Research in Developmental Disabilities 83 (2018) 108–119

Contents lists available at ScienceDirect

Research in Developmental Disabilities

journal homepage: www.elsevier.com/locate/redevdis

Place and community inclusion: Locational patterns of supportive

T
housing for people with intellectual disability and people with
psychiatric disorders
⁎
Yin-Ling Irene Wonga, , Yiyue Huangfub, Trevor Hadleyc
a
School of Social Policy & Practice, University of Pennsylvania 3701 Locust Walk, Philadelphia, PA 19104-6214, USA
b
Department of Sociology, University of Wisconsin-Madison 1180 Observatory Drive, Madison, WI 53706-1393, USA
c
Center for Mental Health Policy and Services Research, University of Pennsylvania 3535 Market Street, 3/F, Philadelphia, PA 19104, USA

A R T IC LE I N F O ABS TRA CT

Number of reviews completed is 2 Purpose: This study examines the locational patterns of publicly-funded supportive housing for
Keywords: people with intellectual disability (people with ID) and people with psychiatric disorders (people
Community inclusion with PD).
Intellectual disability Methods: Administrative data provided housing locations of 4599 people with ID and people with
Psychiatric disorders PD in one urban county and one suburban county in the United States. Census tract data captured
Spatial analysis neighborhood characteristics. Descriptive statistics and spatial analysis were used to analyze the
Supportive housing distribution of supportive housing sites.
Results: People with ID were more dispersed across a larger number of census tracts with smaller
number of residents per tract than people with PD. While spatial dispersion in favor of people
with ID was consistent across both counties, difference in dispersion was more pronounced in the
urban county. People with PD were concentrated in neighborhoods with more socio-economic
disadvantage, more residential instability, and a higher level of race/ethnic diversity than people
with ID.
Conclusion: This study suggests that spatial-analytic method can serve as a useful tool for as-
sessing the extent to which integrated housing is achieved for people with ID and people with PD.
Interpretation of findings should be given due consideration of the policy context and neigh-
borhood characteristics of the study communities.

What this paper adds?

This cross-county, cross-disability comparative study describes a geographic information system method to examine the re-
sidential locations of people with intellectual disability and people with psychiatric disorders living in supportive housing. While the
findings reflect the socio-demographic and housing profiles of the two counties under study and the development of supportive
housing for the two populations in the United States, the spatial-analytic method employed can be broadly applied to other com-
munities to enable policy makers and service administrators to implement housing-siting strategies that facilitate community in-
clusion.

⁎
Corresponding author.
E-mail addresses: ylwong@upenn.edu (Y.-L.I. Wong), hfelaine@gmail.com (Y. Huangfu), thadley@upenn.edu (T. Hadley).

https://doi.org/10.1016/j.ridd.2018.08.009
Received 22 December 2017; Received in revised form 5 July 2018; Accepted 20 August 2018
0891-4222/ © 2018 Elsevier Ltd. All rights reserved.
Y.-L.I. Wong et al. Research in Developmental Disabilities 83 (2018) 108–119

1. Introduction

People living with severe disabilities encounter multiple barriers in maintaining community tenure, which limit their opportunity
to fully engage in daily activities and assume social roles alongside community members without severe disabilities. In the United
States, since the advent of deinstitutionalization in the 1950s and in response to the community support needs of people with
psychiatric disorders (people with PD) and people with intellectual disability (people with ID), supportive housing has evolved as a
vital component of community-based service delivery systems with the goals of facilitating rehabilitation and fostering community
inclusion. Supportive housing refers to a collection of publicly-funded programs that couple housing with supportive services to
enable individuals with disabilities to maintain community tenure (Wong, Filoromo, & Tennille, 2007). Examples of supportive
housing include group residences, half-way houses or transitional housing, and supported independent living apartments (Finkler,
2014).
National trends toward providing housing in small-size supportive housing settings have been documented for people with ID in
the United States (Larson, Salmi, Smith, Anderson, & Hewitt, 2013); however, comparable data are not systematically collected for
state and local mental health systems for people with PD. An earlier study conducted in a metropolitan area comparing the two
populations found that people with ID lived in more integrated housing as indicated by smaller size programs and in neighborhoods
with more favorable features than people with PD (Wong & Stanhope, 2009). This finding poses questions regarding whether this
pattern is also evidenced in other jurisdictions with different socio-economic characteristics. Moreover, in the international context,
developing a method that examines the locational patterns of supportive housing for these two populations could help assess the role
of place in community inclusion in countries that vary in the extent and pace of deinstitutionalization.

2. Background

2.1. Deinstitutionalization of care for people with ID and people with PD

The process of deinstitutionalization originates from the 1950s to 1960s in Western Europe and North America.
Deinstitutionalization refers to the shift in practice of caring for people with disabilities from large scale residential institutions
(including long-term hospitals) to small-scale services in the community to enable them to maintain community tenure and attain
good quality of life (Mansell & Beadle-Brown, 2010; Shen & Snowden, 2014). In half a century after the 1950s, countries across the
globe vary in the degree and pace of which deinstitutionalization are implemented. For example, while deinstitutionalization for
people with ID is considered to be well advanced in Scandinavia, the United States, Canada, United Kingdom, and Australasia, people
with ID are mostly placed in large institutions in Central and Eastern European countries (Mansell & Beadle-Brown, 2010). In the field
of mental health, countries have been classified by timing of adopting the deinstitutionalization policy, including innovators, early
adopters, early majority, late majority and non-adopters (Shen & Snowden, 2014). As community inclusion is considered a key goal
and guiding principle of deinstitutionalization, evaluating the extent to which inclusion is achieved through delivery of community-
based services is a central concern for policymakers and researchers.

2.2. Supportive housing as an integral component of the community-based service delivery systems in fostering community inclusion in the
United States

Community inclusion refers to the opportunity to live in the community and be valued for one’s uniqueness and abilities, re-
gardless of the form and severity of disability an individual may experience (Salzer, 2006). A basic right solidified through legislation
in the United States, community inclusion is a guiding principle of the federal policy framework, stipulated in Title II of the 1990
Americans with Disabilities Act (ADA), affirmed by the 1999 Olmstead Decision, and promulgated in The 2001 New Freedom
Initiative (ADA, 1990; Olmstead et al. vs. L. C. et al., 1999; President George W. Bush’s New Freedom Initiative, 2001). In nearly three
decades, community inclusion as a policy goal has guided the development of state and local service systems for people with ID and
people with PD.
Housing, as a basic right and ubiquitous need, is a vital component of community-based services delivery for people with dis-
abilities. Although a significant portion of people with PD and people with ID can maintain community tenure in private, regular
residences when given access to services such as educational and employment supports, day programs, and case management,
individuals living with more severe disabilities require an intensive level of care provided at their homes.
Regardless of different supportive housing approaches adopted in services for people with ID (IDS) and services for people with
PD (PDS), there is evidence that quality housing coupled with supportive services is associated with positive resident outcomes,
including improving social functioning, reducing institutionalization, and achieving residential stability (Kozma, Mansell, & Beadle-
Brown, 2009; Rog et al., 2014). These findings lend support to the commonly-held assumption that given appropriate services suited
to their support needs, people with severe disabilities may participate as full community members in an integrated and normalized
living environment.

2.3. Major supportive housing programs in IDS and PDS

Since its onset in the 1950s, the process of deinstitutionalization in the United States has been influenced by an array of tech-
nological, legal, financial, and cultural factors (Bagnall & Eyal, 2016; Dear & Wolch, 2014; Mansell & Ericsson, 2013). Although

109
Y.-L.I. Wong et al. Research in Developmental Disabilities 83 (2018) 108–119

community-based IDS and PDS have evolved separately with little systematic exchange between program planners of the two systems
(Bachrach, 1981), similarities can be drawn regarding supportive housing for the two populations.
One common feature is the use of group homes as a key housing and service approach offered to people with ID, which has been,
until recently, a primary approach offered to people with PD (Nelson & Caplan, 2017). In the field of IDS, a group home refers to a
residence shared by two or more individuals, in which residential services staff provide care, supervision and other forms of support
on-site to maintain residents’ community tenure. A group residence is owned, rented or managed by a service provider, or by a
provider’s housing management agent. More than one-quarter of IDS users in the United States are living in congregate home settings
(Larson et al., 2013). In the field of PDS, supportive housing in a local community typically comprises supervised group residences, as
well as single-site and scattered-site apartments coupling long-term rental subsidies with support services offered by mental health
agencies (Nelson & Caplan, 2017).
With an increasing emphasis on community inclusion in the past three decades, supportive housing for both people with ID and
people with PD has focused on normalized living arrangements with in-vivo supports from specialized service providers. The em-
phasis on home-style living with supports, as opposed to segregated, large congregate care facilities, has largely been guided, re-
spectively, by the principles of normalization and social role valorization in IDS (Flynn & Aubry, 1999), and the empowerment/
integration housing model espousing right to home, resident choice and resident control in PDS (Parkinson, Nelson, & Horgan, 1999;
Ridgway & Zipple, 1990). More recently, the movement toward Housing-First as an evidence-based practice to address chronic
homelessness among persons with a mental health and/or substance use condition has spurred the growth of tenant-based, scattered-
site permanent supportive housing in local communities (Dickson-Gomez et al., 2017; Montgomery, Hill, Kane, & Culhane, 2013;
Tsemberis, Gulcur, & Nakae, 2004).

2.4. Integrated housing and neighborhood characteristics associated with community inclusion for people with ID and people with PD

Integrated housing refers to an environment in which housing is physically arranged in a manner that facilitates opportunities for
people with disabilities to engage in community activities and social interaction with neighbors. A key characteristic of integrated
housing is the scattering of housing units so that there is little clustering within one neighborhood. The raison d’être is to avoid
creating physical and “image” barriers for people with disabilities to community engagement (Hogan & Carling, 1992).
The extent to which supportive housing programs are located in integrated housing settings has scantly been examined in prior
research. There is documentation that state and non-state congregate care settings (i.e., group homes) for people with ID are located
in small-size homes. The most recently published nation-wide data found that 61% of people with ID living in group homes were in
small home settings with 6 or fewer residents, compared to 19% in large settings with 16 residents or more (Larson et al., 2013). Data
on program size for people with PD, however, have not been tracked in the mental health service delivery system nation-wide.
Regardless of data availability in the two service systems, program-size data alone provide insufficient information about integrated
housing because the extent of concentration of small-size supportive housing within neighborhoods is not known.
Prior research has examined the role that neighborhood characteristics play in community inclusion of people with PD and people
with ID. A comprehensive literature review on neighborhood effects found that availability of meeting places and amenities for
activity facilitates inclusion, whereas public aggressiveness and tense neighborhood relations erodes the sense of feeling at home for
people with ID (Overmars-Marx, Thomése, Verdonschot, & Meininger, 2014). A longitudinal study in the United Kingdom similarly
found that favorable perceptions of neighborhood characteristics, including safety and less crime, and higher levels of social and civic
participation to be predictors of positive self-rated health for people with ID (Emerson, Hatton, Robertson, & Baines, 2014). Research
on people with PD in supportive housing has consistently shown neighborhood satisfaction, including a sense of safety, security and
calmness, and proximity to amenities and community resources, to be associated with higher level of community inclusion (Chan,
Gopal, & Helfrich, 2014; Ecker & Aubry, 2016, 2017; Timko, 1996; Townley & Kloos, 2011; Wright & Kloos, 2007; Yanos, Felton,
Tsemberis, & Frye, 2007). These study findings across the two disability fields signal the need to examine objective community
indicators that contribute to people with ID’s and people with PD’s perception of their neighborhood and sense of satisfaction.
According to social disorganization theory, neighborhood disadvantage is indicated by low socioeconomic status (SES), re-
sidential instability, and racial heterogeneity (Sampson & Groves, 1989). It has been postulated that low SES limits the resources
available for community-based organizations, such as neighborhood watch groups and mutual help associations, for exercising formal
and informal social controls on community residents. Residentially unstable communities prevent the formation of friendship net-
works and associational ties, which could reduce crimes and increase a sense of safety. Racial heterogeneity promotes heightened
distrust among neighbors, which could lead to segmented and hostile communities. While the concentration of people with ID and
people with PD as large-scale facilities is antithetical to community inclusion, the socio-demographic profile of a neighborhood
represents an important consideration in service providers’ decision when locating residential sites for people with disabilities.

3. Research questions

This study examines the locational patterns of supportive housing for people with ID and people with PD. Based on administrative
data collected in 2011 from the human service systems of two counties in the United States, the current study focuses on the following
questions:

• To what extent are residential sites for people with ID and people with PD spatially dispersed, an indicator of integrated housing?
• Are residential sites for people with ID and people with PD located in neighborhoods with similar or different sociodemographic
110
Y.-L.I. Wong et al. Research in Developmental Disabilities 83 (2018) 108–119

characteristics that affect community inclusion?

• Are there urban-suburban differences in spatial clustering and neighborhood characteristics of residential sites for people with ID
and people with PD?

4. Methods

4.1. Study sites

The study was conducted in two adjoining counties, Montgomery County (MC) and Philadelphia County (PC) in Pennsylvania.
Among the 67 counties in Pennsylvania, MC and PC rank, respectively, third and first in population size; and fourth and first in
population density. Despite similarly high ranking in population density, PC, an urban county, is nearly seven times as dense as MC, a
suburban county. MC has the second highest median household income (US$79,183) in Pennsylvania, whereas PC has the second
lowest median household income (US$37,192). The poverty rate is 26.5% in PC (highest in the state), compared with 6.1% in MC
(second lowest).

4.2. Data collection and measurement

4.2.1. Data access and management

The study was conducted with the approval of the Institutional Review Board of the university with which the first and third
authors are currently affiliated. The research team invited administrators of the departments of IDS and PDS in the two counties to
participate in the study. The following data elements were obtained from the management information systems (MIS): 1) supportive
housing site addresses, including street number, street direction (if any), street name, and zip code; 2) types of program (group homes
and tenant-based supportive housing), and 3) the number of residents per address.
The MIS data were integrated into one common database per each county. Residential site addresses were matched with the
unique identifiers of census tracts in the 2006–2010 American Community Survey (ACS) Five-Year Summary Files. Socio-demo-
graphic and housing variables from the ACS data files for all census tracts of the 2 counties were extracted.

4.2.2. Units of analysis—census tracts

The units of analysis for constructing indicators of integrated housing and neighborhood characteristics are census tracts. A census
tract covers a contiguous area with a population size of between 1200 and 8000 people residing in communities with varying spatial
size according to population density. Because most census tracts are delineated by local residents as part of the Census Bureau’s
Participant Statistical Areas Program, it is appropriate to use census tracts as the units of analysis (U.S. Census Bureau, 2017).

4.2.3. Data processing

The study used a geographic information system to process the residential site data. Using ArcMAP 9.3, a desktop software
package, addresses in the 2 counties were matched with respective counties’ street files. After an address is identified, its “X” and “Y”
coordinates (i.e., specific latitudes and longitudes) were assigned and the address to a census tract identifier was matched. Supportive
housing residents located at all addresses within a census tract were aggregated by disability type and the aggregated data were
merged with socio-demographic and housing data from the ACS file.

4.2.4. Neighborhood characteristics

Socio-demographic and housing variables were extracted from the 2006–2010 ACS files. These included proportions of: 1)
households with income below the poverty line; 2) female-headed households with children age under 6; 3) one-person households;
4) households moved in the last year; 5) vacant housing units; 6) persons unemployed; 7) rental housing; 8) households with public
assistance income; and 9) persons over 25 not completed high school education. In addition, the proportions of persons who were
black, non-Hispanic white, and others (including Hispanics, non-Hispanic Asians and others) were extracted from the ACS file to
compute a diversity index for the race/ethnic composition in each census tract. The index of qualitative variation (IQV) scores
between 0 and 1, with a score of approaching 1 indicating high degree of diversity and a score of 0 indicating no diversity when
everyone in the census tract reported to belong to the same race/ethnic group (Frankfort-Nachmias & Leon-Guerrero, 2011).
Principal component analysis (PCA) with varimax rotation was conducted to identify factors that captured the neighborhood
conditions of the two counties. PCA results (Table 1) indicate that two factors—neighborhood disadvantage and residential in-
stability—captured the neighborhood characteristics of the two counties. For MC, the two factors explain 58.7% of the variance of the
data and the corresponding percentage for PC is 62.4%. There is dissimilarity with regard to one neighborhood characteristic that was
loaded into the two-factor solutions of the two counties. Proportion of vacant housing units was loaded into the “residential in-
stability” factor for MC, but the “neighborhood disadvantage” factor for PC.

4.3. Data analysis

4.3.1. Global patterns of spatial autocorrelation

First, we tested if there is spatial clustering of people with ID and people with PD in supportive housing residential sites in MC and
PC, that is, whether census tracts with high values of people with ID and people with PD tend to locate close to each other. Global

111
Y.-L.I. Wong et al. Research in Developmental Disabilities 83 (2018) 108–119

Table 1
Means, standard deviations, final commonalities, rotated factor loadings, and percent variance explained for the two-factor model.
Montgomery County
Variables Mean Standard deviation Final Communalities Factor 1 disadvantage Factor 2 instability

Proportion of households with income below poverty line 0.056 0.054 0.679 0.669 0.481
Proportion of female-headed households with children age 0.039 0.052 0.537 0.711 0.180
under 6
Proportion of one-person households 0.257 0.101 0.660 0.034 0.812
Proportion of households recently moved (2009-2010) 0.109 0.079 0.486 0.027 0.697
Proportion of vacant housing units 0.046 0.036 0.486 0.411 0.563
Proportion of persons unemployed 0.053 0.024 0.478 0.691 0.009
Proportion of rental housing units 0.244 0.186 0.855 0.417 0.825
Proportion of households with public assistance income 0.014 0.017 0.572 0.753 0.066
Proportion of persons without high school degree 0.075 0.059 0.530 0.680 0.259

Philadelphia County
Variables Mean Standard deviation Final Communalities Factor 1 disadvantage Factor 2 instability

Proportion of households with income below poverty line 0.255 0.157 0.747 0.828 0.249
Proportion of female-headed households with children age 0.125 0.112 0.257 0.491 0.126
under 6
Proportion of one-person households 0.370 0.113 0.707 −0.182 0.821
Proportion of households recently moved (2009-2010) 0.140 0.095 0.468 −0.051 0.682
Proportion of vacant housing units 0.133 0.075 0.469 0.584 0.359
Proportion of persons unemployed 0.134 0.070 0.653 0.780 −0.211
Proportion of rental housing units 0.436 0.187 0.808 0.300 0.847
Proportion of households with public assistance income 0.081 0.066 0.794 0.879 −0.144
Proportion of persons without high school degree 0.214 0.114 0.714 0.818 −0.213

Note. Number of census tracts = 211. Percent variance explained: 58.7%.

Note. Number of census tracts = 377. Percent variance explained: 62.4%.

Moran’s I statistics (Waller & Gotway, 2004) were calculated to detect the spatial clustering of people with ID and people with PD,
respectively, with the null hypothesis that residential sites for people with ID and people with PD in these two counties were
randomly distributed across census tracts. The null hypothesis is to be rejected if z-score is above 1.96 (p < .05).

4.3.2. Identify local hot spots

Although Global Moran’s I detected if there was any clustering of people with ID and people with PD in supportive housing, it did
not provide information about where the clusters occurred, the size of the clusters, and how strong the clusters were. We used the
Getis-Ord Gi* statistic (Getis & Ord, 1992) to gauge the location, size, and magnitude of the clusters. In calculating the Getis-Ord Gi*,
the local sum of people with ID and people with PD, defined as the total numbers of people with ID and people with PD in the focal
census tract and its neighbors, was calculated. The actual local sum was compared with the expected local sum, assuming that each
tract had the average number of people with ID and people with PD in residential sites. In this analysis, hot spots were those areas in
the two counties with Getis-Ord Gi* score above 1.96 (p < .05).

4.3.3. Conceptualization of spatial relationships

There are multiple ways to define whether a census tract is considered a neighbor for the focal census tract. Disk smoothing sets a
circular window of specified radius and an area is considered as a neighbor when the distance between the centroids of this area and
the focal area falls below the search radius (Casper et al., 2000; Wang & Varady, 2005). The first-order neighborhood contiguity rule
assumes spatial auto-correlation only among the areas that share a common boundary with the focal area (Mitra, Buliung, &
Faulkner, 2010). The K nearest neighbor option employs the general rule of thumb to evaluate each study area in the context of a
minimum of eight neighbors (Longley & Batty, 1996). The Delauney triangulation specifies natural neighbors for a set of areas by
creating Voronoi triangles from area centroids and nodes connected by a triangular edge are considered neighbors (Rossen, Khan, &
Warner, 2014).
In our case, there is considerable variation in the size of census tracts in both counties. Using the K nearest neighbor and first-
order contiguity methods will result in very small search areas for pockets of small census tracts and large search areas where the
census tracts are larger. On the other hand, to apply a circular window to all the census tracts in a county universally will result in
larger tracts or tracts on the county borders with few neighbors. To account for the heterogeneity in the size of census tracts, we used
a conceptualization of spatial relationships that takes into account both the size of tracts and the number of neighbors. First, we
calculated the average radius that each census tract has 8 neighbors in each county; then we kept this search radius for census tracts
that have 8 neighbors within this distance and expanded the search radius to incorporate 8 neighbors for those otherwise did not
have. For the smaller census tracts, mostly seen in Center City of PC, as well as Norristown, Abington, and Willow Grove of MC, the
search area was largely defined by the average distance of having 8 neighbors, thus avoiding too small the search areas resulted from
the K nearest neighbors or first-contiguity methods. For the larger census tracts and tracts on the county borders, this ensured every
tract was evaluated in a context of at least 8 neighbors.

112
Y.-L.I. Wong et al. Research in Developmental Disabilities 83 (2018) 108–119

Table 2
Descriptive statistics on the distribution of supportive housing residents in Montgomery County and Philadelphia County.
Number of residents/tracts and rate of residents Number of residents in tracts that had 1 residents or more

County/disability type Number of Number of tracts with Rate of SH residents (per Minimum Maximum Mean
residents residents 10,000)*

Montgomery
PID 479 77 6.06 1 20 6.2
PDD 508 56 6.43 1 63 9.1
Philadelphia
PID 1143 165 7.69 1 29 6.9
PDD 2469 110 16.61 1 104 22.5

The average distance that all the census tracts had 8 neighbors was 13,098 feet (about 2.48 miles) in MC and 5071 feet (about
0.96 mile) in PC. Using the method outlined above, the census tracts in MC had 11 neighboring tracts on average with the highest one
had 21, and the census tracts in PC also had 11 neighboring tracts on average with the highest one had 27.

4.3.4. Mapping analysis

To examine the neighborhood characteristics of supportive housing residential sites in the two counties, maps of hot spot tracts
were overlaid with those tracts that were ranked at the top quartile of scores on neighborhood disadvantage, residential instability,
and race/ethnic diversity, respectively. Given the number of maps involved in cross-county and cross-disability comparisons, only
four maps displaying the overlaps between hot spots and neighborhood disadvantage tracts are included in this paper. Corresponding
maps for residential instability and race/ethnic diversity are available by request to the first author. In all three neighborhood
variables, percentage distributions of hot spot tracts that were in the top quartile were reported. In addition, one hot spot area of
people with PD in MC and one hot spot area of people with ID in PC were featured in order to give an illustration of the characteristics
of communities where there were clustering of residential sites.

5. Results

Table 2 shows the number and rate (per 10,000 persons) of people with ID and PDD in supportive housing in the two study
counties. In MC, there were 479 people with ID and 508 people with PD in supportive housing, indicating a population rate of 6.06
and 6.43, respectively. The corresponding numbers of supportive housing residents (and rates) for people with ID and people with PD
in PC were, respectively, 1143 (7.69) and 2469 (16.61). Of the 211 census tracts in MC, supportive housing residential sites were
located in 77 tracts for people with ID (range from 1 to 20 residents in the 77 tracts with a mean of 6.2 persons) and in 56 census
tracts for people with PD (range = 1–63; mean = 9.1). Of the 377 tracts in PC, supportive housing residential sites were located in
165 tracts for people with ID (range = 1–29; mean = 6.9) and in 110 tracts for people with PD (range = 1–104; mean = 22.5).
The Global Moran’s I for people with ID and people with PD in MC were 0.14 (z-score = 4.56, p < 0.01) and 0.12 (z-score =
4.19, p < 0.01), respectively; and in PC were 0.4 (z-score = 18.94, p < 0.01) and 0.06 (z-score = 3.10, p < 0.05), respectively.
The null hypotheses that people with ID and people with PD were randomly distributed in MC and PC were rejected.
Fig. 1 includes four maps showing the distribution of supportive housing hot spots in MC and PC, marked in red; and census tracts
that were ranked in the top quartile of disadvantage scores, marked in a checkered pattern. The top left map and top right map show
the overlaps between hot spots and disadvantage tracts for people with ID and people with PD in MC, respectively. The bottom left
map and bottom right map show the corresponding patterns for people with ID and people in PD in PC, respectively. Counting the
number of supportive housing residents in the hot spots yielded the following results. In MC, 42.6% of all people with ID were in 4 hot
spot areas comprising 31 tracts, and 49.4% of all people with PD were in 5 hot spot areas comprising 24 tracts. The corresponding
distributions in PC are 53% of people with ID in 4 hot spots comprising 56 tracts and 29.5% of people with PD in 6 hot spots
comprising 30 tracts. Note that the largest hot spot area for people with ID in PC is located in the Northeast region covering 40 census
tracts. Maps indicating the number of people with ID and people with PD in each census tract are available by request to the first
author.
As Fig. 1 shows, neighborhood disadvantage is dispersed across different regions of MC; whereas in PC, it is concentrated around
the interior/middle section of the county (the neighborhoods of Upper North Philadelphia and Lower North Philadelphia). In MC, 9 of
the 31 (29%) hot spot tracts for people with ID were tracts in the top quartile of neighborhood disadvantage, compared to half of the
24 (50%) hot spot tracts for people with PD. In PC, 12 of the 30 (40%) hot spot tracts for people with PD were in the top quartile of
neighborhood disadvantage, whereas none of the hot spot tracts for people with ID in PC were in the top quartile.
Overlaying hot spot tracts with tracts in the top quartile of residential instability yielded the following results. In MC, while half of
PD hot spot tracts were ranked in the top quartile of residential instability, only 13% (4 of 31) of ID hot spot tracts were in the same
category. The corresponding distributions in PC were 70% (21 of 30) for PD hot spot tracts and 12.5% (7 out of 56) for ID hot spot
tracts. In terms of race/ethnic diversity, proportionally more PD hot spot tracts were included in the top quartile than ID hot spot
tracts. In MC, 54% of people with PD hot spot tracts were in the top quartile of IQV scores, compared to 23% of people with ID hot
spot tracts. The corresponding distribution for PC is 30% of PD hot spot tracts, compared to 9% ID hot spot tracts. Recall that cross-

