ResearchPDF Available

NYC Frequent Users (FUSE) Initiative: Evaluation of a Supportive Housing Intervention

Authors:

Abstract

Housing instability/homelessness increases risk for incarceration and, conversely, incarceration increases the risk for homelessness. While residential instability is common after prison or jail, research has identified a subset of persons with repeated episodes of both incarceration and homelessness. They are often ‘frequent users’ of other services as well, especially crisis care services such as provided in emergency room, hospital, detox, residential mental health settings. Most are poorly served by repeated travels through the ‘institutional circuit’ at great human and social cost. The Frequent Users Services Enhancement (FUSE) program was a collaborative policy initiative to address this human and care system challenge. FUSE NYC was a collaboration with the Corporation for Supportive Housing; The New York City Departments of Homeless Services, Correction, Health and Mental Hygiene, and Housing Preservation and Development; The New York City Housing Authority; and ten non-profit providers of housing and services. FUSE provided supportive housing to roughly 200 individuals who were frequently cycling in and out of jails and homeless shelters. Researchers at Mailman School of Public Health directed the evaluation. The evaluation followed a cohort of program participants from recruitment through two years after being placed into supportive housing and a closely matched comparison group. Data were provided by detailed in-person interviews conducted at approximately 6 month intervals, and overtime analysis of administrative data provided by NYC Departments of Corrections and Homeless Services. The evaluation shows that supportive housing significantly reduced cycling between public systems, reduced days spent in jail and shelter, and lessened use of crisis care health services. Cost analyses showed that these service use reductions resulted in significantly lower costs for government and for society as a whole. This report describes the evaluation and outcomes of FUSE and implications for policy and systems change.
FREQUENT USERS SERVICE ENHANCEMENT
‘FUSE’ INITIATIVE
NEW YORK CITY FUSE II
EVALUATION REPORT
Angela A. Aidala, PhD
Columbia University, Mailman School of Public Health
William McAllister, PhD
Columbia University, Interdisciplinary Center for Innovative
Theory & Empirics
Maiko Yomogida, MA
Columbia University, Mailman School of Public Health
Virginia Shubert, JD
Shubert Botein Policy Associates, LLC
NOTE TO READER
For those reading this report on a computer, we have linked the report internally in the following ways:
(a) each section and subsection name listed in Contents has been linked to the beginning of each
section and subsection; (b) each major division of the Executive Summary has been linked to its
corresponding section in the body of the report; (c) text references to sections and subsections have
been linked to the beginning of each section or subsection; and (d) references to specific tables and
figures have been linked to the relevant table or figure. To use the links, which are hidden, place the
cursor over the section, subsection or table/figure number you wish to move to; when the cursor
becomes a hand with an index finger pointing at the link, click and the document will move to the
linked section, subsection, table or figure.
CONTENTS
Page
Acknowledgements/Authors’ Note
Executive Summary
i
I. Introduction
1
A. Incarceration, Homelessness and Frequent Users
1
B. Frequent Users Services Enhancement
2
II. Evaluation Questions and Methods
7
A. Evaluation Questions
7
B. Research Design
7
C. Comparison Group Formation
8
D. Data Collection
11
E. Description of Frequent Users
13
III. Outcomes Analyses
19
A. Housing Outcomes
19
B. Incarceration Outcomes
24
C. Substance Use, Health, Mental Health and Social Support Outcomes
26
D. Crisis Care Medical, Mental Health and Substance Abuse Services Outcomes
28
E. Institutional Trajectories
30
IV. Cost Evaluation
41
A. Background
41
B. Objectives
41
C. Methods
41
D. Cost and Cost Offset Results
46
E. Cost Analysis Discussion
50
V. Evaluation Summary and Implications
53
References
55
Appendices
A. Propensity Score Analysis
B. Excluded Cases
C. Baseline and Follow-up Questionnaire Measures
D. Comparison Group Screening Questionnaire
E. Baseline Questionnaire
ACKNOWLEDGEMENTS
The NYC Frequent User Service Enhancement (FUSE) initiative represents a collaboration
between the New York City Department of Correction (DOC), the NYC Department of
Homelessness Services (DHS), the NYC Housing Authority and the Corporation for Supportive
Housing (CSH). CSH has funded the FUSE evaluation with the support of the Robert Wood
Johnson Foundation, Jacob and Valeria Langeloth Foundation, the JEHT Foundation and the
Open Society Foundation.
This work would not be possible without the dedication and support of our project staff. We
thank Jocelyn Apicello, our initial Project Director, who got us organized and launched, and
served as agency liaison and fieldwork supervisor. We also thank our team of sensitive and
effective interviewers: Anne Bozack, Jeanne Campbell, Nina Chkareuli, Cassandra Melnikow,
Jeffrey Miller, Susannah Slocum and Gloria Thomas. Thanks also to Jennifer Hill, who helped us
think through various statistical issues; Judy (Chung Min) Kim and Martin Schuster, who helped
with data analysis tasks; and Haydee Cespedes, who provided invaluable administrative support.
Very special thanks goes to the housing agencies and programs participating in the FUSE
initiative, and especially to the FUSE housing tenants and other study participants who shared
with us their time, trust and experiences. We also thank CSH staff for their help in facilitating our
work, especially Jacquelyn Anderson and Ryan Moser.
The contents of this report are solely the responsibility of the authors and do not necessarily
represent the official views of New York City’s Department of Corrections, its Department of
Homeless Services or the Corporation for Supportive Housing.
AUTHORS’ NOTE
Angela Aidala and William McAllister designed the study and oversaw its execution. Both also
drafted most of the text and oversaw the substantive data analyses. Maiko Yomogida also
conceptualized these analyses and carried them out; as well, she was the Project Manager. Virginia
Shubert, in conjunction with Aidala, carried out the cost evaluation analyses and drafted that text.
EXECUTIVE SUMMARY
FUSE II Program
Housing instability/homelessness increases risk for incarceration and, conversely, incarceration
increases the risk for homelessness. To address these risks, the Frequent Users Services Enhancement
(FUSE) initiative was developed in a collaboration between the Corporation for Supportive Housing;
The New York City Departments of Homeless Services, Correction, Health and Mental Hygiene, and
Housing Preservation and Development; The New York City Housing Authority; and ten non-profit
providers of housing and services. FUSE provided supportive housing to roughly 200 individuals who
were frequently cycling in and out of jails and homeless shelters. This evaluation follows a subset of
those participants from recruitment through two years after placement into supportive housing. The
evaluation analyzes the experiences of a group of people with complex involvement in multiple public
systems, numerous barriers to housing and complicated histories of behavioral health, physical health
and significant trauma. It shows supportive housing significantly improved their lives by reducing their
cycling between public systems, their days spent in jail and shelter and their use of crisis health
services. These service use reductions resulted in significantly lower costs for government and for
society as a whole. This report describes the intervention, evaluation and outcomes of FUSE II, a
second generation FUSE initiative.
Compared to people with homes, persons without stable housing necessarily live more in public
spaces, where they are more visible to authorities and are often targeted for ‘disruptive’ or ‘quality of
life’ offenses. The war on drugs captures persons for using or possessing even small amounts of
controlled substances, thereby incarcerating millions who struggle with addiction and, often, co-
occurring mental illness. At the same time, prison and jail experience increases the risk of housing loss
and homelessness. While many people experience some form of residential instability after prison or
jail, research has identified a subset of persons with repeated episodes of both incarceration and
homelessness. They are ‘frequent users’ of other services as well, especially crisis care services such as
hospital emergency departments, inpatient and residential mental health and substance abuse treatment
facilities.
Given the success of supportive housing models to improve residential stability and community
integration of persons with histories of homelessness and behavioral health conditions, the
Corporation for Supportive Housing (CSH) launched the Returning Home Initiative to help these
frequent users. Its central premise is that the thousands of people with chronic health and behavioral
health conditions cycling in and out of incarceration and homelessness are poorly served by these
systems and at great public expense. Returning Home argues that establishing permanent supportive
housing as a key component of reentry services for persons with recurring experiences of
homelessness and criminal justice involvement will improve their life outcomes, more efficiently utilize
My life was in turmoil. I was trying to find myself and be somebody other
than me at the same time. I was fighting my addiction but running with the
guys that were getting high. I was fighting the devil. My life was a revolving
door.
--Program participant describing life before FUSE
ii
public resources, and likely save costs in publicly funded crisis care systems, including emergency
medical, mental health and addiction services.
As part of this initiative, CSH, along with New York City’s Departments of Correction and Homeless
Services, established the Frequent User Service Enhancement (FUSE) Program. The FUSE model has
three core elements:
Data-driven problem-solving. City agencies and/or service providers use data to identify a specific
target population of high-cost, high-need individuals who are clients of multiple systems (e.g.,
jail, shelters and hospitals) and whose persistent cycling among these systems indicates the
failure of traditional service approaches. Such data is used also to measure the success of
program efforts on human and public costs and to demonstrate that individuals are able to
avoid cycling among institutions altogether, rather than being off-loaded from one system to
another.
Policy and systems reform. Public systems and policymakers engage in a collective effort to address
the needs of shared clients, shift resources away from costly crisis services towards permanent
housing solutions, and better integrate resources and policies across systems.
Targeted housing and services. Supportive housing permanent housing linked to individualized
supportive services is enhanced with targeted and assertive recruitment through in-reach
into jails, shelters, hospitals and other settings to help clients obtain housing stability and avoid
returns to costly crisis services and institutions
After a promising first generation initiative based on these elements, known as FUSE I, was concluded
in New York City, a second generation program was undertaken, known as FUSE II. This document
reports the initial findings of an evaluation of FUSE II, conducted by researchers from Columbia
University and Shubert Botein Policy Associates.
