Housing Attributes and Serious 2001

Dedicated to
Mental Illness: Implications Evidence-
Based
for Research and Practice Psychiatry

Sandra J. Newman, Ph.D.

Objectives: This paper critically reviews studies of the relationship be- ditions of the dwelling unit or the
tween housing attributes and serious mental illness, highlights important quality of the neighborhood.
gaps in the research, generates hypotheses to be tested, and suggests a This paper, which draws on some of
research agenda. Methods: Studies published between 1975 and March my previous work (3), provides a crit-
2000 were identified through computerized searches, previous literature ical review of this smaller body of re-
reviews, and consultation with mental health and housing researchers. search on the housing setting. The
Criteria for inclusion included the presentation of quantitative evidence, goals of the review were to identify
a systematic sample of known generalizability, and systematic analytic findings, highlight important gaps in
techniques. Results and conclusions: The 32 studies that met these crite- the research, generate hypotheses to
ria relied on one or more of three conceptualizations of the role of hous- be tested, and suggest a research
ing: housing attributes or assessments as an outcome or dependent vari- agenda.
able; housing attributes as inputs or independent variables in a model in
which the outcome pertains to a nonhousing factor, such as a mental Methods
health outcome; or housing as both an input and an outcome. Three stud- The review is limited to studies that
ies found no long-term effect of improved housing adequacy on housing used specific measures of the attrib-
satisfaction above and beyond case management. Three studies found utes of housing and neighborhoods.
better outcomes for settings that have fewer occupants. Another study Much of the literature generally con-
suggested that persons who live in small-scale, good-quality, noninstitu- sidered to fall under the “housing”
tional environments are less likely to engage in disruptive behavior when rubric was therefore excluded, be-
a larger proportion of other tenants also have serious mental illness. The cause it focuses on programmatic in-
strongest finding from the literature on housing as an input and an out- terventions in the service or social en-
come was that living in independent housing was associated with greater vironments of different housing set-
satisfaction with housing and neighborhood. Most of the studies had tings (4–6). The review is also largely
methodological weaknesses, and few addressed key hypotheses. There is limited to research that describes the
a critical need for a coherent agenda built around key hypotheses and for specific service context of the study
a uniform set of measures of housing an input and an outcome. (Psychi- subjects. The types of housing set-
atric Services 52:1309–1317, 2001) tings ranged from group homes to in-
dependent apartments. Including
only studies that described the specif-

