Prevalence and Diagnostic Performance of Isolated and Combined NEXUS Chest CT Decision Criteria

Transcription

1 ORIGINAL CONTRIBUTION Prevalence and Diagnostic Performance of Isolated and Combined NEXUS Chest CT Decision Criteria Ali S. Raja, MD, MBA, MPH, William R. Mower, MD, PhD, Daniel K. Nishijima, MD, Gregory W. Hendey, MD, Brigitte M. Baumann, MD, MSCE, Anthony J. Medak, MD, and Robert M. Rodriguez, MD Abstract Objectives: The use of chest computed tomography (CT) to evaluate emergency department patients with adult blunt trauma is rising. The NEXUS Chest CT decision instruments are highly sensitive identifiers of adult blunt trauma patients with thoracic injuries. However, many patients without injury exhibit one of more of the criteria so cannot be classified low risk. We sought to determine screening performance of both individual and combined NEXUS Chest CT criteria as predictors of thoracic injury to inform chest CT imaging decisions in non low-risk patients. Methods: This was a secondary analysis of data on patients in the derivation and validation cohorts of the prospective, observational NEXUS Chest CT study, performed September 2011 to May 2014 in 11 Level I trauma centers. Institutional review board approval was obtained at all study sites. Adult blunt trauma patients receiving chest CT were included. The primary outcome was injury and major clinical injury prevalence and screening performance in patients with combinations of one, two, or three of seven individual NEXUS Chest CT criteria. Results: Across the 11 study sites, rates of chest CT performance ranged from 15.5% to 77.2% (median = 43.6%). We found injuries in 1,493/5,169 patients (28.9%) who had chest CT; 269 patients (5.2%) had major clinical injury (e.g., pneumothorax requiring chest tube). With sensitivity of 73.7 (95% confidence interval [CI] = 68.1 to 78.6) and specificity of 83.9 (95% CI = 83.6 to 84.2) for major clinical injury, abnormal chest-x-ray (CXR) was the single most important screening criterion. When patients had only abnormal CXR, injury and major clinical injury prevalences were 60.7% (95% CI = 52.2% to 68.6%) and 12.9% (95% CI = 8.3% to 19.4%), respectively. Injury and major clinical injury prevalences when any other single criterion alone (other than abnormal CXR) was present were 16.8% (95% CI = 15.2% to 18.6%) and 1.1% (95% CI = 0.1% to 1.8%), respectively. Injury and major clinical injury prevalences among patients when two and three criteria (not abnormal CXR) were present were 25.5% (95% CI = 23.1% to 28.0%) and 3.2% (95% CI = 2.3% to 4.4%) and 34.9% (95% CI = 31.0% to 39.0%) and 2.7% (95% CI = 1.6% to 4.5%), respectively. Conclusions: We recommend that clinicians check for the six clinical NEXUS Chest CT criteria and review the CXR (if obtained). If patients have one clinical criterion (other than abnormal CXR), they will have a very low risk of clinically major injury. We recommend that clinicians discuss the potential risks and benefit of chest CT in these cases. The risks of injury and major clinical injury rise incrementally with more criteria, rendering the risk/benefit ratio toward performing CT in most cases. If the patient has an abnormal CXR, the risks of major clinical injury and minor injury are considerably higher than with the other criteria chest CT may be indicated in cases requiring greater anatomic detail and injury characterization. From the Department of Emergency Medicine, Massachusetts General Hospital and Harvard Medical School (ASR), Boston, MA; the Department of Emergency Medicine, University of California (WRM), Los Angeles, CA; the Department of Emergency Medicine, University of