University of Southern California Dissertations and Theses

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

(USC Thesis Other)

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

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Copyright 2007 Joseph Michael Guada
THE IMPACT OF FAMILY FACTORS ON THE FUNCTIONING OF AFRICAN-
AMERICAN CONSUMERS LIVING WITH SCHIZOPHRENIA
by
Joseph Guada
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
ii
Dedication
The following dissertation is dedicated to the memory of my parents Dorothy
(Rumfola) Guada and Louis A. Guada both of whom offered me their ongoing love
and support through all of my educational endeavors. I love and miss you, and think
of you every day.
iii
Acknowledgements
I wish to thank John Brekke, Ph.D., Helen Land, Ph.D., and Timothy Biblarz, Ph.D.,
the dissertation committee members, for their ongoing feedback, insights, critiques,
and support throughout the dissertation writing process. I particularly wish to thank
Dr. Brekke for his ever supportive mentoring and knowledge throughout my entire
doctoral studies. I also wish to thank the School of Social Work at the University of
Southern California for its ongoing guidance and for the rigorous education
opportunity it provided me.
The following study was funded by NIMH, F-31 (Grant #:1F31 MH072126-01) with
myself as the Ph.D. candidate and Dr. John Brekke, Ph.D. as sponsor.
iv
Table of Contents
Dedication ii
Acknowledgments iii
List of Tables v
List of Figures vi
Abstract vii
Introduction 1
Chapter 1: First Structural Equation Model 13
Chapter 2: Second Structural Equation Model 75
Study Conclusion 121
Bibliography 137
v
List of Tables
Table 1: Descriptive statistics of sample 26
Table 2: Descriptives of indicators for latent variables 36
Table 3: Pearson’s correlation matrix for indicators used in model
specification
38
Table 4: Summary of goodness-of-fit indices for original and alternative
measurement models
46
Table 5: Summary of goodness-of-fit indices for original and alternative path
models
59
Table 6: Summary of estimates for alternative path model 5 64
Table 7: Descriptive statistics of sample 84
Table 8: Descriptives of indicators of latent variables 95
Table 9: Pearson’s correlation matrix for indicators used in model
specification
97
Table 10: Summary of goodness-of-fit indices for original and alternative
path models
109
Table 11: Summary of estimates for alternative path model 2 111
vi
List of Figures
Figure 1: Measurement model of family variables and consumer psychosocial
functioning
21
Figure 2: Hypothesized structural model for family latent variables on
consumer psychosocial functioning
23
Figure 3: Modified measurement model after initial analyses 48
Figure 4: Alternative path model 1 with family latent variables hypothesized to
have direct effects on consumer psychosocial functioning.
53
Figure 5: Alternate path model 2 54
Figure 6: Alternate path model 3 56
Figure 7: Alternative path model 4 57
Figure 8: Final path model testing family factors on consumer psychosocial
functioning
64
Figure 9: Hypothesized measurement model of family variables and consumer
psychiatric functioning
80
Figure 10: Hypothesized structural model for family latent variables on
consumer psychiatric functioning
82
Figure 11: Final alternative path model including all hypothesized paths 110
Figure 12: Final path model predicting consumer psychiatric functioning 111
vii
Abstract
Background: Many studies have tested the impact of different family factors on the
functioning of consumers living with schizophrenia. Few studies have concurrently
tested these factors together and with a larger sample of poorer African-American
consumers and their families, not typically well represented in the literature. Because
consumer clinical functioning and psychosocial functioning are known to be
minimally correlated, they were treated as two separate domains. Likewise, it is
argued that consumer psychosocial functioning is more germane as an outcome
variable given that most consumers spend more time in the community than in the
past.
Methods: The study took advantage of advanced statistical methodology not
previously used in the family mental health literature in order to concurrently test
several family factors as exogenous variables for consumer functioning. Two SEM
models tested the effect of family factors on consumer psychosocial functioning and
clinical functioning.
Results: More family criticism predicted worse consumer clinical functioning
replicating a finding already seen in the EE literature with substantially smaller
samples of African-Americans. More family contact predicted improved consumer
psychosocial functioning while more family dysfunction predicted poorer consumer
viii
psychosocial functioning. Fewer family resources predicted greater family pressures
while more family pressures predicted greater family dysfunction. Findings suggest
that contact with families was crucial, in and of itself, to consumer well-being in the
community even given that family dysfunction had a concurrently negative effect on
psychosocial functioning.
Conclusion: The study appears to confirm that what affects consumer psychosocial
functioning does not affect clinical psychiatric functioning. These findings are new
for understanding the contextual factors that impact consumer functioning, especially
psychosocial functioning. These findings fill a critical need in the literature for
understanding what contextual factors impact consumer clinical and psychosocial
functioning.
1
Introduction
The hypothesis that the family influences how an individual member behaves
or feels has been an important idea since the beginning of the mental health
movement (Hoenig & Hamilton, 1966; J. P. Leff, 1976). The research on
schizophrenia adds a wealth of information about familial influence on individual
functioning (Brown, Monck, Carstairs, & Wing, 1962; Jablensky et al., 1992; J. P.
Leff, 1976). The most important concept in the literature to date is that of Expressed
Emotion (EE). Nonetheless, other family factors, such as burden, are areas of interest
within the literature (Biegel, Milligan, Putnam, & Song, 1994; Dixon, Adams, &
Lucksted, 2000; Pilling et al., 2002; Stricker, Schulze Monking, & Buchkremer,
1997). The following is a study testing the relationship between family factors and
the functioning of a family member diagnosed with schizophrenia using two
structural equation models.
The family mental health literature - the need for a broader framework of analysis
The two areas that have received the most interest in the family mental health
literature are Expressed Emotion (EE) and caregiver burden. EE is defined as a type
of interaction, which includes hostile and critical comments and an incessant and
intrusive style on the part of a family member attempting to manage the minutia of
the ill family member’s affairs (Butzlaff & Hooley, 1998; Hashemi & Cochrane,
1999a; Kavanagh, 1992). This contact impacts the outcome of how the consumer
does over time, particularly with regards to relapse and rehospitalization, as has been
2
demonstrated in numerous studies (Hashemi & Cochrane, 1999a; J. Leff, Kuipers,
Berkowitz, & Sturgeon, 1985; J. Leff & Vaughn, 1981). Those families with ‘high’
EE are thought of as expressing more Critical Comments (CC), Hostility (H), and/or
Emotional Over-Involvement (EOI) than families rated as ‘low’ in Expressed
Emotion.
Over the course of several decades of research, the EE literature
demonstrated that a higher level of EE predicted poorer consumer functioning. This
relationship held across ethnicity and was assumed to be present despite the level of
care the consumer received (i.e. inpatient or outpatient). Thus, EE is predictive for
rehospitalization and/or relapse regardless of the ethnicity of the family, although the
level of EE that is predictive is not uniform (Jenkins & Karno, 1992; Moline, Singh,
Morris, & Meltzer, 1985). For example, the level of Critical Comments (CC) must be
at a higher level to be predictive for relapse and rehospitalization for African-
American consumers, as compared to European-American consumers, although the
total percentage of families rated as high EE was not different across the two groups
(Moline et al., 1985; Wuerker, Haas, & Bellack, 1999).
Level of family burden is the other substantive area of research for evaluating
the relationship between the family unit and the family member with a major mental
illness (Baronet, 1999; Saldana, Dassori, & Miller, 1999; Stueve, Vine, & Struening,
1997). The overall findings are that most families report some level of burden
regarding managing and interacting with their family member diagnosed with
schizophrenia (Baronet, 1999; Cook, Lefley, Pickett, & Cohler, 1994; Fadden,
3
Bebbington, & Kuipers, 1987; Horwitz & Reinhard, 1995). However, there are
differences in how much burden is reported depending on the ethnicity of the family
(Guarnaccia & Parra, 1996; Horwitz & Reinhard, 1995). For example, the findings
consistently show that African American families reported significantly less burden
when compared to white care-giving families. However, a few studies found no
differences in burden across ethnicity among lower SES families taking care of a
seriously mentally ill family member (Biegel et al., 1994; Song, Biegel, & Milligan,
1997). Nonetheless, it has been repeatedly hypothesized that when African-American
families report substantially lower levels of burden it is because of extended family
networks, a more extensive history of dealing with difficult long-term situations (i.e.
racism), and due to a higher level of a family-centered orientation versus an
individual-centered orientation (Guarnaccia & Parra, 1996; Hill, 1998; McAdoo,
1998).
The findings of the EE literature, that there are essentially few differences
between ethnic groups for EE’s impact on consumer levels of rehospitalization and
relapse, and that of the burden literature, that there are differences in reported levels
of burden across ethnic groups, appears somewhat contradictory. Why would some
ethnic groups report lower levels of burden, but then score at essentially equivalent
levels of EE? Thus, for example, why do African-Americans essentially have
equivalent levels of EE while also reporting lower levels of burden? Given the
findings of the EE literature, one would expect that high-EE families would also
report higher levels of burden. This apparent contradiction both highlights the state
4
of the literature as well as the need for testing these family factors in the same model
so as to assess interactions between them while at the same time testing their impact
on consumer functioning.
Thus far the vast majority of the literature has looked at these variables in
numerous separate studies. Family factors are treated as separate concepts from each
other by virtue of the fact that so few studies have tested possible relationships
between EE and burden, for example (Jackson, Smith, & McGory, 1990; Scazufca &
Kuipers, 1996). The U.S. Department of Health and Human Services, in its
supplement Mental Health: Culture, Race, and Ethnicity (Services, 2001; USDHHS,
2001), identified the critical need to investigate the contextual factors, particularly
those related to the family, impacting the functioning of consumers with a serious
and persistent mental illness. It is clear that the literature about family influences on
consumer functioning needs a broader lens to help explain these influences. That is,
the literature lacks models that concurrently test a variety of variables such as level
of family criticalness, caregiver burden, as well as the family’s overall functioning,
and how these might interact with each other, while also analyzing the impact on
consumer functioning (Kavanagh, 1992; S. King & Dixon, 1996). Given the findings
of similar levels of EE while also having lower levels of burden, African-American
consumers and their families are an ideal population to test these relationships.
However, the specifics of what these relationships are for African-American inner-
city consumers and their families remain unexplored.
5
For the present study, family factors were selected based on both the EE and
burden literatures. Thus, for example, some EE researchers have hypothesized that
the family communication style, interpersonal relationships, and level of dysfunction
contribute to poorer outcomes for consumers (C. M. Anderson, Reiss, & Hogarty,
1986; Bebbington & Kuipers, 1994; Karno et al., 1987; J. Leff & Vaughn, 1981,
1985; McFarlane et al., 1991). Likewise, some caregiver burden researchers have
hypothesized that stressors or pressures linked to such things as resource availability
and role demands contribute to caregiver burden, which in turn can contribute to
consumer functioning (Biegel et al., 1994; Cook et al., 1994; Jones, Roth, & Jones,
1995; Rivera, Torres, & Carre, 1997; Saldana et al., 1999). The following study fills
this gap in the literature by taking a “broader view” by concurrently testing concepts
used in numerous separate studies that have been hypothesized to impact consumer
functioning.
Consumer functioning – the domains of clinical and psychosocial functioning
Historically, the concept of consumer functioning was a sort of umbrella term
which included rehospitalization, relapse, psychosocial functioning and symptom
severity. Each of these was viewed as different sub-parts of the “consumer
functioning” whole (Barrowclough & Tarrier, 1990; Brekke, Kohrt, & Green, 2001;
Brekke, Levin, Wolkon, Sobel, & Slade, 1993; Brekke & Long, 2000; Brekke,
Raine, Ansel, Lencz, & Bird, 1997; Inoue, Tanaka, Shimodera, & Mino, 1997; S.
King & Dixon, 1996; Phillips, Barrio, & Brekke, 2001). Recent studies have
demonstrated, however, that symptom severity is typically only mildly correlated to
6
consumer psychosocial functioning (Brekke & Long, 2000; Brekke et al., 1997;
Inoue et al., 1997; Phillips et al., 2001). Researchers presently view consumer
functioning as a set of distinct but interrelated domains.
Psychosocial functioning is defined here as those areas that require social
skills in order to function and cope with the daily demands of the social environment
(Brekke & Long, 2000; Brekke et al., 1997; Inoue et al., 1997; Phillips et al., 2001).
The arenas that are most typically included are work, independent living skills, and
interpersonal/family relations (Brekke & Long, 2000; Goodman, Sewell, Cooley, &
Leavitt, 1993).
As mentioned above, the findings of high EE on rehospitalization and relapse
are robust. Yet, with the onset of deinstitutionalization more than 40 years ago as
well as substantially shorter hospital stays due to managed care and more restrictive
public funding, the vast majority of consumers diagnosed with schizophrenia spend
less time in a hospital setting (Geller, 2000; Lamb & Bachrach, 2001; Mechanic,
1999). Indeed, the majority live free of relapse for long periods of time and many
live independently from their families of origin. Given the challenges that consumers
with schizophrenia often face regarding psychosocial functioning there is an urgent
need for research to investigate what might contribute to or protect the individual’s
psychosocial functioning. The NIMH prioritized more research into this area given
the serious consequences that impaired psychosocial functioning has for the
individual, the family, and society (Workgroup., 2000). On the other hand, more
7
“traditional” or clinical-oriented outcomes remain important as markers for
psychiatric functioning for the outpatient consumer, too.
Thus, another gap in the literature is the need to distinguish between
different, only mildly correlated constructs for consumer functioning such as clinical
and psychosocial functioning. For instance, in pre-analyses for this study, consumer
clinical functioning and consumer psychosocial functioning were only mildly, albeit
non-significantly correlated (r=-.24, p=.06). Thus, the following study addresses this
gap in the literature by testing numerous family factors (based on the EE and burden
literatures) on both consumer psychosocial and clinical functioning by treating these
as two separate domains of interest.
African-American consumers & their families – a gap in the mental health literature
The proposed study provides data on inner-city African-American consumers
and their families, a section of the general population that has had fewer clinical
protocols targeted toward them as well as have had fewer clinical or research
resources earmarked for them (McAdoo, 1998; D. R. Williams & Fenton, 1999). As
noted above, the NIMH stated that contextual factors hypothesized to contribute to
consumer psychosocial and clinical functioning is also a top priority for future
research (Workgroup., 2000). The social networks, extended family forms, and a
higher sense of a family-orientation, often reported among African-American
families (Alston & Turner, 1994; Billingsley, 1990; Wallace Williams, Dilworth-
Anderson, & Goodwin, 2003; D. R. Williams & Fenton, 1999), offers an ideal
opportunity to assess broader contextual factors, as outlined above, on consumer
8
psychosocial and clinical functioning. And in turn, because of the importance of the
family to African-American consumers, studying these family factors is of critical
importance to designing and implementing services specifically targeted to
consumers from similar communities.
In addition, within the EE literature, there are small samples sizes as well as
few actual studies that included African-American consumers and their families. For
example, only two studies with samples of less than 40 African-American consumers
and their families are available in the EE literatures (Jenkins & Karno, 1992; Moline
et al., 1985). Although sample sizes used in the caregiver burden literature are larger,
these studies have not typically included a concurrent sample of consumers related to
the family members (Baronet, 1999; Fadden et al., 1987; Pickett, Vraniak, Cook, &
Cohler, 1993; Solomon & Draine, 1995; Stueve et al., 1997). The following study
addresses these gaps in the literature by including a larger African-American sample
(n=94) that includes both consumers and their families.
The use of Structural Equation Modeling – extending the mode of analysis
The majority of the EE and caregiver burden literature includes the use of t-
tests, multiple regressions, and a few path analytic models. Although such
methodologies have their advantages and uses, they limit the ability of researchers to
take a broader approach to testing relationships between family and consumers
factors. The present study takes advantage of advanced statistical methodology to
concurrently test more of these relationships through the use of structural equation
modeling (SEM).
9
Specifically, the study involves testing two models: the first SEM includes
latent variables for family functioning and how these are hypothesized to have a
direct and/or indirect effect on the criterion variable of consumer psychosocial
functioning. The second SEM includes the same latent family variables and the
hypothesized direct or indirect effects these have on consumer clinical functioning.
Use of structural equation modeling (SEM) provides several advantages over
multiple regression and path-model analysis: it tests hypotheses at a higher level of
abstraction (latent variables can be included versus only the means and
intercorrelations between specific observed variables), multiple paths can be tested
simultaneously, and an entire model is tested as to how approximately it explains the
covariance relationships across numerous variables (Kline, 2005).
Thus, there is a need in the literature to take advantage of advanced statistical
methodology such as SEM. The following study takes advantage of these
methodologies by concurrently testing these variables at one time as opposed to what
has been done in the past where numerous studies tested these variables separately.
Advanced statistical methodology provides the means to use a “broader lens”.
Summary of the study’s purposes and aims
To summarize, the following study addresses several gaps in the literature in
the following ways:
1.) It includes a broader range of family factors, based on the EE & caregiver
burden literatures. It tests these family factors that have been hypothesized to either
directly and/or indirectly effect consumer functioning. The purpose is to test a
10
broader array of family factors – to broaden “the lens”– by concurrently testing a
number of variables, some of which have thus far been tested across numerous
separate studies.
2.) It includes consumer psychosocial functioning while also including
clinical symptomatology. Thus, it will treat “consumer functioning” as two main
domains of interest: psychosocial functioning, which can be argued is more
applicable to outpatient consumers; and the more “traditional” domain of consumer
clinical functioning.
3.) In addition, very few studies have thus far tested family factors
specifically on consumer psychosocial functioning. This study offers one of the first
studies to do so by testing family contextual factors’ influence on consumer
psychosocial functioning.
4.) The study includes a larger sample of poorer African-American
consumers and their families, a section of the general population not often included
in such studies. The NIHM has made the study of under represented communities
one of its major, top priorities. In addition, the study offers the opportunity to test
previous findings within the EE literature on a larger sample of African-American
consumers and their families.
5.) And finally, the study uses a more advanced statistical methodology
(structural equation modeling - SEM), not often used in the family mental health
literature, to test these relationships. The use of SEM allows the testing of more
11
detailed models, which includes latent variables, while simultaneously testing the
relationship between family and consumer related variables.
Format for presenting the structural equation model findings
The methods, analyses, and results for each structural equation model (SEM)
analyzed for this study are presented in separate sections. Each SEM will be
presented as a complete, but separate article of publishable quality. All references for
each of the separate articles are included in the Bibliography section of the main
paper. An overall conclusion, based on the findings of the analyses of the two
structural equation models, follows the articles.
The first SEM extends a path model that found that the more family contact a
consumer had with the family the better the psychosocial functioning of the
consumer (Guada & Brekke, 2007). The structural equation model will test what
family factors might contribute to or hinder amount of family contact. Of interest is
whether circumstances or functioning of families – a broader range of family factors
than Critical Comments (the sub-construct of EE) or family burden – contributes to
amount of contact between consumer and family. Do such things as the pressures
that families have (including level of burden), the resources families have to cope
with these pressures, and the family’s overall functioning (including communication
and emotional involvement) impact how often consumers interact with their
families?
The second SEM includes consumer outcomes that are conceptually closer to
those of the original EE studies (i.e. rehospitalization and relapse), in this case level
12
of psychiatric symptomatology (i.e. psychiatric symptoms related to anxiety and
psychosis). It tests the direct path from family criticalness to consumer clinical
functioning while also including a broad array of family variables’ (the same as those
included in the first SEM) effect on amount of family criticalness. In addition, the
domains of consumer functioning are represented by latent constructs for
psychosocial functioning (the first SEM) and clinical functioning (the second SEM).
13
Chapter 1: First Structural Equation Model
Introduction
The construct of “consumer functioning” for individuals living with
schizophrenia has traditionally included level of symptoms, rehospitalization, and
social functioning. Each of these were viewed as different sub-parts of the
“consumer functioning” whole (Barrowclough & Tarrier, 1990; Brekke et al., 2001;
Brekke et al., 1993; Brekke & Long, 2000; Brekke et al., 1997; Inoue et al., 1997; S.
King & Dixon, 1996; Phillips et al., 2001). Recent studies have demonstrated,
however, that amount of symptoms is at most only mildly correlated with consumer
psychosocial functioning (Brekke & Long, 2000; Brekke et al., 1997; Inoue et al.,
1997; Phillips et al., 2001). Psychosocial functioning is defined here as a set of skills
that allow the individual to respond to the daily demands and stressors of such things
as work, living independently, and maintaining interpersonal/family relationships
(Brekke & Long, 2000; Brekke et al., 1997; Inoue et al., 1997; Phillips et al., 2001).
Thus, individuals can function at some level within these spheres even given some
level of symptomatology. This suggests that what effects consumer psychosocial
functioning might be unrelated to what was found to affect other aspects of
functioning, such as rehospitalization or relapse, and hence should be a specific area
of interest in the literature.
A top priority of the National Institutes of Mental Health (NIHM) is to
understand the contextual factors that contribute to the psychosocial functioning of
14
people living with schizophrenia (USDHHS, 2001). Poor psychosocial functioning
creates numerous individual, interpersonal, and social problems. Understanding
contextual factors is critically needed in designing both practice models and mental
health policy so as to improve the recovery and quality of life of those living with
schizophrenia (USDHHS, 2001; Workgroup., 2000).
Many researchers assert the importance and centrality of the family in
minority communities, particularly for African-Americans (Alston & Turner, 1994;
Billingsley, 1990; Hines & Boyd-Franklin, 2005; McAdoo, 1998; Turner & Alston,
1994; D. R. Williams & Fenton, 1999). The present study offers the opportunity to
identify contextual relationships between family factors and consumer functioning
with a sample of African-American consumers living with schizophrenia and their
families, a population typically underrepresented in the family mental health
literature (USDHHS, 2001). The family is hypothesized to be of critical importance
to each family member’s functioning. Indeed, some have argued that family factors
should be tested for their effect on consumer psychosocial functioning outcomes in
general (Barrowclough & Tarrier, 1990; Inoue et al., 1997; Stricker et al., 1997).
This makes sense given that with the onset of deinstitutionalization more than 40
years ago as well as substantially shorter hospital stays due to managed care and
more restrictive public funding, the vast majority of consumers diagnosed with
schizophrenia spend less time in a hospital setting (Geller, 2000; Lamb & Bachrach,
2001; Mechanic, 1999). Hence, psychosocial functioning is arguably more
applicable as an outcome variable for the vast majority of consumers today (Brekke
15
& Long, 2000; Goodman et al., 1993). Understanding family contextual factors’
influence on psychosocial functioning for African-American consumers is of
particular importance in order to design and implement practice models and social
policy targeted for this community.
In a preliminary study, Guada & Brekke (2007) tested a path analytic model
that used consumer psychosocial functioning as the outcome variable for a sample
(n=88) of African-American consumers and their families. Critical Comments (CC),
the most predictive sub-construct of Expressed Emotion (EE) for African-American
consumers (Chambless, Steketee, Bryan, Aiken, & Hooley, 1999; Moline et al.,
1985; Van Humbeeck, Van Audenhove, De Hert, & Storms, 2002; Wuerker et al.,
1999), was used as the principle exogenous variable. The model included a direct
path from Perceived Criticism (PC) (Hooley & Teasdale, 1989; Riso, Klein,
Anderson, Ouimette, & Humberto, 1996), as a proxy for CC, on consumer
psychosocial functioning. Likewise, amount of family contact was used as a control
variable. Although a higher level of consumer psychiatric symptomatology was
significantly associated with a higher level of PC, level of PC was not directly
related to level of consumer psychosocial functioning. Interestingly, a higher level of
family contact was significantly related to level of consumer psychosocial
functioning: more contact with the family was associated with better consumer
psychosocial functioning. Thus, a family’s level of criticalness had no direct effect
on consumer psychosocial functioning, but amount of family contact did.
16
The present study continues to focus on the psychosocial domain of
consumer functioning; it also tests what appears to be a novel finding in the literature
– more family contact is related to better consumer psychosocial functioning. It
extends the aforementioned path model by exploring what might contribute to or
hinder the beneficial effects of family contact on consumer psychosocial functioning.
Of interest is whether circumstances or functioning of families contribute to the
amount of contact. Family factors were chosen based on those variables
hypothesized in the EE and caregiver burden literature to affect a family’s
functioning (Jackson et al., 1990; Scazufca & Kuipers, 1996). Do such things as the
pressures that family’s have (including level of burden), the resources family’s have
to cope with these pressures, and the family’s overall functioning (including
communication styles and affective involvement between family members) directly
effect how often consumers interact with their families?
Lastly, the vast majority of research involving family factors on consumer
functioning used statistical methodology such as correlations, multiple regression,
and path analysis. Such methodologies limit the number of variables used and
complexity of models that are tested. Hence, family factors were typically analyzed
separately and across numerous individual studies. This present study used the
advanced statistical methodology of structural equation modeling (SEM) so that
multiple factors could be concurrently tested (i.e. tested in the same model). It
expands the lens of analysis by testing these relationships in one study as opposed to
what has been done before across numerous separate studies.
17
Thus, the present study extends the literature on family influence on
consumer functioning in the following ways. The study focuses on African-
American consumers living with schizophrenia and their families, a community
traditionally underrepresented in the family schizophrenia mental health literature
and a top priority for inclusion in research (USDHHS, 2001; Workgroup., 2000). In
addition, understanding family contextual factors is of vital importance in
understanding the functioning of African-American consumers living in the
community given the importance the family plays within this community. The study
extends the analysis of what appears to be a novel finding (that increased family
contact has a positive direct effect on consumer psychosocial functioning) by
introducing other family factors (e.g. pressures, resources, and overall functioning),
as hypothesized in the caregiver burden and EE literatures, into the model. It focuses
on consumers functioning domain (psychosocial functioning) more applicable to
outpatient consumers (Guada & Brekke, 2007). And it uses a more advanced
statistical methodology (SEM) to concurrently test these relationships with a large
sample of African-American consumers and their families. Findings offer important
information for improving the evidenced-based aspects of present practice models as
well as informing mental health policies that are geared toward assisting similar
communities.
18
Study hypotheses
The present study chose family factors that both the EE and burden literatures
hypothesize to be directly or indirectly important for consumer functioning. For
example, a family’s communication and affective style are hypothesized to directly
effect consumer functioning in the EE literature (C. M. Anderson, Hogarty, & Reiss,
1980; Bebbington & Kuipers, 1994; Dixon et al., 2000; Falloon et al., 1982; Hooley,
1985; J. Leff & Vaughn, 1985; McFarlane et al., 1991). The caregiver burden
literature specifically focuses on amount of family burden and what stressors and
pressures might contribute to this (Baronet, 1999; Demi, Bakeman, Moneyham,
Sowell, & Seals, 1997; Fadden et al., 1987; Solomon & Draine, 1995). The present
study includes family factors that include these variables as indicators for latent
constructs that are hypothesized to impact amount of family contact. Thus, the latent
variable for a family’s overall functioning includes indicators for family
communication and affective styles, amongst other indicators. The latent variable for
family pressures includes level of family burden as well as financial and
discrimination pressures. And the latent variable for family resources includes
indicators for family social support and the family’s sense of well-being.
In addition, the direction of how these latent variables should affect amount
of family contact is based on the EE and caregiver burden literatures. For example,
both the caregiver burden and coping literatures suggest that contact with an ill
family member creates increased levels of burden or stress for the family (Aranda &
Knight, 1997; Baronet, 1999; Land & Hudson, 1997, 2002; McCabe, Yeh, Lau,
19
Garland, & Hough, 2003; Pearlin, Aneshensel, & LeBlanc, 1997; Pinquart &
Sorensen, 2005; Rivera et al., 1997; Saldana et al., 1999; Wallace Williams et al.,
2003). The present study hypothesizes that as families report more pressures (such as
higher levels of burden) the amount of family contact will decrease.
Likewise, EE can be conceived of as a “kind” of family dysfunction:
certainly this is implied in the construct itself as well as the variables proposed as
influencing its amount (such as poor or dysfunctional family communication styles)
(Bebbington & Kuipers, 1994; Boye et al., 1999; Brown, Birley, & Wing, 1972;
Hatfield, 1997a; S. King & Dixon, 1996; J. P. Leff, 1976). Additionally, numerous
theories of clinical family functioning hypothesize that either too much or too little
involvement with a mentally ill family member can have a deleterious effect on the
family, the consumer, or both (C. M. Anderson et al., 1986; Boye et al., 1999; Dixon
et al., 2000; Fadden et al., 1987; Hatfield, 1997b; Lambert, 2003; Minuchin &
Fishman, 2004; Nichols & Schwartz, 2005). Thus, it is hypothesized that as the
amount of family dysfunction increases the amount of family contact will decrease.
Finally, the caregiver burden literature suggests that as the amount of family
resources increase, the level of burden reported by families decreases (Biegel et al.,
1994; Guarnaccia & Parra, 1996; Stueve et al., 1997). Although never tested before,
the literature suggests that as family’s report more resources that they will either
tolerate or seek out more contact with their ill family member. The present study
hypothesizes that as amount of family pressures increases the amount of family
contact also increases
20
As noted before the study will test the following relationships using
structural equation modeling (SEM). Thus, the specific hypotheses for the SEM are
divided into two parts: the Measurement Model and the Structural Model.
Measurement model hypotheses
Hypothesis 1: factor loadings will form independent clusters from each other
(McDonald & Ringo Ho, 2002). That is, at least two indicators will exclusively load
for every latent variable without cross-loading on other latent variables.
Hypothesis 2: the predicted measurement model’s relationship between
latent variables and their indicators will fit the observed data as evidenced by at least
acceptable goodness-of-fit indices. That is, the measurement model will not
significantly differ from the observed data, and other goodness-of-fit indices will be
at least in the “acceptable” fit range. See Figure 1 below.
21
Figure 1: Measurement model of family variables and consumer psychosocial
functioning
Structural model hypotheses
The proposed model can be “deconstructed” to the following specified
relationships between latent variables:
Hypothesis 3a: Family Contact will have a significant, positive direct
association on Consumer Psychosocial Functioning: more contact will predict better
consumer psychosocial functioning.
Hypothesis 3b: Family Dysfunction will have a significant negative direct
association on amount of family contact: more family dysfunction will predict less
family contact.
Family
Contact
Family
Pressures
Family
Resources
Family
Dysfunction
Consumer
Psychosocial
Functioning
22
Hypothesis 3c: Family Pressures will have a significant, negative direct
association on amount of family contact: more family pressure will predict less
family contact.
Hypothesis 3d: Family Resources will have a significant, positive direct
association on amount of family contact: more family resources will predict more
family contact.
Hypothesis 4: The three exogenous family variables (Family Pressures,
Family Resources, and Family Dysfunction) will significantly correlate with each
other.
Hypothesis 5: the predicted structure model’s relationship across latent
variables will fit the observed data as evidenced by at least acceptable goodness-of-
fit indices. That is, the structure model will not significantly differ from the observed
data, and other goodness-of-fit indices will be in at least the “acceptable” fit range.
(Schreiber, Stage, King, Nora, & Barlow, 2006). Please refer to Figure 2 below.
23
Figure 2: Hypothesized structural model for family latent variables on consumer
psychosocial functioning
Methods
Study Context
The sample for this study came from a project that was investigating client
and family outcomes from a home-based intervention for families with a seriously
mentally ill member (schizophrenia spectrum disorders). The intervention and the
research protocol are described elsewhere (Connery & Brekke, 1999). Families were
recruited if their ill family member was receiving services at the participating county
contract mental health facility. The study was approved by the University of
Southern California Institutional Review Board, and the Los Angeles County
Consumer
Psychosocial
Functioning
Family
Contact
+
–
+
–
Family
Pressures
Family
Dysfunction
Family
Resources
24
Department of Mental Health. All study participants signed informed consent to
participate in the study.
Sample
The total project sample consisted of 106 families. In this report the data
from the 94 African-American families who were assessed at baseline before the
family intervention began are presented. A family reporter was selected in
conference with the family and consumer. This family reporter was a member of the
household who filled out all of the family measures at baseline. The vast majority of
family reporters were women (86.2 %) with a mean age of 47 years old (range from
18 to 80). Thirty-six percent were parents to the consumer, 22% were a sibling, while
21% were a child and 7% were a spouse. The remainder (14%) was “other” typically
a close friend or an extended family member (e.g. grandparent, aunt/uncle, or
cousin). There was notable diversity of who was chosen as family member
representative. In most studies involving white consumers and their families it would
be family (parent) to consumer (child) or spouse. The diversity of “types” of
relationships suggests the importance of the extended family form for this sample.
The average age of the consumer was 42 years old (age range of 23 – 64
years of age). Consumers lived with a family member on average 45 (sd=72) days in
the previous six months before entering the study. The distribution for living with
family was skewed: 59 consumers (63%) reported not living with a family member
in the preceding six-month period while 14 consumers (15%) lived with the family
the entire time. Thus, 63% of consumers lived independently of any family members.
25
Nonetheless, the vast majority of consumers lived in the same neighborhood as their
families (e.g. in same building, next street over, on the same block, etc.).
In addition, the average number of days in which the consumer was
hospitalized for psychiatric problems was 8 (sd=25) days for the preceding six-
month period. This variable’s distribution was also skewed with 72 consumers (82%)
having no days on inpatient psychiatric treatment. Thus, the vast majority of
consumers remained out of a hospital and lived continuously in the community for
the previous six months.
As of 1992 the median income for African Americans residing in the
neighborhood from which the sample was drawn was $22,000 with some areas
having as low as $12,000 as the median income (Civil, 1993). The average annual
income for consumers was $12,000. Thus, this was a relatively poor sample of
consumers and their families. This was the most recent data for this community
when enrollment for subjects began (1998). All of the consumers and their families
lived in the same section of a large urban area from which the sample was drawn.
Table 1 summarizes descriptive statistics for the sample below.
26
Table 1: Descriptive statistics of sample
Consumer Family Representative
Age (SD) 43 (42) 47 (49)
Gender (%) Female (56) Female (85)
Relationship Type
(%)
Child (35)
Parent (19)
Sibling (27)
Spouse (6)
Other (13)
Parent (35)
Child (19)
Sibling (27)
Spouse (6)
Other (13)
Income 1027.95* $12K – 22K
Days with family
#
45
NOTE: 63% lived
independently in
preceding 6 month
period.
NA
Days inpatient
#
8 (SD=24)
NOTE: 82% out of
the hospital in
preceding 6 month
period.
NA
Number of
residences
#
2 (SD=1) NA
* Per month
#
In preceding six-month period
Measures
Altogether there are five variables in the structure model (refer to Figure 1).
Four of these variables are latent and have at least two indicators. Three of the latent
variables (Family Pressure, Family Resources, and Family Dysfunction) refer to
27
experiences or functioning to the family unit as a whole. The fourth (an observable
variable) refers to amount of family contact between consumer and family members.
The fifth variable, which is latent (Consumer Psychosocial Functioning), is specific
to the functioning of the consumer.
Family Pressure indicators
Family Pressure refers to both concrete (financial) and subjective (burden)
pressures that a family might face when interacting or managing a family member
diagnosed with a major mental health disorder. A higher score on each of the
indicators demonstrates greater amounts of the pressure being measured. Family
Pressure has three indicators.
The Family Pressure Scale – Ethnic (FPRES-E) is 64-item scale developed to
measure the pressures unique to the life experiences of families of color (McCubbin,
Thompson, & McCubbin, 1996). A higher score indicates greater family pressure,
specifically those unique to families of color such as experiencing personal or
institutional racism. The original internal reliability (Chronbach’s alpha) was .92
(McCubbin et al., 1996). The FPRES-E was the strongest predictor of family
difficulties in a study with minority families (Native Hawaiian) thus suggesting the
validity of the measure (McCubbin et al., 1996). The alpha for this study was .94.
The Family Inventory of Resource Management (FIRM) is a self-
administered 66-item scale designed to both describe and assess a family’s amount
and management of their resources (McCubbin et al., 1996). It was specifically
created for use with samples of ethnic minority families (McCubbin et al., 1996;
28
McCubbin, Thompson, Thompson, & Futrell, 1998). The FIRM is divided into four
sub-scales. Only the Financial Well-Being (FWB) subscale was used and refers to
the family’s ability to meet financial commitments, the adequacy of financial
resources, the ability to assist others financially, and optimism regarding the
financial future of the family. A higher score indicates greater financial well-being.
The Chonbach’s alpha for this sample was .85. The FWB was recoded so that a
higher score indicates greater financial pressures.
The Burden Assessment Scale (BAS) is a 19-item self-administered scale that
measures the caregiver burden of family members with a seriously mentally ill
member of the family (Reinhard, Gubman, Horwitz, & Minsky, 1994). A higher
score signifies greater burden. Chronbach’s alpha for initial studies using the Burden
Assessment Scale were .91 and .89, and data on the validity of the Burden
Assessment Scale are available in Reinhard et al., (1994). The alpha on the Burden
Assessment Scale for this study was .94. The Burden Assessment Scale (BAS) has
been used successfully with African-American samples (Horwitz & Reinhard, 1995).
Family Resources indicators
Family Resources refers to both social, and an over all sense of well-being
that a family has available to help the family manage or cope including dealing with
a family member living with a major mental health disorder. A higher score on each
of the indicators demonstrates greater amounts of the resources available to the
family. Family Resources has four indicators.
29
The Social Support Index (SSI) is a 17-item self-administered scale that
measures to what degree the family finds support and integration within the
community (McCubbin et al., 1996). It uses a 5-point Likert scale ranging from
“Strongly Disagree” to “Strongly Agree”. Chronbach’s alpha for initial studies using
the SSI was .82 (McCubbin et al., 1996). The alpha on the SSI for this study was .75.
In addition, the SSI demonstrated discriminate validity when it negatively correlated
with a scale on family distress with a sample of ethnic minorities (McCubbin,
McCubbin, & Thompson, 1995).
The Family Crisis Oriented Personal Evaluation Scales (F-COPES) is 30-
item self-administered questionnaire designed to assess a family’s ability to problem
solve and strategize ways to cope with stressors (McCubbin et al., 1996). The scale
conceptualizes coping as a family resource akin to other resources such as social
support. The F-COPES consists of five sub-scales. The five subscales refer to five
ways that a family will cope (e.g. acquiring social support, reframing a difficult
situation, seeking spiritual support, seeking and accepting help outside of the family,
and passive appraisal). Two of the five sub-scales were used for the study – the
“Seeking Spiritual Support” and “Acquire & Accept Help” subscales. The validity
was assessed by testing the factor analysis by using randomly split samples. The
factor structure was replicated across the two groups (McCubbin et al., 1996). The
Cronbach’s alpha from the original development for the two subscales had the
following ranges (Seeking Spiritual Support: .79-.95, Acquire & Accept Help: .70-
.78) (McCubbin et al., 1996). The alphas subscales for this sample were the
30
following: Seeking Spiritual Support (.75), Acquire & Accept Help (.60). The ethnic
composition of the original samples used during the design of the scale was not
listed.
The Family Member Well-being Index (FMWB) is an 8-item self-
administered questionnaire that measures how much a family member feels adjusted
regarding issues about health, energy, feelings of concern, cheerfulness, etc.
(McCubbin et al., 1996). Each item is a 10-point Likert scale ranging from “Not
(very much)” to “Very (much)”. The Cronbach’s alpha for the original development
of the subscale was .85. Standard scores, means, and standard deviations were tested
on numerous ethnic groups including African-American families (McCubbin et al.,
1996). The alpha on the FMWB for this study was .74.
Family Dysfunction indicators
The third latent variable is Family Dysfunction. It refers to a family’s
problems to function as a family unit regardless of the presence of stressors or not. A
higher score on each of the indicators demonstrates greater problems with family
functioning in that domain. It is measured by six indicators, which are the sub-scales
of the MFAD.
The McMaster Family Assessment Device (MFAD) is a 53-item self-
administered questionnaire designed to assess families according to the McMaster
Model of Family Functioning (Epstein, Baldwin, & Bishop, 1983). The McMaster
Model of Family Functioning is, “a clinically oriented conceptualization of
families…It describes structural and organizational properties of the family group
31
and patterns of transactions among family members which have been found to
distinguish between healthy and unhealthy families” (Epstein et al., 1983, p. 172). A
higher score indicates more problems/more dysfunction. The McMaster Model of
Family Functioning is made up of six subscales including a seventh subscale for
overall “General Functioning”, which was not used in this study.
Each of the subscales is designed to measure a different dimension of family
functioning (Miller, Ryan, Keitner, Bishop, & Epstein, 2000). Miller et al. (2000)
described the six subscales that constitute the McMaster Model in the following way.
The first subscale is “Problem-solving”, which, as defined by the Model, is “…a
family’s ability to resolve problems at a level that maintains effective family
functioning” (p. 170). The second subscale is “Communication”, which is defined as
the family’s ability to exchange or move important information within a family
system with a focus specifically on “verbal exchange” (p. 171). The third subscale is
“Roles”. The Model defines this as the “recurrent patterns of behavior[s]” (p. 171)
that take place between family members and which influence the functioning of the
family unit; that is, the subscale measures how clear and often family members take
roles in order to problem solve. The fourth subscale is “Affective Responsiveness”,
which refers to the quantity and quality of the emotions experienced and expressed
amongst family members in response to stressors and other stimuli from the
environment. The fifth subscale, “Affective Involvement”, refers to the amount of
interest and focus that family members offer to each other around the specific,
individual interests and needs of family members. The last subscale is “Behavior”.
32
This is defined as “the pattern a family adopts for handling behavior in three types of
situations” (p. 172). The three types of situations are: physically dangerous
circumstances, expressing and meeting psychobiological needs, and situations that
involve interpersonal socializing needs.
The validity of the McMaster Family Assessment Device subscales were
examined in several studies (Epstein et al., 1983; Miller, Bishop, Epstein, & Keitner,
1990). They demonstrated both concurrent and discriminate validity in the expected
directions. The original Chronbach alpha’s for the six subscales ranged from .72 -
.92 (Epstein et al., 1983). Reliability was also analyzed with test-retest (Miller,
Epstein, Bishop, & Keitner, 1985) with results ranging from .66 to .76. The
indicators’ alphas for this sample were the following: Problem Solving (.80),
Communications (.40), Roles (.60), Affective Responsiveness (.77), Affective
Involvement (.78), and Behavior Control (.82). The alpha level for the indicator
Communications was low. However, by dropping one item the alpha increased to
.53. In addition, the ethnic composition of the original samples used during the
design of the scale was not listed (Friedmann et al., 1997; Kabacoff, Miller, Bishop,
Epstein, & Keitner, 1990; C. King, Hovey, Brand, Wilson, & Ghaziuddin, 1997;
Miller et al., 1990).
Family Contact
Contact with families is a simple count variable of number of times that the
consumer had contact with a family member via phone or personal contact. The
amount of contact was measured by several questions asking how much contact, in
33
hours over the previous two months, each subject had with parents or other relatives.
The sum of all contacts (whether initiated by the consumer or family member and
whether the contact was with a parent or with another relative) was used given the
extended family form that the sample presented with. Initially, FC was used as an
indicator for a latent variable called Family Interactions. Initial analyses showed the
beta for FC on Family Interactions was .99. It was decided to simplify the model by
deleting the latent variable and use FC as an observable variable in all subsequent
model analyses.
Consumer Psychosocial Functioning
The Consumer Psychosocial Functioning indicators were selected because
they have been used with success in previous studies on psychosocial rehabilitation
service outcomes (Brekke et al., 2001; Brekke & Long, 2000; Brekke et al., 1997),
and because they were also used to examine variation in cross-ethnic outcomes as
well (Barrio, 2001; Brekke & Barrio, 1997; Phillips et al., 2001). Based on the
conceptualization of psychosocial outcomes for schizophrenia by Brekke and Long
(2000), consumer outcomes for this study were chosen to reflect functional domains
such as work, independent living, and interpersonal relations.
The Role Functioning Scale (RFS) measures psychosocial functional
outcomes. The RFS was created by several state mental health commissions for
outcome evaluations of service provision. The measure was designed as a rapid
assessment tool for clinicians to rate the functioning of their clients (McPheeters,
1984). It is made up of four subscales (Work, Independent Living, Family, and
34
Social), in which raters score subjects on one of seven levels of functioning for the
area in question. A higher level (i.e. score) indicates better functioning in the domain
being measured. The RFS scored the highest on nine of twelve “utility scores” for
use with the chronically mentally ill (e.g. in ease of use, reliability, clinical
usefulness, sensitivity to change, etc.) when used and assessed by a consortium of
evaluators (Green & Gracely, 1987). It has also shown good reliability and
discriminant validity with a sample of African-American clinical and non-clinical
women (Goodman et al., 1993). The Chronbach’s alpha for this sample was 0.84.
The RFS was used successfully in three studies that each had nearly 50% ethnic
minority representation (Brekke & Barrio, 1997; Brekke et al., 2001; Brekke &
Long, 2000; Brekke et al., 1997) (predominantly African-American and Latino). The
ethnic composition of the original samples used during the design of the scale was
not listed. Interrater reliability was established using a protocol detailed in Brekke, et
al. (1993). Ratings achieved kappa .78 for all items (Brekke et al., 1993). Due to
potential confounding problems between the Family Contact (FC) and RFS-Family
subscale (RFS-F), the RFS-F was not included as part of the RFS outcome variable.
Analysis
As Kline (2005) notes, the structural equation model is a combination of both
a structural model (typically involving one or more latent variables) and
measurement model (observables or indicators hypothesized to “indicate” or measure
the latent variables). This SEM was tested by first analyzing the measurement model
and then the structural model using goodness-of-fit indices.
35
Structural equation modelling includes the same assumptions as are present
in multiple regression and path analyses (Loehlin, 1998). The present analyses
involve only interval-level measures. Frequencies, measures of central tendency and
dispersion, as well as correlation matrices, analyses for normality and heteroscedicity
are included for all indicators. These analyses were run using SPSS 11.5 with an
alpha level of .05 for the evaluation of statistical significance. The measurement and
structural models were tested for goodness-of-fit using AMOS 5.0.
Pre-analysis descriptives
A total of 93 cases were used for all analyses. The majority of cases (n=90)
had information from both a family member representative and the consumer related
to that family member. In three cases information was only available for the family
member (n=1) or for the consumer (n=2). Because of the Maximum likelihood
estimation (ML) technique these latter cases could be used for the analyses (Kline,
2005; Loehlin, 1998). Only one case from the original 94 cases had no information
from the consumer or the family member, and was therefore dropped from all
analyses. The issue of missing items was treated with an item by item mean-
imputation. That is, each item for every indicator was evaluated for percent of
missing items across subjects. The percentage of missing data per item ranged from
0 to 7% (Cohen & Cohen, 1983; Little & Rubin, 1987). Then items were added and
averaged for the mean of each indicator. This ensured that there were no missing
values for analyses of the modification indices used during the testing of the model
36
(see below) (Kline, 2005; Loehlin, 1998; McDonald & Ringo Ho, 2002). Table 2
summarizes descriptive statistics for each indicator by latent variable.
Table 2: Descriptives of indicators for latent variables
Family Pressure Indicator Descriptive Statistics
BAS
1
FPSE
2
FWB
3
Mean (SD) 40.73 (16.51) 31.72 (25.74) 21.26 (8.12)
Variance 272.61 662.53 65.89
Skewness .762 1.430 -.054
Kurtosis .811 2.793 -.281
Family Resources Indicator Descriptive Statistics
SSI
4
FMWB
5
FCSPIRIT
6
FCSEEK
7
Mean (SD) 44.03 (7.92) 44.45 (14.18) 16.62 (3.22) 14.50
(3.42)
Variance 62.69 200.99 10.40 11.69
Skewness -.009 -.212 -1.193 -.538
Kurtosis -.080 .396 1.094 .094
Family Dysfunction Descriptive Statistics
FAD
PS
8
FA
DC
9
FAD
RO
10
FAD
AR
11
FAD
AI
12
FAD
BC
13
Mean
(SD)
9.44
(2.31)
13.39
(2.04)
19.21
(2.77)
13.16
(2.64)
15.97
(3.11)
17.17
(3.70)
Variance 5.36 4.15 7.66 6.95 9.67 13.67
Skewness .105 -.994 -.392 -.412 -.419 -.254
Kurtosis .629 1.935 1.940 .270 .924 .074
Family CONTACT Descriptive Statistics
Mean
(SD)
79.01 (111.28)
Variance 12382.36
Skewness 3.112
Kurtosis 12.044
37
Table 2, Continued
Consumer Psychosocial Descriptive Statistics
RFSWORK
14
RFSINDLI
15
RFSSOC
16
Mean
(SD)
1.99 (1.74) 3.93 (1.56) 3.39 (1.97)
Variance 3.01 2.43 3.87
Skewness 1.792 .456 .173
Kurtosis 2.081 -.866 -1.384
1. Burden Assessment Scale
2. Family Pressure Scale – Ethnic (minority)
3. Financial Well-Being Scale
4. Social Support Index
5. Family Member Well-Being
6. Family Coping – Spirituality subscale
7. Family Coping – Seeking advice/help
8. McMaster Family Assessment Device (MFAD) – Problem Solving
9. MFAD – Communication
10. MFAD – Roles
11. MFAD – Affective Responsiveness
12. MFAD – Affective Involvement
13. MFAD – Behavior Control
14. Role Functioning Scale (RFS) - Work
15. RFS – Independent Living
16. RFS – Socialization
(N=93)
In addition, a Pearson’s correlation matrix was run to analyze bivariate
relationships across indicators used in the model specification. Refer to Table 3
below.
38
Table 3: Pearson’s correlation matrix for indicators used in model specification
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
1 1
2 **.47 1
3 *.32 **.33 1
4 **-.34 **-.34 *-.28 1
5 **-.29 **-.29 -.15 **.35 1
6 .06 -.12 -.15 .17 -.02 1
7 .08 .09 .06 **.28 -.03 **.48 1
8 *.26 .11 **.35 *-.21 -.11 -.01 -.12 1
9 -.06 .14 .14 -.12 -.12 .11 *.22 **. 30 1
10 **.30 **.37 **.28 **-.30 **-.30 .12 .12 *.26 **.27 1
11 **.40 **.28 .20 -.20 **-.34 .14 .10 **.39 **.30 **.56 1
12 *.22 .19 *.25 **-.32 **-.46 .10 .06 **.28 *.24 **.53 **.61 1
13 **.33 *.26 *.21 **-.34 **-.35 -.04 -.11 **.43 **.30 **.52 **.55 **.57 1
14 .15 **.31 .11 -.01 .11 -.03 *-.21 .14 -.10 .10 .02 .01 -.04 1
15 -.01 .04 .03 .04 -.03 -.01 *-.23 .06 -.01 -.00 -.02 -.07 -.06 **.35 1
16 .01 .19 .04 .04 .15 .05 -.09 -.04 -.11 -.11 -.06 -.14 .01 **.33 **.45 1
17. -.01 .01 .06 .10 .10 -.19 -.13 -.01 -.18 -.11 **-.29 **-.33 **-.28 **.40 **.32 *.25 1
* p<.05
** p<.01
1. Burden Assessment Scale
2. Family Pressure Scale – Ethnic
3. Financial Well-Being
4. Social Support Index
5. Family Member Well-Being
6. Family Coping Scale – Spirituality subscale
7. Family Coping Scale – Seeking Assistance subscale
8. McMaster Family Assessment Device (MFAD) – Problem Solving subscale
39
Table 3, Continued
9. MFAD – Communication subscale
10. MFAD – Roles subscale
11. MFAD – Affective Responsiveness subscale
12. MFAD – Affective Involvement subscale
13. MFAD – Behavioral Control subscale
14. Family Contact
15. RFS – Work subscale
16. RFS – Independent Living subscale
17. RFS – Social Skills
40
Results
There is no standard way of reporting the results of SEM analyses, which
creates inconsistencies within the literature (McDonald & Ringo Ho, 2002; Schreiber
et al., 2006). The following section follows the guidelines suggested by McDonald
and Ringo Ho (2002). The analysis of the model specification is evaluated in two
primary steps: first that of the measurement model and second that of the structural
(path) model. Thus, results of the measurement model are presented first and then
followed by the results of the structural model.
Measurement model results
The measurement model is treated as any other CFA model: latent variables
(like factors) are assumed to be correlated while indicators are tested (like items) to
load on factors as hypothesized (Loehlin, 1998). The analysis of a measurement
model tests the accuracy of the hypothesized relationships between latent variables
and their indicators (Kline, 2005; Loehlin, 1998; McDonald & Ringo Ho, 2002). It is
important to evaluate and establish the validity of these relationships before
analyzing relationships between latent variables (Kline, 2005; Loehlin, 1998).
Several “conditions” are necessary to meet in order to satisfy this first step: factor
loadings of observables must form independent clusters onto latent variables as
hypothesized; and each latent variable must have at least two independent identifiers
(indicators) that do not cross load onto other latent variables (Kline, 2005; Loehlin,
1998; McDonald & Ringo Ho, 2002). MacDonald & Ho (2002) label this as
“identifiability”. Goodness-of-fit indices are used to evaluate whether the
41
measurement model fits to the observed, that is, actual covariance matrix found in
the data. If the measurement model is a “good” fit, the difference between predicted
and observed is not statistically significant; and the model has good to very good fit
indices (Kline, 2005; Loehlin, 1998). If the results of the initial analysis indicate a
poor fit, then the model must be evaluated as to whether there are problems with the
indicators. Typically there are two main reasons for this problem: the indicators may
not reflect the construct that they are hypothesized to do so by either not loading
significantly (or substantially lower than other indicators for the factor) or represent
a different construct (Kline, 2005; Loehlin, 1998) . Although various SEM programs
offer modification indices that list changes to the model to improve the fit, these
changes are “data driven” and should only be used if there is a theoretical rationale
for doing so (Kline, 2005; Loehlin, 1998; McDonald & Ringo Ho, 2002).
The hypothesized measurement model (see Figure 1) was run resulting with
the following goodness-of-fit indices:
2
=165.943, df=95, p=.000, RMSEA=.09,
CFI=.81, and NFI=.66. The p-value indicates that there was a significant difference
between the predicted covariance matrix to that of the observed.
The Root Mean Square Error of Approximation (RMSEA) is a parsimony
adjusted index that favors models that are more parsimonious – simpler models are
favored over more complex ones (Kline, 2005; Loehlin, 1998). There is no specified
level for the RMSEA that indicates a good fit. However, the general consensus is
that a value of.05 indicates a close approximation, a value of .06 to .09 indicates
reasonable approximation, and value of .10 indicates poor approximation of error
42
(Kline, 2005) . The RMSEA for the hypothesized measurement model falls within
the reasonable approximation range. An advantage for using the RMSEA is that
many software programs provide a 90% confidence level so as to assess whether the
lower bound (.05) and upper bound (.10) of the confidence level exceeds the cut-
off level. If the lower bound value is equal to or lower than .05, then the hypothesis
of a good approximate fit can not be rejected. Likewise, if the upper level is equal to
or greater than .10 we can not reject the hypothesis of poor approximate fit (Kline,
2005). The RMSEA confidence level for the measurement model was .067
(LOWER) - .113 (UPPER). These values indicate that the hypothesis of a good
approximate fit can be rejected due to the lower bound value being greater than .05;
and the hypothesis of a poor approximate fit can not be rejected due to an upper
bound value being greater than .10.
The Comparative Fit Index (CFI) is an index that compares the predicted
model against a “null” or base model that assumes zero population variances across
all observed variables (Kline, 2005; Loehlin, 1998). Generally, most researchers
accept that any value that is.90 indicates a reasonable fit for the researcher’s model
(Kline, 2005). The CFI for the measurement model was .81. This would indicate that
it is not a reasonable fit to the observed data. Finally, the Normed Fit Index (NFI) is
similar to the CFI in that it tests the hypothesized model against a “normed” or
baseline model. If the fit is no better than that of a “null” or baseline (i.e. assuming
no relationships across observed variables and zero correlations) then the value of
the NFI is zero; if the fit is “perfect” than the NFI equals 1. Thus, a NFI closer to the
43
value of 1 indicates a better fit (Kline, 2005; Loehlin, 1998). The NFI for the
measurement model is .66.
Clearly, based on the p-value, the values of the RMSEA lower and upper
bounds, and on the CFI and NFI, this measurement model had a poor goodness-of-
fit. Preliminary EFA of each latent variable showed that there was a problem with
the latent variable Family Resources. Instead of the hypothesized one factor for all
indicators used, there were two factors. The two indicators from the F-COPES
subscales (Spirituality and Seeking Assistance) loaded as a second factor. A second
measurement model was run by creating two separate latent variables from the latent
variable Family Resources. The Social Support Index (SSI) and Family Member
Well-Being (FMWB) were used as indicators for the original Family Resources
latent variable while the two F-COPES subscales formed another latent variable for
Family Coping Strategies. This revised measurement model was inadmissible
because of negative variances for both indicators for the Family Coping Strategies
latent variable. Typically, negative variances result from problems with
multicollinearity across indicators (Kline, 2005). Review of the bivariate correlation
between the two indicators (Spirituality and Seeking Assistance) was r=.48. From a
theoretical perspective it is possible that, given the importance of spirituality for
many African-American families, seeking assistance may involve behaviors typically
associated with spirituality (e.g. seeking pastoral care, prayer, seeking help from
members of their church community, etc.) (Alston & Turner, 1994; McAdoo, 1998;
S. E. Williams & Finger Wright, 1992). Given this theoretical reasoning, the two
44
subscales were combined to form one indicator and the latent variable was dropped
in favor of using an observable variable for that concept.
The goodness of fit indices for this alternate measurement model were

