Radiological errors in the Canadian Journal of Emergency Medicine.

OBJECTIVES To systematically evaluate the accuracy of text descriptions and labeling of radiologic images published in the Canadian Journal of Emergency Medicine (CJEM). Error detection by radiologists and emergency physicians and the clinical significance and educational value of these errors were assessed. Errors were also correlated with radiologist involvement in publication and imaging modality. METHODS Thirty-three issues of CJEM were examined from January 2003 to May 2008. Electronic copies of all radiologic images published were obtained with their caption and description from the text. Identifying information was removed to present images in an anonymous fashion. Images were presented to two radiologists who, working in consensus, critically appraised each image and accompanying text. Images were then presented to two emergency department physicians who, working in consensus, critically appraised each image and accompanying text. All images with errors detected by either radiology or emergency physicians were then discussed to determine if errors would have affected clinical management or educational value. The emergency physicians also identified "underlabeled" images where it was felt that further labeling would enhance their educational value. RESULTS Forty-five articles with 82 images were obtained. At least one error was observed in 18 (40%) articles and 20 (24%) images. Two errors were present in three images, resulting in 23 errors. Of the 23 errors, 17 were image description errors and 6 were labeling errors. Five errors were detected by both radiology and emergency physicians, whereas 15 were detected only by radiologists and 3 were detected only by emergency physicians. Of these errors, 12 (52%) were rated as potentially affecting both clinical management and educational value, 5 (22%) as only affecting educational value, and 6 (26%) as nonsignificant. Radiologists were involved in six articles, including 12 images that contained no errors. There was no official radiologist involvement in 39 articles, including 70 images, 18 (26%) of which contained errors. In addition, 26 images were identified by emergency physicians as potentially benefiting from enhanced labeling to improve educational value. CONCLUSIONS Radiologic images published in the CJEM are generally of high quality; however, 23 errors were found in 82 images, 18 (78%) of which were rated as potentially affecting clinical management, educational value, or both. Radiologist involvement in the publication process may be of assistance as no errors were seen in articles that included radiologists as authors.