Organizing Healthcare for Optimal Acute Ischemic Stroke Treatment.

Stroke is a major health-care problem that represents a leading cause of death and also the top cause of disability in adulthood. In recent years there has been a significant paradigm shift in treatments for acute ischemic stroke to favor earlier reperfusion therapy, mainly using the systemic infusion of recombinant tissue plasminogen activator. Subsequent trials found that combining this treatment with endovascular therapy was effective in selected patients. The increased complexity of acute stroke treatments has resulted in a substantial reorganization of stroke care. This review reports on the evolution of acute ischemic stroke treatment and describes the main organizational models based on the hub-and-spoke system. The lack of evidence for comparisons of the effectiveness of different paradigms means that some decision-analysis models predicting the best organizational pathways are also reported, with a particular emphasis on the workflow timing in the prehospital and in-hospital settings. Major benchmarks and performance measures are also reported, focusing on the timing of interventions and rates of process indicators. Finally, future directions are illustrated, including using telemedicine for stroke, mobile stroke units, and artificial intelligence and automated machines to produce software for detecting large-vessel occlusion.

INTRODUCTION

Stroke is a major health-care problem that represents a leading cause of death and the top cause of disability in adulthood. Acute cerebrovascular disease accounts up to an estimated 13 million events worldwide annually, which mostly comprise ischemic strokes, with more than 5 million fatalities.1 At the same time, about 18 million stroke survivors cannot perform their daily activities independently, and hence they present challenges to their families, societies, and health-care systems. In the US, 7 million people older than 20 years experience a stroke annually, and the overall prevalence of this disease is estimated to be 2.5%.2 In Europe, stroke is the second-most-common single cause of death,3 accounting up to 1.3 million events yearly.4 Projections show that the overall stroke burden will be 35% greater by 2050.5

EVOLUTION OF ACUTE STROKE TREATMENTS

Over the last 20 years there has been a significant paradigm shift in treatments for acute ischemic stroke treatment to favor earlier reperfusion therapy. The “time is brain” rule was assumed in the workflow management of these patients, and all reperfusion treatments were performed under this fundamental criterion.6 The first randomized controlled trials (RCTs) compared the systemic infusion of recombinant tissue plasminogen activator with placebo, and it was found to be effective at achieving good functional outcomes if administered within 3 hours from symptom onset, or subsequently when extended up to 4.5 hours.789 Each stroke center [now called a primary stroke center (PSC)] organized its activity and facilities based on this pharmacological treatment. Some RCTs from 2013 that compared systemic thrombolysis with endovascular treatment (ET) found no differences between the two therapies in their efficacy in obtaining a good functional outcome,101112 while other RCTs from 2015 and following years produced very different results.13141516171819 The key factors for the clinical success were represented by the presence of large-vessel occlusion (LVO) as an eligibility criterion and the use of latest-generation devices. Many meta-analyses have confirmed the pooled results of these trials, and many international guidelines have included the combination of systemic thrombolysis and mechanical thrombectomy as the gold-standard treatment for patients with LVO within 6 hours from symptom onset. In a patient-level meta-analysis using data from five RCTs, the best odds ratio (OR) obtained was 2.49 [95% confidence interval (CI)=1.76–3.53],20 with other study-level meta-analyses finding ORs between 1.56 and 1.71.212223 An overall pooled analysis of previously published meta-analyses comparing the bridging therapy with medical treatment alone found an OR of 1.82 (95% CI=1.53–2.16), as presented in Fig. 1.

Fig. 1

Forest plot of the pooled analysis considering the previous main published meta-analyses comparing bridging therapy with medical treatment alone. CI: confidence interval, ET: endovascular treatmentm, IVT: intravenous thrombolysis, M-H: Mantel-Haenszel method.

