GIS Mapping Evaluation of Stroke Service Areas in Bangkok Using Emergency Medical Services

Abstract

Stroke is a major cause of death in Thailand. It requires a specific time period of 4.5 h from onset to treatment to increase recovery rates, and therefore, it can be categorized as a time-sensitive disease. The objective of this research is to identify whether the service areas of the main existing Emergency Medical Services (EMS) prehospital stroke practices cover all areas in Bangkok. This is determined by using GIS mapping. After verifying the current EMS delivery models, comparisons are drawn to find the travelling time of each model. The conditioning factors for GIS mapping were collected and verified, including the traffic speed and duration spent in each mode of the prehospital stroke process. After inputting all of the data into GIS, the service areas were visualized to show the area serviced in each delivery model. The results also show the different hospital groups, including the service areas for (1) non-network hospitals and (2) hospitals with stroke networks. Suggestions for re-networking and adding more hospitals to the existing networks were also identified using GIS. Regularly updating the service area with up-to-date data in GIS is key to improving stroke service areas.

1. Introduction

According to the World Health Organization (WHO) [1], stroke is one of the top three global causes of death, challenging all governments to act against this rapidly increasing number. In Thailand, according to the Burden of Disease Research Program, which collects and reports Thailand’s deaths by cause every 5 years, stroke is also in the top three leading causes of death in the country [2].

Werner Hacke, M.D et al. (2008) [3] stated that the “golden hour”—the time that will produce excellent outcomes with significantly lower rates of morbidity and mortality—for stroke, has been adjusted from 3 h to 4.5 h for the time of onset to treatment. In this process, the use of recombinant tissue plasminogen activator (rt-PA) has been the standard of care for stroke treatment for several years. However, according to Rajiv Advani et al. (2017) [4], patients treated within one hour of symptom onset can lead to excellent outcomes without any incidence of iatrogenic bleeds. Additionally, the time of arrival at the hospital until rt-PA administration, referred to as the Door-to-Needle time (DTN), requires 1.5–3.0 h, and patients who are presented to the hospital within the first 60 min of onset time have the greatest opportunity to benefit from recanalization therapy (Saver, J. L. et al., 2010) [5]. Thus, the golden hour from onset time to hospital admission should be within 1 h. There have been some studies that focus on how to improve the Emergency Medical Services (EMS) transportation as the highest priority. The prehospital notification procedure (PNP) is also associated with faster in-hospital stroke responses and represents logical targets for EMS quality improvement efforts. (Oostema, J. A. et al., 2014) [6].

In terms of healthcare facilities, hospitals in Thailand are operated by both the public and private sectors. The majority are operated by the Ministry of Public Health (MOPH), while private hospitals are regulated by the Medical Registration Division under the MOPH’s Department of Health Service Support. However, Bangkok as the capital is not the same as other provinces. In Bangkok, the proportion of private hospitals is higher than in other provinces, and public hospitals are prevalently under the supervision of other government units and public organizations, including the military, universities, local governments, and the Red Cross [7]. This variety of hospital management agencies can lead to dissonant healthcare services when transferal or referral among hospitals is needed.

To ensure that people can access proper health care services throughout Thailand, the National Health Security Office (NHSO), one of the agencies under the Department of Health, MOPH, manages and assesses the health care system throughout the country. Health care service provision is divided among 13 NHSO regions [8]. Each NHSO region covers 4–6 provinces, with an estimated 4–6 million people in the economy of scale unit. NHSO Region 13 covers Bangkok, while NHSO Regions 1–12 cover other provinces across Thailand. The benefits of setting up these health care regions are reflected in each region’s ability to self-manage. This allows more flexibility for each region to manage itself and design its own development plan that is more suitable for the respective region, rather than following the same central guideline from MOPH. NHSO Region 13, covering Bangkok, has also set up 22 service plans to take care of each disease, including stroke, with coordination from hospitals and related parties. A stroke service plan is also organized to share information to the relevant partners [9].

Bangkok is the capital and most populous city of Thailand. The city occupies 1568.5 square kilometers in central Thailand and has an estimated population of 10.723 million as of 2021 [10], which is equal to 15.43 percent of the country’s population. Bangkok’s rapid growth, coupled with little urban planning, has resulted in a haphazard cityscape and inadequate infrastructure.

As a megacity with a population of more than 10 million people, Bangkok has its specific characteristics that are different from other provinces of Thailand, including a wide range of incomes, high costs of living, and rushed lifestyles.

The traffic in Bangkok is also well known for its congestion. Data from INRIX Global Traffic Scorecard (2018) [11]—a report compiled by a private company that measures the impact of congestion for car commuters by estimating the total number of hours the average commuter spends in congestion in each city—shows that Bangkok ranks 11th among the most congested urban areas worldwide, with an average congestion level of 23% and 64 peak hours spent in congestion per year. Despite an extensive expressway network, an inadequate road network and substantial private car usage have led to chronic and crippling traffic congestion. This results in travel times from origin to each destination becoming longer than expected.