2
=102.46, df=77, p=.03, RMSEA=.060. Although the results indicated a significant
difference between the predicted and observed covariance matrices, the value of the
RMSEA improved. The confidence level for the second measurement model was
.021 (LOW) - .089 (HIGH). The CFI for the second measurement model was .92 and
the NFI was .77.
A second version of the measurement model was assessed by dropping the
observable variable of Family Spiritualy+Seeking Assistance to see if the model fit
further improved. The goodness-of-fit indices were:
2
=83.501, df=68, p=.097,
RMSEA=.05 (CL: .000 - .083), CFI=.95, NFI=.78. This measurement model had
improved indices including the non-significant p value indicating that the predicted
and observed covariances were not significantly different from each other. Because
this second alternative measurement model was nested in the original version of the
measurement model, the two models could be compared to see whether the dropping
of the observed variable Family Spiritualy+Seeking Assistance was a significant
improvement (Kline, 2005; Loehlin, 1998). That is, when two models are nested, the
difference between their chi-square test statistics can be treated as a separate chi-
square statistic. Thus, using the difference in degree of freedom one can use the chi-
square distribution to test the value of the difference between the two chi-squares
45
against the critical value from the table (Rigdon, 1998; Steiger, Shapiro, & Browne,
1985). If the difference in
2
given the difference in degrees of freedom, is greater
than the critical value, then the alternative model is a significant improvement to the
original model (Kline, 2005; Loehlin, 1998). The change () in
2
was 83.442 and
df was 27. At the probability level of .05, the critical value is 40.11 (the critical
value at the .000 level is 46.96). Thus, the third alternative measurement model
minus the observed variable Family Spirituality+Seeking Assistance had a
significantly improved fit over the original measurement model. Table 4 summarizes
the goodness-of-fit indices for the original measurement model and the two
alternative measurement models (difference in chi-square tests are also noted).
Please refer to Table 4 below.
46
Table 4: Summary of goodness-of-fit indices for original and alternative
measurement models
Path Model

2
(df)
P RMSEA
(90% C.L.)*
CFI
1
NFI
2
Original (see
Figure 1)
165.94
(95)
.000
#
05.
(.07
§
-.11
^
)
.81 .66
Alternative
measurement
model 1
(Family
Spirituality and
Help Seeking
combined to
form
observable
variable)
102.46
(77)
.030
#
.06
(.021-.09)
.92 .77.
Alternative
measurement
model 2
(Family
Spirituality and
Help Seeking
dropped from
model - see
Figure 3)
83.50
(68)
S
.097 .05
(.00-.08)
.95 .78
* 90% RMSEA Confidence Level
1
Any value that is.90 indicates a reasonable fit (Kline, 2005).
2
A value equal to 1 is considered a “perfect” fit (Kline, 2005).
#
A significant p-value indicates that the path model is significantly different from
observed covariance matrix.
§
A lower bound level >.05 indicates that the hypothesis of a good approximate fit
can be rejected (Kline, 2005).
^
An upper limit value >.10 indicates that the hypothesis of poor approximate fit
cannot be rejected (Kline, 2005).
S
A significant improvement over original measurement model.
47
It is important to point out that except for the noted problem with the original
five indicators hypothesized to represent the latent variable Family Resources,
indicators for the other latent variables loaded in the hypothesized way. In addition,
each latent variable had at least two indicators, and there were no substantial cross-
loadings of indicators across latent variables per modification indices used to test for
any cross-loadings. As previously noted, both conditions are important in
establishing the conceptual validity of the measurement model and the simpler model
(alternative measurement model 2) met these necessary conditions of identifiably
(Kline, 2005; Loehlin, 1998; McDonald & Ringo Ho, 2002). Factor loadings for the
alternative measurement model 2 ranged from a low .45 (McMaster Family
Assessment Device – subscale Problem Solving on the Family Dysfunction factor) to
a high of .77 (McMaster Family Assessment Device – subscale Affective
Involvement on the Family Dysfunction factor).
The final measurement model (alternative model 2) used as the basis of the
structural model is provided in Figure 3 below.
48
Figure 3: Modified measurement model after initial analyses
Key for indicators:
1. Burden Assessment Scale
2. Family Pressure Scale – Ethnic
3. Financial Well-Being Scale
4. Social Support Index
5. Family Member Well-Being
6. McMaster Family Assessment Device (MFAD) – Problem Solving
7. MFAD – Communication
8. MFAD – Roles
9. MFAD – Affective Responsiveness
10. MFAD – Affective Involvement
11. MFAD – Behavior Control
12. Role Functioning Scale (RFS) - Work
13. RFS – Independent Living
14. RFS – Socialization
Goodness-of-fit indices:

2
=83.501
df=68
p=.097
RMSEA=.05 (CL: .000 - .083)
CFI=.95
NFI=.78
Family
Resources
10
13 12 14
Family
Contact
11
3
2
1
5 4
9 8
7
6
Family
Pressures
Family
Dysfunction
Consumer
Psychosocial
Functioning
49
Path (structural) model
Once the measurement model has reasonably goodness-of-fit indices, with
any modification principally based on theoretical considerations, then the second
step is to analyze the structural model or the hypothesized relationships between the
latent variables (Kline, 2005; Loehlin, 1998). The relationships between latent
variables are “structured” in that correlations are dropped and directional paths are
added. Certain paths are free to estimate while other paths are set to zero (that is, not
included thus assuming no relationship between the latent variables). The present
structural model is recursive in that all hypothesized relationships between latent
variables are unidirectional and all disturbance terms are uncorrelated (Kline, 2005).
The hypothesized or predicted structural model is then analyzed for goodness-of-fit.
If the predicted structural model has a poor fit with the observed then the researcher
has the option to evaluate how the structural model might be modified. As was the
case with the measurement model, modification can be generated by empirical as
well as theoretical rationales. However, any modification must involve theoretical
rationales given that, when paths are either added or dropped, one is then offering a
new hypothesis regarding the relationships between latent variables (Kline, 2005;
Loehlin, 1998). Depending on whether any modifications are made, hierarchical
relationships are tested to analyse whether the changes between nested models is
significant. Because of the possibility of improved model fit due to more
hypothesized relationships (simply due to chance), more parsimonious models are
favoured over more complex ones (Kline, 2005; Loehlin, 1998).
50
In addition, alternative theoretical models will be analyzed. A favourable
goodness-of-fit does not signify that the proposed model is indeed representative of
the population in question. It merely means that the predicted and observed models
are not significantly different and that the goodness-of-fit indices indicate an at least
acceptable approximation to what is observed. Hence, it is important to consider the
possibility of other theoretically relevant models that might have as good or an
improved goodness-of-fit (Kline, 2005; Loehlin, 1998; McDonald & Ringo Ho,
2002).
The path model (refer to Figure 2 above), as previously presented, tests the
relationships between family factors and amount of family contact. The model
assumed that any effect that family latent variables had on Consumer Psychosocial
Functioning was indirect. Family latent variables might have a direct effect on
Consumer Psychosocial Functioning, but it was decided to first test direct effects on
Family Contact given its previous strong positive direct effect on Consumer
Psychosocial Functioning in the original path-analysis model as noted in Guada &
Brekke (2007).
The goodness of fit indices for the hypothesized path model were
2
=88.63,
df=71, p=.08, RMSEA=.05. The results indicated that there was no significant
difference between the predicted and observed covariance matrices. In addition, the
value of the RMSEA indicated a close approximation to what was observed in the
data. The 90% RMSEA confidence level for the model was .000 (LOW) - .084
(HIGH). Given that the lower bound value was lower than .05 the hypothesis of a
51
good approximate fit could not be rejected. Likewise, the upper level was less than
.10, thus we could reject the hypothesis of poor approximate fit (Kline, 2005). The
CFI was .95 (recall that most researchers accept that any value that is.90 indicates
a reasonable fit for the researcher’s model). The NFI was .78 indicating an
acceptable fit (Kline, 2005; Loehlin, 1998).
Overall the hypothesized path model demonstrated a good fit. However, the
only direct path that was significant in the model was between Family Contact and
Consumer Psychosocial Functioning (b=.005, p=.000). None of the direct paths from
family latent variables to the Family Contact variable were significant. On the other
hand, all of the covariances between the family latent variables were significant. The
correlations for these relationships were: Family Pressure and Family Resources (r=-
.77, p=.001), Family Pressure and Family Dysfunction (r=.58, p=.000), and Family
Resources and Family Dysfunction (r=-.74, p=.000). Clearly, these correlations are
high, which might indicate a problem of colinearity (Kline, 2005; Loehlin, 1998).
Nonetheless, none of the indicators for these family latent variables cross-loaded nor
were any of the error terms correlated in the measurement model. Hence, it appeared
that while the family latent variables were closely related, they were conceptually
distinct enough to remain as separate concepts within the model. Again, the path
model had good to very good goodness-of-fit indices; but because only one direct
path was significant, the model offered little in conceptual information.
The findings of the first path model presented two parallel processes: amount
of family contact was significantly related to consumer psychosocial functioning
52
while family factors were significantly correlated to each other with no direct effect
on amount of family contact. The next model attempted to test whether these two
sets of factors were related in an alternative way. That is, did other family factors
(e.g. Family Pressures, Family Resources, and Family Dysfunction) have a direct
effect on Consumer Psychosocial Functioning like amount of family contact did?
This alternate path model tested direct paths from the family latent variables
to Consumer Psychosocial Functioning (while keeping the originally significant
relationships between Family Contact and Consumer Psychosocial Functioning). The
specific hypotheses for each path model were similar to those of the original path
model. That is, Family Pressures would have a negative direct association with
Consumer Psychosocial Functioning (more Family Pressures would be associated
with poorer Consumer Psychosocial Functioning); Family Resources would have a
positive direct association with Consumer Psychosocial Functioning (more Family
Resources would be associated with improved Consumer Psychosocial Functioning);
and Family Dysfunction would have a direct negative association with Consumer
Psychosocial Functioning (more Family Dysfunction would be associated with
poorer Consumer Psychosocial Functioning). Please refer to Figure 4, which presents
this model in graphic form.
53
Figure 4: Alternative path model 1 with family latent variables hypothesized to have
direct effects on consumer psychosocial functioning.
The goodness of fit indices for the alternative path model were
2
=94.87,
df=71, p=.03, RMSEA=.06. The results indicated that there was a significant
difference between the predicted and observed covariance matrices. In addition, the
value of the RMSEA indicated a close approximation to what was observed in the
data although not an improved value from the original hypothesized path model’s
RMSEA’s of .05. The 90% RMSEA confidence level was .000 (lower bound) - .090
(upper bound). The CFI was .93 and the NFI was .77. Thus, the alternate model had
somewhat poorer fit indices as compared to the original model (e.g. the p-value
indicated a significant difference between the hypothesized relationships and what
was observed in the data). As with the original hypothesized path model, none of the
direct effects from latent family variables to Consumer Psychosocial Functioning
Consumer
Psychosocial
Functioning
Family
Dysfunction
Family
Pressures
Family
Contact
+
_
_
+
Family
Resources
54
were significant although the direct path from Family Dysfunction to Consumer
Psychosocial Functioning was the closest to any level of significance (b=-.18,
p=.145).
Kline (2005) and Loehlin (1998) advocated for parsimonious models to avoid
finding better fitting indices, primarily due to chance, because more complex models
have more modeled paths. Thus, the next step was to model fewer family latent
variables on Consumer Psychosocial Functioning. Alternative path model 2 included
only Family Dysfunction as an exogenous variable while keeping Family Contact as
the other exogenous variable. Family Dysfunction was chosen because it was the
only family factor that approached significance in the previous tested model and
because it was conceptually most similar to what was tested in the EE literature (that
is, some sort of dysfunction in the family impacts consumer functioning). The
relationships between exogenous variables to Consumer Psychosocial Functioning
were the same as hypothesized in previous path models: see Figure 5 below.
Figure 5: Alternate path model 2
Consumer
Psychosocial
Functioning
Family
Dysfunction
Family
Contact
+
_
55
The goodness of fit indices for alternative path model 2 were
2
=34.37,
df=26, p=.13, RMSEA=.06. The results indicated that there was no significant
difference between the predicted and observed covariance matrices. In addition, the
value of the RMSEA indicated a close approximation to what was observed in the
data. The 90% RMSEA confidence level for the model was .000 (LOW) - .108
(HIGH). The high value for the upper bound of the confidence level was over the
>.10 indicating that the hypothesis of poor approximate fit could not be rejected
(Kline, 2005; Loehlin, 1998). The CFI was .96 (a desirable value is.90). The NFI
was .86 indicating an acceptable fit (Kline, 2005; Loehlin, 1998). In addition, the
path from Family Dysfunction to Consumer Psychosocial Functioning was
significant (b=-.145, p=.02) as was the direct effect from Family Contact to
Consumer Psychosocial Functioning (b=.007, p=.000). Thus, the results are
somewhat mixed: the overall fit indices demonstrated an acceptable fit and all direct
effects were significant. However, the RMSEA confidence level showed that the
hypothesis of poor approximate fit could not be rejected.
Clearly, alternative path model 2 offered a “template” to model relationships
that included other family factors. Based on a model by Pearlin (Pearlin, 1991;
Pearlin et al., 1997) – that the resources available to a family can influence its overall
functioning – the family factor Family Resources was added to the model. Family
Resources was modeled as an exogenous variable with a hypothesized direct effect
on Family Dysfunction. Direct paths from Family Contact and Family Dysfunction
to Consumer Psychosocial Functioning remained in the model. The relationships
56
between Family Resources to Family Contact and Consumer Psychosocial
Functioning were set to zero because of findings from the previous path models. See
Figure 6, alternative path model 3 below.
Figure 6: Alternate path model 3
The goodness of fit indices for this third alternate path model were
2
=47.72,
df=42, p=.25, RMSEA=.04. The results indicated that there was no significant
difference between the predicted and observed covariance matrices. In addition, the
value of the RMSEA indicated a close approximation to what was observed in the
data. The 90% RMSEA confidence level for the model was .000 (lower bound) -
.084 (upper bound). The CFI was .98 and the NFI was .84 indicating an acceptable
fit.
The difference in chi-square was analyzed to see if alternative path model 3
was a significant improvement over the original path model in explaining the
observed data. The difference in chi-square was 88.63 (original model) – 47.72
Consumer
Psychosocial
Functioning
Family
Dysfunction
Family
Contact
Family
Resources
+
_
_
57
(alternative model 3) or
2
=40.413. The difference in degrees of freedom was 71
(original model) – 42 (alternative model 3)df =29. The critical value at the p=.05
level for adf is 42.56. Thus, the alternate model 3 was not a significant
improvement over the original model.
Another alternative model was tested. Alternative model 4 was the same as
alternative model 3, but added the final family latent variable (Family Pressures) to
the path model. Family Pressures was treated as an exogenous variable for the other
two family latent variables (Family Resources and Family Dysfunction) (recall that
Family Pressures was non-significant as an exogenous variable for Family Contact
and Consumer Psychosocial Functioning in previous models). It was hypothesized
that as Family Pressures increased, Family Resources would decrease and amount of
Family Dysfunction would increase. See Figure 7, alternative path model 4 below.
Figure 7: Alternative path model 4
Consumer
Psychosocial
Functioning
Family
Contact
Family
Resources
+
+
_
_
_
Family
Pressures
Family
Dysfunction
58
The goodness-of-fit indices for alternative path model 4 were:
2
=95.287,
df=73, p=.04, RMSEA=.06, RMSEA confidence level (.012-.088), CFI=.93,
NFI=.77. Alternative path model 4 was significantly different from the observed
covariance matrix with good to very good values for all other fit indices. The direct
path from Family Pressures to Family Resources was significant (b=-.31, p=.000)
while the direct path from Family Pressures to Family Dysfunction was not
significant.
One last alternative path (alternative model 5) model was tested by setting to
zero the non-significant direct path from Family Pressures to Family Dysfunction
while keeping all other significant paths from the previous model. The goodness-of-
fit indices for this last alternative model were:
2
=95.292, df=74, p=.05,
RMSEA=.056, RMSEA confidence level (.005-.086), CFI=.93, NFI=.77. The fit
indices were not substantially different from alternative path model 4 although the p-
value approached non-significance. Nonetheless, the direct paths between all latent
variables were significant suggesting a plausible path model for explaining
relationships between these variables for this sample.
Was the alternative model 5 a significant improvement from the original
hypothesized path model? The difference in chi-square was 95.292 (alternative
model 5) – 88.63 (original model) or
2
=6.66. The difference in degrees of
freedom was 74 (alternative model 5) – 71 (original model)df =3. The critical
59
value at the p=.05 level for adf is 7.82. Thus, the alternate model 5 was not a
significant improvement over the original model.
Table 5 summarizes the goodness-of-fit indices for the original path model
and the four alternative path models: differences in chi-square tests are also noted.
Table 5: Summary of goodness-of-fit indices for original and alternative path models
Path Model

