Comprehensive Telestroke Network to Optimize Health Care Delivery for Cerebrovascular Diseases: Algorithm Development

Introduction

Stroke is a major source of disability and death in both developed and developing countries [1-4]. The adequate delivery of care for patients with acute stroke symptoms requires the expertise of neurologists and radiologists for timely diagnosis and treatment. Whereas hemorrhagic stroke often requires urgent surgical intervention, ischemic stroke is managed with reperfusion therapies such as thrombolysis with intravenous recombinant tissue plasminogen activator (IV rtPA) as well as early endovascular thrombectomy in the case of large vessel occlusions. These approaches have significantly improved the long-term outcomes of patients with ischemic stroke [5]. Nevertheless, patients frequently do not receive the appropriate treatment either due to the lack of available specialists to perform appropriate clinical assessments or the long distances and prolonged transfer times to stroke care centers [5]. This situation can also occur for patients located in urban areas due to delays in the referral process to health care facilities with the required stroke handling capabilities.

We have developed several strategies to improve the quality of care and speed up the transfer of patients with acute stroke symptoms from urban and rural areas to our hospital, a certified primary stroke center with thrombectomy capabilities. Our experience has shown that, by necessity, robust stroke systems should be able to assist health care providers in real-time scenarios, thus resulting in adequate transfer processes between any level of complexity in a specific health care setting.

The purpose of this work was to establish a protocol for the management of patients with acute stroke symptoms according to the available treatment alternatives at the initial point of care. This protocol was implemented as a web-based telestroke solution and is based on the guidelines from the American Heart Association and American Stroke Association (AHA/ASA) [6] for any patient with acute stroke symptoms (ie, hemorrhagic stroke, acute or chronic ischemic stroke, transient ischemic attack [TIA], stroke mimics, and large vessel occlusions). The authors considered that a detailed description of the workflow protocols, pathways, and clinical and radiological scales used in the design of our telestroke network will help other maturing countries in the development of early-stage health systems assisting patients with acute stroke symptoms.

Methods

The work presented here is part of a larger project with the following objectives: evaluation of mobile systems for head computed tomography (CT) interpretation in acute stroke patients, evaluation of the quality of stroke care in our country from a public health standpoint, and development of a telestroke network system (the subject of this article). This initiative was approved by the Institutional Review Board of our hospital and university.

The latest diagnostic and therapeutic recommendation guidelines from the American Heart Association [6], in addition to several predictors, scales, and scores that allow the assessment of the vascular insult severity and neurological status of the patient (Table 1), were evaluated in order to define the protocols of this system.

Stroke care requires human resources (eg, neurologists, radiologists, or neuroradiologists) to evaluate the risk and eligibility of patients to receive reperfusion therapies (intravenous thrombolysis or endovascular thrombectomy) and to perform invasive treatments when indicated. In addition, technological resources, such as CT, computed tomography angiography (CTA), magnetic resonance imaging (MRI), magnetic resonance angiography (MRA), and the necessary medical supplies and equipment, are needed to administer reperfusion therapies. Different human and technological resources may be available according to a patient’s point of entry into the health care system. Therefore, several possible health care settings were evaluated to define the protocols and algorithms of this system. The algorithms for the clinical workflow of the three health care settings were defined and reviewed by a group of experts in our hospital: a stroke neurologist, a general neurologist, a neuroradiologist, and two physicians from our stroke center.

Results

The following health care settings were identified: primary health care setting, intermediate health care setting, and advanced health care setting. The interaction between these three settings is shown in Figure 1, which also shows the optimal health care settings according to a specific patient diagnosis (eg, intensive care unit [ICU], recovery room, facility with neurology or neurosurgery capabilities, regular hospitalization, or ambulatory care). IV rtPA administration can be provided in the intermediate health care setting or advanced health care setting, while thrombectomy is performed only in the advanced health care setting. In both cases, a judicious risk assessment is needed before the administration of any treatment modality [6].

In all settings, the first step is the acquisition of demographic data followed by a clinical background update, blood glucose registry, physical exam, assessment of the level of consciousness using the Glasgow Coma Scale [7], and assessment of the clinical severity of the ischemic stroke using the National Institutes of Health Stroke Scale (NIHSS) [9]. The next step is to determine the onset time to calculate the therapeutic window time (ie, the time between neurological symptom onset and patient arrival to the emergency room). In some cases, the event may be classified as wake-up or unwitnessed stroke. The therapeutic window time may be “within window” (<6 hours) or “out of window” with two possible ranges (6-24 hours or >24 hours) that determine the differences in patient management. The combination of the NIHSS and Glasgow scores as well as the time from onset to care delivery are critical breakpoints for the decision-making process performed at each of the three health care settings.

