To cite thepresent paper, use exclusively the following reference. AgostiniC, Durieux M, Milot L, Kamaoui I, Floccard B, Allaouchiche B, Pilleul. Intérêtde la double lecture du scanner corps entier dans la prise en chargedes polytraumatisés (full text in English on www.masson.fr/revues/jr).J Radiol 2008;89:325-30.

Trauma related mortality is a significant concern. Eachyear, about 150,000 individuals are victim of polytrauma, and about 9,000will die as a result. Some of the survivors have permanent functionalsequelae causing moderate to severe disability. Most polytraumasin France are from motor vehicle accidents (MVA). The fight againstthis type of trauma has become a public health concern and several measureshave been implemented over the last few years to reduce the numberof MVA victims. In addition to these preventive measures, improvementsmust also be achieved in the medical management of these patients,in order to reduce the mortality rate and severity of related sequelae.

In France, the term polytrauma most frequently describes a “severelyinjured patient with multiple organ injuries with at least one beingpotentially lethal at short to mid term”. This definitionsummarizes well the level of difficulties encountered by the medicalteams involved in the management of these patients. The urgent contextand the multiplicity of organ system involvement complicate theacquisition of a comprehensive work up of all lesions, which maycause some lesions to go undetected at initial evaluation. Theselesions may have dire consequences with increased mortality andmorbidity. Our own clinical experience and data from the literaturehave lead us to conclude that several lesions remain undetected dueto interpretation errors of imaging studies obtained at the timeof admission. In our practice, the imaging work up relies heavilyon whole body CT. The purpose of this study was to determine if standarddouble reading of whole body CT examinations would improve the managementof polytrauma patients.

This study was performed in the surgical emergency and radiologydepartment of the G pavilion of the Hôpital Édouard Herriotin Lyon. In 2005, 421 polytrauma patients were evaluated in thetrauma unit, most following MVA.

The triage of patients in the trauma unit is based on pre-establishedprotocols that are regularly reviewed to ensure optimal patientcare. Prior to their arrival, patients are first evaluated by theSAMU physician and classified into 3 severity groups (I, II andIII), according to the classification described by Kienlen and de LaCoussaye <sup>[1Kienlen, J, de La CoussayeJE. Management of multiple trauma in the emergency room. J Chir1999;136:240-51.

Cliquez ici pour aller à la section Références]</sup>.This allows better and more dedicated management of each individualpatient in terms of resuscitation, work-up and medical needs.

Patients in class I are severely injured with serious hemodynamicdistress defined as a systolic arterial pressure below 80 mm Hgin spite of good vascular filling and/or neurological distressdefined by a Glasgow score below 8 and signs of increased intracranialpressure and/or respiratory distress defined by an oxygensaturation below 90%.

Patients in class II are seriously injured but quite stabilizedby intensive care such as massive vascular loading, vasopressors, andoxygenation and/or with Glasgow score below 8 but withoutsigns of increased intracranial pressure.

Patients in class III are stable and are in no distress.

Class I patients are immediately managed in the trauma unit byattending physicians in each of the relevant medical or surgicalspecialties. The work up is minimized to only detect life threateninginjuries without delaying management. It includes a chest radiograph,a radiograph of the pelvis, and an US of the abdomen; echocardiography,transesophageal echocardiography and transcranial Doppler may beadded based on clinical findings. After completion of this workup, and if the patient remains unstable, the patient is directlytransferred to the surgical theater or angiography suite for immediate intervention:surgery or arterial embolization. If the patient is relatively stable,a whole body CT is obtained to further characterize the injuries.

The work up of class II patients includes a chest radiograph,a pelvic radiograph, radiographs of injured limbs, and a whole bodyCT.

The work up of class III patients includes a chest radiograph,a pelvic radiograph, radiographs of injured limbs, and is complementedby an abdominal US or a whole body CT based on the nature of the trauma.