2
(df)
P RMSEA
(90% C.L.)*
CFI
1
NFI
2
Original (see
figure 2)
88.63 (71) .08 05.
(.000-.084 )
.95 .78
Alternative
path model 1
(see figure 4)
94.87 (71) .03
#
.06
(.000 - .090)
.93 .77.
Alternative
path model 2
(see figure 5)
34.37 (26) .13 .06
(.000- .108
^
)
.96 .86
Alternative
path model 3
(see figure 6)
47.72
(42)
NS
.25 .04
(.000-.084)
.98 .84
Alternative
path model 4
(see figure 7)
95.287
(73)
.04
#
.06
(.012-.088)
.93 .77
Alternative
path model 5
(see Figure 8)
95.292
(74)
NS
.05 .056
(.005-.086)
.93 .77
* 90% RMSEA Confidence Level
1
Any value that is.90 indicates a reasonable fit (Kline, 2005).
2
A value equal to 1 is considered a “perfect” fit (Kline, 2005).
#
A significant p-value indicates that the path model is significantly different from
observed covariance matrix.
60
Table 5, Continued
^
An upper limit value >.10 indicates that the hypothesis of poor approximate fit
cannot be rejected (Kline, 2005).
NS
Not a significant improvement over original path model
In summary, the results showed the following based on a two-step analysis
process.
The originally hypothesized measurement model did not “fit” the data until
two indicators (for the F-COPES subscales on family spirituality and family help-
seeking behaviors) were dropped. This may be a measurement artifact in that there
were psychometric issues with the F-COPES subscales although the F-COPES
received extensive testing with minority populations without any problems
(McCubbin et al., 1996). The revised measurement model, which dropped the F-
COPES subscales, was a significant improvement over the initial model with good to
very good goodness-of-fit indices:
2
=83.501, df=68, p=.097, RMSEA=.05, RMSEA
confidence level (.000 - .083), CFI=.95, NFI=.78. This latter, significantly improved
model was used as the basis of the path (structural) model.
The originally hypothesized path model, where correlated family factors were
modeled as having a direct effect on amount of Family Contact (and Family Contact
was modeled as having a direct effect on Consumer Psychosocial Functioning) had
acceptable to very good goodness-of-fit indices:
2
=88.63 (df=71), p=.08,
RMSEA=05, RMSEA CL (.000-.084), CFI=.95, and NFI=.78. Family Contact had a
positive and significant direct effect on Consumer Psychosocial Functioning.
61
Nonetheless, none of the latent family variables (Family Pressures, Family
Dysfunction, and Family Resources) had a significant direct effect on amount of
Family Contact. Only two of the original hypotheses were confirmed: Family
Contact had a significant, positive direct effect on Consumer Psychosocial
Functioning and family factors were significantly correlated with each other. Despite
the very good goodness-of-fit indices, the original hypothesized path model did not
offer any new conceptual information. However, although no family factors had a
direct effect on amount of Family Contact, their presence in the path model
contributed to the acceptable goodness-of-fit indices suggesting that they had some
important impact on the fit amongst the variables not modeled in the original path
model.
An alternative path model tested family factors modeled as having a direct
effect on Consumer Psychosocial Functioning (rather than on Family Contact) while
keeping the direct path from Family Contact to Consumer Psychosocial Functioning.
The model had average goodness-of-fit indices:
2
=94.87, df=71, p=.03,
RMSEA=05, RMSEA confidence level (.000-.084), CFI=.93, and NFI=.77. None of
the direct paths from the latent family variables to Consumer Psychosocial
Functioning were significant although the direct path from Family Dysfunction to
Consumer Psychosocial Functioning approached significance. Correlations between
family factors remained significant as did the direct path from Family Contact to
Consumer Psychosocial Functioning. Again, this model offered nothing conceptually
new than the originally hypothesized path model.
62
Several other alternative path models were then tested. Alternative model 3
(see Figure 6 above) included a significant positive direct path from Family Contact
to Consumer Psychosocial Functioning (the more the family contact the better the
consumer psychosocial functioning), a significant negative direct path from Family
Dysfunction to Consumer Psychosocial Functioning (the more dysfunctional the
family the worse the consumer psychosocial functioning), and a significant negative
direct path from Family Resources to Family Dysfunction (the fewer the family
resources the worse the family dysfunction). Its goodness-of-fit indices were:

2
=47.72, df=42, p=.25, RMSEA=.04, RMSEA confidence level (000 - .084),
CFI=.98, and NFI=.84. However, alternative path model 3 was not a significant
improvement over that of the original path model. Nonetheless, alternative path
model 3 offered more conceptual information about the relationships across the
variables than did the originally hypothesized path model.
Alternative path model 4 (see Figure 7 above), which added the last latent
family variable Family Pressures as an exogenous variable to the other two family
factors, was tested. The direct path from Family Pressures to Family Resources was a
statistically significant negative direct effect: as Family Pressures increased (such as
level of burden, financial difficulties, and discrimination difficulties) the family’s
resources decreased. Family Pressures did not have a significant direct effect on
Family Dysfunction. Alternative path model 4’s goodness-of-fit indices were:
63

2
=95.287, df=73, p=.04, RMSEA=.06, RMSEA confidence level (.012-.088),
CFI=.93, NFI=.77.
Alternative path model 5 set the direct path from Family pressures to Family
Dysfunction to zero, but kept all other paths from alternative model 4. It had
acceptable goodness-of-fit indices:
2
=95.292, df=74, p=.05, RMSEA=.056,
RMSEA confidence level (.005-.086), CFI=.93, NFI=.77. This final model offered
the most conceptual information as to how factors related to each other while having
acceptable goodness-of-fit indices. The model demonstrated that family factors had
both direct (amount of family contact, amount of family dysfunction) and indirect
(amount of family resources and family pressures) on consumer psychosocial
functioning. Model five demonstrated how each of these variables contributed to
consumer functioning while also demonstrating how family factors interacted with
each other. Please refer to figure 8 for the final path model and Table 6, which lists
the final path estimates.
64
Figure 8: Final path model testing family factors on consumer psychosocial
functioning

