Content uploaded by M. Susan Ridgely
Author content
All content in this area was uploaded by M. Susan Ridgely on Sep 25, 2015
Content may be subject to copyright.
Alcohol and Drug Abuse Treatment of Homeless Persons: Results
from the NIAAA Community Demonstration Program
Robert G. Orwin
Howard H. Goldman
L. Joseph Sonnefeld
M. Susan Ridgely
Journal of Health Care for the Poor and Underserved, Volume
5, Number 4, 1994, pp. 326-352 (Article)
Published by The Johns Hopkins University Press
For additional information about this article
Access Provided by Peking University at 07/27/11 10:05AM GMT
More
http://muse.jhu.edu/journals/hpu/summary/v005/5.4.orwin.html
326________________________________________________________________
Original paper
ALCOHOL AND DRUG ABUSE TREATMENT OF
HOMELESS PERSONS: RESULTS FROM THE NIAAA
COMMUNITY DEMONSTRATION PROGRAM
ROBERT G. ORWIN, PhD
HOWARD H. GOLDMAN, MD, PhD
L. JOSEPH SONNEFELD, ma
M. SUSAN RIDGELY, Msw
NANCY GRAY SMITH, ma
ROBERTA GARRISON-MOGREN, ma
ELLEN O'NEILL, ma
ANNE SHERMAN, BA
Abstract: In a national evaluation, we assessed the implementation and
outcomes of a multisite demonstration program for homeless persons with
alcohol and other drug problems. We developed comprehensive case studies
from data on client characteristics, utilization of services, implementation of
interventions, and community systems of care at nine project sites. Client-
level outcome data were analyzed to estimate the effectiveness of the interven-
tions in a subset of projects with experimental or quasi-experimental evalua-
tion designs. After controlling for baseline predictors, treatment clients in the
majority of sites were significantly more likely than comparison clients to
report improvement on one or more outcome dimensions. On alcohol use, for
example, under conservative assumptions the average treatment client was
drinking less at follow-up than were 57 percent of comparison clients. Analy-
ses of predictor-by-treatment interactions suggested that clients with fewer
Dr. Orwin is research director, Mr. Sonnefeld is senior research associate, Ms. Smith and Ms.
Garrison-Mogren are research associates, and Ms. Sherman is research assistant at R.O. W. Sciences,
Inc., 1700 Research Blvd., Suite 400, Rockville, MD 20850. Dr. Goldman is professor of psychiatry
and director, Ms. Ridcely is associate director, and Ms. O'Neill is research associate at the Mental
Health Policy Studies Program at the University of Maryland School of Medicine.
Submitted January 23,1994; revised May 14,1994; accepted May 15,1994
Journal of Health Care for the Poor and Underserved · Vol. 5, No. 4 · 1994
__________________Orwin, Goldman, Sonnefeld, et al.__________________327
problems benefited most from the interventions. The implementation analysis
yielded a number of lessons for policymakers and program planners.
Key words: Homeless persons; alcoholism; substance abuse; program
evaluation
Alcohol and other drug problems are widespread among homeless
persons. Although precise estimates of prevalence are difficult to obtain,
a recent review indicates that alcohol problems affect between 40 and 50 percent
of all homeless persons, and that other drug problems affect between 28 and 37
percent.1 Taking into account distinctions among symptoms and diagnoses in
previous prevalence studies, an estimated 40 to 45 percent of homeless persons
suffer from Axis I mental disorders, while undifferentiated mental health
problems afflict between 45 and 50 percent. Thus, whatever the total number of
homeless persons in the United States, alcohol and other drug addiction, as well
as mental illness, are among this population's most pressing problems.
In July 1987, Section 613 of the Stewart B. McKinney Homeless Assistance
Act (Public Law 100-77) authorized funds for a new demonstration program for
homeless persons with alcohol and other drug-related problems. In May 1988,
through this authority, the National Institute on Alcohol Abuse and Alcoholism
(NIAAA), in consultation with the National Institute on Drug Abuse, initiated
the Community Demonstration Projects for Alcohol and Drug Abuse Treatment
of Homeless Individuals (hereinafter referred to as the Community Demonstra-
tion Program). The program awarded two-year grants on a competitive basis
to nine community-based projects to implement and evaluate a variety of
interventions. The projects were located in eight cities: Anchorage, AK; Boston,
MA; Los Angeles, CA; Louisville, KY; Minneapolis, MN; New York, NY;
Oakland, CA; and two projects in Philadelphia, PA. Total funding appropriated
in 1987 for the program was $9.2 million for a two-year period. An additional
$4.5 million was appropriated in FY 1989 for seven of the nine projects that were
funded for a third year. These projects continued through September 1991.
There are at least two distinctive features of the structure of this demonstra-
tion program. First, in developing the Request for Applications (RFA), the
NIAAA staff decided not to prescribe that a single treatment protocol be
implemented and evaluated. The decision was based on the scarcity of empiri-
cal knowledge about what types of treatments are effective with homeless
persons who have alcohol and other drug problems. Applicants responding to
the RFA were thus free to choose the nature and scope of the treatment models
to be investigated, and a rich mix of approaches was funded. In addition to
providing treatment, the projects developed innovative approaches to serving
the target population. These ranged from outreach programs in the streets and
homeless shelters to intensive case management and supportive housing ar-
rangements. Thus, in contrast to more traditional demonstration projects, the
Community Demonstration Program was designed to explore the usefulness of
many diverse treatment models rather than to demonstrate a single treatment
328 Alcohol and Drug Abuse Treatment of Homeless Persons
model in different settings. A summary of each of the projects in the Community
Demonstration Program is provided in Appendix A.
A second distinctive feature of the Community Demonstration Program
was the emphasis on evaluation. The RFA required that 25 percent of the grant
award be allocated for evaluation. Applicants were required to conduct a
process evaluation of their interventions and to participate in a national evalu-
ation of the full demonstration program. In addition, applicants were encour-
aged, although not required, to conduct client-level outcome evaluations, and
seven of the nine projects did so. The Community Demonstration Program was
the first national effort to mount a multisite outcome evaluation with this target
population.
This paper summarizes the findings from the national evaluation, including
the number and characteristics of clients served by the program; the services
provided; the extent to which clients completed programs; and the effectiveness
of the interventions in reducing alcohol and other drug use, increasing employ-
ment and economic security, improving physical and mental health status, and
increasing residential stability. It also discusses programmatic lessons from the
implementation of the interventions, and implications for future multisite
research demonstrations designed to serve this target population.
Methods
The national evaluation consisted of a process evaluation and an outcome
evaluation. The process evaluation included all nine projects. Data sources
were project proposals, which described each program as planned; quarterly
reports, which collected aggregate data concerning numbers of clients, client
characteristics, program activities, and systems-level activities in a uniform
format; annual site-visit reports, which contained information on the status of
program implementation and the fit between the planning and implementation
of each program; monthly telephone interviews of project directors and project
evaluators by national evaluation staff; and a special questionnaire on the
system of care, which requested organizational charts and budgetary informa-
tion on the provision of services to the target population by the project and by
local, state, and federal agencies. To assure clarity and consistency of system-
of-care data, these sources were augmented as needed through additional
telephone calls to key informants, including state and local agency officials from
program and budget offices (alcohol and other drug, homeless/housing assis-
tance, and mental health agencies), advocates, demonstration project adminis-
trators, and nonprofit providers.
