Acute ischemic stroke (AIS) leads to severe disability if left untreated, particularly when it is caused by a large vessel occlusion (LVO) [1]. In 2015 however, when several randomized controlled trials showed safety and efficacy of endovascular treatment (EVT) for LVO stroke, the prognosis of these patients has dramatically changed. This evidence has translated into current guidelines, which recommend vascular imaging in all AIS patients upon hospital arrival. If an LVO is identified, EVT is recommended in patients who meet certain clinical inclusion criteria [2]. One of these criteria is the presence of disabling deficits, i.e., a baseline National Institutes of Health Stroke score of 6 or higher [2]. This clinical severity threshold forms the basis for clinical stroke triage tools such as the Los Angeles Motor Scale (LAMS) [3] and the Rapid Arterial Occlusion Evaluation scale (RACE) [4].

Pre-hospital triage of acute ischemic stroke patients: where are we at?

Systems of stroke care have organically grown over many decades, and as such, their set-up is unique in each country and each region. While in developing nations, healthcare systems largely lack the necessary resources and infrastructure to offer EVT at all, most industrial countries have successfully established EVT networks by now. However, EVT-capable centers are mostly clustered in urban regions, while access to EVT in rural areas is either lacking or associated with long transport times. Furthermore, capacities of EVT-capable centers are limited, which often necessitates treatment of EVT-ineligible stroke patients to primary stroke centers (PSCs), i.e., centers that offer intravenous thrombolysis (which is also effective in recanalizing vessel occlusion albeit to a lesser degree), but not EVT. This situation has generated a transport dilemma for acute ischemic stroke patients: When is it reasonable to bypass a nearby primary stroke center, thereby potentially delaying IV thrombolysis, in favor of direct transport to an EVT-capable hospital? Conditional probability modeling suggests that within cities, direct-to-mothership paradigms are generally more time-efficient, while in rural areas with large distances to the next EVT-capable center, drip-and-ship paradigms can be beneficial [5]. Importantly, current triage paradigms will also be affected by the results of trials evaluating the role of IV thrombolysis in patients undergoing EVT: the recently published DIRECT-MT trial for instance showed non-inferior outcomes with EVT as a stand-alone treatment compared to EVT with concurrent in a direct-to-mothership scenario [6]. The results of similar trials like MR CLEAN NO IV (ISRCTN80619088) are expected soon. If EVT these trials show outcomes without concurrent IV thrombolysis to be consistently non-inferior to EVT with concurrent thrombolysis, similar trials in drip-and-ship patients might get triggered. In short, determining the optimal transport paradigm for a given patient is highly complex and depends on various factors, including transport time, workflow efficiency at the primary and comprehensive center, and accuracy of the LVO screening tool used [7]. The common goal of such clinical LVO screening tools is to aid pre-hospital patient triage by identifying AIS patients with LVO, but their accuracy is limited. Depending on the screening tool used, more than 50% of patients could still suffer from an LVO, despite a negative screening result [8]. Thus, oftentimes, LVO patients are not correctly recognized as such and get transported to a PSC, where vascular imaging is performed and shows the occlusion. This relatively common scenario may result in time-consuming interhospital transfers and probably worse outcomes [9, 10], although more recent data from the “Direct Transfer to an Endovascular Center Compared to Transfer to the Closest Stroke Center in Acute Stroke Patients With Suspected LVO” (RACECAT) trial, which compared direct-to-mothership to drip-and-ship transport paradigm, suggests no difference in outcomes between the two transport paradigms [11].

Emerging challenges for stroke patient triage

With EVT indications continuously expanding, we are confronted with additional challenges in pre-hospital stroke triage. The ability of clinical pre-hospital triage tools to identify acute stroke patients who are potential EVT candidates, which was suboptimal to begin with, is likely to drop further once so-called mild or non-disabling LVO strokes, i.e., LVO patients with low baseline NIHSS, become EVT-eligible. At the time being, there is no level 1A EVT recommendation for LVO patients presenting with baseline NIHSS < 6 yet [2]. However, LVO patients with transient deficits and/or so-called minor strokes, i.e., low NIHSS strokes, are at a high risk of early neurological decline and recurrent stroke with subsequent severe disability [12, 13]. This might be part of the reason why the majority of physicians stated in a recent survey that they would offer EVT in low NIHSS LVO stroke [14]. Studies evaluating EVT benefit in these patients are exclusively non-randomized and show conflicting results: While some of them suggest superior outcomes with EVT compared to medical management [15, 16], others found no benefit of EVT [17, 18]. The randomized IN EXTREMIS [19] and ENDO-LOW (NCT04167527) trials which investigate the safety and efficacy of EVT in minor stroke due to LVO, will hopefully provide an answer to this controversy. In light of the recent improvements in thrombectomy technology and the high rate of neurologic deterioration in LVO stroke with low baseline NIHSS (24–38% [20, 21]), it seems overall highly likely that these patients will be increasingly treated with EVT. Since clinical LVO screening tools are based on symptom severity, by definition, they will not be able to detect LVO patients with low NIHSS. As a result, these patients will initially be transported to a PSC, and once the LVO has been diagnosed, need to be re-routed to an EVT-capable center, which leads to substantial time delays. On the other hand, one could argue that LVO patients with low NIHSS probably have good collaterals and are so-called slow progressors, since their brain function is relatively preserved. As such, even initial admission to a PSC followed by interhospital transfer and EVT would still yield a good result. However, this line of reasoning has two major flaws. Firstly, we cannot foresee if and at which point slow progressors turn into fast progressors, which is likely determined by numerous variables such as blood pressure and sympathetic tone, some of which might be unknown in the acute setting. Secondly, outcomes will inevitably be worse with increasing time delays, and even if proportions of good outcome were to be similar, the differences could become more apparent when excellent functional status (mRS 0–1) or cognitive function measures are used as outcomes.

