More than a decade has passed since the World Health Organization redefined its classification of impairment, disability, and health to reflect a biopsychosocial model of disability (
1). As a result, the International Classification of Functioning, Disability and Health (ICF) now emphasizes multiple dimensions of functioning and disability as experienced at the body, individual, and societal levels. As part of this framework, the ICF classifies environmental factors in an effort to capture potential barriers and facilitators to social, physical, and institutional functioning and their role in contributing to disability (
2). The classification promotes the assessment and documentation of the environment’s positive or negative impact on functioning as well as the need for environmental interventions to remediate restrictions to participation in educational, economic, social, cultural, and political activities. Despite advances in policy and research related to environmental barriers and adaptations, significant barriers and knowledge gaps remain (
3).
The paucity of research in this area is likely a contributing factor to the dearth of interventions addressing community participation of persons with serious mental illness. In fact, direct application of environmental factors classified by ICF in mental health research remains limited (
4). This may in part be explained by inadequacies in the conceptualization of the ICF environmental factors and difficulties in measuring them (
5). Notably, however, the International Mental Health Task Force was actively involved in developing the ICF to ensure that all aspects of the revised classification apply to persons with mental illness (
6).
Despite these shortcomings, according to Baumgartner and Susser (
7), the ICF provides a useful framework for research on community integration of persons with serious mental illness, and it offers a conceptual basis for the growing body of research on the relationship between neighborhood characteristics and integration. Although the conceptualization of community integration has evolved over time (
8), we use a broad definition of community integration as “the opportunity to live in the community and to be valued for one’s uniqueness and abilities like everyone else” (
9). Prior studies have shown community integration to be positively associated with working-class neighborhoods and neighborhood racial-ethnic diversity (
10), negatively associated with neighborhood disadvantage and crime (
11,
12), and both positively and negatively associated with immigrant concentration (
11,
13). These studies included a limited set of modifiable socioeconomic neighborhood indicators, relatively small samples of persons with serious mental illness, and select neighborhoods of an urban area and, therefore, have had little influence on policy or neighborhood improvements. Furthermore, earlier studies defined neighborhoods by zip code rather than by a smaller geographic unit, such as the block group used by the U.S. Census Bureau, which is a more appropriate proxy for a neighborhood.
Studies of the environment in which persons with serious mental illness live need to include a more diverse spectrum of factors and a unit of analysis that will enhance interventions and ultimately promote health rather than disability. Thus the purpose of this study was to establish a broader foundation from which to study and develop interventions that remove barriers and activate facilitators of participation by persons with serious mental illness. Specifically, consistent with the ICF’s emphasis on environment, this study used objective measures of environmental factors to determine whether physical and social characteristics and crime among neighborhoods in Philadelphia where adults with serious mental illness live differ from those of neighborhoods representative of the city’s general population. It also examined the relationship between neighborhood concentration of persons with serious mental illness and these neighborhood characteristics.
Methods
Study groups
Data for mental health consumers with serious mental illnesses were obtained under a business associate agreement with Community Behavioral Health (CBH), a publicly run managed care organization that funds behavioral health services for Medicaid recipients in Philadelphia. The data set consisted of 15,246 individuals, ages 18–64, who were eligible for Medicaid throughout 2000 and who had a claim for at least one inpatient or two outpatient Medicaid-reimbursed services between 1997 and 2000 that were related to diagnoses of affective disorder (
ICD−9 code 296) or schizophrenia (code 295). These are common criteria for identifying serious mental illness in Medicaid claims data (
14–
16), and their accuracy has been validated (
17). Geographic information systems (GIS) software was used to match the addresses of individuals with serious mental illness in the CBH data set to the U.S. Census block group in which their residence was located. GIS software was also used to create a comparison group of 15,246 randomly generated address points that were representative of the general Philadelphia population. These points were generated such that the number of points in each block group was proportional to that block group’s share of the overall Philadelphia population. A more detailed description of the procedures used to generate these points is available elsewhere (
15). This study was approved by the institutional review board at the University of Pennsylvania.
Neighborhood-level characteristics
U.S. Census block groups were chosen as the unit of analysis for assessing neighborhood-level characteristics, with Census 2000 identifying 1,816 block groups in the City of Philadelphia. Block groups are geographic units comprising clusters of several street blocks. They are the smallest geography for which the Census Bureau makes available aggregated data and have been previously used as appropriate neighborhood proxies (
18,
19).
Three data sources were used to obtain an array of variables at the block group level. A number of variables were obtained from the Census 2000 data file, and additional variables, also from the year 2000, were extracted from databases that are maintained by the Cartographic Modeling Lab at the University of Pennsylvania. One of these databases provides crime statistics from the Philadelphia Police Department’s incident transmittal system, which includes records of all incidents to which police responded. An additional database integrates data from the Philadelphia Department of Licenses and Inspection, the Philadelphia Revenue Department, the public gas utility (Philadelphia Gas Works), and the Board of Revision of Taxes, the public entity that assesses property values in Philadelphia.
