University of Southern California Dissertations and Theses

Homelessness and substance use treatment: using multiple methods to understand risks, consequences, and unmet treatment needs among young adults

(USC Thesis Other)

Homelessness and substance use treatment: using multiple methods to understand risks, consequences, and unmet treatment needs among young adults

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Copyright 2023 Graham DiGuiseppi

Homelessness and Substance Use Treatment: Using Multiple Methods to Understand Risks,
Consequences, and Unmet Treatment Needs Among Young Adults
by
Graham DiGuiseppi, ScM

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirement of the Degree
DOCTOR OF PHILOSOPHY
(SOCIAL WORK)

May 2023

ii

Dedication
To Crystal, for your never-ending love and support.
And to Zach, for your hope and resilience.

iii

Acknowledgements
First, I would like to acknowledge my academic advisors, Dr. Jordan P. Davis and Dr.
Eric Rice. You have dedicated an inordinate and ridiculous amount of time to help me complete
this work, from helping me devise the most important research questions to ask, work through
the methods to answer them, and interpret what the results mean for science and society. The
generosity you have shown me has undoubtedly helped me grow as a social work scientist. I
hope to pay this forward someday by serving as a mentor to students and aspiring colleagues. My
heartfelt thanks extend to others on my dissertation committee: John Prindle, Ben Henwood,
Stanley Huey, and Joan Tucker. This work would not be the same without your thoughtful
guidance and expertise.
I would like to thank the Suzanne Dworak-Peck School of Social Work for showing
interest in me as a doctoral student and providing financial and academic support throughout my
doctoral program. In particular I would like to thank Dr. Michael Hurlburt for his steadfast
guidance and dedication as the program chair and as the instructor in my grant writing course,
without whom I may not have submitted a grant proposal to the National Institute on Drug Abuse
and received funding for this work. I would also like to thank Malinda Sampson, Ph.D. Program
Manager, and other faculty at the School of Social Work for your helpful guidance, expertise,
and support throughout my doctoral studies.
I would like to thank The National Institute on Drug Abuse, for seeing promise and
providing funding for this work. The Ruth L. Kirchstein National Research Service Award (F31)
is a generous gift to predoctoral students who want to develop their skills to become independent
health scientists, and it has helped me reach my scientific training goals. The F31 award also
provided financial support by helping me compensate research participants, buy textbooks and

iv

software, travel to conferences, and pay for my tuition. I am eternally grateful for this support
and hope to continue to work with NIDA to advance future research dedicate to the prevention
and treatment of substance use problems.
I would like to thank my community research partners who graciously helped facilitate
this research. This includes Daniel Ballin at Covenant House California, Dr. Marvin Belzer and
Dr. Irene Lim at Children’s Hospital Los Angeles; Toni Cooper, Erin Casey, and Heather
Carmichael and the wonderful staff at My Friend’s Place; Michael Dennis and Kathryn
Modisette at the GAIN Coordinating Center, and all of the participating clinical sites who used
the GAIN set of assessments and allowed their de-identified patient data to be used for research
purposes. Your work supporting youth and young adults struggling with homelessness and
addiction has saved countless lives, and I hope that I may someday do a fraction of the good you
have done. I would also like to thank Dr. Carolyn Wong at Children’s Hospital Los Angeles,
who served as a research mentor to me and believed in the value of my work.
Thanks to the graduate students and research staff who helped me carry out qualitative
data analysis: Angeles Sedano, Nina Christie, Laura Petry, Shaddy Saba, and Colin Ring. I
thoroughly enjoyed working with each of you and appreciate your helpful insights while
analyzing the data. I also could not have gotten through this doctoral program without the
support of my fellow students at USC, including but not limited to Jessica Dodge, Nina Christie,
Colin Ring, Sara Semborski, Sheila Pakdaman, Shaddy Saba; and my fellow “math camp
veterans” Leslie Schnyder, Adriane (AJ) Clomax, Erika Salinas, Jiaming Liang, Sara Miller,
Jessenia De Leon, and Stephen Morgano. I am so glad I met all of you and we went through this
journey together.

v

And thanks to my wife, Crystal, for being by my side for the past 15 years. You have
always believed that my work will help someone, someday, somewhere, even though I did not
always believe so. Likewise, I have always believed in you. We would not be where we are
today without each other.
Finally, I would like to thank the young adults and service providers who dedicated their
time and effort to share their stories and experiences with me, without you I would not have been
able to do this work.

vi

Table of Contents
Dedication ....................................................................................................................................... ii
Acknowledgements ........................................................................................................................ iii
List of Tables ............................................................................................................................... viii
List of Figures ................................................................................................................................ ix
Abbreviations .................................................................................................................................. x
Abstract .......................................................................................................................................... xi
Chapter 1: Introduction .................................................................................................................. 1
The Dual Problems of Youth Homelessness and Substance Use Disorder ........................ 1
Utilization of Substance Use Treatment Services ............................................................... 4
Homelessness Among Young Adults in Treatment: An Unexplored Terrain ................... 5
Overview of Dissertation Chapters ..................................................................................... 7
Conceptual Framework ....................................................................................................... 9
Chapter 2 (Study 1): Using Machine Learning to Predict Homelessness Among Young Adults
Entering Substance Use Treatment ............................................................................................... 12
Abstract ............................................................................................................................. 12
Introduction ....................................................................................................................... 14
Methods............................................................................................................................. 19
Results ............................................................................................................................... 28
Discussion ......................................................................................................................... 30
Chapter 3 (Study 2): Examining Bidirectional Associations Between Homelessness and
Substance Use Among Young Adults in Substance Use Treatment............................................. 47
Abstract ............................................................................................................................. 47

vii

Introduction ....................................................................................................................... 49
Methods............................................................................................................................. 54
Results ............................................................................................................................... 61
Discussion ......................................................................................................................... 67
Chapter 4 (Study 3): Barriers and Facilitators of Substance Use Treatment for Young Adults
Experiencing Homelessness: Perspectives from Young Adults and Service Providers .............. 87
Abstract ............................................................................................................................. 87
Introduction ....................................................................................................................... 88
Methods............................................................................................................................. 94
Results ............................................................................................................................. 100
Discussion ....................................................................................................................... 112
Appendix A: Young Adult Qualitative Interview Guide ............................................... 127
Appendix B: Service Provider Qualitative Interview Guide ......................................... 130
Appendix C: Qualitative Codebook ............................................................................... 133
Chapter 5: Conclusion................................................................................................................ 140
Review of Study Findings ............................................................................................... 140
Implications for Treatment-Engaged Young Adults....................................................... 143
Implications for Young Adults Experiencing Homelessness Not Engaged in Treatment
......................................................................................................................................... 144
Charting a Future Research Agenda ............................................................................... 147
Conclusion ...................................................................................................................... 150
References ................................................................................................................................... 151

viii

List of Tables
Table 2.1 Participant Characteristics at Treatment Intake (N = 40,758) ...................................... 37
Table 2.2 Descriptive Statistics for Original and Imputed Data Sets ........................................... 39
Table 2.3 Logistic Regression Predicting Past-Year Homelessness at Treatment Entry ............. 41
Table 2.4 Classification Model Performance Metrics .................................................................. 43
Table 3.1 Participant Characteristics at Treatment Intake (N = 3,717) ........................................ 74
Table 3.3 Model Fit Comparisons for Substance Use Frequency Growth Model ........................ 76
Table 3.4 Model Fit Comparisons for Homeless Days Growth Model ........................................ 77
Table 3.5 Nested Model Comparisons for Unconditional Bivariate LGC-SR Model .................. 78
Table 3.6 Parameter Estimates for Conditional LGC-SR Model ................................................. 80
Table 3.7 LGC-SR Parameter Estimates Stratified by Sex........................................................... 81
Table 3.8 LGC-SR Parameter Estimates Stratified by Race/Ethnicity ......................................... 83
Table 3.9 LGC-SR Parameter Estimates Stratified by Sexual/Gender Minority Status ............... 85
Table 4.1 Characteristics of Young Adult In-depth Interview Participants (N = 22) ................. 116
Table 4.2 Characteristics of Service Providers (N = 16) ............................................................ 117
Table 4.3 Themes and Exemplary Quotes from Young Adults and Providers ........................... 120

ix

List of Figures
Figure 1.1 Dissertation Conceptual Framework ........................................................................... 11
Figure 2.1 Variables Most Strongly Associated with Past-Year Homelessness in Stepwise
Logistic Regression ....................................................................................................................... 44
Figure 2.2 Variables Most Strongly Associated with Past-Year Homelessness in Lasso Logistic
Regression ..................................................................................................................................... 45
Figure 2.3 Variable Importance from Random Forest Gini Index ................................................ 46
Figure 3.1 LGC-SR Model of Homeless Days and Substance Use Frequency ............................ 79
Figure 4.1 Problematic Substance Use Among Young Adults Completing Screening Survey . 118
Figure 4.2 Themes Displayed within an adapted Behavioral Model for Vulnerable Populations
..................................................................................................................................................... 119

x

Abbreviations
AUC Area Under the Curve
GAIN Global Appraisal of Individual Needs
LGC-SR Latent Growth Curve Model with Structured Residuals
ML Machine Learning
MLR Robust Maximum Likelihood
RAM Risk Amplification Model
SEM Structural Equation Modeling
SUD Substance Use Disorder
YAEH Young Adults Experiencing Homelessness

xi

Abstract
An estimated one in 10 young adults (aged 18 to 25) experience homelessness in the
United States each year. Collectively, young adults experiencing homelessness (YAEH) are a
highly vulnerable population at greater risk for poor physical and mental health outcomes. One
such risk relates to substance use and substance use disorders (SUDs), which are exceedingly
high among YAEH when compared to stably housed young adults. More research is needed to
draw attention to the needs of YAEH who suffer from SUDs. One unexplored area to do this is
within substance use treatment settings. Almost no research to date has comprehensively
examined risk and protective factors for homelessness among young adults receiving substance
use treatment. Neither is there a solid understanding of how homelessness and substance use may
relate to each other (bidirectionally) over time among treatment-engaged young adults.
Furthermore, more research is needed to understand the barriers and facilitators of treatment
engagement from the perspective of both YAEH and providers. The following dissertation is
composed of three studies that address these domains of inquiry. Study 1 uses Machine Learning
(ML) to identify risk and protective factors for homelessness in a large sample of young adults
receiving treatment in the U.S. Study 2 examines bidirectional (i.e., reciprocal) associations
between homelessness and substance use during and after treatment. Finally, Study 3 uses
qualitative methods to investigate barriers and facilitators to treatment among YAEH (n = 22)
and service providers (n = 16) in Los Angeles, California. Taken together, this research may
inform efforts to engage YAEH with SUDs in substance use treatment and address vulnerable
young adults’ housing needs during and after treatment.

1

Chapter 1: Introduction
The Dual Problems of Youth Homelessness and Substance Use Disorder
Youth homelessness is a substantial problem in the United States and throughout the
world. In the U.S. alone, it is estimated that 30,090 “unaccompanied” youth under the age of 25
were experiencing homelessness on a single night in January 2022—that is, without a parent or
guardian or a child of their own (de Sousa et al., 2022). A large majority of these youth (91%)
are young adults between the ages of 18 and 25. Young adulthood is a time of great change and
opportunity (Arnett, 2000). However, young adults experiencing homelessness are among the
most vulnerable members of our society and have encountered significant adversity. This calls
into question their chances of securing a healthy and stable future in their transition to adulthood.
Young adults who experience homelessness are more likely than stably housed young adults to
be members of racial, ethnic, and sexual and gender minority groups (i.e., Black or African
American, Hispanic, and lesbian, gay, bisexual, queer and/or transgender) (de Sousa et al., 2022;
Morton et al., 2018), come from low income households, have lower educational attainment
(Brakenhoff et al., 2015; Morton et al., 2018; Shah et al., 2017; van den Bree et al., 2009), and
are more likely to be involved in the child welfare and criminal justice systems (Almquist &
Walker, 2022; Dworsky et al., 2013). Lifetime rates of abuse, neglect, and victimization are
startingly high among young adults experiencing homelessness (DiGuiseppi et al., 2022; Tyler &
Ray, 2019a; van den Bree et al., 2009). This abuse and family dysfunction are often cited as
primary reasons youth leaving or being “thrown out” of home (Embleton et al., 2016).
The Risk Amplification Model (RAM) is the most relevant theoretical model describing
(mostly pathological) development for this population. The RAM was developed by Whitbeck,
Hoyt, and Yoder (1999) and tested empirically using data collected from a sample of adolescents

2

recruited from street and shelter locations in four midwestern states. The RAM is a life course
developmental theory that describes how street-based experiences (i.e., substance use, deviant
subsistence strategies, and non-conventional peer affiliations) amplify pre-existing risks
stemming from youths’ dysfunctional family backgrounds. The authors demonstrated that these
street-based experiences increase the risk for physical and sexual victimization; which in turn,
increase risks for the further development of mental health and substance use problems
(Whitbeck et al., 1999). While this model has been critiqued by Milburn and colleagues (2009)
for its focus on negative developmental processes, the RAM has proven helpful for investigating
adaptive behaviors among young adults experiencing homelessness—namely, substance use and
SUDs—in the present dissertation.
Given such life histories, it may come as no surprise that systematic reviews indicate high
rates of a variety of physical and mental health problems among youth experiencing
homelessness, ranging from dental problems to post-traumatic stress disorder (Edidin et al.,
2012). Substance use is also a concern, with studies generally indicating a higher prevalence of
substance use and substance use disorders among youth experiencing homelessness, relative to
stably housed youth. A recent study by Chassman and colleagues (2022) used reports from 1,426
young adults experiencing homelessness in seven large U.S. cities and revealed high rates of
substance use. Past month rates of heavy drinking (34.0%), marijuana (60.7%),
methamphetamine (17.5%) and injection drug use (8.3%) were much higher than those found in
the general population of young adults in the past year. For comparison, past-month rates of
heavy drinking (binge drinking on 5 or more of the past 30 days, 8.6%) and past-year marijuana
(34.5%), methamphetamine (0.5%) and heroin use (reliable data on injection drug use not
available; 0.2%) were much lower in the general population of young adults (Substance Abuse

3

and Mental Health Services Administration [SAMHSA], 2021). The prevalence of clinically
relevant substance use disorders is also high among young adults experiencing homelessness
(ranging from 40 to 78 percent across studies; Baer et al., 2003; DiGuiseppi et al., 2022;
Whitbeck, 2009), indicating that substance use causes significant consequences and interferes
with daily activities among a sizeable proportion of this population.
The etiology of substance use and SUDs among young adults experiencing homelessness
is not a central question of this dissertation, but deserves discussion. To be clear, substance use is
not universal problem among this population, despite high prevalence rates. Youth who are
younger, female, residing in shelters (vs. unsheltered locations), more engaged in education and
employment, and youth with more conventional and familial social support resources are less
likely to have problems associated with alcohol and other drugs (DiGuiseppi, Davis, Christie, et
al., 2020; Milburn et al., 2009). The RAM can be relevantly applied to understand substance use
among this population, in addition to a sizeable body of qualitative and quantitative studies
investigating this issue. Substance use can be understood as a social learning process. Youth may
be initiated into substance use by family members (including parents) or peers (Melander et al.,
2016; Tyler & Ray, 2019b); these youth may be more likely to continue using into young
adulthood (Mallett et al., 2005, 2010; Whitbeck, 2009). Once on their own, the sheer availability
and access to various types of substances through street networks increases the likelihood that
youth will use these substances (Barman-Adhikari et al., 2018; DiGuiseppi, Davis, Christie, et
al., 2020; Wenzel et al., 2012). Paired with youths’ extensive backgrounds of trauma and
victimization, alcohol and drugs are used as a means to cope and connect with peers
(DiGuiseppi, Davis, Christie, et al., 2020; Heerde & Hemphill, 2016; Tyler et al., 2018;
Whitbeck et al., 1999).

4

Utilization of Substance Use Treatment Services
Such high rates of physical, mental health, and SUD are only made worse by inadequate
access to preventative and routine medical care. Although available research suggests that young
adults experiencing homelessness have inadequate access to medical care in general (Edidin et
al., 2012), previous studies show relatively high lifetime rates of substance use treatment
utilization. For example, 23 percent of men and 15 percent of women in a longitudinal study of
homeless young adults had received inpatient SUD treatment, and 38 percent overall had
received outpatient SUD treatment in their lifetime (Whitbeck, 2009). This may be indicative of
greater involuntary access to treatment through youths’ involvement in the child welfare and
juvenile justice systems. Despite this, studies conducted with young adults currently
experiencing homelessness have identified several barriers to treatment. These include
systematic barriers (i.e., lack of insurance or ability to pay), and personal barriers, such as low
perceived need for treatment (Nyamathi et al., 2007; Rabinovitz et al., 2010; Whitbeck, 2009).
Barriers and facilitators to treatment are important to understand in order to increase young
adults’ access to effective treatment services, whether this is done by removing barriers or
strengthening facilitators. Missing from previous research on this topic is the perspective of
service providers, who are important for referring youth to substance use treatment services or
providing these services directly. Providers may be more attuned to the organizational barriers to
care (e.g., lack of staff training), and may have interesting perspectives on facilitators to
treatment as well (Shadel et al., 2014).
The Behavioral Model for Vulnerable Populations was developed to understand
healthcare service utilization among (primarily adult) homeless populations (Gelberg et al.,
2000). The model presents three types of factors that are important in determining an

5

individual’s use of healthcare services. These are predisposing factors (demographics and
individual factors that predispose one to illness), enabling factors (i.e., insurance coverage and
availability of services) and need factors (perceived need or objective need determined by a
formal diagnosis). The Behavioral Model for Vulnerable Populations builds upon a prior model
of healthcare service use among general population by adding vulnerability factors unique to
individuals experiencing homelessness (e.g., length of time spent homeless, victimization, etc.).
This provides a helpful framework to understand barriers and facilitators to health service use for
young adults experiencing homelessness (Pedersen et al., 2016; Rosa Solorio et al., 2006), but
has not yet been applied to substance use treatment services.
Homelessness Among Young Adults in Treatment: An Unexplored Terrain
According to the most recent data collected in 2020, an estimated 445,000 young adults
in the U.S. received some form of substance use in the past year (SAMHSA, 2021). While this
represents only 4.4 precent of young adults who met past-year criteria for SUD, it is not a trivial
number. Young adults receiving substance use treatment share many of the same vulnerabilities
as young adults experiencing homelessness, including high rates of adverse childhood
experiences, victimization, and criminal justice system involvement (Davis et al., 2020; Perker &
Chester, 2021; Spencer et al., 2021). However, few studies have documented risks for
homelessness among treatment-engaged young adults. This is astounding, given the potential
risks for homelessness among this population, and represents a potential missed opportunity for
identifying individuals in need of housing services. In a study of young adult treatment
admissions from 1992 to 2017 in the U.S., Green and colleagues (2020) reported that 6.5% of
these cases (approximately 206,000 individuals) were residing in shelters or did not have a fixed
address at treatment admission. These cases were more likely to seek treatment for illicit drug

6

use (vs. alcohol or marijuana), be self-referred (vs. criminal justice referral), and be admitted to
inpatient/residual care (vs. outpatient) than young adults with more conventional housing
situations.
Although this work represents an important first step towards describing the unique
characteristics of this population, a more comprehensive examination of important correlates
(i.e., “predictors”) of homelessness among young adults in treatment is needed. These correlates
can be thought of as risk or protective factors identified in the RAM and relevant social
ecological models of development (Haber & Toro, 2004). A social ecological framework
presupposes that the determinants of homelessness exist across multiple domains, ranging from
micro individual-level factors (e.g., demographic characteristics) to broader societal factors such
as systematic racism and lack of affordable housing (Bronfenbrenner, 1979). Altogether, a more
comprehensive understanding of the risk and protective factors for homelessness among young
adults entering treatment will aid in identifying individuals at greater risk for homelessness as
they enter treatment, thereby promoting the possibility for more efficient delivery of housing
services.
Another important set of research questions for young adults in treatment relate to how
homelessness and substance use may relate to one another over time. The temporal associations
between homelessness and substance use have been explored in studies of youth experiencing
homelessness outside the context of treatment. As we have seen from studies using the RAM and
other life course models, homelessness can lead to an amplification of risks over time which may
increase risks for substance use problems and poor overall health (Mallett et al., 2005; Tyler &
Schmitz, 2020; Whitbeck et al., 1999). There is less empirical support, however, for the
alternative hypothesis that developing substance use problems leads to a greater risk of

7

homelessness. These temporal relationships are complex and possibly bidirectional, requiring
sophisticated statistical models (Hodgson et al., 2013). To date, only one study has examined this
empirically within the context of substance use treatment, using a clinical sample of adolescents
(Davis et al., 2019).
One of the aims of this dissertation is to extend this research by examining these
reciprocal associations among young adults in treatment—a population that may experience
homelessness and substance use over time differently than adolescents, by virtue of the
developmental risks and opportunities they are confronted with. Specifically, three hypotheses
will be tested: 1) A risk-enhancing hypothesis that more days spent homeless prior to treatment
will be associated with worse treatment outcomes (more substance use immediately post-
treatment), 2) A hypothesis informed by the relapse prevention model (Larimer et al., 1999),
which assumes that increases in homeless days during or after treatment will be associated with
greater substance use at a later point in time, and 3) a symptom-driven model which states that
homelessness is a consequence of increased substance use during or after treatment, just as
substance use can initiate problems in other life domains like relationships, work, and school.
This work will be extended further by examining whether these reciprocal associations differ for
important subgroups—by sex, race/ethnicity, and sexual and/or gender minority status. This will
be important for examining whether certain groups are at greater risk as a result of the (potential)
bidirectional relationships between substance use and homelessness—before, during, and after
treatment.
Overview of Dissertation Chapters
The overall goal of this dissertation is to understand complex associations between
homelessness and substance use among young adults, with implications for enhanced service

8

delivery inside and out of substance use treatment settings. This work involves two intersecting
populations: 1) young adults currently in treatment and experiencing homelessness (or at risk of
experiencing homelessness), and 2) young adults with SUD who are experiencing homelessness
and not in treatment. This is carried out using multiple, advanced quantitative and qualitative
methods. This dissertation research was funded by a Ruth L. Kirchstein National Research
Service Award from the National Institute on Drug Abuse (Grant #: F31DA053779), which had
three specific aims. These aims culminated in three studies and corresponding dissertation
chapters.
Chapter 2 presents results of Study 1, which uses Machine Learning (ML) classification
models to predict the likelihood of past-year homelessness among a nationwide sample of young
adults entering substance use treatment. ML is helpful for developing predictive models using
large data sets. Predictors include important variables from the RAM and social-ecological
models of homelessness (e.g., substance using peer affiliations, victimization, etc.). Data for this
study come from a recently updated version of The Global Appraisal of Individual Needs
(GAIN), which includes intake assessment data from 40,758 young adults entering U.S.
substance use treatment clinics from 2002 to 2018. Three ML models (logistic regression, Lasso
regression, and Random Forests) were developed (“trained”) using a subset of the data, and
evaluated (“tested”) for their predictive accuracy on another subset of the data.
Chapter 3 presents the results of Study 2, which examines bidirectional associations
between number of days spent homeless and the percentage days young adults’ evidenced
substance use or problems in the past 90 days. Data come from a smaller, longitudinal GAIN
data set that includes follow-up assessments from 3,717 young adults extending out to 12 months
post treatment intake. Latent Growth Curve models with Structured Residuals (LGC-SR) were

9

used to examine between and within-person change and bidirectional associations between
homelessness and substance use before, during, and after treatment. These models were stratified
by sex, race/ethnicity, and sexual and/or gender minority status to investigate potentially
important differences across vulnerable subgroup.
Chapter 4 contains Study 3, which uses qualitative methods (in-depth interviews and
Thematic Analysis) to better understand perceived barriers and facilitators to treatment among
young adults currently experiencing homelessness and service providers working with this
population. Young adults (n = 22) were receiving services (e.g., food, clothing, hygiene and case
management) at one drop-in center in Los Angeles. Service providers (n = 16) were employed in
a variety of roles at agencies providing direct services to young adults experiencing
homelessness in Los Angeles; all had experiencing delivering or referring youth to substance use
treatment.
Chapter 5 integrates findings from these three empirical studies, and outlines implications
for future research, practice, and policy for young adults experiencing homelessness who may
suffer from SUDs.
Conceptual Framework
Given the multiple aims of this dissertation, it may be helpful to present a conceptual
framework showing how these aims relate to one another in a cohesive manner. Figure 1.1
presents such a framework. As previously mentioned, the two populations of interest are young
adults currently engaged in treatment experiencing homelessness or at risk of homelessness, and
currently homeless engaged young adults not engaged in treatment (represented by the two
disjointed rectangles). Studies 1 and 2 have implications for the young adult treatment

10

population. Namely, by identifying those at greater risk for homelessness at treatment admission
(Study 1) and investigating bidirectional associations between homelessness and substance over
time (Study 2). The three hypotheses tested in Study 2 are represented by the dashed boxes and
arrows pointing to their respective cross-lagged paths (from homelessness to substance
use/problems and vice-versa) evaluated in the final LGC-SR model. Study 3 has implications for
young adults experiencing homelessness with substance use disorders who are not currently in
treatment. This population is represented by the separate darker oval in the bottom left portion of
the figure. An arrow points from this section to the larger oval above, indicating that findings
from Study 3 may help understand young adults’ eventual treatment engagement. The theoretical
concepts informing each of these studies are contained within the figure as well, demonstrating
that this overall framework is guided by prior research and theory related to young adult
development, service use, and behavior change.

11
Figure 1.1 Dissertation Conceptual Framework

12

Chapter 2 (Study 1): Using Machine Learning to Predict Homelessness Among Young
Adults Entering Substance Use Treatment
Abstract
Background: In the United States, nearly half a million young adults aged 18 to 25 receive
substance use treatment each year. Young adults in substance use treatment may be at greater
risk for experiencing homelessness, but a comprehensive investigation of risk and protective
factors for homelessness in this population has not been conducted. The current study uses
traditional logistic regression and machine learning classification models to address this
important public health question. Methods: Data come from 40,758 young adults (Mage = 21.4,
SD = 2.4, 34.9% female, 62.6% Non-Hispanic White) receiving substance use treatment in the
United States who completed a Global Appraisal of Individual Needs intake assessment at
treatment entry. Risk and protective factors from previous literature were selected as predictor
variables, and stepwise logistic regression, penalized (Lasso) logistic regression, and random
forest classification models were used to identify significant correlates of homelessness in the
year prior to treatment. Results: Models correctly classified the past-year housing status
(homeless or housed) of about two-thirds of all individuals (64.4% - 66.5%). Models correctly
classified a greater proportion of homeless individuals than housed individuals (sensitivity: 73.7
- 76.4%; specificity: 60.8 - 64.4%). Demographic, familial, mental health, and behavioral
variables were identified as important correlates of past-year homelessness, with some
differences in variable importance across models. Conclusions: Commonly used classification
models identified correlates of homelessness that are consistent with previous literature.
Implications for the use of predictive modeling to identify individuals at greater risk of
homelessness in clinical settings are discussed.

13

Keywords: Young adults experiencing homelessness; machine learning; classification analysis;
substance use treatment

14

Introduction
In 2020, approximately 445,000 young adults aged 18 to 25 (1.3 percent of all young
adults in the U.S.) received substance use treatment in the past year (Substance Abuse and
Mental Health Services Administration [SAMHSA], 2021). Young adults in substance use
treatment report high rates of adverse experiences, including childhood abuse and victimization,
co-occurring mental health disorders, and criminal justice system involvement (Davis et al.,
2020; Perker & Chester, 2021; Spencer et al., 2021). Young adults in treatment may also be at
greater risk for experiencing homelessness than young adults in the general population
(DiGuiseppi, Davis, Leightley, et al., 2020; Green et al., 2020). However, a comprehensive study
of risk and protective factors for homelessness among this population has not been carried out.
Existing studies have tended to rely on traditional analytic methods, such as logistic regression to
identify correlates of homelessness among young adults in the general population. Recent studies
have applied machine learning algorithms to identify predictors of homelessness and substance
use treatment outcomes using large data sets, and have evaluated the predictive performance
(i.e., accuracy) of these models (Barenholtz et al., 2020; Byrne et al., 2020). The present study
adds to this growing body of work by evaluating the use of three classification models to predict
a dichotomous outcome—homelessness in the year prior to treatment—using a large, nationwide
data set of young adults receiving substance use treatment in the United States.
Prevalence and Correlates of Homelessness among Young Adults
Recent estimates indicate that 91 percent of all unaccompanied youth who experience
homelessness in the United States are young adults between the ages of 18 and 24, and one in 10
young adults experience homelessness each year (de Sousa et al., 2022; Morton et al., 2018).
Although captured at a single point in time, 6.4% of young adult substance use treatment

15

admissions from 1992-2017 involved individuals without a fixed address or residing in shelters
(Green et al., 2020). Homelessness among adolescents and young adults is thought to be
associated with multiple social-ecological factors that vary in importance depending on an
individual’s developmental stage (Bronfenbrenner, 1979; Haber & Toro, 2004). In a meta-
analysis of youths’ self-reported reasons for homelessness, top reasons were family conflict
(48%), running away (46%), poverty (30%), abuse (29%), psychosocial health (i.e.,
sexuality/gender issues, mental health, substance use; 26%), and delinquency (20%) (Embleton
et al., 2016). This largely comports with theories such as the Risk Amplification Model, which
posits that family conflict, abuse, and neglect often lead youth to run away or be forced out of
home by caregivers or child protective services (Whitbeck et al., 1999). Once on their own, these
youth are more likely to experience several negative consequences such as physical and sexual
victimization, poor mental health, and substance use problems (Whitbeck et al., 1999).
Studies of young adults in the general population have identified risk and protective
factors for homelessness across multiple social-ecological domains. For example, demographic
and economic variables emerged as important correlates of homelessness in a recent population-
based study by Morton and colleagues (2018). Young adults who were Black/African-American,
lesbian, gay, bisexual and/or transgender (LGBT), without a high school diploma or general
education diploma, unmarried parents, and those with annual household incomes less than
$24,000 were more likely to experience homelessness in the past year (Morton et al., 2018).
Other longitudinal studies have identified risk and protective factors in adolescence that are
prospectively associated with homelessness in young adulthood. For example, familial and peer
factors including adverse childhood experiences (Shelton et al., 2009; Tucker et al., 2021),
multiple runaway episodes (Brakenhoff et al., 2015), family dysfunction (Brakenhoff et al.,

16

2015; Shelton et al., 2009; van den Bree et al., 2009) and peer substance use (Tucker et al., 2021)
in adolescence have been associated with a higher risk of homelessness in young adulthood.
Furthermore, educational and economic variables such as school suspensions and academic
failures (Brakenhoff et al., 2015; Heerde et al., 2020; van den Bree et al., 2009) and economic
instability (Brakenhoff et al., 2015; Shelton et al., 2009) have been identified as risk factors.
Greater victimization (van den Bree et al., 2009), poor mental health (Shelton et al., 2009), and
drug use in adolescence (Shelton et al., 2009) are also uniquely associated with homelessness in
young adulthood. Protective factors associated with a lower risk of homelessness in young
adulthood include demographic and familial factors such as Hispanic ethnicity and having a
permissive parent (Brakenhoff et al., 2015) as well as educational factors like greater academic
achievement and academic aspirations (Shah et al., 2017; Tucker et al., 2021).
To date, only one study has investigated correlates of homelessness among young adults
receiving substance use treatment. Compared to stably housed young adults, young adults
entering substance use treatment facilities from 1992 to 2017 without a fixed address differed in
terms of mental health, substance use, and treatment-related variables (Green et al., 2020).
Specifically, young adults experiencing homelessness appeared more likely to report protective
factors in some domains (initiated substance use after age 14, less likely to have co-occurring
substance use and mental health diagnosis), and risk factors in others (more likely to be admitted
to treatment for illicit drug use [i.e., cocaine/crack, heroin, or methamphetamine] versus
marijuana or alcohol use). Young adults experiencing homelessness were also more likely to
self-refer to treatment than be referred by the criminal justice system, and more likely to be
admitted to inpatient/residential treatment than outpatient treatment. While this study provides
initial evidence of differences between homeless and housed young adults in treatment, more

17

research is needed to identify young adult risk and protective factor for homelessness using a
broader set of predictor variables across multiple social-ecological domains.
The Need for Multiple Statistical Methods
Most existing studies examining correlates of homelessness among young adults use
traditional statistical approaches such as logistic regression. In recent years, however, researchers
have applied supervised machine learning models to investigate homelessness and substance use
treatment outcomes (Barenholtz et al., 2020). Supervised machine learning models are “trained”
on a set of existing cases and then used to “predict” an outcome in a new set of cases, given a set
of variables (James et al., 2021). Some key advantages of using machine learning algorithms are
feature reduction (identifying the most important variables in a large set), evaluating model
accuracy on new testing data sets, and employing models in real-world applications. In a recent
example, Byrne and colleagues (2020) used machine learning to develop a model of
homelessness that correctly classified 84% of cases and estimated opioid overdose mortality
rates using model-based predictions. These model-predicted homeless cases had a 23-fold
increased risk of overdose mortality than stably-housed cases.
Recent studies using machine learning often employ multiple model algorithms, allowing
researchers to compare and contrast results of different modeling strategies, each with their own
unique output and interpretations. For example, Davis and colleagues (2021, 2022) have used
least absolute shrinkage and selection operator (lasso) regression and random forest models to
investigate longitudinal substance use treatment outcomes. The authors found converging
evidence identifying variables that were most strongly associated with relapse to illicit drug use
among individuals across models, and were able to draw implications from both sets of results.
Namely, results of random forest models rank-ordered variables according to their importance in

18

determining the outcome, whereas lasso regression produced estimates of the strength (or
“valence”) of associations, expressed as hazard ratios (Davis et al., 2022). The use of multiple
machine learning algorithms can therefore be applied to identify the most important predictors of
homelessness among young adults entering treatment, and the strength of these associations,
while accounting for other variables. Further work is needed to determine the extent to which
results of different machine learning modeling strategies converge to predict important public
health problems like homelessness.
The Current Study
The purpose of the current study was to compare three different machine learning
classification models (stepwise logistic regression, lasso logistic regression, and random forest)
of young adult homelessness in the year prior to entering substance use treatment. Models were
compared with regards to variable importance and predictive accuracy. Predictor variables were
selected from theoretical concepts from the Risk Amplification Model (Milburn et al., 2009;
Whitbeck et al., 1999) and empirical literature. Variables were also grouped within social-
ecological-developmental domains (e.g., demographics, family and peer relationships) to
determine if features from specific domains had more important contributions to model
performance (Bronfenbrenner, 1979; Haber & Toro, 2004). Although the current study is not
intended to test specific theories, and no hypotheses are made, developing accurate classification
models of homelessness can inform theoretical work by identifying salient risk and protective
factors for homelessness among young adults entering treatment, thereby contributing to future
efforts to provide appropriate services during treatment.

