University of Southern California Dissertations and Theses

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

(USC Thesis Other)

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

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

LONGITUDINAL RELATIONSHIPS OF COGNITIVE DEFICITS, SYMPTOMS,
AND SOCIAL FUNCTIONING OUTCOMES IN COMMUNITY-BASED
PSYCHOSOCIAL REHABILITATION PROGRAMS: MECHANISMS OF
LONGITUDINAL CHANGE

by

Maanse Hoe

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

August 2007

Copyright 2007 Maanse Hoe

ii

DEDICATION

It is with great honor that I dedicate this dissertation to the deceased Dr. John
L. Horn (1929-2006). Dr. Horn was a psychology professor at the University of
Southern California and best known for his groundbreaking work in the field of
psychometrics, the measurement of human cognitive ability (e.g., Cattell-Horn
theory of multiple intelligences).
It was John who first encouraged me to pursue the initial idea of this
dissertation. He also gave me the most valuable lesson of my doctoral education,
which was in his words to me:
In this regard think of a statement by one of my favorite authors, Mark
Twain: "Always do right: it will amaze most, and gratify some," or
words to that effect. Do the job the BSI constructors should have
done; it will amaze most, and gratify some.

iii

ACKNOWLEDGEMENTS
I am sincerely grateful to each member of my dissertation committee: Drs.
John Brekke, John Bola, Bradley Zebrack and Chih-Ping Chou. Their unique
expertise guided me in the completion of this dissertation. Dr. Chou’s substantial
knowledge and experience in research methods was extremely helpful.
Since childhood I have wanted to be a scholar and always dreamed of
meeting a mentor who would be a role model as a scholar and educator. Dr. John
Brekke is the person I envisioned in my dream. He is a thoughtful educator and the
most ardent researcher I have ever met. I deeply thank him for his full support and
guidance during my doctoral education. I also want to thank Drs. John Bola and
Robert Nishmoto for their unfailing support and friendship. I truly respect their
savory world views.
It was my fortune also to have met top researchers: Drs. John Hon, Bengt
Muthén and Jack McArdle. They taught me how to read data, explore stories in the
data and advance the knowledge base as a researcher. Their comments on my
research questions for my dissertation were encouraging and helpful. They deserve
my special thanks. I also cannot forget to thank Dr. Bin Xie for his friendship and for
sharing his knowledge with me.
I spent five years in the School of Social Work at USC, and it will remain one
of the most delightful memories of my life. I fully enjoyed my tenure as a doctoral

iv

student. I thank all the staff and faculty at the school for their support. Although I
cannot mention each one individually, you all know who you are.
Additionally, I wish to extend my thanks to my colleagues who have been
with me during this doctoral program and to the professors who taught me in my
master’s and bachelor’s programs. A special note of recognition should come to Drs.
J. Sam Park, Kyu-Taik Sung, Soo-Hyun You, Tae-Kyun Yoo and Sung-Woo Bae.
The honor of fulfilling my doctoral education also should be shared with my
family. I especially thank my mother, the best teacher in my life, who planted the
seed to achieve such honor.

v

TABLE OF CONTENTS

DEDICATION .............................................................................................................ii

ACKNOWLEDGEMENTS ........................................................................................iii

LIST OF TABLES .....................................................................................................vii

LIST OF FIGURES ....................................................................................................ix

ABSTRACT.................................................................................................................x

CHAPTER I: INTRODUCTION.................................................................................1
1.1 Purpose of the Dissertation ................................................................................1
1.2 Organization of the Dissertation ........................................................................2

CHAPTER II: LITERATURE REVIEW ....................................................................3
2.1 Significance and Background ............................................................................3
2.2 Two Cognitive Rehabilitation Approaches........................................................5
2.3 Cognitive Deficits in Schizophrenia ..................................................................6
2.4 Current Interests in Service Outcome Research...............................................10

CHAPTER III: SPECIFIC AIMS AND HYPOTHESES ..........................................12
3.1 Specific Aim 1 .................................................................................................12
3.2 Specific Aim 2 .................................................................................................14

CHAPTER IV: METHOD .........................................................................................15
4.1 Data Source ......................................................................................................15
4.2 Design ..............................................................................................................16
4.3 Measures ..........................................................................................................17
4.4 Statistical Methods...........................................................................................21
4.5 Overview of Statistical Methods......................................................................24
4.5.1 Latent Growth Curve Analysis..................................................................24
4.5.2 Latent Mean Difference Test ....................................................................30

CHAPTER V: RESULTS ..........................................................................................34
5.1 Description of the Sample................................................................................34
5.2 Preliminary Analyses .......................................................................................37
5.3 Hypothesis Testing...........................................................................................38

vi

5.3.1 Hypothesis 1: Change in psychosocial functioning with a time-varying
covariate of psychiatric symptoms..............................................................38
5.3.2 Hypothesis 2: Change in psychiatric symptoms over 12 months .............40
5.3.3 Hypothesis 3: Change in neurocognition over 12 months ........................42
5.3.4 Hypothesis 4-1: Relationships between change in psychosocial
functioning and psychiatric symptoms .......................................................47
5.3.5 Hypothesis 4-2: Relationships among change in neurocognition, change
in psychiatric symptoms and change in psychosocial functioning .............49
5.3.6 Hypothesis 5: Latent trajectory classes of change in psychosocial
functioning ..................................................................................................52
5.4 Exploratory Analyses of Predictors of Latent Trajectory Classes ...................61
5.4.1 General Growth Mixture Model................................................................61
5.4.2 Mean Difference Tests ..............................................................................62
5.4.3 Logistic Regression Analyses ...................................................................64

CHAPTER VI: DISCUSSION...................................................................................67
6.1 Summary of Findings.......................................................................................67
6.2 Overall Discussion ...........................................................................................70
6.2.1 Restorative Model and Compensatory Model in Cognitive
Rehabilitation Treatments...........................................................................70
6.2.2 Relationships Among Change in Psychosocial Functioning, Change in
Neurocognition and Change in Psychiatric Symptoms...............................73
6.2.3 Who Changes and Who Does Not During Community-Based
Psychosocial Rehabilitation? ......................................................................75
6.3 Implications......................................................................................................77
6.3.1 Implications for Practice ...........................................................................77
6.3.2 Implications for Social Work Research ....................................................79
6.4 Limitations and Suggestions for Further Studies.............................................82

REFERENCES...........................................................................................................85

vii

LIST OF TABLES
Table 1: Measurement protocol.…………………...….…………………………….17

Table 2: Characteristics of the samples: full baseline sample and 12-month
completers……………………………………………………...……………35

Table 3: Correlation and means for psychosocial functioning and symptoms……...38

Table 4: Estimated means and variances of initial level and of rate of change in
psychosocial functioning………………………………………...………….40

Table 5: Estimated means and variances of initial level and of rate of change in
symptoms………………………………………………………….….……..41

Table 6: Zero-order correlations among five neurocognitive tests at baseline and
at twelve months…………………………………………………………….44

Table 7: Comparisons of neurocognitive improvers and non-improvers at
baseline……………………………………………………………...……....46

Table 8: Model comparisons test results …………………………………………...51

Table 9: Estimated means and variances of multiple group growth curve model of
change in psychosocial functioning controlling for change in symptoms......51

Table 10: Test statistics for growth mixture models..……………………………....57
Table 11: Estimated means in two class growth mixture model of the change in
psychosocial functioning……………………………………….…………...58

Table 12: Significance test of the effect of service intensity on the change in
psychosocial functioning in two class growth mixture model………………60

Table 13: Mean differences of individual baseline characteristics by latent
trajectory classes of functional change……………………………………...63

Table 14: Zero-order correlations of study variables and estimated latent trajectory
classes………………...………………..………...………………………….64

viii

Table 15: Exploring potential predictors of latent trajectory classes of functional
change covaried with intensity of treatments (logistic regression results)….65

ix

LIST OF FIGURES
Figure 1: The intervention model of neurocognitive deficits in schizophrenia…........8
Figure 2: Latent growth curve model (unconditional linear change)……………….25
Figure 3: Hypothesized model of functional change with a time-varying
covariate……….….………………………………………………………....39

Figure 4: Hypothesized model of latent mean differences between neurocognition
at baseline and at twelve months………………………..………………......43

Figure 5: Proportion of cognitive improvers and cognitive non-improvers......…….45

Figure 6: Bivariate change model of the effect of symptom change on functional
change…………………………….………………………………………....48

Figure 7a: Estimated mean growth curve in a model with single latent class……....53

Figure 7b: Estimated mean growth curves in a model with two latent classes……..54

Figure 7c: Estimated mean growth curves in a model with three latent classes…….55

Figure 7d: Estimated mean growth curves in a model with four latent classes…..…56

Figure 8: Estimated means and sample means in the two class growth mixture
model of the functional change with a covariate of service intensity……….59

Figure 9: General growth mixture model of the functional change with a covariate
of service intensity…………………………………………………………..62

x

ABSTRACT
The present study investigated longitudinal relationships among cognition,
symptoms and functioning in individuals diagnosed with schizophrenia for better
understating the mechanisms of psychosocial functional change in community-based
psychosocial rehabilitation programs. There are 2.97 million people suffering form
schizophrenia in the United States, and the economic burden of schizophrenia is
approximately $62.7 billion per year. Community-based treatment programs for
individuals diagnosed with schizophrenia have established intensive clinical services
to support and maintain these individuals in communities. The outstanding success
of the programs has resulted in the clinical outcomes of reducing relapse and
hospitalization, improving social functioning and improving housing stability.
However, this success has been controversial because there is no clear understanding
of how individual change is caused by the treatments and of why the change occurs
or does not occur in individuals. This is a gap of knowledge in practice and research.
Two specific aims of this study were 1) to investigate longitudinal
relationships among change in neurocognition, change in psychiatric symptoms and
change in psychosocial functional outcomes given community-based psychosocial
rehabilitation programs; and 2) to explore unobserved heterogeneous sub-groups
(latent classes) in psychosocial functioning change. All research hypotheses were
derived from two cognitive rehabilitation models (restorative and compensatory),
and from Green and Nuechterlein’s model of neurocognition and functional

xi

outcomes. The sample consists of 130 individuals diagnosed with schizophrenia or
schizoaffective disorder from five rehabilitation programs. Subjects were assessed
every six months over 12 months regarding psychosocial functioning and psychiatric
symptoms and neurocognition. Structural equation modeling based on longitudinal
data analysis methods was used (e.g., Growth Mixture Modeling).
Findings revealed three treatment mechanisms of community-based
psychosocial rehabilitation programs: 1) Psychosocial functioning can be improved
through increase in neurocognition (i.e., a therapeutic mediator); 2) Improvement in
psychosocial functioning is not related to change in psychiatric symptoms; and 3)
Effects of service intensity on the functional improvement are more distinctive in
individuals who are younger, have more social contacts, have better neurocognition
and show less symptomatology. The implication of these findings was discussed in
terms of social work practice and research, and suggestions for future research were
presented.

1

CHAPTER I: INTRODUCTION

1.1 Purpose of the Dissertation
This dissertation study aimed to investigate longitudinal relationships among
change in cognitive deficits in schizophrenia, change in psychiatric symptoms and
change in psychosocial functional outcomes in community-based psychosocial
rehabilitation programs. This purpose of the study was derived from two competing
theories of cognitive rehabilitation for those with schizophrenia: a restorative model
and a compensatory model (Bellack, Gold, & Buchanan, 1999; Velligan & Glahn,
2004). Based on these two competing theories, research hypotheses of the
longitudinal relationships among the change in cognitive deficits in schizophrenia,
the change in psychiatric symptoms and the change in psychosocial functional
outcomes were formulated and analyzed. In addition, an analytical aim of this study
was to explore unobserved heterogeneous sub-groups (latent classes) in psychosocial
functioning change, in other words, to identify trajectory group membership in the
different individual trajectories of the unobserved sub-groups in the sample. Data
were collected through a prospective quasi-experimental service outcome study,
which focused on investigating the change of treatment outcomes in community-
based psychosocial rehabilitation programs for those with schizophrenia.

2

1.2 Organization of the Dissertation
This dissertation consists of five chapters. In Chapter I, the purpose of this
study is compactly described. Chapter II presents the significance of the study, the
theoretical as well as empirical backgrounds of the study, and current research
interests relevant to the study topic. In Chapter III, two specific aims of the study are
presented along with research hypotheses and questions. Chapter IV describes
overall research methodology. Especially, an overview of statistical methods applied
in the study is discussed with technical details, such as equations. Chapter V
provides the results, which consist of four parts: description of the sample,
preliminary analysis results, hypothesis testing results and exploratory analysis
results. Chapter VI provides a discussion of the research findings and their
implications. Discussed are three topics, which are two cognitive rehabilitation
models (restorative model and compensatory model), longitudinal relationships
among three domains in schizophrenia (neurocognition, psychosocial functioning
and psychiatric symptoms), and findings of latent trajectory classes of functional
change. This chapter also presents the author’s interpretation of the major findings
regarding two aspects of implication (implication for social work practice and
implication for social work research). In addition, the limitations of this dissertation
study and suggestions for future research are included in Chapter VI.

3

CHAPTER II: LITERATURE REVIEW

2.1 Significance and Background
Schizophrenia is a large problem in the United States. The National Advisory
Mental Health Council at the National Institution of Mental Health reported that 2.8
percent of the adult population in the United States suffered severe mental illness
over a one-year period (Torrey, 2001). There are 2.97 million people suffering from
schizophrenia (Mueser & McGurk, 2004), and the economic burden of schizophrenia
is approximately $62.7 billion per year (Wu et al., 2005). Sevy and Davison (1995)
have asserted that cognitive impairment is a major contributor to the direct and
indirect costs of schizophrenia because cognitive impairment is responsible for the
schizophrenic patient's high utilization of services, low productivity and the burden
placed on the family.
Community-based treatment programs for individuals diagnosed with
schizophrenia have established intensive clinical services to support and maintain
these individuals in communities (Mueser, Bond, Drake, & Resnick, 1998; Scott &
Dixon, 1995). The outstanding success of these community-based treatments has
resulted in the clinical outcomes of reducing relapse and hospitalization, improving
social functioning and improving housing stability (Brekke, Long, & Nesbitt, 1997).
However, the successfulness of community-based treatments has been controversial

4

because there is no clear understanding of how individual change is caused by the
treatments and why the change occurs or does not occur in individuals. This is a gap
of knowledge in practice and research.
Traditionally, risk factors of schizophrenia have been referred to abnormal
brain structure (Pierri et al., 1999; Benes, Kwok, & Vincent, 1998), brain chemistry
(Bell, 1973; Okubo et al., 1997; Seeman et al., 1976), environmental factors
(Mortensen et al., 1999), genes (Gottesman & Bertelsen, 1989; Pulver, 2000),
pregnancy and birth complications (Geddes & Lawrie, 1996; Kendell, Juszczak, &
Cole, 1996), and viruses (Mednick et al, 1988; Mednick, Parnas, & Schulsinger,
1987). For example, Akbarian and colleagues (1993) have demonstrated that an
abnormal distribution of a subset of pyramidal neurons in the frontal and temporal
lobes is consistent with abnormal neuronal migration during brain development.
Pierri and colleagues (1999) reported that positive and negative symptoms originated
from different brain areas, such as frontal, temporal and parietal areas. More recently,
schizophrenia has been viewed as a neurodevelopmental disorder. For example,
Lieberman and colleagues (2001) have argued that the finding of progressive loss of
brain volume in a subgroup of individuals with schizophrenia suggests an additional
neurodegenerative process. One study (Cannon et al., 2002) suggests that most
individuals who developed schizophrenia as adults had been within the normal range
for cognitive functioning and behavior in childhood, but they more commonly had

5

lower mean scores on intelligence tests and neuromotor impairment compared with
non-schizophrenia children.
As knowledge of the role of cognition in schizophrenia increases, cognitive
rehabilitation is receiving focus as a prominent intervention for people diagnosed
with schizophrenia. For example, cognitive rehabilitation has been studied,
demonstrating that neurocognitive deficits or capability affect psychiatric symptoms
and psychosocial functional outcomes (Bellack, Mueser, Wade, Sayers, & Morrison,
1992; Hogarty & Flesher, 1999). However, there has been no consensus on how
cognitive rehabilitation intervenes with cognitive deficits, decreases schizophrenic
symptoms and enhances the psychosocial functioning of those diagnosed with
schizophrenia in community-based psychosocial rehabilitation programs (Green,
1996). Therefore, more research is needed to gain a better understanding of how
cognition affects individuals’ psychosocial functioning in community-based
treatments.