From Gache, K. et al. (2013) [12], EMS is suggested for people who have stroke symptoms, as the time elapsed between the onset of symptoms and brain imaging is significantly shorter in cases of EMS transportation. Patients who called for EMS could access timely brain imaging 2.7 times more often than patients using self-transportation. In Thailand, although each hospital has its EMS system to support patients, the service is still limited to the area that is near to that hospital. The dial 1669, the public emergency hotline for Bangkok residents, is suggested for patients when they have any symptom and need medical service. Patients call and inform the call center. Then, call center will contact nearest and available EMS operator in network to pick up the patient and deliver to nearest hospital in Bangkok area. However, according to Yuwares S. et al. (2014) [13], EMS in Thailand is still limited despite having been established for over 20 years. Most EMS systems have insufficient medical personnel and medical devices, in addition to the lack of knowledge of medical personnel. In emergency rooms, overcrowding is the most common issue, while problems with medical devices, collaborations with other organizations, and communication devices are the main problems in terms of prehospital EMS. Riyapan, S. et al. (2020) [14] also found that stroke patients represent 13.8%, or the 3rd rank, of EMS cases to Siriraj Hospital in Thailand.

Murad A. (2018) [15] also use GIS to determine health access and identify the service area of health centers in Jeddah. However, to date, there is no study that identifies the service area of the Stroke Service Plan of NHSO Region 13, which uses EMS as its current stroke service delivery model. This research aims to use GIS analysis to identify whether or not the service area of the main existing prehospital delivery model for stroke covers all areas in Bangkok. In addition, this study will benefit NHSO Region 13 in identifying areas that are not in its stroke service area, providing suggestions to re-network in order to expand the current service area.

2. Data and Methodology

2.1. Study Area

This study focuses on hospitals located in Bangkok, which are under the NHSO Region 13. However, as most hospitals in Bangkok comprise a high portion of private hospitals, while public hospitals are under the supervision of many government units, it is difficult to obtain information from each hospital separately, including data on available stroke facilities. Thus, the data in this study will come from the Stroke Service Plan Committee of NHSO Region 13, representing the data of Bangkok as in Figure 1. The size of road is also displayed in this figure.

NHSO Region 13 also collects data from hospitals located in Bangkok; however, it still has not been able to acquire data from all hospitals in the city. Problems persist in terms of data linkages between hospitals, even among government hospitals. This study includes the hospitals in the system of NHSO Region 13 and the linkages within the prevalent delivery model in Bangkok. The study will examine the stroke delivery model by EMS from the stroke onset scene until hospital admission. As previously mentioned, in this study, the presented results for the stroke service area will only regard the area of Bangkok city, and will not mention the area of the provinces in the vicinity.

2.2. Data Preparation

In building GIS models of the hospitals, we collected the relevant data focusing on stroke in order to run the delivery models. This is in order to verify that the models can be used to represent actual situations.

The main input data of this study will be the Demand and Supply for stroke treatment services with an appropriate Means of Transportation. This includes the speed and delivery data that was collected and selected for the delivery models.

As per Figure 2, this study will show the current correlation among the three categories of input data and will provide suggestions for adjusting stroke facilities and networks to support the population in Bangkok. The study workflow includes selecting the input data, filling in GIS program and analyzing the result, as in Figure 3.

2.2.1. Data and Sources of Data

Three categories of data, demand, supply, and means of transportation, were selected to analyze the delivery models in this study.

As the patient demand is confidential, the demand in this study is derived from the density of the population in Bangkok. By comparing the demand with the service area of each hospital, we can determine how many people the current hospitals can and cannot serve. The population data is classified by both district level and sub-district level.

For Bangkok, there are 2 sources of population data. The first is the Bangkok Metropolitan Administration (BMA), the local government agency of Bangkok (also called Krung Thep Maha Nakhon in Thai), which officially collects and registers population data in Bangkok based on the number of individuals who have registered their household in Bangkok. This data can be obtained every year, as it is already systematized. The second data source is the National Statistical Office (NSO), placed under the Office of the Prime Minister, which is responsible for conducting the population census—a head count of every individual living in Thailand according to place of residence. NSO also manages statistics and information to be integrated and standardized so as to develop and support the country’s competitiveness. The census is conducted every 10 years and the most updated census data is from 2010 [16].

The land area can also be obtained from BMA. Bangkok occupies 1568.5 square kilometers in the Chao Phraya River delta in central Thailand. It is divided into 50 districts (known in Thai as Khet), which are further subdivided into 180 sub-districts (Khwaeng).