19

Methods
Data Source
Data come from pooled Global Appraisal of Individual Needs-Initial (GAIN-I) data set,
which includes client intake assessments collected from 519 substance use treatment agencies in
43 U.S. states from 2002 to 2018. The GAIN-I is a comprehensive, biopsychosocial assessment
instrument that measures psychosocial functioning in eight areas (background, substance use,
physical health, risk behaviors, mental health, environment, legal, and vocational) (Dennis et al.,
2008). The GAIN-I is a one to two-hour, semi-structured interview that collects client self-report
data with the assistance of trained interviewers and accompanying software. The GAIN-I is used
for treatment planning, program evaluation and research purposes and was collected as part of
routine clinical practice or research studies at each participating site. Pooled GAIN data are
managed by the GAIN Coordinating Center, a division of Chestnut Health Systems, and were
obtained through a data use agreement (Ives et al., 2010). Clients provided their informed
consent to allow their de-identified data to be used for research purposes; informed consent and
data use procedures were approved by each study’s Institutional Review Board. Approval to
conduct secondary data analysis for the current study was also obtained from the University of
Southern California Institutional Review Board.
Participants
The original data set included 50,914 young adults with GAIN-I data (young adults
accounted for 29.8% of all participants in the GAIN-I data set). However, cases were excluded
from the current analyses if they were missing self-report data on sex, race/ethnicity, and
sexual/gender minority status (more information on missing data handling is described below).

20

This resulted in a final analytic sample of 40,758 young adults. Participants in the final analytic
sample received treatment from 405 clinical sites as part of 56 separate studies. The sample was
21 years old on average (SD = 2.4), majority male (65.1%), and non-Hispanic White (62.6%).
Most (77.4%) were involved in the criminal justice system and used substances on an average of
a 31 days (SD = 34.6) in the 90 days prior to treatment. In total, 9,096 young adults (22.3%)
experienced homelessness in the 12 months prior to treatment. Young adults who experienced
homelessness were significantly different from stably housed young adults for a majority of the
variables examined (ps < 0.05). For example, compared to stably housed young adults, homeless
young adults were less likely to be Hispanic, more likely to be sexual/gender minority, and
reported more victimization, worse mental health symptoms, more frequent substance use and
more substance use disorder symptoms. These descriptive statistics and between-group
comparisons are displayed in Table 2.1.
Measures
Homelessness and Housing
Past year homelessness was the primary outcome variable. Participants were asked,
“When was the last time, if ever, that you considered yourself to be homeless?” Response
options included “Never”, “More than 12 months ago”, “4 to 12 months ago”, “1 to 3 months
ago”, “1 to 4 weeks ago”, “3 to 7 days ago” or “Within the past two days.” A dichotomous
variable was created such that 1 = any homelessness experience in the past 12 months, and 0 =
never experienced homelessness or experienced homelessness more than 12 months ago.

21

Demographics
Demographic variables include participant age, gender identity, race, and Hispanic
ethnicity. Race/ethnicity responses were combined to produce six racial/ethnic categories
displayed in Table 2.1. Due to their low sample sizes, Asian, Native American, Native Hawaiian,
Pacific Islander, Alaskan Native and individuals identifying as another group were collapsed into
an Other racial/ethnic category. Sexual/gender minority status is a binary variable (1 = yes, 0 =
no) that was indicated if participants reported a transgender identity, non-heterosexual sexual
orientation, or sexual attraction or behavior with the same or both sexes in the past year (coded
as 1). Participants who reported male or female gender, heterosexual identity, and sexual
attraction and behavior with the same sex (or asexual/celibate) were considered non-
sexual/gender minority (coded as zero).
Family and Peer Relationships
Familial variables included dichotomous variables (1 = yes, 0 = no) indicating family
history of an alcohol or drug problem (“Have any of your blood relatives ever had… problems
with alcohol/drug use?”), currently married or living as married (reference: not currently
married or living as married) and unmarried with children (reference: has no children or married
with children). Deviant peer affiliations were assessed using the Social Risk Index (SRI; Dennis
et al., 2008), which consists of seven items assessing how many people the respondent is socially
involved with who: 1) binge drink, 2) use drugs, 3) get into fights/arguments, 4) are involved in
illegal activities, 5) are employed/in school, 6) in substance use treatment, and 7) in recovery.
Responses to each item include 0 = none, 1 = a few, 2 = some, 3 = most, 4 = all (items 5 through
7 are reverse scored). Responses are summed, with higher scores indicating greater social risk
(range: 0 – 28). The social risk index is a count variable that focuses on unique variance (versus

22

common variance), and therefore calculating the internal consistency (Cronbach’s alpha) is not
appropriate (Modisette & Dennis, 2022).
Education, Employment, and Income
High school graduate is a dichotomous variable that indicated if participants graduated
from high school or completed a general education diploma (1 = yes, 0 = no). Current
employment and school involvement was assessed by asking participants whether or not they
were working full time (“35 hours or more a week”), part time (“less than 35 hours a week”), or
currently “in school or training.” Currently employed or in school was indicated (1 = yes, 0 = no)
by affirmative responses to any of these items. Past 90-day income is a sum of income (in
dollars) participants reported receiving from a variety of sources (wages/salary, family/friends,
supplemental security, disability, unemployment, or other governmental assistance) in the past
90 days. This was rescaled into $500 units for easier interpretation of classification model
results.
Victimization and Criminal Justice Involvement
Lifetime physical, emotional, and sexual victimization were assessed using the General
Victimization Scale (GVS; Titus et al., 2003). The GVS is a count of each victimization type plus
six traumagenic factors (e.g., duration, relationship to perpetrator, negative social reactions to
disclosure of the event), plus four fear of re-victimization items (range: 0-15; Cronbach’s α =
0.88). Criminal behavior in the past 12 months was assessed using the General Crime Scale, a
count of 19 illegal activities engaged in (theft, breaking and entering, assault, drug dealing, etc.;
Cronbach’s α = 0.84) (Dennis et al., 2006). Criminal justice system involvement was assessed by
asking participants if they were currently involved in the criminal justice system in 12 different

23

ways (e.g., awaiting trial, on probation/parole, house arrest, etc.) or currently detained in jail,
detention, or prison. Current criminal justice system involvement is a binary variable indicated (1
= yes, 0 = no) by participants’ affirmative responses to any of these items.
Mental Health
Mental health variables included the Depressive Symptoms Scale (Modisette & Dennis,
2022), a count of 10 DSM-5 depression symptoms in the past 12 months (i.e., “diminished
interest or pleasure in all or almost all, activities most of the day”; range: 0-10; α = 0.90). Post-
traumatic stress disorder (PTSD) symptoms were measured using the Traumatic Stress Scale
(Lauterbach et al., 1997), a count of 12 PTSD symptoms in the past 12 months (e.g., “You had
nightmares about things in your past that really happened?”), adapted from the Civilian
Mississippi scale (range: 0-12; α = 0.94). Furthermore, Co-occurring disorder is a binary
variable also indicated (1 = yes, 0 = no) if participants met past year DSM diagnostic criteria for
a co-occurring mental health and substance use disorder.
Substance Use and Treatment Experiences
Five variables were used to characterize participants’ substance use and treatment
experiences. Age at first substance use is participants’ self-report of how old they were when
they “first got drunk or used any drugs.” Number of previous substance use treatment episodes is
a count of self-reported number of times admitted to inpatient, admitted to outpatient, prescribed
medication to help with withdrawal and cravings, or any other kind of treatment program for
“alcohol or other drug use problems.” Primary substance is a categorical variable indicating
participants’ primary substance (alcohol, cannabis, stimulants, opioids, and other), calculated
based on self-reported symptom severity, days of use, substance used, and recency of use for

24

each substance. Binary “dummy” variables were created to examine the association between
participants’ primary substance and likelihood of past-year homelessness, using cannabis (the
most frequently occurring primary substance) as the reference category. Number of substance
use days is the maximum number of days participants reported using a variety of substances
(alcohol, cannabis, cocaine, opioids, and other drugs) in the past 90 days at intake. Finally,
number of DSM-5 substance use disorder symptoms endorsed in the past year (e.g., “cravings
and urges to use the substance”, “not managing to do what you should at work, home, or school
because of substance use”) served as a measure of problematic substance use (range: 0-12, α =
0.95; Hasin et al., 2013).
Data Analysis
Missing Data
As a preliminary step, GAIN-I data from all young adults aged 18 to 25 (n = 50,914)
were checked for patterns of missingness. Missing data were negligible for a majority of selected
variables, but missingness was relatively high for number of previous treatment episodes (23.7%
missing), general crime scale (20.9% missing), sexual/gender minority (19.7% missing), family
history of a substance use problem (19.5% missing), and the dichotomous variable indicating
current enrollment in school and/or employment (18.0% missing). Data were also checked for
inconsistencies, and nine participants who reported age of first substance use greater than their
current age were recoded as missing for this variable. Following recommendations from the
GAIN Coordinating Center (Ives et al., 2010), a modified hot-deck method was used to impute
missing data. Cases were sorted by sex, race, age, and substance use disorder severity (number of
DSM-5 substance use disorder symptoms), and the modal value of the nearest 20 cases was used
to impute missing data values. We chose not to impute demographic or identity-based variables

25

(sex, race/ethnicity, and sexual/gender minority identity/behavior), and excluded cases with
missing on these variables (n = 10,156, 20.0%). The majority of these cases (96.8%) were
excluded due to missing data on sexual/gender minority identity and behavior, resulting in an
analytic sample of N = 40,758 young adults (see Table 2.2 for a comparison of the original and
imputed data sets).
Developing and Testing Classification Models
Three classification models of past year homelessness were estimated: (1) stepwise
logistic regression; (2) penalized (lasso) logistic regression; and (3) random forest classification.
Logistic regression is a well-established method that uses a maximum likelihood logistic
function to predict the probability (or odds) of a binary outcome occurring, given a set of
predictor variables (James et al., 2021). Stepwise logistic regression adds and removes predictor
variables using sequential forward and backwards selection, using Akaike Information Criterion,
resulting in a reduced set of predictor variables that contribute to optimal model fit. Stepwise
logistic regression was carried out using the stepAIC() function in the MASS R package
(Venables & Ripley, 2002).
Penalized logistic regression is similar to stepwise logistic regression, in that it results in
a reduced set of predictor variables by “shrinking” variable coefficients that do not significantly
contribute to prediction of the outcome (James et al., 2021). The level of shrinkage is determined
by finding an optimal shrinkage factor (lamda) across varying levels of an elastic net mixing
parameter (alpha) that ranges from 0 to 1. Setting alpha equal to 0 is known as ridge regression
(coefficients of less important predictors are shrunk nearly to zero), and setting alpha equal to 1
is known as lasso regression (coefficients for unimportant predictors are shrunk to exactly zero)
(Tibshirani, 1996). Results from five-fold cross-validation showed that an optimal lamda value

26

(𝜆 = 0.01) was produced when alpha was set equal to 1; therefore, lasso logistic regression was
used. A lasso model was then fit using 5-fold cross-validation in R (Jerome Friedman et al.,
2010).
The third model used was a random forest classification model. While logistic regression
and penalized regression are optimal for modeling linear associations between variables, they are
less useful for modeling complex non-linear relationships between variables and ranking
variables by their relative importance. Random forest classification models are an ensemble
machine learning method that aggregate many bootstrapped classification trees to predict a
binary/categorical outcome (Breiman, 2001). Each classification tree performs binary splits of
observations given a set of predictor variables, assigning observations to nodes where they most
likely belong given their status as homeless or housed; this is done internally using a metric
called the classification error rate (James et al., 2021). Each classification tree results in many
splits (“branches”) and corresponding nodes (“leaves”), with the most important predictors
determining splits at the top.
The novelty of random forests relies on randomly sampling a subset of predictor
variables at each split, so that the resulting trees are not predominated by a small set of important
predictor variables. The optimal number of predictors sampled at each split can be adjusted using
a tuning parameter mtry. The optimal mtry value was found using 10-fold cross-validation with
three repeats in R (Kuhn, 2008). We also adjusted the random forest model using different
numbers of classification trees (250, 500, 1,000, and 1,500) and using the model with the lowest
out of bag (OOB) error rate. Finally, we varied the minimum node size, ranging from 20 to 200
cases per node and evaluated the OOB error rate. Allowing more cases per node results in
smaller and more parsimonious classification trees. Mean decrease in the Gini index was used as

27

a measure of variable importance in the random forest. The Gini index is a measure of node
impurity (the ability to correctly classify cases at each split), ranging from zero to one (Menze et
al., 2009). The decrease in Gini index when a given variable is removed from a classification tree
is summed for each variable and averaged across all classification trees (James et al., 2021).
Final models used the randomForest package in R (Liaw & Wiener, 2002).
All initial models were trained using 80 percent of cases randomly selected to comprise
the training data set (n = 32,606) and evaluated on the remaining 20 percent of cases (n = 8,152).
A model-predicted probability of greater than 50 percent (better than random guessing) was used
to determine if a given individual in the training data set was correctly classified. Model
performance metrics included classification accuracy (the proportion of correctly identified
homeless or housed cases), sensitivity (the proportion of correctly identified homeless cases),
and specificity (the proportion of correctly identified stably housed cases) (Kassambara, 2017).
Receiver operating characteristic curves were also plotted to determine the area under the curve
(AUC), or the relative sensitivity and specificity at different predictive probability thresholds. An
AUC of 1 indicates perfect prediction, and an AUC of 0.50 indicates random guessing.
Initial models had good specificity (range: 95-96%) but poor sensitivity (range: 21-24%),
indicating that they performed well when correctly identifying stably housed cases, but poorly
when identifying homeless cases. This is a common problem when the outcome is imbalanced in
the training data set (22% experienced homelessness) (Byrne et al., 2020). Downsampling was
therefore used to make the training data set equally balanced on the homelessness outcome, and
involved retaining all homeless cases in the training data set (n = 7,248) and randomly selected
an equal number of stably housed cases to be retained in the training set (Kuhn, 2008). All
continuous variables in the downsampled training data set were standardized by mean centering

28

and dividing by the standard deviation. Therefore, model coefficients for continuous predictor
variables (e.g., depression symptoms) in the logistic regression results are interpreted as the odds
of experiencing homelessness, given one standard deviation increase in the predictor variable.
Results
Stepwise Logistic Regression
Parameter estimates showing the association between each predictor variable and
experiencing homelessness in the training data set are shown in Table 2.3. Nearly every variable
had a statistically significant association with past year homelessness. To aid in interpretation of
model results, ten variables with the largest unique association with past year homelessness (in
the log odds) are displayed in Figure 2.1. Variables associated with higher odds of homelessness
were: stimulant use as a primary substance (relative to cannabis), victimization, sexual/gender
minority status, multiracial (non-Hispanic) race/ethnicity, and number of substance use disorder
symptoms. Variables associated with lower odds of experiencing homelessness were: currently
working and/or in school, Hispanic/Latinx ethnicity, reporting alcohol as a primary substance
(relative to cannabis), married or living as married, and “other” substance use as a primary
substance (e.g., hallucinogens, inhalants, etc. relative to cannabis). Performance metrics of the
models’ ability to accurately classify homelessness in the test data are displayed in Table 2.4.
The model overestimated a 44.4% prevalence of homelessness (the actual prevalence was
22.7%). Overall classification accuracy was 66.4%, indicating that the model correctly classified
almost two-thirds of all cases in the test data. The model performed better at accurately
predicting homeless cases (sensitivity: 73.7%) than housed cases (specificity: 64.2%), with an
overall AUC of 76.0%.

29

Lasso Logistic Regression
Parameter estimates of the association between each predictor variable and homelessness
in the training data set using lasso logistic regression are also shown in Table 2.3. Note that the
lasso method does not produce confidence intervals or corresponding p values, but rather,
reduces model coefficients for non-significant variables to zero, essentially removing them from
the model. Compared to stepwise logistic regression, the lasso dropped seven more variables
from the model, and added one (co-occurring disorder), resulting in a more parsimonious model.
As shown in Figure 2.2, the predictor variables with the largest unique association with past year
homelessness were stimulant use as primary substance (relative to primarily using cannabis),
victimization, substance use disorder symptoms, depressive symptoms, family history of
substance use problems, sexual/gender minority status, and PTSD symptoms. Currently working
and/or in school, Hispanic/Latinx ethnicity, and alcohol as primary substance (relative to
cannabis) were associated with lower odds of experiencing homelessness. Model performance
was nearly identical to that of the stepwise logistic regression model. Classification accuracy was
66.5%, sensitivity was 73.7%, specificity was 64.4%; furthermore, the AUC was 75.9%.
Random Forest
Results of cross-validation revealed that the optimal random forest model (lowest out of
bag error rate) randomly selected five predictor variables at each split, with 500 trees, and a
minimum node size of 200. Variable importance on the training data set is shown in Figure 2.3.
The 10 most important variables in the model were: number of depression symptoms, co-
occurring disorder, number of PTSD symptom, number of substance use disorder symptoms,
victimization, currently working and/or in school, criminal behavior, past 90-day income, social

30

risk, and age. The final random forest model had a 64.4% classification accuracy, 76.4%
sensitivity, 60.8% specificity, and 68.6% AUC.
Discussion
The current study is the first to use supervised machine learning models to identify
important correlates of homelessness and compare model performance among young adults in
the year prior to entering substance use treatment. This research is important, as treatment-
engaged young adults are a vulnerable population who may be at heightened risk of experiencing
homelessness (DiGuiseppi, Davis, Leightley, et al., 2020; Green et al., 2020). In the current
study, 22% of young adults experienced homelessness in the 12 months prior to treatment. This
is twice as high as past-year homelessness estimates in the general young adult U.S. population
(Morton et al., 2018), and almost four times as high when only considering current homelessness
at a single point in time at treatment entry (Green et al., 2020). Furthermore, the use of multiple
predictive modeling techniques enhances efforts to identify individuals who are at greater risk of
experiencing homelessness at treatment entry. Performance was similar across models, with the
exception that random forest models had lower predictive accuracy across different probability
thresholds (area under the curve). Other differences emerged across models with regard to which
variables had a greater contribution to the outcome. These similarities and differences are
discussed below.
The present study identified important correlates of homelessness across six social-
ecological domains. These correlates are generally consistent with those found in previous
literature, with some notable exceptions. For example, demographic variables such as older age,
multi-racial (non-Hispanic) ethnicity, and sexual/gender minority status were consistently
associated with greater risk of homelessness in the logistic regression models, but appeared to be

31

less important in the random forest model. Such differences between models may in large part be
due to the different algorithms used. While logistic regression functions produce population
estimates of the unique association between each predictor variable and the outcome in an
additive manner, random forest classification models identify variables that are consistently
important in classifying individuals into one of two groups (e.g., by appearing at the top of many
decision trees; James et al., 2021). Further, by aggregating results of hundreds of decision trees,
random forest classification models may be capturing complex interactions between predictor
variables that were not examined in logistic regression models (Breiman, 2001).
Despite differences across models, some notable findings with regard to race and
ethnicity in the logistic regression models are worth discussing. While homelessness is more
prevalent among Black/African American young adults in the general population (Henry et al.,
2020; Morton et al., 2018), the odds of experiencing homelessness for African American young
adults in the current study were no different from non-Hispanic White young adults (a finding
that was consistent across models). This may reflect a higher risk for homelessness among non-
Hispanic White young adults when using clinical samples. Furthermore, consistent with previous
research with adolescents (DiGuiseppi, Davis, Leightley, et al., 2020), logistic regression
indicated that Hispanic/Latinx young adults had lower odds of experiencing homelessness than
non-Hispanic White young adults. While it is not entirely clear why homelessness may be less
prevalent among Hispanic young adults in treatment, greater social support in Hispanic families
and communities may play a protective role (Padilla et al., 2020; Stapley et al., 2021). However,
these results should be considered in light of Hispanic ethnicity not emerging as a particularly
important variable in the random forest model. It is possible that the random forest model was
able to capture the complexity of associations between variables, potentially implicating other

32

factors (discussed below) as more important risk and protective factors for past year
homelessness.
Among the variables in the family and peer relationships domain, variables strongly
associated with past-year homelessness were married or living as married (stepwise logistic
regression model), family history of a substance use disorder (lasso model), and the social risk
index (random forest model). It is likely that all three of these factors are important correlates of
homelessness (Morton et al., 2018; Tucker et al., 2021; Tyler & Schmitz, 2013; Whitbeck et al.,
1999), but their salience appears to depend on the modeling strategy used. Among the variables
in the education, employment and income domain, current involvement in school or employment
had a strong and consistent negative relationship with homelessness across all models. Stable
employment has been associated with successful treatment outcomes (Sahker et al., 2019), and
providing vocational and educational support services are promising intervention component in
samples involving unstably housed individuals and treatment samples (Ferguson & Xie, 2008;
Slesnick et al., 2018; Tanner-Smith et al., 2020). However, due to the cross-sectional nature of
the data, it is unknown whether being in school or employed acts as a protective factor, or
whether young adults experiencing homelessness are less likely to be involved in school or work
due to various barriers (Tierney et al., 2008). What is apparent from these results is that one’s
current activity in school and/or work appears to be more strongly associated with homelessness
than previous academic achievement (i.e., earning a high school diploma), which is a commonly
used predictor of homelessness among young adults (Brakenhoff et al., 2015; Heerde et al.,
2020; Morton et al., 2018; van den Bree et al., 2009). Notably, income emerged as an important
variable in the random forest (but not logistic regression) results. It is unclear why this was not

33

consistent across models, given the seemingly important role of personal income as a
determinant of stable housing (Morton et al., 2018).
Mental health, substance use, and victimization variables had consistently strong
associations with homelessness across all models. In fact, mental health, substance use disorder
symptoms, and victimization appeared among the five most important variables in the random
forest model results. This is not surprising, as high rates of depression and PTSD symptoms,
substance use disorders, and poly-victimization (e.g., physical, sexual, and emotional abuse)
exist among clinical and community samples of youth experiencing homelessness (Bender et al.,
2015; DiGuiseppi et al., 2022). Substance use variables may be particularly relevant correlates of
homelessness, given the current treatment-based sample. Results indicate that overall substance
use disorder severity, as well the primary type of substance used are significant correlates of
homelessness. For example, the stepwise logistic regression model indicated that compared to
young adults entering treatment for cannabis use, young adults entering treatment for stimulant
use had 45% higher odds of experiencing homelessness. This pattern was reversed for
individuals entering treatment primarily for alcohol use, as these individuals had 25% lower odds
of experiencing homelessness relative to individuals entering treatment for cannabis. This may
be reflective of specific types of substance use (stimulants and cannabis) that may be more
prevalent and serve specific functions among young adults who experience homelessness prior to
treatment, versus those who were stably housed. Stimulants have been reported to help youth
experiencing homelessness stay awake at night to avoid victimization and relieve hunger pains,
and cannabis has been cited as a popular alternative to “harder” drugs, as well as for its perceived
medicinal value (Hudson et al., 2009; Paul et al., 2020).

34

Limitations
There are several limitations of the current study. First, this study uses data collected at a
single point in time, and several variables could have co-occurred at the same time as young
adults’ experience of homelessness. Therefore, we cannot make conclusions about temporal
relationships between variables, and opt to frame the variables used to inform our models as
“correlates” rather than “predictors.” Secondly, although great care was taken in selecting
predictor variables from previous literature, it is possible that some important variables were
omitted from our analysis. Notably, we were not able to include a variable indicating
participants’ involvement in the foster care or child welfare system due to a high amount of
missing data on this variable. Third, data were collected via participant self-report, which may be
subject to recall bias. However, the large sample size and use of the GAIN-I as a validated
instrument may have helped reduce this risk.
Implications and Conclusion
Results of the current study have implications for theories of homelessness among young
adults, statistical methods, and clinical practice. Results may inform theories of youth
homelessness, such as the Risk Amplification Model and the Risk Amplification and Abatement
Model (Milburn et al., 2009; Whitbeck et al., 1999), which have largely been developed using
adolescent samples from street and shelter settings. More recent research has identified
significant correlates of homelessness among young adults using community samples (Morton et
al., 2018; Shelton et al., 2009; Tucker et al., 2021), but studies have yet to investigate these
correlates in treatment settings. Our results are consistent with the risk amplification model, as
we provide evidence that homelessness tends to co-occur with a host of adverse experiences
(victimization), mental health symptoms, and substance use problems. Our results also provide

35

support for the role of education and employment as an “abatement” factor, consistent with the
risk amplification and abatement model and empirical research (Milburn et al., 2009; Morton et
al., 2018). However, unlike these theories, we cannot make inferences about longitudinal
developmental processes or treatment outcomes given that these data were cross-sectional (see
Heerde et al., 2021 for an example using a life-course perspective). We conceptualize correlates
in our results as potential risk and protective factors for homelessness that may serve as
important variables in future research and interventions.
Secondly, the current study may inform future research that makes use of multiple
statistical methods and machine learning techniques to predict relevant outcomes in treatment
settings (Barenholtz et al., 2020; Davis et al., 2021, 2022). While logistic regression tended to
include demographic variables among the strongest correlates of homelessness (sexual/gender
minority status, Hispanic ethnicity, marital status), random forest results did not identify these
variables as highly important, and instead identified mental health, substance use disorder
symptoms, victimization, and current involvement in education and/or employment as most
important. Such differences in model results raise questions about who may be at greatest risk of
experiencing homelessness when interpreting the results of different models. We caution against
making recommendations for the “best” classification model to use for this purpose, and
encourage adopters of machine learning algorithms to become familiar with the strengths and
weaknesses of each approach (Barenholtz et al., 2020). For example, although it is assumed that
large sample sizes are needed for machine learning approaches, both logistic regression and
random forest can perform well using relatively small sample sizes (Hosmer & Lemeshow, 2000;
Qi, 2012).

36

Finally, the current study may inform the development of predictive models for practical
use in clinical settings. Perhaps the most exciting potential for the use of these techniques is
making predictions about individuals’ future risk of experiencing homelessness, as this allows
for the opportunity to prevent such homelessness experiences from occurring. Recent research
has used logistic and lasso regression to identify predictors of adolescents’ first episode of
homelessness following treatment, but this model’s performance has not been evaluated in
subsequent research or practice (DiGuiseppi, Davis, Leightley, et al., 2020). Experiencing
homelessness during and after treatment is a profoundly stressful event that can jeopardize an
individuals’ health and recovery from substance use disorders (Whipple et al., 2016; Zhang &
Slesnick, 2018). Continued efforts are needed in treatment settings, where the housing needs of
vulnerable individuals can be assessed, and supportive services can be provided.

37

Table 2.1 Participant Characteristics at Treatment Intake (N = 40,758)
Variable
Stably Housed
n = 31,662 (77.7%)
Homeless
n = 9,096 (22.3%)
Total
(N = 40,758)
Demographics

Age* 21.3 (2.4) 21.7 (2.3) 21.4 (2.4)
Female* 10,111 (31.9%) 9,096 (22.3%) 14,232 (34.9%)
Race/Ethnicity

African American (Non-Hispanic)* 2,987 (9.4%) 638 (7.0%) 3,625 (8.9%)
White (Non-Hispanic)* 19,231 (60.7%) 6,278 (69.0%) 25,509 (62.6%)
Multiracial (Non-Hispanic)* 1,596 (5.0%) 598 (6.6%) 2,194 (5.4%)
Multiracial (Hispanic) 909 (2.9%) 269 (3.0%) 1,178 (2.9%)
Hispanic/Latinx* 5,875 (18.6%) 1,009 (11.1%) 6,884 (16.9%)
Other (Non-Hispanic) 1,064 (3.4%) 304 (3.3%) 1,368 (3.4%)
Sexual/Gender Minority* 2,434 (7.7%) 1,400 (15.4%) 3,834 (9.4%)
Family and Peer Relationships

Family history of AOD problem* 23,106 (73.0%) 7,872 (86.5%) 30,978 (76.0%)
Married or living as married* 3,449 (10.9%) 799 (8.8%) 4,248 (10.4%)
(Unmarried with) one or more children* 7,992 (25.2%) 3,160 (34.7%) 11,152 (27.4%)
Social Risk Index* 11.7 (4.0) 13.4 (4.9) 12.1 (4.3)
Education, Employment & Income

Graduated high school 19,824 (62.6%) 5,625 (61.8%) 25,449 (62.4%)
Currently working and/or in school* 14,522 (45.9%) 2,133 (23.4%) 16,655 (40.9%)
Past 90 day income, in dollars* $1,876.4 ($7,432.1) $1,450.2 ($8,250.0) $1,781.3 ($7,624.2)
Victimization and Criminal Justice Involvement

Victimization* 3.4 (3.6) 5.8 (3.9) 3.9 (3.8)
Criminal Behavior (General Crime
Scale)* 1.0 (2.2) 2.1 (3.2) 1.3 (2.5)
Current criminal justice involvement* 24,738 (78.1%) 6,822 (75.0%) 31,560 (77.4%)
Mental Health

Depressive Symptoms Scale* 3.1 (3.3) 5.5 (3.5) 3.7 (3.5)
Traumatic Stress Scale* 2.6 (3.8) 5.5 (4.7) 3.3 (4.2)
Co-occurring disorder* 13,495 (42.6%) 6,701 (73.7%) 20,196 (49.6%)
Substance Use & Treatment

Age at first substance use* 14.2 (2.9) 13.3 (2.9) 14.0 (2.9)
Number of treatment episodes* 1.0 (2.7) 1.4 (2.5) 1.1 (2.7)
Substance use frequency* 28.6 (33.8) 40.6 (35.6) 31.2 (34.6)
SUD symptoms (Past year)* 5.2 (4.4) 8.0 (3.8) 5.8 (4.4)
Primary Substance Treatment Need

Alcohol* 10,289 (32.5%) 1,664 (18.3%) 11,953 (29.3%)
Cannabis* 9,920 (31.3%) 2,459 (27.0%) 12,379 (30.4%)
Stimulants* 3,624 (11.4%) 2,030 (22.3%) 5,654 (13.9%)

38

Opioids* 6,597 (20.8%) 2,565 (28.2%) 9,162 (22.5%)
Other 1,232 (3.9%) 378 (4.2%) 1,610 (4.0%)
* Significant group differences (p < 0.05) using Chi-square or t-tests

Note. Other (non-Hispanic) includes Asian, Native American, Native Hawaiian, Pacific Islander, Alaskan
Native and individuals identifying as another group; SUD = Substance Use Disorder

39

Table 2.2 Descriptive Statistics for Original and Imputed Data Sets

Original
(N = 50,914)

Imputed
(N = 40,758)
Variable M (SD) or N (%) % Missing M (SD) or N (%)
Demographics

Age 21.4 (2.4) 0.0%

21.4 (2.4)
Female 17,509 (34.4%) 0.2%

14,232 (34.9%)
Race/Ethnicity

African American (Non-Hispanic) 5014 (9.9%) 0.5%

3,625 (8.9%)
White (Non-Hispanic) 31384 (62.0%) 0.5%

25,509 (62.6%)
Multiracial (Non-Hispanic) 2739 (5.4%) 0.5%

2,194 (5.4%)
Multiracial (Hispanic) 1483 (2.9%) 0.5%

1,178 (2.9%)
Hispanic/Latinx 8325 (16.4%) 0.5%

6,884 (16.9%)
Other (Non-Hispanic) 1,715 (3.4%) 0.5%

1,368 (3.4%)
Sexual/Gender Minority 3,851 (9.4%) 19.7%

3,834 (9.4%)
Family and Peer Relationships

Family history of AOD problem 30,454 (74.3%) 19.5%

30,978 (76.0%)
Married or living as married 5,065 (10.1%) 1.4%

4,248 (10.4%)
(Unmarried with) one or more children 13,323 (26.6%) 1.5%

11,152 (27.4%)
Social Risk Index 12.10 (4.36) 12.3%

12.1 (4.3)
Education, Employment & Income

Graduated high school 30,474 (61.0%) 2.0%

25,449 (62.4%)
Currently working and/or in school 17,173 (41.1%) 18.0%

16,655 (40.9%)
Past 90 day income, in dollars $1,898 ($7,532) 11.2%

$1,781.3 ($7,624.2)
Victimization and Criminal Justice Involvement

Victimization 3.94 (3.98) 2.9%

3.9 (3.8)
Criminal Behavior (General Crime Scale) 1.53 (2.78) 20.9%

1.3 (2.5)
Current criminal justice involvement 37,241 (77.0%) 5.0%

31,560 (77.4%)
Mental Health

Depressive Symptoms Scale 3.55 (3.49) 5.1%

3.7 (3.5)
Traumatic Stress Scale 3.18 (4.20) 4.5%

3.3 (4.2)
Co-occurring disorder 23,339 (46.7%) 1.8%

20,196 (49.6%)
Substance Use & Treatment

Age at first substance use 14.06 (2.92) 3.8%

14.0 (2.9)
Number of substance use treatment episodes 1.16 (2.79) 23.7%

1.1 (2.7)
Substance use frequency 31.22 (34.53) 0.01%

31.2 (34.6)
Substance use disorder symptoms (Past year) 5.65 (4.44) 0.2%

5.8 (4.4)
Primary Substance Treatment Need

Alcohol 15,223 (29.9%) < 0.01%

11,953 (29.3%)
Cannabis 15,783 (31.0%) < 0.01%

12,379 (30.4%)
Stimulants 7,173 (14.1%) < 0.01%

5,654 (13.9%)