2.2 Two Cognitive Rehabilitation Approaches
In the cognitive rehabilitation literature, two competing theories of cognitive
rehabilitation exist for those with schizophrenia, which are the restorative model and
the compensatory model (Bellack, Gold, & Buchanan, 1999; Velligan &Glahn,
2004). The restorative model postulates that cognitive deficits can and should be
cured to attain an effective rehabilitation outcome, while the compensatory model

6

focuses on minimizing the resultant disability rather than correcting cognitive
deficits. In other words, the restorative model aims to target and correct cognitive
deficits directly, while the compensatory model purports to develop a bypassing
strategy for helping individuals regardless of eliminating the cognitive deficits.
Consequently, these two competing explanatory models have different stances on
addressing cognitive deficits in those with schizophrenia and imply different clinical
approaches in community-based psychosocial rehabilitation programs. Therefore, it
is critical for schizophrenia researchers and clinicians to test which of the two
models is more appropriate for providing clinical services for those with
schizophrenia. Nonetheless, there has been no empirical test of these models.

2.3 Cognitive Deficits in Schizophrenia
Fundamental aspects of cognition appear to be necessary for effective
everyday functioning (Heaton & Pendleton, 1981; Sevy & Davison 1995). It is
generally accepted that cognitive deficits in schizophrenia are related to functional
outcomes (Green, 1996; Green, Kern, Robertson, Sergi, & Kee, 2000; Pinkham et al.,
2003) and psychiatric symptoms in those with schizophrenia (Bilder et al. 1985;
Strauss 1993). It has been suggested that symptoms and cognition are predictors of
community functioning, although it is not consistently reported whether cognition is
a better predictor of community functioning than symptoms (McGurk & Meltzer,
2000; Norman et al., 1999). Nonetheless, it is assumed that improvements of disease-

7

related cognitive deficits may provide essential building blocks to improve
rehabilitative outcomes in this disorder (Green, Kern, & Heaton, 2004; Green &
Nuechterlein, 1999). Green and Nuechterlein (1999) proposed three models of
neurocognitive interventions based on previous empirical study findings. The three
models are of increasing complexity, which explicates the role of neurocognitive
deficits in schizophrenia in relation to psychopharmacological and
cognitive/behavioral interventions and functional outcomes. The role of
neurocognitive deficits is as a therapeutic mediator between treatments and outcomes,
and psychiatric symptoms are correlated with basic neurocognition in addition to a
possible mediator for functional outcomes. The most comprehensive model of the
three is as follows:

8

Figure 1: The intervention model of neurocognitive deficits in schizophrenia (Green
and Nuechterlein, 1999)

Our previous study (Brekke, Hoe, Long, & Green, 2007) partially supports Green
and Nuechterlein’s model, revealing that neurocognition and social cognition each
positively predicts functional improvements over 12 months in community-based
psychosocial rehabilitation. In addition, neurocognition shows a significant
moderator effect on the relationship between functional change and service intensity
(days of received treatments). Furthermore, biosocial pathways from neurocognition

9

to functional outcomes via social cognition were confirmed in a previous study
(Brekke, Kay, Lee, & Green, 2005).
A growing body of cross-sectional research has reported that impaired
cognition in individuals with schizophrenia is associated with psychosocial
functioning (Brekke, Raine, Ansel, & Bird, 1997; Green, 1993; Green, 1996; Green
et al., 2000; Pinkham et al., 2003). These cognitive deficits typically are seen in
working memory, executive memory, verbal memory and attention (Green, 1996).
For example, some individuals with schizophrenia achieved poor scores on tests of
verbal memory, while performing normally on tests of working memory,
psychomotor speed and attention. Thus, cross-sectional studies have reported
sufficiently the correlations among cognitive deficits in schizophrenia, psychiatric
symptoms and psychosocial functional outcomes. These types of studies, however,
are limited in their ability to answer what and how changes arise during psychosocial
rehabilitation interventions for those with schizophrenia. Furthermore, longitudinal
relationships among cognitive deficits in schizophrenia, psychiatric symptoms and
psychosocial functional outcomes rarely have been rarely studied using samples
from community-based psychosocial rehabilitation programs. Thus, it is necessary to
study the longitudinal relationships to better understand the mechanisms of
psychosocial functional change in community-based psychosocial rehabilitation
programs for people with schizophrenia. The questions asked in this study are: 1)
How do cognitive deficits, symptoms and social functioning change over time? 2)

10

How are their changes related? 3) Who changes and who does not change during
psychosocial rehabilitation interventions? Answering these questions is intended to
help us to predict how much people will benefit from an intervention (Green et al.,
2004).

2.4 Current Interests in Service Outcome Research
Current interests in longitudinal service outcome research involve
investigating how service outcomes change over time, identifying what covariates
affect the change pattern of service outcomes and understanding which people
benefit from treatments under what conditions (Peer, Kupper, Long, Brekke, &
Spaulding, 2007). These unresolved questions entail rather complicated longitudinal
research, and recent methodological innovations in statistics and computer software,
such as growth curve or growth mixture analyses, make it possible to answer those
complex longitudinal research questions. However, empirical studies applying
recently developed longitudinal data analysis techniques have not yet been fully
utilized in studying psychosocial service outcomes of community-based treatment
programs for people with schizophrenia, although the recent methodological
innovations in statistics and computer software permit a researcher to address
unresolved research questions.
However, the well-known heterogeneity of the schizophrenia population
(McLachlan, Welham, & McGrath. 2000) makes it harder to understand which

11

people benefit from treatments under what conditions. The heterogeneity feature in
schizophrenia might lead any statistical method, which assumes a homogenous
population, to produce wrong estimations or, at best, biased aggregated estimations
in data analysis (Nagin, 2005). In addition, the consequence of understanding which
people benefit from treatments under what conditions might be sub-groups in a
population in terms of their differences in the trajectory pattern of service outcomes
over time, which responds to treatment effect. Thus, a longitudinal service outcome
study necessitates the analysis of groups, such as dummy group variable analysis,
multiple-group analysis and unknown group membership analysis (Bollen & Curran,
2006). Dummy group variable analysis and multiple-group analysis are relatively
clear because the group membership in a sample is known. However, either of the
two analyses cannot capture the unobserved heterogeneity of the sample (i.e., latent
classes). Only unknown group membership analyses, such as Growth Mixture
Modeling (GMM: Muthén, 2004), can capture the heterogeneity in a schizophrenia
sample, which sheds light on understanding which people benefit from treatments
under what conditions. By using GMM in Mplus statistical software, the unobserved
heterogeneous subgroups (latent classes) in the longitudinal relationship between
cognitive deficits and psychosocial functional outcomes were explored in this
present dissertation study.

12

CHAPTER III: SPECIFIC AIMS AND HYPOTHESES

This study aimed to examine possible longitudinal relationships among
change in cognitive deficits in schizophrenia, change in psychiatric symptoms and
change in psychosocial functioning. How service intensity affected psychosocial
functioning improvement was tested. In addition, possible predictors of latent
trajectory class membership of functional improvement were explored. Each of three
variables (neurocognition, psychiatric symptoms and psychosocial functioning
outcome) was tested separately to see if they changed over time. Then their
longitudinal relationships (whether or not and how change in a variable is related to
change in another variable over time) were examined. All research hypotheses were
derived from two cognitive rehabilitation strategy models, and from Green and
Nuechterlein’s model of neurocognition and functional outcomes (1999). The
research hypotheses and exploratory questions are described in the following
chapters.

3.1 Specific Aim 1
Specific Aim 1 was to examine relationships among change in cognitive
deficits, change in psychiatric symptoms and change in psychosocial functioning.
Research hypotheses were as follows:

13

Hypothesis 1: Psychosocial functioning will improve over 12 months controlling for
psychiatric symptoms over time (i.e., time-varying covariate).

Hypothesis 2: Psychiatric symptoms will decrease over 12 months.

Hypothesis 3: Neurocognition at 12 months will be higher than that at the baseline.

Hypothesis 4-1: Change in psychiatric symptoms will predict change in
psychosocial functioning.

Hypothesis 4-2: Change in neurocognition will be associated with change in
psychosocial functioning controlling for change in psychiatric symptoms.

Hypothesis 5: There will be more than one latent trajectory class of change in
psychosocial functioning, and effects of service intensity on psychosocial
functioning will be varied within each of the latent classes.

14

3.2 Specific Aim 2
Specific Aim 2 was to explore potential predictors of identified latent
trajectory classes of psychosocial functioning improvement controlling for effects of
service intensity on functional improvement. Research questions were as follows:

Exploratory Analysis 1: Individuals with an increase in neurocognition more likely
will belong to a latent trajectory class of functional improvement controlling for the
effects of service intensity on functional improvement.

Exploratory Analysis 2: Exploring differences in individual baseline characteristics
among identified latent trajectory classes of psychosocial functioning improvement
controlling for effects of service intensity on functional improvement.

Exploratory Analysis 3: Exploring potential predictors of latent trajectory class
memberships of functional improvement controlling for effects of service intensity
on functional improvement.

15

CHAPTER IV: METHOD

4.1 Data Source
The present study was a secondary analysis using existing data collected
through an NIMH-funded, prospective quasi-experimental design study (R01 MH
53282) from 1996 to 2000 at five community-based psychosocial rehabilitation
programs in Los Angeles County (Brekke, 2000). The five programs were part of a
county mandated mental health initiative (Young, Sullivan, Murata, Sturm, &
Koegel, 1998). Theses programs were comprehensive service environments, and
they provided integrated and comprehensive rehabilitative services: mental health
treatment, housing services, social and vocational rehabilitation, substance abuse
treatment and 24-hour crisis response. These sites are similar in their use of intensive
case management to provide comprehensive services to outpatients. Our previous
studies have shown that the programs yielded significant improvements in functional
outcomes over time (Brekke, Green, Wynn, Hoe, & Xie, 2007; Brekke, Hoe, Long,
& Green, 2007). Service characteristics, such as service intensity, were associated
significantly with functional change (Brekke, Long, Nesbitt, & Sobel, 1997).
The parent study aimed to investigate how biopsychosocial capacities and
service characteristics were related to community rehabilitative outcomes.
Psychosocial and functional status data were collected within two weeks of
neurocognition testing. Semi-structured psychosocial interviews were completed at

16

baseline, six months and 12 months by trained interviewers who were blind to the
neurocognition test results. Neurocognitive batteries, however, were administered at
two time points only (at baseline and at 12 months). Complete explanations for the
data-gathering protocols are described in previous reports (Brekke, Raine, &
Thomson, 1995).

4.2 Design
The sample consists of 130 individuals diagnosed with schizophrenia or
schizoaffective disorder who completed baseline neurocognition tests. Thirty-eight
individuals of the 168 participants recruited to the parent study were excluded from
the present study because they did not finish the complete neurocognition tests or
receive treatments in the programs. Diagnoses were determined using clinical
records, the DSM-IV checklist, and collateral reports from the admitting clinician
and on-site psychiatrist. Subjects were included if they met criteria as follows: 1)
diagnosis of schizophrenia or schizoaffective disorder,; 2) residence in Los Angeles
for at least three months before study admission; 3) adults age 18 and older; and 4)
no primary diagnosis of alcohol or drug dependence in the previous six months, no
mental retardation diagnosis and no identifiable neurological disorder. Psychosocial
functioning and psychiatric symptomatology were measured at baseline and
reassessed at six months as well as at 12 months, while neurocognition was
measured at baseline and at 12 months (Table 1). All subjects signed an informed

17

consent under protocols approved by the Institutional Review Boards at the
University of Southern California and VA Greater Los Angeles Healthcare System.

Table 1: Measurement Protocol

Measures Baseline Six Months Twelve Months
1. Functional Outcome X X X
2. Neurocognition X X
3. Symptomatology X X X
Note:
X indicates measurement, and blank means no measurement.

4.3 Measures
All psychosocial variables in the data were gathered in face-to-face
interviews conducted at a place of the subject’s choosing (at a program site or their
residence). The interviewers were master’s-level clinicians trained using a protocol
described in detail (Brekke, Levin, Wolkon, Sobel, & Slade, 1993). They were
trained on the Brief Psychiatric Rating Scale (Lukoff, Lieberman, & Nuechterlein,
1986) using a protocol described in Ventura, Nuechterlein, Subotnik, and Gilbert
(1995). The neurocognitive data came from laboratory-based assessments in a
facility designed for this study.
Functional outcome. The Role Functioning Scale, (RFS; Goodman, Sewell,
Cooley, & Leavitt, 1993), a scale of choice for this population (Green & Gracely,
1987), was used as a functional outcome measure of this study. RFS is composed of
the work, independent living and social functioning of individuals living in the

18

community. RFS is measured by interviewer ratings of work, social functioning and
independent living in accordance with procedures described in Brekke, Levin,
Wolkon, Sobel, and Slade (1993). The RFS provides anchored descriptions of each
level of functioning for each of the three items, aiming to capture both quantity and
quality of functioning in each domain. After interview training, the intra-class
correlations among three interviewers on the RFS items were greater than .80
(Brekke et al. (1993). The global score is the sum of the three items. A principal
components factor analysis of the three items found a single factor with an eigen
value greater than 1 that explained 55% of the item variance and supported the use of
the global score (Brekke & Long, 2000).
Service intensity. Service intensity was measured with a method successfully
used in a previous study of similar sites (Brekke, Long, Nesbitt, & Sobel, 1997).
Intensity was calculated as the number of days that a consumer received a service
contact from the admitting rehabilitation agency in the 365 days subsequent to
admission to the study. The service contact data were collected by staff on a daily
basis for billing and administrative purposes. The units of service data were uploaded
to a billing program that created units of service billing data in 15-minute increments.
The service intensity data came from the programs’ service data that eventually were
transferred to the county billing system. There was large individual variation in the
days and minutes of rehabilitation contact with minutes and days of treatment
received highly correlated (Pearson’s r = .7). Clients were seen on average about

