PEDIATRICS/ORIGINAL RESEARCH
Pediatric Cervical Spine Injury Evaluation After Blunt Trauma: A Clinical Decision Analysis Megan Hannon, MD*; Rebekah Mannix, MD, MPH; Kate Dorney, MD; David Mooney, MD; Kara Hennelly, MD *Corresponding Author. E-mail: [email protected], Twitter: @mhannon19.
Study objective: Although many adult algorithms for evaluating cervical spine injury use computed tomography (CT) as the initial screening modality, this may not be appropriate in low-risk children, considering radiation risks. We determine the optimal initial evaluation strategy for cervical spine injury in pediatric blunt trauma. Methods: We constructed a decision analysis tree for a hypothetical population of patients younger than 19 years with blunt trauma, using 3 strategies: clinical stratification, screening radiographs followed by focused CT if the radiograph result was positive, and CT. For the model inputs, we used the current literature to determine the probabilities of cervical spine injury and estimate the long-term risks of malignancy after CT, as well as test characteristics of radiographic imaging. We used published utilities and conducted 1- and 2-way sensitivity analyses to determine the optimal strategy for evaluation of pediatric cervical spine injury. Results: In our model of a population with blunt trauma, the expected value of a clinical stratification strategy was the highest of the 3 strategies, making it the overall preferred management. One-way sensitivity analysis of several contributing factors revealed that the only independent factor that altered the dominant strategy was the sensitivity of clinical clearance criteria, lowering the threshold at which screening-radiograph strategy is optimal. Within the patient population considered as having non-negligible risk by clinical stratification and thus requiring imaging, the preferred imaging modality was screening radiograph/focused CT. The probability of cervical spine injury above which CT became the preferred strategy was 24.9%. Conclusion: The model highlights that clinical clearance and screening radiographs in a hypothetical trauma pediatric population are preferred strategies, whereas CT scanning is rarely the initial optimal evaluation. [Ann Emerg Med. 2015;65:239-247.] Please see page 240 for the Editor’s Capsule Summary of this article. A feedback survey is available with each research article published on the Web at www.annemergmed.com. A podcast for this article is available at www.annemergmed.com. 0196-0644/$-see front matter Copyright © 2014 by the American College of Emergency Physicians. http://dx.doi.org/10.1016/j.annemergmed.2014.09.002
INTRODUCTION Background Every year, there are approximately 10 million children who experience traumatic injuries, many of whom require specific cervical spine evaluation.1 Although pediatric blunt trauma is common, cervical spine injuries occur in only 1% to 2% of all pediatric patients with blunt trauma.2-4 Despite the rarity of injury, the cervical spine trauma evaluation presents a diagnostic dilemma because the risk of a missed injury has potential for devastating neurologic sequelae.5,6 Initial management strategies in the evaluation of the cervical spine of children with blunt trauma range from clinical “clearance” (ie, use of clinical screening criteria to remove cervical spine precautions) to advanced imaging. Clinical clearance of the asymptomatic pediatric patient is often the first-line
Volume 65, no. 3 : March 2015
management strategy despite only limited data supporting this approach. Some studies have attempted to apply adult approaches to the pediatric population, such as the National Emergency X-radiography Utilization Study (NEXUS).2 The NEXUS study was hampered by a small sample size, especially in the youngest children. The Pediatric Emergency Care Applied Research Network (PECARN) conducted a large, multicenter, case-control study to define clinical clearance criteria specific to children; however, this study has not been validated prospectively.7 Thus, clinicians remain uncertain about which clinical screening criteria, if any, to use when evaluating children for potential cervical spine injury after blunt trauma.8,9 For the symptomatic patient, it is generally agreed that radiographic imaging is a necessary adjunct, but there are no consistent recommendations on indications or modality. Adult
Annals of Emergency Medicine 239
Hannon et al
Pediatric Cervical Spine Injury Evaluation
Editor’s Capsule Summary
What is already known on this topic Computed tomography (CT) is more sensitive than radiography in identifying cervical spine injuries, but carries risk of radiation-induced malignancy. What question this study addressed Clinical decision analysis was used to determine the optimal cervical spine evaluation strategy for maximizing quality of life for children with blunt trauma, balancing risk of missed cervical spine injury with risk of radiation-induced malignancy. What this study adds to our knowledge In a hypothetical cohort of pediatric patients with blunt trauma, using literature-derived probabilities and utilities, clinical clearance and screening plain radiography were preferred over a CT-all strategy. How this is relevant to clinical practice Unlike for adults, for pediatric trauma patients cervical spine plain radiography screening may be preferred over a CT-all strategy.
trauma protocols recommend computed tomography (CT) for symptomatic patients because it is considered both a highly sensitive test and cost-effective approach.10-12 Given lower cervical spine injury rates in and greater sensitivity to radiation in children versus adults, it is unclear which imaging strategy should be applied in pediatrics. In contrast to adult protocols, the American Association of Neurological Surgeons consensus pediatric guidelines for symptomatic patients recommend screening radiographs followed by focused CT if abnormalities are present on the radiograph, though the guidelines also acknowledge the lack of evidence to support the recommendation.13 Given a paucity of definitive evidence in relation to a pediatric clearance strategy, as well as the fact that the majority of children are treated in nonpediatric emergency departments, in which advance imaging is more standard, many children likely undergo CT as the initial screening modality.14-17 The absence of validated clinical clearance protocols and controversy surrounding appropriate imaging modalities lead to wide practice variation. Importance Given the practice variation and the lack of definitive evidence for an approach to screening children for cervical spine injuries, understanding the risks and benefits of different initial screening strategies could guide management decisions. The rarity of cervical spine injuries makes meaningful prospective data 240 Annals of Emergency Medicine
collection challenging. Decision analysis is an objective way to compare the various approaches. Goals of This Investigation We explored several cervical spine evaluation strategies, including clinical stratification and both CT and screening radiographs, with particular attention to the risk tradeoffs between the importance of identifying a significant cervical spine injury and the risk of radiation-induced malignancy. We determined the optimal management strategy for a hypothetical pediatric population with blunt trauma by identifying the risk threshold at which CT imaging becomes the preferred strategy, considering health-related quality of life and risk of radiation.
MATERIALS AND METHODS This study was based on literature review and was exempt from review by our institutional review board. We constructed a decision analysis tree for a hypothetical population of patients younger than 19 years and with blunt trauma, using 3 strategies: clinical stratification, screening radiographs followed by focused CT if the radiograph result was positive, and CT (Figure 1, depicted as square decision nodes). As shown in the decision tree, after the initial choice, physicians observe the outcome of chance events (circles) and ultimately end at one of the possible terminal node (triangles). Utilities are listed at each terminal node and are a numeric measure representative of the quality of life of an individual with a particular state of health (ie, represents an individual with a malignancy or cervical spine injury). This decision tree determines the optimal patientcentered approach for cervical spine evaluation in a population with blunt trauma, with the goal of reducing harm to the individual. In decision analysis, the measure of a given decision is represented as an expected value. Expected values are determined by first identifying the potential value of a given outcome (positive or negative). The value unit reflects the aim of the study, eg, financial or quality of life. In this case, the value was defined as a utility for a given health state, which will be discussed below. Next, the probability that each result will occur from a given course of action is determined, usually from previous literature. One then multiplies the value by the probability of an outcome to calculate an expected value of a particular result. As an example, in our decision tree, the choice to perform a CT results in 2 possible outcomes (branches), negative or positive cervical spine injury result. After this, there are 2 additional possible outcomes or branches: having a malignancy related to radiation or not. Each outcome has a utility (displayed after the terminal node in the decision tree; Figure 1). Using a probability of malignancy of 0.004 (and 1–probability of malignancy¼0.996 as the probability of remaining malignancy free), expected values can be calculated. If the patient had both a cervical spine injury and malignancy, the expected value calculated is 0.002. If the Volume 65, no. 3 : March 2015
Hannon et al
Pediatric Cervical Spine Injury Evaluation
Figure 1. Decision analysis tree using 3 strategies for a hypothetical population of patients younger than 19 years and with blunt trauma: clinical stratification, screening radiographs followed by focused CT if the radiograph result was positive, and CT, each depicted as square decision nodes. After the initial choice, an outcome is observed of chance events (circles). Each branch ends at a possible terminal node (triangles). Utilities are listed at each terminal node. CSI, Cervical spine injury.
