Likelihood and Predictors of Acute Findings on CTA Head and Neck for Neurologic Presentations to the Emergency Department

Abstract

Background and Objectives

CT angiography of the head and neck (CTAHN) is commonly obtained in the emergency department (ED) for patients with acute neurologic symptoms to guide interventions. There are limited data on which clinical characteristics predict these changes. The primary objective of this study was to assess the yield of CTAHN across common neurologic ED presentations and identify patient and clinical factors associated with CTAHN findings that result in acute management change.

Methods

We performed a retrospective cohort study of 1,445 ED patients at an academic Level 1 trauma center who underwent CTAHN in 2023 for 5 common neurologic presentations: stroke code, headache, dizziness, altered mental status, and vision changes. Multivariable logistic regression models with LASSO variable selection were used to identify predictors of acute management change. Odds ratios (OR), 95% CIs, and p values were reported for each independent predictor.

Results

Overall, 216 of 1,445 CTA head and neck sudies (CTAHNs; 14.9%) resulted in acute management change. The highest yield of CTAHN acute management change was for stroke codes (21.2%) and vision changes (14.7%), and the lowest for dizziness (4.0%). For stroke codes, independent predictors included abnormal CT head (OR 2.84, p = 0.0001), focal neurologic deficit (OR 2.2, p = 0.0119), new vs recurring symptoms (OR 2.39, p = 0.0272), maximal symptom severity at onset (OR 3.95, p = 0.0011), and tobacco use (OR 2.02, p = 0.0016). National Institutes of Health Stroke Scale (NIHSS) scores of 10–14 had the highest likelihood of CTAHN acute management change. Previous neurologic disorder was protective (OR 0.66, p = 0.0317). For headache, abnormal CT head (OR 4.98, p = 0.0303) and thunderclap presentation (OR 5.31, p = 0.0276) were associated with CTAHN acute management change. Only univariate trends could be identified for dizziness, altered mental status, and vision changes. CTAHN was most useful in those presenting with vision loss or diplopia, but not cases of blurry vision or positive visual phenomena. Abnormalities on the preceding CT head increased the yield for altered mental status, while CTAHNs were low yield for isolated dizziness.

Discussion

These results highlight specific, obtainable clinical predictors that can guide judicious use of CTAHN in the ED. Recognizing these predictors can inform emergency physicians' and neurologists' clinical judgment, optimize imaging utilization, improve resource allocation, and reduce patient burden.

Introduction

CT angiography of the head and neck (CTAHN) is frequently used as a diagnostic tool in the emergency department (ED). Rapid advancements in CTAHN technology have led to its increased clinical application, exemplified by a 536.4% increase in utilization of CTAHNs between 2014 and 2021 at a Level 1 center, demonstrating the greatest percentage increase over time among all CT modalities.¹ While overall stroke incidence decreased from the early 1990s to 2015,² the proportion of CTAHN studies for ischemic stroke workup in 1 study increased by approximately 140% from 2006 to 2010.³ Despite their diagnostic utility, judicious utilization of CTAHN studies is critical because they are not without risk. CTAHN studies can overestimate the degree of vascular stenosis,⁴ increase the risk of radiation-induced carcinogenesis,⁵ contribute to contrast-induced nephropathy,⁶ add financial burden to the patient, and prolong occupancy of the CT scanner for image rendering and evaluation,⁷ resulting in increased patient wait times and delays in care.

Measuring CTAHN impact across different clinical concerns and identifying patient risk factors that contribute to a higher likelihood of impactful CTAHNs can provide insight into more streamlined ordering practices. In this study, we define significant CTAHN impact as leading to an acute change in management related to the primary presenting symptoms, thereby separating incidental findings on CTAHN—which are frequently detected and may warrant further outpatient workup⁸—from higher acuity findings within the relevant vascular distribution for symptoms. These findings are deemed “CTAHN acute management change.” Five chief concerns were assessed: stroke code, dizziness, headache, altered mental status, and vision changes. These were selected based on their frequent presentation at our ED, large volume of CTAHN studies ordered for them,⁹ and high likelihood of neurologic consultation by the ED for assistance in management. Patient clinical characteristics were then evaluated to determine their relative contributions toward a higher likelihood of CTAHN acute management change, with the ultimate goal of contributing to a predictive model for high-yield CTAHN ordering practices by ED and neurology providers for common neurologic indications.

