Utility of Severity-Based Prehospital Triage for Endovascular Thrombectomy: ACT-FAST Validation Study

Increasing global uptake of endovascular thrombectomy (EVT) for patients with large vessel occlusion (LVO) and limited endovascular-capable comprehensive stroke centers (CSCs) in many regions of the world have created an urgent need for LVO-triage strategies. Initial transport of patients to the nearest primary stroke center without EVT capability and then undertaking secondary interhospital transfer to a CSC delayed onset to arterial access by 114 minutes in EVT trials.¹ Delayed EVT is strongly associated with poorer outcomes in most randomized trials,² and patients requiring secondary transfer have worse outcomes post-EVT.^3,4 While a post hoc analysis of the DEFUSE-3 trial (Endovascular Therapy Following Imaging Evaluation for Ischemic Stroke 3) did not show this time effect, this is likely to be related to bias due to the imaging selection applied after CSC arrival to exclude patients with unfavorable perfusion imaging.⁵

Severity-based clinical triage tools allow paramedics to identify patients likely to require CSC-level care, as LVO and intracranial hemorrhage (ICH) typically have more severe neurological deficits. This allows direct ambulance transport to the nearest CSC, rather than closer non-EVT thrombolysis centers. Avoiding delays to EVT and neurosurgical intervention incurred by secondary transfers may potentially improve patient outcomes.

The use of severity-based triaging has attracted criticism due to the diagnostic inaccuracies inherent in a clinically based system.^6,7 A systematic review conducted for the American Heart Association concluded that triage tools had insufficient specificity and would, therefore, lead to excessive hospital bypass due to false-positive classifications.⁸ This would potentially result in unnecessary resource expenditure, CSC overburdening and delayed thrombolysis. Attempts to improve specificity often adversely affect sensitivity and could lead to more LVO cases being missed. This led to the initial recommendation against severity-based triaging in the 2018 American Heart Association stroke guidelines. Subsequently, the 2019 guidelines gave a class IIb recommendation that encouraged bypass of alteplase-ineligible patients with likely LVO directly to EVT-capable centers, but stated that the benefit was uncertain in alteplase-eligible patients.⁹

To date, many severity-based triage tools have not undergone adequate validation in large prehospital samples. We, therefore, aimed to investigate the effects, both positive and negative, of a severity-based triage strategy in a large real-world paramedic validation study using the ambulance clinical triage for acute stroke treatment (ACT-FAST) severity-based triage algorithm.¹⁰ Our key objectives were to quantify diagnostic accuracy for LVO and the need for CSC-level care, as well as the potential facilitation of more rapid EVT commencement. Concurrently, we aimed to quantify the LVO cases missed, potential delays in thrombolysis due to bypass, and CSC overburdening.

The data that support the findings of this study are available from the corresponding author upon reasonable request.

In this paramedic prehospital validation study, we have shown the utility of the ACT-FAST severity-based algorithm in providing high specificity for recognition of LVO and accurate prediction of the need for CSC-level care. Over 50% of all metropolitan patients in the study undergoing EVT would have benefitted from bypass directly to CSC, with a calculated median time saving to EVT commencement of 52 minutes through avoidance of secondary transfers. In comparison, only a small fraction of patients (3% of all non-LVO ischemic strokes in the study) would have bypassed the closest thrombolysis-capable center with potentially delayed thrombolysis. This delay was calculated to be around 10 minutes if CSC and non-CSC centers have comparable door-to-needle times (a conservative assumption). Using ACT-FAST, CSC overburdening by bypassed patients not requiring CSC expertise would likely be limited to approximately one extra patient per week for each existing EVT center in our region. This suggests a substantial overall benefit of severity-based triage in directing appropriate patients to CSCs and expediting EVT, with little evidence that thrombolysis delays and excessive CSC overburdening would outweigh these advantages.