The process evaluation data were used to develop case studies on each of the
nine projects. The case studies described the operational elements of each
project; explained how the project fit into the community system of services for
homeless persons with alcohol and other drug problems; presented representa-
tive descriptive data on service utilization; documented the process of imple-
mentation (facilitators and barriers—specific changes in operations that poten-
Orwin, Goldman, Sonnefeld, et al. 329
tially affect client outcomes); and summarized what was learned about program
implementation from each project.
The outcome evaluation consisted of an independent analysis of client-level
outcome data from five of the nine projects (in Boston, Los Angeles, Louisville,
Minneapolis, and New York). Although the evaluation designs for seven of the
nine projects included an outcome component designed to assess experimen-
tally or quasi-experimentally the impact of program interventions, two of these
projects were not included in the national evaluation outcome study because
follow-up data on comparison-group clients were unavailable. The outcome
evaluation drew from two primary data sources: the Addiction Severity Index
(ASI)2, a structured 45-minute clinical research interview designed to assess the
severity of problems in seven areas (medical condition, employment/economic
support, alcohol use, other drug use, illegal activity, family relations, and
psychiatric condition) commonly affected in persons who abuse alcohol and
other drugs; and the Treatment/Housing Reporting Form, a standardized
client-level questionnaire devised by the national evaluation team in coopera-
tion with NIAAA and the project evaluators to address questions related to
program completion and housing status.
Improvement was assessed in two ways: likelihood of improvement (whether
treatment-group clients were more likely to improve than comparison-group
clients) and magnitude of improvement (the extent to which treatment-group
clients improved relative to controls, on average). To bracket the uncertainty
resulting from attrition at follow-up, improvement rates were computed and
compared using two different denominators: the total number of clients as-
sessed at intake ("total sample"); and the subset of clients who were assessed at
both intake and follow-up ("follow-up sample"). Each denominator carries a
different assumption about the outcome status of clients who were not fol-
io wed-up. Use of the total sample denominator assumes, in effect, that all clients
who were not followed up did not improve. Although addiction researchers
frequently do assume that individuals lost to follow-up have relapsed3, this
assumption is nonetheless conservative; it underestimates the true improve-
ment rate because it is likely that at least some of those clients who were not
assessed at follow-up did in fact improve. (For example, some clients who
maintained their sobriety may have left the area and so could not be tracked.)
On the other hand, use of the follow-up sample denominator assumes that
attrition was random. That is, it assumes that clients who were not included in
follow-up are just as likely to have improved as clients who were, and that any
differences in improvement rates between the treatment and comparison groups
would also be present in the clients from each group who did not participate in
follow-up. This assumption is more liberal; it almost certainly overestimates the
true improvement rate because it does not account for the fact that clients who
were not assessed at follow-up are likely to be doing worse on the average than
those who were. The actual outcome status of these "lost" clients is by definition
unknown. But because the two approaches are biased in opposite directions,
330 Alcohol and Drug Abuse Treatment of Homeless Persons
taken together they serve to bracket the true value with a lower and upper
bound.
The core procedures of the site-level outcome analyses consisted of three
main phases: design diagnostics (these included baseline nonequivalency analy-
ses, attrition analyses, and power analyses); program completion analysis; and
estimation of intervention effectiveness. In addition to computing unadjusted
differences between treatment and comparison groups, we used hierarchical
multivariate procedures to estimate the effects of the intervention after the
observed variance associated with individual differences between clients on
baseline predictors had been removed. This potentially increased the power of
the test of the intervention and also permitted us to compute an adjusted mean
difference and an estimated standardized effect size on each outcome criterion
for each site.
Seven categories of individual characteristics—or predictor domains—
were identified in the literature as both associated with treatment outcome and
measurable using the ASI. The predictor domains and ASI measures are listed
below:
1. Demographics (age, gender, race, and education);
2. Employment status and level of economic security (existence of a
profession, trade, or skill; employment pattern over past three
years; income from employment in past 30 days; and general
economic security as measured by the combined income in past 30
days from employment, welfare, pension, benefits, social security,
or mate, family, or friends);
3. Residential stability (usual living arrangements over the past three
years);
4. Family and social connections or relationships (marital status;
number of persons with whom free time is generally spent; number
of close friends; and lifetime experience with conflicts with family,
friends, and others);
5. Mental health problems (previous treatment for psychological
problems; prescribed psychotropic drug use in the past or present;
difficulty with violent behavior, depression, anxiety, hallucina-
tions, understanding, concentrating, or remembering; and experi-
ence with suicide attempts or ideation);
6. Alcohol and other drug abuse or dependence (major substance of
abuse; lifetime history of drinking to intoxication; number of times
treated for alcohol abuse; number of times treated for other drug
abuse; number of times overdosed on drugs; and lifetime history of
___________________Orwin, Goldman, Sonnefeld, et al.________________331
drug use including cannabis, cocaine, heroin, methadone, other
opiates, barbiturates, other sedatives, amphetamines, hallucino-
gens, and inhalants); and
7. Multiple problems (the number of problem areas reported, based
on ASI composite scores across the following dimensions: medical
health, alcohol use, other drug use, employment status, legal status,
family and social issues, and psychiatric status).
Given the instability of some of the individual ASI items in representing a
particular predictor domain, certain ASI items were grouped to create more
stable, interpretable indexes. Candidate predictors that did not covary with
program completion or specific outcome criteria were dropped. Keeping the
predictor set small and literature-based served two purposes: It limited the
problem of capitalization on chance, and it permitted the equations to be run on
all sites, including those with small sample sizes.
We used logistic regression to estimate likelihood of improvement, analysis
of covariance (ANCOVA) to estimate magnitude of improvement. In each, a
hierarchical "set" approach4 was employed, with the predictor sets (e.g., age,
race, sex, and education in the demographic set) entered prior to the treatment
variable. The hierarchical approach produces a conservative estimate of the
treatment effect, because variance common to both the predictor set and the
treatment variable is "credited" to the predictors, that is, the predictors are given
causal priority over the treatment in the hierarchy. Therefore, to the extent that
individual differences on predictors are responsible for outcome variance, the
hierarchical approach provides a measure of protection against the misinterpre-
tation of observed group differences as treatment effects.
All models were diagnosed for multicollinearity among predictors, and (in
the case of the ANCOVAs) heterogeneity of regression slopes. Where indicated
by the diagnostics, the model was reduced and the treatment-effect parameter
was reestimated.
The standard core procedures that were conducted on data from each site
were supplemented by site-specific procedures, including analyses using addi-
tional client-level data, and analyses derived from the process evaluation and
projects' final reports. These were used to further analyze and interpret the
results from the client-level outcome data.
For the cross-site synthesis, the site-level effect size estimates for each
outcome dimension were: weighted to reflect differences in sample size across
sites5; averaged across sites; converted to Z-scores; and translated to Cohen's LÕ3
statistic.6 LÕ3 is defined as the percentage of the comparison distribution
exceeded by the median of the treatment distribution.7 It is readily obtainable
from a normal probability table, and is more intuitively meaningful than the
expression of average effect sizes in standard deviation units. Consideration
was given to pooling the client-level data across sites and specifying site as a
332 Alcohol and Drug Abuse Treatment of Homeless Persons
categorical independent variable, but this approach was rejected due to objec-
tions from site-level investigators.
