Master Thesis
Testing and Comparing the Effects of Intelligent
Patient Routing Strategies on the Performance
of Stroke Systems: a Simulation Study
Luuk Platvoet
l.r.platvoet@student.rug.nl
S2767414
MSc Technology & Operations Management
Faculty of Economy & Business
University of Groningen
Supervisor:
dr. ir. D.J. van der Zee
Second assessor:
dr. J.A.C. Bokhorst
3 ABSTRACT
Purpose: Recently, trauma networks face the logistical challenge of organizing the pathway for their patients as efficiently as possible. One of these trauma networks is the stroke pathway. Recent studies have tried to develop and test intelligent patient routing strategies to improve the routing of stroke patients with a large vessel occlusion (LVO) by giving Emergency Medical Services means to make a more informed decision where to transfer the patient. The effects of two intelligent patient routing strategies are tested in this study: prehospital screening scales and guidelines for LVO patient transfer. To improve the generalizability of this study the strategies will be tested in stroke systems with differing characteristics. This is done by changing the door-in door-out (DIDO) times and the regional characteristics of the stroke systems.
Methodology: This study will use a simulation model to test the effects of the strategies in a representation of a realistic environment. The simulation model represents the stroke system in the Northern Netherlands. The model is extended to include the patient routing strategies. By running several experiments the effects the intelligent patient routings strategies have on the performance of the different stroke systems are examined.
Findings: The intelligent patient routing strategies improve outcomes for LVO patients in all stroke systems. Stroke systems should only consider screening scales with both a high sensitivity and specificity. DIDO times and regional characteristics determine the importance of the strategies for a stroke system. Whereas the strategies only show a limited effect on the performance of stroke systems with a fast DIDO time, they show large improvements in systems with a slow DIDO time. More rural regions should look to implement the guidelines of the European Stroke Organization, whereas more urban regions already benefit from the guidelines of the American Stroke Association.
Conclusions: Stroke systems may benefit from implementing intelligent patient routing strategies to increase the performance of their stroke system for LVO patients.
4 PREFACE
I conducted this research in order to obtain my master’s degree for the MSc Technology & Operations Management at the University of Groningen. The process of writing my thesis has been a challenging one and while writing I have had great support from several people I would like to thank.
First of all, I would like to thank my supervisor Durk-Jouke van der Zee for his guidance throughout my entire thesis process. Besides providing very useful feedback during our meetings every week, he was always willing to answer questions and send extra information. Secondly, I want to thank my second supervisor Jos Bokhorst, for providing useful feedback. Furthermore, I would like to thank Maarten Lahr (UMCG) for his help with introducing me to the stroke research field and Willemijn Maas (UMCG) for her help with introducing me to the Plant Simulation models. I also would like to express my appreciation to Maarten Uyttenboogaart (UMCG), who helped me with setting up the experiments and validating the results of my thesis.
5 LIST OF ABBREVIATIONS
CSC Comprehensive Stroke Center
CT Computed Tomography
CTA Computed Tomography Angiogram DIDO Door-In Door-Out
DS Drip-and-Ship
DTN Door-To-Needle
ED Emergency Department
EVT Endovascular Thrombectomy EMS Emergency Medical Services IAT Intra-Arterial Thrombectomy IVT Intravenous Thrombolysis LVO Large Vessel Occlusion
MS Mothership
MSU Mobile Stroke Unit mRS Modified Rankin Scale
NIHSS National Institutes of Health Stroke Scales PSC Primary Stroke Center
RCT Randomized Clinical Trial
6 1. INTRODUCTION
Acute care patients are patients with an injury or illness that is acute and poses an immediate risk to a person’s life or long-term health (Ramanayake et al., 2014). Because of these emergency conditions it is important for these patients to be treated as soon as possible. An important example of acute care can be found in the treatment of stroke patients. Each year over 13.7 million people suffer from a stroke worldwide, of which 5.5 million people do not survive the stroke (World Stroke Organisation, 2019). This makes a stroke one of the most common causes of disability and death, with severe associated socioeconomic consequences (Feigin et al., 2017).
There are two types of stroke which both require a different treatment, ischemic and hemorrhagic. Around 85% of the strokes are ischemic strokes and only 15% of the strokes are hemorrhagic strokes (Mackay & Mensah, 2004). This research focuses on the treatment of ischemic strokes, therefore when referring to a stroke in this research an ischemic stroke is meant. When suffering from an ischemic stroke or brain infarction, a human brain on average loses 1.9 billion neurons per minute (Saver, 2006). This means that the transportation time to the right hospital is crucial in the treatment of strokes, as each minute saved in time to treatment adds over a day of extra healthy life on average (Monks et al., 2017).
The stroke treatment consists of either an Intravenous Thrombolysis (IVT) treatment or an IVT treatment followed by an Endovascular Thrombectomy (EVT) treatment. When the stroke patient suffers from a large vessel occlusion (LVO), an IVT treatment may no longer be sufficient to treat the patient and additional EVT treatment may improve patient outcomes. Whereas many community hospitals acting as Primary Stroke Centers (PSCs) have the capability to treat patients with the IVT treatment, the EVT treatment is only limited to more technologically advanced Comprehensive Stroke Centers (CSC). Currently, guidelines advocate that patients suffering from a stroke are transferred to the nearest stroke center based on transportation time, whether that is a PSC or a CSC. For LVO patients it might be more beneficial to be transferred to a CSC immediately as this might decrease the time it takes until they receive the EVT treatment (Fassbender et al., 2020).
7 Every stroke system has different characteristics. One of the main differences between stroke systems can be found in the door-in door-out (DIDO) times of the PSC. The DIDO time is defined as the total time the patient spends in the PSC, beginning when the patient is first documented as arriving at the referring hospital and ending when the patient is discharged to the care of the transport team (Reimer et al., 2013). Regional characteristics like geography will also have an influence on the performance of the stroke system (Ciccone et al., 2019). With geography the number of stroke centers and the spread of these stroke centers across the region are meant. For example, stroke systems in more rural areas might have a longer time to treatment because of longer transportation times.
Several studies have developed and tested intelligent patient routing strategies in stroke systems, however limited research has been conducted in order to test and compare the effects of these strategies in multiple stroke systems with different characteristics. Therefore, the research objective of this study is:
Identifying the effects of intelligent patient routing strategies on the performance of stroke systems
The performance of the stroke systems will be measured in time to treatment and the probability of a good functional outcome of the patients. This study will contribute to the already existing literature by providing insights on the effects of the intelligent patient routing strategies on a realistic stroke system. To represent a realistic stroke system, a simulation study is used. To compare the effects of the strategies in stroke systems with different characteristics, the experiments are executed in stroke systems with different DIDO times and regional characteristics. The findings of this study can be used to gain insights on how to organize and optimize the routing of stroke systems and which effects the intelligent patient routing strategies can have on the performance of stroke systems.
This study will answer the research objective in a representation of a realistic environment by means of a simulation model. This simulation model is based on the stroke system in the Northern Netherlands and uses historical data of stroke patients in the Northern Netherlands. In this simulation model, experiments will be executed which test the effects of the different intelligent patient routing strategies on the different stroke systems.
8 2. THEORETICAL BACKGROUND
2.1 Acute care networks
Studies have shown that quality of care is higher at trauma centers with appropriate resuscitative, imaging, surgical and critical care facilities (Metcalfe et al., 2016). As a consequence, many countries in the world are now developing acute care networks. Within these trauma networks, hospitals are designated as specialist trauma centers that focus on providing treatment for one specific kind of acute care. By focusing on only one kind of acute care, these trauma centers become specialized in providing treatment for those patients. Research shows an improvement in medical outcomes for patients with myocardial infarction, stroke, cardiac arrest and acute respiratory distress syndrome in the designed trauma networks (Feazel et al., 2015). Because of the specialization towards trauma centers the likelihood of trauma patients requiring treatment in hospitals further away from their homes increases (Metcalfe et al., 2016). This creates the logistical challenge for acute care networks to organize their pathway in such a way that every patient is transferred to the right hospital, as every minute lost decreases the chances of a good outcome for acute care patients (Fassbender et al., 2020).
2.2 Stroke networks
The stroke network is one of these acute care networks for which the challenge of organizing the pathway as efficiently as possible is relevant. Currently, there are two treatments for an ischemic stroke. The first, and most used, treatment is the IVT treatment. This treatment is used to dissolve the blood clot that is causing the stroke. To be effective it needs to be administered within 4.5 hours of the stroke onset (Berkhemer et al., 2015). The second treatment is the EVT treatment, which is also referred to as Intra-Arterial Thrombectomy (IAT). This treatment aims to remove or break up the blood clot by means of a catheter inserted intravenously through the groin. EVT treatment needs to be performed within 6 hours of the stroke onset (Berkhemer et al., 2015).
9 logistics (Lahr et al., 2020a). Since ‘time is brain’ in stroke care, the speed by which a patient with LVO is transferred to a CSC determines the probability of good functional outcome for a patient.
2.3 Dominant patient routing models
There are two models that co-exist together to guide the patient routing for stroke patients, the ‘drip and ship’ (DS) model and the ‘mothership’ (MS) model. Standard guidelines for stroke management recommend transferring a patient to the nearest hospital capable of administering IVT, which can be either a PSC or a CSC (Lahr et al., 2020a). If the nearest stroke center is a PSC, the patient will be transferred there and receive the IVT treatment. The patient will also go through a CTA scan to check for a possible LVO. If an LVO is diagnosed the patient needs to be transferred to the nearest CSC to perform the EVT treatment. This model is called the DS model (Fassbender et al., 2020).
In case the nearest stroke center is a CSC, the patient is transferred to the CSC. In the CSC the patient would first receive the IVT treatment and go through the CTA scan. If an LVO is detected the EVT treatment is performed in the CSC as well. This model is called the MS model (Fassbender et al., 2020). Since most regions only have a few CSCs, transport time to the nearest CSC is on average longer than the transport time to the nearest PSC (Ciccone et al., 2019). Therefore, most patients are currently transferred to the PSC according to the DS model. The advantage of the DS model is that all the stroke patients without an LVO can be treated with the IVT treatment as soon as possible. However, for the patients with an LVO the time to treatment increases as they need to be transferred after receiving the IVT treatment in the PSC to the nearest CSC to receive the EVT treatment. The time to treatment for a patient with an LVO is the time that passes from the stroke onset till the start of the EVT treatment.
2.4 Intelligent patient routing
One of the solutions to improve the routing of LVO patients is to give the Emergency Medical Services (EMS) means to make a more informed decision on which stroke center to transport the patient to. This is referred to as intelligent patient routing. By using intelligent patient routing strategies EMS might be able to send more LVO patients to the MS model and thereby improve the outcomes for LVO patients. The intelligent patient routing strategies which will be considered in this research will be adjusting the guidelines for transferring LVO patients and prehospital screening scales.
2.4.1 Guidelines for transferring patients
10 longer than the travel time to the nearest PSC (Maas et al., 2020). Whereas the guidelines of the American Stroke Association recommend bypassing the PSC if the transportation time to the CSC is less than 15-20 minutes longer (Benoit et al., 2018). By changing the guidelines for LVO patient transfer, more LVO patients might be transferred to the further away CSC. 2.4.2 Prehospital screening scales
Most of the prehospital screening scales which are under development have been derived from the National Institutes of Health Stroke Scale (NIHSS). Since the NIHSS has 15 variables on which the patient should be screened it takes a lot of time before the screening is complete. Therefore, several other scales have been developed based on the variables used by the NIHSS model. These scales could be used to determine whether a patient is suspected to have an LVO and needs to be transferred to a CSC instead of a PSC. The advantage of these scales compared to the NIHSS is that these scales are faster and easier to use for EMS personnel onset. However, the prehospital screening study field is still new. Therefore, most of the prehospital screening scales are in development and the studies in this field focus on the quality of the screening scales. This means that these studies discover the sensitivity and specificity of these screening scales, but these studies do not focus on finding which effects the screening scales have on the eventual outcomes of patients. Therefore, the applicability of the screening scales in an actual stroke system has not been tested yet.
11 2.5 Stroke system characteristics
Ciccone et al. (2019) sought to identify the optimal model for organizing stroke networks. They found no model that was superior and therefore recommended that the choice of model should be determined based on the characteristics of the stroke system. Since every stroke system has different characteristics, the effects of the intelligent patient routing strategies could also differ per stroke system. Even though it is important to take these characteristics into account, most of the studies in these trauma networks do not provide the details of their systems’ characteristics but focus more on the medical perspectives and outcomes (Vali et al., 2017). In this research the stroke system characteristics that will be taken into account are the DIDO times and the regional characteristics, specifically the geography of the region.
2.5.1 Door-in door-out times
Since not all stroke centers are equally technologically advanced DIDO times can differ per system. Only the DIDO times in the PSCs are relevant for stroke systems, since the time to treatment for stroke patients stops after the start of their EVT treatment in the CSC. Even though the DIDO times should be taken into account, limited research has been conducted on the effects of these inhospital efficiencies on the organization of patient routing in a stroke system (Benoit et al., 2018). Stroke systems with PSCs with a slow DIDO time might have a higher need for better patient routing than stroke systems with a fast DIDO time, as having a slow DIDO time will increase the time to treatment for LVO patients who are sent to the DS model significantly. Since every minute counts in the treatment of stroke patients, immediate transport to a further away CSC could then be the better option for the patient. PSCs should be stimulated to make their DIDO times as fast as possible, as this would make direct transfer to a further away CSC a less favorable option and increase the number of patients that will be transferred to the PSC (Venema et al., 2020).
2.5.2 Regional characteristics
Besides the logistical characteristics of the system, there are many regional characteristics within a region which will have an influence on the decision where to transfer the patient e.g., the demography, topography, climate and transportation systems (Ciccone et al., 2019). The most obvious regional characteristic is the geography of the region. Generally, undertreatment of stroke is most pronounced in rural areas (Fassbender et al., 2020), especially CSCs are located almost exclusively in metropolitan centers. Since people in rural areas tend to live further away from these cities the average transportation time from the hospital to the patient and vice versa increases, which influences the chances that the patient is in time for an IVT or EVT treatment. For the stroke systems with higher average transportation times the intelligent patient routing strategies might be more necessary than for stroke systems with lower average transportation times.
2.6 Simulation Study
12 the organization of acute stroke care. While these RCTs will answer which model is the best for that specific region, it is uncertain whether these results may be generalizable to other regions (Maas et al., 2020). The advantage of using a simulation study compared to a RCT is that it is less time consuming and less expensive to experiment with a simulation model than with a real system. Simulation also allows for more comprehensive and detailed analysis (Robinson, 2014). Another strength of a simulation model is its flexibility i.e., the model can be easily adjusted and adapted to test the system in a variety of alternative settings. Therefore, a simulation study is capable of testing and comparing the effects of the intelligent patient routing strategies in stroke systems with different characteristics.
2.7 Summary of findings
13 3. CASE DESCRIPTION
3.1 Regional characteristics
The stroke system in the Northern Netherlands is used as the stroke system to answer the research objective. The Northern Netherlands consists of the provinces of Groningen, Friesland and Drenthe. There is only one CSC in this region, located in the capital city of Groningen (the University Medical Center Groningen (UMCG)), and there are nine community hospitals acting as PSCs (see appendix 1). The area comprises around 1.7 million inhabitants with a mean population density of 209 inhabitants per square kilometers. Compared to other regions in the Netherlands, the Northern Netherlands consists of more rural areas with a relatively low population density. Most of the PSCs and CSCs are located in somewhat larger cities, in which the population density is higher. It takes 1 hour to transport a patient by car from the furthest away PSC (Sneek) to the CSC in Groningen and maximally 1,5 hours by car from the furthest away towns in the region to travel to the CSC. For every town in the area there is a PSC within 30 minutes travel distance by car.
3.2 Stroke system set-up
Currently, the standard guidelines for stroke patient routing are followed in this region. These guidelines state that the patient needs to be transferred to the stroke center which is the nearest. This means that most patients in the provinces of Friesland and Drenthe are routed to the nearest PSC, according to the DS model. In case an LVO is found in the PSC the patients are transported from the PSC to the CSC in Groningen. The patients in the province of Groningen are mostly routed directly to the CSC, according to the MS model. Some patients are already in the hospital when the stroke occurs, however most patients arrive at the hospital by ambulance transportation.
3.3 Pathway description
The patients follow the pathway of the DS and MS model which are displayed in figure 1a and 1b. In the prehospital phase the patient enters the system after the stroke onset. Often, bystanders or the family of the patient call the emergency number (112) for help, the ambulance responds and drives to the location of the patient, picks up the patient on scene and then transports the patient to the nearest stroke center.
14 From this point onward the DS and MS model differ. The patients in the MS model continue in the inhospital phase and receive the EVT treatment as they are already in a CSC. The phases of the EVT treatment are preparing for angiography suite and groin puncture.
If a patient is scanned with an LVO in a PSC, the patient enters the interhospital phase. This means that the patient needs to be transferred to a CSC according to the DS model. The ambulance has to be called to transfer the patient, after which the ambulance responds by driving to the PSC, picking up the patient and transporting the patient to the CSC where the patient is handed over to the personnel of the CSC. After arriving at the CSC, the patient enters the inhospital phase for the CSC. Some patients require additional diagnostics to check if the EVT treatment is necessary. Patients who do not need the additional diagnostics and the patients for whom the EVT treatment is necessary after the diagnostics are brought to the angio suite for EVT treatment.
3.4 Intelligent patient routing
Several changes need to be made to test the intelligent patient routing strategies in the stroke system. First, the DS model and the MS model need to be combined into one new composite organization model. When the models are composed, the routing of LVO patients’ needs to be added to the new model. This has to be integrated after the EMS arrives on scene. At this decision point the patient is transferred to either the DS model or the MS model. Currently, this decision is based on which stroke center is the nearest. This decision will be adjusted for the different intelligent patient routing strategies.
The simplified organization model of the new stroke system with the combined MS and DS model can be found in figure 2.
15
Figure 1a: Drip & Ship model
16
17 4. DESIGN OF THE SIMULATION STUDY
4.1 Simulation study
Simulation can effectively cope with variability and interconnectedness between variables (Robinson, 2014). A lot of uncertainties can be found throughout the whole stroke pathway, for example the time it takes before the ambulance is called for the stroke patient, the time to transport the patient to the hospital and the time it takes to treat the patient. Given the time-sensitive nature of acute stroke interventions, simulation modelling could facilitate insights into the complex interplay of separate elements of the pathway (Lahr et al., 2020a). Previous research has demonstrated that simulation models for representing the acute stroke pathway can be accurately developed within different settings (Lahr et al., 2020b).
4.2 Research objective
The research objective “Identifying the effects of intelligent patient routing strategies on the performance of stroke systems” will be examined by performing several experiments in a simulation model. These experiments represent a scenario in which the intelligent patient routing strategies are added to a stroke system. Then, the stroke system characteristics will be adjusted in the system to be able to compare the effects that the intelligent patient routing strategies would have in a different stroke system.
4.3 Simulation model
The simulation model will be built and extended in Plant Simulation version 15 (Siemens Plant Simulation, 2021). In this simulation model the DS model is extended with the MS model into one new composite organization model, to accurately represent a realistic setting. The simulation model resembles the stroke system in the Northern Netherlands as described in section 3. The simulation model builds upon and extends simulation models which were created by Maas et al. (2020). The new conceptual model of the simulation model, the model codes, the distributions of the time variables and the formulas used for calculating the probability of good functional outcome can be found in appendix 2. In the simulation model, only patients with an LVO exist. Therefore, non-LVO patients do not exist in this model and the effects of the intelligent patient routing strategies on non-LVO patients cannot be measured. In this new model the patients who are already in the stroke center will not be taken into account. These patients will not be affected by the intelligent patient routing strategies, as they are already in the stroke center and cannot be routed differently.
18 4.4 Experiments
The research objective will be examined by executing several experiments. The first experiment will test the new integrated DS and MS model. The results of the new integrated DS and MS model will be used as base setting for the other experiments. The results of the base setting can be used to compare the results of the other experiments to the results of the stroke system in the Northern Netherlands without any implemented strategies.
Experiment 2 and 3 will measure the effects of the intelligent patient routing strategies on the stroke system. Afterwards, experiment 4 and 5 will test these strategies in stroke systems with different characteristics, by changing the DIDO times and the regional characteristics. The experiments with the experimental factors can be found in table 1.