California (DKN), Davis, CA; the Department of Emergency Medicine, San Francisco Fresno Medical Education Program (GWH), San Francisco, CA; the Department of Emergency Medicine, Cooper Medical School of Rowan University (BMB), Camden, NJ; the Department of Emergency Medicine, University of California at San Diego School of Medicine (AJM), San Diego, CA; and the Department of Emergency Medicine, University of California (RMR), San Francisco, CA. Received December 30, 2015; revision received April 18, 2016; accepted April 19, Presented at the Society for Academic Emergency Medicine annual meeting, San Diego, CA, Funded by University of California Center for Health Quality and Innovation (CHQI). The authors have no potential conflicts to disclose. Supervising Editor: Erik P. Hess, MD, MSc. Address for correspondence and reprints: Ali S. Raja, MD; ASRaja@partners.org by the Society for Academic Emergency Medicine ISSN doi: /acem PII ISSN

2 864 Raja et al. PERFORMANCE OF NEXUS CHEST CT DECISION CRITERIA ACADEMIC EMERGENCY MEDICINE 2016;23: by the Society for Academic Emergency Medicine The use of chest computed tomography (CT) in adult blunt trauma evaluation is rising, 1 3 in part due to the increasing implementation of head-topelvis CT (pan-scan) imaging protocols. 4 Although chest CT undeniably detects more injuries than plain chest x-ray (CXR), it also exposes patients to greater radiation risks, 5 higher hospital charges, and the potential for diagnosis of nonemergent, incidental findings that require additional imaging. 6 In our presentation of two NEXUS Chest CT decision instruments (DIs), which allow clinicians to safely forego chest CT in 25% 37% of patients, we emphasized the critical point that these DIs are strictly rule-out criteria. 7,8 Most of the approximately 70% of patients who are not assessed as low risk by the NEXUS Chest CT DIs do not have clinically significant injuries. 7 As with other DIs of this type, 9,10 failing to meet a low-risk classification does not mandate the need for imaging, but opens the opportunity to employ clinical judgment. We have recently described the screening performance of NEXUS Chest criteria for chest imaging in general (combining CXR and CT); 11 however, clinicians often use abnormalities noted on CXR to determine the need for CT and no information yet exists to guide this decision. The focus of this analysis was to provide data to assist decision-making for CT imaging in this group of non low-risk patients, since this is the decision most often encountered. Specifically, we sought to determine: 1) the screening performance of individual NEXUS Chest CT DI criterion for predicting injury seen on CT and 2) the prevalence of injury in patients who exhibit one or more criteria. These data may help guide clinicians in their choices of imaging when patients do not rule-out for injury by NEXUS Chest CT DIs. Of course, these decisions should be made in conjunction with patients whenever possible; while we classify outcomes as major and minor in this article, many patients will likely place significant importance on even minor injuries and may desire imaging for these alone. METHODS Study Design This was an a priori analysis of data from the NEXUS Chest CT study, the details of which are described in detail elsewhere. 12,13 Institutional review board approval was obtained at all study sites. Study Setting and Population Briefly, we conducted the NEXUS Chest CT study at 11 United States urban Level I trauma centers from September 2011 to May 2014 with the following patient inclusion criteria: 1) age over 14 years, 2) emergency department (ED) presentation for blunt trauma occurring within 6 hours of ED arrival, and 3) having chest imaging as part of trauma evaluation. 