In case of supply, this study will rely on the lists and data from hospitals registered in NHSO Region 13. There are a total of 111 registered hospitals, as shown in Figure 1. As each hospital has different facilities for stroke services, in this study, we divided each hospital into categories in order to establish the delivery models. In NHSO Region 13, each hospital develops their own facilities to support stroke services, which can cause difficulty in categorizing the services offered by each hospital. Thus, the Stroke Service Plan Committee of NHSO Region 13 has categorized the hospitals into three levels based on facilities provided by each hospital, which helps to identify the hospitals’ capability to support stroke services. Level 1 includes hospitals with general care for stroke. Level 2 hospitals can provide round-the-clock rt-PA services, while Level 3 hospitals can provide 24-h rt-PA and thrombectomy services [17].

Moreover, in NHSO Region 13, some hospitals have developed their own stroke networks focusing on patient referral and knowledge and resource sharing. Currently, there are 8 networks in NHSO Region 13. In each network, there is 1 main hospital with varying numbers of member hospitals, as shown in Figure 4. Each network will develop its own rules and regulations to support each other’s referral processes. This is in order to shorten patient delivery times and increase efficiency within the network.

To link the demand with the supply, the appropriate means of transportation plays a crucial role. The traffic speed and stroke delivery model data were collected from several sources that could provide reliable and up-to-date data for each delivery model.

Bangkok is well known for its traffic congestion. The traffic’s impact on vehicle speed is critical in the treatment of time-sensitive diseases such as stroke. In this study, traffic data will come from The Intelligent Traffic Information Center Foundation (iTIC) [18], officially approved and registered by the Ministry of Interior as a non-profit organization which collects public traffic data from both government and private sources in order to develop high-quality real-time traffic reports. The service of iTIC was established to minimize traffic congestion, increase road safety, and improve the efficiency of the logistics system in Thailand. iTIC collects and processes public traffic data from both government (CCTV footage) and private sources from taxi, bus, logistic and mobile phone probes. The combined traffic data is processed for added value before being disseminated to the public. This study drew upon iTIC’s probe data from 2017–2019, particularly focusing on speed by route and by period.

The transportation models to deliver stroke patients in Bangkok can be categorized into 2 types as follows. In the first model, the patients will use ambulance services or Emergency Medical Services (EMS) to travel directly to a hospital. In Bangkok, the hospitals with stroke networks will develop prehospital notification procedures (PNP) with main EMS providers, which shortens the time in the emergency department (ED). Non-network hospitals do not have such procedures. For the second model type, the patient will travel to the hospital themselves by any means of transportation [19].

For GIS mapping of the networks, we used GIS data of the Bangkok map and road network from the BMA GIS Data Center [20], which is under the Bangkok Metropolitan Administration (BMA). The BMA GIS Data Center provides GIS data in ShapeFile format with a map scale of 1:20000.

2.2.2. Data Preparation

Data pre-processing is essential for solid modelling. It ensures that the required data is adequate and valid for building the GIS model in question. The objectives of data pre-processing are to delete any erroneous or meaningless data in order to ensure that the data is complete before inputting in the GIS program and to segment the data into groups of interest to the study.

Demand data includes population data and land data in Bangkok at both district and sub-district levels. The population data of BMA comes solely from the people who have registered their households in Bangkok, and does not take into account the many non-registered residents in the city [21]. The data of BMA is collected in sub-district level as it is registered in system. Therefore, for the purpose of this study, NSO data will be used as it includes the non-registered population in its report, and thus more accurately represents the population of Bangkok. However, as the latest data of NSO is from the year 2010, and NSO has announced that it would not conduct a survey in 2020 due to the COVID-19 pandemic [22], in this study, we used data from Worldpopulationreview.com [10], which also forecasts population data for 2020 by using NSO’s data of Bangkok, and already include non-registered populations. This data will complement the limitations of the BMA data that do not include non-registered populations. However, the data of Worldpopulationreview.com is presented in Bangkok city level. To allocate the population from Worldpopulationreview.com into sub-district level, we will allocate this number by using the proportion and ratio of the population in sub-district level from BMA data in 2020. The total population in Bangkok from Worldpopulationreview.com will be multiplied by population proportion in sub-district level from BMA and the total population will be the same as in Worldpopulationreview.com by having the proportion in sub-district level from BMA data.

For supply of stroke service, in addition to the aforementioned levels of hospitals in NHSO Region 13, we can still further classify hospital levels by the availability of stroke facilities including CT Scan/MRI, rt-PA (24-h and non-24 h services) and thrombectomy services, as seen in

Table 1. Number of hospitals disaggregated by the levels of NHSO Region 13 and the rt-PA facilities available.

	Hospital’s Facility for Stroke
Hospital Level By NHSO	A	B		C	D	Total
		B1	B2
	CT Scan/MRI	CT Scan/MRI	CT Scan/MRI	CT Scan/MRI
	rt-PA Service	rt-PA Service	rt-PA Service
	(24 h)	(24 h)	(Not 24 h)
	Thrombectomy
1			8	23	35	66
2		18				18
3	27					27
Total	27	18	8	23	35	111

. The availability of each facility reflects a hospital’s ability to serve stroke patients. Based on the fact that rt-PA is the standard of care for stroke patients, we divided the hospitals into 2 groups: hospitals with rt-PA services (Group A/B) and hospitals without rt-PA services (Group C/D).