A set of 22 variables measuring the crime, physical and structural, economic, housing market, and social characteristics of neighborhoods was selected from these data sources.
Neighborhood concentration of serious mental illness
The CBH data set was matched with general population data from Census 2000 in order to assess the concentration of persons in the data set who were living in each block group in Philadelphia. This was calculated for each block group by using the number of residents in the CBH data set as the numerator and the Census 2000 estimate of the number of residents ages 18 to 64 as the denominator.
Analysis procedures
Thirty-one block groups had populations of zero and were dropped from subsequent analyses. Consequently, two persons in the serious mental illness group and eight address points in the comparison group erroneously linked to these block groups were excluded from the analysis.
To identify latent neighborhood characteristics, we subjected the 22 neighborhood characteristics to principal-components analysis by using varimax rotation. To identify the ideal-factor solution, we included components that met the following three criteria: they satisfied Cattell’s (
20) scree test, they had eigenvalues greater than 1 (the Kaiser [
21] criterion), and they were conceptually comprehensible and interpretable.
After we conducted the principal-components analysis, scores for the retained components were computed for each block group. All records of persons with serious mental illness and of the comparison group were then matched with the component scores for their corresponding block groups, such that all records associated with a particular block group were assigned the same scores.
Independent-sample t tests were used to compare the mean values of each of the principal components of persons with serious mental illness and of the general population comparison points. Ordinary least-squares (OLS) regression was used to assess the strength of the association between neighborhood concentration of persons with serious mental illness and the complete set of principal components after analyses controlled for the proportions of residents in the block group who were ages 18–64, nonwhite, and male. Because neighboring block groups are likely to have similar characteristics, the Moran’s I statistic was used to assess whether there was significant spatial autocorrelation in the OLS residuals. A spatial lag regression model using the Queen weight matrix was then estimated to adjust for spatial dependencies. The results of this model and the OLS model were compared to check for any substantive differences when accounting for potential bias introduced by the presence of spatial autocorrelation. All spatial statistical tests were carried out with the Geoda software (
22).
Results
Neighborhood characteristics
Table 1 presents information about each variable, including data source, unit of measurement, and means and standard deviations for all block groups in Philadelphia.
Table 2 provides the results of the principal-components analysis. By using the criteria established for identifying the ideal-factor solution, a six-component solution explaining 62% of the variance was identified. [A figure illustrating the scree plot of the eigenvalues for all 22 factors is available online as a
data supplement to this article.]
The first and second components were labeled serious crime and delinquent crime, respectively. Both arson and narcotics arrests loaded on the third component, drug-related activity. Although these two indicators may appear unrelated, there is evidence of a higher incidence of arson in areas with elevated drug activity (
23,
24). The fourth component, physical and structural inadequacy, described the physical and socioeconomic environment of neighborhoods. The fifth component—which identified block groups with higher proportions of renter-occupied households, households that had recently moved, and one-person households—was interpreted as an indicator of neighborhood social instability. The final component, labeled social isolation, described neighborhoods that were more commercial and industrial and less residential, thereby potentially providing less opportunity for social interaction than might occur in more residential areas.
Comparison of neighborhood characteristics
Table 3 presents a comparison of neighborhoods of persons with serious mental illness and of the comparison group. The neighborhoods in which persons with serious mental illness resided were found to have significantly higher levels of physical and structural inadequacy. Similarly, persons with serious mental illness lived in neighborhoods with more serious crime, higher amounts of delinquent crime, and far more drug-related activity than the neighborhoods of the comparison group. Persons with serious mental illness lived in neighborhoods with higher amounts of social instability and social isolation than the neighborhoods of the comparison group, but the difference between the neighborhoods was not statistically significant.
On closer inspection, the biggest disparity between the neighborhoods of the two groups was in their physical and structural characteristics. Average component scores for both serious crime and drug-related activity were also markedly higher in the neighborhoods of persons with serious mental illness than in the neighborhoods of the comparison group. By contrast, the difference was not nearly as large for the delinquent-crime component.
Neighborhood and residents with serious mental illness
Table 4 presents the results of the regression analysis assessing the association between neighborhood concentration of persons with serious mental illness and neighborhood characteristics. In the OLS model, Moran’s I (.25) was significantly greater than zero, which indicated the presence of spatial autocorrelation. The Moran’s I value (–.04) for the spatial lag model was slightly negative and suggested less spatial autocorrelation, although the finding was still statistically significant.
There were no other substantive differences between the OLS and the spatial lag models in terms of the relationship between the neighborhood measures and the concentration of persons with serious mental illness. Consistent with the comparison of neighborhoods of persons with serious mental illness and of the general population, both models found that serious crime, drug-related activity, and physical and structural inadequacy were positively associated with neighborhood concentration of persons with serious mental illness, even after the analyses controlled for additional demographic characteristics of the neighborhoods. However, whereas no significant differences were found in social instability and social isolation between neighborhoods of persons with serious mental illness and neighborhoods of the comparison group, the regression analysis showed that higher levels of these indicators were associated with higher neighborhood concentration of persons with serious mental illness.
