Health care accessibility and second homes: A spatial analysis in South-East Norway

(1)

0

Health care accessibility and second homes: A spatial analysis in South-East Norway

Olav Alexander Rønning

Master's Program in Human Geography with specialization in Geographical Information Systems (GIS)

Department of Geography Spring term 2020

Umeå University

(2)

1

1. Introduction

While the government in Norway strives for equity in health and access, factors of urbanization, modernization, and sustainable development may discourage advancement in rural municipalities. In the rural hinterlands, often where the mountain belt resides, this is known to be caused by declining employment-rates in typical rural industries like agriculture and forestry.

A consequence of this has been outmigration and lower numbers in the permanent population in most rural municipalities (Rye et al., 2011). Simultaneously, there is an ongoing trend of more second homes, particularly in rural areas, ascribed to second-home owners living in the cities. More second homes are evident from the country's growth of second homes on par with residential buildings since the 1970s and the existing second-home agglomerations around the major cities (Arnesen et al., 2011). The increased technical standard in second homes, from traditional cottages to high standard recreational homes with electricity- and water-utilities has also increased the year-round frequency of use (ibid). And while the political narrative is urbanization, this increased frequency of use in second homes may support a record of significant rural redistribution with second homes and domestic migration at its core (Ellingsen, 2017). A potential consequence of rural redistribution is skewed supply-demand ratios to critical public services such as hospitals, casualty clinics, and ambulance stations.

The health care system in Norway is a cooperative interaction between the primary- and specialist health care services. Still, it is the ambulatory services (an extension of the hospitals) that make the foundation of the emergency medical service. Response time has traditionally been a key quality indicator for ambulance services, both nationally and internationally, and is used in Norway to assess quality in the emergency health services (Helsenorge, 2018). Cardiac arrest and stroke are just two of the many conditions that require immediate response, where the quickest possible arrangement from the ambulance service can be crucial for the patient's outcome (ibid). As of 2020, response time in ambulatory services in Norway is non-enacted, but there are specific guidelines for how fast the prehospital services should arrive on the scene.

In densely populated areas, these guidelines are a maximum 12 minutes response time in 90%

of cases, with 25 minutes response time 90% of the time in rural areas. A 'densely populated area' was defined as a municipality with 10 000 to 15 000 permanent residents (Helsenorge, 2018). Despite these response times being non-enacted, regional health authorities must serve people in line with the current legislation, meaning a statutory duty to provide quick responses and qualified personnel (Akuttmedisinforskriften, 2015).

(4)

3 Similarly, the Norwegian Research Centre (NORCE) suggested that drivetimes to casualty clinics should be 40 minutes in 90% of cases. Here, statutory drivetimes could be valuable because there is a significant drop in the use of consultations, home visits, and ambulance call- outs per inhabitant as distances increase from casualty clinics (Raknes et al., 2014). They also comprise a source of emergency medical aid when they are nearer than an ambulance station.

The rural redistribution and its effects became more apparent during the COVID-19 pandemic, primarily because the government prohibited the population from using second homes. The government issued the prohibition on the 19^th of March 2020. It lasted until the 20^th of April, which forbid recreational commuting for residents who owned a second home outside of their residential home-municipality. This ban was a maximum precautionary strategy in that potential contamination would happen more frequently in rural communities where the health care system is only capable of handling the permanent residents.

A motivator was that more people spend time in their second homes, strictly linked to high- standard recreational dwellings. According to Statistics Norway, approximately 50% of the Norwegian population state that they have access to a second home (Norway, 2020). Further, 57% of the second-home owners were 60 years or older in 2017 compared to the 22% they made out of the total population (Norway, 2017). Lastly, distance and how quickly it is bridged has, for a long time, been a determinant of access. Measuring distances to second homes with geodata may gain perspective on potentially problematic areas associated with the redistribution and is described to aid health-instances to take quality indicators, such as response time, under revision (Helsenorge, 2018)

In a contemporary debate regarding the enactment of response times in the prehospital services, it has come forward that none of the Norwegian counties fulfilled the aforementioned recommended response times from ambulatory services to residential homes. Also that the resources committed in prehospital services are dependent on the regional health authorities' economy (Helsenorge, 2018). Second homes should arguably be included in this debate as 26 municipalities contain a more significant temporary population than permanent residents.

Arguably, the increase in technical standard and frequency of use has deemed second homes more part of the term 'home', with ~437 000 registered second homes to 5,4 million citizens in 2020. This paper argues for the inclusion of second homes into rural studies and investigates the drivetimes from second homes to casualty clinics, ambulatory stations, and somatic hospitals.

(5)

4

2. Aim

This study aims to calculate travel time in minutes to somatic hospitals, casualty clinics, and ambulance stations from second homes in the South-East Norway Regional Health Authority (see Figure 1). Furthermore, identify regions that break suggested national guidelines in travel time made by the Norwegian Research Centre and the Norwegian Emergency Committee submitted to the Ministry of Health and Care. The proposed guidelines are as follows:

i. NORCE suggested that drivetimes to casualty clinics should be 40 minutes and 60 minutes for the 90^th and 95^th percentile.

ii. The Emergency Committee suggested that response times from ambulance stations should be maximum of 12 minutes in 90% of cases in densely populated areas and 25 minutes in 90% percent of incidents in rural areas.

The research questions are as follows:

• How many second homes hold residence outside of the suggested guidelines for ambulatory services and casualty clinics?

• Can we identify peripheral areas to ambulance stations and casualty clinics?

