Geographic access to and availability of community resources for persons diagnosed with severe mental illness in Philadelphia, USA.
ABSTRACT This study assesses whether there are differences in geographic access to and availability of a range of different amenities for a large group of persons diagnosed with severe mental illness (SMI) in Philadelphia (USA) when compared to a more general set of residential addresses. The 15,246 persons who comprised the study group had better outcomes than an equal number of geographical points representative of the general Philadelphia population on measures of geographic proximity and availability for resources considered important by people diagnosed with SMI. These findings provide support for the presence of geographic prerequisites for attaining meaningful levels of community integration.
- SourceAvailable from: sciencedirect.comHealth & Place 01/2014; · 2.42 Impact Factor
Geographic access to and availability of community resources for persons
diagnosed with severe mental illness in Philadelphia, USA$
Stephen Metrauxa,n, Eugene Brusilovskiyb, Janet A. Prvu-Bettgerc, Yin-Ling Irene Wongd,
Mark S. Salzere
aDepartment of Health Policy & Public Health, University of the Sciences. 600S. 43rd Street, Philadelphia, PA 19104-4495, USA
bDepartment of Rehabilitation Sciences, Temple University 1700N. Broad Street, Suite #304, Philadelphia, PA 19122, USA
cSchool of Nursing, Duke University. 311 Trent Drive, DUMC 3322, Durham, NC 27710, USA
dSchool of Social Policy and Practice, University of Pennsylvania. 3701 Locust Walk, Philadelphia, PA 19104-6214, USA
eDepartment of Rehabilitation Sciences, Temple University 1700N. Broad Street, Suite #304, Philadelphia, PA 19122, USA
a r t i c l e i n f o
Received 19 April 2011
Received in revised form
22 December 2011
Accepted 23 December 2011
Available online 3 January 2012
a b s t r a c t
This study assesses whether there are differences in geographic access to and availability of a range of
different amenities for a large group of persons diagnosed with severe mental illness (SMI) in
Philadelphia (USA) when compared to a more general set of residential addresses. The 15,246 persons
who comprised the study group had better outcomes than an equal number of geographical points
representative of the general Philadelphia population on measures of geographic proximity and
availability for resources considered important by people diagnosed with SMI. These findings provide
support for the presence of geographic prerequisites for attaining meaningful levels of community
& 2012 Elsevier Ltd. All rights reserved.
Much is still to be learned about the quality of neighborhoods in
which persons with severe mental illness (SMI) reside (Yasui and
Berven, 2009; Zippay and Thompson, 2007). Initial studies on this
topic, conducted in the 1970s and 1980s, presented evidence that
persons diagnosed with SMI were, in the wake of deinstitutionaliza-
tion, shunted to inner-city ‘‘service-dependent ghettoes’’ where they
lived in proximity to community mental health services and amidst
concentrated poverty (Dear, 1977). Low socioeconomic status, dis-
crimination, and access to services were cited as factors that shaped
this ecological niche. Much of this ‘‘landscape of despair’’ that Dear
and Wolch (1987) describe has presumably changed with the
emergence of such diverse dynamics as the waning influence of
the psychiatric hospital, the upgrading of many of these erstwhile
service-dependent ghetto areas, and the increasing attention paid to
housing for persons with SMI after homelessness surged in the
1980s (Wolch and Philo, 2000). However, research in mental health
geography has not kept up with these changes (Yanos, 2007), and
has focused inordinately on services configurations at the expense of
more general examinations of mental illness in the community
(Deverteuil and Evans, 2009).
Research that has addressed this agenda confirms that the
housing landscape for persons with SMI has indeed changed, with
residential locations no longer as heavily clustered in poor, inner-
city areas (Metraux et al., 2007; Zippay and Thompson, 2007;
Wong and Stanhope, 2009). However, much of this research is
limited in at least two respects. First, with the exception of
Metraux et al. (2007), the studies focus on housing units that are
part of publicly funded programs that couple community-based
support services with provision of housing. While no precise data
is available, residents of these housing programs represent a small
fraction of persons with SMI living in the community, and the
dynamics of where these housing units are situated differ sub-
stantially from how and where persons find housing on their own.
Second, the studies that have examined neighborhood quality and
persons with SMI largely use aggregated geographical indicators.
These indicators, often taken from census data, are limited by
neighborhood parameters based on the available data and are
often inconsistent with residents’ individual perceptions of their
immediate environs (McWayne et al., 2007; Townley et al., 2008;
Wright and Kloos, 2007; Yanos et al., 2007).
Instead of the neighborhood-level measures that are tradition-
ally used in such geographic analyzes, this study assesses
Contents lists available at SciVerse ScienceDirect
journal homepage: www.elsevier.com/locate/healthplace
Health & Place
1353-8292/$-see front matter & 2012 Elsevier Ltd. All rights reserved.
$The authors gratefully acknowledge that funding for this research was
provided through the National Institute on Disability and Rehabilitation Research
(Grant #H133B080029). NIDRR had no direct involvement in any aspect of
conducting, reporting, or disseminating this research.
nCorresponding author. Tel.: þ1 215 349 7612; fax: þ1 215 349 7614.
E-mail addresses: email@example.com (S. Metraux),
firstname.lastname@example.org (E. Brusilovskiy), email@example.com
(J.A. Prvu-Bettger), firstname.lastname@example.org (Y.-L. Irene Wong),
email@example.com (M. Salzer).
Health & Place 18 (2012) 621–629
relationships between locations of specific amenities, grouped by
resource type, and residential addresses for 15,246 persons in
Philadelphia who were eligible for Medicaid and were diagnosed
with SMI, as well as for an equal number of geographical points that
are representative of the residential distribution for the general
Philadelphia population. Specifically, for each resource type, these
relationships are measured in two ways: geographic access – the
mean distance between addresses and the closest amenity, and
availability – the mean concentration of amenities around addresses.
What results is an analysis of whether or not the residences for the
persons in the study group were, on average, comparable to the set
of comparison points in terms of geographic access to and availability
of amenities in various resource types. These resource types include
supermarkets, public transportation, mental health services and
others that were deemed important based on surveys about living
preferences of persons with SMI.
