Advertisement

The AO spine upper cervical injury classification system: Do work setting or trauma center affiliation affect classification accuracy or reliability?

Open AccessPublished:August 15, 2022DOI:https://doi.org/10.1016/j.injury.2022.08.030

      Abstract

      Purpose

      To assess the accuracy and reliability of the AO Spine Upper Cervical Injury Classification System based on a surgeons’ work setting and trauma center affiliation.

      Methods

      A total of 275 AO Spine members participated in a validation of 25 upper cervical spine injuries, which were evaluated by computed tomography (CT) scans. Each participant was grouped based on their work setting (academic, hospital-employed, or private practice) and their trauma center affiliation (Level I, Level II or III, and Level IV or no trauma center). The classification accuracy was calculated as percent of correct classifications, while interobserver reliability, and intraobserver reproducibility were evaluated based on Fleiss’ Kappa coefficient.

      Results

      The overall classification accuracy for surgeons affiliated with a level I trauma center was significantly greater than participants affiliated with a level II/III center or a level IV/no trauma center on assessment one (p1<0.0001) and two (p2 = 0.0003). On both assessments, surgeons affiliated with a level I or a level II/III trauma center were significantly more accurate at identifying IIIB injury types (p1 = 0.0007; p2 = 0.0064). Academic surgeons and hospital employed surgeons were significantly more likely to correctly classify type IIIB injuries on assessment one (p1 = 0.0146) and two (p2 = 0.0015). When evaluating classification reliability, the largest differences between work settings and trauma center affiliations was identified in type IIIB injuries.

      Conclusion

      Type B injuries are the most difficult injury type to correctly classify. They are classified with greater reliability and classification accuracy when evaluated by academic surgeons, hospital-employed surgeons, and surgeons associated with higher-level trauma centers (I or II/III).

      Keywords

      Introduction

      Upper cervical spine trauma occurs at a disproportionally high rate in geriatric patients.1 Typically, the injury mechanism is low-energy trauma with the most commonly identified fracture being a sagittally-oriented odontoid fracture [
      • Spivak J.M.
      • Weiss M.A.
      • Cotler J.M.
      • Call M.
      Cervical spine injuries in patients 65 and older.
      ]. Atlas fractures are increasingly recognized in elderly patients with a reported 700% increase in injury rate during the last 20 years [
      • Lyons J.G.
      • Mian H.M.
      Epidemiology of atlas fractures in the United States: a 20-year analysis.
      ].
      Given that most upper cervical spine injuries in geriatric patients are caused by ground level falls, and greater than 70% of upper cervical spine injuries result in no neurologic deficit, many of these patients will initially be evaluated at low acuity trauma centers [
      • Malik S.A.
      • Murphy M.
      • Connolly P.
      • O'Byrne J
      Evaluation of morbidity, mortality and outcome following cervical spine injuries in elderly patients.
      ]. Current evidence points to no difference in patient mortality based on whether a spine fracture is transferred to a high-level (level I or II) trauma center or if it is managed at a low-level center [
      • Barmparas G.
      • Cooper Z.
      • Haider A.H.
      • Havens J.M.
      • Askari R.
      • Salim A.
      The elderly patient with spinal injury: treat or transfer?.
      ]. Thus, the work setting (academic, hospital employed or private practice) and trauma center affiliation (levels I, II, III, etc.) of a surgeon may not affect their ability to accurately identify simple atlas or dens injuries given their frequent presentation.
      Upper cervical spine trauma is seen in a bimodal distribution with younger patients presenting to high-level trauma centers predominantly after falls from height or motor vehicle collisions [
      • Wang H.
      • Ou L.
      • Zhou Y.
      • Li C.
      • Liu J.
      • Chen Y.
      • Yu H.
      • Wang Q.
      • Zhao Y.
      • Han J.
      • Xiang L.
      Traumatic upper cervical spinal fractures in teaching hospitals of China over 13 years: a retrospective observational study.
      ]. C2 fractures, and more specifically odontoid fractures, remain the most common upper cervical spine injuries evaluated in younger trauma patients [
      • Wang H.
      • Ou L.
      • Zhou Y.
      • Li C.
      • Liu J.
      • Chen Y.
      • Yu H.
      • Wang Q.
      • Zhao Y.
      • Han J.
      • Xiang L.
      Traumatic upper cervical spinal fractures in teaching hospitals of China over 13 years: a retrospective observational study.
      ]. However, less common fracture presentations including craniocervical junction dislocations and atypical hangman's fractures are almost exclusively seen with motor vehicle collisions or falls from height [
      • Benzel E.C.
      • Hart B.L.
      • Ball P.A.
      • Baldwin N.G.
      • Orrison W.W.
      • Espinosa M.
      Fractures of the C-2 vertebral body.
      ,
      • Labler L.
      • Eid K.
      • Platz A.
      • Trentz O.
      • Kossmann T.
      Atlanto-occipital dislocation: four case reports of survival in adults and review of the literature.
      ]. Thus, these patients are predominantly evaluated at high-level trauma centers. The greater number of Type B and C injuries evaluated by surgeons’ affiliated with level I trauma centers, may provide them an advantage when attempting to classify a variety of upper cervical spine injuries based on the AO Spine Upper Cervical Injury Classification System. Therefore, the purpose of our study was to determine if a surgeons’ work setting or trauma center affiliation affected their classification accuracy, interobserver reliability, or intraobserver reproducibility.