2.2 Definitions and limitations

Travel time and drivetime is used interchangeably throughout this thesis and means the time it takes to traverse a road using a vehicle from point A to point B. Response time differs from travel time as it refers to the time it takes from health authorities receiving a call to the responders are on the scene. The additional time it takes to send out a vehicle and responding to a client is not included, nor the potentially decreased drivetime associated with emergency rescue driving.

This paper investigates distances from second homes in the context of second-home mobility.

This is not to be confused with temporary mobility, which includes all forms of migration and tourism. Other forms of tourism, such as the use of caravans or boats, are not considered.

The inherent scope was to model the circulation attributed to second-home mobility in the residential population to see the space-time pattern. And further, to see how redistribution potentially affects supply-demand ratios to public health services. Circulation-data was not available, which creates an abundance of limitations. Limitations become apparent in future chapters.

(6)

5

3. Previous studies

3.1 Access theory

The accessibility-concept holds many meanings, definitions, and operationalizations in the literature. A standard description defines it as 'the ability to reach,' often divided into locational- accessibility which pertains to proximity, and distance-bridging accessibility: mobility. In other words, how far away and what means of travel (Haugen, 2012). Accessibility is in this thesis defined as the combination of proximity and mobility, or how easily a client can reach a destination. However, several barriers may impede or influence accessibility to health care.

Penchansky and Thomas (1981) were the first to group characteristics of access to health care into dimensions. They pointed to accessibility, availability, acceptability, affordability, and accommodation (Penchansky et al., 1981).

Availability describes the number of local services or providers that clients may choose from (Guagliardo, 2004), e.g., how many individual health care centers (supply) are within a 5- minute walking distance from a demand-point or a client. From a health care perspective, this could also include available physicians or hospital beds offered in each health care center. The author strictly link accessibility (i.e., proximity and mobility) to availability. If the means of mobility is high, such as using a vehicle instead of walking, the area coverage within the specified time frame becomes greater. In turn, more available providers may be located within reach of the client. Thus, the two dimensions are considered simultaneously in the literature, denoted as spatial accessibility. The term is used frequently in urban health-accessibility studies for the reason that more facilities exist, and availability is more relevant (Luo et al., 2009; Luo et al., 2003; Shah et al., 2016).

Of the non-spatial dimensions, affordability relates to the provider's charges and the client's ability and inclination to pay for services (McLaughlin et al., 2002). In contrast, acceptability relates to the comfortability of both client and provider towards each other. It is related to different demographics, such as age, sex, ethnicity, diagnosis, and coverage of the client (ibid).

Lastly, accommodation relates to how the provider-services are organized, including the hour of operation, communication coverage, and function and the client's ability to receive care before the appointment (McLaughlin et al., 2002; Penchansky et al., 1981). Other impediments have been suggested, such as the dimension of awareness attributed to health care (Saurman, 2015). All of these dimensions constitute separate entities that must be interconnected for a true operationalization of access.

(7)

6

3.2 Measuring accessibility

Various methods exist to measure spatial accessibility. Guagliardo (2004) describes four categories appropriate for measuring spatial accessibility to health care: provider-to-population ratios, distance to nearest provider, the average length to a set of providers, and gravity models (Guagliardo, 2004), all of which are applicable in a GIS-environment. The methods associated with each category contain limitations. Still, ultimately gravity models are considered to capture the relationship between the population and the provider more accurately, which is also reflected by the current literature (Frew et al., 2017; Kaur Khakh et al., 2019).

One of the popular gravitational models is the two-step floating catchment area method (2SFCA) developed by Lou and Wang (2003). This approach stems from a family of metrics designed to quantify the dimensions that forego in spatial accessibility. Initially, the method was formed to account for caveats in studies that distinctly considered either accessibility or availability. This is appealing because it regards capacity restrictions in the providers (e.g., available physicians or hospital beds per facility) and allows for local competition between the providers. While a variation of the 2SFCA could be ideal for measuring a second home population, attempting to measure available physicians at a location for two separate communities is yet to be developed. The second home population is a product of the residential population, which requires circulation data to measure.

Moreover, the 2SFCA is criticized for not including personal characteristics of the demand, such as age, social class, economic status, and transportation opportunities that are available (Reshadat et al., 2019). Furthermore, accessibility, as it is defined in this thesis, is affected by traffic density, weather conditions, and time constraints, which are common limitations in 2SFCA-studies (ibid). The temporal dimension is particularly interesting because circulation to second homes tends to be seasonal in Norway. The winter season may bring difficult travel conditions, whereas traffic density decides how quickly it is possible to traverse a line segment at a given time. Opening hours of health care facilities are also influenced by time. In Norway, this has previously been documented in out-of-hours casualty clinics where the available clinics alternate with time (Raknes et al., 2014). Because of this, the 2SFCA is not an ideal approach for measuring the second-home population.

Guagliardo (2004) describes the 'distance to nearest provider' as a commonly used measure of spatial accessibility. Still, it is criticized because it fails to include availability where there is an array of providers (ibid). The argument is that a travel impedance (such as drivetime or distance) is a poor indicator of spatial accessibility as it does not include other dimensions such as those

(8)

7 described by Penchansky and Thomas (1981). However, there is no correct method that accounts for all the dimensions related to access, and the one that should be chosen is circumstantial. The 'distance to nearest provider' method is more appropriate when 'availability' is unavailable and is particularly useful in rural areas when the nearest service provider is the only available option. This method is helpful in identifying the best-case scenario when other barriers do not impede accessibility.

3.3 Second homes and frequency of use

Accessibility analyses have the potential to calculate impediments of access from a population to a service. However, as pointed out by Back & Marjavaara (2017), such studies are often based on census records or government registries which place the population in their residential homes (Back et al., 2017). Census data is inherently excellent for use in the analysis but may be misrepresentative in areas where temporary residents or an invisible population exists (ibid).

In the case of Finland, Adamiak et al. (2017) show that seasonal population-density is more significant in rural areas due to the processes of temporary mobility. This redistribution is linked to second-home tourism in amenity-rich regions, more second homes, and increased year-round frequency of use – mainly in summer months, which is well documented in many countries (Adamiak et al., 2015). More recently, studies call this a phenomenon of temporary 'counter urbanization' as half of the population has access to second homes meaning that half of the population relocates during certain seasons (Adamiak et al., 2017; Ellingsen, 2017). A potential consequence of second-home mobility, and other temporary mobility, is a mismatch of where the population exists in time and how public services are allotted. This was further highlighted during the COVID-19 pandemic, as mentioned in the first chapter, but also in that many municipalities in Norway contain more second homes than residential buildings (Norway, 2020).

The rate at which second homes are used, and at what time has become increasingly important in understanding forms of temporary mobility and to aid in policy planning. Müller et al. (2004) created a theoretical model that groups second homes into types based on a number of variables (Müller et al., 2004). Part of these variables included distance from a primary residence to second homes, change in property value, year of construction, and second-home density. In turn, second homes were grouped based on the variables above to determine the potential frequency of use (ibid). This theoretical model was operationalized and used to map the Swedish second-home landscape by Back & Marjavaara (2017), which is quite significant to local and regional planning, also in that it has an impact on accessibility-studies.

(9)

8 In Norway, the literature focuses on second-home agglomerations in the peripheries of Oslo or designated second-home areas with purpose-built second homes densely constructed together (Arnesen et al., 2011; Borge et al., 2015). These accumulations are associated with more usage and can fulfill daily life functions similar to the primary residence. Their density can identify these groups, comparable to what is described as a purpose-built weekend or purpose-built vacation by Müller (Back et al., 2017; Ericsson et al., 2011; Müller et al., 2004).

3.4 Second homes and health services

Since the new health care system was implemented on the 1^st of January 2012: 'the interaction reform,' municipalities received more responsibility in ensuring its people health care services.

The Municipal Health Care Act, which was revised, states that the municipalities have a statutory duty to provide health care to all residents, including temporary populations staying within the confinements of the municipality (Akuttmedisinforskriften, 2015). The act refers to public organized health care that does not belong to the state or county, but also private health care is regulated to uphold this law. Hospitals and ambulatory services are state-owned, while casualty clinics are the responsibility of municipalities in addition to how these are organized.

Currently, many municipalities incorporate inter-municipal casualty clinics, meaning a collaboration where two or more municipalities share a single casualty clinic. While this is thought to increase the quality of service, it also increases the distance one has to travel to reach a casualty clinic (Raknes et al., 2014). As mentioned in the first chapter, distance is associated with infrequent use (ibid).

Prior to the reform mentioned above, an investigation was carried out by Ellingsen et al. (2010) to assess the state of health care for municipalities related to second homes. Challenges in health care were outlined to be an aging population, an increase in diseases and chronic health conditions, increased use of second homes by the elderly, and growth in rights-awareness (Ellingsen et al., 2010). The assessment showed that few municipalities had attributed any additional cost in civil infrastructure or health services to account for the temporary second- home population (Borge et al., 2015). While this was discussed a decade ago, the COVID-19 pandemic proved that municipalities that host many second homes are not capable of serving as many people as they have in their community (Arnesen, 2020).

(10)

9

4. Study Area

The Southern and Eastern Regional Health Authority is one of four geographically defined areas or health zones in Norway. The aforementioned regional health authority (RHA) is composed of 5 counties: Agder, Innlandet, Vestfold and Telemark, Oslo, and Viken. These counties comprise the study area in which ~245 000 second homes exist, with a state-owned health system subjected to the Norwegian Ministry of Health and Care Services. Approximately 3 million people and an estimated 78 000 employees reside within the RHA as of 2020.

The map in Figure 1 shows the geographic distribution of the study area with the corresponding county- and municipality borders, but also grants an overview of the existing casualty clinics, somatic hospitals, and ambulance stations which are relevant for this thesis. Transparency is added to the symbols so that the co-location of ambulance stations and casualty clinics may be observed. The mountain belt, typically associated with Norwegian rurality, is located to the Western and Northern parts on the map. A few domestic services located outside of the study area are included.

Each RHA is responsible for all public enterprises within the specialist health service. The responsibility entails somatic- and psychiatric hospitals, substance abuse treatment centers, other institutions, and prehospital services. They also have partnerships with the private sector.

The RHA's ensure that the population has their need for specialist health services available. In contrast, the subgroups, or local health authorities (LHA) provide the services in accordance with the Norwegian legislation and the Norwegian Ministry of Health and Care Services. On the other hand, facilities related to primary health care are attributed to the municipalities, and so the organization of health departments is split. The services included in the primary care are casualty clinics, nursing- and care services, and other local health facilities.

The South-East RHA is interesting because it contains above half of the second homes in the country while maintaining the quickest response time out of the four RHA's when measured from primary residences to casualty clinics (Raknes et al., 2014). It also contains over half of the population in Norway, which translates to approximately six people per second home.

(11)

Figure 1: Map of the study area with an overview of somatic hospitals, casualty clinics, and ambulance stations.

(12)

11

5. Material and method

The data material incorporated in this thesis stems from the Norwegian Mapping Authority, Geodata AS, and the Norwegian Research Centre. The data have their origin in Statistics Norway - the Norwegian statistics bureau (Norway, 2020). Administrative borders, such as municipality- and county borders, were extracted from the Norway Digital collaboration as raw data. A subcontractor of Esri, Geodata AS, provided the Norwegian cadaster from 2017. The cadaster contains information related to real estates, such as ownership, the dimensions, and the precise location of each building. The same company supplied a processed network dataset.

Lastly, they provided statistical units which comprise smaller areas of similarity that fit within municipalities (see table 1). The Norwegian Research Centre provided an Excel-document with casualty clinics from 2018, which were geocoded manually.

Table 1: An overview of the data used in this thesis

Data source Date updated Data type

Administrative borders Norway Digital 2020 Polygons

Network dataset Geodata AS 2019 Network data

Statistical block units Geodata AS 2019 Polygons

Second homes Geodata AS 2017 Origins (point)

Ambulance stations Geodata AS 2019 Destinations (point)

Casualty clinics NORCE 2018 Destinations (point)

Somatic hospitals Geodata AS 2019 Destinations (point)

5.1 Network data

Network datasets are used to model transportation systems in a Geographical information system (GIS) (Esri, 2019d). The latter is a software-type that can process geostatistical data and is in this paper used to model drivetime from second homes to three public services. The network dataset is the source feature which contains user-created information about how quickly one may travel, where it is possible to travel, and how. In the GIS-environment, transportation analysis may be conducted, such as the 2SFCA or 'distance to nearest provider' as methods that measure forms of accessibility.

The network dataset used in this thesis was preprocessed and built by Geodata AS and applied in Esri's software: ArcGIS Pro version 2.5.0. The network data can be attained in its raw version

(13)

12 from Norway Digital and is part of the public geographical databank (DOK) that is organized for municipal planners.

In a GIS-environment, road networks are comprised of edges (lines) and junctions (points) that are linked together by attributes and properties (Esri, 2019a). Edges contain characteristics and properties from real-world road-segments, while junctions are imaginary points that determine how they connect: their connectivity policy. For instance, a vehicle may not traverse on pedestrian roads or train tracks.

Attributes in the preprocessed network dataset are the following: tunnel, bridge, road structure, physical barriers, and locked entry. Moreover, incorporated properties are axel load, speed limit, height restrictions, and turn restrictions. Attributes and properties will create complexity and accuracy in the modeled network and influence which paths are traversable in a travel mode or for a specific vehicle. For example, a height restriction or an axel load property may restrict a truck from traversing on particular edges. In turn, this changes the path modeled by the solver (ibid).

Hierarchical structures in networks enable personal preference in the network. These structures are useful for networks that measure travel time, e.g., so highways take priority over local roads when the travel time is equal in both paths (Esri, 2019b). Freeways and highways are grouped as primary roads, major and arterial roads are secondary roads, while local roads contain local streets. The incorporation of hierarchies did not influence the final travel time statistics displayed in the next chapter and emulates the preference of drivers on the road (Esri, 2019b).

Additional restrictions were applied to ferry ways, walkways, one-way, and physical barriers;

segments where cars cannot traverse. U-turns are allowed in the network.

Network analysis layers in ArcGIS are the toolkits required to perform analysis on the network dataset. Two of these are relevant for this paper: the OD cost matrix- and Service Area analysis.

The latter encompasses all available streets from a point with a cutoff time, e.g., 15 minutes. In that case, all roads that may be reached within 15 minutes will be displayed in a coverage area from the point, which allows us to see the drivetime coverage from the facility (Esri, 2019c).

The OD cost matrix solver finds the minimum network travel impedance required to travel from the origin to the nearest destination (ibid). The cost impedance was measured in travel time (minutes) from all origin-points (second homes) to the closest destination (service).

(14)

13

5.2 Origin data

The cadaster contains real estate data that originates in the official government registry. In its spatial form, the cadaster contains an overview of all properties, property boundaries, addresses, and buildings in Norway. The building-part of the cadaster is represented as a point feature class with attributes for individual buildings. These attributes include the 'amount of buildings attached to the address,' 'building category,' and 'building type,' as the main features. This system is also coded according to the function the building should have. The standard of each code, and how second homes may be extracted from the dataset is available from Statistics Norway (Norway, 2020). Codes named: '161', '162', and '163' are the categories that comprise second homes. Buildings classified as: 'permanent residential buildings used as a second home,' 'farmhouse used as a second home,' and 'recreational buildings' are the building categories that comprise these codes. In this selection, 'building types' named 'cottage' are included. As shown in Table 1, the cadaster is from 2017. Because of this, ~429 000 individual second homes are represented in the final selection, compared to the existing ~437 000 second homes that exist in Norway in 2020 (ibid).

5.3 Destination data

Health care services (casualty clinics, ambulance stations, and somatic hospitals) are localized to serve as many of the permanent population as possible within a municipality, but this varies significantly from urban to rural areas (Morken et al., 2012). Many of these services are also co-localized, as observed in Figure 1. Out of these services, casualty clinics have become fewer, and the distances have increased through inter-municipal collaborations, while the adequacy of service may increase (Raknes et al., 2014). Collaborations where local casualty clinics are closed during weekends, night, or during a shift in the day are not accounted for. The consequence, particularly as this thesis calculates a minimized cost impedance in travel time to the nearest facility, is the assumption that local casualty clinics are open. This grants a best- case scenario in the results. A report from 2011 showcases that 40 percent of leisure municipalities have an associated inter-municipal casualty clinic, which likely results in longer travel times during certain days and hours of the week, but this varies significantly between each municipality (Hodne, 2011). A complete list of casualty clinics in South-Eastern Norway is included in this thesis and was supplied by NORCE.

Somatic hospitals and ambulance stations were gathered based on IPER's business register from 2019, which follow the NACE categorization standard. Among economic properties, the accurate location of businesses may be found in this dataset. The data were compared with

(15)

14 available open street map data and information from the Norwegian Directorate for Health in the South-East Region. The business data was manually cleared for duplicates using the information in the attribute table.

5.4 Method

Travel times were calculated using an OD cost matrix analysis layer. According to Esri (2019c), an OD cost matrix calculates the distance from multiple origins to multiple destinations using a solver that minimizes the chosen cost impedance. This finds the least-cost paths along the network measured in travel time. Three OD cost matrix analysis layers were created for the separate health services: a single destination for each origin. In other words, travel time was calculated from the 245 000 second homes in the South-East region to the nearest facility in each of the analysis layers. Each matrix contains no cutoff time and measures distance to the closest destination point in all analysis layers from the same origins. Unlocated second homes located more than 2000 meters from a nearby road were excluded. This was done because it is difficult to assume that an ambulance would be used instead of a helicopter to reach those buildings. The total amount of excluded second homes were 2165, none of which affected the identified areas in the next chapter.

The result from the analysis layers is represented as straight lines from the origin to destination.

A map of those lines is not included in the results. This is because the accumulated attributes in each line are computed from calculated routes, not as the crow flies, which could misrepresent the solver. The OD cost matrix solver grants the same attribute-data as with the 'closest facility' solver, with the exception that OD cost matrices do not compute the shapes of the routes as graphics, only the travel time related to them (Esri, 2019c). The method is particularly useful when the dedicated routes are not needed as graphics, and the result is generated quicker (ibid)

The straight-line outputs from the OD cost matrices contain drivetime attributes. These attributes were joined to the origins using the attribute: 'Source ID.' Each second home, or origin, includes information on the travel time to each of the health services. This information consists of the distances of each route in meters and the time it takes to travel.

Statistical block units were initially created by Statistics Norway to separate municipalities into smaller polygon-units, which allows for more flexibility in geographic analysis. For reference, there are currently 356 municipalities in Norway. Within these, there are 14 000 statistical units, which were used for aggregation.

(16)

15 The point-data, or origins, that contain the results and attributes from the OD cost matrices was aggregated into statistical block units using spatial join, with the merge rule: 'completely within.' This ensures that all points within the confined border of a statistical unit are merged into that unit. The principle is demonstrated in Figure 2:

Figure 2: Illustration of points merging into polygons in a spatial join.

The spatial join also combines the attributes from each point into the polygon. This was done using the median: a measure of central tendency appropriate for positively skewed distributions.

The result of this is median travel time to each of the public services from each statistical unit in the study area.

In addition to the OD cost matrices, a service area analysis layer was used to calculate a 25 minute cutoff from ambulance stations. It uncovers the accurate area coverage of ambulatory services in the study area and is used in conjunction with other results.

5.5 Identifying peripheries

The median travel time in each statistical unit was controlled against the suggested response time to casualty clinics and ambulance stations, as described in the third chapter. Units that fell outside of the proposed time constraints were extracted to identify areas for each category. For ambulance stations, this means all regions, or units, that have a median travel time above or equal to 25 minutes. For casualty clinics, 40 minutes were used. As for somatic hospitals, a suggestion was set to 60 minutes as no guidelines exist. This resulted in 3 separate polygon layers that contain peripheral areas to public services with a basis in second homes. An overlay of these three layers determines areas that are peripheral to two or more of the service providers so that no one area is peripheral to one service alone.

5.6 Kernel density estimation

The kernel density tool was used to calculate the probability density function of second homes.

This creates a smoothed circular surface over each point, or second homes, which diminishes based on the set search radius. The result is mostly determined by the user-specified search radius, as this decides the prominence of each point. A small radius of 500 meters was used to create a detailed pattern, where second-home agglomerations may be dominant. This addition can help visualize agglomerations and their relative location to peripheral areas.

Unit 1: 8 homes Unit 2: 4 homes

Spatial join

(17)

16

6. Results

This chapter will be divided into four sections. The first section showcases the statistical results in travel time from the OD cost matrix for each of the services. The following section shows three products: a hot-spot analysis of second-home owners, service area analysis of ambulance station, and kernel density of second homes. Lastly, municipalities with peripheries of low geographic access are identified and displayed with a final summary.

Statistical results

The resulting statistics from the OD cost matrix is revealed in Table 2. The table is based on travel time from ~245 000 individual second homes, calculated in minutes. In the text below the table, statistics are compared with the suggestions discussed in this thesis.

Table 2: Resulting statistics in the South-Eastern Norway Regional Health Authority in minutes.

Significant numbers related to the descriptions below are highlighted in bold.

Somatic hospitals: The suggestion of 60 minutes as the maximum travel time for the 90^th percentile is false. ~79 000 second homes have more than one hour of travel time to the nearest hospital.

Casualty clinics: The Norwegian Research Centre suggests the maximum travel time for the 90^th and 95^th percentile to be 40 and 60 minutes. Table 2 shows the percentiles of 50- and 78 minutes. In total, ~74 000 second homes have more than 40 minutes to the nearest casualty clinic, and ~28 000 more than one hour.

Ambulance stations: The Emergency Committee suggested a 25-minute response time as the 90^th percentile in acute situations in rural areas. This was measured using a typical drivetime impedance based on speed-limits. The result shows that an estimated ~106 500 second homes are outside of this range.

Median Std. Dev 10^th 90^th 95^th Somatic hospitals 42 min 31 min 13 min 94 min 106 min

Casualty clinics 29 min 21 min 10 min 50 min 78 min Ambulance stations 23 min 17 min 10 min 48 min 58 min

(18)

17

6.1 Hot-spot analysis

This section is a complementary result which shows statistically significant clusters of second- home owners in a hot-spot analysis. While it is not unnatural to find more clusters in urban areas due to the higher population density, it supports that many second-home owners travel short distances to their houses and that the frequency of use may be higher than anticipated.

Figure 3: Hot-spot map of second-home owners in the South East Regional Health Authority.

The green area represent so-called mountain municipalities.

(19)

18

6.2 Service area analysis

This map shows the extent of all accessible paths a vehicle can take in a 25-minute cutoff time from ambulance stations. The resulting service areas show a detailed distribution when there is no traffic, and speed-limits are followed. The result may be used in conjunction with results in the next sections.

Figure 4: Service area analysis from ambulance stations in the study area.

(20)

19

6.3 Kernel density of second homes

The kernel density estimation shows the probability density function of second homes and is displayed in Figure 5. This map includes counties, municipalities, and prehospital services as relational features. The dense areas are shown circular concentrations, whereas the sparse areas are symbolized in white to transparent in the low-end spectrum. The inclusion of this method is for visualization related to the peripheries in the final sections and to see the density of second homes in the region. Second-home agglomerations are also associated with increased year- round use.

Figure 5: Map of the study area with a kernel density estimation and prehospital services.

(21)

20

6.4 Peripheries to health care

The peripheries are used to systemize and display the geographic results from the OD cost matrix. Second homes and their drivetime attributes to the service providers are spatially joined to the statistical units. The product is peripheral areas to a chosen set of service providers. Areas are identified where the median drivetime is above the suggested guidelines mentioned in the second chapter, and where those coincide in an overlay. Essentially, places where the median drivetime from second homes exceed 25 minutes to ambulance stations and 40 minutes to casualty clinics. A second periphery is included to show areas with more than 30 minutes to ambulance stations and 60 minutes to casualty clinics.

Municipalities that contain peripheries are marked with a number on the map with name descriptions. Second homes, and their amount in each municipality, is shown as proportional symbols. The location of these symbols in their respective municipalities is not relevant to where the second homes are located, but merely indicates the amount in each municipality.

Figure 6: Map showing peripheries to prehospital services, and second homes as proportional symbols.

An estimated 20 000 second homes are found within all identified peripheries. Approximately 10 000 of the second homes are located in the peripheries in the outlined map frame.

(22)

21 Figure 7 differs from the previous model in its incorporation of somatic hospitals in the peripheries. Areas identified in 'periphery 1' also have a drivetime ≥ 60 minutes to the nearest somatic hospital. The second periphery adopts the 90^th percentile of somatic hospitals, as shown in Table 2. Periphery 2 shows areas where the median drivetime is ≥ 30 minutes to ambulance stations, ≥ 60 minutes to casualty clinics and ≥ 94 minutes to somatic hospitals. Overall, the peripheries comprise 15 000 second homes, with 8500 located in the map frame.

Figure 7: Map showing peripheral areas to prehospital services and somatic hospitals.

6.5 Understanding the results

In the study area, 79 000 second homes are located more than a 1-hour drive from the nearest hospital. Seventy-four thousand of them are located 40 minutes away from casualty clinics, and more than 106 000 second homes are more than 25 minutes away from ambulance stations.

Approximately 8 500 of these are located within the outlined frame of Figure 7.

The use of peripheries grants an aggregated and generalized view on the map. A single area may be highlighted in red but may contain a concentration of second homes anywhere within the unit. Furthermore, overlaying coincident health institutions removes many peripheries in one category. This is evident from the second-home count found in the peripheries and in the service area analysis in Figure 4, which demonstrates the coverage in a single category.

(23)

22

7. Concluding discussion

Guagliardo (2004) writes about different methods that can be used to measure spatial accessibility and how none of them fully captures the dimensions in his theory related to health care (Guagliardo, 2004). The correct method is circumstantial to the service we measure against, and as described by Haugen (2012): to whom – as not everyone readily requires a service or choose the nearest location (Haugen, 2012).

This multifaceted topic increases in complexity when second-home mobility is introduced.

Large masses migrate to the rural hinterlands on a seasonal basis. This creates a circulation pattern where the demand for services in urban areas becomes lessened, and the market in rural areas increases. This pattern is difficult to capture as it requires circulation data that explain where people travel, how frequent, and at what time. From a geographical standpoint, that information is useful because people are commonly used as a proxy for where services should be located and how meaningful these services should be. If half of the population migrates to rural areas during summer months, it should be paramount to study it, but also model it so that the effects of the demand can be adjusted in the supply. In turn, this makes it so that one may compare the market share in each service provider, which is ideal for enforcing health equity.

This circulatory data is, therefore, a requirement to model Penchansky and Thomas' (1981) availability with the correct demand in space and time, but also to apply more appropriate accessibility-methods such as a variation of the 2SFCA.

Currently, there is little knowledge about second homes in the health sector. This is evidenced by studies that argue for the inclusion of second homes in rural studies in Finland, Sweden, and Norway (Adamiak et al., 2017; Back et al., 2017; Ellingsen, 2017) but also by the recent COVID-19 pandemic where the government prohibited use of second homes outside of the residential municipality as a maximum precautionary strategy.

A strength in this study is the possibility to estimate drivetimes to a temporary population that inherently is not excluded in accessibility analysis. A second strength is the ability to show peripheral areas from second homes to health care in small statistical units identified within municipalities. This limits aggregation and generalization in the dataset, while still providing concrete results. The statistics from the OD cost matrix are comparable to previous studies, such as Raknes' (2014) study on travel time from residential buildings to casualty clinics (Raknes et al., 2014). Here, the median drivetime from residential buildings to casualty clinics

(24)

23 is 11 minutes fewer than what was found in this paper. This used data from 2011, although it does show that the 90^th percentile is similar for both populations.

This study investigated one aspect of accessibility: travel time, which fits into the current debate on statutory response times. It demonstrates that approximately 30% of second homes are outside of the current guidelines made by NORCE and the Emergency Committee and 1 hour away from somatic hospitals. The current policies cannot be enforced without dramatically increasing the resources spent in rural civil infrastructure and public health services. Second homes in areas that collectively are peripheral to all of the service providers discussed in this thesis hold the weakest positions for emergency medical aid. The peripheries contain ~20 000 second homes, which potentially can host 120 000 residents. The potential for redistribution to weakly positioned areas is great, which was further emphasized by the maximum precautionary strategy. At the same time, it not economically viable to supply a temporary population in the same manner as the residential population. This paper therefore argues that second home mobility must be placed in a system so that the appropriate resources may be allotted according to the seasonal redistribution.

7.1 Limitations

The quantitative approach of using 'distance to the nearest provider' as described by Guagliardo (2004) shows the best-case scenario. The method only incorporates Penchansky and Thomas' accessibility, i.e., proximity and mobility, disregarding the other dimensions. In turn, this produces a result that is more attractive than actual conditions, as no other barriers are involved (Penchansky et al., 1981). As shown in the results, 79 000 – 106 000 second homes are outside of the suggested guidelines in each category compared to the existing 245 000 second homes.

The further weakness stems from the inability to consider time, i.e., the temporal dimension.

Because the circulation pattern is unknown, we cannot deduct residents from the permanent population to create a realistic time-space configuration, e.g., for a specific month. This makes it more challenging to include more impediments, such as environmental data for a particular season, or to take seasonal recreational commuting, traffic, into effect. Furthermore, the nearest available service provider may be unavailable during certain hours of the day, as in Norwegian casualty clinics (Raknes et al., 2014).

Moreover, it is difficult to determine if the speed-limits represent how quickly vehicles traverse the network. Ambulance stations can perform emergency rescue driving where they drive

(25)

24 according to conditions but are allowed to break speed limits. Moreover, the calculated travel times are based on distances from second homes, not where incidents occur in reality.

Other weaknesses in this study are not including second-hand on-call arrangements, such as the Red Cross and other volunteer-organizations that may aid ambulatory services with transport.

Casualty clinics by boat or the use of a helicopter are not taken into account. Other flaws relate to the inability to consider the total second home population, as the study has its basis in second homes, not people. On average, there are ~6.1 people per second home in the study area.

7.2 Future studies

Since the last decade, the Norwegian health authorities have discussed equal care for everyone.

One way towards this is to know where people are in space and time and how this affects supply and demand so that the appropriate resources are used efficiently. In the context of rural studies, we need to understand where second-home mobility is happening and, more importantly, the time of occurrence and place this in a system so that the appropriate resources can be spent where residents are located in space and time. A variation of the 2SFCA could be used to model this with the appropriate data.

(26)

25

8. References

Adamiak, C., Pitkänen, K., & Lehtonen, O. J. G. (2017). Seasonal residence and

counterurbanization: the role of second homes in population redistribution in Finland.

82(5), 1035-1050.

Adamiak, C., Vepsalainen, M., Strandell, A., Hiltunen, M., Pitkänen, K., Hall, C., . . . Åkerlund, U. (2015). Second home tourism in Finland – Perceptions of citizens and municipalities on the state and development of second home tourism.

Akuttmedisinforskriften. (2015). Forskrift om krav til og organisering av kommunal

legevaktordning, ambulansetjeneste, medisinsk nødmeldetjeneste. (FOR-2015-03-20- 231). Retrieved from: https://lovdata.no/dokument/LTI/forskrift/2015-03-20-231 Arnesen, T. (2020). Meanwhile in Norway: the coronavirus crisis puts emergency

preparedness in rural mountain municipalities on the agenda. Retrieved from https://www.euromontana.org/en/meanwhile-in-norway-the-coronavirus-crisis-puts- emergency-preparedness-in-rural-mountain-municipalities-on-the-agenda/

Arnesen, T., Overvåg, K., Ericsson, B., & Skjeggedal, T. (2011). Recreation and the making of the multi-house hom. Retrieved from Lillehammer, Norway:

http://distriktssenteret.no/wp-content/uploads/2013/09/notat_second- homes_ostlandsforskning.pdf

Back, A., & Marjavaara, R. J. T. G. (2017). Mapping an invisible population: the uneven geography of second-home tourism. 19(4), 595-611.

Borge, L.-E., Ellingsen, W., Hjelseth, A., Leikvoll, G. K., Løyland, K., & Nyhus, O. H.

(2015). Inntekter og utgifter i hyttekommuner.

Ellingsen, W. (2017). Rural Second Homes: A Narrative of De-Centralisation. 57(2), 229- 244. doi:10.1111/soru.12130

Ellingsen, W., Hodne, T., & Sørheim, S. (2010). Pleie- og omsorgstjenester for hytteboere.

Report 8/2010. Kristiansand: Agderforskning

Ericsson, B., Skjeggedal, T., Arnesen, T., & Overvåg, K. (2011). Second Homes in Norway.

Retrieved from https://distriktssenteret.no/wp-content/uploads/2013/02/second- homes.pdf

Esri. (2019a). Network Elements. Retrieved from

https://desktop.arcgis.com/en/arcmap/latest/extensions/network-analyst/network- elements.htm

Esri. (2019b). About network analysis with hierarchy. Retrieved from

https://desktop.arcgis.com/en/arcmap/latest/extensions/network-analyst/network- analysis-with-hierarchy.htm

Esri. (2019c). Types of network analysis layers. Retrieved from

https://desktop.arcgis.com/en/arcmap/latest/extensions/network-analyst/types-of- network-analyses.htm

Esri. (2019d). What is a network dataset? Retrieved from

https://desktop.arcgis.com/en/arcmap/latest/extensions/network-analyst/what-is-a- network-dataset.htm

(27)

26 Frew, R., Higgs, G., Harding, J., & Langford, M. (2017). Investigating geospatial data

usability from a health geography perspective using sensitivity analysis: The example of potential accessibility to primary healthcare. Journal of Transport & Health, 6, 128-142.

Guagliardo, M. (2004). Spatial accessibility of primary care: Concepts, methods and

challenges. International journal of health geographics, 3, 3. doi:10.1186/1476-072X- 3-3

Haugen, K. (2012). The accessibility paradox : Everyday geographies of proximity, distance and mobility. (Doctoral thesis, comprehensive summary), Umeå universitet, Umeå.

Retrieved from http://urn.kb.se/resolve?urn=urn:nbn:se:umu:diva-50710 DiVA database. (2012:1)

Helsenorge. (2018). Kvalitetsindikator for akuttmedisinske tjenester utenfor sykehus: Tid fra AMK varsles til ambulanse er på hendelsessted. Retrieved from

https://www.helsedirektoratet.no/statistikk/kvalitetsindikatorer/akuttmedisinske- tjenester-utenfor-sykehus/tid-fra-amk-varsles-til-ambulanse-er-p%C3%A5- hendelsessted

Hodne, T. (2011). Second Homes i et lokalt utviklingsperspektiv. Retrieved from

Kaur Khakh, A., Fast, V., & Shahid, R. (2019). Spatial Accessibility to Primary Healthcare Services by Multimodal Means of Travel: Synthesis and Case Study in the City of Calgary. International Journal of Environmental Research and Public Health, 16(2).

doi:10.3390/ijerph16020170

Luo, W., & Qi, Y. (2009). An enhanced two-step floating catchment area (E2SFCA) method for measuring spatial accessibility to primary care physicians. Health & Place, 15(4), 1100-1107. doi:https://doi.org/10.1016/j.healthplace.2009.06.002

Luo, W., & Wang, F. (2003). Measures of Spatial Accessibility to Health Care in a GIS Environment: Synthesis and a Case Study in the Chicago Region. Environment and Planning B: Planning and Design, 30(6), 865-884. doi:10.1068/b29120

McLaughlin, C. G., & Wyszewianski, L. (2002). Access to care: remembering old lessons.

Health services research, 37(6), 1441-1443. doi:10.1111/1475-6773.12171 Morken, T., & Zachariassen, S. M. (2012). Legevaktorganisering i Norge. Rapport fra

Nasjonalt legevaktregister 2012. Retrieved from Bergen:

http://hdl.handle.net/1956/6247

Müller, D. K., Hall, C. M., & Keen, D. (2004). Second home tourism impact, planning and management. In (pp. 15-32). Clevedon: Channel View Publications.

Statistics Norway. (2017). Eldre menn lever ut hyttedrømmen. Retrieved from

https://www.ssb.no/bygg-bolig-og-eiendom/artikler-og-publikasjoner/eldre-menn- lever-ut-hyttedrommen

Statistics Norway. (2020). Bulding Stock. Retrieved from https://www.ssb.no/en/bygg-bolig- og-eiendom/statistikker/bygningsmasse

Penchansky, R., & Thomas, J. W. (1981). The Concept of Access: Definition and Relationship to Consumer Satisfaction. Medical Care, 19(2).

Raknes, G., Marken, T., & Hunskaar, S. (2014). Travel time and distances to Norwegian out- of-hours casualty clinics. Den Norske legeforening(22). doi:doi:

10.4045/tidsskr.14.0437

(28)

27 Reshadat, S., Zangeneh, A., Saeidi, S., Ghasemi, S., R., Rajabi-Gilan, N., & Zakiei, A.

(2019). Inequalities in access to hospitals: a case study in the Islamic Republic of Iran 1997-2012. Eastern Mediterranean Health J, vol 25(2), 119-126.

Rye, J. F., & Gunnerud Berg, N. (2011). The second home phenomenon and Norwegian rurality. Norsk Geografisk Tidsskrift - Norwegian Journal of Geography, 65(3), 126- 136. doi:10.1080/00291951.2011.597873

Saurman, E. (2015). Improving access: modifying Penchansky and Thomas's Theory of Access. Journal of Health Services Research & Policy, 21.

doi:10.1177/1355819615600001

Shah, T. I., Bell, S., & Wilson, K. (2016). Spatial Accessibility to Health Care Services:

Identifying under-Serviced Neighbourhoods in Canadian Urban Areas. PLoS ONE, 11(12), 1-22. doi:10.1371/journal.pone.0168208

(29)

11

Appendix

(30)

11

(31)

11

(32)

11

(33)

12

(34)

13

Health care accessibility and second homes: A spatial analysis in South-East Norway