University of Southern California Dissertations and Theses

The impact of psychosocial rehabilitation services on recovery outcomes for racial/ethnic minorities with severe mental illness

(USC Thesis Other)

The impact of psychosocial rehabilitation services on recovery outcomes for racial/ethnic minorities with severe mental illness

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

THE IMPACT OF PSYCHOSOCIAL REHABILITATION SERVICES ON

RECOVERY OUTCOMES FOR RACIAL/ETHNIC MINORITIES WITH SEVERE

MENTAL ILLNESS

by

Melissa Ann Edmondson

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

December 2012

Copyright 2012 Melissa Ann Edmondson
ii
DEDICATION
The following dissertation is dedicated to my parents, Nancy Edmondson and
Lloyd Edmondson, without whom I would not be the person I am today. Words cannot
express how grateful I am for your unconditional love, guidance and support throughout
my life and academic endeavors. I love you Mom and Dad!
This dissertation is also dedicated to the memory of my grandparents, Mary
DiBartolo, George DiBartolo, Lucille Edmondson and Lloyd Edmondson. I love and
miss you all.
iii
ACKNOWLEDGEMENTS
I wish to thank my dissertation committee members, John Brekke, Ph.D., Karen
Lincoln, Ph.D., John McArdle, Ph.D., and Larry Palinkas, Ph.D., for their continual
support, sharing of knowledge, feedback and constructive criticism throughout the
dissertation process. I especially would like to express my gratitude to my chair, Dr.
Brekke for his guidance, wisdom and support throughout my doctoral education. It has
truly been a privilege to be mentored by you. I would also like to acknowledge all the
professors who have supported me during my doctoral education and the University of
Southern California School of Social Work for the educational opportunities afforded to
me.
I also wish to thank my family, friends, and fellow students who have supported
me throughout my educational endeavors. Special thanks to my fiancé, Antoine Smith;
my parents, Nancy and Lloyd Edmondson; and my siblings, Nanci Edmondson, Jason
Edmondson and Laurie Dorsainvil for their love and support. I would also like to give a
special thanks to my west coast family, the Colemans, who provided me with love,
support, and wisdom throughout this process. To the “Fab Five,” thank you for your
unending and continued friendship, encouragement, inspiration, and support through this
process.
I would like to acknowledge the Council on Social Work Education, Minority
Fellowship Program and the Association for Pan African Doctoral Scholars for their
guidance, educational, emotional and financial support throughout my doctoral program.
I would also like to acknowledge the consumers with severe mental illness that I
have had the privilege of working with as a social work practitioner, administrator, and
iv
doctoral student. You are an inspiration to overcoming challenges in the face of
adversity.
The data analyzed in this dissertation is from an ongoing research study,
“California Mental Health Services Act: Impact on Practice & Organizational Culture in
Public Clinics,” funded by the National Institute of Mental Health (R-01 MH080671)
(Principal Investigators: Joel Braslow, Ph.D., M.D., and John Brekke, Ph.D.).

v
TABLE OF CONTENTS

Dedication ii

Acknowledgements iii

List of Tables vii

List of figures ix

Abstract x

Chapter One: Introduction and Overview of the Three Studies 1
Introduction
Rationale
Organization of Dissertation

Chapter Two (Study 1): Recovery outcomes for Racial/Ethnic Minorities 16
with Severe Mental Illness Receiving Psychosocial Rehabilitation Services
Introduction
Methods
Results
Discussion

Chapter Three (Study 2): An Exploratory Analysis of Predictors of 55
Symptom Trajectories for Individuals with Severe Mental Illness
Introduction
Methods
Results
Discussion

Chapter Four (study 3): An Exploratory Analysis of Predictors of Functioning 88
and Quality of Life for Individuals with Severe Mental Illness
Introduction
Methods
Results
Discussion

Chapter Five: Conclusion 127
Purpose of the Studies
Major Findings
Practice and Research Implications

References 135

vi
Appendices
Appendix A: Example Mplus Scripts for Multilevel Models 150
Appendix B: Example Mplus Scripts for Growth Mixture Modeling 154
vii
LIST OF TABLES

Table 1: Sample Characteristics at Baseline 24

Table 2: Comparison of Baseline Characteristics by Services Received 31

Table 3: Comparison of Baseline Characteristics by Race 33

Table 4: Nonlinear Change Model of Psychotic Symptoms 37

Table 5: Nonlinear Change Model of Depression/Anxiety Symptoms 39

Table 6: Nonlinear Change Model of Overall Functioning 41

Table 7: Nonlinear Change Model of Social Functioning 43

Table 8: Linear Change Model of Satisfaction with Life 45

Table 9: Sample Baseline Characteristics (N = 426) 60

Table 10: Comparison of Models for Psychotic Symptoms 67

Table 11: Standardized Estimates for a Two-Class Mixture Model for Psychotic 70
Symptoms

Table 12: Sample Statistics by Class for Psychotic Symptoms 72

Table 13: Comparison of Models for Depression/Anxiety 75

Table 14: Standardized Estimates for a Two-Class Mixture Model for 76
Depression/Anxiety

Table 15: Sample Statistics by Class for Depression/Anxiety 78

Table 16: Sample Baseline Characteristics (N = 426) 94

Table 17: Comparison Model for Social Functioning 101

Table 18: Comparison of Models for Overall Functioning 103

Table 19: Standardized Estimates for a Three-Class Mixture Model for Overall 106
Functioning

Table 20: Odds Ratios for Likely Class Membership Compared to "Low 108
Difficulty Maintainers" class for Overall Functioning
viii

Table 21: Sample Statistics by Class for Overall Functioning 111

Table 22: Comparison of Models for Satisfaction with Life 113

Table 23: Standardized Estimates for a Two-Class Mixture Model for 115
Satisfaction with Life

Table 24: Sample Statistics by Class for Satisfaction with Life 118

ix
LIST OF FIGURES

Figure 1: Two Level Linear Change Model 29

Figure 2: Two Level Nonlinear Change Model 35

Figure 3: Growth Curve Trajectories for the Two-Class Model of Psychotic 68
Symptoms

Figure 4: Sample Statistics by Class for Satisfaction with Life 74
Depression/Anxiety

Figure 5: Growth Curve Trajectories for the Three-Class Model of 104
Overall Functioning

Figure 6: Growth Curve Trajectories for the Two-Class Model of Satisfaction 117
with Life

x

ABSTRACT
Psychosocial rehabilitation services (PRS) are designed to improve recovery
outcomes (symptoms, functioning and quality of life) for individuals with severe mental
illness. However, the heterogeneity in the course of severe mental illnesses suggests that
there will be variation in response to PRS. Studies have shown that in addition to
treatment effects, demographic and clinical factors may play a role in heterogeneity in
recovery outcomes. In addition, limited research has examined recovery outcomes for
racial/ethnic minorities with severe mental illness despite theory and research suggesting
that cultural factors may influence recovery. Using quantitative data from an ongoing
study funded by the National Institute of Mental Health, the goal of this dissertation was
to understand the impact of PRS on recovery outcomes for racially/ethnically diverse
individuals with a severe mental illness receiving services from five mental health clinics
in Los Angeles County over the course of one year. The dissertation is organized as three
studies. Study 1 examined the differential effects of PRS on recovery outcomes for Euro
American, African American, Latino and multiracial individuals with a severe mental
illness. Studies 2 and 3 explored the demographic, clinical, and service predictors of
heterogeneity in recovery outcomes. Overall, the findings indicated that: 1) psychosocial
rehabilitation services maintained or improved recovery outcomes over time, 2) there was
variation across individuals in the trajectory of recovery outcomes, and 3) recovery
outcomes differed based on a variety of demographic and clinical factors, including race
and ethnicity. Mental health practice and research implications as well as suggestions for
xi
future research for racially/ethnically diverse individuals with severe mental illnesses are
discussed.
1
CHAPTER ONE:
Introduction
Approximately one in seventeen adults in the United States are diagnosed with a
severe mental illness (Kessler, Chiu, Demler & Walters, 2005; National Institute of
Mental Health [NIMH], 2008).

According to NIMH (2008), mental illnesses are “the
leading cause of disability in the U.S.” For example, individuals diagnosed with
schizophrenia may experience a range of functional impairments which can lead to social,
psychological and economic problems including unemployment, homelessness, social
isolation, and family disruption (U.S. Department of Health & Human Services, 1999).
However, the severity and long-term impact of severe mental illness varies widely.
Research on the long-term course of schizophrenia indicates that there is significant
heterogeneity in the course of illness (Drake, McHugo, Xie, Packard, & Helmstetter,
2006; Harding, Brooks, Ashikaga, & Strauss, 1987) suggesting that for some individuals
with schizophrenia there is a better prognosis for recovery (Harrow, Grossman, Jobe, &
Herbener, 2005).

Interventions such as psychosocial rehabilitation services (PRS) for individuals
with severe mental illness, which vary in service intensity and treatment focus, target
multiple indicators of recovery outcomes including symptom remission and improvement
in functioning and quality of life (Liberman, 2008). However, studies examining the
heterogeneity within samples of individuals with severe mental illness receiving
psychosocial rehabilitation services in regards to outcomes are underrepresented in the
existing intervention literature (Peer, Kupper, Long, Brekke, & Spaulding, 2007; Peer &
Spaulding, 2007). Examining heterogeneity within samples in regards to outcomes can
2
help us understand: 1) the type of services that promote recovery outcomes; 2) how
individuals differentially respond to services; and 3) how to modify or tailor services to
meet individual needs especially for those who show a poor response to treatment (Peer
& Spaulding, 2007).
Studies examining outcomes for racial/ethnic minorities with severe mental
illness receiving psychosocial rehabilitation services are also underrepresented. Although
mental health research in other areas such as service utilization and quality of care have
documented racial/ethnic disparities

(U.S. Department of Health and Human Services,
2001), the few studies that have examined treatment outcomes for racial/ethnic minorities
are inconclusive, with some studies reporting worse or poor outcomes for racial/ethnic
minorities (Phillips, Barrio, & Brekke, 2001; Telles et al., 1995), others reporting no
racial/ethnic differences (Bae, Brekke, & Bola, 2004; Jerrell &Wilson, 1997; Phillips et
al., 2001), and still others reporting favorable outcomes (Baker, Stokes-Thompson, Davis,
Gonzo, & Hishinuma, 1999).

A review of culturally relevant treatment practices highlights the importance of
examining differential effects of services for racial/ethnic minorities. Important
protective cultural factors such as a sociocentric orientation that promote family and
social support, and interdependent living for African American, Latino and Asian
consumers may be weakened or enhanced by mental health services (Barrio, 2000).
Other protective factors such as a level of “health paranoia” for African Americans in
response to societal racism and oppression have been misinterpreted as pathology for
African Americans, which has been cited as a contributing factor to the over diagnosis of
African Americans with paranoid schizophrenia in clinical settings (Whaley, 1997;
3
Whaley, 2001). Furthermore, the relationship between a culturally competent provider
and consumer is an important determinant in consumer outcomes (Yamada & Brekke,
2008). Larrison, Schoppelrey, Hack-Ritzo, and Korr (2011) found a direct relationship
between consumer outcomes and provider’s cultural competency for African Americans
with severe mental illness. As such, race/ethnicity and culture appear to impact mental
health services for racial/ethnic minorities with severe mental illness on multiple levels
ranging from diagnosis to outcomes.
Based on these issues, the goal of this dissertation is to understand the impact of
psychosocial rehabilitation services on recovery outcomes of racially/ethnically diverse
individuals diagnosed with a severe mental illness. The specific aims of this study are:
Aim 1: To examine the differential effects of psychosocial rehabilitation on recovery
outcomes for Euro American, African American, Latino and multiracial individuals with
a severe mental illness.
Aim 2: To explore the predictors of heterogeneity in recovery outcomes within a sample
of individuals with a severe mental illness receiving psychosocial rehabilitation services.
Findings from this dissertation will contribute to the small but growing body of
literature that exists regarding the differential impact of psychosocial rehabilitation
services on recovery outcomes for individuals of different racial/ethnic backgrounds.
Findings will also provide insight into the heterogeneity in outcomes for individuals with
severe mental illness receiving psychosocial rehabilitation services. It is important for
intervention researchers and service providers to understand which services work for
whom so that appropriate referrals are made and recovery of individuals with severe
mental illness is promoted.
4
Rationale
The rationale for this dissertation is based on the current state of knowledge on
definitions of recovery; heterogeneity in recovery outcomes; and outcome research on
racial/ethnic minorities with severe mental illness receiving psychosocial rehabilitation
services.
Definitions of Recovery
“Recovery” as a concept has been used to define a process, a mental health
service orientation and an outcome (Andresen, Oades, & Caputi, 2003; Færden, Nesvåg,
& Marder, 2008; Silverstein & Bellack, 2008; Stainsby, Bledin, & Mason, 2008). As a
process, several models have identified key elements in a consumer’s process of recovery
including hope, optimism, empowerment, rebuilding/restructuring of a sense of
self/identity, coping with stigma, finding meaning in life and taking responsibility for
one’s own recovery (Anthony, 1993; Baxter, & Diehl, 1998; Davidson & Strauss, 1992;
Resnick, Rosenheck, & Lehman, 2004; U.S. Department of Health and Human Services,
2005). As a mental health service orientation, recovery oriented mental health services
aim to foster the process of recovery and include services that: 1) are self-directed,
individualized and person-centered, empowering, holistic, non-linear, strengths-based;
(2) incorporate peer support; and (3) foster respect, responsibility, and hope (Bellack,
2006).
As an outcome, several operational definitions of recovery exist in the literature
(Andreasen et al., 2005; Harvey & Bellack 2009; Liberman, Kopelowicz, Ventura, &
Gutkind, 2002). In general, recovery as an outcome has been defined as symptom
remission, improvement in functioning and/or improvement in quality of life over a
5
specified period of time (Bellack, 2006; Leucht & Lasser, 2006; Liberman & Kopelowicz,
2005; Lysaker, Roe & Buck, 2010). For example, Liberman et al. (2002) defines
recovery from schizophrenia as: 1) remission of psychotic symptoms as indicated by a
score of 4 or less on the Brief Psychiatric Rating Scale (BPRS); 2) full- or part-time
engagement in school or work; 3) living independently; and 4) active social relationships
for a sustained period of two years.
The Remission in Schizophrenia Working Group was formed in 2003 to develop a
consensus definition of symptom remission in schizophrenia. The working group
focused on establishing criteria for symptom remission by reviewing the literature in
schizophrenia and using established symptom remission models for mood disorders as a
guide (Andreasen, et al. 2005). The criteria included reaching and maintaining low to
mild levels of positive, negative and disorganized symptoms as indicated by items on
four widely used scales (the Positive and Negative Syndrome Scale, BPRS, Scale for
Assessment of Positive Symptoms and Scale for Assessment of Negative Symptoms) for
a minimum of 6 months. In addition to the criteria for symptom remission, the working
group established that “individuals may remain in remission while experiencing minor
changes in symptoms in the absence of appreciable effects on daily function or subjective
well-being” (Andreasen et al., 2005, p. 447). The working group also concluded that use
of the symptom remission criteria in research and clinical practice would assist in the
development of a consensus definition of recovery that included functional outcome
criteria.
In the most recent definition of functional recovery, Harvey and Bellack (2009)
proposed a level and breadth approach to functional recovery in schizophrenia that
6
provides a continuum of progress and success in instrumental (productive activities such
as work), independent (residential and daily living skills) and social functioning. The
level of accomplishment in functioning ranges from making no attempts toward
improvement to complete success in independent living; employment or education; or
social relationships. The breadth of accomplishment refers to whether or not partial or
full success has been achieved across one or more of the three functional outcomes.
Furthermore, this definition of functional recovery acknowledges the previously
described criteria for symptom remission as a prerequisite to functional recovery. As
such, the timeframe for functional recovery is defined as 6 months after individuals have
developed symptom remission.
Although all of the above definitions of recovery outcomes acknowledge quality
of life as an important indicator of recovery, clear outcome criteria for quality of life has
not been well established. However, a few empirical studies have included quality of life
indicators as a recovery outcome in addition to symptom remission and functional
improvement (Drake et al., 2006; Novick, Haro, Suarez, Vieta, & Naber, 2009). This
dissertation study will focus on examining three recovery outcomes: symptoms,
functioning and quality of life.
Psychosocial Rehabilitation Services and Heterogeneity in Recovery Outcomes
Psychosocial rehabilitation services are designed to assist individuals with severe
mental illness in improving and maintaining symptom remission, functioning and quality
of life, all of the identified indicators of recovery outcomes for individuals with severe
mental illness (Liberman, 2008). However, heterogeneity in the course of severe mental
illness among individuals suggests that differences in response to psychosocial
7
rehabilitation will also occur. In a review of the use of longitudinal analytic methods for
examining treatment effects, Peer et al. (2007) found that antipsychotic medication,
cognitive behavioral therapy and social skills training were shown to be effective in
reducing psychotic symptoms and social skills training was effective in improving social
functioning improvement. However, Peer et al. highlight the juxtaposition of the
dynamic and fluctuating nature of severe mental illness and the lack of use of more
sophisticated longitudinal analytic methods that can describe how improvements occur,
when they occur and heterogeneity within samples in regards to improvement over the
course of treatment. Traditionally, outcome studies have examined outcomes in the
aggregate, such as examining sample means. In contrast, studies exploring heterogeneity
examine the variation in changes among individuals to understand differential responses
to treatment.
Of the studies examining heterogeneity within samples in terms of outcomes,
most have focused on functional outcomes, sampling either inpatient populations (Liu,
Choi, Reddy, & Spaulding, 2011; Peer & Spaulding, 2007; Peer, Strachan, & Spaulding,
2008) or community-based populations (Brekke, Long, Nesbitt, & Sobel, 1997; Brekke,
Hoe, Long, & Green, 2007; Kupper & Hoffman, 2000). In general, these studies have
found distinct groups that exhibit low, moderate or high levels of improvement in
functional outcomes over the course of treatment. Studies focusing on symptom
remission have also found significant heterogeneity, finding upwards of six distinct
groups that respond differentially to antipsychotic treatment (Case et al., 2011; Correll et
al., 2011). There are very few, if any, studies that examine heterogeneity in quality of
life outcomes.
8
When examining heterogeneity within samples in regards to functional outcome
and symptom remission, various clinical, treatment, and demographic variables have
been found to be associated with level of improvement, including service intensity,
history of psychiatric treatment, psychopathology, neurocognition, social cognition,
personality and coping skills, gender and education (Case et al., 2011; Correll et al.,
2011; Brekke, et al., 2007; Kupper & Hoffman, 2000; Peer & Spaulding, 2007; Peer et al.,
2008). In two studies, demographic variables such as age and education were not
associated with level of improvement (Peer et al., 2008; Peer et al., 2007). It is unclear if
other demographic variables such as race/ethnicity play a role in heterogeneity of
outcomes. However, theory regarding racial/ethnic minority mental health and previous
quantitative and qualitative studies with racial/ethnic minorities with severe mental
illness suggest that it may.
Research and Theory on Mental Health Outcomes of Racial and Ethnic Minorities
Mental Health Recovery Outcomes in Psychosocial Rehabilitation Services
Limited research has examined symptoms, functioning and quality of life
outcomes for racial/ethnic minorities with severe mental illness receiving psychosocial
rehabilitation services. Racial/ethnic variation has been found in symptom profile
(Brekke & Barrio, 1997), symptom expression (Yamada, Barrio, Morrison, Sewell, &
Jeste, 2006), and symptom severity (Barrio et al., 2003a; Chang, Newman, D’Antonio,
McKelvey, & Serper, 2011) for African Americans, Latinos, and whites with severe
mental illness receiving psychosocial rehabilitation services. In terms of functioning,
studies have reported mixed results for racial/ethnic minorities receiving psychosocial
rehabilitation services or inpatient treatment (Bae et al., 2004; Baker et al., 1999; DHHS,
9
2001; Jerrell & Wilson, 1997; Phillips et al., 2001; Telles et al., 1995). Phillips et al.
(2001) found that African American and Latinos declined in work functioning compared
with whites but found no differences in independent functioning. Racial/ethnic
differences in subjective quality of and satisfaction with life have also been reported. For
example, two studies found that African Americans with severe mental illness reported
higher subjective quality of life compared with whites (Lehman, Rachuba, & Postrado,
1995; Prince, 2006). However, one study found no racial/ethnic differences in changes
over time in satisfaction with life for African Americans, Latinos and whites receiving
psychosocial rehabilitation services (Bae et al., 2004).
Taken together, this body of research suggests that individuals of different racial
and ethnic groups may differentially respond to treatment depending on the targeted
outcome. Furthermore, theories regarding the mental health of racial and ethnic
minorities offer insight and explanation about differential responses to treatment and
differences in recovery outcomes.
Theories on Minority Mental Health
Among theories regarding the mental health of racial and ethnic minorities, two
competing theories attempt to explain racial and ethnic differences in mental health
outcomes. The minority status theory asserts that racial and ethnic minorities experience
more distress than nonminorities (Mirowsky & Ross, 1980). Furthermore, this distress is
caused by factors associated with minority status such as prejudice and discrimination,
which has a negative effect on mental health (Mirowsky & Ross, 1980; Sue & Chu,
2003). It appears that the minority status theory has not been applied to mental health
outcomes for individual with severe mental illness. However, recent reviews of studies
10
within the general population examining the effects of racism and discrimination on
racial and ethnic minorities have found an effect on a variety of psychological and mental
health outcomes including depression, anxiety, general psychological distress and
substance abuse (Harrell, 2000; Okazaki, 2009). Cokley, Hall-Clark and Hicks (2011)
found that perceived discrimination partially mediated the effect of ethnic minority status
on mental health outcomes for college students. Others have argued that the distress is
caused by factors associated with low socioeconomic status of which minorities are
disproportionately represented. However when socioeconomic status is controlled for,
minorities and nonminorities should have equal distress and similar mental health
outcomes (Mirowsky & Ross, 1980).
A more nuanced theory, the ethnic culture theory does not make an assumption
that all minorities are in poor mental health (Mirowsky & Ross, 1980). Instead it
assumes that mental health varies based on cultural values, traditions, and norms that may
serve as protective factors such a strong family, religious or community support
(Mirowksy & Ross, 1980; Sue & Stanley, 2003).
For example, in a study of symptom severity among individuals with severe
mental illness, Barrio and Brekke (1997) found that Latinos and African Americans had
lower severity of symptoms than whites. Additionally, lower symptom severity was
mediated by sociocentric indicators suggesting that cultural values related to
interdependence served as protective factors.
Furthermore unlike the minority status theory, the ethnic culture theory also takes
into account cultural differences among various racial and ethnic groups. In another
study on symptom severity, Barrio and colleagues (2003a) found that African American,
11
Latino, and Euro American participants with severe mental illness differed by type of
symptom in terms of severity. For example, African Americans reported higher severity
in hallucinatory/suspiciousness than Euro Americans; Euro Americans reported higher
severity in behavioral symptoms than African Americans; and Latinos reported higher
severity in somatic symptoms compared with both racial/ethnic groups. These results
suggest that manifestations of symptoms among different racial and ethnic groups may be
rooted in cultural differences.
Given that the ethnic culture theory is a viable explanation for racial and ethnic
differences in symptoms, this theory can also be extended as a possible explanation for
racial and ethnic differences that may occur in the other two recovery outcomes:
functioning and quality of life. For example, in regards to functioning, high amounts of
family contact have been directly linked to better functioning for African Americans with
severe mental illness (Guada, Hoe, Floyd, Barbour, & Brekke, 2012). This may be
associated with cultural values and norms of interdependence. African American
families have been shown to be highly involved and provide social support to an
individual family member with severe mental illness (Guarnacia, 1998; Horwitz &
Reinhard, 1995). Similarly, in regards to satisfaction with life, two studies that found that
African Americans reported a higher subjective quality of life compared to whites
(Lehman et al., 1995; Prince, 2006). The authors partially contribute these findings to
family support networks.
Given the increasing importance of and advances in defining recovery, the
heterogeneity found within samples of individuals with several mental illness in regards
to outcomes, and theories suggesting differences in outcomes for racial and ethnic
12
minorities with severe mental illness, the purpose of the dissertation is to examine the
impact of psychosocial rehabilitation services on recovery outcomes of racially and
ethnically diverse individuals diagnosed with a severe mental illness.
Organization of Dissertation
The dissertation is organized in a multiple manuscript format as three distinct but
interrelated studies designed to examine recovery outcomes for racial/ethnic minorities
receiving psychosocial rehabilitation services. The purpose of the multiple manuscript
format is to write multiple manuscripts with the intention of expediting publication of
findings. The dissertation study will use quantitative data from an ongoing research
study entitled “California Mental Health Services Act: Impact on Practice &
Organizational Culture in Public Clinics,” funded by the National Institute of Mental
Health (Braslow & Brekke, 2006) and from here on referred to as the parent study. The
parent study was designed to examine the impact of mental health system transformation
on practice and organizational culture in public mental health clinics in Los Angeles over
a three-year period. The parent study collected data on organizational-, provider- and
consumer-level outcomes. The three studies comprising this dissertation will focus on
consumer-level recovery outcomes.
Chapter One is a brief introduction of the rationale and organization of the
dissertation. Chapter Two (Study 1) will address Aim 1 of the dissertation by examining
the differential effects of psychosocial rehabilitation on recovery outcomes for Euro
American, African American, Latino, and multiracial individuals with a severe mental
illness. As operationalized in the existing literature, this study will conceptualize
recovery outcomes as three main domains that encompass the objective and subjective
13
indicators of recovery: symptoms, functioning and quality of life. Although these
outcomes have not been widely research for racial and ethnic minorities with severe
mental illness, the ethnic culture theory suggests that cultural protective factors may help
improve recovery outcomes for racial and ethnic minorities receiving services. As such,
psychosocial rehabilitation services will be hypothesized to differentially impact recovery
outcomes for individuals of different racial and ethnic backgrounds. Using multilevel
modeling, the following hypotheses will be tested:
1. African American, Latino, and multiracial individuals with a severe mental illness
receiving low intensity or high intensity psychosocial rehabilitation services will
have higher rates of symptom remission than Euro American individuals
receiving the same services.
2. African American, Latino, and multiracial individuals with a severe mental illness
receiving low intensity or high intensity psychosocial rehabilitation services will
have a higher rate of functional improvement than Euro American individuals
with severe mental illness receiving the same services.
3. African American, Latino, and multiracial individuals with a severe mental illness
receiving low intensity or high intensity psychosocial rehabilitation services will
have a higher rate of improvement in quality of life than Euro American
individuals with severe mental illness receiving the same services.
Chapter Three (Study 2) and Chapter Four (Study 3) will address Aim 2 of the
dissertation by exploring the demographic, clinical, and service predictors of
heterogeneity in recovery outcomes for individuals with a severe mental illness receiving
psychosocial rehabilitation services. Specifically, Study 2 will explore the heterogeneity
14
in symptoms. Research in heterogeneity in symptoms has focused primarily on response
to antipsychotic medication. Exploring the demographic, clinical, and service factors will
broaden our understanding of the determinants of variation in symptom remission. As
such, Study 2 is exploratory in nature and will address the following research questions:
1. Is there heterogeneity in the trajectory of symptoms for individuals with severe
mental illness receiving psychosocial rehabilitation services?
2. What are the demographic, clinical, and service predictors of symptom remission?
Although demographic characteristics such as age and education have been
explored as determinants in the heterogeneity in functioning, the importance of other
demographic variables such as race/ethnicity is less understood. Exploring the role of
demographic, clinical and service factors will provide a more comprehensive
understanding of the determinants of heterogeneity in functioning for individuals with
severe mental illness. Furthermore, a clear understanding of the main determinants of
variation in subjective recovery outcomes such as quality of life is lacking. Study 3 will
explore the heterogeneity in functioning and quality of life. Similar to Study 2, Study 3 is
also exploratory in nature and will address the following research questions:
1. Is there heterogeneity in the trajectory of functional outcomes among individuals
with severe mental illness receiving psychosocial rehabilitation services?
2. What are the demographic, clinical, and service predictors of functional
outcomes?
3. Is there heterogeneity in the trajectory of quality of life outcomes for individuals
with severe mental illness receiving psychosocial rehabilitation services?
4. What are the demographic, clinical, and service predictors of quality of life
outcomes?
15
Studies 2 and 3 will use Latent Class Growth Mixture Modeling (LCGMM) to
examine the above research questions. LCGMM is a longitudinal data analysis technique
that empirically identifies multiple latent groups, the shape of the trajectory change for
each group, the individuals that are likely to be in each group, and the features that
describe group membership. For example, the analysis may find three latent classes
(groups) of individuals that are low, moderate, or high on symptom remission. One could
then look at the characteristics of the individuals that have been assigned to each latent
class. As Study 2 and 3 are comparable in research design, the methods sections will be
similar.
The final chapter (Chapter Five) will integrate results from the three studies,
including a discussion of convergent/divergent findings and implications for social work,
future research, and practice. The findings from the three studies will have the potential
to inform consumers, researchers, and providers about the effectiveness of psychosocial
rehabilitation services in improving recovery outcomes for racial/ethnic minorities with
severe mental illness.

