Identification of barriers and enablers to rapid diagnosis along the paediatric stroke chain of recovery using Value-Focused Process Engineering

ABSTRACT

Coordinated systems of care are required to improve access to reperfusion therapies in paediatric stroke. A conceptual model was developed to map the process-of-care from symptom onset to confirmation of diagnosis. Value-Focused Process Engineering with event-driven process modelling was used to identify barriers and enablers to timely and accurate paediatric stroke diagnosis. Stakeholder interviews were conducted to inform model design, development, demonstration and validation. Barriers included: (i) ambulance dispatcher failure to allocate high-priority response, (ii) childrens’ exclusion from paramedic clinical practice guidelines, (ii) non-allocation of high triage category on hospital arrival, (iii) absence of emergency department guidelines for focal neurological deficits, and (iv) computed tomography as the first imaging investigation.

Enablers included: (i) public awareness programs, (ii) childrens’ inclusion in prehospital emergency stroke algorithms, (iii) re-organisation of health services, with primary paediatric stroke centres, (iv) implementation of triage and neuroimaging decision support tools, and (iv) rapid stroke MRI imaging protocols.

1. Introduction

Stroke is one of the most time-critical neurological disorders, with evidence in adults supporting use of reperfusion therapies to improve functional outcome (Jauch et al., 2013; Powers et al., 2015). Identification and acute management of stroke is a complex process, involving multiple interacting factors which include actions of the community and health professionals, and utilisation of health-care resources and technology, along a time continuum from symptom onset to diagnosis. These factors, in combination with the need to deliver treatment within predefined time windows, have driven the development of a systems approach to adult stroke care. The Stroke Chain of Recovery is a framework developed by the American Heart Association to improve the efficiency of acute stroke care in the pre-hospital and emergency department settings (Jauch et al., 2013). The benefits of implementing multipoint interventions along the Chain of Recovery at institutional and regional levels, have been demonstrated for reducing time to diagnosis and increasing access to reperfusion therapies (Lindsberg et al., 2006; Meretoja et al., 2012, 2013).

In contrast to adults, stroke is an uncommon cause of acute neurological symptoms in children (Mackay et al., 2014), is not often considered by parents (Mackay et al., 2016) or paediatric physicians (Srinivasan, Miller, Phan, & Mackay, 2009), and ED algorithms which have been recently described (Bernard et al., 2016; Ladner et al., 2015) are not widely implemented in practice. Coordinated systems of pre-hospital and emergency department care are urgently required for children with stroke-like symptoms to improve access to time-critical reperfusion therapies (Jauch et al., 2013; Powers et al., 2015). The relatively infrequent occurrence of childhood stroke makes it difficult to use a conventional data-driven approach to understand reasons for delayed diagnosis, and to improve delivery of care. Alternative approaches are required to identify current barriers to stroke diagnosis in children, and to identify potential enablers to improving access to reperfusion therapies.

Thus, the health system problem that provides research motivation for this study was the lack of understanding of the unacceptably long delays to diagnosis and treatment of childhood stroke, due to a paucity of published data and limited knowledge of the current process of care, while the objective for solution of the identified problem was to improve rapid diagnosis of paediatric stroke. The objective of the study was, therefore, to investigate whether and how Value-Focused Process Engineering (VFPE) (Neiger, Churilov, & Flitman, 2009) can be used to understand and improve the effectiveness and efficiency of the process of acute paediatric stroke care, from symptom onset to radiological confirmation of diagnosis.

This objective was achieved through designing, developing, and validating a conceptual model of the paediatric stroke chain of recovery using VFPE (Sargent, 2013). The broad Design Science research paradigm proposed by Pfeffers and colleagues was adopted to guide the modelling process (Hevner, March, Park, & Ram, 2004; Peffers, Tuunanen, Rothenberger, & Chatterjee, 2007/2008).

The original contribution of this research is two-fold:

1. For the first time, the barriers and enablers for improving rapidity and accuracy of paediatric stroke diagnosis along the Stroke Chain of Recovery are identified;

2. It applies VFPE in the new decision-making and system context of paediatric stroke.

The remainder of the paper is structured as follows: Section 2 provides an overview of Value-Focused Process Engineering methodology; in Section 3 the methods for development, design and validation of the conceptual model are presented; in Section 4 the modelling results are presented, with a description of the process of paediatric stroke care including barriers and enablers identified; Section 5 presents the discussion of the results, and Section 6 summarises the findings, in the context of best practice acute stroke care in adults, and future recommendations to improve accuracy and timeliness of stroke diagnosis in children.

2. Value-Focused Process Engineering: an overview

Value-Focused Process-Engineering (VFPE), first introduced by Neiger et al., was chosen to model the paediatric stroke chain of recovery in this study (Neiger et al., 2009). This methodology combines Value-Focused Thinking (VFT) objective modelling, to identify values, with Event-Driven Process Chains modelling (EPC), to provide a representation of the operation.

VFT is a business objectives modelling technique that guides decisions to improve effectiveness (“doing the right things”), by linking broader decision-makers’ values to individual decision objectives, thereby ensuring that fundamental values drive the entire decision-making process (Keeney, 1992). This technique differs from the usual approach of considering available options and then choosing the alternative that is most likely to result in the desired outcome. Instead, the focus of VFT is on fulfiling values rather than being limited to the alternatives. It provides a logical and explicitly articulated approach for identification and structuring of objectives.