2
=95.292 (df=74)
p=.05
RMSEA=.056
RMSEA confidence level (.005-.086)
CFI=.93
NFI=.77
Table 6: Summary of estimates for alternative path model 5
Regression Weights
Beta S.E. Stand.
Beta
C.R. P
Family
Pressures
Family Resources -.313 .088 -.787 -3.56 .00
Family
Resources
Family Dysfunction -.446 .117 -.744 -3.82 .00
Family
Dysfun
Consumer
Psychosocial
Functioning
-.136 .058 -.309 -2.33 .02
Family
Contact
Consumer
Psychosocial
Functioning
.007 .002 .635 4.47 .00
b=.007
p=.000
b=-.14
p=.02
b=-.45
p=.000
b=-.31
p=.000
Family
Resources
Family
Dysfunction
Family
Pressures Consumer
Psychosocial
Functioning
Family
Contact
65
Table 6, Continued
Variances
Estimate S.E. C.R. P
Family Pressures 132.306 42.601 3.10
6
.002
Family Resources 8.000 4.687 1.70
7
.088
Family Dysfunction 3.362 1.251 2.68
7
.007
Family Contact 12249.214 1806.048 6.78
2
.000
Consumer
Psychosocial
Functioning
.734 .345 2.12
8
.033
Discussion
The study extended the literature on family influences on consumer
functioning in the following ways. The purpose of the study was to see what family
factors that, based on the EE and caregiver literatures, might contribute to the
amount of beneficial contact between consumers and families. It extended a previous
study that found that the more contact a consumer had with the family the better the
consumer functioning (Guada & Brekke, 2007). These factors included Family
Dysfunction, Family Pressures, and Family Resources. It tested a domain of
consumer functioning (psychosocial functioning) more applicable to outpatient
consumers (Guada & Brekke, 2007) and that has not before been extensively tested
with family related variables. It used a more advanced statistical methodology
(SEM) that provided the ability to analyze all the hypothesized relationships
concurrently and thereby assess how well the model “fit” the observed relationships
66
for this sample. Finally, it tested these relationships on a larger sample of African-
American consumers and their families not typically seen, especially in the EE
literature.
The importance of family contact for consumer psychosocial functioning
No matter which family domains were included in the path model, the
relationship between amount of family contact and the consumer’s psychosocial
functioning remained positive and significant. Hence, the more contact that the
consumer had with her or his family, the better the psychosocial functioning was.
This replicated a previous finding from a path model that found that amount of
family contact had a significant impact on consumer psychosocial functioning
(Guada & Brekke, 2007). Ongoing contact is helpful in and off itself. This contact
offers a beneficial effect on the consumer’s ability to work, live independently, and
to socialize.
Although none of the original family factors (Family Dysfunction, Family
Resources, Family Pressures) had direct effects on amount of family contact or on
Consumer Psychosocial Functioning, when Family Dysfunction was modelled as the
only family factor (along with amount of family contact) with a direct effect on
Consumer Psychosocial Functioning, the relationship was significant. The level of
family dysfunction had a negative direct effect on psychosocial functioning: the
more dysfunction the worse the psychosocial functioning. This is reminiscent of how
level of Expressed Emotion (as a specific “kind” of family dysfunction) has a
negative impact on a consumer’s level of relapse and re-hospitalization (Hooley,
67
1985; S. King & Dixon, 1996; J. Leff & Vaughn, 1985). A family’s ability to
problem solve, communicate, be affectively involved with each other, etc. does
effect how well a consumer does in such areas as work, socialization, and
independent living skills. This occurs even given the beneficial significant effect that
amount of family contact had on the consumer’s psychosocial functioning.
Nonetheless, these results show that for these poorer African-American consumers,
at least some interaction or contact with their families – no matter how those families
function – is crucial to the consumer’s well-being out in the community. On the other
hand, even given this important contact, some families with problems of functioning
may need interventions so that any deleterious effects their functioning might have
on consumers are attenuated.
Thus, amount of family contact was consistently associated with improved
consumer psychosocial functioning and regardless of such things as the family’s
level of functioning, the resources available to the family for coping, and the
pressures that families were faced with. What seemed most important was that the
consumer had contact with the family, whether this was with telephone calls or in-
person visits.
Two family contextual processes
The final path model demonstrated that the amount of family pressures and
resources indirectly affected consumer psychosocial functioning by directly affecting
the family’s level of functioning. The model demonstrated that there was a direct,
significant effect from Family Pressures to Family Resources (as amount of Family
68
Pressures went up, the amount of Family Resources went down) and a direct
significant direct effect from Family Resources to Family Dysfunction (as amount of
Family Resources went down the level of Family Dysfunction went up). Thus, the
story told by this model was that as the amount of the family’s pressures increased,
the family’s resources diminished; as the amount of their resources decreased, the
family’s functioning worsened and as the family’s functioning worsened, the
consumer’s psychosocial functioning worsened. At the same time amount of family
contact continued to have a significant beneficial impact on consumer functioning
despite these relationships across the other family factors.
Hence, two family contextual processes differently impacted consumer
psychosocial functioning at the same time. But, these are not contradictory because
even families that have clinical problems (and even if these cause problems for the
consumer) can be a resource for the person living in the community. Ongoing
contact with the family is a resource, in and of itself, for the consumer. This is a
powerful finding since it suggests that the “ties that bind” (as one grandmother put it)
for these poorer African-American consumers and their families transcend both
concrete (e.g. financial, discrimination, social supports) and abstract (e.g. sense of
burden, sense of well-being) barriers to family interactions. White, middle-class
consumers and their families tend to report more difficulties when faced with similar
challenges (Boye et al., 1999; Stueve et al., 1997; Wuerker et al., 1999).
In addition, these findings are reminiscent of findings within the caregiver
burden and family treatment literatures. That is, as families experience more
69
pressures and have fewer resources, their ability to adequately function as a family
diminishes (Hatfield & Lefley, 1993; McCubbin et al., 1995; Pearlin, Mullan,
Semple, & Skaff, 1990; Reinhard et al., 1994). In addition, this adds indirect support
for stress models such as Pearlin’s, which postulates that as the level of stress
multiplies there is a cascading and reverberating effect across several spheres of a
person’s or family’s life. A sense of stress heightens in an exponential process of
decreasing resources and poorer coping.
Practice implications
The study has implications for intervention models. Practice interventions
targeted toward families from poorer, African-American populations should include
issues around discrimination, financial resources, and increasing social networking,
issues not typically included in family-oriented psychoeducation programs, which
have tended to focus solely on communication styles and educating families about
major mental illness (Dixon et al., 2000). These latter topics are obviously as
important for poorer African-American families, but the findings suggest these
programs need to include other issues to assist any families faced with ongoing
pressures and lack of resources. Addressing these issues will help the overall
family’s functioning, which in turn will assist the consumer’s functioning in the
community. More specifically, interventions for consumers and their families should
include increasing a family’s own level of social support & directly assisting with
any concrete pressures (e.g. financial issues or institution-based discrimination).
These interventions should also include decreasing the burden that families
70
experience through the use of psychoeducation, mutual aid groups, respite while
encouraging ongoing contact between consumer & family – this helps the consumer
with those areas of functioning most important for living in the community.
Addressing family issues – most germane to the family’s own needs – will help the
overall family’s functioning, which in turn will assist the consumer both directly &
indirectly. These issues need to be added to present psychoeducation family models
in order to be most effective with families from similar communities.
Thus, from this sample of poorer African-American consumers & their
families a clearer picture emerges about how the family context impacts the quality
of consumers’ lives: despite substantial challenges faced by the families, they are a
major resource for the well-being of these consumers; these are consumers
embedded in an extended family context, that is itself embedded in a surrounding
neighborhood context reminiscent of what is sometimes called in the literature
“familism”.
Familial ties unrelated to spirituality?
Likewise, it is unclear why subscales on spirituality and seeking help were
problematic for the study’s original measurement model given the plethora of
literature regarding the importance of spirituality and mutual help-seeking behaviors
within the African-American community (Alston & Turner, 1994; Brekke & Guada,
2004; McAdoo, 1998; Pickett et al., 1993; Robinson, 1983; Turner & Alston, 1994;
S. E. Williams & Finger Wright, 1992). On the other hand, perhaps a consumer’s
psychosocial functioning or contact with family is unrelated to a family member’s or
71
family unit’s spirituality or religious activities. In this way, one might think of family
contact as unrelated to specific spiritual or religious beliefs and/or practices because
emotional and geographic ties between family members are important regardless of
any religious belief system or practice (Alston & Turner, 1994; McCabe et al., 2003;
Pinquart & Sorensen, 2005; Rivera et al., 1997; Solomon & Draine, 1995). Future
studies are needed to clarity this issue.
Study limitations
The study had several limitations. Firstly, it was cross-sectional in design. It
provided a “snap-shot” of this particular sample just prior to beginning services in a
family based intervention. A longitudinal study offers the advantage of evaluating
any changes in the relationships between exogenous and endogenous variables
across time. Depending on the trajectory of these relationships, services could focus
on either supporting ongoing contact between families and consumers and/or assist
families and consumers to improve interactions if a negative trajectory were present.
It also offers the opportunity to assess in greater detail how family factors (such as
family pressures and overall functioning) change across time and how these changes
directly and/or indirectly effect the psychosocial functioning of outpatient
consumers. Cross-sectional studies are limited in establishing causation.
Longitudinal analyses of these relationships are needed to clarify directionality and
causality (Berry & Feldman, 1985; Kachigan, 1986; Kline, 2005; Knoke, Bohrnstedt,
& Potter Mee, 2002).
72
Another limitation is sample size, which was not ideal for SEM statistical
methodology although the sample was larger than what has been used in EE and
many caregiver burden mental health studies (Baronet, 1999; Butzlaff & Hooley,
1998; Fadden et al., 1987; Hashemi & Cochrane, 1999a; Horwitz & Reinhard, 1995).
Sample sizes of 100 and more (ideally around 200-300) are recommended to
increase power (better able to reject an incorrect model), model specification (paths
may be non-significant due to small sample size as opposed to poor specification),
and decrease potential technical problems with the analyses (Kline, 2005). Thus, the
sample size might account for some of the non-significant findings between family
and consumer functioning variables. Nonetheless, the present study included nearly
100 consumers and nearly 100 family members in the analyses.
Another potential issue is that there were different reporters across the
indicators used for the latent variables. There could be reporter biases with each
indicator that therefore effects the relationships between latent variables
(Association, Association, & Education, 1999; DeVellis, 1991; McIver & Carmines,
1981). If only a family member, consumer, or an interviewer completed the scales,
then relationships between variables may have been different. As previously noted,
some studies show that when multiple members of a family report on the same
phenomenon there is some divergence in the responses (Bogels & Brechman-
Toussant, 2006).
Conclusion
73
The relationships that occurred in this study confirm previous findings while
offering new information that is possibly unique for African-American or other
outpatient consumers.
Because the findings are novel for the consumer outcomes literature, future
studies are needed to test the same model across other ethnic and SES groups to
assess if these findings are unique to poorer African-American families and
consumers or to other ethnic and SES consumers living in the community.
In addition, future studies can test other models involving the same family
factors and consumer outcomes, but extend the analysis by including moderator
effects. For example, future models can test whether the direct effect from family
dysfunction to psychosocial functioning of consumers is moderated by amount of
family contact. It is also possible that the relationship between family contact and
consumer psychosocial functioning is moderated by family dysfunction. These
future studies will clarify whether the current study’s findings (of the beneficial
direct effect of amount of family contact and the concurrent deleterious direct effect
of amount of family dysfunction on consumer psychosocial functioning) is an
artefact of this particular structural model. Despite the novelty of the findings from
this study it is but the first step in “teasing out” or unravelling the puzzle of how
family factors interact with and impact each other while having both direct and
indirect effects on consumer psychosocial functioning.
Finally, because of the importance of amount of family dysfunction for
consumer functioning, it is imperative that future research test the factor structure of
74
the MFAD with African-American families. The use of the MFAD with African-
American families is relatively new and thus there is a need to confirm that the
concepts of clinical dysfunction developed primarily with white samples holds for
African-American samples (especially for the sub-scale on Communications, which
had some psychometric problems).
It is clear that the relationship that the outpatient consumer has with her or his
family is important for the consumer’s overall psychosocial functioning. This
relationship is important even when family dysfunction has a negative effect on the
consumer’s psychosocial functioning. Nonetheless, improving family functioning by
decreasing the pressures that they face will both directly and indirectly assist
outpatient poorer African-American consumers in the critical areas of work,
socialization, and independent living skills. Future studies will assist in determining
whether these relationships are replicated with other African-American samples,
whether these findings change over time, and/or are present in the lives of other
ethnic and SES groups of consumers and their families.
75
Chapter 2: Second Structural Equation Model
Introduction
The concept that family functioning has an etiological impact on consumer
functioning has been a focus of attention since the inception of family therapy (C. M.
Anderson et al., 1986; Brown et al., 1962; J. P. Leff, 1976; Minuchin & Fishman,
2004; Nichols & Schwartz, 2005). One of these theories, that of Expressed Emotion
(EE), is one of the most investigated concepts in the family mental health literature
(Butzlaff & Hooley, 1998; Hashemi & Cochrane, 1999a; Kavanagh, 1992). The most
consistent finding from the EE literature is that EE (or a specific component of it
such as Critical Comments) is predictive of consumer functioning. In this case a
higher level of EE predicts increased levels of relapse and/or rehospitalization
(Bebbington & Kuipers, 1994; Butzlaff & Hooley, 1998; Hooley, 1985; J. Leff &
Vaughn, 1985). Many of the original researchers in EE also hypothesized that
amount of family contact was an important variable for relapse and/or
rehospitalization. The reasoning here was that the more contact a consumer had with
a family with high EE, the more problems the consumer would have with relapse
and/or rehospitalization (Brown et al., 1962; Hooley, 1985; J. P. Leff, 1976).
Of particular note is that the EE findings are robust and are found regardless
of the ethnicity of the family although the level of EE that is predictive is not
uniform (Jenkins & Karno, 1992; Moline et al., 1985). For example, the level of the
component Critical Comments (CC) must be higher for EE to be predictive for
76
African-American consumers (Moline et al., 1985; Wuerker et al., 1999). However,
these findings are based on small samples (less than 40 families), which are the only
studies to specifically investigate Expressed Emotion with African-American
consumers and their families known to the author (Moline et al., 1985; Wuerker et
al., 1999). Clearly, the state of the EE literature in regards to African-American
consumers and their families requires replication with a larger sample.
Another area of interest in the EE literature is testing possible variables that
might have a direct effect on the amount of EE (or CC). One such hypothesis is that
the level of psychiatric symptomatology from the consumer effects the family’s level
of EE (Barrowclough & Hooley, 2003; Boye et al., 1999; Hatfield, 1997a; S. King &
Dixon, 1996). Other researchers have tested to see if there are specific family
variables (e.g. communication style) that predict the level of EE (Kavanagh, 1992).
Few if any studies have tested other kinds of family variables (e.g. overall family
functioning, family resources, family pressures) influence on level of EE or its
component parts. So while the caregiver burden literature demonstrates that issues of
family burden, resources, and pressures are important variables to family functioning
(Baronet, 1999; Demi et al., 1997; Hatfield & Lefley, 1993; Nabors, Seacat, &
Rosenthal, 2002; NAMI, 2003; Saldana et al., 1999) these have only been included
in few studies testing level of EE (Jackson et al., 1990; Scazufca & Kuipers, 1996).
One of the major gaps in the EE literature is its almost exclusive focus on
those either recently hospitalized or discharged and recently diagnosed with
schizophrenia (Butzlaff & Hooley, 1998; Hashemi & Cochrane, 1999b; Hooley,
77
1985; Kavanagh, 1992). Yet, with the onset of deinstitutionalization more than 40
years ago as well as substantially shorter hospital stays due to managed care and
more restrictive public funding, the vast majority of consumers spend less and less
time in a hospital setting (Geller, 2000; Lamb & Bachrach, 2001; Mechanic, 1999).
The question is whether the same sorts of findings are evident amongst outpatient
consumers and their families. Would the level of EE or CC have a deleterious impact
on the functioning of the outpatient consumer?
Some may argue that the influence of the family for consumers living out in
the community is of less salience than for those populations (i.e. inpatient) focused
on in the EE literature. However, research shows that the importance of family for
consumers may vary dependent on ethnicity and/or socio-economic status (Alston &
Turner, 1994; Barrio, 2001; Billingsley, 1990; Brekke & Barrio, 1997; Solomon &
Draine, 1995; Weick & Saleebey, 1995). The literature shows that the family is
deemed of greater importance for Latino, Asian-American, and African-American
communities than what is seen among many people of European descent (Barrio,
2001; Brekke & Barrio, 1997; Guarnaccia & Parra, 1996). Hence, the attitudes,
functioning, and challenges of families could be of greater relevance on the
functioning of outpatient consumers from these communities.
Thus, there are several important “next steps” needed within the EE
literature: there is a need to test EE-related findings with a larger sample of African
American consumers; there is a need to test other family factors’ direct effect, (e.g.
the pressures they face, the amount of family resources available to the family, their
78
overall functioning), on amount of EE or one of its component parts; and there is a
need to test these relationships for a sample of consumers living in the community,
particularly for consumers from communities where their ongoing connection with
the family is very important.
The present study addresses each of these “next steps” in the following ways.
The study includes a larger sample of African-American consumers and their
families than previously used in the EE literature. It tests a wider range of family
factors to see what direct effect these have on the level of family criticalness towards
the consumer. The specific family factors included are the overall functioning of the
family, amount of family resources (i.e. social support and well-being of the family),
and the pressures families are faced with (e.g. burden, financial strains, macro-
systemic pressures unique for ethnic minority families such as ongoing
discrimination), which were selected based on findings and hypotheses from the
caregiver burden literature (Baronet, 1999; Pickett et al., 1993; Solomon & Draine,
1995; Stueve et al., 1997). In addition, the study includes outpatient consumers, not
typically used in the EE literature. To reiterate, because the sample is of African-
American consumers it is argued that family factors are more salient for this
population given the importance of the family for this community (Billingsley, 1990;
Guarnaccia & Parra, 1996; Hines & Boyd-Franklin, 2005; Robinson, 1983; Wallace
Williams et al., 2003). Additionally, the study uses structural equation modeling
(SEM) to test these relationships. SEM is a more advanced statistical methodology
that provides the ability to concurrently test relationships not heretofore used in
79
either the EE or burden literatures (Barrowclough & Hooley, 2003; Butzlaff &
Hooley, 1998; Fadden et al., 1987; Loehlin, 1998).
Although the present study does not test EE directly, it does use the amount
of family criticalness as the endogenous variable for predicting the consumer’s level
of psychiatric symptomatology. As previously noted, Critical Comments was the
most predictive sub-construct of EE for African-American subjects (Moline et al.,
1985; Wuerker et al., 1999). Hooley and Teasdale (1989) developed a two-question
instrument, (Perceived Criticism) that is both briefer and can be used as a proxy for
Critical Comments (CC) (Hooley, 1998; Hooley & Teasdale, 1989). The present
study uses Perceived Criticism (PC) as a proxy for Critical Comments (CC). In
addition, because of the importance of the amount of contact that a consumer had
with his or her family in the original EE literature, the amount of family contact is
included as an exogenous variable for PC. To summarize the two main research
questions were: do family factors (e.g. resources, pressures, level of functioning)
directly affect amount of family criticalness toward the consumer? And, does this
criticalness have a deleterious effect on consumer clinical functioning as already
demonstrated with smaller African-American samples?
The specific hypotheses for the SEM are divided into two parts: the
Measurement Model and the Structural Model.
80
Measurement model hypotheses
Hypothesis 1: factor loadings will form independent clusters from each other
(McDonald & Ringo Ho, 2002). That is, at least two indicators will exclusively load
for every latent variable without cross-loading on other latent variables.
Hypothesis 2: the predicted measurement model’s relationship between
latent variables and their indicators will fit the observed data as evidenced by at least
acceptable goodness-of-fit indices.
Figure 9: Hypothesized measurement model of family variables and consumer
psychiatric functioning
Structural model hypotheses
The structural model hypothesizes that family variables will have a
significant direct effect on amount of Perceived Criticism (i.e. family criticalness
toward the consumer). Perceived Criticism will have a significant direct effect on
Perceived
Criticism
Family
Contact
Family
Pressures
Family
Dysfunction
Family
Resources
Consumer
Psychiatric
Functioning
81
consumer psychiatric functioning. The proposed model can be “deconstructed” to the
following specified relationships between latent variables:
Hypothesis 1: Perceived Criticism (PC) will have a significant direct effect
on consumer psychiatric functioning: the higher the level of PC (more family
criticalness) the more psychiatric symptoms the consumer will have.
Hypothesis 2: Family Dysfunction will have a significant positive direct
effect on PC: the higher the amount of family dysfunction, the higher the level of PC.
Hypothesis 3: Family Pressures will have a positive direct effect on amount
of Perceived Criticism (PC): the higher the level of family pressure, the higher the
level of PC.
Hypothesis 4: Family Resources will have a negative direct effect on amount
of PC: the more family resources there are, the lower the PC.
Hypothesis 5: Family Contact will have a direct and positive effect on PC:
the more the Family Contact, the higher the amount of PC.
Hypothesis 6: The four family variables (Family Pressures, Family
Resources, Family Dysfunction, and Family Contact) will significantly correlate with
each other.
Hypothesis 7: the predicted structural model’s relationship across latent
variables will fit the observed data as evidenced by at least acceptable goodness-of-
fit indices. That is, the structural model will not significantly differ from the
observed data, and other goodness-of-fit indices will be at least in the “acceptable”
fit range (Schreiber et al., 2006).
82
Figure 10: Hypothesized structural model for family latent variables on consumer
psychiatric functioning
Methods
Study Context
The sample for this study came from a project that was investigating client
and family outcomes from a home-based intervention for families with a seriously
mentally ill member (schizophrenia spectrum disorders). The intervention and the
research protocol are described elsewhere (Connery & Brekke, 1999). Families were
recruited if their ill family member was receiving services at the participating county
contract mental health facility. The study was approved by the University of
Southern California Institutional Review Board, and the Los Angeles County
Department of Mental Health. All study participants signed informed consent to
participate in the study.
Family
Pressures
Consumer
Psychiatric
Functioning
Family
Contact
Family
Resources
+
+
+
+
-
Perceived
Criticism
Family
Dysfunction
83
Sample
The total project sample consisted of 106 families. This report presents data
from the 94 African-American families who were assessed at baseline before the
family intervention began. A family reporter was selected in conference with the
consumer. This family reporter was a member of the household who filled out all of
the family measures at baseline. The vast majority of family reporters were women
(86.2 %) with a mean age of 47 years old (range from 18 to 80). Thirty-six percent
were parents to the consumer, 22% were a sibling, while 21% were a child to the
consumer, and 7% were a spouse. The remainder (14%) was “other”, which was
typically a close friend or an extended family member (e.g. grandparent, aunt/uncle,
or cousin). This diversity of relationships is typical for non-white, non-middle class
samples in other studies involving a family member and a person with a major
mental disorder (Guarnaccia & Parra, 1996; Pickett et al., 1993).
The average age of the consumer was 42 years old (age range of 23 – 64
years of age). Consumers lived with a family member on average 45 (sd=72) days in
the previous six months before entering the study. The distribution for living with
family was skewed: 59 consumers (63%) reported not living with a family member
in the preceding six-month period while 14 consumers (15%) lived with the family
the entire time. Thus, 63% of consumers lived independently of any family members.
Nonetheless, the vast majority of consumers lived in the same neighborhood as their
families (e.g. in same building, next street over, on the same block, etc.).
84
In addition, the average number of days in which the consumer was
hospitalized for psychiatric problems was 8 (sd=25) days for the preceding six-
month period. This variable’s distribution was also skewed with 72 consumers (82%)
having no days on inpatient psychiatric treatment. Thus, the vast majority of
consumers remained out of a hospital and lived continuously in the community for
the previous six months.
Consumer mean monthly income was 1036.45 (or approximately $12,432 per
year) making this a poor sample of adults. Families resided in one of the poorest
sections of a major urban area. As of 1992 the median income for African Americans
in this area was $22,000 with some areas having as low as $12,000 as the median
income (Civil, 1993). The 1992 estimated income figures were used because families
were enrolled and data was collected during the later 1990’s. Table 7 summarizes
descriptive statistics for the sample below.
Table 7: Descriptive statistics of sample
Consumer Family Representative
Age (SD) 43 (42) 47 (49)
Gender (%) Female (56) Female (85)
Relationship Type
(%)
Child (35)
Parent (19)
Sibling (27)
Spouse (6)
Other (13)
Parent (35)
Child (19)
Sibling (27)
Spouse (6)
Other (13)
Income 1027.95* $12K – 22K
85
Table 7, Continued
Days with family
#
45
NOTE: 63% lived
independently in
preceding 6 month
period.
NA
Days inpatient
#
8 (SD=24)
NOTE: 82% out of the
hospital in preceding 6
month period.
NA
Number of
residences
#
2 (SD=1) NA
* Per month
#
In preceding six-month period
Measures
Altogether there were six variables in the structure model (refer to Figure 9
and Figure 10). Four of these variables were latent and had at least two indicators.
Three of the latent variables (Family Pressure, Family Resources, and Family
Dysfunction) referred to experiences or functioning of the family unit as a whole.
The scales used as indicators for these latent variables were answered by the family
member representative. Two variables were observable variables, and referred to the
level of family criticalness of the consumer (Perceived Criticism) and amount of
family contact. These were answered by the consumer. The fourth latent variable
(Consumer Psychiatric Functioning) was specific to the psychiatric functioning of
the consumer (level of symptoms). It was also filled out by the consumer with the
assistance of an interviewer.
86
Family Pressure indicators
Family Pressure referred to both concrete (financial) and subjective (burden)
pressures that a family might face when interacting or managing a family member
diagnosed with a major mental health disorder. A higher score on each of the
indicators demonstrated greater amount of the pressure being measured. Family
Pressure had three indicators.
The Family Pressure Scale – Ethnic (FPRES-E) is 64-item scale developed to
measure the pressures unique to the life experiences of ethnic minority families
(McCubbin et al., 1996). A higher score indicates greater family pressure,
specifically those unique to ethnic minority families (e.g. interpersonal and
institutional discrimination). The original internal reliability (Chronbach’s alpha)
was .92 (McCubbin et al., 1996). The FPRES-E was the strongest predictor of family
difficulties in a study with minority families (Native Hawaiian) thus suggesting the
validity of the measure (McCubbin et al., 1996). The alpha for this study was .94.
The Family Inventory of Resource Management (FIRM) is a self-
administered 66-item scale designed to both describe and assess a family’s amount
and management of their resources (McCubbin et al., 1996). It was specifically
created for use with samples of ethnic minority families (McCubbin et al., 1996;
McCubbin et al., 1998). The FIRM is divided into four sub-scales. Only the
Financial Well-Being (FWB) subscale was used and refers to the family’s ability to
meet financial commitments, the adequacy of financial resources, the ability to assist
87
others financially, and optimism regarding the financial future of the family. A
higher score indicates greater resource management. The Chonbach’s alpha for this
sample was .85. The FWB was recoded so that a higher score indicated greater
financial problems.
The Burden Assessment Scale (BAS) is a 19-item self-administered scale that
measures the caregiver burden of family members with a seriously mentally ill
member of the family (Reinhard et al., 1994). A higher score signifies greater
burden. Chronbach’s alpha for initial studies using the Burden Assessment Scale
were .91 and .89, and data on the validity of the Burden Assessment Scale are
available in Reinhard et al., (1994). The alpha on the Burden Assessment Scale for
this study was .94. The BAS has been used successfully with African-American
samples (Horwitz & Reinhard, 1995).
Family Resources indicators
Family Resources referred to both social and an over all sense of well-being
that a family had available to itself to help the family manage or cope including
dealing with a family member diagnosed with a major mental health disorder. A
higher score on each of the indicators demonstrated greater amounts of the resources
available to the family. Family Resources had two indicators.
The Social Support Index (SSI) is a 17-item self-administered scale that
measures to what degree the family finds support and integration within the
community (McCubbin et al., 1996). It uses a 5-point Likert scale ranging from
“Strongly Disagree” to “Strongly Agree”. Chronbach’s alpha for initial studies using
88
the SSI was .82 (McCubbin et al., 1996). The alpha on the SSI for this study was .75.
In addition, the SSI demonstrated discriminate validity when it negatively correlated
with a scale on family distress with a sample of ethnic minorities (McCubbin et al.,
1995).
The Family Member Well-being Index (FMWB) is an 8-item self-
administered questionnaire that measures how much a family member feels adjusted
regarding issues about health, energy, feelings of concern, cheerfulness, etc.
(McCubbin et al., 1996). Each item is a 10-point Likert scale ranging from “Not
(very much)” to “Very (much)”. The Cronbach’s alpha for the original development
of the subscale was .85. Standard scores, means, and standard deviations were tested
on numerous ethnic groups including African-American families (McCubbin et al.,
1996). The alpha on the FMWB for this study was .74.
Family Dysfunction indicators
The third latent variable was Family Dysfunction. It referred to a family’s
problems to function as a family unit regardless of the presence of stressors or not. A
higher score on each of the indicators demonstrated greater clinical problems with
family functioning in that domain. It was measured by six indicators, which are the
sub-scales of the MFAD.
The McMaster Family Assessment Device (MFAD) is a 53-item self-
administered questionnaire designed to assess families according to the McMaster
Model of Family Functioning (Epstein et al., 1983). The McMaster Model of Family
Functioning is, “a clinically oriented conceptualization of families…It describes
89
structural and organizational properties of the family group and patterns of
transactions among family members which have been found to distinguish between
healthy and unhealthy families” (Epstein et al., 1983, p. 172). A higher score
indicates more problems/more dysfunction. The McMaster Family Assessment
Device (MFAD) is made up of six subscales including a seventh subscale for overall
“General Functioning”, which was not used in this study.
Each of the subscales is designed to measure a different dimension of family
functioning. Miller et al. (2000) described the six subscales in the following way.
The first subscale is “Problem-solving”, which, as defined by the Model, is “…a
family’s ability to resolve problems at a level that maintains effective family
functioning” (p. 170). The second subscale is “Communication”, which is defined as
the family’s ability to exchange or move important information within a family
system with a focus specifically on “verbal exchange” (p. 171). The third subscale is
“Roles”. The Model defines this as the “recurrent patterns of behavior[s]” (p. 171)
that take place between family members and which influence the functioning of the
family unit; that is, the subscale measures how clear and often family members fulfill
specific roles in order to problem solve. The fourth subscale is “Affective
Responsiveness”, which refers to the quantity and quality of the emotions
experienced and expressed amongst family members in response to stressors and
other stimuli from the environment. The fifth subscale, “Affective Involvement”,
refers to the amount of interest and focus that family members offer to each other
around the specific, individual interests and needs of family members. The last
90
subscale is “Behavior”. This is defined as “the pattern a family adopts for handling
behavior in three types of situations” (p. 172). The three types of situations are:
physically dangerous circumstances, expressing and meeting psychobiological needs,
and situations that involve interpersonal socializing needs.
The validity of the McMaster Family Assessment Device has been examined
in several studies (Epstein et al., 1983; Miller et al., 1990). It showed both concurrent
and discriminate validity in the expected directions. The original Chronbach alpha’s
for the six subscales ranged from .72 - .92 (Epstein et al., 1983). The reliability was
also analyzed with test-retest (Miller et al., 1985) with results ranging from .66 to
.76. The alphas subscales for this sample were the following: Problem Solving (.80),
Communications (.40), Roles (.60), Affective Responsiveness (.77), Affective
Involvement (.78), and Behavior Control (.82). The alpha level for the sub-scale
Communications was low. However, by dropping one item the alpha increased to
.53. In addition, the ethnic composition of the original samples used during the
design of the scale was not listed (Friedmann et al., 1997; Kabacoff et al., 1990; C.
King et al., 1997; Miller et al., 1990).
Family Contact
Contact with relatives was a simple count variable of number of times that
the consumer had contact with her or his family via phone or personal contact. The
amount of contact that the subject had was measured by several questions asking
how much contact, in hours over the previous two months, each subject had with
parents or other relatives. For the present study the sum of all contacts (whether
91
initiated by the consumer or family member and whether the contact was with a
parent or with another relative) was used given the extended family form suggested
by the diversity of which family member was chosen as the family’s respresentative.
Perceived Criticism
The primary and best means for measuring the construct Expressed Emotion
is the Camberwell Family Interview (CFI) (Chambless et al., 1999; Van Humbeeck
et al., 2002). Chambless and colleagues (1999) argue that Expressed Emotion is
actually an umbrella construct for several sub-constructs. One of these, Critical
Comments (CC), is the most consistently predictive for relapse (Hooley, 1998)
particularly for African-American samples (Moline et al., 1985; Wuerker et al.,
1999). As previously noted, Hooley and Teasdale (1989) developed a two-question
instrument, (Perceived Criticism) that is both briefer and can be used as a proxy for
Critical Comments (CC) (Hooley, 1998). As previously noted, it was used as a proxy
for Critical Comments (CC) in this study.
Perceived Criticism (PC) includes two self-report items, the first of which
asks the subject to assess how critical the family is toward her or him. It uses a 10-
point Likert-type scale where ‘1’ represents “not at all critical” to ‘10’ – “very
critical”. The second item asks the subject to assess her or his own criticalness
toward the family on a similar 10-point scale.
Perceived Criticism (PC) showed strong concurrent validity with the
Camberwell Family Inventory (CFI) while showing discriminant validity with non-
EE instruments (Hooley & Teasdale, 1989; Riso et al., 1996; Van Humbeeck et al.,
92
2002). Van Humbeeck et al. (2002) concluded that PC had the best predictive power
besides the CFI based on a review of the literature for EE and EE-related
instruments. The PC scale was used principally with other diagnostic groups (Riso et
al., 1996), thus, there is a need for its use with a sample of adults with schizophrenia
as well as with an African-American sample (Van Humbeeck et al., 2002). Given
that this study was primarily interested in how consumers perceived their families’
criticalness towards them, only the first item for PC was used in the path model
analyses. Initial studies for the PC scale did not list what the ethnic make-up was of
the original samples.
Consumer Psychiatric Functioning
The primary outcome for the majority of the EE literature has been relapse
and rehospitalization (Barrowclough & Tarrier, 1990; Inoue et al., 1997; Stricker et
al., 1997). Although much of the EE literature includes rehospitalization as an
outcome or control variable for Critical Comments, a previous path analysis that
included number of hospitalizations and Perceived Criticism showed no relationship
for this same sample (Guada & Brekke, 2007). For this study then “clinical” outcome
refers to a consumer’s level of psychiatric symptomatology (primarily psychosis and
anxiety symptoms) and whether Perceived Criticism is predictive for any variation in
this symptomatology.
Colorado Symptom Inventory
The Colorado Symptom Inventory (CSI) was selected because it has been
used with success in previous studies on psychosocial rehabilitation service
93
outcomes (Brekke et al., 1993; Brekke & Long, 2000; Brekke et al., 1997), and
because it was also used to examine variation in cross-ethnic outcomes (Brekke et
al., 1997; Phillips et al., 2001).
The CSI (Shern et al., 1994) assesses symptom severity. It is a multi-item,
self-report scale on which consumers rate themselves on frequencies of their
experiences of specific symptoms. It has two broad categories for symptoms: anxiety
and psychosis. It was modified from a similar scale used in the Denver Community
Mental Health Questionnaire (Shern et al., 1994). In the present study each subscale
was treated as a separate indicator (i.e. CSI-anxiety and CSI-psychosis); a higher
score indicates more symptoms. The alphas for each scale were: Anxiety (=.85) and
Psychosis (=.84).
Analysis
As Kline (2005) notes, the structural equation model is a combination of both
a structural model (typically involving one or more latent variables) and
measurement model (observables or indicators hypothesized to “indicate” or measure
the latent variables). This SEM was tested by first analyzing the measurement model
and then the structural model using goodness-of-fit indices.
Structural equation modelling includes the same assumptions as are present
in multiple regression and path analyses (Loehlin, 1998). The present analyses
involve only interval-level measures. Frequencies, measures of central tendency and
dispersion, as well as correlation matrices, analyses for normality and heteroscedicity
are included for all indicators. These analyses were run using SPSS 11.5 with an
94
alpha level of .05 for the evaluation of statistical significance. The measurement and
structural models were tested for goodness-of-fit using AMOS 5.0.
Pre-analysis descriptives
A total of 94 cases were available for the analyses. The majority of cases
(n=90) had information from both a family member representative and the consumer.
In three cases information was only available for the family member (n=1) or for the
consumer (n=2). Because of the Maximum likelihood estimation (ML) technique
these latter cases could be used for the analyses (Kline, 2005; Loehlin, 1998). One
case had no information from either the consumer or family member and was
dropped from the analyses resulting in an n=93. The issue of missing items was
treated with an item by item mean-imputation. That is, each item for every indicator
was evaluated for percent of missing items across subjects. The percentage of
missing data per item ranged from 0 to 7%, which was in an acceptable range for
missingness (Cohen & Cohen, 1983). Then items were added and averaged for the
mean of each indicator. This ensured that there were no missing values for analyses
of the modification indices used during the testing of the measurement model (Kline,
2005; Loehlin, 1998; McDonald & Ringo Ho, 2002). Table 8 summarizes descriptive
statistics for each indicator by latent variable.
95
Table 8: Descriptives of indicators of latent variables
Family Pressure Indicator Descriptive Statistics
BAS
1
FPS-E
2
FWB
3
Mean (SD) 40.73 (16.51) 31.72 (25.74) 21.26 (8.12)
Variance 272.61 662.53 65.89
Skewness .762 1.430 -.054
Kurtosis .811 2.793 -.281
Family Resources Indicator Descriptive Statistics
SSI
4
FMWB
5
Mean (SD) 44.03 (7.92) 44.45 (14.18)
Variance 62.69 200.99
Skewness -.009 -.212
Kurtosis -.080 .396
Family Dysfunction Descriptive Statistics
MFAD
PS
6
MFAD
C
7
MFAD
R
8
MFAD
AR
9
MFAD
AI
10
MFAD
BC
11
Mean (SD) 9.44
(2.31)
13.39
(2.04)
19.21
(2.77)
13.16
(2.64)
15.97
(3.11)
17.17
(3.70)
Variance 5.36 4.15 7.66 6.95 9.67 13.67
Skewness .105 -.994 -.392 -.412 -.419 -.254
Kurtosis .629 1.935 1.940 .270 .924 .074
Perceived Criticism
Mean (SD) 4.74 (3.45)
Variance 11.70
Skewness .29
Kurtosis -1.36
Family Contact Descriptive Statistics
Mean (SD) 79.01 (111.28)
Variance 12382.36
Skewness 3.112
Kurtosis 12.044
96
Table 8, Continued
Consumer Psychiatric Descriptive Statistics
CSI - A
12
CSI - P
13

