Prehospital stroke scale (FAST PLUS Test) predicts patients with intracranial large vessel occlusion

Abstract Background and Purpose Mechanical thrombectomy (MT) is indicated for the treatment of large vessel occlusion (LVO) stroke. MT should be provided as quickly as possible; therefore, a test identifying suspected LVO in the prehospitalization stage is needed to ensure direct transport to a comprehensive stroke center (CSC). We assume that patients with clinically severe hemiparesis have a high probability of LVO stroke. We modified the FAST test into the FAST PLUS test: The first part is the FAST test and the second part evaluates the presence of severe arm or leg motor deficit. This prospective multicenter study evaluates the specificity and sensitivity of the FAST PLUS test in detecting LVO stroke. Methods Paramedics were trained through e‐learning to conduct the FAST PLUS test. All prehospital suspected stroke patients who were administered the FAST PLUS test were included. Demographics, National Institutes of Health Stroke Scale (NIHSS) score, brain computed tomography (CT), and CT angiography (CTA) were recorded. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and receiver operating curve (ROC) area for LVO were calculated. Results The study included 435 patients. LVO were found in 124 patients (28%). Sensitivity was 93%, specificity was 47%, PPV was 41%, NPV was 94%, and ROC area for ICA/MCA occlusion was 0.65. Intracerebral hemorrhage (ICH) was identified in 48 patients (11%). Conclusion We found that the FAST PLUS test had a high sensitivity for LVO stroke. Of the 435 patients, 41% were all directly transported to a CSC based on positive FAST PLUS test scores and were potential candidates for MT.


| FAST PLUS test
The FAST PLUS test has two parts. The first part is the FAST test, which is employed in all possible cases of stroke occurrence. This test consists of the following items: Facial palsy (0-1), any failure of Arm motor function (0-1), and Speech (scored 0-1). The FAST test is considered positive if the score is at least one.
The second part of the FAST PLUS test evaluates only the presence of severe arm or leg motor deficit (scored 0-1). An NIHSS score of 3 or 4 for arm or leg is considered a severe deficit. The FAST PLUS test results are considered positive when there is a positive general FAST test score and severe paresis of a leg or an arm or both. A completely new version of the Stroke Card was created using the FAST PLUS test criteria ( Figure 1).

| Training of paramedics
In previous practice, paramedics selected suspected stroke patients according to their FAST test results. For this study, paramedics were trained via e-learning to conduct the FAST PLUS test. For their further education, three video recordings were used in order to demonstrate the examination for motor deficit in the lower and upper limbs. The first video shows a patient with complete hemiparesis; the NIHSS score was 4 for both limbs. The second video shows a patient with severe hemiparesis, with an NIHSS score of 3 for both limbs. The third video shows a patient with mild hemiparesis, with an NIHSS score of 2 for both limbs. A certified neurologist performed the NIHSS scoring.

| Study population
Prehospital patients with suspected stroke (FAST test positive) were transported by emergency medical services to one of the three stroke centers in Ostrava (Czech Republic) according to their territory. Mechanical thrombectomy (MT) is provided at two of these centers. The catchment area of the centers was 637,584 inhabitants.
Patients were transported according to the stroke triage protocol established in the Czech Republic.
Inclusion criteria for this study were as follows: (a) Suspected acute stroke patient admitted to one of the three stroke centers; (b) FAST PLUS test evaluation by paramedics; and (c) CT and CTA evaluations.
The exclusion criterion was suspected stroke with more than 12 hr from symptom onset.
The following baseline parameters were recorded: gender, age, FAST PLUS test results, total NIHSS score during admission, NIHSS score for arms, NIHSS score for legs, brain CT results, an occlusion of the middle cerebral artery (MCA) part M1/2 or of the intracranial internal carotid artery (ICA), etiology other than ischemic stroke, onset of stroke within 6 hr, number of patients with systemic thrombolysis, and mechanical recanalization.
A neurologist verified the accuracy of the FAST PLUS test data entered by the paramedics. In addition, a written paper record for each patient was submitted to a neurologist along with the FAST PLUS test results.

| Statistical analysis
Basic descriptive statistics were used for the final evaluation and statistical analysis. The FAST PLUS test was evaluated in terms of sensitivity and specificity calculations with a 95% confidence interval. The area under the receiver operating curve (ROC) was also calculated. Statistical tests were evaluated with a 5% significance level.
The Stata version 14 software was used for the statistical analysis. ROC and areas under the ROC (c-statistics) were calculated as measures of the FAST PLUS test's predictive ability for LVO. An ideal prediction produces a c-statistic of 1.00; precision no better than chance is associated with a c-statistic of ≤0.50.