16
CHAPTER TWO (STUDY 1):
RECOVERY OUTCOMES FOR RACIAL/ETHNIC MINORITIES WITH
SEVERE MENTAL ILLNESS RECEIVING PSYCHOSOCIAL
REHABILITATION SERVICES

Introduction
Increasing focus has been given to the concept of recovery for individuals with
severe mental illness. In general, the concept of recovery suggests that consumers can
lead productive and satisfying lives in spite of the challenges of living with a severe
mental illness. Earlier definitions focused on the journey of recovery and included key
process-oriented elements such as hope, rebuilding a sense of identity, coping with
stigma, finding meaning in life and taking responsibility for one’s own recovery
(Anthony, 1993; Baxter, & Diehl, 1998; Davidson & Strauss, 1992). Definitions of
recovery have evolved to include more objective outcome-related criteria. Stemming
from intervention research, there is a consensus in the research community that recovery,
as an outcome, should include improvement in symptoms and functioning over a period
of 6 months to 2 years (Harvey & Bellack, 2009; Liberman, Kopelowicz, Ventura, &
Gutkind, 2002; Silverstein & Bellack, 2008). In fact, a recent review of long-term studies
indicates that upwards of 40% of individuals with severe mental illness such as
schizophrenia demonstrated recovery in symptoms and functioning (Silverstein &
Bellack, 2008). More recently, studies have included the examination of subjective
outcomes such as quality of life, that complement the more objective outcomes (Drake et
al., 2006; Novick, Haro, Suarez, Vieta, & Naber, 2009; Roe, Mashiach-Eizenberg, &
Lysaker, 2011). Therefore, definitions of recovery outcomes have come to include a
holistic, multidimensional criteria that encompass the objective indicators of a
17
consumer’s improvement in symptoms and functioning as well as the subjective
indicators of a consumer’s quality of life.
With increasing emphasis being placed on the importance of recovery for all
individuals with severe mental illness in practice settings and intervention research, little
focus has been given to what recovery looks like for racial/ethnic minorities in terms of
process and outcomes. However, considering the role that factors related to
race/ethnicity may play in recovery from severe mental illness is particularly important
given the documented racial/ethnic disparities in mental health service areas such as
diagnosis, service utilization and quality of mental health care. For example, African
Americans are diagnosed with schizophrenia in clinical settings at considerably high rates
as well as under-diagnosed with depression (Barnes, 2008; Das, Olfson, McCurtis, &
Weissman, 2006; Whaley, 2001). Research has also shown that African Americans over
utilize emergency room psychiatric services (Kuno & Rothbard, 2005). In contrast,
African Americans and Latinos underutilize outpatient services such as case management
and Assertive Community Treatment (Barrio et al., 2003b; Horvitz-Lennon, Zhou,
Normand, Alegría, & Thompson, 2011). In regards to quality of care, African Americans
with schizophrenia are less likely to be prescribed any antipsychotic medication and less
likely to be given best practice atypical antipsychotics which have less severe side effects
compared with whites (Busch, Lehman, Goldman, & Frank, 2009; Mallinger, Fisher,
Brown, & Lamberti, 2006). African Americans and Latinos are less likely to receive
adequate treatment for depression compared with whites (Das et al., 2006; Lagomasino et
al., 2005).
18
While disparities in recovery for racial/ethnic minorities have not been well
documented, limited research has examined symptoms, functioning and quality of life
outcomes for racial/ethnic minorities with severe mental illness. Racial/ethnic variation
has been found in symptom profile (Brekke & Barrio, 1997), symptom expression
(Yamada, Barrio, Morrison, Sewell, & Jeste, 2006) and symptom severity (Barrio et al.,
2003a; Chang, Newman, D’Antonio, McKelvey, & Serper, 2011) among African
Americans, Latinos, and whites with severe mental illness receiving psychosocial
rehabilitation services. In terms of functioning, studies have reported mixed results for
racial/ethnic minorities receiving psychosocial rehabilitation services or inpatient
treatment (Bae, Brekke, & Bola, 2004; Baker, Stokes-Thompson, Davis, Gonzo, &
Hishinuma, 1999; Jerrell & Wilson, 1997; Phillips, Barrio, & Brekke, 2001; Telles et al.,
1995). Phillips et al. (2001) found that African Americans and Latinos declined in work
functioning compared with whites but found no differences in independent functioning.
In contrast, Bae and colleagues (2004) found that African Americans had a slow rate of
improvement in social functioning; however found no racial/ethnic differences in other
areas of functioning.
Racial/ethnic differences in subjective quality of and satisfaction with life have
also been reported. Two studies found that African Americans with severe mental illness
reported higher subjective quality of life compared with whites (Lehman, Rachuba, &
Postrado, 1995; Prince, 2006). However, one study found no racial/ethnic differences in
changes over time in satisfaction with life for African Americans, Latinos, or whites
receiving psychosocial rehabilitation services (Bae et al., 2004). Taken together, this
19
body of research suggests that individuals of different racial/ethnic groups may
differentially respond to treatment depending on the targeted outcome.
Furthermore, theories regarding the mental health of racial and ethnic minorities
offer insight and explanation about differential responses to treatment and differences in
recovery outcomes. The minority status theory asserts that racial and ethnic minorities
experience more distress than nonminorities (Mirowsky & Ross, 1980). Furthermore,
this distress is caused by factors associated with minority status such as prejudice and
discrimination, which has a negative effect on mental health (Mirowsky & Ross, 1980;
Sue & Chu, 2003). Although minority status and the associated factors of prejudice and
discrimination may also explain the racial and ethnic disparities described earlier, this
theory only provides a partial explanation for differences in mental health outcomes.
Others have argued that the distress is caused by factors associated with low
socioeconomic status of which minorities are disproportionately represented. However
when socioeconomic status is controlled for, minorities and nonminorities should have
equal distress and similar mental health outcomes (Mirowsky & Ross, 1980).
A more nuanced theory, the ethnic culture theory does not make an assumption
that all minorities are in poor mental health (Mirowsky & Ross, 1980). Instead it
assumes that mental health varies based on cultural values, traditions, and norms that may
serve as protective factors such a strong family, religious or community support
(Mirowsky & Ross, 1980; Sue & Chu, 2003).
For example, in a study of symptom severity among individuals with severe
mental illness, Barrio and Brekke (1997) found that Latinos and African Americans had
lower severity of symptoms than whites. Furthermore, lower symptom severity was
20
mediated by sociocentric indicators suggesting that cultural values related to
interdependence served as protective factors.
Furthermore unlike the minority status theory, the ethnic culture theory also takes
into account cultural differences among various racial and ethnic groups. In another
study on symptom severity, Barrio and colleagues (2003a) found that African American,
Latino and Euro American participants with severe mental illness differed by type of
symptom in terms of severity. For example, African Americans reported higher severity
in hallucinatory/suspiciousness than Euro Americans; Euro Americans reported higher
severity in behavioral symptoms than African Americans; and Latinos reported higher
severity in somatic symptoms compared with both racial/ethnic groups. These results
suggest that manifestations of symptoms among different racial and ethnic groups may be
rooted in cultural differences.
Given that the ethnic culture theory is a viable explanation for racial and ethnic
differences in symptoms, this theory can also be extended as a possible explanation for
racial and ethnic differences that may occur in the other two recovery outcomes:
functioning and quality of life. For example, in regards to functioning, high amounts of
family contact have also been directly linked to better functioning for African Americans
with severe mental illness (Guada, Hoe, Floyd, Barbour, & Brekke, 2012). This may be
associated with cultural values and norms of interdependence. African American
families have been shown to be highly involved and provide social support to an
individual family member with severe mental illness (Guarnacia, 1998; Horwitz &
Reinhard, 1995). Similarly, in regards to satisfaction with life, two studies that found that
African Americans reported a higher subjective quality of life compared to whites
21
partially contribute these findings to family support networks (Lehman et al., 1995;
Prince, 2006).
Based on existing definitions of recovery outcomes, the current state of the
literature for racial/ethnic minorities with severe mental illness and the ethnic culture
theory, the purpose of the present study was to examine racial/ethnic differences in
symptoms, functioning and quality of life outcomes over a 1-year period among adults
with severe mental illness receiving psychosocial rehabilitation services.
Hypothesis 1: African American, Latino, and multiracial consumers receiving
psychosocial rehabilitation services will have a higher rate of improvement in symptoms
compared with Euro Americans receiving psychosocial rehabilitation services.
Hypothesis 2: African American, Latino, and multiracial consumers receiving
psychosocial rehabilitation services will have a higher rate of improvement in functional
outcomes compared with Euro Americans receiving psychosocial rehabilitation services.
Hypothesis 3: African American, Latino, and multiracial consumers receiving
psychosocial rehabilitation services will have a higher rate of improvement in quality of
life outcomes compared with Euro Americans receiving psychosocial rehabilitation
services.
Methods
Data used for the present study are from an ongoing research study designed to
examine the impact of mental health system transformation on consumer outcomes,
practice, and organizational culture in public mental health clinics in Los Angeles over a
three-year period (Braslow & Brekke, 2006). Participants in the original study are adults
diagnosed with a severe mental illness recruited upon admission from five mental health
22
clinics providing case management services or Full Service Provider (FSP) services
within the Los Angeles County Department of Mental Health (LACDMH). Based on an
Assertive Community Treatment (ACT) Team Model (Dixon et al., 2010), FSP teams
were staffed by a psychiatrist and other mental health providers such as social workers
and peer providers. Consumer-provider ratios were small, usually no more than 15 to 1.
FSP teams were available 24 hours a day, seven days a week and provided a variety of
services in the field or consumer’s home such as mental health treatment, case
management, housing and employment services.

As such, FSP teams were designed to
provide a higher intensity of services than usual care. A sampling strategy was used to
match FSP and usual care participants on diagnosis, Global Assessment of Functioning
(GAF) scores, and demographic characteristics as they began receiving services and were
recruited into the study.
Participants
The original sample consists of 482 ethnically diverse adults diagnosed with a
severe mental illness. Clinic consumers with a V-code diagnosis derived from the
Diagnostic and Statistical Manual of Mental Disorders (DSM-IV_TR 4
th
Edition) and/or
consumers on conservatorship were excluded from the study (Braslow & Brekke, 2006).
For the present study, the sample included 472 participants at baseline, 369 participants at
6 months and 331 participants at 12 months indicating an attrition rate of 22% at 6
months and 30% at 12 months. Asian American, Native American, and Hawaiian/Pacific
Islander participants were excluded from the present study due to the relatively low
proportion of participants in these racial and ethnic categories. A series of chi-square and
t-tests were conducted on demographic, clinical and outcome variables to test for attrition
23
bias at 12 months. No significant differences were found between participants who
completed data at 12 months and those who did not on baseline age, gender, race, marital
status, education, primary diagnosis, inpatient psychiatric hospitalizations or outcome
variables (p > .05). To deal with missing data, data were analyzed using Mplus 6.1,
which allows for the use of all data available to estimate statistical models (Muthén &
Muthén, 1998-2010).
Participants completed measures on all demographic and outcome variables at
baseline and completed measures on all outcome variables at 6 and 12 months. Missing
data on demographic variables was obtained from the LACDMCH information system
records when available. Clinical characteristics and diagnosis were also obtained from
LACDMH information system records. Sample characteristics are reported in Table 1.
Participants were predominantly racial or ethnic minorities, unmarried and 51.5 percent
were male. The average age was 40 years old with 12 years of education. Thirty-five
percent of participants were diagnosed with a schizophrenia spectrum disorder.
Participants reported low to moderate levels of difficulty with symptoms and functioning.
All study procedures were approved by the University of Southern California
Institutional Review Board.
Measures
All demographic and self-report outcome data were collected from participants by
trained research staff. Diagnostic, clinical, and service characteristics were obtained from
the LACDMH information system records.

24

Table 1
Sample Characteristics at Baseline
Gender
Females 48.3%
Males 51.5%
Transgender .2%
Race
Euro American 37.1%
African American 20.8%
Latino 28.0%
Multiracial 14.2%
Marital Status

Never Married 51.0%
Cohabitating/Married 13.3%
Previously Married 35.7%
Diagnosis
Schizophrenia Spectrum 35%
Mood/Anxiety Disorders 65%
Mean (SD)
Age 40.4 (9.9)
Education in Years 12.2 (2.2)
Hospitalizations
a
.5 (1.2)
SWLS 1.48 (.72)
BASIS-32 1.65 (.79)
Psychosis
b
1.19 (.99)
Depression/Anxiety
c
2.13 (1.00)
Social Functioning
d
1.95 (.97)
Overall Functioning
e
1.72 (.87)
Addictive Behavior
f
1.00 (.82)
Note. N=472. SWLS = Satisfaction with Life Scale.
BASIS-32 = Behavior and Symptom Identification Scale (scale
ranges from 0 to 4).
a
Number of psychiatric inpatient hospitalizations in the past year
b
BASIS-32 Psychosis Subscale
c
BASIS-32 Depression/Anxiety Subscale
d
BASIS-32 Relationship to Self/Others Subscale
e
BASIS-32 Daily Living/Role Functioning Subscale
f
BASIS-32 Impulsive/Addictive Behavior Subscale
25
Covariates.
Race and ethnicity. Race and ethnicity were measured using a self-report
demographic survey. Participants were given the option of selecting more than one racial
or ethnic category. Participants who were characterized as Euro American, African
American, or Latino were included in the final sample. Participants who were
characterized as Euro American endorsed ethnic categories of European decent.
Participants who endorsed black/African American were characterized as African
American. Participants who were characterized as Latino predominantly endorsed
Mexican or Mexican American categories. Participants who endorsed more than one
racial category were categorized as multiracial and also included in the final sample.
Race and ethnicity was dummy coded into three variables: African American, Latino and
multiracial. Euro Americans were the comparison group.
Service intensity. Service intensity was defined as the number of outpatient visits
a participant had and the number of minutes spent on each visit (Brekke, Ansel, Long,
Slade, & Weinstein, 1999). Results of t-tests indicated that the mean number of
outpatient visits and the mean number of minutes spent on each visit was significantly
higher for participants receiving FSP services than for participants receiving case
management services 12 months after admission (p <.0001). In addition as previously
mentioned, FSP teams followed an ACT team model, which was designed to provide
more intense services. As such, service intensity was dichotomized as low intensity (case
management services) and high intensity (FSP services).

26
Demographic/Clinical Characteristics. Demographic and clinical characteristics,
age, education, and diagnosis, were used as covariates in the final analysis. Diagnosis
was dichotomized into schizophrenia spectrum disorders and mood/anxiety disorders.
Outcome Variables. The outcome variables in this study included symptoms,
functioning and quality of life.
Symptoms and functioning. To measure symptoms and functional outcomes,
participants completed the Behavior and Symptom Identification Scale (BASIS-32). The
BASIS-32 is a brief, mental health assessment tool developed to measure change in
symptoms and functioning for individuals with severe mental illness across five domains:
relations to self/others, daily living/role functioning, depression/anxiety,
impulsive/addictive behavior and psychosis. It is a self-report measure that assesses the
level of difficulty in each domain over the past week on a 5-point scale ranging from 0
(no difficulty) to 4 (extreme difficulty) (Eisen & Culhane, 1999). The BASIS-32 has
been validated for use with outpatient consumers (Eisen, Wilcox, Leff, Schaefer, &
Culhane, 1999) and with major U.S. racial/ethnic groups (white, African American,
Latino American and Asian American) (Chow, Snowden, & McConnell, 2001). In
addition, the BASIS-32 has been shown to be sensitive to change over time (Jerrell,
2005).
For preliminary analysis, the average score for the BASIS-32 was used to
measure symptoms and functioning. Cronbach’s alpha for the BASIS-32 was .94. For
the main analysis, the daily living/role functioning subscale and the relationship to
self/others subscale were used to measure overall functioning and social functioning,
respectively. The depression/anxiety and psychosis subscales were used to measure
27
symptomatology. For each participant, the score for each subscale was obtained by
averaging the items in the subscale. Cronbach’s alpha for the 4 subscales ranged
from .74 to .85.
Quality of life. To measure quality of life, participants completed the Satisfaction
with Life Scale (SWLS). The SWLS is an 18-item, self-report scale measuring consumer
satisfaction with life across four domains: living situation, social relationships, work, and
self/present life. The SWLS is scored on a 5-point scale with consumers reporting
satisfaction ranging from 0 (not at all) to 4 (a great deal). The SWLS has been validated
on diverse, outpatient populations with severe mental illness and found to be
psychometrically sound (Lee, Brekke, Yamada, & Chou, 2010; Test, Greenberg, Long,
Brekke, & Burke, 2005). For each participant, the SWLS score was obtained by
averaging the scale items. Cronbach’s alpha for the SWLS was .91.
Data Analysis
The study hypotheses were tested using multilevel modeling. Multilevel
modeling allows for the modeling of group-level change over time and estimates
differences in the rates of change over time (Tabachnick & Fidell, 2007; Ferrer,
Hamagami & McArdle, 2004). Multilevel modeling will address the study hypotheses by
modeling group level rates of change in outcomes for individuals of different racial and
ethnic backgrounds over three time points. Multilevel modeling involves simultaneous
modeling change at two levels. Level one involves modeling the random effects of
individual change over time using the equation below:

28
Y
ij
= β
0i
+ β
1i
t
ij
+ e
ij

where Y
ij
is the outcome score for the i
th
participant at time j.
β
0
is the latent score representing an individual’s initial level (intercept);
β
1
is the latent score representing an individual’s change over time (slope);
t
ij
is the parameter representing time;
e
ij
is the residual error term

Level two involves modeling the fixed effects where the level-one parameters are treated
as dependent variables and regressed on a covariate as illustrated in the equations below:
β
0i
= β
0
+ β
2
(african american
i
) + β
4
(latino
i
) + β
6
(multiracial
i
) + b
0i
β
1i
= β
1
+ β
3
(african american
i
) + β
5
(latino
i
) + β
7
(multiracial
i
) + b
1i

African American, Latino, and multiracial represent dummy-coded race covariates with
Euro American as the comparison group where:
β
0
and β
1
are the regression intercepts
β
2
and β
3
are the regression slopes
b
0i
and b
1i
are disturbance variables

Figure 1 depicts a simplified two-level linear change model demonstrating group change
over one year. In this model, the factor loadings for the latent slope are fixed at baseline
(Y[0]), ½ year (Y[.5]), and 1 year (Y[1]) to indicate linear change over time. The
intercept was fixed at 1 for each time point to represent the initial level of the outcome
variable corrected for measurement error (Kline, 2011). To control for variation in
demographic, clinical, and service intensity characteristics, additional covariates were
added to the simplified model during model testing.
Two criteria were used determine if our hypotheses were supported: model fit and
significant parameter estimates. Criteria one included testing model fit for the level-one
and level-two models using chi-square and Root Mean Square Error of Approximation fit
statistics (RMSEA; Steiger, 1990). The chi-square statistic provides a test of perfect
29

Multi-
racial
Latino
African
American
β
6

β
7

β
4

β
2

β
5

β
0

β
3

β
1

b
0
b
1
Intercept

Slope

1

Y[0]

Y[.5]

Y[1]
Figure 1. Two Level Linear Change Model
1 1 1 1 0 .5
e[0]

e[.5] e[1]
30
model fit where a nonsignificant Chi-Square indicates a good model fit (Kline, 2011).
The RMSEA is a test of close model fit where values below .05, between .05 -.08, and
above .08 indicate a close fit, reasonable fit, and poor fit of the model to the data,
respectively. Additionally, if the upper bound of the 90% confidence interval for the
value of RMSEA is equal to or greater than .10, this is considered a poor fit (Kline, 2010;
Brown & Cudeck, 1993). Once adequate model fit was determined, parameter estimates
for the latent slope and race covariates were examined for significance.
Results
Preliminary Analysis
Service intensity. As participants were not randomized into case management or
FSP services, a series of chi-square tests were conducted to test for baseline differences
between service types (Table 2). Although a matched pair sample design was employed
to match participants on key demographic and clinical characteristics in these service
types, significant differences between the two service types were found on several
variables (Table 2). Furthermore, the current study is based on the assumption that
participants receiving high intensity services (FSP) would show more improvement in
symptoms, functioning and quality of life outcomes than participants receiving low
intensity services (case management services). As such, preliminary analyses were
conducted comparing the rate of change in overall symptoms and functioning (BASIS-
32) and in satisfaction with life (SWLS) for participants receiving high intensity and low
intensity services over a one-year period. A two-level change model was analyzed for
each outcome variable including service intensity as the second-level covariate. In
regards to overall symptoms and functioning, the model fit the data well (χ
2
=3, df =3,
31
Table 2
Comparison of Baseline Characteristics by Services Received
Usual Care
N = 298
Full Service Providers
N =174
Gender (%)*
Females 52.7% 40.8%
Males 47.0% 59.2%
Transgender .3% 0%
Race (%)
Euro American 36.6% 37.9%
African American 20.1% 21.8%
Latino 29.5% 25.3%
Multiracial 13.8% 14.9%
Marital Status (%)
+

Never Married 50.9% 51.2%
Cohabitating/Married 14.3% 11.6%
Previously Married 34.8% 37.2%
Diagnosis (%)
Schizophrenia Spectrum 33.2% 35.0%
Mood/Anxiety Disorders 66.8% 65.0%
Mean (SD) Mean (SD)
Age 40.18 (9.83) 40.68 (10.03)
Education in Years
++
12.09 (2.35) 12.31 (1.90)
Hospitalizations
a
** .40 (.92) .76 (1.52)
SWLS 1.43 (.73) 1.82 (.89)
BASIS-32** 1.72 (.83) 1.52 (.69)
Psychosis
b+++
* 1.26 (1.04) 1.07 (.88)
Depression/Anxiety
c++++
** 2.24 (1.03) 1.94 (.90)
Social Functioning
d++++
* 2.02 (1.01) 1.82 (.89)
Overall Functioning
e++++
* 1.78 (.91) 1.61 (.79)
Addictive Behavior
f+++++
* 1.06 (.84) .90 (.78)
Note. SWLS = Satisfaction with Life Scale. BASIS-32 = Behavior and Symptom
Identification Scale.
a
Number of psychiatric inpatient hospitalizations in the past year
b
BASIS-32 Psychosis Subscale
c
BASIS-32 Depression/Anxiety Subscale
d
BASIS-32 Relationship to Self/Others Subscale
e
BASIS-32 Daily Living/Role Functioning Subscale
f
BASIS-32 Impulsive/Addictive Behavior Subscale

*significant differences at p <.05, **significant differences at p < .01

+
Due to missing data total n = 451

++
Due to missing data total n = 432

+++
Due to missing data total n = 470

++++
Due to missing data total n = 471

32
p=.35; RMSEA = .01, C.I. = .00-.08). There was a significant difference in initial level
of symptom and functioning between participants receiving high intensity services and
participants receiving low intensity services (β=-.213, p=.004) indicating that high
intensity service participants had lower initial levels of difficulty with symptoms and
functioning. However, there were no differences in rates of change in symptoms and
functioning for participants in high intensity services compared with participants in low
intensity services (β=.001, p=.99).
For satisfaction with life, the two-level change model also fit the data well (χ
2
=5,
df =4, p=.34; RMSEA = .02, CI=.00-.07). There was a significant difference in initial
level of satisfaction with life between participants receiving high intensity services and
participants receiving low intensity services (β=.137, p=.004) indicating that high
intensity service participants had higher initial levels of satisfaction with life. Again,
there were no differences in the rates of change for participants in high intensity services
compared with participants in low intensity services (β=.004, p=.96).
Given the differences between service types found at baseline, it is not surprising
that differences were found in initial levels of difficulty with symptoms, functioning and
satisfaction with life. Due to these differences, service intensity was used as a control
variable in the main analysis to account for differences in baseline and any variation in
rates of improvement due to treatment condition.
Baseline comparisons by race. A series of chi-square and analysis of variance
(ANOVA) tests were conducted to compare baseline demographic, clinical and service
characteristics among the four racial/ethnic groups. Baseline demographic, clinical and
service characteristics by race are provided in Table 3. Significant differences in age,
33
Table 3
Comparison of Baseline Characteristics by Race
Euro
American
N = 175
African
American
N= 98

Latino
N= 132

Multiracial
N= 67
Gender
Females 44% 53.1% 45.5% 58.2%
Males 56% 46.9% 53.8% 41.8%
Transgender 0% 0% .8% 0%
Marital Status
Never Married 47.7% 54% 50.8% 56.1%
Cohabitating/Married 11% 9.2% 19% 13.6%
Previously Married 41.3% 36.8% 30.2% 30.3%
Diagnosis**
Schizophrenia Spectrum 26.9% 49% 39.4% 26.9%
Mood/Anxiety Disorder 73.1% 51% 60.6% 73.1%
Services Received
Case Management 62.3% 61.2% 66.7% 61.2%
Full Service Provider 37.7% 38.8% 33.3% 38.8%
Mean (SD) Mean (SD) Mean (SD) Mean (SD)
Age** 42.26 (9.61) 40.58 (9.62) 37.39 (10.11) 40.96 (9.44)
Education in Years* 12.76 (2.26) 11.73 (1.98) 11.73 (2.30) 12.14 (1.68)
Hospitalizations
a
.37 (.78) .53 (1.44) .70 (1.38) .61 (1.24)
SWLS 1.43 (.68) 1.41 (.69) 1.58 (.76) 1.50 (.80)
BASIS-32 1.61 (.78) 1.65 (.82) 1.68 (.82) 1.69 (.70)
Psychosis
b
** 1.03 (.94) 1.46 (1.09) 1.23 (.98) 1.14(.89)
Depression/Anxiety
b
2.14 (.98) 2.14 (1.02) 2.14 (1.07) 2.09(.93)
Social Functioning
d
2.01 (.95) 1.87 (.99) 1.96 (1.03) 2.12 (.91)
Overall Functioning
e
1.70 (.88) 1.80 (.87) 1.80 (.87) 1.79 (.84)
Addictive Behavior
e
.95 (.83) 1.08(.81) 1.00 (.85) 1.01 (.76)
Note. SWLS = Satisfaction with Life Scale. BASIS-32 = Behavior and Symptom Identification Scale.
a
Number of psychiatric inpatient hospitalizations in the past year
b
BASIS-32 Psychosis Subscale
c
BASIS-32 Depression/Anxiety Subscale
d
BASIS-32 Relationship to Self/Others Subscale
e
BASIS-32 Daily Living/Role Functioning Subscale
f
BASIS-32 Impulsive/Addictive Behavior Subscale
*significant differences at p <.05, **significant differences at p < .01

34
education, diagnosis, and psychotic symptoms were found. Consistent with the
disparities literature on diagnosis, chi-square analysis indicated that African Americans
were overrepresented in terms of schizophrenia spectrum disorders. African Americans
were diagnosed with schizophrenia spectrum disorder at a higher percentage (49%) than
the other three racial/ethnic groups (Table 3). Furthermore, an ANOVA test revealed
racial/ethnic differences in baseline psychotic symptoms. Post-hoc comparisons
indicated that African Americans reported significantly higher difficulty with psychotic
symptoms at baseline compared to Euro Americans (p < .05). Due to racial/ethnic
differences at baseline in age, education, diagnosis, and psychotic symptoms, these
variables will be used as control variables in the final analysis.
Main Analysis
In the first step of the analysis, a one-level linear change model was tested to
examine change over time in the outcome variable for the entire sample. When a linear
model did not show a good fit to the data, a nonlinear model was tested by allowing the
factor loading at the 0.5 year time point to vary (Figure 2). This model specifies that the
change in the outcome variable over the 3 time points is curvilinear, that is, not occurring
in a straight line. Based on model fit criteria, either a linear or nonlinear model was
chosen (Model 1). In Model 1, a significant parameter estimate for the latent slope
indicated a significant rate of change in the outcome variable. In the second model tested
(Model 2) the level-two dummy coded race and ethnicity covariates were added. Model
fit was assessed and the parameter estimates for race and ethnicity covariates were
examined for significance. A significant parameter estimate for African American,
Latino, or multiracial race covariates indicated a significant difference in the rate of
35

Multi-
racial
Latino
African
American
β
6

β
7

β
4

β
2

β
5

β
0

β
3

β
1

b
0
b
1
Intercept

Slope

1

Y[0]

Y[.5]

Y[1]
Figure 2. Two Level Nonlinear Change Model
1 1 1 1 0
e[0]

e[.5] e[1]
36
change in the outcome variable as compared with Euro Americans. In the third model
(Model 3) the baseline covariates age, education, diagnosis, psychosis and service
intensity were added to the model as control variables. Again, overall model fit was
assessed and significant parameter estimates were examined.
Symptoms. Hypothesis one stated that African American, Latino, and multiracial
consumers receiving psychosocial rehabilitation services would have a higher rate of
improvement in symptoms compared with Euro Americans receiving psychosocial
rehabilitation services. This hypothesis was not supported for psychotic symptoms. This
hypothesis was partially supported for depression/anxiety symptoms.
Psychotic symptoms. Table 4 shows the model fit and parameter estimates for
change models of psychotic symptoms that were tested. A nonlinear change model of
psychotic symptoms fit the data well (Model 1). The negative latent slope was
significant indicating that participants’ level of difficulty with psychotic symptoms
decreased over time. In other words, participants’ symptoms improved over the one-year
period. In Model 2, three dummy variables were added to represent race and ethnicity for
African American, Latino, and multiracial participants. Although Model 2 also fit the
data well, the model showed no significant differences in the rates of change in psychotic
symptoms for African American, Latino or multiracial participants compared with Euro
American participants. Although not originally hypothesized, the intercept for African
Americans was significant indicating that African Americans have a higher initial level of
psychosis compared with Euro Americans. In Model 3, age, education, diagnosis and
service intensity were added to the model. Again, the model fit well to the data and
African Americans had a significantly higher initial level of psychosis compared with
37
Table 4
Nonlinear Change Model of Psychotic Symptoms
Model 1 Model 2 Model 3
Level 1 β β β
Latent Intercept 1.19*** 1.032*** 1.43***
Latent Slope -.15** -.18** -.11
Level 2 Covariates
African American
a
Intercept
Slope