The whole body CT examinations were performed using a multidetectorrow CT scanner. Our protocol included noncontrast images of thebrain followed by postcontrast images of the chest, abdomen andpelvis. The image acquisition was started 25 seconds after initiationof contrast administration at a rate of 3 ml/sec. Coronal,sagittal and/or oblique multiplanar reformatted imagesalong with 3D images were routinely obtained to better evaluatefractures. Whole body CT examinations were routinely interpretedby an attending radiologist and radiology resident. CT images werereviewed on an independent workstation or a PACS station. The reportwas verbally communicated to the trauma team followed by a writtenreport as soon as possible after completion of the exam.

This is a prospective study performed between January 1^st,2005 and July 31^st, 2005. During this time period, allwhole body CT examinations performed on trauma patients (adults,children) from the trauma unit of the G pavilion were systematicallyreviewed by two senior radiologists with expertise in trauma radiology within12 hours of acquisition. The reviewers included a senior radiologistwith expertise in musculoskeletal and neuroradiology a senior radiologistwith expertise in body imaging. The initial report was made availablein all cases. All examinations were reviewed on PACS workstations. Theaxial images as well as reformatted images generated per protocolwere available in all cases. An itemized review list classifyinglesions per organ systems (brain, abdomen-pelvis, chest, activebleeding, spine, bony pelvis, skull, facial bones, and other locations)was used. This expedited the transcription of the report from thesecond reading.

The first and second readings from all examinations werecompared and all lesions described on the second reading (gold standard)but not on the first reading were catalogued and analyzed. Theselesions were considered as missed on the initial reading. Theselesions were further subclassified into three levels of severity(severe, moderate, and minor) based on the Abbreviated Injury Scale(AIS). Severe lesions corresponded to AIS scores 4 and 5, moderatelesions to AIS scores 2 and 3, and minor lesions to AIS score 1.Lesions not addressed in the AIS were distributed into the 3 groupsafter consensus review by the radiologists and trauma team. Activebleeding, characterized by extravasation of iodinated contrast materialon CT, was categorized as a severe lesion if arterial embolizationwas required.

Results from the double read were provided to the lead traumaphysician. When necessary, patient management was modified basedon the updated imaging data.

The patient population with lesions missed on the first readwas compared to the patient population where no lesion was missedbased on all recorded data. The interobserver correlation was calculated.

A descriptive analysis of the collected data was performed.Quantitative variables were analyzed using the Student t-test or Mann-Whitneytest based on their distribution. Qualitative variables were analyzedusing the Fisher test or Chi 2 test. A p value<0.05 wasconsidered significant. Interobserver correlation was calculated usingthe kappa test. A kappa value of 0.60 was selected as the thresholdbetween fair and good correlation.

A total of 105 patients were included in this study, with78.1% males (n=82). The mean age was 35.7 years(95% confidence interval [CI]: 31.4-40).The mean age of the pediatric population was 10.6 years (95% CI:8.9-12.3), and the mean age of the adult population was 43.6 years(95% CI: 39.2-47.9). The mean ISS (injury severity score)was 27.1 (95% CI: 24.3-29.9). The mean Apache II scorewas 17.4 (95% CI: 15.3-19.5). A total of 10.5% ofpatients were classified as class I (n=11), 28.6% as classII (n=30) and 60.9% as class III (n=64).A total of 70.5% of patients admitted to the trauma unitwere admitted in the hospital for further management (n=74).A total of 12.4% of patients died within the first 28 daysafter trauma (n=13).

1. Number of lesions: lesions were missed in 75patients (71.4%) on the initial interpretation of wholebody CT examinations. A total of 280 lesions were missed from atotal of 765 lesions detected on the second read (36.6%).Of these 280 missed lesions, 31 (11%) were considered severe, 192(69%) moderate and 57 (20%) minor.

Thirty one of 84 (36.9%) severe lesions were initiallymissed, 192 of 567 (33.9%) moderate lesions were initiallymissed and 57 of 114 (50%) minor lesions were initiallymissed.

