patients who received PCT assays and blood cultures (BCs) were classified into four groups according to the BCs: (i) BC negative, (ii) bacteria positive, (iii) fungi-positive, and (iv) combined-positive, and the patients with bacteremia were further subdivided into Gram+ and Gram– bacteremia. Results: The database included 11,219 patients. There were 3,593 patients who met the criteria for the analysis. The PCT concentration differed significantly across BC groups (p < 0.0001). The fluctuation of PCT significantly increased in the BC positive groups (p < 0.0001). According to the receiver operating characteristic (ROC), the optimum cutoff of the fluctuation of PCT was around 8 ng/ml for predicting BSI. Conclusion: Our study indicated that the fluctuation of PCT could be an indicator for screening BSI, but less accurate for Gram-positive infections. With a fluctuation of PCT less than 8 ng/ml, BSI should not be a rational cause for sepsis exacerbating.