Impact of clinical history on radiographic detection of fractures: a comparison of radiologists and orthopedists.

Are the methods used by radiologists in interpreting imaging studies different from those of attending physicians? We evaluated the influence that knowledge of localizing clinical signs has on the accuracy of fracture detection by orthopedic surgeons and compared the results with those of an identical study of radiologists performed earlier. The orthopedists studied had the same degree of experience as the radiologists. Forty radiographs of the extremities were examined twice by seven orthopedic surgeons. In 26 cases, a subtle fracture was present; 14 cases were normal. During one interpretation of the radiographs of each case, the precise location of pain, tenderness, or swelling was provided; during the other, this information was withheld. Analysis of receiver-operating-characteristic parameters indicates that the clues regarding location of trauma facilitate detection of fractures by orthopedists (an 11% improvement in Az, the area under the ROC curve, F[1,12] = 49.67, p less than .001). This finding is similar to the results of the earlier study with radiologists (a 6% improvement in Az, F[1,12] = 14.77, p less than .005). Statistical comparison of the two experiments showed that orthopedists depend on this information much more than do radiologists, demonstrated by a statistically significant prompting-by-specialty interaction (F[1,12] = 5.13, p less than .05). Localization clues improve ability of orthopedic surgeons to detect fractures in the trauma patient even more than they improve the ability of radiologists. The accuracy of the radiologist will suffer less than that of the orthopedist when localization clues are unavailable. Nonetheless, the findings show that localizing clues are important to both orthopedists and radiologists when searching for fractures.