Methods

Study Setting

Adult patients (≥18 years old) presenting to an academic Level 1 trauma center ED who underwent CTA of the head and neck (CTAHN) between January 1, 2023 and December 31, 2023 were included. Patients younger than 18 years or undergoing CT as part of trauma assessment were excluded.

Data Collection

Relevant data were extracted from the electronic medical record (EPIC systems) using the SlicerDicer tool. Study data were collected and managed using REDCap (Research Electronic Data Capture) hosted at Duke University.¹⁰ REDCap is a secure, web-based application designed to support data capture for research studies, providing (1) an intuitive interface for validated data entry; (2) audit trails for tracking data manipulation and export procedures; (3) automated export procedures for seamless data downloads to common statistical packages; and (4) procedures for importing data from external sources.

CTAHN Acute Management Change

CTAHN acute management change was defined as imaging findings related to the indication for study that led to an acute clinical intervention. Examples included thrombolytics or thrombectomy decisions for severe stenosis correlating with symptoms, initiation of dual antiplatelet therapy for large artery atherosclerosis, carotid endarterectomy for moderate-to-high-grade stenosis, or management of stuttering symptoms due to hypoperfusion. Incidental or clinically insignificant findings (e.g., mild atheromatous plaque, ipsilateral/nonrelevant vascular changes, or non-neurologic findings such as lung nodules) were excluded. Radiologist interpretations were used to identify findings, and ED and Neurology documentation was used to confirm resulting management changes.

Statistical Analysis

Patients with non-neurologic indications for CTAHN, such as trauma, were excluded. Patient demographic and clinical characteristics were obtained and reported as continuous and categorical variables. The 5 most frequent neurologic indications for which a CTAHN was obtained were further analyzed. These indications included stroke codes, vision changes, altered mental status, dizziness, and headache. When there were multiple encounters for an individual patient within the time frame studied, only the first encounter was used for analysis. Patient encounters where multiple indications were cited were incorporated into the statistical analysis for the first reported indication.

We first analyzed counts (percentages) of positive CTAHN acute management change for each chief concern. Clinical characteristics evaluated as potential predictors included medical history, previous neuroimaging (CT, CTAHN, MRI, and ultrasound) abnormalities per radiologist impression, history of stroke or other neurologic disorders, tobacco or substance use, abnormal CT head findings (acute, age-indeterminate, or chronic), focal neurologic deficits, symptom onset pattern (maximal severity vs progressive), and select vitals. Chronic white matter changes and mild atherosclerosis were considered normal because of their high prevalence in the general population without meaningful clinical significance.

A LASSO regression model,¹¹ including variables from univariate and multivariate associations, was fit to perform variable selection with CTAHN acute management change as the outcome. The goal of the LASSO model is to perform a linear regression that shrinks the coefficients of less important variables toward zero, thereby increasing accuracy of the model, preventing overfitting, and allowing for better generalizability of findings. The optimal model had the smallest Akaike information criterion. The variables selected by the LASSO model were fit in multivariable regression models. These models were fit assuming that there was not a large amount of missing data. Adjustments for multiple testing were made by indication. Firth penalized logistic regression was used to mitigate bias when dealing with small sample sizes.¹² Odds ratios (ORs), 95% CIs, and p values are reported for all independent variables. National Institutes of Health Stroke Scale (NIHSS) distributions for stroke code patients were visualized with bar graphs, and descriptive analyses were performed for select subsets using Microsoft Excel. All statistical analyses were conducted in SAS 9.4 (SAS Institute, Cary, NC) at a significance level of 0.05.

Standard Protocol Approvals, Registrations, and Patient Consents

This retrospective chart review was HIPAA-compliant and approved by the Duke Health Institutional Review Board, with waiver of written consent.

Data Availability

Anonymized data not published within this article will be made available by request from qualified investigators.