7 Study Protocol Research personnel enrolled patients and obtained physician impressions after imaging was ordered, but prior to image interpretation. Physicians indicated the presence or absence of the following clinical criteria: 1) age greater than 60 years, 2) altered mental status or altered level of consciousness, 3) intoxication, 4) rapid deceleration mechanism, 5) chest pain, 6) distracting injury, 7) chest wall tenderness, 8) sternal tenderness, 9) thoracic spine tenderness, 10) scapular tenderness, and 11) focused assessment with sonography in trauma (FAST) criteria. Low rates of FAST use precluded the use of these criteria in our DIs and primary analyses. During dual-provider criteria assessments, we found acceptable (j > 0.6) inter-rater reliability of all other criteria. 7 We defined the presence of injury according to final attending radiologist chest CT reports (the reference standard). Injuries were then classified by associated s, observation, and hospital admission into expert panel-derived categories of major clinical injury (e.g., pneumothorax necessitating chest tube), minor injury (e.g., multiple rib fracture not requiring surgical ), and insignificant. We deemed the CXR to be abnormal if any injury (including clavicle fracture) or widened mediastinum was reported. To derive our NEXUS Chest CT DIs, we employed binary partitioning techniques. We subsequently validated the two DIs in a separate cohort. 7 Measurements Among 11,477 patients enrolled in the NEXUS Chest CT study, 5,169 patients had chest CT (4,501 with and 668 without CXR); these 5,169 patients formed the cohort for this study. Using their data, we determined the injury and major clinical injury screening performance of the seven criteria comprising the NEXUS Chest DIs: 1) rapid deceleration mechanism, 2) distracting injury, 3) chest wall tenderness, 4) sternal tenderness, 5) thoracic spine tenderness, 6) scapular tenderness, and 7) abnormal CXR (using only the 4,501 patients who had both chest CXR and CT). Considering all criteria as binary, we calculated the sensitivity, specificity, positive predictive value (PPV), negative predictive value, and positive likelihood ratio. For example, screening parameters for injury with criterion 1 would be: true positive = presence of criterion 1 and having injury; true negative = absence of criterion 1 and not having injury; false positive = presence of criterion 1 and not having injury; and false negative = absence of criterion 1 and having injury. In these same groups of 5,169 patients for the six clinical criteria and 4,501 patients for abnormal CXR, we then determined how frequently individual NEXUS Chest CT criteria occurred as a single isolated criterion, as well as the associated prevalence of injury and PPV. Data Analysis Because our original study was powered to derive and validate the NEXUS Chest CT DIs, we did not conduct

3 ACADEMIC EMERGENCY MEDICINE August 2016, Vol. 23, No these secondary analyses using formal hypothesis testing, nor did we set a priori threshold limits for injury or major clinical injury prevalence for the individual criterion. In contrast to directive DI data, we intended this analysis to provide general guidance to assist clinicians in decision-making around imaging, within the context of their trauma practice. We used Research Electronic Data Capture (RedCAP) hosted by the University of California at San Francisco to manage data and Stata v.13.2 (StataCorp) to analyze data and calculate descriptive characteristics. RESULTS Study Subjects Characteristics of the 5,169 