An Intuitive Tool for Screening of Stroke Mimics in the Emergency Department
Background and Purpose—A large number of patients with symptoms of acute cerebral ischemia are stroke mimics (SMs). In this study, we sought to develop a scoring system (FABS) for screening and stratifying SM from acute cerebral ischemia and to identify patients who may require magnetic resonance imaging to confirm or refute a diagnosis of stroke in the emergency setting.
Methods—We designed a scoring system: FABS (6 variables with 1 point for each variable present): absence of Facial droop, negative history of Atrial fibrillation, Age <50 years, systolic Blood pressure <150 mm Hg at presentation, history of Seizures, and isolated Sensory symptoms without weakness at presentation. We evaluated consecutive patients with symptoms of acute cerebral ischemia and a negative head computed tomography for any acute finding within 4.5 hours after symptom onset in 2 tertiary care stroke centers for validation of FABS.
Results—A total of 784 patients (41% SMs) were evaluated. Receiver operating characteristic curve (C statistic, 0.95; 95% confidence interval [CI], 0.93–0.98) indicated that FABS≥3 could identify patients with SM with 90% sensitivity (95% CI, 86%–93%) and 91% specificity (95% CI, 88%–93%). The negative predictive value and positive predictive value were 93% (95% CI, 90%–95%) and 87% (95% CI, 83%–91%), respectively.
Conclusions—FABS seems to be reliable in stratifying SM from acute cerebral ischemia cases among patients in whom the head computed tomography was negative for any acute findings. It can help clinicians consider advanced imaging for further diagnosis.
The treatment decision for patients with acute stroke symptoms is based on pertinent history, brief neurological examination, and head computed tomography (CT). The differentiation of stroke mimics (SMs) from genuine cases of acute cerebral ischemia (ACI) is challenging, given the narrow time window for the administration of intravenous thrombolysis (intravenous tissue-type plasminogen activator [IV-tPA]). In this setting, patients with SM may mistakenly receive IV-tPA. Various surveys have demonstrated that the prevalence of SM among patients presenting to the emergency department with acute stroke symptoms can be as high as 30%.1–5 Administration of IV-tPA to SMs, although relatively safe, is unnecessary and costly, with ≈$5400 increase in total direct and indirect, actual hospital cost.6 There have been several studies to identify the variables that can differentiate ACI from SM1,6–10 (Table I in the online-only Data Supplement). Chang et al8 conducted a retrospective analysis and concluded that patients with symptoms of ACI are more likely to be >50 years of age, have evidence of atherosclerosis on CT angiography, have a history of atrial fibrillation and focal weakness, whereas history of migraine and seizures and presentation of isolated numbness favored a diagnosis of SM.
Several scales have been developed over the years to differentiate strokes from mimics in prehospital and emergency room settings. The Face Arm Speech Test,11,12 the Cincinnati Pre-hospital Stroke Scale,13 and the Los Angeles prehospital stroke screen14 scale are stroke diagnostic tools for ambulance paramedics. Although they have a consistent diagnostic accuracy between 80% and 95%, they have been designed for use by ambulance paramedics in the prehospital setting. On the contrary, the Recognition of Stroke in the Emergency Room (ROSIER) scale, developed by Nor et al15 in the United Kingdom, has been designed for emergency physicians (Table II in the online-only Data Supplement). Similarly, Ali et al7 proposed a predictive tool to differentiate between ACI and SM for patients being evaluated by telemedicine network.
The goal of our study was to develop an intuitive scoring system, with the least number of parameters, for screening and stratifying SM from ACI cases in the emergency setting.
This study was divided into 2 phases. First, a development phase, in which we evaluated our single-center pilot data and a cohort of patients from a different stroke center and constructed an intuitive scoring system: FABS. Second, a validation phase, during which we evaluated consecutive patients with symptoms of ACI in 2 tertiary care stroke centers during a 3-year period. The validation of our scoring system (FABS) is the focus of the present study.