Our study also showed that the 2-examination step ACT-FAST algorithm performed favorably in comparison to more complex triage scales used worldwide, though formal comparison is not possible due to different study populations. Of note, ACT-FAST has been shown in an independent validation sample to be superior to other triage scales for detection of ICA/M1 occlusions in the extended time window.²⁵ Importantly, the simplicity of the algorithmic approach allowed paramedics in this study to be trained solely using an 8-minute video, without the extensive in-person training and regular refresher training required for other tools.

A minority of patients with LVOs and EVT were not identified by the ACT-FAST algorithm, generally in patients with low baseline stroke severity or with stroke subtypes (especially basilar occlusion) poorly recognized by severity-based tools. In current practice, more mildly affected patients with less evidence for EVT (eg, M2-middle cerebral artery or partial occlusions) are not universally recommended treatment, reducing the value of identification by clinical triaging methods. Although some groups have advocated that greater detection of LVO should be preferred due to the powerful treatment effect of EVT,²⁶ there is a strong association between low baseline severity LVOs and better collaterals. These patients are likely to have slow ischemic core growth and, therefore, benefit less from the time savings derived from bypass to CSC.²⁷ Trading off specificity is likely to disproportionately lower positive predictive value and cause over-triage in all but the most well-resourced regions.

The American Heart Association systematic review of severity-based triaging highlighted the criticism of triage tools (excluding ACT-FAST) that did not display both high sensitivity and high specificity.⁸ Our study showed that the sensitivity of the ACT-FAST algorithm is high for ICA/M1/basilar LVOs (82.6%), and that false-negative patients have milder deficits and less certain treatment benefit. While specificity for LVO was 77.9%, the clinically relevant reference is requirement for CSC-level care, which includes ICH and intracranial tumors. Using this expanded definition, the specificity was 89.5%, generating a positive predictive value of 80.0%. As a result, excessive triage to a CSC is likely to be minor, as most CSCs incorporate both EVT and neurosurgical expertise.

Several aspects of our study suggest that prehospital triage is likely to increase the rate of EVT performed. The substantial time saving to EVT achieved by prehospital bypass may place some patients within 6 hours when broader eligibility criteria are applied rather than in the >6-hour imaging-selected time window. There were also patients with rapidly progressing stroke who were excluded from EVT due to extensive established noncontrast CT changes or large estimated ischemic core, and these may have been EVT-eligible if bypass delivered them to a CSC substantially earlier. There were also a small proportion of patients with LVO not referred for EVT by the non-CSC center, despite having no apparent contraindication. Had these patients been bypassed directly to CSC, they may have benefitted from EVT.

We acknowledge that, even with faster delivery of patients to a CSC, not all patients with ACT-FAST positive LVO will be eligible for EVT. However, these patients have severe strokes that may require neurocritical care or hemicraniectomy and, therefore, may benefit from the experience of CSC stroke teams. This is analogous to the currently accepted system of care for thrombolysis, which transports all patients with suspected stroke to a thrombolysis-capable center to receive expert assessment for treatment eligibility, despite only ≈15% of patients with ischemic stroke actually receiving treatment.²⁸ Furthermore, the reason for one-third of LVOs not receiving EVT in our study was premorbid disability, indicating that the ACT-FAST eligibility criteria for EVT were inconsistently followed and could be improved with further paramedic education.

The additional ambulance travel time that is acceptable to reach the CSC is still currently unknown. Current American Heart Association Mission:Lifeline stroke guidelines endorsing severity-based triaging recommend that the total travel time to CSC be no >30 minutes and not risk placing the patient outside the 4.5-hour time window for thrombolysis.²⁹ However, modeling suggests that additional travel time <30 minutes in metropolitan and <50 minutes in rural settings remains beneficial.³⁰ Based on our study, the additional median travel time of 10 minutes in our metropolitan region would easily meet such criteria. In rural and regional areas, bypass may be problematic because of longer transport times to the CSC. Nonetheless, our data indicate that the majority of these patients will subsequently require secondary transfer to the CSC and initialization of transport services could be considered in patients with ACT-FAST positive.