T
hree-fourths of the 4.6 mil- which individuals and under what ic service context allowed comparison
lion people with severe and conditions. Most of this work contin- of housing settings and neighbor-
persistent mental illness now ues the traditional—and impor- hoods with particular features across
live most of their lives in the com- tant—focus on the services and different communities while other
munity. An as yet unmeasured num- treatment interventions that people potentially important factors, such as
ber of these individuals reside in in- with mental illness may require for functioning, were held constant.
dependent housing and receive no community living; however, a small Studies of persons who were domi-
on-site service supports (1,2). Re- body of research considers the hous- ciled and who were previously home-
searchers have devoted substantial ing setting itself—the relationship less were included for the same rea-
attention to understanding which between an individual’s psychologi- son: to examine whether, when other
community settings work best for cal well-being and the physical con- factors are held constant, homeless-
ness is associated with different re-
sponses to features of the physical en-
The author is with the Institute for Policy Studies at Johns Hopkins University, Wyman vironment.
Park Building, 3400 North Charles Street, Baltimore, Maryland 21218 (e-mail, sjn@ Research on the mental health-
jhu.edu). housing nexus is relatively new. Thus
PSYCHIATRIC SERVICES ♦ October 2001 Vol. 52 No. 10 1309
the criteria for inclusion of a study in
this review were the minimum re-
quired to guard against spurious find-
ings: quantitative evidence, a system-
atic sample of known generalizability,
and analytic techniques that meet ba-
sic standards of scientific rigor—for
example, the application of statistical
techniques that are appropriate to the
question posed. It is important to
note that these criteria reduce, but by
no means eliminate, idiosyncratic or
artifactual findings.
Studies conducted between 1975 and
March 2000 were identified through
extensive searches, including com-
puterized searches using MEDLINE
and PsycINFO, an examination of pre-
vious literature reviews, and consulta-
tions with experts in mental health
and in housing research. A total of
280 studies on the broad topic of
housing and mental illness were iden-
tified, and of those, 32 were selected
for review. Studies that met the above
criteria were included even if their
primary goal was not to gain an un-
derstanding of the role of housing in
mental health. Although their weak-
nesses from the housing perspective
are understandable, their inclusion
made it possible to distill what could
and could not be learned from them
about housing and mental health and
to gain insight into appropriate ana-
lytic methods.
Results
The studies reviewed here explicitly
or implicitly relied on one of three
conceptualizations of the role of
housing. The first conceptualization
sees housing attributes—size of
dwelling, physical quality, and neigh-
borhood—as an outcome or a de-
pendent variable to be explained. An
example of a study relying on this
conceptualization would be one that
examined whether persons with seri-
ous mental illness are more likely to
live in distressed neighborhoods, net
of other factors—such as income—
that are usually associated with the
distribution of the general population
in neighborhoods of varying quality.
The second conceptualization sees
housing attributes as inputs or inde-
pendent variables in a model in which
the outcome pertains to a nonhousing
factor—for example, a person’s func-
1310
tioning or length of hospital stay. The
third approach sees housing attrib-
utes as both input and outcome—for
example, the effects of the physical
adequacy of the dwelling on its af-
fordability. A study may encompass
more than one conceptualization.
Table 1 lists the studies included in
this review (7–33) by type of concep-
tualization and summarizes key char-
acteristics of the study samples.
Housing as an outcome
Table 2 summarizes the key design
features of the six studies that exam-
ined housing as an outcome (7–12).
The mean age of the individuals in
the samples was the late 30s; about
half of them had a diagnosis of schiz-
ophrenia. The samples had more
males than females (range of male
subjects, 55 percent to 68 percent).
The proportion of whites in the sam-
ples ranged from 24 percent to 85
percent. Substance abuse was not re-
ported for two studies; in the other
four studies it varied widely.
All of the studies except one fo-
cused on the homeless population,
from individuals who were precari-
ously housed to those who literally
had no place to go at night. Observa-
tion periods ranged from a single
cross-section to a follow-up period of
24 months. Two studies (11,12) exam-
ined the impact of assertive case man-
agement on housing satisfaction,
which allowed the results to be com-
pared. The other four studies ad-
dressed research questions that were
unique to each, so no comparisons
could be made.
In a study of outpatients of a men-
tal health agency in King County,
Washington, Uehara (7) found that
race was more powerful than mental
illness in predicting quality of housing
and neighborhood. This finding is
consistent with that reported by New-
man and Reschovsky (33). However,
the generalizability of the King Coun-
ty study is constrained by a localized
sample, the reliance on cross-section-
al data, the absence of any informa-
tion about support networks or serv-
ice use, and relatively few indicators
of housing quality and community
context.
The results of three studies (9,11,
12) suggest that more adequate hous-
PSYCHIATRIC SERVICES
ing has no long-term effect above and
beyond case management on housing
satisfaction among persons with seri-
ous mental illness. In the study by
Hough and colleagues (9), about 90
individuals in San Diego were ran-
domly assigned to one of four groups.
One group had access to Section 8
rent subsidies and received compre-
hensive case management, the sec-
ond group had access to Section 8
subsidies and received traditional
case management, the third group re-
ceived comprehensive case manage-
ment only, and the fourth group re-
ceived traditional case management
only. (A Section 8 rent subsidy can be
used to rent physically adequate and
affordable housing in the private mar-
ket. The tenant’s out-of-pocket rent
payment is limited to no more than 30
percent of income.)
In the study by Lehman and col-
leagues (11), which was conducted in
Baltimore, 77 individuals in the ex-
perimental group were provided with
assertive community-based clinical
treatment combined with intensive
case management and advocacy; 75
control subjects received “services as
usual.” Goering and colleagues (12),
using a synthetic pre-post design, in-
terviewed a sample of participants in
the Toronto Hostel Outreach Pro-
gram who were provided with as-
sertive case management with dormi-
tory-style accommodations.
The results of these three studies
are not conclusive about whether
housing features affect residents’ sat-
isfaction with housing. Hough and col-
leagues (9) noted that the two types of
case management—traditional and
comprehensive—converged over time,
so that after 12 months any differ-
ences in case management services
were reduced or eliminated. Further-
more, their analysis focused on access
to rental subsidies rather than use of
subsidies. Thus two different sets of
factors could have influenced the re-
sults of the study: factors that affected
whether a rental subsidy was used
and factors that affected the type of
housing and neighborhoods that were
attainable with a Section 8 certificate,
such as ethnic and racial discrimina-
tion and tightness in the housing mar-
ket. The information necessary to un-
derstand the effects of housing attrib-
♦ October 2001 Vol. 52 No. 10
Table 1
Studies included in the literature review, by type of housing conceptualization, and selected characteristics of the study
samples1

Characteristics of study samples

Percent with Percent

Conceptualization Mean age Percent Percent diagnosis of substance Percent
and study (years) male white schizophrenia abusers homeless

Housing as an outcome
Uehara (7) 40 55 85 52 0
Mowbray and Bybee (8) 38 57 59 29 100
Hough et al. (9) 36 67 63 55 100
Hurlburt et al. (10) 67 63 55 54 100
Lehman et al. (11) 38 68 24 45 71 100
Goering et al. (12)2 39 58 70 86 36 100
Housing as an input
Baker and Douglas (13) 56 44 86 65
Coulton et al. (14) 46 48 78 63 0
Linn et al. (15) 45 100 75 0
Nagy et al. (16) 45 48 87 0
Davies et al. (17)3 50 30 90 100 0 0
Depp et al. (18)4 38 35 64
Earls and Nelson (19)4 38 60
Nelson et al. (20) 37 58 100 0 0
Lipton et al. (21) 37 65 82 100
Trute and Segal (22) 0
Segal et al. (23) 0
Dixon et al. (24) 41 67 27 49 73 100
Goldfinger et al. (25)5 38 71 50 50 50 100
Goldfinger et al. (26)6 38 72 45 50 100
Dickey et al. (27)7 37 70 50 72 100
Dickey et al. (28) 37 71 46 70 >70 100
Wood et al. (29)8 66 62 55 100
Newman et al. (30)9 36 42 60 54 26 13
Newman et al. (31) na na na na na na
Housing as an input and an outcome
Hurlburt et al. (32) 67 63 55 54 100
Newman and Reschovsky (33)10 36 42 34
1 Data are based on the total sample, including, where relevant, experimental and control or multiple comparison groups, and were obtained directly
or calculated from group percentages. Articles that contain multiple conceptualizations are presented only once under the first model to which they
apply; thus, Hough et al. (9) is not repeated under Housing as an input, and four studies (6,9,11,30) are not repeated under Housing as an input and
an outcome.
2 Age is median years.
3 The study indicates that subjects were not suffering from alcoholism but does not report on drug abuse.
4 Placed patients were hospitalized before receiving Section 8 placement; it is not clear where patients on the waiting list were housed.
5 The figures for percent white, percent with schizophrenia, and percent substance abusers are approximate.
6 The figure for percent substance abusers is approximate.
7 The figure for percent white is approximate.
8 The figure for percent with schizophrenia is from the original sample of 362 individuals; figures for all the other categories are based on the study sub-

sample of 316 individuals.

9 Percentage of African Americans provided in the paper. The percentage reported in the table assumes that all non-African Americans are white and

is for Hamilton County only (Baltimore not reported). The figures for percent substance abusers and percent homeless are for Baltimore only; the fig-
ures for Cincinnati were not reported.
10 The figure for percent white was for Baltimore only; the figure for Cincinnati was not reported.

utes—that is, whether a person used a
subsidy—was not part of this study.
Lehman and colleagues (11) ac-
knowledged two sources of contami-
nation in their study that introduced
opposing types of bias of unknown ex-
tent. First, funding for their experi-
ment included additional housing op-
tions, which were available to individ-
uals in both the experimental group
PSYCHIATRIC SERVICES ♦ October 2001 Vol. 52 No. 10
and the control group. Most of these
options were ultimately assigned to
those in the experimental group;
therefore, an upward bias in measur-
ing the impacts of assertive communi-
ty-based clinical treatment would be
expected (24). Second, during the
course of the study, the City of Balti-
more funded an intervention similar to
assertive community-based clinical
treatment at another community men-
tal health center. These additional
services were available to individuals
in the control group, so that a down-
ward bias would be expected in meas-
ures of the impacts of such treatment.
The nonexperimental design of the
study by Goering and colleagues (12)
and the fact that persons who
dropped out were less residentially
1311
Table 2
Design features of studies in the literature review that used housing as an outcome
Study Study type
Uehara (7) Cross-sectional
survey
Mowbray Longitudinal,
and Bybee (8) 12 months
Hough et al. (9) Experimental,
18 months
Hurlburt et al. (10) Experimental,
24 months
Lehman et al. (11) Experimental,
12 months
Goering et al. (12) Pre-post,
18 months
1 181 in the experimental groups, 181 in the control groups
2 Viewed as an outcome of access to rental subsidy
3 77 in the experimental group, 75 in the control group
stable than those who remained in
the study call into question the gener-
alizability of the results. The residen-
tial instability of the dropouts may ex-
plain why these results were more
positive than those obtained in the ex-
perimental study by Lehman and col-
leagues (11). Other weaknesses of the
study by Goering and colleagues in-
cluded the reconstruction of prein-
tervention status through retrospec-
tive reporting and measurement of
housing satisfaction for the setting
where the subject had lived the
longest rather than for all settings in
which the individual had lived since
the last interview.
A central issue in the conceptual-
ization of housing as an outcome is
the relative importance of housing
versus services for an outcome such
as housing satisfaction. Therefore,
the apparent lack of association be-
tween housing adequacy and housing
satisfaction above and beyond case
management is worthy of additional
scrutiny. Ideally, the research would
follow an experimental design similar
1312
N Key inputs
517 Clinical status, social skills, demo-
graphics, socioeconomic status,
living within or outside of city limits
163 County of service site, client recruit-
ment site (homeless shelter, mental
health inpatient hospital, existing
community mental health aftercare
caseload), age of client, related vari-
ables such as intensity and duration
of service
3621 Assistance received from case man-
agers in areas such as activities of
daily living, linkage, mental health-
related activities
3621 Access to Section 8 certificate; demo-
graphics; whether client had alcohol
or drug dependence, neither, or both
1523 Housing status, service use
71 Demographics, clinical history, history
of homelessness, physical adequacy
rating of housing, subject satifaction
with housing, service usage, case
manager-client relationship
to that used by Hough and colleagues
(9) but with a focus on residence in
decent, affordable housing and with
clear distinctions between case man-
agement interventions. From a public
policy perspective, the relationship
between housing satisfaction, mental
health outcomes, and service costs
also needs to be understood. Al-
though the quality-of-life benefits of
improved housing satisfaction can be
justified on humanitarian grounds,
public investments based solely on
improved satisfaction will remain dif-
ficult to justify unless it can be
demonstrated that satisfaction is cost-
effectively related to improved men-
tal health outcomes.
Housing as an input
The group of studies based on the
conceptualization of housing as an in-
put is by far the largest. Table 3 sum-
marizes the key design features of the
20 studies in this group (9,13–31). The
characteristics of the samples vary
widely, with the exception of mean
ages, which ranged from 36 to 56
PSYCHIATRIC SERVICES
Key outcomes
Physical safety (unit physically
unsafe or located on unsafe
block); permanency (unit
rented by day or week, short-
term or limited arrangement)
Housing type (independent alone
or with others, supervised de-
dependent, homeless), follow-
up housing status as predicted
by service and other variables
Duration of certificate use2; resi-
dence in independent housing;
satisfaction with residence,
neighborhood, safety
Three housing categories (stable
independent, other consistent,
unstable)
Days homeless, housing satis-
faction
Housing status, stability, and
satisfaction; adequacy of hous-
ing conditions
years but were concentrated in the
late 30s. In most of these studies, the
subjects were more likely to be male
and white and to have a primary diag-
nosis of schizophrenia. Depending on
the definition of homelessness, seven
to 11 of the studies focused on indi-
viduals who were homeless and men-
tally ill. Seven studies reported rates
of substance abuse, which generally
exceeded 50 percent. Study designs
and duration ranged from a single
cross-section to an experimental study
with a 36-month follow-up.
Many studies in the housing-as-in-
put group examined the same ques-
tions, which provided a basis for com-
paring their results. The topic that
generated the largest concentration of
research—five studies—was the ef-
fect on mental health outcomes of ac-
cess to or use of a Section 8 housing
certificate, either by itself or in combi-
nation with a mental health service in-
tervention (9,18,24,29,30). Four stud-
ies examined the effects of neighbor-
hood features on mental health out-
comes (16,22,23,30). Three looked at
♦ October 2001 Vol. 52 No. 10
Table 3
Design features of studies in the literature review that used housing as an input

