This was a single-center retrospective observational study of code stroke patients in the ED from January to December 2019. It was conducted in a tertiary academic hospital with approximately 70,000 annual ED visits.

Fig. 1 illustrates the flow of the code stroke protocol. First, a nurse obtained the patient’s medical history, performed an essential neurological examination to help classify the patient’s condition, and informed the emergency physician that there was a possibility of acute stroke. The emergency physician then confirmed the medical history and performed a neurological examination. If there was a possibility of acute stroke, the emergency physician activated a code stroke. For acute ischemic stroke, the principle of treatment is intravenous injection of a fibrinolytic drug (tPA) within 4.5 hours of the onset of symptoms. However, as the effectiveness of endovascular recanalization therapy using a stent retriever has been demonstrated in large-scale randomized trials, treatment guidelines have been changed such that endovascular recanalization can be attempted up to 24 hours after the onset of stroke symptoms.141516 Accordingly, a code stroke was applied to all patients who visited the emergency room within 24 hours after the onset of stroke symptoms. For patients who recognized symptoms after waking up, a code stroke was activated in those who visited the hospital within 24 hours after falling asleep or within 24 hours of the last time they felt normal. If the time of symptom onset or when they last felt normal was unclear, it was determined by the judgment of the emergency physician whether to activate the code stroke. Patients under the age of 18 years were excluded from the study.

We used data from a single-center prospective stroke registry that included all code stroke patients treated in the ED. Additionally, we reviewed the medical records and analyzed the data retrospectively. Age, sex, past medical history, vital signs (blood pressure, heart rate, respiratory rate, body temperature) at the time of visit, clinical symptoms, last known normal time, symptom onset time, time interval between symptom onset and ED visit, National Institutes of Health Stroke Scale (NIHSS) score, and use of tPA were investigated. A neurological specialist applied the NIHSS after activation of code stroke. If there was no neurological specialist available, a neurological resident performed it. Neurological symptoms are defined as only those symptoms described by the patient or caregiver during history taking, and focal neurological deficits are defined as signs confirmed by the NIHSS. The final diagnosis was based on the primary diagnosis at discharge as identified from the medical record. After all tests, including brain CT and MRI, were completed, a neurosurgeon or neurologist diagnosed whether a stroke had occurred. If the condition was not stroke, the final diagnosis was made after consultation with all relevant departments, including emergency medicine. Cases of cerebral infarction, cerebral hemorrhage, and transient ischemic attack (TIA) were defined as stroke, and all other cases were defined as stroke mimics. Based on the final diagnosis, the patients were classified into a stroke group and a stroke mimic group. The clinical and neurological characteristics of the two groups were compared and analyzed.

We utilized IBM SPSS Statistics version 20.0 (IBM Corp., Armonk, NY, USA) for all statistical methods. If the continuous variables in two groups followed a normal distribution, the Pearson χ² test was used, and if not, the Mann-Whitney U test was used. Independence tests and trend analyses of categorical variables were performed using the χ² test, Fisher’s exact test, and linear-by-linear association. Multivariate logistic regression analysis was performed to identify the variables affecting stroke, and all variables with P < 0.05 in the univariate analysis were included in multivariate analysis. We also constructed a receiver operating characteristic (ROC) curve to evaluate the performance of the multivariable model and calculated the area under the curve (AUC) to estimate the predictive validity. Statistics are expressed as the mean ± SD, and if the data do not follow the normal distribution, the statistics are expressed as the median and interquartile range. In the comparative analysis of neurological tests, odds ratios (ORs) and 95% confidence intervals (CIs) are presented together. A result was considered statistically significant if its P value was less than 0.05.

The present study was approved by the Institutional Review Board (IRB) of Seoul National University Hospital (IRB No. 2101-034-1186). Informed consent was waived by the IRB due to the retrospective nature of the study.

