Stroke Recognition for First Aid Providers: A Systematic Review and Meta-Analysis

Aim To perform a systematic review of the literature on the effectiveness of existing stroke recognition scales used in a prehospital setting and suitable for use by first aid providers. The systematic review will be used to inform an update of international first aid guidelines. Methods We followed the Cochrane Handbook for Systematic Reviews of Interventions methodology and report results according to PRISMA guidelines. We searched Medline, Embase and CENTRAL on May 25, 2020 for studies of stroke recognition scales used by first aid providers, paramedics and nurses for adults with suspected acute stroke in a prehospital setting. Outcomes included change in time to treatment, initial recognition of stroke, survival and discharge with favorable neurologic status, and increased layperson recognition of the signs of stroke. Two investigators reviewed abstracts, extracted and assessed the data for risk of bias. The certainty of evidence was evaluated using GRADE methodology. Results We included 24 observational studies with 10,446 patients evaluating 10 stroke scales (SS). All evidence was of moderate to very low certainty. Use of the Kurashiki Prehospital SS (KPSS), Ontario Prehospital SS (OPSS) and Face Arm Speech Time SS (FAST) was associated with an increased number of suspected stroke patients arriving to a hospital within three hours and, for OPSS, a higher rate of thrombolytic therapy. The KPSS was associated with a decreased time from symptom onset to hospital arrival. Use of FAST Emergency Response (FASTER) was associated with decreased time from door to tomography and from symptom onset to treatment. The Los Angeles Prehospital Stroke Scale (LAPSS) was associated with an increased number of correct initial diagnoses. Meta-analysis found the summary estimate sensitivity of four scales ranged from 0.78 to 0.86. The FAST and Cincinnati Prehospital Stroke Scale (CPSS) were found to have a summary estimated sensitivity of 0.86, 95% CI [0.69-0.94] and 0.81, 95% CI [0.70-0.89], respectively. Conclusion Stroke recognition scales used in the prehospital first aid setting improves the recognition and diagnosis of stroke, thereby aiding the emergency services to triage stroke victims directly down an appropriate stroke care pathway. Of those prehospital scales evaluated by more than a single study, FAST and Melbourne Ambulance Stroke Screen (MASS) were found to be the most sensitive for stroke recognition, while the CPSS had higher specificity. When blood glucose cannot be measured, the simplicity of FAST and CPSS makes these particular stroke scales appropriate for non-medical first aid providers.


Introduction
Stroke is one of the leading causes of death and disability worldwide [1]. The early detection of stroke in the prehospital setting has the potential to improve stroke outcomes by decreasing delays in treatment. A variety of stroke assessment scales have been developed for both in-hospital and prehospital use. Stroke scales designed for the prehospital setting have a lower number of diagnostic criteria, easy-to-identify clinical signs and simplicity of implementation, making them applicable for use by first aid providers and lay persons. In 2015, the International Liaison Committee on Resuscitation (ILCOR) published a Consensus on Science with Treatment Recommendations (CoSTR), suggesting a benefit from the first aid use of stroke recognition scoring systems or scales for individuals with suspected acute stroke [2,3].

Data analysis
Continuous outcomes are reported as mean differences (MD) with 95% confidence intervals (CIs). Dichotomous outcomes are reported as risk ratios (RR) with 95% CIs. There was insufficient data to conduct meta-analyses of effectiveness data. For diagnostic studies, all scales used the same positivity threshold of 'one or greater', which indicates that the person was considered to have a stroke with one or more positive criteria. For each index test, we generated a diagnostic 2 × 2 table (true positives, false positives, true negatives and false negatives) from which we calculated sensitivity and specificity with 95% confidence intervals (CI). When more than one study was identified per scale, we calculated a summary point estimated sensitivity and specificity using a random effects meta-analysis and created Summary Receiver Operating Characteristic (SROC) plots to show the variation in test accuracy estimates across studies with Review Manager 5.3 (RevMan 5.3, The Nordic Cochrane Centre, Copenhagen, Denmark, 2014). Parameter values required by Review Manager Software to construct plots in the SROC space were calculated with MetaDTA: Diagnostic Test Accuracy Meta-Analysis website, version 1.25 (https://crsu.shinyapps.io/dta_ma/) [12].