Implicit to assessing these relationships between residence and
resource types is assessing the geography of the city and its ability
to facilitate geographic access and availability. Philadelphia is typical
of other older, Northeastern and Midwestern US cities in a variety of
characteristics that include high density residential patterns, mixed
land uses, expansive public transportation systems, walkable neigh-
borhoods, and a central business district which acts as a hub for the
city. Philadelphia’s deindustrializing landscape of the 1970s and 80s
(Adams et al., 1991) created socioeconomic conditions conducive to
service-dependent ghetto formation as the non-working poor took
advantage of older, physically deteriorated neighborhoods that
provided both low housing costs and ready access to needed
services (Wolch, 1980). These inner city neighborhoods, while
services rich, were presumed to be isolated from the mainstream
economy and the presence of amenities that are vital elements of
functional community (Wilson, 1996). More recently, however,
forces such as gentrification and immigration have added economic
and cultural heterogeneity to these areas (Simon and Allnut, 2007;
Katz and Lang, 2003).
These trends have been linked to the fracturing of service-
dependent ghetto areas and subsequent increases in homeless-
ness and incarceration among persons with mental illness (Dear
and Wolch, 1987). However, the changing characteristics of these
neighborhoods also offer the opportunity for greater geographical
access to and availability of a wider range of amenities that stand
to improve quality of life and offer a platform for increased
interaction within the community. Here the geography of a city
such as Philadelphia, with its dense land use patterns and
extensive public transportation network, offers particular advan-
tage to persons with psychiatric disabilities and others that
Wolch (1980) collectively referred to as the non-working poor.
Studies in the geography of mental health, in limiting their
focus on access and availability of psychosocial services, and have
overlooked the role of other amenities in facilitating the efforts of
persons with psychiatric disabilities to lead healthy, meaningful
lives in the community. Geographic access to and availability of
resource types matter, not only for quality of life but also as a
fundamental (yet incomplete) prerequisite for community inte-
gration. Geographic access and availability facilitate community
integration on social and psychological dimensions (Wong and
Solomon, 2002). Conversely, unfavorable disparities in these
dynamics would indicate a continuing a priori disadvantage
among this group. Such a disparity would recall the legacy of
the service-dependent ghetto and would raise practical impedi-
ments to realizing other aspects of community integration.
In assessing housing for persons with SMI, researchers have
focused more attention on housing characteristics than on
Thompson, 2007; Wong and Stanhope, 2009). The literature
examining housing preferences among persons with SMI (Piat
et al., 2008; Parkinson et al., 1999; Srebnik et al., 1995; Sylvestre
et al., 2007; Tanzman 1993) is illustrative of this. Here the
primary focus has been on preferences regarding the housing
itself – the nature of its physical attributes, the presence of
cohabiting residents, and the juxtaposition of living arrangements
with services. This narrow focus belies findings by Massey and
Wu (1993) in which mental health consumers, when expressing
their housing preferences, attached substantially higher impor-
tance to convenient location and proximity to mental health
services than did case managers. The study also cited consumers
as identifying transportation as a substantial barrier (where case
managers did not). These findings underscored the importance of
nearby amenities, and the lack of awareness of this dynamic by
Zippay and Thompson (2007) found that nearby amenities did
in fact figure prominently in providers’ decisions on where to
place group residences for persons with SMI. Interviews with
mental health administrators revealed their preference for locat-
ing facilities near commercial districts and mental health services.
Locations with such geographic proximity both facilitated access
and were more likely to meet their agencies’ affordability criteria.
Situating these residences in these types of neighborhoods often
put them in areas with above average rates of poverty, but the
authors also found these locations to be most amenable to
promoting independence and interaction. Such a description, in
which many of the low income neighborhoods in question were
considered vibrant, heterogeneous, and socially engaging for the
residents, contrasts with the notorious service-dependent ghet-
toes depicted in the older literature (Philo, 2005; Milligan, 1996).
More generally, resource access and availability are para-
meters of what Wong and Solomon (2002) describe as the
physical dimension of community integration. This physical
dimension encompasses how an individual ‘‘spends time, parti-
cipates in activities, and uses goods and services in the commu-
nity outside his/her home or facility in a self-initiated manner’’
(18). Assessing the role of amenities in facilitating community
integration is a complex undertaking that involves considering
both the actual physical amenities and consumer attitudes and
preferences towards these amenities. Yanos et al. (2007) combine
such subjective and objective dimensions in what they term
‘‘perceived opportunity’’ for taking advantage of available com-
munity resources. In another, similar conceptualization, Beal et al.
(2005) describe amenities as ‘‘facilitators’’ in their study of the
processes used by persons with SMI in establishing and main-
taining social interactions in their communities.
One of the most innovative studies to focus on the role of place
in community integration is Townley et al. (2008), who use
participatory mapping and GIS methods to integrate physical
and mental experiences of place by people with SMI. Their results
showed that, while home was the most important ‘‘activity
[a] considerable portion of participants also reported that
social/leisure activities (e.g., churches movies, YMCAs, etc)
and activities of daily living (e.g., grocery stores, shopping
centers, and restaurants) were locations where they spend
the most time, deem most important, and obtain a sense of
belonging. (527, emphases in original)
Such amenities become central to the experience of place for
persons with SMI, and represent staging areas which facilitate
S. Metraux et al. / Health & Place 18 (2012) 621–629
Wong et al. (2007), in a fidelity study of supported indepen-
dent living facilities for persons with SMI, also link resource
access and consumer preference to community integration. More
explicitly than the other studies, they cite proximity and avail-
ability of resources as key features of resource access, in con-
sideration of their role in facilitating residents’ participation in
community activities. And while publicly funded housing pro-
grams tend to be located in neighborhoods that are ‘‘resources
rich’’ (Wong et al. 2007; Zippay and Thompson, 2007), the extent
to which others in the population with SMI enjoy similar
geographic access and availability to various types of resources
is unknown. This study, then, examines the extent to which
amenities favored by persons with SMI are in fact accessible
and available to them on a broader scale than previous studies
3. Data and methods
3.1. Selection of study group and comparison addresses
Data on persons with SMI came from an administrative
database on Medicaid recipients maintained by Community
Behavioral Health (CBH), a publicly run managed-care organiza-
tion that funds behavioral health services for Medicaid recipients
in Philadelphia. Records were selected for persons who were
eligible for Medicaid over the course of the entire calendar year
2000 and who had claims records including at least one inpatient
or two outpatient Medicaid-reimbursed claims for services in
which ICD-9 diagnoses of affective disorder (296) or schizophre-
nia (295) were present. These are common criteria for defining
severe mental illness using administrative records (Lurie et al.,
1992; Blank et al., 2002). Addresses were obtained from the
Medicaid eligibility records for 16,439 persons who met these
criteria in the year 2000. Addresses for 15,246 individuals (92.7%)
were successfully geocoded, and this became the study group.