113
Y.-L.I. Wong et al. Research in Developmental Disabilities 83 (2018) 108–119

Fig. 1. Distribution of hot spots areas overlaid with tracts ranked in the top quartile neighborhood disadvantage scores in Montgomery County and
Philadelphia County. Montgomery County people with ID hot spot Montgomery County people with PD hot spot. Philadelphia County people with
ID hot spot Philadelphia County people with PD hot spot.

114
Y.-L.I. Wong et al. Research in Developmental Disabilities 83 (2018) 108–119

county, cross-disability maps displaying the distribution of residential instability and race/ethnic diversity are available by request to
the first author.
Fig. 2 displays zoomed-in maps featuring two hot spots with clustering of people with PD in MC and people with ID in PC,
respectively. The top map highlights the concentration of health (general and mental) and social service facilities in Norristown, a
borough and county seat of MC. Located at the center of the hot spot is Norristown State Hospital (indicated by an orange triangle), a
long-term psychiatric facility that serves residents from five counties comprising the Southeastern Region of Pennsylvania, including
MC and PC. Central Behavioral Health (indicated by an orange circle), located at the lower right edge of the hot spot, is a multi-
service center offering community-based mental health services to residents in MC. The hot spot has a highway (US Hwy 202)
running close to its border, connecting residents to other parts of the county and Southeastern Pennsylvania. In addition, the hot spot
is close to a transportation hub of MC, Norristown Transportation Center (indicated by a green star), an important resource for
Norristown residents. Other facilities located in or close to the hot spot are one psychiatric emergency services center (located
southeast of Norristown State Hospital and indicated by an orange triangle), three general hospitals (indicated by black triangles),
and one generic multi-service center (indicated by a black circle). The Norristown hot spot is a prime example of concentration of
supportive housing residential sites for people with PD around mental health and health facilities with easy access to government
services and public transportation (Dear & Wolch, 2014).
The hot spot featuring clustering of people with ID in PC (bottom map of Fig. 2), located in the Roxborough-Manayunk area,
exemplifies a community that is characteristic of suburban living within an urban county. The West Northwestern Avenue and the
Schuykill River are borders (county lines) dividing PC from MC, with Whitemarsh and Lower Merion, two townships of MC located
outside the area’s northwestern border and western border, respectively. This hot spot is a single-family residential area with a main
thoroughfare, Ridge Avenue, running through the area. Retails stores, groceries, and restaurants (indicated by green circles) are
spotted in three clusters along Ridge Avenue and a cluster in the southern end of the hotspot, close to Schuykill River. In contrast with
the Norristown hot spot, there is no major mental health facility located in the area.

6. Discussion

This study examines locational patterns of supportive housing in two counties in the United States with contrasting socio-de-
mographic and housing profiles. The descriptive statistics (Table 2) indicate that residential sites for people with ID were geo-
graphically more dispersed across a larger number of census tracts with smaller number of residents per tract than people with PD.
Although the pattern of spatial dispersion in favor of people with ID is evident in both counties, the difference in the degree of
clustering is more pronounced in the urban county than in the suburban county.
Using hot spot analysis, we found that supportive housing for people with PD was concentrated in communities with more socio-
economic disadvantage, more residential instability, and a higher level of race/ethnic diversity than people with ID. As is in the case
of spatial distribution, differences in socio-demographic characteristics were more distinct in the urban county than the suburban
county. For example, while 29% of hot spot tracts for people with ID versus 50% of hot spot tracts for people with PD was included in
the top quartile of neighborhood disadvantage in the suburban county, the corresponding difference in the urban county was 0%
(people with ID) versus 40% (people with PD).
What are the factors that might have contributed to the different locational patterns of supportive housing noted in this study?
Although housing and community inclusion of people with ID and people with PD has been examined in their respective fields in
regions where deinstitutionalization has reached a more “advanced” stage, including Western Europe, North America and
Australasia, research that elucidates the similarities and differences of the two processes is still scant (Bagnall & Eyal, 2016;
Braddock, 1992). In the development of residential and community support arrangements, a focus on realizing the principles of
normalization and social role valorization (Flynn & Aubry, 1999), a commitment to full inclusion into the wider community (Racino,
1995), and an emphasis on creating a home-like, least restrictive environment as a right to habilitation (Annison, 2000; Taylor, 2004)
are the key driving forces toward promoting inclusion among people with ID in these regions. Specific to the IDS system in the United
States, comprehensive residential supports are available to enable the majority of people with ID to live with family members or in
their own homes, thereby scaling down the need for placing people with ID in supportive housing. In 2010, while about 30% of
people with ID were in supportive housing nation-wide, the majority (64.5%) were living with their family members or in their own
home (Bagnall & Eyal, 2016).
In contrast, research in the mental health field has showed that choices made by mental health administrators in locating sup-
portive housing with on-site staff supervision were largely guided by considerations of housing affordability and accessibility to
services and community resources (Zippay & Son, 2013; Zippay & Thompson, 2007). A study based on addresses of 152 congregate
residences in 7 states indicated that residences were located in mixed-use, walkable neighborhoods in a variety of small towns and
metropolitan areas, with average poverty levels significantly higher than those of the general population of the states in which the
residences were sited (Zippay & Thompson, 2007). Furthermore, although there has been a shift in the mental health field to use
scattered-site apartments to enable people with PD to live in more integrated setting, inadequate level of housing subsidy and
landlords’ non-acceptance of people with PD as desirable tenants result in the concentration of supportive housing residents in
neighborhoods with socio-economic distress characterized by high-crime and high-poverty rates (Dickson-Gomez et al., 2017).
Specific to the current study, the two counties have contrasting features in terms of housing tenure and housing affordability,
which might further explain the difference in locational patterns of the two populations. MC is a predominantly owner-occupier
suburban county with a homeownership rate of 73%, whereas the housing tenure of residents in PC is mixed, with a slight majority
(54%) of households owning their homes (Pennsylvania Housing Finance Agency, 2012). And despite the substantially higher median

115
Y.-L.I. Wong et al. Research in Developmental Disabilities 83 (2018) 108–119

Fig. 2. Zoom-in map of a people with PD hot spot in Montgomery County and a people with ID hotspot in Philadelphia County.