The FUSE II Intervention
The threshold eligibility criteria for FUSE II participation was four jail and four shelter stays over the
five years prior to admission. These stays were determined by administrative data match between jail
and public shelter records. Additional criteria were used by specific housing providers, based on client
eligibility for available types of housing assistance. Clients either had substance abuse treatment within
the past 12 months, no recent problem alcohol or drug use and expressed readiness for change, or had
a serious psychiatric diagnosis and mental health treatment in the past year. While these criteria were
influenced by NY/NY III criteria (New York State, 2005), not all programs had these entry
requirements. FUSE II leveraged resources from supportive housing production programs in New
York City that were targeted to persons experiencing or at-risk of homelessness with extremely low
incomes. It used funding committed to assisting persons experiencing homelessness, employing
targeted units of existing government funded permanent supportive housing for extremely low income
homeless New Yorkers with diagnoses of serious and persistent mental illness (SPMI)
1
and/or
1
The mental health community no longer uses the term ‘serious and persistent mental illness’. We are using this
term because this was the terminology specified in the NY/NY III applications at the time they were enrolling
clients.
iii
substance use disorder. FUSE II participants received permanent supporting housing in either
scattered-site housing with services provided through mobile case management teams and other staff,
or single site, mixed-tenancy buildings operated by non-profits as special needs housing with onsite
services. Units were subsidized such that the tenant pays no more for rent than 30% of income or of
their housing allowance from benefits. Housing providers were given a one-time $6,500 payment per
client to allow for flexible service funding during the critical period from recruitment and engagement
to linkage with sustainable, comprehensive medical and mental health services and other support
services needed to promote stability and tenant success. Use of this enhancement varied by housing
program, but included spending for clinical supervision; client recruitment and engagement; intensive
case management with lower client-to-case manager ratios; special FUSE II service staff to provide
more intensive support during the first year of housing; and/or additional specialty services as needed.
Evaluation of FUSE II
We designed the evaluation to measure the impact of FUSE II on a number of outcomes consistent
with the intent of the program. We analyze the effect of the intervention on clients’ (1) retention in
permanent housing and avoiding homelessness; (2) criminal justice involvement, including arrests and
returns to jail; (3) problem drinking and drug use; (4) health and mental health; (5) connection with
family and other forms of social support; (6) use of health, mental health and substance abuse services;
(7) over all temporal patterns of institutional involvement beyond participants’ use of individual public
systems, i.e., reduced cycling between institutions. In addition, we analyze the cost of the FUSE II
intervention and possible cost offsets resulting from reduced public expenditures associated with using
shelter, medical, behavioral health and criminal justice systems.
Our basic evaluation design is two-group pre/post, with a comparison group constructed among
FUSE II-eligible individuals who strongly match those receiving the FUSE II intervention. The
intervention group consisted of the 72 people who were provided FUSE II housing and services. We
recruited potential comparison members by working with the housing provider agencies to determine
how they selected individuals for their programs from among the larger FUSE II-eligible population
and mimicking as much as possible the strategy and tactics these agencies used to locate and recruit
person’s eligible for their services. We visited the same few shelters from which the programs recruited
and used a screening questionnaire that covered topics the service providers used to assess suitability
for their specific housing program. Using this approach, we recruited 89 potential comparison group
members who importantly matched those selected by the agencies for the intervention.
We then used propensity score analysis to improve the comparability of the intervention and
comparison groups. This technique allowed us to estimate a “score” representing the probability (i.e.,
“propensity”) of being selected for the FUSE II program for people in both the potential comparison
group and the actual intervention group. The technique estimates this score based on a model that
incorporates many pre-intervention demographic, clinical, experiential and service use variables
thought to affect chances of being selected for the program and/or to affect outcomes. We used this
score to select people for the comparison group who had scores comparable to those in the
intervention group and successfully tested to make sure no strong differences exited between the two
groups. This analysis resulted in a trimmed sample for analysis of 60 intervention group members and
70 strongly matched comparison group members.
iv
We followed participants in both intervention and comparison groups for up to 24 months after
baseline data collection by surveying them at roughly six month intervals. For the intervention group,
we conducted the baseline assessment immediately subsequent to their move into FUSE II supportive
housing; for the comparison group, we conducted the baseline interview at study enrollment, which
was timed to coordinate with intervention group assessments. In addition to these surveys, we used
administrative data from the NYC Departments of Corrections and Homeless Services. An
administrative data match provided information on jail and homeless shelter experience for five years
prior and two years following enrollment in FUSE II or, for comparison group members, the baseline
assessment.
An examination of background characteristics and experiences of the study population found a pattern
of overlapping personal vulnerabilities and experiences of social exclusion including extreme poverty,
minority race/ethnicity, long experience of homelessness, chronic illness, mental health challenges,
substance use problems, lack of family or social connections and histories of victimization. Regarding
the criminal justice profile of frequent users, three-fourths have been incarcerated for drug related
charges, overwhelmingly for possession. However, repeated incarcerations are more often associated
with low-level misdemeanors such as theft of services” (mostly jumping the turnstile for public transit
access), “quality of life” offenses (vagrancy, trespassing, loitering, disorderly conduct, public urination),
and probation or parole violations rather than additional drug convictions. This highlights the need to
understand better how structural factors such as local laws and police practice interact with individual
mental health, addiction, or other vulnerabilities to increase the risk for re-incarceration among the
frequent user population.
We estimated effects of the program by using OLS and logistic regression models that included
theoretically relevant variables or those that previous research suggested mattered for the outcomes of
interest.
Program Effects
Permanent housing. Comparing housing situations of intervention and comparison group members
at 12 and 24 months after baseline show extremely strong support for the effect of FUSE II on obtain-
ing and maintaining permanent housing among program participants. The following results are all
highly statistically significant:
At twelve months, over 91% of FUSE II participants were housed in permanent housing,
compared to the 28% who would have been housed had they not received FUSE II housing
and services.
By 24 months, FUSE II participants experienced a slight drop to 86% who were in permanent
housing. By this point in time, only 42% of comparison group members were in permanent
housing.
The small change over time in the FUSE II participants housing situation speaks well for the
lasting effects of the program. This 24-month analysis suggests it is likely that FUSE II-
induced effects will be sustained past this study’s two year follow-up period.
v
Homeless shelter use. Our analysis shows the FUSE II program effectively reduced homeless shelter
use. These effects are substantively and statistically very strong. Measuring shelter use from housing
placement for FUSE II participants and from study enrollment for comparison group members, the
major findings are:
On average, intervention group members spent 146.7 fewer days in shelter than did
comparison group members.
The percentage of FUSE II participants with any shelter episode over the study period was
reduced on average by 70%.
Incarceration. Results for incarceration show reductions in jail involvement benefiting the
intervention group and most, although not all, results are statistically significant. Measuring from
housing placement and compared to the comparison group, people receiving the intervention had, on
average:
19.2 fewer days incarcerated, a 40% reduction over the comparison group.
Fewer jail admissions over the 24 month follow-up period.
For incarceration and homeless shelter use, effects were also measured from when FUSE II
participants were first enrolled in the program, which, for most, was several months prior to placement
in permanent housing. This drawn out placement process was largely driven by a very slow application
and approval process for Section 8 vouchers. This process required extensive documentation for
people with limited access to personal records and, in a significant number of cases, proof of income
for people with no access to public benefits or employment. For homeless shelter use, results
measured from program enrollment are significant but less robust than results measured from actual
housing placement. However, for incarceration, there is little difference in findings whether measured
from initial program enrollment or from housing placement. It may be that the promise of permanent
housing and/or initial activities by FUSE II program staff to engage clients and connect them to
services in support of the housing placement process contributed to reduced risk of recidivism.
Substance use, mental health and health functioning. Intervention effects on substance use,
health and mental health present a mix of program effects:
The FUSE II program had a significant and positive effect on drug abuse outcomes. The
percentage with any recent use of hard drugs (heroin, cocaine, crack, methamphetamine) is
half as high as the comparison group and current alcohol or substance use disorder is one-
third less for FUSE II participants at follow-up than among comparison group members, des.
This is despite similar histories of chronic, relapsing addiction and recent substance abuse
treatment prior to baseline interview.
Half of all study participants, both FUSE II and comparison group members, screened
positive for a current psychiatric disorder although there were differences in specific mental
health issues.
Compared to the comparison group, the intervention group score significantly lower on a
measure of psychological stress and higher on measures of current family and social support,
factors associated with improved social functioning among those with mental illness
vi
Physical health functioning is lower for intervention group than comparison group members;
however, it is not clear that the difference is sufficiently large to indicate a clinically significant
difference.
Crisis care service use. Of particular importance to public spending is the effect of FUSE II on the
use of ‘crisis care’ health and behavioral health serviced: ambulance rides, emergency department visits,
hospital inpatient stays, inpatient mental health or substance abuse treatment, o: medically supervised
detox. In general, the service use findings suggest a reduction in some but not all categories of service
use resulting from the intervention:
Ambulance rides were significantly less for the intervention group. Comparison group
members had an average of 1.2 ambulance rides; FUSE II participants had fewer than one
ambulance ride (mean 0.67).
Comparison group members spent on average eight days hospitalized for psychiatric reasons,
4.4 days more than intervention group members.
Services use difference was especially strong for AOD residential treatment: people in the
comparison group spent on average almost 10 days in such a facility compared to no days for
those in the intervention group.
Average number of hospitalization days for medical reasons and of emergency room visits for
any reason showed no substantial differences between intervention and comparison groups,
though the slight differences were in the direction the program expected to create.
Mean number of AOD inpatient hospital days and mean number of detoxification days
showed differences that were not in the direction the program expected, but these differences
were not statistically significant. Wide confidence intervals indicate substantial variation in
these outcomes.