For further details on the data sources, outcome measures, and analytic
approach, see the full report on the outcome evaluation.8
Results
Numbers and characteristics of clients served. Approximately 6,762
clients were served in the demonstration, of whom 1,563 were served by those
projects providing primarily extended services. Extended services (as distin-
guished from initial services, such as outreach, sobering, and detoxification)
include residential recovery, sober housing, nonresidential recovery, case man-
agement, and in some cases, shelter. Half (51 percent) of the clients served by
those projects were black, 38 percent were white, and seven percent were
Hispanic. Almost one-third (29 percent) were women. Almost half of the clients
(46 percent) were between the ages of 25 and 34, and more than four-fifths (82
percent) were between age 25 and 54.
Characteristics of services provided. Table 1 illustrates the types of
services that received NIAAA funding for each of the demonstration projects
providing services data in 1990, the year that best represents the fully imple-
mented program. Core services included outreach, shelter, sobering, detoxifi-
cation, residential recovery, sober housing, nonresidential recovery, case man-
agement, and transportation. All nine projects provided at least one of these
core services, and most provided several. Projects differed considerably in their
treatment models, populations, and objectives. Therefore, a project providing
many types of services was not necessarily better implemented or more success-
ful than one providing relatively few. Table 1 provides a snapshot of which
projects provided which services; for detailed demographic and quantitative
services data by project, see the full report on the process evaluation.9 For an
insider's perspective, see the excellent descriptions by the project investiga-
tors.10·11
Program completion. Among the projects included in the outcome study,
56 percent of treatment-group clients either completed or were retained in their
program; 19 percent dropped out; 13 percent were classified as administrative
discharges; and 13 percent left for "other" reasons. Table 2 breaks out the
program completion data by project. Although there was wide variation in
completion/retention rates across sites (43 percent to 83 percent) it would be
inappropriate to infer that some project administrators were more successful
than others in retaining clients. A "successful" retention rate for one type of
intervention or target population might differ considerably from that of another
type; moreover, retention policies and definitions of terms differed across
projects. For example, clients in a case management program might have no
requirement for sobriety and no provision for dropping out except through
TABLE 1
SERVICES PROVIDED IN NIAAA-FUNDED COMPONENTS FOR EACH PROJECT DURING 1990 BASED ON
UTILIZATION DATA FROM QUARTERLY REPORTS
SITE
Anchorage
Boston
Los Angeles
Louisville
Minneapolis
New York
Oakland
Philadelphia
Diagnostic and
Rehabilitation
Center
Horizon House
OUTREACH
SHELTER
X
X
SOBERING
X
RESIDENTIAL
RECOVERY
X
X
SOBER NONRESIDENT- CASE MAN-
HOUSING IAL RECOVERY AGEMENT
X
X
X
TRANS-
PORTATION
X
X
X
X
X
O
2.
a.
3
Q
3
cn
o
3
3
"&
a.
to
«^.
a
w
iE
TABLE 2
COMPLETION/RETENTION RATES AND REASONS FOR CLIENTS LEAVING PROGRAMS
BEFORE COMPLETING TREATMENT (TREATMENT GROUP ONLY)
SITE
Boston *
Los Angeles
Louisville
Minneapolis
New York
COMPLETED OR
RETAINED
48
64
83
43
84
82
66
56
REASONS FOR LEAVING BEFORE COMPLETING TREATMENT
Dropped Out Administrative Other
Discharge
15
30
9
18
26
39
7
23
36
0
0
2
63
0
0
2
2 3
5 7
8 7
37 48
O
g
tu
3
D.
era
C
H
0)
s
"Completed treatment" as used by the Boston project referred to completion of the stabilization phase of the intervention, not
the case management phase. According to the Boston model, clients completed stabilization as a prerequisite for assignment
to case management, after which they received case management services; they did not complete (or faü to complete) case
management.
η
Vl
Vl
1I
Vl
O
3
U3
___________________Orwin, Goldman, Sonnefeld, et al.________________335
death or other extraordinary circumstances. By contrast, clients in a residential
recovery program could be asked to leave after one relapse.
To examine the characteristics of clients that were associated with complet-
ing programs, a discriminant analysis of program completion was performed
on data from each project, using candidate predictors from the literature. On
balance, the results showed little cross-site consistency in the characteristics of
program completers (data not shown). Differences in target populations, in the
interventions, and in retention policies, and the possibility that there were too
few sites for stable commonalities to emerge, may help explain this observation.
Intervention effectiveness. The unadjusted project-level results for the
two alcohol outcome variables from the ASI—alcohol use in the past 30 days,
and the alcohol composite score—are summarized in Table 3. The numbers in
the table can be interpreted as follows: When the total sample was examined in
Boston (for example), 57 percent of the treatment group and 45 percent of the
comparison group demonstrated improvement, and the difference in rates was
significant (p<0.05). As expected, improvement rates in both groups increased
when the analysis was limited to the follow-up sample (83 and 80 percent
respectively), but the difference was no longer significant. However, a compari-
son of mean days of drinking suggested that the average Boston treatment-
group client was drinking significantly less frequently (4.1 days per month) than
was the average comparison-group client (6.4 days per month). The resulting
effect size (d=-0.24) suggests a "small" effect.
Results from the multivariate analyses—estimates of the likelihood and
magnitude of improvement—are shown in Table 4. For the total sample, after
controlling for the effects of the predictors, the treatment-group clients were
more likely than comparison-group clients to report improvement on both
variables in four of five projects. Only one of these differences was statistically
significant, however—in Boston, where the odds of improving were approxi-
mately 1.75 times greater in the treatment group than in the comparison group
(95 percent confidence interval (CI)=I .19,2.59). Though not shown in the Table,
treatment clients in Boston were also more likely to achieve abstinence (odds
ratio (OR)=I .79,95 percent CI=I .21,2.65), as were treatment clients in New York
(OR=2.01, 95 percent CI=1.12,3.63). For the follow-up sample, the treatment
clients were again more likely to report improvement in four projects. However,
only one difference was statistically significant—in Los Angeles, where the
odds of improving were 4.05 times greater in the treatment group (95 percent
0=1.43,11.46). Treatment clients in Los Angeles were also more likely to
achieve abstinence (OR=2.45,95 percent 0=1.06,5.66) (data not shown).
With respect to magnitude of improvement, the differences between ad-
justed means favored the treatment group in four projects on the 30-day variable
and in three projects on the composite score (Table 4). The differences were
statistically significant in Los Angeles where the treatment accounted for
approximately 8 and 11 percent of the outcome variance over and above that
accounted for by the predictors. This corresponds to a "medium" effect size.
TABLE 3
IMPROVEMENT RATES AND UNADJUSTED MEAN OUTCOMES FOR TREATMENT (T) AND COMPARISON (C)
GROUPS: ALCOHOL USE AND ALCOHOL COMPOSITE SCORES
SITE
Boston
Los Angeles %
η
Louisville
%
η
Minneapolis %
η
New York %
ALCOHOL USE IN PAST 30 DAYS
ALCOHOL COMPOSITE SCORES'
"CLffiNTS WHO IMPROVED CLIENTS WHO IMPROVED
Total Sample Follow-up FREQUENCY OF USE" Total Sample Follow-up
Sample _______________________ ___________ SamPle
57
244
45
176
39
129
19
78
15
227
45*
219
34
86
25
36
40*
107
3*
304
ES(d)c
83
167
89
89
52
97
44
34
26
129
80
122
69*
42
47
19
59
73
14
73
4.10 6.41
167 122
2.76
89
2.28
129
9.07
42
9.35 11.26
97 19
13.26 13.27
34 73
1.96
73
,24*
,66*
-.16
.07
51
240
47
168
49
129
27
66
21
225
42
217
41
80
33
36
35
93
11*
290
74
164
93
85
65
97
62
29
37
129
77
120
85
39
63
19
52
64
47
70
SCORE11
.24
158
.15
87
.33
97
..31
31
.06
128
.24
117
.33
42
.40
19
.39
67
.12
71
ES(d)c
-.73
-.25
,32
-.46*
CO
Os
η
O
3"
ο_
su
g.