2. Type of lesions: Missed lesions were distributedas follows based on severity and organ systems (table I):

• severe lesions (n=31):
  • brain: intraparenchymalhematoma (n=7), epidural hematoma (n=1), subduralhematoma (n=2), subfalcine herniation (n=1);
  • facial bones: LeFort type III fracture (n=1);
  • spine: cord compression (n=1);
  • chest: hemomediastinum (n=6), pneumomediastinum (n=9),flail chest (n=3);
  • no severe lesion at the abdomen or pelvis, or bony pelvis andno contrast extravasation requiring arterial embolization was missedat the time of initial interpretation.
• moderate lesions (n=192):
  • brain: skull fracture(n=3), intraparenchymal petechial hemorrhage (n=3),cerebral contusion (n=4), subarachnoid hemorrhage (n=2),intraventricular hemorrhage (n=3), cerebral edema (n=10),pneumocephalus (n=2);
  • facial bones: varied fractures of the zygoma, orbits, temporomandibular joints,paranasal sinuses, temporal bones, hyoid bone, and thyroid cartilage(n=14);
  • spine: vertebral body fracture (n=10), occipital condylefracture (n=1), posterior articular fracture (n=2),pedicle fracture (n=1), transverse process fracture (n=10), spinousprocess fracture (n=7);
  • chest: fractures of 2 or more ribs (n=13), sternalfracture (n=2), single lobe lung contusion (n=8),hemothorax (n=6), pneumothorax (n=8), pleuraleffusion (n=2), pericardial effusion (n=1), pneumopericardium(n=1), atrial thrombus (n=1), active bleeding(n=2);
  • abdomen: splenic contusion (n=1), renal contusion (n=3),adrenal contusion (n=3), hemoperitoneum (n=4),intraperitoneal hematoma (n=6), retroperitoneal hematoma(n=12), pneumoperitoneum (n=1), retropneumoperitoneum(n=1), portal venous gas (n=2), hepatic artery pseudoaneurysm(n=1), active bleeding (n=1);
  • pelvis: acetabular fracture (n=7), inferior pubic ramusfracture (n=4), superior pubic ramus fracture (n=2),parasymphyseal fracture (n=1), articular pelvic fracture(n=1), urethral injury (n=1), sacral fracture(n=4), active bleeding (n=3);
  • limbs: humeral fracture (n=4), clavicle fracture (n=7),scapular fracture (n=4), femoral fracture (n=2),active bleeding (n=1).
• minor lesions (n=57):
  • brain: cephalhematoma (n=5);
  • facial bones: nasal bone fracture (n=3), hemosinus(n=1), foreign body (n=2), orbital emphysema (n=1);
  • spine: foreign body (n=1);
  • chest: rib fracture (n=16); subcutaneous emphysema(n=8), chest wall hematoma (n=1);
  • abdomen: subcutaneous emphysema (n=7), abdominal wallhematoma (n=6), intramuscular hematoma (n=4);
  • pelvis: coccygeal fracture (n=2).
• lesion distribution based on organ systems was as follows:
  • brain: 43 lesions (15.4%);
  • facial bones: 22 lesions (7.9%);
  • spine: 34 lesions (12.1%);
  • chest: 87 lesions (30%);
  • abdomen: 52 lesions (18.6%);
  • pelvis: 24 lesions (8.6%);
  • limbs: 18 lesions (6.4%).

3. Comparison of patient populations: We have compared the populationof patients where lesions were missed at initial interpretationwith the population of patients where no lesion was missed at initialinterpretation, based on collected data. These observations aresummarized in table II.

The interobserver agreement was 63.3% (kappa=0.409;95% CI: 0.35-0.46).

The management of polytrauma patients may be problematicwhen injuries are missed at the time of initial evaluation. Thesemissed lesions may have severe functional consequences and may even causedeath. A number of strategies were proposed to decrease the numberof missed lesions, such as the implementation of pre-establishedstandardized protocols at the time of admission at the trauma roomor repeat clinical and paraclinical evaluations. In spite of these efforts,the incidence of missed lesions remains high requiring the implementation ofadditional measures.