While information on personal characteristics was collected on
this group, only the location points for their addresses are used in
this study’s analyses. More information on the data and the
procedures used to identify the study group is available in
Metraux et al. (2007).
Addresses for this study group were compared to a set of
15,246 geographic points that represent the geographic distribu-
tion of Philadelphia’s overall population. This set was generated
using the Hawth’s Tools Extension to ArcGIS (Spatialecology.com,
2009) and the shapefile of the Philadelphia street network from
the US Census Bureau’s TIGER-Line files (US Census Bureau, 2008).
In this point generation process, locations were randomly
selected from each Philadelphia block group, and the number of
points generated for each census block group in Philadelphia was
proportional to the population of that block group. This means
that the number of geographic points for each block group was
the block group population divided by 99.54, which is the
quotient of the total Philadelphia population (1,517,550) and
the number of observations in the study group (15,246). For each
block group, each allocated point was randomly placed at a
geographic location along a road in a process that simulated a
residence location. The resulting set of comparison points could
then be considered representative of residential locations for the
3.2. Selection of relevant resource types
Amenities were grouped here into resource types, and in order
to determine important resource types, this study used results
from a local survey of persons with SMI. This survey, a version of
the Community Resource Accessibility Index (CRAI) originally
presented in Witten et al., 2003 and modified significantly by
Brusilovskiy and Salzer (2010), was administered to 119 indivi-
duals at two community mental health centers in Philadelphia.
Part of the survey involved asking respondents to rate the
importance of having 47 resource types within 15 min walking
distance or 5 min driving distance from their residence. The
respondents rated the proximity of each resource type on a Likert
scale ranging from not at all important (score of 1) to very
Twenty-one resource types received mean CRAI scores of at
least 3 (somewhat important), and 17 of these resource types for
which data on specific amenities were collected for this study are
shown in Table 1 in order of their perceived importance by CRAI
respondents. Of these 17, data on locations for 13 of these types
were obtained from InfoUSA. InfoUSA is a commercial service
that provides business and consumer databases for commercial
applications. Information collected from this database was com-
pany name, location and the six-digit Standard Industry Classifi-
cation (SIC) code that was used to categorize the data into each of
the ten resource types. When resource types were based on
InfoUSA data, Table 1 lists the SIC codes included as part of each
The resource type rated most important was supermarkets.
Using InfoUSA data and adapting criteria from the California
Center for Public Health Advocacy (2007), the resource type
‘‘supermarkets’’ included only the 125 establishments in the
‘‘Food Market’’ SIC category (541,105) that reported a sales
volume of $1,000,000 or higher; were part of a chain; and had
the word ‘‘supermarket’’ in their name.
Public transportation stops were rated second most important
of the resource types. Data on public transportation stops were
obtained from the Delaware Valley Regional Planning Commis-
sion for the year 2008. Geographic access to these stops was
measured in three ways. The first used the entire set of 9491
public transportation stops run by the Southeast Pennsylvania
Transportation Authority (SEPTA), resulting in a very high density
of stops in essentially all parts of Philadelphia. Two subsets of
these stops were also used as measures: the 136 stops that
accommodated four or more bus and trolley routes, and the 135
subway and regional rail stops.
The next six resource types were related to health and mental
health care. Location data on 40 Philadelphia hospitals, rated the
third most important resource type, came from a list compiled by
the Pennsylvania Department of Health and made available
through the Pennsylvania Spatial Data Access (PASDA) Center at
the Pennsylvania State University. Data on pharmacies (rated
fourth), including chain stores and independent pharmacies;
offices of medical doctors (rated sixth), including general practi-
tioners and specialists; optometrist offices (rated seventh); and
dentist offices (rated eighth) all came from InfoUSA. Information
as to which of these providers accept Medicaid and would thus be
accessible to the study group was unknown, but while Medicaid
was widely accepted at hospitals and pharmacies, large propor-
tions of Philadelphia physicians, dentists and optometrists did not
accept Medicaid coverage.
Mental health services were rated the fifth most important
resource type and this resource type includes both community
mental health centers (CMHC) and mental health (MH) service
providers. Along with CMHCs, the 128 MH services locations
included partial hospitals, and outpatient MH providers who
treated Medicaid-eligible patients in the year 2000. Data for the
locations of these providers came from CBH.
Other resource types where location data came from InfoUSA
were, in ranked order of importance, banks and credit unions;
post offices; department stores; places of worship; dry cleaners
S. Metraux et al. / Health & Place 18 (2012) 621–629
and laundry services; libraries; and a combined resource type of
restaurants and coffee and ice cream shops. Places of worship and
religious organizations, at 2733 locations, encompassed a diverse
set of religions and facilities, but also did not encompass some
types of religious establishments such as mosques, ashrams and
shrines. The other resource type was shopping centers, which
were extracted from the Philadelphia zoning shapefile provided
by the PASDA center. This dataset identified 124 parcels in
Philadelphia that were designated for commercial use with the
codes ‘‘ASC’’ (Area Shopping Centers) and ‘‘NSC’’ (Neighborhood
Four resource types were excluded from this study because the
establishments could not be consistently and reliably categorized
given the available data. Employment opportunities along with job
training and related services were excluded because location data
were unavailable. Smaller grocery stores were omitted because a
careful examination of InfoUSA data showed a number of omissions
and problems with SIC code classifications for this resource type.
Lastly, clothing and shoe stores were excluded because many stores
in that category specialize in a specific type of apparel or style of
clothing which would often be gender-specific.