116
Y.-L.I. Wong et al. Research in Developmental Disabilities 83 (2018) 108–119

home value in MC ($295,200) compared to PC ($132,200), the gross rent estimates of the two counties in 2010 were similar. Given
the lower socio-economic status of PC residents compared to those in MC, the rent burden, measured as a percentage of median rent
to median household income, was substantially higher in PC (46%) than in MC (27%). High rent burden in the urban county posed
particular challenge for scattered-site, tenant-based residents with PD to find affordable and quality housing because of the com-
petitiveness of the rental market among low- and middle-income households (Dickson-Gomez et al., 2017).
By featuring the Norristown hot spot in our analysis, we have showcased the clustering of people with PD in a community with
relatively high level of poverty, low median household income, high renter rate, and low median property value in a suburban
county. In fact, census data indicate that the socio-economic and housing profile of Norristown resembles a typical community in PC
(Data USA, 2017). The Norristown hot spot’s close proximity to both the state psychiatric hospital and a behavioral health center
offering comprehensive outpatient services and day programs for people with PD fits into the prototype of a “psychiatric ghetto”
(Dear & Wolch, 2014). The gravitation of people with PD in pockets of poverty and socio-economic disadvantage has been considered
a legacy of failed deinstitutionalization. However, one might argue that positive features of a self-contained suburban town like the
Norristown hot spot, including availability of community resources within walking distance and access to public transportation,
provides more opportunities for people with PD to interact with persons without disabilities than other suburban neighborhoods
(Metraux, Brusilovskiy, Prvu-Bettger, Wong, & Salzer, 2012).
The location of the Roxborough-Manayunk neighborhood as a hot spot area with clustering of people with ID can be viewed in the
context of a prolonged decline of the middle class in the urban county (Pew Charitable Trusts, 2014). The Pew study reported that 81
percent of the census tracts in PC in 1970 were middle-class neighborhoods, defined as having more than 50 percent of its households
with an income between 67 and 200 percent of regional median household income. The percent of middle-class census tracts had
dwindled to 31 percent in 2010 (p. 8–9). Whereas a number of ID hot spots, including the Roxborough-Manayunk neighborhood,
were among the neighborhoods held to be middle-class at both time points, most of the PD hot spots were located in tracts that either
had never been a middle-class neighborhood or had lost its standing as such during the past four decades.

7. Limitations of the study

Several considerations must be kept in mind when drawing conclusions from the findings of this study. First, the current study has
limited generalizability because it is based on geospatial data from two adjoining counties in one state. Nevertheless, the two study
counties are distinct with regard to their socio-demographic and housing profiles, thus making the analysis a useful case study to
further researchers’ understanding of the geography around mental health and intellectual disability (Park, Radford, & Vickers,
1998). Second, the neighborhood characteristics examined were restricted to an array of variables available in the American
Community Survey. Although the selected variables have been shown to be important correlates of the health and mental health
status in the general population (Mair, Roux, & Galea, 2008; Pickett & Pearl, 2001; Weden, Carpiano, & Robert, 2008), research to
date with regard to their relationships with quality of community life for people with disabilities has not been established. The focus
on comparing group differences in specific neighborhood characteristics might also have overlooked the relevance of neighborhood
typologies as an alternative avenue to identify communities that are conducive of inclusion of individuals with various types of
disabilities (Wong & Stanhope, 2009).
Another important consideration is the limited generalizability of the study findings beyond the United States where the process
of deinstitutionalization is more advanced, particularly for people with ID. In countries where normalization, valued role for people
with ID, and community inclusion are vouched for as policy goals, differences exist in the extent of success in achieving integrated
housing for people with ID. One comparative study of the living circumstances of children and adults with intellectual disabilities in
the United States, Canada, England and Wales, and Australia concluded that the United States is furthest advanced in the provision of
small community settings that are compatible with typical housing architecture (Braddock, Emerson, Felce, & Stancliffe, 2001). As
such, cross-disability spatial analysis of supportive housing in other countries might find more similar locational patterns for people
with ID and people with PD than the differences documented in the current study.

8. Conclusion

Despite the limitations aforementioned, this study demonstrates the applicability of a geographic information system approach for
improving policy makers’ and service administrators’ capacity to monitor the locational patterns of supportive housing and to design
housing-siting strategies that facilitate community inclusion. Although the development of supportive housing for people with ID and
people with PD has been guided by similar values and compatible service philosophies, particularly in countries at more advanced
stage of deinstitutionalization, the locational patterns of supportive housing for the two populations could be different due to policy
mandates, community profiles, and availability of public funding for housing and services. Future works should focus on developing a
nuanced understanding of what constitutes an optimum housing setting from the perspectives of people with disabilities and other
stakeholders, including policy makers, service administrators, and disability rights advocates. It should be cautioned that char-
acteristics of a community that are conducive for inclusion may be different for people with ID and people with PD.

Funding

The contents of this article were developed under the Switzer Research Fellowship awarded to Yin-Ling Irene Wong from the
United States Department of Education (National Institute on Disability and Rehabilitation Research [NIDRR] grant H133F090035).

117
Y.-L.I. Wong et al. Research in Developmental Disabilities 83 (2018) 108–119

Conflict of interest

Yin-Ling Irene Wong declares that she has no conflict of interest. Yiyue Huangfu declares that she has no conflict of interest.
Trevor Hadley declares that he has no conflict of interest.

Ethical approval

All procedures performed in this research study were in accordance with the ethical standards of the University of Pennsylvania
Office of Regulatory Affairs Institutional Review Board and with the 1964 Helsinki declaration and its later amendments. The study
was conducted with the approval of the Institutional Review Board of the University of Pennsylvania.

Informed consent

No informed consent was obtained because the study was based on administrative data with no personal identifiers.

References

Americans with Disabilities Act (1990). Retrieved from https://www.ada.gov/ada_title_II.htm.