The previous paradigm expressed by the “time is brain” rule was changed to “each brain has each time” in 2018 by the publication of two RCTs.2425 These studies increased the therapeutic window for ET in patients selected using neuroimaging at up to 24 hours from symptom onset. A recent meta-analysis confirmed positive intervention outcomes also beyond 6 hours from stroke onset.26 However, there are no significant results available for comparisons between bridging therapy and direct ET, even if combining the two treatments seems to be slightly superior than a single-therapy approach.272829 In particular, these meta-analyses produced ORs between 1.18 and 1.21, and the overall pooled analysis of these studies resulted in an OR of 1.20 (95% CI=1.01–1.47) in favor of the bridging therapy. There are currently some ongoing RCTs attempting to clarify this issue.

ORGANIZATIONAL MODELS FOR ACUTE STROKE TREATMENT

When organizing systems for managing acute ischemic stroke, two main phases have to be considered: the prehospital and in-hospital settings. The first phase requires a prompt activation of emergency medical services to perform an initial clinical evaluation of the patient. A first assessment of suspected stroke is recommended by the main international guidelines. According to the evolution of acute stroke treatment and the indication for endovascular therapy, in recent years several prehospital stroke scales for detecting LVO have been investigated. However, a previous meta-analysis found wide variations in the accuracies of these scales.30 In light of these findings and based on the current guidelines, the use of a prehospital scale for LVO is not recommended when selecting the hospital destination for stroke patients. However, in the case of positive stroke scale (e.g.,the Cincinnati prehospital scale), a prenotification to the transferring hospital is also useful for reducing in-hospital delays.

The evolution of ischemic stroke treatment and the associated increasing procedure complexity determined that the systems of stroke care have been organized to transport patients in the nearest hospital with a initial possibility to perform reperfusion therapy (i.e., a PSC). These centers are organized to manage stroke patients by diagnosing LVO and treating it with intravenous thrombolysis (IVT). For this reason PSCs require the availability of vascular neurologists, brain CT and CT angiography, neurosonology, early rehabilitation, and beds with continuous monitoring.31 Further evidence increased the possibility of treatment with ET, but fewer hospitals are able to perform these procedures [called comprehensive stroke centers (CSCs)]. CSCs require facilities that guarantee the availability of more-complex procedures such as acute ET and neurosurgical interventions. For these reasons, neurosurgeons, neurointerventionalists, or interventional radiologists are needed in these centers. Different national laws and declarations as well as recommendations by scientific societies have defined specific hospital characteristics for classifying PSCs and CSCs by considering structural settings and the presence of certain departments and technological equipments (Table 1).31 In each stroke center the workflow has to be as fast as possible, including the use of multidisciplinary equipment (Fig. 2).

Table 1

Requirements and standards of care for PSCs and CSCs

	Standards of care
PSCs	Multiprofessional competences in hospitals
	Dedicated neurologists and nurses
	One bed with continuous monitoring (at least)
	Early rehabilitation
	IVT
	Neurosurgical on-call availability (also in other hospitals)
	Neurosonology with Doppler ultrasonography and echocardiography
	24/7 availability of CT and CT angiography (at least 16 multislices) and/or MRI (also with DWI) and MR angiography
	Linking with CSCs and rehabilitation units
CSCs	The same as PSCs and 500 admissions/year at least
	24/7 neuroradiology with CT (64 multislices) and CT angiography, MRI with DWI and PWI, and MR angiography
	Endovascular interventional unit
	24/7 neurosurgery
	24/7 vascular surgery
	Cerebral angiography
	Intra-arterial thrombolysis (emergency), mechanical thrombectomy (emergency), and extracranial and intracranial stenting
	Decompressive hemicraniectomy

CSC: comprehensive stroke center, DWI: diffusion-weighted imaging, IVT: intravenous thrombolysis, MR: magnetic resonance, PSC: primary stroke center, PWI: perfusion-weighted imaging.

Fig. 2

Examples of in-hospital workflows in PSCs (A) and CSCs (B). CSC: comprehensive stroke center, ER: emergency department, PSC: primary stroke center, Revasc: revascularization.