Finally, because there are often no distinct physical boundaries
which separate Philadelphia from the surrounding Bucks, Mon-
tgomery and Delaware counties in the north, west, and south of
the city, individuals living in Philadelphia may take advantage of
resources which are outside of the city limits. Therefore, when-
ever possible, locations were acquired for resources not only in
Philadelphia but also the neighboring parts of Bucks, Montgomery
and Delaware counties which lie within half a mile of the city
3.3. Measuring differences in geographic access and availability
The spatial relationship between amenities (categorized by
resource type) and locations of the residential addresses of
persons with SMI and the comparison points was assessed in
two ways. First, geographic access was represented by proximity –
the Euclidean distance between residence location (or comparison
point) and the nearest location for each resource type (e.g.,
supermarket, place of worship, etc.) and was calculated using
the Spatial Analyst Extension to ArcMap (ESRI, 2006). Second,
availability was measured by the concentration of amenities in a
resource type that were located within a half-mile of each
residence location (or comparison point). A half mile can be
viewed as a 10-min walk by a fully abled, ambulatory person
and is considered the outer bounds of what is walkable (Dittmar
and Poticha, 2004; Daisa, 2004). While walkability involves
factors other than distance measures (Frank et al., 2009), distance
is commonly used as a proxy for walkability (Carr et al., 2010).
Euclidean distance, rather than the substantially more compu-
tationally intensive along-road (i.e. network) distance, was used
in our analyzes. However, recent research looking at ways to
measure geographic access to urban health services reported very
high Pearson correlations between the Euclidean and network
distances in urban areas (Apparicio et al., 2008).
To summarize, proximity and concentration of amenities
served as two related but conceptually distinct metrics of com-
munity resource presence: the former a measure of geographic
access and the latter a measure of resource availability. For
certain resource types, only one of those two measures was
presented due to contextual factors. Any concentration measure
Resources with CRAI importance scores of 3 or abovea,b.
Resource importance CRAI
Information source (with SIC codes for Info USA data)
Public transportation stopsc
Community mental health (MH)
centers and MH services providers
Offices of medical doctors
Offices of optometrists
Offices of dentists
3.81 (0.57) 125
3.80 (0.63) 9491
3.79 (0.47) 40
3.78 (0.56) 371
3.70 (0.65) 128
Info USA - 541105 - Food markets
Regional planning commission data
PA Department of Health (PASDA)
Info USA - 591205 - Pharmacies
City of Philadelphia Community Behavioral Health
3.69 (0.61) 6158
3.68 (0.66) 141
3.65 (0.66) 1187
3.53 (0.83) 503
Info USA - 801101 - Physicians and Surgeons
Info USA - 8042XX - Optometrists
Info USA - 802101 - Dentists
Info USA - 602101 - Banks; 602201 - State Commercial Banks; 603501 - Savings or Loan
Associations; 606101 - Credit Unions (CUs); 606102 - Federally Chartered CUs
Info USA - 43XXXX - Post Offices
Info USA - 531102 - Department Stores
Info USA - 866102 - Bible Schools and Study; 866104 - Church Organizations; 866105 - Christian
Science Practitioners; 866107 – Churches; 866110 - Religious Organizations; 866112 -
Info USA - 721101 - Laundries; 721201 - Cleaners; 721501 - Laundries Self-Service; 721603 -
Cleaners - Wholesale
Info USA - 823104 - City Government Libraries; 823106 - Public Libraries
City of Philadelphia Zoning (PASDA)
Info USA - 581206 - Food To Go (Carry-Out); 581208 - Restaurants; 581209 - Delis; 581215 -
Box Lunches; 581217 - Appetizers, Snacks, etc.; 581219 - Sandwiches; 81222 - Pizza; 581224 –
BBQ; 581227 - Italian Food Products; 581229 - Deli-Bakery; 581230 - Restaurants with Food
Delivery; 5813XX - Drinking Places (pubs, bars) , 581203 - Ice Cream Parlors; 581228 - Coffee
Shops; 581214 – Cafes
Info USA - 723101 - Skin Treatments; 723102 – Manicuring; 723106 - Beauty Salons; 723108 -
Make Up Studios; 724101 - Barber Shops
Place of worship
3.48 (0.84) 49
3.26 (0.84) 65
3.24 (1.09) 2733
Dry cleaners & laundry services 3.21 (1.08) 662
Restaurants, and coffee and ice cream
3.19 (0.93) 77
3.14 (0.88) 124
3.09 (0.84) 3958
Barber shops, hair salons, & beauty
3.07 (1.11) 1940
aBased on results from CRAI survey item asking respondents (119 persons diagnosed with SMI living in Philadelphia) to rate the importance of having each resource in
proximity to their residences.
bResource types listed as ‘‘somewhat important’’ (mean rating of 3 or higher) by CRAI respondents but not included here include smaller grocery stores (mean rating
3.50, SD 0.74); clothing and shoe stores (mean rating 3.21, SD 0.83); employment opportunities (mean rating 3.12, SD 1.02); and Job training and related services (mean
rating 3.05, SD 1.05).
cOf 10,804 bus and trolley stops, 136 offered access to 4 or more routes. These multi-route stops were also analyzed as their own separate category.
dThe restaurant and coffee shop resource type was combined with the ice cream shop resource type.
S. Metraux et al. / Health & Place 18 (2012) 621–629
less than one (i.e., where on average there was less than one
amenity of a particular resource type within a half-mile radius of
addresses) was not shown in the results as it was not considered a
valid measure of availability. For example, since there are only 46
post offices in the study area, getting to a post office would be a
geographic access issue and not one of choosing which of the
nearby post offices would be the best to use. The same reasoning
applied to libraries, hospitals, both public transportation subsets
(but not the full set of stops), department stores, and shopping
center resource types. Conversely, concentration would be a
much more suitable measure for something like places of wor-
ship. Here if an individual is Protestant but the place of worship
nearest to her house is a Catholic church, the proximity measure
would not be useful. This problem was ameliorated with the
concentration measure, as having more places of worship within
a half-mile radius would make it more likely that one of the
nearby facilities would have suitable qualities (including denomi-
nation). Place of worship was the only resource type where the
access measure is omitted.
T-tests assessed whether differences in the mean measures of
proximity and concentration differed between the study group
and the comparison points. Parametric t-tests were appropriate,
even though the distributions of some of the amenities examined
here were heavily zero-inflated or skewed, because the large
sample size in our study invalidated the need for the non-
parametric alternatives (Lumley et al., 2002). While demographic
information and other characteristics were reported for the study
group, they were not included in any of the spatial analyzes.