Discussion
Drawing on the ICF’s emphasis on assessing the impact of environmental factors on functioning and disability, this study found that on the whole, adults with serious mental illness lived in neighborhoods that were more disadvantaged than neighborhoods of a comparison group representing the general population, a finding that is consistent with prior research (
25). Specifically, neighborhoods in which adults with serious mental illness resided were found to have higher levels of physical and structural inadequacy, more drug-related activity, and higher levels of crime. Moreover, the regression models indicated that higher levels of physical and structural inadequacy, crime, drug- related activity, social instability, and social isolation were associated with higher neighborhood concentration of persons with serious mental illness. Notably, the associations observed in the regression models between less desirable neighborhood conditions and neighborhood concentration of persons with serious mental illness did not appear to be solely the result of the concentration of persons with serious mental illness in poor or economically disadvantaged neighborhoods. Apart from delinquent crime, all other components were found to have a significant association with neighborhood concentration of serious mental illness, even after controlling for the physical and structural inadequacy component, which included both median income level and median home value.
The divergent results related to social instability and social isolation found by the regression models and the comparison of neighborhoods with or without high concentrations of persons with mental illness bear mentioning. These components retained a significant association with neighborhood concentration of persons with serious mental illness even after the estimation of an additional set of regression models that included a measure of neighborhood population density as a possible confounder. Thus although the study findings provide strong support for the difference in the physical and crime environments of neighborhoods with or without high concentrations of persons with serious mental illness, additional research should examine the social environments of these neighborhoods more closely.
Although this study was unable to directly test the relationship between neighborhood environmental context and the functioning of persons with serious mental illness, its findings are nonetheless concerning in light of emerging evidence that worse neighborhood characteristics impede the community integration of persons with serious mental illness (
11). To date, the still nascent body of research in this area has yet to ascertain the extent to which a worse physical or crime environment may serve as an impediment to community integration of persons with serious mental illness. Thus although it is not possible to draw direct implications from the current study about the impact of neighborhood factors on community integration, the findings presented here provide a broader foundation for future research to identify barriers and facilitators to community participation of persons with serious mental illness.
The need for research on the relationship between neighborhood characteristics and community integration is especially important, given that housing interventions for persons with serious mental illness continue their shift away from congregate housing and toward scattered-site independent housing units in the community (
26). Despite evidence that persons with serious mental illness who live in independent housing have better community integration outcomes (
13), no study to date has examined the interaction between housing type, neighborhood characteristics, and community integration. Additional research in this area could help refine housing approaches to promote better community integration of persons with serious mental illness.
Future studies on the relationship between environmental context and the functioning and participation of persons with serious mental illness could directly test the same set of factors considered in this study. Doing so would result in greater comparability across studies and allow for the identification of concrete factors that could serve as intervention points.
One limitation of this study was its use of randomly distributed geographic points, rather than actual persons, as a comparison group. Although person-level data were available for the serious mental illness group, without knowledge of the demographic characteristics of the comparison group it was not possible to control for person-level factors that might mediate associations between diagnosis of serious mental illness and neighborhood indicators. Moreover, this study did not assess differences between persons with schizophrenia and those with affective disorders. Given that environmental factors may have different influences on these two groups, such a comparison remains an important goal for future research. As an additional limitation, the data used in this study are now more than a decade old. Although we have no reason to believe that the use of more recent data would yield substantively different results, our results should nonetheless be interpreted cautiously in light of the age of the data. Finally, this study did not include a comparison group of adult Medicaid recipients without a serious mental illness diagnosis. Although such a group may have formed an appropriate and important comparison group, the intent of this study was to compare persons with serious mental illness and the normative, general population. We felt that making this comparison was especially important in the context of a community integration framework and the preference for independent, scattered-site housing, insofar as such a comparison might point to opportunities and challenges in the current environment of services for persons with serious mental illness.
Conclusions
This study found differences in physical and crime characteristics among the neighborhoods of a large group of persons with serious mental illness and a comparison group representing the general population. These factors, as well as neighborhood social characteristics, were also associated with neighborhood concentration of persons with serious mental illness. These findings establish the importance of further exploration of the degree to which environmental factors may act as either barriers or facilitators to the functioning and participation of persons with serious mental illness. This is especially true of physical characteristics, for which the largest effects were observed. Studying these factors in the broader context of the ICF could do much to help identify new areas of intervention intended to promote the functioning and integration of persons with serious mental illness.
Acknowledgments and disclosures
The contents of this article were developed under a grant to Temple University from the United States Department of Education (National Institute on Disability and Rehabilitation Research [NIDRR] grant H133B100037, Dr. Salzer, principal investigator). The authors thank the Philadelphia Department of Behavioral Health and Intellectual Disability Services for its support and for data that were integral to this study as well as the Center for Mental Health Policy and Services Research at the University of Pennsylvania for its assistance with this study. The contents of this article do not necessarily represent the policy of the Department of Education or reflect the endorsement of the federal government. NIDRR had no direct involvement in any aspect of conducting, reporting, or disseminating this research.
The authors report no competing interests.