Objective: To determine whether the interpretations of digital radiographs by pediatric critical care physicians displayed on the bedside personal computer differ from the interpretations of images displayed on the diagnostic workstation.
Design: Paired comparison.
Setting: A 38-bed pediatric critical care unit in a 372-bed pediatric university hospital.
Subjects: Four pediatric critical care fellows and four pediatric critical care staff physicians.
Interventions: Eight critical care physicians interpreted 114 radiographs in random order on two separate occasions. Each radiograph was assessed for the presence or absence of five chest abnormalities, the correct or incorrect endotracheal tube position, and the position of central venous catheters. These interpretations were scored against a gold standard.
Measurements and main results: Sensitivity and specificity were calculated for the presence or absence of five chest abnormalities and the identification of correct or incorrect endotracheal tube position. Kappa was calculated to assess agreement in the interpretation of central catheter position. Regarding chest abnormalities, improvement in sensitivity on the diagnostic workstation was statistically significant for one critical care fellow. The specificity on the diagnostic workstation was significantly worse for two critical care fellows and two critical care staff physicians. Regarding endotracheal tube position, improvement in sensitivity on the diagnostic workstation was statistically significant for one critical care staff physician. There were no statistically significant differences between the two viewing modalities for specificity measures. For central venous catheter position, there were no statistically significant differences in the interobserver or intra-observer agreements between the two viewing modalities.
Conclusions: With the exception of diffuse chest abnormalities, pediatric critical care physicians can use the Web-based bedside personal computer for clinical decision-making with the confidence that the decisions will be similar to those made on the diagnostic workstation.