These findings only scratch the surface of the relationship between FUSE II (and programs like it) and
use of medical and behavioral health services. It may be a positive impact of the program that some
kinds of services use increase while others are reduced. A program that stably houses people and
provides them access to a range of client-centered services may be creating the conditions for people
to have unidentified problems become known and at an earlier stage of the problem than would
otherwise have been the case. From this perspective, increases in some kinds of service use might be
expected (and be the kind of effect the program seeks). For example, it may be that hospitalization for
medical reasons increases as people get treatment for ailments postponed or that would otherwise go
unknown. That people in the intervention group are completely able to avoid longer-term residential
AOD treatment may mean that the program effectively helps people sustain recovery or reduce the
severity of relapse experience. All in all, identifying what kinds of services use effects to expect needs
to be subtly scrutinized to understand what constitutes program success regarding particular services.
vii
Trajectory Analyses Findings
Previously reported results concerning incarceration and shelter use show us differences between
comparison and intervention groups by summing information over the follow-up period, e.g., the sum
of the number of days jailed or sheltered. Here we report our analysis comparing over-time patterns of
incarceration and shelter use between the comparison and intervention groups. This gives us evidence
of how the intervention affected people as they were living their lives, month in and month out. To the
extent that the intervention had effects, this shows us when in the follow-up those effects were
occurring, how long they lasted and what preceded and followed these effects.
Trajectory analyses produce classes of people who have similar histories over follow-up. In this instance,
these are trajectories of people who, for consecutive thirty-day periods (which we also call “months”),
had similar histories of being incarcerated or in shelter for at least one day during each thirty-day
periods versus not being in either (or both) situations during the entire thirty-day period. Thus, the
comparison we report here between the intervention and comparison groups is a comparison of the
number and size of classes and their makeup with regard to patterns of incarceration and shelter use
and not being in either institutional setting.
Incarceration. The findings for the incarceration-only trajectories are the following:
The intervention group has two fewer classes compared to the comparison group, suggesting
that the intervention creates more homogeneous histories over follow-up. One way to think
about this is that the intervention changes more chaotic lives into more orderly lives.
The intervention and comparison groups each has a class of people with no incarceration
history over follow-up, and the class is about the same size, representing about half of each
sample.
The intervention group has a class of people (22% of the sample) with only one month
showing any incarceration, sporadically over the follow-up period. That is, but for this one
month with some incarceration experience, these people would have avoided incarceration
entirely. The comparison group, however, does not have this group, i.e., their patterns of
incarceration are more intense.
Overall, the intervention reduced the number of patterns and changed the nature of patterns
of those who had some incarceration. For the most part, individuals stopped cycling through
incarceration (though they may have had one jail episode), and incarceration was pushed to
later in the follow-up period.
Shelter use. The trajectory differences between the two groups are more striking for shelter use:
The intervention group has one less class than the comparison group, again suggesting the
intervention creates more ordered lives, at least regarding use of institutions like shelter).
The overwhelmingly modal class for the intervention group is people with no shelter use
history over the 24 months of follow-up (85%). The comparison group has no such class, but
rather many classes which all begin with people having shelter experience in the first month
but then stopping having shelter experience at different time points, i.e., in months two, eight
viii
and eighteen. Thus, the intervention transformed these comparison group histories of
different lengths of shelter use into histories of no shelter use.
The second largest class for the intervention group is very small, with only 6.7% of the
sample, and groups together people who were in a shelter during only one thirty-day period
but at different times over the follow-up period. By contrast, all comparison group classes are
characterized by people having different continuous months of shelter experience.
The main thrust of these findings is that the intervention virtually eliminated the different
patterns of shelter use found in the comparison group. It created a very large class of people
who no longer use shelter and a smaller class with very sporadic use, which, as a whole,
replaced the comparison group’s patterns of ever increasing contiguous shelter use from the
start of the follow-up period and its patterns of early and late contiguous shelter use.
Incarceration and shelter use and cycling. Here we report trajectory analysis results when we
consider whether people were in jail, shelter, both or neither. Because there are now four situations
people can possibly be in, the results are likely to be more complicated and they are:
The intervention group contains a large class (45%) who had no shelter or incarceration
experience while the comparison group did not have such a class.
The intervention group had a second large class (40%) with one or two months of jail
experience but no shelter episodes, while the comparison group had no such class of sporadic
jail or shelter experience. If we combine these two intervention group classes, fully 85% of
that group had no or a little, very sporadic shelter use, while the jail, shelter or combined use
experienced by the comparison group occurred in long continuous stretches of time over
follow-up.
We can summarize the trajectory analysis for institutional use by saying that it indicates a
strong impact of the intervention on the trajectories that people would have followed but for
the intervention. Those in the comparison group had fairly structured histories of shelter use
and incarceration, with the timing, sequencing and location (i.e., jail, shelter or both) defining
the variation between the classes. Except for a small number of people in the intervention
group, those receiving the FUSE II showed none of this, but rather exhibited histories of
either no or little and sporadic shelter use and incarceration experience.
Cost Evaluation Results
The cost evaluation seeks to answer three questions: (1) what is the cost per participant of the FUSE II
housing and enhanced services intervention; (2) what are the public cost implications of the observed
impact of the intervention on the jail, shelter and medical and behavioral health services use as
estimated by this evaluation; and (3) to what extent do cost reductions in these crisis and acute care
services offset the public costs of the intervention?
We used standard methods of cost analysis to calculate an average per-client, per-year cost of FUSE II
and to monetize service use outcomes reported in the impact sections of the report. These methods
include determining the number of clients served, identifying resources consumed, estimating the cost
per unit of each resource type, calculating the total cost of the intervention, and expressing all costs on
ix
a per client basis. We take a public payor or taxpayer perspective, designed to identify costs incurred by
public agencies, including federal, state and city payors. We also present intervention costs from a
societal perspective that includes all housing costs regardless of who pays, including participant
contributions to rent paid from earned income or from government funded public assistance or
disability benefits (but excluding other costs incurred by study participants such as travel costs or the
value of time spent in program activities).
We tracked NYC jail and municipal shelter use by study group members through the administrative
data obtained from DOC and DHS for the 24 months prior to and following the baseline interview
(typically conducted within one month of housing placement for the intervention group). Data on use
of inpatient and crisis medical and behavioral health services, as well as housing costs incurred by
intervention group members, were collected through the survey of intervention and comparison group
members conducted at baseline and at six-month intervals over follow-up. Cost findings for jail and
shelter use reflect outcomes for the trimmed and balanced sample of study participants 60
intervention group members and 70 comparison group members. Cost findings for self-reported
services use are based on responses from members of the study who completed on average 3.4 follow-
up interviews (i.e., 57 intervention group members and 52 comparison group members).
We obtained FUSE II intervention costs by interviewing program staff at each of the participating
housing provider agencies (who had reviewed cost records for their programs); undertaking a written
survey of provider agencies; reviewing provider agency materials; and interviewing CSH project staff
responsible for FUSE II project implementation and oversight.
As noted previously, we estimate intervention costs based on the 24-month period following each
study participant’s placement in FUSE II supportive housing or study enrollment (for comparison
group members). However, to provide useful cost comparisons, we present annualized intervention
costs for service use variables, expressed as the average or mean cost per person per year. All costs are
adjusted for inflation to reflect 2012 dollars.
Cost and Cost Offset Findings
The major fiscal findings we estimated are:
The annual average cost of the intervention from the payor perspective is $25,157 (2012
dollars, here and throughout), and from the societal perspective is $27,383. For both
perspectives, these costs vary by housing model and by program.
The intervention reduced annual average total costs for inpatient and crisis medical and
behavioral health services by $7,308 per intervention group member over the full 24-month
follow-up period,. The bulk of savings is attributable to reducing psychiatric inpatient days.
Results indicate an intervention effect reducing average total costs for shelter and jail days by
$8,372 per person per 12-month period.
For intervention group members for the 24 months prior to and following study enrollment,
the total per person average cost of shelter and jail days decreased from $38,351 in the 24
months prior to study participation to $9,143 in the 24 months following housing a
$29,208 or 76% reduction. This same cost also went down for the comparison group, but
x
from $38,598 in the two years prior to the study to $25,955 during the 24 follow-up period,
about 33% reduction.
Taking the full public payor intervention cost into consideration, including federal spending
for affordable housing vouchers, the $15,568 reduction in avoidable public costs offsets over
60% of the total public cost for FUSE II housing and services.
Conclusion and Implications
In sum, the intervention had strong positive effects on reducing homeless shelter and jail use,
especially when measured from housing placement. It transformed people’s patterns of institutional
cycling such that only a very small percentage of people in the intervention group had patterns akin to
the heavier use patterns of the comparison group. Rather, the patterns exhibited by the intervention
group show no or extremely infrequent jail or shelter experience.
The FUSE II intervention was highly successful in securing and maintaining permanent housing for
program participants. Rates of 12-month and 24-month success in maintaining housing are higher than
seen in other supportive housing interventions for persons with complex histories of homelessness
and behavioral health needs.
Strong program effects were also apparent for problem alcohol and drug use. Findings are less
consistent regarding mental health outcomes. Rates of current disorder are similar among intervention
and comparison group members. However, differences in psychological stress and in social support
favor FUSE II participants. Other research has shown that such differences are associated with
improved community integration, mental health functioning and quality of life among those with
persistent mental illness.
Findings from the cost evaluation of the FUSE II intervention indicate that removing policy and
system barriers limiting access to housing assistance for persons with criminal convictions,
incorporating housing into reentry services, expanding existing housing resources available for
homeless persons with health and behavioral health challenges, and giving housing providers an
additional onetime $6,500 enhancement per client for more intensive supportive services immediately
post release would result in substantial cost savings to corrections, homelessness and/or health care
systems for persons who would otherwise continue their cycling between jail and crisis care
institutions. FUSE II enhancements were largely used to address a mismatch of resources and system
barriers resulting from funding sources not being directly targeted to frequent users of jail and shelter
services. Future FUSE programs will not need additional enhancements. Every year in the United
States, local jails process an estimated 12 million admissions and releases. Poverty, homelessness,
chronic addiction, persistent mental illness, multiple health problems or all of these are widespread
among the jail population. Since 80% of inmates are incarcerated for less than one month, jails have
little ability to address these deep-seated personal and community challenges. Supportive housing has
been demonstrated to end homelessness for persons with complex needs and to reduce overall public
systems involvement and costs. The FUSE II program results described in this report add to this body
of evidence that supportive housing decreases recidivism, reduces chaotic use of expensive emergency
homeless, health and behavioral health services and improves health care access and outcomes, all
while helping government avoid unproductive spending.
