Combined Predictive Value of Serum Procalcitonin and Platelet Indices in Differentiating Gram-Negative and...

Sepsis is caused by bloodstream infections (BSIs) and potentially results in death. Parameters like platelet (PLT) indices and procalcitonin (PCT) are common inflammatory markers used for the diagnosis and assessment of sepsis. Thus, the present study aimed to explore the predictive value of combined serum PCT and PLT indices to differentiate the host responses of Gram-positive bacteria (GPB) and Gram-negative bacteria (GNB) in BSIs.  Materials and methods: A prospective cohort study was conducted at the Critical Care and General Medicine Department of Kalinga Institute of Medical Sciences, a tertiary care hospital in Bhubaneswar, Odisha, India, by including 385 BSI-diagnosed patients consecutively over the period of two years. All patients were more than 18 years of age and had culture-positive sepsis diagnosed based on the sepsis-3 criteria. Laboratory investigations such as complete blood count, serum PCT, and bacterial cultures for GPB and GNB, were done and correlated with clinical features. In statistical analysis, univariate and multivariate analyses (logistic regression analysis) were done, and a multivariate regression model was proposed by combining PCT and PLT indices for both Gram-positive and Gram-negative BSI cases.  Results: BSI cases were confirmed from culture-positive. Out of 385 BSI cases, 51.9% were Gram-negative and 48.05% were Gram-positive. Escherichia coli and Staphylococcus aureus were the most common organisms. A significantly higher level was observed in hematological parameters such as white blood cell count (WBC) (18.02 cells/µL ± 9.06) and PLT indices; mean platelet volume (MPV) (12.24 fL ± 2.02), PLT (159.11 x 10³ mcl ± 76.7); and platelet distribution width (PDW) (14.13 fL ± 13.03) in GNB. Elevated levels were also observed in PCT (37.20 ± 40.4 µg/L), a key biomarker for bacterial infection and sepsis. Generalized linear model (GLM) based on the combined value of the above significant hematological parameters demonstrates a strong ability to discriminate GNB infection, with an area under the curve (AUC) of 0.8644. The optimal threshold for this model is 0.433, which provides a balanced sensitivity and specificity, both at 0.80. The high value of the AUC indicates the potential utility of the GLM-based prediction tool to classify the GNB infection.  Conclusion: A multi-biomarker linear model, by combining PCT, WBC, MPV, PDW, and PLT-derived ratios, can effectively differentiate Gram-negative from Gram-positive BSIs. Elevated PCT, WBC, and PLT indices indicate Gram-negative infections.