.43**
.05

.33*
.09
Latino
a
Intercept
Slope

.18
.08

.04
.12
Multiracial
a
Intercept
Slope

.10
.03

.04
.03
High Intensity Services
b
Intercept
Slope

-.20*
-.09
Age
Intercept
Slope

0
-.001
Education
Intercept
Slope

-.03
.01
Schizophrenia Spectrum
c
Intercept
Slope

.14
-.09
Latent Intercept Residual Variance .64 .62 .57
Latent Slope Residual Variance .12 .09 .05
Error Variance .36 .32 .33
Model Fit
χ
2
/df/p 1/2/.59 6/5/.29 16/9/.07
RMSEA
90% Confidence Interval
0
0-.08
.02
0-.07
.04
0-.08
R
2
.67*** .67*** .64***
Latent Intercept R
2
.04 .06*
Latent Slope R
2
.01 .11
Note. RMSEA=Root Mean Square Error of Approximation.
a
Comparison group is Euro American.
b
Comparison group is low intensity, case management services.
c
Comparison group is Mood/Anxiety Disorders.
*p < .05, **p < .01, ***p < .001

38
Euro Americans. After controlling for age, education, diagnosis and service intensity, no
significant differences were found for African American, Latino or multiracial
participants compared with Euro American participants.
Depression/Anxiety symptoms. Table 5 shows the model fit and parameter
estimates for change models of depression/anxiety symptoms that were tested. A
nonlinear change model of depression/anxiety symptoms fit the data well (Model 1). The
negative latent slope was significant indicating that participants’ level of difficulty with
depression/anxiety symptoms decreased over time. In Model 2, three dummy variables
were added to represent race and ethnicity for African American, Latino, and multiracial
participants. Although Model 2 also fit the data well, the model showed no significant
differences in the rates of change in difficulty with depression/anxiety symptoms for
African American, Latino, or multiracial participants compared with Euro American
participants. In Model 3, age, education, diagnosis and service intensity were added to
the model as control variables. Again, the model fit well to the data. After controlling
for age, education, diagnosis, and service intensity, significant differences were found
between African American and Euro American participants. Although not hypothesized,
compared with Euro Americans, African Americans had lower initial levels of difficulty
with depression/anxiety as indicated by a significant negative intercept. African
Americans also had a higher rate of change over one year compared with Euro Americans.
In other words, for African Americans difficulty with depression/anxiety symptoms
decreased more rapidly than Euro Americans. No significant differences in the rate of
change in difficulty with depression/anxiety symptoms were found for Latino or
multiracial participants compared with Euro American participants.
39

Table 5
Nonlinear Change Model of Depression/Anxiety Symptoms
Model 1 Model 2 Model 3
Level 1 β β β
Latent Intercept 2.14** 2.14*** 1.52***
Latent Slope -.36*** -.42*** -.49
Level 2 Covariates
African American
a
Intercept
Slope

.003
.14

-.22*
.27*
Latino
a
Intercept
Slope

.004
.06

-.08
.18
Multiracial
a
Intercept
Slope

-.06
.081

-.08
.14
High Intensity Services
b
Intercept
Slope

-.142
.08
Age
Intercept
Slope

.007
.002
Education
Intercept
Slope

-.01
.003
Psychosis
Intercept
Slope

.63***
.-.13**
Schizophrenia Spectrum
c
Intercept
Slope

-.44***
.07
Latent Intercept Residual Variance .64 .64 .21
Latent Slope Residual Variance .12 .12 .08
Error Variance .36 .35 .37
Model Fit
χ
2
/df/p 2/2/.35 5/5/.37 12/10/.32
RMSEA
90% Confidence Interval
.01
.00 - .09
.01
.00 - .07
.02
.00 - .06
R
2
.64*** .66*** .65***
Latent Intercept R
2
.001 .67***
Latent Slope R
2
.023 .26
Note. RMSEA= Root Mean Square Error of Approximation.
a
Comparison group is Euro American.
b
Comparison group is low intensity, case management services.
c
Comparison group is Mood/Anxiety Disorders.
*p < .05, **p < .01, ***p < .001
40
Functioning. Hypothesis two proposed that African American, Latino, and
multiracial consumers receiving psychosocial rehabilitation services would have a higher
rate of improvement in functioning compared with Euro Americans receiving
psychosocial rehabilitation services. This hypothesis was partially supported for overall
functioning and social functioning.
Overall functioning. Table 6 shows the model fit and parameter estimates for
change models of daily living/role functioning that were tested. A nonlinear change
model of daily living/role functioning symptoms fit the data well (Model 1). The
negative latent slope was significant indicating that participants’ level of difficulty with
daily living/role functioning decreased over time. In other words, participants’
functioning improved over the one-year period. In Model 2, three dummy variables were
added to represent race and ethnicity for African American, Latino, and multiracial
participants. Although Model 2 also fit the data well, the model showed no significant
differences in the rates of change in daily living/role functioning for African American,
Latino or multiracial participants compared with Euro American participants. In Model 3,
age, education, diagnosis, and service intensity were added to the model as control
variables. Again, the model fit well to the data. After controlling for age, education,
diagnosis and service intensity, significant differences were found between African
American and Euro American participants. Compared with Euro Americans, African
Americans had lower initial levels of difficulty with daily living/role functioning as
indicated by a significant negative intercept. African Americans also had a higher rate of
change in improvement in daily living/role functioning over one year compared with
Euro Americans. In other words, difficulty with daily living/role functioning decreased
41
Table 6
Nonlinear Change Model of Overall Functioning
Model 1 Model 2 Model 3
Level 1 β β β
Latent Intercept 1.72*** 1.72*** .23***
Latent Slope -.26*** -.32*** .07
Level 2 Covariates
African American
Intercept
Slope

-.11
.19

-.38***
.34**
Latino
Intercept
Slope

.08
.03

0.04
.07
Multiracial
Intercept
Slope

.07
.11

.04
.16
High Service Intensity
Intercept
Slope

-.04
-.09
Age
Intercept
Slope

.003
.00
Education
Intercept
Slope

-.01
.01
Psychosis
Intercept
Slope

.50***
-.13**
Schizophrenia Spectrum
Intercept
Slope

.18*
-.07
Latent Intercept Residual Variance .48 .48 .23
Latent Slope Residual Variance .10 .10 .07
Error Variance .28 .28 .28
Model Fit
χ
2
/df/p 1/2/.64 5/5/.44 9/10/.52
RMSEA
90% Confidence Interval
.00
.00 - .07
.00
.00 - .06
.00
.00 - .52
R
2
.63*** .63*** .63***
Latent Intercept R
2
.01 .52***
Latent Slope R
2
.05 .29
Note. RMSEA= Root Mean Square Error of Approximation.
a
Comparison group is Euro American.
b
Comparison group is low intensity, case management services.
c
Comparison group is Mood/Anxiety Disorders.
*p < .05, **p < .01, ***p < .001
42
more quickly for African Americans than Euro Americans. No significant differences in
the rate of change in improvement in daily living/role functioning were found for Latino
or multiracial participants compared with Euro American participants.
Social functioning. Table 7 shows the model fit and parameter estimates for
change models of social functioning that were tested. A nonlinear change model of
social functioning fit the data well (Model 1). The negative latent slope was significant
indicating that participants’ level of difficulty with social functioning decreased over time.
In other words, participants’ social functioning improved over the one-year period. In
Model 2, three dummy variables were added to represent race and ethnicity for African
American, Latino, and multiracial participants. Although Model 2 also fit the data well,
the model showed no significant differences in the rates of change in relationship
functioning for African American, Latino, or multiracial participants compared with Euro
American participants. In Model 3, age, education, diagnosis, and service intensity were
added to the model as control variables. Again, the model fit well to the data. After
controlling for age, education, diagnosis, and service intensity, significant differences
were found between African American and Euro American participants. Compared with
Euro Americans, African Americans had lower initial levels of difficulty with social
functioning as indicated by a significant negative intercept. African Americans also had
a higher rate of change in improvement in social functioning over one year compared
with Euro Americans. In other words, difficulty with social functioning decreased for
African Americans more rapidly than for Euro Americans. No significant differences in
the rate of change in improvement in relationship functioning were found for Latino or
multiracial participants compared with Euro American participants.
43

Table 7
Nonlinear Change Model of Social Functioning
Model 1 Model 2 Model 3
Level 1 β β β
Latent Intercept 1.95*** 1.93*** 1.58***
Latent Slope -.30*** -.31*** -.24
Level 2 Covariates
African American
a
Intercept
Slope

-.05
.12

-.28**
.26*
Latino
a
Intercept
Slope

.04
.03

-.06
.13
Multiracial
a
Intercept
Slope

.19
-.10

.21
-.01
High Intensity Services
b
Intercept
Slope

-.07
-.07
Age
Intercept
Slope

-.001
.004
Education
Intercept
Slope

-.008
-.005
Psychosis
Intercept
Slope

.57***
-.16**
Schizophrenia Spectrum
c
Intercept
Slope

-.27**
-.001
Latent Intercept Residual Variance .61 .60 .27
Latent Slope Residual Variance .25 .24 .18
Error Variance .33 .33 .34
Model Fit
χ
2
/df/p 1/2/.48 5/5/.37 16/10/.09
RMSEA
90% Confidence Interval
.00
.00 - .08
.01
.00 - .07
.04
.00 - .07
R
2
.67 .67 .64
Latent Intercept R
2
.009 .55***
Latent Slope R
2
.02 .15
Note. RMSEA= Root Mean Square Error of Approximation.
a
Comparison group is Euro American.
b
Comparison group is low intensity, case management services.
c
Comparison group is Mood/Anxiety Disorders.
*p < .05, **p < .01, ***p < .001
44
Quality of life. Hypothesis three stated that African American, Latino, and
multiracial consumers receiving psychosocial rehabilitation services would have a higher
rate of improvement in quality of life compared with Euro Americans receiving
psychosocial rehabilitation services. Table 8 shows the model fit and parameter
estimates for change models of satisfaction with life that were tested. A linear change
model of satisfaction with life fit the data well (Model 1). The negative latent slope was
significant indicating that participants’ satisfaction with life increased over time. In other
words, participants’ satisfaction with life improved over the one-year period. In Model 2,
three dummy variables were added to represent race and ethnicity for African American,
Latino, and multiracial participants. Although Model 2 also fit the data well, the model
showed no significant differences in the rates of change in satisfaction with life for
African American, Latino or multiracial participants compared with Euro American
participants. In Model 3, age, education, diagnosis, and service intensity were added to
the model as control variables. Again, the model fit well to the data; however no
significance differences in the rate of change in satisfaction with life for African
American, Latino and multiracial participants compared with Euro American participants
were found.
Discussion
This study examined racial/ethnic differences in changes in three main recovery
outcomes (symptoms, functioning and quality of life) for a sample of individuals with
severe mental illness receiving low intensity (case management) or high intensity (FSP)
services over a one-year period. The findings are important to our understanding of the

45

Table 8
Linear Change Model of Satisfaction with Life
Model 1 Model 2 Model 3
Level 1 β β β
Latent Intercept 1.49*** 1.46*** 1.59***
Latent Slope .19*** .22*** .68*
Level 2 Covariates
African American
Intercept
Slope

-.05
-.04

-.03
-.12
Latino
Intercept
Slope

.12
-.06

.12
-.10
Multiracial
Intercept
Slope

.07
-.06

.07
-.10
High Intensity Services
Intercept
Slope

.08
.02
Age
Intercept
Slope

-.01
-.004
Education
Intercept
Slope

.02
-.02
Psychosis
Intercept
Slope

-.19***
-.01
Schizophrenia Spectrum
Intercept
Slope

.36***
-.17*
Latent Intercept Residual Variance .30 .29 .25
Latent Slope Residual Variance .01 .01 0
Error Variance .24 .24 .25
Model Fit
χ
2
/df/p 4/3/.25 6/6/.42 18/13/.16
RMSEA
90% Confidence Interval
.03
.00 - .09
.005
.00 - .06
.03
.00 - .06
R
2
.56*** .62*** .56***
Latent Intercept R
2
.02 .23***
Latent Slope R
2
.31
Note. RMSEA= Root Mean Square Error of Approximation.
a
Comparison group is Euro American.
b
Comparison group is low intensity, case management services.
c
Comparison group is Mood/Anxiety Disorders.
*p < .05, **p < .01, ***p < .001
46
impact of mental health services on recovery outcomes in general and more specifically
for individuals of different racial/ethnic backgrounds.
First, results indicated that initial levels of symptoms, functioning, and
satisfaction with life were better for participants receiving FSP services than for those
receiving case management services. This is contrary to what was expected as FSP
services are designed to be more intense and therefore assist consumers with more severe
impairments in symptoms and functioning than case management services. However,
FSP consumers were recruited largely from a homeless population living on the streets of
Los Angeles. It is possible that to survive living on the streets, consumers must maintain
a certain level of functioning and be less impaired by symptoms, which may account for
higher initial levels of overall functioning and lower initial levels of symptomatology.
Findings also indicated that improvement in symptoms, functioning and
satisfaction with life occurred for participants over a one-year period regardless of
intensity of services received. This finding is particularly relevant to demonstrating that
services assist consumers in moving towards recovery in all three areas. Previous studies
have suggested that the three indicators, while complementary, are not synonymous in
terms of measuring recovery (Roe, Eizenberg, & Lysaker, 2011; Drake et al., 2006). A
recent study found a direct relationship between changes in functioning and changes in
life satisfaction, suggesting that the objective and subjective aspects of recovery are
related (Edmondson, Pahwa, Lee, Brekke, & Hoe, 2012). Findings from this study
suggest that services tap into the key elements needed to improve both objective and
subjective indicators of recovery.
47
While it is important to emphasize that the results indicate improvements for the
entire sample with respect to symptoms, functioning, and quality of life, some key
racial/ethnic differences were found. Although no differential rates of change by race
were found for difficulty with psychotic symptoms, African Americans self-reported
higher initial levels of difficulty with psychotic symptoms. Contrary to the ethnic culture
theory, these findings support the minority status theory and are consistent with
interview-rated reports that African Americans express more severe psychotic symptoms
than other racial/ethnic groups (Barrio et al., 2003a; Chang et al 2011). The minority
status theory suggests that prejudice and discrimination would negatively influence
psychotic symptoms (Kaplan, Moffic, & Adams, 1983; Sue & Chu, 2003). Building on
the minority status theory, African Americans in general may maintain a level of healthy
cultural paranoia due to a history of social oppression and racism (Ridley, 1984). African
Americans with severe mental illness, therefore, may be reporting confluent paranoia, a
combination of healthy cultural paranoia and pathological paranoia suggesting more
severe pathology than is actually present (Whaley, 2002). Another possible explanation
is that the higher initial levels of difficulty with psychotic symptoms may reflect the
overrepresentation of African Americans diagnosed with psychotic disorders in the
sample. Despite higher initial levels of difficulty, African Americans rate of
improvement did not significantly differ from Euro Americans.
African American participants also had the lowest rate of a primary diagnosis of
depression. Furthermore, compared with Euro Americans, African American participants
had lower initial levels of difficulty with and higher rates of improvement in
depression/anxiety. Previous literature suggests that African Americans have lower rates
48
of depression compared with other racial/ethnic groups (Dunlop, Song, Lyons, Manheim,
& Chang, 2003; Kessler et al., 2003; Zhang & Snowden, 1999). For African Americans,
lower rates of depression have been attributed to cultural factors such as family and
community support. This supports the ethnic culture theory, which suggests that cultural
values, norms and traditions serve as protective factors in the mental health of racial and
ethnic minorities (Mirowsky & Ross, 1980; Sue & Chu, 2003). Research has
demonstrated that many African Americans receive social support from immediate family,
extended family, fictive kin (unrelated individuals that are given family titles), and
religious institutions creating large social networks of support (Hatchett, Cochran,
Jackson, 1991; Taylor, Chatters, & Celious, 2003; Taylor, Chatters, Hardison & Riley,
2001; Young, Griffith, & Williams, 2003). Furthermore, African Americans have equal
or better outcomes when receiving evidenced-based treatment such as cognitive-
behavioral and interpersonal therapy or treatment for depression in primary care settings
(Miranda et al., 2005). This is promising given that most individuals seek treatment for
depression in primary care settings (Das et al 2006). This study mimics these findings for
African Americans receiving services in yet another treatment setting.
Similarly, findings indicated that African Americans reported lower levels of
difficulty with and higher rates of improvement in overall and social functioning, which
may also be related to community and social supports. African American families have
been shown to be highly involved and provide social support to an individual family
member with severe mental illness (Guarnacia, 1998; Horwitz & Reinhard, 1995). High
amounts of family contact have also been directly linked to better functioning for African
Americans with severe mental illness (Guada, Hoe, Floyd, Barbour, & Brekke, 2012).
49
While beyond the scope of this study, another possible explanation may involve the
structure and components of services such as cultural competency of providers. A recent
study has found a direct link between cultural competency of providers and consumer
outcomes for African Americans with severe mental illness. Larrison, Schoppelrey,
Hack-Ritzo, and Korr (2011) examined the cultural competency of mental health
providers and found that the providers’ level of positive experiences with racial/ethnic
groups different from his/her own was related to whether or not African American
consumers had worse, the same or better symptom and functional outcomes than white
consumers. Future research in this area may provide an explanation for better recovery
outcomes for this population.
Given the higher rates of improvement in symptoms and functioning for African
Americans, one would expect similar results in regards to satisfaction with life. Although
these three outcomes are not synonymous, they have been found to be complimentary.
Improvement in symptoms and functioning has been associated with higher levels of life
satisfaction (Edmondson et al., 2012; Markowitz, 2001). However, African American,
participants did not show significant differences in rates of improvement in overall
satisfaction with life compared with Euro Americans. It is possible that higher rates of
improvement in specific areas of symptoms and functioning did not directly translate to
higher rates of improvement in overall satisfaction with life. Again, it’s important to
reiterate that there was improvement in overall satisfaction with life for African
Americans; however improvement was not significantly higher than that of Euro
Americans. These results are consistent with one other study that found satisfaction with
50
life to be similar across race/ethnicity for individuals receiving psychosocial
rehabilitation services (Bae et al., 2004).
In addition, no differences were found among Latino and multiracial participants
compared with Euro Americans in satisfaction with life nor any of the other recovery
outcome indicators. This finding suggests that, for Latino participants, services are
improving both objective and subjective domains of recovery outcomes at similar rates as
Euro American participants. This finding may be directly related to culturally competent
practices in the treatment settings in the study. At minimum, clinic sites were able to
provide bilingual services for Spanish-speaking consumers. Previous studies
incorporating culturally relevant practices such as providing services in the consumer’s
primary language have demonstrated an improvement in symptoms and functioning for
individuals with severe mental illness (Patterson, et al., 2005; Kopelowicz, et al., 2003).
In contrast Telles et al. (1995) found that a behavioral family intervention had an
iatrogenic effect on symptoms and functioning for unacculturated Spanish-speaking
participants but not for acculturated participants. The authors suggest that these
unexpected findings could be explained by intervention exercises that were not culturally
congruent, creating an intrusive and stressful experience for the participants (Telles et al.,
1995). Examining the relevance and cultural competency of mental health practices for
Latinos with severe mental illness would be particularly helpful in understanding why
services worked equally as well for Latino participants in this study.
There were no significant differences in psychosis, depression, overall
functioning, social functioning, and satisfaction with life for multiracial participants
compared with Euro Americans. Most literature on multiracial identity asserts that issues
51
related to multiracial identity development have a negative impact on psychological well-
being (Shih & Sanchez, 2005). Shih and Sanchez (2005) found mixed support for this
assertion and revealed that clinical samples of multiracial individuals tended to have
worse psychological well-being and adjustment outcomes than nonclinical samples (Shih
& Sanchez, 2005). However, in general these studies have included adolescent samples
and focused on school performance, problem behaviors, depression, and self-esteem.
Findings that adults with severe mental illness that endorsed more than one racial/ethnic
background showed similar outcomes to Euro Americans is novel and warrants further
investigation. With nine million people reporting more than one racial background in the
U.S. (http://2010.census.gov/news/releases/operations/cb11-cn125.html), more
multiracial individuals with severe mental illness may be presenting for treatment and
how services impact treatment outcomes for this population may become increasingly
important.
There were several limitations to the study. As this was not a randomized trail,
study participants were not randomized by service type/intensity and as a result
significant differences were found on several baseline variables. However, service
type/intensity was controlled for in the main analysis. Additionally, because the study
was not randomized, participants were admitted to case management services or Full
Service Provider (FSP) teams based on clinic admission criteria. Furthermore, within
case management or FSP, participants could have received a variety of services based on
clinic recommendations, consumer needs, and/or consumer preferences. For example,
differences in the specific treatments received by consumers in FSP (such as medication
management, individual or group therapy, housing services, employment services, etc.)
52
could explain some of the racial and ethnic differences found in the study. Although it
was beyond the scope of this study, future studies should examine specific treatment
factors within the context of community based psychosocial rehabilitation that may
contribute to differences in outcomes.
Second, the attrition rate in this study after one year was 30%. The current study
used a follow-along design in the context of community-based services where treatment
attrition is typically high, posing a challenge to maintaining the study sample over time.
In general, dropout rates from studies examining community-based psychosocial
rehabilitation programs range from 20%-50%, placing the current study within the
attrition range typically found in this treatment setting (Kurtz, Rose & Wexler, 2011). In
the present study, attrition analysis indicated no differences between participants who
completed the study and participants who did not on demographic, clinical, service and
outcome variables. Although it is possible that differences exist between completers and
non-completers based on variables not included in the study, most variables typically
associated with attrition bias (age, education, number of inpatient hospitalizations,
symptomatology, diagnosis) were not associated with attrition rates in this study (Kurtz et
al., 2011).
Third, symptoms and functioning were measured using a self-report measure,
which may not be as reliable as interview-rated reports. However, the self-report BASIS-
32 has been shown to be as reliable a measure of symptoms and functioning as the
interview-rated version of this survey (Eisen, 1995).
Fourth, the multiracial category used in this sample was heterogeneous.
Multiracial participants represented several different combinations of individuals
53
endorsing more than one racial category. There could be different racial combinations
within the multiracial category that have unique protective or risk factors related to their
specific cultural backgrounds that affect recovery outcomes. However, the sample sizes
were too small to split up the multiracial category used in the study. Nonetheless,
theories and empirical studies related to multiracial participants, racial identity
development and psychological well-being have been applied broadly to individuals of
multiracial backgrounds (Shih & Sanchez, 2005). With numbers of multiracial
individuals increasing and possibly entering the mental health system, future studies
examining recovery outcomes for this population should oversample multiracial
individuals to be able to analyze specific racial and ethnic combinations.
Finally, due to inadequate subsamples of Asian American, Native American and
Hawaiian/Pacific Islander participants, the differential effects of race for other ethnic
groups were not analyzed. However, the relatively small percentage of these groups in
our study represents the percentage of individuals from these racial/ethnic groups being
served in the clinics. Future studies should oversample participants from these
racial/ethnic groups to examine recovery outcomes for these individuals.
Despite these limitations, the present study is novel in that it is one of the few
studies to examine recovery outcome indicators for racial and ethnic minorities with
severe mental illness. The findings demonstrate that 1) psychosocial rehabilitation
services improve both objective and subjective indicators of recovery and 2) race and
ethnicity differentially affected improvement in recovery outcomes for individuals with
severe mental illness from diverse racial and ethnic backgrounds over a one-year period.
Future studies should examine cultural factors that may influence recovery outcomes as
54
well as those specific treatment factors that contribute to enhancing recovery for racial
and ethnic minorities with severe mental illness.

55
CHAPTER THREE (STUDY 2):
AN EXPLORATORTY ANALYSIS OF PREDICTORS OF SYMPTOM
TRAJECTORIES FOR INDIVIDUALS WITH SEVERE MENTAL
ILLNESS

Introduction
Psychosocial rehabilitation services provide treatment to individuals with severe
mental illness such as schizophrenia and mood disorders. Treatments entail a variety of
services including medication management, case management, skills training, and
psychotherapy targeted at helping consumers recover and lead productive lives. One of
the main goals of psychosocial rehabilitation services is to assist consumers in reducing
and maintaining manageable levels of symptoms of severe mental illness. Some studies
have demonstrated that individuals receiving psychosocial treatments such as Assertive
Community Treatment (ACT) report a reduction in symptoms (Dixon et al., 2010).
Furthermore, research on the long-term course of schizophrenia indicates that there is
significant variation in recovery for individuals receiving treatment. For example,
Harding et al. (1987) found that after a 32-year follow up, one-half to two-thirds of
individuals with schizophrenia showed improvement in symptomatology. A recent
review of long-term studies indicates that the research supports these findings with about
half of individuals demonstrating partial or full recovery in symptoms and functioning
(Silverstein & Bellack, 2008). These studies suggest that there is a substantial number of
individuals that continue to experience symptoms or respond poorly to treatment.
With the development of more sophisticated longitudinal data analysis techniques,
researchers have found that there are groups of individuals that respond differentially to
treatment with respect to their symptoms (Nandi, Beard, & Galea, 2009; Peer, Kupper,
56
Long, Brekke, & Spaulding, 2007). Most studies focusing on the course of symptoms for
individuals with schizophrenia have examined response to antipsychotic treatment.
These studies have found distinct subgroups that respond differentially to treatment
ranging from no response to showing rapid improvement in psychotic symptoms (Case et
al., 2011; Correll et al., 2011; Levine & Rabinowitz, 2010; Peer et al., 2007). Levine and
Rabinowitz (2010) found that diagnosis, premorbid functioning, cognitive functioning,
and age of onset were significant determinants in course of symptoms. The existence of
poor responders to antipsychotic treatment and associated determinants suggests that
there is potential for psychosocial treatments to affect the course of symptoms where
antipsychotic treatment cannot. In fact, cognitive-behavioral therapy has been shown to
decrease severity of symptoms for medication-resistant consumers (Dixon et al., 2010).
Studies of heterogeneity in the course of depressive symptoms also show distinct
groups of nonresponders and responders to pharmacotherapy, psychotherapy, and
combination therapy (Stulz, Thase, Klein, Manber, & Crits-Christoph, 2010; Uher et al.,
2009; Uher et al., 2011). A recent review of epidemiological studies in the general
population also found strong evidence for multiple, distinct trajectories in the course of
depressive symptoms including groups with no depressive symptoms, moderate
symptoms, stable symptoms, increasing symptoms and decreasing symptoms over time
(Nandi et al., 2009). In addition, several factors have been associated with different
trajectories including gender, age, race, socioeconomic status, social support, negative
life events and stress, to name a few (Nandi et al., 2009).
The variation in response to pharmacological treatment for psychotic and
depressive symptoms provides an important base for exploring heterogeneity in treatment
57
response to psychosocial rehabilitation services. While it is generally assumed that
psychosocial rehabilitation services improve symptoms, there may be those individuals
that respond poorly or at different rates to treatment. Identifying individuals in a poor
response group and associated characteristics can assist service providers in tailoring
treatment to meet individual service needs. In contrast, identifying individuals that
respond well to treatment and those predictors of symptom reduction or remission can
help providers enhance those factors that promote symptom recovery.
The purpose of this study is to identify heterogeneity in psychotic and depressive
symptoms for a sample of individuals diagnosed with a severe mental illness receiving
psychosocial rehabilitation services. This study will use Latent Class Growth Mixture
Modeling (GMM) to answer the following research questions:
1. Is there heterogeneity in the trajectory of symptoms for individuals with severe
mental illness receiving psychosocial rehabilitation services?
2. What are the demographic, clinical, and service predictors of symptom trajectories?
GMM is a longitudinal data analysis technique that empirically identifies multiple latent
groups (classes), the shape of the trajectory change for each group, the individuals that
are likely to be in each group, and the features that describe group membership. In this
way, GMM is uniquely suited to determine if there are groups of individuals that
differentially respond to treatment in terms of symptoms and the characteristics (or
profiles) of individuals in the different groups.
Methods
Data used in the present study are from an ongoing research study designed to
examine the impact of mental health system transformation on consumer outcomes,
practice, and organizational culture in public mental health clinics in Los Angeles over a
58
three-year period (Braslow & Brekke, 2006). Participants in the original study were
adults diagnosed with a severe mental illness recruited upon admission from five mental
health clinics providing case management services or Full Service Provider (FSP)
services within the Los Angeles County Department of Mental Health (LACDMH).
Based on an Assertive Community Treatment (ACT) Team Model (Dixon et al., 2010),
FSP teams were staffed by a psychiatrist and other mental health providers such as social
workers and peer providers. Consumer-provider ratios were small, usually no more than
15 to 1. FSP teams were available 24 hours a day, seven days a week and provided a
variety of services in the field or consumer’s home such as mental health treatment, case
management, housing and employment services.