This research was approved by institutional review boards of
Temple University, the University of Pennsylvania, the University
of the Sciences, and the City of Philadelphia.
The characteristics of the study group (which do not figure
into the analyses presented here) were reported in detail in
Metraux et al. (2007). To summarize, the majority of the study
group was female (65%); 60% were over age 40 (with one-quarter
over age 50); and just about half were non-Hispanic Black, one-
quarter were non-Hispanic White, and 12% were Hispanic. Almost
three-quarters of the population (73%) were Medicaid-eligible by
virtue of receiving SSI benefits and 14% were Medicaid eligible
through receipt of TANF welfare benefits. Seventy-seven percent
had diagnoses of affective disorders, as compared to 39% who
were diagnosed with schizophrenia (these diagnoses were not
mutually exclusive). Among Medicaid-reimbursed services, one-
third of the group had a record of an inpatient psychiatric stay
and almost all had histories of outpatient care.
Fig. 1 shows the spatial distribution of points for the study
group and the comparison points. While no tests of difference
were performed here, the distributions were clearly different. The
figure indicates that the study group was not as widely distrib-
uted across the city as the comparison points; was more densely
concentrated in an area approximately at the center of the
Philadelphia map, which corresponds to the North Philadelphia
area of the city; and was less densely distributed through
numerous other parts of the city.
Table 2 shows the results of the t-tests comparing the study
group and the set of comparison points in terms of proximity and
concentration. For most resource types, there were highly sig-
nificant differences for both types of measures. For all but four
resource types, differences in proximity favored the study group.
Three of the four exceptions – ‘‘bus stops with 4 or more routes’’
subset of public transportation stops, department stores, and
shopping centers had mean proximity measures that disadvan-
taged the study group, and there was no significant difference in
the proximity of dentist offices. Among the 12 resource types that
had valid concentration measures, seven (supermarkets, public
transportation stops, pharmacies, mental health services, places
of worship, laundry and dry cleaner services, and restaurants) had
concentration measures that favored the study group, while four
resource types (doctor offices, dentist offices, optometrist offices,
and banks and credit unions) were significantly less concentrated,
on average, around the study group locations. There was
no significant difference in concentration for barber shop
Assessing the geographic access findings from Table 2 neces-
sitates considering, for each resource type, both the absolute
mean proximity and the differences in the mean proximity
between the study group and the control points. The shortest
mean proximity to the nearest amenity in a resource type for the
study group was 380 feet (for any type of public transportation
stop), and four other resource types (restaurants, doctor offices,
Fig. 1. Densities of individuals with SMI and randomly generated points.
S. Metraux et al. / Health & Place 18 (2012) 621–629
cleaners and barber shops) had a location, on average, less than
1000 feet from study group residences. These resource types all
had differences in mean proximity between 100 and 200 feet and
that favored the study group. Eight of the resource types had
mean proximities to the nearest amenity of over a half-mile
(42640 feet), and, among these resource types, four (optometrist
offices, department stores, shopping centers, and multi-route bus
stops) disfavored (i.e., were farther for) the study group. For
department stores, optometrist offices and shopping centers,
however, the relative differences, 90, 76, and 75 feet, respectively,
were small relative to the mean proximity to the nearest location.
The subset of bus stops with four or more routes, on the other
hand, had both the farthest mean proximity to the nearest
location and the biggest difference between the means for the
study group and the comparison points.
Looking at concentration, 12 of the 19 resource types examined
here had more than one amenity within a half-mile of the mean
location point for the study group and the comparison points. Beyond
that, the mean concentrations for the study group ranged from under
four for four resource types to over 17 for four others. Four resource
types – religious organizations, restaurants, doctor offices and dentist
offices – had differences in mean concentration that were greater
than one. The two former resource types favored and the two latter
disfavored the study group.
This study found that a large group of Medicaid recipients
diagnosed with SMI had better outcomes, when compared to a
representative distribution of Philadelphia locations, on measures
of geographic proximity and availability for resources considered
to be important to people diagnosed with SMI. These findings are
contrary to the negative locational disparities associated with
persons diagnosed with SMI and living in the community that
were found in prior ‘‘waves’’ of geographical research on mental
illness (Wolch and Philo, 2000). Furthermore, these findings
provide support for the presence of geographic prerequisites for
attaining meaningful levels of community integration. Stated
differently, while meaningful community integration involves
much more than being geographically connected to desirable
community resources, the findings here are encouraging.
The commercial resource types examined here, excepting
shopping centers and department stores (both hardly neighbor-
hood fixtures) were more geographically accessible and available
for the SMI study group. The mean distances to the nearest
restaurants, barber shops, and cleaners – typical neighborhood
fixtures – were all the equivalent of less than three city blocks
away for the study group. And while there was less than 100 feet
difference in the mean distances for the comparison points, when
these differences were cast in terms of the mean distances
themselves, the resulting proportions were substantial. This
suggests that people with SMI in this study were more likely
than the typical Philadelphian to live near a commercial area
featuring clusters of small businesses. Obviously this is not
desirable in all cases, as personal preferences and the qualities
of these commercial areas may not be compatible. Seen in the
aggregate, however, these dynamics appear consistent with the
siting decisions made by mental health administrators (Zippay
and Thompson, 2007). These dynamics are also consistent with
traditional neighborhood design concepts and their emphasis on
spatially integrating housing with surrounding amenities as a
means of community-building, and that have been best known in
the context of the New Urbanism movement (Calthorpe, 2001;
It was not just small neighborhood establishments that were
more available to the study group. Also proximate and more
available to the study group were supermarkets, which represent
a key amenity from the viewpoints of mental health consumers,
advocates for low-income urban neighborhoods (California Center
for Public Health Advocacy, 2007), and public health researchers
(Laraia et al., 2004; Morland et al., 2006; Powell et al., 2007).
Furthermore, mental health facilities, another key amenity
cited by both mental health consumers and mental health
Proximity to and concentrationaof resources.