However, in NHSO Region 13, there are some hospitals that have formed their own networks to transfer patients and knowledge and to build cooperation within the network. In the case of a stroke patient arriving at a network hospital without rt-PA, i.e., Group C/D, that hospital will transfer the patient to the main hospital in its network. In NHSO Region 13, there are 8 stroke networks with different numbers of hospitals in each network, as shown in

Table 2. Hospitals with stroke networks in NHSO Region 13 disaggregated by services provided.

Network		Hospital’s Facility for Stroke
Network No.	Main	A	B		C	D	Total
			B1	B2
		CT Scan/MRI	CT Scan/MRI	CT Scan/MRI	CT Scan/MRI
		rt-PA Service	rt-PA Service	rt-PA Service
		(24 h)	(24 h)	(Not 24 h)
		Thrombectomy
1	King Chulalongkorn Memorial Hospital	2	6	2	4	3	17
2	Phramongkutklao Hospital	2	1	0	0	0	3
3	Lerdsin Hospital	0	0	0	1	2	3
4	Prasat Neurological Institute	2	2	1	3	2	10
5	Bhumibol Adulyadej Hospital	2	2	1	1	2	8
6	Taksin Hospital	0	2	1	3	3	9
7	Siriraj Hospital	0	1	0	4	1	6
8	Faculty of Medicine Vajira Hospital	0	0	0	0	1	1
Total		8	14	5	16	14	57

.

In terms of means of transportation, traffic speed and stroke delivery models will be considered. As, in GIS, the route that the program chooses is the shortest time to travel, the different speed in each road will make the model more solid and practical. In this study, traffic speed data were retrieved from probe data provided by iTIC from the period of 2017–2019. The speed data are collected from specific locations throughout Thailand. This study will focus on 1817 location codes as shown in Figure 5. These data are from iTIC probes located in the Bangkok area only.

iTIC data is recorded every 10 min, starting from 00:00 to 23:50 every day, resulting in a total of 144 data sets per day. From iTIC’s data, we can create a scatter plot showing the traffic speed in each location code. We calculated the average speed while counting only the speed value between µ ± 2σ to prevent outliers. After calculating the average speed in each period, the centered moving average ± 3 periods was used to smooth out the average speed line. Upon comparing the data, especially in peak periods, traffic tended to decrease every year during the period of 2017–2019. To capture the most updated speed data in the delivery models, we opted to use the speed data from 2019.

From the iTIC probe data, there is a significant difference in traffic speed in Bangkok between working days and non-working days. In 2019, the average speed on working days was 33.5 km per hour compared with 39.8 km per hour on non-working days. This study aims to address a critical situation, so only speeds on working days were considered

From the traffic speed data on working days in 2019, as shown in Figure 6, we can see that the lowest speed on each day is during 18:00–18:30 with an average speed of 20.4 km per hour. Thus, to find out the service area in the most critical situation, we will use the speed of 1817 location codes during 18:00–18:30 on working days in 2019 to input into GIS. However, as Kerstin Petzäll et al. (2011) [23] stated that the speed of ambulances will be higher than that of normal cars, we used the data of Narenthorn EMS Center—a prevalent service provider under Rajavithi Hospital—during 2018 to determine the ambulance speed. The data is collected from ambulance’s GPS and inhouse system. It was found that, during the 18:00–18:30 time window, the speed of ambulances will be 30 km per hour which is 50% higher than the normal speed as per the data provided by iTIC. Therefore, in order to establish the tentative stroke service areas, we added only the 80th percentile of the incremental speed to each location code’s speed.

In the Stroke Service Plan of NHSO Region 13, there are stroke protocols for delivering a patient who exhibits stroke symptoms in accordance with varying times from the stroke onset scene to hospital arrival. Patients delivered to hospitals in Groups A/B tend to receive rt-PA treatment within 4.5 h. The prevalent delivery model starts from the departure from the stroke onset scene by EMS or other means of transportation until the arrival at hospitals in Group A/B or Group C/D. It must be noted that Group C/D hospitals would still need to further transfer patients to hospitals in Group A/B to receive rt-PA treatment.

Nevertheless, not all hospitals in Bangkok have the proper equipment to support stroke patients. Moreover, it may be the case that patients would not know which hospitals are fully equipped and ready for their treatment. Additionally, as stroke is a time-sensitive disease, the traffic in Bangkok plays a crucial role in whether patients can arrive within the golden period. This traffic can prolong the travel time to the hospital, especially during peak hours, which can result in a decreased percentage of recovery from stroke symptoms.

For a better recovery rate, the total time from onset until receiving treatment should be within 4.5 h. However, from previous research, the prehospital time which is the golden period for stroke patients would be within 1 h from stroke onset, as shown in Figure 7. The figure will show all stroke delivery models, but self-transportation will be excluded from this study.

