Identifying Low-Risk Trauma Patients for Blunt Cerebrovascular Injury: Evidence From the North Carolina Trauma Registry

Abstract

Background

Blunt cerebrovascular injury (BCVI) is a serious complication of blunt trauma with high rates of morbidity and mortality. There is evidence to support universal screening, but this raises concerns about cost, radiation exposure, and system burden. This study examined the connection between clinical factors at initial presentation and BCVI risk to identify opportunities to improve imaging utilization.

Methods

We analyzed patients with blunt injury from the North Carolina Trauma Registry (2016-2021). Blunt cerebrovascular injury was identified using ICD-10 codes. Clinical risk factors evaluated were low Glasgow Coma Score (GCS), significant supraclavicular injury (above-the-clavicle (ATC) injury), and trauma activation level. Logistic regression was used to identify predictors of BCVI.

Results

Among 198 211 blunt trauma patients, 1336 (0.70%) were diagnosed with BCVI. The highest prevalence was among patients with both a low GCS and significant ATC injury (4.0%), followed by those with only ATC injury (1.5%). Only two cases (0.001%) occurred in patients with neither indicator. The adjusted odds ratios for BCVI were 3.0 (95% CI: 2.6-3.5) for low GCS and 2.8 (95% CI: 2.4-3.2) for level 1 trauma activation. Above-the-clavicle injury was nearly a perfect predictor of BCVI.

Conclusions

Patients without altered mental status or significant injuries above the clavicle had an extremely low risk of BCVI. Although liberal screening strategies remain important to prevent missed injuries, our findings suggest that a clearly defined low-risk subgroup may not require routine CTA screening. Prospective, multicenter validation of contemporary screening criteria is needed to refine risk stratification and optimize cross-sectional imaging utilization.

Level of Evidence

Level III, Prognostic/Epidemiological.

Background

The appropriate management of blunt cerebrovascular injuries (BCVIs) has been widely discussed over the past decade. ¹ Because of the risk of devastating thromboembolic stroke and the poor sensitivity and high false-negative rate of traditional screening methods, broad screening of patients with significant blunt injuries has been recommended with the Eastern Trauma Society conditionally recommending imaging even for lower-risk patients. ² Universal screening using computed tomography angiography (CTA) of the head and neck can be performed alongside existing blunt trauma CT protocols without an additional contrast bolus and with only a marginal increase in radiation exposure and in-scanner time. The cost of CTA head and neck imaging is modest (including technologist and radiologist time, depreciation on equipment and facilities, and image storage), but billed charges vary widely, and patient charges can be substantial, leading to increased financial hardship for patients. ¹ The system burden on interpreting radiologists at a busy trauma center is also substantial and continues to increase. ³

The decision to further evaluate patients with CTA after blunt trauma depends on physical findings and the mechanism of injury. The expanded Denver Criteria (eDC) are commonly used to guide decision-making and include a broad range of risk factors, such as high-energy trauma with cervical spine fractures or facial fractures, to improve early detection of BCVI and reduce the risk of stroke. ⁴ Stroke secondary to BCVI typically occurs within 72 hours, making timely diagnosis essential. ⁵ The current clinical consensus supports early identification of BCVI since outcomes are improved with prompt initiation of anticoagulation or antiplatelet therapy. ⁶ However, field triage guidelines that determine access to trauma centers may rely on subjective interpretations of injury mechanisms, which can result in variability in how high-kinetic injuries are assessed in the pre-hospital and emergency department settings and may contribute to system under- or over-triage of patients.

Whole-body CT (WBCT) is frequently used in trauma; however, some studies report up to 40% of imaging is negative for traumatic injury, which again highlights concerns about unnecessary radiation exposure, health care cost, and over-triage. ⁷ Universal CTA of the head and neck is justified in populations where the BCVI risk is greater than 6%. ⁷ In the cited study, the eDC were found to be the optimal screening criteria, with a BCVI risk of 2.6-6%, which suggests that further refinement in screening criteria may be needed. In our study, we aimed to evaluate our state trauma database for relationships between changes in the Glasgow Coma Score (GCS), significant ICD-10-based supraclavicular injuries, and BCVI in blunt trauma patients to identify opportunities to improve CT imaging utilization at busy trauma centers. These data will provide additional context for risk stratification of patients for future prospective evaluations of BCVI screening protocols.

Methods

We used the NC Trauma Registry (NCTR) from 2016 to 2021 for our analysis. The NCTR has been used since 1987 to collect data on trauma patients in North Carolina. It is a multi-institutional project between all trauma centers in the state. Diagnoses were identified in the registry using ICD-10 codes. Our primary objective was to estimate the risk of BCVI for patients with blunt traumatic injury based on clinical risk factors, including trauma activation level, identified at the time of initial trauma assessment prior to CT imaging.