In light of the previous evidence concerning treatments for acute ischemic stroke, three main organizational models have been introduced: mothership (MS), drip and ship (DS), and drip and drive (DD; also called trip and treat). In the MS paradigm, a patient with suspected stroke is directly transported to the CSC where acute management ensures all diagnostic and therapeutic phases are applied, starting from initial neurological evaluation and ending with mechanical thrombectomy in an angiosuite. All procedures are performed in the same hospital. In the DS model, the patient with suspected stroke is transported to the nearest PSC (called a spoke center) to diagnose the disease and detect its etiology. IVT has to be administered if this is indicated, and if LVO is detected the patient is transported to the CSC of the territorial network (called the hub center) by primary transportation. In this organizational model, we hypothesized that if the place of the rescue and the PSC are separated by more than 60 minutes for transportation by ambulance, helicopter use is indicated and the transport should be centralized to the referring CSC of that territory.31 Finally, the DD paradigm consists of a mobile interventional stroke team comprising interventionalists and radiological technologists that travels from a central hospital (the CSC) to treat stroke patients presenting at peripheral hospitals with interventional capacity.

The DD paradigm is the least common worldwide, while the DS organizational model is the most common. The popularity of the DS paradigm is due to several factors including the rationalization of resources, the distribution of few CSCs and more PSCs, the reduced number of neurointerventionalists required, and the availability of an efficient emergency transportation system (Table 2). Although it remains unclear which type of organizational model is superior in achieving good functional outcomes, the American Heart Association (AHA)/American Stroke Association (ASA) has recommended the MS paradigm if bypassing a PSC for a CSC increases the travel time by less than 15 minutes.32 A recent meta-analysis of observational studies showed that the 90-dayoutcome is better in patients who are directly admitted to CSCs (adjusted relative risk=0.87, 95% CI=0.77–0.98).33 However, a multicenter registry found an association between the DS paradigm and poor functional outcomes.34 These contradictory results indicate that RCTs are needed to provide robust findings for determining the best organizational paradigm, also while applying different instruments for a prehospital triage, such as a stroke scale predicting LVO or a mobile stroke ambulance. Some trials with these aims are ongoing or in a start-up phase (i.e., direct transfer to an Endovascular Center compared to transfer to the closest Stroke Center in acute stroke patients with suspected large vessel occlusion in Catalonia - RACECAT trial, Spain, and mothership versus drip and ship for endovascular treatment in large-vessel occlusion strokes - MODEL trial, in Northern Italy). A critical point in choosing the best organizational paradigm is the financial cost. The optimal economic trade-off might be achieved by distributing the high-level centers according to the travelling distances between different CSCs and PSCs and to the population density in a defined territory or geographical area. The territorial organization and the economic investment for CSCs could start by determining the “isochrones” of the hub centers. PSCs should then be placed along these temporal lines that have a travelling cut-off time of 60 minutes. An example of this geographical distribution is presented in Fig. 3.

Table 2

Advantages and disadvantages of the three organizational models

Model	Advantages	Disadvantages
Mothership	All procedures performed in the same center	Potentially longer onset-to-needle time
	No secondary transfers	Smaller number of CSCs
	Shorter door-to-groin time in LVO	Higher workload
	High-volume centers
Drip and ship	Shorter onset-to-needle time	Longer door-to-groin time in LVO
	Rationalization of the geographical distribution of the PSC and CSC	Cost of medical transfers
	Rapid initial clinical assessment
Drip and drive	No secondary transfers for patients	Long door-to-groin time
	Shorter onset-to-needle time	Different angiosuite settings for neurointerventionalists

CSC: comprehensive stroke center, LVO: large-vessel occlusion, PSC: primary stroke center.

Fig. 3

Distribution of hub centers and spoke centers using “isochrones.” Black lines represent the distance from hub centers for a maximum traveling time of 60 minutes as calculated using a Google Maps application programming interface. □ Comprehensive stroke center, ○primary stroke center.