patient has cervical spine injury but no malignancy, the expected value calculated is 0.6972. We can add these expected values to represent that of the entire CT positive branch as 0.6992. Calculating in a similar fashion, the CT negative-result branch expected value is 0.9984. To obtain the expected value of an entire extended branch that includes several nodes, a process named rolling back is completed. Rolling back is multiple rounds of the aforementioned calculations using probabilities and branch values sequentially. The expected values reported in the results for each strategy are the conglomerate expected utility for the entire branch after rolling back. To examine the effect of certain factors such as radiation risk and varying probability of cervical spine injury on presentation, as well as variability of both clinical and radiographic performance characteristics, we performed sensitivity analyses ranging through the 95% confidence intervals (CIs) extracted from available data around each factor. We constructed the decision analysis model using TreeAge Pro 2014 (TreeAge Software Inc, Williamstown, MA). As described above, we made a decision tree with 3 clinically common initial management plans for our hypothetical population. All sensitivity analyses and threshold acquisition were performed with the TreeAge Pro software. Volume 65, no. 3 : March 2015
For our baseline model inputs, we used standard definitions and data from a combination of recently published studies (Table 1). We defined cervical spine injury as any radiographic finding, excluding anatomic variants independent of neurologic outcome. We did not differentiate between injuries resulting in neurologic deficits or requiring intervention from those with clinically insignificant findings. For our model, we used probabilities obtained from the largest available studies, including NEXUS and PECARN. To estimate the probability of cervical spine injury, we used the NEXUS study. Though the NEXUS entrance criteria required initial imaging, thus likely excluding the lowest-risk patients, we chose this data set because it is the largest pediatric data set and would likely yield the most conservative estimates. In NEXUS, subjects were considered to have non-negligible cervical spine injury risk if any of the following characteristics were present: midline tenderness, altered mental status, intoxication, neurologic deficits, and distracting injury. There were a total of 3,065 pediatric trauma patients with cervical spine imaging, of whom 0.98% (30 subjects) had cervical spine injuries. Overall, there were 2,462 NEXUS-positive patients. All of the cervical spine injuries met at least 1 NEXUS criterion. The clearance criterion was then observed to have 100% sensitivity Annals of Emergency Medicine 241
Hannon et al
Pediatric Cervical Spine Injury Evaluation Table 1. Summary of decision tree inputs.* Probabilities p(CSImain) p(CSINEXUSpos) p(CSINEXUSneg) Radiographic specificity Radiographic sensitivity p(malignancy) Utilities Healthy CSI/radiographþ CSI/radiographþ/malignancy CSI/radiograph– No CSI/radiographþ/malignancy No CSI/radiograph– CSI/CTþ CSI/CTþ/malignancy
Baseline
Source 2
0.98% 1.22% 0.001% 95% 90% 0.4%
Viccellio et al Viccellio et al2 Viccellio et al2 Streitwieser et al,22 Nguyen and Clark,23 Tins and Cassar-Pullicino24 Nigrovic et al18 Berrington de Gonzalez et al26
1 0.7 0.5 0.6499 0.6 0.9999 0.7 0.5
Kaplan Kaplan Kaplan Kaplan Kaplan Kaplan Kaplan
et et et et et et et
al,29 al,29 al,29 al,29 al,29 al,29 al,29
Oladeji Oladeji Oladeji Oladeji Oladeji Oladeji Oladeji
et et et et et et et
al,27 al,27 al,27 al,27 al,27 al,27 al,27
Ravichandran Ravichandran Ravichandran Ravichandran Ravichandran Ravichandran Ravichandran
and and and and and and and
Silver28 Silver,28 Stein et al31 Silver28 Silver,28 Stein et al31 Silver28 Silver28 Silver,28 Stein et al31
p(CSImain), Probability of CSI in the main trauma population; p(CSINEXUSneg), probability of cervical spine injury in the negligible-risk group; p(malignancy), probability of malignancy. *Probabilities of each outcome were obtained from the best available literature that described the particular outcomes or testing characteristics. Utilities representative of health states were obtained from pediatric survey literature and supported by adult data.
(95% CI 87.8% to 100%). In our model, we considered NEXUS positivity to equate to a symptomatic or non-negligible cervical spine injury risk population. Thus, we used 0.98% as the probability of cervical spine injury in the overall trauma population and 1.22% for that in the NEXUS-positive group. To define the characteristics of screening radiographs, we used recently published PECARN data that evaluated the test characteristics of cervical spine radiographs.18 For this retrospective study, the entrance criteria were a pediatric trauma patient with related bony or ligamentous injury and who had cervical spine radiographic imaging. Cervical radiographic sensitivity was 90% (95% CI 85% to 94%), which we used as our baseline model input. PECARN reported that 33% had 2-view radiographs, whereas 67% had greater than or equal to 3-view radiographs. The types of radiographs in the study reflected common practice in the pediatric centers enrolled in PECARN. In general, the literature and the American Association of Neurological Surgeons guidelines support 2-view radiographs for patients younger than 8 years, with an additional odontoid view in those older than 8 years; thus, that is how the model’s radiograph strategy was defined. Additional smaller studies had somewhat lower sensitivities reported, and these data were used in sensitivity analyses.19,20 Specificity obtained from various trauma literature supported a wide range, from 70% to 95%.21-24 We estimated probability of malignancy related to CT radiation exposure from the most recently published data on projected lifetime cancer risk for pediatric patients.25,26 For example, it is projected that malignancy related to radiation exposure from a cervical spine CT is 70/10,000 for a 1-year-old girl and 10/10,000 for a 15-year-old boy. Thus, we chose to average the risks at a conglomerate baseline rate of 40/10,000 for a baseline rate and range through the aforementioned low and high rates with sensitivity analysis. When possible, to be compliant with our objective of creating a tree to minimize harm to the individual, we chose to 242 Annals of Emergency Medicine
incorporate the most conservative data to avoid any missed diagnosis and minimize poor outcomes. In our model, we assigned values (or utilities) to the health outcomes, using a standard scale from 0 (death) to 1 (baseline health, assumed as perfect health for this hypothetical population). Published utility scales were used that combine several major domains: physical, emotional, social, cognitive, and school capabilities. Many are adult survey based, but when pediatric-specific utilities were available, we used them as reviewed below. We assigned cervical spine injury utilities according to publications on the quality of life of pediatric patients with spinal cord injuries (Table 1).27,28 As noted above, we erred conservatively and considered any radiographic cervical spine injury to be equivalent to a patient with severe spinal cord injury, with a baseline utility of 0.7. With delayed diagnosis, we assigned an additional disutility according to the assumption that early detection results in better outcomes. This translated to a utility of 0.65 if a cervical spine injury was missed by a radiograph and 0.6 if a patient had no imaging. These utilities are comparable to those for adult cervical spine injury.29 We estimated the utility of a malignancy to be 0.6, as used in previous pediatric studies using a Markov model for head CT.30,31 If a patient had cervical spine injury and malignancy, then he or she mathematically ought to have a utility of 0.4, but given the later onset of malignancy (likely 10 or more years), we used a higher value of 0.5, which errs on the side of not missing the immediate diagnosis of a spine injury. Sensitivity Analyses Baseline model inputs (Table 1) were used in the basic decision tree. From the literature, we obtained data on the CIs around each model input and performed sensitivity analyses to measure the effect of each contributing factor, including the probability of cervical spine injury in the main trauma Volume 65, no. 3 : March 2015
Hannon et al
Pediatric Cervical Spine Injury Evaluation
Table 2. Results of 1-way sensitivity analysis for the overall trauma population. Model Factor p(CSINEXUSneg) Radiograph sensitivity Radiograph specificity p(malignancy) CSI utility
Baseline
Range
0.001%
0%–0.6%
90% 95% 0.4% 0.6
60%–100% 60%–100% 0.1%–0.7% 0.4–0.9
Source
Effect
Nigrovic et al,18 Buhs et al,20 Mower et al19 Streitwieser et al,22 Nguyen and Clark,23 Tins and Cassar-Pullicino24 Berrington de Gonzalez et al26 Kaplan et al,29 Oladeji et al,27 Ravichandran and Silver28
At 0.03%, the preferred strategy became screening radiograph No change No change No change No change
2
Viccellio et al
population, the probability of CT radiation–related malignancy (probability of malignancy), sensitivity of radiograph imaging, and the utility of cervical spine injury. Additionally, the sensitivity of a clinical clearance rule was evaluated by ranging the probability of cervical spine injury in the negligible-risk group (or NEXUS negative). This variable probability of cervical spine injury in the negligible-risk group was calculated from the CI reported in the NEXUS study. All of the sensitivity analysis parameters were obtained from the primary literature described above and are summarized in Table 2. A separate analysis was performed on the non-negligible risk population branch to understand how changes in probability of malignancy related to CT radiation might affect imaging modality choices. We performed 1-way sensitivity analysis on probability of cervical spine injury in the NEXUSpositive group and 2-way sensitivity analysis ranging both the probability of cervical spine injury and probability of malignancy.