‎View this figure

Health Care Settings

The general diagnostic and treatments steps, within each health care setting, are presented in a simplified workflow shown in Figure 2.

‎View this figure

Primary Health Care Setting

In this setting, diagnostic tools are limited to the physical exam performed by a primary care physician as well as basic blood tests (eg, blood glucose). This setting may include ambulances, which may be a possible entry point to the health care system. The purpose of this setting is to provide an initial clinical assessment and determine the patient’s transfers to a health care center with reperfusion capabilities, according to possible TIA, large vessel occlusion, or ischemic stroke in the anterior or posterior circulation. In this scenario, there are neither imaging facilities (CT or MRI) nor specialized health care personnel. Therefore, different clinical scales are used to assess the patient’s risk at multiple levels and to predict final patient outcomes. For example, patients with possible compromised posterior circulation will be transferred to an advanced health care setting, while those with anterior circulation may be transferred to any intermediate or advanced health care site; this estimation is achieved using ischemic stroke circulation predictors [13,26]. The ABCD2 score [14] is a powerful tool that predicts the subsequent risk of stroke after a TIA, and the Field Assessment Stroke Triage for Emergency Destination score is used to determine the probability of a large-vessel occlusion [16]. The detailed workflow for this setting is shown in Figure 3.

‎View this figure

Intermediate Health Care Setting

At this level of health care facility, head CT must be available to detect hemorrhagic or ischemic stroke; intravenous thrombolysis capabilities are also required at this level.

To diagnose a potential large vessel occlusion and determine if further transfer to the advanced health care setting is necessary, CTA must be available (or contrast head CT if CTA is not available). At the intermediate health care setting, both neurologists and radiologists may be available, but not fulltime. This setting works as a mothership for urgent and priority transfers from the primary health care setting. Possible outcomes include priority transfer to an advanced health care setting for thrombectomy purposes, emergent assessment by the neurology or neurosurgery teams, IV r-TPA administration, or ambulatory care. In this setting, if a patient is eligible for IV r-TPA administration, this is done in situ, to enable early treatment, even if the patient will be transferred to the advanced health care setting. The detailed workflow for this setting is shown in Figure 4.

‎View this figure

Advanced Health Care Setting

In this health care setting, specialized human and technological resources, such as stroke neurologists, neuroradiologists, CT, CTA, MRI, MRA, and the capacity for thrombolysis and mechanical thrombectomy, are available fulltime. Therefore, interfacility transfers are not necessary. In this setting, CT perfusion images are required for patients with wake-up stroke for which MRI is contraindicated. This setting receives transfers from primary health care settings and intermediate health care settings. The possible outcomes are shown in Figure 5. The workflow in the advanced health care setting for patients within the window for thrombolysis (ie, <6 hours) is shown in Figure 6. To simplify the workflow figures and render each figure on a single page, several common procedures in the intermediate and advanced health care settings were arranged in modules presented in Figures 7-10 (ie, hemorrhagic module, TIA or mimic module, intravenous thrombolysis module, endovascular treatment module, and ischemic stroke out of window module).

‎View this figure

‎View this figure

Common Modules

Hemorrhagic Module

If the imaging examination shows a hemorrhagic stroke, the Fisher scale [20] and modified World Federation of Neurosurgical Societies scale [21] are used to evaluate the severity of the subarachnoid hemorrhage. To predict mortality in patients with intracerebral hemorrhage, the intracerebral hemorrhage score [17-19] is used. CT or CTA may reveal active intracranial bleeding; in this case, the patient is referred to the neurosurgery team for urgent care. Otherwise, the patient can be stabilized and treated in the ICU (Figure 7).

‎View this figure

TIA or Mimic Module

If the symptoms are gone or are not consistent with a vascular territory and imaging examination reveals neither a hemorrhagic stroke nor an ischemic stroke, there are two possibilities: the patient is presenting with a stroke mimic or having a TIA. In the later, the ABCD2 score [14] is calculated to evaluate the actual stroke risk. Secondary prevention using statins and antiaggregant therapy is initiated, and the patient is discharged for neurologic outpatient care (Figure 8).

‎View this figure

Ischemic Module

This module consists of two submodules: intravenous thrombolysis module for IV rtPA administration and endovascular treatment module, to evaluate thrombectomy treatment (Figure 9).