DECISION-ANALYSIS MODELS TO PREDICT THE BEST ORGANIZATIONAL PATHWAYS

The current uncertainty concerning the best paradigm for the organizational pathway stimulated the development of different decision-analysis models based on probability theory. Different steps for acute stroke management were considered in each model, with the final endpoint represented by the best clinical outcome. Two main studies provided statistical models to identify the organizational paradigm with the highest probability of achieving good functional outcomes. In the study of Milne et al.,35 the DS model showed effectiveness for any PSCs that are remote from CSCs, regardless of the door-to-needle time (DNT). In the case of the PSC and CSC being in close proximity, the DS paradigm remained effective only if the DNT for the PSC was shorter than 30 minutes. On the other hand, the MS scenario was more effective only if bridging therapy was effectively applied in the CSC. In the model of Holodinsky et al.,36 the DS paradigm was effective if the patients were close to a PSC and this center was more than 45 minutes from a CSC, regardless of the DNT. The same model was effective also when the distance between the PSC and CSC was 10–45 minutes with a DNT of shorter than 30 minutes. Finally, if the DNT was longer than 60 minutes, the MS model was effective if the distance between the PSC and CSC was shorter than 30 minutes.

The results of a previous geographical modelling analysis to quantify the impact of PSC and CSC destination policies showed low time cost values—adding median durations of only 3.1 and 8.3 minutes—when bypassing the closest PSC and CSC, respectively.37 A recent decision-analysis model introduced LVO screening methods and key time metrics,38 and found that the DS paradigm was superior when LVO screening was not performed, while the MS paradigm was superior when LVO positivity was found in formal screening and the additional transport time to a CSC was shorter than 23 minutes. Transferring these results in the real-world setting with a PSC workflow slower than the ideal settings indicates that the MS model is superior to the DS model even when formal LVO screening is available.

In conclusion, different decision-analysis models have yielded results favouring the DS paradigm only if the workflow times in PSCs are optimized with a DNT of shorter than 30 minutes. Fig. 4 summarizes the effectiveness of the two main probabilistic models (MS and DS).

Fig. 4

Probabilistic models of the effectiveness of mothership (A) and drip and ship (B). t1, onset-to-needle time; t2, door-to-needle time; t3, indoor-to-outdoor time; t4, transfer time; t5, door-to-groin time; tx+t2: door-to-groin time. □ Comprehensive stroke center, ○primary stroke center. (A) is better than (B) when t2(B)>30 minutes or t2(B)>60 minutes and t4<30 minutes. (B) is better than (A) when t4 is high or t4>45 minutes and t1 is very low or t4>10 and <45 minutes and t2<30 minutes.

MAJOR BENCHMARKS AND PERFORMANCE MEASURES FOR ACUTE STROKE TREATMENT

A frequent situation involves a goal being identified by a panel of experts and set as a benchmark as an objective to strive for. The key aspect of benchmarking is that it forms part of a comprehensive and participative policy of continuous quality improvement. Benchmarking should be based on voluntary and active collaboration among several organizations to create a spirit of competition, to apply the best practices, and to achieve the goal of reaching the defined benchmark. According to this definition, the organizational model for acute stroke treatment is based on a network in which different hospitals that collaborate so as to create an uninterrupted pathway for managing these patients.

The following major time benchmarks for acute ischemic stroke treatment are all tracked by international standard of recommendations, and they were chosen by a group of experts at the National Institutes of Health; they have subsequently been adopted by the ASA:39

• Door to medical doctor: 10 minutes.

• Door to neurologist/strokologist: 15 minutes.

• Door to imaging: 25 minutes.

• Door to image reading: 45 minutes.

• Door to needle: 60 minutes.

• Door to groin: 120 minutes.

• Groin to start of revascularization: 45 minutes.

• Clinical outcomes for ET of at least 30%, with a score on the modified Rankin Scale of 0–2 at 90 days.