Results

Demographic and Clinical Characteristics

We studied 1,445 patients with a mean age of 62.7 years (SD 16.5). Most were female (54.2%), White (52.9%), and non-Hispanic (90.4%). Detailed baseline clinical characteristics are shown in eTable 1. Most presented as a stroke code (54.1%), which was also the most common indication for encounters with acute management change (76.9%) (Table 1).

Table 1.

CTAHN Indications

	No CTAHN acute management change (N = 1,229)	CTAHN acute management change (N = 216)	Total (N = 1,445)
CTAHN indication (%)
Stroke code	616 (50.1)	166 (76.9)	782 (54.1)
Headache	152 (12.4)	14 (6.5)	166 (11.5)
Dizziness	213 (17.3)	9 (4.2)	222 (15.4)
Altered mental status	78 (6.3)	11 (5.1)	89 (6.2)
Vision changes	81 (6.6)	14 (6.5)	95 (6.6)
Others	89 (7.2)	2 (0.9)	91 (6.3)

Abbreviation: CTAHN = CT angiography of the head and neck.

Stroke Code

Table 2A provides the results from the multivariable logistic regression model for the association between clinical characteristics selected by the LASSO model and CTAHN acute management change for stroke patients. In multivariable analysis, abnormal CTH findings were strongly associated with acute management change (OR 2.84, p = 0.0001). Further analysis using a simple regression analysis, in which age-indeterminate CTH findings served as the reference group for comparison with acute CTH findings and chronic CTH findings, revealed that acute CTH findings had the strongest association with CTAHN acute management change (eTable 2). Focal neurologic deficits (OR 2.2, p = 0.012), new symptoms (OR 2.39, p = 0.027), maximal severity at onset (OR 3.95, p = 0.0011), and tobacco use (OR 2.02, p = 0.0016) were also significant. A previous neurologic disorder was associated with lower odds (OR 0.56, p = 0.032).

Table 2.

Multivariate Associations Between Clinical Characteristics Selected by the LASSO Model and CTAHN Acute Management Change

A. Stroke codes
Variable	OR (95% CI)	p Value
Abnormal CTH, yes vs no	2.84 (1.81–4.45)	<0.0001
Atrial fibrillation, yes vs no	1.42 (0.75–2.7)	0.2798
History of diabetes, yes vs no	1.4 (0.89–2.2)	0.1434
Elevated HR (>100 bpm) at presentation, yes vs no	1.27 (0.7–2.28)	0.4289
History of hyperlipidemia, yes vs no	1.22 (0.78–1.92)	0.3864
HTN (>140/90) at presentation, yes vs no	1.42 (0.88–2.28)	0.1522
Focal neurologic deficit present, yes vs no	2.2 (1.19–4.05)	0.0119
Previous diagnosis of a neurologic disorder, yes vs no	0.56 (0.33–0.95)	0.0317
History of stroke, yes vs no	0.77 (0.43–1.38)	0.3824
New or recurring symptoms, new vs recurring	2.39 (1.1–5.18)	0.0272
Maximal severity at onset or progressive, severe onset vs progressive	3.95 (1.73–9)	0.0011
Tobacco use, yes vs no	2.02 (1.31–3.13)	0.0016

B. Headaches
Clinical characteristic	Or (95% CI)a	p Value
Abnormal CTH, yes vs no	4.98 (1.17–21.29)	0.0303
History of atrial fibrillation, yes vs no	0.63 (0.03–11.59)	0.7580
History of diabetes, yes vs no	0.26 (0.03–2.72)	0.2611
History of hyperlipidemia, yes vs no	0.06 (0–0.76)	0.0304
History of hypertension, yes vs no	5.2 (1.04–25.91)	0.0441
Elevated HR (>100 bpm) at presentation, yes vs no	2.1 (0.35–12.73)	0.4206
HTN (>140/90) at presentation, yes vs no	0.27 (0.06–1.15)	0.0757
Previous neuroimaging abnormality, yes vs no	1.49 (0.31–7.07)	0.6150
Focal neurologic deficit present, yes vs no	1.16 (0.25–5.3)	0.8507
Previous diagnosis of a neurologic disorder, yes vs no	2.98 (0.63–14.04)	0.1679
Other substance use, yes vs no	1.23 (0.18–8.57)	0.8375
New or recurring symptoms, new vs recurring	1.75 (0.33–9.29)	0.5118
Thunderclap or progressive, severe onset vs progressive	5.32 (1.2–23.51)	0.0276
Tobacco use, yes vs no	3.12 (0.78–12.54)	0.1093

Abbreviations: CTAHN = CT angiography of the head and neck; LASSO = least absolute shrinkage and selection operator; HTN = hypertension.