Value-Focused Thinking starts by defining values, which are a qualitative statement of “what we care about” (Keeney, 1992). This is followed by the iterative identification and structuring of objectives, which are more concrete propositions of what the decision-makers are aiming for. The objectives are divided into fundamental (also known as “end”) objectives and means objectives. Fundamental objectives, which characterise the reason for interest in the decision situation, are important “just because”, whereas means objectives are of interest only “because of their implications for the degree to which another more fundamental objective can be achieved” (Keeney, 1992). Fundamental objectives are mutually exclusive and are organised as hierarchies. In contrast, means objectives are the means by which one can achieve the fundamental objectives. Means objective are organised into networks, and are typically described in terms of capabilities, resources and activities (e.g., utilise, ensure, establish and allocate) (Keeney, 1992).

Process modelling with Event-driven Process Chains (EPC) (Scheer, 2002) was originally developed for businesses to visually model and evaluate existing processes, in order to identify time- or cost-inefficient activities, and to improve efficiency (“doing things right”). The model is graphically represented as a chain of process workflow, with a sequence of events and activities (also referred to as functions), connected by logical rules (Neiger et al., 2009). Process chains, which also have applications to health care (Dugas & Dugas-Breit, 2012), can be used to demonstrate the most efficient process, thereby allowing a gap analysis between current and ideal processes to enable redesign of the process (Damelio, 1996). Events are passive elements within the chain, and refer to situations or preconditions that trigger activities, and to situations or conditions that result from activities. Activities (functions) are active elements which need to be executed as part of the business process. Events and functions are linked together into a sequence by logical connectors, which allow parallel branches, decisions, multiple triggers and complex flows (Davis, 2012). The basic rules of the process chain that events and functions are linked together into a chain, with alternating events and functions. The process always begins with at least one external initiating event and ends with a final outcome. Multiple events can be linked to a single function (or vice versa) via the logical rules “and”, “or” and “exclusive or”. Decomposition of the process chain is performed to simplify and clarify complex processes and enable modelling at different levels of detail. Components of the event-driven process chain are outlined in Table 1.

Table 1.

Components of the Event-Driven Process Chain

Object	Description	Representation
Events	Something done
Functions (activity)	Do something
Rules (connectors)	Or, Exclusive Or, And
Control flow links	Demonstrates process flow
Logical connector	Following a function (activity)	Preceding a function (activity)
OR	OR decision One or more possible paths can be followed	OR trigger Any one event or combination of events will trigger the activity
EXCLUSIVE OR	EXCLUSIVE OR decision One and only one possible path can be followed	EXCLUSIVE OR trigger One, but only one of the possible events will be triggered
AND	AND branch The process splits into two or more parallel paths which are independent of each other	AND trigger All events must occur to trigger the activity

Integration of the VFT and EPCs into the VFPE framework is achieved by explicitly linking values and objectives to process flow of individual activities and events using the following set of steps:

1. Identification of values and fundamental objectives using the Value-Focused Thinking framework;

2. Development of higher-level process maps, with key events and activities, to provide an overall representation of a complex process;

3. Linkage of higher-order means objectives to higher-level process maps;

4. Linkage of fundamental objectives to a network of means objectives;

5. Decomposition of the higher-level processes into more detailed event-driven process chains of individual events and activities, in order to clarify the process;

6. Direct linkage of at least one decomposed means objective to every individual activity in the EPC.

VFPE methodology was developed and previously used in the context of human resources management (Neiger et al., 2009) and supply chain risk identification (Neiger, Rotaru, & Churilov, 2009). In this paper, we demonstrate how it can support designing, developing, and validating a conceptual model of the paediatric stroke chain of recovery.

3. Application of VFPE for design, development and validation of the Paediatric Stroke Chain of Recovery model

The VFPE Paediatric Stroke Chain of Recovery model was designed and developed by a paediatric stroke specialist (MTM) and a management science expert experienced in systems modelling research (LC). Construction of the model was an iterative process that required several meetings and discussions with key clinicians involved in the acute system of paediatric stroke care. Interviews were conducted with an Emergency Medical Dispatch customer service manager and two senior paramedics in order to understand the process of pre-hospital emergency care. This required a visit to the Emergency Services Telecommunications Authority (ESTA) in Victoria, Australia, to observe the call-taker and dispatch processes first-hand. Interviews were conducted with two senior triage nurse educators, a paediatric emergency physician (FEB), a paediatric radiologist (AM), and a paediatric anaesthetist (IMcK) at the Royal Children’s Hospital, Melbourne, Australia, in order to understand the processes of emergency department, medical imaging and anaesthetic care. Figure 1 provides an overview of the VFPE modelling process, within the framework proposed by Peffers and colleagues (Peffers et al., 2007/2008). The Stroke Chain of Recovery was used as a starting point for development of the model because it is a widely accepted framework for acute stroke care in adults (Jauch et al., 2013) and it provides a high-level depiction of the process (Figure 2). The model was divided into pre-hospital and emergency department phases with sub-processes for each component of the Stroke Chain of Recovery. The pre-hospital sub-processes broadly corresponded with the Dispatch, Delivery and Door components of the Stroke Chain of Recovery. The emergency department model broadly corresponded with the Data (Clinical and imaging) components of the Stroke Chain of Recovery (Figure 3). Microsoft® Visio® Professional 2010 software was used to create the process maps.