While high sensitivity for LVO detection is paramount, one also has to acknowledge that among all patients who are screened by emergency medical services for suspected acute stroke, the prevalence of LVO is low, at around 5% [22]; i.e., in order not to overwhelm CSCs with non-LVO patients, a reasonably high specificity is needed.

Furthermore, EVT for medium vessel occlusions (MeVO) is moving at a rapid pace [23]. With current best medical management, MeVO stroke frequently results in severe morbidity [24], and many centers around the world therefore routinely treat MeVO strokes with EVT. At the time being, there is still lack of randomized evidence [25], but randomized trials are being planned. Detection of MeVOs, which often result in less severe clinical deficits due to their more distal occlusion location [24], will add additional challenges with regard to pre-hospital EVT triage.

In summary, it seems that with the expanding role of EVT, symptom-based pre-hospital stroke triage for endovascular treatment will soon reach a ceiling. The ability of current clinical triage tools to detect EVT candidates, which is already limited (Fig. 1a), will dramatically drop once low NIHSS LVO strokes become EVT-eligible (Fig. 1b). At the same time, omitting triage and admitting every stroke patient to an EVT-capable center is not a feasible solution either, due to various political and practical reasons, and because such an approach would by far exceed the capacities of most EVT-capable stroke centers. At first glance, simply lowering the symptom severity threshold of current clinical stroke triage tools seems like a convenient solution, since it requires no additional training or ambulance equipment and would allow us to maintain a high sensitivity for LVO strokes, but this comes at a cost of decreased specificity, i.e., a high number of EVT-ineligible patients being erroneously transported to EVT-capable centers (Fig. 1c). Clearly, more accurate solutions for pre-hospital stroke triage are needed.

Fig. 1
figure 1

Pre-hospital EVT triage based on clinical symptom severity scores. a Current pre-hospital EVT triage situation. b Expected challenges with currently established pre-hospital LVO triage algorithms once EVT for low NIHSS LVO stroke has been established. Many LVO patients with low NIHSS will be erroneously identified as EVT-ineligible and transported to a PSC, which means they will have to undergo interhospital transfer in order to receive EVT. c Expected challenges when maintaining currently established clinical pre-hospital LVO triage tools with lower symptom thresholds once EVT for low NIHSS LVO stroke has been established. Lowering symptom severity thresholds of clinical LVO screening tools will maintain sensitivity for detection of EVT candidates, but this comes at a cost of decreased specificity; i.e., many non-LVO patients will get erroneously admitted to a CSC, which might overwhelm CSC capacities

Full size image

Beyond symptom-based stroke triage: new LVO detection technologies

Novel pre-hospital triage technologies that are based on imaging findings or analysis of parenchymal activity, are able to overcome the inherent limitations of symptom-based stroke triage, since they do not rely on clinical deficits (suppl. table 1).

Mobile stroke units (MSU), i.e., ambulances that are equipped with a CT scanner, allow for direct visualization of the occluded blood vessel in the field [26]. At the time being, only few MSUs are equipped with CT angiography capacities, but depending on the demand, this might change soon [27]. LVOs can sometimes be seen even on non-contrast CT when using the right slice thickness and reconstruction kernel, and with newer technologies, better imaging quality, and machine learning–based algorithms, it might also become possible to reliably and consistently detect intracranial occlusions on non-contrast CT [28]. This would obviate the need for CTA capacities and facilitate imaging in the MSU, as non-contrast CT imaging can be performed by trained technicians in many jurisdictions, whereas for contrast-enhanced studies, a physician usually has to be present [29]. That being said, MSU staff requirements, acquisition, and maintenance costs are substantial, and depending on the distance of the MSU home base to the patient’s location, use of an MSU might result in substantially longer travel times compared to a “conventional” ambulance [30].

Smaller, portable devices would constitute a more practical solution, as they could be installed in regular ambulances. First studies on transcranial Doppler ultrasound for LVO detection have shown encouraging results [31, 32], but in practice, performing the exam is often difficult due to an insufficient acoustic window [33]. MeVO detection using ultrasound-based technologies will be even more challenging. Electroencephalography (EEG)-based techniques [34] have also shown promise. As opposed to ultrasound, they rely on parenchymal signal alterations rather than detection of the occlusion site itself. One potential problem of this approach is that relative preservation of brain function, as it is the case in patients with mild clinical deficits, might not result in signal abnormalities that are conspicuous enough to be detected. However, once refined and appropriately validated, one or more of these technologies will hopefully allow for efficient detection of EVT candidates independent of clinical deficits. Cost-effectiveness is another important point to consider when deciding about the optimal triage technology. One advantage of the currently used tools is that they entail no costs for additional equipment. Modeling of pre-hospital stroke triage and transport routes is now possible and could help local hospitals and policymakers to decide about the optimal pre-hospital triage strategy for their local environment given their available resources and keeping all factors, including cost-effectiveness, in mind [35, 36].

Time to think ahead

It is important to note that all of the above-mentioned technologies are still in their infancy and far from being used in clinical routine. However, with constant improvements in EVT technology and expanding treatment indications, it becomes obvious that EVT candidates cannot be reliably distinguished from EVT-ineligible patients solely based on clinical deficits. Pre-hospital stroke triage will need to accommodate to this new reality. Should we wait for definitive evidence of EVT benefit for minor strokes, or should we prepare ahead of time? We personally believe that we should start right now, by developing and testing alternative pre-hospital triage tools, understanding ambulance design, interacting with emergency medical services providers, and augmenting the capacities of EVT-capable stroke centers, since it is likely that more and more acute ischemic stroke patients will become potential EVT candidates.