As such FSP teams were designed to
provide a higher intensity of services than usual care. A sampling strategy was used to
match FSP and usual care participants on diagnosis, Global Assessment of Functioning
(GAF) scores, and demographic characteristics as they began receiving services and were
recruited into the study.
Participants
The original sample consisted of 482 ethnically diverse adults diagnosed with a
severe mental illness. Clinic consumers with a V-code diagnosis derived from the
Diagnostic and Statistical Manual of Mental Disorders (DSM-IV_TR 4
th
Edition) and/or
consumers on conservatorship were excluded from the study (Braslow & Brekke, 2006).
For the present study, the sample included participants who had no missing data on items
used as covariates and for which scores on the psychosis subscale (n = 424) and
depression subscale (n =426) of the Behavior and Symptom Checklist (BASIS-32) could
be computed at a minimum of one time point. Asian American, Native American,
59
Hawaiian/Pacific Islander, and transgender participants were excluded from the study due
to the relatively low proportion of participants in these categories.
Participants completed measures on all demographic and outcome variables at
baseline and completed measures on all outcome variables at six and twelve months.
Missing data on demographic variables was obtained from the LACDMCH information
system records when available. Clinical characteristics and diagnosis were also obtained
from LACDMH information system records. The attrition rate at 12 months was 30%. A
series of chi-square and t-tests were conducted on baseline demographic, clinical and
service variables to test for attrition bias at 12 months. No significant differences were
found between participants who completed the survey at 12 months and those who did
not on age, gender, race, marital status, education, primary diagnosis, inpatient
psychiatric hospitalizations or type of services received (p > .05). To deal with missing
outcome data at different time points, data were analyzed using Mplus 6.1, which allows
for the use of all available outcome data to estimate statistical models (Muthén & Muthén,
1998-2010).
Sample characteristics are reported in Table 9. Participants were predominantly
racial or ethnic minorities, unmarried and had at least 12 years of education. Fifty-one
percent of participants were male and approximately 69.5 percent were at least 35 years
of age. Thirty-four percent of participants were diagnosed with a schizophrenia spectrum
disorder. Participants reported low to moderate levels of difficulty with symptoms. All
study procedures were approved by the University of Southern California Institutional
Review Board.
60
Table 9

Sample Baseline Characteristics (N = 426)
Frequency %
Sex Male 217 50.9%

Female 209 49.1%

Age 18-24 19 4.4%

25-34 111 26.1%

35-44 130 30.5%

45-54 138 32.4%

55+ 28 6.6%

Race Euro American 160 37.6%

African American 82 19.2%

Latino 121 28.4%

Multiracial 63 14.8%

Education 0-11 years 135 31.7%

12 years 171 40.1%

13-22 years 120 28.2%

Marital Status Cohabitating/Married 56 13.1%

Not Married 370 86.9%

Diagnosis Schizophrenia Spectrum 146 34.3%

Mood/Anxiety Disorder 280 65.7%

Hospitalizations* None in the past year 303 71.1%

At least one in the past
year 123 28.9%

Service Intensity Case Management 265 62.2%
Full Service Provider 161 37.8%
Mean SD
Depression/Anxiety**
2.12 0.99
Psychosis*** 1.18 0.97
*Number of psychiatric inpatient hospitalizations in the year prior to services
**BASIS-32 Depression/Anxiety Subscale
***BASIS-32 Psychosis Subscale, N = 424

61
Measures
All demographic and self-report outcome data were collected from participants by
trained research staff. Diagnostic, clinical, and service characteristics were obtained from
the LACDMH information system records.
Symptoms. To measure symptoms, participants completed the Behavior and
Symptom Identification Scale (BASIS-32). The BASIS-32 is a brief, mental health
assessment tool developed to measure change in symptoms and functioning for
individuals with severe mental illness across five domains: relations to self/others, daily
living/role functioning, depression/anxiety, impulsive/addictive behavior and psychosis.
It is a self-report measure that assesses the level of difficulty in each domain over the past
week on a 5-point scale ranging from 0 (no difficulty) to 4 (extreme difficulty) (Eisen &
Culhane, 1999). The BASIS-32 has been validated for use with outpatient consumers
(Eisen, Wilcox, Leff, Schaefer, & Culhane, 1999) and with major racial/ethnic groups
(white, African American, Latino American and Asian American) (Chow, Snowden, &
McConnell, 2001). In addition the BASIS-32 has been shown to be sensitive to change
over time (Jerrell, 2005). The psychosis and depression/anxiety subscales were used to
measure symptomatology. For each participant, the score for each subscale was obtained
by averaging the items in the subscale. Cronbach’s alpha for the psychosis and
depression/anxiety subscales were .74 and .85 respectively.
Covariates. Demographic, clinical, and service characteristics were used as
covariates to predict trajectory of symptoms and class membership. All covariates were
dummy coded for the analysis.

62
Demographic Characteristics. All demographic variables are based on self-
report and include age, gender, race and ethnicity, education, and marital status. Age was
represented by five categories: 18 to 24, 25 to 34, 35 to 44, 45 to 54 and over 54 years of
age. Education was represented by three categories: 0-11 years, 12, and 13 to 22 years of
education. Gender was represented by two categories: male and female. Race and
ethnicity were measured using a self-report demographic survey. Participants were given
the option of selecting more than one racial or ethnic category. Participants who were
characterized as Euro American, African American, or Latino were included in the final
sample. Participants who were characterized as Euro American endorsed ethnic
categories of European decent. Participants who endorsed black/African American were
characterized as African American. Participants who were characterized as Latino
predominantly endorsed Mexican or Mexican American categories. Participants who
endorsed more than one racial category were categorized as multiracial and also included
in the final sample. As such, race and ethnicity were represented by four categories: Euro
American, African American, Latino and multiracial. Marital status was represented by
two categories: not married and cohabitating/married.
Service intensity. Service intensity was defined as the number of outpatient visits
a participant had and the number of minutes spent on each visit (Brekke, Ansel, Long,
Slade, & Weinstein, 1999). Results of t-tests indicated that the mean number of
outpatient visits and the mean number of minutes spent on each visit was significantly
higher for participants receiving FSP services than for participants receiving case
management services 12 months after admission (p <.0001). In addition as previously
mentioned, FSP teams followed an ACT team model, which was designed to provide
63
more intense services. As such, service intensity was dichotomized as low intensity (case
management services) and high intensity (FSP services).
Clinical Characteristics. Participants’ DSM-IV_TR primary mental disorder
diagnosis and history of psychiatric hospitalization were obtained from LADMH
information system records. Diagnosis was collapsed into two categories: schizophrenia
spectrum disorders and mood disorders/anxiety disorders. History of psychiatric
hospitalizations was measured by the number of inpatient psychiatric hospitalization in
the year prior to being admitted for services. Psychiatric hospitalization was categorized
into two categories: no hospitalizations and at least one hospitalization in the year prior to
services.
Data Analysis
The current study used Latent Class Growth Mixture Modeling (GMM) to
identify trajectories of psychotic and depression/anxiety symptoms over a one-year
period. GMM is a person-centered analysis that assumes there is heterogeneity within a
population and that individuals will exhibit different patterns of change. Unlike a
conventional growth model which assumes that all individuals are from one population,
GMM assumes that there are unobserved homogeneous subpopulations that exhibit
different growth trajectories. Based on individual patterns, GMM categorizes individuals
such that individuals within a category are more alike than individuals across categories
(Jung & Wickrama, 2008; Múthen & Muthén, 2000). Data were analyzed using GMM
with Mplus 6.1, which allows for the use of all outcome data available to estimate
statistical models (Muthén & Muthén, 1998-2010).
64
In the first step of model testing, a latent growth curve model was identified that
best fit the data for the outcome variable. This is the conventional growth model in
which a latent intercept (initial level) and latent slope (growth trajectory or rate of
change) of the outcome variables are modeled for the full sample. Linear and nonlinear
models were compared and the best fitting model was used in subsequent analyses. Once
a linear or nonlinear model was chosen, a conditional latent model was specified that
included covariates. In this step, analysis entailed running a series of models with
differing numbers of latent classes (trajectories). As the intercept and slope were the
parameters of interest, the within-class variance for these parameters was fixed to zero.
The models were then compared using the criteria described below to determine the
number of classes that best fit the data (class enumeration).
Class Enumeration. Several criteria have been identified in the literature to
determine the number of latent classes that best fit the data in a growth mixture model.
For the current study, the Akaike’s Information Criterion (AIC: Akaike, 1987), the
Bayesian Information Criterion (BIC: Schwartz, 1978), the Vuong-Lo-Mendell-Rubin
Likelihood Ratio Test (VLMR-LRT: Lo, Mendell & Rubin, 2001) and the Lo-Mendell-
Rubin Adjusted Likelihood Ratio Test (LMR Adjusted LRT: Lo, Mendell & Rubin,
2001) were used in determining the number of latent classes that best fit the data. The
AIC and BIC are used to compare nested models that do not have the same number of
classes. When comparing models, a smaller AIC or BIC indicates a better model. The
VLMR-LRT and LMR Adjusted LRT were also used to assess model fit. These statistics
compare neighboring models such that a significant p-value indicates a better model fit
than a model with one less class. Models with an increasing number of classes were
65
tested until the AIC/BIC ceased to decline and/or the p-values for the VLMR-LRT and
LMR Adjusted LRT became nonsignificant.
Additionally, criteria associated with the probability of individual class
membership were used including posterior probabilities and entropy. Posterior
probabilities give the estimated probability of an individual belonging to a class (Muthén
& Muthén, 2000). The average posterior probability for individuals classified into a
given class should be high while the probability of belonging to the other class(es) should
be low (Muthén, 2001). Entropy is “a summary measure of classification” of individuals
in a class with values ranging from zero to one (Muthén et al., 2002). Values closer to
one suggest clear classification of individuals. Finally, interpretability of the model was
considered in choosing the final model.
Interpretation of the Model. Once a class solution was chosen, parameter
estimates for the intercept, slope and covariates were examined within each class. For
each class, a significant parameter estimate for the intercept and slope indicated a
significant initial level and growth trajectory (rate of change), respectively. Next, the
parameter estimates for each covariate within a class were examined for significance.
For example, a significant estimate for the latent slope for gender in class one would
indicate that compared to males, females had a different rate of change.
Class Membership. The covariates were also examined to determine if they
predicted individual class membership. In a multinomial logistic regression, class
membership is characterized by calculating the odds that participants with certain
characteristics (based on categories within the covariates) will belong to one class
compared with another class in the model. In the current study, the covariates race, age,
66
gender, marital status, education, diagnosis, inpatient hospitalizations and service
intensity were used to predict class membership.
Results
In the full sample latent growth curve model, nonlinear models fit the data best
when analyzed for psychotic and depression/anxiety symptoms. As such, nonlinear
models were used as the basis to conduct the growth mixture models.
Psychotic symptoms
Class Enumeration. A conditional latent class growth mixture model was
conducted to determine the number of latent classes that best described the data. Based
on the criteria shown in Table 10, a two-class model was chosen to best describe the data.
The two-class model fit better than one class as indicated by a lower AIC/BIC and
significant p-values for VLMR-LRT and LMR adjusted LRT. Although the AIC
suggests that a three-class model described the data best, this model was not chosen for
two reasons. First, the AIC was the only statistic that suggested a three-class model.
While the AIC has been shown to overestimate the number of classes, the BIC has been
shown to correctly identify the best fitting model with a range of sample sizes compared
with other statistical information criterion (Nylund, Asparouhov, & Muthén, 2007).
Second, the two-class model provided a more clear distinction for interpretability of
classes as is described in the next section. Additionally, the entropy and average
posterior probabilities indicate a fairly high accuracy of classification for the two-class
model.
Two-Class Model. Figure 3 shows the estimated growth curve trajectories for
each class. The two classes were distinguishable by the initial level of psychotic
67
Table 10

Comparison of Models for Psychotic Symptoms

One Class Two Class Three Class
Number of Parameters 32 49 66
AIC 2855.644 2539.848 2485.449
BIC 2985.235 2738.285 2752.731
VLMR-LRT p-value -- 0.00 0.2398
LMR Adjusted LRT p-value -- 0.00 0.2398
Entropy -- 0.863 0.877
Average Posterior Probabilities for
Most Likely Class Membership
-- .942/.969 .912/.945/.962
Class Sample Sizes -- 109/315 63/199/162
Note. AIC =Akaike’s Information Criterion, BIC = Bayesian Information Criterion, VLMR-
LRT = Vuong-Lo–Mendell–Rubin Likelihood Ratio Test, LMR Adjusted LRT = Lo–
Mendell–Rubin Adjusted Likelihood Ratio Test

68

0.00
1.00
2.00
3.00
4.00
Baseline 6 Months 12 Months
BASIS-32 Psychosis Subscale Mean Score
Figure 3. Growth Curve Trajectories for the Two-Class Model of
Psychotic Symptoms
Low Difficulty (N =315)
Moderate Difficulty (N = 109)
69
symptoms with class one having initial moderate difficulty with psychotic symptoms
(labeled “Moderate Difficulty”) and class two having initial low difficulty with psychotic
symptoms (labeled “Low Difficulty”). This distinction is confirmed by a higher intercept
for the “Moderate Difficulty” class. Additionally, the latent intercept was significant for
both classes (Table 11). Although Figure 3 shows that both classes are decreasing in the
level of difficulty with psychotic symptoms over time, the slope was not significant
(Table 11).
Effects of baseline covariates. Results demonstrated that neither gender, age,
race, education, marital status, hospitalizations in the previous year nor service intensity
had a significant effect on the initial level (intercept) or rate of change (slope) of
psychotic symptoms for either class (Table 11). For the “Low Difficulty” and “Moderate
Difficulty” classes, diagnosis had a significant effect on the slope. Compared with
participants with a mood/anxiety disorder diagnosis, participants diagnosed with a
schizophrenia spectrum disorder had a slower rate of change.
Class Membership. Results from the multinomial logistic regression showed that
the odds ratios for the demographic, clinical and service covariates were not statistically
significant (p > .05). This indicates that the covariates in the model did not predict which
participants would be more or less likely to be in the “Low Difficulty” class compared
with the “Moderate Difficulty” class based on the demographic, clinical, and service
characteristics. However, chi-square tests comparing the classes demonstrated significant
differences between the classes with respect to age, race, diagnosis and services received
(Table 12).
70
Table 11

Standardized Estimates for a Two-Class Mixture Model for Psychotic Symptoms

Low Difficulty Moderate Difficulty

(N=315) (N = 109)

b SE b SE
Estimated Growth Factor
Means
Latent Intercept 6.296* 2.555 19.921** 6.512
Latent Slope -0.319 2.17 -0.507 2.36
Covariates
Female
a

Intercept -0.563 0.316 -0.594 0.339
Slope -0.094 0.325 -0.089 0.309
Age 25-34
b

Intercept -0.194 0.673 -0.201 0.692
Slope -0.017 0.656 -0.015 0.612
Age 35-44
b

Intercept 0.127 0.738 0.148 0.863
Slope -0.153 0.798 -0.160 0.829
Age 45-54
b

Intercept 0.140 0.783 0.133 0.754
Slope -0.219 0.826 -0.188 0.700
Age 55+
b

Intercept -0.057 0.497 -0.046 0.403
Slope 0.210 0.516 0.154 0.386
African American
c

Intercept 0.462 0.372 0.596 0.473
Slope 0.393 0.332 0.456 0.399
Latino
c

Intercept 0.297 0.386 0.308 0.392
Slope 0.574 0.348 0.536 0.309
Multiracial
c

Intercept 0.366 0.404 0.342 0.374
Slope 0.175 0.366 0.148 0.312
Education: 12 years
d

Intercept -0.304 0.441 -0.312 0.453
Slope 0.064 0.471 0.059 0.439
Education: 13+ years
d

Intercept 0.112 0.482 0.112 0.481
Slope -0.026 0.416 -0.023 0.376

71
Table 11 (continued)

b SE b SE
Married
e

Intercept -0.043 0.347 -0.045 0.358
Slope -0.107 0.262 -0.100 0.245
Schizophrenia Spectrum
f

Intercept 0.157 0.360 0.177 0.405
Slope -0.637* 0.320 -0.647* 0.317
Hospitalized in Previous Year
g

Intercept -0.533 0.314 -0.541 0.323
Slope -0.243 0.321 -0.222 0.298
Full Service Provider
h

Intercept 0.026 0.31 0.025 0.299
Slope -0.492 0.315 -0.428 0.295
a
Male= Reference Group

b
Age 18-24 = Reference Group

c
Euro American = Reference Group

d
0-11 years = Reference Group

e
Not Married = Reference Group

f
Mood/Anxiety Disorders = Reference Group

g
No hospitalizations in previous year = Reference Group

h
Case Management = Reference Group

*p < .05, ** < .01

72
Table 12

Sample Statistics by Class for Psychotic Symptoms

Low Difficulty
Moderate
Difficulty
(N=315) (N=109)
Gender

Male 52.7% 46.8%
Female 47.3% 53.0%
Age*

18-24 3.8% 6.3%
25-34 27.0% 23.9%
35-44 26.7% 41.3%
45-54 35.2% 23.9%
55+ 7.3% 4.6%
Race*

Euro American 39.3% 33.1%
African American 16.2% 28.4%
Latino 28.9% 26.6%
Multiracial 15.6% 11.9%
Marital Status

Not Married 86.7% 87.2%
Cohabitating/Married 13.3% 12.8%
Education

0-11 years 29.2% 37.6%
12 years 41.6% 36.7%
13-22 years 29.2% 25.7%
Diagnosis*

Schizophrenia Spectrum 31.0% 43.1%
Mood/Anxiety Disorders 68.9% 56.9%
Hospitalizations
a

None in the previous year 70.2% 74.3%
At least one in the previous year 29.8% 25.7%
Services Received*

Case Management 59.0% 70.6%
Full Service Provider 41.0% 29.0%
Mean Mean
Psychosis
b
at Baseline 0.75 2.41
Psychosis
b
at 6 Months 0.61 2.23
Psychosis
b
at 12 Months 0.6 2.17
a
Number of psychiatric inpatient hospitalizations in the year prior to services
b
BASIS-32 Psychosis Subscale
*significant chi-square difference at p < .05

73
Depression/Anxiety symptoms
Class Enumeration. Table 13 displays the criteria for determining the number of
latent classes for a one-, two- and three-class model in the conditional latent growth
mixture model. Similar to the results for psychotic symptoms, the two-class model was
chosen based on a lower BIC and significant p-values for VLMR-LRT and LMR adjusted
LRT. The entropy and average posterior probabilities indicated a fairly high accuracy of
classification for the two-class model.
Two-Class Model. Figure 4 shows the estimated growth curve trajectories for
each class. Class one, “Low Difficulty Improvers,” had a lower initial level of difficulty
with depression/anxiety than class two, “High Difficulty Improvers.” However, the latent
intercept, which represents the initial level of difficulty with depression/anxiety, is only
significant in the “High Difficulty Improvers” class (Table 14). Parameter estimates
indicated that the latent slope for both classes were significant, suggesting that both
classes are improving as difficulty with depression/anxiety decreased over time (Table
14).
Effects of baseline covariates. Results revealed that race, education, and service
intensity had a significant effect on the intercept (initial level) of depression/anxiety in
both classes (Table 14). Compared with Euro Americans, multiracial participants
reported a higher initial level of difficulty with depression/anxiety. Compared with
participants with less than 12 years of education, participants with 12 years of education
reported a higher initial level of difficulty with depression/anxiety. Participants receiving
FSP services reported a lower initial level of difficulty with depression/anxiety, compared
74

0.00
1.00
2.00
3.00
4.00
Baseline 6 Months 12 Months
BASIS-32 Depression/Anxiety Subscale Mean Score
Figure 4. Growth Curve Trajectories for the Two-Class Model of
Depression/Anxiety

Low Difficulty Improvers (N=221)
High Difficulty Improvers (N=225)
75
Table 13

Comparison of Models for Depression/Anxiety

One Class Two Class Three Class
Number of Parameters 32 49 66
AIC 2983.869 2688.793 2638.998
BIC 3113.611 2887.461 2906.591
VLMR-LRT p-value -- 0.072 0.5673
LMR Adjusted LRT p-value -- 0.074 0.5683
Entropy -- 0.806 0.775
Average Posterior Probabilities for
Most Likely Class Membership
-- .948/.938 .851/.930/.920
Class Sample Sizes -- 221/205 148/121/157
Note. AIC =Akaike’s Information Criterion, BIC = Bayesian Information Criterion, VLMR-
LRT = Vuong-Lo–Mendell–Rubin Adjusted Likelihood Ratio Test, LMR Adjusted LRT = Lo–
Mendell–Rubin Adjusted Likelihood Ratio Test

76
Table 14

Standardized Estimates for a Two-Class Mixture Model for Depression/Anxiety

Low Difficulty
Improvers
High Difficulty
Improvers

(N=221) (N = 205)

b SE b SE
Estimated Growth Factor
Means
Latent Intercept 4.294 2.529 10.074* 4.905
Latent Slope -3.815 ** 1.419 -3.448*

1.413
Covariates
Female
a

Intercept 0.176 0.198 0.206 0.213
Slope 0.009 0.266 0.010 0.271
Age 25-34
b

Intercept -0.377 0.472 -0.420 0.552
Slope 0.412 0.507 0.399 0.508
Age 35-44
b

Intercept -0.309 0.528 -0.370 0.666
Slope 0.842 0.487 0.874 0.555
Age 45-54
b

Intercept -0.392 0.437 -0.488 0.556
Slope 0.493 0.510 0.533 0.584
Age 55+
b

Intercept 0.209 0.322 0.202 0.299
Slope 0.260 0.377 0.219 0.325
African American
c

Intercept 0.264 0.344 0.300 0.357
Slope 0.502 0.298 0.494 0.274
Latino
c

Intercept 0.203 0.272 0.242 0.295
Slope 0.251 0.317 0.260 0.315
Multiracial
c

Intercept 0.711** 0.186 0.446** 0.140
Slope 0.450 0.342 0.245 0.182
Education: 12 years
d

Intercept 0.413* 0.169 0.449* 0.184
Slope -0.418 0.328 -0.394 0.313

(continued)

77
Table 14 (continued)

b SE b SE
Education: 13+ years
d

Intercept 0.055 0.227 0.067 0.280
Slope -0.180 0.313 -0.191 0.325
Married
e

Intercept -0.040 0.229 -0.039 0.228
Slope 0.252 0.257 0.213 0.221
Schizophrenia Spectrum
f

Intercept -0.271 0.185 -0.286 0.192
Slope 0.376 0.265 0.344 0.243
Hospitalized in Previous Year
g

Intercept 0.165 0.310 0.172 0.305
Slope -0.319 0.255 -0.289 0.232
Full Service Provider
h

Intercept -0.494** 0.178 -0.596* 0.247
Slope 0.304 0.238 0.318 0.246
a
Male= Reference Group

b
Age 18-24 = Reference Group

c
Euro American = Reference Group

d
0-11 years = Reference Group

e
Not Married = Reference Group

f
Mood/Anxiety Disorders = Reference Group

g
No hospitalizations in previous year = Reference Group

h
Case Management = Reference Group

*p < .05, ** < .01

78
Table 15

Sample Statistics by Class for Depression/Anxiety

Low Difficulty
Improvers
High Difficulty
Improvers
(N=221) (N=205)
Gender

Male 49.8% 52.2%
Female 50.2% 47.8%
Age

18-24 5.5% 3.4%
25-34 28.5% 23.4%
35-44 29.4% 31.7%
45-54 28.5% 36.6%
55+ 8.1% 4.9%
Race***

Euro American 29.4% 46.3%
African American 20.4% 18.0%
Latino 27.1% 29.8%
Multiracial 23.1% 5.9%
Marital Status

Not Married 84.2% 89.8%
Cohabitating/Married 15.8% 10.2%
Education***

0-11 years 24.9% 39.1%
12 years 48.9% 30.7%
13-22 years 26.2% 30.2%
Diagnosis**

Schizophrenia Spectrum 41.2% 26.8%
Mood/Anxiety Disorders 58.8% 73.2%
Hospitalizations

None in the previous year 65.6% 77.1%
At least one in the previous year 34.4% 22.9%
Services Received

Case Management 49.8% 59.5%
Full Service Provider 50.2% 40.5%
Mean Mean
Depression/Anxiety
a
at Baseline 1.49 2.81
Depression/Anxiety
a
at 6 Months 1.14 2.57
Depression/Anxiety
a
at 12 Months 1.16 2.48

