Accessibility and distribution of healthcare networks are critical for national healthcare, particularly for the elderly population who are at a higher risk of developing multiple chronic illnesses.
Since the restructuring of the Regional Health System (RHS) in 2017, Singapore has implemented health policies by creating an integrated network of acute hospitals, community and primary care providers within a certain geographic region, aiming to provide comprehensive care and a care continuum for patients [4].
However, little emphasis has been given to assessing the accessibility of healthcare services from a geospatial analytical perspective. This research article aims to analyse the geographical accessibility, kernel density clusters and distribution networks of major healthcare services in Singapore, which includes acute and community hospitals, primary care networks (PCN), eldercare and nursing homes, and General Practitioners (GPs) services under the Community Health Assist Scheme (CHAS) for older persons.
With an ageing population and a growing burden of chronic diseases, care delivery in Singapore is becoming increasingly complex [10]. The country’s health system is shifting its focus beyond hospitals to the community with an emphasis on health promotion at all stages of life [7]. This approach is highlighted under the Ministry of Health’s (MOH) strategic planning in preventive care to ensure long-term sustainability, reducing expenditure while providing quality and affordable healthcare for people of all ages.
Within Singapore’s healthcare system, it consists of three Regional Health System (RHS) providers, namely SingHealth, National Healthcare Group (NHG), and National University Health System (NUHS). To be future-ready, RHS providers are increasingly looking into areas of health informatics, geographical and geospatial information to improve health service delivery outcomes, reduce cost, and improve continuity of care within the community [6].
We obtained address information and postal codes of publicly available data on Acute and Community Hospitals, Eldercare Facilities and Nursing Homes, Primary Care Networks (PCN), and General Practitioners (GPs) services under the Community Health Assist Scheme (CHAS) from relevant governmental agencies and other websites.
| Data | Source | Additional Actions Required |
|---|---|---|
| Government Restructured Hospitals | HealthHub | Conversion to SHP Files |
| Community and Rehabilitation Hospitals | HealthHub | Conversion to SHP Files |
| Polyclinics | Professor Kam Tin Seong | |
| Nursing Homes | Professor Kam Tin Seong | |
| Eldercare Facilities | Data.gov.sg | |
| Primary Care Networks Clinics | Agency for Integrated Care (AIC) | Conversion to SHP Files |
| Community Health Assist Scheme Clinics | Agency for Integrated Care (AIC) & Community Health Assist Scheme (CHAS) | |
| Master Plan 2014 Planning Area Boundary (Web) | Data.gov.sg | |
| Master Plan 2019 Subzone Boundary (No Sea) | Data.gov.sg | |
| Resident Population by Planning Area/Subzone, Ethnic Group and Sex, 2015 | Data.gov.sg | |
| Roads Networks in Singapore | Open Street Map | |
| Residential Buildings in Singapore | Open Street Map |
We use R studio to perform geocoding of the addresses and postal codes using SLA API. After which, the data is transformed and converted into the geospatial data point of locations of various healthcare facilities into easily accessible shapefiles for usage in our R programming for geospatial analysis methods. Data scraping from web datatables are also conducted. The data scraped then required manual data cleaning for incorrectly reflected information as certain webpages did not provide correct/updated information.
We used mainly two geospatial methods in our analysis of healthcare service in Singapore.
First method is Kernel Density Estimation (KDE). It allows the evaluation of the clustering patterns and distribution of healthcare facilities in Singapore.
Second method is geographical modelling of accessibility. Visualisation using Hansen, KD2SFCA, SAM and 2SFCA Methods to measure the accessibility of healthcare services in Singapore, and will be analysed in this paper.
It is crucial to assess the healthcare system’s distribution and accessibility in Singapore to ensure that all residents have access to high-quality medical care. This application aims to do just that by providing valuable insights into the distribution of primary and tertiary healthcare facilities in Singapore.
By using geospatial analytics, visualisation and mapping techniques, this application will identify potential gaps or disparities in healthcare accessibility and distribution. We hope that the insights gained from this analysis will help policymakers and healthcare providers to make data-driven decisions to improve the planning and distribution of healthcare services in Singapore.
The map models generated by this application will identify areas with high population densities but low healthcare facility densities, allowing policymakers to allocate resources and invest in healthcare infrastructure in these underserved areas. With these insights, policymakers can ensure that all residents have equal access to high-quality medical care, regardless of their location or socioeconomic status.
This application’s visualisation and mapping capabilities will provide a detailed understanding of the healthcare system in Singapore, making it easier for policymakers and healthcare providers to identify areas for improvement. By utilising data-driven insights, Singapore can continue to be a leader in healthcare and ensure that all residents receive the best possible medical care.
In summary, this application’s persuasive and dominant motivation is to identify potential gaps and disparities in healthcare accessibility and distribution in Singapore, provide valuable insights to policymakers and healthcare providers, and ultimately ensure that all residents have equal access to high-quality medical care.
To acquire accessibility, we need to obtain the road networks in Singapore. This data can be found from Open Street Map, which includes residential walkways and roads used by traffic. Hence we had to perform filtering methods to filter out roads used by vehicle transportation and small roads within living estates of Singapore residential areas.
In addition, we need to derive Origin-Destination (OD) Matrix for each type of medical facility. The process was time consuming but necessary for further data processing.
In order to accomplish this, we utilise QGIS Network Analysis Toolbox 3 (QNEAT3) Network Analysis for OD Matrix Computation between the hospital facilities and road networks points.
Similarly using QGIS, we obtained the public and private residential buildings in Singapore, which is then saved as a shapefile
In this paper, we will be using the URA Master Plan Subzone map provided by Urban Redevelopment Authority (URA) as our reference point for the description of geographical locations within Singapore for easier understanding.
In Figure 4, we had applied Real-Centroid to extract out the centroids of all livable buildings in Singapore in QGIS.
This will help us mathematically estimate and visually understand the demand within different geographical regions based on clusters of points of residents buildings within each region. This further enables us to learn the geographical accessibility of various regions within Singapore. This would be effective in interpreting how the density/cluster sof KDE should be like for each given region.
Looking at Figure 4, we can derive that the largest proportion of citizens live in the Central, followed by Northeast, Southeast and southwest, due to the large clusters of data points overlapping each another on the map. A small group of people is also noted to live close to the north-west region of Singapore. This will help us get a general understanding of the demand and supply, distribution of health service and accessibility of healthcare within certain regions.
We would like to highlight that regions covered by forests have understandably no healthcare facility. As a result, it is important to take this into account when analysing our data for accessibility and distribution of healthcare services.
We analyse the accessibility of eldercare facilities which accHansen values fall within 0 (lower limit) to 1100 (upper limit). We also want to highlight that the ranges for accHansen may vary depending on the healthcare facilities dataset being used.
An efficient way to understand the data is to view the colour distribution. Dark shades represent higher accessibility of the healthcare facilities.
In our analysis of eldercare facilities, we will use the following measures:
From the figure, we can observe that accessibility of eldercare facilities is largely within the Central region of Singapore. However, the Northeast, Southeast and Southwest regions are generally largely accessible for residents as well.
At the top of the Northwest region, there is a lack of eldercare facilities, in which health policy makers could consider building more eldercare facilities in future national budgeting and planning. But the visualisation of the model suggests that eldercare are generally well distributed across Singapore.
In our analysis of PCN clinics, we will use the following measures:
We will also be understanding the accessibility through the colour distribution. Dark shades represent higher accessibility of the healthcare facilities.
We can observe that accessibility of PCN facilities are highly accessible within the Central, Northeast and Southeast regions of Singapore, with reference to the darker regions. Additionally, the Northwest region has acceptable range within the moderate accessibility of PCN clinics for its residents.
In our analysis of CHAS clinics, we will use the following measures:
Similarly, we will be using the colour distribution as a way to access the graph.
From the figure above, it can be observed that accessibility of GP Clinics under CHAS are concentrated in the Central, Northeast, and Southeast regions of Singapore, as indicated by the darker regions. The Northwest region has also acceptable range within the moderate accessibility of CHAS Clinics for its residents.
In our analysis of hospitals, we will use the following measures:
Likewise, the colour distribution will be used for analysis.
From the figure, it can be observed that accessibility of Government Restructured Hospitals are largely accessible within the Central, Northeast, Southwest regions of Singapore, as indicated by the darker colours within these areas. The Northwest region is in an acceptable range.
Upcoming Hospital
The Southeast regions have notably less Government Restructured Hospitals, suggesting residents living near and within the Marine Parade region may be under-served.
MOH has announced that they will be building an Eastern Integrated Health Campus. Therefore, we believe that the accessibility for residents living in the Southeast region will improve once it is built [5].
In our analysis of polyclinics, we will use the following measures:
Darker colours indicate higher accessibility and light colour represent lower accessibility to healthcare facilities within a region.
From the diagram, it can be observed that accessibility of polyclinics is largely accessible within the Central, Northeast, Southwest and East regions of Singapore, as indicated by the dark colours within these areas.
Upcoming Hospital
Northwest has relatively low accessibility. However, the construction of Woodlands Health Campus has been announced by MOH [5]. Therefore, it can be expected that the accessibility for residents living in the Northwest region will improve with the completion of the institution.
In our analysis of nursing homes, we will use the following measures:
As mentioned, darker colours indicate higher accessibility and light colour represent lower accessibility to healthcare facilities within a region.
From the diagram, it can be observed that accessibility of nursing homes is largely accessible within all regions of Singapore, as indicated by the darker shades within the region.
Within this paper, we will be only interpreting the accessibility for Hansen method, understanding that there are also other methods namely KD2SFCA, SAM and 2SFCA. However, their principles are relatively similar to the Hansen methods of accessibility. Despite not elaborating in this paper, we have included them in our RShiny application [1].
We applied Kernel Density Estimation (KDE) to estimate the distribution of healthcare clusters and to understand the distribution and allocation of healthcare facilities within each geographical location. Within our analysis in this paper, KDE is performed using Cross Validated Bandwidth Selection (bw.diggle) and Gaussian kernel.
Cross Validated Bandwidth Selection, as its name implied, uses cross-validation to select a smoothing bandwidth that aid in KDE.
The Gaussian kernel estimates that points are closer to the allocated centre point have greater weight as compare to data point which are further from the middle point
Before proceeding, we would like to highlight KDE values fall within 0 (lower limit) to a higher number, depending on the dataset being used in computation. As such, KDE values may fall below 1, if the facilities are sparse across the region.
Higher KDE values indicate that there is a higher density or clustering of healthcare facilities within the region. Similar to the interpretation of graphs for aalysing accessibility, we can also determine based on the colour.
Yellow - Indicating High Density / Clustering
Pink - Indicating Medium Density / Clustering
Dark Blue - Indicating Low Density / Clustering
In interpreting the clustering of hospitals, we cannot confidently determine how many hospitals are within a certain area as the number of hospitals is relatively small and sparse. This is understandably due to their purpose as being the last stop for medical help, as noted in the planning of Singapore’s healthcare system. Hence, we are primarily relying on visual information to interpret the data as we cannot meaningful classify the KDE values lower than 1.
Based on Figure 13, we can observe that the distribution of Government Restructured Hospitals shows a higher density / cluster of hospitals located within the Central region. Since the Central region has a very high population density based on the number of residential buildings within an area compared to other regions, one can assume that hospitals will cluster within the region.
The information highlights that there is an estimated KDE value to be less than 1 within a given area, as indicated by the low occurrence of yellow hotspots. This suggests that the distribution of hospitals is distributed in a manner that within a given area, a hospital will serve multiple planning subzones. This can potentally explain the long waiting time at hospitals due to the population size each hospital is serving.
However, based on the visualisation model, the other regions do not exhibit a similarly high density of clusters of hospitals. Thus it could mean demand for health service within the region may be relatively lower, even when the population size is considerably high based on the number of residential buildings. Or, healthcare services within the region are more efficient and do not require more hospitals within the same area.
Likewise for the interpretation of clustering of polyclinics, we cannot fully determine the number of hospitals within a given area as small number of polyclinics in Singapore. Hence we will too be relying on visual information to interpret the data.
The graph depicts that a rather even spread of yellow hotspots, suggesting that the distribution of polyclinics is generally good, with comparison with areas of higher population density denoted by residential buildings.
Based on Figure 14, we can observe that the distribution of polyclinics shows a higher density / cluster of hospitals located within areas where there are residential buildings. If we observe the KDE model and density values, it indicates that there is a good distribution of yellow and pink gradients across areas of residential buildings. It suggests that MOH has actually carried out effective planning in distributing polyclinics around Singapore.
We can discover that the distribution of nursing homes shows a higher density value / cluster of hospitals located within the Central region and Northeast region. This could possibly indicate there is higher demand for nursing homes within this area. However, due to Singapore’s small landsize and public transportation, we cannot dismiss that nursing homes could be located in these regions due to historical reasons, rent prices, availability of land and/or other reasons, especially when considering that users of nursing homes are typically elderly living in these facilities for long period of time.
Based on visualisation of nursing homes’ locations, the other regions do not exhibit a similarly high density value / clusters of nursing homes. But if we observe the geographical distribution for elderly aged 65 and above, it coincides with the area with higher values KDE distribution of nursing homes.
Else, it could mean demand for health services within these regions are lower even when the population size is considerably high based on the number of residential buildings or that there are less elderly within these planning areas.
Based on Figure 17, we can observe that the distribution of PCN clinics shows an average cluster size / density value of 20-30 within certain regions, this means that within a given area there are approximately 20-30 PCN Clinics. It also suggests that the PCN Clinics are relatively well distributed in Singapore, with reference to the residential buildings that citizens live in.
Based on Figure 17, the distribution of eldercare facilities can be observed to depict an average cluster size / density value of around 1 to 3 within certain regions. It is uncertain the most suitable number of eldercare facilities within a given area, as it is also determined by the size of the facilities. Assuming that 1 to 3 eldercare service per region is acceptable at the moment, we can assume that this number will rise as the elderly population increases.
In conslusion, our observations indicate that the accessibility and spatial distribution of healthcare service in Singapore are generally well-distributed. Most areas with higher populations generally have better accessibility and distribution of healthcare services such as PCN clinics, CHAS clinics, Government Restructured and Community Hospitals, and eldercare services.
Though there are only a couple of exceptions of regions which healthcare is less accessible, we urge relevant institutions like MOH to continue with their expansion of healthcare facilities in these under-served areas. This belief is supported by the fact that public housings are increasingly built in regions outside of central area (e.g., Punggol, Jurong East, Woodlands), and growing population, especially among elderly.
In Neutens [2015], it gives a detailed overview of accessibility, fairness, and healthcare, highlighting the importance of research from different fields [11]. This study emphasises the need for equal access and distribution in healthcare systems and explores various methods and challenges related to evaluating healthcare accessibility. While this review provides helpful information and forms a solid base for understanding healthcare accessibility, there are several key differences and similarities between Neutens [2015] and our project.
Focus on healthcare accessibility: Both Neutens [2015] and our project mainly aim to evaluate healthcare services’ accessibility and distribution. We both understand the importance of these factors in creating fair healthcare systems and efficiently using medical resources.
Geographical focus for Neutens [2015] provides a wider, global view of healthcare accessibility issues, while our project specifically targets Singapore’s healthcare system. This narrower focus helps us to analyse Singapore’s healthcare infrastructure and its unique challenges more closely.
Neutens (2015) talks about general consequences of healthcare accessibility research for policy and planning. In contrast, our project specifically aims to offer insights and suggestions for Singapore’s policymakers and healthcare providers. By focusing on one country, our project can give more targeted advice for improving healthcare distribution and planning capbilities within Singapore.
As our classmates had similar projects, we spoke about the difficulty in procuring and cleaning the data, and putting the app up on RShiny. We did get good responses from our classmates, whose comments suggest that they too feel that this project is of use to a government who wants to address the problem of accessibility to healthcare services.
To expand on our work, researchers will benefits from including relevant data sets that would take into account various factors such as demand and supply of healthcare service, population size and residential information within each planning subzone, age range, and other socioeconomic factors to understand healthcare access better and people from various demographics groups.
Providing information to these healthcare providers and policymakers so that they are able to utilise data to improve health services and to increase adoption of GIS analytics in health informatics in decision making.
Providing a comparison with other countries to understand how Singapore’s healthcare system fares among other countries so as to compare the accessibility and distribution to other countries and regions. This way, we can effectively compare our healthcare system with countries that have a high reputation for their healthcare service such as Denmark, Norway and Switzerland.
To check on the accuracy of our results, comparing them to real-life data, like how patients actually travel to respective healthcare facilities that are being located, is needed. This assures that our findings are accurate to real world conditions
By making these changes, we think our group’s work will give a more complete and useful look at the accessibility and distribution of healthcare facilities in Singapore, helping everyone from policymakers to the general public.
The authors would like to express our gratitude to Professor Kam Tin Seong from SMU for his patience and guidance in advising us on this project. Moreover he had also provided us with valuable healthcare datasets for us to perform our geospatial analysis.
[1] Group 9. Dr Go Where. Retrieved April 17, 2023 from https://is415-group9.shinyapps.io/IS415-Group9/
[2] Alex Jingwei He and Vivien F. Y. Tang. 2021. Integration of health services for the elderly in Asia: A scoping review of Hong Kong, Singapore, Malaysia, Indonesia. Health Policy 125, 3 (March 2021), 351–362. DOI:https://doi.org/10.1016/j.healthpol.2020.12.020
[3] Tin Seong KAM, Ketan BARSHIKAR, and Shaun Jun Hua TAN. 2012. DIVAD: A Dynamic and Interactive Visual Analytical Dashboard for Exploring and Analyzing Transport Data. International Journal of Computer, Electrical, Automation, Control and Information Engineering 6, 11 (November 2012), 834–1353. DOI:https://doi.org/10.5281/zenodo.1081041
[4] Ministry of Health. REORGANISATION OF HEALTHCARE SYSTEM INTO THREE INTEGRATED CLUSTERS TO BETTER MEET FUTURE HEALTHCARE NEEDS. Ministry of Health. Retrieved April 16, 2023 from https://www.moh.gov.sg/news-highlights/details/reorganisation-of-healthcare-system-into-three-integrated-clusters-to-better-meet-future-healthcare-needs
[5] Ministry of Health. Upcoming and Completed Healthcare Facilities. Ministry of Health. Retrieved April 17, 2023 from https://www.moh.gov.sg/upcoming-and-completed-healthcare-facilities
[6] National Healthcare Group. RHS for the Central Region. National Healthcare Group. Retrieved April 16, 2023 from https://corp.nhg.com.sg/RHSys/Pages/RHS-for-the-Central-Region.aspx
[7] Milawaty Nurjono, Pami Shrestha, Ian Yi Han Ang, Farah Shiraz, Ke Xin Eh, Sue-Anne Ee Shiow Toh, and Hubertus Johannes Maria Vrijhoef. 2020. Shifting care from hospital to community, a strategy to integrate care in Singapore: process evaluation of implementation fidelity. BMC Health Services Research 20, 1 (May 2020), 452. DOI:https://doi.org/10.1186/s12913-020-05263-w
[8] Ritu Sharma and Aineena Hani. 2021. Transformational Shifts in Singapore’s Healthcare Industry - OpenGov Asia. Open Gov. Retrieved April 16, 2023 from https://opengovasia.com/transformational-shifts-in-singapores-healthcare-industry/
[9] Singapore General Hospital. 2018. Cared for, supported by many helping hands. Singapore General Hospital. Retrieved April 16, 2023 from http://www.sgh.com.sg:80/news/regional-health-system/cared-for-supported-by-many-helping-hands
[10] Chin Soo Fang. 2022. S’pore’s population ageing rapidly: Nearly 1 in 5 citizens is 65 years and older. The Straits Times. Retrieved April 16, 2023 from https://www.straitstimes.com/singapore/singapores-population-ageing-rapidly-184-of-citizens-are-65-years-and-older
[11] Tijs Neutens. 2015. Accessibility, equity and health care: review and research directions for transport geographers. Journal of Transport Geography 43, (February 2015), 14–27. DOI:https://doi.org/10.1016/j.jtrangeo.2014.12.006