Construct an Intuitive Scoring System
We previously conducted a single-center pilot study to evaluate the feasibility of rapid short-sequence magnetic resonance imaging (MRI) for the screening of suspected SMs before IV-tPA administration.16 In a 9-month period, we evaluated 35 patients, whose clinical presentations were suspicious for an SM, using a rapid short-sequence MRI (mean age, 49; 78% women; median National Institutes of Health stroke scale [NIHSS], 4). Three patients had a diagnosis of an acute ischemic stroke on MRI and received IV-tPA. We compared the demographics and clinical characteristics of SMs with the acute stroke patients (n, 118) who presented to our center during the same time period. The analysis indicated that SM cohort was younger (P<0.01), less likely to have a history of atrial fibrillation (P<0.001) and more likely to have a history of seizure disorder (P<0.05), and presenting systolic blood pressure <150 mm Hg (P<0.05). Moreover, analysis of initial clinical presentation indicated that SMs are less likely to present with facial droop (P<0.0001) and more likely to have isolated sensory symptoms and within normal limit A1C (P<0.05). Because of the small sample size and limited statistical power of our study, we performed a second analysis of a cohort of patients from Barrow Neurological Institute (Phoenix, Arizona). Barrow Neurological Institute database included 150 consecutively and prospectively collected patient data from 2007 to 2008. We analyzed the association of 19 parameters (Table III in the online-only Data Supplement) with SM. The details of this database have previously been published.8 The parameters were selected based on our previous analysis as mentioned above and the current literature (Table I in the online-only Data Supplement). Our logistic regression model showed that younger age, negative history of atrial fibrillation, history of psychiatric illness, low NIHSS, and absence of facial droop had a significant association (at 95% level) with SM. In addition, the model confirmed that positive history of seizure, normal A1C level, and low-density lipoprotein were significantly (at 90% level) associated with SM.
We constructed an intuitive scoring system, named FABS score, by including variables that were associated with SM at 90% or higher level in our logistic regression model. Therefore, FABS included 6 variables: absence of Facial droop, negative history of Atrial fibrillation, Age, systolic Blood pressure at presentation, history of Seizures, and isolated Sensory deficit without limb weakness at presentation. Facial droop was defined as an acute unilateral nasolabial flattening or paralysis of the lower half of one side of the face. History of atrial fibrillation and seizure were based on reported medical history or previous clinical documentation. Age was dichotomized at 50 years. Systolic blood pressure was considered as the first recorded systolic blood pressure and dichotomized at 150 mm Hg. Limb weakness required a clear report of loss of strength as opposed to more vague terms, such as heaviness, in the absence of definite weakness.
FABS includes 2 components of NIHSS (facial droop and isolated sensory deficit without limb weakness). Therefore, we did not include NIHSS in FABS to minimize redundancy and complexity. History of psychiatric illness was not included in FABS secondary to its subjectiveness and low interinformant reliability.17,18 We also excluded low-density lipoprotein and A1C because the results most likely would not be available in the emergency setting.
We did not have the statistical power to reliably determine the exact size of risk ratios. Also, our goal was to keep the FABS scoring system simple, easy to memorize, and practical. Therefore, we allocated binary values (0 and 1) to each variable, the overall predictive score being the sum.
Validation of FABS Scoring System
We prospectively evaluated consecutive patients presented with stroke-like symptoms within 4.5 hours after symptom onset in 2 tertiary care stroke centers (University of Tennessee Health Science Center, Memphis; and Attikon University Hospital, Athens, Greece) during a 3-year period (2012–2014). Patients who had a clinically related acute finding in their initial noncontract head CT scan (early ischemic changes, cerebral hemorrhage, tumor, etc) were not included in this study. Every patient included in our validation phase had an MRI within the first 24 hours of admission. Patients who could not have MRI were excluded. Every patient in our cohort was evaluated by stroke response team, including the neurology house staff and a vascular neurologist. Stroke response team read all the neurodiagnostic images, including initial CT head. In cases of doubt or discrepancy between the read and the radiology report, the images were discussed with the neuroradiologist on call.
The diagnosis of ACI versus SM was made based on the patient’s presentation, medical history, hospital course, resolution of symptoms, discharge diagnosis, head CT, electroencephalography (if available), in addition to post-thrombolysis diffusion-weighted images (DWIs), apparent diffusion coefficient, and fluid-attenuated inversion recovery. We also examined head magnetic resonance angiography or head CT angiography for any acute artery thrombosis.
The data set comprises baseline characteristics of patients, such as age, sex, medical history of hypertension, hyperlipidemia, atrial fibrillation, seizure disorder, cigarette smoking, as well as initial systolic and diastolic blood pressure, clinical presentation, first recorded NIHSS, neuroimaging findings, and discharge diagnosis. We calculated FABS score for each patient as described above.
Continuous variables are presented as mean±SD (normal distribution) and as median±SD (skewed distribution). Statistical comparisons were performed between SM and confirmed ACI patients using the χ2 test, Fisher exact test, unpaired t test, and Mann–Whitney U test as indicated for dichotomous or continuous variables. Logistic regression was used to examine how well FABS score would predict the probability of being an SM. Sensitivity and specificity measurements were calculated for SM and ACI cases to produce a receiver operating characteristic curve and determine the optimum FABS score threshold for the diagnosis of SM. The Statistical Package for Social Science (SPSS Inc, version 22 for Windows) was used for statistical analyzes of study data.