40

Opioids 10,785 (21.2%) < 0.01%

9,162 (22.5%)
Other 1,948 (3.8%) < 0.01%

1,610 (4.0%)
Homelessness (Past year) 10,833 (22.1%) 3.9% 9,096 (22.3%)
Note. Other (non-Hispanic) includes Asian, Native American, Native Hawaiian, Pacific Islander, Alaskan
Native and individuals identifying as another group

41

Table 2.3 Logistic Regression Predicting Past-Year Homelessness at Treatment Entry

Stepwise Logistic

Lasso
Variable OR 95% CI OR
Demographics

Age 1.12* (1.08, 1.16)

1.02
Female 1.09* (1.02, 1.17)

1.01
Race/Ethnicity (vs. White non-Hispanic)

African American (Non-Hispanic) 0.88 (0.74, 1.02)

Multiracial (Non-Hispanic) 1.30* (1.14, 1.45)

1.06
Multiracial (Hispanic) 0.79* (0.58, 1.01)

Hispanic/Latinx 0.63* (0.51, 0.75)

0.75
Other (Non-Hispanic)

Sexual / Gender Minority 1.32* (1.21, 1.44)

1.16
Family and Peer Relationships

Family history of AOD problem 1.26* (1.16, 1.36)

1.17
Married or living as married 0.76* (0.62, 0.89)

0.95
(Unmarried with) 1+ children 1.13* (1.03, 1.23)

1.11
Social Risk Index 1.15* (1.11, 1.19)

1.11
Education, Employment & Income

Graduated high school 0.88* (0.80, 0.96)

Currently working and/or in school 0.57* (0.49, 0.65)

0.60
Income, past 90 days ($500 units) 0.92 (0.84, 1.00)

Victimization & Criminal Justice Involvement

Victimization 1.32* (1.28, 1.36)

1.32
Criminal Behavior 1.17* (1.13, 1.21)

1.10
Current criminal justice involvement

Mental Health

Depressive Symptoms Scale 1.22* (1.16, 1.28)

1.20
Traumatic Stress Scale 1.16* (1.10, 1.22)

1.15
Co-occurring disorder

1.08
Substance Use & Treatment

Age at first substance use 0.93* (0.89, 0.97)

0.97
Num. substance use treatment episodes 0.97 (0.93, 1.01)

Num. substance use days (Past 90)

Substance use disorder symptoms 1.27* (1.23, 1.31)

1.26
Primary Substance (vs. Cannabis)

Alcohol 0.75* (0.65, 0.85)

0.87
Stimulants 1.45* (1.33, 1.57)

1.43
Opioids 0.85* (0.73, 0.97)

Other 0.79* (0.59, 0.98)

Intercept 1.16* (1.02, 1.30) 0.97

42

Note. Sample includes 14,496 young adults from down-sampled training data set; Lasso
coefficients models blank cells are equal to zero; OR = Odds Ratio; CI = Confidence Interval;
AOD = Alcohol or Drug
* Significant odds ratios at p < 0.50 (not calculated for Lasso models)

43

Table 2.4 Classification Model Performance Metrics

Stepwise
Logistic
Regression
Lasso Logistic
Regression
Random Forest
Classification
Classification Accuracy 66.4% 66.5% 64.4%
Sensitivity 73.7% 73.7% 76.4%
Specificity 64.2% 64.4% 60.8%
Area Under the Curve (AUC) 76.0% 75.9% 68.6%

44

Figure 2.1 Variables Most Strongly Associated with Past-Year Homelessness in Stepwise
Logistic Regression

45

Figure 2.2 Variables Most Strongly Associated with Past-Year Homelessness in Lasso Logistic
Regression

46

Figure 2.3 Variable Importance from Random Forest Gini Index

50 100 150 200 250
Age
Social Risk Index
Income, past 90 days ($500 units)
Criminal behavior
Currently working and/or in school
Victimization
Substance use disorder symptoms
Traumatic Stress Scale
Co-occurring disorder
Depressive Symptoms Scale
Mean Decrease in Gini Index
Random Forest Variable Importance

47

Chapter 3 (Study 2): Examining Bidirectional Associations Between Homelessness and
Substance Use Among Young Adults in Substance Use Treatment
Abstract
Background: Young adults receiving substance use treatment may be at greater risk for
experiencing homelessness, and substance use is highly prevalent among young adults
experiencing homelessness. However, research has yet to examine prospective, bidirectional
relationships between homelessness and substance use among young adults in treatment. The
present study aims to address this and investigate differences across demographic subgroups.
Methods: Latent growth curve models with structured residuals (LGC-SR) were used to
examine prospective, cross-lagged associations between homeless days and frequency of
substance use and associated problems between- and within-person. Models were stratified by
sex (male vs. female), race/ethnicity (non-Hispanic White vs. young adults of color), and sexual
and/or gender minority status (vs. cisgender heterosexual young adults). Results: Unconditional
bivariate LGC-SR models indicated significant decreases in homeless days (𝜇 𝑠𝑙𝑜𝑝𝑒 = -0.19, p =
0.046) and substance use frequency (𝜇 𝑠𝑙𝑜𝑝𝑒 1
= -6.19, p < 0.001) during treatment. Overall, no
significant cross-lagged effects emerged when controlling for baseline covariates (ps > 0.05). For
non-Hispanic White young adults, past 90-day homelessness and substance use were positively
correlated at treatment entry (𝜙 𝑠𝑡𝑎𝑛𝑑𝑎𝑟𝑑𝑖𝑧𝑒𝑑 = 0.12, p = 0.02), and greater substance use
frequency at treatment entry was associated with greater declines in homelessness during and
after treatment (𝜙 𝑠𝑡𝑎𝑛𝑑𝑎𝑟𝑑𝑖𝑧𝑒𝑑 = -0.14, p = 0.04). Conclusions: Despite declines in both of these
outcomes, homelessness and substance use frequency may not be reciprocally related for most
young adults in treatment. Continued efforts are needed to ensure that homelessness and
substance use are equitably reduced across subgroups of young adults in treatment.

48

Keywords: Young adults; substance use treatment; homelessness; housing; relapse

49

Introduction
Substance use and substance use disorders (SUDs) peak during young adulthood. For
example, 15.6 percent of young adults aged 18 to 25 in the United States had a past-year alcohol
use disorder in 2020, and 14.6 percent had an illicit drug use disorder (e.g., marijuana, cocaine,
opioid and prescription drug misuse) (Substance Abuse and Mental Health Services
Administration [SAMHSA], 2021). These prevalence rates are higher than those for adolescents
and older adults. Despite this, only 1.3 percent of young adults overall, and 4.4 percent of young
adults with a substance use disorder, receive substance use treatment each year (SAMHSA,
2021). Homelessness and housing instability are also far too common among this age group,
affecting an estimated one in 10 young adults in the United States each year (Morton et al.,
2018). Studies consistently show higher rates of substance use and SUD among young adults
experiencing homelessness relative to their stably housed peers (Chassman et al., 2022; Edidin et
al., 2012; Greene et al., 1997; Johnson et al., 2005). In addition, research suggests elevated risks
for homelessness among young adults in substance use treatment (DiGuiseppi et al., 2021; Green
et al., 2020). Given the possible co-occurrence of SUD and homelessness among young adults,
more research is needed to understand how these two phenomena relate over time, particularly
among vulnerable young adults in treatment. However, few studies have examined bidirectional
associations between substance use and homelessness among this age group using longitudinal
data. The present study addresses this using a sample of young adults receiving substance use
treatment nationwide, with the goal of informing clinical practice and policies to prevent the co-
occurrence of SUDs and homelessness in this age group.

50

The Co-Development of Substance Use and Homelessness Risks among Young Adults
The association between substance use and homelessness among young adults is a
complicated one that develops over time, and is likely bidirectional. Using a theoretical life stress
framework, researchers have posited that the negative effects of early life stressors, including
family dysfunction, homelessness, and victimization accumulate over time, leading to greater
risk of developing SUDs in young adulthood (Tyler & Schmitz, 2020). According to the Risk
Amplification Model (RAM), a majority of young people experiencing homelessness report
adverse childhood experiences and family dysfunction, which contribute to youths’ initial
experiences of homelessness (Whitbeck et al., 1999). Though substance use may or may not
contribute to young people’s initial experience of homelessness (Mallett et al., 2005), substance
use may increase while homeless due to greater affiliation with substance using peers and as a
coping mechanism (Barman-Adhikari et al., 2018; DiGuiseppi, Davis, Christie, et al., 2020;
Whitbeck et al., 2009). In turn, substance use and exposure to other high-risk situations while
homeless may contribute to increased risk for victimization (e.g., verbal, physical, and sexual
assault) and further progression of substance use and mental health disorders, thereby amplifying
young people’s pre-existing childhood risks (Whitbeck et al., 1999). Left untreated, SUDs may
interfere with young people’s ability to secure and maintain stable housing. However, the few
longitudinal studies available offer mixed support for this, with some reporting that substance
use is negatively associated with long-term housing stability (Rosenthal et al., 2007; Roy et al.,
2014; Slesnick et al., 2013; Tevendale et al., 2011), and others reporting that substance use is not
significantly related to long-term housing outcomes after accounting for structural factors (e.g.,
lack of affordable housing), race, and ethnicity (Braciszewski et al., 2016). Still, other research
suggests that higher “vulnerability scores” (which include substance use as part of standardized

51

housing assessments), as well as race and ethnicity, both play a role in youths’ unsuccessful exits
from housing placements (Eric Rice et al., 2018; Hill et al., 2022; Hsu et al., 2021).
While theoretical models such as the RAM may be helpful for hypothesizing
bidirectional associations between substance use and homelessness over time, it was largely
developed using adolescent samples recruited from street and shelter locations. Prospective
associations between substance use and homelessness may differ for treatment-engaged young
adults. Mainly, receiving treatment can be a positive life event that alters individual trajectories
of substance use, and possibly homelessness (Coren et al., 2016; DiGuiseppi et al., 2021; Zhang
& Slesnick, 2018). Coping skills and behavioral changes acquired during treatment may
decouple problematic associations between substance use and homelessness (Slesnick et al.,
2021). However, few studies have tested this empirically among young adults. In a latent class
growth analysis involving 270 youth experiencing homelessness receiving one of three substance
use interventions, 30% evidenced decreased substance use paired with increased social stability
(including stable housing), and another 32% evidenced low but stable substance use paired with
increasing social stability (Zhang & Slesnick, 2018). This suggests that a majority of young
adults who receive substance use treatment may benefit by both decreasing their substance use
and increasing their housing stability. However, more research is needed to examine the potential
bidirectional relationships between substance use and homelessness over time.
Bidirectional Risks for Substance Use and Homelessness During Treatment
Despite the benefits of treatment, relapse prevention models suggest that stressful life
events such as experiencing homelessness during or after treatment can trigger a relapse to
substance use, as negative affect and other pressures to use may overpower an individuals’
ability to cope (Larimer et al., 1999; Patterson et al., 2022). Another possibility is that

52

homelessness may follow increased substance use during or after treatment. This is consistent
with perspectives that view homelessness as a consequence of substance use disorder symptoms
(Baer et al., 2003; Hasin et al., 2013). Substance use problems may interfere with young adults’
ability to fulfill responsibilities at work, home, or school, or may cause interpersonal problems
which may lead to loss of housing.
To our knowledge, only one study has investigated bidirectional associations between
substance use and homelessness among treatment-engaged youth . Using a large nationwide
sample of over 20,000 treatment-engaged adolescents, Davis and colleagues (2019) found that a
greater number of days spent homeless prior to treatment was associated with within-person
increases in substance use and post-traumatic stress disorder (PTSD) symptoms at the end of
treatment (three months later). These increases in both substance use and PTSD, in turn, were
associated with an increased number of days spent homeless in the nine-month period following
treatment (three and six months later). Thus, more prolonged homelessness experiences prior to
treatment may forebode less desirable treatment outcomes (increases in substance use), which
may in turn lead to more prolonged experiences of homelessness after treatment. Importantly,
Davis and colleagues (2019) used statistical models that disaggregate between-person change
(i.e., average increases in one construct predicting subsequent average increases in another
construct) from within-person change (i.e., change from an individual’s previous mean level
predicting subsequent change in an individual’s mean level of another construct). These two
components of change are important to differentiate, as they may aid our understanding of how
bidirectional risks between homelessness and substance use unfold for treatment-engaged young
adults over time.

53

Examining Subgroup Differences
In addition to establishing overall patterns of bidirectional risks between substance use
and homelessness among treatment-engaged young adults, there is a need to investigate whether
such risks develop differently across demographic subgroups. Mainly, it is well known that
Black, Latino, or sexual and/or gender minority (i.e., lesbian, gay, bisexual, transgender, or
queer) young adults are more likely to experience homelessness than non-Hispanic White and
cisgender heterosexual young adults in the U.S. (Henry et al., 2020; Morton et al., 2018). Black
and Latino young adults are also less likely to have favorable exits from homelessness (Hill et
al., 2022). There is also some evidence that male adolescents are at greater risk of experiencing
homelessness following substance use treatment, compared to female adolescents (DiGuiseppi,
Davis, Leightley, et al., 2020). Such disparities may also be present when examining temporal
associations between homelessness and substance use, but have yet to be examined. This
deserves further investigation in order to address potential disparities among vulnerable
subgroups in treatment.
Study Aims and Hypotheses
The aim of the present study was to extend previous research with adolescents by
examining prospective, bidirectional associations between substance use and homelessness in a
nationwide sample of treatment-engaged young adults. We extend this work further by
examining how these bidirectional risks may unfold differently across subgroups (based on sex,
race/ethnicity, and sexual and/or gender minority status). We proposed three hypotheses. Our
first hypothesis was informed by developmental models of the negative effects of youth
homelessness (Whitbeck, 2009; Whitbeck et al., 1999). Specifically, we hypothesized that more
days spent homeless prior to treatment (reported at treatment entry) would be associated with

54

less desirable treatment outcomes (i.e., increased substance use and problems) at the end of
treatment. Our second hypothesis was informed by relapse prevention models (Larimer et al.,
1999), and posited that increases in homelessness during or after treatment would be associated
with subsequent increases in substance use. Our third hypothesis was guided by substance use
disorder symptom models (Hasin et al., 2013), and assumes that homelessness can be a negative
consequence of problematic substance use. Therefore, we hypothesized that increases in
substance use during or after treatment would be associated with subsequent increases in days
spent homeless. We proposed no hypothesis for our subgroup analyses, and rather left this as
exploratory given the lack of previous research on differences in bidirectional risks during
treatment.
Methods
Data Source and Participants
Data come from pooled Global Appraisal of Individual Needs (GAIN) data set. Pooled
GAIN data include comprehensive intake and follow-up assessment data from individuals
receiving substance use treatment at over 1,500 sites in over 49 states in the U.S. Data for the
current study were restricted to young adults aged 18 to 25 at their initial treatment intake
assessment (N = 3,717). Participants received treatment at 160 clinical sites in 39 U.S. states
from 2005 to 2012. Participant descriptive statistics can be viewed in Table 3.1. The average age
was 20.1 years old (SD = 2.2), and a majority were male (72.1%). The largest racial/ethnic
groups included non-Hispanic White (36.9%), Hispanic/Latinx (33.4%), and non-Hispanic
African American (15.9%). Nearly 80 percent of participants had some involvement in the
criminal justice system at treatment intake, and a majority were receiving treatment on an
outpatient or intensive outpatient basis (87.5%).

55

Measures
Participants completed the GAIN-Initial (GAIN-I) at their treatment intake assessment,
and quarterly follow-ups using the GAIN-M90 at 3-months, 6-months, and 12-months post-
intake. Treatment is typically delivered within the first 12-weeks, represented by the time in
between the initial assessment and the 3-month follow-up. The GAIN-I is a comprehensive
assessment battery used for diagnosis, treatment planning, and program evaluation, relies on self-
report data, and is administered by a trained interviewer (Dennis et al., 2008). The GAIN battery
of measures include consistency checks to ensure measurement validity, and interviewers
complete training, certification, and clinical supervision at the GAIN coordinating center or
locally at their clinical site. The GAIN-I core measures take approximately 60 to 90 minutes to
administer, and the GAIN-M90 is a briefer format containing a subset of the GAIN-I, taking
approximately 45 to 60 minutes to complete. Most items in the GAIN-M90 contain a past 90-day
assessment timeframe, which captures change during the first 12-weeks of treatment, and in 90-
day increments thereafter.
Dependent Variables
Homeless Days. At each assessment, participants were asked, “In the past 90 days, on
how many days have you been homeless or had to stay with someone else to avoid being
homeless?” Response options ranged from 0 to 90 days (see Table 3.2 for descriptive statistics).
Substance Use Frequency. The GAIN Substance Frequency Scale (SFS) was used as a
measure of substance use and associated negative consequences (Ives et al., 2010). The SFS is a
main GAIN scale and represents the average percentage of days (in the past 90) of any alcohol or
other drug use (i.e., cannabis, cocaine, heroin and prescriptions opioids, etc.) and/or problematic

56

use (i.e., two items assessed days getting drunk or high most of the day, and days in which
substance use caused problems or interfered with responsibilities at work, school, or home). The
SFS has evidenced good internal consistency (Cronbach’s α = 0.81), test-retest reliability, and
concordance with timeline follow-back methods and urine tests (Ives et al., 2010; Modisette &
Dennis, 2022).
Demographics and Control Variables
The following variables were measured at the initial intake assessment and were used
descriptively or as covariates in multivariate models. Demographic variables included age, sex,
race, Hispanic ethnicity, and sexual and gender minority status (lesbian, gay, bisexual,
transgender identity and/or behavior). Dichotomous variables were created for sex (0 = male, 1 =
female), sexual/gender minority (1 = sexual/gender minority identity/behavior, 0 = cis-gender
heterosexual identity/behavior), and race/ethnicity (1 = non-Hispanic White, 0 = Black,
multiracial, Hispanic/Latinx, or Asian, Native American or Alaska Native [i.e., young adults of
color]). Criminal Justice Involvement was a dichotomous variable indicating involvement in at
least one of 12 different ways (1 = awaiting trial, on probation/parole, house arrest, or currently
detained in jail, detention, or prison) or 0 = no criminal justice involvement. Level of Care was
recoded as a binary variable indicating treatment care placement (1 = receiving inpatient/residual
treatment, 0 = receiving outpatient, intensive outpatient, or post-residential continuing-care).
Lifetime emotional, physical, and sexual victimization were assessed using the GAIN
General Victimization Scale (GVS; Titus et al., 2003). The GVS is a count of each type of
victimization plus six traumagenic factors (e.g., duration, relationship to perpetrator, negative
social reactions to disclosure of the event) and four fear of re-victimization items (range: 0-15;
Cronbach’s α = 0.88). The GAIN Social Risk Index (SRI; Dennis et al., 2008) includes sum

57

scores for seven items used to measure how many people (0 = none, 1 = a few, 2 = some, 3 =
most, 4 = all) participants associate with who engage in risky substance use behaviors (e.g.,
binge drink, use drugs, are involved in illegal activities) versus prosocial behaviors (employed, in
school, in treatment or recovery). The SRI ranged from 0 to 28, with higher scores indicating
greater social risk (internal consistency is not calculated for this summative scale).
Data Analytic Plan
Bivariate latent growth curve modeling with structured residuals (LGC-SR) was the
primary modeling approach used for examining bidirectional associations between substance and
homelessness over time. The bivariate LGC-SR is a relatively recent model that offers several
advantages over previous methods to investigate bidirectional effects, such as the autoregressive
cross-lagged panel model (Littlefield et al., 2021). The bivariate LGC-SR is advantageous as it
allows for the disaggregation of between- and within-person change by specifying auto-
regressive and cross-lagged paths among the residual item variances, producing estimates of
within-person change over and above the between-person changes captured by the underlying
growth models (Curran et al., 2014). Our model building strategy was guided by Curran and
colleagues (2014) and recent work examining between- and within-person cross-lagged
associations between substance use and homelessness using adolescent GAIN data (Davis et al.,
2019).
We began by modeling change in substance use frequency and homeless days as separate
latent growth curve models. As shown in Table 3.2, the distribution of homeless days was
skewed and kurtotic, reflecting a preponderance of zeros (a large proportion of individuals that
did not experience homelessness at each time point). To address this, we attempted to model
trajectories of homelessness using a two-part latent growth curve model, which estimates

58

separate trajectories for change for the probability of experiencing homelessness (binary
outcome) and change in the number of days experienced homelessness (count or continuous
outcome) (Liu & Powers, 2007). However, as is common with these models (Ferrer et al., 2016),
the two-part model did not converge. Therefore, we modeled homelessness days as a continuous
outcome using a robust maximum likelihood (MLR) estimator, which produces conservative
parameter estimates of latent change over time for continuous outcomes that are not normally
distributed (Lai, 2018). The distribution of the substance use frequency scale was approximately
normal at each time point (see Table 3.2) and was modeled as continuous using a MLR estimator
as well.
A series of models were fit to characterize the functional form of change in each
outcome, starting with no growth (intercept only), linear change, and bi-linear (piecewise)
change models. Previous research using GAIN data with adolescents has found that a bi-linear
model optimally fits substance use frequency data, indicating significant declines in substance
use frequency during treatment, followed by slight increases post-treatment (Davis et al., 2019).
Furthermore, Davis and colleagues (2019) found that a linear model optimally fit homelessness
data, indicating non-significant increases in homelessness days during and after treatment. The
best fitting model in the current study was determined using scaled chi-square difference tests
based on log likelihood and scaling correction factors from the MLR estimator (Satorra &
Bentler, 2001). Model fit was also assessed using other common indices including root-mean-
square error of approximation (RMSEA), comparative fit index (CFI), Tucker-Lewis index
(TLI), standardized root mean square residual (SRMR). Next, autoregressive paths among
residuals were specified to estimate wave-to-wave stability in each outcome that was
unexplained by the underlying growth factors. Models specifying autoregressive paths as freely

59

estimated over time—or constrained to be equal during the post-treatment phase—were
compared.
After identifying the optimal functional form for each growth model separately, a series
of bivariate LGC-SR models were specified to find the best fitting autoregressive, cross-lagged
structure between the time-specific residuals of substance use frequency and homeless days. We
first connected the two growth models by covarying the latent intercept and slope factors and
within-time residual covariances between each outcome; residual covariances were freely
estimated at baseline, and constrained to be equal at the 3-, 6-, and 12-month follow-ups. Cross-
lagged paths were then added by regressing substance use frequency on previous measures of
homeless days, and regressing homeless days on previous measures of substance use frequency.
Next, a series of nested models were carried out, starting with a freely estimated model, and
gradually imposing equality constraints on autoregressive and cross-lagged paths (See Table 3.3
for model comparisons). Because we were interested in differential change during and after
treatment, equality constraints were gradually imposed on autoregressive and cross-lagged paths
in the post-treatment phase (3-month to 12-month follow-up). Mplus version 8.8 was used for all
analyses (Muthén & Muthén, 1998-2022).
Predictors of Latent Growth Processes and Subgroup Differences
After finding the best fitting unconditional bivariate LGC-SR model, demographic
variables, criminal justice involvement, victimization, and social risk were added as exogenous,
time-invariant predictors of latent growth factors. Finally, six additional models were run to test
whether prospective bidirectional associations between homelessness and substance use
frequency differed across subgroups. This involved stratifying the conditional model by sex

60

(male vs. female), race/ethnicity (non-Hispanic White vs. young adults of color), and
sexual/gender minority status (sexual/gender minority vs. cis-gender heterosexual).
Missing Data
Pooled GAIN data include client assessment data from a large number of sites, each with
their own protocol for including optional GAIN items. As shown in Table 3.1, baseline
covariates with the highest rates of missing at intake were level of care (4.6%) and the social risk
index (3.2%). Another source of missingness is due to attrition. Treatment sites that use the
GAIN are encouraged to complete follow-up assessments at the 3-month follow-up, given
important changes in behavior during treatment (83% retained). Accordingly, there was greater
attrition at the 6-month (78% retained) and 12-months follow-ups (49% retained), which was
largely due to lack of funding and resources to carry out these later follow-ups. Multivariate
logistic regression models revealed that non-Hispanic White race (OR = 1.41, 95% CI: 1.22,
1.64) and receiving inpatient/residential level of care (OR = 1.50, 95% CI: 1.14, 1.98) was
associated wth greater odds of missingness at the 12-month follow-up. Conversely, female sex
(OR = 0.82, 95% CI: 0.69, 0.97), greater social risk (OR = 0.98, 95% CI: 0.97, 0.99) and greater
substance use frequency at baseline (OR = 0.98, 95% CI: 0.98, 0.99) were associated with lower
odds of missingness at the 12-month follow-up. We carried out analyses with the conditional
missing at random assumption, by controlling for these covariates in the final conditional models
(Enders, 2022). To address missingness on the baseline covariates, final conditional models used
multiple imputation with 100 data sets using Bayesian analysis in Mplus (Enders, 2022; Muthén
& Muthén, 1998-2022). Sex (n = 1 missing) and sexual/gender minority status (n = 61 missing)
were not imputed, since this resulted in different sample sizes in multiply imputed data sets,

61

preventing stratified subgroup analyses. Therefore, these 61 cases were excluded from the final
conditional models.
Results
Descriptive Statistics of Substance Use Frequency and Homelessness Days
Descriptive statistics for our two outcome variables measured at baseline, and the 3, 6,
and 12-month follow-ups are displayed in Table 3.2. The SFS scale score at baseline was 12.37
(SD = 16.12), indicating that participants used substances or had problems resulting from their
use on an average of 12 percent of days in the past 90 days at treatment intake. This was
followed reductions in substance use frequency that ranged from an average of 5.78 (SD = 11.21)
at the 3-month follow-up to 6.89 (SD = 13.19) at the 12-month follow-up. At baseline, 387
participants (10%) reported at least one day spent homeless in the 90 days prior to treatment
intake, with an overall mean of 3.56 days (SD = 14.64). The average days spent homeless
decreased slightly over the follow-up period from 3.33 days (SD = 14.64) at the 3-month follow-
up to 2.96 days (SD = 14.57) at the 12-month follow-up.
Substance Use Frequency Growth Trajectory
The best fitting unconditional growth model for substance use frequency followed a bi-
linear growth pattern with equality constraints imposed on residuals item variances over time
(scaled 
2
= 66.20, 
2
= 26.87, df = 4, p < 0.001, RMSEA = 0.04, CFI = 0.96, TLI = 0.94, SRMR
= 0.04). Fit was better than a no growth (intercept only) model (scaled 
2
= 640.93, 
2
= 352.16,
df = 8, p < 0.001, RMSEA = 0.11, CFI = 0.36, TLI = 0.52, SRMR = 0.12), and a linear model
with freely estimated residual item variances (scaled 
2
= 465.66, 
2
= 246.38, df = 5, p < 0.001,
RMSEA = 0.11, CFI = 0.55, TLI = 0.46, SRMR =0.09). (A bilinear model with freely estimated

62

residual variances could not be identified.) A comparison of fit indices across models can be seen
in Table 3.3. The model estimated an average of 12 percent of days of substance use and/or
problems in the past 90-days at treatment entry (𝜇 𝑖𝑛𝑡 = 12.37, p < 0.001), followed by a
significant decrease by about half during treatment (𝜇 𝑠𝑙𝑝 1
= -6.24, p < 0.001), and a small, non-
significant increase post-treatment (𝜇 𝑠𝑙𝑝 2
= 0.15, p = 0.14). All growth factor variances were
significant (p < 0.05), indicating unexplained variance in growth factors. There were significant
negative correlations between intercept and slope factors, indicating that those who entered
treatment with greater substance use frequency had steeper decreases in substance use frequency
during treatment (𝜌 [𝑖𝑛𝑡 ,𝑠𝑙𝑝 1]
= -0.79, p < 0.001) and after treatment (𝜌 [𝑖𝑛𝑡 ,𝑠𝑙𝑝 2]
= -0.27, p <
0.001). The correlation between during-treatment and post-treatment slopes was positive and
marginally significant (𝜌 [𝑠𝑙𝑝 1,𝑠𝑙𝑝 2]
= 0.20, p = 0.05). Adding autoregressive paths to the residuals
that were freely estimated over time did not improve model fit (scaled 
2
= 82.10, 
2
= 43.67, df
= 3, p < 0.001, RMSEA = 0.06, CFI = 0.93, TLI = 0.85, SRMR = 0.04), but imposing equality
constraints on autoregressive paths post-treatment resulted in simplified the model without worse
fit (scaled 𝜒 Δ
2
(1) = 0.91, p = 0.34).
Homelessness Growth Trajectory
A linear model of homelessness with equality constraints on residuals over time fit the
data reasonably well (scaled 
2
= 168.78, 
2
= 30.23, df = 7, p < 0.001, RMSEA = 0.03, CFI =
0.87, TLI = 0.89, SRMR = 0.06), and was an improvement over a no growth model (scaled 
2
=
191.71, 
2
= 45.6, df = 8, p < 0.001, RMSEA = 0.04, CFI = 0.80, TLI = 0.85, SRMR = 0.07) and
was not significantly worse fitting than a model without equality constraints on the residuals
(scaled 𝜒 Δ
2
(3) = 4.35, p = 0.11). A bi-linear model appeared to have better model fit (scaled 
2
=

63

41.72, 
2
= 3.30, df = 4, p = 0.51, RMSEA = 0.00, CFI = 1.00, TLI = 1.00, SRMR = 0.02), but
was not identified when specifying autoregressive paths among the residuals. Given this, and the
apparently linear decline when visually inspecting observed mean trajectories, a linear model of
homeless days was retained. Comparative fit indices across these models can be viewed in Table
3.4. The linear growth model of homelessness estimated an initial mean of 3.46 homelessness
days in the past 90 days at treatment entry, and a significant mean decrease in homelessness days
during treatment (𝜇 𝑠𝑙𝑝 1
= -0.22, p = 0.02). Variances for the intercept and slope factors were also
significant. There was a significant negative correlation between the intercept and slope factors
(𝜙 [𝑖𝑛𝑡 ,𝑠𝑙𝑝 ]
= -0.61, p < 0.001), indicating that individuals with more homelessness days prior to
treatment entry tended to have steeper decreases in homeless days during treatment. Adding
autoregressive paths to the residual variances resulted in better fit, relative to a model without
autoregressive paths among the residuals (scaled 
2
= 66.78, 
2
= 25.49, df = 4, p < 0.001,
RMSEA = 0.04, CFI = 0.88, TLI = 0.82, SRMR = 0.05). Finally, constraining the autoregressive
path to be equal post-treatment did not degrade model fit (𝜒 Δ
2
(1) = 0.49, p = 0.48), and this final
univariate LGC-SR model of homelessness was retained.
Unconditional Bivariate LGC-SR Model
After fitting separate unconditional growth curve models of substance use frequency and
homeless days, we next fit an unconditional, bivariate LGC-SR for these two outcomes. Results
of nested model comparisons to arrive at the best fitting model can be found in Table 3.5. The
best fitting model was one in which autoregressive paths for homelessness days and substance
use frequency were constrained to be equal post-treatment, and all cross-lagged paths were
constrained to be equal post-treatment (scaled 
2
= 78.86, 
2
= 25.20(10), p = 0.01, RMSEA =
0.02, CFI = 0.99, TLI = 0.96, SRMR = 0.03). This unconditional model revealed no significant

64

cross-lagged associations between substance use frequency and days spent homelessness, either
between or within-person. Initial substance use frequency appeared to be associated with steeper
declines in days spent homeless between-persons, but this was marginally significant
(𝜙 𝑠𝑡𝑎𝑛𝑑𝑎𝑟𝑑𝑖𝑧𝑒𝑑 = -0.10, p = 0.06).
Final Conditional Model Results
Using multiple imputation and controlling for demographics, criminal justice
involvement, level of care, victimization, and social risk, fit of the final conditional model was
excellent ( 
2
= 51.35[31], RMSEA = 0.01, CFI = 0.99, TLI = 0.97, SRMR = 0.02). Parameter
estimates for between and within-person effects are shown in Table 3.6, and illustrated in Figure
3.1. Our first hypothesis was not supported. The correlation between the homelessness intercept
factor and the substance use frequency slope factor during treatment was not significant
(𝜙 𝑠𝑡𝑎𝑛𝑑𝑎𝑟𝑑𝑖𝑧𝑒𝑑 = 0.06, p = 0.27), suggesting that, between-persons, number of days spent
homelessness at treatment entry was not associated with change in substance use frequency
during treatment. Furthermore, controlling for prior levels of homelessness and substance use
frequency within-person, number of days spent homeless at treatment entry was not associated
with increased substance use frequency (relative to one’s own average) at the end of treatment
(𝜌 HmlsBL → SFS3m
= 0.45, SE = 0.36, p = 0.22).
Our second hypothesis was also not supported. Correlations between slope factors for
homeless days and substance use frequency were not significant, indicating that, between-
persons, change in the number of days spent homeless throughout the study period was not
associated with change in substance use frequency during treatment (𝜙 𝑠𝑡𝑎𝑛𝑑𝑎𝑟𝑑𝑖𝑧𝑒𝑑 = 0.02, p =
0.73) or after treatment (𝜙 𝑠𝑡𝑎𝑛𝑑𝑎𝑟𝑑𝑖𝑧𝑒𝑑 = 0.21, p = 0.15). These cross-lagged associations were

65

not significant within-person either. Change in days spent homeless immediately post-treatment
(relative to one’s own average) were not associated with increases substance use frequency
(relative to one’s own average) at the 6-month follow-up (𝜌 Hmls3m → SFS6m
= -0.004, SE = 0.02, p
= 0.86); nor were days spent homeless at the 6-month follow-up associated with within-person
increases in substance use frequency at the 12-month follow-up (𝜌 Hmls6m → SFS12m
= -0.004, SE
= 0.02, p = 0.86).
Finally, our third hypothesis was not supported. As indicated by the non-significant
correlations between slope factors (all ps > 0.05), between-person change in substance use
frequency during and after treatment was not associated with change in days spent homeless
throughout the study. Furthermore, substance use frequency immediately post-treatment was not
associated with more days spent homeless (within-person) at the 6-month follow-up
(𝜌 SFS3m → Hmls6m
= 0.06, SE = 0.06, p = 0.26); nor was substance use frequency at the 6-month
follow-up associated with greater homeless days (within-person) at the 12-month follow-up
(𝜌 SFS6m → Hmls12m
= 0.06, SE = 0.06, p = 0.26).
Subgroup Differences
Sex
Having examined reciprocal associations between homeless days and substance use
frequency among the entire sample, we next estimated separate models for important subgroups.
Our first subgroup analysis estimated the final conditional model separately for males and
females. As shown in Table 3.7, the model for males (n = 2,641) found no evidence of
prospective, bidirectional relationships between homelessness and substance use frequency either

