Effects of state-wide implementation of the Los Angeles Motor Scale for triage of stroke patients in clinical practice

Background The prehospital identification of stroke patients with large-vessel occlusion (LVO), that should be immediately transported to a thrombectomy capable centre is an unsolved problem. Our aim was to determine whether implementation of a state-wide standard operating procedure (SOP) using the Los Angeles Motor Scale (LAMS) is feasible and enables correct triage of stroke patients to hospitals offering (comprehensive stroke centres, CSCs) or not offering (primary stroke centres, PSCs) thrombectomy. Methods Prospective study involving all patients with suspected acute stroke treated in a 4-month period in a state-wide network of all stroke-treating hospitals (eight PSCs and two CSCs). Primary endpoint was accuracy of the triage SOP in correctly transferring patients to CSCs or PSCs. Additional endpoints included the number of secondary transfers, the accuracy of the LAMS for detection of LVO, apart from stroke management metrics. Results In 1123 patients, use of a triage SOP based on the LAMS allowed triage decisions according to LVO status with a sensitivity of 69.2% (95% confidence interval (95%-CI): 59.0–79.5%) and a specificity of 84.9% (95%-CI: 82.6–87.3%). This was more favourable than the conventional approach of transferring every patient to the nearest stroke-treating hospital, as determined by geocoding for each patient (sensitivity, 17.9% (95%-CI: 9.4–26.5%); specificity, 100% (95%-CI: 100–100%)). Secondary transfers were required for 14 of the 78 (17.9%) LVO patients. Regarding the score itself, LAMS detected LVO with a sensitivity of 67.5% (95%-CI: 57.1–78.0%) and a specificity of 83.5% (95%-CI: 81.0–86.0%). Conclusions State-wide implementation of a triage SOP requesting use of the LAMS tool is feasible and improves triage decision-making in acute stroke regarding the most appropriate target hospital.


Background
Stroke is a frequent cause of disability and death [1] with important medical and economic implications. When stroke is caused by large-vessel occlusion (LVO), trials provide compelling evidence that mechanical thrombectomy (MT) rather than medical treatment alone is most effective. Even so, although an estimated 10 to 25% of patients with stroke have LVO [2,3], only a small minority (fewer than 7%) are treated with MT [2,4,5]. A main reason for this difference is that MT is not available at the many primary stroke centres (PSCs) but is offered only by a few specialized stroke centres (comprehensive stroke centres, CSCs).
Currently, stroke-management guidelines recommend the transport of all patients to the nearest stroke-ready hospital [6], which in most cases is a PSC not offering MT. Therefore, patients with LVO may secondarily be transferred to a CSC for thrombectomy. Importantly, compared with direct referral to a CSC, such interhospital transfers cause pronounced treatment delays ranging from 96 min to 111 min for patients with LVO [7][8][9][10]. Consistent with the "time is brain" concept [8], these delays significantly worsen clinical outcomes [9,11]. On the other hand, not all patients with stroke should be transferred to CSCs because, apart from overwhelming already strained accident and emergency departments, bypassing PSCs could delay the administration of intravenous thrombolysis for most patients.
LVO is mostly associated with more severe stroke symptoms. Therefore, researchers have proposed the use in the field of stroke severity scales aimed at prehospitally detecting LVO; patients with LVO could then profit from direct transfer to a CSC. Recently, many LVO scales, such as the Los Angeles Motor Scale (LAMS) [12], the Rapid Arterial Occlusion Evaluation Scale (RACE) [13][14][15] or the A2L2 test (A, arm; L, leg) [16] have been proposed and studied by EMS in regard to their accuracy of detecting LVO.
Because the LAMS, apart from belonging to the most predictive of those scales [17,18], requires assessment of only three motor symptoms (facial paresis, arm strength, and grip strength) [12,19], it appears most suitable for implementation in prehospital emergency care protocols.
However, information on the effects of implementation of the LAMS for real-life triage of stroke patients is still missing. The aim of this study is to explore the effects of a state-wide triage standard operating procedure (SOP) requesting performance of the LAMS for triage decisionmaking in regard to the most appropriate target hospital.

Patients and study design
This prospective multicentre study, coordinated by the University of the Saarland, Germany, was opened on March 1, 2018, and terminated on June 30, 2018. The trial was conducted in the federal state of Saarland in Germany, a mixed urban and rural state with an area of 1004 sq. mi (2571 km 2 ) and approximately 992,000 inhabitants, in the context of a state-wide network of all certified stroke centres (eight PSCs, two CSCs). All emergency calls in this state are evaluated by a single dispatch centre. Inclusion criteria were suspicion of acute stroke by the EMS personnel on-scene, age of at least 18 years, and willingness to participate. Exclusion criteria were referral modes other than via EMS and critical illness requiring immediate transfer to the nearest intensive care unit.