Figure 1.

Overview of the modelling process

Figure 2.

Improving care along the Paediatric Chain of Recovery

Figure 3.

The Paediatric Stroke Chain of Recovery

Design and development of the conceptual model required regular meetings between the paediatric stroke specialist and the management science expert. Structured walk-throughs of each sub-component of the model were performed, in order to demonstrate that the model met the aims of understanding the process of care, and of improving effectiveness and efficiency along the Paediatric Stroke Chain of Recovery. The management science expert also ensured that the model was correct with respect to (i) all activities being linked to at least one means objective, thereby removing redundant activities which did not fulfil the values of maximising efficiency and effectiveness of the process, and (ii) the rules governing process flow were met.

In order to ensure face validity, each component of the model was evaluated by the clinician stakeholders with content-specific knowledge (listed on page 7) to ensure that each sub-process included all the necessary components. Barriers and enablers to achieving fundamental objectives were identified for each subcomponent of the Paediatric Stroke Chain of Recovery.

4. The VFPE conceptual model of the Paediatric Stroke Chain of Recovery

The fundamental objectives for the paediatric stroke VPFE model were defined as the accurate and timely identification of childhood stroke in the pre-hospital and emergency department settings, i.e., “treating the right patients in the right way” The overall values identified for all components of the process map were to maximise the efficiency and effectiveness of each process.

4.1. The emergency call taker, medical service dispatcher and paramedic model

The pre-hospital model included two sub-processes: (i) the emergency call-taker (ECT) and emergency medical service dispatcher (EMSD) assessment, and (ii) the paramedic assessment (Figure 4). The overall value for the pre-hospital process was to maximise efficiency and effectiveness of pre-hospital emergency services stroke diagnosis. The fundamental objectives were: (i) maximise accuracy (effectiveness) of ECT, EMSD and paramedic stroke recognition, and (ii) maximise timeliness (efficiency) of ECT, EMSD and paramedic stroke recognition in the pre-hospital setting, to ensure children with stroke arrive as quickly as possible at hospital. The high level means objectives were to: (i) minimise time delays at every step in the process by ensuring correct triage and identification of stroke, thereby ensuring rapid ambulance transport, (ii) maximise correct recognition of stroke symptoms by EMS dispatchers to ensure rapid dispatch of the correct type of ambulance, and (iii) maximise correct recognition of stroke symptoms and signs by paramedics at the scene, to ensure the child was rapidly transported to an appropriate stroke centre.

Figure 4.

The EMS Call-taker, EMS dispatcher and paramedic sub-processes

The external initiating events required to start the process were presence of stroke-like symptoms in the affected child, and a parental decision to call an ambulance. The call for ambulance assistance could result in the parent waiting for an ambulance, paramedic assessment and ambulance transport to hospital, or the parent ignoring the advice of the emergency medical service dispatcher and transporting their child by private transport to hospital (Figure 4).

4.1.1. The emergency medical dispatcher sub-process

The EMSD event-driven process chain was decomposed into steps including: (i) call-taker identification of the nature of the emergency (as stated by the carer), (ii) call-taker allocation of a chief complaint determinant code using the Advanced Medical Priority Dispatch System (AMPDS) algorithm (Buck et al., 2009), (iii) application of the stroke diagnostic tool (if the problem was stated as stroke, or the caller used keywords that are relevant symptoms for stroke), (iv) allocation and dispatcher notification of a determinant code, and (v) ambulance dispatch.

4.1.2. The paramedic sub-process

The paramedic event-driven process chain was decomposed into the steps including: (i) paramedic assessment (primary survey, obtaining a history of the presenting problem and determining time of symptom onset), (ii) application of the Melbourne Ambulance Stroke Screen (MASS) (Bray et al., 2005), (iii) stabilisation, (iv) pre-notification, (v) transport, and (vi) arrival at hospital.

4.2. The emergency department model

The emergency department model encompassed five sub-processes which included: (i) triage assessment, (ii) GP clinic assessment, (iii) emergency physician cubicle assessment, (iv) resuscitation room assessment, and (v) neuroimaging (Figure 5). The overall values identified for the emergency department process map were to maximise efficiency and effectiveness of the triage, medical and neuroimaging assessments. The fundamental objectives identified for all five sub-processes were to: (i) maximise accuracy (effectiveness) of clinical and radiological diagnosis of stroke, and (ii) minimise time (efficiency) to clinical and radiological stroke diagnosis. The high-level means objectives were to establish clinical and radiological diagnosis of stroke, and to minimise time delay at every step in the process. The external initiating events required to start the emergency department process chain to commence were the combination of arrival of the child in the emergency department and availability of a triage nurse.

Figure 5.

The emergency department and neuroimaging sub-processes

4.2.1. The triage sub-process

Patients can arrive at triage by ambulance or by private transport. Patients arriving by ambulance move directly to the triage stretcher area, where they are met by the triage nurse. In contrast, patients arriving by private transport present to a concierge desk; they are met by a non-medical/non-nursing staff member with basic training in the identification of sick children. The triage process involves a primary survey, obtaining a brief history and determination of the main complaint. The triage nurse assigns a triage category that then determines acuity of care. Patients assigned a category of 1, 2, 3, 4 and 5 are to be seen immediately, within 10, 30, 60 and 120 min respectively (ACEM 2013). Assignment of categories 1 and 2 results in immediate paging of the resuscitation team, and therefore, more rapid medical assessment. Patients are allocated to a resuscitation bay or cubicle, depending on category assigned and bed availability. It should be noted, however, that allocation to a resuscitation bay or cubicle is of secondary importance, because protocols dictate that the resuscitation team sees category 1 patients immediately, regardless of their location in the department. The triage assessment results in three potential outcomes; triage to an emergency department resuscitation bay, an emergency department cubicle or the onsite GP clinic (Figure 6).

Figure 6.

The triage sub-processes

4.2.2. The emergency department medical assessment sub-processes

The resuscitation process includes obtaining a history, examining the patient and formulating a provisional diagnosis, but there are several notable differences from the cubicle assessment. The resuscitation team includes a senior ED consultant, a registrar, and an emergency nurse. The team moves directly to the patient once they are paged by the triage nurse. The primary survey involves initial assessment and stabilisation of vital signs (airway, breathing and circulation), following the Advanced Paediatric Life Support protocol (APLS) (Advanced Paediatric Life Support, 2018). This is followed by a secondary neurological survey to detect conditions requiring emergency treatment, such as raised intracranial pressure or seizures. It is important to note that the APLS protocol does not include assessment for focal neurological symptoms or signs relevant for stroke. The process ends with a decision whether or not to perform neuroimaging (Figure 7).

Figure 7.

The resuscitation room sub-processes

The emergency physician cubicle assessment results in a decision to request brain imaging if the doctor suspects serious disorders including stroke, or exiting the pathway if the possibility of serious neurological conditions are either not considered, or considered and discounted. Observations of vital signs are taken by the emergency nurse on arrival in the cubicle. An emergency physician obtains a history, examines the patient, and formulates a provisional or differential diagnosis. Depending on the doctor’s level of experience and confidence, they may consult a more senior staff member in the department, or seek external consultation from an inpatient team, such as the neurologist or neurosurgeon on call. This is followed by a decision whether or not to perform investigations including neuroimaging.

The GP clinic assessment results in transfer of the patient to an emergency cubicle and referral to an ED physician if the general practitioner suspects serious disorders including stroke, or discharge if serious disorders are not considered.

4.2.3. The neuroimaging sub-process

The neuroimaging assessment results in radiological confirmation of stroke diagnosis, or exclusion of stroke and exiting the pathway (Figure 8). Once the decision is made by the emergency physician to perform a scan, the duty radiologist is called to discuss the clinical indication for the scan. The clinician’s provisional diagnosis or differential diagnoses, and a statement of how imaging will change management, are important in deciding the urgency of imaging (immediately or next available scanning time during normal working hours), and the most appropriate modality to maximise the diagnostic yield (magnetic resonance (MRI) or computed tomography (CT) imaging). The radiologist may request that the clinician seek further opinions from more senior emergency department staff or inpatient units (such as neurologists or neurosurgeons), if the rationale for urgent imaging is unclear or felt to be unsubstantiated.

Figure 8.

The neuroimaging sub-processes

The next step in the process is to determine whether the child requires anaesthesia to obtain adequate images. This is influenced by the age of the child and requires discussion with the duty anaesthetist. Once again, the clinician’s provisional diagnosis or differential diagnoses, and a statement of how imaging will change management, influence decisions about the urgency and type of anaesthesia. The patient is transferred to the pre-determined scanner once it becomes available and intubated if necessary. The radiologist is paged by the radiographer to review the images and then communicates the findings to the referring clinician. If the decision has been made to perform CT imaging, then a normal scan may be insufficient to confirm diagnosis of an ischaemic stroke; this may result in further discussion about the need for, and urgency of MR imaging, resulting in the same processes of deciding need for anaesthesia. Fundamental and means objectives for the neuroimaging sub-process, are presented in Figure 9. Decomposition of the neuroimaging sub-process, demonstrating the link between means objectives and the event-driven process chain, are presented in the Online Supplemental Figure.

Figure 9.

Neuroimaging objectives

4.3. Validation of the VFPE model

The clinical stakeholders repeatedly evaluated each sub-process of the model to ensure face validity by checking that the model closely represented the current process of acute paediatric stroke care. Binary matrix tables (which followed process flow) were constructed and cross-checked by the management science expert, to ensure that each activity in the process chain was linked to at least one means objective, and that there were no activities that did not achieve an objective. The pre-hospital component of the model required 12 iterations, and the hospital component 10 iterations.

5. Barriers and enablers identified using the VFPE conceptual model along the Paediatric Stroke Chain of Recovery

5.1. The EMS dispatcher assessment

In children with stroke, the correct ESTA call-taker action is to allocate a “Stroke” chief complaint (CC) code 28 using the Advanced Medical Priority Dispatch System (AMPDS) (Buck et al., 2009). The correct EMS dispatcher actions are to allocate a mobile intensive care ambulance (MICA) paramedic unit, with a Code 1 (lights and sirens) priority response. Ambulance dispatch must occur within 150 s of receiving the call, but the call-taker usually continues to gather further information from the caller after dispatch. This information may result in changes to the chief complaint codes, type of ambulance dispatched and priority response.