To examine the impact of different proportions of SM in the treated population on the predictive potential of FABS, simulation was used based on the existing data by bootstrapping data sets with different proportions of SM. Specifically, 10 000 data sets of total subjects were randomly sampled with a replacement for each level of SM (45%, 35%, 30%, 25%, and 20%). The proportions of SM for each score were determined and averaged across the data sets within each SM level. The resulting means were then graphed. The simulations were done in R version 3.2.2.
A total of 784 patients (mean age, 58±15 years; 50% men; 41% SMs) were included in this study. Sixty-seven patients were excluded because they could not have an MRI. Median NIHSS at presentation was 5 points (interquartile range, 1–11). Ten percent of patients were from Attikon University Hospital. Table 1 includes patients’ demographic data. Table 2 presents the distribution of FABS score in patients with ACI and SM. The primary analysis indicated that patients with scores of 0 to 2 were highly likely to experience an ACI and with scores of 5 or 6 were likely to have SM. Receiver operating characteristic curve (area under the curve, 0.95; 95% confidence interval [CI], 0.93–0.98) indicated that FABS≥3 could identify patients with SM with 90% sensitivity (95% CI, 86%–93%) and 91% specificity (95% CI, 88%–93%). The negative predictive value and positive predictive value were 93% (95% CI, 90%–95%) and 87% (95% CI, 83%–91%), respectively. FABS≥4 showed a specificity of 98% (95% CI, 97%–99%) and a positive predictive value of 95% but a low sensitivity of 45% (95% CI, 39%–50%). ACI was diagnosed in 7 (4.6%) patients with FABS≥4. Five patients had posterior circulation stroke, whereas remaining 2 had a small cortical stroke. FABS score performs well at each site. There was no significant difference between sites (Memphis and Athens) in terms of sensitivity and specificity for FABS≥3 or ≥4. Data from the University of Tennessee indicated that for FABS≥3 sensitivity was 91% (95% CI, 87–94) and specificity was 92% (95% CI, 88–95). For FABS≥4, sensitivity was 44% (95% CI, 39–49) and specificity was 98% (95% CI, 97–99). Data from Attikon University Center indicated that for FABS≥3 sensitivity was 84% (95% CI, 63–95) and specificity 80% (95% CI, 66–91). For FABS≥4, sensitivity was 50% (95% CI, 30–70) and specificity was 97% (95% CI, 86–99).
We further analyzed the relative importance of each scoring component and its impact on FABS score. Our analysis (Table 3) indicated that the absence of facial droop alone is highly sensitive (94%; 95% CI, 90%–98%) and reasonably specific (71%; 95% CI, 68%–75%) for the diagnosis of SM. The results of the analysis showed that diagnostic probabilities for SM based on FABS score increase as SM rate increases (Figure). The scores of 3 and 4 show some degree of variability depending on the percentage of SM in the underlying cohort, whereas the scores of 0 to 2 and 5 to 6 show little or no difference in either a low or high risk of SM.
Our results show that FABS is a simple and effective scoring tool for screening and stratifying SM from ACI cases in the emergency setting. Furthermore, our analysis suggests that the probability of SM for each FABS score is a function of SM rate.
FABS scoring system can be applied to the patients who present to the emergency department with stroke-like symptoms when the initial noncontrast head CT scan is negative for any related acute findings. This scale is primarily designed for stroke response team, including neurology house staff, neurologists, and vascular neurologists. The purpose of developing FABS is not to exclude potential ACI candidates from IV-tPA administration rather apply a screening system where selected patients (suspected SM) can be subjected to advanced imaging like DWI for further diagnosis. There have been several studies indicating the feasibility of a rapid short-sequence MRI in acute ACI while the door-to-needle time stays within 60 minutes.16,19,20
The positive conventional likelihood ratio of FABS≥3 and FABS≥4 was 9.9 (95% CI, 7.4–13.3) and 29.8 (95% CI, 14.1–62.7), respectively. FABS≥4 also showed a higher specificity (98%) and a positive predictive value (95%). Therefore, for centers with immediate access to MRI, FABS≥3 can be considered for further imaging to discriminate SM from ACI. However, in centers with delayed access to an MRI, we recommend that FABS≥4 be considered for further imaging. It should also be noted that FABS≥4 cannot rule in patients with SM with 100% probability; therefore, the treatment team has to be vigilant and make sure that further imaging (DWI) does not put the patient outside the treatment time window. In both scenarios, we recommend that the stroke response team and pharmacist accompany the patient to the MRI suite. If the DWI is positive for ACI, the team should stop the MRI and administer tPA in the MRI suite.
On the basis of several published studies,4–6 the average rate of tPA administration to SMs is ≈15%. We can predict that using FABS score in a center with 15% rate of tPA administration to SMs will reduce this misdiagnosis rate to ≈2% (with FABS≥3) and 8% (with FABS≥4). Our previous study6 also indicated that early identification of SM can reduce the length of stay by 1 day. Further studies are needed to evaluate the incremental benefit of FABS in terms of lowering the rate of tPA administration to SMs, reducing the financial cost and emotional burden, as well as lowering the length of hospital stay.