RESULTS Overall, the expected values were high for each major decision node, according to the overall baseline health status of the pediatric population and the low rate of cervical spine injury. The greatest variation was based on the outcome (cervical spine injury or malignancy) (Figure 1). The expected value of each strategy was as follows: clinical stratification yielded 0.996912, screening radiograph/focused CT yielded 0.996888, and CT-all yielded 0.995468. The preferred management plan in this analysis, according to the expected probability of cervical spine injury of 0.98%, was clinical stratification. One-way sensitivity analysis determined the threshold at which the preferred strategy changes with independent alterations of model inputs (Table 2). The probability of cervical spine injury in the NEXUS-positive group was the only independent variable that met a threshold at which an alternate strategy was preferred. At a probability of cervical spine injury in the NEXUS-positive group of 0.03%, which represents a
Figure 2. Once the decision to image has been made, the choice of imaging modality is controversial. To represent the decision, the non-negligible CSI risk group was analyzed. The screening radiograph strategy was superior at a probability of CSI of 1.22% in this subgroup. To understand the effect of the p(CSINEXUSpos), 1-way sensitivity analysis was performed and identified 24.9% to be the p(CSINEXUSpos) at which above CT-all would be preferred. p(CSINEXUSpos), probability of CSI in the NEXUS-positive group. Volume 65, no. 3 : March 2015
Annals of Emergency Medicine 243
Hannon et al
Pediatric Cervical Spine Injury Evaluation
Figure 3. In the non-negligible CSI risk group, inputs with potential significant variability based on circumstance included both the probability of CSI (p[CSINEXUSpos]) and the radiation risk (probability of malignancy). A 2-way sensitivity analysis represented the effect of varying both the p(CSINEXUSpos) and the radiation risk probability of malignancy. As the probability of malignancy decreased or in scenarios suggesting the probability of CSI to be much higher than baseline tree inputs, CT strategy dominated.
sensitivity of NEXUS criteria of 99%, a screening radiograph/ focused CT becomes a preferred management plan. The non-negligible risk population of the clinical stratification branch of the tree had 2 branches of its own, which were analyzed separately: screening radiograph/focused CT and CTall. The expected value of each strategy was as follows: screening radiograph/focused CT yielded 0.996165 and CT-all yielded 0.994749. Thus, the preferred management plan was screening radiograph/focused CT, according to the expected probability of cervical spine injury in the NEXUS-positive group of 1.22%. The probability of cervical spine injury in the NEXUS-positive group above which CT-all strategy became the preferred strategy was 24.9% (Figure 2). A 2-way sensitivity analysis represented the effect of varying both the probability of cervical spine injury in the NEXUS-positive group and the radiation risk probability of malignancy (Figure 3). In the decision tree, we used a probability of CT radiation exposure–related malignancy of 0.4% (40 in 10,000 scans). As expected, reduction in proposed malignancy risk from CT resulted in a shift toward CT use. For example, with a risk of malignancy of 0.1%, a threshold for CT would be reduced to a probability of cervical spine injury in the NEXUS-positive group of 6.5%. As the probability of cervical spine injury in the NEXUS-positive group increased, CT became acceptable at a higher probability of malignancy.
scanner type. For example, the malignancy-related radiation risk for a 15-year-old boy is estimated to be as low as 10/10,000 scans, whereas for a 1-year-old girl it is as high as 70/10,000 scans. However, sensitivity analyses allowed us to vary radiation risks, accounting for both the inaccuracies of estimates and individualization of these variations. Second, the model is limited by the decision to include only the factors that contribute directly to patient health outcomes. Obviously, the scope of any health care issue is complicated by both societal and financial realities. For example, the model identifies screening radiograph/focused CT as the predominant strategy in reduction of patient harm. Other factors that were not included, such as cost, risk of screening a population, and falsepositives implications, may alter this result, given that the expected values are so close. Third, the literature used to generate probabilities is heterogeneous, with variable entrance criteria (ie, patients who were imaged versus all patients versus only those with injuries). No study likely accurately reflects the population who requires cervical spine imaging. However, sensitivity analysis provides insight on how variance of the inputs affects optimal imaging strategy.
DISCUSSION LIMITATIONS There are several limitations to this study. First, radiation risk is variable according to institution, age, sex, body size, and 244 Annals of Emergency Medicine
Through decision analysis, we identified the optimal management strategies for cervical spine evaluation in 2 populations: a main population, inclusive of all pediatric patients with blunt trauma and requiring cervical spine evaluation, and a Volume 65, no. 3 : March 2015
Hannon et al higher-risk population (or non-negligible risk), as defined by NEXUS criteria positivity. The decision tree was created with the best available data. It is representative of the decisionmaking process that occurs in such a clinical circumstance but is not a clinical decision rule and thus may not reflect clinical workflow perfectly. Balancing radiation risk from CT- and health-related outcomes with respect to cervical spine injury, as well as the potential ramifications of a delayed diagnosis, we found that clinical stratification is the preferred strategy, followed by radiograph/focused CT. This is important because the model suggests less aggressive imaging approaches for the majority of patients. The results suggest that on the population level, the risk of CT radiation–related malignancy outweighs the small number of cervical spine injuries identified by advanced imaging in most circumstances. Certainly some factors, such as neurologic deficits or worsening clinical status, would increase the pretest probability and one would opt for CT immediately; however, most individual criteria, such as midline tenderness or mechanism of injury, would not increase the pretest probability of cervical spine injury to the threshold value of 24.9%, above which the CT-all strategy is optimal. Although not developed for this purpose, several studies, including NEXUS, provide data about the rates of cervical spine injury for each clearance criterion.2 This may be a useful way to help provide a basis for stratifying patients. For example, from the NEXUS data, the probability of cervical spine injury if a patient has midline tenderness is 1.75%. If a patient has a neurologic deficit, the probability of cervical spine injury increases to 4.44%. Neither meets the CT threshold as an individual factor. The data on cumulative probability of cervical spine injury with multiple characteristics are not reported in this data set. However, this information provides a sense of the relative risk of individual risk factors for a cervical spine injury. Data are now available on the cumulative ionizing radiation effects on radiosensitive tissues, particularly the thyroid.32,33 Often in a trauma evaluation cervical spine CT is part of a larger evaluation in which both head and chest imaging are considered and thus the cumulative radiation effects must be considered.34 With the above data and reports of cervical spine CT rates increasing in some areas, despite a simultaneous downward trend in other types of scans, clinicians must continue to address the challenge to reduce unnecessary scans.15 This model suggests that unless the probability of cervical spine injury is high, clinical clearance or screening radiographs should predominate the current management strategies when radiation risk is considered. Over time, radiation risk may change and a lower threshold may be considered for CT, as can be appreciated in the 2-way sensitivity analysis (Figure 3). Of course, there needs to be periodic reassessment as imaging technology improves, CT radiation exposure reduces, or alternative imaging options become feasible options (ie, more rapid magnetic resonance imaging that does not require sedation). There are several factors that make this a conservative analysis, consistent with an emphasis on the reduction of individual harm, including using what is probably an overestimate of the probability Volume 65, no. 3 : March 2015
Pediatric Cervical Spine Injury Evaluation of cervical spine injury, as well as equating all cervical spine injury. First, the model was based on estimates from the NEXUS study, a probability of cervical spine injury in the main trauma population of 0.98% and of 1.22% in the NEXUS-positive group (non-negligible risk group). Because the entrance criteria for the NEXUS study required previous cervical spine radiography, this likely overestimates the actual rate of cervical spine injury in the general trauma population who requires cervical spine evaluation. There is no published data set that reliably measures the baseline rate because the “trauma” population requiring cervical spine evaluation is difficult to capture. For a completely asymptomatic, or NEXUSnegative, group, the rate of cervical spine injury is thought to be zero or at least closely approaching zero, according to previous literature.2 We used sensitivity analysis to examine the potential that NEXUS criteria have a lower sensitivity than the chosen baseline rate because it is not well validated in pediatrics. If the sensitivity is as high as 99%, then the screening radiograph/focused CT predominates. CT, however, does not become preferred in the chosen ranges of variables. Second, the utility applied to any cervical spine injury is based on having significant spinal injury and neurologically impairing lifelong disability. Ultimately, this means that any radiographically positive finding was equated to a true neurologically impairing spinal cord injury despite that most patients will have no neurologic sequelae and a low rate of neurosurgical intervention. Furthermore, because prompt diagnosis aids in appropriate treatment and better outcomes, an even lower utility is applied if the diagnosis of cervical spine injury is delayed.35 We recognize that both factors reduce harm in the model but potentially slant the outcomes toward imaging. The model is not applicable to the adult trauma population, which, for many reasons, is quite different. First, largely related to anatomic differences, the rates of cervical spine injury are much higher, ranging up to 20%, and the clinical relevance of such injuries is much higher.10,13 The risk of radiation to the neck and therefore the associated probability of malignancy are much lower.26,36 Also, studies on cervical spine radiographs recognize that sensitivity is much lower in the adult population, as low as 52%,12 making them a much less useful screening test. Finally, often there is greater clinician comfort to complete cervical spine clearance with a cooperative adult, using validated clinical clearance criteria that reliably define a low-risk population. Evaluating a cervical spine for a pediatric patient may be a less common occurrence in nonpediatric trauma centers. All of these factors would alter the analysis greatly if the decision tree were built with adult data as model inputs. An adult decision analysis would identify a much lower CT threshold in both the main and non-negligible risk population. It is important to consider this because most pediatric trauma patients are evaluated in non–Level I general trauma centers, where CT rates are much higher than in Level I pediatric trauma facilities.14 Likely lack of definitive pediatric trauma guidelines for cervical spine evaluation and clinician comfort and familiarity with CTfocused strategies in adult emergency medicine trauma protocols Annals of Emergency Medicine 245
Pediatric Cervical Spine Injury Evaluation contribute to the practice variation among clinicians and facilities. In conclusion, the model highlights that in both the main and non-negligible risk populations, CT scanning is not the optimal initial strategy at the current probabilities of cervical spine injury and radiation risk. The various strategy thresholds support a less CT-focused approach, thus decreasing associated radiation risk. For a patient with a normal neurologic examination result, nontender neck, and otherwise normal mental status, the risk of cervical spine injury is typically not high enough to outweigh the risk of malignancy from a cervical spine CT. Individual risk for a particular patient must be considered and may be much higher and place the patient above the CT threshold from the onset. As with any decision analysis, clinical judgment trumps the universal application of a management strategy, but we provide objective aid to this complicated decision by evaluating the influences of the risk of radiation exposure and the consequences of cervical spine injury to maximize patient health outcome. Supervising editor: Kelly D. Young, MD, MS Author affiliations: From the Division of Emergency Medicine (Hannon, Mannix, Dorney, Hennelly) and the Department of Surgery (Mooney), Boston Children’s Hospital, Boston, MA. Author contributions: MH, RM, and KH conceived the study, designed the decision tree, and performed data analysis. MH, RM, DM, and KH reviewed the tree and inputs and made appropriate improvements. MH drafted the article, and all authors contributed substantially to its revision. MH takes responsibility for the paper as a whole. Funding and support: By Annals policy, all authors are required to disclose any and all commercial, financial, and other relationships in any way related to the subject of this article as per ICMJE conflict of interest guidelines (see www.icmje.org). The authors have stated that no such relationships exist. Publication dates: Received for publication February 27, 2014. Revisions received June 17, 2014; July 30, 2014; and August 18, 2014. Accepted for publication September 2, 2014. Available online October 16, 2014.
