Interobserver reliability of computed tomography to diagnose scaphoid waist fracture union

<i>Purpose : </i>To determine the interobserver agreement and diagnostic performance characteristics of computed tomography (CT) for determining union of scaphoid waist fractures.<br><br><i>Methods :</i> A total of 59 orthopedic and trauma surgeons rated for union a set of 30 sagittal CT scans of 30 scaphoid waist fractures. Of these fractures, 20 were treated nonoperatively, were imaged between 6 and 10 weeks after injury, and were known to have eventually achieved union. Ten were operatively confirmed to be ununited. We rated each scan as united or ununited using a Web-based rating application. We assessed interobserver reliability using Siegel's multirater Kappa. We calculated diagnostic performance characteristics using Bayesian formulas.<br><br><i>Results :</i> The interobserver agreement among 59 raters was substantial. The average sensitivity, specificity, and accuracy of diagnosing union of scaphoid waist fractures on sagittal CT scans were 78%, 96%, and 84%, respectively. Assuming a 90% prevalence of fracture union of the scaphoid, the positive predictive value of a diagnosis of union on sagittal CT scan was 0.99 and the negative predictive value was 0.41.<br><br><i>Conclusions : </i>Our results suggest that CT scans are accurate and reliable for diagnosis of union but inadequate for ruling out nonunion of scaphoid waist fractures between 6 and 10 weeks after injury.<br>