Our primary outcome was a diagnosis of BCVI based on recorded ICD-10 codes in the NCTR. We examined all patients with a blunt traumatic injury and a complete record in the NCTR. All the Denver Criteria were not available because physical exam findings are not included in the NCTR. Consequently, we focused on examining factors included in the registry that were readily available on initial assessment such as injury patterns and presenting characteristics. These included a low GCS, significant injury above the clavicle, and trauma activation level. Trauma activation level was included because it was associated with the severity of injury.

We estimated the relationship between BCVI incidence and these factors to identify patient populations that had a high or low risk of BCVI. A low GCS was defined as a GCS motor score of less than 6, or a low total GCS, defined as GCS <13. Significant injury above the clavicles (ATC) was defined using injuries used in the expanded Denver Criteria and other significant injuries as identified by two emergency physicians and one trauma surgeon. ICD-10 codes were used to identify these injuries in the registry. The Denver Criteria include major laceration of the carotid artery, LeFort II or III fracture, cervical spine (c-spine) subluxation, c-spine fractures extending into the transverse foramen, fractures of C1-C3, diffuse axonal injury, or injury by hanging. Trauma activation level included level 1, level 2, and level 3 (which at some centers may be a trauma consultation).

The prevalence of BCVI was examined among four groups ¹ : No significant ATC injuries and no Low GCS, ² Low GCS only, ³ Low GCS with significant ATC injury, and ⁴ significant ATC injury only. We report likelihood ratios for factors associated with a BCVI.

The association between these indicators and having a BCVI was then modeled using logistic regression. The model was fitted with statistically significant factors identified in our bivariate analysis. Variables were removed stepwise if they did not contribute meaningfully to the multivariate model. We report the final model with adjusted odds ratios and 95% confidence intervals. The final model included significant ATC, low GCS, and trauma activation level. Institutional IRB approval was obtained.

Results

During the study period, 198 211 patients with a blunt injury were included in the analysis. Most patients did not have a low GCS or significant ATC injury (n = 125 296). Otherwise, the most common cohorts were significant ATC injury only (n = 59 236), significant ATC injury plus low GCS (n = 11 476), and low GCS only (n = 4126). A total of 20 731 patients had a level 1 activation with 43 285 patients having a level 2 activation. The remaining patients had either a lower activation level or were only seen in consultation.

Table 1 shows a breakdown of patient cohorts by the risk indicators they met. Overall, 0.70% of patients (n = 1336/198 211) had a BCVI. Blunt cerebrovascular injury was most common among patients with a low GCS and a significant ATC injury at 4.02% (n = 461/11 476), followed by significant ATC injury only at 1.5% (n = 914/59 236). Of the 125 296 patients with no indicators, only two had a BVCI (0.001%). The most common BCVI injuries were vertebral artery (n = 841/1336, 63.0%) and carotid artery (n = 403/1,336, 30.2%), with the remaining patients having a combination of the two. The presence of a BCVI indicator was more common for patients with higher trauma activation levels. For example, for level 1 activations, 71.3% of patients had at least one BCVI indicator, while for level 3 activations, only 53.3% of patients had one.

Table 1.

BCVI Injuries by Criteria

Indicator definition	No BCVI (N)	BCVI Present (N)	Percentage with BCVI (%)
No criteria met	125 296	2	0.002
Low GCS only	4126	0	0
Low GCS + ATC injury	10 467	420	3.8
Significant ATC injury only	58 322	914	1.5
Total	198 211	1336	0.70

Abbreviations: ATC, above the clavicle; BCVI, blunt cerebrovascular injury; GCS, Glasgow Coma Scale.

In total, 1334 patients had a positive indicator and a BCVI, while 72 915 patients had a positive indicator and no BCVI (Table 2). Only two patients had no indicators and a BCVI, and 125 296 had no indicators and no BCVI. The calculated positive likelihood ratio for any risk criteria was 2.7, and the negative likelihood ratio for meeting no risk criteria was 0.002 (Figure 1).

Table 2.

Summary of BCVI Injuries

Indicator definition	No BCVI (N)	BCVI present (N)	Percentage with BCVI (%)
Any exposure (GCSm <6 or significant ATC injury)	72 915	1334	1.80
No risk criteria met	125 296	2	0.001

Abbreviations: ATC, above the clavicle; BCVI, blunt cerebrovascular injury; GCSm, Glasgow Coma Scale—motor component.

Figure 1.

Log-transformed likelihood ratio of BCVI based on exposure to any clinical criteria versus no criteria

Based on multivariate logistic regression analysis, having an ATC injury was a perfect predictor of having a BCVI because very few patients without an ATC injury had a BCVI. Therefore, we only included the GCS indicator and a level 1 activation in our final model. The adjusted odds ratio for having a BCVI among patients who had a “Low GCS” indicator was 3.0 (95% CI: 2.6, 3.5), and for a level 1 activation, it was 2.8 (95% CI: 2.4, 3.2).