FREQUENT USERS SERVICE ENHANCEMENT INITIATIVE (FUSE II)
EVALUATION REPORT
I. INTRODUCTION
A. Incarceration, Homelessness and Frequent Users
Housing instability/homelessness increases risk for incarceration and conversely, incarceration in-
creases the risk for homelessness. Compared to people with homes, those without stable housing
necessarily live more often in public spaces, more visible to authorities and often targeted for
‘disruptive’ or ‘quality of life’ offenses such as loitering, jaywalking, panhandling, public urination and
so forth. The war on drugs captures persons for their using or possessing even small amounts of
controlled substances, thereby incarcerating millions who struggle with addiction and, often, co-
occurring mental illness. Extreme poverty among homeless persons increases risk for incarceration for
minor offenses when resources are unavailable to make bail or pay fines. At the same time, prison and
jail experience increases vulnerability for homelessness. Incarceration disrupts family and community
relationships, limits employment prospects and interrupts and/or disqualifies receipt of public benefits,
all of which increases risk of homelessness. In addition, policies limit access to publicly funded housing
assistance for persons with a history of criminal conviction. With or without legal prohibitions,
landlords discriminate and communities resist providing housing to the formerly incarcerated
(Fontaine & Biess, 2012; Metraux, et al., 2007).
While many people experience some form of residential instability after prison or jail, research has
identified a subset of persons with repeated episodes of both incarceration and homelessness. As well,
they frequently use other services at high levels, especially crisis care services such as hospital
emergency departments, residential treatment facilities and inpatient mental health and substance abuse
services (Burt & Anderson, 2005; Hall, et al., 2009; Culhane, et al., 2007). Given the success of
supportive housing models to improve the residential stability and community integration of persons
with complex histories of homelessness and mental illness (for review see Rogers, et al., 2009), the
Corporation for Supportive Housing launched the Returning Home Initiative. The central premise of
Returning Home is that thousands of people with chronic health and behavioral health conditions
cycle in and out of incarceration and homelessness and are poorly served by these systems at great
public expense and with limited positive outcomes for their lives. It is thought that establishing
permanent supportive housing as a key component of reentry services will improve these people’s life
outcomes; more efficiently utilize public resources; and likely avoid expenses in crisis care systems,
including emergency medical, mental health and addiction services, as well as in correction facilities
and homeless shelters. The Returning Home Initiative works to coordinate resources and policies to
create supportive housing in communities across the United States for persons with high needs and
histories of homelessness who are leaving jail or prison (CSH, 2011).
This document reports the initial findings of an evaluation of an initiative under the umbrella of the
Returning Home Initiative, the New York City Frequent Users Services Enhancement program
(FUSE).
2
Figure 1. Blueprint for FUSE
B. Frequent Users Service Enhancement
As an integral part of the Returning
Home Initiative, the FUSE model as
developed by CSH has three core
elements (CSH, 2011; see Figure 1):
Data-driven Problem-Solving. Data is
used to identify a specific target
population of high-cost, high-
need individuals who are shared
clients of multiple systems and
whose persistent cycling indicates
the failure of traditional service
approaches. Measures of success
focus on human and public costs
and show that individuals are able
to avoid cycling among
institutions altogether, rather than
being off-loaded from one system
to another.
Policy and Systems Reform. Public systems and policymakers engage in a collective effort to address the needs
of shared clients, shift resources away from costly crisis services towards permanent housing solutions,
and better integrate resources and policies across systems.
Targeted Housing and Services. Supportive housing permanent housing linked to individualized supportive
services is enhanced with targeted and assertive recruitment through in-reach into jails, shelters,
hospitals and other settings to help clients obtain housing stability and avoid returning to costly crisis
services and institutions
NYC FUSE I. The first FUSE project was in New York City. In 2006, a Discharge Planning Collaboration
(the Collaboration) was formed that included staff from CSH, service providers and advocates concerned
with the reentry population, and administrators from city agencies including the NYC Department of
Corrections (DOC), the NYC Housing Authority (NYCHA), the NYC Department of Health and Mental
Hygiene (DOHMH), and the NYC Human Resources Administration (HRA). Reentry issues were of
growing concern. Of the roughly 350 former inmates released in NYC every day, over 40% released from jail
were re-incarcerated within 12 months. The Collaboration developed a pilot housing program that targeted
high-needs individuals with multiple incarcerations in Riker’s Island, the City’s jail. Using a data match
between jail and public shelter records, the Collaboration identified individuals who had at least four jail stays
and four stays in the city’s homeless shelter system over the prior five-years and used this “4-4-5 rule” to
determine threshold eligibility for the program (CSH, 2009a; Fontaine, Roman & Burt, 2010).
Working with eight community-based housing and service providers, this initial FUSE project placed 100
formerly incarcerated ‘frequent users’ into permanent supportive housing to try to improve reentry outcomes
and break peoples’ cycling between jail, shelter, emergency health and other public systems. Housing
resources included 50 Section 8 Housing Choice Vouchers and 50 supportive housing placements within
3
single-site developments. Based on a commitment to stabilization and support services, the NYC Housing
Authority created a specialized admission and review process for FUSE tenants that waived the non-violent
and drug-related criminal justice exclusions which are typical barriers for tenants matching the FUSE profile.
Additional vouchers were provided by the NYC Department of Housing Preservation and Development,
which used the minimum for criminal justice exclusions set by the U. S. Department of Housing and Urban
Development. In addition, each housing provider was awarded a one-time funding of $6,500 per tenant to
enhance its customary care services. This allowed providers to actively recruit and assist FUSE clients with
their application and access to supportive housing and to deliver additional acclimatization and stabilization
supports and assistance during their clients’ first year living in the provided housing.
The John Jay College Research and Evaluation Center (at the John Jay College of Criminal Justice/City
University of New York) evaluated this first FUSE program (known as FUSE I). The center used a quasi-
experimental design, creating a comparison group by using jail and shelter administrative data to select
individuals meeting the 4-4-5 criteria (four shelter entries and four jail entries in previous five years) and
“matching” them to program participants on demographics and mental health diagnosis. The John Jay
evaluation of the first year following placement found over 90% housing retention, a 92% reduction of
shelter days and a 53% reduction of jail days; rates of stable housing and avoiding jail were much lower
among comparison group members. The reduced rate of cycling between jail and public shelter indicated cost
offsets to those systems of approximately $3,000 per person, per year (CSH, 2011). These promising
evaluation results led to an expanded program, FUSE II.
NYC FUSE II. CSH continued to work with city agencies and community providers to further develop the
NYC FUSE program. In 2008, it obtained commitments to support an additional 100 units of housing and
enhanced services for FUSE participants. NYCHA and HPD provided 101 units and set aside units from
DOHMH funded supportive housing. The eligibility criteria of four jail and four shelter stays over the last
five years was maintained, determined by DOC and DHS administrative data match. Every quarter, these
agencies generated a replenishing list of approximately 850-1,100 individuals meeting these criteria, and CSH
cross-referenced this list with current jail and shelter census information to locate potential FUSE
participants for program outreach.
The public-private collaborations integral to FUSE I grew stronger in FUSE II. CSH continued its facilitative
role to secure resources, provide training and technical assistance to housing providers and oversee program
implementation. The NYC Housing Authority provided what were referred to as “quasi sponsor-based
Housing Choice Vouchers. This was conceived as a pilot effort using a rider to tie tenant-based vouchers to a
service provider and was one of the first efforts in the country to develop a sponsor-based approach with
Housing Choice Vouchers. These vouchers were classified as tenant-based vouchers, but the tenants accessed
housing through master-lease agreements with the service providers. Similar efforts have been made by some
housing authorities that have more flexible administrative rules than the NYC Housing Authority. In
addition, the NYC Department of Housing Preservation and Development provided tenant-based vouchers,
and the NYC Department of Health and Mental Hygiene funded set-aside units in supportive housing
buildings. Service resources for tenants were provided through contracts from the NY/NY III and High
Service Needs supportive housing production programs. Six of the community-based housing providers who
had participated in the original FUSE initiative continued to recruit, house and provide services to reentry
clients with complex histories of incarceration and homelessness.
4
For FUSE II participants, housing is permanent, not transitional. Units are subsidized through Section 8,
OMH or DOHMH, such that the rent a tenant pays is no more than 30% of income or of housing allowance
from benefits. DHS has formal authority over the FUSE project, linking the target population to permanent
supportive housing. As in FUSE I, housing providers were given a one-time $6,500 payment per client to
allow for flexible service funding during the critical time period from recruitment and engagement to linkage
with sustainable, comprehensive medical and mental health services and other support services needed to
promote stability and tenant success. Uses of the enhancement varied by housing program and included
clinical supervision; client recruitment and engagement; intensive case management with lower client-to-case
manager ratios; special FUSE service staff to provide more intensive support during the first year of housing;
and/or additional specialty services as needed. Table 1 on the next page presents a snapshot of NYC FUSE
II providers’ housing and service delivery models. Additional descriptions of specific housing and service
characteristics and funding sources used by the different agencies serving FUSE II clients can be found in
Section IV, Cost Evaluation.
