Combination of Clinical Symptoms and Blood Biomarkers Can Improve Discrimination Between Bacterial or Viral Community-Acquired Pneumonia in Children

Mejbah U. Bhuiyan; Christopher C. Blyth; Rachel West; Jurissa Lang; Tasmina Rahman; Caitlyn Granland; Camilla de Gier; Meredith L. Borland; Ruth B. Thornton; Lea-Ann S. Kirkham; Andrew Martin; Peter C. Richmond; David W. Smith; Adam Jaffe; Thomas L. Snelling


BMC Pulm Med. 2019;19(71) 

In This Article

Abstract and Introduction


Background: Differentiating bacterial from viral pneumonia is important for guiding targeted management and judicious use of antibiotics. We assessed if clinical characteristics and blood inflammatory biomarkers could be used to distinguish bacterial from viral pneumonia.

Methods: Western Australian children (≤17 years) hospitalized with radiologically-confirmed community-acquired pneumonia were recruited and clinical symptoms and management data were collected. C-reactive protein (CRP), white cell counts (WCC) and absolute neutrophil counts (ANC) were measured as part of routine care. Clinical characteristics and biomarker levels were compared between cases with definite bacterial pneumonia (clinical empyema and/or bacteria detected in blood or pleural fluid), presumed viral pneumonia (presence of ≥1 virus in nasopharyngeal swab without criteria for definite bacterial pneumonia), and other pneumonia cases (pneumonia in the absence of criteria for either definite bacterial or presumed viral pneumonia). The area-under-curve (AUC) of the receiver operating characteristic (ROC) curve for varying biomarker levels were used to characterise their utility for discriminating definite bacterial from presumed viral pneumonia. For biomarkers with AUC > 0.8 (fair discriminator), Youden index was measured to determine the optimal cut-off threshold, and sensitivity, specificity, predictive values (positive and negative) were calculated. We investigated whether better discrimination could be achieved by combining biomarker values with the presence/absence of symptoms.

Results: From May 2015 to October 2017, 230 pneumonia cases were enrolled: 30 with definite bacterial pneumonia, 118 with presumed viral pneumonia and 82 other pneumonia cases. Differences in clinical signs and symptoms across the groups were noted; more definite bacterial pneumonia cases required intravenous fluid and oxygen supplementation than presumed viral or other pneumonia cases. CRP, WCC and ANC were substantially higher in definite bacterial cases. For a CRP threshold of 72 mg/L, the AUC of ROC was 0.82 for discriminating definite bacterial pneumonia from presumed viral pneumonia. Combining the CRP with either the presence of fever (≥38°C) or the absence of rhinorrhea improved the discrimination.

Conclusions: Combining elevated CRP with the presence or absence of clinical signs/symptoms differentiates definite bacterial from presumed viral pneumonia better than CRP alone. Further studies are required to explore combination of biomarkers and symptoms for use as definitive diagnostic tool.