      Methods

      Injury classification description

      A detailed description of this fracture classification has previously been provided [
      • Vaccaro A.R.
      • Karamian B.A.
      • Levy H.A.
      • et al.
      Update on upper cervical injury classifications: the new AO upper cervical spine classification system.
      ]. In brief, injuries are assigned a roman numeral based on their injury location (I. Craniovertebral junction and occipital condyles; II. Atlas and C1-C2 joint; III. Dens, C2 ring, and C2-C3 joint). The injury is then grouped based on its stability (Type A — Stable, purely bony injuries; Type B — Tension band failures or ligamentous injuries without evidence of instability; Type C — Vertebral body translation consistent with a fracture subluxation or dislocation). Neurologic status is classified in the same manner as previous AO Spine injury classifications and injury modifiers are assigned when the injury could potentially alter management decisions. However, neither neurologic status nor injury modifiers were evaluated during this classification (Fig. 1).
      Fig. 1
      Fig. 1Pictorial representation of the AO Spine Upper Cervical Injury Classification. The classification is based on injury location (occipital condyle and craniocervical junction, C1 ring and C1–2 joint, and C2 and C2–3 joint) and injury type (bony, tension band, ligamentous with translation). Permission to use this figure was granted by the AO Foundation©, AO Spine, Switzerland.

      Classification validation

      A total of 275 AO Spine members participated in an international validation of the AO Spine Upper Cervical Injury Classification System. Each participant reviewed a tutorial video in English prior to the validation (https://www.youtube.com/watch?v=KyUYfa_JMb4). All participants then completed a sample validation of three cases with feedback from the instructor, who was an original creator of the classification system. After training completion, the official validation was conducted via a live, online webinar format [
      • Lambrechts M.J.
      • Schroeder G.D.
      • Karamian B.A.
      • Canseco J.A.
      • Oner C.
      • Vialle E.
      • Rajasekaran S.
      • Hazenbiller O.
      • Dvorak M.R.
      • Benneker L.M.
      • Kandziora F.
      • Schnake K.
      • Kepler C.K.
      • Vaccaro A.R.
      AO spine subaxial classification group members. Development of online technique for international validation of the AO spine subaxial injury classification system.
      ]. The validation consisted of 25 computed tomography (CT) videos, which consisted of axial, sagittal, and coronal cuts of upper cervical spine injuries. The video was viewed once by each participant who was provided with associated radiographic key images. An online REDCap survey stored all answers for statistical analysis. Three weeks were given between the first and second assessment. Prior to the second assessment, the cases were re-randomized.

      Gold standard committee

      Members of the AO Spine Knowledge Forum Trauma, which included original creators of the classification system, formed the gold standard committee. All injury classifications reached unanimous consensus from the committee prior to distribution to the study participants. Any injury without consensus agreement was resolved through live webinar meetings where a discussion was held to achieve unanimous agreement.

      Statistics

      Each study participant was grouped based on their work setting (academic, hospital-employed, or private practice) and their trauma center affiliation (Level I, Level II or III, and Level IV or no trauma center). Each study participant was then evaluated based on the percentage of injuries they correctly classified.
      Kappa coefficients (ƙ) were calculated based on the classification agreement between study participants (interobserver reliability) and the same participant (intraobserver reproducibility). All kappa values used Fleiss’ Kappa coefficient, which allows for missed ratings and comparisons between multiple validation members [
      ]. Interpretation of the reliability and reproducibility was based on the Landis and Koch convention, which categorizes Kappa values as “slight” (0–0.2), “fair” (>0.2 - 0.4), “moderate” (>0.4 - 0.60), “substantial” (>0.6 - 0.8), and “excellent” (>0.8–1.0) [
      • Landis J.R.
      • Koch G.G.
      The measurement of observer agreement for categorical data.
      ]. Pearson's chi square test or Fisher's exact test was utilized to screen for potentially relevant associations.