66

between-person or within-person. Homeless days and substance use frequency were also not
reciprocally related over time for females between- or within-person (n = 1,015).
Race / Ethnicity
Final models stratified by race and ethnicity are shown in Table 3.8. For non-Hispanic
White young adults (n = 1,347), homelessness and substance use intercept factors were
positively related (𝜙 𝑠𝑡𝑎𝑛𝑑𝑎𝑟𝑑𝑖𝑧𝑒𝑑 = 0.12, p = 0.02), indicating that, between-persons, more days
spent homeless at treatment entry tended to be associated with greater substance use frequency at
treatment entry. There was also a negative association between the substance use intercept factor
and the homelessness slope factor (𝜙 𝑠𝑡𝑎𝑛𝑑𝑎 𝑟 𝑑𝑖𝑧𝑒𝑑 = -0.14, p = 0.04), indicating that for non-
Hispanic White young adults, greater substance use at treatment entry was associated with a
steeper decline in days spent homeless between-persons throughout the study. For young adults
of color (n = 2,309), there were no significant cross-lagged associations between days spent
homeless and substance use frequency throughout the study, either between- or within-person.
Sexual / Gender Minority Status
Model results stratified by sexual/gender minority identity and/or behavior are shown in
Table 3.9. For cisgender and heterosexual young adults (n = 3,388), there were no significant
cross-lagged associations between homeless days and substance use frequency. There were also
no significant cross-lagged associations for sexual/gender minority young adults (n = 268).
Interestingly, post-treatment homelessness autoregressive parameters were not significant
(𝜌 Hmls3m → Hmls6m
= 0.15, SE = 0.12, p = 0.18), indicating that for sexual/gender minority young
adults, within-person changes in homelessness were unstable after treatment. Furthermore, post-
treatment autoregressive parameters for substance use frequency were significant and negative

67

(𝜌 SFS3m → SFS6m
= -0.92, SE = 0.34, p = 0.01), indicating that within-person increases in
substance use frequency post-treatment were associated with decreased substance use frequency
within-person at the following timepoint.
Discussion
The present study examined prospective, bidirectional associations between
homelessness and substance use and/or problems for a vulnerable population of young adults
receiving substance use treatment in the United States. We found little evidence to support our
hypotheses that these two phenomena are reciprocally related over time for treatment-engaged
young adults, either at a general between-person level, or when examining within-person change.
After controlling for a variety of important factors (e.g., demographics, victimization history,
etc.) there was little overall support for our first “risk-amplification” hypothesis that more days
spent homeless prior to treatment would result in worse treatment outcomes (i.e., greater
substance use frequency at the end of treatment; Whitbeck et al., 1999). This null finding is
inconsistent with previous research with treatment-engaged adolescents, which found that greater
homeless days prior to treatment was related to increased substance use and PTSD symptoms,
(within-person) during treatment (Davis et al., 2019). When examining subgroup differences in
the current study, we actually found evidence to support an opposite hypothesis for non-Hispanic
White young adults only. For this group, it appears that greater substance use frequency at
treatment entry is associated with steeper decreases in days spent homeless during and after
treatment. The reason(s) behind this are unclear, but this may be due to the fact that
homelessness and substance use frequency were strongly related among non-Hispanic White
young adults prior to treatment. Since individuals who entered treatment with higher levels of
homelessness and substance use frequency tended to have the most dramatic declines in these

68

outcomes during treatment, it follows that those who enter treatment with more substance use
and/or problems may also be more likely to attain stable housing over time. This may not occur
for other subgroups for whom substance use and homelessness are less strongly related prior to
treatment. Thus, we can conclude from these results that greater housing stability may be a
secondary benefit of treatment for young adults who enter treatment with higher-than-average
(between-person) substance use and/or problems. In the present study, this happened to be the
case for non-Hispanic White young adults.
We found no support for our second “relapse prevention” hypothesis, which posited that
homelessness during or after treatment would be associated with increased frequency of
substance use and/or problems at the next timepoint (Larimer et al., 1999). Although
homelessness was generally stable post-treatment (as indicted by the significantly positive
autoregressive parameters across groups), there were no significant cross-lagged effects leading
from homelessness to substance use frequency after treatment. Thus, from a within-person
perspective, an increase in days spent homeless (relative to one’s own average) during or after
treatment may not in itself trigger a “relapse” in the form of increased substance use and related
problems. This seems inconsistent with the relapse prevention model, which states that “high-
risk situations” can lead to relapse during or after treatment (Larimer et al., 1999). Clearly,
homelessness is a high-risk situation that can increase the risk for numerous adverse experiences,
including victimization and psychological trauma (Bender et al., 2015; Tyler & Ray, 2019a).
Young adults experiencing homelessness may also be more likely to be exposed to substance-
using peers (DiGuiseppi, Davis, Christie, et al., 2020; Wenzel et al., 2012; Whitbeck et al.,
1999). Despite these risks, we did not find support that acute increases in homeless days was
associated with increased substance use at the next time-point. It may be that the time-lag

69

between assessments was too long to detect significant cross-lagged effects of these outcomes,
(Littlefield et al., 2021). More frequent assessments, such as those using ecological momentary
assessment, may be needed to detect cross-lagged effects (Tyler et al., 2018). However, this may
not be a concern, since significant cross-lagged effects were found in Davis and colleagues’
(2019) study of treatment-involved adolescents. Another more optimistic possibility is that, as a
result of treatment, young adults may have acquired the skills necessary to cope with an acute
increase in homelessness during or after treatment. More research is likely needed to understand
the absence of significant cross-lagged effects that are inconsistent with prevailing theory.
Finally, results did not support a “symptom driven” hypothesis, in which homelessness
follows as a consequence of increased substance use and related problems during or after
treatment. (Although these relationships were marginally significant among non-Hispanic White
and cisgender/heterosexual subgroups.) Our finding is inconsistent with prior work with
adolescents supporting a symptom-driven model, in which increased substance use during
treatment is followed by more homeless days after treatment (Davis et al., 2019). As suggested
by that study, this process may actually begin earlier, with elevated rates of homelessness at
treatment entry. This may initiate a “cascade” of negative effects leading to greater substance use
and PTSD symptoms during treatment, and in turn, further homelessness after treatment.
However, we found no evidence of this negative cascade among young adults. One reason for
this may relate to important developmental differences between these two age groups. Compared
to adolescents, young adults are more developmentally mature (Maggs et al., 2022), and
therefore they may be less susceptible to the negative effects of substance use in other life
domains. Young adults are also less likely to live with parents and other caregivers, and therefore
may be less likely than adolescents to be “kicked out” or run away from home as a result of

70

substance use and family conflict (Tyler & Schmitz, 2013). Adolescents are also more likely to
receive treatment involuntarily (Black & Chung, 2014), which suggests that young adults may
derive more benefits from treatment (i.e., reduced risk of homelessness) if they have more
internal motivation to change. Of course, young adults are tasked with their own unique
responsibilities (e.g., employment, supporting children or a significant other), which may make
homelessness more likely if substance use causes problems in these areas. More research may be
needed to understand why we did not observe significant increases in homelessness following
increased substance use and/or problems, especially given the marginally significant effects
among non-Hispanic White and cisgender Heterosexual young adults.
Strengths and Limitations of the Current Study
The current study is the first of its kind to examine bidirectional associations between
homelessness and substance use frequency and/or problems among young adults during and after
receiving treatment. Strengths of the study include the large, diverse sample, the use of state-of-
the-art quantitative methods to examine complex change processes, and testing relevant
hypotheses derived from theory. We also examined how these processes may unfold over time
across important subgroups of young adults. The current study also has a number of limitations.
This study relied on participant self-report data, which can be subject to recall and social
desirability biases. This is all the more important to consider, given that a majority of
participants had some involvement in the criminal justice system which may have dissuaded
them from accurately reporting their substance use behaviors. There were also high rates of
attrition at the 12-month follow-up, which we attempted to mitigate by controlling for important
variables related to missingness under the conditional missing at random assumption (Enders,
2022). The prevalence and frequency of homelessness was also relatively low, which could have

71

interfered with the ability to detect significant bidirectional effects in relation to substance use.
Our models were also conservative, as we included multiple control variables and used a robust
maximum likelihood estimator to account for non-normal distributions in these outcomes.
Implications for Research and Practice
Despite not finding evidence of cross-lagged effects, results from the present study raise
several implications for clinical practice with young adults in treatment settings. First, substance
use and associated problems, as well as days spent homeless, significantly declined for young
adults during treatment, and these declines were greatest for young adults with higher levels on
each of these outcomes at treatment entry. This underscores the importance of increasing access
and engagement in substance use treatment among at-risk young adults. This may be particularly
true for young adults whose substance use and homelessness are closely related prior to
treatment. In the current study, this was most apparent among non-Hispanic White young adults.
Continued efforts are likely needed to ensure that young adults entering treatment are provided
with comprehensive support services, including housing support. This is all the more important,
given the rising cost of housing (and lack of affordable housing) which has become more dire in
recent years after these data were collected (Up For Growth, 2022).
Study results also raise more questions for future research. Specifically, more research on
the availability of housing support services for young adults entering substance use treatment—
and how effectiveness of stable housing in supporting young adults’ recovery—is needed.
Research indicates that young adults experiencing homelessness are more likely to receive
treatment in inpatient/residential settings (Green et al., 2020). This suggests that residential
treatment may be serving as a temporary shelter for young adults experiencing homelessness.
Broadly, however, it is unclear what kinds of housing options exist for young adults after exiting

72

treatment, and whether these are effective in supporting their recovery. Two existing service
models may be relevant and deserve further research with young adults: recovery residences and
permanent supportive housing. Recovery residences (also called “sober living houses” or
“recovery homes”) are freestanding residential dwellings for individuals exiting substance use
treatment that require abstinence and provide peer and other recovery support services (Mericle,
Patterson, et al., 2022). Permanent supportive housing is an established model for chronically
homeless adults that provides housing and wraparound support services without requiring
abstinence. While these service models seem to be effective in addressing the need for housing
among individuals in recovery from substance use disorders, they have traditionally not been
designed for, and have thus been underutilized by young adults (Mericle, Slaymaker, et al., 2022;
Semborski et al., 2021). Research on the availability and effectiveness of housing services for
young adults in treatment can be complemented by sophisticated methods to evaluate their
effectiveness. Namely, the use of ecological momentary assessment, or employing LGC-SR
model designs with more frequency assessments may help researchers and clinicians better
understand bidirectional relationships between housing and substance use behaviors (Littlefield
et al., 2021). Ultimately, this may inform social-ecological theories of relapse and recovery and
service provision (Karriker-Jaffe et al., 2020).
Conclusion
After accounting for important baseline variables and complex change processes, the
current study found little evidence for prospective, bidirectional associations between
homelessness and substance use among young adults during and after substance use treatment.
One exception may be for young adults entering treatment with above-average levels of
substance use, who tended to be non-Hispanic White. Greater declines in homelessness during

73

and after treatment were observed for these individuals. Overall, results suggest relatively lower
risk of experiencing negative, reciprocal effects of homelessness and substance use during and
after treatment, when compared to a recent study of adolescents (Davis et al., 2019). Other
favorable outcomes include overall decreases in days experiencing substance use and/or
problems and days spent homelessness, suggesting broad benefits of treatment. More work is
needed to engage young adults who at risk for substance use disorder and homelessness in
treatment, and sustained efforts are likely needed to prevent the negative reciprocal effects of
these two outcomes.

74

Table 3.1 Participant Characteristics at Treatment Intake (N = 3,717)
Variable M (SD) or n (%)
n (%)
Missing
Demographics

Age (18 - 25) 20.1 (2.2) 0
Female, n (%) 1,037 (27.9%) 1 (< 1%)
Sexual / Gender Minority 268 (7.3%) 61 (2%)
Race/Ethnicity, n (%)

0
African American (non-Hispanic) 590 (15.9%)

White (non-Hispanic) 1,370 (36.9%)

Multiracial (non-Hispanic) 221 (5.9%)

Multiracial (Hispanic) 190 (5.1%)

Hispanic/Latinx only 1,241 (33.4%)

Asian, Native American or Alaska Native (non-Hispanic) 105 (2.8%)

Criminal justice system involvement, n (%) 2,954 (79.7%) 9 (< 1%)
Level of care

170 (4.6%)
Outpatient or intensive outpatient 3,104 (87.5%)

Residential 247 (6.7%)

Post-residential continuing care (outpatient) 196 (5.5%)

General Victimization Scale (range: 0 - 15) 3.9 (3.4) 13 (< 1%)
Social Risk Index (range: 0 - 28) 12.6 (4.9) 119 (3.2%)

75

Table 3.2 Descriptive Statistics for Homeless Days and Substance Use Frequency.
Assessment N Mean SD Minimum Maximum n (%) ≥ 1
Homeless Days

Baseline 3,695 3.56 14.64 0 90 387 (10%)
3 Month Follow-up 3,059 3.32 15.19 0 90 240 (8%)
6 Month Follow-up 2,890 2.56 13.16 0 90 191 (7%)
12 Month Follow-up 1,804 2.96 14.57 0 90 121 (7%)
Substance Frequency Scale
Baseline 3,716 12.37 16.12 0 83.19 2,561 (31%)
3 Month Follow-up 3,083 5.78 11.21 0 100 1,707 (45%)
6 Month Follow-up 2,907 6.23 12.12 0 100 1,470 (49%)
12 Month Follow-up 1,815 6.89 13.19 0 100 900 (50%)

76
Table 3.3 Model Fit Comparisons for Substance Use Frequency Growth Model

Model 0 Model 1 Model 2 Model 3 Model 4 Model 5 Model 6

No growth
(intercept
only) Linear
Linear
(equal
residuals) Bilinear
a

Bilinear
(equal
residuals)
Bilinear with AR
residual paths
(freed)
Bilinear with AR
residual paths
(equal post-tx)
χ
2
352.16 246.38 332.96 -
26.87
43.67 27.40
df 8 5 8 - 4 3 4
p < 0.001 < 0.001

- < 0.001 < 0.001 0.001
Δ df - 3 3 - 4 n/a 1
c 1.82 1.89 3.11 - 2.4639 1.88 3.24
cd - n/a 5.14 - n/a n/a 7.32
Scaled χ2 640.93 465.66 1035.51 - 66.20 82.10 88.776
Scaled Δ χ2 - n/a 110.79 - n/a n/a 0.91
χ2 diff. test p - n/a 0.00 - n/a n/a 0.34
Comparison - n/a M2 vs. M1

n/a n/a M6 vs. M5
AIC 91,162.46 90,994.78 91,557.01 - 90,597.16 90,615.06 90,619.68
BIC 91,199.78 91,050.77 91,594.33 - 90,659.37 90,683.49 90,681.89
RMSEA 0.11 0.11 0.11 - 0.04 0.06 0.04
CFI 0.36 0.55 0.4 - 0.96 0.93 0.96
TLI 0.52 0.46 0.55 - 0.94 0.85 0.94
SRMR 0.12 0.09 0.15 - 0.04 0.05 0.05
a
Model not identified.

Note. AR = Autoregressive paths; Hmls = Homeless days; SFS = Substance Frequency Scale; post-tx = post-treatment; df = degrees of
freedom; c0 = scaling correction factor for Null Hypothesis; cd = difference test scaling correction; Scaled χ2 = Satorra-Bentler scaled
chi-square difference test; RMSEA = Root Mean Square Error of Approximation; AIC = Akaike Information Criterion; BIC = Bayesian
Information Criterion; CFI = Comparative Fit Index; TLI = Tucker-Lewis Index; SRMR = Standardized Root Mean Square Residual.

77
Table 3.4 Model Fit Comparisons for Homeless Days Growth Model

Model 0 Model 1 Model 2 Model 3 Model 4 Model 5 Model 6

No growth

Linear

Bi-linear
(equal
residuals)
Bi-linear with
AR residuals
(freed)

Linear
(equal
residuals)
Linear with
residuals &
AR paths
(freed)
Linear with
AR paths
(equal post-tx)
χ
2
45.60 32.12 3.30
n/a
a

30.23 25.49 15.63
df 8 5 4 - 7 4 5
p < 0.001 < 0.001 0.51 - < 0.001 < 0.001 0.008
Δ df - 3 3 - 2 n/a 1
c 4.20 3.94 7.16 - 5.58 2.62 4.68
cd - n/a 3.48 - 9.68 n/a 12.92
Scaled χ2 191.71 126.70 23.63 - 168.78 66.78 73.15
Scaled Δ χ2 - n/a 41.72 - 4.35 n/a 0.49
χ2 diff. test p - n/a 0.000 - 0.11 n/a 0.48
Comparison - n/a M2 vs. M1 - M4 vs. M1 n/a M6 vs. M5
AIC 92,664.77 92,605.76 92,501.11 - 92,644.40 92,547.80 92,552.26
BIC 92,702.10 92,661.75 92,563.31 - 92,687.94 92,610.00 92,608.24
RMSEA 0.04 0.04 0.00 - 0.03 0.04 0.02
CFI 0.80 0.85 1.00 - 0.87 0.88 0.94
TLI 0.85 0.82 1.00 - 0.89 0.82 0.93
SRMR 0.07 0.05 0.02 - 0.06 0.05 0.05
a
Model not identified.

Note. AR = Autoregressive paths; Hmls = Homeless days; SFS = Substance Frequency Scale; post-tx = post-treatment; df =
degrees of freedom; c0 = scaling correction factor for Null Hypothesis; cd = difference test scaling correction; Scaled χ2 =
Satorra-Bentler scaled chi-square difference test; RMSEA = Root Mean Square Error of Approximation; AIC = Akaike
Information Criterion; BIC = Bayesian Information Criterion; CFI = Comparative Fit Index; TLI = Tucker-Lewis Index;
SRMR = Standardized Root Mean Square Residual.

78
Table 3.5 Nested Model Comparisons for Unconditional Bivariate LGC-SR Model

Model 1 Model 2 Model 3 Model 4 Model 5
Description
ARs & Crosslags
freely estimated
Hmls ARs
equal (post-tx)
SFS ARs
equal (post-tx)
SFS on Hmls
crosslags
equal (post-tx)
Hmls on SFS
crosslags equal
(post-tx)
Model
Comparisons n/a M2 vs M1 M3 vs M2 M4 vs M3 M5 vs M4
Decision n/a Retain M2 Retain M3 Retain M4 Retain M5
χ
2
11.00 17.90 21.86 23.96 25.20
df 6 7 8 9 10
p 0.09 0.01 0.01 0.00 0.01
Δ df - 1 1 1 1
c0 1.64 2.86 3.48 3.20 3.05
cd - 10.18 7.82 0.96 1.70
Scaled χ2 18.04 51.19 76.07 76.67 76.86
Scaled Δ χ2 - 3.26 3.18 0.62 0.11
χ2 diff. test p - 0.071 0.074 0.430 0.739
AIC 182,923.98 182,955.04 182,978.05 182,976.68 182,974.81
BIC 183,160.37 183,185.21 183,201.99 183,194.40 183,186.32
RMSEA 0.02 0.02 0.02 0.02 0.02
CFI 1.00 0.99 0.99 0.99 0.99
TLI 0.98 0.96 0.95 0.96 0.96
SRMR 0.01 0.02 0.030 0.030 0.030
Note. AR = Autoregressive paths; Hmls = Homeless days; SFS = Substance Frequency Scale; post-tx = post-treatment; df = degrees
of freedom; c = scaling correction factor for Null Hypothesis; cd = difference test scaling correction; Scaled χ2 = Satorra-Bentler
scaled chi-square difference test; RMSEA = Root Mean Square Error of Approximation; AIC = Akaike Information Criterion; BIC =
Bayesian Information Criterion; CFI = Comparative Fit Index; TLI = Tucker-Lewis Index; SRMR = Standardized Root Mean Square
Residual.

79
Figure 3.1 LGC-SR Model of Homeless Days and Substance Use Frequency

Note. Between-person effects (latent growth factor covariances) not shown for clarity; LGC-SR = Latent Growth Curve Model with
Structured Residuals; SFS = Substance Frequency Scale; Hmls = Days spent homeless; SGM = Sexual and/ Gender Minority; LOC =
Level of Care; INT = Intercept; S1 = Slope 1; S2 = Slope 2.

80

Table 3.6 Parameter Estimates for Conditional LGC-SR Model
Estimated Parameter Est. SE Est. (Std.) p
Between-Person Covariances / Correlations

Hmlsint, Hmlslope1 -18.78 7.26 -0.77 0.01
Hmlsint, SFSint 0.68 6.39 0.01 0.92
Hmlsint, SFSslope1 7.66 6.90 0.06 0.27
Hmlsint, SFSslope2 -1.83 2.31 -0.07 0.43
SFSint, SFSslope1 -178.19 10.80 -0.84 < 0.001
SFSint, SFSslope2 -5.91 2.81 -0.15 0.04
SFSint, Hmlsslope1 -2.04 2.02 -0.05 0.31
SFSslope1, SFSslope2 -2.16 3.60 -0.06 0.55
SFSslope1, Hmlsslope1 0.78 2.26 0.02 0.73
SFSslope2, Hmlsslope1 1.41 0.97 0.21 0.15
Within-person autoregressions

HmlsBL → Hmls3m -0.06 0.19 -0.05 0.75
Hmls3m → Hmls6m 0.20 0.06 0.19 0.001
Hmls6m → Hmls12m 0.20 0.06 0.20 0.001
SFSBL → SFS3m -3.22 2.54 0.00 0.21
SFS3m → SFS6m 0.05 0.12 0.04 0.70
SFS6m → SFS12m 0.05 0.12 0.05 0.70
Within-person cross-lagged effects

SFSBL → Hmls3m -0.16 0.35 0.00 0.66
SFS3m → Hmls6m 0.06 0.06 0.04 0.26
SFS6m → Hmls12m 0.06 0.06 0.05 0.26
HmlsBL → SFS3m 0.45 0.36 0.69 0.22
Hmls3m → SFS6m -0.004 0.02 -0.01 0.86
Hmls6m → SFS12m -0.004 0.02 -0.01 0.86
Note. Factor loadings, within-time cross-construct residual covariances, intercepts,
residual variances, and between-person effects of demographic and control variables on
latent growth factors not shown.

81
Table 3.7 LGC-SR Parameter Estimates Stratified by Sex

Males (n = 2,641)

Females (n = 1,015)
Estimated Parameter Est. SE Est. (Std.) p

Est. SE Est. (Std.) p
Between-Person Growth Factor
Covariances / Correlations

Hmlsint, Hmlslope1 -3.76 6.64 -0.39 0.57

-37.35 16.76 -0.86 0.03
Hmlsint, SFSint -6.72 8.38 -0.09 0.42

-15.71 21.21 -0.09 0.46
Hmlsint, SFSslope1 13.80 9.48 0.19 0.15

26.60 21.37 0.15 0.21
Hmlsint, SFSslope2 -1.67 2.94 -0.14 0.57

-2.34 4.50 -0.05 0.60
SFSint, SFSslope1 -179.43 15.13 -0.84 < 0.001

-154.12 24.76 -0.82 < 0.001
SFSint, SFSslope2 -5.77 3.80 -0.17 0.13

-6.17 3.53 -0.12 0.08
SFSint, Hmlsslope1 -0.22 2.52 -0.01 0.93

3.36 6.05 0.07 0.58
SFSslope1, SFSslope2 -2.18 4.34 -0.07 0.62

-6.12 6.13 -0.13 0.32
SFSslope1, Hmlsslope1 -1.23 2.74 -0.05 0.64

-3.62 6.22 -0.08 0.56
SFSslope2, Hmlsslope1 0.51 1.14 0.12 0.65

3.67 2.18 0.31 0.09
Within-person autoregressions

HmlsBL → Hmls3m 0.16 0.14 0.17 0.26

-1.27 1.09 -0.94 0.25
Hmls3m → Hmls6m 0.19 0.09 0.19 0.04

0.21 0.07 0.24 0.004
Hmls6m → Hmls12m 0.19 0.09 0.19 0.04

0.21 0.07 0.21 0.004
SFSBL → SFS3m -3.28 2.85 0.00 0.25

0.18 0.40 0.14 0.65
SFS3m → SFS6m 0.13 0.11 0.10 0.25

-0.38 0.31 -0.36 0.21
SFS6m → SFS12m 0.13 0.11 0.13 0.25

-0.38 0.31 -0.38 0.21
Within-person cross-lagged effects

SFSBL → Hmls3m -0.13 0.32 0.00 0.68

3.74 5.21 1.03 0.47
SFS3m → Hmls6m 0.06 0.06 0.04 0.31

0.06 0.18 0.03 0.72
SFS6m → Hmls12m 0.06 0.06 0.05 0.31

0.06 0.18 0.03 0.72
HmlsBL → SFS3m 0.43 0.40 0.68 0.28

-0.02 0.12 -0.04 0.86
Hmls3m → SFS6m 0.03 0.04 0.03 0.51

-0.04 0.03 -0.11 0.10

82
Hmls6m → SFS12m 0.03 0.04 0.03 0.51 -0.04 0.03 -0.10 0.10
Note. Factor loadings, within-time cross-construct residual covariances, intercepts, residual variances, and between-person
effects of demographic and control variables on latent growth factors not shown.

83
Table 3.8 LGC-SR Parameter Estimates Stratified by Race/Ethnicity

Non-Hispanic White (n = 1,347)

Young Adults of Color (n = 2,309)
Estimated Parameter Est. SE Est. (Std.) p Est. SE Est. (Std.) p
Between-Person Growth Factor
Covariances / Correlations

Hmlsint, Hmlslope1 -25.55 10.67 -0.87 0.02

-16.55 8.86 -0.72 0.06
Hmlsint, SFSint 19.51 8.33 0.12 0.02

-11.24 9.37 -0.08 0.23
Hmlsint, SFSslope1 -1.79 9.61 -0.01 0.85

14.55 10.10 0.11 0.15
Hmlsint, SFSslope2 -4.41 4.92 -0.15 0.37

-0.57 2.36 -0.02 0.81
SFSint, SFSslope1 -187.76 16.31 -0.86 < 0.001

-173.80 16.47 -0.82 < 0.001
SFSint, SFSslope2 -3.83 5.40 -0.09 0.48

-6.41 3.49 -0.16 0.07
SFSint, Hmlsslope1 -5.82 2.79 -0.14 0.04

0.52 2.80 0.01 0.85
SFSslope1, SFSslope2 -0.13 8.77 0.00 0.99

-5.11 3.73 -0.13 0.17
SFSslope1, Hmlsslope1 0.47 3.16 0.01 0.88

0.65 3.18 0.02 0.84
SFSslope2, Hmlsslope1 2.20 1.79 0.28 0.22

0.97 1.14 0.14 0.40
Within-person autoregressions

HmlsBL → Hmls3m -0.21 0.34 -0.19 0.53

-0.02 0.23 -0.02 0.92
Hmls3m → Hmls6m 0.25 0.10 0.25 0.01

0.17 0.08 0.17 0.03
Hmls6m → Hmls12m 0.25 0.10 0.25 0.01

0.17 0.08 0.17 0.03
SFSBL → SFS3m -5.31 39.11 -0.40 0.89

-2.36 1.52 0.00 0.12
SFS3m → SFS6m -0.09 0.30 -0.08 0.76

0.08 0.12 0.06 0.50
SFS6m → SFS12m -0.09 0.30 -0.09 0.76

0.08 0.12 0.08 0.50
Within-person cross-lagged effects

SFSBL → Hmls3m -0.68 4.48 -0.04 0.88

-0.12 0.27 0.00 0.64
SFS3m → Hmls6m 0.15 0.08 0.10 0.07

0.01 0.07 0.01 0.86
SFS6m → Hmls12m 0.15 0.08 0.11 0.07

0.01 0.07 0.01 0.86
HmlsBL → SFS3m 0.17 1.34 0.24 0.90

0.50 0.33 0.83 0.13
Hmls3m → SFS6m 0.00 0.05 0.00 0.97

-0.01 0.03 -0.01 0.85

84
Hmls6m → SFS12m 0.00 0.05 0.00 0.97 -0.01 0.03 -0.01 0.85
Note. Factor loadings, within-time cross-construct residual covariances, intercepts, residual variances, and between-person
effects of demographic and control variables on latent growth factors not shown.

85
Table 3.9 LGC-SR Parameter Estimates Stratified by Sexual/Gender Minority Status

Cisgender Heterosexual (n = 3,388)

Sexual / Gender Minority (n = 268)
Estimated Parameter Est. SE Est. (Std.) p Est. SE Est. (Std.) p
Between-Person Growth Factor
Covariances / Correlations

Hmlsint, Hmlslope1 -17.61 8.20 -0.76 0.03

-9.99 24.40 0.00 0.68
Hmlsint, SFSint -5.71 7.75 -0.04 0.46

40.77 32.04 0.27 0.20
Hmlsint, SFSslope1 14.74 8.28 0.12 0.08

-17.26 29.82 -0.11 0.56
Hmlsint, SFSslope2 -3.30 2.44 -0.15 0.18

4.62 9.41 0.10 0.62
SFSint, SFSslope1 -178.53 12.72 -0.85 < 0.001

-162.48 16.30 -0.72 < 0.001
SFSint, SFSslope2 -5.91 3.09 -0.16 0.06

-8.79 8.86 -0.13 0.32
SFSint, Hmlsslope1 -0.54 2.33 -0.01 0.82

-8.27 8.75 0.00 0.35
SFSslope1, SFSslope2 -1.74 3.75 -0.05 0.64

-18.34 10.89 -0.25 0.09
SFSslope1, Hmlsslope1 -1.68 2.47 -0.05 0.50

4.71 8.60 0.00 0.58
SFSslope2, Hmlsslope1 1.93 1.06 0.30 0.07

-1.91 3.17 0.00 0.55
Within-person autoregressions

HmlsBL → Hmls3m 0.05 0.22 0.05 0.82

-0.72 1.29 -0.64 0.58
Hmls3m → Hmls6m 0.22 0.07 0.22 0.00

0.15 0.12 0.18 0.18
Hmls6m → Hmls12m 0.22 0.07 0.22 0.00

0.15 0.12 0.15 0.18
SFSBL → SFS3m -2.95 2.48 -0.15 0.24

4.11 8.80 0.99 0.64
SFS3m → SFS6m 0.12 0.11 0.09 0.30

-0.92 0.34 -0.80 0.01
SFS6m → SFS12m 0.12 0.11 0.12 0.30

-0.92 0.34 -0.83 0.01
Within-person cross-lagged effects

SFSBL → Hmls3m -0.39 0.35 -0.01 0.27

10.43 19.74 0.78 0.60
SFS3m → Hmls6m 0.10 0.06 0.06 0.08

-0.17 0.22 -0.06 0.43
SFS6m → Hmls12m 0.10 0.06 0.07 0.08

-0.17 0.22 -0.07 0.43
HmlsBL → SFS3m 0.50 0.42 0.73 0.24

-0.20 0.54 -0.57 0.72
Hmls3m → SFS6m 0.01 0.03 0.01 0.83

0.02 0.09 0.05 0.84

86
Hmls6m → SFS12m 0.01 0.03 0.01 0.83 0.02 0.09 0.04 0.84
Note. Factor loadings, within-time cross-construct residual covariances, intercepts, residual variances, and between-person effects of
demographic and control variables on latent growth factors not shown.

87

Chapter 4 (Study 3): Barriers and Facilitators of Substance Use Treatment for Young
Adults Experiencing Homelessness: Perspectives from Young Adults and Service
Providers
Abstract
Introduction: A high prevalence of substance use disorders exists among young adults
experiencing homelessness. To date, however, few studies have documented the perceived
barriers and facilitators to substance use treatment using perspectives from both young adults and
service providers. Methods: In-depth individual interviews were conducted with young adults
experiencing homelessness who met past-year substance use disorder criteria (n = 22) and
homeless youth service providers occupying a variety of roles (n = 16) in Los Angeles,
California. Interviews queried perceived barriers and facilitators to treatment entry and
engagement. Thematic Analysis was used to develop themes which were compared and
contrasted for young adults and providers. Results: Nine themes were identified: 1) lack of
access to treatment, 2) need for low-barrier treatment, 3) rigidity vs. flexibility of treatment, 4)
homelessness and instability, 5) shame and stigma, 6) supportive referrals, 7) systems referrals,
8) safe spaces, and 9) readiness for treatment. Young adults and providers had unique
perspectives stemming from different lived and professional experiences. However, quotes from
both groups contributed to all themes. Conclusions: For young adults experiencing
homelessness, significant barriers to substance use treatment exist across multiple social-
ecological levels (systemic, interpersonal, individual). Facilitators to treatment exist as well, and
should be strengthened to increase access and engagement in care.
Keywords: Homelessness, Young Adults, Substance Use Treatment, Qualitative, Service Use

88

Introduction
It is estimated that one in 10 young adults aged 18 to 25 experience homelessness each
year, which includes “couch surfing” (e.g., temporary living arrangements with family or friends
to avoid being homeless) or explicit homelessness (Morton et al., 2018). Young adults
experiencing homelessness are at greater risk for a variety of physical and mental health
problems, including substance use disorders (Edidin et al., 2012; Kulik et al., 2011). This is not
surprising, as mental health and substance use problems tend to peak during young adulthood
(Substance Abuse and Mental Health Services Administration [SAMHSA], 2021). However,
young adults experiencing homelessness have a higher prevalence of substance use and
substance use disorders than stably housed young adults (Chassman et al., 2022; Nyamathi et al.,
2007; van Leeuwen et al., 2006). This is thought to be due to a greater accumulation of risk
factors, including childhood trauma, greater access to substance use through peer networks, and
using alcohol and other drugs to cope with day-to-day stressors (DiGuiseppi, Davis, Christie, et
al., 2020; Tyler et al., 2018; Whitbeck et al., 1999). This calls for a continued need for
prevention and treatment of substance use disorders for young adults experiencing homelessness.
However, the literature has identified several barriers to accessing and engaging in substance use
treatment for young adults experiencing homelessness (Nyamathi et al., 2007; Rabinovitz et al.,
2010; Whitbeck, 2009). The current study expands on this work, using in-depth interviews with
young adults currently experiencing homeless and homeless youth service providers.
Rates of Substance Use Treatment Service Utilization
Most unaccompanied youth experiencing homelessness receive healthcare services
through hospital emergency departments, and specialty service agencies such as drop-in center
and shelters (de Rosa et al., 1999; Klein et al., 2000; Parast et al., 2019). Oftentimes it is through

89

these agencies, as well as through referrals through the criminal justice and child welfare system,
that youth receive specialized substance use treatment on an inpatient or outpatient basis. Rates
of substance use treatment service utilization among young adults experiencing homelessness
vary across studies. For example, an early study of service use among youth experiencing
homelessness in Los Angeles (77% young adults) found that 10 percent of youth had utilized
substance use treatment services or attended a 12-step program (i.e., Alcoholics Anonymous,
Narcotics Anonymous) since arriving in Los Angeles (de Rosa et al., 1999). Other surveys with
homeless youth and young adults from street-based and agency locations in Los Angeles indicate
that 14 percent of youth received substance use treatment in the past year (Rabinovitz et al.,
2010). In another study by Pedersen and colleagues (2018), 30 percent of youth met criteria for
substance use disorder, but only seven percent had received treatment or counseling for this
(most often provided at drop-in centers). Other studies suggest somewhat higher rates of
treatment. For example, a study of 760 lesbian, gay, and bisexual (LGB) homeless youth and
non-LGB homeless youth aged 14 to 24 in eight U.S. cities reported lifetime treatment rates of
38% for LGB youth and 27% for non-LGB youth (van Leeuwen et al., 2006). In a longitudinal
study of young adults experiencing homelessness in eight Midwestern cities, 23 percent of men
and 15 percent of women had ever received inpatient hospitalization for treatment for substance
use disorder, while 38 percent overall had attended outpatient treatment (Whitbeck, 2009). These
rates of substance use and behavioral health service use are relatively high compared with overall
rates among the general population of young adults (SAMHSA, 2021), but may still be
inadequate given the high prevalence of substance use problems among this population.