5
Table 1. Snapshot of NYC FUSE II Providers’ Housing and Service Delivery Models
Agency
Target
Populationa
Type of
Housing
Funding for
Housing
Case Management
Service Model
Brooklyn
Community
Housing
Services
(BCHS)
SPMIb
community
care
SRO units
(one site)
Project-based
Section 8
Comprehensive Service Model
Therapeutic Case Management
Harm Reduction
Peer Support
CAMBA
SPMIb
community
care
SRO units
(two sites)
Shelter+Care
Project-based
Section 8
Assertive Case Management
Therapeutic Case Management
Interdisciplinary Teams
Common
Ground
NY/NY III
F (recent
AOD
treatment)c
Scattered-site
apartments
NYCHA
quasi
sponsor-
based
Section 8d
Comprehensive Service Model
Strength Based Case
Management
Harm Reduction
Jericho
Project
SPMIb
community
care
SRO units
(five sites)
Project-based
Section 8
Comprehensive Service Model
Palladia, Inc.
NY/NY III
F (recent
AOD
treatment)c
Scattered-site
apartments
NYCHA
quasi
sponsor-
based
Section 8d
Service Brokering
Comprehensive Service Model
Strength Based Case
Management
Interdisciplinary Teams
Harm Reduction
Peer Support
Pathways to
Housing
Axis I
diagnosis
community
care
Scattered-site
apartments
HPD tenant-
based
Section 8
Comprehensive Service Model
Assertive Community Treatment
Interdisciplinary Teams
Pathways Housing First Model
Source: Interviews with project staff and review of program documents
a All programs target recently incarcerated single adults with multiple episodes of homelessness and jail experience.
b SPMI programs are for people with serious and persistent mental illness.
c NY/NY agreements are between the New York City and New York State to provide funding to nonprofit providers
and developers to create supportive housing for homeless people with mental illness and other disabilities. “Category F”
is for homeless single adults who have completed substance abuse treatment.
d Scatter-site sponsor-based Section 8 apartment leases are held by the agencies, who enter into occupancy agreements
with residents.
6
II. EVALUATION QUESTIONS AND METHODS
The FUSE II evaluation reported here was conducted by researchers from Columbia University and Shubert
Botein Policy Associates. In this section, we report the questions that drove the evaluation and the
methodology used to answer those questions. In the following Section III, we report findings regarding
point-in-time and time-aggregated outcomes, and investigate time-patterned outcomes for jail and shelter use.
Section IV of this report examines FUSE II intervention costs and the results of a cost-offset analysis; the
final section summarizes the report and points to policy implications.
A. Evaluation Questions
The evaluation was designed to measure the impact of the second phase of the FUSE initiative on a number
of important outcomes. Specifically, we ask whether or not the intervention positively changed clients’ lives
with regard to their:
avoiding homelessness and retaining housing,
criminal justice involvement, including arrests and returns to jail or prison,
health and mental health and health services utilization,
using hard drugs, problem drinking and engaging in similar high-risk behaviors,
connecting with family and having other forms of social support,
over all temporal patterns of institutional involvement beyond their using individual public systems,
i.e., reduced cycling between institutions.
In addition, we analyze the cost of the FUSE intervention and possible cost offsets from reducing public
expenditures associated with use of shelter, medical and criminal justice systems.
B. Research Design
Our basic study design is a two-group pre/post design with a comparison group constructed among FUSE
II-eligible individuals who strongly match those receiving the intervention. We are interested in estimating the
effects of the FUSE II intervention or “treatment” on those who received the intervention. We do not
estimate effects on the broader population of those who meet program criteria.
To allow enough time to test the program’s effectiveness, we followed participants in both intervention and
comparison groups for up to 24 months after baseline data collection. Data sources included survey
interviews as well as administrative data from the NYC Departments of Corrections and Homeless Services.
Using an extensive questionnaire, we interviewed study participants at baseline, six, twelve, eighteen and
twenty-four months.
2
For the intervention group, we conducted the baseline assessment immediately after
they moved into FUSE II housing; for the comparison group, we conducted the baseline interview at study
enrollment, which was timed to coordinate with intervention group assessments.
2
Not all study participants were interviewed at each of the time periods. Some completed their final interview more than
24 months after baseline, and some were lost to follow-up. See Table 4 for relevant response rates and numbers.
8
An administrative data match provided information on jail and homeless shelter experience for five years
prior to and two years following enrollment in FUSE II, or, for comparison group members, following
baseline assessment. In our analysis, individuals were analyzed as FUSE II intervention group members,
regardless of whether they maintained FUSE II housing or otherwise continued to be part of the FUSE II
program.
C. Comparison Group Formation
Since FUSE II was implementing an already established protocol with a complex process to determine
eligibility and enroll to individuals into the program, a random assignment design with people randomly
assigned to a control group was not feasible. This necessitated our forming a comparison group to address
possible confounders of any intervention effects. Such a comparison group improves our ability to ascertain
if the program caused the result that we see in jail, shelter, health or other outcomes, or if something about
the individuals in FUSE II caused such effects. For instance, perhaps persons in the program were better off
in some way (e.g., higher functioning, more motivated to change, and so forth) than people who weren’t in
the program. As a result, they may have avoided jail or scored better on outcome measures regardless of their
participation in FUSE II. On the other hand, perhaps FUSE II participants were worse off, struggling with
mental health needs that were bound to improve just with the passing of time, again, regardless of their
participation in FUSE II. Thus, to determine the effect of FUSE II, we need to answer: What would have
happened to the people who received the intervention if they had not received the intervention?
Comparison group recruitment. Our recruitment strategy was to mimic as much as possible the strategy
and tactics program agencies used to locate and recruit persons eligible for FUSE II. Thus, the first step in
our strategy for forming this group was to work with the housing provider agencies to determine how they
selected individuals for their programs from among the larger FUSE II-eligible population. Using
information on client selection processes and from the monthly list of FUSE II-eligible people in DHS
homeless shelters, our field staff visited shelters where FUSE II-eligible persons lived to identify potential
study participants who met the 4-4-5 criteria.
To follow as closely as possible agency recruitment efforts, our staff went to the same shelters from which
the programs recruited. While there, they used a questionnaire that covered topics the service providers were
using to assess suitability for their specific housing programs. To be more specific: Informed by eligibility
criteria used by programs that targeted services for persons with a serious persistent mental health diagnosis
or with substance use histories who engaged in or had recently completed a successful course of addiction
treatment, the screening questionnaire included questions on these topics (see Appendix D for Screening
Questionnaire). Thus, to be eligible for the comparison group, in addition to being on the DOC-DHS match
list and meeting the 4-4-5 criteria, people had to meet either additional criteria (A) or (B):
A. They had to have been in been in drug or alcohol treatment in the twelve months prior to the
administration of the screening survey and report not drinking alcohol to the point of intoxication or
using cocaine, crack or heroin in the prior 45 days. They also had to answer “definitely willing” or
“possibly willing to one of the following: In order to get housing, would you be willing to (1)
completely quit using drugs, (2) go to an outpatient substance use program where you would go every
day for counseling and treatment, (3) attend a support group related to alcohol or drug use, or (4) go to
individual alcohol or drug counseling or therapy for alcohol and substance use.
9
B. Potential study participants also had to report if they had ever been diagnosed with a psychiatric
condition, or had mental health treatment or talked to a mental health specialist in the twelve months
prior to administration of the screening questionnaire.
Not all programs used the same criteria. To create a comparison group that was as equivalent as possible to
the intervention group, we utilized enrollment criteria common to all programs. As a proof of concept that
this approach was appropriate, we note that 12 of the people we identified for the comparison group but
before we formally included them in the study were subsequently accepted into the FUSE II intervention.
Using this multi-layered process, we selected individuals for comparison group membership who closely
matched those chosen by housing providers for the FUSE II intervention. By the time the program
admissions window closed in March of 2010, these providers had identified 72 people for FUSE II. These
individuals comprise the intervention or treatment group for the evaluation. During the same time period,
using the procedures just described, we identified 89 persons for the comparison group. We selected a larger
number for the comparison group because we anticipated a larger program population and because we
wanted a larger group from which to select to carry out the second step in forming our comparison group.
Propensity score analysis. Our second step in forming the comparison group was to use propensity score
matching to improve the comparability of the intervention and comparison groups. This technique allows us
to estimate a “score” which represents the probability (i.e., “propensity”) of being selected for the program
for people in both the potential comparison group and the actual intervention group. The basic idea of this
method is to estimate a score based on a model that incorporates pre-intervention demographic, clinical,
experiential and service use variables thought to affect people’s chances of being selected for the program
and/or thought to affect outcomes. We use this score to make the comparison group more similar to the
intervention group than it otherwise would be and, thus, minimize pre-intervention group differences across
relevant characteristics. (See Figure 2 for a diagram of this logic.)
Cases are excluded at both ends of the propensity score distribution to improve
the match between program participants and comparison group members.
Figure 2. Using Propensity Scores to Identify Comparable Cases
10
Table 2 reports the number of cases for each study group resulting from steps one (initial recruitment of
comparison group members) and two (selecting cases using propensity scores). For further details of the
propensity score analysis, see Appendix A.
The propensity score analysis indicates that, in following the same criteria and procedures as the programs to
identify the comparison group, we did a fairly good job. Sixty of 72 people in the intervention group had
propensity scores that overlapped with those of 70 people in the potential comparison group. Using the
constructions of the intervention and comparison groups resulting from this second step, we tested the
results by checking how well the intervention and comparison groups were balanced” on the variables
(“covars”) used to estimate the propensity score. For each covar, the difference between the mean of the
comparison group and the mean of the intervention group is assessed for bias and statistical significance. We
found that once we trimmed the intervention and comparison group cases to eliminate those with no overlap
in propensity scores, the two “trimmed” groups are fairly balanced without further propensity score
adjustment. Appendix A, Table A-2 contains the results of the balance analysis for all the variables that were
initially thought to possibly affect selection into the intervention and/or intervention outcomes. Note in this
table that a range of measures of prior jail experience were examined during the propensity analysis; all
remain balanced in the trimmed sample used for the outcomes analyses.
