Examining Environmental Predictors of Community Participation for Adults with Autism Spectrum Disorder Using Geographic Information Systems (GIS)

Abstract

A geographic information system (GIS) approach systematically assessed whether population density and distribution of community resources contributed to caregiver reported community participation outcomes for 124 adults with autism spectrum disorder (ASD). Regression analyses examined whether GIS measures predicted community participation in areas of social activities and use of services, while also accounting for adult age, conversation ability, and daily living skills (DLS). Results indicated that in addition to person factors of greater DLS and better conversation ability, access to specific community features, such as bus stops, contributed to improved participation. Unexpectedly, population density where one lived made minimal contribution to participation outcomes, except in getting together with friends outside of organized activities.

Introduction

The limited research on adults with Autism Spectrum Disorder (ASD) reports poorer community participation, including social interaction, paid employment, and independent living, compared to both neurotypically developing peers and those with other disabilities (Liptak et al. 2011; Orsmond et al. 2013; Roux et al. 2013). To date, most research examining predictors of community participation has focused on individual factors such as intellectual functioning, language ability, or diagnostic severity (Howlin et al. 2013; Taylor et al. 2015). Consistent with the International Classification of Health, Function, and Disability model (World Health Organization 2002), there is now a call to look beyond individual factors impacting adult outcomes to explore environmental factors that may promote independence and social inclusion in adulthood (Gray et al. 2014; Kirby et al. 2016; Myers et al. 2015). Based on prior research with the neurotypical population, two environmental factors, the population density of where adults live and the accessibility of community resources, may impact community participation by influencing the availability of programs, services, supports, and their frequency or ease of use (Kenney 2012; Mouratidis 2018; Nordbø et al. 2019). Existing literature on adolescents and adults with ASD show participation outcomes are often dependent on access to services (Levy and Perry 2011).

Investigations into environmental factors affecting access to services or participation in community activities are primarily based on self-reported barriers to use, such as a lack of transportation or attitudinal barriers (Badia et al. 2011), and not objective assessments of the community environment. Geographical Information Systems (GIS) are computerized mapping programs used to create, manage, and display geographic data (de Smith et al. 2007) that provide a unique method of analyzing the population density of where one lives and proximity to resources. The purpose of this brief report is to present findings of a study using a GIS approach to systematically examine the contribution of population density and distribution of community resources to community participation outcomes for adults with ASD.

Methods

Participants

For this study, data were used from a larger longitudinal study assessing adult outcomes of 274 individuals who were diagnosed with ASD as children between 1969 and 2000 at a university-based autism center. This larger study surveyed caregivers of these adults with ASD regarding a wide variety of outcomes. A subset of this sample who still resided in the state of North Carolina and had address data available for the adult was used for the current study. Data from 150 surveys were excluded because an exact address could not be identified or the adult currently lived outside of the state. The final sample for the current study included outcomes reported by 124 caregivers of adults with ASD (aged 21–54, M = 34.8 years, see Table 1).

Table 1.

Demographics of the sample of adults with ASD (N = 124)

Demographics
Age	34.8 years (range: 21–54)
Male	81.1%
Race
White	65.1%
African American	31.1%
Other	3.0%
Diagnosed with intellectual disability	56.6%
High school completion status
Regular diploma	44.3%
Certificate of completion	41.5%
Occupational diploma	5.7%
Dropped out/stopped going	4.7%
GED	2.8%
Employment status
Never employed	46%
Currently employed	39%
Previously employed, currently unemployed	15%
Living situation
With relative or guardian	74%
Supervised residential	21%
Independent	4%

Informed consent was obtained for the larger study from all participants included in the present study. The university’s Institutional Review Board approved all aspects of both studies.