^aEstimates computed with the Firth penalized likelihood method.

The Figure shows NIHSS score distribution by acute management change among stroke code patients. NIHSS scores of 10–14 had the highest proportion with acute management change (39.5%, p < 0.0001). Among 136 patients with NIHSS score 0, descriptive analysis revealed that 18 had acute management change, 6 due to monocular vision loss with ipsilateral internal carotid disease, 8 with transient or stuttering symptoms that had resolved at the time of examination, and 4 with posterior circulation presentations not captured by the NIHSS.

Figure. Proportion of CTAHN Acute Management Change by NIHSS Score.

Stacked bar chart showing the proportion of patients with and without CTAHN acute management change based on the NIHSS score. The likelihood of CTAHN acute management significantly varies depending on the NIHSS group (p value < 0.0001). CTAHN = CT angiography of the head and neck; NIHSS = National Institutes of Health Stroke Scale.

Headache

Table 2B provides the results from the multivariable logistic regression model for the association between clinical characteristics selected by the LASSO model and CTAHN acute management change for patients with headache using the Firth penalized likelihood method. Among patients with headache, acute CTH findings (OR 4.98, p = 0.030), thunderclap onset (OR 5.31, p = 0.028), and history of hypertension (OR 5.2, p = 0.044) were associated with acute management change, whereas hyperlipidemia was inversely associated (OR 0.06, p = 0.030).

Dizziness, Altered Mental Status, and Vision Changes

There are no results from a model with variables selected by the LASSO model because an intercept-only model met the optimal criteria for the dizziness, altered mental status, and vision changes cohorts. Variables included in the LASSO model selection are listed in eTables 3–13.

Descriptive analysis was performed to better understand the subset of patients who presented with dizziness and CTAHN acute management (N = 9). In 2 patients, concerning findings on the CTAHN prompted further imaging and were ultimately deemed to be unremarkable or insignificant. In 3 patients, other presenting symptoms alongside dizziness that localized to the anterior circulation warranted intervention within the anterior circulation vasculature on CTAHN (i.e., high-grade stenosis and aneurysm). In 2 patients, recurrent episodes of dizziness in the presence of posterior circulation stenosis raised concern for TIAs for which pharmacologic treatment was initiated. In 1 patient, a posterior fossa dural arteriovenous (AV) fistula was seen, which prompted surgical intervention.

Descriptive analysis was also performed to further stratify presenting symptoms in the vision changes patient subset who had CTAHN acute management change. Of 14 patients, 12 of them presented with vision loss while the other 2 presented with double vision. No patients presenting with blurry vision or positive visual phenomena had CTAHN acute management change.

Discussion

The objective of a stroke code is to determine candidacy for thrombolytics or mechanical thrombectomy. Thereafter, pharmacologic therapy and carotid surgical intervention may be considered. CTAHN is a useful imaging technique for assistance in acute decision making. Between January 2014 and December 2019, there was a tenfold rise in the yearly number of stroke code CTAHNs, but only 1.3% of patients underwent a carotid surgical procedure (endarterectomy or stenting) and only 0.9% were transferred to a stroke center for consideration of endovascular therapy.¹³

The American College of Radiology (ACR) Appropriateness Criteria for cerebrovascular disease define scenarios in which CTAHN is “usually appropriate” (e.g., carotid bruit, TIA, new focal deficit, subarachnoid hemorrhage, aneurysm follow-up, vascular malformation, venous sinus thrombosis, and CNS vasculitis).¹⁴ A retrospective study showed a higher yield when CTA was ordered under the “usually appropriate” indication compared with the “may be appropriate” and “usually not appropriate” groups.¹⁵ However, these criteria do not specify appropriateness in the acute stroke code setting, where patient history, examination, vitals, and preceding CT head are critical.