Discussion

In this study of North Carolina trauma patients, we showed that BCVI was extremely rare in patients without significant injury above the clavicles or in those patients with a normal GCS. For patients who suffered a BCVI, each of the combinations most associated with a BCVI included a significant injury above the clavicle. These data demonstrate the importance of ruling out BCVI in high-risk patients but show that BCVI is very rare in a subset of patients with a low-risk injury pattern.

The debate over universal screening for BCVI has been contentious in the United States. The prevalence of BCVI is relatively low in the US trauma population (0.5-2%) but is associated with substantial morbidity and mortality and is often treatable with timely diagnosis.^8,9 Consequently, with the near-universal use of CT imaging for trauma patients, there has been a reexamination of whether routine injury screening should include contrasted neck imaging to evaluate for BCVI. In 1999, the Denver Criteria were introduced, focusing on identifying high-risk patients to optimize resource utilization, particularly at lower-volume trauma centers.^6,10 The Denver Criteria were expanded in 2012 due to data that the original criteria were missing too many injuries, with some data suggesting that the original screening guidelines missed as many as 20-30% of patients.^8,11

Several studies over the last decade have demonstrated that expanded criteria are necessary to capture these injuries adequately. Two single-institution studies demonstrated that the incidence of diagnosed BVCI increased significantly after the implementation of expanded screening protocols.^12,13 For example, a 2016 study by Geddes et al reported a 2.36% incidence of BVCI during the pre-expansion period and a 2.99% incidence after expansion. ¹² A 2018 review of the National Trauma Data Bank (NDTB) compared the incidence of BCVI in patients who presented pre- and post-expansion of the Denver Criteria. ⁴ The authors found that the incidence of diagnosed BCVI increased significantly (0.19% vs 0.22%, P < .001) after criteria expansion. Similarly, a 2021 study of a national Japanese trauma registry found that broad screening was necessary to achieve a reasonable CTA sensitivity for BCVI. ¹⁴ Like these investigations, our study approximates the expanded criteria by adding significant injuries above the clavicle to the traditional Denver Criteria.

In contrast, other retrospective data have suggested that following the eDC is insufficient. A 2020 study from Virginia evaluated 4659 trauma patients over 2 years, during which all blunt trauma patients received CTA of the neck. They assessed the diagnostic utility of the eDC and ACS TQIP best practice guidelines and found that both were inadequate, missing nearly 20% of patients with a BCVI, including many with higher-grade injuries. ¹⁵ A similar retrospective study from Alabama examined 6800 adult trauma patients and found that only 50% or fewer of patients with BCVI would be appropriately screened based on existing guidelines. ¹⁶ In 2022, Harper et al showed that 16% of patients with BCVI at their institution did not have any Denver Criteria indicators. ¹⁷ All three studies concluded that universal screening was necessary, given the risk of missing injuries.

While our data joins a conflicting body of literature on BCVI screening, the data quality is generally poor, with virtually all data based on retrospective studies. Consequently, the decision to adopt universal screening is complex due to these conflicting data and the potential strain on health care resources. Recent evidence is limited, but studies that use CMS reimbursement data for neck CTA report costs of approximately $700-$900 per study,^18,19 while patient charges range from $3700 to $8000.^5,6 There is also the time “cost” for the radiologist to interpret an additional study. Implementing universal screening would significantly increase radiology interpretation time and patient charges at high-volume centers, leading to one recent neuroradiology study questioning the utility of screening all low-risk patients. ²⁰

However, a recent cost-effectiveness analysis of universal screening provided evidence that universal screening may be the most cost-effective strategy compared to the eDC, especially when the incidence of BCVI is high (greater than 6%). ⁷ However, the strategies are more equivocal when the incidence is closer to accepted reports of 2-3% nationally.

At the very least, our study and previous reports support using broad criteria for BCVI screening. Many studies, including ours, have examined the effectiveness of a specific screening protocol using retrospective injury data from CT imaging following the initial assessment. The burden of BCVI screening increases for the patient after the initial CT scan, as it requires an extra dose of intravenous contrast, which is not needed when performed with standard trauma protocol imaging. ²¹ Consequently, clinicians need guidelines on findings available during their initial trauma assessment. The most evident association with BCVI in this study was injuries above the clavicle, a finding consistent with other studies. ²² Ultimately, our report emphasizes that trauma centers would benefit from prospective risk stratification data to better characterize what predictors are most helpful.

This study is limited by factors inherent to the use of a large, statewide database. The NC Trauma Registry (NCTR) only collects data from registered trauma centers, excluding over 100 non-trauma hospitals in the state. In addition, the retrospective design relies on the accuracy of inputted registry data, which are subject to coding errors and lack post-hospital patient data, including clinical outcomes.

Conclusion

Patients without altered mental status or significant injuries above the clavicle had an extremely low risk of BCVI. Although liberal screening strategies remain important to prevent missed injuries, our findings suggest that a clearly defined low-risk subgroup may not require routine CTA screening. Prospective, multicenter validation of contemporary screening criteria is needed to refine risk stratification and optimize cross-sectional imaging utilization while maintaining patient safety.