Triage SOP based on the LAMS instrument
The stroke triage SOP was a rule for all EMS stations in the entire state. In this SOP, the following factors were considered in triage decision-making: (1) LAMS score (cut-off score of 4) [19]; (2) symptom onset times of 8 h or less or presence of "wake-up" stroke; and (3) qualityof-life aspects, such as severe comorbid conditions and severe prestroke dependency [20].
The triage protocol was set in operation by the central EMS coordinating authority of the state, the Zweckverband für Rettungsdienst und Feuerwehralarmierung, Saar, on May 30, 2015 (VAW MED-012) [20], and was accompanied by state-wide structured training sessions held approximately every 3 months for EMS personnel in the field and every 6 months for dispatch centre personnel in the context of the 30 educational sessions per year required for EMS personnel in the field. Moreover, the protocol was a component of the educational curriculum of the state's EMS school.

Endpoints
The primary endpoint was accuracy of the triage SOP in triaging stroke patients to the appropriate target hospital, PSC versus CSC. Correct triage decision was defined as a decision to transport patients with LVO to the nearest CSC and to transport patients without LVO to the nearest PSC (or to a CSC, if this was the nearest stroke-treating hospital). LVO was defined as occlusion of the intracranial internal carotid artery, the proximal (M1) segments of the middle cerebral artery, or the basilar artery.
Secondary endpoints were performance of the triage protocol with regard to either LVO or intracranial haemorrhage (ICH), the number of secondary interhospital transfers, and the sensitivity and specificity of the LAMS in detection of LVO itself. Documented stroke management metrics included times from call to (1) "on scene", (2) hospital admission ("door"), (3) first neurologist contact, (4) start of non-contrast imaging, (5) start of vascular imaging (in case of ischaemic stroke), (6) needle, (7) door of CSC after secondary transfer, and (8) groin puncture. To determine short-term outcomes, we assessed mRS scores at discharge and mortality rates.

Geocoding
For analysis of the potentially alternative conventional pathway of the guideline-recommended transfer to the nearest stroke-treating hospital, geocoding was performed for each patient based on postal code, as previously described [21]. This approach allowed exact identification of the level of care (PSC or CSC) offered by the nearest stroke-treating hospital to which the individual patient with stroke would normally have been transferred.

Ethical aspects
The study protocol, the informed consent document, and the subject information document were approved by the Ethics Committee of the Medical Association of the Saarland, Germany (AZ-257/17). Informed consent was obtained from all patients or their legal representatives and was documented in the receiving hospitals. The otherwise separated datasets of the EMS and of the hospital were pseudonymized and linked only by the information from (1) the target hospital, (2) the admission date, (3) the patient's sex, and (4) the patient's year of birth.

Statistical analyses
Results are reported according to Strengthening The Reporting of OBservational Studies in Epidemiology (STROBE) guidelines [22]. Group comparisons were analysed with the Mann-Whitney U Test or the Fisher exact test. Statistical analyses were performed with IBM SPSS Statistics for Windows, version 25.0.0.0 (IBM Corporation, Armonk, NY, USA).

Demographic and medical characteristics
After the 4-month study period, the study was terminated with a total of 1123 enrolled patients. Exclusion of patients is specified in Fig. 1. Baseline demographic and medical characteristics of the patients are presented in Table 1. Of the 1123 patients, 644 (57.3%) had experienced an acute ischaemic stroke, 84 patients (7.5%) had experienced a haemorrhagic stroke, and 395 patients (35.2%) had experienced stroke mimics. Of the 644 patients with ischaemic stroke, 129 (20.0%) underwent thrombolysis; and 489 underwent vascular imaging, which showed that 78 (12.1%) had LVO; 53 of these 78 patients (67.9%) were treated with MT.