      Results

      The validation consisted of 25 upper cervical spine injuries, which were evaluated by 275 study participants. Most study participants worked in an academic environment (N = 120, 43.6%) or were hospital-employed (N = 120, 43.6%), while the remainder worked in private practice (N = 35, 12.7%). Additionally, most participants were affiliated with level I trauma centers (N = 192, 69.8%), but 66 (24%) were affiliated with level II/III center and 17 (6.2%) were affiliated with either a level IV trauma center or no trauma center (Table 1). Each injury subtype was validated a minimum of two times with the exception of IB injuries, which could not be evaluated since they were not included in the AO Spine imaging database (Supplemental Data 1).
      Table 1Demographics of the participants in the AO Spine Upper Cervical Injury Classification.
      Survey DemographicsN (%)
      Work Setting# of participants275 (100)
      Academic120 (43.6)
      Hospital Employed120 (43.6)
      Private Practice35 (12.7)
      Trauma Center Level# of participants275 (100)
      Level I192 (69.8)
      Level II49 (17.8)
      Level III17 (6.2)
      Level IV12 (4.4)
      No trauma center5 (1.8)

      Classification accuracy

      When evaluating work setting, study participants affiliated with an academic practice or working in a hospital were significantly more likely to correctly classify type IIIB injuries on assessment one (p1=0.0146) and assessment two (p2=0.0015), compared to those in a private practice. Broader differences emerged when evaluating classification accuracies based on trauma center affiliation. The overall classification accuracy for participants affiliated with a level I trauma center was significantly greater than participants affiliated with a level II/III center or a level IV/no trauma center on assessment one (p1<0.0001) and assessment two (p2=0.0003). On both assessments, surgeons affiliated with a level I or a level II/III trauma center were significantly more accurate at identifying IIIA (assessment 1: p1=0.0138; assessment 2: p2=0.0008) and IIIB injury types (p1=0.0007; p2=0.0064) when compared to surgeons affiliated with a level IV/no trauma center. Although there were statistically greater classification accuracies for level I affiliated spine surgeons when assessing type IC injuries (p1=0.0251) and type IIB injuries (p1=0.0016) on assessment one, these did not persist on assessment two (p2=0.3420 and p2=0.0715, respectively) (Table 2).
      Table 2Proportion and percent of correctly classified injuries based on the surgeons work affiliation and trauma center setting.
      First Assessment
      Work SettingTrauma Center
      Injury ClassificationAcademicC/N (%)Hospital EmployedC/N (%)Private PracticeC/N (%)P-valueLevel IC/N (%)Level II/IIIC/N (%)Level IV/ No traumaC/N (%)P-value
      Overall2268/2811 (80.7)2098/2646 (79.3)570/740 (77.0)0.07413515/4318 (81.4)1137/1491 (76.3)284/388 (73.2)<0.0001
      Indicates statistical significance with P<0.05.
      IA184/227 (81.1)177/210 (84.3)54/59 (91.5)0.1454289/345 (83.8)99/120 (82.5)27/31 (87.1)0.8232
      IC210/225 (93.3)194/213 (91.1)53/60 (88.3)0.4065324/346 (93.6)107/120 (89.2)26/32 (81.2)0.0251
      Indicates statistical significance with P<0.05.
      IIA357/446 (80.0)333/425 (78.4)93/116 (80.2)0.8039549/689 (79.7)186/236 (78.8)48/62 (77.4)0.8922
      IIB318/453 (70.2)287/424 (67.7)83/120 (69.2)0.7237503/694 (72.5)147/240 (61.2)38/63 (60.3)0.0016
      Indicates statistical significance with P<0.05.
      IIC146/226 (64.6)139/213 (65.3)30/59 (50.8)0.1080227/347 (65.4)71/120 (59.2)17/31 (54.8)0.2856
      IIIA382/450 (84.9)358/422 (84.8)97/120 (80.8)0.5222597/690 (86.5)192/239 (80.3)48/63 (76.2)0.0138
      Indicates statistical significance with P<0.05.
      IIIB242/336 (72.0)222/317 (70.0)50/89 (56.2)0.0146
      Indicates statistical significance with P<0.05.
      376/517 (72.7)116/179 (64.8)22/46 (47.8)0.0007
      Indicates statistical significance with P<0.05.
      IIIC429/448 (95.8)388/422 (91.9)110/117 (94.0)0.0625650/690 (94.2)219/237 (92.4)58/60 (96.7)0.3981
      Second Assessment
      Injury ClassificationWork SettingTrauma Center
      AcademicC/N (%)Hospital EmployedC/N (%)Private PracticeC/N (%)P-valueLevel IC/N (%)Level II/IIIC/N (%)Level IV/ No traumaC/N (%)P-value
      Overall1921/2368 (81.1)1795/2295 (78.2)543/741 (73.3)<0.0001
      Indicates statistical significance with P<0.05.
      2980/3725 (80.0)1034/1335 (77.5)245/344 (71.2)0.0003
      Indicates statistical significance with P<0.05.
      IA155/190 (81.6)149/185 (80.5)46/60 (76.7)0.7043243/299 (81.3)87/108 (80.6)20/28 (71.4)0.4542
      IC173/189 (91.5)163/184 (88.6)52/60 (86.7)0.4685270/298 (90.6)92/107 (86.0)26/28 (92.9)0.3420
      IIA317/377 (84.1)295/366 (80.6)96/118 (81.4)0.4462492/593 (83.0)178/214 (83.2)38/54 (70.4)0.0623
      IIB265/381 (69.6)247/366 (67.5)71/117 (60.7)0.2010418/598 (69.9)130/211 (61.6)35/55 (63.6)0.0715
      IIC122/190 (64.2)120/183 (65.6)31/59 (52.5)0.1820187/298 (62.8)69/107 (64.5)17/27 (63.0)0.9501
      IIIA328/378 (86.8)297/365 (81.4)96/119 (80.7)0.0884504/595 (84.7)181/212 (85.4)36/55 (65.5)0.0008
      Indicates statistical significance with P<0.05.
      IIIB220/285 (77.2)203/278 (73.0)52/90 (57.8)0.0015
      Indicates statistical significance with P<0.05.
      333/449 (74.2)118/162 (72.8)24/42 (57.1)0.0604
      IIIC341/378 (90.2)321/368 (87.2)99/118 (83.9)0.1455533/595 (89.6)179/214 (83.6)49/55 (89.1)0.0693
      Abbreviations: Correct (C) – total number of correct responses; (N) - total number of injury films evaluated.
      low asterisk Indicates statistical significance with P<0.05.
      When stratifying by injury location alone and by injury type alone, work setting did not significantly affect the classification accuracy for injury location (p1=0.2435l) or injury type (p1=0.1322) in assessment 1. However, significant differences in injury location accuracy were found on assessment two (p2<0.0001) between groups due to the lower accuracy of private practice surgeons ability to correctly classify location I (p2=0.0038), II (p2=0.0002), or III (p2=0.003) injuries. Private practice surgeons also had a lower classification accuracy when evaluating injury type (p2=0.0029) predominantly due to lower accuracy when classifying type B injuries (p2=0.012). When evaluating trauma centers, surgeons affiliated with level I centers were most likely to correctly identify the injury based on its location (p1=0.0104; p2=0.0006) and injury type (p1<0.0001; p2=0.0280), although the effect sizes were mostly rather small. (Supplemental Data 2).