Experiment nr Model variables Experimental factors Experiments
1 Base setting
Intelligent patient routing strategies
2 Guidelines for patient transfer Transport decision 2. Maximum difference in transportation time PSC-CSC
3 Prehospital screening scales Transport decision 3.1 Sensitivity 3.2 Specificity
Stroke system characteristics
4 DIDO times Arrival to CT PSC Routes through PSC Call for transfer PSC-CSC Response EMS PSC EMS on scene PSC
4.1 Fast DIDO times 4.2 Slow DIDO times
5 Regional characteristics Transport to CSC Transport to PSC Interhospital transport PSC-CSC 5.1 Rural regions 5.2 Urban regions Table 1: Experiments
19 4.5 Experimental design
Experiments will be conducted with a sample size of 100,000 patients, to make sure the 95% confidence interval half width for the mean time to treatment and the probability of a mRS score between zero and two is below 1%.
4.6 Model outputs
The model will show the percentage of patients following the DS model and percentage of patients following the MS model as output. The model will also show the mean time to treatment, and its confidence intervals. The probability of a good functional outcome of the patients is calculated with a mRS score, which is based on the time to treatment, the age of the patient, the NIHSS score of the patient and the CTA collateral grading score. A mRS score between 0 and 2 indicates that the patient will return to functional health within 90 days. Therefore, this model also returns the mean probability that a patient receives a mRS score between 0 and 2 after being treated.
4.7 Data collection
Data for this study has been collected by looking at 287 patients in the Northern Netherlands who were treated with the EVT treatment. These patients were included in the MR CLEAN registry. The patient characteristics and inhospital data consisted of time delays obtained by the MR CLEAN registry. This data was cross-checked by using internal quality databases. Data for the prehospital and interhospital time delays of the ambulance transports, was collected from three EMS and linked to the MR CLEAN registry data. The data in the model was collected by Maas et al. (2020), this study made use of this data and extended it with data that was found in the literature about the intelligent patient routing strategies and stroke characteristics. Subsequently, this data was cross-checked by conducting domain experts. 4.8 Experimental factors
4.8.1 New model
The DS model is extended with the MS model by adding all the process steps of the MS model and its distributions to the simulation model of the DS model. In the new model the decision whether to transfer the patient to a PSC or immediately to a CSC needs to be added after the EMS arrives at the patient on scene. In the first experiment the decision where to transfer the patient is made based on which stroke center is the nearest. This is done by comparing the transportation time to the PSC and to the CSC. The nearest stroke center will be chosen for patient transfer by the simulation. Since all patients in this model are assumed to have an LVO, this decision is made for every patient.
4.8.2 Guidelines for transferring LVO patients
20 maximum difference of 20 minutes. The model will then interpret the transportation time to the PSC as 7 + 20 = 27 minutes, this time is compared to the transportation time to the CSC and when the transportation time to the CSC is less than 27 minutes the patient is transferred to the CSC. If the transportation time to the CSC is more than 27 minutes, the patient is transferred to the PSC. The guidelines of the American Stroke Association and the European Stroke Organization will be used for the experiments and can be found in table 2.
Guidelines for LVO patient transfer Maximum difference
American Stroke Association - 15 minutes - 20 minutes European Stroke Organization - 30 minutes - 45 minutes Table 2: Experiment 2: Guidelines for LVO patient transfer
4.8.3 Prehospital screening scales
In the model every patient is assumed to have an LVO, however most of the time the EMS cannot determine whether the patient has an LVO or not. Therefore, the results of the former two experiments where only possible in an ideal situation where it is always known that a patient has an LVO. Since EMS can use prehospital screening scales to identify whether a patient has an LVO or not, prehospital screening scales will be added to the experiment with the guidelines for LVO patient transfer. Three scales have been found in the literature and will be used for the experiments in this research. A large number of prehospital screening scales is under development, the scales which are used for this research are chosen based on their sensitivity and specificity.
Besides the prehospital screening scales, the possibility of adding an MSU to the stroke pathway is also explored. Adding an MSU would make it possible to immediately scan the patients for an LVO with a CTA scan. It can be assumed that the CTA scan has a sensitivity and specificity of 100% and thus make it possible to always correctly identify where the patient needs to be transferred to. The sensitivity and specificity of the found screening scales and the MSU can be found in table 3.
The first experiment will look at the sensitivity of the screening scale as a decision variable. The prehospital screening scales determine whether a patient is suspected to have an LVO or not. In case of a positive scan the patient is transferred to the nearest stroke center according to the guidelines for LVO patient transfer. The LVO patients with a negative scan are send to the PSC and will therefore go through the DS model.
21 further away CSC. The number of non-LVO patients who are wrongfully sent to the CSC will also be measured.
Prehospital screening scale Sensitivity Specificity
RACE1 77% 75%
C-STAT2 83% 40%
PASS3 71% 84%
MSU 100% 100%
Table 3: Experiment 3: Prehospital screening scales
4.8.4 DIDO times
The DIDO times influence the overall time to treatment of patients in the DS model and can make transportation to the PSC more or less preferable. To test the faster and slower DIDO times for the PSC, all the variables starting from arrival to CT until the handover to EMS for the second transfer need to be modified. In a scenario with fast DIDO times, the average time spent in the hospital will be 50 minutes while a scenario with an average DIDO time of 120 minutes will be considered slow (Holodinsky et al., 2018). Therefore, the experiments that will be conducted set the average DIDO times for PSCs at 50 and 120 minutes. The average DIDO time in the DS model is 93 minutes. Therefore, the DIDO time of the experiment needs to be divided by 93 and that outcome determines the number the process steps which influence the DIDO times in the model need to be multiplied by to adjust the DIDO times in the simulation model. The outcomes can be seen in table 4.
DIDO times Multiplied by
50 minutes 0.538 120 minutes 1.290 Table 4: Experiment 4: DIDO times
4.8.5 Regional characteristics
The regional characteristic of which will be tested is the geography of the region. The geography influences the transportation times to the PSCs and CSCs, which can influence which stroke center is the nearest and thus influence where the patient is transferred to. Several scenarios will be tested, the first scenario will be a stroke system which lies in a more rural region. In a more rural region, the locations of the stroke centers will be more widespread across the region and thus increase the average transportation times. Therefore, the transportation time to the CSC, the transportation time to the PSC and the interhospital transportation time from the PSC to the CSC have to be adjusted. An experiment will be conducted in a region where only the CSCs are located more spread out across the region and an experiment will be conducted in a region where both the PSCs and CSCs are located more spread out across the region. After discussion with a domain expert, the distributions for the transportation times in the model will be increased by 50% for the two experiments.
22 The second scenario will test a stroke system that lies in a more urban region. In a more urban region, the locations of the stroke centers might be less spread out across the regions. Therefore, an experiment is conducted where only the CSC is located more optimally and an experiment is conducted where the PSCs and CSCs are located more optimally. After discussion with the domain expert the transportation time will be decreased by 25%. The experiments can be seen in table 5.