‎View this figure

Intravenous Thrombolysis Module

If the initial imaging examination reveals an ischemic stroke, an imaging vascular evaluation of the anterior and posterior circulation is performed according to the onset time of the ischemic insult (acute, subacute, or chronic). Patients with subacute or chronic lesions are referred to the recovery room if hemodynamically stable; otherwise, they are hospitalized for neurology assessment. For patients with acute lesions in the middle cerebral artery territory, the Alberta Stroke Program Early CT Scan (ASPECTS) is calculated; patients with ASPECTS <6 are referred to the ICU, whereas patients with ASPECTS ≥6 are evaluated in terms of the absolute and relative contraindications for IV rtPA administration (Table 2) [24]. Next, if there is no risk or only a relative risk after judicious medical assessment, intravenous thrombolysis is performed. At the same time, the patient is evaluated for the presence of large vessel occlusions and possible thrombectomy, as indicated in the endovascular treatment module (Figure 9).

Endovascular Treatment Module

Large vessel occlusions are evaluated using contrast CT or CTA. If there are no occlusions, the patient is referred to the recovery room if hemodynamically stable; otherwise, they are hospitalized for continuous neurologic assessment. If a large vessel occlusion is confirmed, a comprehensive risk evaluation should be performed before thrombectomy [24]. Then, if no risks are identified, thrombectomy should be performed as soon as possible. After thrombectomy is performed, the degree of reperfusion is measured by means of the Thrombolysis in Cerebral Infarction score [25], and the patient is referred to the ICU (Figure 9).

Ischemic Stroke Out of Window Module

Patients arriving to an advanced health care setting after a wake-up stroke, unwitnessed stroke, or “out of window” stroke with symptom onset 6-24 hours before first medical contact may benefit from reperfusion therapies only if certain conditions are met (Figure 10). These conditions rely on the infarct volume as quantified by diffusion-weighted MRI or CT perfusion (if MRI is contraindicated), patient age, and stroke severity (NIHSS score).

‎View this figure

Design of Algorithms

The three algorithms work as a handy framework for the most critical steps in the care of patients with ischemic or hemorrhagic stroke. These algorithms were based on decision trees that represent the clinical requirements and specifications of the system and consist of checklists and questionnaires [27] evaluating common physiological variables, the patient’s clinical background, different predictors and scales, and specific laboratory tests according to each algorithm stage. The responses to these questionnaires determine the next step to be performed.

The predictors, scales, and scores implemented in this system allow the assessment of the vascular insult severity and neurological status of the patient. Other factors that determine patient management, either in situ or in a distant health care setting, are shown in Table 1. The clinical background, physical exam, stroke severity, and radiological findings were stored in basic modules within the clinical algorithms. These modules, as implemented in our telestroke system, are shown in Table 3. The software outputs corresponding to specific diagnoses, clinical scenarios, and transfer decisions are shown in Table 4. The checklists for risk assessment before the administration of reperfusion therapies [24] are detailed in Table 2.

Software Development and Validation

The algorithms included in the three health care settings were incorporated into web-based software. Individual user profiles were created for the administrative staff and health care providers, who were assigned specific privileges.

A user-friendly interface reduces human error and assures the completeness and integrity of the information. The questionnaires implemented were straightforward and only required single-click selections instead of free-text typing for easy and rapid data input. The software was developed using the Hypertext Preprocessor and JavaScript languages and could be executed in any web browser on a laptop, tablet, or smartphone.

For data storage, a MySQL 5.1.40 database (Oracle Corporation, Redwood City, CA) was used, wherein sensitive data were encrypted (ie, patient identification). Data were stored in a structured relational database, allowing future evaluation of the system performance as well as a strong foundation for public health policies. The database included the administrative information of each facility in the telestroke network and the possible referral facilities (ie, those that have a given facility that was contracted to receive patients when a transfer is required). This information allowed the rapid selection of the most suitable stroke center according to the patient’s needs after a judicious assessment of the clinical requirements and transfer times. Given that “time is brain,” potential administrative pitfalls between primary health care settings and advanced health care settings also had to be considered for a quick and effective transfer. In our country, patients may be transported to various emergency departments until they are accepted in one of them, producing a critical delay in the required care known as “the death ride.”