Study Study type N Key inputs Key outcomes

Baker and Longitudinal, 844 Housing type, physical conditions, Unmet service needs, change in func-
Douglas (13) nine months adequacy and appropriateness tioning and maladaptive behavior,
for client perceived quality of life
Coulton Longitudinal, 51 Privacy Stable functioning
et al. (14) three months
Linn et al. (15) Longitudinal, 210 Number of occupants, distance Changes in social dysfunction
four months from hospital, rural versus urban,
rating of cost and quality of home
Nagy et al. (16) Cross-sectional 1,148 Number of residents, cost, predominant Activity levels during previous month
survey land use in neighborhood, urban inside and outside home, produc-
versus nonurban tive activity
Davies et al. Cross-sectional 1241 Physical housing conditions, neigh- Psychopathology, social adjustment,
(17) survey borhood conditions global functioning, quality of life
Depp et al. Cross-sectional 462 Living independently in community Symptom severity, use of formal sup-
(18) survey using Section 8 certificate versus ports, size and composition of in-
living in some other arrangement formal support network, community
integration
Earls and Convenience 89 Index of housing quality Emotional well-being
Nelson (19) sample
Nelson et al. Longitudinal, 173 Housing concerns, residence size, Emotional well-being, personal em-
(20) one year privacy powerment
Lipton et al. Experimental, 493 Physical housing adequacy scale, sat- Residential stability, extent of hos-
(21) one year isfaction with and commitment to pitalization, severity of psychopathy,
living arrangement quality of life
Trute and Cross-sectional 2274 Total occupied units, tenure of units, Index of social integration
Segal (22) survey and percentage of crowded units in
census tract
Segal et al. (23) Cross-sectional 397 Four neighborhood types based Index of external social integration
survey on sociopolitical attributes
Dixon et al. Longitudinal, 77 Use of Section 8 certificate Housing arrangement after one year
(24) one year
Goldfinger Experimental, 1185 Residence in evolving consumer house- Clinical status, symptoms, functional
et al. (25) 18 months hold or independent apartment with impairments, social functioning, life
weekly case manager visit satisfaction, substance abuse
Goldfinger Experimental, 1185 Type of residence preferred by client, Housing status, days homeless
et al. (26) 18 months housing recommended by clinician,
social background indicators
Dickey et al. Experimental, 1126 Case management and substance treat- Housing status, housing stability, asso-
(27) 18 months ment service, mental health treatment, ciation between housing stability and
inpatient admission information service use
Dickey et al. Experimental, 1126 Costs of housing, treatment, community Housing costs, housing stability
(28) 18 months support services, and case management
Hough et al. Experimental, 3627 Residence with versus without Section Income, quality of life, length of psy-
(9) 18 months 8 certificate chiatric hospitalizations, levels of de-
pression, health, and functional status
Wood et al. (29) Experimental, 3627 Frequency of family interaction, family Better mental health and housing out-
three years satisfaction scale, perceived availability comes leading to improved family
of family support, mental health relations
status, housing status
Newman et al. Pre-post, 2698 Affordable housing, housing and neigh- Number of hospitalizations, change of
(30) three years borhood problems, housing and lengths of stay in hospital, unmet
neighborhood rating scales service needs
Newman et al. Longitudinal 1539 Amenities in unit; neighborhood quality; Disruptive behavior as measured by
(31) age, structural integrity, and financial maintenance and operating costs of
characteristics of property; proportion housing units
of mentally ill tenants in property
1 78 in an urban clinic, 46 in a rural clinic
2 24 Section 8 placements, 22 on Section 8 waiting list
3 26 in the experimental group, 23 in the control group
4 129 in the California sample, 98 in the Saskatchewan sample
5 63 in the experimental group, 55 in the control group
6 61 in the experimental group, 51 in the control group
7 181 in the experimental groups, 181 in the control groups
8 131 in Hamilton County, 138 in Baltimore
9 Unit of observation is buildings, not people

PSYCHIATRIC SERVICES ♦ October 2001 Vol. 52 No. 10 1313

Table 4
Design features of studies in the literature review that used housing as an input and an outcome