σ
a
era
C
Si
S
I
* p<.05
« Composite scores for each of seven ASI problem areas are based on the sum of several of the individual questions within that problem area.
b Mean outcomes, follow-up sample
' Effect size based on Cohen's d, calculated from the standardized difference between means. Benchmarks: SmaU: d=.20; Medium: d=.50; Large: d=.80
m
"1
Vl
O
3
Vl
TABLE 4
LIKELIHOOD AND MAGNITUDE OF IMPROVEMENT FOR TREATMENT (T) AND COMPARISON (C) GROUPS:
ALCOHOL USE AND ALCOHOL COMPOSITE SCORES
ALCOHOL USE IN PAST 30 DAYS
ALCOHOL COMPOSITE SCORES"
Odds Ratio of Likelihood of
Improvement1'
Magnitude of
Improvement
Total Follow-up
Sample Sample
Adjusted
Means
SITE
Boston
Los Angeles
Louisville
Minneapolis
New York
1.75*
1.70
1.87
0.39*
3.71
1.62
4.05*
1.24
0.52
1.47
T
4.36
2.76a
9.36
13.20
2.48
C
6.09
9.07d
11.20
14.46
1.63
ES
(pv)c
0.01
0.08*
0.00
0.00
0.00
Odds Ratio of Likelihood of
Improvement0
Total Follow-up
Sample Sample
Magnitude of
Improvement
Adjusted
Means
1.41
1.29
1.73
0.92
1.21
1.20
2.86
1.21
1.56
0.74
T
0.25
0.15
0.34
0.32
0.10
C
0.22
0.33
0.36
0.43
0.09
ES
(pv)c
0.00
0.11*
0.00
0.04
0.00
* p<.05
" Composite scores for each ASI problem area are based on the sum of several individual questions within the area.
b The odds ratio expresses the difference between groups in terms of relative odds of improvement, after controlling for predictors. An odds ratio greater
than 1.0 indicates higher likelihood of improvement in the treatment group.
c The effect size expresses the percentage of variance (pv), calculated from the increment in outcome variance accounted for by the treatment, over and above
that accounted for by the predictors. Benchmarks: Small: pv=.01; Medium: pv=.09; Large: pv=.25
d Because the ANCOVA model for 30-day use did not include predictors from the candidate list, the means for this variable are unadjusted.
O
"t
S
3"
s
3
&
3
3
3
OJ
TABLE 5
STATISTICALLY SIGNIFICANT ESTIMATES OF INTERVENTION EFFECTIVENESS (p<.05): LIKELIHOOD AND
MAGNITUDE OF IMPROVEMENT
OJ
OJ
OO
LIKELIHOOD OF IMPROVEMENT
Other drug use
Total Sample
OUTCOME DOMAIN Site
Alcohol use
Boston
New York
Boston
New York
Boston
Odds ratiob
1.79
2.01
Employment/
economic security Los Angeles
New York
Medical status
Psychiatric status
Housing status
Los Angeles
Boston
2.05
3.07
1.58
2.16
5.51
2.86
2.56
Follow-up Sample
Site Odds ratio"
Los Angeles 4.05
Los Angeles
Los Angeles
Los Angeles
Boston
3.88
2.83
4.00
2.22
MAGNITUDE OF IMPROVEMENT·
Site Effect size (pv)°
Los Angeles 0.11
Los Angeles 0.03
Los Angeles
Boston
0.11
0.02
η
O
3*
SU
3
α
>
Vi
η
H
*ι
<Ό
SU
1S
Italics indicate that more than one outcome measure was significant within a given outcome domain. In these cases, the largest effect is displayed.
• Follow-up sample only
b The odds ratio expresses the difference between groups in terms of relative odds of improvement, after controlling for predictors. An odds ratio greater
than 1.0 indicates higher likelihood of improvement in the treatment group.
c The effect size expresses the percentage of variance (pv), calculated from the increment in outcome variance accounted for by the treatment, over and
above that accounted for by the predictors. Benchmarks: Small: pv=.01; Medium: pv=.09; Large: pv=.25
CL
Vl
Vl
'-a
re
i-l
CiI
O
3
CO
___________________Orwin, Goldman, Sonnefeld, et al.________________339
Program-wide (across all five projects) after controlling for the effect of baseline
predictors, the average client in a treatment group reported drinking fewer days
in the 30 days before follow-up than did approximately 57 percent of the
comparison-group clients. For a more detailed presentation of project-level
outcome evaluation results, see the full report on the outcome evaluation.8
Table 5 displays the estimated effects of each project at a glance. All of the
statistically significant estimates of intervention effectiveness are represented,
reported by outcome domain. After controlling for the effects of baseline
predictors, there were statistically significant improvements in three of the five
projects on one or more outcomes under at least one of the follow-up assump-
tions. These projects were in Boston, Los Angeles, and New York. In addition,
one or more interventions were estimated to be effective on every dimension
except medical status.
Client characteristics associated with intervention effectiveness. Analy-
ses of predictor-by-treatment interactions revealed few distinct or conclusive
patterns of client characteristics associated with intervention effectiveness that
were consistent across sites. This mirrors findings on characteristics of program
completers, probably for many of the same reasons. Among the few patterns
that did emerge, the following were observed in three or more sites:
1. Lifetime drug use was the most common predictor of treatment
effects, typically showing a negative association with multiple
outcomes. That is, drug users in the treatment group reported less
improvement than did persons who did not use drugs, relative to
similar clients in the comparison group.
2. Treatment-group clients reporting multiple problems at baseline
improved less on housing outcomes than did those with fewer
multiple problems, relative to similar comparison clients.
3. Treatment-group clients reporting psychiatric problems at baseline
were relatively less likely to report improvement in employment/
economic security than were clients not reporting psychiatric prob-
lems.
Discussion
The NIAAA Community Demonstration Program served over 6,700 home-
less persons with addiction problems through services ranging from short-term
outreach, shelter, and sobering, to longer-term residential recovery, case man-
agement, and sober housing. Among projects with interpretable outcome data,
treatment-group clients were significantly more likely than comparison-group
clients to report improvement on one or more outcome dimensions at the
majority of sites. The evaluation of the demonstration confirms the feasibility
340 Alcohol and Drug Abuse Treatment of Homeless Persons
of implementing and evaluating addiction treatment and recovery programs
for homeless persons, and shows that at least some of the interventions were
effective in improving clients' functioning.
As noted earlier, after controlling for baseline predictors, the average
treatment-group client reported less drinking at follow-up than did approxi-
mately 57 percent of comparison-group clients. However, these numbers reflect
the results from the follow-up samples only (that is, only those clients who were
assessed at both intake and follow-up). The overall pattern of evidence—
particularly the differences between the total-sample and follow-up-sample
improvement and abstinence rates—suggests that the true relative reduction in
drinking was probably larger. This conclusion is based on the fact that analyses
of the follow-up sample typically generated more conservative estimates of
differences in improvement between treatment and comparison groups than
did analyses of the total sample.