In Table 3 (p. 12), we report the balance analysis for a subset of the covariates that were statistically
significant in the model used to estimate the propensity score and for the one covar (never psychiatric
diagnosis) whose mean difference between the comparison and intervention groups remains statistically
significant in the trimmed sample. The table shows no statistically significant differences between the
intervention and comparison group members in the trimmed sample except ‘never had mental health
diagnosis; of the other measures, only ‘no close friends or family contacts’ is even marginally significant. For
example, the mean number of shelter admissions over the 24 month period prior to baseline interview is 2.5
for the intervention group and 2.3 for the comparison group; 22% of intervention group members had been
homeless for five years or more over their lifetime as were 23% of comparison group members.
As another measure of balance, Table 3 also shows the biasstatistic. This is the difference of the means
expressed as a percentage of the square root of the average of the sample variances in the intervention and
comparison groups. The lower the percentage, the less the two groups differ (Rosenbaum & Rubin, 1985). As
Table 3 shows, almost all variables have about a 20% bias or less; the mean and median bias for the data with
trimmed cases is 13.4 and 10.2, respectively, indicating very good balance. In our outcomes analyses, we
adjust for potential residual bias by further covariate adjustment through regression modeling that includes
Table 2. Size of Intervention and
Comparison Groups for Each
Selection Step
Selection Step
Groups
First
Second
Intervention
72
60
Comparison
89
70
11
the variables listed in Table 3. That is, all analyses control for ever psychiatric diagnosis, drug use, history of
homelessness, education and the other variables shown in Table 3.
For all outcome analyses, we use the trimmed sample or a subset of it. However, we use the full intervention
sample of 72 FUSE II participants when we provide descriptive statistics for those the intervention served
(e.g., on average, how long those who received housing stayed housed and similar statistics). In Appendix B,
we show how the intervention participants excluded in the trimmed sample differ from and are similar to the
60 FUSE II intervention group members used in the outcome analyses.
D. Data Collection
We use two data sources to carry out the evaluation. One is an extensive survey of comparison and
intervention group participants based on in-person interviews. (See Appendix E for the Baseline
Questionnaire and Appendix C for a concordance listing conceptual variables, specific measures and sources
for standardized measures used in the questionnaire.) The FUSE II interview includes original items
developed specifically for this evaluation as well as standardized measures and validated assessment tools
measuring:
demographics,
current and recent housing and living arrangements,
residential history for the five years prior to baseline,
health conditions and health functioning,
mental health diagnoses and mental health functioning,
alcohol and substance use,
health, mental health and substance abuse services,
social networks and social support,
need for and use of case management and social services.
We administered the surveys at baseline and at roughly six month intervals over two years, for a total of five
waves of data collection. Table 4 reports the number and percentages of completed surveys at each wave.
Mean number of follow-up interviews was 3.1. An additional 52 interviews were completed beyond the 24
month time period that frames the current study and therefore are not included. Information from these
additional surveys will be available for future analyses.
The second dataset resulted from a data match of administrative records from the Departments of
Corrections and Homeless Services of the City of New York. Measures included dates of admission and
discharge into jail or shelter, length of residence or incarceration, location of facility and, for those
incarcerated, the arrest charge(s).
12
Table 3. Balance of Covariates for Trimmed Data: Covars in the Propensity Score Model
and Covars with Statistically Significant Differences
Covariates
Intervention
Group
Means/
Proportionsa
(n=60)
Comparison
Group
Means/
Proportionsa
(n=70)
% Bias
t-score
p
Number shelter admissions over 24
months before program enrollment
2.47
2.29
7.5
0.420
0.674
Life time homelessness > 5 yearsb
0.22
0.23
-2.8
-0.160
0.872
Veteran
0.03
0.07
-17.0
-0.960
0.341
Physically Disabled
0.58
0.53
11.0
0.620
0.535
Current employment incomec
0.23
0.30
-15.0
-0.850
0.397
Current income from public assistanced
0.63
0.67
-7.9
-0.450
0.652
Didn’t graduate high school
0.40
0.37
5.8
0.330
0.741
Graduated high school/GED
0.48
0.44
8.1
0.460
0.648
Reported health fair or poor
0.32
0.27
9.9
0.560
0.575
Age at first sexual relations with
opposite sex
14.4
14.0
10.2
0.570
0.568
Never psychiatric diagnosise
0.37
0.19
41.0
2.350
0.020
*
Mental health services past 6 monthsf
0.45
0.54
-18.5
-1.050
0.295
Never used hard drugsg
0.17
0.26
-22.1
-1.250
0.214
Past use hard drugsh
0.52
0.41
20.5
1.160
0.246
No close friends or family contactsi
0.03
0.13
-35.2
-1.960
0.052
* p < .05
a Values shown are means for continuous variables or proportions for the one category of dichotomous variables
shown in the table, e.g., 0.22 or 22% of the intervention group had five or more years homeless prior to baseline
interview.
b Self-report of lifetime street or shelter homeless experience since age 18.
c Includes pay for odd jobs, occasional or temporary part-time work (irregular hours).
d Income from SSI, SSDI, TANF, VA or PA/TA (New York State temporary safety net assistance for individuals).
e Self-report never diagnosed with psychiatric disorder, received medications or hospitalized for mental health
problems.
f Received treatment or therapy from mental health professional or supportive counseling six months prior to
baseline.
g Never used cocaine, crack, heroin, or methamphetamine.
h Ever used cocaine, crack, heroin and/or methamphetamine but not within six months of baseline interview.
i No close friends who are not relatives or adult relatives seen at least occasionally or speak to on the phone.
13
The result of this data collection is an extensive set of information that will help us understand the ‘frequent
user’ population and effects of the FUSE II intervention. In addition, availability of a wide range of
information about participants allows us to feel confident in the propensity score analysis, since that method
assumes all relevant variables have been measured, and to include in the outcome analyses variables that
might be thought to affect outcomes independent of the intervention.
Table 4. Completed Surveys for Each Interview Wave
Intervention Group
Comparison Group
Interview Wave Number
N
%a
Mean
months
from
baseline
N
%a
Mean
months
from
baseline
One (baseline)
72
na
na
89
na
na
Two (month 6)
68/69
98.6%
6.5
61/86
70.9%
6.8
Three (month 12)
64/67
95.5%
13.0
56/61
91.8%
13.9
Four (month 18)
58/64
90.6%
19.9
37/56
66.1%
19.3
Five (month 24)
40/56
75.0%
26.2
28/35
80.0%
25.5
Any follow-up interview
68
94.4%
na
66
74.2%
na
a Percent of sample eligible to be interviewed at each wave.
E. Description of Frequent Users
Before discussing outcome effects associated with the FUSE II intervention, we describe the FUSE II sample
of ‘frequent users’ adults with multiple experiences of jail and homeless shelter admission. Table 5 shows a
range of demographic, clinical, service need and service utilization characteristics, as well as pre-baseline
histories of jail and shelter experience. Selected here are characteristics and experiences that other research
has shown to be associated with poor outcomes and recidivism among persons leaving jails (Andrews &
Bonta, 1995; Vera Institute, 2012) and factors increasing risk for homelessness among low income
populations (for review see Apicello, 2010). Table 5 presents means and proportions for the trimmed sample.
Frequent users in both the intervention and comparison group are overwhelmingly male and predominantly
African American or Latino. A substantial proportion do not have a high school diploma or GED. Three-
fourths have had a history of regular full-time employment but current rates of disability are high. Extreme
poverty is the norm. For the great majority, yearly income from all sources is less than $7,500. More than half
are food insecure. Most frequent users of jail and shelters have very limited social networks: Close to 80%
have never been married, the median number of family members with whom they have any contact, see
occasionally or speak with on the phone is two.
Scores on a summary measure of social support (adequacy of emotional, instrumental or informational
support) are low, similar to results for this population when compared to general samples of adults (Messeri,
et al., 1993). The research literature on recidivism and on substance abuse relapse suggests that increased self-
efficacy and positive coping skills predict better outcomes for an individual. We have categorized these
14
measures as ‘dispositions.’ We included these measures in the questionnaire to examine possible differences
in pre-intervention self-motivation to change indicated by positive coping skills compared to ‘emotion
focused coping’ associated with drug and alcohol use and other less effective responses to life challenges.
Substance use is almost universal, and rates of past abuse are high. Almost all (over 90%) report illicit drug
use, most having a history of ‘hard drug’ use, i.e., using heroin, cocaine, crack or, less often,
methamphetamine, and doing so weekly or more often for one or more periods in their lives. For about one-
third the sample, serious addiction challenges continue; others have benefited from treatment or otherwise
reduced or stopped using drugs, other than marijuana.
As research has shown, persons with multiple jail stays and those with multiple homeless shelter stays have
high rates of physical as well as mental health problems (CSH, 2009a). About 70% of program participants
have one or more serious chronic health conditions, including hypertension, cardiovascular disease, asthma,
diabetes, hepatitis and/or epilepsy. (Rates of HIV infection are also high among incarcerated populations, but
in New York City most homeless persons known to be HIV positive are served by a separate system of
AIDS housing resources and service agencies. Thus, very few persons diagnosed with HIV are included in
the FUSE II eligible sample.)
Serious and persistent mental illness characterizes FUSE eligible persons found in jails or shelters. Specific
diagnoses include schizophrenia, bipolar disorder, major depression and post traumatic stress disorder. Study
participants score high on a standardized measure of psychological stress; this is true even for those whose
symptoms do not meet threshold criteria for diagnosis of anxiety disorder.
One characteristic that may distinguish frequent users of jail and homeless shelters from the general jail
inmate population is early exposure to trauma and violence and loss or separation from parents (BJS, 2004;
McDonnell, et al., 2011). Over two-thirds report traumatic or highly stressful events during childhood or
adolescent including physical assault and sexual assault. About half have been victims of or witnessed other
family members violently victimized. More than one in five spent time in foster care.
Another finding regarding early experience is that for 80% of the sample, their first episode of incarceration
preceded their first episode of homelessness. Based on narrative descriptions of reasons for homeless
experience, incarceration was for many a major cause or trigger for housing loss. About one in five study
participants had experienced both incarceration and street or shelter homelessness prior to age 25 years
(Bozack, 2010). Such a lack of family resources increases risk for homelessness among low income persons
generally, especially those with behavioral health issues. The intersection of early exposure to violence, lack of
family/kin supports, jail and shelter experience is worth more investigation.