A total of 409 code stroke patients presented to the ED during the study period. Of these, 284 (69.4%) had strokes, and 125 (30.6%) had stroke mimics. Of the 284 strokes, 189 (46.2%) were ischemic strokes, 47 (11.5%) were hemorrhagic strokes, and 47 (11.5%) were TIAs (Fig. 1). During the study period, none of the code stroke patients were under 18 years of age.

Table 1 describes the baseline characteristics. There was no significant difference in mean age between the two groups, but the number of patients over 50 years old was significantly higher in the stroke group than in the stroke mimic group (247 [87.0%] vs. 98 [78.4%], P = 0.028). The proportion of males was also higher in the stroke group (59.5% vs. 47.2%, P = 0.021). Stroke patients had a significantly more prominent history of hypertension, atrial fibrillation, and smoking, whereas stroke mimic patients had significantly higher cirrhosis, epilepsy, and psychiatric disorders (Table 1). The frequency of atrial fibrillation on the initial electrocardiogram was also higher in the stroke group (11.6% vs. 3.2%, P = 0.005). There were no significant differences in initial vital signs between the two groups except for blood pressure. Both systolic and diastolic blood pressure were significantly higher in the stroke group (Table 1).

The most common cause of stroke mimics was seizure (21.6%), followed by drug toxicity (12.0%), metabolic disorders (11.2%), peripheral vertigo (7.2%), psychological disorders (5.6%), and stroke sequelae (4.8%) (Table 2).

Table 3 presents a comparison of neurological symptoms between the two groups. Dizziness, altered mental status, seizure-like movements, and memory disturbance were significantly more common in the stroke mimic group than in the stroke group. On the other hand, dysarthria and hemiparesis were significantly more common in the stroke group. In addition, the number of neurological symptoms was associated with the likelihood of stroke. Among a set of seven focal neurological symptoms (unilateral limb weakness, unilateral limb sensory change, facial palsy, dysarthria, aphasia, diplopia, and visual disturbance), the code stroke patients showed an increased likelihood of stroke as the number of neurological symptoms increased (P < 0.001).

The initial NIHSS score was significantly higher in the stroke group than in the stroke mimic group (3.0 vs. 2.0, P = 0.003) (Table 4). Analysis of NIHSS scores that were classified into five ranges based on stroke severity demonstrated an increased likelihood of stroke as the NIHSS score increased (P = 0.005). For individual items in the NIHSS, oculomotor disorder, visual disturbance, facial palsy, upper limb monoparesis, hemiparesis, sensory change, dysarthria, and neglect were observed more frequently in the stroke group than in the stroke mimic group (Table 4).

Table 5 shows the number of focal neurological deficits observed in patients and their association with stroke. As in neurological symptoms, code stroke patients tended to increase the likelihood of a stroke as the number of accompanying focal neurological deficits increased among the seven specific neurological deficits (unilateral limb weakness [hemiparesis or upper limb monoparesis], unilateral limb sensory change, facial palsy, dysarthria, aphasia [or neglect], visual field defect, and eye movement disorder) (P < 0.001).

We performed multivariate logistic regression analyses to identify variables affecting stroke (Fig. 2). Consequently, psychiatric disorders, dizziness, altered mental status, and seizure-like movements were associated with stroke mimics. On the other hand, the stroke-related factors were current smoking, elevated systolic blood pressure, atrial fibrillation on the initial ECG, hemiplegia as a symptom, facial palsy as a sign, and an increased number of accompanying focal neurological deficits. On ROC curve analysis, the AUC for predicting stroke was 0.685 (95% CI, 0.637–0.730, P < 0.001) for the number of focal neurological deficits, 0.675 (95% CI, 0.627–0.720, P < 0.001) for facial palsy, and 0.651 (95% CI, 0.603–0.697, P < 0.001) for systolic blood pressure (Fig. 3). The cutoff value of the number of focal neurological deficits was 2 (sensitivity = 62.7%, specificity = 66.4%), and the cutoff value of systolic blood pressure was 153 mmHg (sensitivity = 61.6%, specificity = 63.2%).