REFERENCES 1. Centers for Disease Control and Prevention, National Center for Injury Prevention and Control. Web-based Injury Statistics Query and Reporting System (WISQARS). www.cdc.gov/injury/wisqars. Accessed January 2, 2014. 2. Viccellio P, Simon H, Pressman BD, et al. A prospective multicenter study of cervical spine injury in children. Pediatrics. 2001;108:E20. 3. Platzer P, Jaindl M, Thalhammer G, et al. Cervical spine injuries in pediatric patients. J Trauma. 2007;62:389-396; discussion 94-96. 4. Polk-Williams A, Carr BG, Blinman TA, et al. Cervical spine injury in young children: a National Trauma Data Bank review. J Pediatr Surg. 2008;43:1718-1721. 5. Guidelines for the acute medical management of severe traumatic brain injury in infants, children, and adolescents. J Trauma. 2003;54:S235-310.
246 Annals of Emergency Medicine
Hannon et al 6. Leonard JR, Jaffe DM, Kuppermann N, et al. Cervical spine injury patterns in children. Pediatrics. 2014 Apr 28. [Epub ahead of print]. 7. Leonard JC, Kuppermann N, Olsen C, et al. Factors associated with cervical spine injury in children after blunt trauma. Ann Emerg Med. 2011;58:145-155. 8. Burns EC, Yanchar NL. Using cervical spine clearance guidelines in a pediatric population: a survey of physician practices and opinions. CJEM. 2011;13:1-6. 9. Browne GJ, Lam LT, Barker RA. The usefulness of a modified adult protocol for the clearance of paediatric cervical spine injury in the emergency department. Emerg Med (Fremantle). 2003;15:133-142. 10. Como JJ, Diaz JJ, Dunham CM, et al. Practice management guidelines for identification of cervical spine injuries following trauma: update from the Eastern Association for the Surgery of Trauma Practice Management Guidelines Committee. J Trauma. 2009;67:651-659. 11. Blackmore CC, Ramsey SD, Mann FA, et al. Cervical spine screening with CT in trauma patients: a cost-effectiveness analysis. Radiology. 1999;212:117-125. 12. Holmes JF, Akkinepalli R. Computed tomography versus plain radiography to screen for cervical spine injury: a meta-analysis. J Trauma. 2005;58:902-905. 13. Hadley MNM, Walters B, Grabb PA, et al. Section on Disorders of the Spine and Peripheral Nerves of the American Association of Neurological Surgeons and the Congress of Neurological Surgeons. 2001. http://www.aans.org/Education%20and%20Meetings/Clinical %20Guidelines.aspx. Accessed January 2, 2014. 14. Mannix R, Nigrovic LE, Schutzman SA, et al. Factors associated with the use of cervical spine computed tomography imaging in pediatric trauma patients. Acad Emerg Med. 2011;18:905-911. 15. Larson DB, Johnson LW, Schnell BM, et al. Rising use of CT in child visits to the emergency department in the United States, 1995-2008. Radiology. 2011;259:793-801. 16. Broder J, Fordham LA, Warshauer DM. Increasing utilization of computed tomography in the pediatric emergency department, 20002006. Emerg Radiol. 2007;14:227-232. 17. Adelgais KM, Browne L, Holsti M, et al. Cervical spine computed tomography utilization in pediatric trauma patients. J Pediatr Surg. 2014;49:333-337. 18. Nigrovic LE, Rogers AJ, Adelgais KM, et al. Utility of plain radiographs in detecting traumatic injuries of the cervical spine in children. Pediatr Emerg Care. 2012;28:426-432. 19. Mower WR, Hoffman JR, Pollack CV Jr, et al. Use of plain radiography to screen for cervical spine injuries. Ann Emerg Med. 2001;38:1-7. 20. Buhs C, Cullen M, Klein M, et al. The pediatric trauma C-spine: is the “odontoid” view necessary? J Pediatr Surg. 2000;35:994-997. 21. Brockmeyer DL, Ragel BT, Kestle JR. The Pediatric Cervical Spine Instability Study. A pilot study assessing the prognostic value of four imaging modalities in clearing the cervical spine for children with severe traumatic injuries. Int Soc Pediatr Neurosurg. 2012;28: 699-705. 22. Streitwieser DR, Knopp R, Wales LR, et al. Accuracy of standard radiographic views in detecting cervical spine fractures. Ann Emerg Med. 1983;12:538-542. 23. Nguyen GK, Clark R. Adequacy of plain radiography in the diagnosis of cervical spine injuries. Emerg Radiol. 2005;11:158-161. 24. Tins BJ, Cassar-Pullicino VN. Imaging of acute cervical spine injuries: review and outlook. Clin Radiol. 2004;59:865-880. 25. Pearce MS, Salotti JA, Little MP, et al. Radiation exposure from CT scans in childhood and subsequent risk of leukaemia and brain tumours: a retrospective cohort study. Lancet. 2012;380:499-505. 26. Berrington de Gonzalez A, Mahesh M, Kim KP, et al. Projected cancer risks from computed tomographic scans performed in the United States in 2007. Arch Intern Med. 2009;169:2071-2077. 27. Oladeji O, Johnston TE, Smith BT, et al. Quality of life in children with spinal cord injury. Pediatr Phys Ther. 2007;19:296-300.
Volume 65, no. 3 : March 2015
Hannon et al
Pediatric Cervical Spine Injury Evaluation
28. Ravichandran G, Silver JR. Survival following traumatic tetraplegia. Paraplegia. 1982;20:264-269. 29. Kaplan RM, Bush JW, Berry CC. Health status: types of validity and the index of well-being. Health Serv Res. 1976;11:478-507. 30. Hennelly KE, Fine AM, Jones DT, et al. Risks of radiation versus risks from injury: a clinical decision analysis for the management of penetrating palatal trauma in children. Laryngoscope. 2013;123: 1279-1284. 31. Stein SC, Hurst RW, Sonnad SS. Meta-analysis of cranial CT scans in children. A mathematical model to predict radiation-induced tumors. Pediatr Neurosurg. 2008;44:448-457. 32. Brenner DJ, Hall EJ. Computed tomography—an increasing source of radiation exposure. N Engl J Med. 2007;357:2277-2284.
33. Schneider AB, Ron E, Lubin J, et al. Dose-response relationships for radiation-induced thyroid cancer and thyroid nodules: evidence for the prolonged effects of radiation on the thyroid. J Clin Endocrinol Metab. 1993;77:362-369. 34. Mueller DL, Hatab M, Al-Senan R, et al. Pediatric radiation exposure during the initial evaluation for blunt trauma. J Trauma. 2011;70: 724-731. 35. Ravichandran G, Silver JR. Missed injuries of the spinal cord. Br Med J (Clin Res Ed). 1982;284:953-956. 36. Smith-Bindman R, Lipson J, Marcus R, et al. Radiation dose associated with common computed tomography examinations and the associated lifetime attributable risk of cancer. Arch Intern Med. 2009;169:2078-2086.
Links to Additional Resources
The Annals Web site provides links to helpful resources. Go to the Resources pull-down menu and click on the Smart EM link to take you to Scientific Medicine and Research Translation (SMART EM). This site, created by Annals’ podcast editors David Newman and Ashley Shreves, presents podcasts on numerous topics relevant to emergency medicine. You can visit SMART EM directly at www.smartem.org.