Regarding the criminal justice profile of frequent users, three-fourths have been incarcerated for drug related
charges, overwhelmingly for possession. However, repeated incarcerations are more often associated with
low-level misdemeanors such as shoplifting or “theft of services” (mostly jumping the turnstile for public
transit access), “quality of life” offenses (vagrancy, trespassing, loitering, disorderly conduct, public urination),
and probation or parole violations rather than with repeated drug convictions. This highlights the need to
understand better how structural factors such as local laws and police practice interact with individual mental
health, addiction or other vulnerabilities to increase the risk for re-incarceration among frequent users of jail
and shelter.
15
Table 5. Incarceration, Homelessness, Sociodemographics and Select Clinical and Attitudinal
Characteristics of FUSE II Evaluation Study Participants
Study Participants' Characteristics
Intervention Group
Mean or
Proportiona
(n=60)
Comparison Group
Mean or
Proportiona
(n=70)
Criminal Justice History
Age at first arrest
21.0
22.6
Number jail admissions over 6 months before enrollmentb
2.47
2.97
Number nights in jail over 24 months before enrollmentb
68.9
79.7
Homeless History
Life time homelessness > 5 yearsc
47%
49%
Number shelter admissions over 6 months before enrollmentb
55%
54%
Number nights in shelter over 24 months before enrollmentb
245.5
208.7
Demographics
Current Age
46.0
44.3
Male
88%
87%
Race/ethnicity : Black
58%
66%
Race/ethnicity: Hispanic
22%
23%
Education/Employment/Income
Graduated high school/GED
48%
44%
Ever had full-time job for a year or more
75%
67%
Current income from employmentd
23%
30%
Income from all sources < $7,500 per yr.
75%
61%
Family/Marital/Social Support
Ever placed in foster care or group home
22%
23%
Ever married
23%
19%
No close friends or family contactse
3%
13%
Social support summary scoref
23.7
19.9
16
Table 5. Incarceration, Homelessness, Sociodemographics and Select Clinical and Attitudinal
Characteristics of FUSE II Evaluation Study Participants (cont’d)
Study Participants' Characteristics
Intervention
Group Mean or
Proportiona
(n=60)
Comparison
Group Mean or
Proportiona
(n=70)
Substance Use
Never used hard drugsg
17%
26%
Past use hard drugsh
52%
41%
Problem alcohol usei
37%
34%
Substance abuse services past 6 monthsj
53%
53%
Mental Health
Ever psychiatric diagnosisk
63%
81%*
Mental health services past 6 monthsl
45%
54%
Psychological stress scorem
8.3
7.3
Physical Health
Health rated fair or poor
32%
27%
Number of chronic or infectious illnesses ever diagnosedn
1.4
1.4
Attitudes/Dispositions
Religion or spirituality somewhat or very important
76%
87%
Mastery index (self-efficacy)o
17.2
16.3
Coping: Take action to try to make the problem betterp
78%
67%
Coping: Get help/advice from othersq
63%
53%
Coping: Try to come up with strategyr
75%
71%
Substance abuse treatment readiness scores
35.0
35.8
* p < .05
a Values shown are means for continuous variables or, for dichotomous variables, the percentage with the characteristic.
b During the time period prior to FUSE program enrollment, or for comparison group, prior to baseline interview.
c Self-report of lifetime street or shelter homeless experience since age 18.
d Any income from paid work.
e No close friends who are not relatives or adult relatives seen at least occasionally or speak to on the phone.
f Summary measure of degree and number of people who can be counted on for support in different situations.
(Adapted from Sherbourne & Stewart, 1991). Range is 0-48; higher score indicates more support.
g Never used cocaine, crack, heroin, or methamphetamine.
h Ever used cocaine, crack, heroin and/or methamphetamine but not within six months of baseline interview.
i Positive screen for alcohol abuse or dependence based on Client Diagnostic Questionnaire (CDQ; Aidala, et al., 2002).
j Alcohol or drug abuse treatment or services anytime during six months prior to baseline interview.
17
Table 5. Incarceration, Homelessness, Sociodemographics and Select Clinical and Attitudinal
Characteristics of FUSE II Evaluation Study Participants (cont’d)
k Self-report ever diagnosed with psychiatric disorder, or received medications or hospitalized for mental health
problems.
l Received treatment or therapy from mental health professional or supportive counseling for emotional or psychological
difficulties at any time within six months of baseline interview.
m Perceived Stress Scale (Cohen, et al., 1983). Range is 0-20; higher score indicates more stress.
n Self-report medical provider has diagnosed with asthma, hypertension, diabetes, high cholesterol, heart attack or
stroke, cancer, seizure disorder, sickle cell anemia. Includes four persons with only STIs such as herpes or gonorrhea.
o Mastery/Locus of Control (Pearlin, et al., 1981). Range is 7-28, higher score indicates greater self-efficacy, sense of
control.
p Coping in response to difficult or stressful events: do this medium amount or a lot. (Adapted from Carver, et al., 1989).
q Readiness and Treatment Eagerness Scale (SOCRATES; Miller & Tonigan, 1996). Range is 10-50; higher score
indicates greater readiness.
18
III. OUTCOME ANALYSES
In this section, we describe basic findings for, first, those who received FUSE II housing and services and,
then, the effects of the program on the FUSE II intervention group relative to the comparison group. The
first analyses are simple descriptions of those receiving housing and services; the second are the results of
regression analyses. In the latter, the different outcome variables of interest are regressed on the covariates in
Table 3 and the outcome variable measured at baseline (e.g., analyzing intervention effects on mental health
functioning at follow-up, controlling for baseline mental health functioning score, as well as the Table 3
covariates). Thus, this modeling assumes linear effects of the covars on differences. In some instance, this
assumption is a function of our measuring outcome variables dichotomously or collapsing them into
dichotomies. For continuous variables, our theoretical assumptions were always for linear effects.
In addition to regression analyses of differences between comparison and intervention groups, we also show
results from a “trajectory analysis to test for the effects on temporal patterns of jail and shelter use. For this
analysis, DOC and DHS administrative data is examined using optimal matching to look at sequences of jail
and/or shelter use. This approach allows us to compare post-intervention incarceration and homeless shelter
trajectories of the two groups, thus showing differences between them not only at points-in-time (e.g., at the
end of 24 months) or aggregated over time as in the regression analyses, but in the patterning of incarceration
and homeless shelter use over the post-intervention period.
A. Housing Outcomes
A primary thrust of FUSE II is to help participants achieve housing success and community reintegration
after leaving jail. Specifically, FUSE II-provided housing is the primary “active ingredient” of the initiative.
The argument is that because people have stable and appropriate housing, the kinds of problems that
characterized their lives prior to FUSE II repeated episodes of incarceration, shelter use, emergency
hospitalizations, and problems associated with mental health symptoms and/or addiction would be
reduced. Housing is the central focus of the program’s attempt to improve people’s lives more generally. In
this section, we examine the question: Did frequent users placed in FUSE II housing keep their housing? We
then report effects of the intervention on housing status by comparing results for the intervention group with
the comparison group.
FUSE II participants’ housing retention
.
Table 6 reports rates of housing success or how well FUSE II
participants were able to maintain their housing over the follow-up period. It reports housing at the 12th and
24th months after participants moved into FUSE II-provided permanent supportive housing. Because we are
not, in this analysis, comparing those receiving the intervention with those who did not, we use the full
sample of FUSE participants (72 people), and we use survey data collected over the entire follow-up period.
As Table 6 shows, of the 69 people who received FUSE II housing and services and were not deceased at the
12th month of follow-up, 89.9% were in FUSE II-provided housing at that point-in-time. The comparable
statistic for 24 months is 80.9%. Obviously, these statistics indicate a very small number of people failed to
maintain their FUSE II-housing over the relevant time periods.
20
Table 6. Percent FUSE II Participants Housed in FUSE II-provided Housing
at and over 12 and 24 Months of Follow-up Period
Kinds of Retention in FUSE II-provided Housing for 12 & 12 months
%
N
Housed in FUSE II housing at 12 months
89.9%
69a
Housed in FUSE II housing at 24 months
80.9%
68b
Housed continuously in FUSE II housing over 24 months
47.1%
68b
FUSE II housing continuously or with brief interruption over 24 months
80.9%
68b
a Three participants died over first 12 months of follow-up.
b Four participants died over 24 months of follow-up.
We also looked at whether or not people maintained residence in their FUSE II housing continuously over
the entire follow-up period, or had brief periods when they were staying in other situations. Housed
continuously means that FUSE II housing remained people’s home address throughout the period, did not
enter jail or a homeless shelter for even one night, and were not hospitalized or in a residential treatment
facility for more than 90 days. And housed with brief interruption means they were in one of these
institutional settings during the follow-up time period (in jail or shelter one or more nights or in a health or
other residential treatment facility for more than 90 days) but maintained tenancy and came back to FUSE II-
housing after these institutional or treatment experiences.
The results show that about half the program participants remained continuously in their FUSE II housing.
Over 80% maintained residency with no days away from their FUSE II residence or had only limited
interruptions for a brief jail stay or treatment episode. Rates of 12-month and 24-month success in
maintaining housing are higher than seen in other supportive housing interventions for persons with complex
histories of homelessness and behavioral health needs. Usually, retention in housing over 24 months seldom
exceeds 75% (Malone 2009; Martinez & Burt 2006; Wong 2006). Our findings indicate that whatever issues
arise for FUSE II participants, they tend to return to FUSE II housing, much in the way people ordinarily do
in their lives after stints in hospital or other such settings.
Intervention effects on housing status
.