Since a patient can arrive at any given hospital and may be transferred across several health care settings without receiving adequate stroke care, a “case” starting point was defined as the time when the first medical contact was documented in the last visited hospital until the final patient outcome was reported before the patient’s discharge. Therefore, when a patient is transferred to a second health care setting, all the information for the case is available in the receiving facility given that all data are stored in a server database and shared with all the facilities. This design decision allows common access to the patient’s health condition at any moment from any health care setting while also avoiding the entry of redundant information. Hence, past medical history, current clinical condition, blood test results, imaging evaluations, and procedures are available in real-time for all health care facilities across the whole spectrum of stroke patient care. In addition, this allows transfer reporting to the referral facilities ahead of the patient’s arrival, avoiding prolonged waiting times at emergency departments.

Software validation was performed in different phases. The first phase consisted of a simulation of the test scripts on all possible workflow pathways for each of the health care settings, which was performed to validate adequate software representation of each the algorithms. To validate the software implementation, more than 1211 test scripts were performed covering all possible pathways in each of the predefined health care settings. The second phase consisted of a retrospective registry of cases from our stroke database (nearly 600 patients in the last 5 years). The third phase was the validation of the software by neurology residents, who utilized the software while also performing a usual clinical assessment with printed forms. The final phase is to be performed between different health care facilities to test the performance of our telestroke network with real-life cases and transfers based on the information broadcast. The total number of possible pathways documented after this initial experience was as follows: 10 in the primary health care setting, 39 in the intermediate health care setting, and 1162 in the advanced health care setting.

The final system was named Telestroke-RU (copyright 13-70-240, 03/12/2018 from the National Copyright Office, Colombia) and is available for authorized users [28]. This system is not a product intended for commercial or profit uses and may be used for educational purposes.

Discussion

Principal Findings

The comprehensive telestroke platform developed in this work is the first stage to optimizing health care delivery for patients with stroke symptoms regardless of the entry point into our local emergency network in both urban and rural regions.

This system supports health care personnel by providing adequate stroke care and facilitating the prompt transfer of patients to a more appropriate health care setting according to the specific cerebrovascular disease at presentation. This system facilitates stroke care delivery for patients with acute ischemic stroke within the therapeutic window who are candidates for reperfusion therapies. Therefore, the system will contribute to mitigating the well-known mortality and morbidity associated with stroke.

Further evaluations will be performed to assess the true impact of this tool in terms of reductions in critical time windows, such as the time between symptom onset and reperfusion, door to needle time, primary health care setting to advanced health care setting transfer times, discharge clinical outcomes, accuracy of the final diagnosis, and the clinical outcomes of patients at 30 and 90 days using the modified Rankin scale [29].

Comparison With Prior Work

To the best of our knowledge, in our country, there are no software tools for the assessment and management of patients with stroke symptoms. Worldwide, smartphone apps and web-based tools are available [30-34]. These solutions were designed for acute ischemic stroke care, for triage protocols, and as an aid for transfers when reperfusion therapies are needed (eg, Field Assessment Stroke Triage for Emergency Destination score) [30] as well as the delivery of efficient inpatient or ambulatory stroke care [32,34]. Other tools have been developed with the purpose of the evaluation of specific clinical scales or radiological scores [33]; these tools use the same algorithm independent of the level of resources available at the entry point to the health care system. In contrast, our system integrates 12 clinical and radiological scales, scores, and predictors according to the specific health care setting or the referral facility to provide a specific diagnosis (hemorrhagic stroke; acute, subacute, or chronic ischemic stroke; TIA; stroke mimic). Our system allows either immediate treatment or further transfer to the appropriate health care setting. Since our solution is not integrated with the hospital electronic health record (EHR) system, it can be used in all health care services, independent of the EHR system used at each facility. It is worth mentioning that, in our country, EHR systems are not available at several rural primary health care settings.

Limitations

The continuous improvement of evidence-based stroke care guidelines motivates the continuous review of health care setting algorithms and, therefore, software updates. Further work includes the use of GPS and traffic applications to calculate the actual duration of real-time patient transfers and adequate selection of the referral facility with the shorter transfer time. In the short-term, this system will be migrated to a smartphone app to allow for a greater number of system users in a friendlier interface.

Conclusions

The implementation of this system in a telestroke network contributes to the fulfillment of and adherence to recently published stroke care guidelines, providing evidence-based practice, improving patient outcomes, and supporting the achievement of several requirements to achieve and maintain primary stroke center certification.

This telestroke system allows the assessment of different therapeutic alternatives according to the specific patient’s clinical condition, thus improving efficiency and providing high-quality delivery of care. Finally, the epidemiological information stored in the database will inform public health care policies to design and implement better national policies for remote regions with significant underreporting of acute cerebrovascular diseases.