Annison, J. E. (2000). Towards a clearer understanding of the meaning of “home”. Journal of Intellectual & Developmental Disability, 25(4), 251–262. https://doi.org/10.
1080/13668250020019566.
Bachrach, L. L. (1981). A conceptual approach to deinstitutionalization of the mentally retarded: A perspective from the experience of the mentally ill. In R. H.
Bruininks, C. E. Meyers, B. B. Sigford, & K. C. Lakin (Eds.). Deinstitutionalization and community adjustment of mentally retarded people (pp. 51–67). Washington DC:
American Association on Mental Deficiency.
Bagnall, A., & Eyal, G. (2016). Forever children” and autonomous citizens: Comparing the deinstitutionalizations of psychiatric patients and developmentally disabled
individuals in the United States. In B. L. Perry (Ed.). 50 years after deinstitutionalization: Mental illness in contemporary communities (pp. 27–61). Bingley: Emerald
Group Publishing Limited.
Braddock, D. (1992). Community mental health and mental retardation services in the United: A comparative study of resource allocation. The American Journal of
Psychiatry, 149(2), 175–183. https://doi.org/10.1176/ajp.149.2.175.
Braddock, D., Emerson, E., Felce, D., & Stancliffe, R. J. (2001). Living circumstances of children and adults with mental retardation or developmental disabilities in the
United States, Canada, England and Wales, and Australia. Mental Retardation and Developmental Disabilities Research Reviews, 7, 115–121.
Casper, M. L., Barnett, E., Halverson, J. A., Elmes, G. A., Braham, V. E., Majeed, Z. A., et al. (2000). Women and heart disease: An atlas of racial and ethnic disparities in
mortality (2nd ed.). Morgantown: Office for Social Environment and Health Research West Virginia University.
Chan, D. V., Gopal, S., & Helfrich, C. A. (2014). Accessibility patterns and community integration among previously homeless adults: A Geographic Information
Systems (GIS) approach. Social Science & Medicine, 120, 142–152. https://doi.org/10.1016/j.socscimed.2014.09.005.
Data USA (2017). Socio-economic and housing profile of Norristown, PA, compared to Philadelphia, PA. Retrieved fromhttps://datausa.io/profile/geo/norristownpa/?
compare=philadelphia-pa.
Dear, M. J., & Wolch, J. R. (2014). Landscapes of despair: From deinstitutionalization to homelessness. Princeton: Princeton University Press.
Dickson-Gomez, J., Quinn, K., Bendixen, A., Johnson, A., Nowicki, K., Ko, K. T., et al. (2017). Identifying variability in permanent supportive housing: A comparative
effectiveness approach to measuring health outcomes. The American Journal of Orthopsychiatry, 87(4), 414–424. https://doi.org/10.1037/ort0000232.
Ecker, J., & Aubry, T. (2016). Individual, housing, and neighborhood predictors of psychological integration among vulnerably housed and homeless individuals.
American Journal of Community Psychology, 58(1-2), 111–122. https://doi.org/10.1002/ajcp.12066.
Ecker, J., & Aubry, T. (2017). A mixed methods analysis of housing and neighbourhood impacts on community integration among vulnerably housed and homeless
individuals. Journal of Community Psychology, 45(4), 528–542. https://doi.org/10.1002/jcop.21864.
Emerson, E., Hatton, C., Robertson, J., & Baines, S. (2014). Perceptions of neighbourhood quality, social and civic participation and the self rated health of British
adults with intellectual disability: cross sectional study. BMC Public Health, 14(1), 1252. https://doi.org/10.1186/1471-2458-14-1252.
Finkler, C. (2014). Supportive housing. In A. C. Michalos (Ed.). Encyclopedia of quality of life and well-being research (pp. 2988–2991). Dordrecht, Netherlands: Springer.
Flynn, R. J., & Aubry, T. D. (1999). Integration of persons with development or psychiatric disabilities: Conceptualization and measurement. In R. J. Flynn, & R. Lemay
(Eds.). A quarter century of normalization and social role valorization: Evolution and impact (pp. 3–16). Ottawa: University of Ottawa Press.
Frankfort-Nachmias, C., & Leon-Guerrero, A. (2011). Social statistics for a diverse society (6th edition). Thousand Oak: Pine Forge Press (Sage).
Getis, A., & Ord, J. K. (1992). The analysis of spatial association by use of distance statistics. Geographical Analysis, 24(3), 189–206. https://doi.org/10.1111/j.1538-
4632.1992.tb00261.x.
Hogan, M. F., & Carling, P. J. (1992). Normal housing: A key element of a supported housing approach for people with psychiatric disabilities. Community Mental
Health Journal, 28(3), 215–226. https://doi.org/10.1007/BF00756818.
Kozma, A., Mansell, J., & Beadle-Brown, J. (2009). Outcomes in different residential settings for people with intellectual disability: A systematic review. American
Journal on Intellectual and Developmental Disabilities, 114(3), 193–222. https://doi.org/10.1352/1944-7558-114.3.193.
Larson, S., Salmi, P., Smith, D., Anderson, L., & Hewitt, A. (2013). Residential services for persons with intellectual or developmental disabilities: Status and trends through
2011. Minneapolis: University of Minnesota, Research Training Center on Community Living, Institute on Community Integration.
Longley, P. A., & Batty, M. (1996). Spatial analysis: Modelling in a GIS environment. Hoboken: John Wiley & Sons.
Mair, C. F., Roux, A. V. D., & Galea, S. (2008). Are neighborhood characteristics associated with depressive symptoms? A critical review. Journal of Epidemiology and
Community Health, 62, 940–946. https://doi.org/10.1136/jech.2007.066605.
Mansell, J., & Beadle-Brown, J. (2010). Deinstitutionalization and community living: Position statement of the comparative policy and practice special interest
research group of the international association for the scientific study of intellectual disabilities. Journal of Intellectual Disability Research, 54(2), 104–112.
Mansell, J., & Ericsson, K. (2013). Deinstitutionalization and community living: Intellectual disability services in Britain, Scandinavia and the USA. New York City: Springer.
Metraux, S., Brusilovskiy, E., Prvu-Bettger, J. A., Wong, Y.-L. I., & Salzer, M. S. (2012). Geographic access to and availability of community resources for persons
diagnosed with severe mental illness in Philadelphia, USA. Health & Place, 18(3), 621–629. https://doi.org/10.106/j.healthplace.2011.12.011.
Mitra, R., Buliung, R. N., & Faulkner, G. E. (2010). Spatial clustering and the temporal mobility of walking school trips in the Greater Toronto Area, Canada. Health &
Place, 16(4), 646–655. https://doi.org/10.1016/j.healthplace.2010.01.009.
Montgomery, A. E., Hill, L. L., Kane, V., & Culhane, D. P. (2013). Housing chronically homeless veterans: Evaluating the efficacy of a Housing First approach to
HUD‐VASH. Journal of Community Psychology, 41(4), 505–514. https://doi.org/10.1002/jcop.21554.
Nelson, G., & Caplan, R. (2017). Housing model for people with serious mental illness. In J. Sylvestre, G. Nelson, & T. Aubry (Eds.). Housing, citizenship, and communities
for people with serious mental illness: Theory, research, practice, and policy perspectives (pp. 23–44). Oxford: Oxford University Press.
Olmstead et al. vs. L. C. et al., 527 U.S. 581, 1999. Retrieved from https://supreme.justia.com/cases/federal/us/527/581/case.html.
Overmars-Marx, T., Thomése, F., Verdonschot, M., & Meininger, H. (2014). Advancing social inclusion in the neighbourhood for people with an intellectual disability:
An exploration of the literature. Disability & Society, 29(2), 255–274. https://doi.org/10.1080/09687599.2013.800469.