Regional characteristics Experimental factors Experiments Rural regions
Further away CSC Transport to CSC
Interhospital transport PSC-CSC
+ 50% Further away PSC and CSC Transport to CSC
Transport to PSC
Interhospital transport PSC-CSC
+ 50%
Urban regions
Better location CSC Transport to CSC
Interhospital transport PSC-CSC
- 25% Better location PSC and CSC Transport to CSC
Transport to PSC
Interhospital transport PSC-CSC
23 5. ANALYSIS
In this chapter the results of the experiments will be discussed. The results of all the experiments regarding the time to treatment and mRS score with their confidence intervals and the percentage of patients per model can be found in appendix 3. Since the time to treatment is a major determiner of the probability of good functional outcome, the results in time to treatment will be mentioned in this analysis. The average time to treatment in minutes of the experiments will be abbreviated with (T), the percentage of patients who follow the MS model will be abbreviated with (MS%) and the average transportation times will be abbreviated with (Transp.)
5.1 Base setting
The results of the base setting can be found in table 6. The confidence intervals of the time to treatment and probability of a good mRS score can be found between the squared brackets. In 90.2% of the stroke cases the PSC is the nearest stroke center. Only in 9.8% of the stroke cases the CSC is the nearest stroke center. The total average time to treatment is 238.05 minutes. The average time to treatment in the MS model is 183.17 minutes, while the average time to treatment in the DS model is 244.03 minutes. This means that patients in the MS model receive the EVT treatment about an hour earlier than patients in the DS model. In the base setting the difference in average transportation times between the DS model (Transp. = 10.70) and MS model (Transp. = 13.69) are small.
Experiment 1 Transportation times (min.)
Time to treatment (min.)
mRS Score % Patients per model
Base setting DS: 10.70 MS: 13.69 Total: 238.05 [237.34 - 238.76] 0.512 [0.509 - 0.515] DS: 90.2% MS: 9.8% Table 6: Results experiment 1: Base setting
5.2 Guidelines for transferring LVO patients
24 Experiment 2 Transportation times
(min.)
Time to treatment (min.)
mRS Score 0-2 % Patients per model
Base setting DS: 10.70 MS: 13.69 238.05 [237.34 - 238.76] 0.512 [0.509 - 0.515] DS: 90.2% MS: 9.8% 15 minutes DS: 9.96 MS: 18.37 218.02 [217.31 – 218.72] 0.536 [0.533 – 0.539] DS: 55.0% MS: 45.0% 20 minutes DS: 9.08 MS: 21.18 210.04 [209.34 – 210.75] 0.545 [0.542 – 0.548] DS: 38.5% MS: 61.5% 30 minutes DS: 6.46 MS: 25.26 198.10 [197.40 – 198.80] 0.557 [0.554 – 0.560] DS: 9.2% MS: 92.8% 45 minutes DS: 0.00 MS: 26.40 195.13 [194.77 – 195.48] 0.560 [0.559 - 0.562] DS: 0% MS: 100%
Table 7: Results experiment 2: Guidelines for LVO patient transfer
5.3 Prehospital screening scales
To test the prehospital screening scales the sensitivity and the specificity of the scales have been used. Experiments have been conducted without the guidelines and with the guidelines of a maximum difference of 20 minutes and a maximum difference of 45 minutes. Patients who are scanned positively for an LVO will be transported according to the guidelines for patient transfer. Patients who are negatively scanned for an LVO will be transported to the nearest stroke center.
5.3.1 Sensitivity
The first part of the experiment considers the sensitivity of the screening scales, the results can be found in figure 3a and 3b. The prehospital screening scale with the highest sensitivity (C-STAT) shows the best results in time to treatment, therefore the results of the C-STAT will used for comparing the experiments with each other and with the base setting. Without the guidelines the screening scales (C-STAT: T = 239.09, MS% = 8.2) do not show an improvement when comparing the results to the base setting (T = 238.05, MS% = 9.8). This makes sense, since when implementing the screening scales only the patients who are scanned positively are assumed to have an LVO. When implementing the guidelines of 20 minutes (T = 215.80, MS% = 51.1) and 45 minutes (T = 203.39, MS% = 83) together with the screening scales, the results show large improvements compared to the base setting. The differences between the results of the three screening scales grows when implementing the screening scales with higher guidelines. Therefore, the sensitivity of the screening scales increases in importance when implementing higher guidelines. The results of the MSU are the same results as the results obtained in experiment 2, since the MSU replicates an optimal situation where every patient is known to have an LVO.
5.3.2 Specificity
25 to the MS model if the guidelines of 45 minutes (C-STAT: Transp. = 15.01, MS% = 60) are compared to the guidelines of 20 minutes (C-STAT: Transp. = 10.64, MS% = 37) and 0 minutes (C-STAT: Transp. = 2.73, MS% = 6). This increase in transportation time means that implementing the guidelines would result in a longer time to treatments for non-LVO patients. Whereas the guidelines show an effect on the increase in transportation time, the differences in specificities of the screening scales show little effect on the increase in transportation time. These specificities do account for the number of patients who are wrongfully sent to the CSC. Having a low specificity would mean sending a lot of patients wrongfully to the CSC, hereby increasing their time to treatment and also filling the capacity of the CSC with non-LVO patients. Therefore, stroke systems should look for screening scales with a high specificity. The base setting and MSU results are the same for this experiment as both send 100% of the patients to right DS model.
5.4 DIDO
The first part of the experiment with DIDO times tests the intelligent patient routing strategies in a stroke system with PSCs with a DIDO time of 50 minutes, the results can be found in figure 5a and 5b. The number of patients that are sent to the DS and MS model is the same as for both experiments, as the DIDO times did not influence the decision where to transfer the patient in this experiment. When looking at the results, the intelligent patient routing strategies have a very limited effect. Only small differences can be found between the different intelligent patient routing strategies. The faster DIDO times have decreased the time to treatment for the patients in the DS model. Therefore, the difference in average time to treatment between the DS model (C-STAT: T = 200.77) and the MS model (T = 182.96) has declined.
The second part of the experiment tests the intelligent patient routing strategies in a stroke system with PSCs with a DIDO time of 120 minutes, the results can be found in figure 6a and 6b. Contrary to the experiment with DIDO times of 50 minutes, the time to treatment in the DS model (T = 271.25) has now increased compared to the time to treatment in the MS model (T= 182.79). Without the guidelines for patient transfer, a large number of patients is transferred to the DS model (MS%: 9.8). All these patients are affected by the very slow DIDO times. The differences between the experiments with a DIDO time of 50 minutes (C-STAT: T = 199.31) and a DIDO time of 120 minutes (C-STAT: T = 264.10) are significant in the experiment without the guidelines. When implementing the guidelines, the results of the experiments are brought a bit closer, especially with the guidelines of 45 minutes (C-STAT: DIDO = 50: T = 196.05; DIDO = 120: T = 208.05). This can be related to the increase in number of patients who are transferred to the MS model when increasing the guidelines from 0 minutes (C-STAT: MS% = 8.2) to 45 minutes (C-STAT: MS% = 83.0%). Since only the patients in the DS model are affected by the DIDO times, decreasing the number of patients that go through the DS model also decreases the average time to treatment significantly.