Feasibility and performance of the LAMS-based triage SOP in triage of stroke patients
This study demonstrated the feasibility of state-wide implementation of a LAMS-based triage SOP with high adherence. Only 60 of 1123 patients (5.3%) had no documented LAMS score, and 941 of the 1123 patients (83.8%) were triaged according to the LAMS score ( Table 1). The triage SOP allowed triage to the appropriate hospital with a sensitivity of 69.2% (95% confidence interval (95%-CI): 59.0-79.5%) and a specificity of 84.9% (95%-CI: 82.6-87.3%) ( Table 2). In contrast, if the current guidelines had been adhered to (transfer to the nearest stroke-treating hospital), patients with LVO would have been correctly triaged with a sensitivity of 17.9% (95%-CI: 9.4-26.5%) and a specificity of 100% (95%-CI: 100-100%).
Similar results were obtained when the evaluation included ICH as a condition for triage to a CSC (Table 3). In accordance with this high triage accuracy, only 14 of the 78 patients with LVO (17.9%) required secondary transfer.

Stroke management metrics and short-term outcomes
Stroke management metrics are displayed in Table 4. The median call-to-needle times for patients transferred to a PSC were shorter than those for patients transferred  to a CSC, and, conversely, call-to-groin puncture times were longer for patients transferred to a PSC than for those transferred to a CSC (Table 4).

Discussion
The current guideline-recommended practice of transferring each patient to the nearest stroke-treating hospital, usually a hospital not offering thrombectomy, may delay or even preclude thrombectomy for patients with LVO. The results show that state-wide implementation of an SOP requesting the use of LAMS for triage decisions is feasible with high adherence and that this intervention can improve triage decision-making in regard to the appropriate target hospital in clinical practice.
Our evaluation of the effects of the state-wide implementation of an EMS SOP based on the LAMS achieved a sensitivity of 69.2% and a specificity of 84.9% in transferring patients with LVO to the appropriate target hospital. In accordance with such a high accuracy of triage decisions, we observed a low rate of secondary transfers (only 17.9% of LVO patients). In contrast, the conventional, guidelinerecommended approach of transferring every patient to the nearest stroke-treating hospital, as determined by geocoding for each of the patients, would have achieved a sensitivity of only 17.9% and a specificity of 100%, a finding supporting the benefit of using the LAMS-based SOP. (The high specificity achieved by the modelled conventional pathway can be explained by the fact that "non-LVO" patients would always arrive at the "correct" hospital: the  Similar values were obtained when ICH was included as a "target condition" for triage to a CSC. Although increasing evidence indicates that the "time-is-brain" concept is also valid for patients with haemorrhagic stroke and that these patients could also benefit from rapid  specialist treatment at a CSC [12,23], evidence from randomized studies regarding primary transfer to CSCs is still scarce [24]. Regarding the LAMS itself, this instrument detected LVO with a sensitivity of 67.5% and a specificity of 83.5% and detected either LVO or ICH with a sensitivity of 56.6% and a specificity of 86.7%. Calculating the accuracy of a variety of alternative LAMS cut-off scores confirmed that the cut-off score of 4 is appropriate. Thus, these are within the range of results of most earlier validation studies on the LAMS. E.g., a subgroup of 94 patients from the FAST-MAG trial who underwent vascular imaging found that the LAMS detected LVO with a sensitivity of 76% and a specificity of 65% [12], and two European studies reporting sensitivities of 63 and 38%, and specificities of 84 and 93%, respectively [17,18].
However, these earlier studies did not use their LAMS results for triaging patients in clinical practice. At the same time, the results also corroborate the conclusions of previous meta-analyses indicating that all of the LVO scales tested still miss a substantial proportion of LVOs [25,26].
While the strengths of this study are the pseudonymized linkage of separated prehospital and in-hospital data sets, and the high degree of data completeness, one limitation of the study is the lack of a control group. The alternative of retrieving historical data before implementing the SOP may, however, have caused a bias by secular effects, and randomization was not possible in this state-wide EMS rule. However, geocoding allowed us to exactly identify the level of care offered by the nearest stroke-treating hospital in each patient if the conventional practice had been applied. Furthermore, in this real-life-study, vascular imaging has not been performed in all patients admitted to the ten hospitals, thus resulting in omission of 24% of stroke patients from evaluation. In the future, triage accuracy may be further enhanced by improved scales, by the inclusion of additional factors such as screening tools for treatment eligibility or for mimicking conditions [27], or by additional telemedical consultation with the CSC team [16].

Conclusion
The prehospital identification of stroke patients with LVO that should be directly transported to a thrombectomy capable centre is an unsolved medical problem. Here, we show for the, to our knowledge, first time that a state-wide EMS SOP for stroke management requesting determination of the LAMS score is feasible with high adherence and can indeed be beneficial in triage decision-making regarding the most appropriate target hospital.