patients who underwent chest CT are shown in Table 1. Across the 11 study sites, rates of chest CT performance ranged from 15.5% to 77.2% (median = 43.6%). Of the 5,169 patients, 15 (0.3%) did not have all seven criteria collected. However, as we were analyzing each criterion separately, these patients were still included in the analysis. Injury Prevalence We found injuries in 1,493 of the 5,169 patients (28.9%) who had chest CT imaging, including 269 (5.2%) with major clinical injury (Table 2). Of the 4,501 patients undergoing both CT and CXR, 1,355 (30.1%) sustained injuries with 264 patients (5.9%) having major clinical injury. Among 688 patients with only chest CT, 138 (20.7%) had injury and five patients (0.7%) had major clinical injury (Figure 1). Screening Performance of NEXUS Chest CT Criteria Screening performance for individual criteria are provided in Table 3. Abnormal CXR had the best screening Table 1 Patient Characteristics (N = 5,169) Characteristic Age (y) 45 (29 61) Injury Severity Score (ISS)* 6 (1 14) Glasgow Coma Scale 15 (14 15) Sex Male 3,274 (63.8) Female 1,857 (36.2) Mechanism of injury MVC 2,199 (42.5) Fall 1,067 (20.6) MCA 645 (12.5) PMV 573 (11.1) Bike accident 313 (6.1) Assault 189 (3.7) Other and unknown 251 (4.8) Admitted to hospital 3,087 (59.8) Survived to hospital discharge 2,895 (93.8) Hospital length of stay (d) 4 (2 8) Data are reported as median (IQR) or number (%). IQR = interquartile ratio; MVC = motorized vehicle collision; MCA = motorcycle accident; PMV = pedestrian struck by motorized vehicle. *ISS was performed in 3805/5169 (73.6%) patients other patients had no (or very minor) injuries. Table 2 Patient Injuries in 5,169 Patients Major injuries (371 injuries in 264 patients) Pneumothorax with chest tube 190 Hemothorax with chest tube 72 Spinal fracture: received surgical 39 stabilization Pulmonary contusion: received 27 mechanical ventilation Aortic or great vessel injury: no surgery 11 but observed > 24 h Aortic or great vessel injury: underwent 10 surgery Ruptured diaphragm 4 Scapular fracture: received surgical 4 Sternal fracture: received surgical 3 Mediastinal or pericardial hematoma: 3 received drainage procedure Bronchial injury: received surgical 1 Esophageal injury received surgical 1 Other major thoracic injury 6 Minor injuries (2,486 injuries in 1,493 patients) Multiple rib fracture: no surgical 950 Pulmonary contusion: no mechanical 441 ventilation but observed > 24 h Pneumothorax: no chest tube but 260 observed > 24 h Sternal fracture: no surgical 259 Spinal fracture: no surgical 222 Scapular fracture: no surgical 140 Hemothorax: no chest tube but 78 observed > 24 h Mediastinal or pericardial hematoma: 71 no surgical but observed > 24 h Pneumomediastinum without 54 pneumothorax: no surgical but observed > 24 h Esophageal injury: no surgical 1 but observed > 24 h Other minor thoracic injury 10 performance characteristics for injury and major clinical injury of the seven NEXUS Chest CT criteria. Of the clinical characteristics, distracting injury and rapid deceleration mechanism were the most common criteria found in isolation, occurring in 11.5 and 9.0% of patients, respectively. Isolated sternal, scapular, and chest wall tenderness were very uncommon, all occurring in < 1% of patients. In Table 4 we present the prevalence and predictive value of injury and major clinical injury when criteria appear in isolation or in combinations (only those that occurred in >3% of patients are presented). Patients who exhibited an abnormal CXR but none of the other criteria had an injury prevalence of 60.7% (95% confidence interval [CI] = 52.2% to 68.6%) and a major clinical injury prevalence of 12.9% (95% CI = 8.3% to 19.4%). Patients with any single clinical criterion other than abnormal CXR had much lower injury and major clinical injury prevalences of 16.8% (95% CI = 15.2% to