None of the projects demonstrated improvement in physical health. This is
not surprising, because chronic medical conditions may be the most difficult to
affect in a short period, particularly among older clients such as chronic public
inebriates. Although staff at several projects expected to see overall improve-
ments in functionality and quality of life, including physical health, none
specified the improvement of physical health asa direct objective or hypothesis,
as they did for some of the other outcome domains (e.g., sobriety).
The analysis of client characteristics associated with intervention effective-
ness suggests that clients with fewer problems are the most likely to benefit from
the interventions and therefore may be better candidates for treatment than
clients with more problems. Because the findings are based on interaction
effects, they represent the relative improvement of the few-problem versus
many-problem clients over and above their expected relative levels of improve-
ment in the absence of the intervention. This is consistent with previous
findings reported in the literature and supports the mainstream view within the
treatment field.
Of the three case management interventions—Boston, Louisville, and Min-
neapolis—only one (Boston) showed indications of effectiveness relative to the
comparison group, and there the effects were ambiguous because of uncertainty
over the status of clients who were not followed-up. These findings merit
comment, particularly because case management has been advocated by pro-
gram designers as a way of extending alcohol and drug treatments to improve
outcomes in people who do not respond well to traditional treatments.1215 The
present findings do not provide unambiguous evidence that case management
per seis ineffective, or even that it was ineffective in these projects. Our analysis
of all three case management projects and related literature on case manage-
ment with mentally ill persons who are homeless suggests that a number of
factors may have been responsible for the absence of more differences between
the treatment and comparison groups.16 These include: bias from the differen-
tial attrition of case management clients who are "doing less well," who are
easier to track than comparable comparison-group clients; lack of sufficient
___________________Orwin, Goldman, Sonnefeld, et al.________________341
intervention intensity to give case management a "good test;" lack of distinction
between treatment and comparison groups because comparison-group clients
also receive some level of case management; contextual factors related to the
network of services in the broader community into which case managers
attempt to link clients; low statistical power and selection bias in the Louisville
and Minneapolis projects; and measurement issues. Although some of the
problems may be unique to case management, the pattern of biases and
weaknesses that tend to diminish or reverse the apparent effects of a service
innovation is a familiar one in the history of comparative outcome evaluations.17
Finally, it is important to note that the requirements outlined in the RFA for
participation in the national evaluation of the Community Demonstration
Program were nonspecific. There were no initial requirements for standardized
instruments for data collection, intervals for data collection, or the selection of
comparison groups. Participation in the national evaluation outcome study was
strictly voluntary, and sites were free to submit data in whatever format they
chose, although the seven projects that developed outcome studies did agree to
use the ASI and other national evaluation instruments. Therefore, certain
limitations in the data and consequent constraints on their interpretation were
probably inevitable. The main limitations were as follows:
1. In general, comparisons of treatment and comparison groups must
be made with an awareness of general cautions about the
nonequivalence of groups and differential attrition. For most
projects, including those that randomly assigned clients to groups,
initial nonequivalences and subsequent attrition suggest that some
observed differences may not have been entirely due to the inter-
vention. The use of a hierarchical modeling approach to estimate
treatment effects reduced to some degree the bias from
nonequivalence.
2. Low follow-up rates were the single biggest threat to the integrity
of evaluation designs and the main impediment to drawing unam-
biguous conclusions about the effectiveness of interventions. Of 11
statistically significant likelihood-of-improvement effects, only three
were robust across the two different assumptions about clients who
were not followed-up: housing status in Boston, and employment
and psychiatric status in Los Angeles. Attempts were made to
bracket the uncertainty with multiple estimates and to reduce it
analytically where possible, with partial success. Ultimately, though,
several conclusions about intervention effectiveness hinged on
what was assumed about the outcome status of clients who were
not followed-up.
342 Alcohol and Drug Abuse Treatment of Homeless Persons
3. The wide variety of treatment interventions, client characteristics,
and treatment settings, in conjunction with the small number of
projects, limited the ability of the national evaluation team to
identify effective interventions or to draw broad generalizations
from the data. Because of the small number of projects, there are
few replications of treatment/setting combinations and many
"gaps" in the range of possible models, so that any differences
between projects in outcomes are difficult to attribute to any one
cause or circumstance among the many ways in which projects
vary.
The above limitations, though not trivial, need to be put in perspective.
Important goals were realized from this first effort to mount a multisite client
outcome study on homeless persons with alcohol and other drug problems.
First, the demonstration projects confirmed the feasibility of implementing
comprehensive alcohol and other drug abuse services for this population.
Second, the fact that most sites were able to collect and submit the standardized
data proved that this difficult at-risk population can be tracked and assessed on
policy-relevant outcome dimensions. Third, the outcome results indicate that
at least some of the interventions were effective in improving clients' lives.
Moreover, the results reported above are conservative, in that the scope of the
outcome study was restricted to components of the interventions that were
experimentally (or quasi-experimentally) tested against a comparison group.
We did not evaluate components provided to all clients (the lack of a comparison
group made cause-and-effect relationships too difficult to interpret), yet many
of these components may also have been effective in producing positive
outcomes. The data indicate that on average, comparison-group clients did tend
to improve between baseline and follow-up, which provides support for this
hypothesis.
Programmatic lessons. The implementation analysis yielded several im-
portant lessons for policymakers and program planners:
1. Successful implementation requires extensive program planning,
model development, and start-up time, especially when integrat-
ing services provided by diverse agencies with different ideologies,
treatment practices, and organizational cultures. In support of
such forward planning, funding agencies should require specific
strategies for facility development, such as an assessment of com-
munity acceptance, strategies to prevent or resolve opposition,
contingency plans for establishing an alternative site, and evidence
of an awareness of the complex process of acquiring and renovating
a facility.
________________Orwin, Goldman, Sonnefeld, et al.________________343
2. Gaps in the community system of services (e.g., lack of affordable
housing, shortage of jobs) must also be recognized and overcome
with creative strategies for combining and integrating services.
Creative strategies that were successful in the Community Demon-
stration Program included, for example, working to develop Ox-
ford-type housing (that is, resident-governed alcohol- and drug-
free housing) for project clients, collaborating with state housing
authorities to establish possible alternatives, and providing incen-
tives to employers to hire clients. Programs also must be prepared
to confront resistance to change. Community educational strate-
gies aimed at the public, at government, and at service providers
have helped to reduce resistance.
3. Clients' needs for shelter, sustenance, and security should be met
before treatment is addressed. Similarly, the limitations of the
target population must be recognized, with services tailored to
their experiences and capacities. Basic training in personal hy-
giene, child care, literacy, and job readiness may be more appropri-
ate at first than, for example, GED or vocational training.
4. It is also necessary to combine structure and flexibility, particularly
in residential programs. That is, successfully implemented pro-
grams were those that met individual client needs (e.g., through
use of individualized client contracts) within a structure that in-
cluded rules and predictable consequences for specific behaviors.
5. The availability of transportation services to link residences with
treatment centers is essential. These services offer safety to clients
travelling to programs located in unsafe neighborhoods and help to
assist clients who must travel long distances. Alternative options
include providing recovery programming in residential facilities
or providing companion(s) to travel with clients.