Among all the stroke diagnostic tools, the ROSIER scale,15 similar to the FABS, has been designed for an emergency room setting. In the ROSIER scale, arm weakness, leg weakness, speech disturbance, facial weakness, and visual field defect predict stroke, whereas seizures, confusion, or loss of consciousness predicts nonstroke diagnosis. The ROSIER scale screens for ischemic and hemorrhagic strokes and has a sensitivity of 93% and a specificity of 83%. Although the ROSIER score is a well-designed score and has a good sensitivity and specificity, it has several differences from the FABS score. First, atrial fibrillation was neither tested nor included in the development of ROSIER score. Our analysis indicated that atrial fibrillation has a strong association with ischemic stroke and transient ischemic attack. Second, ROSIER score is designed for emergency department physicians to screen for stroke (both ischemic and hemorrhagic) and refer patients to a stroke center before initial neurodiagnostic imaging. However, FABS score is designed for stroke response team and should be applied after a negative head CT scan for any acute change. FABS score screens for SMs among candidates for intravenous thrombolysis to select those that need to be subjected to further confirmatory diagnostic imaging (eg, DWI). In a future study, we plan to compare the performance of FABS with ROSIER score in an imaging selected population.
Our SM rate (41%) was higher than a similar large study performed by the National Institutes of Health (NIH) Stroke Program.9 They analyzed >8000 stroke codes where the SM rate was 30%. We think one of the main reasons for our higher SM rate was the exclusion of patients who had acute findings on their initial head CT scan (early ischemic changes, intracerebral hemorrhage, etc). We excluded those patients because, in many CT-based centers, the head CT is done quickly, and the question of SMs versus stroke usually is raised when the initial head CT scan is negative for any acute finding. Therefore, our approach better replicates the actual practice. However, NIH program is an MRI-based program, and they usually skip the head CT scan.
Our study similar to the other studies showed that SMs are younger, less likely to have a history of hypertension or atrial fibrillation, and more likely to have a history of psychiatric illness and seizure disorders.5,6,9 Our study also confirmed that the absence of facial droop alone is highly sensitive and specific for the diagnosis of SM.7,16
Our study had some limitations. We did not separate anterior versus posterior circulation strokes. Patients with posterior circulation stroke not always present with typical focal neurological deficits like facial droop.21 Excluding posterior circulation strokes from our multicentric data should further increase the sensitivity and specificity of FABS. The latter can be considered for future studies. Moreover, our cohort represents the patient population from only 2 stroke programs (a total of 6 hospitals). Therefore, other studies with different patient population are needed to independently evaluate the diagnostic yield of FABS scoring system. Another limitation of our study is that we could not measure how excluding the patients who did not undergo brain MRI might have biased our sample. However, it should be mentioned that FABS is not useful in patients who are unable to undergo brain MRI. The clinical utility of this scoring system depends on subsequent MRI for suspected patients with SM for further confirmation of ACI.
We think that further optimization can be performed to have a score that is more accurate and can be used as a metric from the electronic medical data on arrival of a patient to the emergency department. However, this optimization should be achieved with a larger cohort of patients. In addition, full optimization will produce complex equations that may be best used if integrated into specialized applications.
In conclusion, our study provides a simple scoring tool, FABS score, for stratifying SM from ACI cases in the emergency setting. On the basis of the score, potential SM cases can be subjected to further advanced imaging (DWI) before being considered for IV-tPA administration. Further studies are necessary for the external validation of this novel scoring tool.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.116.013842/-/DC1.
- Received May 13, 2016.
- Revision received July 6, 2016.
- Accepted July 7, 2016.
- © 2016 American Heart Association, Inc.
- Hand PJ,
- Kwan J,
- Lindley RI,
- Dennis MS,
- Wardlaw JM
- Kothari RU,
- Brott T,
- Broderick JP,
- Hamilton CA
- Tsivgoulis G,
- Alexandrov AV,
- Chang J,
- Sharma VK,
- Hoover SL,
- Lao AY,
- et al
- Tsivgoulis G,
- Zand R,
- Katsanos AH,
- Goyal N,
- Uchino K,
- Chang J,
- et al
- Ali SF,
- Viswanathan A,
- Singhal AB,
- Rost NS,
- Forducey PG,
- Davis LW,
- et al
- Harbison J,
- Hossain O,
- Jenkinson D,
- Davis J,
- Louw SJ,
- Ford GA
- Kidwell CS,
- Starkman S,
- Eckstein M,
- Weems K,
- Saver JL