Study Study type N Key inputs Key outcomes

Depp et al. Cross-sectional 461 Descriptive data not controlled Housing satisfaction, safety, comfort, prox-
(18) survey in analysis imity to hospital, roominess, quiet,
privacy, affordability, Section 8 rating
Uehara (7) Cross-sectional 517 Clinical status, social skills, demo- Physical safety (unit physically unsafe or
survey graphics, socioeconomic status, located on unsafe block), permanency
living inside or outside of city limits (unit rented by day or week, short-
term or limited arrangement)
Hurlburt et al. Experimental, 3622 Access to Section 8 certificate versus Living arrangement (for example, inde-
(32) 24 months no access to Section 8 certificate pendent housing, shelter, street), resi-
dential stability, time required to
achieve stable housing
Hough et al. Experimental, 3622 Experimental versus control Section 8 certificate use, duration of
(9) 18 months conditions certificate use, residence in indepen-
dent housing, satisfaction with resi-
dence, neighborhood, safety
Newman Pre-post, 2693 Affordable housing (30 percent of in- Physical housing conditions (problems
et al. (30) 36 months come or less), housing problems with space or privacy, plumbing, floors,
(including interior, rats, heating, ceilings, walls, windows, rats, heat,
exterior), 10-point housing rating furnishings, exterior), neighborhood
scale, neighborhood problems problems (street noise, traffic, security,
(including traffic, security, crime, upkeep, crime, availability or conven-
availability of selected services), ience of services)
10-point neighborhood rating scale
Newman and Cross-sectional 2674 Use of Section 8 certificate Neighborhood quality (median owner-
Reschovsky survey occupied house value, median rent,
(33) median income, percent blacks in tract,
percent owner-occupied, percent va-
cant, percent boarded up, percent
crowded, percent built before 1940,
number of Part I crimes)
1 24 Section 8 placements, 22 on Section 8 waiting list
2 181 in the experimental groups, and 181 in the control groups
3 131 in Hamilton County, 138 in Baltimore
4 137 in Baltimore, 130 in Cincinnati