19

twice a week with an average treatment day representing more than two hours of
service contact.
Neurocognition. A latent variable is constructed using five neurocognitive
measures, which represent five neurocognitive ability domains. The five domains are
verbal fluency, immediate (working) memory, secondary (episodic) memory,
sustained attention and mental flexibility. These five domains were derived from five
neurocognitive tests. Each of the five measures can be used as a single variable or as
separate factors when multiple indicators are available (Nuechterlein et. al., 2004).
However, the author in the present study uses a latent variable using the five
measures as indicators because this latent variable may reveal larger relationships
with functional outcomes than do individual neurocognitive constructs (Green, Kern,
Robertson, Sergi, & Kee, 2000).
The five tests were the Controlled Oral Word Association Test (Lezak, 1995),
the Digit Span Distractibility Test (Oltmanns & Neale, 1975), the California Verbal
Learning Test (Delis, Kramer, Kaplan, & Ober, 1987), the Degraded-Stimulus
Continuous Performance Test (Nuechterlein & Asarnow, 1992) and perseverative
errors from the Wisconsin Card Sorting Test (Heaton, 1981). The Controlled Oral
Word Association Test (Lezak, 1995) requires the subject to name as many words as
possible that begin with the letters F, A and S. The score is the sum of all acceptable
words produced in three one-minute trials. This word fluency test appears to be a
sensitive indicator of frontal functioning. The Digit Span Distractibility Test

20

(Oltmanns & Neale, 1975) is an audiotape measure of short-term verbal recall with
and without distraction. The mean of correct responses in the two conditions yields
an indicator of immediate or working memory. The California Verbal Learning Test
(Delis et al., 1987) uses the oral presentation of 16 items and allows the examination
of several aspects of verbal memory. This study used an overall recall score as a
measure of secondary memory. The Degraded-Stimulus Continuous Performance
Test is a computer-based measure of focused, sustained attention with a fast-paced
visual vigilance task. The sensitivity score measures an overall ability to discriminate
signal from noise. The Wisconsin Card Sorting Test (WCST) is sensitive to activity
in the prefrontal cortex and measures executive functioning, including the ability to
attain and shift a cognitive set. Several indices can be produced by the WCST, and
the score for perseverative errors was used.
Psychiatric symptomatology. Severity of psychiatric symptoms will be
assessed using the Brief Psychiatric Rating Scale (BPRS; Lukoff, Nuechterlein, &
Ventura, 1986; Ventura, Green, Shaner, & Liberman, 1993), which internationally is
accepted for rating the presence and severity of psychiatric symptoms. The BPRS is
a semi-structured instrument with a seven-point rating scale. Signs and symptoms are
rated along severity and impairment in functioning dimensions. BPRS reliability of
raters at UCLA is excellent immediately after training (median intra-class correlation
(ICC) = .82, n = 87). The longitudinal factor structure of the BPRS was confirmed in
Long and Brekke (1999).

21

Other measures. Individual background information, such as medication, was
obtained from the Community Adjustment Form (CAF; Test et al., 1991). The CAF
provided data regarding medication, social contacts on the Strauss and Carpenter
Outcome Scale (Strauss & Carpenter, 1972) and the Demographic Interview Form.
Social cognition was the sum of correct responses to three scales measuring visual
and auditory perception of six basic emotions (happy, angry, afraid, sad, surprised
and ashamed). The three scales were the Facial Emotion Identification Test (Kerr &
Neale, 1993), the Voice Emotion Identification Test (Kerr & Neale, 1993), and the
Videotape Affect Perception Test (Bellack, Blanchard, & Mueser, 1996).

4.4 Statistical Methods
All hypotheses in the present study were tested using the following data
analyses methods:
Hypothesis 1: Psychosocial functioning will improve over 12 months controlling for
psychiatric symptoms over time—tested using latent growth curve models, including
time-varying covariates (Duncan, Duncan, & Strycker, 2006).

Hypothesis 2: Psychiatric symptoms will decrease over 12 months—tested using
unconditional latent growth curve models (Bollen & Curran, 2006).

22

Hypothesis 3: Neurocognition at 12 months will be higher than that at baseline—
tested using a latent means difference test (Chen, Sousa, & West, 2005).

Hypothesis 4-1: Change in psychiatric symptoms will predict change in
psychosocial functioning—tested using bivariate growth curve models (Bollen &
Curran, 2006).

Hypothesis 4-2: Change in neurocognition will be associated with change in
psychosocial functioning controlling for change in psychiatric symptoms—tested
using multiple group growth curve models (Duncan, Duncan, & Strycker, 2006).

Hypothesis 5: There will be more than one latent trajectory of classes of change in
psychosocial functioning, and the effects of service intensity on psychosocial
functioning will vary within each of the latent classes—tested using growth mixture
models (Muthén et al., 2002).

Three exploratory analyses were conducted using general growth mixture
models (Muthén, 2002), ANOVA and logistic regressions. Amos version 7.0
(Arbuckle, 2006) and Mplus version 4.2 (Muthén & Muthén, 2006) were used. All
parameter estimates for models were obtained using Full-Information Maximum
Likelihood Estimation (FIMLE). FIMLE includes subjects with missing data in the

23

estimation procedure by breaking the likelihood function down into components
based on patterns of missing data (Arbuckle, 1996). The adequacy of model fit was
assessed by the chi-square statistic ( χ2), the Comparative Fit Index (CFI), the
normed fit index (NFI) and Steiger’s Root Mean Square Error of Approximation
(RMSEA), as multiple fit indices are recommended to evaluate the overall goodness-
of-fit of models (Kline, 2005). Following conventional recommendations (Arbuckle,
2006), a value greater than 0.95 for CFI and NFI indicates an acceptable model, and
values of the RMSEA less than 0.05 were indicative of acceptable model fit.

24

4.5 Overview of Statistical Methods
4.5.1 Latent Growth Curve Analysis
Latent Growth Curve Modeling (LGCM) is a state-of-the-art analytic tool in
longitudinal data analysis. LGCM is an analytical framework that integrates two
statistical techniques: 1) individual growth modeling (e.g., hierarchical linear
modeling); and 2) structural equation modeling (SEM). Traditionally, measurement
change was analyzed using Repeated Measures ANOVAs (RMANOVA) because it
was simple, flexible and relatively easy for analyzing longitudinal data but was
limited in addressing individual change (Bryk & Raudenbush, 2002). For example,
RMANOVA cannot unravel significant aspects of individual change, such as initial

level and rate of change; moreover, it requires the unrealistic assumption of
comparable growth across all subjects,

wherein heterogeneity of growth is more
likely. On the contrary, LGCM permits researchers to estimate random intercepts and
random slopes for each individual in the sample and therefore generate different
growth trajectories over time (Bollen & Curran, 2006). By incorporating random
coefficients as latent variables into SEM, researchers are capable of specifying
patterns of growth trajectories of outcomes, estimating parameter coefficients in
models of individual growth and testing the goodness of fit of the models.
A latent growth curve model is represented in Figure 2. This model has three
repeated measures (Y1 to Y3), two latent variables ( α and β) and three measurement
errors (e1 to e3). The two latent variables, which often are called growth factors,

25

represent an unobserved trajectory, which is not directly observed (in this sense, they
are latent) but is inferred from the repeated measures observed (i.e., Y1 to Y3). Thus,
a latent growth curve model is a factor analysis for analyzing panel data (see detailed
discussion in Bollen and Curran, 2006).

β α
Y1 Y2 Y3
e3 e2
e1
12
1
0
6
1
1

Figure 2: Latent growth curve model (unconditional linear change)

26

The latent growth curve model in Figure 2 is a pictorial representation of the
following equations:
y
it
= α
i
+ β
i
λ
t
+ ε
it
(G1),
α
i
= µ
α
+ ζ
αi
(G2),

β
i
= µ
β
+ ζ
βi
(G3),
where y
it
refers to multiple indicators of two growth factors ( α and β), which are
repeated measures of y for every individual i over all time points t; latent variable α
i

are factor scores for every individual i, which in turn represent the initial level (or
initial status) of a growth trajectory of y; β
i
are factor scores for every individual i,
which in turn represent rate of change (or growth rate) of a growth trajectory of y; λ
t
are time scores, which represent factor loadings between the two growth factors ( α
and β) and indicators (repeated measures of y); ε
it
, ζ
αi
, and ζ
βi
are residuals for each
equation, which represent inter-individual differences in the growth trajectory of y.
In equation (G2) and (G3), µ
α
is the mean of initial level in the growth trajectory of y,
and µ
β
is the mean of rate of change in the growth trajectory of y. Equation G1 is a
two-factor confirmatory factor analysis in that there are two latent factors (i.e., α and
β); factor loadings (i.e., time score λ
t
) are constrained to represent linear change in
Figure 2. Matrix expression of the linear change model is as follows:

27

⎟
⎟
⎟
⎠
⎞
⎜
⎜
⎜
⎝
⎛
+
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
⎟
⎟
⎟
⎠
⎞
⎜
⎜
⎜
⎝
⎛
=
⎟
⎟
⎟
⎠
⎞
⎜
⎜
⎜
⎝
⎛
3
2
1
3
2
1
12 1
6 1
0 1
i
i
i
i
i
i
i
i
y
y
y
ε
ε
ε
β
α
(G4),
where the first column elements of the second matrix (1, 1, 1) refer to constrained
factor loadings for α
i,
and the second column elements of the second matrix (0, 6, 12)
refer to constrained factor loadings ( λ
t
) for β
i,
which indicate measurement time
points (i.e., baseline, at six months and at 12 months in the context of the current
study). This linear growth model is called an unconditional linear growth curve
model in that there are no covariates (or predictors), which account for the two
growth factors ( α and β).
Unconditional linear growth curve models were used to test Hypothesis 2 in
the current study and to conduct preliminary analyses of psychosocial functioning
change. Hypothesis 1 was tested using an unconditional linear growth curve model
with time-varying covariate of psychiatric symptom change, which is represented in
the following equation:
y
it
= α
i
+ β
i
λ
t
+ γ
t
ω
it
+ ε
it
(G5),
where ω
it
refers to repeated measures of psychiatric symptom change over three time
points (baseline, at six months and 12 months). As described above, α
i
and β
i
are
latent variables, while ω
it
are observed measures for every individual i over all time
points t. Thus γ
t
are regression coefficients at each time point (a pictorial

28

representation of this model is in Figure 3 in Chapter V, p. 39). This unconditional
linear growth curve model with time-varying covariate of psychiatric symptom
change was applied simultaneously to two sub-samples of neurocognition change
(Hypothesis 4-2), which is called multiple group growth curve analysis (see detail in
Duncan, Duncan, & Strycker, 2006). For Hypothesis 4-1, a bivariate growth change
model was used (see detail in Bollen & Curran, 2006) in which growth factors in
psychiatric symptom change were specified to predict growth factors in psychosocial
functioning change (a pictorial representation of this model is in Figure 6 in Chapter
V).
Hypothesis 5 was tested using growth mixture models (GMM) (Muthén,
2002). This approach is useful to test multiple growth curve analysis when group
membership is unknown (Bollen & Curran, 2006). If k unknown groups (which are
called classes) in a growth trajectory are assumed, classes are included in a latent
growth curve analysis:
y
it
= α
i
+ β
i
λ
kt
+ ε
it
(G6),
α
i
= µ
αk
+ γ
αk
x
i
+ ζ
αi
(G7),

βi = µ
βk
+ γ
βk
x
i
+ ζ
βi
( G8),
where subscript k indicates latent classes (unknown group membership), and x
i
refers
to covariates (or predictors) of growth factors. GMM is a model-based clustering

29

technique using a posterior probability structure. In the probability structure,
Equation G6 is rewritten as follows:

] [
1
k
it
k
t
k
i
k
i
K
k
k
i it
p y ε λ β α + + =
∑
=
(G9),

where
k
i
p is the probability that the ith subject belongs to the kth class with all
k
i
p ≥
0, and 1
1
=
∑
=
K
k
k
i
p . Each class is clustered based on its probability resulting in
differentiate growth trajectories. Selection of the best number of latent classes was
followed using the recommendations in Muthén (2004), which are the Bayesian
Information Criterion (BIC) approach (smaller value indicates a better model),
Entropy (higher value indicates a better model) and a likelihood ratio test for mixture
analysis, such as the Lo-Mendell-Rubin Likelihood Ratio Test (see Li, Duncan,
Duncan, & Hops, 2001; Rodriguez, Moss, & Audrain-McGovern, 2005, for detailed
discussion).
Extended growth mixture analysis and general growth mixture models
(Muthén, 2004) were used to explore further latent classes in the growth trajectory of
psychosocial functioning change. In the current study, possible predictors (e.g.,
neurocognition) of latent trajectory membership in psychosocial functioning change
were explored. General growth mixture models use a multinomial logistic regression

30

model for the K class, when covariates were specified to predict K class membership.
Following Muthén (2001), a predictor of z is introduced as follows:

P(c
ik
= 1|z
ik
) =
∑
=
+
+
K
k
z
z
ik ck ck
ik ck ck
e
e
1
γ µ
γ µ
(G10),
where 1 =
ik
c if individual i belongs to class k and 0 otherwise. The multinomial
logistic model uses the standardization of zero coefficients for the last class, 0 =
ck
µ ,
0 =
ck
γ . Thus, the model produces the logit for the odds of class k relative to class K
(Duncan, Susan, Strycker, Okut, & Li, 2002; Muthén, 2001):
logit c = log[P(c
ik
= 1|z
ik
)/P(c
K
= 1|z
ik
)] = µ
ck
+ γ
ck
z
ik
(G11),
with K = 2 (i.e., binary classes), this multinomial logistic regression model becomes
a regular logistic regression in which
ck
γ is the increase in the log odds of being in
the alternative class versus the reference class (the last class by default in Mplus) for
a unit increase in z.

4.5.2 Latent Mean Difference Test
Hypothesis 3 is tested using the latent means difference test in the mean and
covariance structure (MACS) analysis framework (Deshon, 2004; Little, 1997;
Ployhart & Oswald, 2004; Vandenberg & Lance, 2000). The MACS is superior to

31

traditional ANOVA or T-tests in testing individual differences and group mean
differences (Ployhart & Oswald, 2004). Ployhart and Oswald summarize the
advantages of the MACS framework. The most important advantage of MACS
analysis is that it permits researchers to test simultaneously group mean differences
and measurement invariance within a single integrated statistical framework. It
should be noted that any traditional general linear model, such as ANOVA, requires
assessing measurement invariance between groups before testing group mean
differences. Measurement invariance is defined as “the equivalence of measured
constructs in two or more independent groups to assure that the same constructs are
being assessed in each group” (Chen, Sousa, & West, 2005, p. 472), and it is a
logical prerequisite for addressing group differences analyzed using observed scores
of measurement (Byrne & Watkins, 2003; Horn & McArdle, 1992). If there is no
evidence indicating the presence of measurement invariance, findings of differences
or similarities between individuals and groups cannot be interpreted clearly (Horn &
McArdle, 1992).
The measurement relationship between k items and g groups can be
expressed in the following equations (Vandenberg & Lance, 2000):

g
k
g g
k
g
k
g
k
δ ξ τ + Λ + = Χ , (M1),
g g
X
g g
X
g
δ
Θ + Λ Φ Λ =
∑
'
, (M2),

32

where
g
k
Χ is the vector of items consisting of the composite measure;
g
k
Λ is the
matrix of regression weights (factor loadings) relating
g
k
Χ to the latent variable of
g
ξ ;
g
k
τ indicates the vector of regression intercepts (or item means); and
g
k
δ is the
vector of unique factors. Following classical test theory E(
g
ξ ,
g
k
δ ) = 0, equation M1
becomes equation M2, which represents multiple group common factor analysis
(Joreskog, 1979). In the context of the current study, X is a composite measure of
neurocognition; k is five in that the neurocognition has five cognitive tests; and g is
two in that the five tests are measured twice (at baseline and at 12 months). Latent
mean differences of the neurocognition between two time points can be analyzed
using scalar invariance (DeShon, 2004; Ployhart & Oswald, 2004).
Scalar invariance (Meredith, 1993) requires three equality constraints across
groups in equation M2: 1) equal factor loadings ( Λ
1
= Λ
2
= … = Λ
G
); 2) equal
uniqueness ( Θ
1
= Θ
2
= … = Θ
G
); and 3) equal item intercepts ( τ
1
= Θ τ
2
= …
= τ
G
). When this scalar invariance model is supported by the data, latent mean
differences can be analyzed. In latent mean difference tests, latent means of one
group is set to zero as a reference group, and other groups’ latent means are denoted
as deviations from the reference group’s latent mean (Ployhart & Oswald, 2004). In
the context of the current study, the latent mean of neurocognition at baseline is set
to zero, and the latent mean of neurocognition at 12 months is tested to determine
whether the latent mean at 12 months is significantly different from zero. In addition,

33

the latent mean difference model is compared with an equal latent mean model,
which represents no change in neurocognition between baseline and 12 months.

34

CHAPTER V: RESULTS

5.1 Description of the Sample
Sample characteristics are presented in Table 2. Of the 130 subjects at
baseline, 105 subjects returned at 12 months, producing a 19.2% attrition rate. One-
sample mean tests were used to compare baseline scores of the study variables in
Table 2 between all subjects (henceforth, “originals”) and subjects who did not drop
at 12 months (“completers”). Both originals were different from completers in
regards to all study variables. However, there were no significant differences found,
suggesting that attrition effects had no impact.
Subjects were recruited at five community-based rehabilitation programs,
which consisted of three different types of community based programs: 1)
psychosocial rehabilitation clubhouses (n = 48); 2) community living programs (n
=20); and 3) case management programs (n = 62). Possible site effects of the three
program types were tested using independent-samples means tests and chi-square
difference tests. The sample of originals at baseline (n=130) were used for these
analyses. Only social cognition among all variables in Table 2 shows significant
mean differences among the three program types (F(2,119) = 3.285, p = .04).

35

Table 2: Characteristics of the samples: full baseline sample and 12-month
completers

Originals (N =130) Completers (N =105)
Gender (%)
Male
Female

89 (68.5)
41 (31.5)

72 (68.6)
33 (31.4)
Age
Range

18-62 years

21-62 years
Mean 37.98 38.19
SD 9.02 9.26
Ethnicity (%)
White

57 (43.8)

48 (45.7)
African-American 51 (39.2) 38 (36.2)
Latino 14 (10.8) 12 (11.4)
Asian 4 (3.1) 4 (3.8)
Other 4 (3.1) 3 (2.9)
Mean (SD) Mean (SD)
Education 11.89 years (1.81)
a
12.04 years (1.84)
Length of illness 13.98 years (10.01)
b
13.86 years (10.08)
d

Age of onset 23.91 years (8.94)
b
24.17 years (8.79)
d

Psychosocial
functioning
1

8.26 (3.55) 8.39 (3.67)
Neurocognition
(factor score)
-0.16 (5.51) 0.46 (4.97)
Social cognition
2
37.83 (9.50) 38.33 (9.46)
e

Symptomotogy
3
39.26 (10.22) 39.38 (10.63)
Social contact
4
1.91 (1.36) 1.90 (1.36)
Days of treatments 88.46 (55.68) 95.60 (54.02)
Days of medication
in previous six months
146.64 (63.09)
c
142.20 days (65.99)
f

Note:
1
Role Functioning Scale; total of social, work, and independence subscales.

2
Social cognition; visual and auditory perception of emotion
3
Expanded Brief Psychiatric Rating Scale
4
Social contact scores on the Strauss and Carpenter Outcome Scale
Due to missing responses,
a
n = 125.
b
n = 123.
c
n = 129.
d
n = 101.
e
n = 99.
f
n = 104

36

The highest mean score of 40.34 was found in the psychosocial rehabilitation
clubhouse, indicating that on average consumers in the clubhouse were more likely
to recognize persons’ emotions correctly than those in the other two types of
programs. The site effects, therefore, were not considered in analyzing the research
hypotheses of change in neurocognition, symptoms and psychosocial functioning
during the period of 12-month rehabilitation.
Selective non-response causes biased results. To avoid the effect of the
possible selective non-response in the sample, missing patterns in three repeated
measurements of psychosocial functioning (RFS) were analyzed before testing
change in RFS. There were four missing patterns in which 101 subjects had no
missing value on the three measurements, but 29 subjects across the other three
patterns did not have a rating score on at least one of the three measurements over
the 12 months. One-way ANOVA was used to test whether the four missing patterns
were associated significantly with scores of the RFS at baseline. No significant
association was found, indicating that the data had no statistically significant selective
non-response in measuring RFS over the three measurements. In other words, missing
data on the RFS was assumed to be missing at random (MAR), which is required to
produce unbiased parameter estimations in longitudinal data analyses (Allison, 2001;
Little & Rubin, 2002).

37

5.2 Preliminary Analyses
An unconditional latent growth curve model of change in psychosocial
functioning was examined with Amos 7 using maximum likelihood estimation
through iteration. The unconditional model had an excellent fit with the data ( χ
2
(3) =
1.240, CFI = 1.000, NFI = .994, RMSEA = .000). The mean value of the initial level
was significantly different from zero (C.R. = 26.815, p < .001), and the mean value
of the rate of change also was significantly different from zero (C.R. = 5.34, p
< .001). These results indicate that psychosocial functioning is significantly
improved across consumers in community-based rehabilitation programs.
Furthermore, the variance of the mean of the initial level is significantly larger than
zero (C.R. = 6.195, p < .001), and that of the mean of the rate of change also is
significant (C.R. = 3.735, p < .001), indicating that large interpersonal variations
exist along the overall mean trajectory of functional change. This finding means that
a consumer’s initial levels of psychosocial functioning are substantially different,
and their rates of functional change are not the same. There was, however, no
significant covariance between the initial level and the rate of change (C.R. = .915, p
< .36). This non-significant correlation suggests that community-based rehabilitation
programs help consumers improve their psychosocial functioning regardless of their
baseline level of functioning. Table 3 shows zero-order correlations, means and
variances of the measure of psychosocial functioning (RFS).

38

Table 3: Correlations and means for psychosocial functioning and symptoms
1 2 3 4 5 6
1. RFS
1
at baseline -
2. RFS at 6 months .73 -
3. RFS at 12 months .64 .80 -
4. BPRS
2
at baseline -.22 -.26 -.29 -
5. BPRS at 6 months -.19 -.35 -.39 .44 -
6. BPRS at 12 months -.15 -.18 -.29 .37 .49 -
Mean 8.26 9.68 10.13 39.26 37.23 36.33
Standard Deviation 3.56 4.10 4.72 10.23 8.96 9.88
Variance 12.66 16.79 22.31 104.63 80.31 97.55
Note:
1
RFS; total of Role Functioning Scale (social, work, and independence)
2
BPRS; total of Expanded Brief Psychiatric Rating Scale

5.3 Hypothesis Testing
5.3.1 Hypothesis 1: Change in psychosocial functioning with a time-varying
covariate of psychiatric symptoms

The statistical model for Hypothesis 1 is in Figure 3. This model shows that
psychosocial functioning is improved over 12 months controlling for psychiatric
symptoms over time (i.e., time-varying covariate). Zero-order correlations of the
model variables and their means are shown in Table 3.

39

Rate of change
Initial level
Functioning T1 Functioning T2
Functioning T3
e1 e2
e3
0
Symptom T1 Symptom T2
Symptom T3
1
1 1
6
12
1
1
1

Figure 3: Hypothesized model of functional change with a time-varying covariate

The hypothesized model in Figure 3 was constrained to have no covariance
between the initial level and the rate of change because the covariance was not
significant in the unconditional model of the functional change in the preliminary
analysis. In addition, chi-square model difference tests indicated that there was no
significant difference between the model without the covariance and a model with
the covariance ( χ
2
difference
= .003, df
differences
= 1, p = .95); the former was more
parsimonious and selected.

40

Testing the hypothesized model revealed that it fit excellently into the data
( χ
2
(2) = .320, CFI = 1.000, NFI = .999, RMSEA = .000). The means of the initial
level and that of the rate of change were both significant in Table 4. Thus, the data
supported the first hypothesis, suggesting that consumer psychosocial functioning
was improved over the 12 months regardless of their symptom levels when
community-based rehabilitation was received.

Table 4: Estimated means and variances of initial level and of rate of change in
psychosocial functioning

Estimate S.E. C.R.
Means
Initial level

7.791***

1.647

4.731
Rate of change .454* .227 2.002
Variances
Initial level

10.446***

1.665

6.275
Rate of change .057** .017 3.256
** p < .05, *** p < .001

5.3.2 Hypothesis 2: Change in psychiatric symptoms over 12 months
An unconditional latent growth curve model of change in psychiatric
symptoms was examined by analyzing three repeated measurements of BPRS.
Correlations among the repeated measurements and observed means of BPRS at each
time point are presented in Table 3. Values of observed means decreased over time,
suggesting that consumer psychiatric symptoms over the 12-month rehabilitation
period decreased over time (Hypothesis 2 in the current study). This research

41

hypothesis was supported by the fact that the unconditional latent growth curve
model fit very well into the data ( χ
2
(3) = 1.140, CFI = 1.000, NFI = .980, RMSEA
= .000). Parameter estimations in the model are presented in Table 5, where the mean
of the initial level in BPRS and its variance are significantly different from zero
(both p < .001). The rate of change in BPRS (p < .005) and its variance (p < .001) are
significantly different from zero. As seen in Table 5, the estimate of the rate of
change was -0.249, indicating consumer psychiatric symptoms on average across
subjects in the sample decreased by -0.249 per month over the 12-month
rehabilitation period. This finding indicates that community-based psychosocial
rehabilitation programs are effective in reducing consumer psychiatric symptoms.

Table 5: Estimated means and variances of initial level and of rate of change in
symptoms

Estimate S.E. C.R.
Means
Initial level

39.192***

.855

45.852
Rate of change -0.249* .089 -2.779
Variances
Initial level

58.317***

12.744

4.576
Rate of change .313* .149 2.101
* p < .05, *** p < .001

However, an additional analysis of change in a negative symptom construct
of BPRS revealed that there was no significant decrease over the 12 months. The
estimated mean of the rate of change in the negative symptom construct was not

42

significantly different from zero (C.R. = -1.884, p = .06), although the unconditional
growth curve model of the construct fit the data well ( χ
2
(3) = .796, CFI = 1.000, NFI
= .986, RMSEA = .000).

5.3.3 Hypothesis 3: Change in neurocognition over 12 months
Figure 4 shows the statistical model for Hypothesis 3. Hypothesis 3 indicates
that neurocognition at 12 months will be higher than that at baseline. Zero-order
correlations among five neurocognitive tests at baseline and at 12 months are
presented in Table 6. The hypothesized model fit the date very well ( χ
2
(42) = 48.003,
CFI = .987, NFI = .909, RMSEA = .033, TLI = .983). In addition, this hypothesized
model was significantly different from a competing model in which latent means
were constrained to be the same at baseline and at 12 months ( χ
2
differences
= 6.718, df

differences
= 1, p = .01). The estimated latent mean of neurocognition at 12 months was
1.19 when the mean of neurocognition at baseline was constrained to zero (McArdle
& Woodcock, 1997; Ployhart & Oswald, 2004). This mean difference was
significant (C.R. = 2.575, p = .01), revealing that consumer neurocognition increased
over 12 months in community-based rehabilitations.

43

Neurocogniton T1
Vigilance T1
Immedeiate
memory T1
Perseverative
errors T1
Secondary
memory T1
e5 e4
e3
e2
.54
.81
Verbal
fluency T1
e1
.63
.58
.55
Neurocogniton T3
Vigilance T3
Immedeiate
memory T3
Perseverative
errors T3
Secondary
memory T3
e10 e9
e8
e7
Verbal
fluency T3
e6
.57
.83
.66
.61
.58
.58
.56
.40
.61
.57
.93

Figure 4: Hypothesized model of latent mean differences between neurocognition at
baseline and at 12 months. Note: all values are standardized coefficients

44

Table 6: Zero-order correlations among five neurocognitive tests at baseline and at
12 months

Variables 1 2 3 4 5
At baseline
1. Secondary memory T1

-

2. Vigilance T1 .29 -
3. Verbal fluency T1 .33 .29 -
4. Immediate memory T1 .38 .36 .55 -
5. Perseverative errors T1 .31 .36 .29 .52 -
At 12 months
1. Secondary memory T3

-

2. Vigilance T3 .42 -
3. Verbal fluency T3 .34 .34 -
4. Immediate memory T3 .41 .48 .55 -
5. Perseverative errors T3 .24 .46 .29 .45 -
Note:
T1 = baseline; T3 = 12 months

Two factor scores of neurocognition at baseline and at 12 months were
estimated using the regression imputation method in Amos 7 (Arbuckle, 2006). This
regression imputation method generated the two factor scores based on the model of
Hypothesis 3 in Figure 4, assuming that the population means and covariances of all
variables in the model were equal to their maximum likelihood estimates. After
producing factor scores of neurocognition, a group variable was created by
subtracting individual factor scores of neurocognition at baseline from those of
neurocognition at 12 months. This variable indicates change scores of
neurocognition between baseline and 12 months. Individuals whose change score
was greater than zero were defined as neurocognition improvers (henceforth

45

“improvers”), while those whose change score was zero or less than zero were
defined as neurocognition non-improvers (henceforth “non-improvers”). The
proportion of these two groups is presented in Figure 5. Seventy-six subjects
(58.46 %) showed neurocognitive improvement over 12 months, while 54 subjects
(41.54 %) did not. Comparisons of improvers and non-improvers in terms of study
variables are presented in Table 7. Only three variables (length of education, baseline
psychosocial functioning and baseline social cognition) indicate significant mean
differences between improvers and non-improvers. Baseline characteristics of
treatment intensity (days of treatments), symptomatology and medication usage were
not associated among the two groups (i.e., neurocognition change over 12 months).

Figure 5: Proportion of cognitive improvers and cognitive non-improvers

46

Table 7: Comparisons of neurocognitive improvers and non-improvers at baseline
Improvers in
neurocognition
(n = 76)
Non-improvers
in neurocognition
(n = 54)

df
5

F p

Age(year) 36.85 39.57 129 2.19 .090
Education (year) 12.16 11.50 124 4.21 .042
Length of illness
(year)
12.86 15.62 122 2.27 .134
Age of onset (old) 23.98 23.80 122 .01 .910
Psychosocial
functioning
1

8.72 7.61 129 3.14 .079
Neurocognition
(factor score)
2.03 -3.26 129 37.25 .000
Social cognition
2
39.86 34.80 121 8.87 .004
Symptomotogy
3
38.34 40.55 129 1.48 .225
Social contact
4
1.92 1.89 129 .02 .895
Days of treatments 90.68 85.33 129 .29 .591
Days of medication
in previous 6 months
140.64 156.13 128 2.05 .155
Note:

1
Role Functioning Scale; total of social, work and independence subscales.
2
Social cognition; visual and auditory perception of emotion
3
Expanded Brief Psychiatric Rating Scale
4
Social contact scores on the Strauss and Carpenter Outcome Scale
5
df: total of degree of freedom, which is a sum of between-group df (1) and within-
group df, which is varied due to missing in variables.

