Use of radiography in acute knee injuries: need for clinical decision rules.

OBJECTIVES: To study: 1) the efficiency of the current use of radiography in acute knee injuries, 2) the judgments and attitudes of experienced clinicians in their use of knee radiography, and 3) the potential for decision rules to improve efficiency.
METHODS: This two-stage study of adults with acute knee injuries involved: 1) a retrospective review of all 1,967 patients seen over a 12-month period in the EDs of one community and two teaching hospital, and 2) a prospective survey of another 1,040 patients seen by attending emergency physicians. The prospective survey assessed each clinician's estimate of the probability of a knee or patella fracture; 120 patients were independently assessed by two physicians.
RESULTS: Of the 1,967 patients seen in the first stage, 74.1% underwent radiography but only 5.2% were found to have fractures. Of the 1,727 knee and patella radiographic series ordered, 92.4% were negative for fracture. In the second stage, experienced physicians predicted the probability of fracture to be 0 or 0.1 for 75.6% of the patients. The kappa value for this response was 0.51 (95% CI 0.34 to 0.68). The physicians also indicated that they would have been comfortable or very comfortable in not ordering radiography for 55.5% of the patients. The area under the receiver operating characteristics curve for the physicians' prediction of fracture was 0.87 (95% CI 0.82 to 0.91), reflecting good discrimination between fracture and nonfracture cases. Likelihood ratios for the physicians' prediction ranged from 0.09 at the 0 level to 42.9 at the 0.9-1.0 level.
CONCLUSIONS: Emergency physicians order radiography for most patients with acute knee injuries, even though they can accurately discriminate between fracture and nonfracture cases and expect most of the radiographs to be normal. These findings suggest great potential for more efficient use of knee radiography, possibly through the use of a clinical decision rule.