90

Barriers and Facilitators to Substance Use Treatment
Despite the need for intervention among a large minority of young adults experiencing
homelessness, a number of barriers to accessing and engaging in health care services (including
substance use treatment) have been identified in the literature. Many barriers are systemic in
nature, and include lack of available services and burdensome paperwork and administrative
hurdles, leading to long waiting times for inpatient or outpatient care (Hudson et al., 2009;
Rabinovitz et al., 2010). Interviews with youth and young adults have also revealed a lack of
awareness of available treatment services and how to access them, lack of transportation, and
lack of insurance or ability to pay (Bozinoff et al., 2017; Hudson et al., 2009; Nyamathi et al.,
2007; Rabinovitz et al., 2010; Rosa Solorio et al., 2006; Whitbeck, 2009). Other barriers are
interpersonal, and include negative previous experiences and perceived discrimination and
judgment from providers, as well as perceived disapproval from peers (Derbyshire et al., 2006;
Ensign, 2004; Hudson et al., 2010; Nyamathi et al., 2007; Rabinovitz et al., 2010; Whitbeck,
2009). A third set of barriers identified in the literature appear to exist within the individual, and
include lack of internal motivation, and lack of perceived need for treatment or problem
recognition (Hudson et al., 2009; Nyamathi et al., 2007; Whitbeck, 2009). This is illustrated by
the idea endorsed by some young adults that treatment will be ineffective if individuals are not
motivated to change their behavior. Low perceived need for treatment is common among young
adults in general (SAMHSA, 2021), but research with young adults experiencing homelessness
suggests that other more systemic barriers are at play, given that a lower proportion of young
adults who perceived a need for treatment actually received it, when compared to young adults
who perceived a need for other services such as mental health counseling and housing services
(Pedersen et al., 2018).

91

In addition to these barriers, factors that facilitate, or increase access and engagement in
substance use treatment services have been identified. Interviews with young adults reveal that a
support person who could listen to young adults’ concerns could help facilitate youths’
engagement in treatment. In focus groups with young adults experiencing homelessness,
participants valued support from people who were not in a paid staff role, but rather, a friend or
close contact (Nyamathi et al., 2007). Natural mentors with lived experience of substance use
disorder, similar to a sponsor in 12-step programs, were preferred by young adults as well
(Nyamathi et al., 2007). Research on facilitators of young adults’ general health-seeking
behavior suggests that social support from a variety of sources is beneficial, whether from
agency staff, homeless peers, family, friends or significant others (Hudson et al., 2010). Young
adults also prefer services that are easily accessible, offer a wide range of alternative substance-
free activities (e.g., art, music, sports), and providers who are caring, nonjudgmental, and
respectful of young adults’ autonomy (Hudson et al., 2008; Nyamathi et al., 2007; Thompson et
al., 2006; Whitbeck, 2009). Some young adults have expressed that providers who were
supportive of their continued use of cannabis would be beneficial, as cannabis could be used as a
stress reliever and alternative to “harder” drugs like methamphetamine and opiates (Nyamathi et
al., 2007; Paul et al., 2020). Being court mandated to treatment through the criminal justice
system has also been presented as a facilitator to treatment (Nyamathi et al., 2007), but others
have expressed that one’s internal motivation to change is necessary, whether treatment is
mandated or voluntary (Hudson et al., 2009; Nyamathi et al., 2007). Major life events, such as a
having a child, negative consequences of substance use, and being tired of a drug using lifestyle
have been expressed as facilitators to enhance young adults’ motivation to engage in treatment
(Hudson et al., 2009; Nyamathi et al., 2007).

92

Some research in this area has relied on established theoretical models to better
understand barriers and facilitators to service use among youth experiencing homelessness.
Solorio and colleagues (2006) and Pedersen and colleagues (2016), respectively, extended the
Behavioral Model for Vulnerable Populations to examine barriers and facilitators to mental
health and drop-in center service use. The Behavioral Model for Vulnerable Populations
presented predisposing factors (demographics and individual characteristics that predispose one
to illness), enabling factors (personal resources such as insurance coverage, social support), and
need factors (perceived need for care or an objective medical diagnosis) as determinants of
health care service use among adults experiencing homelessness (Gelberg et al., 2000). The
model adds vulnerability factors to each of these domains (e.g., length of time spent homeless as
a predisposing factor) relevant for understanding service use among people experiencing
homelessness. Although facilitators would traditionally fit within the category of enabling
factors, barriers and facilitators could exist across all of these domains. For example, perceived
need for substance use treatment could be a barrier or facilitator to using services, depending on
one’s perception (Pedersen et al., 2018).
Including Service Provider Perspectives
This body of research is helpful for understanding barriers and facilitators to substance
use treatment engagement from young adults themselves. However, few studies have
incorporated the perspectives of service providers. Provider perspectives are important because
they often have extensive experience delivering services and may have a different perspective on
the barriers and facilitators to service delivery than young adults (Kuusisto & Lintonen, 2020).
For example, surveys with substance use treatment providers in the U.S. tend to emphasize
systemic barriers, such as the need for additional training, expansion of services and resources,

93

and better coordination across service systems (Ashford et al., 2018). One study of homeless
youth service providers in Los Angeles revealed widespread interest in providing smoking
cessation services to youth, but identified multiple system-level barriers to service provision,
including lack of financial, personnel, and knowledge resources to deliver these services within
agencies serving homeless youth (Shadel et al., 2014). Although patients and providers may have
different perspectives, themes related to certain topics may be similar. In a qualitative study with
both patients and providers, seven themes were derived to describe success in opioid use disorder
treatment (Hooker et al., 2022). Patients and providers agreed on five of the seven themes (e.g.,
improved psychological well-being), while two themes were uniquely generated from interviews
with patients (tapering off of medications and reducing stigma).
The Current Study
Although there is notable body of work on perceived barriers and facilitators to substance
use treatment engagement among young adults experiencing homelessness, much of this research
is now dated, and does not include the perspective of providers. A contemporary understanding
of the barriers and facilitators to treatment from the perspective of young adults and providers
may inform efforts to remove these barriers and improve access to services. Therefore, the goal
of the current study was to better understand barriers and facilitators to engagement in substance
use treatment for young adults experiencing homelessness who meet criteria for substance use
disorder. Qualitative, in-depth interviews were conducted with young adults currently
experiencing homelessness and service providers occupying a variety of roles within the
homeless youth service system in Los Angeles, California. Themes were developed and for these
two groups, and were combined if they converged, or compared and contrasted if they diverged.

94

Methods
Participants
Participants included 22 young adults experiencing homelessness and 16 service
providers and in Los Angeles, California. Young adult participants who provided qualitative, in-
depth interview were recruited from one drop-in center in Hollywood, California. Study
eligibility criteria for young adults were: 1) currently receiving services at the drop-in center; 2)
aged 18 to 25; 3) screening positive for problematic use of alcohol, cannabis or other substances
in the past 12 months; and 4) being able to read and converse in English. Demographic
characteristics for young adult participants are included in Table 4.1. Most (71.4%) were male
and the average age was 23.4 years old (SD = 1.5). Participants were 18.3 years old, on average
(SD = 2.6) when they first experienced homelessness, and were homelessness for an average of
3.5 years (SD = 2.8) in their lifetime. Seven participants (31.8%) reported ever receiving
substance use treatment; four of these received treatment in the past 12 months, and three
received treatment more than 12 months ago.
To be eligible, service providers must have met two criteria: 1) served a role as a health
or social service professionals providing direct services to young adults experiencing
homelessness, and 2) had experience referring or providing substance use treatment and
counseling services to young adults experiencing homelessness. Service providers were
employed at five organizations in Los Angeles. Locations included a drop-in center in
Hollywood (n = 4 providers), a drop-in center in the beachside city of Venice (n = 2), a drop-in
center/shelter in east Los Angeles (n = 1), a shelter in Hollywood (n = 2), and an adolescent and
young adult medical clinic at a large Children’s Hospital (n = 7). As shown in Table 4.2, service

95

providers occupied a variety of roles and educational levels, and had 12 years’ experience, on
average (SD = 3.8) providing direct services to young adults experiencing homelessness.
Procedures
Young adult participants were enrolled in person by the lead author at one drop-in center
serving youth experiencing homelessness in Los Angeles, California during a two-week period
in April 2022. Young adults signed a consent form before completing a screening questionnaire
to determine eligibility. In total, 50 young adults completed the screening questionnaire; three
young adults declined to complete the screening questionnaire for reasons unspecified. Forty
young adults (80% of those screened) met criteria for problematic use of at least one substance,
and 22 (44%) met criteria for problematic use of two or more substances. As shown in Figure
4.1, the most common problematic substances were cannabis (64%), alcohol (36%), and
stimulants (22%). Twenty-three of the 40 eligible young adults completed an in-depth individual
interview with the study author. However, one qualitative interview was discarded due to the
participant being non-verbal during the interview, resulting in a total sample of 22 young adults.
Among those eligible, reasons for not completing the interview included not having enough time
to complete the interview, or the study author not having enough gift cards to compensate them
on that particular day. None of the young adults who were eligible refused to participate in the
interview. Eligible participants completed an additional consent form, and a brief demographic
survey before beginning the interview. All interviews were conducted in a private location at the
drop-in center by the lead author. Interviews were 30 minutes on average, ranging from seven to
50 minutes. Participants were compensated with a $10 Target gift card for completing the
screening questionnaire, and a $35 Target gift card for completing the individual interview. All

96

procedures were approved by the University of Southern California Institutional Review Board
and drop-in center administrators.
Recruitment of service providers was performed by the lead author by contacting
administrative staff from health and social service agencies providing services to youth
experiencing homelessness in Los Angeles. Administrators were asked if they would be willing
to allow their staff providers to participate in the study; all agreed to distribute a recruitment
letter advertising the study to their staff. Interested providers were put in touch with the lead
author and were asked screening questions via email to determine eligibility (all interested
providers except one were deemed eligible). Providers then signed an online consent form and
scheduled the interview with the lead author to be conducted online via Zoom videoconferencing
software. These interviews were conducted from October 2021 through January 2022.
Measures
Screening Questionnaires
A screening questionnaire was used to determine young adult participant eligibility, and
assess young adults’ age and past-year substance use disorder criteria using the Tobacco,
Alcohol, Prescription Medication, and other Substance Use (TAPS) tool, which has
demonstrated adequate sensitivity (>70%) for identifying a DSM-5 substance use disorder in
adults (McNeely et al., 2016). The TAPS is comprised of two parts: the first part contains four
items assessing past year frequency of use of tobacco, binge alcohol use (5 or more drinks for
men or 4 or more drinks for women in a single day), illicit drug use (“marijuana, cocaine or
crack, heroin, methamphetamine [crystal meth], hallucinogens, ecstasy/MDMA”), and
prescription medication misuse. Participants who indicate any past year use of each substance are

97

asked a series of follow-up questions for each substance in part two (i.e., strong urges or desire to
use the substance, failed attempts to quit or cut down, and concern from others about their use).
Participants who were aged 18 to 25 scored 2 or more on the TAPS for alcohol, and cannabis or
1 or more on the TAPS for other substances (e.g., stimulants, prescription medication misuse,
etc.) were eligible to participate in the in-depth individual interview. All participants who were
approached at the drop-in center met eligibility for currently receiving drop-in center services,
and competency in reading and conversing in English was determined by the first author.
Service providers were screened for eligibility via their responses to two screening
questions, asking 1) if they had experiencing providing professional services (e.g., “social or
health care services”) to young adults experiencing homelessness; and 2) if they had “direct or
indirect” experience providing substance use treatment services to young adults experiencing
homelessness (this included direct service provision or referral to substance use treatment
services). Service provider participants were eligible if they answered “yes” to both questions.
Demographic and Background Experiences Questionnaire
Demographics and background experiences were assessed prior to taking part in the in-
depth interviews. Service providers reported their demographics (age, sex, gender identity,
race/ethnicity), education (highest degree received), current role at their organization (e.g., social
worker, nurse, physician), and length of experience providing direct services to young adults
experiencing homelessness. Young adults provided similar information, as well as current school
and employment, living situation (number of nights spent in various settings in the past 14 days),
and homelessness experiences (age when first experienced homelessness, total time spent
homeless). Young adults also reported any previous experience receiving substance use

98

treatment using an item from the National Survey on Drug Use and Health (Administration,
2021).
In-Depth Interview Guides
In-depth interview guides for providers and young adults were developed by the lead
author in consultation with two co-authors (BH and JT) and a drop-in center staff member who
was not enrolled as a research participant. Questions were similar for both providers and young
adults, and addressed a range of research questions related to homelessness, housing, and
substance use (See Appendix 3.1 and 3.2). Interview questions relevant to the current study
included questions about perceived barriers (“what gets in the way?”) and facilitators (“what
helps?”) of young adults’ entry and engagement in substance use treatment. All questions were
asked in an open-ended manner, with the interviewer probing further or asking clarifying
questions when necessary (Padgett, 2012).
Qualitative Data Analysis
The present study utilized Thematic Analysis (TA) as the main analytic approach. TA is a
flexible approach used to generate themes from qualitative data. We used Braun and Clarke’s
definition of a theme as “a pattern of shared meaning organized around a central concept” (Braun
& Clarke, 2022, p. 77). TA was carried out in several steps. First, audio recordings of interviews
were transcribed with identifying information removed. Following this, thematic analysis began
by coding the data using Atlas.ti (version 22.0.10.0). Coding is a process used in many
qualitative approaches that involves assigning meaning labels to excerpts of text (Braun &
Clarke, 2022). Provider interview transcripts were coded first, as these were completed prior to
young adult interviews. Initial codes were deductive if derived from topical domains from the

99

interview guide (i.e., “barriers to treatment”), or inductive if they arose from initial and
successive readings of the text (i.e., “trauma-informed care”). Initial rounds of coding the
transcripts were carried out by a team of six coders including the lead author and graduate
students in psychology and social work.
Coders conducted multiple rounds of coding and discussed and agreed upon definitions
of codes until a working codebook was developed (see Appendix 3.3). The codebook was used
to guide successive coding of transcripts, with minor adjustments made to codes and their
definitions if necessary. The codebook was later applied to code young adult interviews, since
their interview questions mirrored those of providers. Some codes were added or edited while
coding to reflect different concepts that emerged in young adult transcripts (e.g.,
“food/nutrition”, “sleep”, and “spirituality” codes were added). Overall, the existing codebook
worked well when coding young adult interview transcripts. All service provider interviews were
double coded by the lead author and at least one research assistant; 10 young adult transcripts
(45%) were co-coded by the lead author and a research assistant, with the rest coded by the lead
author. Coders met after coding each interview to discuss their codes and come to an agreement
if discrepancies arose during coding.
With all transcripts coded, the lead author followed Braun and Clarke’s (2022) process of
developing themes. This included generating initial themes by reading through relevant coded
excerpts and finding similarities in meaning. For example, to generate themes related to
treatment barriers, coded excerpts tagged “barriers to treatment”, “discrimination”,
“effectiveness/need for treatment”, “need for freedom/autonomy”, “readiness for treatment”,
“shame/stigma”, and “access to treatment” were queried in Atlas.ti. Excerpts from young adults
and providers were analyzed separately, and it was determined if coded excerpts contributed to

100

similar themes for both groups. Memos were written by the lead author in order to document
analytic decisions and thematic development ideas as they arose. Thematic maps were also
drawn to find interconnections between candidate themes and refine them. Initial themes were
evaluated using the criteria of meaningfulness, coherence, and clear boundaries (Braun & Clarke,
2022). This was followed by reviewing, revising, and developing candidate themes by evaluating
them in light of all relevant coded excerpts. Theme definitions were written in order to define the
central organizing concept and boundary for each theme. Themes were finalized through the
process of writing the results, and any similarities or differences between young adults and
providers were described as a final step.
Results
In-depth interviews with young adults and providers led to the development of nine
themes related to barriers and facilitators to treatment. Most themes included information related
to barriers and facilitators, depending on the context. Quotes from both young adults and
providers contributed to all themes, but any differences between the groups in relationship to the
themes are noted below. Themes and exemplary quotes are displayed in Table 4.3. Minor edits
were made to some quotes for clarity. An illustration of how themes fit within an adapted
Behavioral Model for Vulnerable Populations is displayed in Figure 4.2.
Lack of Access to Treatment
Lack of access to treatment was described as an important barrier by young adults and
providers. Participants mentioned long wait times, lack of insurance, lack of affordable
treatment, and burdensome paperwork, making treatment harder to access. Young adults tended
to highlight long wait times, lack of preferred treatment (“I tried to find substance abuse

101

treatment for marijuana, but nobody does that. So, I still smoke weed” [25-year-old young adult,
never received treatment]), and lack of affordable treatment. Once treatment became available,
young adults were often no longer interested:
“But the amount of time that it takes for me to actually ask for the help and then receive
the help, that amount of time is like so fucking far apart (coughs). That like, I might not
want it anymore when the fucking shit comes around, you know? Because like maybe my
circumstances change. Get in a better spot, whatever, you know? Who knows, you
know?” (24-year-old male, received treatment more than 12 months ago).
These issues were largely echoed by providers, who appeared to have more experience
with the challenges associated with navigating entry into the treatment system. One of the main
issues raised by providers was that the severity of young adults’ substance use often required a
more intensive level of care (i.e., detox and inpatient or residential treatment), but there often
were not enough “treatment beds” available. Paired with the need to secure insurance coverage
for young adults, providers described the process of trying to access treatment as “grueling.”
Many providers mentioned that the process for accessing residential treatment involved having to
keep “checking in” to see if treatment beds were available:
“The nature of those kind of more residential program is there almost exclusively have
long waits to get in. And you have to kind of be able to check in every day. And it makes
it nearly impossible for homeless youth who are just trying to survive, to do what they
need to do, you know, to actually get in and then stay in that, you know, detox and then
residential program. So I think that system ultimately just fails these young adults, in
terms of, you know, helping them move along” (male physician).

102

Other issues raised exclusively by providers included a lack of developmentally
appropriate treatment. Many providers believed that adolescents and young adults have unique
developmental needs that are often not a good fit for the “more structured,” “abstinence only”
approaches that are found in older adult treatment programs. Difficulty finding insurance
coverage for young adult immigrants who were undocumented was also a barrier mentioned
exclusively by providers.
Need for Low-Barrier Treatment
To address the challenges associated with accessing traditional treatment options,
participants described the need for more open and accessible services that could facilitate
treatment engagement. One young adult mentioned the need for more outreach to help increase
young adults’ awareness of treatment services, specifically mentioning harm reduction services
(needle exchange programs) that don’t require abstinence. A majority of quotes contributing to
this theme came from providers, though, who believed that more accessible, “low-barrier”
treatment options would help facilitate treatment. This involved offering more open and
accessible treatment services (open “24 hours”) where young adults could “come and go” with
low paperwork requirements, in settings where many young adults already receive services (i.e.,
drop-in centers). Both young adults and providers mentioned that providers should “meet clients
where they’re at” by coming out to spaces where young adults already receive services. Having a
“one-stop-shop” setting for services would reduce the fragmented nature homeless services,
because it would build upon the trust that providers had already established with young adults:
“So having like, a one, one-stop-shop kind of place where you're able to access that
treatment in a place where you have already built rapport, you already have trust, rather
than it being like, ‘we refer you here, we refer you there.’ Because that's a lot of how the

103

homeless service system works in general, is you know, you finally build trust with these
people. And then it's, you know, we're sending you to the next person, and then you're
just getting jumped all around. And it's hard to build trust and connection that way” (male
intake worker/case manager).
Rigidity vs. Flexibility of Treatment
Substance use treatment programs (particularly inpatient programs) were described by
providers as having rigid rules and expectations that can result in disengagement from treatment
or not wanting to enter treatment in the first place. Although a majority of quotes for this theme
came from providers, one young adult who had previous experience in the foster care system
described how lack of autonomy in treatment may be a barrier for young adults who perceive
treatment as “systematic… like when you grow up in a system, it’s like broken people feel like,
they won’t have a voice or like people are going to tell them like, ‘you need to do this, you need
to be doing that’” (22-year-old female, never received treatment). Providers described how rigid
rules in residential treatment programs (e.g., young people’s phones were taken away) and other
restrictions on their freedoms deter young adults from treatment. Providers also believed that
“abstinence only” approaches to treatment were not appropriate or effective for most youth:
“But I think for a lot of our youth just the thought of ‘I can never have another beer or
another, you know, can't use marijuana anymore at all, for the rest of my life.’ That's a lot
to ask. Maybe needed at some point, but as a starting place. It's a lot to ask” (male social
worker).
In contrast, participants stated that a more flexible approach was needed, which could act
to facilitate treatment. Ideally, this would involve more transparency and choice in the types of
activities young adults could engage in during treatment. One young adult stated that he would

104

be willing to listen to a counselor’s advice regarding “activities that could take [my] mind off of
marijuana use,” but did not want to be “sitting down in a room talking for hours” about how
marijuana is “a problem” (25-year-old male, never received treatment). Providers also suggested
treatment programs should be more flexible when working with young adults with complex
challenges (e.g., co-occurring substance use and mental health disorders), and those who were
late or missed appointments, since young adults experiencing homelessness often have
transportation difficulties and competing needs, including multiple service appointments required
by housing, criminal justice, and child welfare systems them may be involved in.
Homelessness and Instability
Participants described how homelessness itself is a barrier to treatment. Young adults
described being trapped in a negative cycle of homelessness and substance use, making
engagement in treatment more difficult:
“I think it goes hand in hand with like, the fact they don’t have stable housing. They just
keep drinking more. But if they keep drinking, they can’t get stable housing. It just keeps
going back and forth” (19-year-old transgender young man).
Providers described how the experience of homelessness creates more barriers to
treatment engagement—“they have much less privacy, to make appointments, to attend
appointments, to do video appointments, to store medications to help them reduce use. You
know, without that basic level of security—housing—everything becomes more challenging”
(male physician). The basic need for housing was thought to be related to future success in
treatment: “if there is no change in the [housing] situation, nothing’s going to change” (24-year-
old male, never received treatment).

105

Furthermore, young adults’ and providers agreed that current and future opportunities for
housing could factor into their decision to engage in treatment—particularly residential
treatment. One young adult who had completed residential treatment described how his uncertain
housing situation after treatment was “the hardest part,” because he was preoccupied with his
“next move” (23-year-old male, received treatment more than 12 months ago). Securing housing
after treatment was described by providers as a serious challenge, as discharging young adults
from treatment without housing could undo any positive gains made during treatment. One
provider described how young adults’ precarious housing situation can influence their decision to
engage in treatment, particularly for those residing in shelters who are referred to residential
treatment by shelter staff:
“So it becomes an issue of ‘Okay, well, if you send me away to this treatment facility,
when I get released, Where will I go?’ So it's like, ‘I'd rather stay here and secure my bed
with the shelter, secure my housing, then go off to a treatment facility and then be back
homeless. Because If I'm gonna just leave the facility and go back to the street, then
going to rehab was pointless because I'm going to be right back exposed to the
environment that I was in, with the using, and the encampments, and being overwhelmed,
and having to self-medicate and cope’” (female substance use counselor, shelter).
Shame and Stigma
Shame and stigma associated with substance use were described as important barriers to
treatment by young adults and providers. Young adults and providers both described how some
individuals may avoid reporting their substance use to providers, or avoid seeking treatment, if
they anticipate being judged or stigmatized. One young adult remarked that young people who
use drugs “really do care about what people think,” and if they anticipate judgement, “it makes

106

them realize they can be doing better than what they are doing, and a lot of people don’t want to
think about that. So they’ll avoid it as long as they can” (24-year-old male, never received
treatment). Young adults and providers strongly advocated for a de-stigmatized approach to
substance use and treatment, specifically stating that stigmatizing language during treatment can
be counterproductive:
“You know they would say, ‘You can't help people who don't want to help themselves.’
Like, that person doesn't need to hear that. People don't understand that's like a mean
thing to tell someone. Like, ‘Oh, if you don't want to help yourself...’ But nobody wants
to be on the street, nobody wants to be addicted to substances, you know? Like, and
there's this whole concept that we or they or whoever—I’ll say myself included—brought
it upon ourselves, you know?” (23-year-old male, received treatment more than 12
months ago).
Providers described their previous experience with “old school” models of treatment that
are more “punishment oriented, same oriented” (female social work supervisor) and how stigma
has manifest in their work with other providers. Providers recounted instances in which other
providers held negative assumptions about individuals “not being ready” for treatment, or had
unrealistic expectations for young adults (e.g., “Why don’t they just stop using?” [female social
worker, children’s hospital]). Providers described how reacting negatively when a young person
relapses can contribute to stigma. One provider (a male physician) recommended the need for
more training to help staff approach relapse as a learning opportunity, and praise young adults
for their effort and progress, rather than shame them. Welcoming young adults to treatment with
genuine care and working collaboratively to develop treatment goals were described as more

107

effective and humane approaches than requiring total abstinence, which can make youth want to
“run away” [from treatment] (female physician).
Supportive Referrals
Young adults and providers expressed that having support persons they could talk to
about treatment could facilitate referral and treatment engagement. Support persons could
provide multiple benefits. They could help improve young adults’ self-esteem, which may
motivate young adults to seek treatment if the benefits of treatment are clearly explained to them:
by “showing people how to love themselves, letting them know that they’re worthy. And… then
inviting them to treatment after that” (25-year-old male, never received treatment). Young adults
stated that diverse types of support providers could be helpful, including family members, staff,
or individuals with lived experience of substance use who could serve in a mentoring role. The
main prerequisites appeared to be the presence of social supports who “really care”, are “good
listeners”, are dependable and available in times of need, have high expectations, treat young
adults “like family,” and are not just doing it as part of their “job.”
Providers tended to raise the importance of support persons in cases where young adults
are referred to treatment from drop-in centers and shelters, and often recalled best practices from
their own experience. Providers stated that referral and engagement in treatment is more
successful when young adults have a connection with their providers. One provider explained the
process of encouraging reluctant young adults to attend outpatient substance use counseling
while they reside in shelters:
“Yeah, so we just try to, always try to, paint it in a way that it's doable as well, right? A
lot of people have that impression that ‘All right, I'm gonna go into rehab, and be stuck in
there and not be able to go out.’ In this case, it's like, ‘No, it's just classes’ I think it's

108

twice a week now. Before COVID, I think it was three times a week. You go there, you
have a group and you learn what you learn, right? And I always tell everyone, ‘Whatever
group it is, you go, and if you feel you know, everything, well then share what you know.
You know, if they talk about 20 things and you learn two things, well, then you learn two
things, and those are two things you're going to use the rest of your life’” (male intake
worker/case manager, drop-in center/shelter).
Another provider who worked at a youth shelter described that referring young adults to
treatment requires careful planning and support, and assurance that they can return to the shelter
and receive services after treatment—rather than simply referring youth to treatment without any
support:
“Yeah, usually, I mean, any referrals would be with the agreement of the youth, because
to refer them somewhere they don't want to go is not gonna be helpful. So it'd be a
discussion and connection with the whatever resource that is outpatient or inpatient. And
we make sure they had transportation, that it was set up, that it was manageable for them.
And if it was inpatient, then we would probably help them transport and help them with
the referral process to take them there. And if they would, if they stay in the program, we
would then support them by visiting with them as much as possible” (Male social worker,
shelter).
Systems Referrals
Aside from being referred to treatment by staff at drop-in centers and shelters, referrals to
treatment through housing placements, child welfare and the criminal justice systems were
common treatment facilitators. Some young adults with previous treatment experience simply
stated that they received treatment through these sources, but it is unknown what impact these

109

had on them since they did not elaborate further. One young adult (24-year-old male) described
his experience of “not willingly” being “forced” into treatment while incarcerated. Another
young woman who was referred to treatment through supportive housing said, “I find it useful.”
Providers appeared to believe that it was unlikely that young adults would have voluntary
started treatment if it were not for these referrals, which—in addition to referrals from shelters—
were typically mandated as part of other behavioral health programs or court requirements. As
one provider put it, “I don't know if I've ever heard anyone say, ‘Yeah, I want to get into the
substance abuse treatment program, because I just need to clean up.’ (Laughs) You know what I
mean? It's just not… I haven't heard that. It's usually, it has to be forced or mandated” (female
nurse). Interestingly, some providers stated that receiving substance use treatment while
incarcerated resulted in positive changes among their young adult clients. Whether these positive
changes persisted long-term was unclear:
“Um it's been super unfortunate this has been the case, but we've had a lot of success with
um people going to (local jail) to get their mental health care. And from that, their
substance use will clear. So, they'll stop using for the most part… And so, we've had the
most success with communicating with the social workers who work inside (local jail)
about treatment plans, and recommendations, and discharge plans as well, and getting
people linked with services once they're released. Unfortunately, if there's no housing
there, when they're released, they go back to the streets and the risk of relapse is high, it's
almost guaranteed” (female social worker/program manager).
Safe Spaces
Young adults and providers stressed how important the social and spatial environment
was for accessing treatment services. One young adult stated that all people are “seeking

110

acceptance,” and the ideal treatment environment was somewhere where young adults could
“relax, chill, talk about some things... have a good positive environment” (24-year-old male,
never received treatment). Young adults preferred interactions with others that were “nice,
supportive and have like… more one-on-one resources with people that they feel comfortable
going to,” rather than authoritative, “tough love” interactions that many young adults were
accustomed to (23-year-old male, received treatment more than 12 months ago). Participants
described the ideal treatment setting as similar to a drop-in center, where young adults felt safe,
supported, and comfortable, and can engage in relaxing activities such as making art.
Providers noted that young adults can feel uncomfortable in a new treatment setting
where they haven’t established rapport with providers and staff. One provider clearly described
the benefits of safe spaces:
“So it seems that being viewed as a safe space really helps with trust. And if your clinical
space is viewed as a safe space, patients are more likely to make an effort to go there,
because they want safety anyway. And they're seeking safety, they're seeking a sense of
stability. So if you provide that even for, you know, an hour at a time, that itself is an
incentive to come in and try to get help” (male physician).
Readiness for Treatment
Readiness for treatment was an important factor that could be a barrier or facilitator to
treatment. Young adults expressed doubt that every young person experiencing homelessness
with a substance use problem “really wants the help,” and agreed with that treatment would not
be effective if it was “forced” on them:

111

“So... really, if they're trying to get the help, I would say that anyone that needs the help,
wants the help, is willing for the help and actually is... say you're the help, they're willing
to reach out to you for the help... I think ‘yes’, that would be very beneficial on them, but
forcing the help towards people, I feel like ‘no.’ They… that's something they need to
make themselves” (21-year-old male, never received treatment).
Treatment was viewed as beneficial if the “people who need it… [are] actually willing to
try” (Young Adult 7). Young adults suggested the idea that lack of readiness to engage in
treatment was more common among those who perceived benefits of substance use, particularly
among those who used cannabis to cope with various life stressors. Young adults also suggested
that individuals would be more willing to engage in treatment if the right incentives were
offered, or as one young adult put it, “if there was a promise on the other side that like, ‘Oh, if
you do A, B, and C, then here’s the incentive” (23-year-old male, received treatment more than
12 months ago). For one young adult, his incentive to attend treatment was the promise of a
housing placement: “Housing was my biggie because I just wanted to get housing. So I stopped
that substance abuse shit” (25-year-old male, received treatment more than 12 months ago).
Quotes from providers generally agreed with those of young adults. Providers described
that young adults who are “in the throes of their addiction” or are in a “pre-contemplation stage,”
have difficulty accepting treatment or counseling (female, social work supervisor). Providers
suggested that individual readiness for treatment was one of the most important—perhaps the
most important—factor determining treatment engagement. One provider summarized this, by
acknowledging that structural barriers to treatment do exist and are important, but an individual’s
internal barriers are important to consider as well: “there are treatment centers, and yet, there
probably are not enough… a lot of them, you have to have MediCal [insurance], that’s a

112

barrier…but, it is a lot of getting someone willing to do it” (male psychotherapist). Providers
recounted various factors which could affect an individual’s readiness for treatment. These
included major life changes, traumatic events, and harms associated with substance use. For
example, overdosing on alcohol or other drugs, pregnancy or the birth of a child, and hope
associated with a new job can contribute to one’s motivation to change and readiness for
treatment. As one provider stated, “if they have a realization that like, ‘this is harming me,’ I
think it's really the most successful precipitator to getting treatment” (male, intake worker/case
manager).
Discussion
There are many barriers to substance use treatment for young adults experiencing
homelessness. The present study investigated these barriers—as well as facilitators of
treatment—among a sample of young adults who met criteria for substance use disorder and
currently experiencing homelessness in Los Angeles. A diverse group of service providers, each
with 12 years’ experience providing services to this population, on average, were also
interviewed to understand their perspectives of these barriers and facilitators. The barriers and
facilitators identified were largely consistent with previous research conducted with young adults
experiencing homelessness (Hudson et al., 2008; Nyamathi et al., 2007; Rabinovitz et al., 2010).
There were systemic barriers to care, such as lack of available treatment options and difficulty
obtaining insurance coverage. There were organizational-level barriers, such as rigid rules,
expectations, and program structures that discouraged young adults from seeking treatment.
There were interpersonal barriers, including shame and stigma, which could be enacted by
service providers and internalized by young adults (Hudson et al., 2008). And finally, there were
individual-level barriers, namely low internal readiness for treatment. One unique theme that

113

emerged from this research was the impact of homelessness itself as a barrier to treatment.
Homelessness not only made it more difficult for young adults to carry out activities necessary
for treatment (attending appointments, et cetera), but also acted as a barrier if young adults
believed that treatment would interfere with their current housing placement or would not help
with their long-term housing goals.
This study also added to existing evidence on perceived facilitators of substance use
treatment in this population. Consistent with previous research (Hudson et al., 2010; Nyamathi et
al., 2007), young adults and providers identified the need for support persons who express
concern about young adults’ harmful substance use behavior, and are willing to support them as
they participate in treatment. Referrals through various “systems” (juvenile justice system, foster
care system, housing placements, etc.) also appeared to facilitate treatment, although it is unclear
how successful this was in decreasing substance use problems long-term. As a whole, young
adults and providers agreed with the idea that treatment would not be successful without “buy
in” from young adults, highlighting the importance of individual readiness for treatment. In fact,
this intuitive idea has some empirical support (Wild et al., 2016). One’s readiness for treatment
likely interacts with other facilitators at the organizational level, such as the degree to which
treatment programs and providers are flexible when working with young adults, and degree to
which treatment settings are perceived to be “safe spaces,” which is more likely when programs
adopt a de-stigmatized approach to treatment (Slesnick et al., 2008; Society for Adolescent
Health and Medicine [SAHM], 2018).
While young adults and providers generally agreed on the above themes, some important
differences between the groups are worth noting. Young adults and providers have different lived
experiences, social roles, and expectations, which influence their views on treatment. Given their

114

current unstable situation, young adults were generally most concerned with getting their basic
needs met through services provided at the drop-in center, and two-thirds had no direct
experience receiving substance use treatment. Importantly, not all young adult participants
acknowledged that they personally needed or would benefit from treatment, despite meeting
criteria for probable SUD. Given this, most young adults’ beliefs about barriers and facilitators
tended to be based on their preconceived notions of treatment, and ideas regarding what might
work for “other” young adults experiencing homelessness. Still, those who had been through
substance use treatment provided observations that were largely consistent with those of
providers (e.g., lack of available treatment, shame and stigma in treatment). The unique
perspectives of young adults were most pronounced when describing homelessness as a barrier
to treatment, when stating their preferences for ideal social support persons, and the need to offer
more incentives to engage in treatment.
In contrast to young adults, service providers had years of specialized training and
experience working with diverse groups of youth experiencing homelessness. As such, they
tended to have a more “behind the scenes view” of the service system. This led not only to
highly detailed descriptions of the systemic, interpersonal, and individual barriers to treatment,
but also proposing more concrete solutions, which may be most relevant for provider audiences.
This was particularly true when describing the need for more flexible, low-barrier treatment
options; the need for more provider training to avoid stigma; and best practices for supportive
referrals. Providers’ experience working with a large number of young people over the years also
made them attuned to the diverse perspectives and needs of youth (i.e., their internal readiness
for treatment). Although they may not have had lived experience of homelessness, providers
were deeply committed to understanding and advocating for the needs of youth.

