Background: Radiology is an essential part of neonatal intensive care. Interpretation of chest radiographs frequently contributes to respiratory management of neonates, but there has been little assessment of the consistency of this interpretation.
Objective: To assess the inter- and intra-observer variability for the reporting of atelectasis and/or consolidation in neonatal chest radiographs.
Materials and methods: A total of 585 chest radiographs from the 220 babies ventilated in our nursery over a 2-year period were coded by two radiologists for generalised, lobar and segmental atelectasis and/or consolidation. Two months later one of the radiologists re-coded a random sample of these films (n = 117, 20 %). Agreement was assessed by the kappa statistic and by proportions of agreement for normality and abnormality.
Results: The reported incidence of focal atelectasis was low (5-6 %). Focal changes of any nature were found in 21-26 % of films. Inter-observer agreement was fair to moderate (kappa = 0.25-0.44). Intra-observer agreement was mostly moderate to good (kappa = 0.38-0.66).
Conclusion: The poor inter-observer agreement for the diagnosis of pulmonary parenchymal abnormalities on chest radiographs of neonates receiving intensive care suggests that abnormalities should be described rather than diagnoses given or that a list of differential diagnoses be offered. When research involves radiographic interpretation, the potential lack of consistency in reporting abnormalities must be borne in mind.