| RE SULTS
Over the 10-month study period, 1,605 patients with suspected stroke were transported to stroke centers by emergency medical services.  Men formed 51% of the study population; the average age was 73 years (median 74). Baseline data are shown in Table 1

| D ISCUSS I ON
Our study found that the specificity and PPV of the FAST PLUS test were 47% and 41%. This corresponds to the results of the published tests G-FAST (39%) and CPSSS (40%) ( Table 4) (Katz et al., 2015;Scheitz et al., 2017).
In fact, our study found that a simple test such as the presence of hemiparesis can identify 41% (PPV = 41%) of patients with LVO.
Practically, when the test is applied to a population with a 28% prevalence of LVO, four of 10 FAST PLUS test positive patients are directly transported to a comprehensive stroke center could be expected to have LVO. The acceptability of this number depends on several considerations, such as the capacity of prehospital services and CSCs, because 60% of patients would not benefit from such transport.
The real value of LVO prevalence is not yet precisely known; it ranges from 4.7% to 24% (Dozois et al., 2017;Rai et al., 2017). As PPV strongly depends on LVO prevalence, we provided expected PPV for populations with 10% and 20% LVO prevalence (Table 5). Such a prevalence could be expected if the FAST PLUS test is applied more generally to a less select population. In such populations, PPV would decrease to 30% and 16%, such patients may still benefit from direct transport to a CSC.
The presented test has a higher sensitivity and NPV than other tests (Table 4) (Hastrup et al., 2016;Katz et al., 2015;Lima et al., 2016;McMullan et al., 2017;Nazliel et al., 2008;Pérez de la Ossa et al., 2014;Scheitz et al., 2017;Singer et al., 2005). We found that the FAST PLUS test had a high sensitivity of 93% and high NPV of 94%. Therefore, if the direct transport to CSC is selected, the majority of patients with LVO occlusion could be identified.
In practice, the FAST PLUS test, with its high sensitivity and NPV, is suitable for "mothership" transport systems in areas with a short distance between the PSC and a CSC that has a sufficient capacity for systemic thrombolysis in patients without LVO.
There are 32 PSC (1 per 300,000 inhabitants) and 12 CSC (1 per 900,000) in the Czech Republic. The FAST PLUS test seems suitable for countries with similar networks of stroke care, without delaying systemic thrombolysis or overloading the CSC (Tomek et al., 2017 (Fischer et al., 2005;Kalita et al., 2013). The FAST PLUS test evaluates each item with only a two-degree scale (yes/no), in contrast to most tests using scales with three or more degrees (  (Hastrup et al., 2016;Katz et al., 2015;Lima et al., 2016;Nazliel et al., 2008;Scheitz et al., 2017;Singer et al., 2005). The best prehospital data are currently provided by the RACE test: 357 patients, which is the most extensive of all available tests. The LAMS test has been used in prehospital care for stroke identification, but not for predicting LVO (Nazliel et al., 2008). In a pilot study of the Cincinnati test, untrained paramedics tested only 58 patients (Table 4) (McMullan et al., 2017;Pérez de la Ossa et al., 2014).
Our study has several limitations.
Only 47% of the stroke patients were given a FAST PLUS test.
We cannot rule out selection bias or the higher prevalence of LVO than could be expected when the test is applied in the future. Patients were enrolled only after admission to hospital, which explains the low percentage of stroke mimics (2%).
Another limitation of our study was that paramedics were trained only once via e-learning. Three video recordings with different levels of hemiparesis were presented, without further testing. The training was not obligatory, so not all paramedics were trained. A more thorough education process could lead to better results.
The clinical impact of "mothership" or "drip-and-ship" transport systems has not yet been assessed. According to published studies, secondary transport significantly prolongs the time from stroke onset to recanalization (Mørkenborg, Steglich-Arnholm, Holtmannspötter, & Krieger, 2015;Zhao et al., 2017).
The influence of the FAST PLUS test or other triage stroke tests on the time from stroke onset to MT or to systemic thrombolysis in non-LVO patients was not studied and needs to be evaluated in further studies.

| CON CLUS ION
We found that the FAST PLUS test is highly sensitive to the presence of LVO; the majority of patients with LVO could be identified.
However, PPV is moderate, so less than half of the patients identified by FAST PLUS test could eventually have LVO. Nevertheless, the test is very simple, and its results could be improved by better training of paramedics.

ACK N OWLED G M ENTS
The authors thank Jarmila Lakomá (Angels Initiative, European Stroke Organisation) and Petr Jaššo (Emergency Health Services, Ostrava) for their help with organizing the education of paramedics.
The authors also thank Hana Tomášková for statistical analysis and Anne Johnson for grammatical corrections.

CO N FLI C T O F I NTE R E S T
The authors report no conflicts of interest. The authors themselves are responsible for the content and composition of the manuscript.