The stroke Joint Commission provided some core stroke measures related to the overall care of stroke patients, such as venous thromboembolism prophylaxis, discharged on antithrombotic therapy, anticoagulation therapy for atrial fibrillation, and thrombolytic therapy.40 However, this commis sion also elucidated some performance measures for PSCs and CSCs that could have benchmarks:

1) Rate of patients for whom the score on the National Institutes of Health Stroke Scale (NIHSS) is recorded on arrival.

2) Rate of eligible patients treated with IVT within 60 minutes of arrival.

3) Rate of patients who arrive by 3.5 hours and are treated by 4.5 hours.

4) Rate of patients with any score on the modified Rankin Scale at 90 days.

5) Rate of patients with stroke severity measurement on arrival using the NIHSS.

6) Rate of patients with a symptomatic intracranial hemorrhage treated with both IVT and ET within 36 hours.

7) Median time to recanalization therapy for ET patients.

8) Rate of patients treated with ET exhibiting at least thrombolysis in cerebral infarction (TICI) 2b recanalization.

FUTURE DIRECTIONS

The increasing burden of ischemic stroke will increase the number of patients eligible for ET, and the health-care organizational model has to consider this changing epidemiology. Rapid technological developments will make it possible to also apply artificial intelligence (AI) and automated machines to produce software to promptly detect LVO and brain mismatch areas.41 Early in 2018 the FDA approved an AI algorithm for use in a clinical decision support system in a triage setting. This algorithm, called Viz. AI, is able to analyze CT and CT angiography scans, and can automatically identify suspected LVO strokes and promptly alert the on-call stroke team, including that in other hub centers. Another application is the development of software that automatically calculates mismatch areas, also when no radiologists or neuroradiologists expert in CT perfusion are present in PSCs. Finally, some software and apps provide platforms with text messaging and calling for facilitating communication between clinical teams in hub-and-spoke networks. However, the ability of these AI systems to provide useful information is crucially dependent on the correctness of input data and the method used to instruct the model.

In addition to these promising applications, the use of mobile stroke units and CT ambulances have shown positive results in previous studies, by shortening time metrics and increasing the number of patients eligible for acute ischemic stroke treatment and concomitant transportation to the mostappropriate center for receiving the acute treatment.42 The use of helicopters mitigates larger distances between PSCs and CSCs. The application of telemedicine for stroke provides the possibility of treating patients with IVT by a rapid strokologist evaluation for intravenous tissue plasminogen activator eligibility in centers without expertise and all of the required facilities. This application could also facilitate the sharing of neuroimaging facilities that would improve the ability to detect LVO or the volume of the damaged ischemic area, using also apps derived from AI as described above. Previous AHA/ASA guidelines have recommended the use of telemedicine for stroke, identifying also the quality of application and measurable outcomes.43 The future use of telemedicine for stroke to cover extensive geographical areas will increase the application of IVT when vascular neurologists or strokologists are not present, as well as the use of software or apps based on AI. The use of these technologies will reduce temporal delays in acute stroke management, as well as the effective distance between centers with different facilities.

Regarding the organizational model, the MS and DS paradigms could also coexist in the same geographical area. A mixed organizational model considering urban and rural areas could be implemented based on which performs better (e.g., MS in urban areas and DS in rural areas), based on information about traffic levels that potentially slow the transportation of patients. However, the feasibility of such a mixed organizational network system needs to be validated in a future RCT.

CONCLUSIONS

Bridging therapy in acute ischemic stroke involving the administration of IVT followed by mechanical thrombectomy has become the current standard of care for patients with LVO.44,45 However, the best organizational paradigm for this acute treatment remains to be defined, and the current international guidelines do not recommend bypassing PSCs since this might result in lower-accuracy screening tools being used to detect LVO.46 In light of these observations, the DS model remains today the paradigm that shows the best ratio of clinical effectiveness to financial cost. Further studies are required to better understand the best health-care organizational model for ischemic stroke patients.