Abstract
BACKGROUND: Discrimination of bacterial and viral etiologies of childhood community-acquired pneumonia (CAP) is often challenging. Unnecessary antibiotic administration exposes patients to undue risks and may engender antimicrobial resistance. This study aimed to develop a prediction model using epidemiological, clinical and laboratory data to differentiate between bacterial and viral CAP. METHODS: Data from 155 children with confirmed bacterial or mixed bacterial and viral infection (N = 124) and viral infection (N = 31) were derived from a comprehensive assessment of causative pathogens [Partnerships for Enhanced Engagement in Research-Pneumonia in Pediatrics (PEER-PePPeS)] conducted in Indonesia. Epidemiologic, clinical and biomarker profiles (hematology and inflammatory markers) were compared between groups. The area under the receiver operating characteristic curve (AUROC) for varying biomarker levels was used to characterize performance and determine cut-off values for discrimination of bacterial and mixed CAP versus viral CAP. Diagnostic predictors of bacterial and mixed CAP were assessed by multivariate logistic regression. RESULTS: Diarrhea was more frequently reported in bacterial and mixed CAP, while viral infections more frequently occurred during Indonesia's rainy season. White blood cell counts (WBC), absolute neutrophil counts (ANC), neutrophil-lymphocyte ratio (NLR), C-reactive protein (CRP), and procalcitonin (PCT) were significantly higher in bacterial and mixed cases. After adjusting for covariates, the following were the most important predictors of bacterial or mixed CAP: rainy season (aOR 0.26; 95% CI 0.08-0.90; p = 0.033), CRP ≥5.70 mg/L (aOR 4.71; 95% CI 1.18-18.74; p = 0.028), and presence of fever (aOR 5.26; 95% CI 1.07-25.91; p = 0.041). The model assessed had a low R-squared (Nagelkerke R(2) = 0.490) but good calibration (p = 0.610 for Hosmer Lemeshow test). The combination of CRP and fever had moderate predictive value with sensitivity and specificity of 62.28 and 65.52%, respectively. CONCLUSION: Combining clinical and laboratory profiles is potentially valuable for discriminating bacterial and mixed from viral pediatric CAP and may guide antibiotic use. Further studies with a larger sample size should be performed to validate this model.