      Interobserver reliability

      Surgeons in an academic practice had substantial interobserver reliability on both assessments (ƙ1=0.644; ƙ2=0.650), while private practice surgeons had moderate reliability on both assessments (ƙ1=0.593; ƙ2=0.515). Hospital-employed physicians had substantial reliability on assessment one (ƙ1=0.624) and moderate reliability on assessment two (ƙ2=0.599). When evaluating injury subtypes based on work setting, the largest differences between academic and private practice work settings was identified for type IIIB injuries. When analyzing surgeons’ interobserver reliability based on their trauma center affiliation, surgeons affiliated with level I centers had substantial reliability on both assessments (ƙ1=0.655; ƙ2=0.630), while surgeons affiliated with a level II/III (ƙ1=0.578; ƙ2=0.589) or level IV/no trauma center (ƙ1=0.556; ƙ2=0.494) had moderate reliability. Similar to our findings based on work setting, the largest differences in reliability for surgeons based on their trauma center affiliation was for type IIIB injuries (Table 3).
      Table 3Interobserver reliability of the AO Spine Upper Cervical Injury Classification System based on the surgeons work setting and trauma center affiliation.
      First Assessment
      Injury ClassificationWork SettingTrauma Center Level
      Academic (ƙ)Hospital Employed (ƙ)Private Practice (ƙ)I (ƙ)II/III (ƙ)IV/No Trauma (ƙ)
      Overall0.6440.6240.5930.6550.5780.556
      IA0.7320.7600.7780.7550.7550.674
      IC0.8730.8600.8020.8890.7980.773
      IIA0.5910.5720.6180.6100.5440.528
      IIB0.4840.4690.4530.5140.3930.340
      IIC0.4690.4440.3670.4720.4070.301
      IIIA0.7030.6890.6250.7130.6390.633
      IIIB0.5650.5310.3800.5780.4390.360
      IIIC0.8310.7810.7300.8120.7510.790
      Second Assessment
      Injury ClassificationWork SettingTrauma Center Level
      Academic (ƙ)Hospital Employed (ƙ)Private Practice (ƙ)I (ƙ)II/III (ƙ)IV/No Trauma (ƙ)
      Overall0.6500.5990.5150.6300.5890.494
      IA0.7280.6990.6310.7250.7070.493
      IC0.8770.8290.7570.8520.8220.781
      IIA0.6340.5750.5390.6200.5790.444
      IIB0.4800.4340.3550.4710.3860.399
      IIC0.4720.4930.3510.4720.4680.377
      IIIA0.7100.6530.5770.6810.6700.508
      IIIB0.5900.5310.3600.5620.5080.343
      IIIC0.8220.7620.6140.7820.7340.710
      When sub-analyzing the injuries based on injury location and injury type, academic surgeons (ƙ1=0.862; ƙ2=0.860) and hospital employed surgeons (ƙ1=0.855; ƙ2=0.829) had excellent reliability for injury location on both assessments, while private practice surgeons had excellent reliability on assessment one (ƙ1=0.819) and substantial reliability on assessment two (ƙ1=0.706). On analysis of injury types, each group had moderate reliability on assessment 1 with only academic surgeons reaching substantial reliability on assessment two (ƙ1=0.605) (Supplemental Data 3). Some differences emerged when evaluating trauma center affiliation, with only surgeons at level I (ƙ1=0.870; ƙ1=0.842) or level II/III (ƙ1=0.825; ƙ1=0.807) centers having excellent reliability when classifying injury location. When evaluating injury type, only level I trauma center affiliated surgeons reached substantial reliability on assessment one (ƙ1=0.616), while all groups achieved moderate reliability on assessment two (Supplemental Data 3).