Volume 65, no. 3 : March 2015
Annals of Emergency Medicine 247
Pediatric Cervical Spine Injury Evaluation After Blunt Trauma: A Clinical Decision Analysis Megan Hannon, MD*; Rebekah Mannix, MD, MPH; Kate Dorney, MD; David Mooney, MD; Kara Hennelly, MD *Corresponding Author. E-mail: [email protected], Twitter: @mhannon19.
Study objective: Although many adult algorithms for evaluating cervical spine injury use computed tomography (CT) as the initial screening modality, this may not be appropriate in low-risk children, considering radiation risks. We determine the optimal initial evaluation strategy for cervical spine injury in pediatric blunt trauma. Methods: We constructed a decision analysis tree for a hypothetical population of patients younger than 19 years with blunt trauma, using 3 strategies: clinical stratification, screening radiographs followed by focused CT if the radiograph result was positive, and CT. For the model inputs, we used the current literature to determine the probabilities of cervical spine injury and estimate the long-term risks of malignancy after CT, as well as test characteristics of radiographic imaging. We used published utilities and conducted 1- and 2-way sensitivity analyses to determine the optimal strategy for evaluation of pediatric cervical spine injury. Results: In our model of a population with blunt trauma, the expected value of a clinical stratification strategy was the highest of the 3 strategies, making it the overall preferred management. One-way sensitivity analysis of several contributing factors revealed that the only independent factor that altered the dominant strategy was the sensitivity of clinical clearance criteria, lowering the threshold at which screening-radiograph strategy is optimal. Within the patient population considered as having non-negligible risk by clinical stratification and thus requiring imaging, the preferred imaging modality was screening radiograph/focused CT. The probability of cervical spine injury above which CT became the preferred strategy was 24.9%. Conclusion: The model highlights that clinical clearance and screening radiographs in a hypothetical trauma pediatric population are preferred strategies, whereas CT scanning is rarely the initial optimal evaluation. [Ann Emerg Med. 2015;65:239-247.] Please see page 240 for the Editor’s Capsule Summary of this article. A feedback survey is available with each research article published on the Web at www.annemergmed.com. A podcast for this article is available at www.annemergmed.com. 0196-0644/$-see front matter Copyright © 2014 by the American College of Emergency Physicians. http://dx.doi.org/10.1016/j.annemergmed.2014.09.002
INTRODUCTION Background Every year, there are approximately 10 million children who experience traumatic injuries, many of whom require specific cervical spine evaluation.1 Although pediatric blunt trauma is common, cervical spine injuries occur in only 1% to 2% of all pediatric patients with blunt trauma.2-4 Despite the rarity of injury, the cervical spine trauma evaluation presents a diagnostic dilemma because the risk of a missed injury has potential for devastating neurologic sequelae.5,6 Initial management strategies in the evaluation of the cervical spine of children with blunt trauma range from clinical “clearance” (ie, use of clinical screening criteria to remove cervical spine precautions) to advanced imaging. Clinical clearance of the asymptomatic pediatric patient is often the first-line
Volume 65, no. 3 : March 2015
management strategy despite only limited data supporting this approach. Some studies have attempted to apply adult approaches to the pediatric population, such as the National Emergency X-radiography Utilization Study (NEXUS).2 The NEXUS study was hampered by a small sample size, especially in the youngest children. The Pediatric Emergency Care Applied Research Network (PECARN) conducted a large, multicenter, case-control study to define clinical clearance criteria specific to children; however, this study has not been validated prospectively.7 Thus, clinicians remain uncertain about which clinical screening criteria, if any, to use when evaluating children for potential cervical spine injury after blunt trauma.8,9 For the symptomatic patient, it is generally agreed that radiographic imaging is a necessary adjunct, but there are no consistent recommendations on indications or modality. Adult
Annals of Emergency Medicine 239
Hannon et al
Pediatric Cervical Spine Injury Evaluation
Editor’s Capsule Summary
What is already known on this topic Computed tomography (CT) is more sensitive than radiography in identifying cervical spine injuries, but carries risk of radiation-induced malignancy. What question this study addressed Clinical decision analysis was used to determine the optimal cervical spine evaluation strategy for maximizing quality of life for children with blunt trauma, balancing risk of missed cervical spine injury with risk of radiation-induced malignancy. What this study adds to our knowledge In a hypothetical cohort of pediatric patients with blunt trauma, using literature-derived probabilities and utilities, clinical clearance and screening plain radiography were preferred over a CT-all strategy. How this is relevant to clinical practice Unlike for adults, for pediatric trauma patients cervical spine plain radiography screening may be preferred over a CT-all strategy.
trauma protocols recommend computed tomography (CT) for symptomatic patients because it is considered both a highly sensitive test and cost-effective approach.10-12 Given lower cervical spine injury rates in and greater sensitivity to radiation in children versus adults, it is unclear which imaging strategy should be applied in pediatrics. In contrast to adult protocols, the American Association of Neurological Surgeons consensus pediatric guidelines for symptomatic patients recommend screening radiographs followed by focused CT if abnormalities are present on the radiograph, though the guidelines also acknowledge the lack of evidence to support the recommendation.13 Given a paucity of definitive evidence in relation to a pediatric clearance strategy, as well as the fact that the majority of children are treated in nonpediatric emergency departments, in which advance imaging is more standard, many children likely undergo CT as the initial screening modality.14-17 The absence of validated clinical clearance protocols and controversy surrounding appropriate imaging modalities lead to wide practice variation. Importance Given the practice variation and the lack of definitive evidence for an approach to screening children for cervical spine injuries, understanding the risks and benefits of different initial screening strategies could guide management decisions. The rarity of cervical spine injuries makes meaningful prospective data 240 Annals of Emergency Medicine
collection challenging. Decision analysis is an objective way to compare the various approaches. Goals of This Investigation We explored several cervical spine evaluation strategies, including clinical stratification and both CT and screening radiographs, with particular attention to the risk tradeoffs between the importance of identifying a significant cervical spine injury and the risk of radiation-induced malignancy. We determined the optimal management strategy for a hypothetical pediatric population with blunt trauma by identifying the risk threshold at which CT imaging becomes the preferred strategy, considering health-related quality of life and risk of radiation.
MATERIALS AND METHODS This study was based on literature review and was exempt from review by our institutional review board. We constructed a decision analysis tree for a hypothetical population of patients younger than 19 years and with blunt trauma, using 3 strategies: clinical stratification, screening radiographs followed by focused CT if the radiograph result was positive, and CT (Figure 1, depicted as square decision nodes). As shown in the decision tree, after the initial choice, physicians observe the outcome of chance events (circles) and ultimately end at one of the possible terminal node (triangles). Utilities are listed at each terminal node and are a numeric measure representative of the quality of life of an individual with a particular state of health (ie, represents an individual with a malignancy or cervical spine injury). This decision tree determines the optimal patientcentered approach for cervical spine evaluation in a population with blunt trauma, with the goal of reducing harm to the individual. In decision analysis, the measure of a given decision is represented as an expected value. Expected values are determined by first identifying the potential value of a given outcome (positive or negative). The value unit reflects the aim of the study, eg, financial or quality of life. In this case, the value was defined as a utility for a given health state, which will be discussed below. Next, the probability that each result will occur from a given course of action is determined, usually from previous literature. One then multiplies the value by the probability of an outcome to calculate an expected value of a particular result. As an example, in our decision tree, the choice to perform a CT results in 2 possible outcomes (branches), negative or positive cervical spine injury result. After this, there are 2 additional possible outcomes or branches: having a malignancy related to radiation or not. Each outcome has a utility (displayed after the terminal node in the decision tree; Figure 1). Using a probability of malignancy of 0.004 (and 1–probability of malignancy¼0.996 as the probability of remaining malignancy free), expected values can be calculated. If the patient had both a cervical spine injury and malignancy, the expected value calculated is 0.002. If the Volume 65, no. 3 : March 2015
Hannon et al
Pediatric Cervical Spine Injury Evaluation
Figure 1. Decision analysis tree using 3 strategies for a hypothetical population of patients younger than 19 years and with blunt trauma: clinical stratification, screening radiographs followed by focused CT if the radiograph result was positive, and CT, each depicted as square decision nodes. After the initial choice, an outcome is observed of chance events (circles). Each branch ends at a possible terminal node (triangles). Utilities are listed at each terminal node. CSI, Cervical spine injury.