The model demonstrated that all three outcomes are dependent on how the problem is stated by the caller. Three barriers were identified by the model. Firstly, accuracy of call-taker diagnosis, and incorrect allocation of an alternative chief complaint code are possible if the parent does not suspect stroke, or they suspect stroke but do not state this to the call-taker. In turn, this may result in dispatch of an ambulance (rather than MICA) paramedic, or a lower priority response. There may be similar consequences if the parent does not state one of four AMPDS designated stroke symptoms: (i) sudden speech problems, (ii) weakness or numbness of one side of the face, arm or leg, (iii) loss of balance or incoordination, and (iv) severe headache.

Secondly, the AMPDS protocols only allow a one-way shunt away from the Stroke Chief Complaint code 28 to other codes (but not vice versa). The possibility of stroke needs to be recognised by the call-taker within 150 s and the system does not allow reallocation of a Chief Complaint code 28 if the ESTA call-taker has already allocated another code, even if further questioning reveals symptoms suggestive of stroke.

Thirdly, the AMPDS manual for Code 28 has the following axiom: “some younger people have strokes (often fatal) from a ballooned blood vessel called a berry aneurysm that expands and then breaks. This condition is present from birth (congenital). Early symptoms include sudden headache” This statement only refers to the possible occurrence of haemorrhagic stroke, and therefore, may result in the Call-taker failing to consider ischaemic stroke in a child.

5.2. The paramedic assessment

In children with stroke, correct paramedic actions include (i) identification of neurologically relevant signs and symptoms, (ii) consideration of stroke as a cause for the symptoms, (iii) pre-notification, and (iv) rapid transport to hospital. The paramedics should spend no more than 20 min assessing and stabilising the child at the scene.

A major barrier to improving accuracy and timeliness of paramedic diagnosis of paediatric stroke was identified by the model. Ambulance Victoria has developed Clinical Practice Guidelines (CPGs) for specific adult and paediatric disorders, to assist the paramedic with evaluation, stabilisation and urgency of transportation to hospital. The Continuous Tonic-Clonic Seizures (CPG P0703) is the only specific neurological CPG for children. In contrast, there is an adult Suspected Stroke or TIA Clinical Practice Guideline (CPGA0711) (Ambulance Victoria, 2006), which incorporates the Melbourne Ambulance Stroke Screen (MASS) screen (Bray et al., 2005). This CPG provides paramedics with a framework for assessing adults with suspected stroke and provides a list of potential mimic diagnoses to consider. The CPG recommends transport to the nearest hospital providing thrombolysis if the patient is stable, there are no comorbidities, onset time is less than 4.5 h, and transport time is less than 1 h. The exclusion of children from the adult Stroke CPG means that stroke is less likely to be identified, and the child is less likely to be rapidly transported to a stroke-capable centre.

5.3. The triage nurse assessment

The category assigned by the triage nurse determines acuity of care. The triage nurse sub-process is one of two major rate-limiting steps once the child arrives at hospital. In children with stroke, the correct action is to allocate a category 1 or 2, which results in assessment within 10 minutes by the resuscitation team (consisting of senior emergency department staff), and typically transfer to the resuscitation room. Failure to allocate categories 1 or 2 potentially results in wait times ranging from 30 minutes to several hours, and assessment by less experienced emergency staff. Barriers to allocation of a high triage code may include how the problem is presented to the triage nurse by the parent, the nurse’s knowledge of symptoms relevant for stroke, and lack of knowledge that stroke occurs in children.

If the parent provides a history that is not suggestive of stroke, or they do so and the triage nurse does not appreciate the significance of this history, then the child may be allocated a lower triage category and be allotted a cubicle or GP clinic assessment, resulting in a longer waiting time.

It was also clear, from interviews of senior triage nurses, that children arriving to the ambulance bay were more likely to be seen immediately by the nurse than children presenting to the concierge desk. Ambulance-transported children were more likely to be allocated a high category, and to be transferred to the resuscitation room.

5.4. The ED physician assessment

There were three pathways to clinical assessment in the conceptual model but the resuscitation sub-process is the only one that achieves both fundamental objectives. Children waiting for GP clinic or emergency cubicle assessments are often sent to the waiting room, resulting in longer delays. Being seen by a GP or junior emergency staff increases the risk of a missed diagnosis of stroke. These outcomes are likely to be associated with longer delays to diagnosis because cases usually require discussion with more senior emergency physicians. In contrast, a child assigned a high category will be seen rapidly by a pre-determined team, consisting of senior clinicians who follow predefined APLS protocols when assessing the child.

Potential barriers to accurate diagnosis of stroke in the emergency process include (i) failure to identify symptoms or signs relevant for stroke, or (ii) failure to consider stroke in the differential diagnosis of focal neurological deficits. Both factors influence decisions about the need for investigations. Another important barrier is that the APLS manual does not consider focal neurological deficits – and their potential relevance for stroke – in the secondary neurological survey module. Instead, the module is confined to the assessment and management of children with altered conscious state, raised intracranial pressure or seizures (Advanced Paediatric Life Support, 2018). Clinical tools to assist the emergency physician to identify stroke do not currently exist for the paediatric population, and there are no protocols to assist clinicians with decisions about the most appropriate type of brain imaging.

5.5. The neuroimaging sub-process

The neuroimaging sub-process is the second major rate-limiting step along the Paediatric Stroke Chain of Recovery. The clinician must correctly differentiate between ischaemic and haemorrhagic stroke subtypes, to decide the most accurate diagnostic imaging modality (CT for haemorrhagic stroke and MRI for ischaemic stroke). There are two potential pathways to stroke diagnosis in the conceptual model, but only one achieves the fundamental objective of accurate diagnosis of ischaemic stroke, due to the poor sensitivity of CT for detection of early acute infarction.

