Inflammatory Markers Preceding the Microbiological Result in 53,033 Blood Cultures

Authors:
Open Access
Publisher website
Google Scholar
Abstract
Inflammatory biomarkers are read at the bedside as one value against one threshold, although serial determinations can be summarized in several ways, and the summary being tested is seldom stated. We measured the extent to which the discrimination attributed to a biomarker depends on the summary is read. We analyzed 53,033 blood cultures from 16,431 patients at one hospital between January 2020 and August 2026, anchoring each culture to its microbiological result and deriving, for eight biomarkers, the first value before the result, the maximum before it, the value nearest it and the amplitude between determinations. The outcomes were pathogen growth, excluding coagulase-negative staphylococci, and bacterial group. In the 35,150 cultures with a procalcitonin determination preceding the result, procalcitonin provided the best discrimination on every metric: area under the curve, 0.766 (95% CI 0.756–0.776) for the maximum before the result, 0.738 (0.728–0.748) for the first value, and 0.773 (0.764–0.784) nearest the result. The advantage of the maximum grew once sampling density and interval length were held constant. The neutrophil-to-lymphocyte ratio performed better before the result than near it, 0.707 versus 0.620, and fell from 0.738 on surgical wards to 0.512 in hematology, whereas procalcitonin varied little between services. Sensitivity at 0.5 ng/mL ranged from 88.0% for Enterobacterales to 55.6% for other Gram-positive organisms. Combining biomarkers improved the area by, at most, 0.015. Which number is read changes the answer as much as which marker is chosen. A procalcitonin level below 0.5 ng/mL argues strongly against Gram-negative bacteremia and far more weakly against enterococcal, other Gram-positive, and fungal bloodstream infection.

This publication has 25 references indexed in Scilit: