Reliability of CXR for the diagnosis of bronchopulmonary dysplasia

Abstract

Background. Bronchopulmonary dysplasia (BPD) continues to be prevalent, despite new treatment, in part because of increased survival in less mature infants. Investigations of new treatments have been hampered by a lack of universally accepted diagnostic criteria. Radiographic scoring systems have been developed to provide objective assessment of lung injury and risk for chronic lung disease. Objective. We sought to test the reliability of a recently reported system using chest radiography as the main tool for diagnosis of BPD. Materials and methods. One hundred chest radiographs, half demonstrating BPD and the other half without BPD, were analyzed by pediatric radiologists and by a neonatologist, using the Weinstein score (1–6, depending on increasing radiographic severity). The reliability of this scoring system was tested by kappa (k) statistics. Results. Reliability at the lowest threshold (dividing score 1 from score ≥ 2) was unacceptably low in this population. Reliability increased with inclusion of higher BPD scores in the comparison groups: 1–3 versus 4–6. Conclusion. Using the chest radiograph for the prediction of BPD is not reliable between different observers except at the two extremes of the disease.