a
BASIS-32 Depression/Anxiety Subscale, **p<.01, **p<.001
79
with participants receiving case management services. In terms of the slope (rate of
change), there were no significant effects due to baseline covariates.
Class Membership. A multinomial logistic regression was conducted using the
“Low Difficulty Improvers” class as the reference group. Results revealed that race and
education predicted class membership. Multiracial participants were not likely to be in
the “High Difficulty Improvers” class (OR = .11; CI = .02, .56; p = .008) compared with
the “Low Difficulty Improvers” class. Participants with 12 years of education were also
not likely to be in the “High Difficulty Improvers” class (OR = .32; CI = .14, .70; p
= .005). Furthermore, chi-square tests comparing the classes demonstrated significant
differences between the classes with respect to race, education and diagnosis (Table 15).
Discussion
This study revealed heterogeneity in the trajectory of psychotic and
depression/anxiety symptoms in a sample of individuals diagnosed with a severe mental
illness receiving psychosocial rehabilitation services over a one-year period. Although it
is generally assumed that psychosocial rehabilitation services improve symptoms for
consumers, results from this study indicate that there are differences in response to
treatment depending on the targeted symptoms.
First, results revealed two trajectories of psychotic symptoms that included (1)
consumers that reported low difficulty with psychotic symptoms and maintained this
level of symptoms over one year (“Low Difficulty” class) and (2) consumers that
reported moderate difficulty with psychotic symptoms and maintained this level of
symptoms over one year (“Moderate Difficulty” class). The main factor distinguishing
these two classes was the intercept (initial level of symptoms). The “Low Difficulty” and
80
“Moderate Difficulty” classes did not show any significant differences in growth
trajectories (rate of change) except in regards to diagnosis. In both groups, participants
with a schizophrenia spectrum diagnosis showed a greater rate of change compared to
participants with a mood/anxiety disorder. Although neither class showed significant
differences in terms of the factors predicting class membership in the multinomial logistic
regression, classes significantly differed with respect to age, race, diagnosis, and services
received in the chi-square analysis.
Second, two trajectories of depression/anxiety symptoms were identified that
included (1) consumers that reported low difficulty with depressive/anxious symptoms
but continued to improve over one year (“Low Difficulty Improvers”) and (2) consumers
that reported high difficulty with depressive/anxious symptoms that improved over one
year (“High Difficulty Improvers”). For both classes, race, education, and service
intensity variables had an effect on the initial level of depression/anxiety. Multiracial
participants (compared with Euro Americans) and participants with 12 years of education
(compared with those with less than 12 years of education) had a higher initial level of
difficulty with symptoms whereas participants receiving FSP services (compared with
those receiving case management services) had a lower initial level of difficulty with
symptoms. In addition, race and education predicted class membership such that
multiracial participants and participants with 12 years of education were less likely to be
“High Difficulty Improvers.”
In terms of psychotic symptoms, it is not surprising that two groups characterized
by low and moderate initial levels of symptoms emerged. Consumers with severe mental
illness participating in psychosocial rehabilitation services tend to enter treatment at a
81
point when psychotic symptoms are relatively stable. Although symptoms did not
significantly improve, they also did not worsen over the one-year period. This is
particularly relevant for the “Low Difficulty” participants as they were able to maintain
this level of symptom difficulty with continued services. For the “Moderate Difficulty”
participants it may be that they have plateaued at this level of difficulty with psychotic
symptoms or that more time is needed to see significant change. For both classes,
diagnosis was the only variable that significantly affected the rate of change for both
classes. Given that psychotic symptoms are one of the hallmark features of schizophrenia
spectrum disorders, it follows that participants with this diagnosis had a slower rate of
change than participants with mood/anxiety disorders.
Demographic, clinical and service variables did not predict which participants
would be more or less likely to be in the “Low Difficulty” group compared with the
“Moderate Difficulty” group. Recent studies with inpatient samples examining
heterogeneity in symptom trajectories have found that other variables such as
neurocognitive functioning predict symptom class membership such that higher
neurocognitive functioning predicts membership in the best treatment response groups
(Levine & Rabinowitz, 2010; Levine, Rabinowitz, Case, Ascher-Svanum, 2010). It may
be that the two groups differed on neurocognitive variables; however this information
was not available in the present study. Nonetheless, further post-hoc analysis of the
covariates used in the present study showed that the “Low Difficulty” and “Moderate
Difficulty” groups differed in terms of age, race, diagnosis and services received. Of
particular note is the higher percentage of participants in the “Moderate Difficulty” group
diagnosed with a schizophrenia spectrum disorder compared with the “Low Difficulty”
82
group. Again, the disproportionate representation of participants with a schizophrenia
spectrum disorder in the group exhibiting more difficulty with psychotic symptoms is not
surprising given the diagnostic and symptom criteria associated with schizophrenia
spectrum disorders.
While it appears that participants maintained stable levels of difficulty with
psychotic symptoms, this was not the case with depression/anxiety symptoms. Two
distinct classes were identified that improved over time but differed in initial level of
difficulty with symptoms (“Low Difficulty Improvers” and “High Difficulty Improvers”.)
In both classes, multiracial participants had a higher initial level of difficulty with
depression/anxiety compared with Euro Americans. Paradoxically, multiracial
participants were not likely to be in the “High Difficulty Improvers” class. However,
these two seemingly conflicting results provide important information about multiracial
participants with severe mental illness and their difficulty with depression/anxiety
symptoms. There were fewer multiracial participants who were “High Difficulty
Improvers,” however within this group these participants still had higher initial levels of
depression compared with Euro Americans. Within the “Low Difficulty Improvers” class,
of which multiracial participants compose nearly one-quarter, multiracial participants still
reported higher levels of difficulty with depression/anxiety symptoms compared with
Euro Americans.
While few studies exist that have analyzed multiracial adults with severe mental
illness, research on adolescents in the general population have reported higher levels of
depression/anxiety in multiracial adolescents compared to monoracial adolescents (Chen
& Lively, 2009; Shih & Sanchez, 2005; Udry, Li, & Hendrickson-Smith, 2003).
83
Furthermore, depression and anxiety have been linked to challenges of racial identity and
perceived racial discrimination (Coleman & Carter, 2007; Jackson, Yoo, Guevarra, &
Harrington, 2012; Lusk, Taylor, Nanney, & Austin, 2010; Sanchez, Shih, & Garcia,
2009). In support of a minority status theory, this research suggests that multiracial
individuals may be experiencing negative mental health outcomes due to factors
associated with their status as racial and ethnic minorities, such as prejudice and
discrimination (Kaplan, Moffic, & Adams, 1983; Sue & Chu, 2003). In the present study,
those exhibiting severe difficulty with challenges related to minority status may be
represented in multiracial individuals with severe mental illness such as those in our
sample. In contrast, the ethnic culture theory suggests that multiracial participants may
have values, norms or traditions associated with their specific cultures that service as
protective factors in mental health outcomes (Mirowsky & Ross, 1980; Sue & Chu, 2003).
In fact, resiliency factors related to social relationships and positive psychological
adjustment have been identified in studies with multiracial participants (Bonam & Shih,
2009; Jackson et al., 2012; Lusk et al., 2010; Shih & Sanchez, 2005). These resiliency
factors may account for the higher likelihood of being in the “Low Difficulty” with
depression/anxiety group for multiracial participants in the sample.
Compared with participants with less than 12 years of education, participants with
12 years of education were not likely to be in the “High Difficulty Improvers” class. This
is consistent with studies showing a negative association between education and
depressive symptoms and/or major depression (Adler, et al., 1994; Everson, Maty,
Lynch, Kaplan, 2002; Kessler et al 2003). However, we also found that participants with
12 years of education had a higher initial level of difficulty with symptoms in both
84
classes. One plausible explanation for this relationship is that other factors have modified
the relationship between depression and education for individuals with severe mental
illness. Researchers have suggested that the relationship between education and insight
might increase the risk of depression (Wiffen, Rabinowitz, Lex, & David, 2010). A few
studies have found that education is positively associated with insight into mental illness
for individuals with schizophrenia spectrum disorders (MacPherson, Jerrom, & Hughes,
1996; Wiffen et al. 2010). Furthermore, those with greater insight have been found to
have higher levels of depression and anxiety (Drake et al., 2004; Mintz, Dobson, &
Romney, 2003; Schwartz, 2001; Wiffen et al, 2010). It is plausible that a higher
educational level is related to better insight, which in turn leads to higher levels of
depression. Therefore, insight might be a key mediating factor that accounts for the
relationship between depression and education and partially explain the results in our
study. It is unclear, however, as to why there were no differences in depression/anxiety
found for participants with more than 12 years of education (compared to those with less
than 12 years of education).
Finally, participants receiving FSP services reported a lower initial level of
difficulty with depression/anxiety, compared with participants receiving case
management services. As FSP provides more intense services than case management, we
would expect consumers with more severe symptomatology to receive FSP services.
However, FSP services largely recruited consumers with severe mental illness who were
homeless. It is possible that to survive living on the streets, consumers must maintain a
certain level of functioning and be less impaired by symptoms, which may account for
the lower initial level of difficulty with depression/anxiety in FSP participants.
85
This study has several implications for mental health practice and recovery for
individuals with severe mental illness. First, findings demonstrate that psychosocial
rehabilitation services were able to assist consumers in (1) managing and maintaining
low/moderate levels of difficulty with psychotic symptoms regardless of demographics,
clinical or service-related characteristics and (2) managing and lessening difficulty with
depression/anxiety symptoms over a one year course of treatment. Stabilizing and
managing symptoms is one important factor in facilitating other aspects of recovery such
as improving functioning and overall quality of life (Harvey & Bellack, 2009). Second,
multiracial individuals are one of the fasting growing populations in the U.S. (Humes,
Jones, Ramirez, 2011) and as such, mental health providers may need to address the
connection between depression and issues surrounding racial identity in the treatment of
depression for individuals with severe mental illness. Finally, mental health providers
may need to assess educational level as a risk factor for depression.
This study has several limitations. As the present study was not randomized,
participants were admitted to case management services or Full Service Provider (FSP)
teams based on clinic admission criteria. Furthermore, within case management or FSP,
participants could have received a variety of services based on clinic recommendations,
consumer needs, and/or consumer preferences. For example, differences in the specific
treatments received by consumers in FSP (such as medication management, individual or
group therapy, housing services, employment services, etc.) could explain the
classification of groups of symptom trajectories found in the study. Given that previous
research has shown differential response to anti-psychotic medications (Case et al., 2011;
Correll et al., 2011; Levine & Rabinowitz, 2010; Peer et al., 2007), examining treatment
86
specific factors is particularly important. Although it was beyond the scope of this study,
future studies should examine specific treatment factors within the context of community
based psychosocial rehabilitation that may contribute to differences in outcomes.
Second, the attrition rate in this study after one year was 30%. The current study
used a follow-along design in the context of community-based services where treatment
attrition is typically high, posing a challenge to maintaining the study sample over time.
In general, dropout rates from studies examining community-based psychosocial
rehabilitation programs range from 20%-50%, placing the current study within the
attrition range typically found in this treatment setting (Kurtz, Rose & Wexler, 2011). In
the present study, attrition analysis indicated no differences between participants who
completed the study and participants who did not on demographic, clinical, service and
outcome variables. Although it is possible that differences exist between completers and
non-completers based on variables not included in the study, most variables typically
associated with attrition bias (age, education, number of inpatient hospitalizations,
symptomatology, diagnosis) were not associated with attrition rates in this study (Kurtz et
al., 2011).
Third, this study takes place over the first year that consumers received
psychosocial rehabilitation services. More time may be needed to see additional
improvement in managing psychotic and depression/anxiety symptoms as well as
delineate additional trajectories of symptoms. In addition, although negative symptoms
often persist for individuals with schizophrenia spectrum disorders, the BASIS-32 does
not measure negative symptoms and therefore change in these symptoms could not be
assessed.
87
Lastly, the racial/ethnic categories used in this sample were heterogeneous. For
example, multiracial participants represented several different combinations of
individuals endorsing more than one racial category. There could be different racial
combinations within the multiracial category that have unique protective or risk factors
related to their specific cultural backgrounds that affect recovery outcomes. As such it is
possible that multiracial participants classified in different groups of depression/anxiety
trajectories differed by their racial and ethnic combinations. However, the sample sizes
were too small to split up the main categories used in the study. Nonetheless, theories
and empirical studies related to multiracial participants, racial identity development and
psychological well-being have been applied broadly to individuals of multiracial
backgrounds (Shih & Sanchez, 2005). With numbers of multiracial individuals
increasing and possibly entering the mental health system, future studies examining
recovery outcomes for this population should oversample multiracial individuals to be
able to analyze specific racial and ethnic combinations.
Future studies should also examine the role that racial/ethnic identity plays in
changes in symptomatology for racial/ethnic minorities. In addition, the link between
depression and education through insight should be examined. Finally, future studies
should explore the impact of psychosocial rehabilitation services on other aspects of
recovery such as functioning and quality of life for individuals with severe mental illness.

88
CHAPTER FOUR (STUDY 3):
AN EXPLORATORY ANALYSIS OF PREDICTORS OF FUNCTIONING
AND QUALITY OF LIFE TRAJECTORIES FOR INDIVIDUALS WITH
SEVERE MENTAL ILLNESS

Introduction
Because severe mental illnesses are associated with a range of impairments in
functioning and quality of life (Kessler et al., 2003; Rosa et al., 2010; U.S. Department of
Health & Human Services, 1999), key goals of mental health services include improving
community functioning and quality of life to assist consumers in their recovery from
mental illness. Functioning refers to the ability to take care of one’s self through
independent living skills and work/school roles (role functioning) as well as the ability to
maintain/foster social relationships (social functioning). Definitions of quality of life
found in the literature are more complex. While some definitions include aspects of
functioning, definitions and measures of subjective quality of life are also used with
individuals with severe mental illness to assess quality of life from the consumer’s
perspective (Zissi & Barry, 2006). Furthermore, subjective quality of life is often
assessed using measures of life satisfaction or well-being.
Some studies have demonstrated that individuals with severe mental illnesses
receiving mental health services show improvement in functioning and subjective quality
of life (Aikens, Kroenke, Nease, Klinkman, & Sen, 2008; Brekke, Long, Nesbitt, & Sobel,
1997; Drake, McHugo, Xie, Fox, Packard, & Helmstetter, 2006; Dixon et al., 2010;
Edmondson, Pahwa, Lee, Ho & Brekke, 2012; Mauskopf, Simeon, Miles, Westlund, &
Davidson, 1996; Papakostas, Petersen, Mahal, Mischoulon, Nierenberg, & Fava, 2004;
Rosenheck et al., 1998; Vittengl, Clark, & Jarrett, 2004). However treatment outcome
89
studies suggest that there is substantial variation in functional outcomes for individuals
with severe mental illness receiving mental health services. Studies on individuals with
mood disorders have found that even when individuals have recovered from symptoms,
impairments in functioning remain for a percentage of consumers (Tohen et al., 2000;
Vittengl et al., 2009). Studies on individuals with schizophrenia have found
heterogeneity in functional outcomes in inpatient settings (Silverstein et al., 2006; Liu,
Choi, Reddy, & Spaulding, 2011; Peer & Spaulding, 2007; Peer, Strachan, & Spaulding,
2008). For example, Peer and Spaulding (2007) found two groups of individuals with
schizophrenia spectrum disorders that differed in terms of overall functioning and change
in functioning over time. The high functioning group and low functioning group differed
on clinical factors such as psychiatric treatment history, neurocognition, and history of
negative symptoms. This research suggests that there is a substantial percentage of
individuals with severe mental illnesses that continue to struggle with functioning despite
receiving services.
Research also suggests that some individuals with severe mental illness continue
to struggle with improving subjective quality of life (Drake et al., 2006; Novick, Haro,
Suarez, Vieta, & Naber, 2009; Xie, McHugo, Helmstetter, & Drake, 2005). At 3-year
follow-up, Novick et al. (2009) found that only 27% of individuals with schizophrenia
receiving mental health services reported an “adequate quality of life.” Furthermore
while studies have shown improvements in overall life satisfaction over time
(Edmondson et al., 2012; Lasalvia & Ruggeri, 2006; Papakostas et al. 2004; Rosenheck
et al., 1998), other studies have found no changes (Brekke & Long, 2000; Tempier et al.,
1997).
90
Psychosocial rehabilitation services seek to improve functioning and subjective
quality of life through a multitude of services for individuals with severe mental illness.
However, research on psychosocial rehabilitation service outcomes has often failed to
specifically examine the heterogeneity in response to treatment for individuals with
severe mental illness. In a review of the use of longitudinal analytic methods for
examining treatment effects, Peer, Kupper, Long, Brekke, and Spaulding (2007) highlight
the juxtaposition of the dynamic and fluctuating nature of severe mental illness and the
lack of longitudinal research to examine changes over the course of treatment. As such,
the current study seeks to shift the current research by using sophisticated data analysis
techniques such as Latent Class Growth Mixture Modeling (GMM) to answer more
complex outcome research questions. The use of GMM will facilitate questions such as:
“Is there heterogeneity in response to treatment?” “When do changes occur?” and “For
whom do changes occur?” GMM is a longitudinal data analysis technique that
empirically identifies multiple latent groups (classes), the shape of the trajectory change
for each group, the individuals that are likely to be in each group, and the features that
describe group membership. In this way, GMM is uniquely suited to determine if there
are groups of individuals that differentially respond to treatment in terms of functioning
and quality of life and the characteristics (or profiles) of individuals in the different
groups.
As such, the purpose of this study is to identify heterogeneity in functioning and
quality of life outcomes for a sample of individuals diagnosed with a severe mental
illness receiving psychosocial rehabilitation services. Using GMM, this study will
answer the following research questions:
91
1. Is there heterogeneity in the trajectory of functioning for individuals with severe
mental illness receiving psychosocial rehabilitation services?
2. What are the demographic, clinical, and service predictors of functioning trajectories?
3. Is there heterogeneity in the trajectory of subjective quality of life for individuals with
severe mental illness receiving psychosocial rehabilitation services?
4. What are the demographic, clinical, and service predictors of subjective quality of life
trajectories?
Methods
Data used in the present study are from an ongoing research study designed to
examine the impact of mental health system transformation on consumer outcomes,
practice, and organizational culture in public mental health clinics in Los Angeles over a
three-year period (Braslow & Brekke, 2006). Participants in the original study were
adults diagnosed with a severe mental illness recruited upon admission from five mental
health clinics providing case management services or Full Service Provider (FSP)
services within the Los Angeles County Department of Mental Health (LACDMH).
Based on an Assertive Community Treatment (ACT) Team Model (Dixon et al., 2010),
FSP teams were staffed by a psychiatrist and other mental health providers such as social
workers and peer providers. Consumer-provider ratios were small, usually no more than
15 to 1. FSP teams were available 24 hours a day, seven days a week and provided a
variety of services in the field or consumer’s home such as mental health treatment, case
management, housing and employment services.

As such FSP teams were designed to
provide a higher intensity of services than usual care. A sampling strategy was used to
match FSP and usual care participants on diagnosis, Global Assessment of Functioning

92
(GAF) scores, and demographic characteristics as they began receiving services and were
recruited into the study.
Participants
The original sample consisted of 482 ethnically diverse adults diagnosed with a
severe mental illness. Clinic consumers with a V-code diagnosis derived from the
Diagnostic and Statistical Manual of Mental Disorders (DSM-IV_TR 4
th
Edition) and/or
consumers on conservatorship were excluded from the study (Braslow & Brekke, 2006).
For the present study, the sample included participants who had no missing data on items
used as covariates and for which scores on the daily living/role functioning and relation
to self/others subscales of the Behavior and Symptom Checklist (BASIS-32) and
Satisfaction with Life Scale (SWL) could be computed at a minimum of one time point.
Asian American, Native American, Hawaiian/Pacific Islander, and transgender
participants were excluded from the study due to the relatively low proportion of
participants in these categories.
Participants completed measures on all demographic and outcome variables at
baseline and completed measures on all outcome variables at six and twelve months.
Missing data on demographic variables was obtained from the LACDMCH information
system records when available. Clinical characteristics and diagnosis were also obtained
from LACDMH information system records. The attrition rate at 12 months was 30%. A
series of chi-square tests were conducted on baseline demographic, clinical and service
variables to test for attrition bias at 12 months. No significant differences were found
between participants who completed the survey at 12 months compared with those who
did not on age, gender, race, marital status, education, primary diagnosis, inpatient
93
psychiatric hospitalizations or type of services received (p > .05). To deal with missing
outcome data at different time points, data were analyzed using Mplus 6.1, which allows
for the use of all available outcome data to estimate statistical models (Muthén & Muthén,
1998-2010).
Sample characteristics are reported in Table 16. Participants were predominantly
racial or ethnic minorities, unmarried and had at least 12 years of education. Fifty-one
percent of participants were male and 69.5 percent were at least 35 years of age. Thirty-
four percent of participants were diagnosed with a schizophrenia spectrum disorder.
Participants reported low to moderate levels of difficulty with symptoms. In addition, a
previous study found that participants maintained a low/moderate level of difficulty with
psychotic symptoms and that difficulty with depression/anxiety symptoms decreased over
the same one-year period of the current study (See chapter three). All study procedures
were approved by the University of Southern California Institutional Review Board.
Measures
All demographic and self-report outcome data were collected from participants by
trained research staff. Diagnostic, clinical, and service characteristics were obtained from
the LACDMH information system records.
Functioning. To measure functioning, participants completed the Behavior and
Symptom Identification Scale (BASIS-32). The BASIS-32 is a brief mental health
assessment tool developed to measure change in symptoms and functioning for
individuals with severe mental illness across five domains: relations to self/others, daily

94
Table 16

Sample Baseline Characteristics (N = 426)
Frequency %
Sex Male 217 50.9%

Female 209 49.1%

Age 18-24 19 4.4%

25-34 111 26.1%

35-44 130 30.5%

45-54 138 32.4%

55+ 28 6.6%

Race Euro American 160 37.6%

African American 82 19.2%

Latino 121 28.4%

Multiracial 63 14.8%

Education 0-11 years 135 31.7%

12 years 171 40.1%

13-22 years 120 28.2%

Marital Status Cohabitating/Married 56 13.1%

Not Married 370 86.9%

Diagnosis Schizophrenia Spectrum 146 34.3%

Mood/Anxiety Disorder 280 65.7%

Hospitalizations* None in the previous year 303 71.1%

At least one in the previous
year 123 28.9%

Service Intensity Case Management 265 62.2%
Full Service Provider 161 37.8%
Mean SD
Overall Functioning**
1.72 0.87
Social Functioning***
1.95 0.97
Satisfaction with Life**** 1.50 0.73
*Number of psychiatric inpatient hospitalizations in the year prior to services

**BASIS-32 Daily Living/Role Functioning Subscale
***BASIS-32 Relation to Self/Others Subscales
****Satisfaction with Life Scale

95
living/role functioning, depression/anxiety, impulsive/addictive behavior, and psychosis.
It is a self-report measure that assesses the level of difficulty in each domain over the past
week on a 5-point scale ranging from 0 (no difficulty) to 4 (extreme difficulty) (Eisen &
Culhane, 1999). The BASIS-32 has been validated for use with outpatient consumers
(Eisen, Wilcox, Leff, Schaefer, & Culhane, 1999) and with major U.S. racial/ethnic
groups (white, African American, Latino American and Asian American) (Chow,
Snowden, & McConnell, 2001). In addition, the BASIS-32 has been shown to be
sensitive to change over time (Jerrell, 2005). The daily living/role functioning subscale
was used to measure overall functioning. The relation to self/others subscale was used to
measure social functioning. For each participant, the score for each subscale was
obtained by averaging the items in the subscale. Cronbach’s alpha for the daily
living/role functioning and relation to self/others subscales were .82 and .85, respectively.
Subjective Quality of Life. To measure subjective quality of life, participants
completed the Satisfaction with Life Scale (SWLS). The SWLS is an 18-item, self-report
scale measuring consumer satisfaction with life across four domains: living situation,
social relationships, work, and self/present life. The SWLS is scored on a 5-point scale
with consumers reporting satisfaction ranging from 0 (not at all) to 4 (a great deal). The
SWLS has been validated on diverse, outpatient populations with severe mental illness
and found to be psychometrically sound (Lee, Brekke, Yamada, & Chou, 2010; Test,
Greenberg, Long, Brekke, & Burke, 2005). For each participant, the SWLS score was
obtained by averaging the scale items. Cronbach’s alpha for the SWLS was .91.

96
Covariates. Demographic, clinical, and service characteristics were used as
covariates to predict trajectory of functioning and subjective quality of life as well as
class membership. All covariates were dummy coded for the analysis.
Demographic Characteristics. All demographic variables were based on self-
report and include age, gender, race and ethnicity, education, and marital status. Age was
represented by five categories: 18 to 24, 25 to 34, 35 to 44, 45 to 54 and over 54 years of
age. Education was represented by three categories: 0-11 years, 12, and 13 to 22 years of
education. Gender was represented by two categories: male and female. Race and
ethnicity were measured using a self-report demographic survey. Participants were given
the option of selecting more than one racial/ethnic category. Participants who were
characterized as Euro American, African American, or Latino were included in the final
sample. Participants who were characterized as Euro American endorsed ethnic
categories of European decent. Participants who endorsed black/African American were
characterized as African American. Participants who were characterized as Latino
predominantly endorsed Mexican or Mexican American categories. Participants who
endorsed more than one racial category were categorized as multiracial and also included
in the final sample. As such, race and ethnicity were represented by four categories: Euro
American, African American, Latino and multiracial. Marital status was represented by
two categories: not married and cohabitating/married.
Service intensity. Service intensity was defined as the number of outpatient visits
a participant had and the number of minutes spent on each visit (Brekke, Ansel, Long,
Slade, & Weinstein, 1999). Results of t-tests indicated that the mean number of
outpatient visits and the mean number of minutes spent on each visit was significantly
97
higher for participants receiving FSP services than for participants receiving case
management services 12 months after admission (p < .0001). In addition, as previously
mentioned, FSP teams followed an ACT team model, which was designed to provide
more intense services. As such, service intensity was dichotomized as low intensity (case
management services) and high intensity (FSP services).
Clinical Characteristics. Participants’ DSM-IV_TR primary mental disorder
diagnosis and history of psychiatric hospitalization were obtained from LADMH
information system records. Diagnosis was collapsed into two categories: schizophrenia
spectrum disorders and mood disorders/anxiety disorders. History of psychiatric
hospitalizations was measured by the number of inpatient psychiatric hospitalizations in
the year prior to being admitted for services. Psychiatric hospitalizations were
categorized into two categories: no hospitalizations and at least one hospitalization in the
year prior to services.
Data Analysis
The current study used Latent Class Growth Mixture Modeling (GMM) to
identify trajectories of psychotic and depression/anxiety symptoms over a one-year
period. GMM is a person-centered analysis that assumes there is heterogeneity within a
population and that individuals will exhibit different patterns of change. Unlike a
conventional growth model which assumes that all individuals are from one population, a
growth mixture model assumes that there are unobserved homogeneous subpopulations
that exhibit different growth trajectories. Based on individual patterns, GMM categorizes
individuals such that individuals within a category are more alike than individuals across
categories (Jung & Wickrama, 2008; Múthen & Muthén, 2000). Data were analyzed
98
using GMM with Mplus 6.1, which allows for the use of all outcome data available to
estimate statistical models (Muthén & Muthén, 1998-2010).
In the first step of model testing, a latent growth curve model was identified that
best fit the data for the outcome variable. This is the conventional growth model in
which a latent intercept (initial level) and latent slope (growth trajectory or rate of
change) of the outcome variables are modeled for the full sample. Linear and nonlinear
models were compared and the best fitting model was used in subsequent analyses. Once
a linear or nonlinear model was chosen, a conditional latent model was specified that
included covariates. In this step, analysis entailed running a series of models with
differing numbers of latent classes (trajectories). As the intercept and slope were the
parameters of interest, the within-class variance for these parameters was fixed to zero.
The models are then compared using the criteria described below to determine the
number of classes that best fit the data (class enumeration).
Class Enumeration. Several criteria have been identified in the literature to
determine the number of latent classes that best fit the data in a growth mixture model.
For the current study, the Akaike’s Information Criterion (AIC: Akaike, 1987), the
Bayesian Information Criterion (BIC: Schwartz, 1978), the Vuong-Lo-Mendell-Rubin
Likelihood Ratio Test (VLMR-LRT: Lo, Mendell & Rubin, 2001) and the Lo-Mendell-
Rubin Adjusted Likelihood Ratio Test (LMR Adjusted LRT: Lo, Mendell & Rubin,
2001) were used in determining the number of latent classes that best fit the data. The
AIC and BIC are used to compare nested models that do not have the same number of
classes. When comparing models, a smaller AIC or BIC indicates a better model. The
VLMR-LRT and LMR Adjusted LRT were also used to assess model fit. These statistics
99
compare neighboring models such that a significant p-value indicates a better model fit
than a model with one less class. Models with an increasing number of classes were
tested until the AIC/BIC ceased to decrease and/or the p-values for the VLMR-LRT and
LMR Adjusted LRT became nonsignificant.
Additionally, criteria associated with the probability of individual class
membership were used including posterior probabilities and entropy. Posterior
probabilities give the estimated probability of an individual belonging to a class (Muthén
& Muthén, 2000). The average posterior probability for individuals classified into a
given class should be high while the probability of belonging to the other class(es) should
be low (Muthén, 2001). Entropy is “a summary measure of classification” of individuals
in a class with values ranging from zero to one (Muthén et al., 2002). Values closer to
one suggest clear classification of individuals. Finally, interpretability of the model was
considered in choosing the final model.
Interpretation of the Model. Once a class solution was chosen, parameter
estimates for the intercept, slope and covariates were examined within each class.
Additionally, the characteristics of each class were examined. For each class, a
significant parameter estimate for the intercept and slope indicated a significant initial
level and growth trajectory (rate of change), respectively. Next, the parameter estimates
for each covariate within a class were examined for significance. For example, a
significant estimate for the latent slope for gender in class one would indicate that
compared to males, females had a different growth trajectory or rate of change.
Class Membership. The covariates were also examined to determine if they
predicted individual class membership. In a multinomial logistic regression, class
100
membership is characterized by calculating the odds that participants with certain
characteristics (based on categories within the covariates) will belong to one class
compared with another class in the model. In the current study, the covariates race, age,
gender, marital status, education, diagnosis, inpatient hospitalizations and service
intensity were used to predict class membership.
Results
In the full sample latent growth curve model, a nonlinear model fit the data best
when analyzed for daily/living role functioning and relation to self/others. As such,
nonlinear models were used as the basis to conduct the growth mixture models for the
functioning outcome variables. A linear model fit the data best when analyzed for
satisfaction with life. Therefore, a linear model was used to conduct the growth mixture
models for the satisfaction with life variable.
Social Functioning
Class Enumeration. A conditional latent class growth mixture model was
conducted to determine the number of latent classes that best described the data. Table
17 presents a comparison of one-, two- and three-class models of relationship functioning.
It is unclear from the statistics which model best describes the data. The BIC favors a
two-class model. However, according to the VLMR-LRT and LMR Adjusted LRT p-
values, a one-class model fits the data better than a two-class model. Finally, the AIC
suggests a three-class model; however the AIC tends to overestimate the number of
classes (Nylund et al., 2007). As such, it is unclear if there are latent classes of
participants that differ in social functioning in the data or if the data represents one
homogeneous population.
101
Table 17

Comparison of Models for Social Functioning

One Class Two Class Three Class
Number of Parameters 32 49 66
AIC 2934.109 2680.737 2634.256
BIC 3063.851 2879.404 2901.849
VLMR-LRT p-value -- 0.380 0.222
LMR Adjusted LRT p-value -- 0.383 0.222
Entropy -- 0.839 0.849
Average Posterior Probabilities
for Most Likely Class
Membership
-- .949/.965 .919/.943/.959
Class Sample Sizes -- 209/217 194/79/153
Note. AIC =Akaike’s Information Criterion, BIC = Bayesian Information Criterion, VLMR-
LRT = Vuong-Lo–Mendell–Rubin Adjusted Likelihood Ratio Test, LMR Adjusted LRT =
Lo–Mendell–Rubin Adjusted Likelihood Ratio Test

102
Overall Functioning
Class Enumeration. A conditional latent class growth mixture model was
conducted to determine the number of latent classes that best described the data. Based
on the criteria shown in Table 18, a three-class model was chosen to best describe the
data. The three class model fit the data better than one-, and two-class models as
indicated by a lower AIC and BIC and significant p-values for VLMR-LRT and LMR
adjusted LRT. Although the AIC suggests that a four-class model described the data best,
this model was not chosen for two reasons. First, the AIC was the only statistic that
suggested a four-class model. While the AIC has been shown to over estimate the
number of classes, the BIC has been shown to correctly identifying the best fitting model
with a range of sample sizes compared with other statistical information criterion
(Nylund, Asparouhov, & Muthén, 2007). Second, the three-class model provided a more
clear distinction for interpretability of classes based on different intercepts and slopes as
described in the next section. Additionally, the entropy and average posterior
probabilities indicate a fairly high accuracy of classification for the three-class model.
Three-Class Model. Figure 5 shows the estimated growth curve trajectories for
each class. Class one, the “Low Difficulty Maintainers,” maintained a low level of
difficulty with functioning over the course of one year. Class two, the “Moderate
Difficulty Improvers,” had an initial moderate level of difficulty with functioning and
improved over the one-year. Class three, the “High Difficulty Nonimprovers,” had an
initial high level of difficulty with functioning, which did not significantly improve over

103
Table 18

Comparison of Models for Overall Functioning

One
Class
Two
Class Three Class Four Class
Number of Parameters 32 49 66 83
AIC 2694.891 2438.411 2369.063 2364.456
BIC 2824.633 2637.079 2636.656 2700.974
VLMR-LRT p-value -- 0.002 0.0082 0.3472
LMR Adjusted LRT p-value -- 0.003 0.0086 0.3507
Entropy -- 0.770 0.861 0.848
Average Posterior
Probabilities for Most Likely
Class Membership
-- .934/.936 .927/.945/.965 .881/.931/.985/.939
Class Sample Sizes -- 223/203 195/161/70 185/118/37/86
Note. AIC =Akaike’s Information Criterion, BIC = Bayesian Information Criterion,
VLMR-LRT = Vuong-Lo–Mendell–Rubin Adjusted Likelihood Ratio Test, LMR
Adjusted LRT = Lo–Mendell–Rubin Adjusted Likelihood Ratio Test

104

0.00
1.00
2.00
3.00
4.00
Baseline 6 Months 12 Months
BASIS-32 Overall Functioning
Subscale Mean Score
Figure 5. Growth Curve Trajectories for the Three-Class Model of
Overall Functioning
Low Difficulty Maintainers (N = 161)
Moderate Difficulty Improvers (N = 195)
High Difficulty Nonimprovers (N=70 )
105
the course of one year. Parameter estimates for the latent intercept and slope for each
class are reported in Table 19.
Effects of baseline covariates. Results demonstrated that gender, age, race, and
education had a significant effect on the initial level (intercept) of functioning (Table 19).
Across classes, participants who were African American or multiracial had a lower initial
level of difficulty with functioning compared with Euro Americans. Females also
had a lower initial level of difficulty with functioning compared with males. Participants
25 to 34 and 45 to 54 years old had a high initial level of difficulty with functioning
compared with participants 18 to 24 years old. In addition, participants with 12 years of
education and 13 or more years of education had a higher initial level of difficulty with
functioning compared with participants with less than 12 years of education.
In terms of growth trajectory or rate of change (slope), race and number of
previous psychiatric hospitalizations had a significant effect on functioning. Across the
three classes, African Americans had a higher growth trajectory or rate of change in
functioning compared with Euro Americans. Additionally, participants who had at least
one psychiatric hospitalization in the year prior to starting services had a slower growth
trajectory or rate of change in functioning compared with participants who had no
psychiatric hospitalizations in the previous year.
Class Membership. Table 20 presents the results from the multinomial logistic
regression used to predict class membership with the “Low Difficulty Maintainers” class
used as the comparison class. Demographic covariates revealed several interesting
patterns. With respect to age, participants 25 to 34, 35 to 44, and 45-54 years old were
not likely to be in the “Moderate Difficulty Improvers” class. Participants 45-54 years
106
Table 19