urban versus rural location (15–17),
three focused on the number of resi-
dents in the dwelling (15,16,20), and
three examined the effects of
dwelling quality (13,21,30).
The core question of the housing-
as-input conceptualization is whether
the housing setting has therapeutic
benefits that operate independently of
the type, array, and intensity of servic-
es provided. The two studies that ad-
dressed this question most directly
produced different results. Baker and
Douglas (13) analyzed data from a
random sample of 844 clients in com-
munity support services programs in
New York State (excluding New York
City). The subjects were stratified by
time in the program, but they were
not randomly assigned to housing.
Most of the data were obtained from
questionnaires completed by case
managers at baseline and at nine-
month follow-up.
1314
The authors found that clients who
had a mental illness and also were liv-
ing in physically inadequate housing
were more likely to manifest mal-
adaptive behavior, regardless of the
amount of support services they re-
ceived. Physical inadequacy was
measured by a 5-point rating scale of
neighborhood, residence exterior, res-
idence interior, and client’s personal
property. Living in adequate housing
was associated with an increase in res-
idents’ functioning over the nine-
month study period. This finding is
inconsistent with that of Hough and
colleagues (9), in which access to ad-
equate housing was not associated
with a housing outcome—housing
satisfaction.
Hough and colleagues (9) also found
that access to rental subsidies was not
significantly associated with clinical
outcomes, including level of depres-
sion, time spent in a mental health set-
PSYCHIATRIC SERVICES
ting during the 60 days before each in-
terview, or self-reported functioning
(34). However, obtaining independent
housing—with or without a rent sub-
sidy—was associated with significantly
lower levels of depression and less
time spent in a mental health setting in
the previous 60 days. The physical ad-
equacy of this housing—the housing
indicator in Baker and Douglas (13)—
was not reported.
The design used to test the ques-
tion of the net benefits of the housing
setting was much stronger in the
study by Hough and colleagues than
in the Baker and Douglas study, be-
cause the subjects were randomly se-
lected for both the housing interven-
tion and the case management inter-
vention. Unlike Baker and Douglas,
Hough and colleagues found no sig-
nificant relationship between access
to adequate housing and clinical out-
comes of persons with serious mental
♦ October 2001 Vol. 52 No. 10
illness above and beyond case man-
agement services. Although actual at-
tainment of independent housing was
associated with better mental health
outcomes, selection bias may weaken
the generalizability of this result, be-
cause the randomization was based
on Section 8 access, not use.
Again, both studies have weakness-
es from the specific perspective of
understanding the role of housing
features in mental health outcomes,
separate from other influences. The
Baker and Douglas study was not a
randomized trial, so it is not possible
to determine whether the results
were attributable to housing. The au-
thors also relied primarily on simple
correlational analyses. Most impor-
tant, nearly two-fifths of the individu-
als in the nine-month follow-up sam-
ple who were described as having
“markedly deteriorated” were exclud-
ed from the analysis. The results of
the study would undoubtedly have
been weaker had the entire sample
been included.
Beyond the issue of not distinguish-
ing access to rental subsidies from
their actual use, Hough and col-
leagues also relied solely on correla-
tional analysis, making it unclear
whether improved clinical status fa-
cilitated independent living or whether
independent living, along with case
management, improved clinical sta-
tus. The finding of an association be-
tween attainment of independent
housing and better clinical outcomes
must be confirmed by reanalyzing
the data with a different statistical
modeling approach. If the associa-
tion is confirmed, further testing
with a similar design but implement-
ed in different housing markets and
avoiding contamination of the case
management interventions would be
warranted.
Findings converge in the three
studies that examined the effects of
number of occupants in a housing set-
ting on mental health outcomes such
as social functioning and well-being.
Linn and colleagues (15), Nagy and
colleagues (16), and Nelson and col-
leagues (20) reported that individuals
with mental illness do better in set-
tings that have fewer occupants.
After setting size and other factors
are controlled for, tenant mix also ap-
PSYCHIATRIC SERVICES ♦ October 2001 Vol. 52 No. 10
pears to play a role. Tenants who
lived in settings that had a larger pro-
portion of other tenants with mental
illness had better clinical outcomes
(31). This result might appear to con-
tradict the model of housing that
seeks to normalize housing for indi-
viduals with mental illness because
such housing leads to better out-
comes. However, in the study in ques-
tion (31), the sample of dwellings was
limited to housing that provided
small-scale, good-quality, noninstitu-
tional environments in a community
setting, with off-site services avail-
able to assist with independent living.
Findings on neighborhood attributes
and serious mental illness suggest
that diverse, somewhat disorganized
neighborhoods have salutary effects
on persons with mental illness, be-
cause these types of neighborhoods
are likely to be more welcoming (22,
23,30). If confirmed, this finding
could be useful for developing hous-
ing policies, including the targeting of
subsidies to certain neighborhoods.
Housing as an input
and an outcome
Table 4 summarizes the key features
of the six studies that examined hous-
ing as an input and an outcome (7,9,
18,30,32,33). Because most of these
studies include analyses of the other
two conceptualizations of the role of
housing or neighborhood, they have
already been discussed and therefore
receive briefer mention here. Sample
members were typically in their late
30s. In two of the studies the samples
were composed entirely of individuals
who were homeless and mentally ill
(9,32). In most of the studies males
outnumbered females, with the pro-
portion of male subjects ranging from
35 percent to 67 percent.
The majority of subjects—from 52
percent to 64 percent—had a diagno-
sis of schizophrenia. In four of the six
study samples (7,9,30,32) whites were
in the majority—from 60 percent to
85 percent; in the other two studies
(18,33) blacks constituted the largest
group. Substance abuse data were re-
ported in only two studies (30,32).
The strongest finding from the
studies that examined housing as an
input and an outcome is that living in
independent housing is associated
with greater satisfaction with housing
and neighborhood (9). Confirmation
of this finding is warranted. These
positive associations between a hous-
ing feature—independent housing—
and housing outcomes—satisfaction
with housing and neighborhood—
contrast with the insignificant associ-
ations found in the same study be-
tween comprehensive case manage-
ment or traditional case management
and all quality-of-life domains. As
noted above, individuals in the study by
Hough and colleagues (9) were not ran-
domly assigned independent housing
status; rather, they were randomly as-
signed to groups that did or did not
have access to independent housing
though Section 8 subsidies. Further-
more, the two case management in-
terventions were not very different