      Intraobserver reproducibility

      Injury classification reproducibility was on average substantial for academic surgeons (ƙ=0.73 ± 0.18), hospital employed surgeons (ƙ=0.69 ± 0.20), and private practice surgeons (ƙ=0.67 ± 0.21). Reproducibility was also on average substantial for surgeons affiliated with a level I trauma center (ƙ=0.73 ± 0.17) and level II/III center (ƙ=0.66 ± 0.25), but it was moderate for surgeons at a level IV/no trauma affiliation (ƙ=0.60 ± 0.15). When assessing the amount of surgeons who obtained excellent intraobserver reproducibility, academic surgeons (p = 0.0445) and surgeons affiliated with a level I trauma center (p = 0.0324) were the most likely to achieve excellent reproducibility (Table 4).
      Table 4Intraobserver reproducibility of the AO Spine Upper Cervical Injury Classification System based on surgeon work setting and trauma center affiliation.
      Level of AgreementWork SettingTrauma Center Level
      AcademicN (%)Hospital EmployedN (%)Private PracticeN (%)P-valueIN (%)II/IIIN (%)IV/No TraumaN (%)P-value
      Slight (<0.2)1 (1.1)3 (3.5)1 (3.9)0.43582 (1.4)3 (5.8)00.2799
      Fair (0.2–0.40)4 (4.3)5 (5.8)1 (3.9)0.89317 (5.0)2 (3.9)1 (7.7)0.7411
      Moderate (>0.40–0.60)14 (15.1)14 (16.1)4 (15.4)0.981514 (9.9)13 (25.0)5 (38.5)0.0023
      Indicates statistical significance with P<0.05.
      Substantial (>0.60–0.80)30 (32.3)35 (40.2)14 (53.9)0.120657 (40.4)16 (30.8)6 (46.2)0.3954
      Excellent (>0.80–1.0)44 (47.3)30 (34.5)6 (23.1)0.0445
      Indicates statistical significance with P<0.05.
      61 (43.3)18 (34.6)1 (7.7)0.0324
      Indicates statistical significance with P<0.05.
      low asterisk Indicates statistical significance with P<0.05.

