Inter-observer variability in chest radiograph reading for diagnosing acute lung injury in children

Abstract

Acute lung injury (ALI), including its most serious form called acute respiratory distress syndrome (ARDS), is a devastating disease that can occur at any age. ALI/ARDS accounts for only 5-8% of admissions to pediatric intensive care units (PICUs) but is fatal in 30-60% of cases. International multicenter prospective studies are needed to better understand pediatric ALI/ARDS. However, a reproducible definition of ALI/ARDS is crucial to ensure that study populations are homogeneous. We designed a retrospective review to test the inter-observer variability of chest radiograph interpretation for presence of the American-European Consensus Conference (AECC) radiographic criterion for ALI/ARDS. The medical files of 24 children ventilated for ALI/ARDS in our PICU between January 1993 and December 2002 were reviewed. Five pediatric radiologists and five pediatric intensivists interpreted one frontal chest radiograph (FCR) per patient taken on the day of ALI/ARDS diagnosis. Each reader indicated whether the radiograph showed the AECC radiographic criterion for ALI/ARDS. Data analysis involved comparing each reader to all the others based on the raw agreement and Kappa coefficient (kappa). Features in the 24 patients were consistent with earlier studies. Global inter-observer agreement beyond chance was fair (kappa = 0.29 +/- 0.02) among the five radiologists (kappa = 0.26 +/- 0.05) and among the five intensivists (kappa = 0.29 +/- 0.05). Thus, considerable inter-observer variability occurred in assessing the radiographic criterion for ALI/ARDS, as previously shown in adults. Given the low incidence of ALI/ARDS in children, this variability may have a large impact in studies of pediatric ALI/ARDS.