Comparison of clinical performance of cranial computed tomography rules in patients with minor head injury: a multicenter prospective study.

OBJECTIVES: The objective was to compare the predictive performance of three previously derived cranial computed tomography (CT) rules, the Canadian CT Head Rule (CCHR), the New Orleans Criteria (NOC), and National Emergency X-Ray Utilization Study (NEXUS)-II, for detecting clinically important traumatic brain injury (TBI) and the need for neurosurgical intervention in patients with blunt head trauma.
METHODS: This was a prospective, multicenter, observational cohort study of patients with blunt head trauma from June 2008 to May 2009. The historical and physical examination components of the CCHR, NOC, and NEXUS-II were documented on a data collection form and the performance of each of the three rules was compared. Patient eligibility for each specific rule was defined exactly as previously described for each specific rule. To compare the three decision rules in terms of sensitivity and specificity, an intersection cohort satisfying inclusion criteria of all three decision rules was derived. The primary outcome was clinically important TBI, and the secondary outcome was neurosurgical intervention. The sensitivity and specificity of each rule were calculated with 95% confidence intervals (95% CIs). We also calculated the potential reduction rate in cranial CT scan utilization realized by theoretical implementation of these rules.
RESULTS: A total of 7,131 patients were prospectively enrolled, including 692 (9.7%) with clinical TBI. Among the enrolled population, patients eligible for CCHR, NOC, and NEXUS-II totaled 696, 677, and 2,951, respectively. The sensitivity and specificity for clinically important brain injury were as follows: CCHR, 112 of 144 (79.2%, 95% CI = 70.8% to 86.0%) and 228 of 552 (41.3%, 95% CI = 37.3% to 45.5%); NOC, 91 of 99 (91.9%, 95% CI = 84.7% to 96.5%) and 125 of 558 (22.4%, 95% CI = 19.0% to 26.1%); and NEXUS-II, 511 of 576 (88.7%, 95% CI = 85.8% to 91.2%) and 1,104 of 2,375 (46.5%, 95% CI = 44.5% to 48.5%). The sensitivity and specificity for neurosurgical intervention were as follows: CCHR, 100% (95% CI = 59.0% to 100.0%) and 38.3% (95% CI = 34.5% to 41.9%); NOC, 100% (95% CI = 54.1% to 100.0%) and 20.4% (95% CI = 17.4% to 23.7%); and NEXUS-II, 95.1% (95% CI = 90.1% to 98.0%) and 41.4% (95% CI = 39.5% to 43.2%). Among the enrolled population, intersection patients of CCHR, NOC, and NEXUS-II totaled 588. The sensitivity and specificity for clinically important brain injury were as follows: CCHR, 73 of 98 (74.5%, 95% CI = 64.7% to 82.8%) and 201 of 490 (41.0%, 95% CI = 36.6% to 45.5%); NOC, 89 of 98 (90.8%, 95% CI = 83.3% to 95.7%) and 112 of 490 (22.9%, 95% CI = 19.2% to 26.8%); and NEXUS-II, 82 of 98 (83.7%, 95% CI = 74.8% to 90.4%) and 172 of 490 (35.1%, 95% CI = 30.9% to 39.5%). The potential reduction in emergency CT scans by using these decision rules would have been higher with the NEXUS-II rule (39.6%, 95% CI = 37.8% to 41.4%) than with the CCHR rule (27.0%, 95% CI = 23.7% to 30.3%) or NOC rule (20.2%, 95% CI = 17.2% to 23.3%).
CONCLUSIONS: For clinically important TBI, the three cranial CT decision rules had much lower sensitivities in this population than the original published studies, while the specificities were comparable to those studies. The sensitivities for neurosurgical intervention, however, were comparable to the original studies. The NEXUS-II rule showed the highest reduction rate for CT scans compared to other rules, but failed to identify all undergoing neurosurgical intervention for their original inclusion cohort.