A review of the underlying causes for these missed lesions showsthat a large portion are due to errors in the initial interpretationof whole body CT examinations. This study was conducted as partof a process to improve the care of our polytrauma patients. Thepurpose was to assess the relevance of a procedure that could reducethe morbidity and mortality in these patients.

We have hypothesized that routine double reading of imaging studiesin polytrauma patients would reduce the number of missed lesions.In our department, the imaging work up of polytrauma patients reliesheavily on whole body CT. This is why a double reading program wasimplemented for whole body CT examinations in polytrauma cases.All lesions missed on the initial interpretation were recorded andanalyzed to determine the value of this program.

We have observed a high number of lesions that were missed onthe initial interpretation with about 280 missed lesions from atotal of 765 lesions (fig. 1), including a non-negligible numberof severe lesions (11%). About three quarters of patientshad at least one lesion that was missed. If the error rate is definedas the proportion of examinations with at least one interpretationerror, then, the error rate in our series would be 71.4% withan interobserver agreement of 0.41.

The error rate in the interpretation of imaging studies reportedin the literature is much lower, between 0.34-10.6% for radiographs <sup>[2Halsted MJ, Kumar H, PaquinJJ, Poe SA, Bean JA et al. Diagnostic errors by radiology residentsin interpreting pediatric radiographs in an emergency setting. PediatrRadiol 2004;34:331-6.

Cliquez ici pour aller à la section Références]</sup>, <sup>[3Brunswick JE, IlkhanipourK, Fuchs S, Seaberg D. Emergency medicine resident interpretationof pediatric radiographs. Acad Emerg Med 1996;3:790-3.

Cliquez ici pour aller à la section Références]</sup> and1.5-38.7% for CT <sup>[4Lal NR, Murray UM, EldevikOP, Desmond JS. Clinical consequences of misinterpretations of neuroradiologicCT scans by on-call radiology residents. AJNR Am J Neuroradiol 2000;21:124-9.

Cliquez ici pour aller à la section Références]</sup>, <sup>[5Alfaro D, Levitt MA, EnglishDK, Williams V, Eisenberg R. Accuracy of interpretation of cranialcomputed tomography scans in an emergency medicine residency program.Ann Emerg Med 1995;25:169-74.

Cliquez ici pour aller à la section Références]</sup>. This may be due to the fact thatnone of these studies evaluated whole body CT examinations in polytraumapatients. Most studies evaluated standard radiographs <sup>[6Gratton MC, Salomone JA 3rd,Watson WA. Clinically significant radiograph misinterpretationsat an emergency medicine residency program. Ann Emerg Med 1990;19:497-502.

Cliquez ici pour aller à la section Références]</sup>, <sup>[7Nitowski LA, O’ConnorRE, Reese CL 4th. The rate of clinically significant plain radiographmisinterpretation by faculty in an emergency medicine residencyprogram. Acad Emerg Med 1996;3:782-9.

Cliquez ici pour aller à la section Références]</sup>, <sup>[8Lufkin KC, Smith SW, MatticksCA, Brunette DD. Radiologists’ review of radiographs interpretedconfidently by emergency physicians infrequently leads to changesin patient management. Ann Emerg Med 1998;31:202-7.

Cliquez ici pour aller à la section Références]</sup>, <sup>[9McLauchlan CA, Jones K, GulyHR. Interpretation of trauma radiographs by junior doctors in accidentand emergency departments: a cause for concern? J Accid Emerg Med1997;14:295-8.

Cliquez ici pour aller à la section Références]</sup>, <sup>[10Cascade PN, Kazerooni EA,Gross BH, Quint LE et al. Evaluation of competence in the interpretationof chest radiographs. Acad Radiol 2001;8:315-21.