47

5.3.4 Hypothesis 4-1: Relationships between change in psychosocial functioning
and psychiatric symptoms

Hypothesis 4-1 is that change in psychiatric symptoms will predict change in
psychosocial functioning. This hypothesis was modeled using a bivariate latent
growth curve model (Bollen & Curran, 2006) in which two growth factors (i.e.,
initial level and rate of change) in symptom change (BPRS) were specified to predict
the initial level and the rate of change in functioning change (RFS). This model is
presented in Figure 6. A path of interest in terms of Hypothesis 4-1 was a path from
the rate of change in the symptom change to the rate of change in the functional
change.
The hypothesized model fit the data well ( χ
2
(13) = 6.526, CFI = 1.000, NFI
= .976, RMSEA = .000). The initial level in the symptom change significantly
predicted the initial level of functioning improvements ( β = -0.148, C.R. = -2.947, p
= .003) as well as the rate of change in the improvements ( β = -0.015, C.R. = -2.927,
p = .003). However, the rate of change in the symptom change did not significantly
predict the rate of change in functioning improvements ( β = -0.148, C.R. = -2.947,
p = .003). Therefore, Hypothesis 4-1 was not supported.

48

Rate of change
F
Initial level
F
Functioning T1 Functioning T2 Functioning T3
e1 e2
e3
Di
Ds
Rate of change
S
Initial level
S
Symptom T1 Symptom T2 Symptom T3
e6 e5
e4

Figure 6: Bivariate change model of the effect of symptom change on functional
change

Note:
1. A measure of “functioning” was the Role Functioning Scale; a measure of
“symptom” was the BPRS
2. T1 = baseline; T2 = six months; T3 = 12 months

49

5.3.5 Hypothesis 4-2: Relationships among change in neurocognition, change in
psychiatric symptoms and change in psychosocial functioning

Hypothesis 4-2 stated that change in neurocognition will be associated with
change in psychosocial functioning controlling for change in psychiatric symptoms.
This hypothesis was examined by testing whether the model in Figure 3 (change in
psychosocial functioning with a time-varying covariate of psychiatric symptoms)
fitted simultaneously into the two sub-samples of neurocognition change (improvers
and non-improvers). This analysis was conducted using multiple group latent growth
curve analyses (Bollen & Curran, 2006) in which three nested models were
compared.
The first model was a model with free parameters on the time-varying effect
of psychiatric symptoms on functioning change (“no constraint model”), which
suggests that change in neurocognition influenced the relationship between change in
psychosocial functioning and change in psychiatric symptoms. The second model
was a growth curve model of the psychosocial functioning change in which time-
varying effects of psychiatric symptoms on the functioning change were constrained
to be equal between the group of improvers and that of non-improvers (i.e., the
model of Hypothesis 4-2). The last model was equally constrained on two growth
factor means between improvers and non-improvers (“equal mean constraint model”)
in addition to the equality constraints on time-varying effects of psychiatric
symptoms on functioning change in the second model.

50

Model comparison results are presented in Table 8. The hypothesized model
was not statistically different from the unconstrainted model in chi-square difference
tests, resulting in the selection of the hypothesized model over the unconstrainted
model based on the principle of parsimony (Kline, 2005). Furthermore, the
hypothesized model fit significantly better into the data than the equal mean
constraint model (p = .003). Therefore, the hypothesized model was chosen over the
other two models. Table 9 shows estimated parameters in the hypothesized model.
The estimated mean of the rate of change in improvement of psychosocial
functioning was significantly different from zero in a group of neurocognitive
improvers (estimate = .496, C.R. = 2.188, p = .029), while the estimated mean of the
rate of change was not significantly different from zero in a group of neurocognitive
non-improvers (estimate = .363, C.R. = 1.481, p = .139). The estimated mean of the
initial level of functional change was significant with both groups. Individuals whose
neurocognition increased over 12 months in community-based psychosocial
rehabilitation obtained significant improvement in their psychosocial functioning,
but those whose neurocognition did not increase did not show significant
improvement in psychosocial functioning. In addition, functional improvement was
not affected by change in symptomatology. These results demonstrate that
neurocognition is an important driving force for psychosocial functioning
improvement in community-based psychosocial rehabilitation programs.

51

Table 8: Model comparison test results

Models χ
2
df CFI NFI RMSEA
1. No constraint model 3.128 7 1.000 .988 .000
2. Hypothesized model 3.961 10 1.000 .985 .000
3. Equal mean constraint model 15.410 12 .985 .942 .047

Model differences
between

∆ χ
2

∆df

p

Model 1 & 2 .833 3 .842
Model 2 & 3 11.449 2 .003
Note:
1. CFI = comparative fit index; NFI = normed fit index; RMSEA = root-mean-square
error of approximation.
2. ∆ χ
2
= differences in chi-squares of models; ∆df = differences in degree of
freedom of models.

Table 9: Estimated means and variances of multiple group growth curve model of
change in psychosocial functioning controlling for change in symptoms

Neurocognitive improvers Neurocognitive non-improvers
Estimate S.E. C.R. Estimate S.E. C.R.
Means
Initial level

8.291***

1.648

5.030

7.101***

1.759

4.037
Rate of
change
.496* .227 2.188 .363 .245 1.481
Variances
Initial level

11.021***

2.340

4.710

7.872***

2.161

3.642
Rate of
change
.059* .019 3.149 .025 .016 1.613
* p < .05, *** p < .001

52

5.3.6 Hypothesis 5: Latent trajectory classes of change in psychosocial functioning

Hypothesis 5 was that there will be more than one latent trajectory class of
change in psychosocial functioning, and the effects of service intensity on
psychosocial functioning will be varied within each of the latent classes. An
unconditional latent growth curve model of RFS fit excellently into the data, which
was previously noted in the preliminary analyses. Four growth mixture models (from
a model with one latent class to a model with four classes) were specified using the
unconditional growth curve model, and all four growth mixture models were
compared. Estimated latent growth curves for each of the four models are presented
in Figure 7 (7a to 7d).

53

Figure 7a: Estimated mean growth curve in a model with single latent class
Note:
Dotted lines indicate observed individual growth curves, and solid color lines
indicate estimated mean growth curves for each class.

54

Figure 7b: Estimated mean growth curves in a model with two latent classes
Note:
Dotted lines indicate observed individual growth curves, and solid color lines
indicate estimated mean growth curves for each class.

55

Figure 7c: Estimated mean growth curves in a model with three latent classes
Note:
Dotted lines indicate observed individual growth curves, and solid color lines
indicate estimated mean growth curves for each class.

56

Figure 7d: Estimated mean growth curves in a model with four latent classes
Note:
Dotted lines indicate observed individual growth curves, and solid color lines
indicate estimated mean growth curves for each class.

57

As seen in Table 10, the model with the two latent classes (GMM
2
) fits the
data best, as the Lo-Mendell-Rubin Likelihood Ratio Test (LMR LRT) indicates that
only the two-class model was significantly different from the model with one class (p
= .0003).

Table 10: Test statistics for growth mixture models
Models
Free
parameters
Log
likelihood
BIC Entropy
LMR
LRT
VLMR
LRT
GMM
1
7 -884.739 1803.552 NA NA NA
GMM
2
10 -870.289 1789.253 0.88 0.0003 0.0002
GMM
3
13 -855.549 1774.377 0.84 0.51 0.49
GMM
4
16 -854.756 1787.394 0.85 0.07 0.07
GMMWC
1
10 -883.125 1814.926 NA NA NA
GMMWC
2
15 -864.090 1801.194 0.83 0.05 0.04
GMMWC
3
20 -849.550 1796.451 0.83 0.13 0.12
GGMM
2
17 -856.154 1795.056 0.85 0.03 0.03
Note:
BIC = Bayesian Information Criterion; LMR LRT = Lo-Mendell-Rubin Likelihood
Ratio Test for k (H
0
) versus k-1 classes; VLMR LRT = Vuong-Lo-Mendell-Rubin
Likelihood Ratio Test for k (H
0
) versus k-1 classes; GMM
k
= k-classes Growth
Mixture Model; GMMWC
k
= k-classes Growth Mixture Model with a covariate;
GGMM
k
= k-classes General Growth Mixture Model with a covariate and predictors;
NA = not applicable for the analysis.

In addition, the two-class model had the highest entropy value (0.88) among
the four models, and BIC values dropped sharply from the model with one class
compared to the model with two classes, indicating that the model with two classes
was the best fit model for the data. The average latent class probabilities for the most

58

likely class membership were respectable at 0.928 and 0.982 for class one to class
two, denoting that individuals in the sample were well clustered into the two groups
by the two-class model. In the two-class model, the first latent class with 107
subjects (82%) showed a significant average rate of change (Estimate = .172, z =
4.959, p < .01), while the second class with 23 subjects (18%) did not have a
significant average rate of change (Table 11).

Table 11: Estimated means in two-class growth mixture model of the change in
psychosocial functioning

Latent class one (n=107) Latent class two (n=23)
Estimate S.E. z Estimate S.E. z
Means
Initial level

7.011***

.238

29.47

14.300***

.674

21.20
Rate of change .172** .035 4.96 .124 .086 1.44
** p < .01, *** p < .001

A growth mixture model using a growth curve model with the covariate of
amount of treatment (i.e., service intensity) identified latent classes underlying the
change in functional outcomes. A growth mixture model with two classes (Growth
Mixture Model with a covariate; GMMWC
2
) was the best fit model (see Figure 8).
The data favored GMMWC
2
over GMMWC
3
(i.e., three-class model) in that
Bayesian Information Criterion (BIC) values dropped sharply from the model with
one class to the model with two classes (Table 10). Furthermore, LMR LRT indicate

59

that only the two-class model was significantly different from the model with one
class (p = .05), and the Vuong-Lo-Mendell-Rubin Likelihood Ratio Test (VLMR
LRT) also indicates that only the two-class model was significantly different from
the model with one class (p = .04). The average latent class probabilities for the most
likely class membership were respectable at 0.907 and 0.971 for class one to class
two.

Figure 8: Estimated means and sample means in the two-class growth mixture model
of the functional change with a covariate of service intensity

Estimated means and sample means of the identified two latent classes are
plotted in Figure 8. Thirty-two subjects of the total sample (24.5%) belonged to
latent class one and 98 (75.5%) belonged to class two. The first class indicates that
the amount of treatment significantly affects the initial level (β = -.039, z = -3.630, p
< .001) and rate of change (β = 0.003, z = 3.006, p < .01) in functional

60

improvements over 12 months. However, there were no significant effects of the
amount of treatment on the rate of change in functional improvements in the second
class. These results are presented in Table 12.

Table 12: Significance test of the effect of service intensity on the change in
psychosocial functioning in two-class growth mixture model

Latent class one
(n=32)
Latent class two
(n=98)

β
S.E. z
β
S.E. z
Effects of
service intensity on
Initial level

-.039***

.011

-3.63

-.001

.005

-.20
Rate of change .003** .001 3.01 .000 .000 -.06
** p < .01, *** p < .001

In sum, growth mixture model analyses support the fifth hypothesis in that
there is more than one latent trajectory class of change in psychosocial functioning,
and the effects of service intensity on psychosocial functioning vary within each of
the latent classes. The mixture analyses revealed that there are two latent trajectory
classes of psychosocial functioning change in which the effects of service intensity
on the rate of change of functional improvement over 12 months vary across the two
classes.

61

5.4 Exploratory Analyses of Predictors of Latent Trajectory Classes

5.4.1 General Growth Mixture Model
Potential predictors of the two-class growth mixture model of functional
change, which was presented in the previous chapter, are explored using three
statistical tests: a General Growth Mixture model (GGMM), means difference tests
and a logistic regression. Figure 9 shows the GGMM
2
in which a group of
neurocognition (i.e., improvers and non-improvers) and symptomatology at baseline
is specified to predict class memberships of the two-class growth mixture models of
psychosocial functioning change over 12 months with a covariate of service intensity.
The GGMM
2
was significantly better than the one-class growth mixture model (p
= .029 at LMR LRT presented in Table 10). The group of neurocognition change
significantly predicted class memberships (β = 1.605, z = 2.873, p < 0.05),
indicating that individuals with increases in neurocognition were about five times
more likely to belong to the first class with functional change controlling for service
intensity compared with the second class with non-significant functional change
controlling for service intensity (odds ratio [OR] = 4.978; 95 % confidence interval
[CI] = 1.666 – 14.877). Contrariwise, the level of symptomatology did not
significantly predict class memberships (β = -.049, z = -1.711, p > .05).

62

Rate of change
Initial level
Functioning T1 Functioning T2
Functioning T3
e1 e2
e3
0
6
12
1
1
1
1 1
1
Latent
class
Group of
neurocognition change
Baseline
symptomatology
Service intensity
Di
1
Dr
1

Figure 9: General Growth Mixture Model of functional change with a covariate of
service intensity

5.4.2 Mean Difference Tests
Sample characteristics at baseline were tested to determine differences in the
two latent classes of the two-class growth mixture model of functional change using
means difference tests. Results are presented in Table 13, which shows significant

63

means differences between the two latent classes in age, years of education,
psychosocial functioning level, neurocognition, social cognition, symptomatology
and degree of social contacts. Individuals in the first class with significant functional
change are younger by about five years, more educated, have higher psychosocial
functioning, better neurocognition, better social cognition, less symptom levels and
more social contacts than those in the second class with non-significant functional
change.

Table 13: Mean differences of individual baseline characteristics by latent trajectory
classes of functional change

Class 1
(n = 32)
Class 2
(n = 98)
df
5
F p
Age (year) 34.75 39.04 129 5.660 .019
Education (year) 12.53 11.67 124 5.662 .019
Length of illness (year) 12.66 14.45 122 .759 .385
Age of onset (old) 22.09 24.54 122 1.798 .182
Psychosocial functioning
1
9.45 6.81 129 25.298 .000
Neurocognition 1.80 -2.39 129 19.808 .000
Social cognition
2
38.30 34.61 121 7.570 .007
Symptomotogy
3
32.41 38.39 129 6.065 .015
Social contact
4
2.53 1.70 129 9.500 .003
Days of treatments 93.25 86.90 129 .312 .577
Days of medication
in previous six months
146.68 146.62 128 .000 .996
Note:

1
Role Functioning Scale; total of social, work, and independence subscales.
2
Social cognition; visual and auditory perception of emotion
3
Expanded Brief Psychiatric Rating Scale
4
Social contact scores on the Strauss and Carpenter Outcome Scale
5
df: total of degree of freedom, which is a sum of between-group df (1) and within-
group df, which is varied due to missing variables.

64

Potential predictors of latent class memberships in the two-class growth mixture
model of functional change were explored using a series of logistic regressions.
Zero-order Pearson correlations among the study variables are presented in Table 14.

