Study design
The project is divided into three stages:
Stage 1: First, based on existing scores that have been successfully applied in previous studies,28 we develop a dedicated LVO screening tool suitable for use during telephone calls in close collaboration with dispatchers at the emergency control centre. The identification of stroke patients as candidates for thrombectomy presenting a left-hemispheric syndrome is based on the directed assessment of the presence of a right arm paresis in combination with at least one of the following cortical symptoms: gaze deviation to the left, and/or aphasia, and/or erroneous performance of a complex task (earlobe-test: the fingers of the unaffected arm are required to reach to the contralateral earlobe). In the case of a right-hemispheric syndrome, the score assesses the presence of left arm paresis in combination with gaze deviation to the right, and/or neglect and or erroneous performance of the earlobe-test (without crossing the midline).28 The test result is dichotomous (positive when unilateral brachial paresis and congruent cortical deficits are present, otherwise negative). The identification of cortical deficits is also a central element of successfully utilised pre-hospital screening tools such as the RACE score19 or the Vision, Aphasia, and Neglect (VAN) scale.29
Second, we test the score’s feasibility by applying it in a simulation study, including 48 volunteers without a medical background, eight actors, and 24 medical staff (dispatchers at emergency control centres, EMS personnel, and nurses). Currently, there is little experience in using such scores during the emergency call involving LFR. We aim to achieve high sensitivity (≥0.75) and specificity (≥0.7) in the range of previously published analyses of pre-hospital stroke scales.30 Followed by the prospective validation of the LVO score, where EMS dispatchers apply the LVO score in real-life emergency calls to screen patients with suspected stroke but still alert the usual rescue chain in parallel (N=500). The research staff assesses the validity of the LVO score.
For the implementation of the score, we develop training seminars adapted to each occupational group. The effectiveness of the EMS professionals’ training is evaluated in EMS dispatchers, emergency doctors, and emergency rescue staff. Effectiveness will be tested by means of an online pre-/post-test, including case-based examples and knowledge testing, before, directly after, and three months after the training. The test comprises general questions on stroke recognition, management, treatment, as well as specific questions about the intervention itself (application of the LVO score in the field). The results allow for both descriptive and quantitative statistical analysis. In addition, all the emergency professionals involved in the rescue procedure have access to the LVO score via a mobile app, specifically created for the score assessment at the emergency site and the consolidation of knowledge by text and video explanations.
Stage 2: Consecutively, we performed a controlled study with a stepped-wedge design and clusters of six emergency control centres for gradual implementation of the intervention procedure. The intervention and control groups each include 125 patients (N=250). Enrolment began in early summer 2021. During the control phase, emergency control centres carry out emergency interrogation and dispositions in accordance with local standards. Subsequently, the six participating emergency districts move gradually from the control to the intervention phase (stepped-wedge design). Control data is available from the control phase of this design and existing retrospective datasets from patients who received MT in the Department of Neuroradiology, University Medical Centre Freiburg, between 2012 and 2020. A retrospective application of the LVO screening tool on historical cohorts is not feasible, as these examination steps are neither part of the routine clinical examination nor of frequently used stroke-screening tools, such as the National Institutes of Health Stroke Scale (NIHSS). 90 days post-stroke, a study staff member blinded to group allocation will enquire the mRS by phone from patients or their guardians (unblinded treatment and blinded endpoint evaluation).
Stage 3: A formative process evaluation assesses the intervention’s feasibility. Interviews start at two time-points, during and at the end of the intervention phase, with EMS dispatchers, emergency doctors and EMS staff, air rescue staff, and LFR. The feedback from the first interview round will be used to optimise the intervention, if necessary. A workshop will be held for all the professionals involved in the study to discuss both the data and their experience with applying the procedure. This workshop aims to devise improvements for the implementation of the intervention during stage 2. We interview LFR about their level of psychological stress during the emergency call and the clarity of the EMS dispatchers’ instructions. The survey conducted among EMS professionals focuses on the implementation and feasibility of the LESTOR approach, benefits and undesirable effects of the intervention.
Statistical analysis
Stage 1: During the validation of the new LVO score, we determine the test’s performance, e.g. sensitivity, specificity and the percentage of correct classifications. Inter-observer reliability is assessed by Krippendorff’s Alpha and Fleiss’ Kappa. We assess the internal validation by performing cross-validation, and the external validation by comparison of the score results with vascular imaging. The knowledge gained by EMS professionals from dedicated training is evaluated in a pre-test/post-test design.