      Discussion

      An ideal injury classification should be reliable and reproducible, although underlying demographics may alter the classification reliability. The frequency at which a surgeon evaluates spine injuries, and their associated exposure to upper cervical spine injuries, may be one factor that impacts classification reliability. Improving our understanding of how demographic factors affect classification accuracy and reliability may provide an opportunity to provide global education, thereby, improving classification generalizability. The results of our study suggest that academic surgeons and hospital-employed surgeons generate high classification accuracies with greater interobserver reliability than private practice surgeons. Additionally, surgeons affiliated with a high-level trauma center (I or II/III) have the greatest interobserver reliability and classification accuracy. Type IIIB injuries are the most difficult injuries to accurately and reliably classify.
      When evaluating a surgeon's work setting and trauma center affiliation, granular analysis of specific injury types demonstrated that IIIB injuries resulted in the most disparity in classification accuracy and surgeon reliability. These injury types are visualized on CT scans as complex C2 coronal fractures with variable extension into the pars or lamina depending on the mechanism of injury as depicted by Benzel [
      • Benzel E.C.
      • Hart B.L.
      • Ball P.A.
      • Baldwin N.G.
      • Orrison W.W.
      • Espinosa M.
      Fractures of the C-2 vertebral body.
      ]. In this classification, the fracture characteristics are linked to the mechanism of injury, which includes an extension with axial load variant. This is more commonly labeled an atypical hangman's fracture (IIIB injury as classified by the AO Spine Upper Cervical Injury Classification System due to a tension band failure mechanism). The remaining injury mechanisms described by Benzel include flexion with axial load, flexion distraction, and hyperextension with axial load injuries, which predominantly result in AO Spine Type C injuries, due to vertebral body translation resulting from either intervertebral disk ruptures or avulsion fractures of the longitudinal ligaments [
      • Benzel E.C.
      • Hart B.L.
      • Ball P.A.
      • Baldwin N.G.
      • Orrison W.W.
      • Espinosa M.
      Fractures of the C-2 vertebral body.
      ]. Since nearly all IIIB injuries are the result of high energy mechanism trauma (motor vehicle collisions or falls from height), these injury types are unlikely to be encountered by a private practice surgeon or surgeons without a high-level trauma center affiliation, likely resulting in their lower classification accuracy and reliability [
      • Benzel E.C.
      • Hart B.L.
      • Ball P.A.
      • Baldwin N.G.
      • Orrison W.W.
      • Espinosa M.
      Fractures of the C-2 vertebral body.
      ]. IIIA injuries were also commonly incorrectly classified by surgeons affiliated with low-level trauma centers, although this is likely because they were frequently classified as IIIB injuries.
      The accuracy and reliability of applying the AO Spine Upper Cervical Injury Classification System to atlas injuries was lower than occipital condyle/craniovertebral junction or C2 vertebrae injuries, even though the injury incidence of atlas fractures is increasing [
      • Lyons J.G.
      • Mian H.M.
      Epidemiology of atlas fractures in the United States: a 20-year analysis.
      ]. Although this may be partially due to the inherent complexity of atlas fractures given that the reliability of the Gehweiler classification (a descriptive classification for atlas injuries) was recently demonstrated as moderate (k = 0.50), our study suggests that minimal differences in classification accuracy or reliability exist for IIA or IIC injuries [
      • Laubach M.
      • Pishnamaz M.
      • Scholz M.
      • Spiegl U.
      • Sellei R.M.
      • Herren C.
      • Hildebrand F.
      • Kobbe P.
      Interobserver reliability of the Gehweiler classification and treatment strategies of isolated atlas fractures: an internet-based multicenter survey among spine surgeons.
      ]. Instead type IIB injuries, which can be treated with divergent management pathways (conservative versus operative), may be uncommonly encountered and require additional trauma experience or tailored education to optimize classification accuracy [
      • Dickman C.A.
      • Greene K.A.
      • Sonntag V.K.
      Injuries involving the transverse atlantal ligament: classification and treatment guidelines based upon experience with 39 injuries.
      ]. Additionally, if surgeons choose to supplement CT scans with magnetic resonance imaging (MRI), additional information may be obtained which could improve the classification accuracy [
      • Schoenfeld A.J.
      • Bono C.M.
      • McGuire K.J.
      • Warholic N.
      • Harris M.B.
      Computed tomography alone versus computed tomography and magnetic resonance imaging in the identification of occult injuries to the cervical spine: a meta-analysis.
      ]..
      Although the goal of AO Spine is to improve accessibility of the injury classification to an international group of spine surgeons, some of which have limited access to MRI, an MRI may supplement the CT scan and improve injury diagnosis and potentially classification accuracy [
      • Schoenfeld A.J.
      • Bono C.M.
      • McGuire K.J.
      • Warholic N.
      • Harris M.B.
      Computed tomography alone versus computed tomography and magnetic resonance imaging in the identification of occult injuries to the cervical spine: a meta-analysis.
      ,
      • Ogbole G.I.
      • Adeyomoye A.O.
      • Badu-Peprah A.
      • Mensah Y.
      • Nzeh D.A.
      Survey of magnetic resonance imaging availability in West Africa.
      ,
      • Volpi G.
      Radiography of diagnostic imaging in Latin America.
      ]. In the event that MRI scans become more accessible to low income global regions, MRI scans may be a useful supplement to CT scans, without requiring any alteration to the AO Spine Upper Cervical Injury Classification System since the current classification accounts for ligamentous disruption. However, in order to standardize the assessment across all international regions, MRI scans were not provided during this validation assessment.
      Some additional limitations were present that merit discussion. First, study participants were solely comprised of AO Spine members who may have a better understanding of AO Spine principles than the non-AO spine surgeons. Second, type IB injuries were not included in the assessment because this injury type was not available in the AO Spine imaging database. Additional differences in classification accuracy and reliability based on surgeons’ work settings and trauma center affiliation may exist for this injury type. Finally, the training session was limited to English so differences in fluency may have altered the validation accuracy and reliability.