26 experiments, since the MSU sends 100% of the patients to the MS model and thus LVO patients would not be affected by the DIDO times.
5.5 Regional characteristics
The first experiment tests the intelligent patient routing strategies in a more rural region. The results of the experiments in a more rural region can be found in figures 7a-7f. In more rural regions, the average time to treatment (C-STAT: T = 255.78) is higher than in the base setting (T = 238.05). What can also be seen is that the guidelines of 45 minutes no longer send every patient to the MS model (MSU: MS% = 90.8) and the guidelines of 45 minutes provide the most improvements in time to treatment. The prehospital screening scales show little differences in results, however the screening scale with the highest sensitivity (C-STAT) provides the best outcomes again. Changing the guidelines does improve the performance of the stroke systems.
The experiments show a faster time to treatment in the experiments with the guidelines of 0 minutes when only the CSC is further away (C-STAT: T = 255.31) compared to the results of the stroke system where both the CSC and the PSCs are further away (C-STAT: T = 255.78). However, the results only differ slightly. When using the guidelines of 20 minutes the time to treatment is actually slower when only the CSC is further away (C-STAT: T = 241.43) compared to the system where both the PSC and the CSC are further away (C-STAT: T = 239.01). This is explained by the fact that the average time to treatment in the MS model (C-STAT: T = 195.73) is more than 60 minutes shorter than the average time to treatment in the DS model (C-STAT: T = 260.08). When the transportation times to the PSC are increased, the CSCs become the nearest stroke center more often. Therefore, more patients are transferred to the faster MS model and this increase outweighs the increase in transportation times for the patients who are sent to the DS model. This effect can be seen mostly in experiments with 20 minutes where the difference in number of patients sent to the MS model is high when comparing the results of the experiment with the increase in transportation time to the CSC (C-STAT: MS% = 22.5) to the results of the experiment with the increase in transportation time to the PSC and CSC (C-STAT: MS% = 33.3).
27
Figure 3a: Results Time to Treatment experiment 3.1: Prehospital screening scales: Sensitivity + Guidelines for LVO patient transfer Figure 3b: Results % of patients MS Model experiment 3.1: Prehospital screening scales: Sensitivity + Guidelines for LVO patient transfer
28
Figure 5a: Results Time to treatment experiment 4.1: DIDO times: 50 minutes + Intelligent patient routing strategies Figure 5b: Results % of patients MS model experiment 4.1: DIDO times: 50 minutes + Intelligent patient routing strategies
29
Figure 7a: Results Time to treatment experiment 5.1.1: Rural regions: Guidelines 0 minutes Figure 7b: Results Time to treatment experiment 5.1.2: Rural regions: Guidelines 20 minutes Figure 7c: Results Time to treatment experiment 5.1.3: Rural regions: Guidelines 45 minutes
30
Figure 8a: Results time to treatment experiment 5.2.1: Urban regions: Guidelines 0 minutes Figure 8b: Results time to treatment experiment 5.2.2: Urban regions: Guidelines 20 minutes Figure 8c: Results time to treatment experiment 5.2.3: Urban regions: Guidelines 45 minutes
31 6. DISCUSSION
Recently, stroke systems face the challenge of organizing their pathway as efficiently as possible. This study shows that the MS model provides much better outcomes for LVO patients compared to the DS model. Since every minute counts for stroke patients, the intelligent patient routing strategies that are tested in this research could be a solution to improve the performance of stroke systems for LVO patients.
Intelligent patient routing strategies
The first intelligent patient routing strategy is the implementation of guidelines for LVO patient transfer. The results of this experiments show that the expectation of Fassbender et al. (2020) that direct transfer to a CSC could improve patient outcome was correct. The guidelines of the American Stroke Association and the European Stroke Organization both show improvements in patient outcome for LVO patients. In the stroke system of the Northern Netherlands the guidelines of 45 minutes show the largest improvement, because then every patient is transferred to the MS model. However, to implement these guidelines the EMS should know whether a patient has an LVO or not. Therefore, the guidelines can only be implemented in combination with a form of prehospital testing.
When implementing the guidelines together with the prehospital screening scales, the outcomes improve for the LVO patients. The sensitivity of the screening scales should be taken into account. A higher sensitivity means sending more LVO patients to the MS model and therefore the higher sensitivity the better the outcomes for LVO patients. The screening scales chosen in this research did not show large differences in results, this could be explained by the sensitivities which were not very far apart from each other. The differences between the results of the screening scales grew when the guidelines for LVO patient transfer were implemented, therefore the sensitivity of the scale becomes more important when implementing the scales together with higher guidelines.
Venema et al. (2020) stated that the harm of delaying potential IVT for non-LVO patients should also be taken into account when implementing prehospital screening scales. This harm is the added transportation time because of the transfer to the further away CSC. Non-LVO patients suffer from this harm if they are positively scanned for an LVO because of the specificity of the screening scales. The lower the specificity, the more patients are wrongfully sent to the CSC. Implementing higher guidelines for LVO patient transfer increases the added transportation time to the CSC for non-LVO patients.
32 The best results are achieved when a MSU is added to the model. An MSU would have a sensitivity and a specificity of 100% and would therefore be beneficial for both the LVO as well as the non-LVO patients. However, adding the MSUs to stroke systems is still in early development and more research needs to be conducted before it is possible to add MSUs to the stroke system. The costs for implementing an MSU are very high and only one stroke patient can be transferred at a time. Therefore, stroke systems would require significant financial resources to be able to have the capacity to transport all the stroke patients with an MSU (Ciccone et al., 2019).
Stroke system characteristics
The results show that Venema et al. (2020) was right when stating that stroke systems should stimulate their PSCs to focus on decreasing their DIDO times. The results of this study show that a stroke system with PSCs with a DIDO time of around 50 minutes has a small difference in average time to treatment between the DS and the MS model. Because of this small difference in time to treatment, improving the routing of patients has a smaller effect on patient outcome. In contrast, stroke systems with PSCs with a slow DIDO time should pay more attention to the routing of patients and should look to implement the patient routing strategies. Because of the slow DIDO times, the difference in time to treatment when comparing the MS model to the DS model is high. In these stroke systems it is important to send as many LVO patients to the MS model as possible, as every LVO patient who is sent to the DS model is affected by the DIDO times. Therefore, implementing the intelligent patient routing strategies could prove to be a good solution for these systems.
Implementing the intelligent patient routing strategies results in better outcomes in a more rural region as well as a more urban region. In more rural regions, the guidelines of 45 minutes show the largest increase in outcomes for LVO patients. Using smaller guidelines would only result in a minor increase of LVO patients who are routed to the MS model. Even though the guidelines of 45 minutes show the best results in the urban regions, the guidelines of 20 minutes could also be used by the stroke system in the Northern Netherlands and even more urban regions. Since the guidelines of 45 minutes show an increase in transportation time for non-LVO patients who are wrongfully transported to the CSC and therefore have a negative effect on the time to treatment for non-LVO patients.