Procedure

Caregivers completed an Autism in Adulthood Survey assessing adult outcomes in various domains, including community participation. The survey included 87 questions assessing demographic information, current developmental level including conversation ability and independent living skills, educational background, current living situation, social and recreational activities in the past year, and service use in the past 2 years. Person level predictors of interest in the current study included adult age, conversation ability, and daily living skills (DLS) score, as measured by the Waisman Activities of Daily Living Scale (Maenner et al. 2013). Conversation ability was assessed using a question from the National Longitudinal Transition Study 2 (NLTS2; Cameto et al. 2000), a survey of adolescent and adult outcomes for individuals who received school-based special education services. This measure has been used in other studies of outcomes for adults with ASD (Roux et al. 2013). Specifically, caregivers were asked “How well does your adult son/daughter carry on a conversation?” Ordinal responses included “Has no trouble,” “Has a little trouble,” “Has a lot of trouble,” or “Doesn’t carry on a conversation at all,” in which a higher score indicated greater conversational impairment. Questions of interest in the current study related to community participation included social and recreation activities (six yes/no questions), and total number of services used in the last 2 years since use of resources in the community is an aspect of community participation. Social and recreational activity questions in the Autism in Adulthood survey also closely matched information queried in the NLTS2 regarding activities with friends, volunteering, attending religious activities, attending social skills groups, and taking classes (art, music, dance, computers). These aspects of participation were of particular interest as living in a more urban or rural area could affect availability and accessibility of social opportunities, supports and services.

Environmental Measures

Accessibility

Accessibility to community features was measured two ways: (1) proximity, or closest distance to the identified feature, and (2) density, or number of features within a specified distance, from the participant’s home location. Using ArcGIS software v.10.3, each participant’s home address was plotted using the corresponding latitude and longitude coordinates and added onto a pre-generated county map of North Carolina retrieved from the NC Map Database. Although caregivers completed the survey, if the adult was not living with the caregiver at the time of the survey, the actual home location where the adult lived was used.

Locations representing access to different types of resources in the community related to healthcare, transportation, daily living, and social/recreational activities were included in the analysis. Data layers containing the location of five types of community features (grocery stores, coffee shops, hospitals, bus stops, and religious organizations) were constructed in GIS and added to the maps of the community areas where participants lived (see Fig. 1). Data layers for these community features were created using InfoUSA which identified the presence of business locations within the Standard Industrial Classification (SIC) code for each category of community features for a specified radial distance from a given address, and provided corresponding address and latitude/longitude coordinates. Bus stops were obtained from various county transit systems when available or constructed manually through Google Maps.

Fig. 1.

Example of participant home location mapped in ArcGIS with community features within 0.5, 1, and 2 mile radii

Population Density

Point population density by square mile was determined using a pre-generated 2010 Census tract data map retrieved from www.census.gov. Census tracts, as defined by the U.S. Census Bureau (2019), “generally have a population size between 1200 and 8000 people, with the optimum size of 4000 people.” Tracts of different square miles were created using the GIS population density function (Fig. 2). Point population density was measured and recorded for each participant’s address.

Fig. 2.

Population density map in GIS with participant home locations, and an example of a densely populated area zoomed in

Analysis

Regression analyses examined whether GIS measures of population density and accessibility to identified community features predicted community participation and total service use variables, while also accounting for person variables. All analyses were completed using SPSS v.26. We expected that measures of community features would have high collinearity, we therefore adopted a two-step analysis strategy. For the initial analysis for each dependent variable, backward regression was run, thereby identifying the GIS variables that accounted for the most unique variance. This was done to focus the analyses on those community features that accounted for unique variance in our community participation measures. After these variables were identified, hierarchical analyses were conducted. In the first step of each hierarchical regression, person variables were entered as predictors; these person variables included (1) age, (2) current conversation ability as a proxy for current functioning level, and (3) adult DLS. Next, population density was entered to account for broad factors associated with the number of persons in the environment. Finally, the significant variables from the original backward regression analyses were entered. By doing this, we tested whether these GIS variables had a significant impact on the dependent variables above and beyond the effects of the person variables and population density.

The social community participation outcome variables were dichotomous, indicating whether the adult with ASD did or did not participate in that activity in the past 12 months. Therefore, the analysis here was similar to that used for the regression analyses above but was completed using hierarchical logistic regression analyses. For these analyses, all predictor variables were standardized so that the odds ratio would represent the change in odds in the outcome variable that was associated with a 1 standard deviation change in the predictor variable. By standardizing predictor variables and reporting the odds ratio on these standardized predictor variables this analysis provides a measure of effect associated with a standard deviation change on a predictor variable as when standardized coefficients are reported (Kaufman 1996; Menard 2004, 2011).

