6-week radiographs unsuitable for diagnosis of suspected scaphoid fractures

INTRODUCTION: Six week follow-up radiographs are a common reference standard for the diagnosis of suspected scaphoid fractures. The main purpose of this study was to evaluate the interobserver reliability and diagnostic performance characteristics of 6-weeks radiographs for the detection of scaphoid fractures. In addition, two online techniques for evaluating radiographs were compared. MATERIALS AND METHODS: A total of 81 orthopedic surgeons affiliated with the Science of Variation Group assessed initial and 6-week scaphoid-specific radiographs of a consecutive series of 34 patients with suspected scaphoid fractures. They were randomized in two groups for evaluation, one used a standard website showing JPEG files and one a more sophisticated image viewer (DICOM). The goal was to identify the presence or absence of a (consolidated) scaphoid fracture. Interobserver reliability was calculated using the multirater kappa measure. Diagnostic performance characteristics were calculated according to standard formulas with CT and MRI upon presentation in the emergency department as reference standards. RESULTS: The interobserver agreement of 6-week radiographs for the diagnosis of scaphoid fractures was slight for both JPEG and DICOM (k = 0.15 and k = 0.14, respectively). The sensitivity (range 42–79 %) and negative predictive value (range 79–94 %) were significantly higher using a DICOM viewer compared to JPEG images. There were no differences in specificity (range 53–59 %), accuracy (range 53–58 %), and positive predictive value (range 14–26 %) between the groups. CONCLUSIONS: Due to low agreement between observers for the recognition of scaphoid fractures and poor diagnostic performance, 6-week radiographs are not adequate for evaluating suspected scaphoid fractures. The online evaluation of radiographs using a DICOM viewer seem to improve diagnostic performance characteristics compared to static JPEG images and future reliability and diagnostic studies should account for variation due to the method of delivering medical images. LEVEL OF EVIDENCE: Diagnostic level II.