To calculate the service areas in this study, we measured the EMS travel period from the same starting point to the same destination points. Zhang S. et al. (2018) [24] studied prehospital models with and without prehospital notification procedures (PNP) and found that there is a significant difference in terms of the duration in emergency departments (ED) for PNP and non-PNP models. Thus, our study also includes the duration in ED so that we can compare the models in the same situation. Tennyson J. et al. (2019) [25] also found that the time in ED or time from Door-to-Activation (DTA) is about 15 min. Thus, as previously mentioned, the golden prehospital time for stroke is 1 h from stroke onset. In this case, the onset until stroke activation time will be 75 min.

However, in terms of time consumption until stroke activation, the initial time spent in the delivery model will start when EMS providers receive a call from the patient and the subsequent delivery of the patient to the hospital. From the data of Narenthorn EMS Center during 2017–2018, the total time after receiving the call from a patient until the service provider is ready to depart from the stroke onset scene with the patient is 28.05 min during 18:00–19:00, which is the time period that is of interest to this study. Sangob B. et al. (2017) [26] also reported that, in the case that the patient has to be transferred to another hospital, Door-In to Door-Out (DIDO) time in the first hospital before transferring to the next hospital is 13.47 min. Zhang S. et al. [24] also stated that the ED duration for hospitals with PNP is 9.20 min, whereas for non-PNP hospitals, the ED process would take 16.1 min. As previously mentioned, the hospitals with stroke networks will develop PNPs with EMS providers, which will shorten the time in ED, while hospitals without networks would not have such procedures. Similar to previous studies, we filled all mentioned times in each node to find out the time allowance, as shown in Figure 8, and used GIS analysis to draw the stroke service areas in Bangkok. For this study, we measured the service areas for 111 hospitals without stroke networks, and 8 stroke hospital networks using EMS in the Bangkok area.

2.3. Analysis Techniques

To measure or analyze the accessibility to health care services or hospitals, GIS spatial analysis can be used to enhance the data analysis and visualize the results as graphical representations. There are several appropriate applications in the GIS analysis for various objectives. From the collected GIS spatial data, EMS providers can search for the location of the nearest hospital, but not the fastest route where traffic is concerned. However, Murad A. (2007) [27] studied the health service area of a private hospital in the city of Jeddah, Saudi Arabia, by using GIS functions, including network analysis and overlay analysis, in addition to extending the application to the health planning field

In this study, the hospital stroke service area will also be analyzed with spatial data by using ArcGIS program to analyze stroke network data and represent network requirements. The program can be used to analyze route planning, facility location, and network-related problems. ArcGIS will help find the quickest route from the input data provided. Cost–benefit analysis can be calculated from ArcGIS, but is not presented in this study. In this study, network analysis and overlay analysis will be applied.

Network analysis can solve questions related to linear networks. This spatial analysis technique uses network data to calculate distances between points or nodes in the network. Common applications are route finding, route planning, closest facility identification by travel time or distance, and calculation of service areas. In most GIS programs, the network analysis module consists of several modeling functions, including finding the shortest path, service area model, allocation model, location-allocation model, and spatial interaction model. The study has used the allocate and service area functions that are available within ArcGIS Network Analyst for evaluating the stroke service area. In this study, network analysis will define and analyze the spatial distribution of stroke service supply, develop stroke service accessibility models, and calculate hospital service areas.

Overlay analysis will calculate the overlapping service areas, apply the selected hospital to calculate the size of its served demands and compare the service areas between network and non-network hospitals. An overlay operation is much more than a simple merging of line work; all the attributes of the features taking part in the overlay are carried through, from the study, the service area of each network for both hospital group A/B and group C/D, which are overlaid to create a new polygon dataset. In this study, we will use overlay analysis to combine the input data of several datasets into one, finding specific service areas that have a combined value and match with the delivery model we specify. The result will be shown in the service area of each model and hospital in each sub-district level of Bangkok.

The result of spatial data in GIS is a tabulated area where all areas in the input that completely contained the same value are presented. The service area of each hospital will be tabulated and overlayed so that we know the total service area in Bangkok by model. In terms of calculating population, after finding the result from GIS, we use the proportion of service area covered in each sub-district, multiplied by the calculated population in each sub-district, to find the served population of each hospital or delivery model.

After we obtained the speed data of 1817 location codes in Bangkok and calculated the data as previously mentioned, we filled in the speed of each location code in GIS. However, as all location codes do not cover all streets and alleys in Bangkok, we used the “Near” function in GIS to allocate traffic in order to ensure that all streets had an allocated traffic speed. Network analysis was used to find the service areas for hospitals in both Group A/B and Group C/D. At the same time, we also used the closest facilities to find the hospital in Group A/B nearest to the Group C/D hospitals. After we determined the service area in all models of interest, we used the overlay tool to measure the entire service area and to calculate the size of the population served.