This study shows that stroke systems should focus on decreasing the transportation time to the CSC, by locating it more optimally or adding another CSC to the stroke system. By decreasing the transportation time to the CSC, the CSC becomes the more favorable transfer option more often. Another possibility could be to only make the PSCs with the fastest DIDO times available for transfer of LVO patients. This would not only decrease time to treatment for patients who are transferred to the DS model, but also decrease the number of LVO patients who are transferred to the DS model.
Limitations
33 measured. The simulation model only included patients with an LVO, patients without an LVO are not represented in the simulation model. This study tried to include the non-LVO patients by looking at the increase in transportation time for these patients, however further research should focus on finding the effects the strategies have on the non-LVO patients’ outcomes. Secondly, the stroke patients who are already in the hospital have not been taken in to account. Sometimes a stroke occurs when a patient is already hospitalized. These patients were represented in the initial model of Maas et al. (2020), however this study has chosen to remove these patients from the model. This choice was made because these patients will not be affected by the patient routing strategies since no routing decision needs to be made for these patients. Lastly, the simulation model is based on the stroke system in the Northern Netherlands. This research tries to make the results more generalizable by testing and comparing the intelligent patient routing strategies in stroke systems with different characteristics. Therefore, these results can be used as guidelines for others stroke system. However, other stroke systems should look carefully at their own characteristics to see if they match the characteristics of the stroke systems in this research.
34 7. CONCLUSION
Stroke systems may benefit from implementing intelligent patient routing strategies to improve the performance of their stroke system for LVO patients. In this study, the implementation of intelligent patient routing strategies shows an increased outcome for all the stroke systems with different characteristics. Guidelines for LVO patient transfer increase the number of LVO patients who are transferred to the MS model and thus improve the outcome for LVO patients. Because it is necessary for the guidelines of LVO patients transfer to know whether a patient has an LVO or not, EMS should receive means to determine on scene whether a patient has an LVO or not. Prehospital screening scales could be used as means, however currently there is no screening scale which provides a high enough sensitivity and specificity to also limit the consequences for non-LVO patients. Once a sufficient screening scale is developed, implementing this screening scale together with the guidelines for patient transfer would result in large improvements for LVO patient outcome. A higher sensitivity of the scale results in better outcomes for LVO patients. In an optimal future situation EMS would use an MSU to pick up the patient, diagnose the patient and transfer the patient to the right stroke center. Other stroke systems or acute care networks who want to implement intelligent patient routing strategies can use these findings and fit the intelligent patient routing strategies into their system. However, stroke systems need to know the DIDO times of their PSCs and the spread of their PSCs and CSCs across their region to establish the best strategy to use for their region. For stroke systems with a fast DIDO time implementing the intelligent patient routing strategies is less of a necessity, because of the small differences in time to treatment between the DS and MS model. Slow DIDO times have a large influence on the performance of the stroke system. Therefore, stroke systems in which a large number of the patients are transferred to the DS model should either focus on decreasing the DIDO times or implementing the intelligent patient routing strategies. More rural regions with larger transportation times should look to implement the guidelines of 45 minutes of the European Stroke Organization, while more urban regions with smaller transportation times should implement the guidelines of 20 minutes of the American Stroke Association.
Future research
36 REFERENCES
Abilleira, S., Pérez de la Ossa, N., Jiménez, X., Cardona, P., Cocho, D., Purroy, F., ... & Ribó, M. (2019). Transfer to the local stroke center versus direct transfer to endovascular center of acute stroke patients with suspected large vessel occlusion in the Catalan territory (RACECAT): study protocol of a cluster randomized within a cohort trial. International Journal of Stroke, 14(7), 734-744.
Benoit, J. L., Khatri, P., Adeoye, O. M., Broderick, J. P., McMullan, J. T., Scheitz, J. F., ... & Eckman, M. H. (2018). Prehospital triage of acute ischemic stroke patients to an intravenous tPA-ready versus endovascular-ready hospital: a decision analysis. Prehospital Emergency Care, 22(6), 722-733.
Chalos, V., LeCouffe, N. E., Uyttenboogaart, M., Lingsma, H. F., Mulder, M. J., Venema, E., ... & Coutinho, J. M. (2019). Endovascular treatment with or without prior intravenous alteplase for acute ischemic stroke. Journal of the American Heart Association, 8(11), e011592.
Ciccone, A., Berge, E., & Fischer, U. (2019). Systematic review of organizational models for intra-arterial treatment of acute ischemic stroke. International journal of stroke, 14(1), 12-22.
Feazel, L., Schlichting, A. B., Bell, G. R., Shane, D. M., Ahmed, A., Faine, B., ... & Mohr, N. M. (2015). Achieving regionalization through rural interhospital transfer. The American journal of emergency medicine, 33(9), 1288-1296.
Feigin, V. L., Mensah, G. A., & Norrving, B. (2015). The global burden of stroke. Neuroepidemiology, 45(3), 143-145.
Hastrup, S., Damgaard, D., Johnsen, S. P., & Andersen, G. (2016). Prehospital acute stroke severity scale to predict large artery occlusion: design and comparison with other scales. Stroke, 47(7), 1772-1776.
Holodinsky, J. K., Williamson, T. S., Demchuk, A. M., Zhao, H., Zhu, L., Francis, M. J., ... & Kamal, N. (2018). Modeling stroke patient transport for all patients with suspected large-vessel occlusion. JAMA neurology, 75(12), 1477-1486.
Jumaa, M. A., Castonguay, A. C., Salahuddin, H., Shawver, J., Saju, L., Burgess, R., ... & Zaidi, S. F. (2020). Long-term implementation of a prehospital severity scale for EMS triage of acute stroke: a real-world experience. Journal of neurointerventional surgery, 12(1), 19-24.
37 Kim, C. H., Jeon, J. P., Kim, S. E., Choi, H. J., & Cho, Y. J. (2018). Endovascular treatment with intravenous thrombolysis versus endovascular treatment alone for acute anterior circulation stroke: a meta-analysis of observational studies. Journal of Korean Neurosurgical Society, 61(4), 467.
Lahr, M. M., van der Zee, D. J., Luijckx, G. J., Vroomen, P. C., & Buskens, E. (2013). A simulation-based approach for improving utilization of thrombolysis in acute brain infarction. Medical care, 51(12), 1101-1105.
Lahr, M. M., Maas, W. J., van der Zee, D. J., Uyttenboogaart, M., & Buskens, E. (2020). Rationale and design for studying organisation of care for intra-arterial thrombectomy in the Netherlands: simulation modelling study. BMJ open, 10(1), e032754.
Lahr, M. M., van der Zee, D. J., Luijckx, G. J., & Buskens, E. (2020). Optimising acute stroke care organisation: a simulation study to assess the potential to increase intravenous thrombolysis rates and patient gains. BMJ open, 10(1).
Mackay, J., & Mensah, G. A. (2004). The atlas of heart disease and stroke. World Health Organization.
