Validation of a simplified prediction rule to identify etiology in children with pneumonia

INTRODUCTION: Identifying on admission those children with bacterial pneumonia could reduce inappropriate antibiotic use. The BPS (Bacterial Pneumonia Score) is a clinical prediction rule that accurately identifies children with bacterial pneumonia. Because the interpretation of chest X-ray included in this model could be considered difficult, a simplified version was developed, but this version has not yet been validated in a different population. OBJECTIVE: To validate a simplified clinical prediction rule to identify children with an increased risk of having bacterial pneumonia. METHODS: Children aged under 5 years, hospitalized for pneumonia (viral or bacterial) were included. On admission, axillary temperature, age, absolute neutrophil count, bands, and chest radiograph were evaluated. RESULTS: We included 168 patients (23 with bacterial pneumonia and 145 with viral pneumonia). Those with bacterial pneumonia showed a score higher than those with viral pneumonia (5.3 ± 2.5 vs. 2.6 ± 2.02; p <0.001). A score =3 points was identified as the optimum cutoff value to predict bacterial pneumonia (aucROC= 0.79; 95% IC: 0.68-0.90), and was more frequent among patients with bacterial than viral pneumonia (19/23 vs. 42/145, p= 0.003; OR: 4.8; CI95%: 1.4-17.6), achieving 82.6% sensitivity, 50.3% specificity, 20.9% positive predictive value, 94.8% negative predictive value, 1.66 positive likelihood ratio and 0.35 negative likelihood ratio. CONCLUSIONS: The evaluated simplified prediction rule showed a limited diagnostic accuracy on identifying children with bacterial pneumonia, being less accurate than the BPS.