Our current study has several limitations. It was observational in nature with time benefits derived from modeling. However, analyses were performed at the individual patient level where possible using a highly accurate model of ambulance transport estimation validated against prospective real-world travel times. Further definitive group data on time savings will be provided from ongoing randomized controlled trials of severity-based triaging (eg, RACECAT [Direct Transfer to an Endovascular Center Compared to Transfer to the Closest Stroke Center in Acute Stroke Patients With Suspected Large Vessel Occlusion], URL: https://www.clinicaltrials.gov; Unique identifier: NCT02795962; TRIAGE-STROKE [Treatment Strategy in Acute Ischemic Large Vessel Stroke], URL: https://www.clinicaltrials.gov; Unique identifier: NCT03542188). We calculated time differences from using a bypass strategy from scene of ambulance attendance rather than emergency services activation, as the time from emergency activation to scene arrival would not change under either strategy. We also calculated potential thrombolysis delay using all patients with ACT-FAST positive transported to non-CSCs because of the very low number (n=3) of false-positive cases that actually received thrombolysis. We additionally acknowledge that true-positive EVT cases may also potentially receive delayed thrombolysis should they be eligible. However, given the greater disability benefit per minute of faster thrombectomy versus thrombolysis, this would not outweigh the net benefit of bypass. The estimated time-saving to commencement of EVT using ACT-FAST bypass is potentially conservative if prehospital notification of ACT-FAST-positive status, permitting early activation of neurointerventional services, and other workflow gains reduce door-to-arterial access time for direct-presenting patients.

Our results are based on current systems of care in Melbourne, Australia and may not be generalizable to other regions. Travel distance between CSC and non-CSCs in other localities will alter assumptions about time savings, while diagnostic accuracy may vary with the expertise of local emergency personnel or where an alternative triage tool is used. In our region, emergent neurosurgical services are limited to CSCs, and we considered patients with ICH and intracranial tumor to be appropriate for CSC-level care under a bypass strategy. We base this on the ability to provide faster emergent neurosurgical intervention at the CSC, better facilities for neurocritical care monitoring, avoidance of immediate or delayed secondary transfers and preliminary evidence that patients with ICH have better outcomes at neurosurgical centers, regardless of whether they receive surgery.³¹ However, we acknowledge that it is not universally agreed that all ICH or tumor patients will benefit from neurosurgical services, and this would lower the quoted specificity for patients needing CSC-level care. There also appeared to be a bias towards recording ACT-FAST status in patients presenting with more severe clinical deficits, as evidenced by the LVO prevalence of 25.5% in the study compared with previous estimates of ≈15% in our region.³² This indicates that there may have been under-representation of more mildly affected patients.

Alternatives to severity-based triaging for expediting EVT include prehospital imaging though mobile stroke units, or even mobile neurointerventional teams able to perform EVT at the local stroke center.³³ Resourcing and cost limitations mean these options are not feasible in all regions and cannot operate at all hours. A previous comparative study found that LVO triage using a mobile stroke unit was significantly better than the Los Angeles Motor Scale.¹⁹ However, these approaches may be complementary. Widespread use of the ACT-FAST algorithm in our paramedic workforce provides important support in redirecting the Melbourne Mobile Stroke Unit service to patients most likely to benefit, indicating potential for synergy of prehospital solutions. Another strategy of creating more CSCs was recently compared with using severity-based triage across the United States, with the finding that severity-based triage would increase access to EVT for a substantially larger proportion of patients with stroke compared with a more costly increase in EVT centers.³⁴ As such severity-based triage tools currently remain the least expensive and most generalizable option for prehospital triage.

In this real-world paramedic validation study of the severity-based ACT-FAST triage algorithm, adopting a prehospital bypass to CSC strategy would substantially reduce the need for secondary ambulance transfers and allow faster EVT by an estimated 52 minutes in a metropolitan setting. Based on the ACT-FAST diagnostic accuracy, the relative proportion of inappropriately bypassed patients is modeled to be low under conservative assumptions, with an even smaller proportion receiving potentially delayed thrombolysis by ≈10 minutes. Our study, therefore, suggests a substantial overall benefit of EVT triage using the ACT-FAST algorithm. Further studies using ACT-FAST to bypass patients to a CSC are now underway.