Cliquez ici pour aller à la section Références]</sup>, <sup>[11Gatt ME, Spectre G, PaltielO, Hiller N, Stalnikowicz R. Chest radiographs in the emergencydepartment: is the radiologist really necessary? Postgrad Med J 2003;79:214-7.

Cliquez ici pour aller à la section Références]</sup> orCT examinations of specific body parts: brain <sup>[4Lal NR, Murray UM, EldevikOP, Desmond JS. Clinical consequences of misinterpretations of neuroradiologicCT scans by on-call radiology residents. AJNR Am J Neuroradiol 2000;21:124-9.

Cliquez ici pour aller à la section Références]</sup>, <sup>[5Alfaro D, Levitt MA, EnglishDK, Williams V, Eisenberg R. Accuracy of interpretation of cranialcomputed tomography scans in an emergency medicine residency program.Ann Emerg Med 1995;25:169-74.

Cliquez ici pour aller à la section Références]</sup>, <sup>[12Roszler MH, McCarroll KA,Rashid T, Donovan KR, Kling GA. Resident interpretation of emergencycomputed tomographic scans. Invest Radiol 1991;26: 374-6.

Cliquez ici pour aller à la section Références]</sup>, <sup>[13Wysoki MG, Nassar CJ, Koenigsberg RA,Novelline RA et al. Head trauma: CT scan interpretation by radiologyresidents versus staff radiologists. Radiology 1998; 208:125-8.

Cliquez ici pour aller à la section Références]</sup>, <sup>[14Heng, RC, Bell KW. Interpretingurgent brain CT scans: Does review by a radiology trainee make adifference in accuracy? Australas Radiol 2001;45:134-40.

Cliquez ici pour aller à la section Références]</sup>, <sup>[15Arendts G, Manovel A, ChaiA. Cranial CT interpretation by senior emergency department staff.Australas Radiol 2003; 47:368-74.

Cliquez ici pour aller à la section Références]</sup>,cervico-thoracic and thoracic regions <sup>[16Fletcher BD, Glicksman AS,Gieser P. Interobserver variability in the detection ofcervical-thoracic Hodgkin’s disease by computed tomography.J Clin Oncol 1999;17:2153-9.

Cliquez ici pour aller à la section Références]</sup>, <sup>[17Compère V, GenevoisA, Le Corre A, Hellot MF, Bourguignon N et al. Influence of medicalspeciality and experience on interpretation of helicoidal thoraciccomputed tomography in blunt chest trauma. Intensive Care Med 2003; 29:770-3.

Cliquez ici pour aller à la section Références]</sup>,abdomen and abdominopelvic regions <sup>[18Bechtold RE, Chen MY, OttDJ, Zagoria RJ et al. Interpretation of abdominal CT: analysis oferrors and their causes. J Comput Assist Tomogr 1997;21:681-5.

Cliquez ici pour aller à la section Références]</sup>, <sup>[19Yoon LS, Haims AH, BrinkJA, Rabinovici R, Forman HP. Evaluation of an emergency radiologyquality assurance program at a level I trauma center: abdominaland pelvic CT studies. Radiology 2002;224:42-6.

Cliquez ici pour aller à la section Références]</sup> orother body parts <sup>[20Velmahos GC, Fili C, VassiliuP, Nicolaou N, Radin R, Wilcox A. Around-the-clock attendingradiology coverage is essential to avoid mistakes in the care oftrauma patients. Am Surg 2001;67: 1175-7.

Cliquez ici pour aller à la section Références]</sup>, <sup>[21Carney E, Kempf J, DeCarvalhoV, Yudd A, Nosher J. Preliminary interpretations of after-hoursCT and sonography by radiology residents versus final interpretationsby body imaging radiologists at a level 1 trauma center. AJR AmJ Roentgenol 2003;181:367-73.