Resource TypeProximity Concentration
(Distance to nearest amenity in feet) (Number of amenities within ½ mile)
SMI group meanRandom group meant-value SMI group mean Random group meant-value
All access points
Bus stops with 4 or more routes
Subway or regional rail stops
Community mental health (MH) centers & MH providers
Offices of medical doctors
Offices of Optometrists
Offices of dentists
Banks & credit unions
Places of worship
Dry cleaners & laundry services
Restaurants, bars, & ice cream & coffee parlors
Barber shops, hair salons, & beauty shops
aAccessibility measure for religious organizations and all concentration measures with mean less than one are not considered valid measures.
S. Metraux et al. / Health & Place 18 (2012) 621–629
administrators (Zippay and Thompson, 2007), were also more
available to the study group. This greater proximity to mental
health facilities provides a degree of continuity with the old
mental health geographies, however the availability of other
amenities speaks against the spatial isolation that was character-
istic of the service-dependent ghetto. This will be addressed in
more detail shortly.
There were mixed results in the other health provider cate-
gories – doctor, dentist, and optometrist offices, hospitals, and
pharmacies – as well as in the public transportation measures.
These instances of mixed results underscore some of the chal-
lenges to interpreting the proximity and concentration measures
used here. First, while proximity and concentration measures
usually were consistent, as seen with the health provider resource
types this was not always the case. These measures represent
different dynamics, and it is unclear how to assess the compara-
tive utility of each measure when they diverge, especially as the
relative significance of proximity and concentration will vary
with different resource types.
The mixed findings related to public transportation highlight
the impact of qualitative factors on these resource types. In this
case, the three different measures used in this study (all stops,
subway stops, and multi-route bus stops) were borne of the
difficulty in identifying the most useful set of locations for this
resource type. As the results show, in a city such as Philadelphia,
with most of its area consisting of dense, heterogeneous land uses
serviced by an extensive public transportation system, most
addresses are close to numerous transit stops. Some stops are
more useful than others, and in an effort to identify such stops the
other two measures were also included. What the optimal
relationship between residence and transit stop would be remains
unclear, and conclusions about this can differ based on the
measure used. Regardless of this, these results show no clear
disadvantage in geographic access and availability of public
transportation for the study group.
These results suggest that persons in the study group were not
more isolated from the amenities examined here as compared to
addresses that represented Philadelphia’s general population.
More generally, this would imply that the residences of a large
part of the Philadelphia’s population diagnosed with SMI were not
disadvantaged in terms of their relationships with their preferred
amenities. Put in the context of older mental health geographies,
the changes that Dear and Wolch (1987) saw as transforming the
service-dependent ghetto led to positive externalities for its
erstwhile residents along with the higher rates of homelessness,
incarceration, and other collateral damage that also followed in
Three limitations to this study modify the positive nature of
these findings. The first is that the study group was not repre-
sentative of adults in Philadelphia diagnosed with SMI. The
majority of the study group were both predominantly poor, based
on the categories by which they are eligible for Medicaid, but also
relatively stable, based on their having maintained Medicaid
eligibility for at least one year. Specifically, the study group
excluded persons with SMI who had income above the poverty
level (i.e., those likely to work), as well as those with income
below the poverty level but who did not participate in Medicaid
or participated only sporadically (less than a year). All persons in
the study group maintained Medicaid eligibility for at least one
year and almost three-quarters were Medicaid eligible by virtue
of their SSI receipt, indicating a regular income source and
relative economic stability. Destabilizing phenomena such as
extreme poverty, homelessness (and more generally, residential
instability), and incarceration all would have made maintaining
Medicaid eligibility for one year either difficult or impossible.
There is no way of quantifying the number of people excluded by
this selection criteria, but it is clear that the study group over-
represented persons with SMI whose living situations, judged by
their ongoing receipt of medical and financial assistance, indi-
cated a reduced degree of volatility.
Despite such qualifications, the size of the study group, with
records for over 15,000 persons (approximately 7% of Philadel-
phia’s overall adult poverty population), would mitigate against
excessive selectivity. Furthermore, it is inherently worthwhile to
examine the existence of locational disparities among those
impoverished persons with SMI who presumably would have
been best able to integrate into the community along other
dimensions as well. However, those excluded from the study
group because they were unable, unwilling or ineligible (not all
poor persons are eligible for Medicaid) to maintain Medicaid
eligibility would likely have had very different outcomes than
those in the study group with regards to the measures presented
here. This omitted group, by virtue of a mix of social, economic
and mental health issues, also would have faced the most
substantial barriers to more general community integration, and
would have been most prone to the negative outcomes associated
with the transformation of service-dependent ghetto areas. While
the needs of this group should not be ignored, they should also
not be construed as representative of the more general population
diagnosed with SMI in Philadelphia.
The second limitation is that the data examined here only
provided a limited amount of contextual information on ame-
nities themselves and the neighborhoods in which they were
located. Regarding the former, consumer preferences were taken
into account in choosing specific resource types, however quali-
tative differences within these categories were not assessed. For
example, a resource type such as ‘‘restaurants’’ included fast food
and haute cuisine, and one type is unlikely to satisfy a preference
for the other. Furthermore, places such as drinking establish-
ments may facilitate interaction among community residents, or
they may be regarded as nuisances and detract from a neighbor-
hood’s livability. Examining differences within resource types,
using approaches such as those found in Lewis et al. (2005) and
Sister, Wolch and Wilson (2010), would provide opportunities to
address questions related to the quality of specific locales in a
more general category, and add an additional level of nuance to
the results found in this study.
There are other neighborhood-related factors that are also not
addressed in this study. This includes omitted resource types such
as parks, libraries, adult education centers, and social service
centers. As mentioned, we limited our choice of resource types to
those chosen as most important by a local survey of persons with
SMI. While consumer preference provides a solid rationale for
choice of resource types, the categories chosen here are clearly
not exhaustive. Furthermore, the presence of other neighborhood
factors – crime, pollution, housing quality – that can also speak to
differential neighborhood quality are not addressed here but
represent important vectors for continuing this research focus
on spatial aspects of community integration.