6. A variety of lessons were learned about implementing case man-
agement programs. First, successful implementation of case man-
agement requires clear and realistic goals. Second, case managers
must have an appropriate educational and experiential background;
matching the gender and ethnicity of case managers and clients
may be desirable where possible. Third, there is a need to focus on
staff development and to directly address conflicting views on the
philosophy and goals of case management. Fourth, the domains of
alcohol and other drug abuse, mental health, physical health, and
social and economic problems may be independent, so that focus-
ing case management services on one problem may not necessarily
affect change in another. Finally, program administrators must
344 Alcohol and Drug Abuse Treatment of Homeless Persons
actively support case managers to sustain morale and retain good
staff. Lack of administrative support for the realities of doing the
job, pressure on case managers to make unworkable systems work
for the clients, and pressure to save costs in treatment can drain the
morale of case managers. Effective strategies for supporting case
managers in this study included: providing extra time to help
nonprofessional-level case managers with especially complex cli-
ent situations; thoroughly training staff members, including hold-
ing practice sessions to prevent staff from reverting to more tradi-
tional roles and behavior; providing time within weekly staff
meetings for the staff to vent frustrations and feelings in a
nonjudgmental environment; providing work time for case man-
agers to attend Al-Anon meetings; using case management aides to
help with tasks; and offering competitive salaries and career lad-
ders to enhance recruitment and retain staff.
Research implications. Finally, a tremendous amount was learned about
the difficulties faced by field researchers in conducting quantitative outcome
evaluations on this population, as well as about the difficulties faced by national
evaluators in assisting from afar. This knowledge has been put to use in the
design of subsequent multisite evaluations of demonstration projects designed
to serve homeless persons, both inside and outside of NIAAA. The main lessons
and implications for future multisite research demonstrations on the target
population are as follows:
1. Research demonstration programs should run for a minimum of
five years. It has been repeatedly shown, in this demonstration
program as in others, that innovative programs take time to de-
velop and mature.91819 Extensive planning before implementing a
program is invaluable but does not obviate the need for a matura-
tion period to overcome start-up barriers, learn what must be
learned, and fine-tune the intervention.
2. Consideration should be given to mandating randomized designs
or, short of that, mandating an assignment process based on clients'
need for treatment, as determined by their scores on pretreatment
measures. This permits analyses that can correctly adjust for
nonequivalences and produce unbiased estimates of the treatment
effect, even without random assignment.20
3. Consideration should also be given to requiring sufficiently large
samples to provide adequate power to detect treatment effects of
interest and to allow for a realistic attrition rate in all power
calculations.
___________________Orwin, Goldman, Sonnefeld, et al.________________345
4. Proactive technical assistance should be offered to project evalua-
tors during the design phase and throughout implementation. This
technical assistance should include training in the use of all data
collection instruments, for both process and outcome. Ultimately,
the quality of any national evaluation is dependent on the compe-
tence and cooperation of both programmatic and research staff
members.
5. The importance of adequate client tracking and follow-up cannot
be overemphasized. Numerous techniques have been developed
and successfully used by field evaluators in tracking homeless
persons, including those with alcohol and other drug problems
(J.D. Wright, personal communication).16 These techniques should
be made available to all project evaluators as part of the technical
assistance package.
6. When developing RFAs and selecting awardees, consideration
should be given to limiting the variability of treatment interven-
tions across participating sites in the demonstration. As noted
above, the very wide variety of treatment interventions, client
characteristics, and treatment settings substantially complicated
efforts to synthesize the information across projects.
7. Consideration should be given to utilizing and developing better
program theory. A clearly articulated program theory, even the
quick sketch represented by a logic model21, can greatly assist the
planning and assessment of program implementation, guide the
development and selection of appropriate process and outcome
measures, and provide a basis for understanding why a program
that appears to be effective works—thus making it possible to
strengthen and replicate the effect in other projects.
The authors wish to thank the Community Demonstration Program grantees and our colleagues
at R.O.W. Sciences (Margaret Blasinsky), CLEW Associates (Friedner Wittman), and NIAAA
(Robert Huebner, Peggy Murray, and Jack Scott) for their comments, corrections, and editorial
assistance on an earlier draft of this paper. This paper is adapted from research supported by a
contract from NIAAA. The views expressed here do not necessarily represent the views of the
funders.
REFERENCES
1. Lehman AF, Cordray DS. Prevalence of alcohol, drug, and mental disorders among the
homeless: A quantitative synthesis. Contemp Drug Issues (In press).
2. McLellan AT, Luborsky L, Cacciola J, et al. Guide to the addiction severity index: Background,
administration, and field testing results. Philadelphia, PA: Veterans Administration Medical
Center, 1988.
3. Singer JD, Willett JB. Modeling the days of our lives: Using survival analysis when designing
andanalyzinglongitudinalstudiesof duration and timingof events. Psych Bull 1991;110(2):268-
346 Alcohol and Drug Abuse Treatment of Homeless Persons
90.
4. Cohen J, Cohen P. Applied multiple regression /correlation analysis for the behavioral
sciences. Hillsdale, NJ: Lawrence Erlbaum, 1975.
5. Hedges LV. Estimation of effect size from a series of independent experiments. Psych Bull
1982;92(2):490-9.
6. Cohen J. Statistical power analysis for the behavioral sciences. 2nded. New York: Academic
Press, 1988.
7. Lipsey MW. Design sensitivity: Statistical power for experimental research. Newbury Park,
CA: Sage, 1990.
8. National Institute on Alcohol Abuse and Alcoholism. Community demonstration projects for
alcohol and drug abuse treatment of homeless individuals. Vol. I: Outcome evaluation report.
Rockville, MD: R.O.W. Sciences, Inc., 1992.
9. National Institute on Alcohol Abuse and Alcoholism. Community demonstration projects for
alcohol and drug abuse treatment of homeless individuals. Vol. II: Case studies of nine
community demonstration grants. Rockville, MD: R.O.W. Sciences, Inc., 1992.
10. Argeriou M, McCarty D, eds. Treating alcoholism and drug abuse among homeless men and
women: Nine community demonstration grants. Alcohol Treat Q 1990;7(1) Special issue.
11. National Institute on Alcohol Abuse and Alcoholism. Community demonstration projects for
alcohol and drug abuse treatment of homeless individuals. Vol. Ill: Innovative strategies for
treating alcohol and drug abuse problems among homeless men and women. Rockville, MD:
R.O.W. Sciences, Inc., 1992.
12. Birchmore Timney C, Graham K. A survey of case management practices in addictions
treatment. Alcohol Treatment Q 1989;6(3/4):103-27.
13. Perl H, Jacobs ML. Case management models for homeless persons with alcohol and other
drugproblems: An overview of the NIAAA Research Demonstration Program. In: Ashery RS,
ed. Progress and issues in case management. NIDA Research Monograph no. 127. Rockville,
MD: U.S. Department of Health and Human Services, 1992:208-22.
14. Sullivan WP, WoIk JL, Hartmann, DJ. Case management in alcohol and drug treatment:
Improving client outcomes. Families Society: J Contemp Human Serv 1992 Apr;73:195-203.
15. Willenbring ML, Ridgely MS, Stinchfield R, et al. Application of case management in alcohol
and drug dependence: Matching techniques and populations. Rockville, MD: National
Institute on Alcohol Abuse and Alcoholism, 1991.
16. Orwin RG, Sonnefeld LJ, Garrison-Mogren R, et al. Pitfalls in evaluating case management
programs: Lessons from the NIAAA Community Demonstration Program. Eval Rev
1994;18(2):153-207.