As seen in Figure 8, the service areas of hospitals in Group A/B, including those with and without stroke networks, were calculated, starting from onset time until activation in the emergency department, then deducting (1) the time spent from first contact until the EMS service departs from the onset scene, and (2) the time in the emergency department. The difference of times in the emergency department is derived from whether or not the hospital has a PNP in place. Non-network hospitals will not have any PNP, while hospitals with stroke networks will.

For both network and non-network hospitals in Group C/D, the service areas were also calculated from onset time until activation in the emergency department. The time spent from first contact until the EMS service departs from the onset scene and the time spent in the emergency department are still factors to deduct with the varying times spent in ED, which again depends on whether or not the hospital has a PNP in place. The additional transfer times, or DIDO, in Group C/D hospitals were considered in this delivery model. Moreover, the total delivery time must include 2 legs of travel after deducting the in-and-out time from a Group C/D hospital and the referral time as the 2nd leg of travel. The time of the 2nd leg of a non-network hospital in Group C/D will be calculated by using the closest facility to the nearest Group A/B hospital, while the time of the 2nd leg for a hospital with a stroke network will be the time between the Group C/D hospital and the main hospital in the same network. For both models, the remaining time after deductions will be the service area of the hospitals in Group C/D. If the time allowance is zero, this means that the hospital in Group C/D has no service area.

3. Results

3.1. GIS Mapping of Stroke Service Areas in Bangkok

After adding all data in GIS and comparing each model and time period, the service areas could be portrayed as GIS maps. Suggestions will also be shown using GIS.

3.1.1. EMS Service to Hospitals without Stroke Networks

Stroke Service Area of Hospitals in Group A/B

From the 53 hospitals in this group, the service area covers 1391 square kilometers, or 88.69% of the total area of Bangkok, as shown in Figure 9.

Stroke Service Area of Hospitals in Group C/D

From the 58 hospitals in this group, the service area covers 936 square kilometers, or 59.67% of the total area of Bangkok, as shown in Figure 10.

Overlayed Stroke Service Area of Hospitals in Bangkok

From the 111 hospitals without networks in NHSO Region 13, the service area covers 1392 square kilometers, or 88.72% of the total area of Bangkok, as shown in Figure 11.

3.1.2. EMS Service to Hospitals with Stroke Networks

Stroke Service Areas of Hospitals in Group A/B in Each Network

From the 35 hospitals in this group, the service area covers 1418 square kilometers, or 90.38% of the total area of Bangkok, as shown in Figure 12.

Stroke Service Areas between Group C/D Hospitals and the Main Hospital in Each Network

From the GIS analysis, as hospitals in Group C/D will refer patients to the main hospital in their respective networks, there are some networks where a hospital in Group C/D has no service area, or is out of the range of the main hospital. From the 30 hospitals in this group, the service area covers 427 square kilometers, or 27.20% of the total area of Bangkok, as shown in Figure 13.

Overlayed Stroke Service Area of 8 Stroke Networks

From the 65 hospitals with networks in NHSO Region 13, the service area covers 1418 square kilometers, or 90.38% of the total area of Bangkok, as shown in Figure 14.

3.1.3. Overlayed Stroke Service Area for Network and Non-Network Hospitals in Bangkok

From the 111 hospitals in the Stroke Service Plan of NHSO Region 13, the stroke service area covers 1418 square kilometers, or 90.42% of the total area of Bangkok, as shown in Figure 15.

3.1.4. Suggestions to Utilize Current Hospitals in NHSO Region 13

In order to increase the service area for NHSO Region 13, we can utilize the current members of the Stroke Service Plan of NHSO 13 as follows.

Change All Non-Network Hospitals in Group A/B to Network Hospitals

From the current 35 hospitals in Group A/B, the service area covers 1418 square kilometers, or 90.38% of the total area of Bangkok. If all non-network hospitals in Group A/B were to become network hospitals, the service area would still cover 1418 square kilometers, which is the same as in the current situation, as shown in Figure 16.

Change All Non-Network Hospitals in Group C/D to Network Hospitals with Patient Transferal to the Nearest Hospital in Group A/B

From the current 30 hospitals with networks in Group C/D, the service area covers 427 square kilometers, or 27.20% of the total area in Bangkok. However, if all 58 non-network hospitals in Group C/D can be improved to become network hospitals with stroke patients being transferred to the nearest hospital in Group A/B, the service area would cover 1245 square kilometers, or 79.37% of the total area of Bangkok, as shown in Figure 17.

3.2. GIS Data Analysis

After GIS data has been visualized as a map, descriptive statistics can help to describe, display, and summarize the data in a meaningful way. Descriptive statistics are used in this research to summarize the GIS analysis including data of the serviced areas and the population served. The results of the descriptive statistics show the areas that the services cover (square kilometer) and the population served (million people). In

Table 3. GIS results by model.