Mean (SD) 14.55 (5.87) 18.85 (8.87)
Variance 34.51 78.60
Skewness .491 .962
Kurtosis -.655 -.003
1. Burden Assessment Scale
2. Family Pressure Scale – Ethnic (minority)
3. Financial Well-Being Scale
4. Social Support Index
5. Family Member Well-Being
6. McMaster Family Assessment Device (MFAD) – Problem Solving
7. MFAD – Communication
8. MFAD – Roles
9. MFAD – Affective Responsiveness
10. MFAD – Affective Involvement
11. MFAD – Behavior Control
12. Colorado Symptom Inventory (CSI) – Anxiety subscale
13. CSI – Psychosis subscale
In addition, a Pearson’s correlation matrix was run to analyze bivariate
relationships across indicators used in the model specification. Please refer to Table
9 for specific bivariate relationships below.
97
Table 9: Pearson’s correlation matrix for indicators used in model specification
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
1 1
2 **.47 1
3 **.32 **.33 1
4 *.26 .11 **.35 1
5 -.06 .15 .15 **.31 1
6 **.33 **.37 **.31 *.28 *.28 1
7 **.38 *.27 .19 **.37 **.33 **.55 1
8 *.24 *.21 *.26 **.32 **.28 **.57 **.64 1
9 **.34 *.25 .20 **.46 **.30 **.53 **.58 **.63 1
10 **-.34 **-.34 **-.28 *-.21 -.11 **-.28 *-.22 **-.33 **-.37 1
11 **-.29 **-.29 -.15 -.12 -.13 **-.33 **-.35 **-.49 **-.36 **.35 1
12 .15 **.31 .11 .14 -.10 .11 .00 .03 -.05 -.01 .11 1
13 -.04 .02 -.07 -.15 -.06 .12 -.09 -.15 -.07 .03 .03 -.15 1
14 -.18 .02 -.03 .04 .11 .08 .06 .02 .02 .06 .19 .12 **.27 1
15 -.17 .09 -.08 -.08 .02 .09 .00 -.11 -.00 .03 .21 .06 **.36 **.81 1
** Correlation significant at the 0.01 level (2-tailed).
* Correlation significant at the 0.05 level (2-tailed).
1. Burden Assessment Scale
2. Family Pressures Scale – Ethnic
98
Table 9, Continued
3. Financial Well-being
4. McMaster Family Assessment Device (MFAD) – Problem Solving
5. MFAD – Communication
6. MFAD – Roles
7. MFAD – Affective Responsiveness
8. MFAD – Affective Involvement
9. MFAD – Behavior Control
10. Social Support Index
11. Family Member Well-Being
12. Family Contact
13. Perceived Criticism
14. Colorado Symptom Inventory (CSI) – Anxiety
15. CSI – Psychosis
99
Results
The following section follows the guidelines suggested by McDonald and
Ringo Ho (2002) because there is no standard way of reporting the results of SEM
analyses. The analysis of the model specification is evaluated in two primary steps:
first that of the measurement model and second that of the structural (path) model.
Thus, results of the measurement model are presented first and then followed by the
results of the structural model.
Measurement model results
The measurement model is treated as any other CFA model: latent variables
(like factors) are assumed to be correlated while indicators are tested (like items) to
load on factors as hypothesized (Loehlin, 1998). Goodness-of-fit indices are used to
evaluate whether the measurement model fits to the observed, that is, actual
covariance matrix found in the data. If the measurement model is a “good” fit, the
difference between predicted and observed is not statistically significant (Kline,
2005; Loehlin, 1998). If the results of the initial analysis indicate a poor fit, then the
model must be evaluated as to whether there are problems with the indicators.
Typically there are two main reasons for this problem: the indicators may not reflect
the construct that they are hypothesized to do so by either not loading significantly
(or substantially lower than other indicators for the factor) or represent a different
construct (Kline, 2005; Loehlin, 1998). The measurement model requires that
specific beta weights as well as the possibility of cross-loadings are analyzed.
Although various SEM programs offer modification indices that list changes to the
100
model to improve the fit, these changes are “data driven” and should only be used if
there is a theoretical rationale for doing so (Kline, 2005; Loehlin, 1998; McDonald &
Ringo Ho, 2002).
To reiterate each indicator must load on the hypothesized latent variable it is
meant to represent (with few to no cross-loadings) and each latent variable must have
at least two indicators exclusive to it (Kline, 2005; Loehlin, 1998; McDonald &
Ringo Ho, 2002). MacDonald & Ringo Ho (2002) label this as “identifiability”.
The hypothesized measurement model (see Figure 9) was run resulting with
the following goodness-of-fit indices:
2
=98.51, df=77, p=.05, RMSEA=.06,
CFI=.95, and NFI=.81. Factor loadings ranged from a low .39 (McMaster Family
Assessment Device – subscale Communication on the Family Dysfunction factor) to
a high of .93 (Colorado Symptom Index – Psychosis subscale on the Consumer
Psychiatric Functioning factor). The p-value indicated that there was a significant
difference between the predicted covariance matrixes to that of the observed.
The Root Mean Square Error of Approximation (RMSEA) is a parsimony
adjusted index that favors models that are more parsimonious – simpler models are
favored over more complex ones (Kline, 2005; Loehlin, 1998). There is no specified
level for the RMSEA that indicates a good fit. However, the general consensus is
that a value of.05 indicates a close approximation, a value of .06 to .09 indicates
reasonable approximation, and.10 indicates poor approximation of error (Kline,
2005) . The RMSEA for the hypothesized measurement model falls within the
reasonable approximation range. An advantage for using the RMSEA is that many
101
computer programs provide a 90% confidence level so as to assess whether the lower
bound of confidence (.05) and upper bound (.10) of the confidence level exceeds
the cut-off level. If the lower bound value is equal to or lower than .05, then the
hypothesis of a good approximate fit can not be rejected. Likewise, if the upper level
is equal to or greater than .10 we can not reject the hypothesis of poor approximate
fit (Kline, 2005). The RMSEA confidence level for the measurement model was .01
(LOWER) - .09 (UPPER). These values indicate that the hypothesis of a good
approximate fit can not be rejected due to the lower bound value being less than .05.
In addition, the hypothesis of a poor approximate fit can be rejected because the
upper bound value was less than .10.
The Comparative Fit Index (CFI) is an index that compares the predicted
model against a “null” or base model that assumes zero population variances across
all observed variables (Kline, 2005; Loehlin, 1998). Generally, most researchers
accept that any value that is.90 indicates a reasonable fit for the researcher’s model
(Kline, 2005). The CFI for the measurement model was .95. This indicated that it
was a reasonable fit to the observed data. Finally, the Normed Fit Index (NFI) is
similar to the CFI in that it tests the hypothesized model against a “normed” or
baseline model. If the fit is no better than that of a “null” or baseline (i.e. assuming
no relationships across observed variables and zero correlations) then the value of
the NFI is zero; if the fit is “perfect” than the NFI equals 1. Thus, a NFI closer to the
value of 1 indicates a better fit (Kline, 2005; Loehlin, 1998). The NFI for the
measurement model is .81.
102
Despite the p-value being equal to .05 that indicated that the model was
significantly different from the observed model, the other goodness-of-fit indices
indicated that the hypothesized measurement model had a relatively good fit to that
of the data. In addition, indicators for the latent variables loaded in the hypothesized
way. Each latent variable had at least two exclusive indicators to itself, and there
were no substantial cross-loadings of indicators across latent variables per
modification indices used to test for these. As previously noted, both conditions are
important in establishing the conceptual validity of a measurement model (Kline,
2005; Loehlin, 1998; McDonald & Ringo Ho, 2002). The tested measurement model
(as seen in Figure 9) was used as the basis for the structural model because the basic
requirements of identifiably were met.
Path (structural) model
The second step is to analyze the structural model or the hypothesized
relationships between the latent variables (Kline, 2005; Loehlin, 1998). The
relationships between latent variables are “structured” in that correlations are
dropped and directional paths are added. Certain paths are free to estimate while
other paths are set to zero (that is, not included thus assuming no relationship
between the latent variables). The present structural model is recursive in that all
hypothesized relationships between latent variables are unidirectional and all
disturbance terms are uncorrelated (Kline, 2005). The hypothesized or predicted
structural model is then analyzed for goodness-of-fit. If the predicted structural
model has a poor fit with the observed then the researcher has the option to evaluate
103
how the structural model might be modified. As was the case with the measurement
model, modification can be generated by empirical as well as theoretical rationales.
However, any modification must involve theoretical rationales given that, when
paths are either added or dropped, one is then offering a new hypothesis regarding
the relationships between latent variables (Kline, 2005; Loehlin, 1998). Depending
on whether any modifications are made, hierarchical relationships are tested to
analyse whether the changes between nested models are significant. Because of the
possibility of improved model fit due to more hypothesized relationships (simply due
to chance), more parsimonious models are favoured over more complex ones (Kline,
2005; Loehlin, 1998).
In addition, alternative theoretical models were analyzed. A favourable
goodness-of-fit does not signify that the proposed model is indeed representative of
the population in question. It merely means that the predicted and observed models
are not significantly different and that the goodness-of-fit indices indicate an at least
acceptable approximation to what is observed. Hence, it is important to consider the
possibility of other theoretically relevant models that might have as good or an
improved goodness-of-fit (Kline, 2005; Loehlin, 1998; McDonald & Ringo Ho,
2002).
Recall that the main interest was to test what might predict Perceived
Criticism from variables measuring family related functioning, resources, and
pressures. The model was designed to include family factors previously tested in the
caregiver burden literature. The model also included factors similar to those assumed
104
to influence EE: these factors represent family functioning issues such as
communication and affective involvement between family members (Jackson et al.,
1990; Scazufca & Kuipers, 1996). Because the EE literature viewed amount of
family contact on EE as important, the observed variable Family Contact was
included as another exogenous variable on level of family criticalness.
The second major research question was: does this criticalness have a
deleterious effect on consumer clinical functioning as already demonstrated with
smaller African-American samples? It was expected that Perceived Criticism would
have a direct effect on Consumer Psychiatric Functioning given what the EE
literature has demonstrated regarding EE’s and Critical Comment’s predictive impact
on relapse and rehospitalization, particularly for African-American samples. The
initial hypothesized model assumed that any effect that family latent variables had on
Consumer Psychiatric Functioning was indirect with Perceived Criticism mediating
these relationships. Family latent variables might have a direct effect on Consumer
Psychiatric Functioning, but it was decided to first test direct effects on Perceived
Criticism (PC) given the findings of the EE literature for Critical Comments (using
PC as a proxy). In addition, the relationships between family latent variables were
modeled as correlations because the main question was what directly effects amount
of PC.
When the path model was initially run one of the indicators for Consumer
Psychiatric Functioning (CSI – Psychosis) had a negative variance (
2
=-6.40). This
sort of finding may be due to multicollinearity between the two indicators for the
105
factor that they were hypothesized to represent (Kline, 2005; Loehlin, 1998). The
correlation between CSI – Anxiety and CSI – Psychosis was r=.81. This
demonstrated a clear multicollinearity problem (Kachigan, 1986; Knoke et al., 2002).
However, the CSI – Anxiety indicator had a loading of .60 indicating that it
adequately loaded on its hypothesized factor separate from the other indicator (CSI –
Psychosis). In addition, the CSI was designed to measure two broad areas of
symptomatology (anxiety and psychosis) and thus can be used as two separate
subscales (Shern et al., 1994). Smaller negative variances can be treated as sampling
error if the confidence interval around the negative variance includes a positive value
(J. C. Anderson & Gerbing, 1988). The standard error of the variance for CSI –
Psychosis was 16.783. Thus, the range, plus or minus, around the variance included
positive values and indicates a probable sampling error issue. Anderson and Gerbing
(1988) recommend that the negative variance is set at an arbitrary number close to
zero to “correct” the problem. The variance for indicator CSI – Psychosis was set to
.01. This constraint was used in all subsequent path model analyses.
The goodness of fit indices for the path model were
2
=103.54, df=82, p=.05,
RMSEA=.05. The results indicated that there was a significant difference between
the predicted and observed covariance matrices. Nonetheless, the value of the
RMSEA indicated a close approximation to what was observed in the data. The 90%
RMSEA confidence level for the model was .00 (LOW) - .08 (HIGH). The lower
bound value was lower than .05; thus, the hypothesis of a good approximate fit could
not be rejected. Likewise, the upper level was less than .10, which meant that the
106
hypothesis of poor approximate fit could be rejected (Kline, 2005). The CFI was .95
(recall that most researchers accept that any value that is.90 indicates a reasonable
fit for the researcher’s model). The NFI was .80 indicating an acceptable fit (Kline,
2005; Loehlin, 1998). Despite the p-value being equal to .05, the other goodness-of-
fit indices indicated that the hypothesized measurement model had a relatively good
fit to that of the data.
The direct path from Perceived Criticism to Consumer Psychiatric
Functioning was significant (b=.933, p=.000). Thus, with an increase in PC the
symptomatology of the consumer also increased, which is conceptually similar to
what is reported in the EE literature. Three correlations between family factors were
significant: Family Pressures and Family Contact (r=33, p=.03), Family Pressures
and Family Dysfunction (r=.55, p=.01), and Family Pressures and Family Resources
(r=-.76, p=.004). However, all of the direct paths from family factors (i.e. Family
Pressures, Family Resources, Family Contact, and Family Dysfunction) to Perceived
Criticism were non-significant. Thus, none of the family factors directly contributed
to level of family criticalness although the higher the amount of Perceived Criticism
was related to higher levels of psychiatric symptoms for this sample.
As noted previously, McDonald and Ho (2002) note the importance of testing
alternate path models that could just as easily explain the data from a conceptual
point of view than the initially hypothesized path model. The study offered the
opportunity to test for possible direct effects on consumer clinical functioning from a
number of family factors not heretofore tested. Thus, an alternative path model
107
introduced a direct path between family latent variables on Consumer Clinical
Functioning while keeping PC as an exogenous variable for Consumer Clinical
Functioning. The hypothesis for each of the direct paths between family factors on
consumer psychiatric functioning were: as Family Contact, Family Pressures, and
Family Dysfunction increased, level of Consumer Psychiatric Functioning would
worsen (i.e. symptoms would increase); and as Family Resources increased,
Consumer Psychiatric Functioning would improve (i.e. symptoms would decrease).
As noted in the previous study (see Chapter 1 above) amount of family contact had a
significant direct beneficial effect on consumer psychosocial functioning. Would
amount of family contact have a different or similar direct effect on this second area
of consumer functioning? Would amount of family contact have the expected
deleterious direct effect (i.e. more “exposure” to the “toxic” effect of highly critical
families) on amount of consumer psychiatric functioning as is often hypothesized in
the EE literature? In addition, only significant correlations between family factors
noted in the original hypothesized path model were included in this model.
The goodness of fit indices for the alternative path model A were
2
=103.31,
df=64, p=.08, RMSEA=.05 (CL: LOW=.00, HIGH=.80), CFI=.95, and NFI=.80. The
results indicated that there was not a significant difference between the predicted and
observed covariance matrices (unlike the original path model). The value of the
RMSEA indicated a close approximation fit to the observed data. The lower bound
value was lower than .05; thus, the hypothesis of a good approximate fit could not be
rejected. Likewise, the upper level was less than .10, which meant that the hypothesis
108
of poor approximate fit could be rejected (Kline, 2005). Finally, the values of the
CFI and the NFI indicated an acceptable fit (Kline, 2005; Loehlin, 1998).
The direct path from Perceived Criticism to Consumer Psychiatric
Functioning remained significant (b=.98, p=.000) as was the case with the original
path model. As with the original hypothesized model, none of the family factors,
including Family Contact, had a significant direct effect on Consumer Clinical
Functioning. Thus, none of the family factors had any impact on clinical functioning;
on the other hand, amount of family criticalness remained a significant direct effect
on clinical functioning. Although the model did not offer anything conceptually new
it did, on the other hand and most significantly, replicate with a larger, outpatient
African-American sample what has been found with smaller, inpatient African-
American samples in the EE literature (Moline et al., 1985; Wuerker et al., 1999). A
summary of the goodness-of-fit indices for the two tested models is provided in
Table 10 below.
109
Table 10: Summary of goodness-of-fit indices for original and alternative path
models
Path Model

2
(df)
p RMSEA
1
(90% C.L.)*
CFI
2
NFI
3
Original (see
figure 2)
103.54
(82)
.05
#
.05
(.00 - .80)
.95 .80
Alternative
path model
(see figure 3)
103.31
(64)
.08 .05
(.00- .80)
.95 .80
1
A value of.05 indicates a close approximation, a value of .06 to .09 indicates
reasonable approximation, and.10 indicates poor approximation of error (Kline,
2005).
2
Any value that is.90 indicates a reasonable fit (Kline, 2005)
3
A value equal to 1 is considered a “perfect” fit (Kline, 2005)
* 90% RMSEA Confidence Level. NOTE: a lower bound value <.05 indicates that
the hypothesis of a good approximate fit can not be rejected; an upper level <.10
indicates that the hypothesis of poor approximate fit can be rejected (Kline, 2005).
#
A significant p-value indicates that the path model is significantly different from
observed covariance matrix.
In summary, the original path model, which hypothesized that family factors
had direct effect on amount of family criticalness, offered nothing conceptually
informative because hypothesized family factors had no direct effect on amount of
family criticalness. Likewise, family factors had not direct effect on level of
consumer psychiatric symptomatology. At the same time, several family factors were
significantly correlated to each other. Most importantly, amount of family
criticalness did have a significant direct effect on amount of consumer psychiatric
symptoms. Thus, the story that this final model tells is that the amount of family
criticalness a consumer perceives impacts that individual’s psychiatric
symptomatology. The model confirms with a larger sample what has already been
110
found within the EE literature. Please refer below to Figure 11 for the alternative
path model as well as Table 11 for summary of the path model’s estimates. Figure 12
shows the final path model predicting for consumer clinical functioning.
Figure 11: Final alternative path model including all hypothesized paths
Family
Pressures
Family
Dysfunction
Consumer
Psychiatric
Functioning
Family
Contact
Family
Resources
b=.38
p=.000
b=-.002
p=.94
b=.34
p=.66
b=.39
p=.38
b=.75
p=.48
Perceived
Criticism
r=.35
p=.01
r=.-.75
p=.000
r=.54
p=.004
r=-.79
p=.003
111
Figure 12: Final path model predicting consumer psychiatric functioning
Goodness of fit indices:

2
=103.31 (df=64)
p=.08
RMSEA=.05 (CL: LOW=.00, HIGH=.80)
CFI=.95
NFI=.80
Table 11: Summary of estimates for alternative path model 2
Regression Weights
Beta S.E. Standard
.
Beta
C.R. p
Family
Pressures
Consumer Clinical
Functioning
.755 1.70 .34 .44 .66
Family
Resources
Consumer Clinical
Functioning
.723 1.02 .75 .71 .48
Family
Dysfunc
Consumer Clinical
Functioning
1.138 1.29 .39 .88 .38
Family
Contact
Consumer Clinical
Functioning
-.002 .02 -.02 -.07 .94
Perceived
Criticism
Consumer Clinical
Functioning
.984 .25 .38 4.00 .00
0
Consumer
Psychiatric
Functioning
b=.38
p=.000
Perceived
Criticism
112
Table 11, Continued
Variances
Estimate S.E. C.R. p
Family Pressures 14.73 6.99 2.11 .035
Family Resources 79.01 30.01 2.63 .008
Family Dysfunction 8.50 2.00 4.26 .000
Family Contact 12249.21 1806.05 6.78 .000
Perceived Criticism 10.69 1.58 6.78 .000
Consumer Clinical
Functioning
52.91 15.42 3.43 .000
Discussion
Expressed Emotion (EE) studies tested EE or sub-constructs of EE on
consumer outcomes such as hospitalization or relapse and typically for consumers
newly hospitalized or recently discharged (Barrowclough & Hooley, 2003;
Bebbington & Kuipers, 1994). EE studies that included African-Americans in their
samples had small sample sizes (Moline et al., 1985; Wuerker et al., 1999). The
present study tested conceptually similar relationships for a larger sample of African-
American consumers who were also outpatient. Although these consumers lived out
in the community (typically not with the family), the perceived criticalness of the
family was tested to see if this directly effected the psychiatric functioning of these
consumers. Many studies demonstrate that the family is very important within the
African-American community and so the study sought to test if the kind of
relationship between criticalness (the most predictive sub-component of EE in
studies involving African-American consumers) and consumer psychiatric
functioning demonstrated in the EE literature existed for this sample, too. In
113
addition, the model included family factors (such as family pressures, the family’s
overall functioning, the amount of resources available to the family, and amount of
family contact), based on both the EE and burden literatures, that might directly
affect the amount of family criticalness toward the consumer. Structural equation
modeling offered the advantage of concurrently testing all of these relationships and
to analyze whether the hypothesized model “fit” the data.
Impact of perceived family criticalness on clinical functioning
There were several consistent findings from the analyses. The results
demonstrated a clear relationship between Perceived Criticism (PC) and Consumer
Psychiatric Functioning across both path models tested. The level of symptoms for
the consumer increased as the level of PC increased. Thus, for these outpatient
consumers, when they perceived the level of the family’s criticalness increase, they
also reported increased symptoms of anxiety and psychosis. This is conceptually
similar to what was found in the EE literature where Critical Comments was
predictive for increased levels of relapse or rehospitalization particularly for African-
American consumers (Bebbington & Kuipers, 1994; Butzlaff & Hooley, 1998;
Hashemi & Cochrane, 1999b; Moline et al., 1985).
Family factors unrelated to perceived criticism or clinical functioning
The other consistent finding from this study was that none of the family
variables or amount of family contact had any effect on either amount of family-
criticalness or on the consumer’s psychiatric functioning. This indicates that the
relationship between family-criticalness and consumer psychiatric functioning is
114
independent of other factors related to the family such as the amount of pressures
that they have, their level of overall functioning, the amount of resources available to
the family, or how much contact occurs between consumer and family. One might
expect, for example, that as the amount of family functioning problems increased,
the consumer would report greater amounts of family criticalness or more symptoms,
but this was not the case. Similarly, one might expect that as the amount of pressures
increased or the fewer the resources available to the family, that the consumer would
experience higher levels of family criticalness or psychiatric symptoms; but, this also
was not the case. The findings indicate that aspects of family functioning and well-
being had no direct association with amount of family criticalness or level of
psychiatric symptoms. Although the family is viewed as important for most African-
Americans, factors associated with these consumers’ families did not contribute to
the consumer’s perception of family criticalness or level of psychiatric symptoms.
Of particular note is that the amount of contact was unimportant to both the
level of family criticalness and to the level of psychiatric symptomatology. Again,
this seems to contradict many of the earlier assertions or findings within the EE
literature that suggest too much contact with the family (particularly if the family
was rated as high EE) had a deleterious impact on the consumer (Brown et al., 1972;
Hoenig & Hamilton, 1966; J. Leff & Vaughn, 1985). In fact, even given a moderate
level of PC (m=4.66, sd=3.42), family contact had no contributing influence on
family criticalness perceived by the consumer or level of psychiatric
symptomatology. This finding may be due to this being a sample of outpatient
115
consumers who were heterogeneous in age and length of illness compared to most of
the samples used in the EE literature (which have tended to focus on recently
hospitalized or discharged consumers or on consumers recently diagnosed with a
schizophrenic-related disorder) (Bebbington & Kuipers, 1994; Butzlaff & Hooley,
1998). This in itself is an important finding as it suggests that family factors salient
for level of symptoms for inpatient consumers has little to no bearing for outpatient
consumers, even for a sample of African-American adults where family is
considered of greater importance.
Summary of findings & implications
The level of perceived family criticalness is clearly and significantly related
to the amount of psychiatric symptoms reported by the consumer. This relationship
remained robust regardless of any other family related factors. For this outpatient
sample of African-American consumers, with varying ages and life-experiences with
the illness, the process of family criticalness on level of symptoms is independent of
any other family influences. Likewise, it demonstrates for an outpatient sample that
the process or perception of family criticalness is a significant factor in amount of
symptoms experienced by the consumer.
The study also offers the first use of the Perceived Criticism scale with a
sample of consumers diagnosed with schizophrenia. It demonstrates its utility with
consumers with a major mental illness other than depression (Hooley & Teasdale,
1989; Riso et al., 1996). As to whether PC measures actual or “only” perceived
criticism from the family remains to be evaluated in future studies (although Hooley
116
& Teasdale argue that the perception of criticalness is more important to the
consumer than the actual amount of criticalness) (Hooley & Teasdale, 1989).
The findings suggest that an important component of treatment for symptom
stabilization involves working with consumers’ perceptions of their families’
criticalness or supportiveness. Consumers may need help with assessing how
accurate these perceptions are: depending on the accuracy of such perceptions
consumer might benefit from skills training in self-assertion to confront or cope with
family criticalness or assistance with reviewing and then replacing inaccurate
cognitions with more accurate ones (Kuipers et al., 2006; Pfammatter, Junghan, &
Brenner, 2006; Pilling et al., 2002). Social workers can also work with families that
are assessed as critical and offer such families the sort of psychoeducation and
communication training present in many family psychoeducation interventions.
Future studies need to evaluate the accuracy of these perceptions of criticalness and
then focus on interventions to address this issue with the consumer and/or the family.
Amongst the family factors, family contact was only significantly correlated
with amount of family pressures: as amount of family contact increased so too did
family pressures increase. Thus, for this sample, it was the family that was
negatively impacted by amount of family contact not the consumer. This suggests
that there is something about the contact that is associated with higher levels of
family pressures. This may be related to the sorts of findings from care-giver burden
studies that show that sometimes contact with an ill family member can create its
own sense of stress or pressures for the family (Aranda & Knight, 1997; Baronet,
117
1999; Boye et al., 1999; Connel & Gibson, 1997; Fadden et al., 1987; Hatfield &
Lefley, 1993; Land & Hudson, 2002; McCabe et al., 2003; Saldana et al., 1999;
Stueve et al., 1997; Theis, Moss, & Pearson, 1994). This finding is interesting given
the amount of research that has been done to develop and implement family
psychoeducation models to improve family functioning in some way (e.g.
communication style, interpersonal interactions, knowledge of illness) in order to
improve the consumer’s level of symptoms, relapse, or rehospitalization
(Barrowclough & Tarrier, 1990; Dixon et al., 2000; J. Leff et al., 1985; Telles et al.,
1995). This is not to imply that family issues are in and of themselves less salient for
the outpatient consumer as already noted. However, it does suggest that family
interventions should also focus on concerns especially germane to the family’s own
specific issues of pressure (e.g. reducing stressors, improving overall coping within
the family, etc.) (Hatfield, 1997a; Hatfield & Lefley, 1993; Lefley, 1998).
Limitations
The study was cross-sectional in design. It provides a “snap-shot” of this
particular sample just prior to beginning services in a family based intervention.
Causation, the directionality of relationships between variables, can not be
determined at this level of analysis. A longitudinal study offers the advantage of
testing causation as well as any changes in the relationships between exogenous and
endogenous variables across time. Depending on the trajectory of these relationships,
services could focus on supporting consumers with managing family criticalness
and/or inaccurate perceptions of criticalness while assisting families with those
118
concerns more germane to them (e.g. family pressures such as care-giver burden,
financial strains, family coping strategies, etc.).
Another limitation was sample size. Despite the fact that the sample is larger
than what has been used in EE studies involving African-American consumers and
families (Baronet, 1999; Butzlaff & Hooley, 1998; Fadden et al., 1987; Hashemi &
Cochrane, 1999a; Horwitz & Reinhard, 1995), it’s size is not considered optimal for
SEM analyses. Sample sizes of 100 and more (ideally around 200-300) are
recommended to increase power (better able to reject an incorrect model), model
respecification (paths may be non-significant due to small sample size as opposed to
poor specification), and decrease potential technical problems with the analyses
(Kline, 2005). Thus, for example, sample size may account for non-significant
findings between family factors and consumer psychiatric functioning. In addition,
sample size may account for the significant p-value demonstrating that the
hypothesized measurement model and the observed was significantly different.
Nonetheless, the present study included nearly 100 consumers and nearly 100 family
members in the analyses.
Another potential limitation was that different reporters were used across the
indicators for the latent variables. There might be reporter biases with each indicator
that, in turn, effected relationships across the model variables (Association et al.,
1999; DeVellis, 1991; McIver & Carmines, 1981). Some studies show that when
multiple members of a family report on the same phenomenon there is some
divergence in the responses. For example, in childhood anxiety problems, children
119
tend to over-report family difficulties while their parents tend to under- report
difficulties (Bogels & Brechman-Toussant, 2006). Thus, in this study for example,
the consumer might be biased to over report levels of criticalness. Nonetheless, some
studies, as noted above, suggest that it is the perception of criticalness on the part of
the consumer that is more important than the actual amount in effecting consumer
outcomes (Hooley, 1998; Hooley & Teasdale, 1989).
Conclusion
It is clear that the amount of criticalness that a consumer perceives on the part
of his or her family impacts the level of psychiatric symptoms. The finding confirms
for a larger sample of African-American consumers living out in the community
what has been found with smaller samples of inpatient African-American consumers
(Moline et al., 1985; Wuerker et al., 1999). Future studies are needed to test the
accuracy of the consumer’s perception of level of family criticalness. Such studies
will help identify how interventions should be tailored so that consumers can better
cope with family criticalness and/or develop more accurate assessments of this.
Although family variables did not impact either the amount of family
criticalness or consumer psychiatric functioning, the study showed that contact
between consumer and family had a negative impact on the level of a family’s
pressures. Thus, it was the family that had an associated increase in pressures to
more family contact; there were no direct effects between amounts of family contact
on amount of consumer symptomatology. Interventions tailored for both families and
consumers should focus on what can assist the family to feel fewer pressures while
120
interacting with their family member. This can include some of the elements
proposed in many family psychoeducation interventions (e.g. educating families
regarding diagnosis, course of illness, resources, etc.), but tailoring the interventions
specific to the needs and concerns that families might have as opposed to the clinical
symptoms that the consumer displays. That is, following the lead of organizations
such as the National Alliance for the Mentally Ill (NAMI), services would involve
support, respite, and encouragement in a non-threatening or non-pathologizing way
(Hatfield, 1997a, 1997b; Lefley, 1998; NAMI, 2003).
The relationships found in this study confirm previous findings for African-
American outpatient consumers in the EE literature. Future studies are needed in
determining whether these relationships are replicated with other comparable African
American samples, whether these relationships change over time, and whether there
are similar or different relationships in the lives of other outpatient ethnic and SES
groups of consumers and their families.
121
Study Conclusion
General summary
The following study included two analyses of structural equation models that
tested the relationship between family latent variables and consumer functioning for
a sample of poor, inner-city African-American consumers and their families. The
purpose of the study was to broaden, both conceptually and methodologically, what
was already tested in the family mental health literature.
The study sought to extend the literature in several ways. First, it tested
similar concepts from the EE literature (i.e. family criticalness as a proxy for Critical
Comments) and the caregiver burden literatures with a larger, outpatient sample of
both African-American consumers and their families, a section of the general
population not often included in such studies. Likewise, the family mental health
literature demonstrates that the family is very important within the African-American
community; thus, this sample was seen as an ideal means of testing contextual
factors on consumer functioning, a priority for the National Institutes of Mental
Health (Services, 2001). In addition, understanding these family factors will help
inform both intervention and policy targeting the needs of similar communities.
Secondly, it concurrently introduced a broader array of family variables than
what was previously tested in either the EE or caregiver burden literatures while
including the more traditional variables such as Critical Comments (from the EE
literature) and burden. These latent variables included Family Dysfunction, Family
Pressures, and Family Resources. Thus, the study offered a broader “lens” to test
122
these relationships in one study what has already been tested across numerous,
separate studies.
Thirdly, it addressed an important gap in the literature regarding consumer
outcomes. The vast majority of EE studies only included relapse and/or
rehospitalization as outcomes. The present study, while including an outcome
variable conceptually similar to the typical outcome in the EE literature (consumer
psychiatric functioning), also included the important domain of consumer
psychosocial functioning. As previously argued, consumer psychosocial functioning,
which pertains to the areas of work, independent living skills, and interpersonal
socialization, is a more pertinent outcome for the vast majority of consumers who
spend increasingly more time in the community versus in a hospital setting. The two
functioning domains (clinical and psychosocial functioning) were treated as separate
areas of interest because of recent findings that the two are typically only mildly
correlated at most. Thus, this study tested family factors on an area not extensively
addressed in the family mental health literature (psychosocial functioning) for adults
living with schizophrenia while also attempting to replicate previous findings from
the EE literature with a larger African-American sample (clinical functioning). In
addition, the study tested what appears to be a novel finding – that contact between
consumer and family is specifically beneficial to the psychosocial functioning of the
consumer.
Lastly, it tested these relationships using a more advanced form of statistical
analysis than previously used in the literature – structural equation modeling. The
123
use of structural equation modeling allowed the testing of more detailed models,
which included latent variables, and the simultaneous analysis of the relationships
between these variables. It also offered the advantage of concurrently testing all of
these relationships and to analyze whether the hypothesized model “fit” the data.
The first structural model for this study tested the direct effect of family
variables, such as Family Pressures, Family Dysfunction, and Family Resources, had
on the amount of contact between the consumer and the family. It likewise tested the
direct effect that family contact had on level of consumer psychosocial functioning.
The second structural model for this study included family factors (Family Pressures,
Family Dysfunction, Family Resources, and Family Contact) that might directly
affect the amount of family criticalness toward the consumer. It also included an
analysis of the direct effect that family criticalness had on the consumer’s level of
psychiatric symptomatology.
Summary of findings - structural equation model involving consumer psychosocial
functioning
No matter which family domains were included in the path model, the
relationship between amount of family contact and the consumer’s psychosocial
functioning remained positive and significant. Hence, the more contact that the
consumer had with her or his family, the better the psychosocial functioning was.
This replicated a previous finding from a path analysis model that found that amount
of family contact had a significant impact on consumer psychosocial functioning
(Guada & Brekke, 2007).
124
On the other hand, none of the family factors had a significant direct effect on
amount of family contact when all of them were included in the structural model. A
second model tested whether family factors had a significant direct effect on level of
consumer psychosocial functioning. Once again none of these had a direct effect
although amount of family dysfunction approached significance. When amount of
family dysfunction was modelled as the only family factor (along with amount of
family contact) with a direct effect on level of consumer psychosocial functioning,
the relationship was significant. The level of family dysfunction had a negative direct
effect on psychosocial functioning: the more dysfunction the worse the psychosocial
functioning. This is reminiscent of how level of Expressed Emotion (as a specific
“kind” of family dysfunction) has a negative impact on a consumer’s level of relapse
and re-hospitalization (Hooley, 1985; S. King & Dixon, 1996; J. Leff & Vaughn,
1985). A family’s ability to problem solve, communicate, be affectively involved
with each other, etc. does effect how well a consumer does in such areas as work,
socialization, and independent living skills. This occurs even given the beneficial
significant effect that amount of family contact had on consumer psychosocial
functioning.
In the past practitioners, researchers, and theorists hypothesized that families
judged as “dysfunctional” (such as those with a high Expressed Emotion profile)
were “toxic” for the consumers and should therefore be avoided (unless the family
received some sort of treatment) (Barrowclough & Hooley, 2003; Bebbington &
Kuipers, 1994; Butzlaff & Hooley, 1998). But, these results show that amount of
125
family contact is important in and of itself, and even with those families who scored
higher on amount of family dysfunction. Hence, for these poorer African-American
consumers, at least some interaction or contact with their families – no matter how
those families function – is crucial to the consumer’s well-being out in the
community. On the other hand, even given this important contact, some families with
problems of functioning may need interventions so that any deleterious effects their
functioning might have on consumers are attenuated. Although contact, in and of
itself, is beneficial for the consumer, nonetheless the quality of the contact might
improve if the level of family functioning is improved. As family functioning
improves, then consumer psychosocial functioning improves, which probably will
improve the quality of contact between consumer and the family.
The study also suggests ways that might improve family functioning as
demonstrated by its relationships with other family factors. Given that family
functioning had a significant direct effect on consumer functioning, the question
remained whether any of the other family factors (i.e. amount of family resources,
amount of family pressures) had a direct effect on level of family functioning. The
final path model demonstrated that there was a direct, significant effect from the
amount of pressures that a family copes with to the amount of resources available to
the family (as amount of family pressures went up, the amount of family resources
went down) and a significant direct effect from amount of available family resources
to the family’s level of dysfunction (as amount of family resources went down the
level of family dysfunction went up). This finding is reminiscent of findings within
126
the caregiver burden and family treatment literatures. That is, as families experience
more pressures and have fewer resources, their ability to adequately function as a
family diminishes (Demi et al., 1997; Nabors et al., 2002; Pearlin et al., 1997;
Pearlin & Schooler, 1978; Rivera et al., 1997). In addition, this adds indirect support
for stress models such as Pearlin’s, which postulates that as the level of stress
multiplies there is a cascading and reverberating effect across several spheres of a
person’s or family’s life. A sense of stress heightens in an exponential process of
decreasing resources and poorer coping (Pearlin, 1991; Pearlin et al., 1997). The
finding also highlights that, although amount of family resources and amount of
family pressures did not have a direct effect on the psychosocial functioning of the
consumer, they did have an indirect effect that was mediated by level of family
dysfunction.
There was one other interesting finding that occurred during the analysis of
this structural model’s measurement model. There was a problem with the originally
conceived latent variable for amount of Family Resources. Two indicators, one
regarding spirituality (F-COPES subscale Spirituality), and one regarding help-
seeking behaviors (F-COPES subscale Seeking Help) loaded as a separate factor.
Several variants of the measurement model including these indicators (first as a
separate factor, then summed and used as an observable variable) proved
problematic with resultant poor goodness-of-fit indices. Once these indicators were
dropped, the resultant measurement model had significantly improved goodness-of-
fit indices. Why indicators involving spirituality and seeking help should be
127
problematic given the plethora of literature of the importance of spirituality and
mutual help-seeking behaviors within the African-American community was baffling
(Alston & Turner, 1994; Brekke & Guada, 2004; McAdoo, 1998; Pickett et al., 1993;
Robinson, 1983; Turner & Alston, 1994; S. E. Williams & Finger Wright, 1992).
However, it seems to suggest that amount of family contact was unrelated to a
family’s spirituality or help seeking behaviors. In this way, it appears that emotional
and geographic ties between family members was important regardless of any
religious belief or practices or help seeking behaviors (Alston & Turner, 1994;
McCabe et al., 2003; Pinquart & Sorensen, 2005; Rivera et al., 1997; Solomon &
Draine, 1995). Future studies are required to assess whether this situation is
replicated or a statistical artifact of this particular sample.
Thus, the story told by this first structural model was that as family pressures
increased, family resources diminished; as family resources decreased, the family’s
functioning worsened & as family functioning worsened, consumer psychosocial
functioning got worse. Nonetheless, family contact continued to have a significant
beneficial impact on consumer functioning despite these relationships across the
other family factors. This is a powerful finding since it suggests that the emotional
connections between poorer African-American consumers and their families
transcend both familial concrete (e.g. financial, discrimination, social supports) and
abstract (e.g. sense of burden, sense of well-being) barriers (Boye et al., 1999;
Stueve et al., 1997; Wuerker et al., 1999). At the same time, factors that impact a
family’s ability to function, such as the daily pressures it faces or amount of
128
resources available to it, indirectly effect consumer psychosocial functioning through
the family’s overall functioning.
Summary of findings - structural equation model involving consumer psychiatric
functioning
The second structural equation model tested family factors (the same factors
used in the first SEM) on family criticalness (while testing the impact of family
criticalness on the level of psychiatric symptoms for the consumer). The main
finding was that family factors did not have a direct effect on level of family
criticalness. Alternative path models demonstrated that these same family factors had
no direct effect on the consumer’s psychiatric functioning, too. These findings are
curious because one might expect that as the amount of family functioning problems
increased, family pressured increased, and/or family resources decreased that the
consumer would report greater amounts of family criticalness or more symptoms, but
this was not the case. Thus, regardless of family factors, both the amount of family
criticalness and the level of consumer symptomatology were unaffected.
Nonetheless, the direct effect from family criticalness to the level of the
psychiatric symptoms of consumers was robust in that it was significant in every
path model analyzed regardless of the presence or absence of other variables. More
perceived family criticalness was associated with more psychiatric symptoms for this
sample of outpatient African-American consumers replicating what was found in the
EE literature for inpatient African-American consumers and their families (Butzlaff
129
& Hooley, 1998; Hashemi & Cochrane, 1999b; Jones et al., 1995; Kavanagh, 1992;
Moline et al., 1985; Wuerker et al., 1999).
As noted in the previous section of this study, the finding is also interesting
given the amount of research that has been done to develop and implement family
psychoeducation models to improve family functioning in some way (e.g.
communication style, interpersonal interactions, knowledge of illness)
(Barrowclough & Tarrier, 1990; Dixon et al., 2000; J. Leff et al., 1985; Telles et al.,
1995). These intervention models implied that something about the family required
treatment so as to change the level of hostility, critical comments, and/or emotional
over-involvement; changes in one or more of these areas would mean that the
consumer experienced improved clinical functioning. But, this study demonstrated
that family factors such as amount of pressures, amount of resources, level of
dysfunction, and amount of family contact had no relationship with level of family
criticalness although family criticalness did predict higher amounts of psychiatric
symptoms, at least for these outpatient consumers.
Of particular note is that level of psychiatric symptomatology was unaffected
by amount of family contact unlike what was found in the first SEM for consumer
psychosocial functioning. Family contact had neither a deleterious nor a beneficial
effect. This was true even given a moderate level of family criticalness. Again, this
seems to contradict many of the earlier assertions or findings within the EE literature
that suggest too much contact with the family (particularly if the family was rated as
130
high EE) has a deleterious impact on the consumer (Brown et al., 1972; Hoenig &
Hamilton, 1966; J. Leff & Vaughn, 1985).
The only area that amount of family contact did have a significant positive
correlation was with amount of family pressures. The more family contact the family
had with the consumer the greater the family pressures. Clearly, it was the family
that was negatively impacted in some way by amount of family contact not the
consumer. This may be related to the sorts of findings from care-giver burden studies
that show that sometimes contact with an ill family member can create its own sense
of stress or pressures for the family (and possibly vice-versa) (Aranda & Knight,
1997; Baronet, 1999; Boye et al., 1999; Connel & Gibson, 1997; Fadden et al., 1987;
Hatfield & Lefley, 1993; Land & Hudson, 2002; McCabe et al., 2003; Saldana et al.,
1999; Stueve et al., 1997; Theis et al., 1994).
Summary of overall findings
The study demonstrated that two processes occurred in regards to what
family factors affected the two consumer functioning domains: what impacts the
ability to work, live independently, and socialize with others is not what impacts the
amount of symptoms that a consumer has. Whereas level of family functioning and
amount of family contact had direct effects, and amount of family pressures and
family resources had indirect effects (mediated by family functioning) on consumer
psychosocial functioning, none of these had a direct or indirect effect on consumer
clinical functioning. Clearly, family factors had a substantially more important role
for consumer psychosocial functioning than for clinical functioning. These results
131
are new findings in that family factors were shown to be related to consumer
psychosocial functioning in ways unlike to that of consumer clinical functioning. It
demonstrates that contextual factors affect this crucial area of functioning for people
living with schizophrenia: families are in some way crucial to the well-being of their
ill family members in regards to the abilities to work, live more independently, and
to interact and relate to others in the community. In addition, the findings from the
second model replicated the findings in the EE literature for African-American
consumers with an outpatient sample of African-American consumers and their
families. That is, critical comments, as represented by more perceived criticism,
negatively impacts amount of symptoms for anxiety and psychosis regardless of the