17. CollinsRC,KinneyPF. HeadStartresearchandevaluation:Backgroundinformation. Vienna,
VA: Collins Managements Consulting, Inc., 1989.
18. Goldman HH, Morrisey JP, Ridgely MS, et al. Lessons from the Program on Chronic Mental
Illness. Health Aff 1992;ll(3):51-68.
19. Ridgely MS, Willenbring ML. Application of case management to drug abuse treatment:
Overview of models and research issues. In: Ashery RS, ed. Progress and issues in case
management. NIDA Research Monograph no. 127. Rockville, MD: U.S. Department of Health
and Human Services; 1992:12-33.
20. Trochim WM. The regression-discontinuity design. In: Sechrest L, Perrin E, Bunker J, eds.
Research methodology: Strengthening causal interpretations of nonexperimental data. Wash-
ington, DC: U.S. Department of Health and Human Services; 1990:119-40.
21. Wholey JS. Evaluation: Promise and performance. Washington, DC: Urban Institute, 1979.
APPENDIX A
PROJECT-LEVEL DESCRIPTIONS
PROJECT: TREATING HOMELESS AND MENTALLY ILL SUBSTANCE ABUSERS
INALASKA
Orwin, Goldman, Sonnefeld, et al.
347
Sponsor
Program compo-
nents
Outcome evaluation
design
Distinctive features
Salvation Army Clitheroe Center, Anchorage, AK
Diagnostic Screening Center (DSQ, sobering, shelter, motivational inter-
views, residential treatment, and transitional living centers.
None
This project was designed to address the "revolving door" issues of ineffec-
tive treatment for homeless chronic public inebriates and those with dual
diagnoses with the introduction of the DSC—a drop-in center for clients. At
the DSC, clients could sleep off their intoxication and receive assessment,
information and referrals, in a nonthreatening environment. From there,
clients could gain access to a multitude of services leading to a continuum of
care.
Difficulties encoun-
tered
The plan for a continuum of services was never fully implemented during the
grant period due to neighborhood opposition to the DSC facility; an insuffi-
cient program design; insufficiently trained staff; and cramped, temporary
facilities. Linkages with Anchorage Community Mental Health Center,
which would havepermitted treatment options for dually diagnosed clients,
were never developed as planned. Project leadership eventually was able to
overcome barriers to leverage support from service providers, community
leaders, and state and local officials.
PROJECT:
Sponsor
STABILIZATION SERVICES FOR HOMELESS SUBSTANCE ABUSERS
Massachusetts Department of Public Health, Division of Substance Abuse
Services, Boston, MA
Program compo-
nents
Outcome evaluation
design
Features
Case management, post-detoxification residential substance abuse pro-
gramming (stabilization) in shelter settings, transitional and extended sober
housing.
Clients were randomly assigned to case management or customary aftercare
after three weeks at a stabilization site. Receipt of case management and
entry into the comparative phase of the study were contingent on completion
of the stabilization program. Clients were assessed at intake, and at one,
three, six, and nine months.
This program demonstrated the feasibility of incorporating recovery-ori-
ented activities, including case management, into homeless shelters. The
project also enhanced system linkages by coordinating services among
detoxification facilities, stabilization sites, and recovery homes throughout
the state.
Difficulties encoun-
tered
There were no major implementation problems. As is often the case,
however, the introduction of a new program into existing facilities posed
challenges. These included the necessity of maintaining program protocols
that differed from the host program's established way of operating. Staff
turnover was also a problem, both of project staff (case managers) and staff
at the detoxification centers and stabilization sites, causing some discontinu-
ity in the program. Monitoring staff turnover and providing repeated
orientations at the detoxification centers and at the four stabilization sites
proved important in minimizing disruption.
348
Alcohol and Drug Abuse Treatment of Homeless Persons
PROJECT:
Sponsor
Program
nents
compo-
Outcome evaluation
design
SOBER TRANSmONAL HOUSING AND EMPLOYMENT PROJECT (STHEP)
Los Angeles County Department of Health Services, Office of Alcohol
Programs (OAP), Los Angeles, CA
A two-phase residential program involving a 90-day clinical treatment
program and a subsequent 120-day recovery program. The first phase
involved assessment, individual counseling, self-help, literacy classes, pre-
employment training, and public speaking. The second phase provided
clients time to develop a foundation in sobriety, housing and employment
assistance, and opportunities to develop social support networks.
To test the effectiveness of the two-phase model, a comparison group
received the firstphaseonly. In a modified randomization scheme, individu-
als were placed on a waiting list and were eligible for randomization into the
treatment group for up to 45 days. The randomization procedure itself took
place when there was an open bed because an individual either completed
Phase 1 or left prior to completion. Individuals whose names were not
selected after 45 days of residency were, by default, assigned to the compari-
son group. Women and Hispanics were given priority admission to the
treatment group and were not always randomized. Clients were assessed at
intake and at 10 months.
Features
Difficulties encoun-
tered
PROJECT
Sponsor
Program
nents
compo-
Outcome evaluation
design
Rather than developing service components that would duplicate existing
services, OAP chose to focus on building the capacity of the larger system to
respond to the needs of the target population, with a specific emphasis on
housing resources and vocational opportunities. STHEP did this by in tegra t-
ing services of two existing treatment programs, while functioning as an
advocate in the housing, employment, and alcohol and other drug recovery
sectors and providing the technical assistance necessary to promote cross-
sector initiatives. Notably, the linkage component of STHEP was able to
enhance available housing resources by facilitating the development of 300
additional beds in alcohol- and drug-free living centers.
Problems were related primarily to implementing a program that requires
cooperation among organizations that operate under different philosophies
and that use different strategies to promote recovery. The project also
experienced delays in renovations and state licensing of facilities, project
director turnover, and problems in engaging and retaining women and
Hispanics in the program.
PROJECT CONNECT FOR HOMELESS ALCOHOL AND DRUG ABUSERS
Jefferson Alcohol and Drug Abuse Center, Seven Counties Services, Inc.,
Louisville, KY
Sobering-up station, intensive case management, work-adjustment training,
jail liaison, access to housing, alcohol and other drug treatment, and health
and social services.
At first, the intervention was made available to all eligible clients. In the
second year, a comparison group was formed that received the same types
of treatment, housing, employment, and other community services as the
treatment group, but did not receive intensive case management. "High-
risk" clients continued to be automatically referred to the treatment group,
Orwin, Goldman, Sonnefeld, et al.
349
Features
Difficulties encoun-
tered
while all others were randomly assigned to the treatment or comparison
group. Clients were assessed at intake, and at eight and 24 months.
Through a 20-bed sobering-up station, the project offered a point of entry for
inebriated homeless men to receive case management by which to link them
with needed services in the community. A jail liaison helped to identify,
target, and refer public inebriates to the project. Work-adjustment training
also was provided, when appropriate. A broader goal of the program was
to increase cooperation and formal linkages among agencies in Louisville
that provide services to homeless individuals.
There was some difficulty in negotiating a consistent approach to recovery
because of differences in philosophy andexpectations among the sponsoring
agencies. Project Connect demonstrated the importance of clear contracts
and specific role definitions and job responsibilities. Other problems were
related to start-up and initial community acceptance of the sobering-up
station.