115

Conclusion and Future Directions for Research and Practice
More research and practical work is needed to determine how to best intervene to remove
barriers and enhance facilitators of substance use treatment. Results suggest that such
interventions to increase service utilization may be most effective if applied to multiple social-
ecological levels (Sallis et al., 2008). Policy interventions could target systemic barriers, by
increasing access and availability to appropriate services across service settings, and by
transforming substance use treatment systems to be more flexible. This may be achieved by
providing more funding for substance use treatment services, including harm reduction.
Interventions targeting interpersonal-level interactions could involve training providers across
service settings to utilize trauma-informed care (SAHM, 2018), and utilizing peers, case
managers, and other natural support persons to help engage young adults in treatment (Barker &
Maguire, 2017; Bassuk et al., 2016; Dang et al., 2014; Kidd et al., 2019). Individual levels
interventions may involve the use of screening, brief intervention, and referral to treatment
(SBIRT) to identify young adults who may benefit from treatment, enhance internal motivation
for change, and troubleshoot barriers to care (Brolin et al., 2022). Research shows that substance
use treatment can effectively reduce substance use problems in this population (Coren et al.,
2016; DiGuiseppi et al., 2021; Zhang & Slesnick, 2018). Removing barriers to care is now
desperately needed.

116

Table 4.1 Characteristics of Young Adult In-depth Interview Participants (N = 22)
M or n SD or % Range
Age 23.4 1.5 19 - 25
Birth Sex

Male 15 71.4%

Female 6 28.6%

Gender Identity (if different from birth sex)

Transgender (female to male) 1 4.8%

Genderqueer / non-binary 3 14.3%

Race/Ethnicity

Non-Hispanic White 5 22.7%

Hispanic, Latino/a 2 9.1%

Black / African American 3 13.6%

Asian 1 4.5%

Native Hawaiian / Pacific Islander 1 4.5%

Other race or ethnicity 3 13.6%

More than one race or ethnicity 7 31.8%

Sexual Orientation

Straight or heterosexual 13 61.9%

Lesbian or gay 2 9.5%

Bisexual 4 19.0%

Queer, pansexual or questioning 2 9.5%

Education

Less than high school diploma 4 18.2%

High school diploma / GED 10 45.5%

Some college 5 22.7%

Associates or Bachelor's degree 3 13.6%

Currently in School 0

Employment

Employed part-time 2 9.1%

Unemployed, not looking for work 5 22.7%

Unemployed, looking for work 15 68.2%

Age first homelessness 18.3 2.4 13 - 23
Years spent homeless (lifetime) 3.3 2.6 0.4 - 11
Received Substance Use Treatment

Never 15 68.2%

Yes, in the past 12 months 4 18.2%

Yes, more than 12 months ago 3 13.6%
Note. 1 participant declined to report their sex, gender, and sexual orientation.

117

Table 4.2 Characteristics of Service Providers (N = 16)
M or n SD or % Range
Age 42.5 11.8 25 - 62
Gender

Male 7 44%

Female 9 56%

Race/Ethnicity

White 9 56%

Hispanic or Latino/a 3 19%

Black / African American 2 13%

Asian 1 6%

Multiracial 1 6%

Education

Bachelor's degree 3 19%

Master's degree 8 50%

Doctoral degree (PhD, EdD, etc.) 1 6%

Professional degree (M.D., PsyD, etc.) 4 25%

Current Professional Role

Intake worker or case manager 5 31%

Social worker 5 31%

Nurse 1 6%

Physician 3 19%

Substance use counselor 1 6%

Administrator 2 13%

Other (program manager, psychologist,
supervisor, therapist, psychotherapist) 5 31%

Experience in role

Years 11.6 8.9 1 - 33
Months 3.8 3.8 0 - 10
Note. Participants could indicate more than one professional role

118

Figure 4.1 Problematic Substance Use Among Young Adults Completing Screening Survey

Note. N = 50; Tobacco, Alcohol, and Cannabis Use Disorder indicated by TAPS subscale scores
≥ 2; Problematic use of stimulants, heroin, and prescription medications indicated by TAPS
subscale scores ≥ 1.
60%
36%
64%
22%
6%
12%
12%
6%
80%
44%
Tobacco
Alcohol
Cannabis
Stimulants (cocaine, methamphetamine)
Heroin
Prescription opiate pain reliever
Prescription sedative (e.g., Xanax)
Prescription stimulant (e.g., Adderall)
1+ SUD (excluding Tobacco)
2+ SUDs (excluding Tobacco)
Percentage of Young Adults

119
Figure 4.2 Themes Displayed within an adapted Behavioral Model for Vulnerable Populations

120

Table 4.3 Themes and Exemplary Quotes from Young Adults and Providers
Theme Young Adult Service Provider
Theme 1: Lack of Access
to Treatment
But the amount of time that it takes
for me to actually ask for the help
and then receive the help, that
amount of time is like so fucking far
apart (coughs). That like, I might not
want it anymore when the fucking
shit comes around, you know?
Because like maybe my
circumstances change. Get in a better
spot, whatever, you know? Who
knows, you now? (24-year-old male,
received treatment more than 12
months ago)

I realized once upon a time, I’m
addicted to marijuana. And I tried to
find substance abuse treatment for
marijuana, but nobody does that. So,
I still smoke weed (25-year-old
male, never received treatment).
Also, like when somebody is ready,
and they're like, "I, I'm done. this is
exhausting. I want treatment. help
me get into treatment." And we call
like, the substance abuse hotline to
like, do that assessment. And you go
through this, like, grueling process
of them asking, you know, so many
questions. And then it's like, a
waiting game of if there's going to
be a bed, and then sometimes, most,
in my experience, there's never been
a bed in that moment. And so then
that moment has passed. And so
then you either have to start over
again. Or if the person is like,
"Okay, I'll keep checking in. I'll see
if there's a bed." Maybe it lasts like a
week or two. And then maybe they
fall back out of like interest in in
pursuing that because now it's
become like a little bit more
difficult… finding a bed and when
that person was ready, compared to
like, when there's a bed ready, those
don't ever seem to match up (female
social work supervisor).

121

Theme 2: Need for
accessible ("low-barrier")
treatment options
Um... I don't know, because I know,
there's just this one spot here that the
clinic, they, I would go by every
time, they do testing and things like
that. But they also, I know they also
give like people clean needles and
things like that. Even though I know
that, you know, because they know
people are still going to do the drugs,
but they want them to at least be
injecting themselves with
instruments that are clean and not
dirty or sanitary. I kind of like that
approach, even though it's not like, a
like, "Oh, we're gonna try and stop
this" per se. But they're, they are
decreasing a lot of other health
hazards that, you know, come with
those things (23-year-old male, never
received treatment).
If there was like, one physical space
where it's like, they can come to get
services, maybe like you get here,
but there's like a built in substance
abuse program where it's like, "Hey,
I know we've been talking about
this, there's someone like right down
the hall that can, that can start
having these conversations with you,
like, would you just want to meet
them today?" (male intake
worker/case manager).

We've tried it, where we had them
come to us, but it just, it was just too
much of a barrier, I think. We're a
little bit further east than where a lot
of our shelters are located. And so
it's a little bit of a walk if they're
going to walk over, which is the
primary form of transportation for
themselves. And then bus, you
know, fare, may be out of their
reach. So, yeah, it's just much easier
to go to them and see them in their
habitat, you know, where they're
currently residing (female
physician).

122

Theme 3: Rigidity vs.
Flexibility of Treatment
Programs
Because um, a lot, a lot of it is also
like, they might feel systematic. I
feel like that plays a huge part…
Like, like when you grow up in a
system, because it's like broken
people feel like, it's like, they won't
have a voice or like people are going
to tell them like "you need to do this,
you need to be doing that." (22-year-
old female, never received treatment)

Like if you can somehow, or
somehow they get a schedule where
I can meet somebody here at (drop-
in center) once a week I would be
interested. Just to talk. Not to take
no... Like, I would take, listen to
their advice, but I'm not gonna take it
if it's them telling you that, "it's a
problem." But no, I would just yeah,
have somebody to vent to. Yeah, I
would do that (25-year-old male,
never received treatment).
I mean I think there's something to
be said about having structure.
Because then you know what the
rules are, what your expectations
are, but they can't be so freakin
rigid. That I think we have to
remember how do we want to live
our lives? Do we want to be able to
access a fridge and get a get a cup of
milk? Or am I going to get written
up for that? (female social worker).

But I think for a lot of our youth just
the thought of "I can never have
another beer or another, you know,
can't use marijuana anymore at all,
for the rest of my life." That's a lot
to ask. Maybe needed at some point,
but as a starting place, it's a lot to
ask (male social worker).

But yeah, I feel like flexibility is like
when places and like the
expectations of somebody who is
not able to maybe hold appointments
and hold and like readiness, you
know, like when they're ready to
take the next step to wanting to be or
at least, make sort some sort of an
attempt towards sobriety. I feel like
most of the programs are just very
rigid, and it's hard to move that's like
a big barrier (female social work
supervisor).

123

Theme 4: Homelessness
and Instability
If there is no change in the [housing]
situation, nothing’s going to change
(24-year-old male, never received
treatment).

I think that what makes it hard to go
into treatment is that there's no like
promise on the other side. Like, you
know, even, even a like a certificate
or something is like a cool sense of
accomplishment. But when, when
you get in there, and you're so used
to like not attending to your
responsibilities, or maybe even
before you go. That like, "Okay, like
what if I like go put all this like work
in, or sober time, or workbooks or
whatever the fuck you have to do in
rehab?" But then the whole time
you're trying to, you can't think
about like yourself, and what you're
feeling and all that, because you had
to think about what your next move
is. Like, where are you going to go?
Where are you going to stay? Where
are you going to sleep? Who's going
to be there for you? Is this going to
change my life immediately? Or like,
whatever? That's the hardest part
(23-year-old male, received
treatment more than 12 months ago).
So it becomes an issue of "Okay,
well, if you send me away to this
treatment facility, when I get
released, Where will I go?" So it's
like, "I'd rather stay here and secure
my bed with the shelter, secure my
housing, then go off to a treatment
facility and then be back homeless.
Because If I'm gonna just leave the
facility and go back to the street,
then going to rehab was pointless
because I'm going to be right back
exposed to the environment that I
was in, with the using, and the
encampments, and being
overwhelmed, and having to self-
medicate and cope" (female
substance use counselor).

124

Theme 5: Shame and
Stigma
As far as the person touching it, they
really do care about what people
think… Well a person about drugs
would think, if I'm high off sherm,
coke or whatever, I would see a
[drop-in] center like this, and
physically able to see it walking past.
I will look at it like "Damn. I'm
hungry or something." Now you
because of the way you look,
because of how high you are. And
they really don't know like, you
really a person that need help. You
know, there is people out there like
that. Luckily, there's good people
like this, but places, East LA, West
LA, Long Beach, stuff like that.
That's where they gonna look at you
like you're a problem, they're not
going to try to help you (24-year-old
male, never received treatment).
There's a very, like old school model
to treatment. And unless you're
going to some probably like newer,
like, like, I don't know... Even the
model that I worked in, when I was
in a treatment facility for minors,
like, it's very... it's very old school.
It's like, definitely more like
punishment oriented, shame oriented
(female social work supervisor).

We still hear a lot of that language
and like, "Oh, they're addicted, you
know, they're drug users, drug
abusers um... Why don't they just
stop?" So even among providers
there needs to be a shift in the way
that we think about it, and the way
that we treat, um, our young people,
as they're really trying the best they
can. And educating ourselves about
like, what impacts, you know,
chronic substance use has on the
brain and the body (female social
worker).
Theme 6: Supportive
Referrals
So showing people how to let go the
negativity that they feel for their own
life. Helping them love themselves,
letting them know that they're
worthy of so. And... then then
inviting them to treatment after that
(25-year-old male, never received
treatment).

So, my auntie, she told me, "Hey, I
want you to clean up and you know
put everything together. Because
you're just losing it." And that was a
truth. Like, in those days, I was
literally, like, you know? I don't like
thinking about those things (24-year-
old male, received treatment in the
past 12 months).
We got that with, especially with,
like the two guys that are going at
the moment, we let them know, and
"Oh, no, I don't have an issue. I'm
not an addict," and things of that
nature. You know, we spoke to them
after a while, and they're like,
"Okay, I'll do it. I'll try it out."
Right? I think having that
connection. You know, having
someone who they respect being
able to tell them upfront, "Hey, you
know, you need this support." I
think, personally, I think that's what
makes a huge difference. Okay.
Rather than just, you know, the
stranger came in, "Hey, you should
be going to [substance use] classes if
you're not gonna listen" (male intake
worker/case manager).

125

Theme 7: Systems
Referrals
Interviewer: So, have you ever
received any treatment or counseling
for substance use before?

24-year-old male: Not willingly. You
could say that I was forced. I was in
camp. They told me I had to go to
substance abuse classes like I was a
drug head or something. So I'm like,
if it's for the good program or getting
my points up, and staying out the
box, I do that. So, I guess for a while
like two months, I think. They had
me do the drug abuse class.

Interviewer: But you didn't think you
needed it?

24-year-old male: Hell no. Hell no.
You know, a certain percentage end
up in jail, probably to their benefit,
because they tend to do better when
they're, you know, incarcerated in
terms of getting sober in treatment.
So it's not always a bad thing, in
terms of accessing services. I don't
have a gut feeling whether in the
long run, that actually ends up
helping them or not, but sometimes,
kids that are spiraling so fast, that,
you know, nothing's gonna help
them unless they can get out of that.
And sometimes jail is the best (male
physician).
Theme 8: Safe Spaces It's a safe space. That's what I love
about (drop-in center). I think, I
think really like, what they're doing
here, I think we should have more of
that, like giving youth or people in
homelessness an outlet. That's, that's
really all I think can help fix things,
because it's distracting them from
that. And they're enjoying it. Yeah,
especially with the (name of
program), like painting, like giving
them an opportunity to come out
here and paint... Like I think that's
great (23-year-old female, never
received treatment).
So it seems that being viewed as a
safe space really helps with trust.
And if your clinical space is viewed
as a safe space, patients are more
likely to make an effort to go there,
because they want safety anyway.
And they're seeking safety, they're
seeking a sense of stability. So if
you provide that even for, you
know, an hour at a time, that itself is
an incentive to come in and try to
get help (male physician).

126

Theme 9: Readiness for
Treatment
I'm gonna say this--Yes, I do believe
that. I believe it [treatment] would be
very beneficial. And the only
problem is this--do they really want
the help? That's my hardest thing to
think about is like, if they really
want the help--and it's hard for me to
actually figure out who wants the
help or not. (21-year-old male, never
received treatment).

Yeah, housing was my biggie
because I just wanted to get housing.
So that stopped that substance abuse
shit (25-year-old male, received
treatment more than 12 months ago).

Yeah, I felt like, I got what I needed
out of it. I am questioning on going
back, because there was one time I
started drinking, and I swear, I
almost died. And I have blacked out
and it was my first time and I
panicked and everything. And when
they took me to the hospital, I didnn't
even like, remember them, like
taking blood samples or nothing.
That was very scary (22-year-old
female).
But I don't know, it, the big thing
that you learn when you work in this
space is that you can't force anyone
to do anything. They have to come
to the realization themselves that
they want help. And they have to be
committed and ready to do that. And
our job is not to push them to do
that, not to force them to do that.
But it's to be there and guide them
along the path to themselves
realizing that they need assistance.
Because it's not going to work, if
you're forcing people to get clean.
Just not going to work (male intake
worker/case manager).

Well, certainly people will have
ambivalence about giving up a
substance. Especially if the
substance has helped them cope for
a long time. People who feel like
they have very little, are naturally
afraid of giving up something what
they do have (male physician).

I think most importantly, like youth
have to lean in, and to lean in they
have to want to (female social
worker / program manager).

Everyone's journey is different, but
whenever they have, if they have a
realization that like, "this is harming
me." Um, I think it's really the most
successful precipitator to getting
treatment (male intake worker/case
manager).

127

Appendix A: Young Adult Qualitative Interview Guide
Introduction (before recording has started)
Thank you for participating in this study. The goal of this study is to understand
substance use and substance use treatment needs and experiences among young adults who are
experiencing homelessness or housing instability. You were asked to take part in an individual
interview that is expected to take up to 1 hour.
During this interview, we’d like to talk about your experience with housing instability,
substance use, and substance use treatment. We’d also like to hear your thoughts about how
these issues may affect young adults in particular, and your ideas for what young adults
experiencing homelessness might want out of treatment, or what treatment might be most
effective for this group.
After starting the recorder I will say my name, the date, and the number linked to your
interview, and will ask you to share your preferred gender pronouns. As a heads-up, I will
sometimes ask you to define words that might seem obvious to make sure everything is clear.
My preferred gender pronouns are ____ /____. Is it okay if I start recording?
**BEGIN RECORDING**
Interviewer: State your name, date, and interviewee’s PID
Domain: Homelessness experiences
1. Can you tell me about your experience with homelessness or housing instability? From
the first time up to now?
a. How long have you been homeless or unstably housed?
b. What first contributed to you becoming homeless?
c. What is contributing to your current challenges finding stable housing?
d. How is the experience of being homeless or unstably housed different for young
adults, compared to adolescents? Compared to older adults? How so?
Domain: Substance use
Now I’d like to talk about substance use. I’m interested in the role that substance use plays in the
lives of unstably housed young adults. I’m interested in the reasons why unstably housed young
adults might use alcohol or drugs, what kinds of drugs are most often used, and how alcohol or
drug use might be helpful or harmful for unstably housed young adults. You can speak from your
own experience or from what you’ve seen and observed.
1. What kinds of substances do you typically use?
a. E.g., Alcohol, cannabis/marijuana, methamphatamines, opioids, etc.?
b. Do you have a drug of choice?
c. Are you interested in changing your substance use – if so, in what ways?

128

d. What kinds of drugs are most often used among the people you’ve met (friends,
associates, etc.)?

2. What are some good things about substance use?

3. What are some not so good things about substance use?

4. What do you think are the main reasons why unstably housed young adults might use
alcohol or drugs?
a. For social reasons (or because other people around you are using)?
b. To cope with your problems?
c. For “fun” or to enhance your experience?
d. Out of boredom?

5. Thinking about your most recent homelessness experience, did your substance use
change (before and after)? Did substance use have anything to do with it?

6. Do you think substance use plays an important part in someone getting stable housing?
a. Does it make it harder or easier to attain stable housing?
b. Do you think young adults’ substance use behaviors change when they become
stably housed?
c. IF decline to answer: what about for other young adults experiencing
homelessness?

Domain: Treatment experiences and beliefs
Now I’d like to talk about substance use treatment. I define treatment as “as any treatment or
counseling received for alcohol or illicit drug use, or for medical problems due to use of alcohol
or illicit drugs.” I’d like to hear about your experience with treatment, or what you’ve heard
about it. I will also ask about your thoughts on what might be the best kinds of treatment for
unhoused young adults (if any).
7. Have you ever received treatment for substance use problems?
Ask if participant has received treatment:
8. What was the treatment/counseling experience like for you?
a. Who, what, where, when, why did you receive treatment?
b. Was it helpful? Not helpful?
c. Would you do anything to change it or make it better?

9. How did you decide to enter treatment?
a. Was there anything that made it harder to start treatment?
i. What made it easier for you to start treatment?

129

b. Was there anything that made it harder for you to continue and keep coming back
to treatment?
i. Was there anything that made it easier for you to continue and keep
coming back to treatment?

10. Would you recommend treatment for other young adults who are experiencing
homelessness?
a. Why or why not?
Ask if participant has NOT received treatment:
8. When you hear someone talk about substance use treatment, what does that look like to
you?
a. How do you imagine it?

9. Do you think treatment or counseling would be helpful for you?
a. Or other young adults experiencing homelessness?
b. Why or why not?

10. What makes it harder for young adults experiencing homelessness to start or keep going
to treatment?
a. What might get in the way?

11. What would make it easier for young adults experiencing homelessness to start or keep
going to treatment?
a. What might help?

Ask of all (received treatment or not):
12. What would an ideal substance use treatment or counseling program for unstably housed
young adults look like?
a. Who might be involved?
b. What kinds of activities would you do?
c. How could you best be supported?
d. How do you think young adults who use substances can reduce their use and
maintain their sobriety over time if they are unstably housed?
Domain: Emerging adulthood goals
13. What’s the next step for you?
a. Where do you see yourself in the near future – say, 6 months from now?
b. How about the longer-term future – say, 5 years from now?
Thinking back through what we talked about today, is there anything else you’d like to add?
Thank you.

130

Appendix B: Service Provider Qualitative Interview Guide
Introduction (before recording has started)
Thank you for participating in this study. The goal of this study is to understand
substance use and substance use treatment needs and experiences for young adults who are
experiencing homelessness or housing instability. As a service provider, you were asked to take
part in an individual interview that is expected to take up to 1 hour.
During this interview, I’d like to talk about your experience providing services to
unstably housed young adults with substance use behaviors that may be causing problems for
them, and also the role of substance use treatment in the lives of these young adults. In particular,
we’d like to hear your thoughts about how these issues may affect young adults (compared to
other age groups), and your ideas for what young adults experiencing homelessness might want
out of treatment, or what kinds of treatment might be most effective for this group.
After starting the recorder I will say my name, the date, and the number linked to your
interview. Is it okay if I start recording?
**BEGIN RECORDING**
Interviewer: State your name, date, and interviewee’s PID
Domain: Service provision for unstably housed young adults & substance use
2. Can you tell me a little bit about your experience providing services to young adults who
are homeless or unstably housed?
a. Is the experience of homelessness different for young adults, compared to
adolescents or older adults? How so?
Domain: Substance use
Now I’d like to talk about substance use among your clients. I’m interested in the role that
substance use plays in the lives of homeless or unstably housed young adults. I’m interested in
your thoughts on the reasons why unstably housed young adults might use alcohol or drugs, what
kinds of drugs are most often used, and how alcohol or drug use might be helpful or harmful for
unstably housed young adults.
1. For the youth you work with now, what kinds of substances do they seem to be using
most often?
a. E.g., Alcohol, cannabis/marijuana, methamphatamines, opioids, etc.?

2. What do you think are the main reasons why unstably housed young adults might use
alcohol or drugs?
a. For social reasons (or because other people around them are using)?
b. To cope with their problems?
c. Out of boredom?

131

d. For “fun” or to enhance their experience?

3. In your experience, what role does substance use play in the lives of unstably housed
young adults?
a. Does substance use contribute to homelessness or unstable housing, or is it more
often the case that some young adults develop problematic substance use after
becoming homeless?

4. What role does substance use play in young adults’ transitions to stable housing?
a. In what ways does substance use make it harder or easier for them to attain stable
housing?
b. In what ways might substance use habits change when they attain stable housing?
Domain: Treatment experiences and beliefs
Now I’d like to talk about substance use treatment. We define treatment as “as any treatment or
counseling received for alcohol or illicit drug use, or for medical problems due to use of alcohol
or illicit drugs.” I’d like to hear about your experience with treatment among the unstably housed
young adults you have worked with. I will also ask about your thoughts on what might be the
best kinds of treatment for unhoused young adults.
5. Could you tell me about the common treatment or counseling practices provided to the
unstably housed young adult clients you’ve worked with?
a. Who, when, where, what kinds of treatments are delivered?
b. Any manualized treatments used?
c. IF not delivered in-house, what is the referral process like?

6. In your experience, what are some of the key barriers that get in the way of young adults
entering treatment?
a. Client barriers vs. systemic barriers?
b. What helps unstably housed young adults begin treatment? (i.e., facilitators?)

7. What are the barriers that get in the way of young adults staying engaged (keep coming
back) to treatment?
a. Client barriers vs. systemic barriers?
b. What helps unstably housed young adults stay engaged (keep coming back) to
treatment?

8. Some define treatment success as total abstinence and others define it as a reduction in
harmful substance use. What does substance use treatment “success” look like for you, as
a service provider?
a. What are the factors that lead to treatment success for unstably housed young
adults?

132

b. What about young adults who self-resolve their substance use without the help of
professional treatment? What factors might be responsible for their recovery?

9. What would an ideal substance use treatment or counseling program for unstably housed
young adults look like?
a. Who might be involved?
b. What activities might providers and clients engage in?
c. How could they best be supported?

10. What about housing during treatment? Should they be integrated? What happens when
treatment is over but housing is still needed?
Domain: Emerging adulthood goals
Now I’d like to talk about young adult goals and developmental trajectories for these youth that
you’ve worked with. Some have called this period of life between ages 18 and 24 as “emerging
adulthood.”

11. Are there different pathways you’ve seen among the young adults’ you’ve worked with?
For example, do some young adults achieve sobriety and stable housing, while others
continue to have problematic patterns of use (and possibly, more housing instability)?
a. What might contribute to these different pathways?
Thinking back through what we talked about today, is there anything else you’d like to add?

Thank you.

133
Appendix C: Qualitative Codebook
Code
Definition
12 step / mutual aid groups Use or discussion of 12-step fellowship groups (AA, NA, etc.)
access to treatment

adverse experiences (general) Code whenever a general adverse experience is mentioned (not as a reason for drug
use, not overdose, not general family abuse) --> use other pre-existing codes for
those.
Example: Pt 10 mentioned getting shot while playing basketball, derailing his
basketball career.
barriers to housing Barriers to housing (not substance use related). For substance use, use "substance use
as a barrier to housing".
barriers to treatment

biology / genetics / physical dependence Mention of genetic predisposition / family history, or other biological factors related
ot substance, including physiological dependence on certain drugs, tolerance,
withdrawal, etc.
boredom Boredom or lack of meaningful activities to occupy one's time. May be a reason for
substance use.
chronic homelessness

counseling Participants' description of specific counseling strategies used (i.e., topics to focus
on, etc.).
COVID-19 Any mention of the COVID-19 pandemic. How it affects treatment or daily life for
YAEH.
cultural competency

defining treatment "success" Code participant responses to the following question, "What does treatment success
look like for you?"

134
discrimination Perceived or actual discrimination due to one's housing status, race/ethnicity, or other
characteristics.
Discrimination may be overt or implicit.
education / employment

effect of housing on substance use How does substance use change AFTER young adults attain some sort of housing?
(Can be shelter, transitional living, or more pemanent housing)
effectiveness / need for treatment Is treatment needed, and does it work? Any mention of how effective treatment is (or
isn't).
This may overlap with "recovery process"
emerging adult development / personal
goals
Any mention of developmental issues and needs specific to emerging adulthood or
among "Transition Age Youth"/TAY (age 18 - 25). May be in response to the final
question about emerging adult development.
environmental/neighborhood factors Factors in young adults' surrounding environment (community, neighborhood,
physical surroundings, etc.) that could be serve as risk or protective factors for
substance use and/or housing outcomes.
family relationships Discussion related to YAEH's family members (most often their parents or
caregivers). May be discussed in terms of contributing reasons for homelessness
(caregivers kicked them out) or supports during recovery, etc.
flexibility This is an in-vivo code from provider 2, who stated that an ideal treatment program
would be "flexible" in order to accomodate YAEH's needs. May also relate to "low-
barrier" services (i.e., "meeting youth where they're at")
food / nutrition

good quote Code good quotes here.
harm reduction Strategies or practices focused on reducing harms associated with substance use.
Some examples include counseling YAEH's on safer/less risky ways of using
substances, or describing the need for programs/practices that are more tolerant of
YAEH's use. May want to file this under a broader "types of services offered"
code.
hope

135
housing during or after treatment YAEH's need for housing during or after treatment. May be in response to the
interview question asking about this. Note that this is different from "barriers to
treatment" or "barriers to housing", because it is focused on programs/policies
addressing YAEH's need for housing specifically during or after treatment.
ideal treatment May be in response to question, "What do you think is an ideal substance use
treatment program for YAEH?" Do not code "types of treatment services offered"
when respondents are describing their ideal treatment. However, maybe code "ideal
treatment" if young adults are asked, "What does treatment look like to you? How do
you imagine it?"
initiation/progression of substance use Description of how substance use behaviors develop over time, can include initiation
of substance use context.
This is more specific than the "reasons for substance use" code, because it focuses on
development.
meeting goals In-vivo code from provider 2, who described that client's meeting their goals could be
a successful treatment outcome. May be in response to the question, "How do you
define treatment success?"
mental health Mental health issues relating to YAEH's homelessness, substance use, or treatment
experiences. May be in the context of a coversation about co-occurring disorders or
poor mental health as a risk factor for homelessness, or a consequence of
homelessness.
need for freedom / autonomy YAEH's needs. May be a barrier to YAEH engaging in treatment programs that have
a lot of rules, or may be an important developmental factor to consider for YAEH in
general.
negative consequences of housing Consider merging with "effect of housing on substance use"? then renaming?
negative consequences of substance use Negative consenquences of use.
For negative consequences on attaining or maintaing housing, use "substance use as a
barrier to housing" code instead.

136
overdose Mention of a drug overdose due to excessive use of opioids, alcohol, or other drugs.
For other, more general consequences of substance use, use "consequences of use"
code instead.
parenting Having children or parenting (young adult as the parent). Parenthood may be a factor
supporting YAEH's recovery, or possibly an additional stressor.
peer / social support Can include peer support in the context of treatment counseling.
Broadened this code to also include other forms of social support in general that
young adults' mentioned in their interviews (does not necessarily need to be a
provider).
For provider support, code "relationships with providers".
peer influence Peer influences on substance use or other behaviors. This may be a more specific
"reason for substance use."
permanent supportive housing (PSH) /
Housing First
Any reference to Permanent Supportive Housing or the "Housing First" Model.
Emphasis is placed on providing people with long-term housing first, without making
it contingent on receive treatment.
policy/practice shortcomings Description of policies or practices related to homeless youth services that are
detrimental to youth or need improvement. Services can include housing or substance
use treatment services, etc.
positive consequences of substance use Any positive effects of substance use, whether planned or unplanned. May be in
response to the question "What are some good things about your substance use?"
previous treatment experience(s) Code any previous experiences (or not) in substance use treatment/counseling.
provider work experiences May be in response to the first question, "Tell me about your experience providing
services to YAEH."
provider's counseling skills A provider's skill level (good or bad). This may have an impact on young adults'
treatment outcomes. This is different from provider work experiences.
psychotropic medications

race, racism, racial disparities Code only when mentioning racial / ethnic disparities?
Possible overlap with "cultural competency" code and "discrimination" code.