In our study, 166 of 782 stroke code CTAHNs (∼21.2%) demonstrated CTAHN acute management change. Abnormal CT head nearly tripled the likelihood, focal neurologic deficit doubled it, new vs recurrent symptoms doubled it, maximal severity of symptoms at onset increased it fourfold, and tobacco use doubled it (Table 2A). Acute CT head findings had the strongest association compared with chronic and age-indeterminate findings (eTable 2). These features should increase suspicion for actionable CTAHN findings, including thrombolysis eligibility, large vessel occlusion, intracranial atherosclerosis, or vascular malformations. By contrast, a previous neurologic disorder was independently associated with 44% lower odds of acute management change, likely reflecting stroke mimics (e.g., migraine and seizure with Todd paralysis), although clinical acumen should be used in these instances. Other clinical history (atrial fibrillation, diabetes, hyperlipidemia, hypertension, or previous stroke) did not meaningfully change likelihood (Table 2A).

NIHSS scores also correlated with CTAHN yield (Figure). Scores of 10–14 had the highest rates of acute management change, whereas very low (0–4) or very high (≥25) scores had the lowest. Yield for very high scores may be confounded by patients with altered mental status who were unable to comply with NIHSS testing, resulting in high scores. A descriptive analysis of patients with NIHSS score 0 revealed cases with monocular vision loss, resolved or stuttering symptoms, or posterior circulation symptoms, which still led to CTAHN acute management change. Thus, although extremes of NIHSS are less predictive, select patients with NIHSS score 0 warrant CTAHN.

CTAHN use for headache has risen substantially in recent years despite declining positivity rates.¹⁶ In 1 study, only 54 of 729 patients with acute severe headache who had a normal neurologic examination and CT head had a vascular abnormality, many of which were incidental aneurysms.¹⁷ This highlights the need for clearer ordering parameters.

In our cohort, 14 of 166 headache CTAHNs (∼8.4%) led to acute management change (Table 1). Abnormal CT head, thunderclap onset, and history of hypertension each increased the likelihood fivefold, whereas hyperlipidemia was associated with a markedly lower likelihood (∼94% lower odds) (Table 2B). These findings may reflect acute parenchymal changes secondary to vascular abnormalities such as aneurysm formation and rupture in the setting of chronic hypertension^18,19 vs hyperlipidemia's association with nonvascular headache syndromes.²⁰ Other clinical features were not associated with CTAHN acute management change (Table 2B).

Previous studies suggest limited diagnostic value of CTAHN in isolated dizziness, with low yield and less cost-effectiveness compared with conventional MRI.^21-23 In our study, only 9 of 222 dizziness CTAHNs (∼4.0%) led to acute management change. Of these, acute management change consisted of further imaging ultimately found to be unremarkable, anterior circulation intervention for confounding presenting symptoms (double vision, weakness, etc.), pharmacologic treatment of presumed recurrent TIAs with posterior circulation stenosis, and 1 case of intervenable posterior fossa dural AV fistula. Therefore, the percentage of cases of acute management change attributable to primary dizziness is even lower. No clinical variables reliably predicted yield (eTables 8 and 9). Existing decision rules may help reduce low-yield CTA use,²⁴ and further predictive models would be beneficial.

Altered mental status (AMS) often has non-neurologic causes, and the yield of CT head in this setting is low.²⁵ For CTAHN, previous studies reported that approximately one-third resulted in nonroutine result communication, although many were incidental.⁹

In our study, 11 of 89 AMS CTAHNs (∼12%) led to acute management change. Abnormal CT head was the strongest predictor (10x likelihood on univariate analysis; eTable 10). Of the 11 patients who had CTAHN acute management change, 4 had acute CTH abnormalities, 2 had age-indeterminate CTH abnormalities, and 3 had only chronic CTH abnormalities, likely representing patients with multivascular ischemia, hemorrhage, mass effect, or other parenchymal or intraventricular injuries reflected clinically by altered mental status.²⁶ Other clinical variables were not predictive (eTables 10 and 11).