Standardized Estimates for a Three-Class Mixture Model of Overall Functioning
Trajectories

Low Difficulty
Maintainers
Moderate
Difficulty
Improvers
High Difficulty
Nonimprovers

(N = 161) (N = 195 ) (N = 70)

b SE b SE b SE
Estimated Growth
Factor Means
Latent Intercept .531 .542 3.093*** .562 6.569*** 0.782
Latent Slope -2.19 1.203 -2.098* .965 -2.134 1.26
Covariates
Female
a

Intercept -.493*** .098 -.431*** .087 -.427*** .096
Slope .109 .215 .112 .223 .106 .210
Age 25-34
b

Intercept .754*** .187 .573*** .134 .503*** .121
Slope .038 .436 .034 .391 .029 .328
Age 35-44
b

Intercept .136

.172 .112 .138 .144 .179
Slope .513 .418 .502 .400 .613 .497
Age 45-54
b

Intercept .728*** .202 .630*** .160 .502*** .137
Slope .099 .462 .102 .473 .077 .362
Age 55+
b

Intercept .121 .106 .119 .103 .156 .132
Slope -.100 .252 -.116 .294 -.146 .357
African American
c

Intercept -.406*** .102 -.398*** .102 -.505*** .127
Slope .477* .224 .555* .245 .671* .340
Latino
c

Intercept -.001 .134 -.001 .104 -.001 .129
Slope .164 .275 .151 .248 .178 .302
Multiracial
c

Intercept -.229* .107 -.235* .109 -.278* .133
Slope .219 .191 .266 .224 .301 .267

107
Table 19 (continued)

b SE b SE b SE
Education: 12 years
d

Intercept 1.024*** .138 .831*** .082 .668*** .073
Slope -.142 .269 -.137 .259 -0.105 .200
Education: 13+ years
d

Intercept .803*** .157 .670*** .100 .523*** .085
Slope .291 .264 .287 .267 .214 .195
Married
e

Intercept .179 .140 .136 .105 .112 .085
Slope .164 .230 .148 .211 .116 .169
Schizophrenia
Spectrum
f

Intercept .178 .118 .145 .099 .130 .087
Slope -.041 .236 -.040 .227 -.034 .196
Hospitalized
g

Intercept -.051 .113 -.044 .097 -.046 .101
Slope -.575** .210 -.589** .192 -.591** .223
Full Service Provider
h

Intercept .039 .106 .033 .089 .024 .065
Slope -.058 .216 -.058 .215 -.040 .149
a
Male= Reference Group
b
Age 18-24 = Reference Group
c
Euro American = Reference Group
d
0-11 years = Reference Group
e
Not Married = Reference Group
f
Mood/Anxiety Disorders = Reference Group
g
No hospitalizations in previous year = Reference Group
h
Case Management = Reference Group
*p < .05, **p < .01, ***p<.001

108

Table 20
Odds Ratios for Likely Class Membership Compared to "Low Difficulty
Maintainers" class for Overall Functioning

Odds Ratio 95% Confidence Intervals
Female
a

Moderate Difficulty Improvers 1.90 .98, 3.67
High Difficulty Improvers 9.62 2.29, 40.45
Age 25-34
b

Moderate Difficulty Improvers .10** .02, .56
High Difficulty Improvers .11 .01, 1.49
Age 35-44
b

Moderate Difficulty Improvers .17* .03, .93
High Difficulty Nonimprovers .87 .063, 12.15
Age 45-54
b

Moderate Difficulty Improvers .13* .03, .71
High Difficulty Nonimprovers .04* .001, .91
Age 55+
b

Moderate Difficulty Improvers .24 .03, 1.68
High Difficulty Nonimprovers 1.69 .09, 30.85
African American
c

Moderate Difficulty Improvers 1.57 .63, 3.82
High Difficulty Nonimprovers 9.50** 1.92, 46.99
Latino
c

Moderate Difficulty Improvers .83 .39, 1.81
High Difficulty Nonimprovers 2.40 .56, 10.28
Multiracial
c

Moderate Difficulty Improvers 2.11 .75, 5.92
High Difficulty Nonimprovers 9.73* 1.47, 64.59
Education: 12 years
d

Moderate Difficulty Improvers .09*** .03, .22
High Difficulty Nonimprovers .01** .002, .03
Education: 13+ years
d

Moderate Difficulty Improvers .13*** .05, .33
High Difficulty Nonimprovers .01*** .001, .04

109
Table 20 (continued)

Odds Ratio 95% Confidence Intervals
Married
e

Moderate Difficulty Improvers .51 .19, 1.40
High Difficulty Nonimprovers .18* .03, .93
Schizophrenia Spectrum
f

Moderate Difficulty Improvers .49* .25, .96
High Difficulty Nonimprovers .10*** .03, .34
Hospitalized in Previous Year
g

Moderate Difficulty Improvers 1.43 .68, 3.03
High Difficulty Nonimprovers 1.64 .49, 5.51
Full Service Provider
h

Moderate Difficulty Improvers .99 .54, 1.82
High Difficulty Nonimprovers .15** .05, .47
a
Male= Reference Group
b
Age 18-24 = Reference Group
c
Euro American = Reference Group
d
0-11 years = Reference Group
e
Not Married = Reference Group
f
Mood/Anxiety Disorders = Reference Group
g
No hospitalizations in previous year = Reference Group
h
Case Management = Reference Group
*p < .05, **p < .01, ***p<.001

110
old were also not likely to be in the “High Difficulty Nonimprovers” class. Education
showed a similar pattern. Participants with 12 years of education and 13 or more years of
education were not likely to be in the “Moderate Difficulty” or the “High Difficulty
Nonimprovers” classes compared with the “Low Difficulty Maintainers” class.
Participants who were married were not likely to be in the “High Difficulty
Nonimprovers” class. In terms of race, African Americans were likely to be in the “High
Difficulty Nonimprovers” class compared with the “Low Difficulty Maintainers” class.
Diagnosis and service intensity also significantly predicted class membership.
Participants diagnosed with a schizophrenia spectrum disorder were not likely to be in the
“Moderate Difficulty Improvers” or “High Difficulty Nonimprovers” class. Finally,
participants receiving FSP services were not likely to be in the “High Difficulty
Nonimprovers” class. Furthermore, chi-square tests comparing the classes demonstrated
significant differences between the classes with respect to age, race, gender, education,
diagnosis, and service intensity (Table 21).
Subjective Quality of Life
Class Enumeration. A conditional latent growth mixture model analysis was
conducted to determine the number of latent classes that best described the data. Based
on the criteria shown in Table 22, a two-class model was chosen to best describe the data.
The two class model fit the data better than one class as indicated by a lower AIC and
BIC and significant p-values for VLMR-LRT and LMR adjusted LRT. Although the
AIC suggests a three-class model, the AIC has been shown to over estimate the number
of classes (Nylund, Asparouhov, & Muthén, 2007). Additionally, the entropy and
111
Table 21

Sample Statistics by Class for Overall Functioning

Low Difficulty
Maintainers
Moderate
Difficulty
Improvers
High Difficulty
Nonimprovers
(N = 161) (N = 195 ) (N = 70)
Gender***

Male 64.0% 48.7% 27.1%
Female 36.0% 51.3% 72.9%
Age***

18-24 1.2% 7.1% 4.2%
25-34 34.2% 24.1% 12.9%
35-44 26.1% 23.6% 60.0%
45-54 33.5% 38.5% 12.9%
55+ 5.0% 6.7% 10.0%
Race**

Euro American 44.0% 40.0% 15.7%
African American 13.7% 20.0% 30.0%
Latino 31.1% 24.1% 34.3%
Multiracial 11.2% 15.9% 20.0%
Marital Status

Not Married 83.9% 87.2% 92.9%
Cohabitating/Married 16.1% 12.8% 7.1%
Education***

0-11 years 6.8% 36.4% 75.7%
12 years 62.1% 31.8% 12.9%
13-22 years 31.1% 31.8% 11.4%
Diagnosis**

Schizophrenia Spectrum 42.2% 33.3% 18.6%
Mood/Anxiety Disorders 57.8% 66.7% 81.4%
Hospitalizations

None in the previous year 71.4% 69.7% 74.3%
At least one in the previous
year 28.6% 30.3% 25.7%
Services Received***

Case Management 54.0% 59.5% 88.6%
Full Service Provider 46.0% 40.5% 11.4%

112
Table 21 (continued)
Mean Mean Mean
Functioning at Baseline 1.12 1.92 2.54
Functioning at 6 Months .82 1.66 2.44
Functioning at 12 Months .83 1.69 2.38
a
BASIS-32 daily living/role functioning subscale
**significant chi-square differences at p<.01
*** significant chi-square differences at p<.001
113

Table 22

Comparison of Models for Satisfaction with Life

One Class Two Class Three Class
Number of Parameters 31 48 65
AIC 2379.936 2168.356 2108.608
BIC 2505.624 2362.969 2372.147
VLMR-LRT p-value -- 0.025 0.254
LMR Adjusted LRT p-value -- 0.026 0.254
Entropy -- 0.767 0.813
Average Posterior Probabilities
for Most Likely Class
Membership
-- .941/.926 .916/.919/.905
Class Sample Sizes -- 260/166 76/220/130
Note. AIC =Akaike’s Information Criterion, BIC = Bayesian Information Criterion, VLMR-
LRT = Vuong-Lo–Mendell–Rubin Adjusted Likelihood Ratio Test, LMR Adjusted LRT =
Lo–Mendell–Rubin Adjusted Likelihood Ratio Test

114
average posterior probabilities indicate a fairly high accuracy of classification for the
two-class model.
Two-Class Model. Figure 6 shows the estimated growth curve trajectories for
each class. Class one, the “Nonimprovers” showed no change in satisfaction with life
over the course of a year. In contrast, class two, the “Improvers,” showed a slow
improvement in satisfaction with life over a one-year period. Parameter estimates
presented in Table 23 indicate that the slope for the “Improvers” was significant.
Effects of baseline covariates. Results revealed that age, race, and diagnosis had
a significant effect on the intercept for the “Nonimprovers” and the “Improvers.” For
both classes, participants age 25-34, 35-44, and 45-54 had a higher initial level of
satisfaction with life compared with participants age 18-24. Additionally, participants
with a diagnosis of schizophrenia spectrum disorder had a higher initial level of
satisfaction with life compared with participants with a mood/anxiety disorder diagnosis
in both classes. In the “Nonimprover” class, multiracial participants had a higher initial
satisfaction with life compared with Euro Americans. In the “Improvers” class, African
American participants had a lower initial level of satisfaction with life compared with
Euro Americans.
Participants in the “Nonimprovers” class did not differ in their of rate of change
(slope) based on any demographic, clinical or service characteristics. In the “Improvers”
class, participants that were married had a slower rate of change compared with those
who weren’t married.

115
Table 23

Standardized Estimates for a Two-Class Mixture Model for Satisfaction with Life

Nonimprovers Improvers

(N=260) (N =166 )

b SE b SE
Estimated Growth Factor
Means
Latent Intercept 1.673 1.149 4.693** 1.423
Latent Slope 2.965 2.047 4.397* 1.771
Covariates
Female
a

Intercept .076 .145 .073 .136
Slope -.346 .285 -.319 .279
Age 25-34
b

Intercept .949** .293 .873*** .218
Slope -.658 .786 -.587 .692
Age 35-44
b

Intercept .702* .324 .605** .233
Slope -.223 .828 -.186 .681
Age 45-54
b

Intercept .645* .292 .629* .254
Slope -.369 .801 -.349 .743
Age 55+
b

Intercept .180 .172 .169 .152
Slope -.312 .469 -.285 .417
African American
c

Intercept -.349 .188 -.425* .197
Slope -.279 .303 -.330 .355
Latino
c

Intercept .061 .156 .060 .154
Slope -.346 .324 -.333 .323
Multiracial
c

Intercept .380* .166 .215 .112
Slope .146 .366 .081 .205
Education: 12 years
d

Intercept -.230 .183 -.226 .171
Slope -.027 .358 -.026 .342
Education: 13+ years
d

Intercept .096 .168 .087 .157
Slope -.369 .345 -.327 .310

116
Table 23 (continued)

b SE b SE
Married
e

Intercept -.033 .100 -.057 .172
Slope -.390 .219 -.653* .269
Schizophrenia Spectrum
f

Intercept .637*** .104 .601*** .129
Slope -.485 .253 -.444 .257
Hospitalized in Previous Year
g

Intercept .216 .159 .204 .162
Slope .034 .306 .031 .281
Full Service Provider
h

Intercept .236 .169 .219 .143
Slope .051 .311 .046 .281
a
Male= Reference Group

b
Age 18-24 = Reference Group

c
Euro American = Reference Group

d
0-11 years = Reference Group

e
Not Married = Reference Group

f
Mood/Anxiety Disorders = Reference Group

g
No hospitalizations in previous year = Reference Group

h
Case Management = Reference Group

*p < .05, ** < .01

117

0.00
1.00
2.00
3.00
4.00
Baseline 6 Months 12 Months
Satisfaction with LIfe Mean Scores
Figure 6. Growth Trajectories for the Two-Class Model of Satisfaction
with Life
Nonimprovers (N=260)
Improvers (N= 166)
118
Table 24

Sample Statistics by Class for Satisfaction with Life

Nonimprovers Improvers
(N=260) (N =166 )
Gender

Male 51.9% 49.4%
Female 48.1% 50.6%
Age***

18-24 0.8% 16.9%
25-34 26.9% 24.7%
35-44 34.6% 24.1%
45-54 31.2% 34.3%
55+ 6.5% 6.6%
Race***

Euro American 39.2% 35.0%
African American 13.8% 27.7%
Latino 26.2% 31.9%
Multiracial 20.8% 5.4%
Marital Status***

Not Married 95.0% 74.1%
Cohabitating/Married 5.0% 25.9%
Education

0-11 years 33.9% 28.3%
12 years 36.5% 45.8%
13-22 years 29.6% 25.9%
Diagnosis

Schizophrenia Spectrum 35.0% 33.1%
Mood/Anxiety Disorders 65.0% 66.9%
Hospitalizations

None in the previous year 70.8% 71.7%
At least one in the previous year 29.2% 28.3%
Services Received

Case Management 60.8% 64.5%
Full Service Provider 39.2% 35.5%
Mean Mean
SWLS
a
at Baseline 1.17 2.01
SWLS
a
at 6 Months 1.27 2.18
SWLS
a
at 12 Months 1.26 2.29
a
Satisfaction with Life Scale
***p < .001

119
Class Membership. Results from the multinomial logistic regression revealed
that those who were married were likely to be “Improvers” (OR = 7.02; CI = .46, 22.53, p
= .001). Participants between the ages of 25 and 34 (OR = .08; CI = .01, .94; p = .044),
35 and 44 (OR = .06, CI = .01, .58, p = .015) or 45 and 54 year of age (OR = .09; CI
= .01, .98; p = .048) were also not likely to be “Improvers”. Trending results also
indicate that participants 55 years of age or older were not likely to be “Improvers” (OR
= .11, CI = .01, 1.19, p = .069). Furthermore, chi-square tests comparing the classes
demonstrated significant differences between the classes with respect to race, age, and
marital status (Table 24).
Discussion
This study revealed heterogeneity in the trajectory of overall role functioning and
subjective quality of life in a sample of individuals diagnosed with a severe mental illness
receiving psychosocial rehabilitation services over a one-year period. Although it is
generally assumed that psychosocial rehabilitation services improve functioning and
quality of life for consumers, findings indicated that there is variety in response to
treatment depending on the targeted outcome.
Functioning
First, the number of classes for a model of social functioning could not be
empirically determined based on fit indices. However, a previous study with this sample
found improvement in social functioning for the entire sample over the same one-year
period (See chapter two).
Second, results revealed a three-class model of overall functioning that included
(1) consumers who reported minimal difficulty with functioning and maintained this level
120
of functioning (“Low Difficulty Maintainers”) over a one-year period; (2) consumers
who reported moderate difficulty with functioning and improved over the course of a
year (“Moderate Difficulty Improvers”); and (3) consumers who reported a high level of
difficulty with functioning and did not significantly improve (“High Difficulty
Nonimprovers”) over one year. These results are consistent with studies in inpatient
setting that have found significant heterogeneity in changes in functioning (Silverstein et
al., 2006; Peer & Spaulding, 2007; Peer, Strachan, & Spaulding, 2008).
Demographic, clinical and service characteristics such as age, education, race,
marital status, diagnosis and service intensity also had an influence on initial level of
functioning, changes in functioning and class membership. Given these associations,
demographic and clinical characteristics may be useful for identifying a focus for
intervention for participants entering treatment with a goal of improving functioning. For
example, more educated participants were not likely to be in the “Moderate Difficulty
Improvers or “High Difficulty Nonimprovers” classes. This suggests that more educated
participants were likely to be in the class that maintained low levels of difficulty with
functioning over the course of treatment. For participants with a goal of improving
functioning, psychosocial rehabilitation services that target education may also assist
consumers with overall functioning.
We also found that African Americans had an initial lower level of difficulty with
and a higher rate of change in functioning indicating that their difficulty with functioning
decreased at a faster rate compared with Euro Americans over the year. These results
may be explained by an ethnic culture theory (Mirowsky & Ross, 1980). Ethnic culture
theory assumes that mental health outcomes vary based on cultural values, traditions and
121
norms that may serve as protective factors such a strong family, religious, or community
support (Mirowsky & Ross, 1980; Sue & Chu, 2003). Therefore, lower levels of
difficulty and faster rates of change for African Americans may be related to the
importance of social support found in many African American families. More
specifically, studies have shown that African American families provide support and are
highly involved in the lives of individual family members with severe mental illness
(Guarnacia, 1998; Horwitz & Reinhard, 1995). Furthermore, family involvement has
been directly linked to functional outcomes. Guada, Hoe, Floyd, Barbour, and Brekke
(2012) found that high amounts of family contact were associated with better functioning
for African Americans with severe mental illness.
In contrast, participants with at least one hospitalization in the year prior to
receiving services had a lower rate of change in functioning indicating that difficulty with
functioning decreased at a slower rate compared with participants with no hospitalization
in the previous year. Participants with psychiatric hospitalizations may be experiencing
more challenges adjusting to community life and living more independently. These
participants may need additional support when receiving services in the community.
Subjective Quality of Life
Finally, results revealed a two-class model of satisfaction with life that included
(1) consumers who reported no change in satisfaction with life over a one-year period
(“Nonimprovers”) and (2) consumers who reported a slow, but steady improvement in
satisfaction with life over the course of a year (“Improvers”). These findings may help to
reconcile previous findings that have showed either no change or significant change in
satisfaction with life for individuals with severe mental illness (Brekke & Long, 2000;
122
Edmondson et al., in press; Lasalvia & Ruggeri, 2006; Papakostas et al. 2004; Rosenheck
et al., 1998; Tempier et al., 1997). Some researchers have argued that an individual’s
subjective quality of life such as satisfaction with life remain constant due to a
readjustment of expectations or adaptation to life circumstances as goals are achieved
(Drake et al, 2006; Diener, 2005). However recent empirical research has found that
adaptation does not play a role in measurement of subjective quality of life in the
majority of individuals with severe mental illnesses as well individuals in the general
population (Evans, 2007). It is possible that the sample used for this study represents
both individuals that adjust their perceptions of satisfaction as well as those who become
more satisfied as life circumstances improve such as functioning and symptoms. As
previously stated, a subpopulation of the sample used for this study showed improvement
in functioning. Additionally, changes in satisfaction with life have been negatively
associated with depression such that as depressive symptoms decrease, satisfaction with
life increases (Berlim & Fleck, 2007; Katschnig, Krautgartner, Beate, Schrank &
Angermeyer, 2006). A previous study using this sample has shown that depressive
symptoms decreased over the same one-year period (See chapter two).
Similar to functioning, demographic and clinical characteristics such as age, race,
marital status and diagnosis affected initial level of life satisfaction, change in satisfaction
and/or class membership. Participants who were married were likely to be “Improvers.”
Participants in the older age categories (25 year of age and older) were not likely to be
“Improvers” compared with participants in the youngest age category (18-24 years old).
However, participants in the older age categories (25-54 years of age) had a higher initial
level of satisfaction with life compared to the youngest participants.
123
In a study of individuals with schizophrenia, Mares, Young, McGuire, and
Rosenheck (2002) found that being married was associated with better subjective quality
of life. Marriage may provide a source of social support for individuals with severe
mental illness that accounts for a higher subjective quality of life. For participants in the
current sample, having spousal support on their road to recovery may be a key factor in
improving subjective quality of life. While it should also be noted that “Improvers” who
were married had a slower rate of change (compared with those who were not), these
married participants did improve over time.
In regards to age, findings might reveal information about the older adults as well
as the young adults in the sample. The finding that older adults are initially more
satisfied with life is consistent with recent research on individuals with severe mental
illness (Marwaha et al., 2008; Priebe et al., 2010). Adaptation to having a mental illness
is one explanation offered for such findings (Marwaha et al, 2008). Older adults may
have had more time to acquire the skills necessary to cope with a mental illness, improve
various aspects of their lives and improve overall quality of life. On the other hand if
younger adults are comparatively less satisfied, this may be due to the effects of
experiencing their first challenges in coping with mental illness and therefore feeling
more loss and consequently more dissatisfied. However, young adults also may have the
most room for improvement thus explaining why older participants were not likely to be
in the “Improvers” class.
Conclusion
The present study has implications for mental health practice and recovery for
individuals with severe mental illness. Although no group of individuals were identified
124
that declined in functioning and satisfaction with life, we did find subsamples of
participants that remained at their level of functioning/subjective quality of life or that
improved over the past year. Functioning and subjective quality of life are two outcomes
important to improving recovery for individuals with severe mental illness receiving
psychosocial rehabilitation services. Findings from this study demonstrate that individual
and clinical characteristics are related to heterogeneity found in these key areas of
recovery. These findings support the idea that mental health interventions should be
tailored to individual needs and that factors such as age, education, and marital status
may affect these individual needs.
This study has several limitations. As the present study was not randomized,
participants were admitted to case management services or Full Service Provider (FSP)
teams based on clinic admission criteria. Furthermore, within case management or FSP,
participants could have received a variety of services based on clinic recommendations,
consumer needs and/or consumer preferences. For example, differences in the specific
treatments received by consumers in FSP (such as medication management, individual or
group therapy, housing services, employment services, etc.) could explain the group
classifications of functioning or quality of life trajectories found in the study. Although it
was beyond the scope of this study, future studies should examine specific treatment
factors within the context of community based psychosocial rehabilitation that may
contribute to differences in outcomes.
Second, the attrition rate in this study after one year was 30%. The current study
used a follow-along design in the context of community-based services where treatment
attrition is typically high, posing a challenge to maintaining the study sample over time.
125
In general, dropout rates from studies examining community-based psychosocial
rehabilitation programs range from 20%-50%, placing the current study within the
attrition range typically found in this treatment setting (Kurtz, Rose & Wexler, 2011). In
the present study, attrition analysis indicated no differences between participants who
completed the study and participants who did not on demographic, clinical, service and
outcome variables. Although it is possible that differences exist between completers and
non-completers based on variables not included in the study, most variables typically
associated with attrition bias (age, education, number of inpatient hospitalizations,
symptomatology, diagnosis) were not associated with attrition rates in this study (Kurtz et
al., 2011).
Third, the racial/ethnic categories used in this sample were heterogeneous. For
example, multiracial participants represented several different combinations of
individuals endorsing more than one racial category. There could be different racial
combinations within the multiracial category that have unique protective or risk factors
related to their specific cultural backgrounds that affect recovery outcomes. As such it is
possible that a multiracial category consisting of one type of racial combination (i.e.,
African American and Euro American only) could have predicted class membership.
However, the sample sizes were too small to split up the main categories used in the
study. Nonetheless, theories and empirical studies related to multiracial participants,
racial identity development and psychological well-being have been applied broadly to
individuals of multiracial backgrounds (Shih & Sanchez, 2005). With numbers of
multiracial individuals increasing and possibly entering the mental health system, future
126
studies examining recovery outcomes for this population should oversample multiracial
individuals to be able to analyze specific racial and ethnic combinations.
Finally, this study takes place over the first year that consumers received
psychosocial rehabilitation services. More time may be needed to see additional
improvement in functioning or satisfaction with life well as delineate trajectories of social
functioning. Second, a subscale of the BASIS-32 was used to measure overall
functioning. Studying functioning in specific areas such as work and independent living
may have revealed more nuanced results of heterogeneity and factors associated with
functioning. Future studies should examine heterogeneity in specific areas of functioning
and related factors. Additionally, studies should examine other factors that could affect
change in subjective quality of life such as readjustment of life expectations. Finally,
treatment outcome research should assess the impact that co-occurring changes in
symptoms, functioning, and quality of life have on each other when examining recovery
for individuals with severe mental illness.

