Derivation of a clinical prediction score for the diagnosis of clinically significant symptomatic carotid artery disease.

Abstract

Emergent vascular imaging identifies a subset of patients requiring immediate specialized care (i.e. carotid stenosis > 50%, dissection or free-floating thrombus). However, most TIA patients do not have these findings, so it is inefficient to image all TIA patients in crowded emergency departments (ED). Our objectives were to derive and internally validate a clinical prediction score for clinically significant carotid arterydisease in TIA patients. This was a planned secondary analysis of a prospective cohort study from 14 Canadian EDs. Among 11555 consecutive adult ED patients with TIA/minor stroke symptoms over 12 years, 9882 had vascular imaging and were included in the analysis. Our main outcome was clinically significant carotid artery disease, defined as extracranial internal carotid stenosis ≥ 50%, dissection, or thrombus in the internal carotid artery, with contralateral symptoms. Of 9882 patients, 888 (9.0%) had clinically significant carotid artery disease. Logistic regression was used to derive a 13-variable reduced model. We simplified the model into a score (Symcard [Symptomatic carotid artery disease] Score), with suggested cut-points for high, medium, and low-risk stratification. A substantial portion (38%) of patients were classified as low-risk, 33.8% as medium risk, and 28.2% as high risk. At the low-risk cut-point, sensitivity was 92.9%, specificity 41.1%, and diagnostic yield 1.7%. This simple score can predict carotid artery disease in TIA patients using readily available information. It identifies low-risk patients who can defer vascular imaging to an outpatient or specialty clinic setting. Medium-risk patients may undergo imaging immediately or with slight delay, depending on local resources. High-risk patients should undergo urgent vascular imaging.