RESUMO
OBJECTIVES: To develop a mathematical model to predict the probability of having community-acquired pneumonia and to evaluate an already developed prediction rule that has not been validated in a clinical scenario. METHODS: Children who presented with fever and had presumptive clinical diagnosis of pneumonia were evaluated in 4 institutions of different complexity during 1 year. The variables assessed were sex, age, respiratory rate, days with fever, maximum body temperature, presence of tachypnea, cough, chest pain, intercostal retraction, nasal flaring, abdominal pain, vomiting, grunting, rales, decreased breath sounds, wheezing, fatigue, loss of appetite, loss of sleep, and season of the year. The chest radiographs were photographed and then interpreted by 2 pediatric radiologists. RESULTS: A total of 257 children were evaluated: 179 (69%) had clinical and radiological diagnosis of community-acquired pneumonia, and 78 (30%) had no radiological confirmation. A total of 96 photographs were recorded, and in 64 of the cases, there was agreement in the diagnosis between the evaluating pediatrician and the radiologists (kappa index = 0.68).With the calculated probabilities, it was possible to build a receiving operating characteristic curve and, based on the estimated coefficients we calculated, a value associated to the probability of having pneumonia. CONCLUSIONS: We developed a model including 5 variables of high level of sensitivity for the diagnosis of pneumonia. To use it, it would be useful to apply the appropriate software. In addition, we validated a clinical prediction rule of 4 variables that proved to have 93.8% sensitivity to diagnose pneumonia in children with a fever and localized rales, or decreased breath sounds, or tachypnea, or any combination of these 4 variables.