127
CHAPTER FIVE: CONCLUSION
The purpose of this dissertation was to understand the impact of psychosocial
rehabilitation services on three major recovery outcomes (symptoms, functioning, and
quality of life) for racially and ethnically diverse individuals diagnosed with a severe
mental illness. Three studies were conducted employing either a hypothesis driven,
variable-centered data analysis technique or an exploratory person-centered data analysis
technique. Study 1 used multilevel modeling to examine the rates of change in outcomes
and to test if rates of change differed for racial/ethnic minorities. Because this approach
is variable-centered, the analysis concentrated on describing the relationship between
covariates and outcome variables. Furthermore, it assumed that the sample came from a
homogeneous population that can be represented by a single rate of change. It also
assumed that covariates (such as race and ethnicity) that impact the single rate of change
do so in the same way for every individual (Jung & Wickrama, 2008). In contrast, Study
2 and Study 3 used Growth Mixture Modeling (GMM) to explore the heterogeneity in
symptoms, functioning, and quality of life. GMM is a person-centered approach that
analyzes the relationship among individuals in the sample. GMM classifies individuals
into groups by how similar or different their individual rates of change are in relationship
to each other and uses covariates to predict group membership (Jung & Wickrama, 2008).
In this way, GMM assumed heterogeneity in the population from which the sample was
drawn and allowed for the identification of subpopulations. In the context of the data
analysis techniques used, this chapter will highlight the major findings and the
race/ethnicity findings of the three studies. Practice and research implications as well as
suggestions for future research will also be discussed.
128
Summary of Findings
Findings from the three studies revealed that psychosocial rehabilitation services
such as case management and assertive community treatment-based models of care
maintained and/or improved difficulty with symptoms, functioning and quality of life for
individuals with severe mental illness. Furthermore consumer characteristics such as race
and ethnicity were shown to have an effect on outcomes in various ways. In general,
these characteristics affected initial levels of outcomes and changes in improvement over
a one-year period, as well as predicted heterogeneity in this sample of individuals with
severe mental illness.
Major Findings
In Study 1 findings indicated that consumers improved in all areas of recovery
outcomes (psychotic symptoms, depression/anxiety symptoms, overall functioning, social
functioning, and quality of life) for individuals receiving psychosocial rehabilitation
services. As mentioned earlier, analysis in Study 1 assumed a homogeneous sample of
individuals with severe mental illness. As there is a dearth of treatment outcome research
examining heterogeneity in outcomes (Peer, Kupper, Long, Brekke, & Spaulding, 2007),
variation in recovery outcomes for this sample was explored. Therefore, unlike Study 1,
Study 2 and Study 3 were exploratory in nature and sought to determine if there was
heterogeneity in changes in symptoms, functioning, and quality of life. Studies 2 and 3
were analyzed under the assumption that the sample was not homogeneous but rather
consisted of subpopulations that differed in initial levels of or rates of changes in
outcomes. This assumption held true for four of the five outcomes analyzed (psychotic
129
symptoms, depression/anxiety symptoms, overall functioning and quality of life).
Heterogeneity in social functioning could not be determined.
For psychotic symptoms and depression/anxiety, there was heterogeneity in initial
levels of difficulty with symptoms but not in rates of improvement. Two groups of
participants emerged that differed by initial level of difficulty with psychotic symptoms
(low and moderate). A similar pattern emerged for depression/anxiety (low initial
difficulty and high initial difficulty groups). Regardless of group classification,
participants improved in terms of depression/anxiety. Although the rate of change slope
for psychotic symptoms was in the direction of improvement for both groups, the change
was not significant.
For overall functioning and quality of life outcomes, the sample showed
heterogeneity in terms of initial level and change over the course of the year. With
respect to overall functioning, results revealed three classes of consumers that differed in
initial levels of difficulty (low, moderate and high). Furthermore, only those that
exhibited moderate initial levels of difficulty showed improvement in overall functioning.
Finally, analysis on the quality of life outcome (satisfaction with life) revealed a group of
consumers that showed improvement in overall satisfaction with life and one that did not.
Race/Ethnicity Findings
Study 1 hypothesized that there would be racial/ethnic differences in improvement
in psychotic symptoms, depression/anxiety symptoms, overall functioning, social
functioning, and quality of life for individuals receiving psychosocial rehabilitation
services. Additionally, racial/ethnic differences in initial levels of these outcomes were
examined. Study 2 and Study 3 included race/ethnicity in exploratory analyses as
130
covariates to predict initial levels of or rates of change in recovery outcomes as well as
latent group membership. All race/ethnicity analyses were based on a comparison of a
racial/ethnic minority group (African American, Latino, or multiracial) to a majority
group (Euro American). In general findings indicated that Study 2 and Study 3 replicated
some of the race/ethnicity findings of Study 1 as well as revealed patterns not previously
found.
African American Participants. In Study 1, the hypothesis of racial/ethnic
differences in recovery outcomes was partially supported for African Americans. African
Americans reported lower initial levels and higher rates of improvement in difficulty with
depression/anxiety symptoms, overall functioning, and social functioning. African
Americans also reported higher initial levels of psychotic symptoms but no differences in
rates of change. In examining the same outcomes for African Americans in Study 2 and
Study 3, only racial/ethnic differences for overall functioning were found. Across the
three latent groups of functioning, African Americans had a lower initial level and higher
rates of improvement in difficulty with overall functioning. These results mimicked the
results from Study 1. Furthermore, due to the exploratory nature of the analyses used in
Study 3, some significant patterns were uncovered that were not found in Study 1. For
overall functioning, African Americans were likely to be in the “High Difficulty
Nonimprovers” class for functioning.
Finally, Study 1 revealed no differences in initial levels or rates of improvement in
satisfaction with life for African Americans. However, Study 3 revealed racial/ethnic
differences for African Americans. As mentioned in the major findings, two latent
groups were found that differed in the rate of change in satisfactions with life
131
(“Improvers” and “Nonimprovers”). African Americans in the “Improvers” group had a
lower initial level of satisfaction with life compared with Euro Americans.
Latino Participants. In Study 1, Latino participants showed no differences in
initial levels of or rates of improvement in difficulty with symptoms, functioning or
quality of life outcomes. These findings also held for Study 2 and Study 3. In addition,
being Latino did not predict latent group membership in any of the recovery outcome
analyses in Study 2 or Study 3.
Multiracial Participants. As with Latino participants, multiracial participants
showed no differences in initial levels of or rates of improvement in difficulty with
symptoms, functioning or quality of life outcomes in Study 1. However Study 2 and
Study 3 revealed significant results for multiracial participants in regards to
depression/anxiety symptoms, overall functioning, and satisfaction with life. In terms of
depression/anxiety symptom, multiracial participants had a higher initial level of
difficulty in both latent groups found (“Low Difficulty Improvers” and “High Difficulty
Improvers”). Findings also indicated that multiracial participants were less likely to be in
the “High Difficulty Improvers” group. In terms of functioning, Study 3 found that
multiracial participants had a lower initial level of difficulty with functioning across the
three functioning groups found. Study 3 also found that multiracial participants in the
“Nonimprover” group had a higher initial level of satisfaction with life.
Practice and Research Implications
It is generally assumed that psychosocial rehabilitation services will improve a
variety of outcomes for individuals with severe mental illness. While Study 1 seems to
support this finding, a closer exploration of the variation in response to treatment in Study
132
2 and Study 3 reveal a more complex picture. While significant improvement in
symptoms, functioning and quality of life occurred for a portion of the sample, others
maintained baseline levels of these recovery outcomes. Furthermore, all studies found
significant racial/ethnic differences in recovery outcomes. These findings have several
implications for mental health practice and research.
Implications for Mental Health Practice and Treatment Outcome Research
Given the heterogeneity in initial difficulty with and changes in outcomes, results
suggest that mental health practice and services should be modified to meet individual
service needs. This supports some of the hallmark features of recovery-oriented services
dictating that services should be individualized and person-centered (Bellack, 2006). In
the same vein, treatment outcome research that uses person-centered analyses such as
growth mixture modeling focuses on individual outcomes and gives us a more nuanced
analysis of how consumers respond to treatment. This study is one of few to examine the
three major recovery outcomes over a one-year period. Future studies should examine
heterogeneity in recovery outcomes for at least two years to determine if recovery
outcomes continue to improve, are maintained or worsen over the long-term.
In addition, each recovery outcome was measured using subscales of the BASIS-
32. While the BASIS-32 has been validated for use with racially/ethnically diverse
individuals with severe mental illness, future studies may consider using scales created to
measure each individual outcome such as the Positive and Negative Symptom Scale
(PANSS; Kay, Fiszbein, & Opler, 1987) for psychotic symptoms, the Hamilton
Depression Scale (HAMD; Hamilton, 1960) for depression or the Role Functioning Scale
133
(RFS; McPheeters, 1984) for functioning. Additionally, the PANSS measures negative
symptoms of schizophrenia, which was lacking in the analyses of symptoms.
Implications for Practice and Research with Racial/Ethnic Minorities
Given the racial/ethnic differences were found in initial levels of and changes in
recovery outcomes, mental health services should be more culturally relevant, focusing
on tailoring services to the unique cultural needs of each individual. In this regard,
services should be well suited to promote recovery in racial/ethnic minorities. For
example, mental health service providers working with African American consumers to
improve functioning may want to examine the social/family supports that have aided the
consumer in improving functioning in the past. For multiracial participants, mental
health participants may want to explore protective or risk factors related to racial identity
development that may impact levels of depression/anxiety. Research with racial/ethnic
minorities with severe mental illness should examine cultural factors, such as social
support and racial/ethnic identity, that may influence recovery outcomes as well as those
specific treatment factors that contribute to enhancing recovery for racial/ethnic
minorities with severe mental illness.
Conclusion
The concept of recovery from severe mental illness has become an increasing area
of interest for consumers, mental health providers, and researchers. More recently,
researchers have begun to explore and examine recovery racial/ethnic minorities. This
study contributes to the small but growing body of literature on recovery outcomes for a
racially/ethnically diverse population and has the potential to inform consumers,
134
researchers, and providers about the impact of psychosocial rehabilitation services on
recovery for individuals with severe mental illness.
135
REFERENCES

Adler, N. E., Boyce, T., Chesney, M. A., Cohen, S., Folkman, S., Kahn, R. L., & Syme, S.
L. (1994). Socioeconomic status and health: The challenge of the gradient.
American Psychologist, 49(1), 15-24. doi:10.1037/0003-066X.49.1.15

Aikens, J. E., Kroenke, K., Nease, D. E., Klinkman, M. S., & Sen, A. (2008).
Trajectories of improvement for six depression-related outcomes. General
hospital psychiatry, 30(1), 26−31. doi:10.1016/j.genhosppsych.2007.10.003

Akaike, H. (1987). Factor analysis and AIC. Psychometrika, 52, 317–332.

Andreasen, N. C., Carpenter, W. T., Kane, J. M., Lasser, R. A., Marder, S. R., & Weinber,
D. R. (2005). Remission in schizophrenia: Proposed criteria and rationale for
consensus. American Journal of Psychiatry, 162, 441-449.

Andresen, R., Oades, L., & Caputi, P. (2003). The experience of recovery from
schizophrenia: Towards an empirically validated stage model. Australian and
New Zealand Journal of Psychiatry, 37(5), 586-594. doi:10.1046/j.1440-
1614.2003.01234.x

Anthony, W. A. (1993). Recovery from mental illness: The guiding vision of the mental
health service system in the 1990s. Psychosocial Rehabilitation Journal, 16(4),
11–23.

Armour, M. P., Bradshaw, W., & Roseborough, D. (2009). African Americans and
recovery from severe mental illness. Social Work in Mental Health, 7(6), 602-622.
doi:10.1080/15332980802297507

Bae, S., Brekke, J. S., & Bola, J. R. (2004). Ethnicity and treatment outcome variation in
schizophrenia: A longitudinal study of community-based psychosocial
rehabilitation interventions. Journal of Nervous and Mental Disease, 192(9), 623-
628. doi:10.1097/01.nmd.0000140035.36075.7f

Baker, F. M., Stokes-Thompson, J., Davis, O. A., Gonzo, R., & Hishinuma, E. S. (1999).
Two-year outcomes of psychosocial rehabilitation of black patients with chronic
mental illness. Psychiatric Services, 50(4), 535-539.

Barnes, A. (2008). Race and hospital diagnoses of schizophrenia and mood disorders.
Social Work, 53(1), 77-83. doi: 10.1093/sw/53.1.77

Barrio, C. (2000). The cultural relevance of community support programs. Psychiatric
Services, 51, 879-884.

136
Barrio, C. Yamada, A. M., Atuel, H., Hough, R. L., Yee, S., Berthot, B., Russo, P. A.
(2003a). A tri-ethnic examination of symptom expression on the positive and
negative syndrome scale in schizophrenia spectrum disorders. Schizophrenia
Research, 60, 259– 269. doi:10.1016/S0920-9964(02)00223-2

Barrio, C., Yamada, A. M., Hough, R. L., Hawthorne, W., Garcia, P., & Jeste, D. V.
(2003b). Ethnic disparities in use of public mental health case management
services among patients with schizophrenia. Psychiatric Services, 54(9), 1264-
1270.

Baxter, E. A., & Diehl, S. (1998). Emotional stages: Consumers and family members
recovering from the trauma of mental illness. Psychiatric Rehabilitation Journal,
21(4), 349-355.

Bellack, A. (2006). Scientific and consumer models of recovery in schizophrenia:
Concordance, contrasts, and implications. Schizophrenia Bulletin, 32(3), 432–442,
doi:10.1093/schbul/sbj044

Berlim, M. T., & Fleck, M. P. A. (2007). Quality of life and major depression. In
M. S. Ritsner, & A. G. Awad (Eds.), Quality of life impairment in schizophrenia,
mood and anxiety disorders: New perspectives on research and treatment (pp.
241-242). New York, NY, US: Springer Science + Business Media

Bonam, C. M., & Shih, M. (2009). Exploring multiracial individuals comfort with
intimate interracial relationships. Journal of Social Issues, 65(1), 87-103.
doi:10.1111/j.1540-4560.2008.01589.x

Braslow, J. & Brekke, J. (2006). California MHS Act: Impact on practice and
organizational culture in public clinics. (National Institute of Mental Health: R-
01 MH080671). University of California Los Angeles and University of Southern
California.

Brekke, J. S., Ansel, M., Long, J., Slade, E., & Weinstein, M. (1999). Intensity and
continuity of services and functional outcomes in the rehabilitation of persons
with schizophrenia. Psychiatric Services, 50(2), 248-256.

Brekke, J. S., & Barrio, C. (1997). Cross-ethnic symptom differences in schizophrenia:
The influence of culture and minority status. Schizophrenia Bulletin, 23(2), 305-
316.

Brekke, J. S., Hoe, M., Long, J., & Green, M. F. (2007). How neurocognition and social
cognition influence functional change during community-based psychosocial
rehabilitation for individuals with schizophrenia. Schizophrenia Bulletin, 33(5),
1247-1256. doi:10.1093/schbul/sbl072.

137
Brekke, J. S., & Long, J. D. (2000). Community-based psychosocial rehabilitation and
prospective change in functional, clinical, and subjective experience variables in
schizophrenia. Schizophrenia Bulletin, 26(3), 667-680.

Brekke, J. S., Long, J. D., Nesbitt, N., & Sobel, E. (1997). The impact of service
characteristics on functional outcomes from community support programs for
persons with schizophrenia: A growth curve analysis. Journal of Consulting and
Clinical Psychology, 65(3), 464-475.

Browne, M. W., Cudeck, R., 1993. Alternative ways of assessing model fit. In M.W.
Browne, R. Cudeck, K. A. Bollen, & J. S. Long, (Eds.), Testing structural
equation
models (154-162). Newbury Park, CA: Sage.

Busch, A. B., Lehman, A. F., Goldman, H., & Frank, R. G. (2009). Changes over time
and disparities in schizophrenia treatment quality. Medical Care, 47(2), 199-207.

Case, M., Stauffer, V. L., Ascher-Svanum, H., Conley, R., Kapur, S., Kane, J.M., . . .
Kinon, B. J. (2011). The heterogeneity of antipsychotic response in the treatment
of schizophrenia. Psychological Medicine, 41, 1291-1300.
doi:1O.1017/S0033291710001893

Chang, N., Newman, J., D’Antonio, E., McKelvey, J., & Serper, S. (2011). Ethnicity and
symptom expression in patients with acute schizophrenia. Psychiatry Research,
185, 453-455. doi:10.1016/j.psychres.2010.07.019

Cheng, S., & Lively, K. J. (2009). Multiracial self-identification and adolescent
outcomes: A social psychological approach to the marginal man theory. Social
Forces, 88(1), 61-98.

Chow, J. C., Snowden, L. R., & McConnell, W. (2001). A confirmatory factor analysis of
the BASIS-32 in racial and ethnic samples. The Journal of Behavioral Health
Services & Research, 28(4), 400-411. doi:10.1007/BF02287771

Cokley, K., Hall-Clark, B., & Hicks, D. (2011). Ethnic minority-majority status and
mental health: The mediating role of perceived discrimination.
Journal of Mental Health Counseling. 33(3), 243-263.

Coleman, V. H., & Carter, M. M. (2007). Biracial self-identification: Impact on trait
anxiety, social anxiety, and depression. Identity: An International Journal of
Theory and Research, 7(2), 103-114. doi:10.1080/15283480701326018

Correll, C. U., Caňas, F., Larmo, I., Levy, P., Montes, J.-M., Fagiolini, A., … Zink, A.
(2011). Individualizing antipsychotic treatment selection in schizophrenia:
Characteristics of empirically derived patient subgroups. European psychiatry, 26,
3-16.
138
Das, A. K., Olfson, M., McCurtis, H. L., & Weissman, M. M. (2006). Depression in
African Americans: Breaking barriers to detection and treatment. The Journal of
Family Practice, 55(1), 30-39.

Davidson, L., & Strauss, J. S. (1992). Sense of self in recovery from severe mental illness.
British Journal of Medical Psychology, 65(2), 131-145.

Diener, E. (2000). Subjective well-being: the science of happiness and a proposal for a
national index. American Psychologist, 55, 34–43. doi:10.1037/0003-
066X.55.1.34

Dixon, L. B., Dickerson, F., Bellack, A. S., Bennett, M., Dickinson, D., Goldberg,
R.W., . . . Kreyenbuhl, J. (2010). The 2009 schizophrenia PORT psychosocial
treatment recommendations and summary statements. Schizophrenia Bulletin,
36(1), 48–70. doi:10.1093/schbul/sbp115

Drake, R. E., McHugo, G. J., Xie, H., Fox, M., Packard, J., & Helmstetter, B. (2006).
Ten-year recovery outcomes for clients with co-occurring schizophrenia and
substance use disorders. Schizophrenia Bulletin, 32(3), 464-473.
doi:10.1093/schbul/sbj064

Drake, R. J., Pickles, A., Bentall, R. P., Kinderman, P., Haddock, G., Tarrier, N., &
Lewis, S. W. (2004). The evolution of insight, paranoia and depression during
early schizophrenia. Psychological Medicine: A Journal of Research in
Psychiatry and the Allied Sciences, 34(2), 285-292.

Dunlop, D. D, Song, J., Lyons, J. S., Manheim, L. M., & Chang, R.W. (2003).
Racial/Ethnic differences in rates of depression among preretirement adults.
American Journal of Public Health, 93(11), 1945-1952.

Edmondson, M., Pahwa, R., Lee, K. K., Hoe, M., Brekke, J. S. (2012). A dual change
model of life satisfaction and functioning for individuals with schizophrenia,
Schizophrenia Research, doi:10.1016/j.schres.2012.04.014

Eisen, S.V. (1995). Assessment of subjective distress by patients' self-report versus
structured interview. Psychological Reports, 76, 35-39.

Eisen, S.V., & Culhane, M. A. (1999). Behavior and symptom identification scale
(BASIS-32). In M. E. Maruish (Ed.), The use of psychological testing for
treatment planning and outcomes assessment (2nd ed.) (pp. 759-790). Mahwah,
NJ, US: Lawrence Erlbaum Associates Publishers.

Eisen, S. V., Wilcox, M., Leff, H. S., Schaefer, E., & Culhane, M. A. (1999). Assessing
behavioral health outcomes in outpatient programs: Reliability and validity of the
BASIS-32. The Journal of Behavioral Health Services & Research, 26(1), 5-17.
doi:10.1007/BF02287790
139
Evans, S. (2007). Quality of life in mental health services In M. S. Ritsner, & A. G.
Awad (Eds.), Quality of life impairment in schizophrenia, mood and anxiety
disorders: New perspectives on research and treatment (pp. 333-353). New York,
NY, US: Springer Science + Business Media.

Everson, S. A., Maty, S. C., Lynch, J. W., & Kaplan, G. A. (2002). Epidemiologic
evidence for the relation between socioeconomic status and depression, obesity,
and diabetes. Journal of Psychosomatic Research, 53, 891– 895.

Færden, A., Nesvåg, R., & Marder, S. R. (2008). Definitions of the term 'recovered' in
schizophrenia and other disorders. Psychopathology, 41(5), 271-278,
doi:10.1159/000141921

Ferrer, E., Hamagami, F. & McArdle, J. J. (2004). Modeling latent growth curves with
incomplete data using different types of structural equation modeling and
multilevel software. Structural Equation Modeling, 11(3), 452-483.

Guada, J., Hoe, M., Floyd, R. Barbour, J., & Brekke, J. S. (2012). How family factors
impact psychosocial functioning for African American consumers with
schizophrenia. Community Mental Health Journal, 48, 45–55. doi:
10.1007/s10597-010-9365-6

Guarnaccia, P. J. (1998). Multicultural experiences of family caregiving: A study of
African American, European American, and Hispanic American families. In H. P.
Lefley (Ed.), Families coping with mental illness: The cultural context (pp. 45-61).
San Francisco, CA, US: Jossey-Bass.

Hamilton, M. (1960). A rating scale for depression. Journal of Neurology, Neurosurgery,
and Psychiatry, 23, 56–62.

Harding, C. M., Brooks, G. W., Ashikaga, T., & Strauss, J. S. (1987). The Vermont
longitudinal study of persons with severe mental illness: I. methodology, study
sample, and overall status 32 years later. American Journal of Psychiatry, 144(6),
718-726.

Harrell, S.P. (2000). A multidimensional conceptualization of racism-related stress:
Implications for the well-being of people of color. American Journal of
Orthopsychiatry, 70(1), 42-57. doi: 10.1037/h0087722

Harrow, M., Grossman L., Jobe, T. H., Herbener, E. S. (2005). Do patients with
schizophrenia ever show periods of recovery? A 15-year multi-follow-up study.
Schizophrenia Bulletin, 31, 723–734. doi:10.1093/schbul/sbi026

Hatchett, S. J., Cochran, D. L., & Jackson, J. S. (1991). Family life. In J. S. Jackson (Ed.),
Life in black America (pp. 46-83). Newbury Park, CA: Sage.

140
Harvey, P. D., & Bellack, A. S. (2009). Toward a terminology for functional recovery in
schizophrenia: Is functional remission a viable concept? Schizophrenia Bulletin,
35(2), 300-306. doi:10.1093/schbul/sbn171

Horvitz-Lennon, M., Zhou, D., Normand, S. T., Alegría, M., & Thompson, W. K. (2011).
Racial and ethnic service use disparities among homeless adults with severe
mental illnesses receiving ACT. Psychiatric Services, 62(6), 598-604.

Horwitz, A. V., & Reinhard, S. C. (1995). Ethnic differences in caregiving duties and
burdens among parents and siblings of persons with severe mental illnesses.
Journal of Health and Social Behavior, 36(2), 138-150.

Humes, K. R., Jones, N. A., & Ramirez, R. R. (2011). Overview of race and
Hispanic origin: 2010 (Census 2010 Brief No. 02). Retrieved from
http://www.census.gov/prod/cen2010/briefs/c2010br-02.pdf

Jackson, K. F., Yoo, H. C., Guevarra, R., & Harrington, B. A. (2012). Role of identity
integration on the relationship between perceived racial discrimination and
psychological adjustment of multiracial people. Journal of Counseling Psychology,
59(2), 240-250. doi:10.1037/a0027639

Jerrell, J. M. (2005). Behavior and symptom identification scale 32: Sensitivity to change
over time. The Journal of Behavioral Health Services & Research, 32(3), 341-
346. doi:10.1007/BF02291833

Jerrell, J.M. & Wilson, J.L. (1997). Ethnic differences in the treatment of dual mental and
substance disorders. Journal of Substance Abuse Treatment, 14(2), 133-140.

Jung, T., & Wickrama, K. A .S. (2008). An introduction to latent class growth
analysis and growth mixture modeling. Social and Personality Psychology
Compass, 2(1). 302-317. doi:10.1111/j.1751-9004.2007.00054.x

Kaplan, H. B., Moffic, H. S., & Adams, G. L. (1983). Minority groups
and mental health. American Journal of Social Psychiatry, 3(2), 5-10.

Katschnig, H., Krautgartner, M., Beate, Schrank, B., & Angermeyer, M.C. (2006) In H.
Katschnig, H. Freeman, & N. Sartorius (Eds.), Quality of Life in Mental Disorders
(2nd ed.) (pp. 129-140). New York: John Wiley & Sons Ltd.

Kay, S. R., Fiszbein, A., & Opfer, L. A. (1987). The positive and negative syndrome
scale (PANSS) for schizophrenia. Schizophrenia Bulletin, 13(2), 261-276

Kessler, R. C., Chiu, W. T., Demler, O., & Walters, E. E. (2005). Prevalence, severity,
and comorbidity of twelve-month DSM IV disorders in the National Comorbidity
Survey Replication (NCS-R). Archives of General Psychiatry, 62(6), 617-27.

141
Kessler, R. C., Berglund, P., Demler, O., Jin, R., Koretz, D., Merikangas, K. R., . . .
Wang, P. S. (2003). The epidemiology of major depressive disorder results from
the national comorbidity survey replication (NCS-R). Journal of the American
Medical Association, 289(23), 3095-3105.

Kopelowicz, A., Zarate, R., Smith, V. G., Mintz, J., & Liberman, R. P. (2003). Disease
management in Latinos with schizophrenia: A family-assisted, skills training
approach. Schizophrenia Bulletin, 29(2), 211-228.

Kuno, E., & Rothbard, A. B. (2002). Racial disparities in antipsychotic prescription
patterns for patients with schizophrenia. American Journal of Psychiatry, 159(4),
567-572.

Kupper, Z. & Hoffman, H. (2000). Course patterns of psychosocial functioning in
schizophrenia patients attending a vocational rehabilitation program.
Schizophrenia Bulletin, 26(3), 681-698.

Kurtz, M. M., Rose, J., & Wexler, B. E. (2011). Predictors of participation in community
outpatient psychosocial rehabilitation in schizophrenia. Community Mental Health
Journal, 47(6), 622-7. doi: 10.1007/s10597-010-9343-z

Lagomasino, I. T., Dwight-Johnson, M., Miranda, J., Zhang, L., Liao, D., Duan, N., &
Wells, K. B. (2005). Disparities in depression treatment for Latinos and site of
care. Psychiatric Services, 56(12), 1517-1523.

Larrison, C. R., Schoppelrey, S. L., Hack-Ritzo, S., & Korr, W. S. (2011). Clinician
factors related to outcome differences between black and white patients at
CMHCs. Psychiatric Services, 62(5), 525-531.

Lasalvia, A., & Ruggeri, M. (2006). Quality of life in mental health services research. In
H. Katschnig, H. Freeman, & N. Sartorius. (Eds.), Quality of Life in Mental
Disorders (2nd ed.) (pp. 257-268) New York: John Wiley & Sons Ltd.

Lee, K. K., Brekke, J. S., Yamada, A., & Chou, C. (2010). Longitudinal invariance of the
satisfaction with life scale for individuals with schizophrenia. Research on Social
Work Practice, 20(2), 234-241. doi: 10.1177/1049731509347860

Lehman, A. F., Rachuba, L. T., & Postrado, L. T. (1995). Demographic influences on
quality of life among persons with chronic mental illnesses. Evaluation and
Program Planning, 18(2), 155-164.

Leucht, S. & Lasser, R. (2006). The concepts of remission and recovery in schizophrenia.
Pharmacopsychiatry, 39(5), 161-170. doi: 10.1055/s-2006-949513.

142
Levine, S. Z., & Rabinowitz, J. (2010). Trajectories and antecedents of treatment
response over time in early-episode psychosis. Schizophrenia Bulletin, 36(3),
624–632. doi:10.1093/schbul/sbn120

Levine, S. Z., Rabinowitz, J., Case, M., & Ascher-Svanum, H. (2010). Treatment
response trajectories and their antecedents in recent-onset psychosis: A 2-year
prospective study. Journal of Clinical Psychopharmacology, 30(4), 446-449.

Liberman, P. L. (2008). Recovery from disability: Manual of psychiatric rehabilitation.
Arlington, VA: American Psychiatric Publishing Inc.

Liberman, R. P., & Kopelowicz, A. (2002). Recovery from schizophrenia: A challenge
for the 21st century. International Review of Psychiatry, 14(4), 245-255. doi:
10.1080/0954026021000016897

Liberman, R. P., & Kopelowicz, A. (2005). Recovery from schizophrenia: A concept in
search of research. Psychiatric Services, 56(6), 735-742.

Liberman, R. P., Kopelowicz, A., Ventura, J., & Gutkind, D. (2002). Operational criteria
and factors related to recovery from schizophrenia. International Review of
Psychiatry, 14, 256-272.

Liu, N.H., Choi, K., Reddy, F. & Spaulding, W.D. (2011). Heterogeneity and the
longitudinal recovery of functioning during inpatient psychiatric rehabilitation for
treatment-refractory severe mental illness. American Journal of Psychiatric
Rehabilitation, 14(1), 55-75. doi: 10.1080/15487768.2011.546293.

Lo, Y., Mendell, N., & Rubin, D. (2001). Testing the number of components in a normal
mixture. Biometrika, 88, 767-778.

Lusk, E. M., Taylor, M. J., Nanney, J. T., & Austin, C. C. (2010). Biracial identity and
its relation to self-esteem and depression in mixed black/white biracial individuals.
Journal of Ethnic & Cultural Diversity in Social Work, 19(2), 109.

Lysaker, P. H., Roe, D., & Buck, K. D. (2010). Recovery and wellness amidst
schizophrenia: Definitions, evidence, and the implications for clinical practice.
Journal of the American Psychiatric Nurses Association, 16(6), 36-42. doi:
10.1177/1078390309353943

MacPherson, R., Jerrom, B., & Hughes, A. (1996). Relationship between insight,
educational background and cognition in schizophrenia. British Journal of
Psychiatry, 168(6),, 718-722. doi:10.1192/bjp.168.6.718

Mallinger, J. B., Fisher, S. G., Brown, T., & Lamberti, J. S. (2006). Racial disparities in
the use of second-generation antipsychotics for the treatment of schizophrenia.
Psychiatric Services, 57, 133-136.
143

Mares, A. S., Young, A. S., McGuire, J. F., & Rosenheck, R. A. (2002). Residential
environment and quality of life among seriously mentally ill residents of board
and care homes. Community Mental Health Journal, 38(6), 447-58.

Markowitz, F. E. (2001). Modeling processes in recovery from mental illness:
Relationships between symptoms, life satisfaction, and self-concept. Journal of
Health and Social Behavior, 42(1), 64-79.

Marwaha, S., Johnson, S., Bebbington, P., Angermeyer, M. C., Brugha, T., Azorin, J., . . .
Toumi, M. (2008). Correlates of subjective quality of life in people with
schizophrenia: Findings from the EuroSC study. Journal of Nervous and Mental
Disease, 196(2), 87-94. doi:10.1097/NMD.0b013e318162aa9c

McPheeters, H.L. (1984) Statewide mental health outcome evaluation: A perspective of
two southern states. Community Mental Health Journal, 20, 44-55.

Mintz, A. R., Dobson, K. S., & Romney, D. M. (2003). Insight in schizophrenia: A meta-
analysis. Schizophrenia Research, 61(1), 75-88. doi:10.1016/S0920-
9964(02)00316-X

Miranda, J., Bernal, G., Lau, A., Kohn, L., Hwang, W., & LaFromboise, T. (2005). State
of the science on psychosocial interventions for ethnic minorities. Annual Review
of Clinical Psychology, 11, 113–142.

Mirowsky, J., & Ross, C. E. (1980). Minority status, ethnic culture, and distress: A
comparison of blacks, whites, Mexicans, and Mexican Americans. American
Journal of Sociology, 86(3), 479-495.

Muthén, B. (2001). Latent variable mixture modeling. In G. A. Marcoulides, & R. E.
Schumacker (Eds.), New Developments and Techniques in Structural Equation
Modeling (pp. 1-33). Lawrence Erlbaum Associates.

Muthén, B., Brown, C. H., Booil Jo, K. M., Khoo, C.-C., Wang, C.-P., … Liao, J. (2002).
General growth mixture modeling for randomized preventive interventions.
Biostatistics, 3(4), 459–475.

Muthén, L. K., & Muthén, B. O. (1998-2010). Mplus User’s guide (Sixth Edition). Los
Angeles, CA: Muthén & Muthén.

Muthén, B. O., & Muthén, L. K. (2000). Integrating person-centered and variable-
centered analyses: growth mixture modeling with latent trajectory classes.
Alcoholism: Clinical and Experimental Research, 24(6), 882-891.