Accessibility to emergent imaging, particularly MR imaging, is another major barrier to rapid radiological diagnosis of stroke. During the iterative process of developing the model, the senior radiologist interviewed stated that the four key factors influencing decisions about timing and type of brain imaging were: (i) the provisional clinical diagnosis, (ii) involvement of an inpatient specialty unit (such as neurologists or neurosurgeons), (iii) an explanation from the clinician of how imaging would change clinical management and (iv) the need for general anaesthesia. The radiologist was more likely to agree to an urgent MRI if a senior clinician or inpatient specialist team provided a clearly articulated provisional diagnosis.

Anaesthesia is another important barrier which adversely impacts the ability to achieve the fundamental objective of timely diagnosis of stroke. The senior anaesthetist interviewed also stated that prioritisation of stroke over other emergency cases is strongly influenced by the clinician’s explanation of exactly how urgent imaging will change clinical management.

5.6. Summary of the barriers to rapid diagnosis of childhood stroke

In summary, the three barriers identified by the model in the pre-hospital sub-processes were (i) Emergency Medical Services (EMS) call-taker failure to allocate a Stroke Chief Complaint code, (ii) EMS dispatcher failure to allocate a Code 1 (lights and sirens) priority response, and (iii) exclusion of children from paramedic clinical practice guidelines. The five barriers identified by the model in the hospital sub-processes were (i) triage nurse failure to allocate a high triage category, (ii) failure of emergency physicians to recognise symptoms or signs of stroke, (iii) absence of guidelines for the assessment of children with focal deficits in the Advanced Paediatric Life Support Guidelines, (iv) failure of radiologists to agree to MRI imaging for possible AIS, and (v) failure of anaesthetists to allocate a high category (priority) for young children requiring anaesthesia.

5.7. Enablers to rapid diagnosis of childhood stroke

The conceptual model allowed for the identification of specific intervention targets and enablers to overcome barriers and inefficient processes and provided a framework to implement multimodal interventions along the Paediatric Stroke Chain of Recovery. Intervention targets and enablers were identified as follows:

1. Improving awareness of childhood stroke through targeted public education campaigns;

2. Inclusion of children in adult AMPDS stroke algorithms;

3. Inclusion of children in the adult stroke paramedic CPG, or development of a childhood specific Brain Attack CPG, which includes use of a pre-hospital paediatric stroke recognition tool (see point vii below), pre-notification and bypass to designated centres (see point iv below);

4. Reorganisation of paediatric health-care services, with accreditation of primary and comprehensive paediatric stroke care centres with thrombolysis and endovascular treatment capabilities. These centres could include established adult comprehensive stroke centres (CSC) with paediatric ED units, or tertiary paediatric hospitals meeting criteria for CSCs;

5. Development, validation and implementation of a brain attack triage screening tool to assist emergency nurses to identify possible stroke;

6. Addition of a Stroke (or “Brain attack”) protocol to the secondary neurological survey section of the Advanced Paediatric Life Support manual;

7. Development, validation and implementation of a paediatric stroke decision support tool to assist emergency physicians with identification of stroke, differentiation of stroke subtypes, and selection of the most appropriate brain imaging;

8. Development of acute stroke imaging protocols, which recommend MRI for suspected ischaemic stroke and CT for suspected haemorrhagic stroke;

9. Development of paediatric sedation protocols for children requiring urgent brain imaging.

6. Summary of the findings in the context of adult best practice and current paediatric practice

The conceptual model has improved understanding the paediatric Stroke Chain of Recovery through identification of key barriers and enablers along the current process of care (Figure 2). It also illustrated ineffective or inefficient sub-processes (such as emergency department GP assessments), and inefficient activities within each sub-process (such as CT imaging for identification of ischaemic stroke). These redundant sub-processes and activities should be removed in an ideal model of care.

The modelling demonstrated that actions of carers and health professionals have important downstream effects on the process-of-care. Firstly, in the pre-hospital setting, if the parent does not suspect stroke, or mention the key symptoms of stroke, then EMS call-taker diagnosis may not allocate the correct chief complaint code, and the EMS dispatcher may not allocate a MICA ambulance with a Code 1 (lights and sirens) response. Secondly, if the paramedic attending the scene does not suspect stroke, then pre-notification and rapid transport to a stroke capable centre are less likely. The knock-on effect continues following arrival at hospital. If the child does not arrive by ambulance, then they are less likely to be seen immediately by the triage nurse. If the triage nurse does not allocate a high category, then the child is less likely to go to the resuscitation room, thereby delaying time to assessment and reducing the likelihood assessment by a senior ED clinician. The more senior the ED clinician, the greater their experience, and the more likely the radiologist will agree to urgent imaging, and the anaesthetist to triaging a higher anaesthetic category.

6.1. Moving interventions up the stroke chain of recovery

The outcomes from recently published endovascular trials (Berkhemer et al., 2015; Campbell et al., 2015; Goyal et al., 2015; Jovin et al., 2015, Saver et al., 2015), highlight the ever-widening gap between rapid diagnosis and service delivery in adult versus paediatric stroke. Adult interventional trials have driven improvements in coordinated systems of care along the Stroke Chain of Recovery, with the focus on achieving efficiency targets of treatment within predefined time windows. For example, an analysis of acute stroke care systems across 304 US hospitals found that rapid triage/stroke team notification, single-call activation system, and storage of tPA in the emergency department, were independently associated with shorter door to needle times. Hospitals using a greater number of strategies had shorter door to needle times (Xian et al., 2014).