In a final limitation, the geographic features of Philadelphia,
discussed earlier, which stand to facilitate geographic access and
availability to persons with limited means of mobility, also run
the risk of being simplified in these analyzes. Densities in the city
vary and are correlated with socioeconomic status, with the
central city areas surrounded by considerably less dense neigh-
borhoods in far Northeast or Northwest sections of the city. These
areas have considerably more suburban characteristics—fewer
businesses, more space between residences, less walkability, and
less access to public transit routes. In contrast, individuals whose
residence locations are examined here lived mainly in the denser
North and West Philadelphia neighborhoods–areas which are
considerably poorer, more affordable, and less likely to oppose
S. Metraux et al. / Health & Place 18 (2012) 621–629
residences for persons with psychiatric disabilities (Metraux
et al., 2007). Thus, while the findings from this study support
parity in geographic access to and availability of amenities, they
should not be construed to suggest that persons with psychiatric
disability enjoy complete and equal integration into all areas of
The unique features of Philadelphia geography also limit the
generalizability of these findings. Insofar as they share such
characteristics, these findings would likely be replicated in other
‘‘rust belt’’ US cities. But just as the concept of service-dependent
ghetto is limited in its transferability (Milligan, 1996), the levels
of geographic access and availability of amenities for persons with
SMI will differ in locations with other geographic patterns. Given
this, the generalizability of this study to locations and populations
outside of Philadelphia is limited. Early studies on the geography
of mental illness focused on urban centers that were, on one hand,
services rich but nonetheless shunted and concentrated persons
with SMI into less desirable neighborhoods. Several decades later,
this study now presents evidence that persons with SMI have
access to resources that are comparable or more advantageous
than those of a representative set of addresses in Philadelphia.
Studies specific to other localities, and other types of localities,
would be needed to determine whether such findings apply more
These analyzes looked at diagnosis of SMI as the common
factor in the study group, comparing it to a set of comparison
points that are geographically representative of the general
Philadelphia population. These comparison points do not repre-
sent actual people, so factors influencing residential location such
as income, race and ethnicity, and family status, could not be
controlled for. The SMI diagnoses among the study group mem-
bers are not monolithic, and diversity of specific diagnoses that
fall under this rubric may influence residential opportunities and
preferences. However, no significant or substantial differences
emerged when those in the study group diagnosed with schizo-
phrenia (DSM IV diagnosis of 295) were compared to the rest of
the group for the mean distances and concentrations that are
assessed in Table 2.
The resources data used in this study reflected a static ‘‘snap-
shot’’ of neighborhood amenities, and did not capture the
dynamic nature of neighborhoods. Persons were selected for
the study group based on their Medicaid records in 2000, the
comparison point data was selected based on 2000 Census data,
and much of the amenities data used here were current in 2000.
A prominent exception to the latter, however, was the InfoUSA data,
which were from a 2003 dataset. The extent to which the study
group and comparison point data were temporally ‘‘mismatched’’ to
the amenities data should not be substantial or systematic. A related
concern is that this study, as a snapshot, cannot take into account
neighborhood change over time, including (but not limited to)
The findings in this study show that persons diagnosed with
SMI living in the community were not disadvantaged with respect
to geographic access to and availability of a variety of resources
that matter to them and that may potentially serve as facilitators
of community integration. As such, it is an indicator of spatial
integration that is both prerequisite to deeper levels of commu-
nity integration and underscores the need for incorporating more
spatial aspects into research that examines the extent to which
community integration is being realized. There are numerous
directions for building upon this research, such as looking more
directly at the relationships between accessible and available
resources and measures of community participation, and the role
of neighborhood organization as building off of the former, and
acting as a mediating factor for the latter. Much remains to be
learned about the role of the built environment on community
integration, to which this paper provides an initial step.
The authors wish to thank the Philadelphia Department of
Behavioral Health and Intellectual disability Services for their
support and for data that was integral to this study. Additionally,
we would like to thank the anonymous reviewers for their
insightful and challenging comments.
Adams, C., Bartelt, D., Elesh, D., Goldstein, I., Kleniewski, N., Yancey, W., 1991.
Philadelphia: Neighborhoods, Division, and Conflict in a Postindustrial City.
Temple University Press, Philadelphia.
Apparicio, P., Abdelmajid, M., Riva, M., Shearmur, R., 2008. Comparing alternative
approaches to measuring the geographical accessibility of urban health
services: distance types and aggregation-error issues. International Journal
of Health Geographics 7, 7.
Beal, G., Veldhorst, G., McGrath, J.L., et al., 2005. Constituting community: creating
a place for oneself. Psychiatry 68 (3), 199–211.
Blank, M.B., Mandell, D.S., Aiken, L., Hadley, T.R., 2002. Co-occurrence of HIV and
serious mental illness among Medicaid recipients. Psychiatric Services 53 (7),
Brusilovskiy, E. & Salzer, M.S. (2010). Perceived community resource availability
among individuals with psychiatric disabilities. Presented at the Annual
Meeting of the American Public Health Association, Denver, November 6–10.
California Center for Public Health Advocacy, 2007. Searching for Healthy Food.
The Food Landscape in California Cities and Counties: Detailed Methodology.
Last accessed 9/5/10.
Carr, L.J., Dunsiger, S.I., Marcus, B.H., 2010. Walk ScoreTMas a global estimate of
neighborhood walkability. American Journal of Preventive Medicine 39 (5),
Calthorpe, P., 2001. The Regional City: Planning for the End of Sprawl. Island Press,
Daisa, J.M., 2004. Transit parking and residential development. In: Dittmar, H.,
Ohland, G. (Eds.), The New Transit Town: Best Practices in Transit-Oriented
Development. , Island Press, Washington, DC.
Dear, M.J., 1977. Psychiatric patients and the inner city. Annals of the Association
of American Geographers 67, 588–594.
Dear, M.J., Wolch, J.R., 1987. Landscapes of Despair: From Deinstitutionalization to
Homelessness. Polity Press, Oxford.
Deverteuil, G., Evans, J., 2009. Landscapes of despair. In: Brown, T., McLafferty, S.,
Moon, G. (Eds.), A Companion to Health and Medical Geography, Wiley-
Blackwell, Oxford, pp. 278–300.
Dittmar, H., Poticha, S., 2004. Defining transit oriented development: the
new regional building block. In: Dittmar, H., Ohland, G. (Eds.), The New
Transit Town: Best Practices in Transit-Oriented Development, Island Press,
ESRI, 2006. ARCGIS Spatial Analyst. author, Redlands CA.
Frank, L.D., Sallis, J.F., Saelens, B.E., et al., 2009. The development of a walkability
index: application to the neighborhood quality of life study. British Journal of
Sports Medicine 44, 924–933.