presence of any other family factors for outpatient consumers as is seen with
inpatient samples.
A clearer picture about how the family context impacts the quality of
consumers’ lives emerges from this study. These findings fill a critical need in the
literature for understanding what contextual factors impact consumer clinical and
psychosocial functioning. The findings demonstrate the importance of family factors
on consumer functioning as well as how these family factors interact with each other.
It does this for a sample of outpatient consumers as well as for a sample of African-
American consumers and their families. This understanding provides critical
information for both social work clinicians and policy makers.
Thus, interventions for consumers & their families should focus on:
increasing a family’s own level of social support & directly assisting with any
132
concrete pressures (e.g. financial issues or institution-based discrimination);
decreasing the burden that families experience through psychoeducation, mutual aid
groups, respite, and encouraging ongoing contact between consumer & family. The
findings suggest that family interventions should focus less on changing the family
for the consumer’s sake as much as providing education, emotional support,
resources and other necessary aids for the family’s sake. Family interventions that
focus on these family-centered specific issues are more likely to both directly and
indirectly help the consumer with those areas of functioning most important for
living in the community (i.e. work, independent living skills, and socialization).
Addressing family issues – most germane to the family’s own needs – will support
and/or improve the overall family’s functioning, which in turn will assist the
consumer both directly & indirectly (Hatfield, 1997a; Hatfield & Lefley, 1993;
Lefley, 1998). These findings tend to support the sorts of programmatic efforts
advocated by national organizations such as the National Alliance of the Mentally Ill
(2003).
In addition, family focused interventions should assist families to diminish or
avoid any criticism of the consumer given its potential negative impact on
symptoms. Consumer focused interventions should assist consumers to identify
perceptions of family criticalness. Workers can help consumers to clarify the
accuracy of these perceptions and then work on assertiveness, cognitive
restructuring, or a combination of both to help the consumer’s coping abilities.
Reducing a perception of criticalness will then help reduce the amount of psychiatric
133
symptoms. Thus, a holistic intervention should target those factors that impact
consumer clinical functioning (such as reducing critical comments) and those factors
that can impact consumer psychosocial functioning (ongoing contact between
consumer and family, the family’s functioning and well-being), Interestingly, many
of these of issues were not typically included in family-oriented psychoeducation
programs, which tended to focus solely on communication styles and educating
families about major mental illnesses (C. M. Anderson et al., 1986; Bae & Kung,
2000; Dixon et al., 2000; Fadden, 1998; J. Leff & Berkowitz, 1996; J. Leff &
Vaughn, 1981; McFarlane et al., 1991).
Research such as this can help social work practice by identifying those
factors helpful to consumer functioning, identifying those factors helpful to families,
identifying how these may influence each other, and identifying key strategies for
implementing these factors into existing practice models. Likewise, research such as
this can help social work educators and students by addressing with quantitative
methodologies person-in-environment concerns; demonstrating the importance of
focusing on more than individuals in practice, and helping to identify evidence-based
features of interventions already used in the field.
Study limitations
As previously addressed in Chapters I and II above, the following study has
several limitations. It was cross-sectional in design and thus can not establish
direction of causality across variables. Thus, for example, although a clear
relationship was established between Perceived Criticism and Consumer Clinical
134
Functioning, it could be argued that the level of Consumer Clinical Functioning had
the direct effect on amount of family criticalness. The direction of causation is best
established through longitudinal statistical techniques (Kachigan, 1986; Knoke et al.,
2002). In addition, a longitudinal study offers the advantage of evaluating any
changes in the noted relationships between exogenous and endogenous variables
across time. It offers the opportunity to assess in greater detail how family factors
(such as family pressures, family resources, and overall functioning) change across
time and how these changes directly and/or indirectly effect consumer functioning.
In addition, sample size was not ideal for SEM statistical methodology.
Despite the fact that the sample was larger than what has been used in EE and many
caregiver burden studies involving African-American consumers and their families
(Baronet, 1999; Butzlaff & Hooley, 1998; Fadden et al., 1987; Hashemi & Cochrane,
1999a; Horwitz & Reinhard, 1995), it’s size is not considered optimal for SEM
analyses. Sample sizes of 100 and more (ideally around 200-300) are recommended
(Kline, 2005). Thus, the sample size might account for some of the non-significant
findings between family and consumer functioning variables. Nonetheless, the
present study included nearly 100 consumers and as well as their family members in
the analyses.
Likewise, the study used different reporters across the indicators. There could
be reporter biases with each indicator that therefore effects the relationships between
latent variables (Association et al., 1999; DeVellis, 1991; McIver & Carmines,
1981). If only a family member, consumer, or an interviewer completed the scales,
135
then relationships between variables may have been different. As previously noted,
some studies show that when multiple members of a family report on the same
phenomenon there is some divergence in the responses (Bogels & Brechman-
Toussant, 2006).
Future directions and conclusion
The study offers a number of directions for future research. Because of the
findings are novel for the consumer outcomes literature, future studies are needed to
test the same model with other African-American samples and across other ethnic
and SES groups to assess if these findings are replicable and/or are unique to poorer
African-American families and consumers living in the community. In addition,
future longitudinal analyses offer the possibility to test relationships across time to
see how relationships across variables change or endure.
In addition, future studies can test other models involving the same family
factors and consumer outcomes, but extend the analysis by including moderator
effects. For example, future models can test whether the direct effect from family
dysfunction to psychosocial functioning of consumers is moderated by amount of
family contact. These future studies can provide the opportunity to clarify the details
of what processes are similar and different for the two areas of consumer functioning
used in this study. The findings from this study are but the first step in “teasing out”
or unravelling the puzzle of how family factors interact with and impact each other
while having both direct and indirect effects on consumer psychosocial and
consumer clinical functioning.
136
The ability to widen the lens of analysis by including both consumer and
family variables typically tested across many separate studies offers the promise of
providing more specific information about the contextual factors that influence
consumer functioning. Such information will only improve the kinds of services and
policies that target both consumers living with schizophrenia and their families.
137
Bibliography
Alston, R. J., & Turner, W., L. (1994). A family strengths model of adjustment to
disability for African American clients. Journal of Counseling and
Development, 7(March/April), 378-383.
Anderson, C. M., Hogarty, G. E., & Reiss, D. J. (1980). Family treatment of adult
schizophrenic patients: a psycho-educational approach. Schizophrenia
bulletin,, 6(3), 490-505.
Anderson, C. M., Reiss, D. J., & Hogarty, G. E. (1986). Schizophrenia and the
Family. New York, New York: Guilford Press.
Anderson, J. C., & Gerbing, D. W. (1988). Structural equation modeling in practice:
a review and recommended two-step approach. Psychological Bulletin,
103(3), 411-423.
Aranda, M. P., & Knight, B. G. (1997). The influence of ethnicity and culture on the
caregiver stress and coping process: A sociocultural view and analysis.
Gerontologist, 37, 342-354.
Association, A. E. R., Association, A. P., & Education, N. C. o. M. i. (1999).
Standards for educational and psychological testing (2nd ed.). Washington,
DC: Amercian Educational Research Association.
Bae, S.-W., & Kung, W. W.-M. (2000). Family intervention for Asian Americans
with a schizophrenic patient in the family. American journal of
orthopsychiatry, 70(4), 532-541.
Baronet, A.-M. (1999). Factors associated with caregiver burden in mental illness: a
critical review of the research literature,. Clinical psychology review, 19(7),
819-841.
Barrio, C. (2001). Culture and schizophrenia: A cross-ethnic growh curve analysis.
Journal of Nervous and Mental Disease, 189(10), 676-684.
Barrowclough, C., & Hooley, J. M. (2003). Attributions and expressed emotion: A
review. Clinical psychology review, 23, 849-880.
138
Barrowclough, C., & Tarrier, N. (1990). Social functioning in schizophrenic patients:
I. The effects of expressed emotion and family intervention. Social
Psychiatry and Psychiatric Epidemiology, 25, 125-129.
Bebbington, P., & Kuipers, L. (1994). The predictive utility of expressed emotion in
schizophrenia: an aggregate analysis. Psychological Medicine, 24, 707-718.
Berry, W. D., & Feldman, S. (1985). Multiple Regression in Practice. Newbury
Park, CA: Sage Publications.
Biegel, D. E., Milligan, S. E., Putnam, P. L., & Song, L.-Y. (1994). Predictors of
burden among lower socioeconomic status caregivers of persons with chronic
mental illness. Community mental health journal, 30(5), 473-494.
Billingsley, A. (1990). Understanding African-American family diversity. In J.
Dewart (Ed.), The state of black America. New York: National Urban
League, Inc.
Bogels, S. M., & Brechman-Toussant, M. L. (2006). Family issues in child anxiety:
Attachment, family functioning, parental rearing and beliefs. Clinical
psychology review, 26, 834-856.
Boye, B., Bentsen, H., Notland, T. H., Mumkvold, O. G., Lersbryggen, A. B.,
Oskarsson, K. H., et al. (1999). What predicts the course of expressed
emotion in relatives of patients with schizophrenia or related psychoses?
Social Psychiatry and Epidemiology, 34, 35-43.
Brekke, J., & Barrio, C. (1997). Cross-ethnic symptom differences in schizophrenia:
the influence of culture and minority status. Schizophrenia bulletin, 23(2),
305-316.
Brekke, J., & Guada, J. (2004). The Inner-City African-American Family & Chronic
Mental Illness: Evidence for Strengths. Under revision.
Brekke, J., Kohrt, B., & Green, M. (2001). Neuropsychological functioning as a
moderator of the relationship between psychosocial functioning and the
subjective experience of self and life in schizophrenia. Schizophrenia
bulletin, 27(4), 697-708.
Brekke, J., Levin, S., Wolkon, G., Sobel, G., & Slade, E. (1993). Psychosocial
functioning and subjective experience in schizophrenia. Schizophrenia
bulletin, 19, 559-608.
139
Brekke, J., & Long, J. D. (2000). Community-based psychosocial rehabilitation and
prospective change in functional, clinical, and subjective experience variables
in schizophrenia. Schizophrenia bulletin, 26(3), 667-680.
Brekke, J., Raine, A., Ansel, M., Lencz, T., & Bird, L. (1997). Neuropsychological
and psychophysiological correlates of psychosocial functioning in
schizophrenia. Schizophrenia bulletin, 23(1), 19-28.
Brown, G. W., Birley, J. L. T., & Wing, J. K. (1972). Influence of family life on the
course of schizophrenic disorders: a replication. British journal of psychiatry,
121, 241-258.
Brown, G. W., Monck, E. M., Carstairs, G. M., & Wing, J. K. (1962). Influence of
family life on the course of schizophrenic illness. British journal of
preventive and social medicine, 16, 55-68.
Butzlaff, R. L., & Hooley, J. M. (1998). Expressed emotion and psychiatric relapse:
a meta-analysis. Archives of General Psychiatry, 55(June), 547-552.
Chambless, D. L., Steketee, G., Bryan, A. D., Aiken, L. S., & Hooley, J. M. (1999).
The structure of expressed emotion: a three-construct representation.
Psychological Assessment, 11(1), 67-76.
Civil, M. C. f. t. I. T. E. M. R. t. t. A. L. A. (1993). Disturbance Ethnicity and
Economic Status of South Central Los Angeles [Online]. Retrieved 2001,
December 13, from
http://www.usc.edu/isd/archives/cityinstress/medical/part11.html
Cohen, J., & Cohen, P. (1983). Applied multiple regression/correlation analysis for
the behavioral sciences (2nd ed.). Hillsdale, NJ: Erlbaum.
Connel, C. M., & Gibson, G. D. (1997). Racial, ethnic and cultural differences in
dementia caregiving: review and analysis. Gerontologist, 37, 355-364.
Connery, L., & Brekke, J. (1999). A home-based family intervention for ethnic
minorities with a mentally ill member. Alcoholism Treatment Quarterly,
17(1/2), 149-167.
Cook, J. A., Lefley, H. P., Pickett, S. A., & Cohler, B. J. (1994). Age and family
burden among patients of offspring with severe mental illness. American
journal of orthopsychiatry, 64(3), 435-447.
140
Demi, A., Bakeman, R., Moneyham, I., Sowell, R., & Seals, B. (1997). Effects of
resources and stressors on burden and depression of family members who
provide care to an HIV infected woman. Journal of Family Psychology, 11,
35-48.
DeVellis, R. F. (1991). Scale Development: Theory & Applications. Newbury Park,
CA: Sage Publications.
Dixon, L., Adams, C., & Lucksted, A. (2000). Update on family psychoeducation for
schizophrenia. Schizophrenia bulletin, 26(1), 5-20.
Epstein, N. B., Baldwin, L. M., & Bishop, D. S. (1983). The McMaster Family
Assessment Device. Journal of marital and family therapy, 9(2), 171-180.
Fadden, G. (1998). Research update: Psychoeducation family interventions. Journal
of family therapy, 20, 293-309.
Fadden, G., Bebbington, P., & Kuipers, L. (1987). The burden of care: the impact of
functional psychiatric illness on the patient's family. British journal of
psychiatry, 150, 285-292.
Falloon, I. R., Boyd, J. L., McGill, C. W., Razani, J., Moss, H. B., & Gilderman, A.
M. (1982). Family management in the prevention of exacerbations of
schizophrenia - a controlled study. The New England journal of medicine,
306(24), 1437-1440.
Friedmann, M. S., McDermut, W. H., Solomon, D. A., Ryan, C. E., Keitner, G. I., &
Miller, I. W. (1997). Family functioning and mental illness: a comparison of
psychiatric and nonclinical families. Family process, 36, 357-367.
Geller, J. L. (2000). The last half-century of psychiatric services reflected in
Psychiatric Services. Psychiatric services, 51, 41-67.
Goodman, S. H., Sewell, D. R., Cooley, E., L., & Leavitt, N. (1993). Assessing
Levels of Adaptive Functioning: the Role Functioning Scale. Community
mental health journal, 29(2), 119-131.
Green, R. S., & Gracely, E. J. (1987). Selecting a Rating Scale for Evaluating
Services to the Chronically Mentally Ill. Community mental health journal,
23(2), 91-102.
141
Guada, J., & Brekke, J. (2007). A Path-analytic Assessment of the Relationships
between Perceived Criticism and outcomes for African-American Consumers
with Schizophrenia. University of Southern California, Los Angeles.
Guarnaccia, P. J., & Parra, P. (1996). Ethnicity, social status, and families'
experiences of caring for a mentally ill family member. Community Mental
Health Journal, 32(3), 243-260.
Hashemi, A. H., & Cochrane, R. (1999a). Expressed emotion and schizophrenia: a
review of studies across cultures. International Review of Psychiatry, 11,
219-224.
Hashemi, A. H., & Cochrane, R. (1999b). Expressed emotion and schizophrenia: A
review of studies across cultures. International Review of Psychiatry, 11(2/3),
219-224.
Hatfield, A. B. (1997a). Families of adults with severe mental illness: new directions
in research. American journal of orthopsychiatry, 67(2), 254-260.
Hatfield, A. B. (1997b). Working collaboratively with families. Social Work in
Health Care, 25(3), 77-85.
Hatfield, A. B., & Lefley, H. P. (1993). Surviving mental illness: stress, coping, and
adaptation. New York: Guilford Press.
Hill, R. B. (1998). Enhancing the resilience of African American families. Journal of
human behavior in the social environment, 1(2/3), 49-61.
Hines, P. M., & Boyd-Franklin, N. (2005). African-American Families. In M.
McGoldrick, J. Giordano & N. Garcia-Preto (Eds.), Ethnicity and Family
Therapy (3rd ed., pp. 87-100). New York, NY: Guildford Press.
Hoenig, J., & Hamilton, M. W. (1966). The schizophrenic patient in the community
and his effect on the household, The International journal of social
psychiatry. 12, 3(165-176).
Hooley, J. M. (1985). Expressed emotion: a review of the critical literature. Clinical
psychology review, 5, 119-139.
Hooley, J. M. (1998). Expressed emotion and locus of control. The Journal of
Nervous and Mental Disease, 189(6), 374-378.
142
Hooley, J. M., & Teasdale, J. D. (1989). Predictors of relapse in unipolar
depressives: Expressed emotion, marital distress, and perceived criticism.
Journal of Abnormal Psychology, 98(3), 229-235.
Horwitz, A. V., & Reinhard, S. C. (1995). Ethnic differences in caregiving duties and
burden among parents and siblings of persons with severe mental illnesses.
Journal of health and social behavior, 36(2), 138-150.
Inoue, S., Tanaka, S., Shimodera, S., & Mino, Y. (1997). Expressed emotion and
social function. Psychiatry Research, 72, 33-39.
Jablensky, A., Sartorius, N., Ernberg, G., Anker, M., Korten, A., Cooper, J. E., et al.
(1992). Schizophrenia: manifestations, incidence and course in different
cultures: A World Health Organization ten-country study (Vol. 97). U.S.:
Cambridge: University Press.
Jackson, H. J., Smith, N., & McGory, P. (1990). Relationship between expressed
emotion and family burden in psychotic disorders: an exploratory study. Acta
Psychiatry Scandinavia, 82, 243-249.
Jenkins, J. H., & Karno, M. (1992). The meaning of expressed emotion: theoretical
issues raised by cross-cultural research. American Journal of Psychiatry,
149(1), 9-21.
Jones, S. L., Roth, D., & Jones, P. K. (1995). Effect of demographic and behavioral
variables on burden on caregivers of chronic mentally ill persons. Psychiatric
services, 46(2), 141-145.
Kabacoff, R. I., Miller, I. W., Bishop, D. S., Epstein, N. B., & Keitner, G. I. (1990).
A psychometric study of the McMaster Family Assessment Device in
psychiatric, medical, and nonclinical samples. Journal of family psychology,
3(4), 431-439.
Kachigan, S. K. (1986). Statistical Analysis: An Interdisciplinary Introduction to
Univariate & Multivariate Methods. New York, New York: Radius Press.
Karno, M., Jenkins, J. H., de la Selva, A., Santana, F., Telles, C., Lopez, S., et al.
(1987). Expressed emotion and schizophrenic outcomes among Mexican-
American families. The Journal of Nervous and Mental Disease, 175(3), 143-
151.
143
Kavanagh. (1992). Recent developments in expressed emotion and schizophrenia.
British Journal of Psychiatry, 160, 601-620.
King, C., Hovey, J. D., Brand, E., Wilson, R., & Ghaziuddin, N. (1997). Suicidal
adolescents after hospitalization: parent and family impacts on treatment
follow-through. Journal of the American Academy of Child & Adolescent
Psychiatry, 36(1), 85-93.
King, S., & Dixon, M. J. (1996). The influence of expressed emotion, family
dynamics, and symptom type on the social adjustment of schizophrenic
young adults. Archives of General Psychiatry, 53(December), 1098-1104.
Kline, R. B. (2005). Principles and Practice of Structural Equation Modeling (2 ed.).
New York: The Guilford Press.
Knoke, D., Bohrnstedt, G., & Potter Mee, A. (2002). Statistics for Social Data
Analysis (4 ed.). Belmont, CA: Wadsworth/Thomas Learning.
Kuipers, E., Garety, P., Fowler, D., Freeman, D., Dunn, G., & Bebbington, P. (2006).
Cognitive, emotional, and social processes in psychosis: Refining cognitive
behavioral therapy for persistent positive symptoms. Schizophrenia bulletin,
32(Supplement 1), S24-S31.
Lamb, H. R., & Bachrach, L. L. (2001). Some perspectives on deinstitutionalization.
Psychiatric services, 52, 1039-1045.
Lambert, M. J. (2003). Bergin and Garfield's Handbook of Psychotherapy and
Behavior Change (5 ed.). New York: Wiley.
Land, H., & Hudson, S. (1997). Methodological considerations in surveying Latina
AIDS caregivers: Issues in Sampling and measurement. Social Work
Research, 21(4), 233-246.
Land, H., & Hudson, S. (2002). HIV Serostatus and factors related to physical and
mental well-being in Latina family AIDS caregivers. Social science and
medicine, 54, 147-159.
Leff, J., & Berkowitz, R. (1996). Working with families of schizophrenic families.
The Psychotherapy Patient, 9(3/4), 185-211.
144
Leff, J., Kuipers, L., Berkowitz, R., & Sturgeon, D. (1985). A controlled trial of
social intervnetion in the families of schizophrenic patients: two year follow-
up. British Journal of Psychiatry, 146, 594-600.
Leff, J., & Vaughn, C. (1981). The role of maintenance therapy and relative's
expressed emotion in relapse of schizophrenia: a two-year follow-up. British
Journal of Psychiatry, 139, 102-104.
Leff, J., & Vaughn, C. (1985). Expressed Emotion in Families: Its Significance for
Mental Illness. New York: Guilford.
Leff, J. P. (1976). Schizophrenia and sensitivity of the family environment.
Schizophrenia bulletin, 2(4), 566-574.
Lefley, H. P. (1998). Families, culture, and mental illness: Constructing new
realities. Psychiatry: Interpersonal & Biological Processes, 61(4), 335-355.
Little, R. J. A., & Rubin, D. B. (1987). Statistical analysis with missing data. New
York: Wiley.
Loehlin, J. C. (1998). Latent Variable Models - An Introduction to Factor, Path, and
Structural Analysis (3rd ed.). Mahwah, NJ: Lawrence Erlbaum Associates,
Publishers.
McAdoo, H. P. (1998). African-American families: strengths and realities. In H. I.
McCubbin, E. A. Thompson, A. I. Thompson & J. A. Futrell (Eds.),
Resiliency in ethnic minority families: African-American families (Vol. 3, pp.
17-30). Thousand Oaks, CA: Sage Publications.
McCabe, K. M., Yeh, M., Lau, A., Garland, A., & Hough, R. (2003). Racial/ethnic
differences in caregiver strain and perceived social support among parents of
youth with emotional and behavioral problems. Mental Health Services
Research, 5(3), 137-147.
McCubbin, H. I., McCubbin, M. A., & Thompson, A. I. (1995). Resiliency in ethnic
families: A conceptual model for predicting family adjustment and
adaptation. In H. I. McCubbin, A. I. Thompson & A. I. McCubbin (Eds.),
Resiliency in ethnic minority families: Native and immigrant American
families (Vol. 1, pp. 3-48). Madison, WI: University of Wisconsin System.
145
McCubbin, H. I., Thompson, A. I., & McCubbin, M. A. (1996). Family assessment:
resiliency, coping and adaptation - inventories for research and practice.
Madison, Wisconsin: University of Wisconsin Publishers.
McCubbin, H. I., Thompson, E. A., Thompson, A. I., & Futrell, J. A. (1998).
Resiliency in Ethnic Minority Families: African-American Families (Vol. 3).
Thousand Oaks, CA: Sage Publications.
McDonald, R. P., & Ringo Ho, M.-H. (2002). Principles and practice in reporting
structural equation analyses. Psychological Methods, 7(1), 64-82.
McFarlane, W. R., Deakins, S. M., Gingrich, S. L., Dunne, E., Horen, B., &
Newmark, M. (1991). Multiple-family psychoeducation group treatment
manual. New York, New York: Biosocial Treatment Division New York
State Psychiatric Institute.
McIver, J. P., & Carmines, E. G. (1981). Unidimensional scaling. Newbury Park,
CA: Sage Publications.
McPheeters, H. L. (1984). Statewide mental health outcome evaluation: a
perspective of two southern states. Community Mental Health Journal, 20(1),
44-55.
Mechanic, D. (1999). Mental Health and Social Policy: the Emergence of Managed
Care (4 ed.). Boston: Allyn & Bacon.
Miller, I. W., Bishop, D. S., Epstein, N. B., & Keitner, G. I. (1990). A psychometric
study of the McMaster Family Assessment Device in psychiatric, medical,
and nonclinical samples. Journal of family psychology, 3(4), 431-439.
Miller, I. W., Epstein, N. B., Bishop, D. S., & Keitner, G. I. (1985). The McMaster
Family Assessment Device: reliability and validity. Journal of marital and
family therapy, 11(4), 345-356.
Miller, I. W., Ryan, C. E., Keitner, G. I., Bishop, D. S., & Epstein, N. B. (2000). The
McMaster Approach to families: theory, assessment, treatment and research.
Journal of family therapy, 22, 168-189.
Minuchin, S., & Fishman, H. C. (2004). Family Therapy Techniques. Cambridge,
MA: Harvard University Press.
146
Moline, R. A., Singh, S., Morris, A., & Meltzer, H. Y. (1985). Family expressed
emotion in relapse in schizophrenia in 24 urban American patients. American
Journal of Psychiatry, 142(9), 1078-1081.
Nabors, N., Seacat, J., & Rosenthal, M. (2002). Predictors of caregiver burden
following traumatic brain injury. Brain injury, 16(12), 1039-1050.
NAMI. (2003). Family to Family :recent studies of Family-to-Family. Retrieved
12/22/2006, from http://www.nami.org/Template.cfm?Section=Family-to-
Family&template=/ContentManagement/ContentDisplay.cfm&ContentID=3
6366
Nichols, M. P., & Schwartz, R. C. (2005). Family Therapy: Concepts and Methods
(7 ed.). Boston, MA: Allyn & Bacon.
Pearlin, L. I. (1991). The study of coping: Problems and directions. In J. Eckenrode
(Ed.), The Social Content of Coping (pp. 261-276). New York: Plennum.
Pearlin, L. I., Aneshensel, C. S., & LeBlanc, A. J. (1997). The forms and
mechanisms of stress proliferation: The case of AIDS caregivers. Journal of
Health and Social Behavior, 38(3), 223-236.
Pearlin, L. I., Mullan, J., Semple, S. J., & Skaff, M. M. (1990). Caregiving and the
stress process: An overview of concepts and their measures. The
Gerontologist, 30, 538-594.
Pearlin, L. I., & Schooler, C. (1978). The structure of coping. Journal of Health and
Social Behavior, 19, 2-21.
Pfammatter, M., Junghan, U. M., & Brenner, H. D. (2006). Efficacy of psychological
therapy in schizophrenia: Conclusions from meta-anayses. Schizophrenia
bulletin, 32(Supplement 1), S64-S80.
Phillips, E. S., Barrio, C., & Brekke, J. (2001). The impact of ethnicity on
prospective functional outcomes from community-based psychosocial
rehabilitation for persons with schizophrenia. Journal of Community
Psychology, 29(6), 657-673.
Pickett, S. A., Vraniak, D. A., Cook, J. A., & Cohler, B. J. (1993). Strength in
adversity: blacks bear burden better than whites. Professional psychology:
research and practice, 24(4), 460-467.
147
Pilling, S., Bebbington, P., Kuipers, E., Garety, P., Geddes, J., Orbach, G., et al.
(2002). Psychological treatments in schizophrenia: I. Meta-analysis of family
intervention and cognitive behavior therapy. Psychological Medicine, 32,
763-782.
Pinquart, M., & Sorensen, S. (2005). Ethnic differences in stressors, resources, and
psychological outcomes in family caregiving: a meta-analysis. The
Gerontologist, 45(1), 90-106.
Reinhard, S. C., Gubman, G. D., Horwitz, A. V., & Minsky, S. (1994). Burden
Assessment Scale for families of the seriously mentally ill. Evaluation and
program planning, 17(3), 261-269.
Rigdon, E. (1998). Model Comparison in SEM. Retrieved 12/20/2006, from
http://www2.gsu.edu/~mkteer/nested.html#refs
Riso, L. P., Klein, D. N., Anderson, R., L., Ouimette, P. C., & Humberto, L. (1996).
Covergent and Discriminant Validity of Perceived Criticism from Spouses
and Family Members. Behavior Therapy, 27, 129-137.
Rivera, R., Torres, M., & Carre, F. J. (1997). Role burdens: The impact of
employment and family responsibilities in the Health status of Latino women.
Journal of Health Care for the Poor & Underserved, 8, 99-113.
Robinson, C. R. (1983). Black women: a tradition of self-reliant strength. Women &
Therapy, 2(2-3), 135-144.
Saldana, D. H., Dassori, A. M., & Miller, A. L. (1999). When is caregiving a
burden? Listening to Mexican American women. Hispanic journal of
behavioral sciences, 21(3), 282-301.
Scazufca, M., & Kuipers, E. (1996). Links between expressed emotion and burden of
care, in relatives of patients with schizophrenia. British Journal of
Psychiatry, 168, 580-587.
Schreiber, J. B., Stage, F. K., King, J., Nora, A., & Barlow, E. A. (2006). Reporting
structural equation modeling and confirmatory factor analysis results: a
review. The Journal of Educational Research, 99(61), 323-337.
148
Services, U. S. D. o. H. a. H. ((2001)). Mental Health: Culture, Race, and
Ethnicity—A Supplement to Mental Health: A Report of the Surgeon General.
Rockville, MD: U.S. Department of Health and Human Services, Substance
Abuse and Mental Health Services Administration, Center for Mental Health
Services.
Shern, D. L., Wilson, N. Z., Coen, A. S., Patrick, D. C., Foster, M., Bartsch, D. A., et
al. (1994). Client outcomes II: Longitudinal client data from the Colorado
Treatment Outcome Study. The Milibank Quarterly, 72(1), 123-148.
Solomon, P., & Draine, J. (1995). Subjective burden among family members of
mentally ill adults: relation to stress, coping, and adaptation. American
Journal of Orthopsychiatry, 65(3), 419-427.
Song, L., Biegel, D. E., & Milligan, S. E. (1997). Predictors of depressive
symptomatology among lower social class caregivers of persons with chronic
mental illness. Community mental health journal, 33(4), 269-286.
Steiger, J. H., Shapiro, A., & Browne, M. W. (1985). On the multivariate asymptotic
distribution of sequential Chi-square statistics. Retrieved 3, 50
Stricker, K., Schulze Monking, H., & Buchkremer, G. (1997). Family interaction and
the course of schizophrenic illness - results of a multivariate prospective
study. Psychopathology, 30, 282-290.
Stueve, A., Vine, P., & Struening, E. L. (1997). Perceived burden among caregivers
of adults with serious mental illness: comparison of black, hispanic, and
white families. American journal of orthopsychiatry, 67(2), 199-209.
Telles, C., Karno, M., Mintz, J., Paz, G., Arias, M., Tucker, D., et al. (1995).
Immigrant families coping with schizophrenia: behavioral family intervention
versus case management with a low-income Spanish-speaking population.
British Journal of Psychiatry, 167, 473-479.
Theis, S. L., Moss, J. H., & Pearson, M. A. (1994). Respite for caregivers: an
evaluation study. Journal of community health nursing, 11(1), 31-44.
Turner, W., L., & Alston, R. J. (1994). The role of the family in psychosocial
adaptation to physical disabilities for African American families. Journal of
the National Medical Association, 86(12), 915-921.
149
USDHHS. (2001). Mental Health: Culture, Race, and Ethnicity-A Supplement to
Mental Health: A Report of the Surgeon General. Rockville, MD: U.S.
Department of Health and Human Services, Substance Abuse and Mental
Health Services Administration, Center for Mental Health Services.
Van Humbeeck, G., Van Audenhove, C., De Hert, M., Pieters, G., & Storms, G.
(2002). Expressed emotions: a review of assessment instruments. Clinical
Psychology Review, 22, 321-341.
Wallace Williams, S., Dilworth-Anderson, P., & Goodwin, P. Y. (2003). Caregiver
role strain: the contribution of multiple roles and available resources in
African-American women. Aging & Mental Health, 7(2), 103-112.
Weick, A., & Saleebey, D. (1995). Supporting family strengths: Orienting policy and
practice toward the 21st century. Families in Society, 76(3), 141-149.
Williams, D. R., & Fenton, D. T. (1999). The mental health of African Americans:
findings, questions, and directions. In I. L. Livingston (Ed.), Handbook of
black American health - the mosaic of conditions, issues, policies, and
prospects (pp. 253-268). Westport, CN: Greenwood Press.
Williams, S. E., & Finger Wright, D. (1992). Empowerment: the strengths of black
families revisited. Journal of multicultural social work, 2(4), 23-36.
Workgroup., N. A. M. H. C. B. S. (2000). Translating Behavioral Science Into
Action: Report of the National Advisory Mental Health Council Behavioral
Science Group (No. NIH Publication No. 00-4699). Bethesda, MD: National
Institutes of Health/National Institute of Mental Health.
Wuerker, A. K., Haas, G. L., & Bellack, A. (1999). Racial and gender differences in
Expressed Emotion and interpersonal control in families of persons with
schizophrenia. Family Process, 38(4), 477-496.