Cliquez ici pour aller à la section Références]</sup>. One study evaluated whole bodyCT in a non-trauma patient population and reported a discordancerate of 37% for oncologic patients <sup>[22Gollub MJ, Panicek DM, BachAM, Penalver A, Castellino RA. Clinical importance of reinterpretationof body CT scans obtained elsewhere in patients referred for careat a tertiary cancer center. Radiology 1999;210:109-12.

Cliquez ici pour aller à la section Références]</sup>. Itmay be that the large number of images and the acute clinical contextincrease the error rate.

Our study design is different from other studies since we havenot compared the interpretations from physicians of different qualification(resident versus attending, radiologist versus surgeon, trauma ofemergency room physician). In our study, both CT interpretationswere performed by senior radiologists. The only difference was thatthe initial interpretation was contemporaneous to the patient admissionand performed by the on-call radiologist of the day whereas thesecond interpretation was performed by a trauma radiologist. Wehave not noted any difference in the quality of interpretationsbetween both groups. The error rates on initial interpretation betweenthe radiologists of pavilion G and other radiologists were identical.The large number of missed lesions on the initial interpretationsis thus not related to the radiologist’s experience.

Interpretation of a whole body CT represents a considerable amountof work. The urgent and stressful clinical context creates a disruptivework environment that interferes with the radiologist’sability to concentrate. Also, the patient’s functionalprognosis and even survival may depend on the speediness of theinterpretation, therefore reducing the amount of time available tothe radiologist to generate a report, time that could otherwisebe available under different clinical circumstances. Also, a numberof CT examinations on polytrauma patients are preformed at night,a period when attentiveness and ability to concentrate are reduced.

These considerations probably provide a basis for understandingwhy a large number of lesions were missed at the time of initialinterpretation in our study (fig. 2).

Thirty percent of lesions initially missed on whole body CT werelocated at the chest level, with over 50% of severe lesionsinitially missed involving the chest (n=18) (table II).Additional time should thus be allocated to the review of chest CTimages since the largest proportion of missed lesions, especiallysevere lesions, involved the chest (fig. 3).

Also, missed lesions were more frequent in adult patients thanin younger or pediatric patients (patients in the second populationwere generally younger than patients in the first population [26.2versus 39.5 years; p<0.001], with larger number ofpediatric patients [40% versus 14.7%; p=0.01]).

A comparison of injury severity between both patient populationsdemonstrates that severity scores were higher in the patient populationwith missed lesions (mean ISS of 30.7 versus 18; p<0.001 – meanApache II score of 19 versus 13.5; p<0.001). The percentageof patients admitted to the intensive care unit was higher in thispopulation (80% versus 46.7%; p=0.0017).On the other hand, there was no significant difference between bothgroups for the mortality rate at 28 days (p=0.34).

As such, more severely injured patients have more lesions onwhole body CT and more of these lesions are missed at the time ofinitial interpretation. The fact that mortality rates at 28 daysare not significantly different eliminates the hypothesis of a causalrelation between missed lesion and patient death.

The CT examinations of severely injured patients should be morecarefully reviewed, and double reading should be considered.

We have defined all lesions that were described on the secondread but not on the initial read as missed lesions. We have viewedthe second read as the gold standard. It might have been preferableto obtain a second read from radiologists with expertise in eachof the organ systems covered on the whole body CT examinations andcompare these to the initial interpretations. However, given the circumstancessurrounding the second interpretation, we believe that the second readsobtained in our study are probably very close to the reality comparedto the initial interpretations (review by an experienced radiologist,in a calm environment, with possibility to consult with colleagues).

Management of polytrauma patients is complex. Advances havealready been achieved with the implementation of standardized protocols,but the incidence of missed injuries remains high. Our results indicatethat a large number of injuries are missed at the time of initialinterpretation of whole body CT examinations.

As such, double reading of these CT examinations is part of aglobal process to improve patient management and quality of care.The CT examinations of severely injured patients should be morecarefully reviewed since lesions are more frequently missed in thispatient population. Finally, additional time should be allocated tothe review of chest CT images since the largest proportion of missedlesions, especially severe lesions, involve the chest.