Comparison of the Canadian C-Spine rule and NEXUS decision instrument in evaluating blunt trauma patients for cervical spine injury.

Concerned about missing potentially catastrophic neurologic injury, emergency physicians have typically made liberal use of radiographic imaging to evaluate blunt trauma patients for cervical spine injuries. This practice subjects large numbers of patients to imaging, with its associated cost, time expenditure, and radiation exposure, in order to detect injury in a small minority. Consequently, decision instruments that allow clinicians to safely reduce cervical spine imaging have the potential to be of significant value. One existing instrument, the National Emergency X-Radiography Utilization Study (NEXUS) low-risk criteria, has been shown in prospective application to more than 34,000 patients to have a sensitivity of 99.6% for detecting clinically important cervical spine injury.1 However, in this issue of Annals, Canadian researchers, seeking to develop their own decision instrument, report the NEXUS instrument to have a sensitivity of less than 93% when retrospectively applied to their patient population.2 These results are inconsistent with the voluminous data collected during the development and validation of the NEXUS instrument and are in conflict with the large body of literature that investigated similar criteria before the performance of the NEXUS trial. Furthermore, the reported 7% miss rate with the NEXUS criteria is inconsistent with clinical experience and existing medical literature, which, except for the rare cases presented in the original NEXUS report, is virtually devoid of reports of missed injury. The discrepancy between these 2 studies reflects, in part, a natural asymmetry between the processes of “validating” and “invalidating” a decision instrument. Validation studies are quite vulnerable to misclassification errors, and when such errors occur, an instrument can easily appear to have been “invalidated.” For example, failure to detect important clinical findings because of inadequate evaluations or the use of surrogate variables can cause high-risk patients to be misclassified as low risk. If injured high-risk patients are misclassified in this manner, the reported sensitivity and negative predictive value will decrease. In contrast, if uninjured patients who exhibit high-risk criteria are misclassified, the reported specificity may be falsely increased. Misclassification can also decrease the instrument’s interrater reliability, as reflected by measures such as the statistic. Compared with the performance of the NEXUS decision instrument documented in the NEXUS report, the Canadian study reports a large decrease in sensitivity, a small but real decrease in negative predictive value, and an increase in the measured specificity.1,2 This pattern suggests the presence of misclassification errors in the Canadian study, and likely results from the study’s retrospective methodology and use of surrogate variables. Consequently, the Canadian article tells us little about the true performance of the NEXUS instrument, but does serve as an important warning regarding the use of decision instruments in general. Clinicians who wish to use a given instrument must understand the definitions used by the instrument, and they must perform careful assessments in determining the classification of individual patients. Failure to use a decision instrument properly can produce inadequate and misleading assessments, can produce misclassification of risk status, and can have potentially devastating consequences. Bearing these concerns in mind, clinicians should retain confidence in the reliability of the NEXUS cervical spine criteria. This instrument has already undergone validation in a large prospective study involving a wide range of institutions, clinical settings, and cliniT R A U M A / E D I T O R I A L