4 866 Raja et al. PERFORMANCE OF NEXUS CHEST CT DECISION CRITERIA 11,477 patients enrolled in derivation and validation cohorts 6,308 had chest x- ray (CXR) only Not included in this analysis 5,169 had chest CT 269 (5.2%) had major injury 1,493 (28.9%) had major or minor injury 4,501 had CXR and chest CT 264 (5.9%) had major injury 1,355 (30.1%) had major or minor injury 668 had chest CT only 5 (0.7%) had major injury 138 (20.7%) had major or minor injury Figure 1. Study enrollment. Table 3 Individual Criteria Screening Performance for Injury Sensitivity Specificity PPV NPV LR(+) Abnormal CXR MCI 73.7 ( ) 83.9 ( ) 22.1 ( ) 98.1 ( ) 4.58 ( ) Major or minor 49.6 ( ) 94.3 ( ) 79.9 ( ) 80.4 ( ) 8.72 ( ) Distracting injury MCI 67.4 ( ) 55.3 ( ) 7.6 ( ) 96.9 ( ) 1.51 ( ) Major or minor 55.0 ( ) 57.8 ( ) 34.6 ( ) 76.0 ( ) 1.30 ( ) CW Tenderness MCI 50.2 ( ) 74.3 ( ) 9.6 ( ) 96.5 ( ) 1.96 ( ) Major or minor 43.7 ( ) 79.9 ( ) 46.9 ( ) 77.8 ( ) 2.18 ( ) Rapid deceleration MCI 40.1 ( ) 61.8 ( ) 5.4 ( ) 95.0 ( ) 1.05 ( ) Major or minor 41.2 ( ) 62.9 ( ) 31.1 ( ) 72.5 ( ) 1.11 ( ) T-spine tenderness MCI 21.7 ( ) 79.8 ( ) 5.5 ( ) 94.9 ( ) 1.08 ( ) Major or minor 24.2 ( ) 81.3 ( ) 34.4 ( ) 72.5 ( ) 1.29 ( ) Sternal tenderness MCI 20.2 ( ) 85.5 ( ) 7.1 ( ) 95.2 ( ) 1.39 ( ) Major or minor 24.0 ( ) 89.0 ( ) 46.9 ( ) 74.3 ( ) 2.18 ( ) Scapular tenderness MCI 10.9 ( ) 94.9 ( ) 10.4 ( ) 95.1 ( ) 2.14 ( ) Major or minor 8.8 ( ) 96.0 ( ) 47.5 ( ) 72.2 ( ) 2.26 ( ) Data are reported as % (95% CI). CW = chest wall; CXR = chest x-ray; LR(+) = positive likelihood ratio; MCI = major clinical injury; NPV = negative predictive value; PPV = positive predictive value; T = thoracic. 18.6%) and 1.1% (95% CI = 0.1% to 1.8%), respectively. Injury and major clinical injury prevalence increased with the presence of more criteria (see Table 3 for all injury as well as only major injury prevalence, stratified according to NEXUS CT DI criteria). In the analysis of combinations of criteria that may be associated with missed injury on CXR, distracting injury and rapid deceleration were, again, the most common solo criteria. We found very low rates of major clinical injury with one criterion and two criteria (Table 4).

5 ACADEMIC EMERGENCY MEDICINE August 2016, Vol. 23, No Table 4 Predictive Performance of Individual and Combinations of Criteria Prevalence of Major Clinical Injury [95% CI] Positive for 0 of 7 criteria* 2 of 813 (0.2%) [0.1% 0.7%] Positive for any 1 or more of 7 criteria* 162 of 3,688 (4.4%) [3.8% 5.1%] Only distracting injury 8 of 592 (1.4%) [0.6% 2.6%] Only rapid deceleration 1 of 464 (0.2%) [0.0% 1.2%] Only rapid deceleration + distracting injury 15 of 351 (4.3%) [ %] Normal CXR + 1 other criterion* 17 of 1,523 (1.1%) [0.1% 1.8%] Normal CXR + 2 other criteria* 33 of 1,040 (3.2%) [2.3% 4.4%] Normal CXR + 3 other criteria* 14 of 513 (2.7%) [1.6% 4.5%] Abnormal CXR alone* 18 of 140 (12.9%) [8.3% 19.4%] Abnormal CXR + 1 other criterion 47 of 254 (18.5%) [14.2% 23.7%] Abnormal CXR + 2 other criteria 62 of 251 (24.7%) [19.8% 30.4%] Prevalence of Major or Minor Injury [95% CI] 56 of 813 (6.9%) [5.3% 8.8] 1,299 of 3,688 (35.2%) [34.8% 35.6%] 89 of 592 (15.0%) [12.4% 18.1%] 63 of 464 (13.6%) [10.7% 17.0] 81 of 351 (23.1%) [ %] 256 of 1,523 (16.8%) [15.2% 18.6%] 265 of 1,040 (25.5%) [23.1% 28.0%] 179 of 513 (34.9%) [31.0% 39.0%] 85 of 140 (60.7%) [52.2% 68.6%] 166 of 254 (65.4%) [59.35% 71.0%] 205 of 251 (81.7%) [76.45% 86.1%] LR(+) for Major or Minor Injury [95% CI] 0.17 [ ] 1.26 [ ] 0.44 [ ] 0.39 [ ] 0.74 [ ] 0.50 [ ] 0.84 [ ] 1.32 [ ] 3.59 [ ] 4.38 [ ] [ ] CXR = chest x-ray; LR(+) = positive likelihood ratio. *7 criteria = above 6 criteria + abnormal CXR. Calculations for these criteria were based on all patients who had both CXR and CT (n = 4,501). 