patient has cervical spine injury but no malignancy, the expected value calculated is 0.6972. We can add these expected values to represent that of the entire CT positive branch as 0.6992. Calculating in a similar fashion, the CT negative-result branch expected value is 0.9984. To obtain the expected value of an entire extended branch that includes several nodes, a process named rolling back is completed. Rolling back is multiple rounds of the aforementioned calculations using probabilities and branch values sequentially. The expected values reported in the results for each strategy are the conglomerate expected utility for the entire branch after rolling back. To examine the effect of certain factors such as radiation risk and varying probability of cervical spine injury on presentation, as well as variability of both clinical and radiographic performance characteristics, we performed sensitivity analyses ranging through the 95% confidence intervals (CIs) extracted from available data around each factor. We constructed the decision analysis model using TreeAge Pro 2014 (TreeAge Software Inc, Williamstown, MA). As described above, we made a decision tree with 3 clinically common initial management plans for our hypothetical population. All sensitivity analyses and threshold acquisition were performed with the TreeAge Pro software. Volume 65, no. 3 : March 2015
For our baseline model inputs, we used standard definitions and data from a combination of recently published studies (Table 1). We defined cervical spine injury as any radiographic finding, excluding anatomic variants independent of neurologic outcome. We did not differentiate between injuries resulting in neurologic deficits or requiring intervention from those with clinically insignificant findings. For our model, we used probabilities obtained from the largest available studies, including NEXUS and PECARN. To estimate the probability of cervical spine injury, we used the NEXUS study. Though the NEXUS entrance criteria required initial imaging, thus likely excluding the lowest-risk patients, we chose this data set because it is the largest pediatric data set and would likely yield the most conservative estimates. In NEXUS, subjects were considered to have non-negligible cervical spine injury risk if any of the following characteristics were present: midline tenderness, altered mental status, intoxication, neurologic deficits, and distracting injury. There were a total of 3,065 pediatric trauma patients with cervical spine imaging, of whom 0.98% (30 subjects) had cervical spine injuries. Overall, there were 2,462 NEXUS-positive patients. All of the cervical spine injuries met at least 1 NEXUS criterion. The clearance criterion was then observed to have 100% sensitivity Annals of Emergency Medicine 241
Hannon et al
Pediatric Cervical Spine Injury Evaluation Table 1. Summary of decision tree inputs.* Probabilities p(CSImain) p(CSINEXUSpos) p(CSINEXUSneg) Radiographic specificity Radiographic sensitivity p(malignancy) Utilities Healthy CSI/radiographþ CSI/radiographþ/malignancy CSI/radiograph– No CSI/radiographþ/malignancy No CSI/radiograph– CSI/CTþ CSI/CTþ/malignancy
Baseline
Source 2
0.98% 1.22% 0.001% 95% 90% 0.4%
Viccellio et al Viccellio et al2 Viccellio et al2 Streitwieser et al,22 Nguyen and Clark,23 Tins and Cassar-Pullicino24 Nigrovic et al18 Berrington de Gonzalez et al26
1 0.7 0.5 0.6499 0.6 0.9999 0.7 0.5
Kaplan Kaplan Kaplan Kaplan Kaplan Kaplan Kaplan
et et et et et et et
al,29 al,29 al,29 al,29 al,29 al,29 al,29
Oladeji Oladeji Oladeji Oladeji Oladeji Oladeji Oladeji
et et et et et et et
al,27 al,27 al,27 al,27 al,27 al,27 al,27
Ravichandran Ravichandran Ravichandran Ravichandran Ravichandran Ravichandran Ravichandran
and and and and and and and
Silver28 Silver,28 Stein et al31 Silver28 Silver,28 Stein et al31 Silver28 Silver28 Silver,28 Stein et al31
p(CSImain), Probability of CSI in the main trauma population; p(CSINEXUSneg), probability of cervical spine injury in the negligible-risk group; p(malignancy), probability of malignancy. *Probabilities of each outcome were obtained from the best available literature that described the particular outcomes or testing characteristics. Utilities representative of health states were obtained from pediatric survey literature and supported by adult data.
(95% CI 87.8% to 100%). In our model, we considered NEXUS positivity to equate to a symptomatic or non-negligible cervical spine injury risk population. Thus, we used 0.98% as the probability of cervical spine injury in the overall trauma population and 1.22% for that in the NEXUS-positive group. To define the characteristics of screening radiographs, we used recently published PECARN data that evaluated the test characteristics of cervical spine radiographs.18 For this retrospective study, the entrance criteria were a pediatric trauma patient with related bony or ligamentous injury and who had cervical spine radiographic imaging. Cervical radiographic sensitivity was 90% (95% CI 85% to 94%), which we used as our baseline model input. PECARN reported that 33% had 2-view radiographs, whereas 67% had greater than or equal to 3-view radiographs. The types of radiographs in the study reflected common practice in the pediatric centers enrolled in PECARN. In general, the literature and the American Association of Neurological Surgeons guidelines support 2-view radiographs for patients younger than 8 years, with an additional odontoid view in those older than 8 years; thus, that is how the model’s radiograph strategy was defined. Additional smaller studies had somewhat lower sensitivities reported, and these data were used in sensitivity analyses.19,20 Specificity obtained from various trauma literature supported a wide range, from 70% to 95%.21-24 We estimated probability of malignancy related to CT radiation exposure from the most recently published data on projected lifetime cancer risk for pediatric patients.25,26 For example, it is projected that malignancy related to radiation exposure from a cervical spine CT is 70/10,000 for a 1-year-old girl and 10/10,000 for a 15-year-old boy. Thus, we chose to average the risks at a conglomerate baseline rate of 40/10,000 for a baseline rate and range through the aforementioned low and high rates with sensitivity analysis. When possible, to be compliant with our objective of creating a tree to minimize harm to the individual, we chose to 242 Annals of Emergency Medicine
incorporate the most conservative data to avoid any missed diagnosis and minimize poor outcomes. In our model, we assigned values (or utilities) to the health outcomes, using a standard scale from 0 (death) to 1 (baseline health, assumed as perfect health for this hypothetical population). Published utility scales were used that combine several major domains: physical, emotional, social, cognitive, and school capabilities. Many are adult survey based, but when pediatric-specific utilities were available, we used them as reviewed below. We assigned cervical spine injury utilities according to publications on the quality of life of pediatric patients with spinal cord injuries (Table 1).27,28 As noted above, we erred conservatively and considered any radiographic cervical spine injury to be equivalent to a patient with severe spinal cord injury, with a baseline utility of 0.7. With delayed diagnosis, we assigned an additional disutility according to the assumption that early detection results in better outcomes. This translated to a utility of 0.65 if a cervical spine injury was missed by a radiograph and 0.6 if a patient had no imaging. These utilities are comparable to those for adult cervical spine injury.29 We estimated the utility of a malignancy to be 0.6, as used in previous pediatric studies using a Markov model for head CT.30,31 If a patient had cervical spine injury and malignancy, then he or she mathematically ought to have a utility of 0.4, but given the later onset of malignancy (likely 10 or more years), we used a higher value of 0.5, which errs on the side of not missing the immediate diagnosis of a spine injury. Sensitivity Analyses Baseline model inputs (Table 1) were used in the basic decision tree. From the literature, we obtained data on the CIs around each model input and performed sensitivity analyses to measure the effect of each contributing factor, including the probability of cervical spine injury in the main trauma Volume 65, no. 3 : March 2015
Hannon et al
Pediatric Cervical Spine Injury Evaluation
Table 2. Results of 1-way sensitivity analysis for the overall trauma population. Model Factor p(CSINEXUSneg) Radiograph sensitivity Radiograph specificity p(malignancy) CSI utility
Baseline
Range
0.001%
0%–0.6%
90% 95% 0.4% 0.6
60%–100% 60%–100% 0.1%–0.7% 0.4–0.9
Source
Effect
Nigrovic et al,18 Buhs et al,20 Mower et al19 Streitwieser et al,22 Nguyen and Clark,23 Tins and Cassar-Pullicino24 Berrington de Gonzalez et al26 Kaplan et al,29 Oladeji et al,27 Ravichandran and Silver28
At 0.03%, the preferred strategy became screening radiograph No change No change No change No change
2
Viccellio et al
population, the probability of CT radiation–related malignancy (probability of malignancy), sensitivity of radiograph imaging, and the utility of cervical spine injury. Additionally, the sensitivity of a clinical clearance rule was evaluated by ranging the probability of cervical spine injury in the negligible-risk group (or NEXUS negative). This variable probability of cervical spine injury in the negligible-risk group was calculated from the CI reported in the NEXUS study. All of the sensitivity analysis parameters were obtained from the primary literature described above and are summarized in Table 2. A separate analysis was performed on the non-negligible risk population branch to understand how changes in probability of malignancy related to CT radiation might affect imaging modality choices. We performed 1-way sensitivity analysis on probability of cervical spine injury in the NEXUSpositive group and 2-way sensitivity analysis ranging both the probability of cervical spine injury and probability of malignancy.