Here we analyze whether or not FUSE II had its intended housing
effects by comparing the intervention and comparison groups. In this and subsequent regression analyses, the
different outcome variables of interest
3
are regressed on the covariates in Table 3 and on the outcome
variable measured at baseline. Thus, this modeling assumes linear effects of the covars on differences
between the two groups. In some instance, this assumption is a function of our measuring outcome variables
dichotomously or collapsing them into dichotomies.
4
For continuous variables, our theoretical assumptions
were always for linear effects.
3
Outcome variables are measured either continuously or dichotomously; thus, in this analysis, we can measure the mean
as the outcome of interest for all comparisons.
4
For all collapsed variables, we expected them to be affected linearly in their initial measurement, thus collapsing did no
harm to our expectation of linear effects. That is, we expected those in the intervention group to score lower (or higher),
relative to the comparison group, on all categories of variables with three or more categories.
21
Table 7 presents results for effects on housing.
5
It compares whether or not members of each group were
housed in permanent housing at the 12th and 24th months after they were placed in housing (intervention
group) or after they were first interviewed for this study (comparison group). Note that FUSE II participants
could be in permanent housing provided by another housing program or in community housing, not
necessarily the FUSE II housing into which they were placed.
The estimates in the table (also shown in Figure 3, next page) suggest extremely strong support for the effect
of FUSE II on obtaining and maintaining permanent housing among program participants. At twelve
months, over 91% of FUSE II participants are housed in permanent housing, compared to the 28% that
would have been housed had they not received FUSE II housing and services.
*** p < .001
By 24 months, this 63% difference has dropped to 43%, mostly because comparison group members
obtained housing (this has increased to 42%) rather than FUSE II participants’ housing situation having
changed (a slight drop to 86%). As we discuss later in Section V, the high rate of housing placement among
the comparison group could be due to the NY/NY III initiative, the largest offering of supportive housing in
New York City at the time, since units were available at the same time as FUSE II was initiated (New York
State, 2005). Nonetheless, the intervention results are highly statistically significant. The small change in the
FUSE II participants housing situation speaks well for the lasting effects of the program. We conducted this
study over 24 months to more strongly test whether or not FUSE II’s effects would last past the more
common one year follow-up period. This analysis indicates that they did, suggesting that it is likely that FUSE
II-induced effects will be sustained past this study’s two year follow-up period.
5
For this analysis, we used both survey and administrative data, hence the Ns are different from those in previous and
subsequent analyses. Also, the administrative exit reason for five comparison group members indicated the person was
leaving for housing at a point three months prior to their 12th or 24th month over follow-up and had no jail or shelter
experience after that exit. In these instances, the person was coded as housed. We did this to avoid a missing data code
for these people. Note that, from the perspective of testing FUSE II, this coding makes it harder to find significant
differences between the intervention and comparison groups.
Table 7. Intervention Effects for Housing
Permanent Housing Measures
Intervention
Group
Comparison
Group
Difference of
Means
95%
Confidence
Interval for
Difference of
Means
%
N
%
N
In permanent housing at 12 months
91.2%
57
28.3%
53
62.9%***
56.5%
69.3%
In permanent housing at 24 months
85.5%
55
42.2%
45
43.2%***
33.9%
52.6%
22
Intervention effects on homeless shelter use. Here we report results for shelter use outcomes, using
administrative data maintained by the Department of Homeless Services. The analysis evaluates program
effects from two time points: program enrollment and housing placement. Program enrollment is the point at
which one of the agencies formally accepted the person into the program. At this point, the person had yet to
complete application procedures, be judged eligible for housing assistance by City agencies and, for scatter-
site programs, locate an apartment with a landlord willing to accept FUSE II clients. This process could take
some time. As a result, the elapsed time between when the person was enrolled in the program and when he
or she was placed into permanent housing varies across individuals, with a minimum of 11 days and a
maximum of 20 months. In addition to procedural and landlord factors, these differences were caused by
client issues (e.g., acquiring appropriate identification documents), unexpected system or agency challenges
(e.g., government funding cuts, agency staff changes), or both. The average time elapsed was 180 days. Note
that for comparison group members, the date of ‘enrollment’ and ‘placement’ is the date of enrollment into
the study, indicated by completion of the baseline interview.
On the next page, Table 8 reports outcomes for shelter use for the following outcomes:
Number of days in shelter, number of episodes and percent having any episode over 24 months of
follow-up from enrollment or placement. Note: Due to Department of Homeless Services
procedures, people had to leave shelter for more than 30 days for a shelter episode to conclude.
Number of days and episodes and percent having any episode over the last three months of the 24
month follow-up period from enrollment or placement.
We show these particular outcomes to give a sense of how the program performed over the entire
follow-up period and to observe how people were performing at the end of the period. This allows for
the possibility that it may take some people more time for the program to be effective. (We also capture
this effect later in the report when we identify overtime trajectories of incarceration and homeless shelter
use.) We estimated effects for many different time-based outcomes for days and episodes of each
situation and, in general, the results we report here were true for these other outcomes as well.
Figure 3. Intervention Effects for Housing
23
Panel B. Homeless Shelter Use: From Program Enrollment
Day and Episode Measures
Intervention
Group Mean
(n = 60)
Comparison
Group Mean
(n = 70)
Difference
of Means
95% Confidence Interval for
Difference of Means
Number of days over 24 month follow-up
68.2
161.9
-93.7***
-113.3
-74.1
Number of episodes over 24 month follow-up
3.0
8.7
-5.7***
-7.5
-3.9
Any episodes over 24 month follow-up
60.0%
81.4%
-21.4%***
-27.5%
-15.4%
Number of days over last 3 months of follow-up
7.7
24.5
-16.8***
-25.0
-8.5
Number of episodes over last 3 months of follow-up
0.3
1.3
-1.0***
-1.4
-0.6
Any episodes over last 3 months of follow-up
10.0%
32.9%
-22.9%***
-30.8%
-15.0%
* p < .05 ** p < .01 *** p < .001
Table 8. Intervention Effects for Homeless Shelter Use
Panel A. Homeless Shelter Use: From Housing Placement
Day and Episode Measures
Intervention
Group Mean
(n = 60)
Comparison
Group Mean
(n = 70)
Difference of
Means
95% Confidence Interval for
Difference of Means
Number of days over 24 month follow-up
15.2
161.9
-146.7***
-166.1
-127.3
Number of episodes over 24 month follow-up
0.8
8.7
-7.9***
-9.4
-6.3
Any episodes over 24 month follow-up
11.7%
81.4%
-69.8%***
-75.8%
-63.8%
Number of days over last 3 months of follow-up
1.8
24.5
-22.7***
-28.5
-16.9
Number of episodes over last 3 months of follow-up
0.1
1.3
-1.2***
-1.5
-0.8
Any episodes over last 3 months of follow-up
3.3%
32.9%
-29.5%***
-35.6%
-23.5%
24
These results report means and differences based on models that controlled for all baseline covariates
thought to affect selection into treatment and outcomes (i.e., the variables in Table 3) plus the variable at
baseline that measures the same phenomenon as the outcome of interest, e.g., measuring days in shelter
over 24 months of the study period, controlling for days in shelter during the 24 months prior to
baseline.
These results show that FUSE program effectiveness in reducing homeless shelter use are substantively
and statistically very strong. For shelter use measured from housing placement, all results are significant
at p < .001 and, for the most part, the absolute differences are large. For instance, the number of days in
shelter over 24 month follow-up was, on average, 146.7 days less for those in the intervention group
than for those in the comparison group, and the percentage of those with any episode was reduced on
average by 69.8%. Not surprisingly, the effects measured from program enrollment were less strong.
Since the major element of the program was housing, and since there was often substantial time elapsed
between program enrollment and actually securing housing, it is to be expected that the greatest impact
would be when people actually moved into their permanent housing. Nevertheless, the findings are
generally robust. Reporting the same effects we just noted, over the entire follow-up period, days in
shelter from program enrollment were reduced by 93.7 days and percentage of people with any shelter
episode was reduced by 21.4%. All comparisons of homeless shelter use show statistically significant
differences whether measured from initial program enrollment or housing placement.
B. Incarceration Outcomes
Table 9 reports effects for jail incarceration over the 24 months of follow-up from initial program
enrollment and from placement in FUSE housing. The table reports the number of days and episodes
and percent of intervention and group members who have had any episode and the number of days and
episodes and percent having any episode over the last three months of the 24 month follow-up period.
For the most part, the results for incarceration show reductions in jail involvement benefiting the
intervention group, though results are not always statistically significant. Measuring from housing
placement, we find that people receiving the intervention had on average 19.2 fewer days incarcerated,
40% less than the comparison group. They also had fewer jail admissions and a smaller percentage had
any episodes in jail over the 24 month follow-up period or during the last three months of this period.
Most but not all comparisons are statistically significant at p < .05. In contrast to the findings for shelter
use, there is little difference in this effect whether we measure from initial program enrollment or from
housing placement. The large confidence intervals indicate substantial variation in incarceration
outcomes. The baseline and outcome distributions for the variables reported in the table suggest that the
intervention had its greatest effect on those in the middle of the distribution at baseline and least effect
on persons with the greatest number of days in or admissions to jail prior to enrollment. In the
following section, we examine different temporal patterns of jail (and shelter) involvement over follow-
up to better specify the subset of program participants who were less successful in avoiding continuing
jail involvement.
25
Table 9. Intervention Effects for Jail Incarceration
Panel A. Incarceration: From Housing Placement
Day and Episode Measures
Intervention
Group Mean
(n = 60)
Comparison
Group Mean
(n = 70)
Difference of
Means
95% Confidence Interval
for Difference of Means
Number of days over 24 month follow-up
28.4
47.6
-19.2**
-31.0
-7.3
Number of episodes over 24 month follow-up
1.0
1.6
-0.6*
-1.1
-0.1
Any episodes over 24 month follow-up
46.7%
51.4%
-4.8%
-12.0%
2.4%
Number of days over last 3 months of follow-up
7.9
11.0
-3.2***
-7.1
0.8