Table 14: Zero-order correlations of study variables and estimated latent trajectory classes

Variables 1 2 3 4 5 6 7 8 9 10 11 12
1. Trajectory class
-
2. Age
-.21* -
3. Education
.21* .17 -
4. Length of illness
-.08 .56** -.02 -
5. Age of onset
-.12 .38** .19* -.55** -
6. Functioning
1

.41** -.04 .25** -.08 .07 -
7. Neurocognition
.37** -.14 .30** -.06 -.06 .24** -
8. Social cognition
2

.24** -.09 .29** -.02 -.04 .34** .55** -
9. Symptomotogy
3
-.21* .03 .05 .16 -.15 -.22* -.11 -.21* -
10. Social contact
4

.26** -.11 -.02 -.10 -.02 .53** .04 .07 -.04 -
11. Days of medication
.05 -.09 .04 -.12 -.02 -.13 .03 -.02 .10 .03 -
12. Days of treatments
.00 -.06 -.06 .08 -.13 .02 -.11 .01 -.07 .01 -.10 -
Note:

p < .05, ** p < .01
1
Role Functioning Scale; total of social, work, and independence subscales
2
Social cognition; visual and auditory perception of emotion
3
Expanded Brief Psychiatric Rating Scale
4
Social contact scores on the Strauss and Carpenter Outcome Scale

5.4.3 Logistic Regression Analyses
A logistic regression, which produced the maximum number of significant
predictors, was selected. Data revealed a logistic model with eight independent

65

variables in which Nagelkerke R
2
indicated that 41.3 % of the variance in the
dependent variable (i.e., latent class memberships) was accounted for by the logistic
model. In addition, the Hosmer and Lemeshow Goodness-of-Fit Test showed that the
logistic model fit well into the data ( χ
2
(8) = 4.534, p = .806). Estimated parameters
of the selected logistic model are presented in Table 15.

Table 15: Exploring potential predictors of latent trajectory classes of functional
change covaried with intensity of treatments (logistic regression results)

B S.E. Wald Df Sig. OR
3
MEF
4

Age -.085 .040 4.449 1 .035 .919 -.012
Gender -.403 .534 .571 1 .450 .668 -.059
Education .279 .160 3.028 1 .082 1.321 .039
Length of illness .045 .036 1.507 1 .220 1.046 .006
Social contact
1
.448 .188 5.681 1 .017 1.566 .063
Neurocognition .177 .065 7.433 1 .006 1.194 .025
Symptomotology
2
-.065 .030 4.592 1 .032 .937 -.009
Day in medication in
previous 6 months
-.002 .004 .141 1 .707 .998 -.000
Note:
1
Social contact scores on the Strauss and Carpenter Outcome Scale
2
Expanded Brief Psychiatric Rating Scale
3
Odds Ratio
4
Marginal effects at means of independent variables

The significant predictors were age, social contact, neurocognition and
symptomatology. Individuals who were younger (OR = .919; 95 % CI = .849 – .994),
had more social contacts (OR = 1.566; 95 % CI = 1.083 – 2.264), had better
neurocognition at baseline (OR = 1.194; 95 % CI = 1.051 – 1.357) and who had less
symptomatology (OR = .937 95 % CI = .884 – .994) were more likely to be in the

66

first class with significant functional change related with service intensity. Gender,
education, length of illness and days of medication in the previous six months did not
significantly predict the first class. The marginal effects of predicators are presented
in column eight in Table 15, indicating the percent change of the probability of being
the first class with significant functional change related with service intensity. For
example, an individual who is younger by one year has a 1.2 % increase in the
probability of being the first class.

67

CHAPTER VI: DISCUSSION

6.1 Summary of Findings
This dissertation study examined the relationships among change in
psychosocial functioning, change in neurocognition and change in psychiatric
symptoms. Six research hypotheses were tested using a latent means difference test
and latent growth curve analyses. In addition, exploratory analyses were conducted
using growth mixture analyses, observed means difference tests and logistic
regression analyses.
Major findings are summarized as follows:
1. Psychosocial functioning on average improved significantly over 12
months across consumers when community-based rehabilitation was received, and
there were large interpersonal variations in functional change (results of preliminary
analyses).
2. Psychosocial functioning improved over 12 months regardless of
psychiatric symptom level difference at baseline, six months and 12 months (result
of testing Hypothesis 1).
3. Consumers’ overall psychiatric symptoms significantly decreased over the
12-month rehabilitation period (result of testing Hypothesis 2). However, negative
symptoms did not decrease over the 12 months.

68

4. Consumers’ neurocognition in the 12-month rehabilitation period
improved (result of testing Hypothesis 3). When subjects were grouped into
improvers and non-improvers using change scores of neurocognition, 76 subjects
(58.46 %) showed neurocognitive improvement over 12 months, while 54 subjects
(41.54 %) did not. Furthermore, three variables among 13 study variables (presented
in Table 2) (i.e., length of education, baseline psychosocial functioning and baseline
social cognition) indicated significant means differences between improvers and
non-improvers.
5. The rate of change of psychiatric symptoms decreases over the 12-month
rehabilitation period but did not significantly predict the rate of change in
functioning improvements. This suggests that the current study data does not
support Hypothesis 4-1 regarding change in psychiatric symptoms but will predict
change in psychosocial functioning.
6. Individuals whose neurocognition increased over 12 months in
community-based psychosocial rehabilitation showed significant improvement in
functional change controlling for change in psychiatric symptoms, but those whose
neurocognition did not increase did not show significant improvement in functional
change (result of testing Hypothesis 4-2).
7. A growth mixture model generated two latent trajectory classes of
psychosocial functional change over the 12 months. The first latent class with 107
subjects (82%) showed a significant average rate of change, while the second class

69

with 23 subjects (18%) did not have a significant average rate of change. A growth
mixture model with a covariate of service intensity also produced two latent
trajectory classes of functional change. However, when adding the covariate, it
yielded only 32 subjects (24.5 %), which were classified into the first group in which
service intensity significantly and positively predicted the rate of change in
functional improvement. On the contrary, for 98 subjects (74.5 %) belonging to the
second class, service intensity did not predict significantly the rate of change in
functional change (result of testing Hypothesis 5). Furthermore, there was no
significant improvement in psychosocial functioning in the second class.
8. There are three novel findings from the exploratory analyses.
First, a general growth mixture model revealed that individuals whose
neurocognition increased were much more likely to belong to a latent trajectory class
(odds ratio = 4.978) with significant functional improvement controlling for service
intensity compared with the other latent trajectory class with non-significant
functional improvement controlling for service intensity. Second, the two trajectory
classes showed significant mean differences in the following baseline sample
characteristics: age, years of education, psychosocial functioning, neurocognition,
social cognition, symptomatology and degree of social contacts. Individuals in the
first class with significant functional change were younger by about five years, more
educated, and had higher psychosocial functioning, better neurocognition, better
social cognition, lower symptom levels and more social contacts than those in the

70

second class with non-significant functional change. Last, logistic regression
analyses indicated that individuals at baseline who were younger, had more social
contacts, had better neurocognition and who had less symptomatology were more
likely to be in the first class with significant functional change related with service
intensity. Gender, education, length of illness and days of medication in the previous
six months, however, did not predict significantly the first class.

6.2 Overall Discussion

6.2.1 Restorative Model and Compensatory Model in Cognitive Rehabilitation
Treatments

This dissertation study investigated indirectly two rival models in cognitive
rehabilitation treatments (CRT): the restorative model and the compensatory model.
It was examined whether neurocognition was improved in community-based
rehabilitation programs (Hypothesis 3). Findings indicate that there was a
statistically significant mean difference in the latent variable of neurocognition,
which was constructed using five measures: 1) secondary memory (Controlled Oral
Word Association Test); 2) vigilance (Digit Span Distractibility Test); 3) verbal
fluency (California Verbal Learning Test); 4) immediate memory (Degraded-
Stimulus Continuous Performance Test); and 5) perseverative errors (Wisconsin
Card Sorting Test). These findings indicate that neurocognition can be improved
given community-based psychosocial rehabilitation, suggesting that the current study

71

data favor the restorative model in cognitive rehabilitation treatments (CRT) over the
compensatory model.
The restorative model and the compensatory model pose different stances on
the assumptions of CRT. CRT has two key assumptions of how CRT works for
individuals diagnosed with schizophrenia. The first assumption is that cognitive
deficits are crucial factors in psychosocial functioning. Neurocognitive research in
schizophrenia increasingly demonstrates the evidence of the relationship between
many neurobiological measures and social functioning, which support the first
assumption. The second assumption is that cognitive deficits must be reduced to
achieve effective rehabilitation. This second assumption differentiates the two rival
models: The restorative model focuses on remediating observed impairments in
cognitive functioning, while the compensatory model emphasizes bypassing
(compensating for) the impairments.
From a theoretical point of view, the second assumption of the cognitive
rehabilitation strategy is very clear, but empirical testing of the assumption is rather
difficult as it is hard to make an obvious conclusion. It is not clear that a desired
outcome is achieved either by the remedy of an intended cognitive deficit or by the
adaptation of functioning, when a particular cognitive intervention is successful
(Bellack & Muser, 1993). Consequently, there have been inconsistent research
findings regarding the second assumption. For example, two studies (Bellack, Muser,
Morrison, & Tierney, 1990; Green, Satz, Ganzell, & Vaclav, 1992) indicated that

72

performance on the Wisconsin Card Sorting Test, a neurocognitive measure of
executive functioning, could be improved through repeated practice, such as
instructional modifications. These studies, thus, support the remediation model.
However, meta-analytic review by Kurtz and colleagues (2001) has concluded that
Wisconsin Card Sorting Test performance cannot be remediated. In a randomized
study conducted by Benedict and colleagues (1994), of 38 outpatients with chronic
schizophrenia assigned to either attentional training in which subjects were given 15
hours of repeated practice on computer-administered vigilance tasks of graduated
difficulty or to a no-treatment control group, findings indicate that the teaching of
compensation strategies may have more impact on cognitive deficits than repeated
practice approaches. In addition to the inconsistent findings of treatment effect on
cognitive deficits, a lack of research exists in whether CRT is effective in
community-based rehabilitation programs as it is difficult to examine using well-
controlled small sample studies.
This current study concludes that neurocognition can be improved given
community-based psychosocial rehabilitation. However, this conclusion should be
interpreted considering the following: Only 76 (58.46 %) out of 130 subjects showed
neurocognitive improvement. Post hoc tests of observed mean differences showed
that secondary memory and verbal fluency among five neurocognitive measures
improved over 12 months. Vigilance, immediate memory and perseverative errors
(Wisconsin Card Sorting Test) did not show significant mean differences between

73

baseline scores and scores at 12 months. These findings suggest that improvements
in neurocognition might happen only to subgroups of individuals; particularly, the
domain of neurocognition can be improved only given specific cognitive
rehabilitation training (e.g., training for increasing perseverative errors). It should be
noted that the subjects in the current study were not assigned to specific cognitive
rehabilitation training. Consequently, it can be concluded that improvements in
neurocognition might be generated to some degree without curing impairments in
neurocognition for people with schizophrenia participating in community-based
rehabilitation programs.

6.2.2 Relationships Among Change in Psychosocial Functioning, Change in
Neurocognition and Change in Psychiatric Symptoms

There were significant improvements in psychosocial functions, and this
functional improvement was not affected by changes in psychiatric symptoms
(Hypothesis 1 and Hypothesis 4-1), although overall psychiatric symptoms decreased
over the 12-month rehabilitation period (Hypothesis 2). These results have been
corroborated in previous studies (e.g., Brekke & Long, 2000; Green et al., 2004). In
Brekke and Long (2000), longitudinal P-technique factor analysis revealed that there
were three outcome domains in longitudinal recovery in schizophrenia, which are the
clinical (BPRS scores), functional (RFS) and subjective experience domains, and
that the correlations among the three domains were weak (the relationships between

74

clinical change and functional change were weakest), suggesting that the three
domains might have somewhat distinct processes of recovery during rehabilitation.
Their findings are supported by the findings in this present study.
It should be noted that this dissertation study did not aim to test whether
interventions of a cognitive rehabilitation can reduce cognitive deficits. Rather this
study aimed to examine whether cognitive deficits must be cured to achieve
improvement in psychosocial functioning, which is the essential argument of the
second assumption of cognitive rehabilitation treatments. Therefore, the most
interesting finding was produced by testing Hypothesis 4-2, which sought to examine
whether change in neurocognition was associated with change in psychosocial
functioning controlling for change in psychiatric symptoms. This was analyzed using
multiple group growth curve analyses. As seen in Table 9, there is significant
improvement in psychosocial functioning for a group of neurocognitive improvers,
but there is no significant functional improvement for a group of neurocognitive non-
improvers. In other words, individuals whose neurocognition increased over 12
months in community-based psychosocial rehabilitation obtained significant
improvement in functional change controlling for change in psychiatric symptoms,
but those whose neurocognition did not increase did not show significant
improvement in functional change. This novel finding indicates that cognitive
deficits must be reduced to achieve improvement in psychosocial functioning,
suggesting that the restorative strategy would be better than the compensatory

75

strategy in applying cognitive rehabilitation treatments in community-based
rehabilitation programs. Furthermore, a growth mixture model with two classes
revealed that individuals with increased neurocognition were about five times more
likely to belong to the significant functional improvement class controlling for
service intensity compared with the non-significant functional change class (odds
ratio [OR] = 4.978; 95 % confidence interval [CI] = 1.666 – 14.877). This finding
provides additional support for the application of the restorative model to
community-based rehabilitation.

6.2.3 Who Changes and Who Does Not During Community-Based Psychosocial
Rehabilitation?

A growth mixture model generated two latent trajectory classes of
psychosocial functional change over the 12-month period. The first latent class with
107 subjects (82%) showed a significant average rate of change, while the second
class with 23 subjects (18%) did not have a significant average rate of change
(Figure 7b). When service intensity was added to the two-class growth mixture
model, only 32 subjects (24.5 %) were classified into the first group in which service
intensity was significantly and positively related with the rate of change in functional
improvement (result of testing Hypothesis 5). On the contrary, 98 subjects (74.5 %)
belonged to the second class in which service intensity did not significantly predict
the rate of change in functional change. These results show that there are latent sub-

76

groups that respond better to community-based psychosocial rehabilitation. In
general, this rehabilitation produces consumers’ functional improvements (107
subjects showed functional improvements), but service intensity was related with
only 32 consumers in their functional improvements. Thus, the functional
improvement in 75 consumers was not related with the amount of treatments
received. These improvements might be explained by other factors, which were not
investigated in this current study.
In exploratory analyses, potential predictors of functional improvements
related with service intensity were found: Individuals at baseline who were younger,
had more social contacts, had better neurocognition and who had less
symptomatology were more likely to be in a latent subgroup in which significant
functional improvement was related with service intensity (as presented in Table 15).
This interesting finding suggests a better understanding of whose psychosocial
functioning can be improved in what conditions. That is, it is concluded that
consumers participating in community-based psychosocial rehabilitation are more
likely to achieve functional improvement if they are younger, have more social
contacts, have better neurocognition and show less symptomatology as the amount of
treatments increase over 12 months. However, gender, education, length of illness
and days of medication in the previous six months were not related to the
relationship between amount of treatments and functional improvements.