Abstract (if available)

Abstract Background: Many studies have tested the impact of different family factors on the functioning of consumers living with schizophrenia. Few studies have concurrently tested these factors together and with a larger sample of poorer African-American consumers and their families, not typically well represented in the literature. Because consumer clinical functioning and psychosocial functioning are known to be minimally correlated, they were treated as two separate domains. Likewise, it is argued that consumer psychosocial functioning is more germane as an outcome variable given that most consumers spend more time in the community than in the past.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

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

PDF

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

PDF

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

PDF

The role of protective factors on outcomes for Latinos with schizophrenia

PDF

Occupation in the lives of adults with schizophrenia: creations of hope

PDF

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

PDF

Cardiovascular disease risk factors and cognitive function

PDF

Family environment as a moderator of the association between theory of mind and social functioning in people with schizophrenia

PDF

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

PDF

School social dynamics as mediators of students personal traits and family factors on the perpetration of school violence in Taiwan

PDF

Testing biopsychosocial models integrating electrodermal activity, neurocognition, and social cognition to predict the functional outcomes among people with schizophrenia at three time points ove...

PDF

An exploratory study of the role of religion and religious involvement among Latinos with schizophrenia and family caregivers

PDF

Effects of social capital on the psychosocial adjustment of Chinese migrant children

PDF

The impact of family structure and family process on adolescent sexual behavior: race and gender variations

PDF

The influences of anxiety, coping, and social support on physical functioning among heart failure patients

PDF

Impact of required parental involvement on African American male students and families: a qualitative study of the USC-NAI program

PDF

Gendered issues in alcohol abuse and dependence among HIV-positive African Americans

PDF

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

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

A longitudinal study of job resources, burnout, and turnover among social workers: implications for evidence-based organizational practices

Asset Metadata

Creator Guada, Joseph (author)

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

School School of Social Work

Degree Doctor of Philosophy

Degree Program Social Work

Publication Date 07/26/2007

Defense Date 06/08/2007

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

Tag African-American,families,OAI-PMH Harvest,psychosocial functioning,schizophrenia,SEM

Language English

Advisor Brekke, John S. (committee chair), Biblarz, Timothy J. (committee member), Land, Helen (committee member)

Creator Email guada@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m690

Unique identifier UC1211556

Identifier etd-Guada-20070726 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-523883 (legacy record id),usctheses-m690 (legacy record id)

Legacy Identifier etd-Guada-20070726.pdf

Dmrecord 523883

Document Type Dissertation

Rights Guada, Joseph

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