Metcalfe, D., Perry, D. C., Bouamra, O., Salim, A., Woodford, M., Edwards, A., ... & Costa, M. L. (2016). Regionalisation of trauma care in England. The bone & joint journal, 98(9), 1253-1261.
Monks, T., Van der Zee, D. J., Lahr, M. M., Allen, M., Pearn, K., James, M. A., ... & Luijckx, G. J. (2017). A framework to accelerate simulation studies of hyperacute stroke systems. Operations Research for Health Care, 15, 57-67.
Ramanayake, R. P. J. C., & Sudeshika Ranasingha, S. L. (2014). Management of emergencies in general practice: role of general practitioners. Journal of family medicine and primary care, 3(4), 305.
Reimer, A. P., Hustey, F. M., & Kralovic, D. (2013). Decreasing door-to-balloon times via a streamlined referral protocol for patients requiring transport. The American journal of emergency medicine, 31(3), 499-503.
Robinson, S. (2014). Simulation: the practice of model development and use. Palgrave Macmillan.
Saver, M. D., & Jeffrey, L. S. (2006). Time is brain—quantified. Stroke, 37(1), 263-266. Vali, Y., Rashidian, A., Jalili, M., Omidvari, A. H., & Jeddian, A. (2017). Effectiveness of
38 Venema, E., Burke, J. F., Roozenbeek, B., Nelson, J., Lingsma, H. F., Dippel, D. W., & Kent, D. M. (2020). Prehospital triage strategies for the transportation of suspected stroke patients in the United States. Stroke, 51(11), 3310-3319.
Siemens Plant Simulation. Plant Simulation Online tutorial. Accused 20th of January 2021. Retrieved https://plant-simulation.de/schulungen/tutorial/
39 APPENDICES
Appendix 1: Stroke centers in the Northern Netherlands
40 Appendix 2: Plant Simulation model, model codes and model distributions
New Plant Simulation model
41 Model codes
For patients with a stroke the time variables; ‘time from stroke onset to 911 call’, ‘EMS response’ and ‘EMS on scene’. The EMS Transport variable is adjusted to include the decision where to transfer the patient. In the new model this decision is based on the nearest stroke center. The code of this decision is:
-- Distribution for patients number
TransportTimes := z_demp(57, sorting_Transport)
The sorting of patients determines that there are 122 patients (based on the 122 data entries in the transportation file delivered by W. Maas), which all have an equal chance of being
chosen. This can be seen in figure a3, which goes on till 122 patients.
Figure a3: Table sorting_Transport
-- Distribution for transport times
Transport_Times["NumTranspPatient", NumStrokePatient] := TransportTimes
Transport_Times["Transport time PSC", NumStrokePatient] := Transport_info["Transport PSC", TransportTimes]
Transport_Times["Transport time CSC", NumStrokePatient] := Transport_info["Transport CSC", TransportTimes]
42
Figure a4: Table TransportTimes
43
Figure a5: Table Transport_Times
This is an example of the eventual table Transport_Times. This table will be filled with all the 100.000 patients. For every patient a new random patient number will be generated on which the transport times to PSC and CSC are based.
d := Transport_Times["Transport time PSC",NumStrokePatient] d1 := Transport_Times["Transport time CSC", NumStrokePatient]
d: is the value which stands for transport time to PSC and d1 is the value which stands for transport time to CSC. After the transportation times to the PSC and CSC are determined, the patient is transferred according to the following code. In case the LVO scale = 0, patient is moved to the DS model and if d < d1 the patient is also moved to the DS model. In case, the LVO scale = 1 and d < d1 is not true the patient is transferred to the MS model.
@.scale :=z_demp(55, Sorting_LVO)
-- If scales says negative
switch @.scale
-- Patient transported to PSC and into DS model
case 0
@.MSModel := 0
@.move(Arrival_to_CT)
44 @.Patient_ID_DS := NumDSPatient
@.Delay := @.Delay + d @.TranSp_delay := d
--NIHSS --> NIHSS score of 0 = 1-15, NIHSS score of 1 =
>15
@.NIHSS_DS := z_demp(50, NIHSS_score_DS)
--age --> age score 0 <= 70, 1>70
@.age_DS := z_demp(51, age_score_DS)
--Collaterals, difference between NIHSS score 0 and score
1 -->
--Collaterals score 0 = filling >50% or 100% of occluded
area
--Collaterals score 1 = filling absent or <50% of occluded
area Switch @.NIHSS_DS case 0 @.Collaterals_DS := z_demp(52, CollateralsAndNIHSS_score0_DS) case 1 @.Collaterals_DS := z_demp(53, CollateralsAndNIHSS_score1_DS) end
-- If scale says positive else
-- Transport times to PSC shorter than transport times to CSC -- Transport to PSC and into DS model
if d + 0 <= d1 @.MSModel := 0 @.move(Arrival_to_CT) NumDSPatient := NumDSPatient + 1 @.Patient_ID_DS := NumDSPatient @.Delay := @.Delay + d @.TranSp_delay := d
--NIHSS --> NIHSS score of 0 = 1-15, NIHSS score of 1 =
>15
@.NIHSS_DS := z_demp(50, NIHSS_score_DS)
--age --> age score 0 <= 70, 1>70
@.age_DS := z_demp(51, age_score_DS)
45
1 -->
--Collaterals score 0 = filling >50% or 100% of occluded
area
--Collaterals score 1 = filling absent or <50% of occluded
area Switch @.NIHSS_DS case 0 @.Collaterals_DS := z_demp(52, CollateralsAndNIHSS_score0_DS) case 1 @.Collaterals_DS := z_demp(53, CollateralsAndNIHSS_score1_DS) end
-- Transport to CSC and into MS model else @.MSModel := 1 @.move(Arrival_to_CT1) NumMSPatient := NumMSPatient + 1 @.Patient_ID_MS := NumMSPatient @.Delay_MS := @.Delay @.Delay_MS := @.Delay_MS + d1 @.TranSp_delay_MS := d1
--NIHSS --> NIHSS score of 0 = 1-15, NIHSS score of 1 =
>15
@.NIHSS_MS := z_demp(50, NIHSS_score_MS)
--age --> age score 0 <= 70, 1>70
@.age_MS := z_demp(51, age_score_MS)
--Collaterals, difference between NIHSS score 0 and score
1 -->
--Collaterals score 0 = filling >50% or 100% of occluded
area
--Collaterals score 1 = filling absent or <50% of occluded
46 end
end end
In this code the experiments of the guidelines and the experiments with the prehospital screening scales can be coded.
The experiments with the prehospital screening scale are coded in the @.scale :=z_demp(55, Sorting_LVO).
The sorting_LVO table looks like this, for example with the LAMS sensitivity
Figure a6: Sorting_LVO sensitivity table
When experimenting with the specificity:
Figure a7: sorting_LVO specificity table
And for all the other experiments where it was assumed that it was known that the patient had an LVO:
Figure a8: Sorting_LVO base setting
For the experiments with the guidelines the code
else
-- Transport times to PSC shorter than transport times to CSC -- Transport to PSC and into DS model