137
readiness for treatment/recovery YAEH's readiness (of motivation) to initiate or follow through with treatment or
recovery, or a reduction in substance use. This may be a facilitator or barrier to
treatment engagement.
reasons for homelessness A more general code of the reasons leading up to young adults' first experience of
homelessness, continuation of homelessness, or return to homelessness.
Reasons may include family conflict, abuse, LGBTQ identity, desire for freedom,
etc.
reasons for substance use A more general code of the reasons why young adults' may use substances.
May be in response to the question, "What are some of the main reasons YAEH use
substances?"
recovery process Any description of the process of recovery from substance use and/or addiction. This
may also include "natural recovery" achieved without help from treatment providers.
It may also include reasons for quitting or reducing one's substance use.
recovery role models Discussion of (older?) adult role models who serve as models for recovery. May be
discussed as part of an ideal treatment for YAEH, etc. Consider merging with peer
support code?
referral to treatment Referrals to treatment to an external service provider or treatment facility. Referral to
treatment may be a process that is described.
relationships with providers Any mention of YAEH's supportive relationships with providers. Can also include
negative relationships or the absence of a relationship.
resilience

safety YAEH's perceived safety in different environments. Consider merging with
"environmental/neighborhood factors" code.
self-awareness This is an in-vivo code from provider participant 1, used to describe self-awareness
and self-determination as a successful treatment outcome.
This may be used to code other descriptions of YAEH's self-awareness or insights
about their behavior or housing situation.

138
self-medication/medicinal use Substance use as self-medication or coping with homelessness, traumatic experiences
or poor mental and physical health.
May be in response to the question, "What are the main reasons why YAEH may use
substances?"
May also refer to self-reported medicinal use or having a medical marijuana card.
shame / stigma

sleep

social isolation / need for community Any example of social isolation. Need for community may be coded here as well, as
these two may often be related.
specific drug prevalence The number or percentage of young adults experiencing homelessness that use
certain substances (or substance use in general). This is not the same thing as what
types of substances YAEH most commonly use. Only code if respondent mentions a
specific number or percentage.
spirituality

substance use as a barrier to housing The effect that substance use has on housing stability (either to attain or maintain
housing). May be in response to the question, "Does substance use contribute to
homelessness or unstable housing?"
systems involvement (e.g., foster care,
justice system)
Discussion of young adults' current or previous involvement in the foster care,
juvenile justice, or criminal justice system. Experience may be an additional risk
factor or a recovery factor.
trauma informed care

treatment facilitators

trust Includes anything that positively contributes to young adults' willingness or actual
engagement in treatment.
types and patterns of homelessness Discussion of various forms of homelessness (couch surfing, living on the streets,
etc.), or patterns of homelessness over time. Use "chronic homelesness" as a more
specific code, and also "reasons for homelessness" as a more specific code. Re-
evaluate if there is significant overlap.

139
types of housing resources available /
used
Any description of the types of housing resources avialable to YAEH (e.g., shelter,
rapid rehousing, etc.) For young adults, use of these services can be coded.
Permanent supportive housing (PSH) or discussion of "housing first" should be coded
with its own code.
types of substance use Common types of substances used (i.e., alcohol, marijuana, etc.), methods of use
(drinking, smoking, injecting, etc.), or typical patterns of use.
Use this when coding response to "What types of substances do YAEH use most
often?"
Use "recovery process" to code decrease in substance use over time (naturally, or a
result of treatment).
types of treatment/counseling services
available/received
Types of substance use treatment or counseling services offered or available to young
adults. May be in response to the question "What treatment or counseling practices
are provided to the young adults you work with?"
Can also include types of services that are NOT offered, by the respondent believes
should be offered. Can include other services that are not substance use treatment
specific (e.g., case management), if they have the potential to have an impact on
young adults' substance use behaviors.

140

Chapter 5: Conclusion
The overarching goal of this dissertation was to better understand complex issues related
to substance use and homelessness among young adults. Nationwide, it is estimated that one in
10 young adults experience homelessness each year (Morton et al., 2018). Such experiences
range from “couch surfing” at a friend or relative’s home, to more entrenched experiences of
unsheltered homelessness (e.g., sleeping on sidewalks, in cars, or abandoned buildings, etc.). A
large minority of young adults who experience homelessness also meet criteria for clinically
relevant substance use disorders (Baer et al., 2003; Whitbeck, 2009). Additionally, nearly half a
million young adults in the U.S. receive some form of substance use treatment each year
(SAMHSA, 2021). To date, however, research has yet to investigate how these two public health
problems—homelessness and substance use—intersect among young adults, with implications
for the substance use treatment system. Therefore, this dissertation was carried out with two
intersecting populations in mind: 1) young adults with substance use disorders who are currently
experiencing homelessness and not engaged in treatment, and 2) young adults currently receiving
substance use treatment and are experiencing (or at risk of experiencing) homelessness. In this
conclusion chapter, I will briefly review study findings and their implications. I will close by
making recommendations for policy and practice and outline a future research agenda to address
the dual problems of homelessness and substance use disorder among young adults.
Review of Study Findings
Study 1 used three different Machine Learning algorithms to predict (or “classify”) young
adults according to their housing status (homeless or stably housed) in the year prior to entering
substance use treatment. Using intake assessment data from over 40,000 young adults entering
treatment in the U.S. from 2002 to 2018, models identified significant risk and protective factors

141

for homelessness that were largely consistent with existing theories of youth homelessness.
Important risk and protective factors emerged across several domains, including mental health
(PTSD and depression symptoms), substance use (stimulant use, SUD symptoms), familial and
background variables (e.g., family history of substance use problems, victimization history), and
demographics (e.g., sexual/gender minority status). Currently working or being in school
emerged as an important protective factor strongly associated with being stably housed, and
Hispanic ethnicity was identified as an important protective factor in the logistic regression
models, but not the Random Forest classification model. Models were also evaluated based on
their accuracy in predicting homelessness on a testing data set for which young adults’ housing
status was known. Accuracy was comparable across models, which were able to correctly
classify about two-thirds of participants as homeless or housed.
Study 2 examined prospective, bidirectional relationships between days spent homeless
and frequency of substance use and/or problems among 3,717 young adults receiving treatment
up to 12-months post-treatment entry. Contrary to previous research with adolescents (Davis et
al., 2019), results provided little overall evidence for bidirectional (i.e., reciprocal, cross-lagged)
effects during and after treatment. However, upon examining these associations across
demographic subgroups, a significant between-group cross-lagged effect emerged for non-
Hispanic White young adults (but not young adults of color). Non-Hispanic White young adults
who entered treatment with relatively high substance use frequency (compared to the average
non-Hispanic White young adult) tended to have greater decreases in days spent homeless going
forward. Substance use and homelessness were also significantly associated among this group at
treatment intake, which could have partially explained this cross-lagged effect. Specifically,
given overall decreases in both homelessness and substance use frequency during treatment, it

142

could be that non-Hispanic White young adults who had more frequent substance use and/or
problems prior to treatment attained more stable housing as a secondary benefit of treatment. It
does not appear that housing stability differed by race/ethnicity per se, in Study 2, because
race/ethnicity was not associated with young adults’ initial homelessness or change over time
(regression results of demographic predictors on latent growth factors not shown). This is
partially consistent with Study 1, which did not suggest differences in the odds of experiencing
homelessness in the past-year when comparing Black / African American young adults to non-
Hispanic White young adults. However, a significantly lower risk of homelessness was observed
among Hispanic/Latinx young adults (compared with non-Hispanic White young adults).
Therefore, it is possible that trajectories of homelessness could have been different for
Hispanic/Latinx young adults in Study 2 (compared to non-Hispanic White young adults), but
these subgroup differences were not examined. Rather, Hispanic/Latinx young adults were
included in the “young adult of color” subgroup. Altogether, study 2 poses a nuanced set of
findings with implications that should be thoughtfully communicated. Overall, homelessness and
substance use may not be reciprocally related over time for young adults engaged in treatment,
because treatment-engaged young adults may have more internal and external resources to
prevent such negative cascades than adolescents in treatment or non-treatment engaged youth.
Study 3 used qualitative methods to investigate barriers and facilitators to substance use
treatment among young adults with substance use disorders currently experiencing homelessness
(n = 22) and service providers working with this population (n = 16) in Los Angeles, California.
Thematic analysis revealed nine themes, which could act as barriers or facilitators, depending on
the context. These themes reflect facilitators and barriers that exist across predisposing
(homelessness and instability, internalized shame and stigma), systemic (access to treatment,

143

systems referrals), organizational (rigidity vs. flexibility of treatment programs, safe spaces),
interpersonal (supportive referrals), and individual levels (readiness for treatment). These themes
and their relation to one another are depicted within an adapted Behavioral Model for Vulnerable
Populations in Figure 4.2. Excerpts from interviews with young adults and providers contributed
to all themes, but some differences in perspective between the two groups are noted.
Implications for Treatment-Engaged Young Adults
Findings from this dissertation have implications for clinical practice with young adults
receiving substance use treatment in the U.S. The first issue relates to identifying young adults in
treatment who are currently experiencing homelessness, or who are at-risk of experiencing
homelessness. Study 1 raises the possibility of using Machine Learning algorithms to predict an
individual’s risk of homelessness as they enter treatment as a function of other relevant data
collected during their intake assessment. Arguably, simply asking young adults about their
current housing status as they enter treatment is a more efficient way of doing this, and housing
support services could then be provided to that individual. The value of the algorithms developed
in Study 1 then, is more theoretical and methodological in nature. This work could also serve as
a starting point for developing algorithms to predict an individual’s future risk of experiencing
homelessness, which would prove to be highly useful in homelessness prevention efforts
(DiGuiseppi, Davis, Leightley, et al., 2020; Morton, 2020). Although we cannot say for certain,
it is likely that the same risk and protective factors associated with experiencing homelessness in
the year prior to treatment continue to play a role in young adults’ future risk of homelessness
during and after treatment.
This dissertation also highlights the importance of homelessness prevention as a standard
practice within substance use treatment settings, whether algorithms are used for this purpose or

144

not. Morton and colleagues (2018) have raised the need for more “upstream” interventions to
prevent homelessness in youth-serving institutions (e.g., schools, child welfare agencies). This
research suggests that substance use treatment facilities can be added to this list. When
considering young adults’ future risk for homelessness during and after treatment, results from
Study 2 should be kept in mind. Perhaps counterintuitively, results of Study 2 demonstrate that
between- or within-person increases in substance use during or after treatment may not
necessarily lead to a greater risk of homelessness at a later point in time. In fact, individuals with
higher-than-average substance use frequency prior to treatment may actually benefit from greater
reductions in days spent homeless during and after treatment (i.e., more housing stability). This,
along with the general decline in homelessness throughout treatment, suggests that substance use
treatment programs may be directly or indirectly addressing vulnerable young adults’ need for
stable housing during and after treatment. Results also suggest that more could be done, since an
average decrease of less-than-half a day spent homeless throughout the study period is quite
modest—after beginning with an overall average of three-and-a-half days spent homelessness in
the 90 days prior to treatment entry. In sum, more work in substance use treatment settings is
needed to provide access to comprehensive services that can help support young adults' recovery,
including housing support services. This may involve greater collaboration and referrals across
service systems, leveraging housing supports within young adults’ familial and social support
networks, and modifying existing housing models (i.e., recovery residences and permanent
supportive housing) to be more appropriate for young adults.
Implications for Young Adults Experiencing Homelessness Not Engaged in Treatment
Although available research suggests that young adults experiencing homelessness are
more likely to have lifetime experience in substance use treatment than the general population of

145

young adults (de Rosa et al., 1999; Rabinovitz et al., 2010; van Leeuwen et al., 2006; Whitbeck,
2009), research also indicates significant barriers to treatment among young adults currently
accessing homeless youth services. Findings from Study 3 have implications for engaging these
young adults in treatment, whether this involves engagement in the “traditional” treatment
system at specialty clinics, or in innovative service models outside of traditional treatment
settings. Interviews with young adults and providers suggest that some young adults may require
a higher level of care in inpatient/residential treatment settings, and for these young adults, a
drastic increase in available services is needed. The need for more services is likely even more
pronounced in areas outside of Los Angeles County, a particularly service-rich area where Study
3 took place. This will likely involve more funding for treatment and expanded access to
inpatient units and specially trained staff. It will also require better coordination between service
providers in the homeless service sector and substance use treatment settings, and greater
flexibility among substance use treatment providers to literally “meet clients where they’re at” in
drop-in centers, shelters, and street locations—and to be more flexible with young adults’ harm
reduction goals. Furthermore, continued support is needed throughout treatment, both in the form
of social support and supportive housing after residential treatment is completed. This last point
is an area where findings from Study 2 and Study 3 converged.
For young adults who do not require inpatient treatment, in-depth interviews from Study
3 suggest the possibility for a range of innovative options to address young adults’ substance use.
One evidence-based option is to implement screening, brief intervention, and referral to
treatment (SBIRT) protocols in popular service settings, such as drop-in centers and shelters.
This may address the issues of raising individual awareness and perceived need for treatment,
and may help facilitate external referrals or continued conversations about substance use with

146

dedicated staff. Some challenges with this approach, however, are that young adults experiencing
homelessness have a variety of competing needs in addition to substance use (domestic violence,
mental health problems, food insecurity, etc.), and SBIRT may not address all barriers to
treatment. Another challenge is the amount of staff training and resources required to implement
and sustain SBIRT protocols (Babor et al., 2017). Still, the sheer number of individuals identified
who may benefit from a brief motivational intervention and some follow-up sessions may be
worth the effort. A recent pilot study delivering group-based motivational interviewing to young
adults in drop-in centers showed significant changes in past 30-day alcohol use, motivation to
change drug use, and self-efficacy to use condoms compared to a control group (Tucker et al.,
2017). The benefits of SBIRT interventions for low-income youth in community settings may
also spillover to positive outcomes in other areas as well, such as greater engagement in
education and employment (Brolin et al., 2022).
Other “low-barrier” options may exist and may be developed and implemented within
drop-in center and shelter locations, including the use of peer-delivered interventions, and
providing more traditional outpatient counseling to groups and individuals by a specialist on site.
Young adults and providers stressed the importance of delivering appropriate care within “safe
spaces,” underscoring the value of delivering interventions in settings where youth feel
comfortable and may already have established relationships. When externally referring young
adults to inpatient or outpatient treatment, it is important to reduce as many barriers as possible
by ensuring that lack of transportation, high perceived cost, and discomfort within clinical
settings don’t get in the way. It is also important to enhance facilitators such as social support,
and internal readiness for treatment. This may involve being clearer about the tangible short and
long-term benefits of treatment, and by providing more incentives to engage in treatment.

147

Regarding this last point, contingency management interventions—which provide monetary and
other types of rewards for negative drug tests and other positive behaviors during treatment
(Babor et al., 2017), were raised by one provider as a promising treatment approach. However,
contingency management interventions have yet to be evaluated with young adults experiencing
homelessness.
Charting a Future Research Agenda
Although this dissertation sought to address several previously unanswered research
questions, findings also raise more questions for future research. First, from an epidemiological
perspective, there is a need for updated prevalence estimates of substance use disorders among
young adults experiencing homelessness. Although there have been recent studies documenting
the prevalence of various types of substance use among young adults experiencing homelessness
in several U.S. cities (Chassman et al., 2022), estimates of clinically significant substance use
disorders are outdated (Baer et al., 2003; Johnson et al., 2005; Kipke et al., 1997). Many of these
estimates use older diagnostic criteria for substance use disorders, and would benefit from using
the most up-to-date conceptualization of substance use disorder using the fifth edition of the
Diagnostic and Statistical Manual of Mental Disorders (Hasin et al., 2013). This would help
provide a better sense of the common types of problems substance use may cause for young
adults experiencing homelessness, which could inform intervention development.
Secondly, more research is needed on innovative models of service delivery for young
adults experiencing homelessness who are not engaged in treatment. As described throughout
this dissertation, this should ideally involve “low-barrier” approaches that utilize harm reduction,
peer support, and trauma-informed care. While the research base on interventions developed
specifically for this population continues to grow (Coren et al., 2016; Slesnick et al., 2021;

148

Tucker et al., 2020; Zhang & Slesnick, 2018), there continues to be a need to assess how these
programs can be implemented and sustained more broadly, and in ways that are effective and
acceptable for young adults and service providers. Furthermore, more research is needed on the
degree to which engagement in other types of services, such as mental health treatment, may
improve substance use disorder symptoms. This raises the important question of the degree to
which intervention content should be specifically focused on substance use. For example, are
specialized substance use interventions really needed if young adults are able to resolve their
substance use problems by engaging in interventions with a broader focus on mental health and
wellness? It is an open question whether focusing exclusively on substance use related content
may dissuade some young adults who are more interested in making improvements in other life
domains. An exclusive focus on substance use may also unintentionally add further stigma and
discourage treatment engagement. The prevalence of substance use among this population
suggests that some specific substance use content is needed. However, it seems likely that
acceptable interventions need to be holistic and address young adults’ diverse needs, perhaps by
incorporating “non-traditional” therapeutic activities such as art, music, and physical exercise,
which are favored by youth (Nyamathi et al., 2007, 2011). These issues will be important for
future intervention development research.
Results from the current dissertation and recent research by Green and colleagues (2020)
will be helpful for providing estimates on the prevalence of homelessness among young adults in
treatment. However, more research is needed to understand the diverse housing situations and
housing services available to this group during and after treatment. In all likelihood, connections
to supportive housing services will vary across clinical sites and geographic locations. For
instance, young adults receiving treatment may be at greater risk of experiencing homelessness

149

in metropolitan areas with higher housing costs and lower availability of affordable housing.
There may also be fewer supportive housing services at treatment sites in rural areas. This
highlights the importance of policies and practices intended to foster cross-collaboration between
behavioral health and housing support services, and the provision of greater funding for both
treatment and supportive housing at federal, state, and local levels. Future policy efforts may
benefit from broad survey research on the types of supportive housing services available through
substance use treatment settings and after treatment. Some important work has been carried out
on recovery residences that provide housing for adults after treatment (Mericle, Patterson, et al.,
2022), but more developmentally appropriate options may be needed for young adults. In
addition, more research on the conditions necessary for recovery and “recovery capital” among
young adults is needed (Cano et al., 2017; Vilsaint et al., 2017). In addition to stable housing,
this likely includes a broad array of social support, education and employment support, and
continued access to physical and mental health care.
One final issue ripe for future research relates to methods. Specifically, longitudinal
researchers seeking to understand the role of housing and homelessness during after treatment
should make use of more frequent and valid assessments to capture important relationships
between homelessness, substance use, and other indicators of recovery over time. While some
interesting work using retrospective qualitative interviews has shed light on youths’ diverse
“pathways” in and out of homelessness (Mallett et al., 2005; Tyler & Schmitz, 2013),
quantitative methods should make greater of assessment methods such as ecological momentary
assessment, which may help better capture these relationships in real time. Some recent research
has shown that EMA is feasible to use with young adults experiencing homelessness, and can

150

raise awareness of specific situational and contextual risks that need can be intervened on
(Henwood et al., 2019; Suchting et al., 2020; Tyler et al., 2018; Tyler & Olson, 2018).
Conclusion
This dissertation adds to the growing body of evidence on the prevalence and correlates
of homelessness, and relationships between homelessness and substance use outcomes among
young adults receiving substance use treatment. It also adds to the body of evidence on the
barriers and facilitators to treatment shared by young adults currently experience homeless and
their service providers. Results indicate that young adults entering substance use treatment are
more likely to have some experience with homelessness than young adults in the general
population, but their likelihood of experiencing homelessness may decrease slightly during and
after treatment. Sustained attention is needed to implement policies and practices that address
vulnerable young adults’ need for housing during and after treatment. Innovative strategies and
solutions are also needed to engage young adults who are currently experiencing homelessness in
programs and services that can reduce substance use harms, and support them on their path to
leading healthy, stable, and productive lives as young adults.

151

References
Administration, S. A. and M. H. S. (2021). Key substance use and mental health indicators in the
United States: Results from the 2020 National Survey on Drug Use and Health (HHS
Publication No. PEP21-07-01-003, NSDUH Series H-56). https://www.samhsa.gov/data
Almquist, L., & Walker, S. C. (2022). Reciprocal associations between housing instability and
youth criminal legal involvement: a scoping review. Health and Justice, 10(1), 1–14.
https://doi.org/10.1186/S40352-022-00177-7/TABLES/3
Arnett, J. J. (2000). Emerging adulthood. A theory of development from the late teens through
the twenties. The American Psychologist, 55(5), 469–480.
http://www.ncbi.nlm.nih.gov/pubmed/10842426
Ashford, R. D., Brown, A. M., & Curtis, B. (2018). Systemic barriers in substance use disorder
treatment: A prospective qualitative study of professionals in the field. Drug and Alcohol
Dependence, 189, 62–69. https://doi.org/10.1016/J.DRUGALCDEP.2018.04.033
ATLAS.ti Scientific Software Development GmbH [ATLAS.ti 22 Windows]. (2022). ATLAS.ti
Scientific Software Development GmbH [ATLAS.ti 22 Windows]. https://atlasti.com
Babor, T. F., del Boca, F., & Bray, J. W. (2017). Screening, Brief Intervention and Referral to
Treatment: implications of SAMHSA’s SBIRT initiative for substance abuse policy and
practice. Addiction (Abingdon, England), 112 Suppl 2, 110–117.
https://doi.org/10.1111/ADD.13675
Baer, J. S., Ginzler, J. A., & Peterson, P. L. (2003). DSM-IV alcohol and substance abuse and
dependence in homeless youth. Journal of Studies on Alcohol, 64(1), 5–14.
Barenholtz, E., Fitzgerald, N. D., & Hahn, W. E. (2020). Machine-learning approaches to
substance-abuse research: emerging trends and their implications. Current Opinion in
Psychiatry, 33(4), 334–342. https://doi.org/10.1097/YCO.0000000000000611
Barker, S. L., & Maguire, N. (2017). Experts by Experience: Peer Support and its Use with the
Homeless. Community Mental Health Journal, 53(5), 598–612.
https://doi.org/10.1007/s10597-017-0102-2
Barman-Adhikari, A., Craddock, J., Bowen, E., Das, R., & Rice, E. (2018). The Relative
Influence of Injunctive and Descriptive Social Norms on Methamphetamine, Heroin, and
Injection Drug Use Among Homeless Youths: The Impact of Different Referent Groups.
Journal of Drug Issues, 48(1), 17–35.
https://doi.org/http://dx.doi.org/10.1177/0022042617726080
Bassuk, E. L., Hanson, J., Greene, R. N., Richard, M., & Laudet, A. (2016). Peer-Delivered
Recovery Support Services for Addictions in the United States: A Systematic Review.
Journal of Substance Abuse Treatment, 63, 1–9.
https://doi.org/10.1016/J.JSAT.2016.01.003

152

Bender, K., Brown, S. M., Thompson, S. J., Ferguson, K. M., & Langenderfer, L. (2015).
Multiple Victimizations Before and After Leaving Home Associated With PTSD,
Depression, and Substance Use Disorder Among Homeless Youth. Child Maltreatment,
20(2), 115–124. https://doi.org/10.1177/1077559514562859
Black, J. J., & Chung, T. (2014). Mechanisms of change in adolescent substance use treatment:
How does treatment work? In Substance Abuse (Vol. 35, Issue 4, pp. 344–351). Routledge.
https://doi.org/10.1080/08897077.2014.925029
Bozinoff, N., Small, W., Long, C., DeBeck, K., & Fast, D. (2017). Still “at risk”: An
examination of how street-involved young people understand, experience, and engage with
“harm reduction” in Vancouver’s inner city. The International Journal on Drug Policy, 45,
33–39. https://doi.org/10.1016/J.DRUGPO.2017.05.006
Braciszewski, J. M., Toro, P. A., & Stout, R. L. (2016). Understanding the Attainment of Stable
Housing: A Seven-Year Longitudinal Analysis of Homeless Adolescents. Journal of
Community Psychology, 44(3), 358–366. https://doi.org/10.1002/JCOP.21773
Brakenhoff, B., Jang, B., Slesnick, N., & Snyder, A. (2015). Longitudinal predictors of
homelessness: findings from the National Longitudinal Survey of Youth-97. Journal of
Youth Studies, 18(8), 1015–1034. https://doi.org/10.1080/13676261.2015.1020930
Braun, V., & Clarke, V. (2022). Thematic Analysis: A Practical Guide. SAGE Publications Ltd.
Breiman, L. (2001). Random Forests. Machine Learning 2001 45:1, 45(1), 5–32.
https://doi.org/10.1023/A:1010933404324
Brolin, M., Reif, S., Buell, J., Whitcher, H., Jaghoo, S., & McNeil, P. (2022). Screening and
Brief Intervention With Low-Income Youth in Community-Based Settings. The Journal of
Adolescent Health : Official Publication of the Society for Adolescent Medicine, 71(4S),
S65–S72. https://doi.org/10.1016/J.JADOHEALTH.2022.06.001
Bronfenbrenner, U. (1979). The ecology of human development: Experiments by nature and
design. Harvard University Press.
Byrne, T., Baggett, T., Land, T., Bernson, D., Hood, M. E., Kennedy-Perez, C., Monterrey, R.,
Smelson, D., Dones, M., & Bharel, M. (2020). A classification model of homelessness
using integrated administrative data: Implications for targeting interventions to improve the
housing status, health and well-being of a highly vulnerable population. PloS One, 15(8).
https://doi.org/10.1371/JOURNAL.PONE.0237905
Cano, I., Best, D., Edwards, M., & Lehman, J. (2017). Recovery capital pathways: Modelling the
components of recovery wellbeing. Drug and Alcohol Dependence, 181, 11–19.
https://doi.org/10.1016/j.drugalcdep.2017.09.002

153

Chassman, S., Barman-Adhikari, A., Hsu, H. T., Ferguson, K. M., Narendorf, S. C., Maria, D. S.,
Shelton, J., Petering, R., & Bender, K. (2022). Prevalence and Correlates of Illicit Substance
Use Among Young Adults Experiencing Homelessness in Seven Cities Across the United
States: Https://Doi.Org/10.1177/00220426211049355, 2022(4), 488–508.
https://doi.org/10.1177/00220426211049355
Coren, E., Hossain, R., Pardo, J. P., & Bakker, B. (2016). Interventions for promoting
reintegration and reducing harmful behaviour and lifestyles in street-connected children and
young people. Cochrane Database of Systematic Reviews, 1, 1–153.
https://doi.org/10.1002/ebch.1923
Curran, P. J., Howard, A. L., Bainter, S. A., Lane, S. T., & McGinley, J. S. (2014). The
separation of between-person and within-person components of individual change over
time: a latent curve model with structured residuals. Journal of Consulting and Clinical
Psychology, 82(5), 879–894. https://doi.org/10.1037/a0035297
Dang, M. T., Conger, K. J., Breslau, J., & Miller, E. (2014). Exploring Protective Factors among
Homeless Youth: The Role of Natural Mentors. Journal of Health Care for the Poor &
Underserved, 25(3), 1121–1138. https://doi.org/10.1353/hpu.2014.0133
Davis, J. P., Christie, N. C., Dworkin, E. R., Prindle, J., Dumas, T. M., DiGuiseppi, G., Helton, J.
J., & Ring, C. (2020). Influences of victimization and comorbid conditions on latency to
illicit drug use among adolescents and young adults. Drug and Alcohol Dependence, 206.
https://doi.org/10.1016/j.drugalcdep.2019.107721
Davis, J. P., Diguiseppi, G., de Leon, J., Prindle, J., Sedano, A., Rivera, D., Henwood, B., &
Rice, E. (2019). Understanding pathways between PTSD, homelessness, and substance use
among adolescents. Psychology of Addictive Behaviors : Journal of the Society of
Psychologists in Addictive Behaviors, 33(5). https://doi.org/10.1037/adb0000488
Davis, J. P., Eddie, D., Prindle, J., Dworkin, E. R., Christie, N. C., Saba, S., DiGuiseppi, G.,
Clapp, J. D., & Kelly, J. F. (2021). Sex Differences in Factors Predicting Post‐treatment
Opioid Use. Addiction, add.15396. https://doi.org/10.1111/add.15396
Davis, J. P., Rao, P., Dilkina, B., Prindle, J., Eddie, D., Christie, N. C., DiGuiseppi, G., Saba, S.,
Ring, C., & Dennis, M. (2022). Identifying individual and environmental predictors of
opioid and psychostimulant use among adolescents and young adults following outpatient
treatment. Drug and Alcohol Dependence, 233, 109359.
https://doi.org/10.1016/J.DRUGALCDEP.2022.109359
de Rosa, C. J., Montgomery, S. B., Kipke, M. D., Iverson, E., Ma, J. L., & Unger, J. B. (1999).
Service utilization among homeless and runaway youth in Los Angeles, California: rates
and reasons. Journal of Adolescent Health, 24(6), 449–458. https://doi.org/10.1016/S1054-
139X(99)00040-3
de Sousa, T., Andrichik, A., Cuellar, M., Marson, J., Prestera, E., Rush, K., & Abt Associates.
(2022). The 2022 Annual Homelessness Assessment Report (AHAR) to Congress.

