Objectives: To develop and validate a prediction model for the early identification of children with pneumonia in emergency care settings.
Methods: This was a diagnostic study with derivation and validation of multivariate logistic regression models. This study was carried out on children aged 1 month-16 years presenting with fever and cough visiting the pediatric emergency department of the Erasmus MC-Sophia, Rotterdam, the Netherlands (derivation population, n=504); Pediatric Assessment Unit, University Hospitals Coventry and Warwickshire NHS Trust, UK (validation population 1, n=420); Accident & Emergency Department, Queen's Medical Centre, Nottingham, UK (validation population 2, n=366). The outcome pneumonia is defined by the presence of pulmonary consolidations on chest radiograph or follow-up.
Results: In population 1, 78 (16%) children were diagnosed with pneumonia; in populations 2 and 3, this number was 58 (14%) and 27 (7%), respectively. A simple clinical decision rule on the basis of ill appearance, tachypnea, decreased oxygen saturation, and elevated serum C-reactive protein categorized children as being at a low (<5%) or a high (>16%) risk of pneumonia. The rule yielded a discriminative value of 0.79 (0.69-0.89) in the Rotterdam population and was validated well in the other two populations.
Conclusion: The risk of pneumonia can be assessed using three key clinical characteristics - overall assessment of the severity of illness, breathing rate, and oxygen saturation. Serum C-reactive protein contributes to the prediction of an intermediate risk of pneumonia in children. Children with a low risk of developing pneumonia with vital signs in the normal range can be discharged with effective safety netting, requiring neither antibiotics nor radiographs.