Results
For the literature search and study selection, an updated search strategy from 2014 to 2019 and a rerun search strategy from 2019 to 2020 identified 1814 unique titles/abstracts. In addition, we identified new studies and 24 from the previous 2015 search results for the 2015 ILCOR CoSTR for first aid stroke assessment [2]. Based on title and abstract screening, we excluded 1768 studies (reviewer agreement was 95.15%, Kappa = 0.44). Of the 78 full-text articles reviewed, a further 54 were excluded (reviewer agreement was 99.87%, Kappa = 0.79). We ultimately included a total of 24 studies (Figure 1). recognition scales evaluated in these studies are described in Table 2     1. Verbal instruction and sensory, Close your eyes! Grip your hand! (n-paretic side); 2. GCS < 7 or suspected head injury exclusion original paper; 3. Seizure at onset, can be transported to arrive within two hours of onset, time since symptom onset < 2 hours, GCS < 10, blood glucose > 4 mmol/L, symptoms of the stroke have resolved; 4. Blood glucose > 3.5 mmol/L, history of seizure; 5. History of seizure, time since symptom onset < 24 hours, at baseline, patient is not wheelchair bound or bedridden, age > 45 years, blood glucose 2.8 to 22.2 mmol/L; 6. History of seizure, time since symptom onset < 24 hours, at baseline, patient is not wheelchair bound or bedridden, blood glucose 3.3 to 22.2 mmol/L; 7. History of seizure, at baseline, patient is not wheelchair bound or bedridden, blood glucose 2.8 to 22.2 mmol/L, age limit = 40 years; 8. Age > 18 years, intended for use, only in conscious people, i.e. alert or aroused by stimulation; 9. Time of onset less than 2 hours, blood glucose measurement inside the range of 4-17 mmol/L.

Risk of bias within studies and certainty of the evidence
An overview of the assessment of the overall certainty of evidence, using ROBINS-I assessment tool for nonrandomized studies of interventions studies and QUADAS-2 for diagnostic studies is provided in Tables 3, 4 respectively. Overall, the certainty of evidence was moderate to very low across all outcomes, primarily due to risk of bias, indirectness and imprecision. A detailed overview of GRADE assessments per outcome can be found in Appendix C.

Study findings on stroke assessment scale effectiveness
For the critical outcome "time to treatment", we identified four observational studies [20,24,29,34] evaluating four different stroke scales (KPSS, LAPSS, OPSS, FASTER). For the KPSS, one retrospective observational study [34], enrolling 430 participants with suspected acute stroke in the prehospital setting, showed an association between the use of KPSS and an increase in the number of patients whose time from symptom onset to hospital arrival was within 3 hours.  [29], enrolling 861 prehospital participants with acute suspected stroke, showed an association between use of the OPSS and increased proportion of patients with a time from symptom onset to hospital arrival within 3 hours when using the OPSS, compared with not using the OPSS (RR We did not identify any comparative studies evaluating stroke recognition for the outcomes of "favorable neurologic status" or "survival with favorable neurologic outcome". For the outcome of recognition of stroke (diagnostic studies, outcome defined as correct stroke diagnosis), we identified 19 observational studies [13-19, 21, 22, 25-32, 35, 36] including a total of 8153 participants, evaluating nine different screening tools (FAST, LAPSS, OPSS, CPSS, ROSIER, MASS, BEFAST, Med-PACS, Pre-HAST) ( Table 5). The reported prevalence, sensitivity, specificity, positive and negative likelihood ratio for each scale are reported in Table 5.   [22,32], but these provided incomplete data and could not be included in the meta-analysis. The diagnostic accuracy of the MASS was assessed by low certainty evidence from two prospective observational studies [14,15] and one retrospective observational study [28]. These three studies included a total of 981 participants suspected of having a stroke.   Studies of stroke assessment scales can be divided into subgroups based on whether the scale includes blood glucose measurement or not. In the nine diagnostic studies that used stroke scales with blood glucose measurement (LAPSS, OPSS, ROSIER, MASS, Med-PACS) [14-18, 26, 28, 29, 36], the reported sensitivities ranged from 0.74 to 0.97, compared with 0.80 to 1.00 in the 14 studies of stroke scales that did not include blood glucose measurement (FAST, CPSS, Pre-HAST, BEFAST) [13-15, 17, 19, 21, 25-28, 30, 31, 35, 36]. The reported specificities from studies with stroke scales including blood glucose measurement (LAPSS, OPSS, ROSIER, MASS, Med-PACS) ranged between 0.18 and 0.86 compared with 0.26 to 0.55 in the studies that used scales without blood glucose measurement (PreHAST, FAST, CPSS, BEFAST). The comparison of Summary Receiver Operating Characteristics (SROC) curve between stroke scales with blood glucose measurement and stroke scales without blood glucose measurement is presented in Figure 3. The first comparison covers all studies ( Figure 3A, 3B); the second covers only the scores assessed by more than one study ( Figure 3C, 3D).