      Conclusion

      Type B injuries may be more complex injury patterns that are more often correctly categorized when evaluated by academic surgeons, hospital-employed surgeons, and surgeons associated with higher-level trauma centers (I or II/III). Additional education on how to correctly classify complex injury types (Type B and C injuries) may improve the overall classification accuracy, reliability, and generalizability of the AO Spine Upper Cervical Injury Classification System, especially for surgeons who do not frequently encounter these injury types.
      AO Spine Upper Cervical Injury Classification International Members
      Tabled 1
      Dewan AsifSachin BorkarJoseph Bakar
      Slavisa ZagoracWelege WimalachandraOleksandr Garashchuk
      Francisco Verdu-LopezGiorgio LofresePragnesh Bhatt
      Oke ObadaserayeAxel PartenheimerMarion Riehle
      Eugen Cesar PopescuChristian KonradsNur Aida Faruk Senan
      Adetunji ToluseNuno NevesTakahiro Sunami
      Bart KuipersJayakumar SubbiahAnas Dyab
      Peter LoughenburyDerek CawleyRené Schmidt
      Loya KumarFarhan KarimZacharia Silk
      Michele ParolinHisco RobijnAl Kalbani
      Ricky RasschaertChristian MüllerMarc Nieuwenhuijse
      Selim AyhanShay MenachemSarvdeep Dhatt
      Nasser KhanSubramaniam HaribabuMoses Kimani
      Olger AlarconNnaemeka AlorDinesh Iyer
      Michal ZigaKonstantinos GousiasGisela Murray
      Michel TriffauxSebastian HartmannSung-Joo Yuh
      Siegmund LangKyaw LinnCharanjit Singh Dhillon
      Waeel HamoudaStefano CarnesecchiVishal Kumar
      Lady Lozano CariGyanendra ShahFuruya Takeo
      Federico SartorFernando GonzalezHitesh Dabasia
      Wongthawat LiawrungrueangLincoln LiuYounes El Moudni
      Ratko YurakHéctor AceitunoMadhivanan Karthigeyan
      Andreas DemetriadesSathish MuthuMatti Scholz
      Wael AlsammakKomal ChandrachariKhoh Phaik Shan
      Sokol TrunguJoost DejaegherOmar Marroquin
      Moisa Horatiu AlexandruMáximo-Alberto Diez-UlloaPaulo Pereira
      Claudio BernucciChristian HohausMiltiadis Georgiopoulos
      Annika HeuerAhmed Arieff AtanMark Murerwa
      Richard LindtnerManjul TripathiHuynh Hieu Kim
      Ahmed HassanNorah FosterAmanda O'Halloran
      Koroush KabirMario GanauDaniel Cruz
      Amin HenineJeronimo MilanoAbeid Mbarak
      Arnaldo SousaSatyashiva MunjalMahmoud Alkharsawi
      Muhammad MirzaParmenion TsitsopoulosFon-Yih Tsuang
      Oliver RisenbeckArun-Kumar ViswanadhaSamer Samy
      David OroscoGerardo Zambito-BrondoNauman Chaudhry
      Luis MarquezJacob LepardJuan Muñoz
      Stipe CorlukaSoh ReubenAriel Kaen
      Nishanth AmparSebastien BigdonDamián Caba
      Francisco De MirandaLoren LayIvan Marintschev
      Mohammed ImranSandeep MohindraNaga Raju Reddycherla
      Pedro BazánAbduljabbar AlhammoudIain Feeley
      Konstantinos MargetisAlexander DurstAshok Kumar Jani
      Rian Souza VieiraFelipe SantosJoshua Karlin
      Nicola MontemurroSergey MlyavykhBrian Sonkwe
      Darko PerovicJuan LouridoAlessandro Ramieri
      Eduardo LaosUri HadesbergAndrei-Stefan Iencean
      Pedro NevesEduardo BertoliniNaresh Kumar
      Philippe BancelBishnu SharmaJohn Koerner
      Eloy Rusafa NetoNima OstadrahimiOlga Morillo
      Kumar rakeshAndreas MorakisAmauri Godinho
      P KeerthivasanRichard MengerLouis Carius
      Rajesh Bahadur LakheyEhab ShibanVishal Borse
      Elizabeth BoudreauGabriel LacerdaPaterakis Konstantinos
      Mubder Mohammed SaeedToivo HasheelaSusana Núñez Pereira
      Jay ReidlerNimrod RahamimovMikolaj Zimny
      Devi Prakash TokalaHossein ElgafyKetan Badani
      Bing Wui NgCesar Sosa JuarezThomas Repantis
      Ignacio Fernández-BancesJohn KleimeyerNicolas Lauper
      Luis María Romero-MuñozAyodeji YusufZdenek Klez
      John AfolayanJoost RutgesAlon Grundshtein
      Rafal ZaluskiStavros I StavridisTakeshi Aoyama
      Petr VachataWiktor UrbanskiMartin Tejeda
      Luis MuñizSusan KaranjaAntonio Martín-Benlloch
      Heiller TorresChee-Huan PanLuis Duchén
      Yuki FujiokaMeric EnercanMauro Pluderi
      Catalin MajerVijay Kamath

      Declaration of conflict of interest

      This study was organized and funded by AO Spine through the AO Spine Knowledge Forum Trauma, a focused group of international Trauma experts. AO Spine is a clinical division of the AO Foundation, which is an independent medically guided not-for-profit organization. Study support was provided directly through the AO Spine Research Department.

      Acknowledgements

      The authors of the manuscript would like to thank Olesja Hazenbiller for her assistance in developing the methodology and providing support during the validation. We would also like to thank Hans Bauer, senior biostatistician at Staburo GmbH for his assistance with the statistical analysis. This study was organized and funded by AO Spine through the AO Spine Knowledge Forum Trauma, a focused group of international Trauma experts. AO Spine is a clinical division of the AO Foundation, which is an independent medically guided not-for-profit organization. Study support was provided directly through AO Network Clinical Research.