RESULTS Overall, the expected values were high for each major decision node, according to the overall baseline health status of the pediatric population and the low rate of cervical spine injury. The greatest variation was based on the outcome (cervical spine injury or malignancy) (Figure 1). The expected value of each strategy was as follows: clinical stratification yielded 0.996912, screening radiograph/focused CT yielded 0.996888, and CT-all yielded 0.995468. The preferred management plan in this analysis, according to the expected probability of cervical spine injury of 0.98%, was clinical stratification. One-way sensitivity analysis determined the threshold at which the preferred strategy changes with independent alterations of model inputs (Table 2). The probability of cervical spine injury in the NEXUS-positive group was the only independent variable that met a threshold at which an alternate strategy was preferred. At a probability of cervical spine injury in the NEXUS-positive group of 0.03%, which represents a
Figure 2. Once the decision to image has been made, the choice of imaging modality is controversial. To represent the decision, the non-negligible CSI risk group was analyzed. The screening radiograph strategy was superior at a probability of CSI of 1.22% in this subgroup. To understand the effect of the p(CSINEXUSpos), 1-way sensitivity analysis was performed and identified 24.9% to be the p(CSINEXUSpos) at which above CT-all would be preferred. p(CSINEXUSpos), probability of CSI in the NEXUS-positive group. Volume 65, no. 3 : March 2015
Annals of Emergency Medicine 243
Hannon et al
Pediatric Cervical Spine Injury Evaluation
Figure 3. In the non-negligible CSI risk group, inputs with potential significant variability based on circumstance included both the probability of CSI (p[CSINEXUSpos]) and the radiation risk (probability of malignancy). A 2-way sensitivity analysis represented the effect of varying both the p(CSINEXUSpos) and the radiation risk probability of malignancy. As the probability of malignancy decreased or in scenarios suggesting the probability of CSI to be much higher than baseline tree inputs, CT strategy dominated.
sensitivity of NEXUS criteria of 99%, a screening radiograph/ focused CT becomes a preferred management plan. The non-negligible risk population of the clinical stratification branch of the tree had 2 branches of its own, which were analyzed separately: screening radiograph/focused CT and CTall. The expected value of each strategy was as follows: screening radiograph/focused CT yielded 0.996165 and CT-all yielded 0.994749. Thus, the preferred management plan was screening radiograph/focused CT, according to the expected probability of cervical spine injury in the NEXUS-positive group of 1.22%. The probability of cervical spine injury in the NEXUS-positive group above which CT-all strategy became the preferred strategy was 24.9% (Figure 2). A 2-way sensitivity analysis represented the effect of varying both the probability of cervical spine injury in the NEXUS-positive group and the radiation risk probability of malignancy (Figure 3). In the decision tree, we used a probability of CT radiation exposure–related malignancy of 0.4% (40 in 10,000 scans). As expected, reduction in proposed malignancy risk from CT resulted in a shift toward CT use. For example, with a risk of malignancy of 0.1%, a threshold for CT would be reduced to a probability of cervical spine injury in the NEXUS-positive group of 6.5%. As the probability of cervical spine injury in the NEXUS-positive group increased, CT became acceptable at a higher probability of malignancy.
scanner type. For example, the malignancy-related radiation risk for a 15-year-old boy is estimated to be as low as 10/10,000 scans, whereas for a 1-year-old girl it is as high as 70/10,000 scans. However, sensitivity analyses allowed us to vary radiation risks, accounting for both the inaccuracies of estimates and individualization of these variations. Second, the model is limited by the decision to include only the factors that contribute directly to patient health outcomes. Obviously, the scope of any health care issue is complicated by both societal and financial realities. For example, the model identifies screening radiograph/focused CT as the predominant strategy in reduction of patient harm. Other factors that were not included, such as cost, risk of screening a population, and falsepositives implications, may alter this result, given that the expected values are so close. Third, the literature used to generate probabilities is heterogeneous, with variable entrance criteria (ie, patients who were imaged versus all patients versus only those with injuries). No study likely accurately reflects the population who requires cervical spine imaging. However, sensitivity analysis provides insight on how variance of the inputs affects optimal imaging strategy.
DISCUSSION LIMITATIONS There are several limitations to this study. First, radiation risk is variable according to institution, age, sex, body size, and 244 Annals of Emergency Medicine
Through decision analysis, we identified the optimal management strategies for cervical spine evaluation in 2 populations: a main population, inclusive of all pediatric patients with blunt trauma and requiring cervical spine evaluation, and a Volume 65, no. 3 : March 2015
Hannon et al higher-risk population (or non-negligible risk), as defined by NEXUS criteria positivity. The decision tree was created with the best available data. It is representative of the decisionmaking process that occurs in such a clinical circumstance but is not a clinical decision rule and thus may not reflect clinical workflow perfectly. Balancing radiation risk from CT- and health-related outcomes with respect to cervical spine injury, as well as the potential ramifications of a delayed diagnosis, we found that clinical stratification is the preferred strategy, followed by radiograph/focused CT. This is important because the model suggests less aggressive imaging approaches for the majority of patients. The results suggest that on the population level, the risk of CT radiation–related malignancy outweighs the small number of cervical spine injuries identified by advanced imaging in most circumstances. Certainly some factors, such as neurologic deficits or worsening clinical status, would increase the pretest probability and one would opt for CT immediately; however, most individual criteria, such as midline tenderness or mechanism of injury, would not increase the pretest probability of cervical spine injury to the threshold value of 24.9%, above which the CT-all strategy is optimal. Although not developed for this purpose, several studies, including NEXUS, provide data about the rates of cervical spine injury for each clearance criterion.2 This may be a useful way to help provide a basis for stratifying patients. For example, from the NEXUS data, the probability of cervical spine injury if a patient has midline tenderness is 1.75%. If a patient has a neurologic deficit, the probability of cervical spine injury increases to 4.44%. Neither meets the CT threshold as an individual factor. The data on cumulative probability of cervical spine injury with multiple characteristics are not reported in this data set. However, this information provides a sense of the relative risk of individual risk factors for a cervical spine injury. Data are now available on the cumulative ionizing radiation effects on radiosensitive tissues, particularly the thyroid.32,33 Often in a trauma evaluation cervical spine CT is part of a larger evaluation in which both head and chest imaging are considered and thus the cumulative radiation effects must be considered.34 With the above data and reports of cervical spine CT rates increasing in some areas, despite a simultaneous downward trend in other types of scans, clinicians must continue to address the challenge to reduce unnecessary scans.15 This model suggests that unless the probability of cervical spine injury is high, clinical clearance or screening radiographs should predominate the current management strategies when radiation risk is considered. Over time, radiation risk may change and a lower threshold may be considered for CT, as can be appreciated in the 2-way sensitivity analysis (Figure 3). Of course, there needs to be periodic reassessment as imaging technology improves, CT radiation exposure reduces, or alternative imaging options become feasible options (ie, more rapid magnetic resonance imaging that does not require sedation). There are several factors that make this a conservative analysis, consistent with an emphasis on the reduction of individual harm, including using what is probably an overestimate of the probability Volume 65, no. 3 : March 2015
Pediatric Cervical Spine Injury Evaluation of cervical spine injury, as well as equating all cervical spine injury. First, the model was based on estimates from the NEXUS study, a probability of cervical spine injury in the main trauma population of 0.98% and of 1.22% in the NEXUS-positive group (non-negligible risk group). Because the entrance criteria for the NEXUS study required previous cervical spine radiography, this likely overestimates the actual rate of cervical spine injury in the general trauma population who requires cervical spine evaluation. There is no published data set that reliably measures the baseline rate because the “trauma” population requiring cervical spine evaluation is difficult to capture. For a completely asymptomatic, or NEXUSnegative, group, the rate of cervical spine injury is thought to be zero or at least closely approaching zero, according to previous literature.2 We used sensitivity analysis to examine the potential that NEXUS criteria have a lower sensitivity than the chosen baseline rate because it is not well validated in pediatrics. If the sensitivity is as high as 99%, then the screening radiograph/focused CT predominates. CT, however, does not become preferred in the chosen ranges of variables. Second, the utility applied to any cervical spine injury is based on having significant spinal injury and neurologically impairing lifelong disability. Ultimately, this means that any radiographically positive finding was equated to a true neurologically impairing spinal cord injury despite that most patients will have no neurologic sequelae and a low rate of neurosurgical intervention. Furthermore, because prompt diagnosis aids in appropriate treatment and better outcomes, an even lower utility is applied if the diagnosis of cervical spine injury is delayed.35 We recognize that both factors reduce harm in the model but potentially slant the outcomes toward imaging. The model is not applicable to the adult trauma population, which, for many reasons, is quite different. First, largely related to anatomic differences, the rates of cervical spine injury are much higher, ranging up to 20%, and the clinical relevance of such injuries is much higher.10,13 The risk of radiation to the neck and therefore the associated probability of malignancy are much lower.26,36 Also, studies on cervical spine radiographs recognize that sensitivity is much lower in the adult population, as low as 52%,12 making them a much less useful screening test. Finally, often there is greater clinician comfort to complete cervical spine clearance with a cooperative adult, using validated clinical clearance criteria that reliably define a low-risk population. Evaluating a cervical spine for a pediatric patient may be a less common occurrence in nonpediatric trauma centers. All of these factors would alter the analysis greatly if the decision tree were built with adult data as model inputs. An adult decision analysis would identify a much lower CT threshold in both the main and non-negligible risk population. It is important to consider this because most pediatric trauma patients are evaluated in non–Level I general trauma centers, where CT rates are much higher than in Level I pediatric trauma facilities.14 Likely lack of definitive pediatric trauma guidelines for cervical spine evaluation and clinician comfort and familiarity with CTfocused strategies in adult emergency medicine trauma protocols Annals of Emergency Medicine 245
Pediatric Cervical Spine Injury Evaluation contribute to the practice variation among clinicians and facilities. In conclusion, the model highlights that in both the main and non-negligible risk populations, CT scanning is not the optimal initial strategy at the current probabilities of cervical spine injury and radiation risk. The various strategy thresholds support a less CT-focused approach, thus decreasing associated radiation risk. For a patient with a normal neurologic examination result, nontender neck, and otherwise normal mental status, the risk of cervical spine injury is typically not high enough to outweigh the risk of malignancy from a cervical spine CT. Individual risk for a particular patient must be considered and may be much higher and place the patient above the CT threshold from the onset. As with any decision analysis, clinical judgment trumps the universal application of a management strategy, but we provide objective aid to this complicated decision by evaluating the influences of the risk of radiation exposure and the consequences of cervical spine injury to maximize patient health outcome. Supervising editor: Kelly D. Young, MD, MS Author affiliations: From the Division of Emergency Medicine (Hannon, Mannix, Dorney, Hennelly) and the Department of Surgery (Mooney), Boston Children’s Hospital, Boston, MA. Author contributions: MH, RM, and KH conceived the study, designed the decision tree, and performed data analysis. MH, RM, DM, and KH reviewed the tree and inputs and made appropriate improvements. MH drafted the article, and all authors contributed substantially to its revision. MH takes responsibility for the paper as a whole. Funding and support: By Annals policy, all authors are required to disclose any and all commercial, financial, and other relationships in any way related to the subject of this article as per ICMJE conflict of interest guidelines (see www.icmje.org). The authors have stated that no such relationships exist. Publication dates: Received for publication February 27, 2014. Revisions received June 17, 2014; July 30, 2014; and August 18, 2014. Accepted for publication September 2, 2014. Available online October 16, 2014.
