RESUMO
BACKGROUND: The etiology of pneumonia is important to indicate antibiotics. A clinical prediction score (RP) has been designed, although the radiological interpretation is not easy. OBJECTIVE: To design a simple prediction score (PRs) to identify etiology in children with pneumonia, including radiological patterns, clinical and laboratory features. METHODS: Cross sectional study. We prospectively included children under 5 years hospitalized for pneumonia with microbiological evidence (2007-2008). According to the RP, were allocated 3 points when the temperature value was ≥ 39 º C, 2 when the patient age was ≥ 9 months, 2 when the number of neutrophils was 8000/mm3 and 1 when the immature neutrophils number was ≥ 5%. Radiography was evaluated as one point when consolidation was diagnosed and 0 point when pleural effusion or other infiltrations were present. RPS range was from 0 to 9 points. We determined the best cutoff for predicting bacterial pneumonia (ROC) and was calculated based on the same sensitivity (S), specificity (Sp), positive predictive value (PPV) and negative (NPV) and positive likelihood ratio ( RVP) and negative (NLR). RESULTS: 196 patients (viral: 82%, bacteria: 18%), 8.7 ± 10 months. We identified a score ≥ 3 (auROCc = 0.87 95% CI 0.81 to 0.94) as the best point to predict bacterial pneumonia (S: 88.6%, E: 68.9%, PPV: 38.3 %, NPV: 96.5% RVP: 2.85; RPN: 0.17). CONCLUSION: The PRs showed an acceptable performance, but less sensitive than the original score to predict bacterial pneumonia. Although this tool may be easily applied, it should be validated in future studies.