Guidelines for reorganisation of adult stroke care systems, which include bypass of acute stroke-ready hospitals or primary stroke centres in favour of comprehensive stroke centres with the capacity to offer endovascular treatments, have been recently proposed (Smith & Schwamm, 2015). Improved triage of patients managed in mobile stroke units, to hospital with specialised facilities for management of ischaemic or haemorrhage stroke, have demonstrated how changing practice at one step along the Stroke Chain of Recovery results in downstream benefits along the continuum of care (Wendt et al., 2015). Implementation of parallel work flow arrangements, such as administration of tPA in the scanning suite whilst setting up the process for acquiring multiphase CT angiography or perfusion, and transferring the patient to the angiography suite whilst post-processing and interpreting imaging studies, have also improved efficiencies (Goyal et al., 2015, Menon, Campbell, Levi, & Goyal, 2015).

6.2. Improving awareness of paediatric stroke symptoms, primary stroke centres and code stroke protocols

Improving stroke recognition among caregivers, EMS dispatchers, paramedics, triage nurses and emergency physicians requires education about common presenting symptoms which include sudden onset focal weakness, visual or speech disturbances, limb incoordination or ataxia, altered mental status, and headache or seizures with additional neurological symptoms. Stroke symptoms are broadly similar to adults, with the exception being the higher rates of seizures observed in children. Ischaemic stroke and haemorrhagic stroke have different presenting features in children. Ischaemic stroke is more likely to present with focal neurological symptoms (limb or facial weakness, speech disturbance, limb incoordination and ataxia), whereas haemorrhagic stroke is more likely to present with non-focal deficits (headache, vomiting and altered mental state) (Yock-Corrales, Mackay, Mosley, Maixner, & Babl, 2011). The different clinical features of ischaemic and haemorrhagic stroke subtypes impact on the selection of first neuroimaging investigation to confirm diagnosis and to guide emergency management.

Paediatric stroke is infrequently diagnosed within the six-hour time window for reperfusion therapies (Mallick et al., 2015; Rafay et al., 2009; Srinivasan et al., 2009). The literature suggests that in-hospital factors are greater contributors to delayed diagnosis than pre-hospital factors, representing an important difference to adults (Mallick et al., 2015; Stojanovski et al., 2017). The conceptual model demonstrated that failure to allocate a high triage category and poor accessibility of emergent imaging, particularly MRI, were the two most important barriers following arrival at hospital, to the rapid diagnosis of ischaemic stroke. Previous research has also shown that non-diagnostic neuroimaging was the most important factor contributing to delayed diagnosis of ischaemic stroke (Mallick et al., 2015), largely related to selection of CT as the first diagnostic imaging modality. This is because non-contrast computed tomography (CT) is relatively insensitive to the early detection of ischaemia, confirming stroke diagnosis in only 16% to 66% of cases (Mallick et al., 2015; Srinivasan et al., 2009; Yock-Corrales et al., 2011). Therefore, focusing initially on the development of rapid MRI protocols in children with suspected ischaemic stroke is likely to be the most effective change to local systems of acute stroke care. This approach has been adopted at French, Canadian and U.S centres. (Ladner et al., 2015; Shack et al., 2017; Tabone et al., 2017) and some consensus-based paediatric stroke guidelines (Medley et al., 2019) recommend magnetic resonance imaging (MRI) as first imaging modality for suspected ischaemic stroke (Medley et al., 2019).

Recently published paediatric stroke guidelines allow for off-label use of thrombolysis (Ferriero et al., 2019; Medley et al., 2019) but improving access to reperfusion therapies will require changes in health-care systems and work flow practices. The last few years have seen the emergence of primary paediatric stroke centres driven by the (now-closed) Thrombolysis in Paediatric Stroke trial. Infrastructure and personnel requirements for primary paediatric stroke centres have been proposed (Bernard et al., 2014). A study from the Vanderbilt University Medical Centre has demonstrated the feasibility and benefits of developing a Code Stroke protocol (Ladner et al., 2015), resulting in improved the delivery of care including selection of MRI as the first imaging modality in 76% of children, and higher treatment rates with reperfusion therapy than those reported in the paediatric literature. The protocol included assessment of the child within 15 min of Code activation, insertion of an intravenous line and collection of baseline bloodwork, group paging alerts of MRI technicians and neuroradiologists with interruption of routine scanning to accommodate urgent imaging of children with suspected stroke, and use of a rapid (12 to 14 min) stroke MRI protocol.

Another recent publication from Paris, France provides further evidence for the feasibility and benefits of developing of a regional paediatric stroke alert protocols to improve access to reperfusion therapies at two paediatric hospitals (Tabone et al., 2017). Several meetings involving a multidisciplinary working group of paediatric and adult clinicians, radiologists and interventionalists were required to build the protocol, which was adapted from best practice adult stroke guidelines (Jauch et al., 2013). One hospital was co-located with a comprehensive adult stroke centre and the other hospital was 3 kilometres away. Therefore, individual hospital protocols varied slightly according to staffing and facility capabilities specific to each centre but MRI was considered mandatory due to the high rates of stroke mimics in children. Paramedics were educated about paediatric stroke and trained in usage of the FAST scale and children were transported to the closest of the two centres in working hours, and preferentially to one centre after hours. A case-by-case collaborative decision-making process involving paediatric neurologists, vascular neurologists, interventional and diagnostic neuroradiologists was used to select each child for treatment. In the first few years following implementation of the Code Stroke protocol, training of the team and modification of the protocol occurred mainly by continuous evaluation of the process of care for each treated child (Tabone et al., 2017). This study demonstrates the importance of a multidisciplinary approach, with collaboration between paediatric and adult centres, to develop paediatric code stroke protocols, where team members have clearly defined roles based on their particular areas of expertise.

