Development and validation of a clinical prediction rule to distinguish bacterial from viral pneumonia in children

Background. Distinguishing bacterial from viral pneumonia on admission to the hospital could guide the decision of whether or not to use antibiotics. We developed and validated a clinical prediction rule to distinguish bacterial from viral pneumonia in hospitalized children. Methods. We enrolled consecutive children, aged 1 month to 5 years, admitted to two tertiary children's hospitals in whom a bacterial or viral etiology for pneumonia was identified. Data from 175 children at one hospital were used to develop a clinical prediction rule or Bacterial Pneumonia Score (BPS). Data from 136 children at the second hospital were used to validate it. Based on receiver operating characteristic (ROC) curve analyses and multivariable logistic regression, significant clinical and laboratory predictors, along with the radiographic score for each participant, were included as factors in the BPS. The main outcome measure was the sensitivity and specificity of the BPS for bacterial pneumonia. Results. The BPS (possible range: -3 to 15; auROCc=0.996,95%CI: 0.99-1.0) was developed by attributing 3 points for axillary temperature >= 8,000 cells/mm(3), 1 point for bands >= 5%, and -3 to 7 points for the chest X-ray findings. A BPS >= 4 predicted bacterial pneumonia with a sensitivity of 100%, specificity of 93.8%, positive predictive value of 75.8%, and negative predictive value of 100%. Conclusions. The BPS accurately identifies hospitalized children's risk of bacterial pneumonia, helping clinicians determine those not likely to benefit from antibiotic therapy.