      Appendix. Supplementary materials

      References

        • Spivak J.M.
        • Weiss M.A.
        • Cotler J.M.
        • Call M.
        Cervical spine injuries in patients 65 and older.
        Spine (Phila Pa. 1976; 19 (1994 Oct 15): 2302-2306https://doi.org/10.1097/00007632-199410150-00009
        • Lyons J.G.
        • Mian H.M.
        Epidemiology of atlas fractures in the United States: a 20-year analysis.
        J Craniovertebr Junct Spine. 2022; 13 (Jan-MarEpub 2022 Mar 9): 85-93https://doi.org/10.4103/jcvjs.jcvjs_164_21
        • Malik S.A.
        • Murphy M.
        • Connolly P.
        • O'Byrne J
        Evaluation of morbidity, mortality and outcome following cervical spine injuries in elderly patients.
        Eur Spine J. 2008; 17: 585-591https://doi.org/10.1007/s00586-008-0603-3
        • Barmparas G.
        • Cooper Z.
        • Haider A.H.
        • Havens J.M.
        • Askari R.
        • Salim A.
        The elderly patient with spinal injury: treat or transfer?.
        J Surg Res. 2016; 202 (May 1Epub 2015 Dec 30): 58-65https://doi.org/10.1016/j.jss.2015.12.032
        • Wang H.
        • Ou L.
        • Zhou Y.
        • Li C.
        • Liu J.
        • Chen Y.
        • Yu H.
        • Wang Q.
        • Zhao Y.
        • Han J.
        • Xiang L.
        Traumatic upper cervical spinal fractures in teaching hospitals of China over 13 years: a retrospective observational study.
        Medicine (Baltimore). 2016; 95 (Oct): e5205https://doi.org/10.1097/MD.0000000000005205
        • Benzel E.C.
        • Hart B.L.
        • Ball P.A.
        • Baldwin N.G.
        • Orrison W.W.
        • Espinosa M.
        Fractures of the C-2 vertebral body.
        J Neurosurg. 1994; 81 (Aug): 206-212https://doi.org/10.3171/jns.1994.81.2.0206
        • Labler L.
        • Eid K.
        • Platz A.
        • Trentz O.
        • Kossmann T.
        Atlanto-occipital dislocation: four case reports of survival in adults and review of the literature.
        Eur Spine J. 2004; 13: 172-180https://doi.org/10.1007/s00586-003-0653-5
        • Vaccaro A.R.
        • Karamian B.A.
        • Levy H.A.
        • et al.
        Update on upper cervical injury classifications: the new AO upper cervical spine classification system.
        Clin Spine Surg. 2021 Jul 7; (Epub ahead of print)https://doi.org/10.1097/BSD.0000000000001215
        • Lambrechts M.J.
        • Schroeder G.D.
        • Karamian B.A.
        • Canseco J.A.
        • Oner C.
        • Vialle E.
        • Rajasekaran S.
        • Hazenbiller O.
        • Dvorak M.R.
        • Benneker L.M.
        • Kandziora F.
        • Schnake K.
        • Kepler C.K.
        • Vaccaro A.R.
        AO spine subaxial classification group members. Development of online technique for international validation of the AO spine subaxial injury classification system.
        Global Spine J. 2022; (Apr 2721925682221098967)https://doi.org/10.1177/21925682221098967
      1. (Intrarater reliability)Gwet KL D'Agostino RB Sullivan L Massaro J Wiley encyclopedia of clinical trials. John Wiley & Sons, Inc, Hoboken, NJ2008: 473-485
        • Landis J.R.
        • Koch G.G.
        The measurement of observer agreement for categorical data.
        Biometrics. 1977; 33 (Mar): 159-174
        • Laubach M.
        • Pishnamaz M.
        • Scholz M.
        • Spiegl U.
        • Sellei R.M.
        • Herren C.
        • Hildebrand F.
        • Kobbe P.
        Interobserver reliability of the Gehweiler classification and treatment strategies of isolated atlas fractures: an internet-based multicenter survey among spine surgeons.
        Eur J Trauma Emerg Surg. 2020; (Sep 12)https://doi.org/10.1007/s00068-020-01494-y
        • Dickman C.A.
        • Greene K.A.
        • Sonntag V.K.
        Injuries involving the transverse atlantal ligament: classification and treatment guidelines based upon experience with 39 injuries.
        Neurosurgery. 1996; 38 (JanPMID: 8747950): 44-50https://doi.org/10.1097/00006123-199601000-00012
        • Schoenfeld A.J.
        • Bono C.M.
        • McGuire K.J.
        • Warholic N.
        • Harris M.B.
        Computed tomography alone versus computed tomography and magnetic resonance imaging in the identification of occult injuries to the cervical spine: a meta-analysis.
        J Trauma. 2010; 68 (Jan109-13; discussionPMID: 20065765): 113-114https://doi.org/10.1097/TA.0b013e3181c0b67a
        • Ogbole G.I.
        • Adeyomoye A.O.
        • Badu-Peprah A.
        • Mensah Y.
        • Nzeh D.A.
        Survey of magnetic resonance imaging availability in West Africa.
        Pan Afr Med J. 2018; 30 (Published 2018 Jul 31): 240https://doi.org/10.11604/pamj.2018.30.240.14000
        • Volpi G.
        Radiography of diagnostic imaging in Latin America.
        Nucl Med Biomed Imaging. 2016; 1: 10-12