77

6.3 Implications
6.3.1 Implications for Practice
Social workers in community-based psychosocial rehabilitation should
understand the mechanisms of treatment effects and recovery in the rehabilitation of
schizophrenia. The clinical realities of the longitudinal nature of schizophrenia entail
the mechanisms of treatment effects and recovery in the rehabilitation of
schizophrenia (Peer, Kupper, Long, Brekke, & Spaulding, 2007). These longitudinal
mechanisms of treatment effects and recovery in the rehabilitation of schizophrenia
are sine qua non in social work practice theories, which in turn should be embedded
in social work interventions. Findings from this dissertation study demonstrate three
treatment mechanisms in the rehabilitation of schizophrenia, denoting that
community-based psychosocial rehabilitation programs for people diagnosed with
schizophrenia can improve consumers’ psychosocial functioning. The three
treatment mechanisms are as follows: 1) Psychosocial functioning can be improved
through increases in neurocognition (i.e., a therapeutic mediator); 2) Improvement in
psychosocial functioning is not related with change in psychiatric symptoms; and 3)
Effects of service intensity on functional improvement are more distinctive in
individuals who are younger, have more social contacts, have better neurocognition
and show less symptomatology.
These treatment mechanisms suggest two interesting clinical implications for
community-based psychosocial rehabilitation. First, clinicians and administrators

78

working with people diagnosed with schizophrenia need to use cognitive remediation
interventions actively to improve consumers’ psychosocial functioning. For example,
cognitive remediation therapy in schizophrenia is effective in improving working
memory and cognitive flexibility, which in turn are associated with social
functioning improvements (Wykes et al., 2007). Another example is cognitive
enhancement therapy, which has been shown to be effective on neurocognition,
processing speed

composites and social cognition (Hogarty et al., 2004). These
treatments can be beneficial to consumers in community-based psychosocial
rehabilitation. Second, clinicians and administrators need to understand person-
environment interactions, which are denoted by heterogeneous unobserved
subgroups of individuals regarding treatments. This understanding may lead to
individually tailored interventions, which can maximize treatment effects and
minimize treatment resources. For example, community-based rehabilitation for
people diagnosed with schizophrenia can improve psychosocial functioning through
increasing service intensity (more days of treatment) only for consumers who are
younger, have more social contacts, have better neurocognition and show less
symptomatology. Thus, focus needs to be placed on other service characteristics
rather than service intensity to improve psychosocial functioning in consumers who
are older, have less social contacts, have lower neurocognitive ability and have
severe symptoms.

79

The current most appealing intervention model in social work practice for
people with schizophrenia is the biopsychosocial model (Drake, Green, Mueser, &
Goldman, 2003), but the social work profession has relegated biological factors to
merely a theoretical background in its practices, overemphasizing the psychological
and social aspects of human behavior (Farmer & Pandurangi, 1997). As Farmer and
Pandurangi (1997) pointed out, social workers in community-based psychosocial
rehabilitation must understand the contributions of the biological revolution and the
neurosciences to the rehabilitation of schizophrenia. As claimed in Saleebey (1992)
and Hogarty (1991), integrating biological knowledge, such as neurocognition, into
social work should be seen not as succumbing to the medical model but rather as
expanding the repertoire of skills for working better with the most difficult client
populations with schizophrenia.

6.3.2 Implications for Social Work Research
It should be noted that the utilization of modern longitudinal methods is
required to address the complexity of biopsychosocial models of treatment and
recovery in schizophrenia (Peer, Kupper, Long, Brekke, & Spaulding, 2007, p. 6).
Peer et al. (2007) point out that the complicated nature of the longitudinal recovery
process in schizophrenia reflects the varying degrees of dysregulation within the
person and between the person and their environments. To understand the
complicated nature of the rehabilitation of schizophrenia, social work research needs

80

more attention on innovative longitudinal data analysis methods. Nonetheless,
empirical studies using innovative longitudinal data analysis methods have not been
fully conducted in social work research (Fraser, 2004), although the innovations
permit social work researchers to understand better the mechanisms of social work
interventions. These longitudinal data analysis methods are cutting-edge so as to be
able to answer the questions of intervention mechanisms of how, when, by what, to
whom and under what conditions do cognitive deficits, symptoms and social
functioning change over time. For example, findings of latent subgroups of clients
will help researchers and clinicians to predict to some degree who will and will not
change given social work interventions.
Latent trajectory subgroups in psychosocial functional change during
rehabilitation were explored using growth mixture models in the present study. The
latent trajectory subgroups are interesting in that: 1) The subgroups capture
heterogeneity in individuals diagnosed with schizophrenia; 2) The subgroups suggest
that clinicians need to differentiate treatment strategies to different subgroups of
individuals to make their intervention more effective; and 3) The subgroups imply
that social work researchers need to pay more attention to interactions between
persons and treatments rather than those between variables and treatments.
A better understanding of interactions between persons and treatments
might give rise to better social work evaluation research in addressing the
effectiveness of social work interventions. It is often hard to conclude whether being

81

unsuccessful is caused by failure in intervention theories or by failure in
implementation of the theories when a social work intervention is evaluated to be
ineffective. For example, an intervention theory might be effective only for one
group of individuals, and a way of implementation might be effective only for one
group of individuals. In this example, there are interactions between persons and
treatments, which are hard to analyze by explaining relationships between variables
(e.g., treatment elements) and variables (e.g., outcomes) as in a variable-centered
approach, such as regression analysis. The interactions between persons and
treatments require a person-centered approach, such as cluster analysis, latent class
analysis and finite mixture analysis (Muthén & Muthén, 2000).
On one hand, variable-centered approaches, which are relatively common in
social work research, are used to explain relationships among variables. Examples
are regression analysis, factor analysis and structural equation modeling. On the
other hand, person-centered approaches aim to explore types of individuals (Horn,
2000) or “to group individuals into categories, each one of which contains
individuals who are similar to each other and different from individuals in other
categories” (Muthén & Muthén, 2000). Thus, person-centered approaches are able to
answer different questions than the more traditional variable-centered approach
(Greenberg, Speltz, Deklyen, & Jones, 2001). However, by maintaining the integrity
of individual cases, these two approaches can be used simultaneously. Horn clearly
explained this point by stating:

82

For any sample of m different variables obtained on a sample of n different
persons, there is an identity transformation between classifications of persons
(types) that are defined with person-centered methods and classifications of
variables (factors) defined with variable-centered methods: types are
manifested in factors and factors expressed in types (2000, p.924).

Growth mixture modeling in this current study, for example, revealed the
relationship between treatment intensity and improvement in psychosocial
functioning as well as how change in neurocognition affects the relationship
(variable-centered approach), while simultaneously exploring latent subgroups of
individuals, which are differentiated in the relationships (person-centered approach).
This innovative finding is more informative for clinicians in community-based
psychosocial interventions compared with possible findings using only either a
variable-centered or person-centered approach.
In conclusion, further social work research needs to apply an integrated
approach of person-centered analyses into variable-centered analyses, which can
increase knowledge of social work interventions and provide better clinical
implications. Growth mixture modeling used in the present study is an integrated
approach of person-centered research and variable-centered research (Muthén &
Muthén, 2000).

6.4 Limitations and Suggestions for Further Studies
The growth mixture model with two classes (Figure 7b) was selected as the
best model in the current study based on the criteria of three indices: 1) BIC; 2)

83

entropy; and 3) Lo-Mendell-Rubin Likelihood Ratio Test. However, the growth
mixture model with three latent classes (Figure 7c) might have better represented the
reality of psychosocial functioning improvements in community-based rehabilitation,
although this model was rejected by the Lo-Mendell-Rubin Likelihood Ratio Test
(Table 10). In addition, the Vuong-Lo-Mendell-Rubin Likelihood Ratio Test
supports the three-class model. A sample of 130 subjects might not have enough
statistical power to produce stable latent class memberships. Consequently, the
findings regarding growth mixture models should not be generalized without the
limitation of the respect of the small sample size of the current study. It is highly
recommended that these findings be replicated with a larger sample size.
The current study had only two repeated measurements of neurocognition,
which limits the testing of the latent growth curve model of change in neurocognition.
Three repeated measurements of neurocognition might permit more rigorous tests of
possible relationships among change in neurocognition, change in psychiatric
symptoms and change in psychosocial functioning. Future research needs to
incorporate theories of directions in the relationship between neurocognitive change
and symptom change. In addition, domain-specific relationships among
neurocognitive change, symptom change and functional change are needed as this
might be most beneficial to community-based psychosocial rehabilitation.
Future research needs to explore potential predictors that might affect
neurocognition as well as affect psychosocial functioning. Medalia and Richardson

84

(2005) investigated predictors of a good response to cognitive remediation
interventions, and reported that treatment intensity, type of cognitive interventions
and therapist qualifications were key predictors. These predictors also might be key
predictors of improvement in psychosocial functioning. Thus, integrating
investigations of predictors on cognitive improvement and functional improvements
might lead to a better understanding of treatment mechanisms in community-based
rehabilitation.

85

REFERENCES

Akbarian, S., Bunney, W. E., Potkin, S. G., Wigal, S. B., Hagman, J. O., Sandman, C.
A., et al. (1993). Altered distribution of nicotinamide-adenine dinucleotide
phosphate-diaphorase cells in frontal lobe of schizophrenics implies
disturbances of cortical development. Archives of General Psychiatry, 50(3),
169-177.

Allison, P. D. (2001). Missing data. Thousand Oaks, CA: Sage.

Arbuckle, J. L. (1996). Full information estimation in the presence of incomplete
data. Advanced structural equation modeling: Issues and techniques, 243–
277.

Arbuckle, J. L. (2006). AMOS (Version 7.0) [Computer Program]. Chicago: SPSS.

Arbuckle, J. L. (2006). AMOS 7.0 User's Guide. Chicago: SPSS.

Bell, D. S. (1973). The experimental reproduction of amphetamine psychosis.
Archives of General Psychiatry, 29, 5–40. .

Bellack, A. S., Blanchard, J. J., & Mueser, K. T. (1996). Cue availability and affect
perception in schizophrenia. Schizophrenia Bulletin, 22, 535-544.

Bellack, A. S., Gold, J. M., & Buchanan, R. W. (1999). Cognitive rehabilitation for
schizophrenia: Problems, prospects, and strategies. Schizophrenia Bulletin,
25(2), 257-274.

Bellack, A. S., & Mueser, K. T. (1993). Psychosocial treatment for schizophrenia.
Schizophrenia Bulletin, 19(2), 317-336.

Bellack, A. S., Mueser, K. T., Morrison, R. L., Tierney, A., & Podell, K. (1990).
Remediation of cognitive deficits in schizophrenia. American Journal of
Psychiatry, 147(12), 1650-1655.

Bellack, A. S., Mueser, K. T., Wade, J., Sayers, S., & Morrison, R. (1992). The
ability of schizophrenics to perceive and cope with negative affect. British
Journal of Psychiatry, 160, 473-480.

86

Benedict, R. H., Harris, A. E., Markow, T., McCormick, J. A., Nuechterlein, K. H.,
& Asarnow, R. F. (1994). Effects of attention training on information
processing in schizophrenia. Schizophr Bull, 20(3), 537-546.

Benes, F. M. (1998). Model generation and testing to probe neural circuitry in the
cingulate cortex of postmortem schizophrenic brain. Schizophrenia Bulletin,
24(2), 219-230.

Bilder, R. M., Mukherjee, S., Rieder, R. O., & Pandurangi, A. K. (1985).
Symptomatic and neuropsychological components of defect states.
Schizophrenia Bulletin, 11(3), 409-419.

Bollen, K. A., & Curran, P. J. (2006). Latent curve models: A structural equation
perspective. Hoboken, NJ: John Wiley & Sons, Inc.

Brekke, J. S. (2000). Predicting psychosocial rehabilitative service outcomes:
Response to reviewers (No. R-01MH53282): National Institute of Mental
Health.

Brekke, J. S., Green, M. F., Wynn, J., Hoe, M., & Xie, B. (2007). How
neurocognitive and psychophysiological factors impact functional
rehabilitative change in schizophrenia. Schizophrenia Bulletin, 33(2), 584-
584.

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, doi:10.1093/schbul/sbl072 sbl072.

Brekke, J. S., Kay, D. D., Lee, K. S., & Green, M. F. (2005). Biosocial pathways to
functional outcome in schizophrenia. Schizophrenia Research, 80(2-3), 213-
225.

Brekke, J. S., Levin, S., Wolkon, G., Sobel, G., & Slade, B. (1993). Psychosocial
functioning and subjective experience in schizophrenia. Schizophrenia
Bulletin, 19(3), 599-608.

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

87

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
& Clinical Psychology, 65(3), 464-475.

Brekke, J. S., Raine, A., Ansel, M., Lencz, T., & Bird, L. (1997).
Neuropsychological and psychophysiological correlates of psychosocial
functioning in schizophrenia. Schizophrenia Bulletin, 23(1), 19-28.

Brekke, J. S., Raine, A., & Thomson, C. (1995). Cognitive and psychophysiological
correlates of positive, negative, and disorganized symptoms in the
schizophrenia spectrum. Psychiatry Research, 57, 241-250.

Byrne, B. M., & Watkins, D. (2003). The issue of measurement invariance revisited.
Journal of Cross-Cultural Psychology, 34(2), 155-175.

Cannon, T. D., Bearden, C. E., Hollister, J. M., Rosso, I. M., Sanchez, L. E., &
Hadley, T. (2000). Childhood cognitive functioning in schizophrenia patients
and their unaffected siblings: a prospective cohort study. Schizophrenia
Bulletin, 26(2), 379-393.

Chen, F. F., Sousa, K. H., & West, S. G. (2005). Testing measurement invariance of
second-order factor models. Structural Equation Modeling, 12, 471–492.

Delis, D. C., Kramer, J. H., Kaplan, E., & Ober, B. A. (1987). California Verbal
Learning Test (Research ed.). Cleveland, OH: The Psychological Corporation.

Deshon, R. P. (2004). Measures are not invariant across groups without error
variance homogeneity. Psychology Science, 46(1), 137-149.

Drake, R. E., Green, A. I., Mueser, K. T., & Goldman, H. H. (2003). The history of
community mental health treatment and rehabilitation for persons with severe
mental illness. Community Mental Health Journal, 39(5), 427-440.

Duncan, T. E., Duncan, S. C., & Strycker, L. A. (2006). An Introduction to Latent
Variable Growth Curve Modeling: Concepts, Issues, and Applications
(Second ed.). Mahwah, NJ: Lawrence Erlbaum Associates.

Duncan, T. E., Susan, S. C., Strycker, L. A., Okut, H., & Li, F. (2002). Growth
mixture modeling of adolescent alcohol use data. Retrieved from Oregon
Research Institute Web site: http://www.ori.org/methodology.

88

Farmer, R. L., & Pandurangi, A. K. (1997). Diversity in schizophrenia: Toward a
richer biopsychososocial understanding for social work practice. Health and
Social Work, 22(2), 109-116.

Fraser, M. W. (2004). Intervention research in social work: Recent advances and
continuing challenges. Research on Social Work Practice, 14(3), 210-222.
Geddes, J. R., & Lawrie, S. M. (1996). Obstetric complications and schizophrenia: A
meta-analysis. Obstetrical and Gynecological Survey, 51, 392-393.

Gottesman, I. I., & Bertelsen, A. (1989). Confirming unexpressed genotypes for
schizophrenia. Risks in the offspring of Fischer's Danish identical and
fraternal discordant twins. Archives of General Psychiatry, 46(10), 867-872.

Green, M. F. (1993). Cognitive remediation in schizophrenia: Is it time yet?
American Journal of Psychiatry, 150, 178-187.

Green, M. F. (1996). What are the functional consequences of neurocognitive
deficits in schizophrenia. The American Journal of Psychology, 153(3), 321-
330.