6 clinical criteria = sternal tenderness, scapular tenderness, chest wall tenderness, thoracic spine tenderness, distracting injury, rapid deceleration. Calculations for these criteria were based on all patients who had CT (n = 5,169). DISCUSSION Like several other clinical decision rules, the NEXUS Chest CT DIs are solely intended to allow practitioners to rule out injury and forego imaging. Most patients, however, do not rule out for injury under strict rule application. To provide useful assistive information for decision-making regarding imaging in these non lowrisk cases, we have presented screening performance and injury prevalence data for adult blunt trauma patients who have one or more NEXUS Chest CT DI criteria. Overall, most patients had one or more criteria. However, when examining this group as a whole, we found that over one-third had injury and 4.4% had major clinical injury. We found that, expectedly, CXR abnormality was the single most important screening criterion, with the best injury predictive value of the seven NEXUS criteria. We also determined that criteria infrequently occurred in isolation (<20% of cases) and that all isolated clinical criteria (those other than abnormal CXR) were associated with relatively low injury prevalence (PPV < 20%). Although we noted incremental rises in injury prevalence when going from zero to three criteria, by far the largest jump in prevalence occurs with the abnormal CXR criterion as demonstrated in Table 4. To implement these assistive data, we recommend that clinicians check for the six clinical criteria and review the CXR (if obtained). If patients have no criteria, they have very low risk of injury (especially major injury) per NEXUS Chest CT and would likely not benefit diagnostically from chest CT. If they have one clinical criterion (other than abnormal CXR), they will still have a low likelihood of injury and major clinical injury. We recommend that clinicians discuss the potential risks and benefit of chest CT with both patients and trauma surgeons in these cases. The risks of injury and major clinical injury rise incrementally with more criteria, rendering the risk/benefit ratio toward performing CT in most cases. Very importantly, if the patient has an abnormal CXR, the risks of injury and major clinical injury are considerably higher than with the other criteria in these cases, chest CT is indicated to allow for greater anatomic detail and injury characterization. Because PPV will vary with prevalence of injury in populations, clinicians should consider our overall major clinical injury rate of 5.2% and all injury rate of 28.9% and shift these PPV point estimates up or down according to their site injury prevalence. Moreover, with regard to acceptable injury miss rates, practitioners should apply our findings of this analysis in the context of their individual thresholds and practice standards. Others have sought to develop DIs to guide the ordering of chest CT in blunt trauma. In a single-site study, Brink et al. 14 derived a rule of nine factors for the prediction of injury on thoracic CT. However, some of the factors in their predictive rule (especially thoracic spine x-rays) are not always performed and take extra time, and so this rule is of limited utility for immediate triage and screening for the need for chest CT.

