The aim of this study was to evaluate procalcitonin (PCT) diagnostic accuracy in discriminating gram-negative (GN) from gram-positive (GP) bloodstream infections and determining the relationship between PCT levels, infection sites, and pathogen types.Methods:
Clinical and laboratory data were collected from patients with blood culture (BC)-positive sepsis between January 2014 and December 2015. PCT levels at different infection sites were compared, as was the presence of GN and GP bloodstream infection. A receiver operating characteristic (ROC) curve was generated to assess diagnostic accuracy.Results:
Of the 486 monomicrobial BCs, 254 (52.26%) were positive for GN bacteria (GNB), and 202 (42.18%) for GP bacteria (GPB). Median PCT levels were higher in BCs positive for GN (2.42 ng/ml, IQR: 0.38–15.52) than in those positive for GPB (0.49 ng/ml, IQR: 0.13–5.89) (P < 0.001). In the ROC analysis to differentiate between GNB and GPB, the area under the curve was 0.628 (95% CI: 0.576–0.679). When the cutoffs for PCT were 10.335 and 15.000 ng/ml, the specificity of GNB infection was 80.2% and 84.2%, respectively. PCT levels caused by GNB differed between Escherichia coli and Acinetobacter baumanni/Burkholderia cepacia, Klebsiella pneumonia and Acinetobacter baumanni. PCT levels caused by GPB differed between Staphylococcus epidermidis/Staphylococcus aureus and Staphylococcus hominis/Staphylococcus haemolyticus, Enterococcus faecium and Enterococcus faecalis/S.hominis/S. haemolyticus. Among patients with known infection sites, there were statistical differences in PCT levels between abdominal infection and pneumonia/infective endocarditis, urinary tract infection and pneumonia/catheter-related infection/infective endocarditis.Conclusion:
PCT can distinguish between GNB and GPB infection, as well as between different bacterial species and infection sites.