6.3. Limitations of the conceptual model

The main limitation is the difficulty validating conceptual models, when compared to computational models or quality improvement methods. Demonstrating the validity of a conceptual model requires determination that theories and assumptions underlying the model are correct, and that the structure and logic are reasonable for the intended purpose of the model (Sargent, 2013). Face validation and structured walk-throughs were performed but further work is required to determine whether the conceptual model is reasonable to solve the problem of delayed diagnosis of childhood stroke.

The conceptual model was developed for an Australian tertiary paediatric hospital. Some components of the model development are generic, such as public education and emergency medical dispatcher use of the AMPDS algorithm (developed for the US health-care system but adopted for use by many other countries including Australia), and use of prehospital stroke recognition tools. Other components, such as the nursing triage and emergency department assessments, are specific to our institution. The purpose of the study was to apply a generic value-focused process-engineering modelling technique to a specific paediatric health-care setting, to map the system of acute stroke care. The modelling methodology was demonstrated to improve understanding of barriers and enablers to more effective and efficient care. As a result, this generic methodology can be applied to evaluate other specific health-care systems which may have different processes.

6.4. Conclusions and future directions

Identification and acute management of paediatric stroke is a complex process, which involves multiple interacting factors such as actions of the community and health professionals, and utilisation of health-care resources. Changing practice at one step along the Paediatric Stroke Chain of Recovery will result in downstream benefits along the continuum of care. As a consequence, each component of the Chain cannot be viewed in isolation.

We have demonstrated how VFPE can be used to understand and improve the effectiveness and efficiency of the process of acute paediatric stroke care, from symptom onset to radiological confirmation of diagnosis. The conceptual model has allowed identification of specific barriers and enablers along the Paediatric Stroke Chain of Recovery and provides key stakeholders with a framework to implement multimodal interventions, as presented in Figure 2. Potential further steps are to:

1. Develop a simulation model from the conceptual model to test the model in a real-world dataset of paediatric brain attacks, where patient flow can be tracked through the model, to determine whether the logic is correct and necessary accuracy is maintained. This may allow comparison to existing adult stroke simulation models, for a gap analysis of current paediatric and the adult gold standard of acute stroke care;

2. Develop paediatric stroke recognition tools, which take into account differences between children and adults, to assist health professionals in the pre-hospital and emergency department settings with the accurate identification of childhood stroke;

3. Develop decision support tools, to identify children with higher likelihood of AIS who require immediate imaging, in contrast to children with less time critical conditions where imaging can be delayed to reduce pressure on busy paediatric medical imaging departments;

4. Develop rapid brain imaging protocols, which take into account the need for sedation in younger children;

5. The conceptual model needs to be externally validated, and tested in real-world settings which include other health-care systems, using quantifiable metrics to demonstrate performance improvement at key steps identified by the model along the paediatric stroke chain of recovery. Measurable outcomes in the prehospital setting could include improved accuracy of EMS call-taker and paramedic stroke diagnosis, higher rates of lights and sirens response and prenotification. In the ED setting they could include, increased selection of MRI as the first imaging modality in suspected AIS and increased rates of thrombolysis usage. Time-based metrics could include time from paramedic arrival at the scene to hospital arrival, from hospital arrival to radiological confirmation of diagnosis, and from hospital arrival to commencement of thrombolysis (door to needle time). Improving acute paediatric stroke care requires consideration of the fundamental objective, common to all stakeholders, of treating the right patients in the right way (to improve outcomes) versus separate, potentially conflicting means objectives for individual stakeholders and departments. The positive or negative impact of the acute stroke protocol on individual stakeholder departments (e.g., emergency or medical imaging) should also be quantified pre- and post-protocol implementation. Metrics could include length of stay in the ED, length of stay in hospital, increased utilisation of MRI or decreased utilisation of non-diagnostic CT. If the protocol resulted in higher rates of reperfusion therapy then the economic impact of treatment could be measured using the Child Health Utility 9D (CHU-9D) (Stevens, 2012), and hospitals linked services health utilisation and cost data;

6. The VFEDP model could also be compared to other methodologies such as those used in adult stroke care. For example, prior to commencement of the NINDS tPA trial, the Total Quality Improvement (TQI) method – which assumes that a highly variable process adds extra steps and time – was used to improve the efficiency of processes of care (Tilley et al., 1996). The TQI method involved use of step-by-step flow charts to identify hospital-specific variability and inefficiencies prior to implementation of the trial protocol, followed by structured walk throughs with team members in order to improve efficiencies in each component of the process of care, and feedback after each participant enrolment to the trial to encourage and motivate clinicians (Tilley et al., 1996).

Disclosure statement

No potential conflict of interest was reported by the authors.

Supplementary material

Supplemental data for this article can be accessed here.