Variability in the interpretation of chest radiographs for the diagnosis of pneumonia in children

Abstract

Although chest radiography is commonly used to establish the diagnosis of pneumonia in children, the reliability of radiographic findings among radiologists is not well described. We sought to evaluate the inter-rater and intra-rater reliability of radiographic features commonly described by radiologists in childhood pneumonia. Prospective case-based study. One hundred and ten radiographs of children evaluated in a pediatric emergency department for suspicion of pneumonia were interpreted by six radiologists at two academic children's hospitals. Radiologists were blinded to the clinical history. Reliability of standardized radiographic features was evaluated using the kappa statistic. The radiographic finding of an alveolar infiltrate demonstrated substantial reliability among radiologists (κ = 0.69). The presence of 'any infiltrate' and pleural effusion demonstrated moderate reliability (κ = 0.47 and k=0.45, respectively). Other radiographic features were less reliable: air bronchograms (κ = 0.32), hilar adenopathy (κ = 0.21), and interstitial infiltrate (κ = 0.14). Similarly, the finding of alveolar infiltrate demonstrated substantial intra-rater reliability upon review of ten duplicate radiographs, whereas interstitial infiltrate was less reliable. The radiographic finding of an alveolar infiltrate is very reliable among pediatric radiologists, whereas the finding of an interstitial infiltrate is less reliable.