REFERENCES 1. Centers for Disease Control and Prevention, National Center for Injury Prevention and Control. Web-based Injury Statistics Query and Reporting System (WISQARS). www.cdc.gov/injury/wisqars. Accessed January 2, 2014. 2. Viccellio P, Simon H, Pressman BD, et al. A prospective multicenter study of cervical spine injury in children. Pediatrics. 2001;108:E20. 3. Platzer P, Jaindl M, Thalhammer G, et al. Cervical spine injuries in pediatric patients. J Trauma. 2007;62:389-396; discussion 94-96. 4. Polk-Williams A, Carr BG, Blinman TA, et al. Cervical spine injury in young children: a National Trauma Data Bank review. J Pediatr Surg. 2008;43:1718-1721. 5. Guidelines for the acute medical management of severe traumatic brain injury in infants, children, and adolescents. J Trauma. 2003;54:S235-310.
246 Annals of Emergency Medicine
Hannon et al 6. Leonard JR, Jaffe DM, Kuppermann N, et al. Cervical spine injury patterns in children. Pediatrics. 2014 Apr 28. [Epub ahead of print]. 7. Leonard JC, Kuppermann N, Olsen C, et al. Factors associated with cervical spine injury in children after blunt trauma. Ann Emerg Med. 2011;58:145-155. 8. Burns EC, Yanchar NL. Using cervical spine clearance guidelines in a pediatric population: a survey of physician practices and opinions. CJEM. 2011;13:1-6. 9. Browne GJ, Lam LT, Barker RA. The usefulness of a modified adult protocol for the clearance of paediatric cervical spine injury in the emergency department. Emerg Med (Fremantle). 2003;15:133-142. 10. Como JJ, Diaz JJ, Dunham CM, et al. Practice management guidelines for identification of cervical spine injuries following trauma: update from the Eastern Association for the Surgery of Trauma Practice Management Guidelines Committee. J Trauma. 2009;67:651-659. 11. Blackmore CC, Ramsey SD, Mann FA, et al. Cervical spine screening with CT in trauma patients: a cost-effectiveness analysis. Radiology. 1999;212:117-125. 12. Holmes JF, Akkinepalli R. Computed tomography versus plain radiography to screen for cervical spine injury: a meta-analysis. J Trauma. 2005;58:902-905. 13. Hadley MNM, Walters B, Grabb PA, et al. Section on Disorders of the Spine and Peripheral Nerves of the American Association of Neurological Surgeons and the Congress of Neurological Surgeons. 2001. http://www.aans.org/Education%20and%20Meetings/Clinical %20Guidelines.aspx. Accessed January 2, 2014. 14. Mannix R, Nigrovic LE, Schutzman SA, et al. Factors associated with the use of cervical spine computed tomography imaging in pediatric trauma patients. Acad Emerg Med. 2011;18:905-911. 15. Larson DB, Johnson LW, Schnell BM, et al. Rising use of CT in child visits to the emergency department in the United States, 1995-2008. Radiology. 2011;259:793-801. 16. Broder J, Fordham LA, Warshauer DM. Increasing utilization of computed tomography in the pediatric emergency department, 20002006. Emerg Radiol. 2007;14:227-232. 17. Adelgais KM, Browne L, Holsti M, et al. Cervical spine computed tomography utilization in pediatric trauma patients. J Pediatr Surg. 2014;49:333-337. 18. Nigrovic LE, Rogers AJ, Adelgais KM, et al. Utility of plain radiographs in detecting traumatic injuries of the cervical spine in children. Pediatr Emerg Care. 2012;28:426-432. 19. Mower WR, Hoffman JR, Pollack CV Jr, et al. Use of plain radiography to screen for cervical spine injuries. Ann Emerg Med. 2001;38:1-7. 20. Buhs C, Cullen M, Klein M, et al. The pediatric trauma C-spine: is the “odontoid” view necessary? J Pediatr Surg. 2000;35:994-997. 21. Brockmeyer DL, Ragel BT, Kestle JR. The Pediatric Cervical Spine Instability Study. A pilot study assessing the prognostic value of four imaging modalities in clearing the cervical spine for children with severe traumatic injuries. Int Soc Pediatr Neurosurg. 2012;28: 699-705. 22. Streitwieser DR, Knopp R, Wales LR, et al. Accuracy of standard radiographic views in detecting cervical spine fractures. Ann Emerg Med. 1983;12:538-542. 23. Nguyen GK, Clark R. Adequacy of plain radiography in the diagnosis of cervical spine injuries. Emerg Radiol. 2005;11:158-161. 24. Tins BJ, Cassar-Pullicino VN. Imaging of acute cervical spine injuries: review and outlook. Clin Radiol. 2004;59:865-880. 25. Pearce MS, Salotti JA, Little MP, et al. Radiation exposure from CT scans in childhood and subsequent risk of leukaemia and brain tumours: a retrospective cohort study. Lancet. 2012;380:499-505. 26. Berrington de Gonzalez A, Mahesh M, Kim KP, et al. Projected cancer risks from computed tomographic scans performed in the United States in 2007. Arch Intern Med. 2009;169:2071-2077. 27. Oladeji O, Johnston TE, Smith BT, et al. Quality of life in children with spinal cord injury. Pediatr Phys Ther. 2007;19:296-300.
Volume 65, no. 3 : March 2015
Hannon et al
Pediatric Cervical Spine Injury Evaluation
28. Ravichandran G, Silver JR. Survival following traumatic tetraplegia. Paraplegia. 1982;20:264-269. 29. Kaplan RM, Bush JW, Berry CC. Health status: types of validity and the index of well-being. Health Serv Res. 1976;11:478-507. 30. Hennelly KE, Fine AM, Jones DT, et al. Risks of radiation versus risks from injury: a clinical decision analysis for the management of penetrating palatal trauma in children. Laryngoscope. 2013;123: 1279-1284. 31. Stein SC, Hurst RW, Sonnad SS. Meta-analysis of cranial CT scans in children. A mathematical model to predict radiation-induced tumors. Pediatr Neurosurg. 2008;44:448-457. 32. Brenner DJ, Hall EJ. Computed tomography—an increasing source of radiation exposure. N Engl J Med. 2007;357:2277-2284.
33. Schneider AB, Ron E, Lubin J, et al. Dose-response relationships for radiation-induced thyroid cancer and thyroid nodules: evidence for the prolonged effects of radiation on the thyroid. J Clin Endocrinol Metab. 1993;77:362-369. 34. Mueller DL, Hatab M, Al-Senan R, et al. Pediatric radiation exposure during the initial evaluation for blunt trauma. J Trauma. 2011;70: 724-731. 35. Ravichandran G, Silver JR. Missed injuries of the spinal cord. Br Med J (Clin Res Ed). 1982;284:953-956. 36. Smith-Bindman R, Lipson J, Marcus R, et al. Radiation dose associated with common computed tomography examinations and the associated lifetime attributable risk of cancer. Arch Intern Med. 2009;169:2078-2086.
Links to Additional Resources
The Annals Web site provides links to helpful resources. Go to the Resources pull-down menu and click on the Smart EM link to take you to Scientific Medicine and Research Translation (SMART EM). This site, created by Annals’ podcast editors David Newman and Ashley Shreves, presents podcasts on numerous topics relevant to emergency medicine. You can visit SMART EM directly at www.smartem.org.
Volume 65, no. 3 : March 2015
Annals of Emergency Medicine 247