Katz, B., Lang, R.E., 2003. Redefining Urban and Suburban American: Evidence from
Census 2000. The Brookings Institute, Washington DC.
Kelbaugh, Douglas, 1997. Common Place: Toward Neighborhood and Regional
Design. University of Washington Press.
Laraia, B.A., Siega-Riz, A.M., Kaufman, J.S., Jones, S.J., 2004. Proximity of super-
markets is positively associated with diet quality index for pregnancy.
Preventive Medicine 39, 869–875.
Lewis, L.B., Sloane, D.C., Nascimento, L.M., et al., 2005. African Americans’ access to
healthy food options in South Los Angeles restaurants. American Journal of
Public Health 95 (4), 668–673.
Lumley, T., Diehr, P., Emerson, S., Chen, L., 2002. The importance of the normality
assumption in large public health data sets. Annual Review of Public Health
Lurie, N., Popkin, M., Dysken, D., Moscovice, I., Finch, M., 1992. Accuracy of
diagnoses of schizophrenia in Medicaid claims. Hospital and Community
Psychiatry 43 (1), 69–71.
Massey, O.T., Wu, L., 1993. Important characteristics of independent housing for
people with mental illness: perspectives of case managers and consumers.
Psychosocial Rehabilitation Journal 17 (2), 81–93.
McWayne, C., McDermott, P., Fantuzzo, J., Culhane, D.P., 2007. Employing com-
munity data to investigate social and structural dimensions of urban neigh-
borhoods: an early childhoold example. American Journal of Community
Psychology 39 (1-2), 47–60.
S. Metraux et al. / Health & Place 18 (2012) 621–629
Metraux, S., Caplan, J.M., Klugman, L., Hadley, T.R., 2007. Assessing residential
segregation among Medicaid recipients with psychiatric disability in Phila-
delphia. Journal of Community Psychology 35 (2), 239–256.
Milligan, C., 1996. Service dependent ghetto formation—a transferable concept?
Health & Place 2 (4), 199–211.
Morland, K., Diez Roux, A.V., Wing, S., 2006. Supermarkets, other food stores, and
obesity: the atherosclerosis risk in communities study. American Journal of
Preventive Medicine 30 (4), 333–339.
Newman, S.J., 2001. Housing attributes and serious mental illness: implications for
research and practice. Psychiatric Services 52 (10), 1309–1317.
Parkinson, S., Nelson, G., Horgan, S., 1999. From housing to homes: a review of the
literature on housing approaches for psychiatric consumer/survivors. Cana-
dian Journal of Community Mental Health 18, 145–164.
Philo, C., 2005. The geography of mental health: an established field? Current
Opinion in Psychiatry 18, 585–591.
Piat, M., Lesage, A., Boyer, R., et al., 2008. Housing for persons with serious mental
illness: consumer and service provider preferences. Psychiatric Services 59 (9),
Powell, L.M., Auld, M.C., Chaloupka, F.J., O’Malley, P.M., Johnston, L.D., 2007.
Associations between access to food stores and adolescent body mass index.
American Journal of Preventative Medicine 33 (4 Suppl), S301–7.
Simon, Allnut, 2007. Philadelphia, 1982–2007: toward the postindustrial city.
Pennsylvania Magazine of History and Biography 131 (4), 395–444.
Sister, C., Wolch, J., Wilson, J., 2010. Got green? addressing environmental justice
in park provision. GeoJournal 75, 229–248.
Spatialecology.com (2009). Overview of Hawth’s Analysis Tools for ARCGIS. /http://
www.spatialecology.com/htools/overview.phpS. Last accessed 9/5/10.
Srebnik, D., Livingston, J., Gordon, L., King, D., 1995. Housing choice and commu-
nity success for individuals with serious and persistent mental illness.
Community Mental Health Journal 31, 139–152.
Sylvestre, J., Nelson, G., Sabloff, A., Peddle, S., 2007. Housing for people with
serious mental illness: a comparison of values and research. American Journal
Community Psychology 40, 125–137.
Tanzman, B., 1993. An overview of surveys of mental health consumers’ preferences for
housing and support services. Hospital and Community Psychiatry 44, 450–455.
Townley, G., Kloos, B., Wright, P.A., 2008. Understanding the experience of place:
expanding methods to conceptualize and measure community integration of
persons with serious mental illness. Health & Place 15, 520–531.
United States Census Bureau, 2008. TIGER/Line shapefiles: Technical documenta-
tion. author, Washington DC.
Wilson, W.J., 1996. When Work Disappears. Alfred A. Knopf, Inc., New York.
Witten, K., Exeter, D., Field, A., 2003. The quality of urban environments: mapping
variation in access to community resources. Urban Studies 40 (1), 161–177.
Wolch, J.R., Philo, C., 2000. From distributions of deviance to definitions of
difference: past and future mental health geographies. Health & Place 6,
Wong, Y.L.I., Solomon, P.L., 2002. Community integration of persons with psy-
chiatric disabilities in supportive independent housing: a conceptual model
and methodological considerations. Mental Health Services Research 4 (1),
Wong, Y.L.I., Filoromo, M., Tennille, J., 2007. From principles to practice: a study of
implementation of supported housing for psychiatric consumers. Administra-
tion and Policy in Mental Health and Mental Health Services Research 34 (1),
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 and Medicine 68, 1376–1387.
Yanos, P.T., 2007. Beyond landscapes of despair: the need for new research on the
urban environment, sprawl, and the community integration of persons with
severe mental illness. Health & Place 13, 672–676.
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.
Yasui, N.Y., Berven, N.L., 2009. Community integration: conceptualization and
measurement. Disability and Rehabilitation 31 (9), 761–771.
Wolch, J.R., 1980. Residential location of the service-dependent poor. Annals of the
Association of American Geographers 70 (3), 330–341.
Wright, P.A., Kloos, B., 2007. Housing environment and mental health outcomes:
a levels of analysis perspective. Journal of Environmental Psychology 27 (1),
Zippay, A., Thompson, A., 2007. Psychiatric housing: locational patterns and
choices. American Journal of Orthopsychiatry 77 (3), 392–401.
S. Metraux et al. / Health & Place 18 (2012) 621–629