FIGURE 3: Summary receiver operating characteristics (SROC) plot of strokes scales with and without glucose measurement
A-SROC of stroke scales with glucose measurement; B-SROC of stroke scales without glucose measurement; C-SROC of stroke scales with more than one study per scale with glucose measurement; D-SROC of stroke scales with more than one study per scale without glucose measurement.

Discussion
We identified and systematically reviewed studies of accuracy for prehospital stroke recognition tools that are applied in the prehospital setting and potentially suitable for use by first aid providers. We consider an ideal stroke assessment tool for first aid to be one that is easily understood and remembered, has a high sensitivity for detecting stroke and can be completed in minimal time. Because the home use of blood glucose measurement devices is increasingly common in populations at risk for acute stroke, we included prehospital stroke scales that incorporate blood glucose measurements but evaluated them separately for accuracy. In this systematic review, three of the four included scales (KPSS, FASTER and OPSS) showed an association between prehospital use and a decreased time from stroke onset to treatment [20,29,34]. Unfortunately, it was not possible to perform a meta-analysis for this outcome due to the limited number of studies.
In terms of definitive stroke diagnosis or therapy administration, using a stroke recognition scale in the prehospital setting does not seem to increase the proportion of patients with confirmed stroke diagnosis. However, patients with confirmed stroke were promptly admitted to a hospital and received treatment more quickly.
For accuracy of recognition of stroke we pooled the data from the 17 diagnostic studies of FAST, CPSS, LAPSS, and MASS individually to calculate a summary estimated sensitivity and specificity [14-19, 21, 22, 25-28, 30-32, 35, 36]. Other scales that were only assessed by a single study were not included [13,17,21,29,36]. We considered both the FAST and CPSS to be stroke assessment tools that a first aid provider would find easy to understand, remember and to use. These two stroke scales are supported by multiple studies with a large total number of participants but do not include a blood glucose measurement. . Thus, whilst FAST appears to be a more sensitive scale for the recognition of stroke, it is less specific than the CPSS. The high risk of bias and between-study heterogeneity did not allow us to determine summary estimates sensitivity and specificity of scales with and without blood glucose measurement. Some studies reported a high specificity of above 0.90, where the proportion of missed stroke patients assessed by the scale ranged from 10 to 22% [16,18]. Prehospital stroke recognition scales should not be interpreted as confirmatory diagnostic tests but only as a screening test. Most of the studies only assessed the outcomes of true positive patients, however it would be of value to know the impact of the scale on those who were false negatives.
Two stroke assessment scales that include blood glucose measurement in their eligibility criteria (MASS and LAPSS) were evaluated by multiple studies and included 981 patients for MASS and 2692 patients for LAPSS ( Table 5). We found these scales had similar sensitivities for stroke identification as for scales without blood glucose measurement, but increased specificities ( Figure 3C, 3D). We recognize that many first aid providers may not have access or the skills to use a properly calibrated glucometer. Local guidelines would need to determine the benefit of increased specificity of stroke scales that include glucose measurement compared with using simpler stroke scales that do not require glucose measurement.
Three systematic reviews analyzed stroke recognition instruments in the prehospital setting [37][38][39]. Brandler et al. in 2014 included studies in which the scales were used by paramedics or emergency medical technicians (EMTs) and included scales requiring blood glucose measurement [37]. The authors concluded that LAPSS performed more consistently and that LAPSS and CPSS had similar diagnostic capabilities. Our systematic review includes all of the studies evaluated by Brandler et al. and adds new data from 16 more publications. Additionally, we report diagnostic accuracy of scales that require blood glucose measurement separately from those without glucose measurement, to help identify appropriate scales for use by first aid providers. A systematic review by Rudd et al. in 2016 included all studies in which the scales were administered face-to-face by any prehospital or hospital clinician to identify adults suspected of stroke [38]. Eleven studies included in this systematic review were also included in our review, but 10 studies did not meet our inclusion criteria (seven papers and three abstracts). The authors concluded that available data do not allow a strong recommendation to be made about the superiority of a particular stroke recognition scale evaluated. Zhelev et al. in a Cochrane review in 2019 analyzed prehospital stroke scales as screening tools for early identification of stroke and transient ischemic attack [39]. They included in a "prehospital setting subgroup" all studies where the scale had been used in the prehospital setting regardless of the background and training of the person performing the assessment, and only evaluated diagnostic accuracy. The author concluded, "in the field, CPSS had consistently the highest sensitivity but was less specific than most of the scales". In our systematic review, we have focused on the scales that can potentially be used by trained first aid providers or lay persons in a prehospital setting. We attribute our inclusion and exclusion criteria to any differences in our results. Lastly, our systematic review is not limited to a diagnostic accuracy review. We also evaluated the influence of stroke scale use on the time to treatment and the rate of stroke diagnosis.
Our review has some limitations. First, only four stroke scales were investigated by more than a single study, for which a large number of participants have been included (FAST, CPSS, LAPSS, MASS). Six scales (FASTER, OPSS, KPSS, ROSIER, BEFAST, Med-PACS) were only investigated in single studies, including between 250 and 600 participants [13,17,20,21,29,34,36]. The PreHAST scale provided the highest sensitivity (1.00, 95% CI [0.87-1.00]), but was only evaluated in a single study, with 69 participants [13]. The prevalence of stroke/TIA ranged from 23% [28] to 88% [16] (Table 5), reflecting differences in population and patient selection that may affect sensitivity and specificity estimates. Second, the accuracy of the scales for identifying people with stroke/TIA may also be affected by confounders such as differences in age, sex, the proportion of patients with ischemic stroke, hemorrhagic stroke or TIA (  [40]. Third, the overall Kappa for the review of titles/abstracts was moderate (Kappa = 0.44). This reflects difficulties in correctly identifying observational studies of stroke recognition in adults in a prehospital setting. However, based on a subsequent review of reference lists, we did not identify any additional articles that were missed during the review process. Finally, the risk of bias is serious or moderate in four of six studies due to possible confounding, missing data and the different time interval definitions for the outcome "time to treatment" [20,29,33,34]. Risk of bias is high or unclear in most of the diagnostic studies for patient selection or quality of the reference standard, and most of the studies failed to include all eligible consecutive participants. The methodology used by the studies is often different, measurement of the time to treatment is not the same, and the method and the length of training used to teach the score varied between studies. There is a high level of between-study heterogeneity, and therefore we must interpret the summary estimate result with caution.