6 868 Raja et al. PERFORMANCE OF NEXUS CHEST CT DECISION CRITERIA We must emphasize that although we enrolled all patients with blunt trauma, our DIs and assistive information for selective CT are primarily meant for use in the hemodynamically stable blunt trauma patient, in whom there is time to obtain and interpret a CXR. Several investigators have reported that major, polytrauma patients benefit from head-to-pelvis CT we do not recommend use of our DIs in these critically ill patients. 15,16 It is also important to note that the injuries seen on CT may have been detected with further clinical observation, by repeat CXR, or using other diagnostic modalities, especially extended FAST, which has excellent sensitivity for pneumothorax. 17 In this study we found that most patients with injury on CT also had an injury or other abnormality identified on CXR. The abnormal CXR, however, may not specifically identify all injuries. In a previous analysis of occult injuries (injuries seen on CT but not on CXR), we reported low sensitivity of CXR for identifying thoracic injuries, such as sternal fracture and pulmonary contusion. 18 The higher sensitivity of CXR that we report now reflects our use of a definition of a completely normal CXR as a binary criterion in this analysis any injury seen on CXR would make that CXR abnormal and a true positive in terms of identifying injury. In the former study, abnormal CXRs did not identify all specific thoracic injuries and therefore showed lower sensitivity. For example, a CXR may identify rib fractures but fail to detect pulmonary contusion and sternal fracture. In the former study, that CXR would miss sternal fracture and pulmonary contusion, and in this analysis it would be considered to detect injury. Overall, the CXR may be adequate for cases in which it identifies an injury such as a pneumothorax, but clinicians should consider chest CT in cases that require more detailed characterization of injuries beyond the scope of the CXR. We encountered wide site variability in the ordering of chest CT for blunt trauma, a finding that decision rule authorities have stated argues for a need for decision rule and assistive criteria. 19 This variability likely arises because clinicians, departments, and institutions have different thresholds for missed injury those desiring extreme diagnostic sensitivity may routinely order panscan in nearly all major trauma patients, while others wishing to diagnose only clinically relevant injuries may be more selective in their use of advanced imaging. 20 Presenting this analysis, we do not intend to take sides in this ongoing debate. Rather, we aim to allow clinicians to generate more informed, evidence-based shared decisions for selective imaging by clinicians and patients, through discussions of risk of injury given patient-specific factors. Along with the presentation of radiation risk and cost information, provision of likelihood of injury data such as ours should also advance the movement toward shared patient physician decision-making with regard to CT. We have demonstrated that most patients wish to be informed of the risks and costs of CT and that they would often choose to forego chest CT when the chances of finding life-threatening injury are low. 21 In this regard, trauma providers may present alert, lowerrisk patients with estimates of the likelihood that chest CT will show an injury and then discuss the risk/benefit ratios of chest CT directly with the patients involved. Of course, patients risk thresholds may vary while some may desire imaging only if it is likely to pick up major clinical injury, others may want imaging even if the only injuries likely to be picked up are minor. Finally, we do not intend our decision assistive data to replace clinical judgment. Clinicians may have other reasons to perform chest CT and may place considerable value on negative studies, which may allow them to shorten observation times or even facilitate discharge. We recommend that trauma providers use our criteria to determine likelihood of injury detection percentages and, with their patients, balance these against the costs and radiation risks of CT. LIMITATIONS All of the limitations we described in our DI derivation and validation article, most notably elements of spectrum and workup bias, also apply to this work. 7 Clinicians may disagree with our clinical categorization of injuries. The low numbers of patients who had only one criterion led to relatively wide CIs surrounding the corresponding point estimates of likelihood of injury. For clinical practicality reasons (ease of use with the NEXUS Chest DIs), we chose to only evaluate our NEXUS Chest CT DI criteria in this analysis. It is possible that other criteria may have had better injury prediction