154

Dennis, M. L., Chan, Y. F., & Funk, R. R. (2006). Development and validation of the GAIN
Short Screener (GSS) for internalizing, externalizing and substance use disorders and
crime/violence problems among adolescents and adults. The American Journal on
Addictions, 15 Suppl 1(Suppl 1), s80–s91. https://doi.org/10.1080/10550490601006055
Dennis, M. L., White, M., Titus, J. C., & Unsicker, J. (2008). Global Appraisal of Individual
Needs: Administration Guide for the GAIN and Related Measures.
https://chestnut.org/li/gain
Derbyshire, P., Muir-Cochrane, E., Fereday, J., Jureidini, J., & Drummond, A. (2006).
Engagement with health and social care services: Perceptions of homeless young people
with mental health problems. In Health and Social Care in the Community (Vol. 14, Issue 6,
pp. 553–562). https://doi.org/10.1111/j.1365-2524.2006.00643.x
DiGuiseppi, G. T., Davis, J. P., Christie, N. C., & Rice, E. (2020). Polysubstance use among
youth experiencing homelessness: The role of trauma, mental health, and social network
composition. Drug and Alcohol Dependence, 216.
https://doi.org/10.1016/j.drugalcdep.2020.108228
DiGuiseppi, G. T., Davis, J. P., Leightley, D., & Rice, E. (2020). Predictors of Adolescents’ First
Episode of Homelessness Following Substance Use Treatment. Journal of Adolescent
Health, 66(4). https://doi.org/10.1016/j.jadohealth.2019.11.312
DiGuiseppi, G. T., Ring, C. R., Rice, E. R., & Davis, J. P. (2022). Sex differences in poly-
victimization among youth experiencing homelessness prior to substance use treatment.
Child Abuse & Neglect, 129, 105670. https://doi.org/10.1016/J.CHIABU.2022.105670
DiGuiseppi, G. T., Tucker, J. S., Prindle, J. J., Henwood, B. F., Huey, S. J., Rice, E. R., & Davis,
J. P. (2021). Comparing the effectiveness of three substance use interventions for youth
with and without homelessness experiences prior to treatment. Journal of Consulting and
Clinical Psychology, 89(12), 995–1006. https://doi.org/10.1037/CCP0000704
Dworsky, A., Napolitano, L., & Courtney, M. (2013). Homelessness During the Transition From
Foster Care to Adulthood. American Journal of Public Health, 103(S2), S318–S323.
https://doi.org/10.2105/AJPH.2013.301455
Edidin, J. P., Ganim, Z., Hunter, S. J., & Karnik, N. S. (2012). The Mental and Physical Health
of Homeless Youth: A Literature Review. Child Psychiatry & Human Development, 43(3),
354–375. https://doi.org/10.1007/s10578-011-0270-1
Embleton, L., Lee, H., Gunn, J., Ayuku, D., & Braitstein, P. (2016). Causes of Child and Youth
Homelessness in Developed and Developing Countries. JAMA Pediatrics, 170(5), 435.
https://doi.org/10.1001/jamapediatrics.2016.0156
Enders, C. K. (2022). Applied Missing Data Analysis (2nd ed.). The Guilford Press.
Ensign, J. (2004). Quality of Health Care: The Views of Homeless Youth. Health Services
Research, 39(4p1), 695–708. https://doi.org/10.1111/j.1475-6773.2004.00253.x

155

Eric Rice, Monique Holguin, Hsun-Ta Hsu, Matthew Morton, Phebe Vayanos, Milind Tambe, &
Hau Chan. (2018). Linking Homelessness Vulnerability Assessments to Housing
Placements and Outcomes for Youth. Cityscape, 20(3), 69–86.
https://www.jstor.org/stable/26524872#metadata_info_tab_contents
Ferguson, K. M., & Xie, B. (2008). Feasibility Study of the Social Enterprise Intervention With
Homeless Youth. Research on Social Work Practice, 18(1), 5–19.
https://doi.org/10.1177/1049731507303535
Ferrer, E., Conger, R. D., & Robins, R. W. (2016). Longitudinal Dynamics of Substance Use and
Psychiatric Symptoms in Count Data with Zero Inflation. Multivariate Behavioral
Research, 51(2–3), 279–295.
https://doi.org/10.1080/00273171.2016.1144501/SUPPL_FILE/HMBR_A_1144501_SM74
09.DOCX
Gelberg, L., Andersen, R. M., & Leake, B. D. (2000). The Behavioral Model for Vulnerable
Populations: application to medical care use and outcomes for homeless people. Health
Services Research, 34(6), 1273–1302. http://www.ncbi.nlm.nih.gov/pubmed/10654830
Green, B., Lee Kim, F., & Dean, D. (2020). Substance Use and Treatment Among Youth
Experiencing Homelessness: 1992–2017. Journal of Adolescent Health, 67(6), 786–792.
https://doi.org/10.1016/j.jadohealth.2020.04.019
Greene, J. M., Ennett, S. T., & Ringwalt, C. L. (1997). Substance use among runaway and
homeless youth in three national samples. American Journal of Public Health, 87(2), 229–
235.
Haber, M. G., & Toro, P. A. (2004). Homelessness among families, children, and adolescents: an
ecological-developmental perspective. Clinical Child and Family Psychology Review, 7(3),
123–164. http://www.ncbi.nlm.nih.gov/pubmed/15645705
Hasin, D. S., O’Brien, C. P., Auriacombe, M., Borges, G., Bucholz, K., Budney, A., Compton,
W. M., Crowley, T., Ling, W., Petry, N. M., Schuckit, M., & Grant, B. F. (2013). DSM-5
Criteria for Substance Use Disorders: Recommendations and Rationale. The American
Journal of Psychiatry, 170(8), 834. https://doi.org/10.1176/APPI.AJP.2013.12060782
Heerde, J. A., Bailey, J. A., Toumbourou, J. W., Rowland, B., & Catalano, R. F. (2020).
Longitudinal Associations Between Early-Mid Adolescent Risk and Protective Factors and
Young Adult Homelessness in Australia and the United States. Prevention Science, 1–11.
https://doi.org/10.1007/s11121-020-01092-9
Heerde, J. A., & Hemphill, S. A. (2016). Sexual Risk Behaviors, Sexual Offenses, and Sexual
Victimization Among Homeless Youth: A Systematic Review of Associations With
Substance Use. Trauma, Violence, and Abuse, 17(5), 468–489.
https://doi.org/10.1177/1524838015584371

156

Henry, M., Watt, R., Mahathey, A., Ouellette, J., Sitler, A., & Associates, A. (2020). The 2019
Annual Homeless Assessment Report (AHAR) to Congress.
https://files.hudexchange.info/resources/documents/2019-AHAR-Part-1.pdf
Henwood, B. F., Redline, B., Dzubur, E., Madden, D. R., Rhoades, H., Dunton, G. F., Rice, E.,
Semborski, S., Tang, Q., & Intille, S. S. (2019). Investigating health risk environments in
housing programs for young adults: Protocol for a geographically explicit ecological
momentary assessment study. JMIR Research Protocols, 8(1).
https://doi.org/10.2196/12112
Hill, C., Hsu, H., Holguin, M., Morton, M., Winetrobe, H., & Rice, E. (2022). An examination of
housing interventions among youth experiencing homelessness: an investigation into
racial/ethnic and sexual minority status. Journal of Public Health, 44(4), 834–843.
https://doi.org/10.1093/PUBMED/FDAB295
Hodgson, K. J., Shelton, K. H., van den Bree, M. B. M., & Los, F. J. (2013). Psychopathology in
Young People Experiencing Homelessness: A Systematic Review. American Journal of
Public Health, 103(6), e24–e37. https://doi.org/10.2105/AJPH.2013.301318
Hooker, S. A., Sherman, M. D., Lonergan-Cullum, M., Nissly, T., & Levy, R. (2022). What is
success in treatment for opioid use disorder? Perspectives of physicians and patients in
primary care settings. Journal of Substance Abuse Treatment, 0(0), 108804.
https://doi.org/10.1016/J.JSAT.2022.108804
Hosmer, D. W., & Lemeshow, S. (2000). Applied Logistic Regression (2nd ed.). John Wiley &
Sons, Inc.
Hsu, H. T., Hill, C., Holguin, M., Petry, L., McElfresh, D., Vayanos, P., Morton, M., & Rice, E.
(2021). Correlates of Housing Sustainability Among Youth Placed Into Permanent
Supportive Housing and Rapid Re-Housing: A Survival Analysis. The Journal of
Adolescent Health : Official Publication of the Society for Adolescent Medicine, 69(4), 629–
635. https://doi.org/10.1016/J.JADOHEALTH.2021.03.022
Hudson, A. L., Nyamathi, A., Greengold, B., Slagle, A., Koniak-Griffin, D., Khalilifard, F., &
Getzoff, D. (2010). Health-seeking challenges among homeless youth. Nursing Research,
59(3), 212–218. https://doi.org/10.1097/NNR.0B013E3181D1A8A9
Hudson, A. L., Nyamathi, A., Slagle, A., Greengold, B., Griffin, D. K., Khalilifard, F., Gedzoff,
D., & Reid, C. (2009). The Power of the Drug, Nature of Support, and Their Impact on
Homeless Youth. Journal of Addictive Diseases, 28(4), 356–365.
https://doi.org/10.1080/10550880903183026
Hudson, A. L., Nyamathi, A., & Sweat, J. (2008). Homeless Youths’ Interpersonal Perspectives
of Health Care Providers. Issues in Mental Health Nursing, 29(12), 1277–1289.
https://doi.org/10.1080/01612840802498235
Ives, M., Funk, R., Ihnes, P., Feeney, T., & Dennis, M. (2010). Global Appraisal of Individual
Needs (GAIN) Evaluation Manual.

157

James, G., Witten, D., Hastie, T., & Tibshirani, R. (2021). An Introduction to Statistical
Learning with Applications in R (2nd ed.). Springer.
Jerome Friedman, Trevor Hastie, & Rob Tibshirani. (2010). Regularization Paths for
Generalized Linear Models via Coordinate Descent - PubMed. Journal of Statistical
Software, 33(1), 1–22. https://doi.org/10.18637/jss.v033.i01.
Jessica A. Heerde, Jennifer A. Bailey, & etc. (2021). Life-course predictors of homelessness
from adolescence into adulthood: A population-based cohort study. Journal of Adolescence,
91. https://www-clinicalkey-com.libproxy2.usc.edu/#!/content/journal/1-s2.0-
S0140197121000865
Johnson, K. D., Whitbeck, L. B., & Hoyt, D. R. (2005). Substance Abuse Disorders Among
Homeless and Runaway Adolescents. Journal of Drug Issues, 35(4), 799–816.
http://www.ncbi.nlm.nih.gov/pubmed/21533015
Karriker-Jaffe, K. J., Witbrodt, J., Mericle, A. A., Polcin, D. L., & Kaskutas, L. A. (2020).
Testing a Socioecological Model of Relapse and Recovery from Alcohol Problems.
Substance Abuse : Research and Treatment, 14. https://doi.org/10.1177/1178221820933631
Kassambara, A. (2017). Machine Learning Essentials: Practical Guide in R. Statistical tools for
high-throughput data analysis (STHDA).
Kidd, S. A., Vitopoulos, N., Frederick, T., Daley, M., Peters, K., Clarc, K., Cohen, S., Gutierrez,
R., Leon, S., & McKenzie, K. (2019). Peer Support in the Homeless Youth Context:
Requirements, Design, and Outcomes. Child and Adolescent Social Work Journal, 36(6),
641–654. https://doi.org/10.1007/s10560-019-00610-1
Kipke, M. D., Montgomery, S. B., Simon, T. R., & Iverson, E. F. (1997). "Substance
abuse" disorders among runaway and homeless youth. Substance Use & Misuse,
32(7–8), 969–986. http://www.ncbi.nlm.nih.gov/pubmed/9220564
Klein, J. D., Woods, A. H., Wilson, K. M., Prospero, M., Greene, J., & Ringwalt, C. (2000).
Homeless and runaway youths’ access to health care. The Journal of Adolescent Health :
Official Publication of the Society for Adolescent Medicine, 27(5), 331–339.
https://doi.org/10.1016/S1054-139X(00)00146-4
Kuhn, M. (2008). Building Predictive Models in R Using the caret Package. Journal of
Statistical Software, 28(5), 1–26. https://doi.org/10.18637/JSS.V028.I05
Kulik, D. M., Gaetz, S., Crowe, C., & Ford-Jones, E. (2011). Homeless youth’s overwhelming
health burden: A review of the literature. Paediatrics & Child Health, 16(6), e43–e47.
https://doi.org/10.1093/PCH/16.6.E43
Kuusisto, K., & Lintonen, T. (2020). Factors predicting satisfaction in outpatient substance abuse
treatment: a prospective follow-up study. Substance Abuse Treatment, Prevention, and
Policy, 15(1). https://doi.org/10.1186/S13011-020-00275-5

158

Lai, K. (2018). Estimating Standardized SEM Parameters Given Nonnormal Data and Incorrect
Model: Methods and Comparison. Https://Doi.Org/10.1080/10705511.2017.1392248,
25(4), 600–620. https://doi.org/10.1080/10705511.2017.1392248
Larimer, M. E., Palmer, R. S., & Marlatt, A. (1999). Relapse Prevention. An Overview of
Marlatt’s Cognitive-Behavioral Model. Alcohol Research & Health, 23(2), 151–160.
https://pubmed.ncbi.nlm.nih.gov/10890810/
Lauterbach, D., Vrana, S., King, D. W., & King, L. A. (1997). Psychometric properties of the
civilian version of the Mississippi PTSD Scale. Journal of Traumatic Stress, 10(3), 499–
513. https://doi.org/10.1023/A:1024801607043
Liaw, A., & Wiener, M. (2002). Classification and Regression by randomForest. R News, 2(3),
18–22. https://CRAN.R-project.org/doc/Rnews/
Littlefield, A. K., King, K. M., Acuff, S. F., Foster, K. T., Murphy, J. G., & Witkiewitz, K.
(2021). Limitations of Cross-Lagged Panel Models in Addiction Research and Alternative
Models: An Empirical Example Using Project MATCH. Psychology of Addictive
Behaviors, 36(3), 271–283. https://doi.org/10.1037/adb0000750
Liu, H., & Powers, D. A. (2007). Growth Curve Models for Zero-Inflated Count Data: An
Application to Smoking Behavior. Http://Dx.Doi.Org/10.1080/10705510709336746, 14(2),
247–279. https://doi.org/10.1080/10705510709336746
Maggs, J. L., Calhoun, B. H., & Allen, H. K. (2022). Substance use across adolescence and early
adulthood: Prevalence, causes, developmental roots, and consequences. In L. J. Crockett, G.
Carlo, & J. E. Schulenberg (Eds.), APA handbook of adolescent and young adult
development. (pp. 541–556). American Psychological Association.
https://doi.org/10.1037/0000298-033
Mallett, S., Rosenthal, D., & Keys, D. (2005). Young people, drug use and family conflict:
Pathways into homelessness. Journal of Adolescence, 28(2), 185–199.
https://doi.org/10.1016/j.adolescence.2005.02.002
Mallett, S., Rosenthal, D., Keys, D., & Averill, R. (2010). Moving Out, Moving On: Young
People’s Pathways In and Through Homelessness. Routledge.
McNeely, J., Wu, L. T., Subramaniam, G., Sharma, G., Cathers, L. A., Svikis, D., Sleiter, L.,
Russell, L., Nordeck, C., Sharma, A., O-Grady, K. E., Bouk, L. B., Cushing, C., King, J.,
Wahle, A., & Schwartz, R. P. (2016). Performance of the Tobacco, Alcohol, Prescription
Medication, and Other Substance Use (TAPS) Tool for Substance Use Screening in Primary
Care Patients. Annals of Internal Medicine, 165(10), 690–699. https://doi.org/10.7326/M16-
0317
Melander, L. A., Tyler, K. A., & Schmitz, R. M. (2016). An Inside Look at Homeless Youths’
Social Networks: Perceptions of Substance Use Norms. Journal of Child & Adolescent
Substance Abuse, 25(1), 78–88. https://doi.org/10.1080/1067828X.2014.918003

159

Menze, B. H., Kelm, B. M., Masuch, R., Himmelreich, U., Bachert, P., Petrich, W., &
Hamprecht, F. A. (2009). A comparison of random forest and its Gini importance with
standard chemometric methods for the feature selection and classification of spectral data.
BMC Bioinformatics, 10(1), 1–16. https://doi.org/10.1186/1471-2105-10-213/TABLES/4
Mericle, A. A., Patterson, D., Howell, J., Subbaraman, M. S., Faxio, A., & Karriker-Jaffe, K. J.
(2022). Identifying the availability of recovery housing in the U.S.: The NSTARR project.
Drug and Alcohol Dependence, 230.
https://doi.org/10.1016/J.DRUGALCDEP.2021.109188
Mericle, A. A., Slaymaker, V., Gliske, K., Ngo, Q., & Subbaraman, M. S. (2022). The role of
recovery housing during outpatient substance use treatment. Journal of Substance Abuse
Treatment, 133. https://doi.org/10.1016/J.JSAT.2021.108638
Milburn, N. G., Rice, E., Rotheram-Borus, M. J., Mallett, S., Rosenthal, D., Batterham, P., May,
S. J., Witkin, A., & Duan, N. (2009). Adolescents Exiting Homelessness Over Two Years:
The Risk Amplification and Abatement Model. Journal of Research on Adolescence, 19(4),
762–785. https://doi.org/10.1111/j.1532-7795.2009.00610.x
Modisette, K. C., & Dennis, M. L. (2022). Norms, characteristics, and psychometrics using the
GAIN 2018 Data Set.
Morton, M. H. (2020). The Complex Predictors of Youth Homelessness. Journal of Adolescent
Health, 66(4), 381–382. https://doi.org/10.1016/J.JADOHEALTH.2020.01.003
Morton, M. H., Dworsky, A., Matjasko, J. L., Curry, S. R., Schlueter, D., Chávez, R., & Farrell,
A. F. (2018). Prevalence and Correlates of Youth Homelessness in the United States.
Journal of Adolescent Health, 62(1), 14–21.
https://doi.org/10.1016/j.jadohealth.2017.10.006
Muthén, L. K., & Muthén, B. O. (n.d.). Mplus User’s Guide (8th ed.). Muthén & Muthén.
Nyamathi, A., Hudson, A., Mutere, M., Christiani, A., Sweat, J., Nyamathi, K., & Broms, T.
(2007). Drug use and barriers to and facilitators of drug treatment for homeless youth.
Patient Preference and Adherence, 1, 1–8.
Nyamathi, A., Slagle, A., Thomas, A., Hudson, A., Kahilifard, F., Avila, G., Orser, J., &
Cuchilla, M. (2011). Art messaging to engage homeless young adults. Progress in
Community Health Partnerships : Research, Education, and Action, 5(1), 9–18.
https://doi.org/10.1353/CPR.2011.0012
Padgett, D. K. (2012). Qualitative and Mixed Methods in Public Health. SAGE Publications, Inc.
Padilla, J., Jager, J., Updegraff, K. A., McHale, S. M., & Umaña-Taylor, A. J. (2020). Mexican-
origin family members’ unique and shared family perspectives of familism values and their
links with parent-youth relationship quality. Developmental Psychology, 56(5).
https://doi.org/10.1037/DEV0000913

160

Parast, L., Tucker, J. S., Pedersen, E. R., & Klein, D. (2019). Utilization and Perceptions of
Drop-in Center Services Among Youth Experiencing Homelessness. The Journal of
Behavioral Health Services & Research, 46(2), 234–248. https://doi.org/10.1007/s11414-
018-9632-9
Patterson, J. G., MacIsco, J. M., Glasser, A. M., Wermert, A., & Nemeth, J. M. (2022).
Psychosocial factors influencing smoking relapse among youth experiencing homelessness:
A qualitative study. PloS One, 17(7). https://doi.org/10.1371/JOURNAL.PONE.0270665
Paul, B., Thulien, M., Knight, R., Milloy, M. J., Howard, B., Nelson, S., & Fast, D. (2020).
“Something that actually works”: Cannabis use among young people in the context of street
entrenchment. PloS One, 15(7). https://doi.org/10.1371/JOURNAL.PONE.0236243
Pedersen, E. R., Tucker, J. S., Klein, D. J., & Parast, L. (2018). Perceived Need and Receipt of
Behavioral Health Services at Drop-In Centers among Homeless Youth. Health Services
Research, 53(6), 4609–4628. https://doi.org/10.1111/1475-6773.12990
Pedersen, E. R., Tucker, J. S., & Kovalchik, S. A. (2016). Facilitators and Barriers of Drop-In
Center Use Among Homeless Youth. Journal of Adolescent Health, 59(2), 144–153.
https://doi.org/10.1016/J.JADOHEALTH.2016.03.035
Perker, S. S., & Chester, L. E. H. (2021). The Justice System and Young Adults With Substance
Use Disorders. Pediatrics, 147(Suppl 2), S249–S258. https://doi.org/10.1542/PEDS.2020-
023523H
Qi, Y. (2012). Random forest for bioinformatics. In Ensemble machine learning (pp. 307–323).
Springer.
Rabinovitz, S., Desai, M., Schneir, A., & Clark, L. (2010). No Way Home: Understanding the
Needs and Experiences of Homeless Youth in Hollywood.
Rosa Solorio, M., Milburn, N. G., Andersen, R. M., Trifskin, S., & Gelberg, L. (2006). Health
care service use among vulnerable adolescents. Vulnerable Children and Youth Studies,
1(3), 205–220. https://doi.org/10.1080/17450120600973437
Rosenthal, D., Rotheram-Borus, M. J., Batterham, P., Mallett, S., Rice, E., & Milburn, N. G.
(2007). Housing stability over two years and HIV risk among newly homeless youth. AIDS
and Behavior, 11(6), 831–841. https://doi.org/10.1007/S10461-007-9235-6
Roy, É., Robert, M., Fournier, L., Vaillancourt, É., Vandermeerschen, J., & Boivin, J. F. (2014).
Residential trajectories of street youth-the Montréal Cohort Study. Journal of Urban
Health : Bulletin of the New York Academy of Medicine, 91(5), 1019–1031.
https://doi.org/10.1007/S11524-013-9860-5
Sahker, E., Ali, S. R., & Arndt, S. (2019). Employment recovery capital in the treatment of
substance use disorders: Six-month follow-up observations. Drug and Alcohol Dependence,
205. https://doi.org/10.1016/J.DRUGALCDEP.2019.107624

161

Sallis, J. F., Owen, N., & Fisher, E. B. (2008). Ecological Models of Health Behavior. In K.
Glanz, B. Rimer, & K. Viswanath (Eds.), Health Behavior: Theory, Research, and Practice
(4th ed., pp. 465–482). Jossey-Bass.
Satorra, A., & Bentler, P. M. (2001). A scaled difference chi-square test statistic for moment
structure analysis. Psychometrika 2001 66:4, 66(4), 507–514.
https://doi.org/10.1007/BF02296192
Semborski, S., Redline, B., Madden, D., Granger, T., & Henwood, B. (2021). Housing
interventions for emerging adults experiencing homelessness: A scoping review. Children
and Youth Services Review, 127. https://doi.org/10.1016/J.CHILDYOUTH.2021.106081
Shadel, W. G., Tucker, J. S., Mullins, L., & Staplefoote, L. (2014). Providing smoking cessation
programs to homeless youth: The perspective of service providers. Journal of Substance
Abuse Treatment, 47(4), 251–257. https://doi.org/10.1016/J.JSAT.2014.05.009
Shah, M. F., Liu, Q., Mark Eddy, J., Barkan, S., Marshall, D., Mancuso, D., Lucenko, B., &
Huber, A. (2017). Predicting Homelessness among Emerging Adults Aging Out of Foster
Care. American Journal of Community Psychology, 60(1–2), 33–43.
https://doi.org/10.1002/AJCP.12098
Shelton, K., Taylor, P., Bonner, A., & van den Bree, M. (2009). Risk Factors for Homelessness:
Evidence From a Population-Based Study. Psychiatric Services, 60(4), 465–472.
https://doi.org/10.1176/appi.ps.60.4.465
Slesnick, N., Chavez, L., Bunger, A., Famelia, R., Ford, J., Feng, X., Higgins, S., Holowacz, E.,
Jaderlund, S., Luthy, E., Mallory, A., Martin, J., Walsh, L., Yilmazer, T., & Kelleher, K.
(2021). Housing, opportunities, motivation and engagement (HOME) for homeless youth
at-risk for opioid use disorder: study protocol for a randomized controlled trial. Addiction
Science & Clinical Practice, 16(1). https://doi.org/10.1186/S13722-021-00237-7
Slesnick, N., Glassman, M., Garren, R., Toviessi, P., Bantchevska, D., & Dashora, P. (2008).
How to open and sustain a drop-in center for homeless youth. Children and Youth Services
Review, 30(7), 727. https://doi.org/10.1016/J.CHILDYOUTH.2007.12.004
Slesnick, N., Guo, X., Brakenhoff, B., & Feng, X. (2013). Two-year predictors of runaway and
homeless episodes following shelter services among substance abusing adolescents. Journal
of Adolescence, 36(5), 787–795. https://doi.org/10.1016/J.ADOLESCENCE.2013.06.007
Slesnick, N., Zhang, J., & Yilmazer, T. (2018). Employment and Other Income Sources Among
Homeless Youth. Journal of Primary Prevention, 39(3), 247–262.
https://doi.org/10.1007/s10935-018-0511-1
Society for Adolescent Health and Medicine. (2018). The Healthcare Needs and Rights of Youth
Experiencing Homelessness. Journal of Adolescent Health, 63(3), 372–375.
https://doi.org/10.1016/j.jadohealth.2018.06.012

162

Spencer, A. E., Valentine, S. E., Sikov, J., Yule, A. M., Hsu, H., Hallett, E., Xuan, Z.,
Silverstein, M., & Fortuna, L. (2021). Principles of Care for Young Adults With Co-
Occurring Psychiatric and Substance Use Disorders. Pediatrics, 147(Suppl 2), S229–S239.
https://doi.org/10.1542/PEDS.2020-023523F
Stapley, E., Vainieri, I., Li, E., Merrick, H., Jeffery, M., Foreman, S., Casey, P., Ullman, R., &
Cortina, M. (2021). A Scoping Review of the Factors That Influence Families’ Ability or
Capacity to Provide Young People With Emotional Support Over the Transition to
Adulthood. Frontiers in Psychology, 12. https://doi.org/10.3389/FPSYG.2021.732899
Suchting, R., Businelle, M. S., Hwang, S. W., Padhye, N. S., Yang, Y., & Santa Maria, D. M.
(2020). Predicting Daily Sheltering Arrangements among Youth Experiencing
Homelessness Using Diary Measurements Collected by Ecological Momentary Assessment.
International Journal of Environmental Research and Public Health, 17(18), 6873.
https://doi.org/10.3390/ijerph17186873
Tanner-Smith, E. E., Nichols, L. M., Loan, C. M., Finch, A. J., & Moberg, D. P. (2020).
Recovery High School Attendance Effects on Student Delinquency and Substance Use: the
Moderating Role of Social Problem Solving Styles. Prevention Science, 21(8), 1104–1113.
https://doi.org/10.1007/s11121-020-01161-z
Tevendale, H. D., Comulada, W. S., & Lightfoot, M. A. (2011). Finding shelter: Two-year
housing trajectories among homeless youth. Journal of Adolescent Health, 49(6), 615–620.
https://doi.org/http://dx.doi.org/10.1016/j.jadohealth.2011.04.021
Thompson, S. J., Mcmanus, H., Lantry, J., Windsor, L., & Flynn, P. (2006). Insights from the
street: Perceptions of services and providers by homeless young adults. Evaluation and
Program Planning, 29, 34–43. https://doi.org/10.1016/j.evalprogplan.2005.09.001
Tibshirani, R. (1996). Regression Shrinkage and Selection Via the Lasso. Journal of the Royal
Statistical Society: Series B (Methodological), 58(1), 267–288.
https://doi.org/10.1111/j.2517-6161.1996.tb02080.x
Tierney, W. G., Gupton, J. T., & Hallett, R. E. (2008). Transitions to Adulthood for Homeless
Adolescents: Education and Public Policy. https://eric-ed-
gov.libproxy2.usc.edu/?id=ED503645
Titus, J. C., Dennis, M. L., White, W. L., Scott, C. K., & Funk, R. R. (2003). Gender differences
in victimization severity and outcomes among adolescents treated for substance abuse.
Child Maltreatment, 8(1), 19–35. https://doi.org/10.1177/1077559502239612
Tucker, J. S., D’Amico, E. J., Ewing, B. A., Miles, J. N. V., & Pedersen, E. R. (2017). A group-
based motivational interviewing brief intervention to reduce substance use and sexual risk
behavior among homeless young adults. Journal of Substance Abuse Treatment, 76, 20–27.
https://doi.org/10.1016/j.jsat.2017.02.008

163

Tucker, J. S., D’amico, E. J., Pedersen, E. R., Rodriguez, A., & Garvey, R. (2020). Study
protocol for a group-based motivational interviewing brief intervention to reduce substance
use and sexual risk behavior among young adults experiencing homelessness. Addiction
Science & Clinical Practice, 15(1). https://doi.org/10.1186/S13722-020-00201-X
Tucker, J. S., Davis, J. P., Perez, L. G., Klein, D. J., & D’Amico, E. J. (2021). Late Adolescent
Predictors of Homelessness and Food Insecurity During Emerging Adulthood. Journal of
Adolescent Health. https://doi.org/10.1016/J.JADOHEALTH.2021.10.035
Tyler, K. A., & Olson, K. (2018). Examining the Feasibility of Ecological Momentary
Assessment Using Short Message Service Surveying with Homeless Youth. Field Methods,
30(2), 91–104. https://doi.org/10.1177/1525822X18762111
Tyler, K. A., Olson, K., & Ray, C. M. (2018). Understanding the Link between Victimization
and Alcohol Use among Homeless Youth Using Ecological Momentary Assessment.
Socius: Sociological Research for a Dynamic World, 4, 237802311877983.
https://doi.org/10.1177/2378023118779832
Tyler, K. A., & Ray, C. M. (2019a). A Latent Class Analysis of Lifetime Victimization Among
Homeless Youth. Journal of Interpersonal Violence, 886260519834090.
https://doi.org/10.1177/0886260519834090
Tyler, K. A., & Ray, C. M. (2019b). Risk and Protective Factors for Substance Use Among
Youth Experiencing Homelessness. Children and Youth Services Review, 107.
https://doi.org/10.1016/J.CHILDYOUTH.2019.104548
Tyler, K. A., & Schmitz, R. M. (2013). Family histories and multiple transitions among homeless
young adults: Pathways to homelessness. Children and Youth Services Review, 35(10),
1719–1726. https://doi.org/10.1016/j.childyouth.2013.07.014
Tyler, K. A., & Schmitz, R. M. (2020). Childhood Disadvantage, Social and Psychological
Stress, and Substance Use Among Homeless Youth: A Life Stress Framework. Youth and
Society, 52(2), 272–287. https://doi.org/10.1177/0044118X18767032
Up For Growth. (2022). Housing Underproduction in the U.S.
van den Bree, M. B. M., Shelton, K., Bonner, A., Moss, S., Thomas, H., & Taylor, P. J. (2009).
A Longitudinal Population-Based Study of Factors in Adolescence Predicting
Homelessness in Young Adulthood. Journal of Adolescent Health, 45(6), 571–578.
https://doi.org/10.1016/j.jadohealth.2009.03.027
van Leeuwen, J. M., Boyle, S., Salomonsen-Sautel, S., Baker, D. N., Garcia, J. T., Hoffman, A.,
& Hopfer, C. J. (2006). Lesbian, gay, and bisexual homeless youth: an eight-city public
health perspective. Child Welfare, 85(2), 151–170.
http://www.ncbi.nlm.nih.gov/pubmed/16846110
Venables, W. N., & Ripley, B. D. (2002). Modern Applied Statistics with S (4th ed.). Springer.

164

Vilsaint, C. L., Kelly, J. F., Bergman, B. G., Groshkova, T., Best, D., & White, W. (2017).
Development and validation of a Brief Assessment of Recovery Capital (BARC-10) for
alcohol and drug use disorder. Drug and Alcohol Dependence, 177, 71–76.
https://doi.org/10.1016/j.drugalcdep.2017.03.022
Wenzel, S., Holloway, I., Golinelli, D., Ewing, B., Bowman, R., & Tucker, J. (2012). Social
networks of homeless youth in emerging adulthood. Journal of Youth and Adolescence,
41(5), 561–571. https://doi.org/10.1007/s10964-011-9709-8
Whipple, C. R., Jason, L. A., & Robinson, W. L. V. (2016). Housing and abstinence self-efficacy
in formerly incarcerated individuals. Journal of Offender Rehabilitation, 55(8), 548–563.
https://doi.org/10.1080/10509674.2016.1229713
Whitbeck, L. B. (2009). Mental health and emerging adulthood among homeless young people.
In Mental health and emerging adulthood among homeless young people. Psychology
Press.
Whitbeck, L. B., Hoyt, D. R., & Yoder, K. A. (1999). A Risk-Amplification Model of
Victimization and Depressive Symptoms Among Runaway and Homeless Adolescents.
American Journal of Community Psychology, 27(2), 273–296.
https://doi.org/10.1023/A:1022891802943
Whitbeck, L. B., Rose, T., & Johnson, K. D. (2009). Social networks: Friends and families at
home and on the streets. In Mental Health and Emerging Adulthoodhood among Homeless
Young People (pp. 187–201). Psychology Press, Taylor & Francis Group.
Wild, T. C., Yuan, Y., Rush, B. R., & Urbanoski, K. A. (2016). Client Engagement in Legally-
Mandated Addiction Treatment: A Prospective Study Using Self-Determination Theory.
Journal of Substance Abuse Treatment, 69, 35–43.
https://doi.org/10.1016/J.JSAT.2016.06.006
Zhang, J., & Slesnick, N. (2018). Substance use and social stability of homeless youth: A
comparison of three interventions. Psychology of Addictive Behaviors, 32(8), 873–884.
https://doi.org/10.1037/adb0000424

Abstract (if available)

Abstract An estimated one in 10 young adults (aged 18 to 25) experience homelessness in the United States each year. Collectively, young adults experiencing homelessness (YAEH) are a highly vulnerable population at greater risk for poor physical and mental health outcomes. One such risk relates to substance use and substance use disorders (SUDs), which are exceedingly high among YAEH when compared to stably housed young adults. More research is needed to draw attention to the needs of YAEH who suffer from SUDs. One unexplored area to do this is within substance use treatment settings. Almost no research to date has comprehensively examined risk and protective factors for homelessness among young adults receiving substance use treatment. Neither is there a solid understanding of how homelessness and substance use may relate to each other (bidirectionally) over time among treatment-engaged young adults. Furthermore, more research is needed to understand the barriers and facilitators of treatment engagement from the perspective of both YAEH and providers. The following dissertation is composed of three studies that address these domains of inquiry. Study 1 uses Machine Learning (ML) to identify risk and protective factors for homelessness in a large sample of young adults receiving treatment in the U.S. Study 2 examines bidirectional (i.e., reciprocal) associations between homelessness and substance use during and after treatment. Finally, Study 3 uses qualitative methods to investigate barriers and facilitators to treatment among YAEH (n = 22) and service providers (n = 16) in Los Angeles, California. Taken together, this research may inform efforts to engage YAEH with SUDs in substance use treatment and address vulnerable young adults’ housing needs during and after treatment.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Understanding emotional regulation and mood of young adults in the context of homelessness using geographic ecological momentary assessment

PDF

Substance use disorder treatment clinician and director challenges working with collaborative justice court and child welfare systems: A multisite qualitative study

PDF

Tobacco use change among formerly homeless supportive housing residents: socioecological barriers and facilitators to cessation

PDF

Couch-surfing among youth experiencing homelessness: an examination of HIV risk

PDF

Pain and multi-morbidity among veterans: theory-guided, data-driven, and narrative approaches

PDF

The impact of childhood trauma on substance use and mental health during the SARS-CoV-2 pandemic among young adults

PDF

Technology enhanced substance use disorder treatment

PDF

Integrative care strategies for older adults experiencing co-occurring substance use and mental health disorders (I-CARE)…

PDF

Engaging homeless men and shelter providers to adapt an existing evidence-based HIV prevention intervention

PDF

Impact of change in sexual identity on mental health risks among sexual minority adolescents

PDF

Transitional housing and wellness center: a holistic approach to decreasing homelessness and mental illness in the Black community

PDF

Social network engagement and HIV risk among homeless former foster youth

PDF

A qualitative study of street fentanyl in Dayton, Ohio: drug markets, trajectories, and overdose risk reduction

PDF

Discrimination at the margins: perceived discrimination and the role of social support in mental health service use for youth experiencing homelessness

PDF

Stigma-free pregnancy: a recruitment and retention strategy for healthcare systems to engage pregnant women with substance use disorder in collaborative care

PDF

Interagency collaboration: cultivating resources and strengthening substance use service delivery in child welfare

PDF

Social network norms and HIV risk behaviors among homeless youth in Los Angeles, California

PDF

Social network and contextual influences on substance use and HIV risk behavior among young men who have sex with men

PDF

Essays on the U.S. market for substance use treatment and the impact of Medicaid policy reform

PDF

The interplay between social connection and substance use

Asset Metadata

Creator DiGuiseppi, Graham (author)

Core Title Homelessness and substance use treatment: using multiple methods to understand risks, consequences, and unmet treatment needs among young adults

School Suzanne Dworak-Peck School of Social Work

Degree Doctor of Philosophy

Degree Program Social Work

Degree Conferral Date 2023-05

Publication Date 04/03/2023

Defense Date 03/27/2023

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag Homelessness,machine learning,OAI-PMH Harvest,qualitative,service use,substance use,substance use treatment,Young adults

Format theses (aat)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Davis, Jordan (committee chair), Henwood, Benjamin (committee member), Huey, Stanley (committee member), Prindle, John (committee member), Rice, Eric (committee member)

Creator Email diguisep@usc.edu,grahamdiguiseppi@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC112922127

Unique identifier UC112922127

Identifier etd-DiGuiseppi-11557.pdf (filename)

Legacy Identifier etd-DiGuiseppi-11557

Document Type Dissertation

Format theses (aat)

Rights DiGuiseppi, Graham

Internet Media Type application/pdf

Type texts

Source 20230404-usctheses-batch-1015 (batch), University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright. The original signature page accompanying the original submission of the work to the USC Libraries is retained by the USC Libraries and a copy of it may be obtained by authorized requesters contacting the repository e-mail address given.

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA

Repository Email cisadmin@lib.usc.edu