from one another, especially after the
first year.
Conclusions
The majority of the 32 studies re-
viewed here suffer from one or more
methodological weaknesses. In addi-
tion, a number of the analyses were
not grounded in a conceptual frame-
work that provided a clear set of hy-
potheses to be tested. Furthermore,
most of the studies relied on correla-
tional analysis, which cannot establish
causation, that is, what affects what.
Even if a positive correlation is the re-
sult of causality, correlation cannot
determine direction or magnitude.
The generalizability of even the ran-
domized experiments is in doubt, be-
cause of such factors as difficulty in
implementing the intervention as
specified, attrition bias, and inappro-
priate or unknown sample definition.
Problems with the analyses can be
overcome by reanalyzing the data in a
different way. However, it is not pos-
sible to correct for contamination in-
troduced by the way in which an ex-
periment is implemented.
Over and above these methodolog-
ical weaknesses, no set of theories ap-
pears to be guiding this work, nor is
there consistency in the methods or
measures used. It is not much of an
exaggeration to say that, with few ex-
ceptions, each study appears to be
starting over.
As a result, much remains un-
known. This body of research has not
1315
demonstrated which housing attrib-
utes or factors are critical to a mental-
ly ill person’s capacity to live inde-
pendently, it has not described the
types of residential alternatives that
are most effective for persons with se-
rious mental illness, it has not identi-
fied specific housing attributes that
can be systematically associated with
the best type of residential settings,
and it has not produced any agree-
ment on the most appropriate way to
conceptualize and measure the effec-
tiveness of the housing setting.
Deinstitutionalization is in its
fourth decade, and a focus on home-
lessness is in its third. It is fair to ask
why a systematic body of knowledge
about housing and mental illness has
not yet been compiled. In their land-
mark article on the effects of deinsti-
tutionalization, Braun and colleagues
(35) offered three possibilities. First,
many mental health professionals
have strong views about what is and is
not effective. These views may affect
the design of a study, the way in which
it is conducted, and the documenta-
tion of the research and therefore the
reliability of the results.
Second, virtually every aspect of re-
search in the area of housing for peo-
ple with mental illness is inherently
complex, from observing and measur-
ing an individual’s psychiatric status
to ensuring that the intervention be-
ing tested is what it purports to be.
Finally, carrying out experimental
studies in this area in ethically appro-
priate ways is a formidable challenge.
Goldman and Morrissey (36) offer an
additional explanation: much mental
health policy and practice have not
considered housing as a potentially
key component of a system of mental
health care.
Part of the problem can be attrib-
uted to the field of housing research,
which does not yet have a strong the-
oretical base or accepted measures
for future work. Moreover, the work
that has been done typically applies to
cross-sections of the general popula-
tion rather than to persons with seri-
ous mental illness.
To ensure that the research mis-
takes of the past 25 years are not re-
peated in the next 25 years, several is-
sues must be addressed. First, there
is a critical need for a coherent re-
1316
search agenda built around key hy-
potheses of housing and mental ill-
ness. This agenda should be organ-
ized and implemented in a linear way,
so that initial studies build theory and
develop valid and reliable measures
and later studies apply these theories
and measures to test behavioral and
structural relationships. Although
much of the agenda will require new-
ly designed studies, there are some
opportunities for reanalyzing existing
data sets with more sophisticated sta-
tistical methods that account for such
problems as selectivity and attrition
bias.
Future research would also be aid-
ed substantially by the establishment
of a basic set of measures of housing
as an input and an outcome, akin to
the minimum data set on long-term
care that was developed 20 years ago
(37,38). The resulting agenda and re-
search should be disseminated to as
broad an audience as possible. The
establishment of a National Academy
of Sciences panel to study housing
and mental illness would be one way
of raising awareness of the issues.
Another concern is the significant
methodological challenge of design-
ing studies that address the two key
policy questions: the relative effects
on mental health outcomes of hous-
ing and services and of different
housing-and-service bundles. Al-
though randomized experiments are
considered the gold standard for
studying effects, the complexity of the
effects in the area of housing and
mental illness strains the limits of this
methodology. Part of the problem is
the need for multiple experimental
and control conditions. For example,
a complete test of the relative roles of
housing and services in mental health
outcomes would require a service in-
tervention and its control, a housing
intervention and its control, and an
additional nonintervention group.
The difference between access to and
attainment of independent housing
adds another layer of complexity to
the design. Also at issue is the ability
to implement the experiment as de-
signed without contamination, bias
from differential attrition, and other
threats to generalizability.
This discussion of the problems in-
herent in the experimental method is
PSYCHIATRIC SERVICES
not meant to discourage its applica-
tion. Rather, it is a call for extra pre-
cautions to ensure that the experi-
mental design is consistent with the
policy questions being asked, that the
interventions are implemented as de-
signed, and that the experimental
sites are not unique, so that results
are generalizable. It also suggests that
opportunities for nonexperimental
studies should be pursued whenever
it is possible to use sophisticated sta-
tistical techniques to simulate experi-
mental conditions. Such an approach
has been frequently used in sociology
and economics, and it has been
strengthened substantially by the
merging of data from administrative
records and sample surveys.
More sophisticated analytic meth-
ods are also required. The pristine
conditions of the laboratory cannot be
applied to social science studies,
which makes correlational analysis in-
herently inadequate. All research de-
signs, whether nonexperimental, qua-
si-experimental, or experimental, typ-
ically require multivariate modeling.
Such models need to account for a
comprehensive set of characteristics,
selection bias, and possible effects of
unmeasured or unobserved attributes.
Turning these recommendations
into reality will, of course, require sig-
nificant funding from government
agencies and private foundations.
The National Institute of Mental
Health (NIMH) and the Center for
Mental Health Services supported
some of the studies reviewed here;
however, funds have been limited,
and the continuity of interest re-
quired to build a coherent body of
knowledge has been lacking. The re-
cent release of the Surgeon General’s
report on mental illness (39) and the
NIMH report (40), which identifies
“contextual influences on mental ill-
ness and its care” as a prime target for
intensified study, raises the hope that
continuous interest and support may
finally be forthcoming. ♦
Acknowledgments
This research was supported by contract