Results

Participant addresses were located across the state of North Carolina including urban and rural areas (Fig. 2). GIS mapping determined how accessible each participant’s home location was to the five community features. Two accessibility measures were entered for each feature: (1) distance in miles to the closest feature in each category, and (2) number of features within each category at a specified distance, specifically bus stops within 0.5 miles, grocery stores within 1 mile, coffee shops within 1 mile, religious organizations within 2 miles, and hospitals within 2 miles (see example in Fig. 1). These distances represent the individual’s immediate and extended area that may be within walking distance to features, or accessible by transportation and reached within a short time. Table 2 presents the range of distances and densities of each feature found for the sample.

Table 2.

Average proximity of community features from participant home locations

Closest features	Distance in miles (range)
Religious center	0.84 (0.05–6.07)
Grocery stores	0.90 (0.03–3.22)
Coffee shops	2.66 (0.16–16.6)
Hospitals	4.65 (0.59–20.88)
Bus stops	11.93 (0.06–122.88)

Logistic regression analyses examined whether GIS measures of population density and accessibility to identified community features predicted adult outcomes. First, we examined the relation between GIS variables and community participation in social activities. For all analyses, we first entered three person factors (DLS, conversation ability, age), then we entered population density, followed by community feature variables. For volunteering or engaging in community service activities, person factors were significantly related to volunteering, χ²(3) = 11.23, p = 0.01, with better DLS associated with greater probability of volunteering (Odds Ratio [OR] = 1.72). This indicates that as DLS increases by 1 standard deviation, the odds of volunteering increases by 72%. Population density was not related to volunteering, χ²(1) = 1.19, p = 0.28, OR = 1.23. Finally, one GIS variable, the number of bus stops within a half mile, added marginally to the model, χ²(1) = 3.02, p = 0.08, OR = 1.50, with more bus stops leading to more volunteering. Next, for attending religious services, person factors were significantly related χ²(3) = 11.42, p = 0.01, with DLS accounting for most of this relation (B = + 0.75, p = 0.003, OR = 2.12) with higher DLS associated with increased attendance of religious services. Population density was not related to religious service attendance, χ²(1) = 0.49, p = 0.48, OR = 0.87. Two GIS variables, the number of bus stops within a half mile and the number of religious organizations within 2 miles, added significantly to the model, χ²(2) = 16.72, p < 0.001. The number of religious organizations was positively related to attending services (B = + 1.04, p = 0.002, OR = 2.83), while the number of bus stops was negatively related to attending religious services (B = – 0.91, p = 0.003, OR = 0.40).

Next, two activities related to friends were examined, getting together with friends outside of organized activities, and being invited to friend’s social activities. Both of these activities were related to person factors, χ²(3) = 27.77, p < 0.001 and χ²(3) = 14.28, p = 0.003, respectively with those who had better DLS being more likely to get together with friends (B = + 0.97, p = 0.002, OR = 2.64) and those with better conversation ability more likely to be invited to friends’ activities (B = + 0.81, p = 0.004, OR = 2.26). Population density was related to getting together with friends, χ²(1) = 5.21, p = 0.02 OR = 1.65 but was not related to being invited to friends′ social activities, χ²(1) = 2.18, p = 0.14, OR = 1.34. Each was also related to GIS variables, χ²(2) = 14.37, p < 0.001 and χ²(2) = 10.65, p = 0.005, respectively. Getting together with friends was positively related to number of bus stops within a half mile (B = + 0.98, p = 0.01, OR = 2.66) and was negatively related to the number of religious organizations within two miles (B = – 1.1, p = 0.002, OR 0.33). Being invited to other friends′ activities was positively related to the number of bus stops within a half mile (B = + 0.79, p = 0.02, OR = 2.21) and was negatively related to the number of grocery stores within a mile (B = – 0.71, p = 0.01, OR = 0.49). Attending social skills groups or taking lessons (e.g., art, music, dance, computers) was not predicted by either population density or accessibility to resources.