Conclusions
The use of stroke recognition scales in the prehospital setting should be encouraged. They assist in the detection of the presence of stroke and reduce the time from symptom onset to definitive treatment.
There are many stroke scales available for use in the prehospital environments and the selection of which scale to use remains complex. This review has shown that the use of the FAST and OPSS stroke recognition scales increases the proportion of stroke patients who receive therapy in the first hours following the onset of stroke. Furthermore, FAST and MASS are the scales with the highest sensitivity, while CPSS is the scale with the highest specificity. When blood glucose measurement is possible in the prehospital setting, LAPSS and MASS are scales with sensitivities similar to that for CPSS and FAST but provide greater specificity for the recognition of stroke.

ILCOR FATF CoSTR Systematic Review on Stroke Recognition
The results of the search strategy for the 2015 ILCOR First Aid Task Force (FATF) systematic review on stroke recognition are presented in Table 6.

Global Search Strategy from Inception to May 2020
The results of the global search strategy are presented in Table 11.

APPENDIX B: Characteristics of excluded studies
The characteristics of excluded studies are presented in Table 12.        Explanations: a-Fast is integrated in a specific protocol call "rapid ambulance protocol" and compared with PCDs and ED doctor's diagnosis of stroke. No information about confounding factors.   Explanations: a-three studies have high risk of bias for patient selection, one has high risk of bias for reference standard, two have moderate risk of bias for reference standard and one has moderate risk of bias for flow and timing; b-One study includes the FAST in a protocol and does not test FAST only (Bergs, 2010), c-Inconsistency is considered as serious due to differences in study cohorts, qualification and training of test administrators, and differences in the reference standard.   [18].
Explanations: a-Very serious risk of bias due to high risk for patient selection (4/5) and reference standard, Moderate risk of bias due to reference standard (2/5) and flow and timing (1/5  Explanations: a-High risk of bias for patient selection (9 studies on 10) and unclear risk of bias for reference standard (8 studies on 10) and for flow and timing (9 studies on 10)  Explanations: a-serious risk of bias due to patient selection and unclear risk of bias due to reference standard and flow and timing TIA: Transient ischemic attack Additional Information Disclosures Human subjects: All authors have confirmed that this study did not involve human participants or tissue. Animal subjects: All authors have confirmed that this study did not involve animal subjects or tissue.

Conflicts of interest:
In compliance with the ICMJE uniform disclosure form, all authors declare the following: Payment/services info: Funding for the submission fees to CUREUS.COM was provided by the French Red Cross and the Global First Aid Reference Center. No other funding was requested or required for this review. Financial relationships: All authors have declared that they have no financial relationships at present or within the previous three years with any organizations that might have an interest in the submitted work. Other relationships: All authors have declared that there are no other relationships or activities that could appear to have influenced the submitted work.