Model		Type of Result
		By Area		By Population
		km2	%	Million	%
Non-Network	A/B	1391	88.69%	9.90	93.96%
	C/D	936	59.67%	8.47	80.40%
	All	1392	88.72%	9.91	93.99%
Network	Main/A/B	1418	90.38%	9.92	94.13%
	C/D	427	27.20%	5.15	48.89%
	All	1418	90.38%	9.92	94.14%
Combine	All	1418	90.42%	9.92	94.17%

, the results show that the service area of hospitals in Group A/B is significantly higher than that of Group C/D when calculated from the time spent for patients who go to Group C/D hospitals, taking into account the DIDO time and the two required travel periods. Regarding the area and population, the percentage of the population served in all models is higher than the percentage of the serviced area. This means that the service areas are in areas with high density populations.

When considering the service area in each sub-district, in

Table 4. Number of sub-districts in range of service areas covered.

% of Service Area Covered	Non-Network				Network				All
	No. of Hospital			%	No. of Hospital			%	No. of Hospital	%
	A/B	C/D	All	All	A/B	C/D	All	All
0	0	8	0	0%	0	37	0	0%	0	0%
1–20%	1	4	1	1%	0	20	0	0%	0	0%
21–40%	1	10	1	1%	2	5	2	1%	2	1%
41–60%	1	8	1	1%	0	4	0	0%	0	0%
61–80%	4	15	4	2%	3	8	3	2%	3	2%
81–100%	173	135	173	96%	175	106	175	97%	175	97%
Total	180	180	180	100%	180	180	180	100%	180	100%

, most sub-districts have stroke service areas which cover about 81–100%, accounting for 97% of Bangkok when combining both non-network and network hospitals’ service areas altogether. Moreover, there are no sub-districts that have no service area at all. However, there are two sub-districts that have the lowest percentage of service areas, which is equal to 21–40% of the total area in each sub-district.

After overlaying the service areas and comparing the service density by using the number of hospitals that have service areas in each sub-district, we can see the result in the highest proportion that, in the non-network model, 41–60 hospitals cover 110 sub-districts, or 61% of the total number of sub-districts, while in the network model, 21–40 hospitals cover 146 sub-districts, or 81% of the total number of sub-districts. When both models are combined, 41–60 hospitals cover 97 sub-districts, or 54% of the total number sub-districts, as seen in

Table 5. Number of hospitals covered in each sub-district.

No. of Hospital	Non-Network				Network				All
	No. of Sub-District			%	No. of Sub-District			%	No. of Sub-District	%
	A/B	C/D	All	All	A/B	C/D	All	All
0	0	8	0	0%	0	37	0	0%	0	0%
1–20	33	172	29	16%	37	143	33	18%	23	13%
21–40	101	0	39	22%	143	0	146	81%	32	18%
41–60	46	0	110	61%	0	0	1	1%	97	54%
61–80	0	0	2	1%	0	0	0	0%	28	16%
81–100	0	0	0	0%	0	0	0	0%	0	0%
>100	0	0	0	0%	0	0	0	0%	0	0%
Total	180	180	180	100%	180	180	180	100%	180	100%

. However, there are still hospitals in Group C/D that have zero service area in both network and non-network models.

We then compared the service areas of non-network and network models. When comparing only the network hospitals with each other, we can see in

Table 6. Area comparison by non-network and network models.

No. of Hospital	Service Area (km2)
	Hospital Group A/B				Hospital Group C/D
	No. of Hospital	Non Network	Network	% Change	No. of Hospital	Non Network	Network	% Change
1	11	4573	6685	46%	7	257	254	−1%
2	4	1731	2511	45%	0	0	0
3	1	338	482	43%	3	142	0	−100%
4	6	3007	4117	37%	5	429	93	−78%
5	6	1798	2674	49%	3	316	296	−6%
6	4	1675	2104	26%	6	284	212	−25%
7	2	873	1099	26%	5	364	347	−5%
8	1	440	570	30%	1	47	100	110%
Total	35	14,435	20,242	40%	30	1841	1303	−29%

that for Group A/B hospitals, the service area is increased significantly—about 40%—as there are higher times in emergency rooms. However, when comparing the total time allowance for hospitals in Group C/D, there is a significant decrease of about 29% as any patients would need to be transferred to the main hospital, which is sometimes not in the range of the service area.

4. Discussion

In stroke treatment, time is brain. Thus, time is a main consideration for the treatment process, as the objective is to increase the recovery rate of the stroke patient. For this purpose, the time until the patient can receive rt-PA should be shortened as much as possible. In terms of prehospital time, from this study, when hospitals that can administer rt-PA, i.e., Group A/B, can be changed to network hospitals, the service area of the individual hospital itself is increased, but it will not increase the total aggregate service area as a whole, as the service areas are overlapping. The benefit for changing hospitals in group A/B from non-network to network hospitals will reflect in the increased service area of the nearest hospital in Group C/D that is within its network. For Group C/D hospitals, changing non-network hospitals to network hospitals would be beneficial when they refer a patient to the nearest hospital, not the main hospital in its network. In this case, hospitals without rt-PA, i.e., Group C/D, should consider re-networking to connect with the nearest hospital within range, or develop a model to deliver the patient to another hospital in the same network in order to shorten the time until rt-PA administration. In terms of policy, the Stroke Service Plan of NHSO Region 13 should include other hospitals that have yet to join the Plan, especially in the areas that are still not covered in the Bangkok area. The Stroke Service Plan must also promote and support the hospitals with rt-PA to join the network as, as the GIS data revealed, it would take only 35 out of the 53 hospitals in Group A/B to be able to increase the service area by 40% or almost 6000 square kilometers. For Group C/D, the Stroke Service Plan should reconsider and address the issue that hospitals in each network should be in the range of service so that no hospitals in Group C/D would have zero service coverage. The density of service is another factor that should be considered in the Stroke Service Plan as, from the GIS results, in some sub-districts, there is an intense supply of service, while there is a supply shortage in others. It is practical to develop the delivery model and utilize the current hospitals that are still not included in the Service Plan. Therefore, NHSO Region 13 still needs to invite other hospitals in Bangkok to join the Plan as soon as possible. Moreover, the public relation to develop the usage of EMS service by calling 1669 should be promoted to increase the stroke service area.

In this study, we considered only the demand and supply in the Bangkok area. We can see from the results that most of the unserved areas are along the borderline of Bangkok, which means that the patients in these areas tend to encounter difficulties when experiencing stroke symptoms that necessitate travelling to a hospital within a stroke network. In reality, such patients can choose to go to hospitals in adjacent provinces, which would be closer than going back to access services in Bangkok. However, most people still prefer to travel to Bangkok to receive health care services, due to the hospitals’ reputation and readiness. However, people in the vicinity of Bangkok also want to receive services in the Bangkok area. Future research may consider studying the areas in adjacent provinces in order to determine a more comprehensive picture of services. However, we would still need to consider the service providers and delivery models, as the additional hospitals to be studied would fall under different NHSO regions.

The result of this study is shown in served and unserved areas which is in a dichotomous manner. We cannot imply that if that area is already a stroke service area, a patient in that area will be safe. The result should be considered with other factors to strengthen any policy for stroke. This will help minimize any potential negative impacts.

Today, stroke response is considered increasingly important. There are many new delivery models to reduce the time to receive rt-PA in order to increase the recovery rate for the patient. For example, there are ambulances equipped to deliver prehospital thrombolysis, and mobile stroke units in which CT scans can be performed. However, these models are still not prevalent and are limited due to high costs.

Further studies on stroke service areas should include new delivery models and more connected networks in case there is a significant volume or other factors that would impact the area of interest. Moreover, to avoid the uncertainty of the factors in the model, alternative approaches that give the result beyond a dichotomous manner can be considered. Consideration of EMS network and location of ambulance station can strengthen the study. Leira et al. (2006) [28] studied the aerial interhospital transport of patients by helicopter. This approach could improve current clinical trial recruitment in rural areas. Thus, the application of aerial mean might be a new tool that can be adapted in such an area that has heavy traffic as Bangkok. In order to build on the research to establish a more solid service area, updated times in each node of the model, as in Figure 8, and up-to-date data from new studies, especially in Bangkok area, may be used to more accurately determine the area serviced in Bangkok circumstances. Considering traffic is a factor for stroke service areas, updated speed data will strengthen the delivery models. In addition, new tools that can predict traffic more accurately and in a more timely manner should be included in new studies. Additional stroke delivery models that include self-transportation, mobile stroke units, and aerial interhospital transport can be studied to compare with the EMS model of transportation, which can help NHSO to design its service plan policies and procedures. In this study, we use open data only for speed data which we consider important for Bangkok, the city with traffic jams. However, there might be other open data or big data that relate to the service area for health, such as the hospital supply and usage rate for health facility. Further study depends on the data sharing in each location. Cost–benefit analysis in each delivery model can be an important factor to consider the model which can be added in next study. Moreover, sharing of data on the number of hospitals with facilities for stroke treatment is key to expanding service areas in a way that requires lower investment. Moreover, regularly reviewing the service areas for both non-network and network hospitals is crucial to increase health care services and support the people in the areas served. Further research in suburban or rural areas would reveal different results due to the fewer numbers of well-equipped hospitals for stroke and different delivery models.

5. Conclusions

This paper identifies stroke service areas by utilizing data about the service demand of the population, different stroke delivery models using EMS as the means of transportation, road traffic speed, and rt-PA service supply. The study considered the time spent in each node along the chain and service areas were drawn for both non-network and network hospital models in order to conduct comparative analysis. The results from this study show the current service area by non-network and network in each hospital type. The service area for a hospital equipped with rt-PA is larger than a hospital without rt-PA. The stroke hospital network and the process to refer patient between hospitals should be re-arranged to increase the service area of Bangkok as a whole. Hospitals without rt-PA should refer patients to nearest hospital with rt-PA, not the main hospital in its network. Data and facility sharing between hospitals are keys to increase stroke service area.