144
Nandi, A., Beard, J. R., & Galea, S. (2009). Epidemiologic heterogeneity of common
mood and anxiety disorders over the life course in the general population: A
systematic review. BMC Psychiatry, 9. doi:10.1186/1471-244X-9-31

National Institute of Mental Health. (2008). The numbers count: mental disorders in
America. Retrieved from NIMH website:
www.nimh.nih.gov/health/publications/the-numbers-count-mental-disorders-in-
america/index.shtml#Intro

Novick, D., Haro, J. M., Suarez, D., Vieta, E., & Naber, D. (2009). Recovery in the
outpatient setting: 36-month results from the schizophrenia outpatients health
outcomes (SOHO) study. Schizophrenia Research, 108(1-3), 223-230.
doi:10.1016/j.schres.2008.11.007

Nylund, K. L., Asparouhov, T., & Muthén, B. (2007). Deciding on the number of classes
in latent class analysis and growth mixture modeling: A Monte Carlo simulation
study. Structural Equation Modeling: A Multidisciplinary Journal, 14, 535–569.

Okazaki, S. (2009) Impact of racism on ethnic minority mental health.
Perspectives on Psychological Science 4 (1), 103-107. doi: 10.1111/j.1745-
6924.2009.01099.x

Papakostas, G. I., Petersen, T., Mahal, Y., Mischoulon, D., Nierenberg, A. A., & Fava, M.
(2004). Quality of life assessments in major depressive disorder: A review of the
literature. General Hospital Psychiatry, 26(1), 13-17.
doi:10.1016/j.genhosppsych.2003.07.004

Patterson, T. L., Bucardo, J., McKibbin, C. L., Mausbach, B. T., Moore, D., Barrio, C.,
Goldman, S. R., & Jeste, D. V. (2005). Development and pilot testing of a new
psychosocial intervention for older Latinos with chronic psychosis. Schizophrenia
Bulletin, 31(4), 922–930. doi:10.1093/schbul/sbi036

Peer, J. E., Kupper, Z., Long, J. D., Brekke, J. S., & Spaulding, W. D. (2007). Identifying
mechanisms of treatment effects and recovery in rehabilitation of schizophrenia:
Longitudinal analytic methods. Clinical Psychology Review, 27, 696-714.
doi:10.1016/j.cpr.2007.01.004

Peer, J. E. & Spaulding, W. D. (2007). Heterogeneity in recovery of psychosocial
functioning during psychiatric rehabilitation: An exploratory study using latent
growth mixture modeling. Schizophrenia Research, 93, 186-193.
doi:10.1016/j.schres.2007.03.002

Peer, J., Strachan, E., & Spaulding, W. S. (2008). Heterogeneity in behavioral treatment
response in severe mental illness. Journal of Nervous of Mental Disease, 196(3),
198-206. doi: 10.1097/NMD.0b013e318165c7d2

145
Phillips, E. S., Barrio, C., & Brekke, J. S. (2001). The impact of ethnicity on prospective
functional outcomes from community-based psychosocial rehabilitation for
persons with schizophrenia. Journal of Community Psychology, 29(6), 657-673.
doi:10.1002/jcop.1041

Priebe, S., Reininghaus, U., McCabe, R., Burns, T., Eklund, M., Hansson, L., . . . Wang,
D. (2010). Factors influencing subjective quality of life in patients with
schizophrenia and other mental disorders: A pooled analysis. Schizophrenia
Research, 121(1-3), 251-258. doi:10.1016/j.schres.2009.12.020

Prince, J. D. (2006). Ethnicity and life quality of recently discharged inpatients with
schizophrenia. American Journal of Orthopsychiatry, 76(2), 202-205. doi:
10.1037/0002-9432.76.2.202

Resnick, S. G., Rosenheck, R. A., & Lehman, A. F. (2004). An exploratory analysis of
correlates of recovery. Psychiatric Services, 55(5), 540-547.
doi:10.1176/appi.ps.55.5.540

Ridley, C. R. (1984). Clinical treatment of the nondisclosing black client: A therapeutic
paradox. American Psychologist, 39(11), 1234-1244. doi: 10.1037/0003-
066X.39.11.1234

Roe, D., Mashiach-Eizenberg, M., & Lysaker, P. H. (2011). The relation between
objective and subjective domains of recovery among persons with schizophrenia-
related disorders. Schizophrenia Research, 131, 133–138.
doi:10.1016/j.schres.2011.05.023

Rosa, A. R., Reinares, M., Michalak, E. E., Bonnin, C. M., Sole, B., Franco, C., . . . Vieta,
E. (2010). Functional impairment and disability across mood states in bipolar
disorder. Value in Health, 13(8), 984-988. doi:10.1111/j.1524-
4733.2010.00768.x

Rosenheck, R., Tekell, J., Peters, J., Cramer, J., Foutana, A., Xu, W., . . . Charney, D.
(1998). Does participation in psychosocial treatment augment the benefit of
clozapine? Archives of General Psychiatry, 55, 618 – 625.

Sanchez, D., T., Shih, M., & Garcia, J. A. (2009). Juggling multiple racial identities:
Malleable racial identification and psychological well-being. Cultural Diversity &
Ethnic Minority Psychology, 15(3), 243-254. doi:10.1037/a0014373

Schwartz, G. (1978). Estimating the dimension of a model. The Annals of Statistics, 6,
461-464.

Schwartz, R. C. (2001). Self-awareness in schizophrenia: Its relationship to depressive
symptomatology and broad psychiatric impairments. Journal of Nervous and
Mental Disease, 189(6), 401-403. doi:10.1097/00005053-200106000-00010
146
Shih, M., & Sanchez, D. T. (2005). Perspectives and research on the positive and
negative implications of having multiple racial identities. Psychological Bulletin,
131(4), 569–591. doi: 10.1037/0033-2909.131.4.569

Silverstein, S. M., & Bellack, A. S. (2008). A scientific agenda for the concept of
recovery as it applies to schizophrenia. Clinical Psychology Review, 28(7), 1108-
1124. doi:10.1016/j.cpr.2008.03.004

Silverstein, S. M., Hatashita-Wong, M., Wilkniss, S., Bloch, A., Smith, T., Savitz, A., . . .
Terkelsen, K. (2006). Behavioral rehabilitation of the "treatment-refractory"
schizophrenia patient: Conceptual foundations, interventions, and outcome data.
Psychological Services, 3(3), 145-169. doi:10.1037/1541-1559.3.3.145

Spaniol, L., Wewiorski, N. J., Gagne, C., & Anthony, W. A. (2002). The process of
recovery from schizophrenia. International Review of Psychiatry, 14(4), 327-336.
doi:10.1080/0954026021000016978

Stainsby, M., Bledin, K., & Mason, O. J. (2008). Models of change in schizophrenia: One
for all, or all for some? Current Psychiatry Reviews, 4(4), 196-208.
doi:10.2174/157340008786576599

Steiger, J. H. (1990). Structural model evaluation and modification: An interval
estimation approach. Multivariate Behavioral Research, 25, 173-180.

Stulz, N., Thase, M. E., Klein, D. N., Manber, R., & Crits-Christoph, P. (2010).
Differential effects of treatments for chronic depression: A latent growth model
reanalysis. Journal of Consulting and Clinical Psychology, 78(3), 409-419.
doi.org/10.1037/a0019267

Substance Abuse and Mental Health Services Administration. (2005). National consensus
conference on mental health recovery and systems transformation. Rockville,
MD: Department of Health and Human Services.

Sue, S., & Chu, J.Y. (2003). The mental health of ethnic minority groups: Challenges
posed by the supplement to the Surgeon General's report on mental health.
Culture, Medicine and Psychiatry. 27(4), 447-465.

Tabachnick, B. G., & Fidell, L. S. (2007). Using multivariate statistics (5th ed.). Boston:
Allyn and Bacon.

Taylor, R. J., Chatters, L. M., & Celious, A. (2003). Extended family households among
black Americans. African American Research Perspectives, 1, 133-151.

Taylor, R. J., Chatters, L. M., Hardison, C. B., & Riley, A. (2001). Informal social
support networks and subjective well-being among African Americans. Journal of
Black Psychology, 27(4), 439-463. doi:10.1177/0095798401027004004
147

Telles, C., Karno, M., Mintz, J., Paz, G., Tucker, D., Arias, M., & Lopez, S. R. (1995).
Immigrant families coping with schizophrenia: behavioural family intervention v.
case management with a low-income Spanish-speaking population. British
Journal of Psychiatry, 167(4), 473-479.

Tempier, R., Mercier, C., Leouffre, P., & Caron J. (1997). Quality of life and social
integration of severely mentally ill patients: A longitudinal study. Journal of
Psychiatry and Neuroscience, 22(4), 249-255.

Test, M. A., Greenberg, J. S., Long, J. D., Brekke, J. S., & Burke, S. S. (2005). Construct
validity of a measure of subjective satisfaction with life of adults with serious
mental illness. Psychiatric Services, 56(3), 292-300.

Tohen, M., Hennen, J., Zarate, C. M., Jr., Baldessarini, R. J., Strakowski, S. M., . . .
Cohen, B. M. (2000). Two-year syndromal and functional recovery in 219 cases
of first-episode major affective disorder with psychotic features. American
Journal of Psychiatry, 157(2), 220–8.

Udry, J. R., Li, R. M., & Hendrickson-Smith, J. (2003). Health and behavior risks of
adolescents with mixed-race identity. American Journal of Public Health, 93(11),
1865-1870. doi:10.2105/AJPH.93.11.1865

Uher, R., Muthén, B., Souery, D., Mors, O., Jaracz, J., Placentino, A., . . . McGuffin, P.
(2009). Trajectories of change in depression severity during treatment with
antidepressants. Psychological Medicine, 40(8), 1367-77.
doi:10.1017/S0033291709991528

Uher, R., Mors, O., Rietschel, M., Rajewska-Rager, A., Petrovic, A., Zobel, A., . . .
McGuffin, P. (2011). Early and delayed onset of response to antidepressants in
individual trajectories of change during treatment of major depression: A
secondary analysis of data from the genome-based therapeutic drugs for
depression (GENDEP) study. Journal of Clinical Psychiatry, 72(11), 1478-1484.
doi:10.4088/JCP.10m06419

U.S. Department of Health and Human Services. (1999). Mental Health: A Report of the
Surgeon General. Rockville, MD: Substance Abuse and Mental Health Services
Administration, Center for Mental Health Services. Retrieved from
http://www.surgeongeneral.gov/library/mentalhealth/home.html

U.S. Department of Health and Human Services. (2001). Mental health: Culture, race,
and ethnicity- A supplement to mental health: A report of the surgeon general.
Rockville, MD: U.S. Department of Health and Human Services, Substance
Abuse and Mental Health Services Administration, Center for Mental Health
Services. Retrieved from http://www.mentalhealth.samhsa.gov/cre/toc.asp

148
U.S. Department of Health and Human Services. (2005). National Consensus Conference
on Mental Health Recovery and Systems Transformation. Rockville, MD:
Substance Abuse and Mental Health Services Administration.

Vittengl, J. R., Clark, L. A., & Jarrett, R. B. (2004). Improvement in social-interpersonal
functioning after cognitive therapy for recurrent depression. Psychological
Medicine: A Journal of Research in Psychiatry and the Allied Sciences, 34(4),
643-658. doi:10.1017/S0033291703001478

Vittengl, J. R., Clark, L. A., & Jarrett, R. B. (2009). Deterioration in psychosocial
functioning predicts relapse/recurrence after cognitive therapy for depression.
Journal of Affective Disorders, 112(1-3), 135-143.
doi:10.1016/j.jad.2008.04.004\

Whaley, A. L. (1997). Ethnicity/Race, paranoia, and psychiatric diagnosis: Clinician bias
versus sociocultural differences. Journal of Psychopathology and Behavioral
Assessment, 19(1), 1-20.

Whaley, A. L. (2001). Cultural mistrust and the clinical diagnosis of paranoid
schizophrenia in African American patients. Journal of Psychopathology and
Behavioral Assessment, 23(2), 93-100.

Whaley, A. L. (2002). Confluent paranoia in African American psychiatric patients: An
empirical study of Ridley’s typology. Journal of Abnormal Psychology, 111(4),
568-577. doi:10.1037/0021-843X.111.4.568

Wiffen, B. D. R., Rabinowitz, J., Lex, A., & David, A. S. (2010). Correlates, change and
‘state or trait’ properties of insight in schizophrenia. Schizophrenia
Research, 122(1-3), 94-103. doi:10.1016/j.schres.2010.03.005

Xie, H., McHugo, G. J., Helmstetter, B.S., & Drake, R. E. (2005). Three-year recovery
outcomes for long-term patients with co-occurring schizophrenic and substance
use disorders. Schizophrenia Research, 75, 337– 348.
doi:10.1016/j.schres.2004.07.012

Yamada, A. M, Barrio, C., Morrison, S. W., Sewell, D., Jeste, D. V. (2006) Cross-
ethnic evaluation of psychotic symptom content in hospitalized middle-aged and
older adults. General Hospital Psychiatry, 28, 161– 168.
doi:10.1016/j.genhosppsych.2005.12.003

Yamada, A. M., & Brekke, J. S. (2008). Addressing mental health disparities through
clinical competence not just cultural competence: The need for assessment of
sociocultural issues in the delivery of evidence-based psychosocial rehabilitation
services. Clinical Psychology Review, 28, 1386–1399.
doi:10.1016/j.cpr.2008.07.006

149
Young, J. L., Griffith, E. E. H., & Williams, D. R. (2003). The integral role of pastoral
counseling by African-American clergy in community mental health. Psychiatric
Services, 54(5), 688-692.

Zhang, A. Y., & Snowden, L. R. (1999). Ethnic characteristics of mental disorders in five
U.S. communities. Cultural Diversity and Ethnic Minority Psychology, 5(2), 134-
146.

Zissi, A., & Barry, M. M. (2006). Well-being and life satisfaction as components of
quality of life in mental disorders. In H., Katschnig, H., Freeman, & N. Sartorius
(Eds.), Quality of Life in Mental Disorders (2nd ed.) (pp. 33-44)., New York:
John Wiley & Sons Ltd.

150
APPENDIX A: EXAMPLE MPLUS SCRIPTS FOR MULTILEVEL MODELS

Script for Modeling Linear Growth (Level 1): Psychotic Symptoms Outcome

TITLE: Psychosis Linear Growth Model
DATA:
FILE mhsa.dat;

VARIABLE:
NAMES = fsp sex age race educ dxcat4 psychos1 psychos2 psychos3;
USEVARIABLES = psychos1 psychos2 psychos3;
MISSING ARE ALL (999);

ANALYSIS: ESTIMATOR = ML;

MODEL: !Default analysis is Mean structure
i BY psychos1 - psychos3 @1; !Intercept fixed at 1
s BY psychos1 @ 0
psychos2 @ .5
psychos3 @ 1; !Slopes fixed at baseline, 1/2 year and 1 year
[i s]; !Latent variables have means
psychos1 psychos2 psychos3 (Ve); !Equal error variance
[psychos1 - psychos3 @0]; !No additional intercepts

OUTPUT:
SAMPSTAT RESIDUAL STANDARDIZED;

151
Script for Modeling Nonlinear Growth (Level 1): Psychotic Symptoms Outcome

TITLE: Psychosis Nonlinear Growth Model
DATA:
FILE mhsa.dat;

VARIABLE:
NAMES = fsp age race educ dxcat4 psychos1 psychos2 psychos3;
USEVARIABLES = psychos1 psychos2 psychos3;
MISSING ARE ALL (999);

ANALYSIS: ESTIMATOR = ML;

MODEL: !Default analysis is Mean structure
i BY psychos1 - psychos3 @1; !Intercept fixed at 1
s BY psychos1 @ 0
psychos2 * .5
psychos3 @ 1; !Slopes fixed at baseline and 1 year; ½ year allowed to vary
[i s]; !Latent variables have means
psychos1 psychos2 psychos3 (Ve); !Equal error variance
[psychos1 - psychos3 @0]; !No additional intercepts

OUTPUT:
SAMPSTAT RESIDUAL STANDARDIZED;

152
Script for Modeling Nonlinear Growth (Level 1) with Race Covariates (Level 2):
Psychotic Symptoms Outcome

TITLE: Psychosis Nonlinear Growth Model with Race Dummy Covariates
DATA:
FILE mhsa.dat;

VARIABLE:
NAMES = fsp age race educ dxcat4 psychos1 psychos2 psychos3;
USEVARIABLES = psychos1 psychos2 psychos3 afam latino multir;
MISSING ARE ALL (999);

DEFINE: afam = 0; IF (race==1) THEN afam = 1; !Creating dummy coded variable
latino = 0; IF (race==2) THEN latino = 1; !Creating dummy coded variable
multir = 0; IF (race==3) THEN multir = 1; !Creating dummy coded variable

ANALYSIS: ESTIMATOR = ML;

MODEL: !Default analysis is Meanstructure
i BY psychos1 - psychos3 @1; !Intercept fixed at 1
s BY psychos1 @ 0
psychos2 * .5
psychos3 @ 1; !Slopes fixed at baseline and 1 year; ½ year allowed to vary
[i s]; !Latent variables have means
psychos1 psychos2 psychos3 (Ve); !Equal error variance
[psychos1 - psychos3 @0]; !No additional intercepts

i on afam latino multir; !Modeling level 2 fixed effects
s on afam latino multir; !Modeling level 2 fixed effects

OUTPUT:
SAMPSTAT RESIDUAL STANDARDIZED;

153
Script for Modeling Nonlinear Growth (Level 1) with All Covariates (Level 2):
Psychotic Symptoms Outcome

TITLE: Psychosis Nonlinear Growth Model with All Level 2 Covariates
DATA:
FILE mhsa.dat;

VARIABLE:
NAMES = fsp age race educ dxcat4 psychos1 psychos2 psychos3;
USEVARIABLES = fsp age educ dxcat4 psychos1 psychos2 psychos3 afam latino
multir;
MISSING ARE ALL (999);

DEFINE: afam = 0; IF (race==1) THEN afam = 1; !Creating dummy coded variable
latino = 0; IF (race==2) THEN latino = 1; !Creating dummy coded variable
multir = 0; IF (race==3) THEN multir = 1; !Creating dummy coded variable

ANALYSIS: ESTIMATOR = ML;

MODEL: !Default analysis is Meanstructure
i BY psychos1 - psychos3 @1; !Intercept fixed at 1
s BY psychos1 @ 0
psychos2 * .5
psychos3 @ 1; !Slopes fixed at baseline and 1 year; ½ year allowed to vary
[i s]; !Latent variables have means
psychos1 psychos2 psychos3 (Ve); !Equal error variance
[psychos1 - psychos3 @0]; !No additional intercepts

i on fsp age educ dxcat4 afam latino multir; !Modeling level 2 fixed effects
s on fsp age educ dxcat4 afam latino multir; !Modeling level 2 fixed effects

OUTPUT:
SAMPSTAT RESIDUAL STANDARDIZED;

154
APPENDIX B: EXAMPLE MPLUS SCRIPTS FOR GROWTH MIXTURE
MODELIING

Script for Initial Linear Growth Model: Satisfaction with Life Outcome

TITLE: Satisfaction with Life - Linear Growth Model

DATA:
FILE mhsa2.dat;

VARIABLE:
NAMES = sid fsp sex age age25_34 age35_44 age45_54 age55p afam latino multi
married educ educ12y educmo12 dxcat4 inpt2 swl1 swl2 swl3;

USEVARIABLES = swl1 swl2 swl3;

MISSING ARE ALL (999);

ANALYSIS: ESTIMATOR = ML;

MODEL: !Default analysis is Meanstructure
i s | swl1@0 swl2@.5 swl3@1; !Intercept fixed at 1
!Slopes fixed at baseline, 1/2 year and 1 year
swl1 swl2 swl3 (Ve); !Equal error variance

OUTPUT: SAMPSTAT STANDARDIZED tech1;

155
Script for One-Class Growth Mixture Model: Satisfaction with Life Outcome

TITLE: Satisfaction with Life - One Class GMM

DATA:
FILE mhsa2.dat;

VARIABLE:
NAMES = sid fsp sex age age25_34 age35_44 age45_54 age55p afam latino multi
married educ educ12y educmo12 dxcat4 inpt2 swl1 swl2 swl3;

USEVARIABLES = fsp sex age25_34 age35_44 age45_54 age55p afam latino multi
married educ12y educmo12 dxcat4 inpt2 swl1 swl2 swl3;

MISSING ARE ALL (999);
IDVARIABLE = sid; !Identifying cases by ID number
CLASSES = C(1); !Defining number of classes

ANALYSIS:
TYPE = MIXTURE; !Missing is default
ESTIMATOR = MLR;
STARTS = 50 5; !Adjusted for successful convergence
STITERATIONS = 20; !Adjusted for successful convergence
CONVERGENCE = 0.00005;
COVERAGE = 0.10;

MODEL:
%OVERALL%
i s | swl1@0 swl2@.5 swl3@1; !Intercept fixed at 1
!Slopes fixed at baseline, 1/2 year and 1 year
swl1 swl2 swl3 (Ve); !Equal error variance
i-s@0;!Within class variance fixed to zero
i s ON fsp sex age25_34 age35_44 age45_54 age55p afam latino multi married educ12y
educmo12 dxcat4 inpt2; !Regressing intercept and slope on covariates

OUTPUT: SAMPSTAT STANDARDIZED tech1 tech8;

PLOT: TYPE = PLOT3; !Obtaining plot for means and individual values
SERIES = swl1-swl3 (s);

156
Script for Two-Class Growth Mixture Model: Satisfaction with Life Outcome

TITLE: Satisfaction with Life -Two Class GMM

DATA:
FILE mhsa2.dat;

VARIABLE:
NAMES = sid fsp sex age age25_34 age35_44 age45_54 age55p afam latino multi
married educ educ12y educmo12 dxcat4 inpt2 swl1 swl2 swl3;

USEVARIABLES = fsp sex age25_34 age35_44 age45_54 age55p afam latino multi
married educ12y educmo12 dxcat4 inpt2 swl1 swl2 swl3;

MISSING ARE ALL (999);
IDVARIABLE = sid; !Identifying cases by ID number
CLASSES = C(2); !Defining number of classes

ANALYSIS:
TYPE = MIXTURE; !Missing is default
ESTIMATOR = MLR;
STARTS = 500 20; !Adjusted for successful convergence
STITERATIONS = 20; !Adjusted for successful convergence
CONVERGENCE = 0.00005;
COVERAGE = 0.10;

MODEL:
%OVERALL%
i s | swl1@0 swl2@.5 swl3@1; !Intercept fixed at 1
!Slopes fixed at baseline, 1/2 year & 1 year
swl1 swl2 swl3 (Ve); !Equal error variance
i-s@0; !Within class variance fixed to zero
i s ON fsp sex age25_34 age35_44 age45_54 age55p afam latino multi married educ12y
educmo12 dxcat4 inpt2; !Regressing intercept and slopes on covariates
c#1 ON fsp sex age25_34 age35_44 age45_54 age55p afam latino multi married educ12y
educmo12 dxcat4 inpt2; !Multinomial logistic regression with latent class as dependent
variables and covariates as independent variables

OUTPUT: SAMPSTAT STANDARDIZED CINTERVAL tech1 tech8 tech11;
PLOT: TYPE = PLOT3; !Obtaining plot for means and individual values
SERIES = swl1-swl3 (s);

SAVEDATA: File = \\vmware-host\Shared Folders\Desktop\CLASSoutput2swl;
save = cprobabilities; !Saving posterior probabilities and class assignments for each case
157
Script for Three-Class Growth Mixture Model: Satisfaction with Life Outcome

TITLE: Satisfaction with Life – Three-Class Model with Covariates
DATA:
FILE mhsa2.dat;

VARIABLE:
NAMES = sid fsp sex age age25_34 age35_44 age45_54 age55p afam latino multi
married educ educ12y educmo12 dxcat4 inpt2 swl1 swl2 swl3;

USEVARIABLES = fsp sex age25_34 age35_44 age45_54 age55p afam latino multi
married educ12y educmo12 dxcat4 inpt2 swl1 swl2 swl3;

MISSING ARE ALL (999);
IDVARIABLE = sid; !Identifying cases by ID number
CLASSES = C(3); !Defining number of classes

ANALYSIS:
TYPE = MIXTURE; !Missing is default
ESTIMATOR = MLR;
STARTS = 1000 20; !Adjusted for successful convergence
STITERATIONS = 20; !Adjusted for successful convergence
CONVERGENCE = 0.00005;
COVERAGE = 0.10;

MODEL:
%OVERALL%
i s | swl1@0 swl2@.5 swl3@1; !Intercept fixed at 1
!Slopes fixed at baseline, 1/2 year and 1 year
swl1 swl2 swl3 (Ve); !Equal error variance
i-s@0; !within class variance fixed to zero
i s ON fsp sex age25_34 age35_44 age45_54 age55p afam latino multi married educ12y
educmo12 dxcat4 inpt2; !Regressing intercept and slope on covariates
c#1 ON fsp sex age25_34 age35_44 age45_54 age55p afam latino multi married educ12y
educmo12 dxcat4 inpt2;
c#2 ON fsp sex age25_34 age35_44 age45_54 age55p afam latino multi married educ12y
educmo12 dxcat4 inpt2; !Multinomial logistic regression with latent class as dependent
variables and covariates as independent variables

OUTPUT: SAMPSTAT STANDARDIZED CINTERVAL tech1 tech8 tech11;
PLOT: TYPE = PLOT3; !Obtaining plot for means and individual values
SERIES = swl1-swl3 (s);

SAVEDATA: File = \\vmware-host\Shared Folders\Desktop\CLASSoutput3swl;
save = cprobabilities; !Saving posterior probabilities and class assignments for each case

Abstract (if available)

Abstract Psychosocial rehabilitation services (PRS) are designed to improve recovery outcomes (symptoms, functioning and quality of life) for individuals with severe mental illness. However, the heterogeneity in the course of severe mental illnesses suggests that there will be variation in response to PRS. Studies have shown that in addition to treatment effects, demographic and clinical factors may play a role in heterogeneity in recovery outcomes. In addition, limited research has examined recovery outcomes for racial/ethnic minorities with severe mental illness despite theory and research suggesting that cultural factors may influence recovery. Using quantitative data from an ongoing study funded by the National Institute of Mental Health, the goal of this dissertation was to understand the impact of PRS on recovery outcomes for racially/ethnically diverse individuals with a severe mental illness receiving services from five mental health clinics in Los Angeles County over the course of one year. The dissertation is organized as three studies. Study 1 examined the differential effects of PRS on recovery outcomes for Euro American, African American, Latino and multiracial individuals with a severe mental illness. Studies 2 and 3 explored the demographic, clinical, and service predictors of heterogeneity in recovery outcomes. Overall, the findings indicated that: 1) psychosocial rehabilitation services maintained or improved recovery outcomes over time, 2) there was variation across individuals in the trajectory of recovery outcomes, and 3) recovery outcomes differed based on a variety of demographic and clinical factors, including race and ethnicity. Mental health practice and research implications as well as suggestions for future research for racially/ethnically diverse individuals with severe mental illnesses are discussed.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Sociocultural influences on mental health functioning: implications for the design of community-based services

PDF

Community integration of individuals with serious mental illness: a network perspective from India and United States

PDF

Working alliance, hope, and functional outcome for individuals with schizophrenia: mechanisms of influence

PDF

Asian Americans served at community mental health agencies for a diagnosis of schizophrenia

PDF

The impact of family factors on the functioning of African-American consumers living with schizophrenia

PDF

Longitudinal relationships of cognitive deficits, symptoms, and social functioning outcomes in community-based psychosocial rehabilitation programs: mechanisms of longitudinal change

PDF

Mental health service use by children and youth in the child welfare system: a focus on need and predisposing factors and caregiver type

PDF

The role of protective factors on outcomes for Latinos with schizophrenia

PDF

Mindfulness and resilience: an investigation of the role of mindfulness in post-9/11 military veterans' mental health-related outcomes

PDF

Mindfulness-based self-regulation for psychotic disorders: a feasibility study

PDF

Neighborhood context and adolescent mental health: development and mechanisms

PDF

Unleashing prevention and reaching the masses with a positive mental wellness museum

PDF

Interdisciplinary community collaborative to reduce adverse outcomes with teen pregnancies

PDF

One size fits none: theorizing weight management in the everyday lives of adults with serious mental illness

PDF

Implementation of peer providers in integrated health care settings

PDF

Multilevel influences on organizational adoption of innovative practices in children's mental health services

PDF

Cultural resources and health among Asian Americans: results from the National Latino and Asian American Study

PDF

Trauma-related treatment gains among women with histories of interpersonal violence and co-occurring mental health and substance abuse disorders

PDF

The influence of contextual factors on the processes of adoption and implementation of evidence-based substance use prevention and tobacco cessation programs in schools

PDF

Immigrants at a loss: the need for services that promote child well-being among Latino families with child welfare contact

Asset Metadata

Creator Edmondson, Melissa Ann (author)

Core Title The impact of psychosocial rehabilitation services on recovery outcomes for racial/ethnic minorities with severe mental illness

School School of Social Work

Degree Doctor of Philosophy

Degree Program Social Work

Publication Date 08/31/2012

Defense Date 07/30/2012

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

Tag mental health outcomes,minorities,OAI-PMH Harvest,recovery,schizophrenia,severe mental illness

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Brekke, John S. (committee chair), Lincoln, Karen D. (committee member), McArdle, John J. (committee member), Palinkas, Lawrence A. (committee member)

Creator Email mae0210@gmail.com,medmonds@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-94336

Unique identifier UC11289278

Identifier usctheses-c3-94336 (legacy record id)

Legacy Identifier etd-EdmondsonM-1178.pdf

Dmrecord 94336

Document Type Dissertation

Rights Edmondson, Melissa Ann

Type texts

Source 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 a...

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