98M00212501D with the Center for
Mental Health Services of the Substance
Abuse and Mental Health Services Ad-
ministration. The author acknowledges
the encouragement and support of Wal-
♦ October 2001 Vol. 52 No. 10
ter Leginski, Ph.D., Fran Randolph,
Dr.P.H., and Pam Fisher, Ph.D., and the
research assistance of Amy Robie and
Sally Katz.
References
1. Kessler R, Berglund P, Leif P, et al: The 12-
month prevalence and correlates of serious
mental illness, in Mental Health, United
States, 1996. Edited by Manderscheid RW,
Sonnenschein MS. DHHS pub (SMA)96-
3098. Rockville, Md, Center for Mental
Health Services, 1996
2. Skinner E, Steinwachs D, Kasper J: Family
perspectives on the service needs of people
with serious and persistent mental illness:
characteristics of families and consumers.
Innovations and Research 1:23–30, 1992
3. Newman SJ: Housing and Mental Illness: A
Critical Review of the Literature. Washing-
ton, DC, Urban Institute, 2001
4. Cournos F: The impact of environmental
factors on outcome in residential programs.
Hospital and Community Psychiatry 38:
848–852, 1987
5. Bond G, Witheridge T, Wasmer D, et al: A
comparison of two crisis housing alterna-
tives to psychiatric hospitalization. Hospital
and Community Psychiatry 40:177–183,
1989
6. Lehman AF, Slaughter J, Myers C: Quality
of life in alternative residential settings.
Psychiatric Quarterly 62:35–49, 1991
7. Uehara ES: Race, gender, and housing in-
equality: an exploration of the correlates of
low-quality housing among clients diag-
nosed with severe and persistent mental ill-
ness. Journal of Health and Social Behavior
35:309–321, 1994
8 Mowbray CT, Bybee D: The importance of
context in understanding homelessness and
mental illness: lessons learned from a re-
search demonstration project. Research on
Social Work Practice 8:172–199, 1998
9 Hough RL, Wood PA, Hurlburt MS, et al:
Using Independent Housing and Support-
ive Services With the Homeless Mentally
Ill: An Eighteen-Month Outcome Study.
Draft, 1995
10. Hurlburt MS, Hough RL, Wood PA: Ef-
fects of substance abuse on housing stabili-
ty of homeless mentally ill persons in sup-
ported housing. Psychiatric Services 47:
731–736, 1996
11. Lehman AF, Dixon LB, Kernan E, et al: A
randomized trial of assertive community
treatment for homeless persons with severe
mental illness. Archives of General Psychi-
atry 54:1038–1043, 1997
12. Goering P, Wasylenki D, Lindsay S, et al:
Process and outcome in a hostel outreach
program for homeless clients with severe
mental illness. American Journal of Or-
thopsychiatry 67:607–617, 1997
13. Baker F, Douglas C: Housing environments
and community adjustment of severely men-
tally ill persons. Community Mental Health
Journal 26:497–505, 1990
PSYCHIATRIC SERVICES ♦ October 2001 Vol. 52 No. 10
14. Coulton CJ, Holland TP, Fitch V: Person-
environment congruence and psychiatric
patient outcome in community care homes.
Administration in Mental Health 12:71–88,
1984
15. Linn MW, Klett CJ, Caffey EM: Foster
home characteristics and psychiatric pa-
tient outcome. Archives of General Psychi-
atry 37:129–132, 1980
16. Nagy MP, Fisher GA, Tessler RC: Effects
of facility characteristics on the social ad-
justment of mentally ill residents of board-
and-care homes. Hospital and Community
Psychiatry 39:1281–1286, 1988
17. Davies MA, Bromet EJ, Schulz SC, et al:
Community adjustment of chronic schizo-
phrenic patients in urban and rural settings.
Hospital and Community Psychiatry
40:824–830, 1989
18. Depp FC, Dawkins JE, Selzer N, et al:
Subsidized housing for the mentally ill. So-
cial Work Research and Abstracts 22:3–7,
1986
19. Earls M, Nelson G: The relationship be-
tween long-term psychiatric clients’ psy-
chological well-being and their perceptions
of housing and social support. American
Journal of Community Psychology 16:279–
293, 1988
20. Nelson G, Hall GB, Walsh-Bowers R: The
relationship between housing characteris-
tics, emotional well-being, and the person-
al empowerment of psychiatric consu-
mer/survivors. Community Mental Health
Journal 34:57–69, 1998
21. Lipton FR, Nutt S, Sabatini A: Housing the
homeless mentally ill: a longitudinal study
of a treatment approach. Hospital and
Community Psychiatry 39:40–45, 1988
22. Trute B, Segal SP: Census tract predictors
and the social integration of sheltered care
residents. Social Psychiatry 11:153–161,
1976
23. Segal SP, Silverman C, Baumohl J: Seeking
person-environment fit in community care
placement. Journal of Social Issues 45:49–
64, 1989
24. Dixon L, Krauss N, Myers P, et al: Clinical
and treatment correlates of access to Sec-
tion 8 certificates for homeless mentally ill
persons. Hospital and Community Psychia-
try 45:1196–1200, 1994
25. Goldfinger SM, Schutt RK, Tolomiczendo
GS, et al: Housing persons who are home-
less and mentally ill: independent living or
evolving consumer households? In Mental-
ly Ill and Homeless: Special Programs for
Special Needs. Edited by Breakey W,
Thompson J. Australia, Harwood Academic
Publishers, 1997
26. Goldfinger SM, Schutt RK, Tolomiczendo
GS, et al: Housing placement and subse-
quent days homeless among formerly
homeless adults with mental illness. Psychi-
atric Services 50:674–679, 1999
27. Dickey B, Gonzalez O, Latimer E, et al:
Use of mental health services by formerly
homeless adults residing in group and inde-
pendent housing. Psychiatric Services 47:
152–158, 1996
28. Dickey B, Latimer E, Powers K, et al:
Housing costs for adults who are mentally
ill and formerly homeless. Journal of Men-
tal Health Administration 24:291–305,
1997
29. Wood PA, Hurlburt MS, Hough RL, et al:
Longitudinal assessment of family support
among homeless mentally ill participants in
a supported housing program. Journal of
Community Psychology 6:327–344, 1998
30. Newman SJ, Reschovsky JD, Kaneda K, et
al: The effects of independent living on
persons with chronic mental illness: an as-
sessment of the Section 8 certificate pro-
gram. Milbank Quarterly 72:171–198, 1994
31. Newman SJ, Harkness J, Galster G, et al:
Bricks and behavior: the development and
operating costs of housing for persons with
mental illness. Real Estate Economics
29:277–304, 2001
32. Hurlburt MS, Wood PA, Hough RL: Pro-
viding independent housing for the mental-
ly ill: a novel approach to evaluating long-
term longitudinal housing patterns. Journal
of Community Psychology 24:291–310,
1996
33. Newman SJ, Reschovsky JD: Neighbor-
hood locations of Section 8 housing certifi-
cate users with and without mental illness.
Psychiatric Services 47:392–397, 1996
34. Making a Difference: Interim Status Re-
port of the McKinney Demonstration Pro-
gram for Homeless Adults With Serious
Mental Illness. Rockville, Md, Center for
Mental Health Services, 1994
35. Braun P, Kochansky G, Shapiro R, et al:
Overview: deinstitutionalization of psychi-
atric patients, a critical review of outcome
studies. American Journal of Psychiatry
138:736–749, 1981
36. Goldman H, Morrissey J: The alchemy of
mental health policy: homelessness and the
fourth cycle of reform. American Journal of
Public Health 75:727–731, 1985
37. National Committee on Vital and Health
Statistics: 1980 Long-Term Health Care
Minimum Dataset. Hyattsville, Md, Na-
tional Center for Health Statistics, Public
Health Service, 1980
38. National Committee on Vital and Health
Statistics: 1987 Annual Report. Hyattsville,
Md, National Center for Health Statistics,
Public Health Service, 1988
39. Mental Health: A Report of the Surgeon
General. Rockville, Md, US Department of
Health and Human Services, 1999
40. Translating Behavioral Science Into Action.
Bethesda, Md, National Institute of Mental
Health, 2000
1317