The diagnostic accuracy of biomarkers for the prediction of bacteremia in patients with suspected infection: a prospective observational study

Rapid recognition of bacteremia is important for critical care, especially in patients with suspected bloodstream infections. Procalcitonin and presepsin are widely used biomarkers in point-of care medical testing for identifying infectious diseases and sepsis; however, the diagnostic accuracy for the prediction of bacteremia is not well established. Therefore, this study aimed to evaluate the diagnostic accuracy of procalcitonin and presepsin for the prediction of bacteremia in patients with suspected bacteremia. We performed a prospective observational study at our hospital. A total of 210 patients (307 samples) who had been admitted from December 2014 through September 2016 with a suspected infection were included. Presepsin and procalcitonin were tested simultaneously with blood cultures and routine laboratory tests. One hundred and four blood samples were obtained at the emergency room (ER). Others were obtained during hospital admission. Blood cultures were positive in 34 samples; 25 samples were obtained in the ER. Presepsin and procalcitonin levels were significantly higher in patients with positive blood cultures than in those with negative blood cultures (1028.5 pg/mL vs. 485.0 pg/mL, P < 0.001 and 4.53 ng/mL vs. 0.33 ng/mL, P < 0.001, respectively). For predicting bacteremia, receiver operating characteristic curve analysis for presepsin showed an area under the curve (AUC) of 0.718 and negative predictive value (NPV) of 95%. The analysis for procalcitonin showed an AUC of 0.778 and NPV of 94.8%. C-reactive protein tests and the quick Sequential Organ Failure Assessment score in the ER failed to be useful tools for predicting bacteremia. Based on our results, procalcitonin and presepsin showed good diagnostic accuracy and NPV for predicting bacteremia among patients with suspected infection. Therefore, these biomarkers may be useful for ruling out bacteremia in patients with suspected infection.

Shiino Y, et al