performance. It is likely that the screening and predictive performance of CXR was inflated because of temporally proximate readings of CXR and CT when radiologists see injuries on initial CXRs, they may be more likely to look for those (and other) injuries on subsequent imaging. Finally, we conducted our work at urban, academic Level I trauma centers. Patient characteristics, injury prevalence, and imaging protocols may differ at dissimilar hospitals and these differences may affect criteria screening performance. CONCLUSIONS In this multicenter analysis of NEXUS Chest CT criteria for adult blunt trauma victims, we found that an abnormal chest x-ray was the best screening parameter for thoracic injury. Patients with an abnormal chest x-ray alone have a moderate risk of both injury and major clinical injury. The presence of any of the other NEXUS Chest CT criteria, in isolation, is associated with a low rate of major clinical injury. Rates of injury and major clinical injury increase with the presence of more criteria. References 1. Korley FK, Pham JC, Kirsch TD. Use of advanced radiology during visits to US emergency departments for injury-related conditions, JAMA J Am Med Assoc 2010;304: Larson DB, Johnson LW, Schnell BM, Salisbury SR, Forman HP. National trends in CT use in the emergency department: Radiology 2011;258:

7 ACADEMIC EMERGENCY MEDICINE August 2016, Vol. 23, No Broder J, Warshauer DM. Increasing utilization of computed tomography in the adult emergency department, Emerg Radiol 2006 Oct;13: Salim A, Sangthong B, Martin M, Brown C, Plurad D, Demetriades D. Whole body imaging in blunt multisystem trauma patients without obvious signs of injury: results of a prospective study. Arch Surg 2006;141:468 73; discussion Committee to Assess Health Risks from Exposure to Low Levels of Ionizing Radiation, National Research Council. Health Risks from Exposure to Low Levels of Ionizing Radiation: BEIR VII Phase 2. Washington, DC: The National Academies Press, Prabhakar AM, Le TQ, Abujudeh HH, Raja AS. Incidental findings and recommendations are common on ED computed tomography angiography to evaluate for aortic dissection. Am J Emerg Med 2015;33: Rodriguez RM, Anglin D, Langdorf MI, et al. NEXUS chest: validation of a decision instrument for selective chest imaging in blunt trauma. JAMA Surg 2013;148: Rodriguez RM, Langdorf MI, Nishijima D, et al. Derivation and validation of two decision instruments for selective chest CT in blunt trauma: a multicenter prospective observational study (NEXUS Chest CT). PLoS Med 2015;12:e Hoffman JR, Mower WR, Wolfson AB, Todd KH, Zucker MI. Validity of a set of clinical criteria to rule out injury to the cervical spine in patients with blunt trauma. National Emergency X-Radiography Utilization Study Group. N Engl J Med 2000;343: Kuppermann N, Holmes JF, Dayan PS, et al. Identification of children at very low risk of clinicallyimportant brain injuries after head trauma: a prospective cohort study. Lancet 2009;374: Raja AS, Lanning J, Gower A, et al. Prevalence of chest injury with the presence of NEXUS chest criteria: data to inform shared decision making about imaging use. Ann Emerg Med 2015 [Epub ahead of print]. 12. Rodriguez RM, Hendey GW, Marek G, Dery RA, Bjoring A. A pilot study to derive clinical variables for selective chest radiography in blunt trauma patients. Ann Emerg Med 2006;47: Rodriguez RM, Hendey GW, Mower W, et al. Derivation of a decision instrument for selective chest radiography in blunt trauma. J Trauma 2011;71: Brink M, Deunk J, Dekker HM, et al. Criteria for the selective use of chest computed tomography in blunt trauma patients. Eur Radiol 2010;20: Huber-Wagner S, Lefering R, Qvick LM, et al. Effect of whole-body CT during trauma resuscitation on survival: a retrospective, multicentre study. Lancet 2009;373: Jiang L, Ma Y, Jiang S, et al. Comparison of wholebody computed tomography vs selective radiological imaging on outcomes in major trauma patients: a meta-analysis. Scand J Trauma Resusc Emerg Med 2014;22: Raja AS, Jacobus CH. How accurate is ultrasonography for excluding pneumothorax? Ann Emerg Med 2013;61: Langdorf MI, Medak AJ, Hendey GW, et al. Prevalence and clinical import of thoracic injury identified by chest computed tomography but not chest radiography in blunt trauma: multicenter prospective cohort study. Ann Emerg Med 2015;66: Reilly BM, Evans AT. Translating clinical research into clinical practice: impact of using prediction rules to make decisions. Ann Intern Med 2006;144: Tillou A, Gupta M, Baraff LJ, et al. Is the use of pan-computed tomography for blunt trauma justified? A prospective evaluation. J Trauma 2009;67: Rodriguez RM, Henderson TM, Ritchie AM, et al. Patient preferences and acceptable risk for computed tomography in trauma. Injury 2014;45: