Applying Environmental Context to Rehabilitation Research Using Geographic Information Systems and Global Positioning Systems Geospatial Technologies

Abstract

Background:

The International Classification of Functioning, Disability, and Health includes important considerations of environmental context in understanding disability, but the environmental impact is often difficult to measure.

Purpose:

Demonstrates the use of Geographic Information Systems (GIS) and Global Positioning Systems (GPS) in rehabilitation research in assessing accessibility and participation; describes how to use these methods, and presents several considerations in using GIS and GPS in research.

Method:

Using methods from public health and medical geography, this article describes how to apply GIS and GPS technologies to rehabilitation research to measure community participation and accessibility to resources.

Findings:

Directions for using ArcGIS functions and case examples joining these mapping technologies with rehabilitation measures are provided.

Conclusions:

Together with traditional measures, these technologies may provide rehabilitation researchers a more comprehensive approach to assessing accessibility and participation.

The 2013 Robert Wood Johnson Foundation city maps predicting variations in life expectancy by zip code provided powerful visual evidence of environmental factors impacting health outcomes (https://www.rwjf.org/en/library/infographics/new-orleans-map.html#/embed). As far back as 1854, public health research has been interested in mapping disease prevalence and environmental conditions precipitating diagnosis at the population and community level. These associations have led to notable findings and prevention measures to improve public health. John Snow, a physician in England, mapped out the locations of the local water wells and cholera deaths on Broad Street in London to support his argument that a contaminated pump was the source of the disease. Through the power of mapping, he convinced the local council to remove the handle to prevent its use. Similarly, Geographic Information Systems (GIS) mapping has been traditionally used for disease mapping that has tremendous value in communicating risk and resulting policy recommendations (Cromley & McLafferty, 2011). GIS is a computerized mapping system that creates, manages, analyzes, and integrates data by geographic location (de Smith et al., 2020). Recently, COVID-19 researchers have utilized GIS technologies to track and predict the disease’s spread, and to inform policy and allocate resources (for a review, see Franch-Pardo et al., 2020). Along with public health research, medical geography uses geographic techniques of mapping and estimation of exposure or risk through spatial analysis in GIS to identify incidents of increased or decreased risk of disease that may be associated with the environmental conditions of where one is geographically located. Environmental factors of climate, air quality, distribution of health resources, or proximity to health-promoting factors are some areas of interest under study to understand what is happening, and where it is happening with the hopes of improving health and mortality outcomes for those affected (de Smith et al., 2020; Rushton, 2003).

In rehabilitation, the International Classification of Functioning, Disability and Health (ICF) also identifies the environment, in conjunction with the person, as contextual factors influencing one’s level of body function, activity, and participation (World Health Organization [WHO], 2002). The ICF model has been an important tool in rehabilitation research and practice in understanding how changes at any level of the model can impact the experience of disability for the individual, and provides a rich picture of the social construct of disability and function where the environment plays a critical role in how much an individual is able to participate. The range of environmental factors affecting an individual’s experience of disability can include social support, availability of transportation and assistive technology, access to healthcare, attitudinal barriers and stigma, and physical barriers in the home or community, among others. These environmental factors are classified in the ICF into five domains, or chapters: (a) products and technology; (b) the natural environment and human-made changes to the environment; (c) support and relationships; (d) attitudes; and (e) services, systems, and policy (WHO, 2002). Each of these aspects of the individual’s environment may benefit from using different assessment approaches to capture the complexity of variables influencing participation.

For example, rehabilitation researchers may use self-report survey or interview responses to assess perceived barriers to participation, and the role of technology, assistive devices, or others in participation. Participant completed time travel diaries can assess frequency of locations visited and activity time, as well as transportation systems used. Surveys may also query the degree of choice or control in community activities or the importance of specific locations in the community, and what features or social aspects of the location contribute to its importance. These measures assess important subjective aspects of participation from the person’s perspective, but may not capture more objective measures of environmental factors also impacting participation and use of community resources.

Borrowing from the mapping techniques of medical geography and public health research, this article presents GIS mapping and Global Positioning System (GPS) technology as additional measures of the environmental context in rehabilitation research. GIS and GPS technologies can integrate temporal and spatial considerations of environmental factors impacting disability and function, with a focus on assessing accessibility to resources and participation outcomes. For people with disabilities, GIS mapping of both density of individuals with disabilities and availability of frequently used rehabilitation services in Australia and the United States was used to inform policy recommendations for resource disbursement and prioritization of services (Gao et al., 2019; Lakhani et al., 2019; Walker et al., 2016).

Taking a more individual perspective, rehabilitation research using GIS and GPS methods with older adults with disabilities noted determinants such as car ownership, health status, and geographic distance play a significant role in access to care (Comber et al., 2011), while population density, street connectivity, and slope were associated with walking for physical activity in the community (Gell et al., 2015). These factors highlight the value of assessing an individual’s community mobility when engaging with the environment as an aspect of participation (Chaix et al., 2012). GIS and GPS offer a unique approach for assessing objective characteristics of the community where one lives, mobility patterns at the individual and group level, and participation within environmental attributes or constraints for people with disabilities. For example, Ben-Pazi and colleagues (2013) noted marked differences in GPS tracking of community mobility patterns and participation activities between an adolescent with cerebral palsy (CP) and her similarly aged sister, while Doherty and colleague (2014) used GIS and GPS to better understand mobility and accessibility patterns for children with CP. GPS measures of speed, average time outdoors, and number of trips and number of tilts for wheelchair users combined with GIS travel maps of locations visited provided significant information on community activities and participation (Harris et al., 2010). Each of these studies also used survey and interview data to provide the personal and social context of GPS and GIS findings, and support a multimodal approach to assessment to better understand the complexity of objective and subjective factors influencing participation.

This article shares examples of opportunities for using GIS and GPS in rehabilitation research as part of a multimodal approach to assessing accessibility and participation, describes how to use these methods, and presents several considerations in using GIS and GPS in research. The three main technologies of interest in the current article are (a) using GIS as a measure of resource accessibility; (b) using GIS as a measure of community participation; and (c) using GPS as a measure of community participation. Although there are specific spatial analysis methods that can be used to analyze geographic data (i.e., spatial autocorrelation, spatial panel data models, Bayesian inference; see Anselin, 1995; Fischer & Getis, 2009; MacNab, 2003), the purpose of this article is to introduce how to construct environment variables in rehabilitation research using GIS mapping technology that can be used in traditional descriptive and predictive analyses. Subjective data, such as perceptions of access to health services, can be integrated into environmental mapping data using more advanced spatial techniques in GIS such as Geographically Weighted Regression (GWR), but this is outside the scope of this article.

RESOURCE ACCESSIBILITY: GIS MEASURES OF PROXIMITY AND DENSITY

Measuring and understanding access to resources is a primary application of using spatial data in rehabilitation research. Of interest may be access to resources that may improve activity or participation, or conversely, exposure to resources in the environment that may present a greater barrier, or risk of experiencing limitations in these areas. One way to objectively assess geographic accessibility to resources is using GIS to measure the proximity and density of the resources of interest for a specific location. Public health research has used GIS to evaluate the role of the built environment, including availability of transportation, recreation, shopping, education, and health facilities in a neighborhood on various health outcomes (Cromley & McLafferty, 2011).

GIS measures of accessibility to positive or negative neighborhood features in one’s community have also been associated with health and participation outcomes. For example, proximity to alcohol outlets was identified as a risk factor for adolescents engaging in risky behavior (Mason & Mennis, 2010). GIS measures of accessibility to resources for individuals in supportive housing found proximity to community features such as libraries, parks, police stations, and grocery stores was associated with more positive community integration outcomes and more involvement in leisure activities (Chan et al., 2014; Mason et al., 2009; Pearce et al., 2006). In these examples, distance measures of proximity or density of identified features in the community were correlated with other measures such as reports of resource or service use to examine how environmental attributes influence behavior.

More recently, GIS has been used to measure availability of services by describing the spatial organization of where individuals with the same heath condition live relative to the locations of needed health or rehabilitation resources in the community (Gao et al., 2019; Lakhani et al., 2019). Establishing these metrics provide useful data for recommendations in rehabilitation service provision and policy related to distribution of needed services. In mental health research, Walker and colleagues (2016) used GIS to map the discrepancy in service needs for children with mental health disorders and the availability of evidence-based practices that could dramatically improve their health outcomes. GIS was used not only in conducting the accessibility analysis, but to visually show the presence of “service deserts,” indicating the lack of trained providers in the geographic area to community stakeholders and policy makers (Walker et al., 2016, p. 851). The authors noted these areas of need were more evident through mapping these discrepancies than presenting tables of computational results.

Measuring Participation With GIS

Studying these “neighborhood effects,” or the surrounding community features of the built environment around one’s home location, has been associated with various health outcomes such as physical activity level, obesity, and mental health (Beyer et al., 2014; Galvez et al., 2013; Wen & Kowaleski-Jones, 2012). The majority of an individual’s employment, social, shopping, and health related activities, however, occur outside of one’s immediate neighborhood, lending support for examining the impact of exposure to areas outside of the residential neighborhood (Kerr et al., 2011; Vallee et al., 2011; Zenk et al., 2011). To broaden the scope of analysis, GIS can be used to create maps of an individual’s “activity space,” or the measure of spatial extent of regular activities reported or recorded in the community (Zenk et al., 2011). The activity space captures the spatial presence, or scope and direction of individuals’ activities in the community (Figure 1). In some cases, it is important to use activity space rather than residential neighborhood to more accurately assess environmental exposure. Using the activity space measure further represents opportunities for participation in the community areas where the individual most frequently travels and engages (Brusilovskiy et al., 2016). Activity spaces can be constructed in GIS through individuals’ self-report of locations visited or frequented in the community, for example, through travel diaries or participatory mapping techniques (Chan et al., 2014; Shaw & Wong, 2011; Townley et al., 2009), or from location data gathered through GPS tracking. In social and behavioral science, these activity space measures of individuals’ presence in the community and opportunity for participation have been correlated with other community, health, mental health, and service use outcomes (Chan et al., 2014; Nemet & Bailey, 2000; Townley et al., 2009; Vallee et al., 2011).

Figure 1.

GIS (Geographic Information Systems) map displaying one standard deviation ellipse activity spaces for three participants based on locations traveled during a 1-week period, with a comparison of social impairment based on the Social Responsiveness Scale (SRS). In this example, Participant 2 has the lowest social impairment score and the largest activity space, as indicated in GIS by the size of the home icon, while Participant 1 had the smallest activity space, visiting just two locations during the week.

Measuring Participation With GPS

While community participation is an important rehabilitation goal, existing measures typically rely on self-report surveys completed by the individual or caregiver regarding the different types of community activities engaged in, and frequency of these activities in a given time period. These methods may be limited by what individuals recall or attend to. An alternative approach to measuring participation for people with disabilities in rehabilitation research is through the use of another geographic based technology, GPS tracking. GPS tracking can be used to measure how individuals with disabilities navigate their environment, the number of locations visited, distance traveled in the community, time spent away from home, and where individuals are spending time in the community. In addition to the examples of rehabilitation research with older adults with disabilities and individuals with CP noted above, GPS tracking has studied participation in individuals with multiple sclerosis, traumatic brain injury, below the knee amputation, stroke, and schizophrenia (Brusilovskiy et al. 2016; Doherty et al., 2014; Jayaraman et al., 2014; McCluskey et al., 2012; Neven et al., 2013). GPS devices can also be paired with accelerometers to measure where in the community individuals with disabilities are most physically active (Jayaraman et al., 2014), or paired with experience sampling methodology (ESM)/ecological momentary assessment (EMA) to provide contextualized, real-time self-reports of an individual’s emotions, cognitions, and behaviors within their environment (Epstein et al., 2014; Mitchell et al., 2014).

Detailed descriptions of the methods of using GIS and GPS technology in relation to accessibility and participation are included below, with examples of key outcomes measured and a review of three case examples.

METHODS

Resource Accessibility: GIS Accessibility Analysis

Geospatial Databases

Critical in GIS research is creating maps of the geographic area of interest in a geodatabase, which stores and organizes data files used for building the map. Maps are constructed in GIS by adding data layers to the database. For example, a base layer could contain the layer of the state, county, or city boundaries, another layer could contain all the geographic features of the land, another the footprints of buildings, or roads and transportation networks. Another layer could contain all of the hospitals in the region, or all of the parks and green spaces. All layers added align through corresponding latitude and longitude geographic coordinates of the database. Beyond the spatial information contained, these data layers can also include quantitative and qualitative information about the attributes of interest, which are then joined to the spatial data in GIS through the latitude and longitude coordinates.

Data Layers

GIS data layers can be obtained through many sources. Government resources at the national, state, county, and city level are now creating GIS data layers representing different community attributes that are often available for download by the public. For example, U.S. Census data GIS layers include information regarding population density and demographic characteristics of the area at a census tract, block group, or block level. Demographic characteristics, such as data on gender, income, employment and disability status, household size, primary languages spoken, and racial and ethnic descriptors describing the community are also available and linked to the census data layers. Police or public safety municipals may have data layers of crime activity, while local transportation departments may have bus or other public transportation networks mapped. Other layers that may be available through government resources of interest to the rehabilitation community include hospitals, acute care centers, community health centers, long-term care centers, schools, universities and colleges, parks/recreation areas, trails and greenways, libraries, public services, land use, farmer’s markets, public services, and prisons. Additionally, Google API allows for the query of place information on a variety of categories, such as establishments, prominent points of interest, geographic locations, and more.

Creating Data Layers

When data layers are not available for a specific community feature of interest, it is also possible to create the data layer by “geocoding” identified locations. Geocoding involves finding the latitude and longitude coordinates when an address or location is known. Locations that are classified as a business by a Standard Industrial Classification (SIC) code can be identified through platforms such as InfoUSA that can provide the names, addresses, and latitude/longitude coordinates for selected categories (e.g., coffee shops) within a user specified distance from a chosen address. InfoUSA generates a list of identified business locations which can be exported into a spreadsheet. InfoUSA requires a paid subscription, but if the researcher’s organization does not have access or the feature is not classified as a business, such as bus stops or parks, locations can also be identified using online mapping interfaces such as Google Maps, which will provide addresses and geographic coordinates. Once the locations are geocoded in a spreadsheet, they can then be imported into a geodatabase created in the mapping software. Of note, for each new data layer added or created, a consistent geographic coordinate system must be selected and applied to project the data for points to correctly align in GIS. For example, in our studies, we used the “WGS 1984” datum. Based on the authors’ familiarity with its use, ArcGIS will be the mapping software used in examples in this article. In ArcGIS once a spreadsheet is uploaded as a.csv file with latitude and longitude coordinates, it can be linked or “joined” as a table to an existing data layer, or converted to a “shapefile” in ArcCatalog as a new data layer (see Table 1 for further details). Participant data is also added in GIS, either as its own data layer, or similarly linked or joined as a table to an existing data layer. External data, such as demographic characteristics or rehabilitation outcomes measured (e.g., employment status, quality of life, depressive symptoms) can also be integrated into GIS by linking or joining the data with the participant’s geographic location to visually display the range of outcomes by community location or proximity to resources. These maps in GIS can visually display how rehabilitation outcomes differ in the community area (Figure 1).

TABLE 1.

A Description of the GIS Functions Referenced in the Article, Corresponding Commands in ArcGIS, and Video Examples

GIS Function Name	ArcGIS Description, from http://desktop.arcgis.com/en/arcmap/	Commands/Path in ArcGIS	Example of Video Demonstration
Join	Transfers attributes from one layer or table to another based on spatial and attribute relationship	In ArcMap, right click on the layer data and select “Joins and Relates” and choose “Join.” Locate the field within data layer you will be joining, ensure that your excel/table is selected, and locate the variable with which you will be joining the data and table.	https://youtube.com/watch?v=oqdj8GceGcg
Shapefile	ESRI vector data storage format for storing the location, shape, and attributes of geographic features. Geographic features in a shapefile can be represented by points, lines, or polygons (areas). It is stored as a set of related files and contains one feature class.	In ArcCatalog, locate the desired .csv file (e.g., coffee shops). Right click, choosing “Create features class from xy table.” Set geographic coordinates to match geospatial database. Select a location to save shapefile (typically where other shapefiles are stored).	https://youtu.be/rQCZHgtTm9Y
Convex hull	Creates a feature class containing polygons which represent a specified minimum bounding geometry enclosing each input feature or each group of input features.	In ArcMap, using the ArcToolbox, select “Data management tools” followed by “Features,” then “Minimum bounding geometry.” Select shapefile to use and change “shape” option to “convex hull.”	https://youtu.be/FuJGKTWvw5A
SDE	Creates standard deviational ellipses to summarize the spatial characteristics of geographic features: central tendency, dispersion, and directional trends.	In ArcMap, using ArcToolbox, select “Special statistics tools,” followed by “Measuring geographic distributions,” then “Directional distribution (standard deviation ellipse).” Select type of standard deviation ellipse (e.g., 1).	https://youtu.be/wkMUTBNZ-el
Activity Space Area	Provides a standardized area for conducting and comparing analyses	Open the attribute table of the activity space (convex hull or SDE) by right clicking on the activity space layer on the right hand of the screen. Select the option “Add a field,” and create a name for the new variable (e.g., Act_Sp). Select “Float” to allow the most decimals and accurate calculation. On the new Field header, right click and select “Calculated geometry” and select unit of measure (e.g., meters, square miles).	https://youtu.be/mLM2QU255Nc
Multiple Ring Buffer Analysis	Creates multiple buffers at specified distances around the input features.	In ArcMap, using ArcToolbox, select “Analysis tools,” followed by “Proximity,” then “Multiple ring buffer.” Then select your input feature (shape file), and make sure the buffer unit is your desired unit (e.g., miles). Select unit inputs (distance from the input feature the ring(s) are) by entering in the box labeled distances, then selecting the plus sign, which should populate the ring distance in the blank space(s) below.	https://youtu.be/AUEaAEjZlu8
XY Path Line	Creates connection lines between two points.	Construct a table with start and end x, y coordinates for each point/location and add to map. In ArcMap, using ArcToolbox, select “Data Management Tools,” then “Features,” then “XY to line”. Select your table as the input, then the start and end x and y fields (column heading) respectively.	https://www.youtube.com/watch?v=n4upO6_0vFs

Access to Resources

Once participants’ home locations and desired community features are added to the map, measures of geographic accessibility can be completed using proximity (distance) or density (concentration) measures. For proximity, distance can be measured using Euclidean (straight line) distance, or using network path, which takes into account road and transportation networks in GIS. There is some support that both proximity measures are highly correlated (r > 0.90; Boscoe et al., 2012), although using network distance takes into consideration travel time, which can be especially important in rural and mountainous landscapes (Apparicio et al., 2008). Density (concentration) measures provide a count of the number of given features in a predetermined area, such as 1 mile from a participant’s home, school, or other location mapped in the database. Density measures are completed in ArcGIS by using the “Multiple Ring Buffer” analysis function (see Table 1). Once calculated, proximity and density measures can then be entered as variables into traditional analysis.

GIS Activity Space

A critical measure of community participation research is geographic scope of mobility (Brusilovskiy et al., 2016). Geographic scope of mobility comprises distance traveled, time, and activity space. GIS-created activity spaces visually present a measure of the area of a participant’s movement in the community, and allow for examination of patterns and distinctions. There are three types of activity space measurements: convex hull, concave hull, and standard deviational ellipses (SDE). Convex hull and concave hull are similar in that they create a perimeter boundary around all data points known as a “minimum bounding polygon” (Figure 2). Concave hull allows the polygon to have angles greater than 180 degrees as might be needed if there are large bodies of water, forests, or road configurations that create holes within the map data that could skew measures of the activity area. Both concave and convex hulls exhibit the points, distance traveled, and shape of the spatial extent of total activities in the community. In contrast, a one SDE (1SDE) method represents 68% of a person’s activity data points. This method gives an average sense of the space a participant occupies, working under the assumption that a person does not go to the same locations every day.

Figure 2.

Overlay of a one standard deviation ellipse activity space in Geographic Information Systems (GIS) and corresponding convex hull activity spaces representing different types of activity locations Participant 3 visited (social, daily living, vocational) that contribute to the spatial extent. In this example, daily living locations were largely contributing to the spatial extent of the one standard deviation ellipse activity space.

Figure 1 shows 1SDE’s to compare total activity spaces between participants as a general representation of the participant’s space occupied in the community, while Figure 2 compares a 1SDE of total activity space and the convex hull approach to display the area occupied by separate types of activity categories (e.g., social, health, vocational). Comparing how much space and directionality the convex hull of each activity type contributes to the overall 1SDE activity space visually displays what types of activities account for the final activity space, and how a participant’s community participation is stratified. Directions on how to generate convex hull and 1SDE activity spaces in ArcGIS, and calculate activity space areas are provided in Table 1.

Regardless of method, activity spaces can be created based on time period (e.g. day, week) or activity type, as noted above. Viewing the activity space for each day can be useful in identifying patterns of activity and travel in the community, to note when during the week individuals are active, and variations in typical routines of travel. Variations and/or similarities between days can be extrapolated, allowing a better understanding of participation. The display of convex hulls for different days or activity types and corresponding points can also be color coordinated to enhance the display and allow for comparisons, set in “Properties,” through “Symbology.” These results also allow identification of patterns, and variances in patterns within and between participants such as differences based on mobility level, employment status, or use of public or private transportation. In all cases, the area of each activity space can be calculated in ArcGIS (see Table 1) and then used as a descriptive variable or outcome measure in further analysis. One final note, in all cases, more than two data points are needed to create activity spaces, otherwise the result will be a straight line with no area to measure.

Creating .csv Files and Plotting Points

To create activity spaces, location names, addresses, latitude and longitude coordinates, and purpose distinctions can be compiled in an excel spreadsheet, similar to creating the data for the accessibility spreadsheets. For a composite map or activity space, duplicates should be removed to represent unique locations visited in the community. The final version that includes all locations should be saved as a .csv file. For Figure 2, separate .csv files were created for each distinct activity type (daily living; social; work) with the home location included in each file. These were then individually created as shapefiles using the latitude/longitude coordinates consistent with the corresponding geographic coordinate system (e.g., WGS 1984) and added as layers on top of the base map.

GPS Participation Tracking

The data for creating activity spaces as a measure of participation can be drawn from travel diaries or participatory mapping. These methods capture the participant’s subjective perception of community and important locations. GPS tracking devices offer an objective method of assessing participation during a determined study period. GPS trackers record the latitude and longitude coordinates of locations visited within the community environment, as well as the time stamp of each point. In addition to mapping activity spaces and accessibility to resources, specific paths of travel can be mapped in GIS, and distinguished by day of the week or activity type (Figure 3).

Figure 3.

Geographic Information System (GIS) map of travel paths for Participant 3 using Global Positioning System (GPS) tracking for a 7-day period, with colors used to distinguish days of the week. In this example, the map shows the participant was fairly active during the week, but did not leave home on the weekend.

While the use of GPS tracking is emerging in research with disability populations, there are a number of factors to consider including type of tracking device, frequency of data points collected, accuracy of devices, battery life, and duration of study period. Since many of these considerations relate to the type of tracking device selected, a comparison of three main types of devices----GPS data loggers, personal GPS tracking devices, and smartphone applications----is provided, with information regarding the other factors noted contained within the descriptions.

Data Loggers

GPS data loggers can be given to participants at the beginning of the study period to record and store geographic data within the device. When the participant returns the device after the study period, researchers can retrieve the data gathered and upload it into study computers. The primary advantages of using data loggers include the accuracy of the data collected, data storage capacity, and battery life. Unlike other GPS devices, data loggers have an extended battery life. This may diminish concerns of missing data due to depleted batteries, as well as the need to learn and maintain charging schedules, which may be an important consideration for older adults and individuals with cognitive impairment. However, since the data cannot be viewed until after the study period ends, there is the possibility of missing multiple days of data if the participant forgets to carry the device, which will be unknown to the researcher until the device is returned. In addition, data loggers are generally somewhat small (similar to a deck of cards or garage door opener) but may not easily fit into one’s pocket.

Personal Tracking Devices

In contrast to the data loggers, personal GPS devices provide real-time tracking of the participant’s movement throughout the community, which can be viewed through an online secured website from the GPS carrier. Personal GPS devices are small and meant to be carried by the individual in a pocket, or by clipping them to a belt loop, key ring, backpack, or purse. They can also be slipped in a pouch which can be carried or clipped on the person. The availability to view whether participants are actively carrying and using the devices can be a critical advantage in attempting comprehensive data collection. These tracking devices can also send real-time e-mail or text notifications when the battery is low, or when participants leave or enter designated areas, such as around the home or a specific community resource. Aside from the ability to check in on data collection, real-time tracking also allows the researcher to review the data collected during the study period and prior to the device being returned to prepare for a follow-up visit if applicable. Another advantage is personal GPS devices are becoming more inconspicuous, often looking like key fobs or fitness trackers, and most mapping interfaces to view the tracking output are integrated with familiar platforms such as Google Maps API that can decode coordinate points into semanticized and contextualized locations.

Despite these advantages, some considerations of personal trackers include battery life, frequency of time points, and data accuracy. Most current GPS tracking devices run off of 4G data networks, and similar to cell phones, battery life decreases more quickly with activity. When the individual is inside a building, the tracking device can either go to sleep and conserve the battery until the signal is emitted again when leaving the building, or continue to track minute movements while inside a building, draining the battery before the end of the day and potentially not collecting data. A primary factor in battery life is the frequency of data tracking. Some providers allow for specification of how frequently data will be recorded, for example, in increments ranging from every 30 seconds, 1–2 minutes, 10–20 minutes, or tracking a specific number of times per day. Researchers will need to decide what frequency of tracking is sufficient for the purposes of the study, which may need to include consideration of the frequency of any other tracking devices used, such as heart rate or accelerometry, if the data will be paired. In addition to these concerns, devices often need to be charged every night, and following charging, remembered to be taken when leaving home the next day.

A second consideration is the accuracy of the data. Unlike data loggers, GPS signals actively emit signals to various carriers’ satellites which can at times result in errant data points or points “bouncing” around the actual location. Most devices provide a measure for accuracy of the tracking, for example, within 10 feet, but this may be in ideal conditions, in suburban or flat areas, and with clear weather. Finally, tracking devices may lose accuracy of tracking over time, even with available updates from the provider, and replacement devices may need to be purchased during the project period.

Smartphone Applications

Similar to personal tracking devices, GPS tracking through smartphones allows for real-time tracking of locations visited through an inconspicuous device that can easily be carried in one’s pocket, purse, or backpack. Considerations of battery life and remembering to bring the device are similar challenges, but a significant strength of app-based trackers such as Google Timeline or smartwatches is the accuracy and continuity of data tracking. However, if creating separate user accounts specifically for research use, confidentiality and Institutional Review Board approval of capturing home address locations as protected health information (PHI) on a less secure platform than an encrypted private service may present challenges. Assurances of keeping other identifying information and research data separate from geographic PHI may help. Privacy considerations in using tracking devices are described below.

Considerations and Potential Limitations

GIS and GPS in rehabilitation research has potential benefits, but there are also many considerations in planning and designing a research project using these methods. For GIS, the main considerations are interrelated factors of cost, time, and expertise. Further considerations when using GPS are the reliability of the device and data, as previously noted, and privacy.

Cost

The primary budget considerations when considering using GIS are related to software. GIS software such as ArcGIS may be acquired or accessed through an affiliated university, if available, or downloaded through open source platforms. Free alternatives to ArcGis are available. Popular programs include QGIS, DIVA-GIS, and Geoda (Anselin, 2014; Anselin et al., 2006; Graser, 2013). These options have most of the functionality of ArcGIS required for rehabilitation research; however, software choice should be determined based on the intended and required use. As with other software, updates may be needed regularly, and care must be given in transferring maps-in-process to updated versions to maintain links to data sources or layers constructed. An additional cost that may be incurred is related to obtaining needed data layers. Although many data layers are available for free, other layers may be nonexistent or restricted to certain organizations that may sell access for a price.

For GPS, financial considerations include the cost of the GPS tracking device(s), SIM cards if not included, and monthly service costs for devices that use real-time tracking. GPS data loggers can cost $30-$80, while real-time GPS tracking devices can range from approximately $40-$150. Although tracking apps such as Google Timeline are free, if providing smartphones, watches, or fitness trackers for use as part of the study instead of participant devices, those costs must be included in the budget. In addition, both the GPS tracking devices and smartphones have monthly service costs. For example, service for GPS devices can range from $13 to $25/month. In some cases rates are dependent on the frequency of tracking needed, with higher frequency corresponding with higher rates, but rates may be lower if prepaying for service over a 6- or 12-month period. A final consideration is staffing costs, to employ research assistants (RA) on projects, or consulting with experts. Working with GPS and GIS can be a time intensive project resulting in the need for RAs and consequently staff funding.

Time

GPS tracking requires more contact points (check-ins, follow-up interviews when applicable) than a survey would typically necessitate, increasing the role of research staff. With check-ins, staff may help remind the participant of charging the device when needed in a GPS study, or track down specific locations identified by participants in travel diaries or participatory mapping for GIS only studies. Additionally, the GPS device is not always reliably accurate and may record random pings, creating conflicting data. Using concurrent travel diaries, EMAs, and/or a post-tracking participant interview is recommended to improve the accuracy of the data collected (McCluskey et al., 2012). These supplementary features can add burden to the participant, including increased time and potential transportation costs if follow-up is needed. Increased time and responsibility of the research staff for cross-referencing, interviewing, transcribing, and/or data cleaning should also be projected.

Cleaning the data and extracting the relevant data points can also be time consuming. For GIS, available data may not be as specific as needed (i.e., categorizing convenience stores with the SIC for “grocery stores”) or flawed with duplicates, which take time to clean to have an accurate representation of features intended to be measured in the geographic area. With GPS, time and training may be needed when deciphering between random GPS pings and purposeful movements within a general area, although newer devices offer built-in algorithms in the tracking software that calculate and differentiate location stops versus travel.

Expertise

Expertise in GIS serves to minimize many challenges affecting efficiency and productivity in the research. A quick search of a university’s course catalog can determine if academic courses in GIS are available for research staff to take to gain familiarity and expertise with the mapping software. Researchers may benefit from taking these courses if offered at their university, as they can better train their RAs and troubleshoot potential problems. It is also helpful to search university websites with “GIS” or “GPS” as search terms to identify potential collaborators or consultants from other departments who are using these methods, or other sources of assistance, for example, through the library or statistical/research supports on campus. Consulting costs can be added to the budget when anticipated, and can allow for more advanced spatial analysis of data.

Many GIS software companies also offer free online classes and online or printed guides that take the reader step by step through using the software. Environmental Systems Research Institute (ESRI) has a series of self-study GIS Tutorial workbooks to learn the basics of map building, using ArcGIS tools, and analyzing spatial data, with accompanying exercises to practice (see Allen, 2016; Gorr & Kurland, 2016). YouTube offers brief videos that can guide researchers through using specific functions of mapping software and step-by-step directions to follow (see Table 1 for examples). In our studies, watching a variety of available video demonstrations was more helpful for the RA than reading instructions written for more advanced ArcGIS users.

Privacy

Privacy concerns at the individual and IRB level for collection of PHI of home address and routine locations visited should also be considered. IRBs are becoming more accustomed to many wearable devices in rehabilitation and health research such as accelerometers, FitBits, cameras, and smartphones with apps that include geotagged information. While this is an evolving field, recent examples of COVID-19 contact tracing apps or devices reflect emerging technology where health overrides privacy concerns. Efforts to increase privacy and security include accessing tracking data through a secured login that is password protected. Some GPS tracking companies, such as Pocketfinder, include a proprietary web interface featuring military grade encryption, and data captured is no longer visible after 2 months.

Following the recommendations of Nebeker et al. (2016) on improving the privacy, informed consent, and data management of PHI collected in research by wearable devices, including GPS trackers, the consent form should clearly state what type of data is being collected and tracked, how frequent this information is collected, who has access to view and retrieve the data, and the option to remove, stop, or leave the device at home for any activities they do not want captured. If a follow-up interview is completed, the researcher can allow the participant to omit locations or activities they do not want recorded. Providing an example of the output of how the information will be used as part of the consent process, such as a GPS destination and mobility map or GIS activity space map, is also recommended to increase transparency in the study process. Of note, in the same study, for the majority of participants privacy was not the main concern, but discomfort in wearing devices around the wrist or neck (Nebeker et al., 2016).

Case Examples

The case examples presented in Figures 1–3 are based on GPS and GIS data from three young adults with autism spectrum disorder and IQ’s over 70 who participated in 1-hour interviews before and after the 1-week GPS tracking period. All participants lived with their parents in a moderately sized southern city built around a military base. Participant 1 was a 29-year-old female who did not drive, and was not working. Figure 1 shows she visited just two locations during the 1-week study period and had the smallest activity space of the group (0.012 sq mi). The GPS data also recorded that she spent an average of less than 1 hour a day away from home during the week, leaving just 1 day to go to a gas station and to the movie theater. She scored an 86 on the Social Responsiveness Scale (SRS), which assesses social skill impairment in areas of social awareness, communication, motivation, and cognition, as well as restricted interest and repetitive behaviors for individuals on the autism spectrum. Scores of 76 or higher indicate severe social impairment associated with autism. When asked if she felt a part of her community, she responded that she felt part of her online community only.

Participant 2 was a 29-year-old male who drove his family’s car independently and was not working but was involved with Vocational Rehabilitation and hoping to get a job soon. He visited 11 unique locations during the week and had the largest activity space (34.54 sq mi) of the group. GPS data showed he spent an average of slightly under 3 hours per day away from home. He spent time out in the community every day except Sunday. When examining his locations, unlike the other participants, he visited slightly more social/recreational locations (6) than daily living locations (5). He had the lowest social impairment score of the group (SRS = 78). He reported he did feel a part of his community, in a small way, and had a small number of friends.

Participant 3 was a 28-year-old male who was not driving but was working part-time as a food sanitation worker at the military base, 6 hours per day. He visited seven unique locations during the week and also had a large activity space (26.23 sq mi) based on traveling to his work and several daily living locations (Figure 2). He spent the most time away from home, an average of over 6 hours per day, largely accounted for by his work schedule, but did not leave home on the weekend (Figure 3). He had the highest social impairment (SRS = 90) of the group. He stated he felt like he existed in the community, but did not feel a part of it.

DISCUSSION

Similar to our case examples, for many adults with physical, developmental, and psychiatric disabilities, community participation, whether through employment or even leisure- and recreation-based activities is limited (Frisch & Msall, 2013; Gray et al., 2014; Yanos et al., 2012). Questions remain regarding how where one lives impacts activity and participation for people with disabilities, and how one’s environment plays a role in these experiences. Evaluating the impact of neighborhood characteristics is particularly important for individuals with disabilities who struggle with social isolation and meaningful community involvement (Byrne et al., 2013). The size of community where one lives and geographic accessibility to resources may be important environmental considerations that can shape participation and service use, but are rarely studied. This article introduces how rehabilitation researchers can assess the environmental context where individuals live using GIS and GPS mapping technologies, and how to incorporate specific measures of environmental variables. Although this article focused on how consideration of accessibility to resources may relate to service use, employment, community participation, and health outcomes in people with disabilities, applications of mapping technology continue to grow. Improving health and participation outcomes means revealing where environmental resources are underutilized or simply unavailable, thereby contributing to disparities in health and mental health recovery at the individual and community level.

When access is not available, rehabilitation counselors have long advocated for physical changes in the environment to increase environmental accessibility for people with disabilities. This increase in environmental accessibility is associated with increased participation. An ongoing challenge in the field is to find valid and meaningful measures of community participation for adults with disabilities. Current measures may not capture how individuals are spending their time in the community, how much time individuals are away from home, how far individuals are traveling in the community for different types of activities, and what types of activities are most important. Combining GPS tracking and GIS mapping with measures of survey, interview, and travel diary information can help address the challenges of capturing environmental context in rehabilitation research. GPS data trackers, loggers, or apps can be used to objectively capture time spent away from home, distance traveled, and number of locations visited, while GIS mapping of the home locations and availability of community resources and services describe the characteristics of the environment where activity occurs. While research is limited, GIS and GPS technologies show promise as valid and systematic measures of environmental variables related to community participation and access to resources (Brusilovskiy et al., 2016; Chan et al., 2014).

Implications for Research

These methods combine with other measures to provide a rich, more complete picture of rehabilitation outcomes. Rehabilitation researchers studying both person and environmental factors contributing to disability have more options in assessing the environmental context through these methods. The case examples demonstrate the complexity of factors that contribute to participating in the community, and feeling part of it, such as transportation, social function, and being present in the community area. GIS and GPS data helped inform the size of one’s spatial presence and time spent in the community and visually allowed comparison of these factors to facilitate understanding different aspects of community integration. For example, Participant 3 was the only participant employed and spent the most time in the community and away from home, but he did not feel part of the community. A closer examination of his data revealed he primarily spent time at work and running daily living errands as a passenger with others. Although Participant 2 was not employed yet, he was able to drive independently, spent time in the community nearly every day, and spent the most time at social/leisure locations, which may have contributed to him feeling more part of his community. These examples support the need for multiple data sources and a multimodal assessment approach to understand the complex objective and subjective factors contributing to participation.

Further investigation is needed using these technologies to determine whether greater accessibility to resources will correspond with more time spent in the community, higher service use, and better quality of life, and whether functional outcomes moderate this relationship. Increased understanding of the specific role of the environment and specific resources in the community that are significant to participation can lead to recommendations for policy makers and service providers regarding aspects of the environment that can be changed to improve community activity and participation for adults with disabilities.

Limitations

The information provided is intended to guide researchers in visualizing rehabilitation data by mapping locations where outcomes and activities occur, and incorporating outcomes generated by spatial data measures into traditional analyses. It is important to note, however, using spatial data violates the traditional statistical assumptions of normality and equality of variance. Waldo Tobler’s first law of geography is “everything is related to everything else,” but near things are more related than distant things (Tobler, 1970; p. 234). Instead of assuming variables are independent, there is the assumption of spatial dependence and spatial autocorrelation, which is the correlation between the value of a variable at two different locations. Similar proximity to the same resources would likely have more similar than varying results, and observations with high or low values of a variable tend to cluster in a specific area. Consideration of these potential limitations is needed in the analysis plan and interpretation.

Conclusions

The use of GIS and GPS in health research is expanding to improve understanding of health and mental health outcomes. Rehabilitation researchers can benefit from these technologies and incorporate consideration of additional environmental factors that may be impacting resource accessibility and community participation for people with disabilities. Knowing activity patterns and actual use of specific resources captured by GPS tracking data and visually mapped in GIS can assist in service planning and advocacy for individuals in the community. With continued use, GIS and GPS technologies hold potential for identifying factors related to the environment that promote service use, community functioning, and quality of life.

Biography

Dara V. Chan, ScD, CRC, is an Associate Professor in the Division of Clinical Rehabilitation and Mental Health Counseling at The University of North Carolina at Chapel Hill. She earned her doctorate in Rehabilitation Sciences from Boston University. Her research focuses on using mixed-methods, including GIS and GPS technologies, to assess community integration and participation patterns in adults with disabilities, and how accessibility to resources impacts participation.

Adam Mann, MS, is a graduate of the Clinical Rehabilitation of Mental Health Counseling Program at the University of North Carolina at Chapel Hill. Currently, he is a clinical psychology graduate student at the University of Toledo working under the mentorship of Dr. Kim Gratz.

Sucharita Gopal is a professor in the Department of Earth & Environment at Boston University. Her research is multidisciplinary dealing with spatial analysis and modeling, GIS, data mining and informationvisualization, fuzzy inference, and artificial intelligence (AI). She has applied spatial analysis and GIS to address a variety of problems in public health including healthcare accessibility, malaria incidence, Spatio-temporal analysis of opioids distribution, and climate change impacts on health and wellness. She has published in other fields including urban planning, marine biology, sustainable finance, and development. Sucharita Gopal