CTAHN for visual symptoms has been less studied, although angiographic modalities have demonstrated utility for carotid stenosis in central retinal artery occlusion.^27-30 One study showed that altered vision, visual field irregularities on physical examination, abnormal pupil reactivity, or abnormal optic discs were positive risk indicators for abnormal neuroimaging, although CTAHN was not independently assessed.³¹

In our cohort, 14 of 95 vision change CTAHNs (∼14.7%) led to acute management change. Of these, 12 were in patients with vision loss and 2 with double vision. There were none for patients with blurry vision or positive visual phenomena, supporting a lower likelihood of vascular contributors in these cases. Previous neuroimaging abnormality increased yield sixfold and diabetes nearly fourfold, while other variables were not predictive (eTables 12 and 13).

Several limitations were identified in the acquisition of these data. Data were gathered from a single academic, tertiary care center, which may not be representative of other patient populations. Our study relied on a retrospective appraisal of clinical notes, in which documentation may not be consistent with the patient's true clinical history or presentation. We relied on impressions of imaging studies by the reading radiologists, and these images were not manually reviewed by us. Lack of LASSO models for the altered mental status, dizziness, and vision change subsets requires judicious interpretation of findings because univariate associations are more likely to be correlative in nature and confounded by other variables.

Unlike previous studies evaluating nonroutine result communication for CTAHN in the ED,⁹ we narrowed positivity by excluding incidental findings and focusing on acute, clinically relevant findings that led to management change. However, scenarios existed in which CTAHN prompted further diagnostics (e.g., MRI) that ultimately revealed no significant pathology. Therefore, some CTAHN acute management changes may represent false positives.

CTAHN studies should be strongly considered in the stroke code setting. Clinicians should particularly note new rather than recurring symptoms, maximal severity at onset, focal neurologic deficits, NIHSS scores 10–14, acute findings on preceding CT head, and patient tobacco use, whereas symptoms attributable to a previous neurologic disorder may warrant deferring CTAHN. For headache in the ED, thunderclap presentation, history of hypertension, and acute CT abnormalities increase CTAHN acute management change pretest probability. CTAHN is not generally recommended for isolated dizziness, although cases with additional neurologic deficits may warrant imaging. In altered mental status, CTAHN should be considered when preceding abnormalities on the CT head are identified. For vision changes, CTAHN is justified in cases of vision loss or diplopia, especially with previous abnormal neuroimaging or patient diabetes, but it is not indicated for blurry vision or positive visual phenomena.

TAKE-HOME POINTS

• → Specific clinical factors significantly predict relevant, acute findings on stroke code CT angiography head and neck studies (CTAHN), namely acute findings on the preceding CT head scan, focal neurologic deficits, new or maximally severe symptoms at onset, and tobacco use.

• → National Institutes of Health Stroke Scale (NIHSS) scores between 10 and 24 in the stroke code setting had the highest yield of acute CTAHN findings, and NIHSS scores of 0 still yielded positives on the CTAHN in cases of monocular vision loss, stuttering symptoms, and posterior circulation signs.

• → CTAHN studies were useful for cases of altered mental status when acute CTH findings were present, and only vision loss or diplopia led to actionable CTAHN findings while blurry vision or positive visual phenomena did not.

• → CTAHN studies were not useful in cases of isolated dizziness at the emergency department.

• → Clinical judgment guided by targeted patient factors can improve CTAHN study yield and reduce unnecessary scans.

Author Contributions

R. Gaini: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data. S.L. Reyes: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design. T. Muqeem: drafting/revision of the manuscript for content, including medical writing for content; study concept or design. S. Unnithan: drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data. H. Al-Khalidi: drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data. P.G. Kranz: drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data. J.C. Morgenlander: drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data.

Study Funding

This publication was made possible (in part) through financial support from the Donald B. Sanders Residents and Fellows Research Fund. The Duke BERD Methods Core's support of this project was made possible (in part) under Grant Number UL1TR002553 from the National Center for Advancing Translational Sciences (NCATS) of the NIH, and NIH Roadmap for Medical Research. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of NCATS or NIH.

Disclosure

The authors report no relevant disclosures. Full disclosure form information provided by the authors is available with the full text of this article at Neurology.org/cp.