Stage 2: Ordinal logistic regressions to predict mRS will be conducted based on the following variables:
-
Independent variables: group assignment (intervention vs. control)
-
Confounders (control variables): e.g. age, severity of stroke (NIHSS) at onset, anamnestic data, such as atrial fibrillation, diabetes mellitus, and the location of the occlusion of the affected arteries.
We will calculate an adjusted odds ratio for a change in the mRS. Favourable clinical outcome is defined as mRS ≤2 (similar to Goyal et al.).8 For the sensitivity analysis, we will use different cut-off values and calculate 95% confidence intervals.
We will model a cluster membership corresponding to recommendations in the literature.31 We will perform imputation of missing values for the predictor variables, but not for the outcome variables. Missing values might occur during three different phases: First, during pre-hospital care, LVO patients might not be recognised as such in the emergency control centre and/or on site. Possible scenarios include, e.g. language barriers or atypical presentations of stroke. In these cases, the pre-hospital LVO scoring might be lacking. Second, missing data might originate during hospital care due to an early patient transfer to external hospitals after MT. As a study nurse follows up patients stringently on time, these cases will be identified quickly and missing data can be tracked. Third, information on the clinical outcome might not be obtained in cases where the systematic follow-up by our study nurse is not successful. Based on our experience during the last >10 years, we expect a high rate of complete follow-up data of >90-95%.
We will perform a complete case analysis if the proportion of missing data is below 5%. Above 5% missing data, multiple imputations with an expectation maximisation algorithm will be conducted.
A power analysis shows that testing for the statistically significant superiority of the intervention over the control group requires far more cases (N=664) than available in our target region during the study period. We therefore conduct a controlled study to explore achievable effects and practicability under routine care conditions. Due to limitations of the sample size, we will explore the clinical significance of the intervention effects and refrain from testing for the statistically significant superiority of the intervention over the control group.
Furthermore, we calculate a propensity score, which represents the probability of a patient with a particular set of initial values receiving the intervention. This enables the monitoring of group differences that may occur in a non-randomised allocation. A cost analysis will be carried out as the first part of the health economic evaluation. For this purpose, the Medical Controlling Department at the University Medical Centre Freiburg will calculate the costs incurred during an in-patient stay for both patients in the intervention and the control group. This data is available, as the University Medical Centre serves as a reference hospital in the maintenance and further development of the German Diagnosis Related Groups (G-DRG) system via the InEK Institute [Institute for Hospital Remuneration Systems].
To determine the costs of the intervention, the training costs and the averaged expenses for the additional helicopter missions are used: if the LVO score, applied by EMS dispatchers, falsely indicates a patient having LVO, the air rescue might have to return to base without a patient, resulting in additional costs. Hence, we will compare the costs of the intervention and control group.
As a final step, a cost-effectiveness analysis will be performed, whereby the primary endpoint (functional capacity, documented according to the mRS) will be expressed in relation to the total costs. Secondary endpoints are, amongst others, length of pre-hospital, in-hospital and, if applicable, inter-hospital treatment. We also asses the following:
-
clinical criteria (e.g., difference between NIHSS at admission and discharge, length of hospital stay, number of hemicraniectomies and mortality)
-
imaging criteria at initial imaging:
-
ASPECTS (Alberta stroke programme early CT score)
-
ischemic core volume (ml)
-
perfusion-lesion volume (ml)
-
mismatch (core/penumbra)
-
early signs of ischemia
-
occlusion side
-
extra- and intracranial location of the occluded segment
-
TICI (thrombolysis in cerebral infarction) score
-
bleeding type, side, diameter (mm), and location
-
imaging criteria at follow-up:
As logistics in stroke care is complex, we will systematically record time metrics regarding both the pre-hospital (e.g., emergency call to hospital admission), and intra-hospital processes (e.g., door to recanalisation). This might streamline the management of LVO stroke patients.
Stage 3: Trained interviewers conduct guided, semi-structured telephone interviews. The interviews allow a broad exploration of diverse aspects of a phenomenon when there are few theoretical and empirical findings within a particular research field, as in the present study.32 Interview questions deal with the implementation of the intervention, its barriers, any unfavourable side effects, problems, and solutions. The interviews are recorded on tape and transcribed. The qualitative content analysis is performed according to Mayring.33