Objectives This study aimed to determine the history features, signs and symptoms identified during a code stroke that correlate with the diagnosis of ischemic stroke or transient ischemic attack (TIA). We investigated the rate of stroke mimics and aimed to develop a clinical prediction model.

Methods Consecutive code stroke presentations to a primary stroke centre were recruited. Patient characteristics, medical history, signs or symptoms on activation of code stroke were collected from the medical record. Diagnosis of ischemic stroke was determined by radiographic evidence of infarction. Univariate analysis and multivariable logistic regression analysis were used to determine the features that predict ischemic stroke/TIA versus mimic.

Results Among 493 code strokes, 64.5% were mimics. The most commonly diagnosed mimics were migraine, peripheral vertigo and seizure. Upper limb sensory change (OR 3.27 [95% CI, 1.75-6.11]), hemiplegia (OR 2.70 [95% CI, 1.65-4.43]), dysphasia (OR 2.62 [95% CI, 1.56-4.40]) and history of atrial fibrillation (OR 2.01 [95% CI, 1.14-3.54]) or hypertension (OR 1.77 [95% CI, 1.10-2.83]) are highly predictive of stroke/TIA. Headache (OR 0.40 [95% CI, 0.23-0.69]) is predictive of a mimic. Dizziness and vertigo were more common in stroke mimics. C-statistic for the study models ranged from 0.70 to 0.76.

Conclusion Objective signs such as unilateral motor weakness and dysphasia are highly predictive of ischemic stroke/TIA whereas symptoms of headache and dizziness are suggestive of stroke mimic. Stroke mimic rate is influenced by local prevalence and threshold for code stroke activation. Incorporating positive and negative predictive features may improve future stroke prediction tools.