118
Y.-L.I. Wong et al. Research in Developmental Disabilities 83 (2018) 108–119

Park, D. C., Radford, J. P., & Vickers, M. H. (1998). Disability studies in human geography. Progress in Human Geography, 22(2), 208–233. https://doi.org/10.1191/
030913298672928786.
Parkinson, S., Nelson, G., & Horgan, S. (1999). From housing to homes: A review of the literature on housing approaches for psychiatric consumers/survivors. Canadian
Journal of Community Mental Health, 18(2), 145–164. https://doi.org/10.7870/cjcmh-1999-0008.
Pennsylvania Housing Finance Agency (2012). Pennsylvania housing availability & affordability ReportPennsylvania: Housing Finance Agency.
Pickett, K. E., & Pearl, M. (2001). Multilevel analyses of neighbourhood socioeconomic context and health outcomes: a critical review. Journal of Epidemiology and
Community Health, 55(2), 111–122. https://doi.org/10.1136/jech.55.2.111.
President George W. Bush’s New Freedom Initiative. (2001). Retrieved from http://govinfo.library.unt.edu/mentalhealthcommission/background.html.
Racino, J. A. (1995). Community living for adults with developmental disabilities: A housing and support approach. Research and Practice for Persons With Severe
Disabilities, 20(4), 300–310. https://doi.org/10.1177/154079699602000406.
Ridgway, P., & Zipple, A. M. (1990). The paradigm shift in residential services: From the linear continuum to supported housing approaches. Psychosocial Rehabilitation
Journal, 13(4), 11–31. https://doi.org/10.1037/h0099479.
Rog, D. J., Marshall, T., Dougherty, R. H., George, P., Daniels, A. S., Ghose, S. S., et al. (2014). Permanent supportive housing: Assessing the evidence. Psychiatric
Services, 65(3), 287–294. https://doi.org/10.1176/appi.ps.201300261.
Rossen, L. M., Khan, D., & Warner, M. (2014). Hot spots in mortality from drug poisoning in the United States, 2007–2009. Health & Place, 26, 14–20. https://doi.org/
10.1016/j.healthplace.2013.11.005.
Salzer, M. S. (2006). Psychiatric rehabilitation skills in practice: A CPRP preparation and skills workbook. Columbia: United States Psychiatric Rehabilitation Association.
Sampson, R. J., & Groves, W. B. (1989). Community structure and crime: Testing social-disorganization theory. The American Journal of Sociology, 94(4), 774–802.
https://doi.org/10.1086/229068.
Shen, G. C., & Snowden, L. R. (2014). Institutionalization of deinstitutionalization:a cross-national analysis of mental health system reform, 2018 a cross-national
analysis of mental health system reform. International Journal of Mental Health Systems, 8, 47. http://www.ijmhs.com/content/8/1/47.
Taylor, S. J. (2004). Caught in the continuum: A critical analysis of the principle of the least restrictive environment. Research and Practice for Persons With Severe
Disabilities, 29(4), 218–230. https://doi.org/10.2511/rpsd.29.4.218.
Timko, C. (1996). Physical characteristics of residential psychiatric and substance abuse programs: Organizational determinants and patient outcomes. American
Journal of Community Psychology, 24(1), 173–192. https://doi.org/10.1007/BF02511886.
Townley, G., & Kloos, B. (2011). Examining the psychological sense of community for individuals with serious mental illness residing in supported housing en-
vironments. Community Mental Health Journal, 47(4), 436–446. https://doi.org/10.1007/s10597-010-9338-9.
Tsemberis, S., Gulcur, L., & Nakae, M. (2004). Housing first, consumer choice, and harm reduction for homeless individuals with a dual diagnosis. American Journal of
Public Health, 94(4), 651–656. https://doi.org/10.2105/AJPH.94.4.651.
U.S. Census Bureau (2017). 2010 participant statistical areas program: Participant information. Retrieved fromhttps://www.census.gov/geo/partnerships/psap_overview.
html.
Waller, L. A., & Gotway, C. A. (2004). Applied spatial statistics for public health data, vol. 368. Hoboken: John Wiley & Sons.
Wang, X., & Varady, D. P. (2005). Using hot-spot analysis to study the clustering of section 8 housing voucher families. Housing Studies, 20(1), 29–48. https://doi.org/
10.1080/0267303042000308714.
Weden, M. M., Carpiano, R. M., & Robert, S. A. (2008). Subjective and objective neighborhood characteristics and adult health. Social Science & Medicine, 66(6),
1256–1270. https://doi.org/10.1016/j.socscimed.2007.11041.
Wong, Y.-L. I., Filoromo, M., & Tennille, J. (2007). From principles to practice: A study of implementation of supported housing for psychiatric consumers.
Administration and Policy in Mental Health and Mental Health Services Research, 34(1), 13–28. https://doi.org/10.1007/s10488-006-0058-y.
Wong, Y.-L. I., & Stanhope, V. (2009). Conceptualizing community: A comparison of neighborhood characteristics of supportive housing for persons with psychiatric
and developmental disabilities. Social Science & Medicine, 68(8), 1376–1387. https://doi.org/10.1016/j.socscimed.2009.01.046.
Wright, P. A., & Kloos, B. (2007). Housing environment and mental health outcomes: A levels of analysis perspective. Journal of Environmental Psychology, 27(1),
79–89. https://doi.org/10.1016/j.jenvp.2006.12.001.
Yanos, P. T., Felton, B. J., Tsemberis, S., & Frye, V. A. (2007). Exploring the role of housing type, neighborhood characteristics, and lifestyle factors in the community
integration of formerly homeless persons diagnosed with mental illness. Journal of Mental Health, 16(6), 703–717. https://doi.org/10.1080/09638230701496378.
Zippay, A., & Son, E. (2013). Community approaches to the siting of psychiatric housing. Journal of Community Practice, 21(1-2), 87–104. https://doi.org/10.1080/
10705422.2013.788338.
Zippay, A., & Thompson, A. (2007). Psychiatric housing: Locational patterns and choices. The American Journal of Orthopsychiatry, 77, 392–401. https://doi.org/10.
1037/0002-9432.77.3.392.

119