PROJECT
Sponsor
Program compo-
nents
COMMUNITY TREATMENT FOR THE CHRONIC PUBLIC INEBRIATE
Chemical Health Division, Hennepin County Community Services Depart-
ment, Minneapolis, MN
Intensive case management, intermediate case management
Outcome evaluation
design
The original design called for individuals to be screened for program
eligibility after referral from the county detoxification center or other agen-
cies and then to be assigned to one of three groups: intensive case manage-
ment, intermediate case management, or a control group. Because the
intermediate case management project was to be implemented about three
to four months ahead of the intensive case management project, the evalua-
tion design was modified so that randomization occurred before screening.
Clients were assessed at intake, and at three, six, nine, 12, 18, 24, and 36
months.
Features
Difficulties encoun-
tered
The goal of intensive case management in this project was to improve client
functioning and the quality of life without necessarily expecting abstinence.
Intensive case management included aggressive outreach, "enforced con-
tact," and continuous care provided in the context of along-term relationship
with a case manager. Native Americans were a focus of this study because
they are a significant proportion of the target population in Minneapolis, and
little was known about why they exclude themselves from most service
settings. The research study included a cost-effectiveness component, which
indicated that case management of this target population can yield consid-
erable savings to local government.
Although there was overall county government support for the intervention,
the project was opposed by a small but vocal group of dissenters from within.
The program model was considered innovative and therefore open to
question. Adherence to Al-Anon and principles of the 12-step approach to
recovery was a strongforce to contend with, and "turf" protection was noted
in the struggle toward implementation. These ideological conflicts not only
occurred at the system level but also affected the project at the operational
level, in staff training and retention difficulties. In addition, the work of case
350
Alcohol and Drug Abuse Treatment of Homeless Persons
PROJECT
Sponsor
Program
nents
Outcome evaluation
design
managers was made difficult by the lack of community treatment resources
for the target population.
OUTREACH AND ENGAGEMENT FOR HOMELESS ALCOHOUC WOMEN
Women In Need, Inc. (WIN), New York, NY
compo- Outreach-engagement teams, first working in two midtown welfare hotels,
and when they were closed by New York City midway through the project,
in a family shelter complex; acupuncture therapy for relief of withdrawal
symptoms and maintenance in recovery; GED/literacy training; housing/
employment counseling; child care assistance; family life counseling related
to issues of child welfare and advocacy.
The treatment group in this study consisted of those women with alcohol and
other drug problems living in the targeted welfare hotels or shelter who
received outreach and engagement services. After implementation, a
nonequivalent comparison group was established, consisting of women
who came to the WIN treatmentclinic through means other than contact with
the outreach workers, regardless of their place of residence. Clients were
assessed at intake, and at six, 12,18, and 24 months.
Through significant changes in staffing and project management, as well as
other program adjustments, the project demonstrated that a program can be
established within a temporary housing facility to provide outreach and
engagement (into treatment) services to women with alcohol and other drug
problems. It further demonstrated the difficulty of implementing new
services in a turbulent environment and the feasibility of doing so if leader-
ship, persistence, support, and guidance are available.
When confronted with the extreme challenges of alcohol and other drugs
within welfare hotels and shelters, project staff, trained in social welfare, fell
back on the more familiar and basic areas of support—housing and child
care. Although this support was importan t, the project was required to focus
on alcohol and other drug problems. This focus could be accomplished only
with external pressure and internal reorganization. In addition, as members
of an outside agency moving on-site to the welfare hotels and shelters, the
project staff had difficulty obtaining basic supports, including space, heat,
and electricity.
PROJECT: COMPREHENSIVE HOMELESS ALCOHOL RECOVERY SERVICES
Sponsor Department of Alcohol and Drug Programs, Alameda County Health Care
Services Agency, Alameda County, CA
Features
Difficulties encoun-
tered
Program compo- Alcohol Crisis Center (ACQ, residential and nonresidential recovery ser-
nen ts vices, community linkages and education focused on improving services for
the target population
Outcome evaluation None
design
Features
Based on a social-community orientation, this project offered homeless
persons with alcohol and other drug problems a wide range of recovery and
alcohol- and drug-free housing options within the county system of care.
The entry point was primarily the ACC, where clients could sober up and
Orwin, Goldman, Sonnefeld, et al.
351
Difficulties encoun-
tered
PROJECT
receive information and referrals in a nonthreatening environment. The
project also worked to establish formal linkages within and outside of the
county among agencies serving the target population.
There were significant delays in establishing the ACC due to problems with
the contracted provider. Many additional problems were related to manag-
ing a comprehensive service delivery network. Working within a large
county system also set a unique context for the project. Barriers in imple-
menting the ACC and challenges in administrating the CHARS project
resulted in several changes in project design during the grant period. These
included administrative-level changes as well as alterations of specific
program components.
FAMILY TREATMENT FOR HOMELESS ALCOHOL AND DRUG-
ADDICTED WOMEN AND THEIR PRESCHOOL CHILDREN
Sponsor
Program compo-
nents
Outcome evaluation
design
Features
Difficulties encoun-
tered
PROJECT:
Diagnostic and Rehabilitation Center, Philadelphia, PA
Six-month residence in drug-free environment, outpatient treatment, case
management, social services.
The original design included random assignment to one of two conditions:
a drug-free residential setting for women with their children and extensive
outpatient treatment (treatment group), and outpatient treatment only (com-
parison group). However, the nonresidential outpatient group was never
formed due to extreme no-show rates, so the comparative design was
abandoned. Consequently, the project was not included in the national
evaluation outcome study.
Designed to break the cycle of homelessness and alcohol and other drug
problems amongmothers, this was the first treatment program in the city to
admit women with their children. The project provided group housing and
case management for the residents supplemented by outpatient treatment
services located off-site. When the need was identified, transportation was
provided between the two sites.
Problems were related to meeting the immense needs of the clients for basic
care of themselves and their children, and frustrations over the fate of the
families after they left the project, because of the lack of low-income housing
in Philadelphia. Additional difficulties included (as noted above) the
inability to attract women to the outpatient group, despite repeated efforts.
COMPREHENSIVE REHABILITATION SERVICES FOR DUALLY
DIAGNOSED HOMELESS INDIVIDUALS
Sponsor
Program compo-
nents
Outcome evaluation
design
Horizon House Rehabilitation Services, Philadelphia, PA
Residential placement and case management and rehabilitation services to
homeless individuals with serious mental illness and alcohol and other drug
problems (individuals with a dual diagnosis).
Aquasi-experimen tal design included twononequivalentcomparison groups:
programs for dually diagnosed homeless individuals, run by two drug
rehabilitation agencies in Philadelphia: one lower demand and one higher
demand than the treatment group (a low-demand residence initially makes
few demands on its residents but gradually adjusts and heightens expecta-
352
Alcohol and Drug Abuse Treatment of Homeless Persons
tions). The project was not included in the national evaluation outcome
study because follow-up data on comparison-group clients were unavail-
able.
Features
Difficulties encoun-
tered
This project focused on an extremely challenging population. Homeless
individuals who have been dually diagnosed often are unwelcome in the
mental health service system, the recovery services system, and in the service
system for homeless individuals. This project coordinated services for
clients through a system of case management within a structured group
residence providing psychosocial rehabilitation.
Project managers had difficulty in implementing the program because of the
great challenge of overcoming fragmentation among the service systems for
this population—both externally in the Philadelphia human services envi-
ronment and internally within Horizon House itself. Internal tensions
mirrored the fragmentation in the external environment. Recovery services
were not administratively supported within the mainstream of psychosocial
rehabilitation services at Horizon House. There also was considerable
change in the leadership and management of the project, and difficulties in
recruiting and retaining case managers to work in the project.
