Accuracy of interpretations of emergency department radiographs: Effect of confidence levels

We conducted a prospective study to assess the relationship between the interpretive agreement rate for emergency department radiographs and the degree of interpretive confidence. We hoped to identify a subset of ED radiographs that did not require mandatory review by a radiologist. For each of the 1,872 plain radiographs studied, emergency physicians assigned a confidence level to the interpretation before comparing it with the radiologist's dictated report. A second radiologist was consulted to resolve disagreements. The overall rate of interpretive agreement was 94.6%. Agreement varied significantly ( P < .001) as a function of confidence level and by type of radiograph, but not by training level. Agreement varied significantly ( P < .001) as a function of confidence level for some types of radiographs (eg, chest, extremities) and for some types of radiographic findings (acute positive). No subset of radiographs had 100% agreement. Treatment was potentially altered in 38 patients as a result of the interpretive disagreement that occurred in 2% of studied radiographs. We conclude that the interpretive agreement rate increases in relation to interpretive confidence but that confidence levels cannot safely exclude certain radiographs from mandatory review by a radiologist.