Finally, we examined whether GIS variables were related to total number of types of services used. Here, the person variables explained 18% of the variance in services used, with conversation ability accounting for most of this effect (β = – 0.41, p < 0.001). Population density was not related to number of services used, F(119) = 0.15, p = 0.90. The only GIS variable that explained additional variance was the density of bus stops within a half mile of the person’s residence which accounted for 4% of the variance in service use, F Change (1,118) = 5.63, p = 0.02.

Discussion

Examining the role of environmental factors in community participation outcomes is a promising new line of research, as they may be more amenable to change (Myers et al. 2015; Tobin et al. 2014). The current study considered the influence of both population density and accessibility to resources within the community as predictors of community participation. Current findings show that while person factors of better conversation ability and stronger DLS significantly predicted better participation outcomes, population density largely did not. Overall, higher daily living skills was a significant predictor for many aspects of community participation. Additionally, those from more and less densely populated areas had similar outcomes across participation in social activities and service use. While an urban setting may seem to provide more opportunities for social engagements, as seen in the follow-up of well-known case study Donald T., adults with ASD can be well integrated into small, rural communities where they are known and accepted by community members (Donavan and Zucker 2010).

Accessibility to specific features in the community, particularly access to public transportation and religious organizations, however, was significantly related to a variety of outcomes; in most cases those with better access to these community features had better outcomes than those with poorer access. For example, greater access to religious organizations in the community was related to increased involvement in community activities of attending religious services, while greater access to public transportation was related to volunteering, being invited to friends′ activities, and getting together with friends outside of organized activities or groups. Accessibility to public transportation was also a significant predictor of a greater number of support services in the last two years.

The findings from the current study suggest it is not population density but access to specific community features such as public transportation and religious organizations that may be notable environmental considerations impacting social participation and service use for adults with ASD. While the specific access to bus stops may provide the means to reach locations in the community associated with better inclusion (Sherman and Sherman 2013), overall it is unclear if the adults with ASD in the current study actually used public transportation to access any of these resources, which is a limitation of the study. In both cases, the current findings regarding the availability of public transportation and religious organizations may be more descriptive of the community itself (Barton and Gibbons 2017; Des Rosiers et al. 2010), as the majority of study findings related to the density of these features in the area, not how close they were to one’s home. Further research is needed into what specific resources adults are using in the course of social and recreational activities in the community, where in the community meaningful social activity occurs, and how adults are accessing services.

Limitations

Limitations of this study include the small sample from a specific geographic area, and the subsequently limited number of community features examined. In addition, based on the availability of addresses, the current sample under-represents adults in more restricted residential placement, but does reflect the focus of the study on adults who were living in the community. Findings related to the presence of different types of community features and the impact on adult community participation may not be consistent with communities in other geographic areas. For example, because data analysis was limited to individuals living in one southern state where the prevalence and culture of attending faith-based organizations is high, it is not known whether the impact of religious organizations would persist if examined in other geographic areas. Future research is needed comparing the impact of similar features across different geographic areas. In addition, there may be other community factors that significantly influence the outcomes studied that were not included in the current study due to the limited number of predictor variables that could be entered into the multiple regression models and collinearity of environmental measures. Although we attempted to include community features that represent key aspects of the environment for adults with ASD related to transportation, leisure, social, and health, it was not a comprehensive list. Future investigations using larger samples will permit a broader examination of community factors to overcome these limitations.

Conclusion

Consideration of accessibility to community features may be an important factor in assessing the environment where participation occurs and whether there is a person-environment fit that promotes thriving in the community for adults with ASD (Tint et al. 2016). As the numbers of adults with ASD increases and families focus on identifying factors associated with improved quality of life, these findings suggest that communities that provide opportunities for community participation via public transportation, and potentially religious organizations, may be important components in supporting positive adult outcomes. In addition to these community features, the daily living skills necessary to utilize these features is an important target for intervention.