Abstract
Only a small proportion of blood cultures routinely performed in emergency department (ED) patients is positive. Multiple clinical scores and biomarkers have previously been examined for their ability to predict bacteremia. Conclusive clinical validation of these scores and biomarkers is essential.This observational cohort study included patients with suspected infection who had blood culture sampling at ED admission. We assessed 5 clinical scores and admission concentrations of procalcitonin (PCT), C-reactive protein (CRP), lymphocyte and white blood cell counts, the neutrophil-lymphocyte count ratio (NLCR), and the red blood cell distribution width (RDW). Two independent physicians assessed true blood culture positivity. We used logistic regression models with area under the curve (AUC) analysis.Of 1083 patients, 104 (9.6%) had positive blood cultures. Of the clinical scores, the Shapiro score performed best (AUC 0.729). The best biomarkers were PCT (AUC 0.803) and NLCR (AUC 0.700). Combining the Shapiro score with PCT levels significantly increased the AUC to 0.827. Limiting blood cultures only to patients with either a Shapiro score of ≥4 or PCT > 0.1 μg/L would reduce negative sampling by 20.2% while still identifying 100% of positive cultures. Similarly, a Shapiro score ≥3 or PCT >0.25 μg/L would reduce cultures by 41.7% and still identify 96.1% of positive blood cultures.Combination of the Shapiro score with admission levels of PCT can help reduce unnecessary blood cultures with minimal false negative rates.The study was registered on January 9, 2013 at the 'ClinicalTrials.gov' registration web site (NCT01768494).