Green, M. F., Kern, R. S., & Heaton, R. K. (2004). Longitudinal studies of cognition
and functional outcome in schizophrenia: implications for MATRICS.
Schizophrenia Research, 72(1), 41-51.

Green, M. F., Kern, R. S., Robertson, M. J., Sergi, M. J., & Kee, K. S. (2000).
Relevance of neurocognitive deficits for functional outcome in schizophrenia.
In T. Sharma & P. D. Harvey (Eds.), Cognition in Schizophrenia (pp. 178-
192). New York: Oxford University Press.

Green, M. F., & Nuechterlein, K. H. (1999). Should schizophrenia be treated as a
neurocognitive disorder? Schizophrenia Bulletin, 25(2), 309-319.

Green, M. F., Satz, P., Ganzell, S., & Vaclav, J. F. (1992). Wisconsin Card Sorting
Test performance in schizophrenia: remediation of a stubborn deficit.
American Journal of Psychiatry, 149(1), 62-67.

Green, R. S., & Gracely, E. J. (1987). Selecting a rating scale for evaluating services
to the chronically mentally ill. Community Mental Health Journal, 23, 91-102.

89

Greenberg, M. T., Speltz, M. L., Deklyen, M., & Jones, K. (2001). Correlates of
clinic referral for early conduct problems: Variable-and person-oriented
approaches. Development and Psychopathology, 13(02), 255-276.

Heaton, R. K. (1981). A manual for the Wisconsin Card Sorting Task. Odessa, FL:
Psychological Assessment Resources.

Heaton, R. K., & Pendleton, M. G. (1981). Use of neuropsychological tests to predict
adult patients' everyday functioning. Journal of Consulting and Clinical
Psychology, 49(6), 807-821.

Hogarty, G., E. (1991). Social work practice research on severe mental illness:
Charting a future. Research on social work practice, 1, 5-31.

Hogarty, G. E., & Flesher, S. (1999). Practice principles of cognitive enhancement
therapy for schizophrenia. Schizophrenia Bulletin, 25(4), 693-708.

Hogarty, G. E., Flesher, S., Ulrich, R., Carter, M., Greenwald, D., Pogue-Geile, M.,
et al. (2004). Cognitive enhancement therapy for schizophrenia: effects of a
2-year randomized trial on cognition and behavior. Archives of General
Psychiatry, 61(9), 866-876.

Horn, J. L. (2000). Comments on integrating person-centered and variable-centered
research on problems associated with the use of alcohol. Alcoholism: Clinical
and Experimental Research, 24(6), 924-930.

Horn, J. L., & McArdle, J. J. (1992). A practical and theoretical guide to
measurement invariance in aging research. Experimental Aging Research,
18(3), 117-144.

Joreskog, K. G. (1979). A general approach to confirmatory maximum likelihood
factor analysis. In J. Magidson (Ed.), Advances in factor analysis and
structural equation models (pp. 21-40). Cambridge, Massachusetts: Abt
Books.

Kendell, R. E., Juszczak, E., & Cole, S. K. (1996). Obstetric complications and
schizophrenia: a case control study based on standardised obstetric records.
British Journal of Psychiatry, 168(5), 556-561.

90

Kerr, S. L., & Neale, J. M. (1993). Emotion perception in schizophrenia: Specific
deficit or further evidence of generalized poor performance? Journal of
Abnormal Psychology, 102(2), 312-318.

Kline, R. B. (2005). Principles and practice of structural equation modeling (Second
ed.). New York: Guilford Press.

Kurtz, M. M., Moberg, P. J., Gur, R. C., & Gur, R. E. (2001). Approaches to
cognitive remediation of neuropsychological deficits in schizophrenia: A
review and meta-analysis. Neuropsychology Review, 11(4), 197-210.

Lambert, M., & Naber, D. (2004). Current issues in schizophrenia: overview of
patient acceptability, functioning capacity and quality of life. CNS Drugs,
18(Suppl 2), 5-17.

Lezak, M. D. (1995). Neuropsychological Assessment (3rd ed.). New York: Oxford
University Press.

Li, F., Duncan, T. E., Duncan, S. C., & Hops, H. (2001). Piecewise growth mixture
modeling of adolescent alcohol use data. Structural Equation Modeling, 8(2),
175-204.

Lieberman, J., Chakos, M., Wu, H., Alvir, J., Hoffman, E., Robinson, D., et al.
(2001). Longitudinal study of brain morphology in first episode
schizophrenia. Biological Psychiatry, 49(6), 487-499.

Little, R. J. A., & Rubin, D. B. (2002). Statistical analysis with missing data (2nd
ed.): Wiley-Interscience.

Little, T. D. (1997). Mean and covariance structures (MACS) analysis of cross-
cultural data: practical and theoretical issues. Multivariate Behavioral
Research, 32, 53-79.

Long, J. D., & Brekke, J. S. (1999). Longitudinal factor structure of the Brief
Psychiatric Rating Scale in schizophrenia. Psychological Assessment, 11(4),
498-506.

Lukoff, D., Liberman, R. P., & Nuechterlein, K. H. (1986). Symptom monitoring in
the rehabilitation of schizophrenic patients. Schizophrenia Bulletin, 12(4),
578-602.

91

Lukoff, D., Nuechterlein, K., & Ventura, J. (1986). Manual for the Expanded Brief
Psychiatric Rating Scale. Schizophrenia Bulletin, 12, 594-602.

McArdle, J. J., & Woodcock, R. W. (1997). Expanding test-retest designs to include
developmental time-lag components. Psychological methods, 2(4), 403-435.

McGurk, S. R., & Meltzer, H. Y. (2000). The role of cognition in vocational
functioning in schizophrenia. Schizophrenia Research, 45(3), 175-184.

McLachlan, G., Welham, J., & McGrath, J. (1998). Heterogeneity in schizophrenia:
a mixture model analysis based on age-of-onset, gender and diagnosis.
Schizophrenia Research, 29(1), 25-25.

Medalia, A., & Richardson, R. (2005). What predicts a good response to cognitive
remediation interventions? Schizophrenia Bulletin, 31(4), 942.

Mednick, S. A., Machon, R. A., Huttunen, M. O., & Bonett, D. (1988). Adult
schizophrenia following prenatal exposure to an influenza epidemic. Archives
of General Psychiatry, 45, 189-192.

Mednick, S. A., Parnas, J., & Schulsinger, F. (1987). The Copenhagen high-risk
project, 1962-86. Schizophrenia Bulletin, 13(3), 485-495.

Mortensen, P. B., Pedersen, C. B., Westergaard, T., Wohlfahrt, J., Ewald, H., Mors,
O., et al. (1999). Effects of family history and place and season of birth on
the risk of schizophrenia. The New England Journal of Medicine 340(8), 603-
608.

Mueser, K. T., Bond, G. R., Drake, R. E., & Resnick, S. G. (1998). Models of
community care for severe mental illness: a review of research on case
management. Schizophrenia Bulletin, 24(1), 37-74.

Mueser, K. T., & McGurk, S. R. (2004). Schizophrenia. Lancet, 363(9426), 2063-
2072.

Muthén, B. O. (2001). Second-generation structural equation modeling with a
combination of categorical and continuous latent variables: New
opportunities for latent class/latent growth modeling. New methods for the
analysis of change, 291–322.

92

Muthén, B. O, Brown, C. H., Masyn, K., Jo, B., Khoo, S.-T., Yang, C.-C., et al.
(2002). General growth mixture modeling for randomized preventive
interventions. Biostatistics, 3(4), 459-475.

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

Muthén, B. O. (2002). Beyond SEM: General latent variable modeling.
Behaviormetrika, 29(1), 81–117.

Muthén, L. K., & Muthén, B. O. (2004). Mplus User's Guide (3rd ed.). Los Angeles,
CA: Muthén & Muthén.

Muthén, L. K., & Muthén, B. O. (2006). Mplus (Version 4.2) [Computer program].
Los Angeles, CA: Muthén & Muthén.

Nagin, D. S. (2005). Group-Based Modeling of Development. Cambridge MA:
Harvard University Press.

Norman, R. M. G., Malla, A. K., Cortese, L., Cheng, S., Diaz, K., McIntosh, E., et al.
(1999). Symptoms and cognition as predictors of community functioning: A
prospective analysis. The American Journal of Psychiatry, 156(3), 400-405.

Nuechterlein, K. H., & Asarnow, R. F. (1992). Manual and computer program for the
UCLA Continuous Performance Test. Los Angeles: unpublished.

Nuechterlein, K. H., Barch, D. M., Gold, J. M., Goldberg, T. E., Green, M. F., &
Heaton, R. K. (2004). Identification of separable cognitive factors in
schizophrenia. Schizophrenia Research, 72(1), 29-39.

Okubo, Y., Suhara, T., Suzuki, K., Kobayashi, K., Inoue, O., Terasaki, O., et al.
(1997). Decreased prefrontal dopamine D 1 receptors in schizophrenia
revealed by PET. Nature, 385(6617), 634-636.

Oltmanns, T. F., & Neale, J. M. (1975). Schizophrenic performance when distractors
are present: Attentional deficit or differential task difficulty? Journal of
Abnormal Psychology, 84, 205-209.

93

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,
doi:10.1016/j.cpr.2007.01.004.

Pierri, J. N., Chaudry, A. S., Woo, T.-U. W., & Lewis, D. A. (1999). Alterations in
chandelier neuron axon terminals in the prefrontal cortex of schizophrenic
subjects. American Journal of Psychiatry, 156(11), 1709-1719.

Pinkham, A. E., Penn, D. L., Perkins, D. O., & Lieberman, J. (2003). Implications
for the neural basis of social cognition for the study of schizophrenia.
American Journal of Psychiatry, 160(5), 815-824.

Ployhart, R. E., & Oswald, F. L. (2004). Applications of mean and covariance
structure analysis: Integrating correlational and experimental approaches.
Organizational Research Methods, 7(1), 27.

Pulver, A. E. (2000). Search for schizophrenia susceptibility genes. Biological
Psychiatry, 47(3), 221-230.

Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models (Second ed.).
Thousand Oaks, CA: Sage Publications

Rodriguez, D., Moss, H. B., & Audrain-McGovern, J. (2005). Developmental
heterogeneity in adolescent depressive symptoms: Associations with smoking
behavior. Psychosomatic Medicine, 67(2), 200-210.

Saleebey, D. (1992). Biology's challenge to social work: Embodying the person-in-
environment perspective. Social work, 37(2), 112-118.

Scott, J. E., & Dixon, L. B. (1995). Assertive community treatment and case
management for schizophrenia. Schizophrenia Bulletin, 21(4), 657-668.

Seeman, P., Lee, T., Chau-Wong, M., & Wong, K. (1976). Antipsychotic drug doses
and neuroleptic/dopamine receptors. Nature, 261(5562), 717-719.

Sevy, S., & Davidson, M. (1995). The cost of cognitive impairment in schizophrenia.
Schizophrenia Research, 17(1), 1-3.

94

Strauss, J. S., & Carpenter, W. T. (1972). The prediction of outcome in
schizophrenia: I. Characteristics of outcome. Archives of General Psychiatry,
27, 739-746.

Strauss, M. E. (1993). Relations of symptoms to cognitive deficits in schizophrenia.
Schizophrenia Bulletin, 19(2), 215-231.

Test, M. A., Knoedler, W. H., Allness, D. J., Burke, S. S., Brown, R. L., & Wallisch,
L. S. (1991). Long-term community care through an assertive continuous
treatment team. In C. Tamminga & S. Schultz (Eds.), Advances in
neuropsychiatry and psychopharmacology, volume 1: Schizophrenia
research. New York: Raven Press.

Torrey, E. F. (2001). Surviving Schizophrenia: A Manual for Families, Consumers,
and Providers: Quill.

Vandenberg, R. J., & Lance, C. E. (2000). A review and synthesis of the
measurement invariance literature: Suggestions, practices, and
recommendations for organizational research. Organizational Research
Methods, 3(1), 4-70.

Velligan, D. I., & Glahn, D. (2004). Treating the cognitive deficits of schizophrenia.
Current Science, 2, 73-77.

Ventura, J., Green, M. F., Shaner, A., & Liberman, R. P. (1993). Training and
quality assurance with the Brief Psychiatric Rating Scale: "The drift busters."
International Journal of Methods in Psychiatric Research, 3(4), 221-244.

Ventura, J., Nuechterlein, K. H., Subotnik, K., & Gilbert, E. (1995). Symptom
dimensions in recent onset schizophrenia: The 24-item expanded BPRS, The
5th International Congress on Schizophrenia Research (Vol. 15, pp. 22). Hot
Springs, Virginia, April 10

Wu, E. Q., Birnbaum, H. G., Shi, L., Ball, D. E., Kessler, R. C., Moulis, M., et al.
(2005). The economic burden of schizophrenia in the United States in 2002.
Journal of Clinical Psychiatry, 66(9), 1122-1129.

Wykes, T., Reeder, C., Landau, S., Everitt, B., Knapp, M., Patel, A., et al. (2007).
Cognitive remediation therapy in schizophrenia. The British Journal of
Psychiatry, 190(5), 421-427.

95

Young, A. S., Sullivan, G., Murata, D., Sturm, D., & Koegel, P. (1998).
Implementing publicly funded risk contracts with community mental health
organizations. Psychiatric Services, 49(12), 1579-1584.

Abstract (if available)

Abstract The present study investigated longitudinal relationships among cognition, symptoms and functioning in individuals diagnosed with schizophrenia for better understating the mechanisms of psychosocial functional change in community-based psychosocial rehabilitation programs. There are 2.97 million people suffering form schizophrenia in the United States, and the economic burden of schizophrenia is approximately $62.7 billion per year. Community-based treatment programs for individuals diagnosed with schizophrenia have established intensive clinical services to support and maintain these individuals in communities. The outstanding success of the programs has resulted in the clinical outcomes of reducing relapse and hospitalization, improving social functioning and improving housing stability. However, this success has been controversial because there is no clear understanding of how individual change is caused by the treatments and of why the change occurs or does not occur in individuals. This is a gap of knowledge in practice and research.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

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

PDF

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

PDF

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

PDF

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

PDF

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

PDF

The effects of physical functioning and public stigma on psychological distress as mediated by cognitive and social factors among Korean survivors of childhood cancer

PDF

Acculturation team-based clinical program: pilot program to address acculturative stress and mental health in the Latino community

PDF

Developmental trajectories of sensory patterns in young children with and without autism spectrum disorder: a longitudinal population-based study from infancy to school age

Asset Metadata

Creator Hoe, Maanse (author)

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

School School of Social Work

Degree Doctor of Philosophy

Degree Program Social Work

Publication Date 07/26/2007

Defense Date 06/07/2007

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

Tag cognitive rehabilitation,functional outcome,growth mixture analysis,latent trajectory class,OAI-PMH Harvest,schizophrenia,treatment mechanism

Language English

Advisor Brekke, John S. (committee chair), Bola, John (committee member), Chou, Chih-Ping (committee member), Zebrack, Bradley (committee member)

Creator Email hoe@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m681

Unique identifier UC1228081

Identifier etd-Hoe-20070726 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-521946 (legacy record id),usctheses-m681 (legacy record id)

Legacy Identifier etd-Hoe-20070726.pdf

Dmrecord 521946

Document Type Dissertation

Rights Hoe, Maanse

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Repository Name Libraries, University of Southern California

Repository Location Los Angeles, California

Repository Email cisadmin@lib.usc.edu